Old Chinese
A NEW RECONSTRUCTION
1
1
Oxford University Press is a department of the University of Oxford.
It furthers the Universitys objective of excellence in research, scholarship,
and education by publishing worldwide.
Oxford NewYork
AucklandCape TownDar es SalaamHong KongKarachi
Kuala LumpurMadridMelbourneMexico CityNairobi
New Delhi Shanghai Taipei Toronto
With officesin
ArgentinaAustriaBrazil ChileCzech RepublicFranceGreece
GuatemalaHungaryItalyJapanPolandPortugalSingapore
South KoreaSwitzerlandThailandTurkeyUkraineVietnam
135798642
Printed in the United States of America
on acid-freepaper
{ CONTENTS }
Acknowledgmentsxi
1 Introduction 1
1.1 What is Old Chinese? 1
1.1.1 The traditional approach to Old Chinese reconstruction 2
1.1.2 A broader approach 3
1.2Methodology 4
1.2.1 The nature of linguistic reconstruction 4
1.2.2 Our approach to reconstructing Old Chinese 6
1.3 Plan of the book 8
6 Conclusion 316
6.1 What kind of language was Old Chinese? 316
6.2 Dialect differences in Old Chinese 319
6.3 Known issues 320
6.3.1 *-a VS. *-A, *-ak VS. *-Ak, *-aj VS. *-Aj 320
6.3.2 *s- VS.*S- 321
6.3.3 Unexplained phonological gaps 321
6.3.4 Unknown functions of the *-s suffix 321
6.4 General directions for future research 322
6.5 Old Chinese in broad comparative context 323
We are grateful for the support for our long-distance collaboration on this project pro-
vided by our respective institutions in Ann Arbor and Paris:the University of Michigan
and the Centre de Recherches Linguistiques sur lAsie Orientale. We would like to thank
the University of Michigans Department of Asian Languages and Cultures, Department
of Linguistics, Center for Chinese Studies, and especially the Information Technology
office (LSAIT) and the Language Resource Center (LRC) of the College of Literature,
Science, and the Arts. Our project would hardly have been possible without the gen-
erous and competent assistance of the LSAIT and the LRC, whose staff (especially
Reid Paxton, Julie Evershed, and Jan Stewart) helped us set up a large database of Old
Chinese words, maintained on a server in LRC, which we are able to access and update
in real time from any internet connection.
We also thank the following individuals who have provided us various kinds of help
and support:Michelle Abud, Akitani Hiroyuki , Wolfgang Behr, Chn Jin
, Katia Chirkova, Michel Ferlus, Paul Goldin, Zev Handel, Guillaume Jacques,
Li Gulng , Alain Lucas, Nohara Masaki , Weera Ostapirat, Pn
Wyn , Thomas Pellard, Phm Th Thu H , Qi Xgu ,
Martha Ratliff, Axel Schuessler, Adam Smith, Tor Ulving, Wng Hngzh , and
Crispin Williams. Of course they bear no responsibility for any errors.
Finally, we would like to acknowledge our profound intellectual debt to the late Jerry
Norman, Andr-Georges Haudricourt, and our teachers Nicholas C. Bodman and L
Guyo.
Old Chinese
{1}
Introduction
This book presents the Baxter-Sagart reconstruction of Old Chinese, a product of several
years collaboration between William H.Baxter and Laurent Sagart to produce an improved
linguistic reconstruction of the phonology, morphology, and lexicon of Old Chinese, the
language of the earliest Chinese classical texts (early first millennium bce) and the ancestor
of later varieties of Chinese. We have both written on Old Chinese reconstruction in the past
(especially Baxter 1992; Sagart 1999c; Baxter and Sagart 1998). However, newly available
evidence has now made it possible to take a new approach to the problem and to achieve
significant improvements over previous research. The Baxter-Sagart reconstruction is the
result of this new approach. Acompanion document giving proposed reconstructions for
over 4,000 individual Old Chinese lexical items is available online at http://ocbaxtersagart.
lsait.lsa.umich.edu/.
Reconstructing the linguistic features of this 3,000-year-old language is not just a
matter of satisfying the curiosity of historical linguists; in fact, it is crucial to interpret-
ing the foundational texts of Chinese civilization. With an ancient text from Greece or
Rome, written in an alphabetic script, the words are usually easy to identify, and one
can begin to interpret the text without worrying excessively about how it was actually
pronounced. But when reading early Chinese texts, reconstructing pronunciation often
plays a crucial role in the initial process of identifying the words of a text themselvesa
fact recognized for centuries by Chinese scholars.
We use the term Old Chinese in a broad sense to refer to varieties of Chinese used
before the unification of China under the Qn dynasty in 221 bce. The earliest
written records in Chinese are oracular inscriptions on bones and shells from about
1250 bce (in the late Shng dynasty, which was overthrown by the Zhu in
1045 bce), so this is an interval of about 1,000years. Obviously, there must have
been many varieties of Chinese during this period, widely distributed in time and
space. In principle, we would like to reconstruct the entire linguistic history of this
period in all its complexity, but we can be certain that much of this information has
been irretrievablylost.
2 Old Chinese
The most important early work on reconstructing Old Chinese was done by Chinese
scholars in the Qng dynasty (16441911), who focused on early pronunciation
as a philological tool for interpreting canonical texts from the Old Chinese period.
Using both early rhymes and the phonetic patterns in the writing system, they were
able to identify words in classical texts that may have been confused or written in
nonstandard ways in the course of textual transmission.
The Swedish sinologist Bernhard Karlgren (18891978) combined the Qng
scholars findings with the notation and some of the techniques of the linguistics
of his time and produced a phonetic reconstruction of Old Chinese (summarized in
Karlgren 1954; his term for it was Archaic Chinese), relying on three main kinds
of evidence:
Karlgrens approach, which became traditional, was to propose pronunciations for Old
Chinese that were both consistent with these three kinds of evidence and (to one degree
or another) plausible for a natural language. Baxters Handbook of Old Chinese phonol-
ogy (1992) is in this tradition. The traditional approach has achieved important results,
but it suffers from several limitations:
1. Evidence from outside Chinese written records (e.g., the spoken forms of
modern dialects and early Chinese loanwords in other languages) was largely
ignored.
2. Most researchers relied on a traditional analysis of the rhymes of the Shjng
based on the work of the Qng philologists, instead of directly examining the
rhyme evidence itself.
3. For the most part, the analysis of the phonetic elements of the Chinese script
was based not on the actual scripts of the pre-Qn period but rather on the script
that has been standard since the Hn dynasty (206 bce 220 ce) or on the
Shuwn jiz , a dictionary completed in 100 ce that attempts
to analyze each character of the script established in the Qn dynasty (221207
bce). This procedure is obviously anachronistic.
4. Old Chinese has mostly been treated as a single homogeneous synchronic system;
little has been said about linguistic variation in the Old Chinese period.
5. The focus has been on phonetic reconstruction, with relatively little
attention to morphology or to the syntactic and semantic properties of the
reconstructedforms.
In our new reconstruction we still rely heavily on the three kinds of evidence used in tra-
ditional reconstructions. But a number of recent developments have now made it feasible
to take a broader approach. Modern dialectsespecially those dialects that are likely to be
most informative about Old Chinese, such as those of the Mn group, spoken in and near
Fjin provinceare much better documented now than in the past, so it is no longer
necessary or even acceptable to rely on the written remains of Middle Chinese alone, as
Karlgren did, as a surrogate for later forms of Chinese.
We also have much better documentation of and research on languages of the
Kra-Dai (= Tai-Kadai),2 Hmong-Mien, Tibeto-Burman, and Vietic families that
4 Old Chinese
preserve early loanwords from Chinese. These loanwords frequently give us informa-
tion that would be difficult or impossible to recover from modern dialects alone, or
from Middle Chinese written sources, but that should be accounted for in reconstruct-
ing Old Chinese.
Another recent development is that dramatic archeological discoveries in China
are producing a growing and diverse corpus of excavated documents3 from the
pre-Qn period, before the script was unified and standardized, especially docu-
ments on bamboo strips from the Gudin archeological site in Hbi
province (see GD) and similar documents acquired in the 1990s by the Shnghi
Museum (published in SB). Previously, the only substantial documents from the
Old Chinese period were rather formulaic and restricted in content: the Shng
oracle-bone inscriptions, and ceremonial inscriptions on bronze vessels. Many
common words simply do not appear in these documents, so Karlgrens reliance
on the later standard script is understandable. But the recently discovered docu-
ments, written on silk or on strips of bamboo or wood, are far richer and more
diverse, representing a variety of domains including philosophy, history, mythol-
ogy, law, divination, and medicine. Some overlap with previously known texts of
the received tradition, but many were previously unknown or known only by name.
Moreover, the characters in these texts, many also previously unknown, give pre-
cious evidence about pre-Qn pronunciations. By contrast, many characters in the
standard script first came into use during Qn and Hn and do not represent Old
Chinese phonology atall.
We also believe that the reconstruction of Old Chinese phonology has progressed to
the point that it is productive to use the techniques of internal reconstruction to recon-
struct the morphology of Old Chinese. Karlgren laid some of the groundwork for this
approach in his paper Word families in Chinese (1933), in which he grouped together
words that were similar in sound and meaning as a preliminary step to identifying roots
and morphological processes. But the inadequacies of his phonological reconstruction
made it difficult to identify the patterns involved. Improved phonological reconstruc-
tions, along with evidence from dialects and early loanwords, are now making it pos-
sible to identify morphological processes with more precision.
Finally, through improved reconstructions, we are now better able to understand
some of the information in early texts, both implicit and explicit, about early Chinese
and its dialectal variants.
1.2Methodology
should not be found in Chinese dialects or in early Chinese loanwords into other
languages. Thus our reconstructions are subject to falsification by either existing or
newly discovered evidence.
It follows that the work of reconstruction can never be considered complete as long
as new data may become available. It is somewhat unfortunate that the English verb
reconstruct seems to fall into the class of what Zeno Vendler (1957) called accom-
plishment verbsthose that involve both a process and an endpoint. With accom-
plishment verbs (or, better, verbal expressions) such as run a mile, it makes sense
to ask questions such as How long did it take you to run a mile? Accomplishment
verbs contrast with activity verbs such as run; for these, it is normal to ask, How
long did you run? but not How long did it take you to run? (unless an endpoint
is presupposed, as when a person runs a specific distance every day). In the case of
reconstruct, it sounds normal to ask, How long did it take you to reconstruct Old
Chinese? so reconstruct Old Chinese is treated as an accomplishment verb. Indeed,
Karlgren seems to have believed that he had finished reconstructing his Archaic
Chinese when he published his Grammata serica (1940).4
But in our view, this understanding of reconstruction is misleading. On the one hand,
since some information about Old Chinese has necessarily been irretrievably lost, the
process of reconstructing it will never be entirely complete. At the same time, as more
evidence becomes available, and as more people study it, existing reconstructions will
need to be modified. We believe that the basic hypotheses of our reconstruction are suf-
ficiently stable that it makes sense to publish our results now. But we are making the
reconstructions themselves available on a public website so that they may be revised as
necessary in the light of new evidence or arguments.
It is worth emphasizing that it is as important for hypotheses to be able to predict
what will not be observed as it is to predict what will be observed. If a reconstruction
cannot account for certain examples, it is usually possible to incorporate ad hoc com-
plications in the reconstruction to account for them; but if the reconstruction becomes
so powerful that it is consistent with all imaginable observations, it loses its predictive
power. It is often better to leave some phenomena unexplained until a pattern emerges
that will explainthem.
Our ultimate goal is not just to reconstruct Old Chinese in the narrow sense but also to
reconstruct everything we can about the linguistic history of the pre-Qn period. What
was the early phonological system, and how did it change? What morphological pro-
cesses were there? What dialect distinctions were there, and how did they develop?
How did the semantic and grammatical properties of each lexical item evolve? What
languages was Chinese in contact with, and what were the linguistic results of these
contacts? For that matter, how was the linguistic history of Chinese related to the his-
tory of Chinese speakers themselves? To answer these questions we think all available
evidence should be used, not just the traditional triad describedabove.
Introduction7
Our use of each of the main kinds of evidence is summarized in Chapter2. First,
though, some remarks about our notation are inorder.
When citing examples, we generally cite pronunciations from modern standard
Mandarin, Middle Chinese (MC), and reconstructed Old Chinese (OC). Mandarin pro-
nunciations are given in pnyn romanization. Middle Chinese pronunciations are given
in a conventional transcription, explained in more detail in Chapter2. These Middle
Chinese transcriptions are not phonetic reconstructions but conventional representa-
tions of the information about pronunciation given in Middle Chinese written sources.
Accordingly, they are not preceded by asterisks; for typographical convenience, and
to emphasize the fact that they are not reconstructions, they are restricted to ordinary
ASCII characters (in italic type), rather than the International Phonetic Alphabet.
Old Chinese reconstructions are preceded by asterisks and do use the International
Phonetic Alphabet. Hyphens indicate morpheme boundaries; angle brackets around pre-
vocalic *-r- indicate that it is an infix (see section 3.3.2.6). As further explained in the
discussion of root structure in section 3.3.1, Old Chinese words could have phonetic
material before the main syllable. In some cases this presyllabic material can be recog-
nized as a synchronic prefix, so we separate it from the main syllable with a hyphen, as
in this pair showing the stative-intransitivizing prefix*N-:
(1) hu < xwae < *qra flower (n.) (from Hn times on, generally
written as5)
hu < hwae < *N-qra flower (v.); flowery (adj.)
However, not all presyllables can be analyzed as synchronic prefixes at the Old
Chinese stage. Some of them may have been prefixes at an earlier time but must be ana-
lyzed as part of the root at the Old Chinese stage; others may have been part of the root
all along. In either case, if we are not confident that a presyllable is a synchronic prefix,
we write a period after it instead of a hyphen:
(2) qin < tshen < *s.ni[] thousand
(3) sh < zyit < *m.li[t]fruit;full6
As noted above, we will never have full information about all the words of Old
Chinese; in particular, the reconstructions of individual words are sometimes under
determined by the evidence. For example, in most cases, we can tell from Middle
Chinese forms whether a word did or did not have *-r- before the main vowel in Old
Chinese, but in some contexts the reflexes with and without *-r- are the same. Thus
while OC *ka and *kra are still distinguished in Middle Chinese as kjang and kjaeng,
respectively, OC *ka and *kra both became MC kjo. Similarly, MC ki can reflect either
OC *k or *kr, and MC kje can reflect either OC *kaj or *kraj. We consider it desir-
able to represent such uncertainty in our notation, so in such cases we write *k(r)a,
*k(r), and *k(r)aj, meaning that for all we know, there could have been an *-r- before
the vowel in Old Chinese (but not implying that there is any positive evidence for*-r-).
Similarly, in our reconstruction, MC k- can come from either OC *k- or OC *C.q-
(where *C is an unspecified consonant). In many cases, we have evidence to choose one
8 Old Chinese
or the other reconstruction, but often we do not, and in such cases, in order to reflect this
uncertainty, we write MC k- < OC *[k]-. In general, the notation *[X] means either
*X, or something else that has the same Middle Chinese reflex as *X. In some cases,
the identity of the main vowel is also unclear. The bracket notation is especially frequent
in words with final -n in Middle Chinese, because although there is strong evidence that
MC final -n can reflect either OC *-r or *-n (see section 5.5.1), in particular words it is
often difficult to decide which coda to reconstruct. As a concrete example of uncertainty
in particular reconstructions, we reconstruct j < kje odd number as *[k](r)aj, indi-
cating that we do not know from currently available evidence whether the MC initial
k- is of velar or uvular origin or whether there was an *-r- before the vowel ornot.7
Finally, in paleographic discussions, we adopt the convenient convention (as in Qi
Xgu8 1988, 2000)that a character in curly brackets refers to the word the character now
represents in the standard script, not to the written character itself. Thus {} wn hear
refers to the word now written as , not the character itself; in the Old Chinese
period, {} wn was written in a variety of ways, but was not one ofthem.
Chapter2 discusses in more detail the kinds of evidence on which our reconstruction
is based. Chapter3 gives a brief summary of the history of proposed reconstructions of
Old Chinese down to the present time and summarizes the innovations of the present
one. The core of the book is Chapter4, on the reconstruction of Old Chinese syllable
onsets, and Chapter5, on the reconstruction of Old Chinese rhymes. Chapter6 identifies
some known problems in the reconstruction and topics for future research.
{2}
Each rhyme is further divided into homonym groups. Example (4)below gives the
entire first homonym group of the first rhyme of the pngshng section in the Wng
Rnx Qiyn, containing entries for the two characters and.
(4)
In each homonym group, the first entry is provided with a spelling in the system
called fnqi , a device for indicating the pronunciation of a word by giving two
characters:one representing a word that has the same initial consonant as the word being
spelled, and a second in which everything but the initial consonant is the same. These two
characters are followed by either fn (as in the Wng Rnx version of the Qiyn) or
qi (as in the Gungyn) to identify them as a fnqi spelling.
For example, here is the fnqi spelling in the entry for pictured in example (4),
specifying the pronunciation of the two homonyms and:
(5)
d hng fn
The first character d identifies the initial consonant; the second character hng
specifies the rest of the syllable; and the third character is what tells us that is
a fnqi spelling. In our ASCII-friendly notation for Middle Chinese (further explained
below), d is tok, and hng is huwng. To get the Middle Chinese pronunciation of
and , we put the initial consonant of tok together with everything but the initial
consonant of huwng: t(ok) + (h)uwng=tuwng. This indicates that tuwng is
the Middle Chinese pronunciation of the words in this homonym group, that is, of both
and.
If in turn we look up the initial speller d, we find the followingentry:
(6)
(7)
(8),
Hu gng zh j nin, bi Sng sh y Hung.6
In the last year of Duke Hu, he defeated the Sng troops at Hung.
(9)
Bi Sng, b mi fn, bi t y. Hu fngc.
[In the phrase] defeat Sng [troops], [ is read] [MC
p(jit) + (m)aejH=paejH]; it means to defeat another; [examples]
below are like this (JDSW222).
The reason for the annotation is that according to the Middle Chinese sources, the
character bi has two readings:MC paejH when it means to defeat (transitive) and
MC baejH when it means to suffer defeat (intransitive). The comment tells the reader
that in this passage in the Zu zhun, bi is to be read as MC paejH, because it is used
TABLE2.1 d, du, and d in Middle Chinese, selected dialects, and loanwords from
Chinese
Middle Chinese tok ta tok
Mandarin [t] [tuo] [t]
Cantonese [tk] [t] [tk]
Szhu [t] [t] [t]
Sino-Korean tk [tk] ta [ta] tk [tk]
Sino-Japanese toku ta toku
Sino-Vietnamese c [k D1] a [ A1] c [ak D1]
12 Old Chinese
The Middle Chinese sources define an abstract space of possible syllables, and it is
usually easy to determine from the written sources where a particular form belongs in
this space. AChinese syllable is traditionally analyzed as consisting of three parts:an
initial (shngm ), a final (ynm ), and a tone (shngdio ); the
phonological space can therefore be thought of as three dimensional. We will use the
traditional terminology of initial, final, and tone in describing Middle Chinese; but
note that the final, as traditionally defined, is not the same thing as the rhyme:the
final includes some elements before the main vowel, which in a different analysis
would be considered part of the syllableonset.
Even if it is problematic to decide exactly how this syllable space might correspond
to the pronunciation of any particular spoken variety of Middle Chinese, in most cases
we can be confident that the distinctions it includes are not artificial and existed in some
variety of Chinese at some time; they are therefore relevant to the reconstruction of Old
Chinese. Our Middle Chinese notation is intended to represent the position of each syl-
lable within that abstractspace.
Generally, our Middle Chinese transcription follows these principles:
2. When the pronunciation of an element of the Middle Chinese system is more or less
unproblematic, its transcription is chosen accordingly. For example, the words
whose MC transcription begins with t- probably really did begin with [t]in most
varieties of Chinese in the Middle Chinese period; those ending in -p, -t, -k, -m, -n,
-ng probably really did end in [p], [t], [k], [m], [n], and []; and soforth.
3. In cases where pronunciations are unclear, are difficult to represent in ASCII
symbols, or may have differed from dialect to dialect, we use conventional
symbols chosen for their mnemonic value. For example, we write gu
country as MC kwok. Now in most varieties of Middle Chinese the word
probably really did begin and end with [k], and the -w- is probably realistic
for most dialects as well. In proposed reconstructions of Middle Chinese, the
word is usually reconstructed something like [kwk] or [kwk]. But we have
no convenient ASCII equivalent for [] or []; besides, how do we know that
the vowel was unroundedand which dialect would we be talking about
anyway? Although the notation kwok may not exactly match any Middle
Chinese dialect, it does have the virtue that it uses easily recognizable
symbols and is easy to connect mnemonically with other known forms of
this etymon, as shown in Table2.2.
4. As far as possible, our notation is designed so as to make it easy to identify what
seem to be natural classes in the Qiyn system. For example, one such natural
class is the so-called division-II (rdng ) finals. Although this name
originates with the Late Middle Chinese rhyme tables, the division-II finals are
easy to define using distributional criteria based on the Qiyn itself, without
reference to the rhyme tables:they are the finals in the Qiyn system that occur
with retroflex initials such as tr- and tsr- but not with palatal initials such
as tsy- and ny-.8 In our notation, all division-II syllables are recognizable from
the fact that they are written with either -ae- or -ea- as main vowel, with no
preceding -j- or -y-. Some varieties of Chinese around 601 ce probably really
did have two contrasting vowels here, which are sometimes reconstructed as []
and [], but there were probably other varieties that did not distinguish them.
Our -ae- is mnemonic for [], and -ea- is easy to relate to both -ae- and[].
The names of the tones were evidently chosen to be both descriptive of the tones
they denote and examples of them; thus png < bjaeng itself is a pngshng word,
shng ~ shng < dzyangX is a shngshng word, and so forth. To preserve this ico-
nicity in modern pronunciation, when is used as the name of the Middle Chinese
tone, it is conventionally given the pronunciation shng (rather than shng, the regular
modern Mandarin reflex of MC dzyangX), because most shngshng words have tone 3
in Mandarin. (However, shngshng words that had voiced obstruent initials in Middle
Chinese, including dzyangX itself, regularly have Mandarin tone 4.) The rshng
category includes all and only those words that ended in -p, -t, or -k; these codas have
been lost in Mandarin but are preserved in some southern dialects, such as Cantonese.
Note that the four tones of Middle Chinese do not correspond in a simple way to
the four tones of modern Mandarin:Middle Chinese pngshng words regularly have
Mandarin tone 1 or tone 2; shngshng words have tone 3 or tone 4; qshng words
have tone 4; and rshng words may have tone 1, 2, 3, or 4 in a pattern that is rather
irregular in standard Mandarin.
system, that of Dng Shngsh and L Rng (1981); the Chinese name of each initial is
a word that had that initial in Middle Chinese. The capitalized notations in the second
column are cover symbols for the sets of initials listed to their right. We call initials
of the types P- and K- grave initials; the others are acute initials. The numbers in
square brackets in Table2.3 refer to notes after the table. (A more detailed discussion of
the Middle Chinese initial consonants, and other variants of the traditional terminology,
may be found in Baxter 1992:4561.)
Lists of Middle Chinese initials also frequently include Fi, F, Fng, and
Wi, but these are simply the labiodental counterparts of p- ( Bng), ph- ( Png),
b- ( Bng), and m- ( Mng), respectively. Although the four names for labiodental
initials are frequently used with reference to Middle Chinese, to use them for initial
consonants of the Qiyn system is anachronistic:the labiodentals had not yet devel-
oped in the Early Middle Chinese of the Qiyn system, so we do not distinguish them
in our notation.
Notes on Table2.3:
[1] Y.R. Chao (1941) pointed out that n- and nr- are in complementary distribution
and proposed that they need not be distinguished; accordingly, sometimes the name
N is used for both our n- and our nr-. However, the distinction is usually maintained in
fnqi spellings, and we retain it in our notation.
[2] The initials dzr- and zr- were not distinguished in the Gungyn, and hence zr-
has no traditional name. However, the Wng Rnx Qiyn does have a distinct initial
zr-, and we make the distinction in our notation.
[3] Nowadays, the more common pronunciation of the character is probably chn
(meaning meditation; Chn or Zen Buddhism), but as the name of this initial, Dng and
L (1981) give the other pronunciation shn (meaning abdicate).
16 Old Chinese
[4] Before the analysis of fnqi spellings showed them to be distinct initials, MC y-
( Y) and hj- ( Yn) were treated as a single initial with the traditional name Y.
Since in the rhyme tables y- ( Y) is always placed in division IV (sdng ) and
hj- ( Yn) in division III (sndng ), a common alternative terminology refers
to y- ( Y) as Y s and to hj- ( Yn) as Y sn; we will sometimes
use these termsalso.
[5] According to current dictionaries, the character is pronounced x in Mandarin,
not the q that might be expected from its Middle Chinese reading khej; some diction-
aries do include q as an older reading.
[6] The traditional initials Xi and Yn are in complementary dis-
tribution, which is why they are both written with h- in our notation, but the
traditional terminology distinguishes them. In context, they can always be dis-
tinguished in our notation as well:all cases of h- before a division-III final rep-
resent Yn (including those written with -i- rather than -j-, e.g., yn
< hwinX fall down). All cases of h- before other finals represent Xi. In lists and
tables such as this one, we distinguish them by writing h- for Xi and hj- for
Yn, even though an h- representing Yn is not always followed by -j- in spelling.
-uwng, -uwngX, -uwngH, and -uwk are treated as a set, as if -uwk were -uwng in a fourth
tone. Another analysis would be to say that Middle Chinese had only three tones but that
stop-final syllables had no tonal distinctions.
Notes on Table2.5:
[1]There is no separate shngshng rhyme corresponding to Dng; a few words
with the final -owngX are included, with special annotations, in the Zhng rhyme
(which otherwise includes only words with -jowngX); evidently, words with the final
-owngX were considered too few to justify a separaterhyme.
18 Old Chinese
[2]Afew rhymes with the coda -j occur only in qshng; this is explained in section
5.5.2.2.
[3] Although -won and -on are separate rhymes in the other tones, there is no separate
rhyme for -ot; the few words in -ot are included in the m rhyme with words in-wot.
[4] The Qiyn does not have separate rhymes for -an vs. -wan, -at vs. -wat,
or -a vs. -wa, but the Gungyn separates them, as in the table. After labial initials, there
is no contrast between -an, -at, -a on the one hand and -wan, -wat, -wa on the other,
and the Qiyns fnqi spellings usually imply the former:e.g., m < mat end of
a branch is spelled in the Wng Rnx Qiyn manuscript as , i.e., m(ak) +
(k)at=mat. But the Gungyn usually treats labial-initial syllables with these finals as
if they had -w- and puts them in the rhymes with -wan, -wat, -wa. Our notation follows
the Qiyn rather than the Gungyn on this; thus we write m as MC mateven
though the Gungyn uses m < mat as the name of the rhyme containing words with
-wat, as if it were mwat.
The division-II finals and their Gungyn rhymes are listed in Table2.6. Note that
they all have vowels written as -ae- or -ea-, with no preceding -j- (or-y-).
The division-IV finals and their Gungyn rhymes are listed in Table2.7.
For the division-III rhymes we give two tables:Table2.8 presents finals with vocalic
(or zero) codas; Table2.9 presents finals with nasal or stop codas. In our Middle Chinese
notation, the division-III finals are those that are spelled with -i- as the main vowel, or
with -j- or -y- to the left of the main vowel, or both. Some finals are spelled with both
-j- before the vowel and -i- as all or part of the main vowel (possibly with an intervening
-w-). This notation is used to distinguish the so-called chngni repeated-initial
finals, which call for special comment.
An example of a rhyme with chngni finals is the Zh rhyme, in which there
are two homonym groups for syllables with MC initial p-:one with the head word
bi < MC pje pillar and one with the head word bi < MC pjie low, humble. The
term chngni repeated initial refers to the fact that rhymes like this include more
than one syllable with the same initial consonantin this case, p-. The phonetic nature
of the contrast between these syllables has long been debated, but in the rhyme tables,
bi < pje is placed in division III, and bi < pjie in division IV. In our notation,
the division-IV chngni finals are those like -jie that are spelled with both prevocalic
-j- and -i-; division-III chngni finals like -je have either provocalic -j- or -i- but not
both. The distinction is limited to syllables with grave initials, that is, types P- and K-.
(For a more detailed discussion of the chngni distinctions see Baxter 1992:7581.)
The grave-initial syllables in the rhymes Yu and Qng are also placed in
division-IV in the rhyme tables, and we write the finals of these syllables as -jiw
and -j(w)ieng, respectively; these can be considered division-IV chngni syllables
in a broad sense, even though there are no contrasting division-III syllables in the
samerhyme.
Note on Table2.8:
[1]There are only a few marginal words with the finals -ja and -jwa; they are of late
origin and cannot be reconstructed for Old Chinese.
20 Old Chinese
TABLE2.9 Division-III finals of Middle Chinese and their Gungyn rhymes (nasal and
stopcodas)
-X -H -p, -t, -k
-juwng tuwng tuwngX suwngH -juwk 'uwk
-jowng tsyong tsyongX yowngH -jowk tsyowk
tsyin tsyit
-in, -win, -jin [1] tsyinX tsyinH -it, -wit-, -jit
tsrin tsrit
-win, -jwin [1] tsywin tsywinX tsywinH -wit, -jwit zywit
-jun mjun mjunX mjunH -jut mjut
-j+n xj+n 'j+nX xj+nH -j+t xj+t
-jon, -jwon ngjwon ngjwonX ngjwonH -jot, -jwot ngjwot
-jen, -jwen, -jet, -jwet,
sjen sjenX sjenH sjet
-jien, -jwien -jiet, -jwiet
-jang, -jwang yang yangX yangH -jak, -jwak yak
-jaeng, -jwaeng kaeng kaengX 'jaengH -jaek maek
-jeng, -jieng, -jek, -jiek,
tshjeng dzjengX kjiengH sjek
-jwieng -jwiek
-ing, -wing tsying tsyingX tsyingH -ik, -wik tsyik
-im, -jim tshim tshimX tshimH -ip, -jip tship
-jem, -jiem yem yemX yemH -jep, -jiep yep
-jaem ngjaem ngjaemX ngjaemH -jaep ngjaep
-jom bjom bjomX bjomH -jop bjop
Note on Table2.9:
[1] The Qiyn has a single rhyme Zhn that the Gungyn split into two rhymes,
Zhn and Zhn. The general pattern in the Gungyn is that words with -in and
grave-initial words with the division-III chngni final -win are included in Zhn,
while acute-initial words in -win and division-IV chngni words in -jwin are placed
in Zhn.
The Middle Chinese notation presented above is a slightly modified version of the
notation introduced in Baxter (1992:2785). The differences between that notation and
the present one are exactlythese:
Many pre-Qn texts contain rhymed passages, but the primary corpus traditionally used
in Old Chinese reconstruction is the Shjng , the earliest collection of Chinese
The evidence for Old Chinese21
poetry, almost all of which employs rhyme. One of the requirements for an adequate
reconstruction of Old Chinese is that it should account for which words rhyme with
each other in the Shjng andequally importantlywhich words do not rhyme with
eachother.
Like other early canonical Chinese texts, the Shjng is generally read from a text
that uses some version of the script that has been standard for about two millennia.
Even though many of the words long ago fell out of use in speech, dictionaries give
modern pronunciations for every character, and these are the pronunciations used in
recitation. When reading aloud, it is easy to notice that the original rhyme words do
not always rhyme anymore. Consider Ode 8, reproduced below as read with mod-
ern pronunciation; rhyme words are in small capitals and are transcribed in Middle
Chinese on theright.
From the structure of the poem, it is easy to identify the rhyme words:they are the
only words that change from one stanza to the next. Stanzas 2 and 3 still rhyme in both
Middle and modern Chinese, but in the first stanza, ci < MC tshojX and yu <
MC hjuwX are not a good rhyme in either Middle or modern Chinese. This rhyme is by
no means exceptional, though, and it follows a regular pattern; there are many parallel
examples. An adequate reconstruction of Old Chinese must be able to account for such
rhymes.
When interpreting rhymes or other poetic devices as evidence about pronuncia-
tion, one must not be too naive:literature is subject to its own conventions and cannot
always be assumed to reflect spoken language in a simple way. The intended audience
for works of literature frequently does not coincide with the population speaking any
22 Old Chinese
single dialect, and compromises of one kind or another are to be expected (see Baxter
1992:8797 for further discussion). But Old Chinese rhyming shows no signs of the
elaborate and often artificial conventions that came to be associated with Chinese
poetry in the Tng dynasty (618907) and later, when rhyming standards were
strictly defined in reference books and observed regardless of the poets own dialect
background. We have no evidence of rhyme books or other prescriptive literature
about poetry from the Old Chinese period. Of course, at any period, newly composed
poetry may be influenced by existing poetry: even in pre-Qn times, for example,
poets might sometimes have imitated rhymes in well-known poetry, even if they did
not reflect their own pronunciation. But on the whole, it is safe to assume that Old
Chinese rhymes were largely based on the phonology of actual speech.
As mentioned in section 1.1.1, the foundations for analyzing Old Chinese rhyming
were laid by Chinese scholars of the Qng dynasty; the culmination of this research was
the work of Wng Ninsn (17441832) and Jing Yugo (?1851).
Their practice, still widely followed today, was to identify groups of words that rhymed
in Old Chinese (ynb ), defining them in terms of the rhymes of the Gungyn,
and using the names of Gungyn rhymes as labels for these groups.
Several versions of the traditional rhyme categories are in use today. Probably
the most influential version is that given in Wng Ls (1999) textbook Gdi
Hny(Ancient Chinese). This version is summarized in Table 2.10.
Wng L did reconstruct phonetic values for each rhyme group, although the group
names are most frequently used without them; his rhyme reconstructions are included
in the table (in quotation marks, so as not to be confused with our reconstructions). As
is traditional, the rhymes are arranged into three categories:ynshng , those with
no coda or a vocalic coda; rshng , those with voiceless stop codas *-k, *-t, or
*-p; and yngshng , those with nasal codas. The division into three categories
(ji, y, and bng, or A, B, and C) is based on the position of articulation of the
coda, if any:in Wng Ls reconstruction, the Acategory have zero or velar codas, the B
category have *-i or dental codas, and the C category have labialcodas.
Wng Ls system of rhyme categories is very little changed from that of the Qng
scholars. The major revision is one proposed by Wng L himself ([1937] 1958): to
divide the traditional Zh group into a Zh group (number 18 in Table2.10), recon-
structed with a front vowel, and a Wi group (number 21 in Table 2.10), recon-
structed with a non-front vowel (Wng Ls *-ei and *-i, respectively).
The Qng scholars analysis is justly considered a major intellectual achievement,
and it is often assumed that the task of identifying Old Chinese rhyme categories is now
complete. Proposed reconstruction systems (such as that of Fang-kuei Li 1971 and the
proposals of Pulleyblank 19771978) are usually not based on the early rhymes them-
selves but rather on the traditional analysis of them. This tendency is reinforced by the
fact that the traditional analysis seems to work well in the sense that when reading early
poetry, one rarely encounters rhymes that are inconsistent withit.
But while the traditional analysis is generally consistent with the rhymes that occur,
it does not adequately explain why certain rhymes do not occur. It is easy to notice
TABLE2.10 Old Chinese rhyme groups according to Wng L, with his reconstructions
ynshng rshng yngshng
(zero or vocalic coda) (voiceless stop coda) (nasal coda)
jili 1. Zh *- 2. Zh *-k 3. Zhng *-ng
type A 4. Yu *-u 5. Ju *-uk 6. Dng [*-ung]a
7. Xio *- 8. Yo *-k
9. Hu *-o 10. W *-ok 11. Dng *-ong
12. Y *-a 13. Du *-ak 14. Yng *-ang
15. Zh *-e 16. X *-ek 17. Gng *-eng
18. Zh *-ei 19. Zh *-et 20. Zhn *-en
yli 21. Wi *-i 22. W *-t 23. Wn *-n
type B
24. G *-ai 25. Yu *-at 26. Yun *-an
cases where words rhymed in Old Chinese but do not rhyme now, as with Ode 8, quoted
above; they are obvious when the Shjng is read aloud. But it is harder to identify cases
where words that do rhyme now did not rhyme in Old Chinese:to do this, one has to
check the whole corpus to verify that certain types of rhymes are absent.
Cases where words no longer rhyme result from splits in Old Chinese rhymes, as
when ci < MC tshojX and yu < MC hjuwX, which rhyme in Old Chinese in
Ode 8.1, went into different rhymes in Middle Chinese. Cases where words rhyme now
but did not rhyme then result from mergers of Old Chinese rhymes. Since the splits are
more easily noticed than the mergers, scholars who investigated Old Chinese rhyming
initially had the impression that Old Chinese made far fewer rhyme distinctions than
later varieties of Chinese. Over time, their analysis gradually became more fine-grained,
but recent research makes it clear that it was not fine-grained enough.
For example, the early Qng scholar G Ynw (16131682) identified
only ten rhyme groups for Old Chinesecompared with thirty in Wng Ls list. As
mentioned above, the Qng scholars used the rhymes of the Gungyn as the units
of their analysis. G Ynws third rhyme group includes words from the Gungyn
rhymes Y (MC -jo), Y (MC -ju), and Hu (MC -uw), among others. He
was presumably influenced by the fact that there are Shjng rhymes between the
Gungyns Y and Y rhymes, and also between the Y and Hu rhymes,
which may have given the impression that these three Gungyn rhymes all rhymed
together in Old Chinese.
It was apparently Jing Yng (16811762) and Dun Yci (1735
1815) who noticed that those words of the Gungyns Y rhyme (MC -ju) that
rhyme in Old Chinese with the Gungyns Y rhyme (MC -jo) are a separate group
from those that rhyme with the Gungyns Hu rhyme (MC -uw); moreover, words
from the Gungyns Y and Hu rhymes do not rhyme with each other. So they
separated G Ynws third rhyme group into a Y rhyme group (our *-a, number
12 in Table2.10) and a Hu rhyme group (our *-o, number 9 in Table2.10). G
Ynw had been confused by the fact that there had been a partial merger of the Old
Chinese Y rhyme group (*-a) and the Hu rhyme group (*-o), words from both
groups winding up in the Gungyns Y rhyme (MC -ju).11
In fact, Old Chinese rhyme distinctions that were later lost through mergers are
extremely difficult to detect inductively, just by going through the Shjng text, unless
one begins with some hypotheses about where the distinctions might be. Suppose we
have a group of just fifteen rhyme words, and we want to check whether it should be
divided into two groups, but we have no idea which words were in which original group.
There are 16,368 ways to divide a group of fifteen rhyme words into two rhyme groups,
so checking all the possible solutions one by one would be a huge task.12 As a practi-
cal matter, we will make progress with the problem only if we start with some specific
hypothesis about how rhyme groups should be divided. In the case of G Ynws
rhyme group 3, the fact that the Gungyn distinguished Y (MC -jo) from Hu
(MC -uw) probably suggested to Jing Yng and Dun Yci the hypothesis that these
The evidence for Old Chinese25
two rhymes were distinguished in Old Chinese rhyming as well; when checked, this
hypothesis turned out to betrue.
Recent progress beyond the traditional analysis of Old Chinese rhyming has been
driven by hypotheses suggested not by the explicit arrangement of the Gungyn itself,
but by patterns in the distribution of phonological elements in the Middle Chinese
system. For example, the distribution of Middle Chinese prevocalic -w- suggested to
Jaxontov (1960b) that some cases of MC -w- came from Old Chinese labialized ini-
tials like *k-, but that other cases had to come from the diphthongization of original
rounded vowels:e.g., tun round; plenty < MC dwan < OC *-on, contrasting with
syllables like tn altar < MC dan < OC *-an. We call this the rounded-vowel
hypothesis. Both tun and tn are included in the traditional Yun rhyme
group of Old Chinese (number 26 in Table2.10), but the rounded-vowel hypothesis
says that they must have had different main vowels, and thus predicts that they should
not rhyme with each other in Old Chinese. Acareful analysis of the rhyme evidence
shows that this prediction is correct (Baxter 1992:370389).13 Similar distributional
arguments led to the front-vowel hypothesis, which predicts the existence of still
more rhyme distinctions that were overlooked in the traditional analysis; these predic-
tions are also supported by the rhyme data. (Both hypotheses are discussed in more
detail in section 5.2.1.)
Let us recapitulate. In using rhyme evidence to reconstruct Old Chinese, discov-
ering which rhymes do not occur is as important as discovering which rhymes do
occur. Discovering which rhymes do occur is easy; they are in the corpus for us to
see. But discovering which rhymes do not occur is a more subtle matter. It is dif-
ficult to discover such distinctions inductively just by looking through the corpus;
we have to be guided by some kind of hypothesis that suggests what distinctions to
look for. The Qng philologists relied on hypotheses suggested by the arrangement
of the Gungyn, and discovered more and more rhyming distinctions as time went
on; but there is no reason to assume that they discovered them all. Analyzing the
distribution of phonological elements in Middle Chinese has suggested hypotheses
about additional rhyme distinctions, which are confirmed by an analysis of the rhyme
data. An adequate reconstruction of Old Chinese phonology must account for these
distinctionsalso.
A limitation in the use of rhymes as evidence is that only a minority of Old
Chinese words are actually attested as rhymes. Rhymed passages in newly discov-
ered documents can resolve some questions, but even if a word is attested as a rhyme,
it may be difficult to reconstruct its pronunciation if there are too few examples. In
part this is because the poems originate at different times and places and do not all
reflect the same phonological system. (The fact that the traditional rhyme groups
are insufficiently fine-grained can give the impression that Shjng rhyming is more
uniform than it actuallyis.)
Based on the use of phonetic elements in the script, and on graphic variants occur-
ring in received texts, the Qng scholars also made contributions in understanding the
26 Old Chinese
As just noted, rhyme evidence is available only for a minority of Old Chinese words;
moreover, rhymes tell us nothing about syllable onsets. Apotentially more compre-
hensive source of evidence is the Chinese script, in which the vast majority of words
are written with a phonetic element:either by itself, in so-called loan characters (jiji
z ), or with a semantic element added, in phonetic compounds (xngshng z
or xishng z ). For convenience, we will say that words written with
the same phonetic element, with or without an added semantic element, have xishng
connections.
Since the Chinese script has remained frozen without major changes for the last
two thousand years, sound changes in the intervening period have often disrupted
the relations of phonetic similarity on which the original choice of phonetic ele-
ments was based. But this very fact means that the phonetic elements preserved in
the script can help us reconstruct earlier pronunciations. Evidence from the script
is especially important because of the three kinds of evidence used in traditional
approaches to reconstructionMiddle Chinese, Old Chinese rhymes, and the
Chinese scriptonly two, Middle Chinese and the script, tell us anything about
syllable onsets.
As with rhymes, in using evidence from the script to reconstruct Old Chinese, it is
important to take into account not only which words were written with the same pho-
netic element, but also which words were not written with the same phonetic element.
Traditional Chinese philologists noticed cases where initial consonants that were very
different in later pronunciation were apparently treated as interchangeable in the early
script.For example, Zng Ynqin ([1927] 1972)observed that words with the Middle
Chinese initials y- (traditional name: Y s= Y) and d- (traditional name:
Dng) are often interchanged in different versions of the same text and are often written
with the same phonetic element. Here are some examples:
These facts are often summarized in the saying, familiar to Chinese paleographers,
that Y s g gu Dng ([The initial] Y s in ancient times
went back to [the initial] Dng). Similar reasoning led scholars to conclude that
the Middle Chinese initials ny- (traditional name: R), nr- ( Ning), and n- (
N) all had the same origin, because the same phonetic element could be used to
write any ofthem.
Such observations gave the impression that Old Chinese had a simpler system of
initial consonants than Middle Chinese:while traditional Chinese terminology identi-
fies some forty Middle Chinese initials (the exact number depending on how one counts
them), Wng L identifies only thirty-two initial consonants for Old Chinese (1999:689
700); H Lny gets by with only nineteen (1998:1). As with rhymes, it is easy to notice
cases where initials that seem to be interchangeable in Old Chinese split into two or
more different initials later. But the reverse situation, where several Old Chinese initials
have merged into one, is not so easily noticed, and such cases have been discovered
only gradually.
For example, Pulleyblank (19621963) observed that Middle Chinese d- had two
different Old Chinese sources that are kept quite distinct in the writing system; we now
reconstruct them as *d- and *l-. Here are some minimal contrasts:
The phonetic elements used to write MC d- < OC *d- are generally separate from those
used to write MC d- < OC *l-, and follow a different pattern:phonetics used to write
MC d- < *d- are also used to write MC dzy- < *d-, MC t- < *t-, and MC tsy- < *t-,for
example, but rarely MC y-. Phonetics used to write *l-, on the other hand, are frequently
used for MC y- < *l-, sy- < *l -, and th- < *l -, but rarely for MC dzy-, t-, or tsy-. Karlgren
and most of his successors had overlooked this distinction, and it is not widely recog-
nized by paleographers either.
Just as there was more than one Old Chinese source for MC d- ( Dng), there was
also more than one source for MC y- ( Y s= Y), and the early script does not
confuse them. We reconstruct both *l- and *- (a voiced uvular stop) as sources for MC
y-, as illustrated by the minimal pairsbelow:
MC y- < *- can generally be distinguished from MC y- < *l- because the two have
different xishng connections:y- < *- tends to have xishng connections with velars
and laryngeals, while y- < *l- has xishng connections with MC sy- < *l -, MC d- < *l-,
and MC th- <*l -.
Old Chinese *- is one of a set of uvular stops that we reconstruct (modifying a
proposal by Pn Wyn 1997); details are given in section 3.1.2 (see also Sagart and
Baxter 2009). The connections with velars result from the fact that when preceded by
a presyllabic consonant, Old Chinese uvulars have velar reflexes in Middle Chinese:
Both these words are commonly written as in early texts and are presumably derived
from the same root, though at this point we cannot specify what the preinitial consonant
in *C.qrawas.
The uvular hypothesis also accounts for other distinctions among phonetic elements
that have not been previously noticed.For example, gng < kuwng work and
gng < kuwng father; prince are homonyms in Middle Chinese, and have previously
been reconstructed as homonyms in Old Chinese as well; but in pre-Qn documents, the
words written with gng and the words written with gng do not overlap at all
(see Bi Yln 2008:254257), and this fact calls for some explanation. We account for
this by reconstructing
(21)
gng < kuwng < *ko work
gng < kuwng < *C.qo father; prince.
This reconstruction also explains why gng < *C.qo was used as phonetic in these
examples, each reconstructed with a uvular initial:
(22)
wng < 'uwngH < *qo-s earthen jar
, rng < yowng < *[](r)o contain14
Paleographers believe that gng < *C.qo was the original graph for {} wng
< 'uwngH < *qo-s earthen jar (J Xshng 2010:8384). As for rng < *[](r)o
contain, in the standard script it appears to be composed of min house; build-
ing and g valley; but the original phonetic element was gng < *C.qo:the
Shuwn jiz cites a gwn ancient script form , composed of min and
gng (SWGL 3236a), and forms from bronze inscriptions confirm this (GG 6.803804);
in the early script, the shapes and are often interchangeable.
The example of rng < *[](r)o, where the original phonetic gng <
*C.qo is not apparent from the standard character, illustrates a crucial problem with
traditional reconstructions of Old Chinese based on the standard script. That script took
more or less its present form during the Qn and Hn dynasties (221207 bce and 206
bce220 ce, respectively), more than a thousand years after the beginning of the Old
Chinese period. How safe is it to use this script as a guide to Old Chinese phonology?
The evidence for Old Chinese29
Until recently the answer has not been clear:as noted in section 1.1.2 above, the only
available large samples of pre-Qn writing were the oracular inscriptions of the late
Shng dynasty and bronze inscriptions. Both kinds of texts tend to be short and formu-
laic, and the domain of their content is limited; many Old Chinese words are simply
not attested in them. This is probably why Old Chinese reconstruction in the past has
generally relied on the phonetic elements of the standard script, not the scripts of the
pre-Qn period. Even though Karlgren was familiar with the scripts of the oracular and
bronze inscriptions, he deliberately excluded from his Grammata serica recensa (GSR)
those pre-Qn characters that did not have counterparts in the later standard script (GSR,
p.5):for example, for {} rng contain, he included the character (GSR 1187a)
but not the earlier, more revealing form and did not recognize the connection of these
graphs to gng (GSR 1173a).
From the scripts of recently discovered pre-Qn documents, we now have more and
better evidence about Old Chinese phonology than was available to earlier researchers,
and we can see the importance to Old Chinese reconstruction of careful paleographic
research. In some cases, the standard script simply fails to give enough information; in
other cases it is positively misleading, because it reflects the phonology of Qn and Hn
rather than Old Chinese. We give two examples here; additional examples are given in
section 3.4below.
It has been difficult to reconstruct the syllable onset of sh < syet set up. The
Middle Chinese initial sy- ( Shn sn or Sh) shows several different patterns of
xishng connections, indicating that it had several Old Chinese sources. In our recon-
struction they include *n -, *l -, and (before front vowels) *-, plus the clusters *s.t- and
*s.t-; the reconstruction depends on xishng connections and dialect reflexes. The
examples below show our reconstructions of some cases of MC sy- and the xishng
connections on which they arebased.
MC Baxter-Sagart
(23) sh indulgent syoH *n a-s
r as, like, if nyo *na
(24) sh convey (v.) syu *l o
y enjoy yu *lo
(25) sh circumstances, setting syejH *et-s
y to plant ngjiejH *et-s
(26) sh write syo *s-ta
zh boil, cook tsyoX *[t]a
But unlike these examples, the standard character sh < syet set up has no
informative xishng connections to tell us which of the various possible Old Chinese
sources of MC sy- should be reconstructed for it.15 In an important paper, Qi Xgu
(1985) showed that in early documents, {} sh set up was written with graphs
30 Old Chinese
ancestral to y < MC ngjiejH to plant (see also Qi Xgu 1998). This discovery
allows us not only to identify the initial consonant of {} sh as *- but also to recog-
nize its etymological relationship to sh < MC syejH circumstances, setting. The
relevant reconstructions are givenbelow.
Having clarified the reconstruction of the verb sh < syet < *et set up, we can now
see that sh < syejH < *et-s circumstances, setting is simply the noun derived from it
by adding the very productive suffix *-s, one of whose main functions is to derive a noun
from a verb (see section 3.3.2.7). This in turn considerably clarifies the semantics of
sh, which has proved a challenge for translators. The Mathews dictionary (1943) defines
sh as Power; influence; authority; strength. Aspect, circumstances, conditions; but
it is difficult to tell from this list of possible English translations how the various senses
are connected semantically. From the connection with sh < *et set up, we can see
the common thread: sh is basically the way things are set up. It can refer to the way
nature has set things up, including such things as terrain and weather; although these are
beyond human control, they can be exploited to advantage by those who know how to
recognize and use them. In the Hn Fi z (third century bce), it can also
refer to the way things are set up by human agents such as rulers, so that things will hap-
pen in the desired way as an apparently natural consequence. An excerpt from Hn Fi
z:Nn sh in which both sh (the verb) and sh (the noun)
occur suggests that readers of the time would have been aware of the etymological connec-
tion betweenthem:
(28)
,
W su wi yn sh (*et-s) zh, yn rn zh su sh (*et) y.
The setup (*et-s) of which Iam speaking refers to what is set up (*et)
bymen.
In the standard script, the character gi < kojX change appears to be a phonetic
compound with j < kiX sixth heavenly stem; self as phonetic, and that is what the
Shuwn says (SWGL 1337b). But in excavated documents, we find that it is consis-
tently written with the phonetic s < ziX sixth earthly branch instead of j (Wi
Cd 2009), as in these examples from the Gudin and Shnghi Museum strips:16
(29){} gi
The evidence for Old Chinese31
For comparison, here are examples of the characters for j and s from the Gudin
and Shnghi Museum strips; in their ancient forms they are much less similar than in
their present forms:17
(30){} j
{} s
By the way, the character s < ziX sixth earthly branch is now written differently
from y < yiX cease, already, but this practice did not begin until the Tng dynasty
(J Xshng 2010:1020). Our reconstructions are as follows:
(31)
j < kiX < *k(r) sixth heavenly stem
s < ziX < *s-[] sixth earthly branch
y < yiX < *(r) cease; already
gi < kojX < *C.q change(v.)
The Middle Chinese initial k- in gi < kojX comes from *C.q-, just as in gng
< kuwng < *C.qo father, prince, discussed above. The replacement of the phonetic
element s < ziX < *s-[] by j < kiX < *k(r), with original *k-, reflects the
sound change *C.q- > *k-, which evidently had already happened by late Hn and is
reflected in the Shuwn.
The patterns of phonetic elements in the script indicate that, far from being simpler
than those of Middle Chinese, the syllable onsets of Old Chinese must have been con-
siderably more complex. In general, we want to keep our reconstructions constrained as
much as possible, so that they have some predictive value:a reconstruction that allowed
all conceivable complex combinations of consonants might be able to account for all
the existing xishng relationships, but it is equally important to make predictions about
what combinations should not occur. We would like to account for all the xishng pat-
terns that actually occur, and nomore.
But the facts point unavoidably to the conclusion that Old Chinese had some quite
complex word-initial consonant combinations. In the majority of cases, words written
with the same phonetic element at least have Middle Chinese initials with the same posi-
tion of articulation. But there is a considerable residue of characters that seem to violate
this principle. Consider the following pair ofwords:
character for elbow; in the character zhu, the cn on the right is a reduced form
of ji, and the semantic element ru flesh has been added on the left (J Xshng
2010:991; see our more detailed discussion in section 4.4.4.1).
So it appears that ji originally represented {} zhu < trjuwX elbow and was
used as a loan character to write {} ji < kjuwX nine. But it is difficult to reconcile
the Middle Chinese initials tr- and k-, which are usually not written with the same
phonetic element. We have no simple Old Chinese consonant that could be the origin
of both MC tr- and MC k-; to account for this connection, we have to reconstruct some-
thing more complex. Our reconstructionsare
Reconstructing combinations like *t-k- may seem ad hoc:by piling up initial con-
sonants, one could account for almost any xishng connection, and an unrestrained
use of this move would diminish the predictive value of the reconstruction. But in
this case both Tibeto-Burman comparisons and the morphological patterns of Old
Chinese indicate that we are on the right track. Tibeto-Burman shows both unpre-
fixed forms such as Written Tibetan khru cubit and prefixed forms such as rGyal-
rong /t kru/ elbow (Hung Lingrng and Sn Hngki 2002:661); and other
evidence suggests that Old Chinese had a prefix *t-, one of whose functions was to
mark inalienable possession (see section 3.3.2.4 below), which is appropriate for
body part terms. Asimilar e xampleis
Here, MC initial tsyh- would usually be reconstructed with a dental *t-, and this is
the traditional solution:Karlgren reconstructed ch as *ti g. But while this is con-
sistent with Middle Chinese, it ignores Mn dialect forms for tooth with velar initials,
such as Fzhu /ki 3/ and Ximn /ki 3/, which appear to represent an unprefixed
form*k (or possibly *, see section 3.1.7.2).
We will see below that both modern dialects and early Chinese loanwords in other
languages support the idea that considerable complexity must be reconstructed for Old
Chinese in word-initial position.
It has long been recognized that the Mn dialects reflect features that cannot be explained
in terms of the Middle Chinese of the Qiyn. But there are at least two other dialect
groups that also retain distinctions lost in Middle Chinese and thus also provide inde-
pendent evidence for the reconstruction of Old Chinese: these are the Hakka (Kji
) dialects of southeastern China and the Wxing or Xinghu dialects
ofHnn.
The evidence for Old Chinese33
view of the limited degree of Hakka diversity, even in the core Hakka regions, not much
earlier than that date. The Hakka features that cannot be derived from Middle Chinese
include a tonal distinction in syllables with resonant initials that largely corresponds to
the distinction between voiced and voiceless resonants reconstructed by Norman for
Proto-Mn (Norman1989).
The Wxing or Xinghu dialects are spoken in western Hnn. They are sufficiently diver-
gent from other Chinese dialects that for a time there was even a debate about whether they
were Sinitic or not (Wng Fsh 1982). Although definitely part of Chinese, like the Mn
group they show a number of features that cannot be derived from the Middle Chinese sys-
tem. Several varieties of Wxing have now been documented, but so far little research has
been done on their linguistic history. Some features appear to be shared with Mn, but they
may be shared retentions rather than shared innovations, so it would be premature to put
them together with Mn in a separate subgroup.
The main language families that received early loanwords from Chinese are Vietic,
Hmong-Mien (known in Chinese as Mio-Yo ), and Kra-Dai.
The Vietic languages include Vietnamese (VN), Mng, and a number of languages
such as Rc, Sch, and Thavng, spoken by isolated upland groups in Vietnam and Laos.
Vietic is one of the subdivisions of Mon-Khmer, the eastern branch of Austroasiatic
according to the traditional classification scheme for that family.
Extensive contacts between speakers of Vietic and Chinese probably began around
the late third century bce. Around the time of the fall of the Qn dynasty, the Chinese
general Zho Tu established the Nnyu kingdom and ruled over portions
of modern Gungdng, Gungx, and northern Vietnam; in the second century bce, this
kingdom became a vassal state of the Western Hn dynasty (Aurousseau 1923). His king-
dom probably included speakers of both Kra-Dai and Vietic languages, so it is likely that
some of the earliest Chinese loans to Kra-Dai and Vietic reflect the phonological system
of the language of Zho Tus administration.
Chinese occupation of Vietnam continued until 938 ce; during the entire period,
Chinese loanwords were continually introduced into Vietic, and especially into
Vietnamese, forming a succession of layers, each reflecting phonological characteristics
of Chinese at the time the borrowings were made. In Vietnamese, two main periods
may be distinguished through tone correspondences. As shown by Maspero (1912) and
Haudricourt (1954a), popular words borrowed in the earlier part of the Chinese occupa-
tion period exhibit these correspondences:20
The evidence for Old Chinese35
Later during the occupation period, the correspondences of the Chinese shng and
q tones were inverted.
The so-called Sino-Vietnamese pronunciation is a system of pronunciation of
Chinese characters used by Vietnamese literati after the end of the Chinese occu-
pation to read documents aloud in Chinese. Its nature is different from a layer of
loanwords in the spoken language, yet it presumably reflects the reading pronuncia-
tion of Chinese characters in use at the Vietnamese capital in the early tenth century,
during the final years of the occupation. Its tone correspondences to Chinese are
those in Table2.12. Sino-Vietnamese readings are not directly relevant to the recon-
struction of Old Chinese, and neither are the borrowings made in the later period,
but the borrowings of the earlier layer, characterized by the tone correspondences in
Table2.11, are relevant. Neither period is entirely homogeneous in terms of sound
correspondences.
(35)
*[l]o > duwng > tng bronze, copper, pHM *d copper
and
Sagart (1999c:3031) gave evidence that these two changes occurred in mainstream
Eastern Hn Chinese before79ce.
The upper dates of Proto-Hmong-Mien are more difficult to determine. Based on the
sound correspondences for onsets presented in c hapter4 of this book, Chinese loans into
Proto-Hmong-Mien sometimes appear to have slightly more archaic features than those
into Vietic or Proto-Mn. For instance, in at least one case, Proto-Hmong-Mien shows
an unaffricated reflex of an Old Chinese nonpharyngealized alveolar stop (see example
(734) in section 4.5.2.2). Since loans to Vietic were made possible by the Qn conquest
of the south, this may point to an upper date sometime before the late third century bce
for Proto-Hmong-Mien.
The Kra-Dai21 languages are spoken in southern China and southeast Asia; the center
of greatest diversity is in southern China, and the extension of the family into south-
east Asia is relatively recent. While the Kra-Dai languages, and especially the Tai and
Kam-Sui subgroups, have had sustained contact with Chinese since the establishment of
the Nnyu kingdom of Zho Tu in the Gungdng and Gungx region in about 206
bce, so far we have reconstructions only for subgroups of Kra-Dai:Proto-Kra (Ostapirat
2000), Proto-Hlai (Ostapirat 2004; Norquest 2007), Proto-Kam-Sui (Thurgood 1988;
Ostapirat 2006), and Proto-Tai (Li 1977; Pittayaporn 2009); so the current state of
Kra-Dai reconstruction has only allowed us to make occasional use of early Chinese
loans in individual Kra-Dai languages.
Among these, Lakkia (Lakkja) must be singled out. Lakkia, a Kra-Dai language
spoken in China by a group officially classified as Yo , is known to be the only
language in the family that simplifies clusters having a nasal as their second element
by preserving their first consonant and transferring nasality onto the main vowel,
The evidence for Old Chinese37
even if there was a nasal coda (Solnit 1988, Edmondson and Yang 1988,L-Thongkum
1992). Examples:
Quite apart from what can be inferred from their rhymed passages and character usages,
there are early Chinese texts that discuss language and pronunciation explicitly. The Fng
yn of Yng Xing (53 bce18 ce), the Shuwn jizof
X Shn (58147 ce), and the Sh mng (ca. 200 ce) of Li X
are well known, but the commentarial literature from Hn times onward is also a rich
source of remarks on both meaning and pronunciation. Many such comments from the
Eastern Hn period (25220 CE) are conveniently available in Coblin (1983), but many
others remain to be discovered and studied. Sometimes, remarks that seem cryptic at
first make more sense if examined in the light of recent work on Old Chinese reconstruc-
tion. Examples to be discussed later include a description of the feature we reconstruct
as pharyngealization, in the commentary by H Xi (129182) on the Gngyng
zhun (section 4.1.1), and comments by Zhng Xun (127200) in his
38 Old Chinese
(38)
*[][r]> *n > hjun > yn say
*[] > hjuwX > yu have,exist
That is, the two words are quite similar in Old Chinese pronunciation:they have the
same onset and the same vowel, and differ only in that yn ended in *-r (which later
became [n]), while yu ended in a glottal stop *-. The fact that one could be substi-
tuted for the other supports our reconstruction of the onset *[]- and the main vowel *
in both. But what about the difference in thecodas?
In several of the cases where yn is said to mean yu have, it is clear that
the next word begins with OC *n-, which suggests that the original glottal stop *- of
yu < *[] may have assimilated phonetically to the following nasal:*[]
n- > *[]n n-. For example, one of the cases cited by Wng Ynzh occurs in a quo-
tation from the philosopher Shn Do (fourth century bce), found in the com-
mentary to Xnz by the Tng-time scholar Yng Ling .23 Yng Ling
quotes Shn Do as follows:
(39)
yn nng r hi w nng, z luny.
(40)
yn yu nng r hi w nng zh rn, z b lun y.
It means that if one has ability and does harm to those who lack
ability, then there will necessarily be disorder.
The evidence for Old Chinese39
(41)
yu nng
hjuwX nong
*[] *n() ~ *n24
have ability
(42)
yn nng
hjun nong
*[]n < *[][r] *n() ~ *n
say ability
That is, the sequence *[] n- has been replaced by *[]n n-, a natural phonetic
assimilation.25 Wng Ynzh goes on to cite a second example of the same substitution
( yn nng meaning yu nng have ability) in the Xnz itself (the R
xio chapter). Similarly, he cites a passage in the Qn sh chap-
ter of the Shng sh where yn < hjun < *[][n]has been substituted for
yu before the nasal initial of *[n]a[n] > nyen > rn so, thus; (adv. suffix). Here
yn < MC hjun and yn < MC hjun are homonyms.26
In the cases discussed so far, the substitution of yn < *[]n (perhaps from
earlier *[][r]) for yu < *[] has occurred before MC n- or ny-, both of which
(usually) reflect OC initial *n-. But such substitutions occur before other consonants
as well. In another passage from the Qn Sh chapter of the Shng sh, Wng Ynzh
says that thephrase
(43)
yn li
hjun loj
should be understoodas
(44)
hu li
hwok loj
there are some whocome
The characters yu and hu are similar in pronunciation and presumably come from
related roots; they are frequently interchanged in early texts. Our reconstructionsare:
But why would yn < *[]n (< *[][r]) be substituted for either yu or
hu before li? As explained in section 4.5.2.4, we believe that li originally had
a preinitial *m-, which helps account for the fact that li is the phonetic element in
the character mi < meak < *m-rk wheat.
We can then understand the substitution of yn for yu or hu as an
assimilation to the nasal preinitial of li < *m.rk. In fact, all the examples of this
substitution that we know of happen before words with nasal initials or preinitials.27
These substitutions give us information about both main vowels (* in both cases) and
preinitials.
Note, however, that this interpretation presupposes that the *-r coda of yn has
already changed to [n]. The examples cited apparently originate in Hn-dynasty texts;
we would predict that such substitutions of for should probably not occur in
pre-Qn excavated documents, written at a time when *-r and *-n were probably still
distinct.
It is likely that the abundant Chinese commentarial literature includes many more
comments like these that may be helpful in reconstructing the Old Chinese phonologi-
cal system or in choosing reconstructions for particular words. As our reconstructions
become more precise, we should be able to make better sense of these comments; so
they are not to be neglected as a form of evidence for Old Chinese reconstruction.
2.7Tibeto-Burman
We accept that Chinese (= Sinitic) and the languages called Tibeto-Burman are part of
a larger family called Sino-Tibetan. Our approach more or less presupposes that Sinitic
itself is a valid taxon within Sino-Tibetan, but the subgrouping of the family as a whole
is not yet clear; in other words, we are not certain whether Sino-Tibetan splits cleanly
into these two branches or whether its phylogeny is more complex. In either case, prog-
ress in understanding either Sinitic or Tibeto-Burman is likely to help us better under-
stand the otherjust as in Indo-European, knowing the location of accents in Greek
and Sanskrit helped explain consonant alternations in Germanic (the Grammatische
Wechsel resulting from the operation of Verners Law). It is perfectly legitimate to take
hints from Tibeto-Burman (or anywhere else) when formulating hypotheses about Old
Chinese; as a matter of fact, Starostins hypothesis that OC *-r occurred as a syllable
coda (contrasting with both *-n and *-j), which we accept here, was initially suggested
by comparisons with Tibeto-Burman words with final [r](Starostin 1989:338341; see
section 5.5.1).
But it would be a mistake to use Tibeto-Burman evidence to test hypotheses about
Old Chinese. The fact that [r], [n], and [j] codas contrast in some Tibeto-Burman lan-
guages may raise the question of whether Old Chinese had a similar contrast, but only
evidence from within Chinese (not excluding words in other languages borrowed
from Chinese) can answer this question. Similarly, the fact that some Tibeto-Burman
The evidence for Old Chinese41
Old Chinese words had an obligatory main initial consonant and sometimes a preinitial
(discussed in detail in c hapter4). Our reconstruction includes a number of innovations
in both main initials and preinitials.
We reconstruct significantly more forms with preinitial elements than in earlier
reconstructions. In his Archaic Chinese, Karlgren already reconstructed a small number
of forms with initial consonant clusters in order to account for xishng connections, as
in the following examples:
(46) Karlgrens *pli t writing brush (> pit > b; our *p.[r]ut);cf.
Karlgrens *bli wt law, rule (n.) (> lwit > l; our *[r]ut)
(47) Karlgrens *mk black (> xok > hi; our *mk);cf.
Karlgrens *mk ink (> mok > m ink; our *C.mk)
(48) Karlgrens *pli m kind, class (> phimX > pn; our *pr[]m);cf.
Karlgrens *bli m look down at (> lim > ln; our *(p.)rum)
However, when one takes into account modern dialects and early Chinese loan
words into other languages, it becomes clear that many more forms must be recon
structed with complex syllable onsets.For example, Karlgren reconstructed w
roof tile simply as *ngwa, Baxter (1992) as *ngraj (equivalent to *raj
in our current notation); but we must reconstruct initial *C.- (with unspecified
preinitial *C) to account for upper-register tones in Mn and Hakka dialects and in
an early loan in Vietnamese:
(49) *C.ra[j] > ngwaeX > w roof tile; pMn *h-; Mixin Hakka
/a3/; cf. VN ngi [i B1] with high-registertone
It is also clear that there was considerable dialect diversity in the treatment of pre
initial elements, as we see from the following adjacent entries in the Shuwn for writ
ing brush (see section 4.4.4.4 for more examples):
An overview of the reconstruction 43
Note that part of the variation involves whether the preinitial is treated as tightly or
loosely attached. In these circumstances it is sometimes necessary to reconstruct more
than one form for certainetyma.
The main innovations in our reconstruction of syllable onsets are summarized in the
remainder of this section.
Type-A syllables in Old Chinese are those that give rise to the division-I, division-II, and
division-IV syllables of Middle Chinese;1 all other syllables, traditionally designated as
division-III, are type B (the terminology is due to Pulleyblank 19771978). The syllables of
Middle Chinese are divided roughly half-and-half into the two types. In his Archaic Chinese
reconstruction, Karlgren reconstructed type-B syllables with a prevocalic semivowel *-i -
(medial yod) and reconstructed type-A syllables without this element. This solution was
followed by Dng Tngh (1948); Li (1971) and Baxter (1992) did the same, but wrote the
medial yod as *-j-. This solution came to be regarded as unsatisfactory for various reasons,
and alternatives were proposed (see section 4.1.1). In our reconstruction we follow Norman
(1994) in reconstructing the type-A syllables with pharyngealized onsets:thus
(51) *ka > kang > gng guiding rope of net (division-I, typeA)
*ka > kjang > jing boundary (division-III, typeB)
(52) *kra > kaeng > gng change (v.) (division-II, typeA)
*[k]ra > kjaeng > jng hill; capital city (division-III, typeB)
(53) *te > tengX > dng cauldron (division-IV, typeA)
*te > tsyengX > zhng arrange; orderly (division-III, typeB)
This reconstruction helps to explain two important facts that were unexplained under
previous hypotheses:(1)the tendency of vowels to be lowered in type-A syllables, and
(2)the resistance of initials in type-A syllables to palatalization. See section 4.1.1 for
further discussion.
For reasons set out in Sagart and Baxter (2009) and already briefly broached in
section 2.3 above, we have added a set of uvular and labiouvular initials to the
44 Old Chinese
A theory that Old Chinese had contrasting velars and uvulars has to explain why
Middle Chinese velars and laryngeals, their normal reflexes, frequently co-occur in
phonetic series. Pns explanation is that the two kinds of sounds are phonetically
close enough to be written with the same phonetic element. In our view, in such
cases, the Middle Chinese velars are the regular reflexes of Old Chinese uvulars
with certain preinitials, as shown by examples in which Middle Chinese words with
velar and laryngeal initials share the same root:for example, 'jaengX > yng
shadow (n.) and kjaengX > jng bright; image. In Sagart and Baxter (2009),
we identified the condition for the evolution of OC uvulars to MC velars as being
the presence of a loosely attached preinitial:our reconstructions were *qra and
*C.qra for and . Here, for reasons detailed in note 24 of chapter4, we
modify this claim and propose that the conditioning factor was a tightly attached
preinitial:we now reconstruct as *qra > 'jaengX and as *C.qra > kjaengX.
Areconstruction that would assign a velar initial to the latter and a uvular one to the
former would not be able to express the root they share. For further discussion and
examples, see section 4.4.5.1.
Here again, in practice, the usefulness of xishng connections in determining the
nature, velar or uvular, of a word with a Middle Chinese velar initial is mitigated by the
fact that characters created after the change of Old Chinese uvulars to velars follow
ing tight preinitials may well have received velar phonetics even though their original
Old Chinese initial may have been uvular, or vice versa. As a result, the testimony of
xishng connections must be weighed against other types of evidence:textual, paleo
graphic, comparative, etymological,etc.
Third, our view of the development of OC *- in type-B syllables differs from
Pns. According to him, type-B *- evolves to MC hj- (the initial Y sn =
Yn). Since the majority of Middle Chinese words with this initial are labialized
(hku ), this line of reconstruction entails treating labialization as secondary
in many words with the initial Y sn (= Yn). In Pns view, this is due
to a tendency to labialization inherent in consonants articulated with the back of the
tongue, parallel to the tendency of back vowels to be rounded. Thus Pn reconstructs
yng < MC hjwaengX long (time) as *ra and y < MC hju go; at as *a;
he supposes that these forms acquired secondary labialization to *ra, *a before
developing their Middle Chinese reflexes (Pn 1997:20). In our view, with very few
exceptions, the Old Chinese source of MC hj- is labiouvular *-, and the labializa
tion in most MC words with hj- is original. Our reconstructions for yng and
y are *[]ra and *(r)a, respectively. We also reconstruct a nonlabialized *-,
but based on xishng and word-family contacts, we think that this initial evolves to
MC y-, the traditional initial Y s = Y (Sagart and Baxter 2009). See sec
tion 4.3.3 for details and examples.
Finally, we claim that OC *q- and *- preceded by nasal prefixes *N- or *m-
merged with OC *-, evolving to MC ng-. This is detailed in sections 4.4.1.2, 4.4.1.3,
4.4.2.2, and 4.4.2.3. The uvular and nonuvular sources of MC ng- can sometimes be
46 Old Chinese
distinguished in the script. This is the case of w < *C.a five and w <
*[m].qa seventh earthly branch; when used as phonetics (apart from some charac
ters of late origin); see section 4.4.2.2 for discussion.
The principal Middle Chinese reflexes of the uvular and labiovular consonants we
reconstruct, when not preceded by any preinitials, are shown in Table3.1.
We now turn to the main new hypotheses concerning preinitials.
Norman reconstructed a series of voiceless resonants *mh-, *nh-, *lh-, etc. for Proto-Mn
to explain distinctions in tonal development; in the Northern Mn dialects,2 *lh- also has
a sibilant reflex different from that of *l-. These do not correspond to our Old Chinese
voiceless resonants *m ()-, *n()-, *l ()-, etc., which have different reflexes; rather, they
reflect Old Chinese voiced resonants with a tightly attached voiceless preinitial conso
nant:for example, *k.r- becomes Proto-Mn *lh-. In some cases, the particular preinital
can be identified from other evidence (such as early loanwords or evidence from the
script); otherwise, we write it as *C. See the discussion in sections 4.4.4.4 and 4.4.5.4.
It was mentioned in section 2.4.1 that Norman (1973, 1974a) reconstructed a set of
softened stops for Proto-Mn, which have distinctive segmental reflexes in Northern
Mn dialects. Norman (1986) suggested that these might have originated as prenasalized
obstruents, but we find this solution unsatisfactory and instead attribute the softening to
voicing and/or lenition of the main syllable initial when preceded by a loosely attached
presyllablethat is, in intervocalic position. For example:
(57) Old Chinese Normans Proto-Mn Middle Chinese
*C.t- > *-t t-
*C.d- > *-d d-
An overview of the reconstruction47
As also mentioned in section 2.4.1, some Mn dialects show a contrast between aspi
rated and unaspirated initials corresponding to the voiced stops and affricates of Middle
Chinese; Norman accordingly reconstructed an aspiration distinction in voiced stops
and affricates in Proto-Mn. Our hypothesis is that the aspirated reflexes are from Old
Chinese tightly attached preinitials other than*N.:
Our explanation for the coexistence in Mn dialects of unaspirated and aspirated stops cor
responding to the voiced stops of Middle Chinese is that Mn dialects were subject to two
waves of devoicing. The early devoicing affected Old Chinese voiced stops and voiceless
stops preceded by *N, which had merged with the voiced stops by the time of Proto-Mn.
This first devoicing produced voiceless unaspirated reflexes but did not affect voiced initials
with tightly attached preinitials. Then these preinitials were lost, exposing the voiced syllable
initial, which became breathy and underwent a second wave of devoicing that produced aspi
rated reflexes; for details, see Table4.9 in section 4.2.1.1.
In both native words and words borrowed very early from Chinese, Vietnamese shows
a phenomenon of spirantization of initial obstruents analogous to the softening in
Northern Mn dialects: where Vietnamese shows initial spirantization, other closely
related Vietic languages have a presyllable that has been lost in Vietnamese, and it is
widely agreed that the Vietnamese spirantization applied to consonants when they were
in intervocalic position. For example:
Where such presyllables occur in loans from Chinese, we take this as evidence of a
presyllable in Old Chinese aswell:
The earliest Chinese loanwords were probably into Proto-Vietic rather than into
Vietnamese itself, which would not yet have become a separate language. It appears
that all presyllables in these early Chinese loanwords were treated in Vietnamese as if
they were loosely attached.
A number of early Chinese loans into Hmong-Mien show prenasalization or the reflex
of prenasalization in those languages; we take this as evidence of a nasal preinitial in
the Chinese source:
The reconstruction of *N-r()- and *m-r()- solves a number of puzzles about initial
consonants:
Middle Chinese Tsy- initials regularly reflect Old Chinese nonpharyngealized alveolar
initials *t-, *t-, *d-, and *n-; another source of Tsy- initials is the palatalization of non
pharyngealized velar initials before the front vowels *i and *e. However, there are also
contacts between velars and palatals in syllables with other vowels, which have long
resisted satisfactory explanation.3 We now attribute this palatalization to a preinitial *t
before the main velar initial:
(63) *t-[k]() or *t- > tsyhiX > ch front teeth, Proto-Mn *khiB.
The fact that stop preinitials are attested in early loans to Vietic lends plausibility to this
solution. See section 4.4.4 for discussion.
TABLE3.2 Rhymes reconstructed for Old Chinese
*- *-j *-w *-n *-m *- *-r *-t *-p *-k *-wk
*i -ij -iw -in -im -i -ir -it -ip -ik -iwk
*u -u -uj -un -um -u -ur -ut -up -uk
* - -j -n -m - -r -t -p -k
*e -e -ej -ew -en -em -e -er -et -ep -ek -ewk
*o -o -oj -on -om -o -or -ot -op -ok
*a -a -aj -aw -an -am -a -ar -at -ap -ak -awk
50 Old Chinese
3.2Rhymes
Old Chinese words consisted of a root plus possible affixes. Word roots were either
monosyllables consisting of a full syllable () or disyllables consisting of a full syl
lable preceded by a minor syllable (), that is, having the structure .. Minor syl
lables were reduced in comparison with full syllables, in terms both of the number
of structural positions they allowed and of the number of phonemes that contrasted
in each position.
A full main syllable included five structural positions, each of which could be filled
by different sets of phonemes (Figure3.1):
Onset Rhyme
The first position, Ci (initial), was obligatorily filled. All Old Chinese consonants
except *j and *w could occur in it. Afull table of the possible initial consonants is given
in Table4.1 in c hapter4.
The second position, Cm (medial), was either left empty or filled by *r. The *r, if
present, could be part of the root, or it could be an infix *<r>, which occupied the same
position. We write infixed *<r> between angled brackets to distinguish it from root *r
in the medial position.
The third position, V (vowel nucleus), which formed the syllables peak, was obliga
torily filled by one of the six vowels *i, *e, *, *a, *u, and *o. The phonetic quality of
the vowel we write as * is uncertain:it may have been a mid central vowel [], a high
central vowel [], or even a high back unrounded vowel[].4
The fourth position, Cc (coda), was either left empty or filled by a consonant, which
could be any one of the following:*m, *n, *, *r, *j, *w, *p, *t, *k, or *wk. Note that
there was no coda*w.
The last position, Cpc (postcoda), was either empty or filled by a glottal stop. The
glottal stop could only follow a sonorant (one of the six vowels or a sonorant coda):there
were no sequences like *-k, *-p, *-t, or *-wk. Examples of unaffixed monosyllabic
words follow:
In addition to the main syllable, certain words were preceded by extrasyllabic seg
mental material, forming a minor syllable or presyllable before the main syllable, and/or
followed by a final *-s. Our provisional hypothesis is that final *s was always a morpho
logical suffix, so we write it with a preceding hyphen. As for presyllables, sometimes
we can identify them as prefixes, sometimes not. Where we cannot, the presyllables
could be (1)as yet unrecognized prefixes, (2)known prefixes with unknown functions,
or (3)part of the root. When we cannot tell whether a preinitial is a prefix or not, we
separate it from the main syllable with a period instead of a hyphen and treat it, at least
provisionally, as part of the root. Thus we write a hyphen after the preinitial *s-in
because we take the *s- to be the valency-increasing prefix (see section 3.3.2.3 below)
applied to the basicroot
Cpi Vpi
Further research may show that some presyllables now treated as part of the root
were synchronically affixes in Old Chinese; others may have functioned as affixes
at some earlier stage but must be treated synchronically as part of the root in Old
Chinese. An analogy would be words such as English believe and German glauben
believe. Originally, these are from the same Proto-Germanic root *laub- with dif
ferent prefixes, but synchronically they must now be considered part of the root.
Still other presyllables may never have been prefixes atall.
In minor syllables there were two structural positions:Cpi (preinitial consonant) and
Vpi (preinitial vowel) (Figure3.2):
The first position, Cpi, must be filled by a consonant. The inventory of possible Cpis
was limited:we have evidence for *p, *t, *k, *r, *s, *m, and *N; of these, *N may be
regarded as a positional allophone of either *n or *. The evidence for *r as a preinitial
consonant is very limited. We reconstruct preinitial*rin
to account for the Middle Chinese initial l- in l < luX, which normally goes back to
OC *r-, and for the use of ngjo > y fish (n.) as phonetic.
At times, the comparative evidence indicates that a preinitial consonant was present, but
the consonant cannot be identified; in such cases we write it as *C. Still other consonants
may have occurred in this position; we cannot provide a definitive list at this time. However,
there is no evidence that minor syllables could include pharyngealized consonants.
The second position, Vpi (preinitial vowel), was either left empty (in which case the
minor syllables peak was a consonant),5 or filledby*.
We call preinitials with an empty Vpi position tightly attached and those in which
the Vpi position is filled with * loosely attached. Tightly attached preinitials formed
tight clusters with the major syllables obligatory Ci consonant, and these clusters were
simplified in different ways in Middle Chinese, while loosely attached preinitials were
lost in Middle Chinese, as a rule (at times influencing the major syllables initial before
disappearing, as in NorthernMn).
Minor syllables did not occur freely outside of feet formed with a following major
syllable; however, a few highly common function words are structurally identical with
minor syllables with // and can be considered as such. Examples follow:
An overview of the reconstruction 53
Root
()
Onset Rhyme
Cpi Ci
(e)
(r) V (Cc) ( )
One notes that *t did not bear stress in the Odes (Kennedy 1939). This behavior
is normal if one thinks of the above forms as minor syllables. Nevertheless we treat
them here as a particular type of full syllables.
The structure of Old Chinese roots is summarized in Figure3.3.
Here are some examples of disyllabic roots with tightly attached preinitials:
(72)
*t.l u[n]> tsyhwen > chun stream, river (the Middle Chinese final
is irregular; we would expect tsywin)
*k.to > *to > tsyowngX > zhng seed; borrowed into
Proto-Vietic as *k-co
*C.pran > paenX > bn plank,board
Here are some examples of disyllabic roots with loosely attached preinitials:
There were no tones in Old Chinese (see the discussion in section 5.1). Major sylla
bles were stressed, and any minor syllables were unstressed. This gave disyllabic words
an iambic, weak-strong, rhythm. In verse, disyllables and monosyllables alike counted
for one foot:a four-foot verse could consist of any combination of monosyllabic and
disyllabicwords.
3.3.2 AFFIXATION
In many Old Chinese words, affixes can be identified. Old Chinese had several prefixes
and suffixes and one infix. All were derivational; some were quite productive. Prefixes
were attached before the roots first segment, and suffixes after the last one. The infix
was inserted between the Ci and the V in a major syllable, that is, in the syllables
54 Old Chinese
medial position. We separate suffixes and prefixes from roots by means of a hyphen, and
isolate an infix within a root by using angled brackets <>. Most prefixes were added to
monosyllabic roots (see examples in the detailed discussion below), but there are some
examples of prefixed disyllabicroots:
There are also examples of doubly prefixed roots (see discussion in section4.6):
The list of Old Chinese affixation processes is still an open one, but some are already
well supported. We list thembelow.
Prefixes were consonants, each having short (*C) and long (*C) variants that appar
ently had the same morphological functions. The short variants behaved phonologically
as tightly attached minor syllables, the long ones as loosely attached ones. The evidence
for these variants is comparative (see section 4.2 below). The conditions of occurrence
of short and long variants are not understood; for the moment we treat them as free vari
ants. Short variants were more common. The syllabic status of words with two prefixes
is uncertain, as is that of prefixed disyllabicroots.
For further discussion and examples, see Sh Jir [Sagart] and Bi Ypng [Baxter]
(2010) and Sagart and Baxter (2012).
See also the double-prefixed form *m-s-t > dzong > cng additional floor, field,
discussed in section4.6.
Prefix *m2- occurs in names of human bodyparts:
For more discussion of the *s- prefix in Old Chinese, see Sagart and Baxter (2012).
(105) *t-nip, *t-n[i]p > tsyip, nep > zh afraid, to fear (for initial develop
ments, see section 4.4.4.4)
(106) *n<r>ep > nrjep > ni unable towalk
*t-nep > tep paralysed, unable tomove
(114) *m-riw-s > mjiwH > mi lie, error (< twist the truth)
*k-riw > kjiw > ji tie around, strangle
(115) *a > ngang > ng high, lifthigh
*k-a > kang > gng lifthigh
(116) *[r][]p-s (dial. >) *rup-s > lwijH > li tear(n.)
*k-rp > khip > qweep
Suffix *-s3 derives verbs of outwardly directed action out of verbs of inwardly
directed action or stativeverbs.
When two Old Chinese words share the same root and have different affixes, they can be
said to belong to the same word family. In this section we illustrate the notion of word
family with three examples.
The root *t ascend occurs bare in theverb:
(138) *s-t > tsong > zng to increase (v.t.) (< cause to ascend).
Out of this last formwhether before or after the shift of *s-t- to ts- is not known
a stative verb increased, augmented, doubled, high was derived through prefixation
of*N-:
We reconstruct *m-s-t with prefixed *m1c-, an additional floor being then conceived
of as a means of increasing the space within a house. In the Mn dialects this word has
60 Old Chinese
displaced *li > den > tin as field (Norman 1996:31):Ximn /tsan 2/, Chozhu
/tsa 2/, Fzhu /tsei 2/, Jinu /tsai 5/, Yngn /tsi 2/. Note the aspirated ini
tials, implying pMn *dzh-, the expected reflex of OC *m-s-t-. The semantic evolution
appears to have been from means of increasing the size of a field, to added terraced
field above an existing field to terraced field to field.
A root *pan divide is in evidence in a set of words with the core meaning to
divide. The bare root does not occur as a word, but its nominal derivative with suffixed
*-s1 is seenin
(141) *pan-s > panH > bnhalf
The original verb root was also given a distributive meaning by means of infixed
*<r1>:
(142) *p<r>an > paen > bn divide, distribute
From a related root (see section 3.3.4 below), we have another word family with the
verb *pan-s > phanH > pn divide and a derived noun with the instrumental
prefix*m1c-:
The root *tru center is a nominal root, appearing in its bare formin
(144) *tru > trjuwng > zhng center
Out of this noun a verb was derived by means of suffix*-s2:
(145) *tru-s > trjuwngH > zhng hit the center
Further, through the addition of the *N- prefix, a new verb, stative and intransitive, was
derived:
(146) *N-tru-s > drjuwngH > zhng middle (of three brothers, of three
months) (< placed in the center); cf. pHM *ntro A center,
middle
The nasal prefix we suppose is directly attested in the Proto-Hmong-Mien form, even
though one would expect tone C corresponding to Chinese qshng, rather than toneA.
3.3.4 RELATEDROOTS
It is not uncommon for two independently reconstructible roots to have similar mean
ings and similar, but not identical, pronunciations. Thus next to root *t ascend,
discussed in the preceding section, we have a root *t rise:
The word was used in particular with reference to steam and has come to be used as
a transitive verb to steam. The root occurs with *s-in
An overview of the reconstruction61
(148) *s-t > sying > shng ascend (e.g., a reception hall; a chariot)
Out of an *s-suffixed form of the verb to steam, the name of a steaming instrument
was derived through the addition of prefixed *S-:6
Thus we have two roots, *t and *t, differing in the pharyngealization of the
initial. There are other cases where roots that are semantically very similar contrast in
the presence or absence of aspiration:
Another such example is the pair *pan-s > panH > bn half and *pan-s>
phanH > pn divide in examples (141) and (143)above.
In still other cases, there appears to be an alternation between final *- and final
*-k:e.g., *ta and *tak, both to place:
Phonological alternations such as these are not productive in Old Chinese. We take
them to be what remains of morphological alternations that were once productive in
languages ancestral to Old Chinese. They are similar to English sets such as whole,
heal, and health:
(152) whole < Old English hl sound, healthy < Proto-Germanic *hail-az
heal < Old English hlan to make whole or sound < Proto-Germanic
*hail-jan
health < Old English hl soundness of body < Proto-Germanic
*hail-ia
These all come from the same Proto-Germanic root *hail- meaning sound in body,
but in modern English they must probably be considered three synchronically different
roots that are related only historically.
Similarly, we may expect to find explanations for Old Chinese related roots such as
*ta and *tak in a wider comparative context. For example, alternations between final
*- and final *-k in verbs are reminiscent of some aspects of the alternation between
stem Aand stem B in Kuki-Chin languages (for instance So-Hartmann 2009:71 for Daai
Chin). More work is needed in this domain.
62 Old Chinese
As outlined in section 2.3, one of our main motivations for developing a revised recon
struction has been the discovery of large numbers of pre-Qn documents whose script is
more relevant to Old Chinese reconstruction than the standard script on which previous
reconstructions have largely been based. In using these pre-Qn documents, we have
been led to a new understanding of the nature of the pre-Qn script, which is relevant to
the problem of reconstructing Old Chinese.
For one who knows modern Chinese, it is easy to think anachronistically about the
early Chinese script and to imagine that it is essentially no different from the script
in use over the last 2,000years. To learn the standard script, it is necessary to memo
rize several thousand characters, whose connection to pronunciation has become less
and less direct as time has passed. It is important to remember that the standard script
has been maintained with very strong institutional and cultural support, which was not
present to the same degree in Old Chinese times. The standard script has been in use
over a vast and linguistically diverse area, and its mastery has been one of the keys to
government employment. It has been a centrally defined standard, supported by widely
distributed reference works and other texts. The circulation of these texts was eventu
ally facilitated by the use of paper (much less expensive than earlier writing materials),
by printing, and by the development of a flourishing commercial market in books. The
cultural prestige associated with the lettered class has reinforced these tendencies. All
these factors have worked to maintain the standard script with minimal changes, in spite
of the fact that it is difficult to learn and inconvenient in many respects.
None of these institutional and cultural supports were present in the pre-Qn period.
There were widely shared conventions, of course, or else the script would not have
been able to function efficiently as a means of communication, but there was no cen
trally defined standard. Literacy would have been an important qualification for some
jobs but was probably less important as a key to prestige and high-status employment
than it became later. Political fragmentation probably meant that the circulation of texts
was generally confined to smaller areas. Books were bulkier and more expensive than
they became later, so compared with imperial times, the familiarity with existing texts
probably presented less resistance to innovation than it became later. The body of texts
regarded as canonical was also much smaller, and what was canonical was probably
more often recited and listened to than written and read. There was more variation from
scribe to scribe (as we can clearly see from recently excavated Warring States docu
ments); and finally, calligraphy, with its well-defined styles and models for emulation,
was less developed as a distinctart.
The pre-Qn script was thus more fluid and less resistant to adaptation and modifica
tion than the script of the imperial period. Still, the writing system had to function for
communication among large numbers of people, and the script must have been shaped
by the functions that it served. Scripts are synchronic systems, with some of the same
kinds of constraints that shape spoken languages. What, then, were the constraints that
shaped the structure of pre-Qn writing?
An overview of the reconstruction 63
First, the script had to be learnable. Just as no spoken language uses more than about
a hundred distinct phonemes, there must have been a practical limit, imposed by memory
limitations, on the number of signs that had to be learned and remembered. Without the
powerful institutional supports that arose later, the number of elements that needed to
be memorized outright was probably considerably smaller than in the imperial system.
Second, all other things being equal, there must have been a limit to the complexity
of the relationship between the elements of the script and the corresponding elements
of the spoken language. If linguistic changes complicated this relationship, the pre-Qn
script was relatively free to adapt in response. If, because of a change in pronunciation,
a phonetic element became less suitable to write a certain word, it was likely to be
replaced by another; and the force of resistance to such modifications must have been
considerably weaker than it later became.
As an example, we can take the history of ways to write the word {} wn hear.
In oracle-bone and early bronze inscriptions, the character used has a person and anear:
(153)
The person is said to be kneeling and holding his hands to his face (Y Xngw 1941,
quoted in GG 9.585). In excavated Warring States documents, {} wn is commonly
written with the character hn dusk, as a phonetic loan; our reconstructionsare:
(154) *mu[n]> mjun > wn hear(v.)
*m u[n]> *xun > xwon > hn dusk,dark
The use of *m u[n]dusk to write *mu[n] hear is based on the phonetic similar
ity of *m- and *m - and the presence of the same rhyme *-u[n] in both words. We also
find hn with r ear added; this is the form , which the Shuwn lists as an
ancient character (SWGL 5356a).
But at some point, the initial *m - of hn changed to a fricative [h]or [x] (> MC x-),
disturbing the phonetic similarity with {} *mu[n]; and in the Qn script (e.g., in the
documents from Shuhd ; see J Xshng 2010:877) we find a new phonetic
compound, with mn gate, door as phonetic:
(155) *mu[n]> mjun > wn hear(v.)
*m[r]> mwon > mn gate,door
Early in the Old Chinese period, mn < *m[r]would not have been a proper pho
netic for wn < *mu[n], because its main vowel was different (* *u); the codas
were probably different as well (*-r *-n), although this is not certain. But as the
result of later sound changes, final *-r changed to *-n in most dialects, and the contrast
between *-un and *-n was lost after labial onsets, so at that point mn became a
suitable phonetic for {} wn hear; and because of the late change of *m() to *x(),
hn < xwon < *m u[n] would have become a less suitable phonetic. This example
shows how the early script, with fewer institutional barriers to innovation, was able to
adapt to these sound changes, replacing the phonetic element hn with the phonetic
elementmn.
64 Old Chinese
But even in the pre-Qn period, there were forces that would have resisted excessive
fluidity in the script. Written texts were intended to be read and used by different people
at different points in space and time. Records from earlier periods still needed to be read;
and texts did circulate from place to place within certain geographic domains. But the
domains were smaller than in imperial times:the reach of government administration
was not as far, and wide circulation of texts must have been less common than it later
became. And in fact, paleographers have identified significant regional differences in
the script of pre-Qn documents.
Not only would the total number of written signs have been limited, but there
would have been a tendency to avoid having to make graphic distinctions that were
too subtle:if two graphic elements were similar in shape, they sometimes merged
altogether, becoming a single element with more than one functionor perhaps they
were never clearly distinguished in the first place. An example of this in the mod
ern script is the fact that the shape of the abbreviated form of ru flesh has
merged with yu moon, even though the two elements have very different func
tions; context provides enough redundancy that no confusion results. Similarly, in
the pre-Qn script, yu moon and x night are not clearly distinguished (J
Xshng 2010:565).
We can assume that the structure of the pre-Qn script was shaped by a kind of
homeostasis among these various forces. The graphic signs could not become so numer
ous as to be too difficult to learn, remember, produce, and recognize, but there must
have been enough of them to represent the necessary distinctions. Signs whose connec
tion to the spoken language became too indirect tended to be replaced by others whose
structure was more transparent. Yet modifications of the script cannot have been too
frequent or capricious, or else documents written at one time and place would become
too difficult to read at another.
The system was word-based in the sense that a given graphic token in a given
text represented a word-length unit, rather than a single phonetic segment as in
alphabetic systems. But from the standpoint of how graphs were learned, remem
bered, retrieved, and read, we suggest that the system was primarily syllable-based,
not word-based. Rather than memorizing something on the order of 5,000 to 10,000
graphs roughly corresponding to morphemes, which is what is required for literacy
today, a person learning a variety of the pre-Qn script would have learned a set of
about a thousand graphic elements, each of which could be used as a phonetic sign
to represent a certain type of syllable. Many of these elements originated as picto
grams, and there were a certain number of graphs constructed by combining graphs
based on their denotations (the so-called huy semantic compounds). But
it was the availability of a set of about a thousand phonetic elements that gave the
system the power to represent the full range of vocabulary of the spoken language.
When a phonetic element by itself would be ambiguous, it could be supplemented
by adding a semantic element to the graph, as r ear was sometimes added to
hn < *m u[n]when it represented {} wn < *mu[n] to hear (see above).
An overview of the reconstruction65
In most cases, each phonetic element represented a type of syllable with a certain
position of articulation in the onset, a certain main vowel, and a certain coda:for
example, the element p skin came to represent syllables of the general shape
*P(r)aj; here *P represents any oral labial stop, pharyngealized or not. There might
or might not have been a prevocalic *-r-, and there might or might not have been a
presyllable. In recently excavated documents, according to Bi Yln (2008:127),
the graph , with no added elements, is used to represent all the followingwords:
Similarly, the graph stood for syllables of the type *P(r)ak; according to Bi Yln
(2008:187), this graph was used, by itself, to represent all thesewords:
These two phonetic elements seem to have been more or less interchangeable:the word
now writtenas
is usually written in the Gudin and Shnghi Museum texts as *(r)aj. (In fact,
{} *(r)aj-s duty; justice is just the nominal derived from {} *(r)aj proper;
should by means of the suffix *-s.) But we also find *(r)aj-s duty; justice written as
66 Old Chinese
At the same time, there were some syllable types that apparently had no proper pho
netic element of their own, and for which the usual criteria for a phonetic match had to
be relaxed.For example, there seems to have been no separate phonetic for syllables of
the form *n()er, as in theword
(160) *ner > nej > n pickled meat with bones in it (also read *ner >nye).
So the phonetic element nn < *nar, whose usual function is to represent syllables of
the form *nar, was used, together with flesh, to represent {}*ner.
Similarly, the element dn < tan < *C.tar single normally represents syllables
of the shape *Tar. But the Sh j:Xingn li zhun gives this
disyllabic word as the name of a kind of horse ridden by the Xingn:
The commentaries tell us this word should be read as MC ten-hej; from this we can
reconstruct the word (probably borrowed from the Xingn language) as *ter.ge for
(late) Old Chinese. Here, the character dn < tan < *C.tar was used to write the syl
lable *ter, in spite of the difference in main vowel, because no better phonetic element
was available.
We say that the phonetic elements in such examples are used faute de mieux, for
want of a better one:they reflect the fact that although the use of a set of phonetic ele
ments for syllable types was basic to the writing system, the coverage of the set of pos
sible syllables was uneven. The precision with which phonetic elements represented
syllable types thus varied from one region of the space of syllable types to another,
and the criteria for a phonetic match were not always consistent.For example, we find
An overview of the reconstruction67
a larger degree of latitude in the use of phonetic elements for syllables with final *-m
and *-p, which are relatively infrequent, than for syllables with final *- and *-k, for
which many choices were available. In general, however, the practice was to pick the
most suitable phonetic element that was available.
There was a tendency over time to make the system more precise:for example, in the
early script it appears that yun was used for both *Qan and *Qen syllabletypes:
Eventually, though, yun seems to be used for *Qan and hun for*Qen.
Such developments tended to make the system of phonetic elements more precise
than it may have been at the earliest stage. But there were also forces working to reduce
the precision of the system, as the script came to be more standardized from the Qn
unification (221 bce) onward. Changes in pronunciation would sometimes make words
written with the same phonetic element sound less similar than they had earlier, but now
the script was much less likely to adapt to changes in pronunciation (as it had earlier by
writing {} wn hear with mn as phonetic rather than hn).
The result was that the connection between pronunciations and phonetic elements
became less and less direct as time passed, resulting in the system we have now.
Moreover, when new phonetic compounds were created, the criteria for phonetic simi
larity would have been looser if they were defined by analogy to already existing pho
netic compounds. As a result, characters of late origin are generally less informative
than earlier ones about pronunciationespecially about Old Chinese pronunciation.
{4}
We define a word onset in Old Chinese as the part of an Old Chinese word that precedes
the main vowel:it includes
1. any presyllabic material that may be present, which can contain up to two
consonants, including prefixes, with or without the vowel*;
2. the initial of the main syllable;and
3. medial *-r- (which in some cases is an infix), if present.
The only obligatory element within an onset was the initial of the main syllable. The
consonants that could occupy the main-syllable initial position (with or without a pre-
initial) are shown in Table4.1below.
In the next section we discuss some of the main developments affecting the evolution
of these consonants.
4.1.1 PHARYNGEALIZATION
Middle Chinese syllables can be divided into two main types, for which Pulleyblank
(19771978) coined the terms type A and type B. In traditional terms, type-A
syllables are those in divisions I, II, or IV; type-B syllables are those in division III.
In our Middle Chinese notation, type-B or division-III syllables can be recognized
by the fact that they include -i-, prevocalic -j-, or both, or have an initial consonant
spelled with -y-; type-A syllables are those that have neither -i- nor prevocalic -j- and
no -y- in the initial.
The Old Chinese origins of the distinction between type-A and type-B syllables
have been debated for decades. Karlgren (1940) reconstructed type-B syllables with a
medial yod *-i - before the vowel, and type-A syllables without this yod; this reconstruc-
tion became traditional for a time (Li 1971 and Baxter 1992 have *-j- instead of *-i -).
But Pulleyblank (19621963:99), noting that foreign loans from Chinese showed no
sign of Karlgrens yod, argued for a vowel-length distinction instead:he proposed that
Old Chinese onsets 69
Plain: (type B) p t ts k k q q
p t ts s k k q q
b d dz g g
m n l r
m n l r
pharyngealized p t ts k k q q *
(type A): p t ts s k k q q
b d dz g g
m n l r
m n l r
*rare
the type-B syllables had distinctively long vowels, which diphthongized after the Old
Chinese period, giving Karlgrens yod as a by-product. Zhngzhng Shngfng (1987)
and Starostin (1989) also attributed the A/B distinction to vowel length, but reconstructed
long vowels in type A and short vowels in type Bthe reverse of Pulleyblanks pro-
posal. Both Zhngzhng and Starostin cited comparisons with certain Tibeto-Burman
languages that appear to have long vowels corresponding to Chinese type Aand short
vowels corresponding to Chinese typeB.1
Meanwhile, Pulleyblank (1973, 19771978) abandoned his previous vowel-length
solution and proposed instead that type-A syllables were characterized by stress
on the second mora of the syllable (which he indicated by an acute accent over the
vowel), and type-B syllables by stress on the first mora (indicated by a grave accent).
Finally, Norman (1994), drawing an analogy to the pervasive contrast between hard
and soft consonants in Russian, reconstructed pharyngealization in Chinese type-A
syllables, and proposed that nonpharyngealized syllables subsequently palatalized.
(Pharyngealization is a secondary articulation of either consonants or vowels, in which
the pharynx is constricted by retracting the root of the tongue; the emphatic conso-
nants of many varieties of Arabic are pharyngealized.) In our current reconstruction,
we adopt Normans pharyngealization hypothesis on the grounds that it has the most
explanatory power.2
The various interpretations of the type-A/type-B distinction, and the corresponding
notations, are summarized in Table4.2, using the Middle Chinese minimal pair mng
< meng inscription (type A) and mng < mjieng name (typeB).
The diversity of phonetic interpretations of type-A and type-B syllables in Old
Chinese illustrates the fact that it is often easier to reconstruct the existence and distribu-
tion of phonological distinctions than to reconstruct their phonetic nature. The evidence
for a distinction between type-A and type-B syllables is overwhelming, but the evidence
70 Old Chinese
for any particular phonetic interpretation of that distinction is much more elusive. Our
choice of pharyngealization as the feature characterizing type-A syllables is based on
the following considerations:
Reconstructing the relevant feature as pharyngealization accounts for all these facts in a
natural way, more naturally than competing proposals.
More recently, Ferlus (2009b) proposed that the distinction between type-A and
type-B syllables comes instead from an Old Chinese contrast between disyllabic words
(with preinitial material), giving type A, and monosyllabic words, with no preinitial
material, giving type B.Ferlus argues that when preinitials were lost, the Old Chinese
distinction was replaced by a strong/weak contrast among initial consonants:strong con-
sonants developed tense voice, while weak consonants developed lax or breathy voice.
These voice qualities in turn led to diphthongizations of the type seen in Mon-Khmer
Old Chinese onsets 71
languages with voice registers, with tense voice lowering vowels and lax or breathy
voice raisingthem.
There are problems with this account. When we reconstruct Old Chinese preinitials
on the basis of comparative evidence, we find no tendency for Chinese type-A words
to have presyllables or for type-B words to lack them. As shown below in section 4.2,
direct evidence for preinitials comes primarily from early loans to Vietic and Lakkia
and from Normans Proto-Mn softened initials (e.g., pMn *-p-) or his voiced aspirates
(e.g., pMn *bh-). We list below examples of words showing strong evidence for preini-
tials despite being typeB:
(167) *to > tuwX > du bushel; ladle, pMn *t-; VN u /w B1/
bushel, with nonspirantized initial
(168) *tsik > tset > ji joint, pMn *ts-; VN tt /tet D1/ new year
festival, with nonspirantized initial (VN /t/ < Proto-Vietic *ts-; see
Ferlus1992)
(169) *k[e][n] > kenX > jin cocoon, pMn *k-; VN kn /kn B1/
cocoon, with nonspirantized initial
(170) *kr[e][t]-s > keajH > ji mustard plant, pMn *k-; VN ci /ki C1/
cabbage, with nonspirantized initial
72 Old Chinese
(171) *tem > temX > din black spot, pMn *t-; VN m /om B1/
spot, with nonspirantized initial
(172) *brak > baek > bi white, pMn *b-; VN:bc /ak D2/ silver,
with nonspirantized initial
The Gngyng zhun has little of the added narrative that makes the Zu zhun
commentary so interesting as a historical text, but it does contain very
explicit discussions of the text of the Chnqi itself, and often attempts to explain why
one word was used in the text instead of another. One passage comments on the follow-
ing text from the Chnqi (Duke Xun , year8):
(174)
In winter, in the tenth month, on the day jchu [twenty-sixth in the
sexagenary cycle], our dukes consort Qng Xing5 was to be buried.
It rained and she could not be buried; but on the day gngyn [the next
day], at noon, they succeeded in buryingher.
In a similar passage elsewhere in the Chnqi (Duke Dng , year 15), in which a burial
is also postponed because of rain, the text says, ni k zng then they suc-
ceeded in burying him, with ni instead of r. The Gngyng zhun addresses the
meaning of r and ni and tries to explain why sometimes one was used and some-
times the other. The commentary takes the form of alternating questions and answers:
(175)
What does [r]mean?
There was a difficulty.
What does [ni]mean?
There was a difficulty.
Why [does the text] sometimes say [r] and sometimes say
[ni]?
With [ni] the difficulty is greater than with [r].
Old Chinese onsets 73
The idea seems to be that both r and ni are used as adversative adverbs (but, or
instead)referring in this case to the fact that there was a difficulty with the burial
but that ni is somehow stronger. Such early metalinguistic comments about pre-Qn
texts are interesting in themselves, but what interests us as far as pronunciation is con-
cerned is what H Xi, the Hn commentator, says about the pronunciation of r
andni:
(176)
yn ni zh, ni r shn; yn r zh, wi rqin.
When ni [*n] is spoken, it is inside and deep; when r
[*n] is spoken, it is outside and shallow.
the earliest stage of Old Chinese and that it persisted for a millennium or more, this
simply reflects our lack of evidence for the nature of the distinction during the earli-
est period. It is quite possible that the pharyngealization that led to the changes listed
above actually existed only for a short time, and, being typologically unusual, was
rather unstable and soon led to further changes.
While languages with pharyngealization tend to concentrate in the Caucasus and
among the Afro-Asiatic and Salishan languages, the feature has been observed in at least
one Tibeto-Burman language:the Northern Qing of Hngyn in Schun (Evans
2006a, 2006b). Amis and Atayal, two Austronesian languages of Tiwn, also have
pharyngealized consonants (Maddieson and Wright 1991). Further, Norman (1994:403,
n.9), citing Jakobson ([1931] 1971), underlines the similarity between the pharyngeal-
ization contrast he proposes for Old Chinese and the process of syllabic harmony in
certain Turkic languages whereby a pairing of consonants distinguished by palataliza-
tion, indissolubly linked with vowel harmony, is observed.
On the whole, we reconstruct Old Chinese pharyngealization in the Middle Chinese
syllables of divisions I, II, and IV, and no pharyngealization in syllables of division
III. But there are exceptions among words with Old Chinese nonpharyngealized sibi-
lant initials followed by medial *-r-:OC *sr-, *tsr-, *tsr-, *dzr- (and other onsets that
merged into this set before the Middle Chinese period). These words sometimes emerge
in Middle Chinese as division-III words with retroflex sibilant initials, as expected with
nonpharyngealized onsets:
But some such words show retroflex sibilants with division-II finals instead:
This is because words like those in examples (180), (181), and (182) shifted
from the division-III category to division II at a rather late date:that is, they lost the
feature represented in our notation as MC -j-. We call this change Tsrj- > Tsr- (see
Baxter 1992:267269). Loss of -j- in these forms was presumably motivated in part
by the phonetic awkwardness of combining retroflexion and palatalization.
We can tell that words like these lost -j- because we often have alternative fnqi
spellings that reflect the stage where -j- was still present.For example, although
shng is spelled in the Gungynas
The change Tsrj- > Tsr- also explains the final -aeng in the Gungyns reading sraeng,
which is otherwise unaccounted for and has generally been regarded as irregular, since
shng rhymes as *-e, but -aeng ordinarily comes only from the Old Chinese rhyme
*-a. But MC -jaeng is the regular reflex of *-re,asin
The MC -aeng in shng is thus the regular result of the change Tsrj- > Tsr-: *sre
> srjaeng > sraeng.
Another exception to the general pattern of nonpharyngealized words going to
Middle Chinese division IIIis
Instead of MC sam, we would expect OC *s.rum > srim; and we see this regular
developmentin
(187) *srum > srim > shn the constellation Orion (named for the three
stars in Orionsbelt)
But as a numeral, *s.rum has MC s- instead of the expected sr-, probably influenced
by the following number s < sijH four. As is well known, the Middle Chinese final
-am is also unexpected; we have no clear explanation for this irregularity.
The principal effect of pharyngealization on initial consonants was to block pala-
talization in alveolar stops and *n-, as well as in velars and laterals (see below). In
addition, pharyngealized velars retracted to uvulars after the Old Chinese period. As
mentioned above, Norman (1994:404) shows that by the time of the early Buddhist
transcriptions (c. 200400 ce), Old Chinese velars in pharyngealized syllables had
become uvular.7 There is evidence that in Middle Chinese times, type-A velars were
still more retracted than their type-B counterparts:in the sixth century ce, the authors
of the Qiyn and the Jngdin shwn used different fnqi spellers for the
type-A and type-B velars.
The distinction between type-A and type-B velars disappeared in modern Chinese
dialects, but loans to Proto-Hmong-Mien and Proto-Hmongic normally reflect Old
Chinese type-A velars as uvulars. Examples follow:
(188) *ka-s > kuH > g old (not new); cf. pHM *quoHold
*ko > khuwngX > kng hollow, empty; hole; cf. pHmong
*qh Bhole
*s-kra-s > kaeH > ji go (as a bride) to ones new home; send
(ones daughter) as a bride; cf. pHmong *qua C marry off (ones
daughter)
76 Old Chinese
The limited evidence at hand does not allow us to determine the Hmong-Mien reflexes
of original Old Chinese uvulars with any degree of confidence.
For the lowering of vowels after Old Chinese pharyngealized consonants, see section
5.3.1, in the discussion of the development of rhymes.
4.1.2 PALATALIZATION
(189) *t- > tsy- *ti[n]> tsyin > zhn true, real
*d- > dzy- *dAk > dzyek > sh stone
*n- > ny- *n[u]p > nyip > r enter
*n- > sy- *ni[] > syin > shn body; self
MC tsy-, dzy-, ny-, and sy- are reasonably interpreted phonetically as [t], [d], [], and
[], respectively. These palatal consonants clearly existed already in the pronunciations
of Zhng Xun (127200 ce) and Yng Sho (? 140206 ce), although other
Chinese pronunciations of the same epoch still had nonpalatalized alveolar stops in
type-B words (Coblin 1983:55). The change to palatal consonants is regularly reflected
in Normans Proto-Mn:
Early Chinese loans to Vietnamese show either palatal reflexes, as in (191), or nonpala-
talized ones, as in (192) and (193):
(191) *k.to > *to > tsyowngX > zhng seed; cf. VN
ging8 [zwB]
(192) *tAk > tsyhek > ch foot (measure); cf. VN thc
[tk D1] meter
(193) *tok > tsyowk > zh torch; cf. VN uc [ukD1]
candle
(194) *to[n] > tsyhwen > chun bore through; cf. pHM *chuen to
thread
Old Chinese onsets 77
(195) *m-t-s > tsyhingH > chng steelyard; cf. pHM *nthjuH
balance
(197) *N.r- > y- *N-ru >*lu > yuwX > yu tenth earthly
branch
*m.r- > y- *m-r > *l > ying > yng fly (n.)
We might expect *N.r- and *m.r- to become MC sy-, but we have no clear examples.
Likewise, nonpharyngealized *- evolved to MC y- irrespective of the following
vowel, as in *(r)a-s > yoH > y participate in. Acorresponding development is
seen in Proto-Mn and Hmong-Mien:
(198) *a > yang > yng sheep, pMn *io A, pHM *ju Asheep/goat
In addition, Old Chinese nonpharyngealized velar stops *k- and *g- and the nasals
*- and *- generally became Middle Chinese palatals when immediately followed by
front vowels (Pulleyblank 19621963:100). This change is known as the first palatal-
ization of velars.9 Examples:
(211) *[q]ij > *xij > syijX > sh excrement (see discussion of this item
in sections 4.3.2 and 5.5.5.1).
Although the labiouvulars *q- and *q- did not undergo palatalization before
front vowels, *- did:thus we have *- > hj(w)- before nonfront vowels, and also
before*r-:
(214) *rewk > *xewk > syak > shu melt, infuse
This word is perhaps etymologically related to *m-r[e]wk > yak > yo medicinal
plant.
Old Chinese onsets 79
Schuessler (2010:35) dates the first palatalization of velars to the earlier part of the
Hn dynasty. He argues that velars palatalized before *i earlier than before *e. Among
Chinese dialects, only the earliest layer of Mn shows substantial evidence of having
escaped this change:
Proto-Mn must therefore have branched off before the first palatalization of velars.
Evidence is insufficient to determine whether the first palatalization of velars affected
the donors to Proto-Hmong-Mien and Vietnamese.
However, as mentioned in section 3.1.9, not all cases of Middle Chinese palatals
with connections to Old Chinese velars can be attributed to front vowels, since superfi-
cially similar developments are also observed in words with nonfront vowels:
We have concluded that examples like these illustrate an entirely distinct process. We
explain the palatal variants by supposing tight consonant clusters of preinitial *t- plus
velar stop (cf. section 4.4.4). We suppose that clusters of preinitial *t- plus a velar stop
first simplified to alveolar stops, then palatalized like the original alveolars:
Actually, these developments are not limited to velar initials; parallel examples are
found with uvularsalso:
(221) *t-[k]<r>u > *tru > trjuwX > zhu elbow; the element on the
right was originally
*[k]u > kjuwX > ji nineitself originally a depiction of an
elbow (see J Xshng 2010:348349,991)
(222) *t-nrep > *trep > trjep > zh hanging ears (used as N.Pr.);
phoneticin
*n<r>ep > nrjep > ni unable towalk
4.1.3 RETROFLEXION
The Old Chinese consonant system had no retroflex initials; retroflex consonants
appeared in the course of evolution to Middle Chinese through reduction of certain
Old Chinese onsets with initial or medial *r. Examples of alveolar initials followed by
medial *r follow:
(224) *s-ro > srjuX > sh count (v.); also *s-ro-s > srjuH > sh
number(n.)
Retroflex sibilant initials could also come from Old Chinese clusters with *s- preini-
tial and medial *-r-. Examples:
(225) *s-trek > *tsrek > tsreak > z demand payment; require
*s-trek-s > *tsrek-s > tsreaH > zhi debt;cf.
*m-trek > *m-drek > *drek > dreak > zh blame, punish
(226) *s-t<r>or > *ts<r>or > *ts<r>oj > tsrhjweX > chui to mea-
sure; to estimate; same phoneticasin
*[t]or > tsyhwenX > chun topant
(227) *s-rak > (srjak >) sraewk > shu first day of month; same
phoneticasin
*rak > ngjaek > n go against
(228) *s-qrt > tsreat > zh strip (n.), tablet;cf.
*qrt > *rt > 'it > y second heavenlystem
Old Chinese onsets 81
(229) *s-[l]<r>a > *s-d<r>a > *dzra > dzrjo > ch hoe;cf.
*[l]<r>a > *dra > drjo > ch remove
(230) *s-l <r>o > *s-t<r>o > *tsro > tsrhaewng > chung
window;cf.
*l o > *to > thuwng > tng penetrate
As in most reconstructions of Old Chinese, beginning with Karlgren (1940) and Dng
Tngh (1948), we project the Middle Chinese three-way manner distinction among
stops and affricatesvoiceless unaspirated, voiceless aspirated, and voicedback onto
Old Chinese; for example:
(233) *[N-t]rok > *N-drok > draewk > zhu muddy, pHM *o C
muddy (original voiceless initial retained)
(234) *N-tse-s > *N-dze-s > dzjengH > jng clean (adj.), pMien
*ndz C clean (reflects *N-ts- > *N-dz-)
(235) *N-ka > *ga > *ga > hwang > hung yellow (from
*ka > kwang > gung light, brightness), pMien *gi
Abright, implying pHM *kj- (and perhaps representing earlier
82 Old Chinese
The loose preinitials *m- and *N- also show up as prenasalization in loans to
Hmong-Mien, but they did not have this voicing effect in Middle Chinese, evidently
because the nasals in them were not in direct contact with the initial.
In Middle Chinese, secondarily voiced stops and affricates behave like their
originally voiced counterparts. Just as original *- and *g- merged as MC h, so did
*N.q- and *N.k-, *m.q-, and *m.k-. But just as nonpharyngealized *- and *g-
remained distinct, as MC y- and g-, respectively, so nonpharyngealized *N.q- and *m.q-
became MC y-, while *N.k- and *m.k- became MC g-. The developments of nonuvular
initials with nasal preinitials are summarized in Table4.3.
Note that most of the time, preinitial *N and *m with voiceless aspirated initials
(e.g., *N.p- and *m.p-) have the same Middle Chinese reflexes as with voiceless
unaspirated initials (e.g., *N.p- and *m.p-). But uvulars develop differently. Nasal
preinitials simply voiced an unaspirated uvular stop:*N.q-, *m.q- > *- > MC y- (see
sections 4.4.1.1 and 4.4.2.1), but nasal preinitials before aspirated or voiced uvu-
lars become MC ng-, as shown in Table4.4 (for details see sections 4.4.1.2, 4.4.1.3,
4.4.2.2, and 4.4.2.3).
The examples below illustrate these developments.
(239) *m-qa > *a > nguX > w resist; crosswise; the character is a
pictogram that originally represented
*t.qa > *ta > tsyhoX > ch pestle
(240) *m-qj > *j > ngj+jX > y ant; the phoneticis
*C.qj > *kj > khj+jX > qhow
(241) *m-<r>a > *ra > ngae > y tooth; phoneticin
*m-q(r)a > yoX > y give; for; and, pMn *o Bgive
(242) *N-(r)aj-s > *(r)aj-s > ngjweH > wi false;cf.
*(r)aj > hjwe > wi make, do, actas
The change to a nasal was blocked in aspirated pharyngealized onsets with *-r-,
like *m.qr- and *N.qr-:with these onsets, evolution was to h-, the same as for
*r-:
Middle Chinese, and not detectable from the study of phonetic series. Because of the great
time depths involved, and because of their abundance, these loanwords are stratified:each
layer corresponds to a different phase of expansion of the Chinese world, and can be char-
acterized by a specific set of sound correspondences with the Chinese donor. Accordingly,
the earliest layers of Chinese loanwords to Hmong-Mien or Vietic can be used as a sur-
rogate for Chinese pronunciations of the end of the Old Chinese period.12
By establishing sound correspondences among Middle Chinese, Proto-Mn, and the
earliest loans to Hmong-Mien and Vietic, we bring the reconstruction of Old Chinese
onsets closer to standard comparative practice. Evidence from Proto-Mn and the early
loans to Hmong-Mien and Vietic is especially valuable for what it tells us about com-
plex onsets, on which the testimony of Middle Chinese and of xishng series is both
limited and hard to interpret. The following sections examine the evidence for distinc-
tions in complex onsets that can be gathered from Proto-Mn and from early loans to
Hmong-Mien and Vietic.
4.2.1 PROTO-MN
alone, ignoring ancient languages such as Greek and Sanskrit. But a reconstruction
should be based on correspondences among all the attested daughter languages, ancient
and modern. The dialect research of recent decades has now made it feasible to follow
thispath.
Jerry Normans research reconstructing Proto-Mn (beginning with Norman 1973,
1974a, and 1981)has shown that Mn dialects preserve considerable complexity in syl-
lable onsets that is absent in Middle Chinese and difficult to reconstruct from other
evidence. This information has generally not been systematically used in reconstructing
Old Chinese, but our reconstruction incorporates it, relying not just on Middle Chinese,
but rather on correspondences between Middle Chinese and Mn. Future dialect research
should make it possible to do this more precisely and comprehensively, including other
dialects that preserve ancient distinctions not found in Middle Chinese. The next sec-
tions give examples of such distinctions.
correspond to MC b-. Examples of the correspondences for types 3 through 6 are given
in Table4.8.
Other evidence (especially from early Chinese loans to other languages, see below)
indicates that these distinctions are not of recent origin, and we have taken them into
account in reconstructing Old Chinese onsets. Our main hypotheses (using labial stops
as an example)are:
3. Proto-Mn *b- can reflect OC *b()- or any labial stop preceded by *N.-.
Evidently, a tightly bound nasal presyllable *N.- had already voiced following
obstruents and disappeared before the Proto-Mn stage, because in Mn dialects,
OC *N.p()- and *b()- have the same reflexes.
4. Proto-Mn softened stops (types 3 and 6)result from the lenition of a stop or
affricate in intervocalic position after a loosely attached preinitial:thus pMn
*-p- < OC *C.p()-, pMn *-b- < OC *C.b()-.
5. Proto-Mn voiced aspirates (type 5)result from voiced stops or affricates
preceded by a tightly bound presyllable other than *N.-, e.g., pMn *bh- < OC
*C.b()-; or from an *m- preinitial before any stop or affricate, e.g., OC *m.p()-
or *m.p()-.
a. The change of *b- to *p- probably involved an intermediate stage *b- > *b-, omittedhere.
prenasalized at some point; see the examples in Table4.10. (The Yo voiced stops and
affricates like those in Table 4.10 reflect Proto-Hmong-Mien prenasalized stops and
affricates.)16
However, we also find Hmong-Mien prenasalization corresponding to Proto-Mn
voiced aspirates (type 5 in Table4.8, like *bh-), as shown in Table4.11; this suggests
that it is not prenasalization per se that is responsible for softening.
From a phonetic point of view, we think it is more plausible to attribute the soften-
ing to lenition in intervocalic position (a possibility also mentioned by Norman) than
to prenasalization as such. The softening in Northern Mn is similar to the synchronic
morphophonemic changes that affect medial consonants in certain compound words in
Fzhu (Fng izhn1998):
Old Chinese onsets 89
In our system, initials with a loosely bound presyllable *C.p()- or *C.b()- produce Proto-
Mn softened *-p- or *-b-, respectively, while onsets like *C.b()-, with a tightly bound pre-
syllable (other than *N; see Table4.9) produce Proto-Mn *bh-, as do *m.p()- and *m.p()-.
In either case, the presyllabic consonant *C.- could be a nasal or something else.For exam-
ple, Norman (1986:383) cites a word for cockroach (see example (246) below) where
Northern Mn dialects reflect Proto-Mn *-dzt D in Normans reconstruction (1981:60),
while several other Mn dialects and Cantonese show a presyllable with /k/, implying per-
haps OC *k-dz-. It is not clear what character (if any) corresponds to this etymon:
an Old Chinese loose presyllable when there is additional evidence for it outside of
Northern Mn, since in those cases borrowing from a dialect with voiced obstruents
actually is a realistic possibility.
TABLE4.15 Normans Proto-Mn *l and *lh (tone *A) in three Northern Mn dialects, Mixin
(Hakka), and Old Chinese
PMn MC Lindncn Shbi Hpng Mixin OC
also that Proto-Mn voiceless resonants like *lh- < OC *C.r()- generally have the same
tonal development as Proto-Mn *bh- < *C.b()-, which supports the idea that the voice-
less *C.- before the resonant was still there in Proto-Mn.
This hypothesis also helps account for interesting properties of Normans voiceless
nasals in Southern Mn, described in Norman (1973). While Normans plain nasal ini-
tials *m, *n, and * tend to denasalize to /b/, /l/, and /g/ in Ximn and Chozhu, his
voiceless nasals *mh, *nh, and *h tend to escape this change. Moreover, the nasality
from a voiceless nasal spreads rightward, nasalizing the adjacent vowel, as shown in
Table4.16.
Our interpretation is that denasalizations *m > /b/, *n > /l/-, and * > /g/ affected
nasals that were word-initial, with no presyllable, but that this development was blocked
in onsets like *C.m()-, where a preinitial consonant was present. We assume that the
(necessarily nonnasal) preinitial both prevented anticipation of the opening gesture of
the velum and shortened the nasal consonant, leading to a spill of nasality onto the adja-
cent vowel. This is similar to the situation in Lakkia, where nasality spreads rightward in
clusters having a nasal as their second element (see sections 2.5.3 and 4.2.2.3). See also
Michaud, Jacques, and Rankin (2012), where the same idea is presented, for examples
in other languages of Asia, Europe, and America.
Although China is still quite diverse linguistically, in ancient times it must have been even
more so. In particular, the population of China south of the Yngz was originally
non-Chinese speaking, and Chinese was introduced only gradually (Gernet 1990:2526).
Considering the minority languages that still remain in the area, it is plausible that lan-
guages of at least the following families were spoken in the area south of the Yngz, and
would have been in contact with Chinese from an early period:Kra-Dai (= Tai-Kadai);
Hmong-Mien, also called Mio-Yo; Austroasiatic; and Tibeto-Burman. Obviously,
Chinese was also in contact from an early date with non-Chinese languages to the north;
these languages too may preserve early loanwords that could be helpful in reconstructing
Old Chinese. Here we focus on Chinese loanwords in Hmong-Mien, Vietic, and Kra-Dai,
taking advantage of important recent research on these families.20 In general, those early
Chinese loans that can be identified confirm the idea that Old Chinese syllable onsets
were rather complex, including presyllabic consonants and minor presyllables.
4.2.2.1 Vietic
Work by Ferlus (1976, 1982)has shown that Proto-Vietic, the language ancestral to
a subgroup of Austroasiatic languages including Vietnamese, Mng, Thavng, and
Rc, allowed presyllables with a minor vowel; these are preserved in some languages,
94 Old Chinese
4.2.2.2 Hmong-Mien
There is agreement among researchers on Hmong-Mien languages that
Proto-Hmong-Mien possessed a three-way contrast in initial stops and affricates
of voiceless unaspirated, voiceless aspirated, and voicedthe same as Middle
Chinesebisected by a prenasalization contrast, prenasalized vs. nonprenasal-
ized:altogether a six-way distinction:*p-, *ph-, *b-, *mp-, *mph-, *mb-. This pat-
tern is well preserved in Hmongic, but has undergone a mutation in Mienic according
to the reconstruction of L-Thongkum (1993), which Ratliff (2010) follows on this
point (see Table4.20).
Old Chinese onsets 95
TABLE4.19 Vietic correspondences of Old Chinese tightly and loosely bound presyllables
MC pMn OC Rc Vietnamese
*N-tru-s > drjuwngH > zhng middle (of brothers) *ntro center, middle
*N-tse > dzjeng > qng clear (weather) *ntshji clear
*m-bi[t]-s > bjijH > b nose *mbruiH nose
*m-to-s > dzyuH > sh tree *ntjuH tree
*N.p[u][t]-s > pj+jH > fi boil (v.) *mpuiH boil (v.i.)
*N.tsu > tsawX > zo early *ntsi ouX early
*N-grut > hweat > hu slippery *uat smooth/slippery
*m-tsaw > tsawX > zo bleach; wash *ntswX wash (hands)
*m-t-s > tsyhingH > chng steelyard *nthjuH balance (n.)
*m-dra > drjoX > zh ramie; flax *nduH ramie/hemp
Because early Hmong-Mien speakers had a set of prenasalized initials, they were
able to use them to render Chinese words with nasal preinitials. The Hmong-Mien lan-
guages do not, however, show any distinction between our two nasal preinitials *N and
*m. Moreover, even though Ratliff distinguishes between tightly and loosely attached
prenasals in Proto-Hmong-Mien (2010:12, 209), this distinction does not consistently
correspond to the Chinese distinction between tightly and loosely attached nasal pre-
initials.For example, all the examples in Table4.21 have tightly attached preinitials in
Proto-Hmong-Mien.
96 Old Chinese
4.2.2.3 Lakkia
In section 2.5.3, we cited the observation that Lakkia simplifies inherited consonant
clusters having a nasal as their second element by preserving the first consonant and
transferring the nasality onto the vowel (Solnit 1988, Edmondson and Yang 1988,
L-Thongkum 1992). This behavior also affects early Chinese loanwords, allowing us to
reconstruct stop preinitials we would otherwise not be able to detect:
We reconstruct a loosely attached cluster in (247) urine because the Proto-Mn initial
is *n; if the Proto-Mn initial had been *nh, we would have reconstructed *k.n-. In
Table4.23 below, we show that in Chinese loanwords, Lakkia also simplifies clusters of
two obstruents by retaining the first one; within Chinese, however, the coronal conso-
nant in a cluster wins, no matter what its position.
That the preinitials revealed by Lakkia were present in the Chinese donor language
is clearly demonstrated by the converging testimony of other conservative languages,
Old Chinese onsets 97
As the preceding sections have shown, the evidence from Chinese dialects and
early loanwords can be combined to advance our understanding of Old Chinese
onset distinctions. In general, Proto-Mn allows us to distinguish between single-
ton onsets, tight cluster onsets, and loose cluster onsets, with two provisos, already
underlined:
1. Old Chinese stops or affricates with preinitial *N- are treated by Proto-Mn as
singleton voiced stops or affricates:for example, *N.p()- merges with *b()-.
2. Old Chinese voiceless stops preceded by voiceless preinitials in tight clusters
are not distinguished from the corresponding singleton onsets:that is, OC
*t-,*C.t-, *k.t-, and *s.t- all go to Proto-Mn *t-, *ts-, or *t-, depending on the
presence or absence of pharyngealization and of medial*-r-.
Examples:
TABLE4.24 Vietnamese distinction between Old Chinese voiceless obstruent initials with and
without a tightly attached nonnasal presyllable
Chinese Vietnamese
*tek > tsyek > zh single chic [tik D1]
not spirantized
*tem > temX > din black spot m [om B1]
*s-trek > tsreak > z blame dc [zk D1]
*k.te > tsyeX > zh paper giy [zi B1]
spirantized
*C.te > tsyeng > zhng 1rst (month) ging [zi A1]
*C.taw > taw > do knife dao [zu A1]
TABLE4.25 Manner distinctions for labial stop initials in Middle Chinese, Proto-Mn,
Vietnamese, and Proto-Hmong-Mien, and corresponding Old Chinese onsettypes
MC pMn VN pHM prenasalized pHM not prenasalized
p- *p b- [] H (no examples) OC *p()-
p- *p v- [v]H (no examples) OC *C.p()-
p- *-p v- [v] H OC *m.p()- OC *C.p()-
ph- *ph ph- [f] H OC *m.p()- OC *p()-,*C.p(), *C.p()
b- *b b- [] L OC *N.p()-, *N.p()-, *N.b()- OC *b()-
b- *bh v- [v] L OC *m.p()-, *m.p()-, *m.b()- OC *C.b()-
b- *-b v- [v] L OC *m.b()- OC *C.b()-
Old Chinese onsets 99
4.3 Singletononsets
Singleton onsets are those that occur without presyllables. The general pattern of devel-
opment is illustrated with labial initials in Table4.26; detailed tables of reflexes for each
type of singleton onset are given at the end of each subsection below. In these tables, a
dash indicates that we lack clear examples; in the Vietnamese column, H repre-
sents high-register tones (ngang, hi, or sc); L represents low-register tones (huyn,
ng, or nng). If reflexes are enclosed in brackets, this means that our hypotheses would
lead us to expect such reflexes, but we know of no actual examples. (This does not apply
to the brackets around IPA [International Phonetic Alphabet] symbols in the Vietnamese
columns, which are simply phonetic transcriptions of the Vietnamese orthography.) In
the Middle Chinese column, E and W indicate the assumed reflexes of an eastern
and a western dialect, respectively.
Voiceless unaspirated obstruents generally remain as such in Middle Chinese, Mn, and
early loans. In Middle Chinese, when not followed by *-r-, nonpharyngealized alveolar
stops *t-, *t-, and *d- (but not the sibilant obstruents *ts-, *ts-, *s-, or *dz-) are palatal-
ized to MC tsy-, tsyh-, and dzy-, respectively (see section 4.1.2):
(258) *tek > tsyek > zh single, VN chic [cik D1] classifier for
vehicles
*tAk-s > tsyaeH > zh roast, broil, pHM*ciC
*ti[n]> tsyin > zhn true, real, pMn *tinA
When followed by *-r- (including the *<r> infix), both pharyngealized and nonpha-
ryngealized alveolar obstruents (including sibilants) become retroflex tr-, tsr-, etc. in
Middle Chinese (see section 4.1.3):
Old Chinese *p()-, *t-, and *ts()- were borrowed into Proto-Vietic as *p-, *t-, and
*ts- respectively, but in Vietnamese, Proto-Vietic *p- and *t- became b- [] and - [],
respectively, with high-register tone and implosive articulation (Ferlus 1982). Examples:
The gap left by the Vietnamese change *t- > - [] was then filled when Proto-Vietic
*ts- (and also *s-) became t-[t]:
(261) *tsik > tset > ji joint of bamboo; pMn *ts-; VN tt [tet D1] (<
Proto-Vietic *ts-) New Year festival
*[ts]en-s > tsjenH > jin arrow; VN tn [ten A1] (< Proto-Vietic
*ts-) arrow
(In early Vietnamese loans, it is not unusual for Chinese qshng to be reflected as tone
*A, as in arrow.)
In both Middle Chinese and Proto-Mn, the distinction between the voiceless unaspi-
rated uvular stops *q-, *q-, *q-, and *q- on the one hand and glottal stops *-, *-,
*-, and *- on the other is lost. As for Proto-Hmong-Mien and early loans to Vietic,
we cannot determine from present evidence whether they maintained this distinction or
also lost it. It is possible that Proto-Tai (PT, Pittayaporn 2009)reflected *q- (etc.) as *k-
and *- (etc.) as *-, but the evidence is unbalanced, and there are few clear examples
of the former development:
(262) *qre > 'wea > w frog, PT *krwe Asmall frog (Chinese also
has *m-qre > hwea frog with the animal prefix *m-, support-
ing the uvular initial)
*(r)k > 'ik > y bosom, PT *k D chest
*ew > 'jiew > yo waist, PT *je:w Awaist
*e[n]-s > 'enH > yn swallow (n.), PT *e:n B swallow
*a-s > 'angH > ng basin, PT *a: B basin
*un > 'won > wn warm; gentle, PT *un Bwarm
*<r>em-s > 'jemH > yn satiated, PT *i:m B satiated
initial '- have word-family contacts with Middle Chinese velars, we normally recon-
struct a uvular initial, assuming that uvulars shifted to velars after a tightly attached
nonnasal preinitial:*C.q- > k-, etc.24 Examples:
However, word-family contacts between MC initial '- and k- can also result from the
prefixation of *k- to a root beginning in OC *-. In the following set, contra Sagart and
Baxter (2009), we reconstruct *- rather than *q-, because if the main syllable had been OC
*quj, we would expect them to be written with the phonetic (also used to write*kuj):
In general, we reconstruct MC initial '- as *- rather than *q- in words that lack word-
family contacts with velars or uvulars and are written with phonetic elements used only
for MC initial '-:
(268) *tk > tok > d obtain; pMn *t-; pHM *tuk get25
(269) *tsu > tsjuwX > ji wine; pMn *ts-; pHmong*cowB
(270) *srik > srit > sh louse; pMn *- (pMn reflexes of *sr- fluctuate
between *- and*s-.)
(271) *sraj > srae > sh sand, pMn*s-
(272) *ka-s > kuH > g old (not new); pHM *quoHold
(273) *k(r)[]m > kim > jn metal, bronze; pMn *k-; VN kim [kim A1]
metal, needle; pHM *kjemgold
102 Old Chinese
Old Chinese voiceless aspirated obstruents generally remain as such in later reflexes,
with the exceptions noted below. In Middle Chinese, prevocalic *-r- blocks palatal-
ization, and causes retroflexion in a preceding alveolar, as with voiceless unaspirated
obstruents. When not followed by *-r-, nonpharyngealized alveolar stops *t- etc.
palatalize in Middle Chinese, Proto-Mn and Hmong-Mien, but sibilants such as
*ts- do not. Unlike with *k-, there is no evidence that OC *k- palatalizes before a
frontvowel:
The aspirated uvular stops *q()()- become MC x-, presumably by way of a frica-
tive like [] or[x]:
(281) *[q]ij > *xij > syijX > sh excrement; also read *qij > xjij >
sh moan, with irregular failure to palatalize26
(282) *[q](r)Ak > *xek(?) > syek > sh raincoat of straw; the phoneticis
*[q](r)Ak > syek > sh red, which is probably relatedto
*[t-q](r)Ak > tsyhek > ch red;cf.
*qrak > xaek > h red,fiery
(283) *qj > xj+jX > x swine; Jinu, Shbi, Jinyng /ky 3/;
Zhngh /kui 3/; Chngn /ku3/
(284) *qra > *ra > xuX > h tiger; Zhnqin, Jinu /ku 3/,
Jinyng /ko3/.
(285) *qu > xaw > ho weed (v.), pMn *kh-: Ximn /kau 1/; the
phonetic is said to be an abbreviationof
*qu > xawX > ho good; also written
*qu > xaw > ho weed (v.) whose phoneticis
*q(r)u > xjuw > xi rest(v.)
(286) *q<r>ep > xjaep > xi flank, side of the body; Jinyng, Shbi
/ke 7/; Show, Hpng, Chngn /kie 7/, all wing.27
This alternation between x- and thr- from OC *qr- is similar to the dialectal alterna-
tion between x- and th-/trh- reflexes of voiceless resonants *l , *r, *n (see section 4.3.5.1
104 Old Chinese
below). With the latter there are grounds to suppose that the x- reflexes are western and
the th- or trh-, eastern. We suppose that this is also the pattern for the reflexes of OC*qr-.
More examples of Old Chinese voiceless aspirated obstruents are listed below:
The reflexes of aspirated singleton onsets are summarized in Table 4.28. (Here
and below, a superscript F in the Middle Chinese column means before a front
vowel.)
Old Chinese singleton voiced stops and affricates mostly remain voiced in Middle
Chinese, Proto-Mn, and Hmong-Mien. In Proto-Mn, they evolve into Normans
plain (unaspirated, unsoftened) voiced stop series *b-, *d-, etc. (Norman 1973),
which in most modern Mn dialects devoice into unaspirated voiceless stops. In
Vietnamese, Old Chinese voiced stops and affricates first devoiced into voiceless
unaspirates with low tones:*p- L, *t- L, etc. Then *p- and *t- became implosives
[] and [] (orthographic b and ). In Middle Chinese, nonpharyngealized alve-
olar and velar stops were palatalized under the same conditions as for voiceless
stops; and alveolar obstruents became retroflex before *-r-. Here is an example of
a palatalizing*g-:
(307) *gij > dzyijX > sh look, see; simplex of the causativeverb
*s-gij-s > zyijH > sh show (v.), used as phonetic for velar-initial
words, such as
*[g]rij > gij > q (place name)
This merger occurred in the context of the late Old Chinese retraction of pharyngeal-
ized velars (see section 4.1.1), illustrated by Proto-Hmong-Mienin:
(314) *a > yang > yng sheep; pMn *io A; pHM *ju A
sheep/goat
because it is is phoneticin
In contrast, *la > yang > yng brighta homonym of sheep in Middle Chinese
has xishng contacts to MC d-, th-, sy-, dr-, a pattern characteristic of Old Chinese
laterals: accordingly we reconstruct as *la. Although sheep and bright
are both common phonetic elements, they write mutually exclusive sets of words in
paleographical materials (Bi Yln 2008:8286, 265268) and their respective pho-
netic series host mostly nonoverlapping word families.
In Mn, OC nonpharyngealized *g- and (probably) *g- go to pMn*g-:
But it appears that pharyngealized *g-, *g-, *-, *- become Proto-Mn *-:29
(323) (*rp) > *rp > (dial.) *rk > yik > y wing; pMn*zitD
(This word evidently involves a dissimilation of the initial *- and a dialectal change
from *-p to *-k; see section 5.7; Proto-Mn *-it often corresponds to MC -ik.30)
(324) *[r][o]m > yem > yn salt (n.), pMn *z-; the phonetic in this
characteris
*[k]ram > kaem > jin inspect;cf.
*C.[g]r[o]m > heam > xin salty31
(338) *bak > bjak > f bind (v.); VN buc [uk D2] to tie,bind
(339) *bre > bjaeng > png even (adj.); pMn *b-; VN bng [a A2]
even, level; pMien *be Alevel
(340) *brak > baek > bi white; pMn *b-; VN bc [k D2] silver;
pMien *bk D white
(341) *dAk > dzyek > sh stone; pMn *diokD
108 Old Chinese
(342) *dro()-s > drjuH > zh stop (v.); pMn *diu C; VN [ C2]
tostop
(343) *dzij > dzij > c rice or millet cake, pMn*dz-
(344) *dzo > dzuwng > cng collect; thicket, pMn*dz-
(345) *g(r)u > gjuw > qi come together; mate(n.)
(346) *gij > gjwij > ku mallow
(347) *[](r)a-s > hjuH > y taro (Colocasia antiquorum?), pMn
*io C; pHM*wouH
(348) *ra[n]-s > hjwenH > yun wall around a courtyard;
pMn*yanC
Example (349) child cannot be from *ne, as the character is the head of a
phonetic series with mostly *- words, such as the following, perhaps from a
relatedroot:
Old Chinese onsets 109
In Middle Chinese, OC *l- palatalizes to y-, *l- becomes d-, and both *l- and *l-
become dr- when followed by medial/infixed *r. We reconstruct laterals in the following
based on xishng contacts or loan graphs:
(351) *lj > yij > y level, peaceful; written in early textsas
*l j > syij > sh corpse (for the rhyme, see section 5.5.5.1)
Scattered forms in Vietnamese, Hmong-Mien and Proto-Mn retain the original laterals:
(356) *lot > ywet > yu exuviae of insects or reptiles, VN lt [lot D2]
to skin; to throwoff
(357) *li > den > tin field; to hunt, pHM *lji field
(358) *lrak > draek > z marsh; moisture, Ximn /la 8/ moist
(359) *liw > dew > tio branch (n.), shoot (n.), Ximn /liau 2/ a
longstrip
Old Chinese laterals are most faithfully preserved in the Wxing dialect of
northwest Hnn. There, modern laterals frequently appear corresponding to OC *l-,
*lr-, and *lr-, as in the following examples from the Gzhng dialect (data from W
and Shn 2010:15,24):
Since Wxing and Mn reflect the change *l- > y- but not (or not always) the change
*l- > d- or *lr- > dr-, it would appear that the change *l- > y- was the first to occur.
110 Old Chinese
The earliest evidence for *l- > d- is from the first century ce (for details see Sagart
1999c:3031).
These changes removed all laterals from the consonant inventory of mainstream
Chinese:then at some point before Middle Chinese, *r()- shifted to l-, filling the
gap. Because the Wxing and Mn dialects branched off before the gap was cre-
ated, these dialects also preserve nonlateral reflexes of Old Chinese *r. Wxing
examples:
These reflexes may be compared with the Mn treatment of OC *m.r-, *N.r- and *r-
as Proto-Mn *z- (see sections 4.4.1.4 and 4.4.2.4), and with the Northern Mn treatment
of *C.r- as [s], probably via an earlier [z] (section 4.4.5.4). As a singleton consonant,
however, *r()- did become a lateral [l] in Mn dialects.32
Hmong-Mien retains *r()- as [r]; Vietnamese has orthographic r- [z] with low-register
tone. Additional examples of Old Chinese voiced resonants:
(370) *maj > ma > m rub, grind; pMn *m-; VN mi [mi A2] to file,
sharpen,whet
(371) *mraj > meaX > mi buy; pMn *m-; pHM *mj X buy33
(372) *mra > maeX > m horse; pMn *m-; pHmong *mjn B
horse
(373) *ma-s > mjangH > wng look at from a distance; pHM *maH
lookat
(374) *nar > nan > nn difficult; pMn *n-; VN nn [nn A2] difficulty
(375) *ni[j]-s > nyijH > r two; pMn*-
(376) *nam > nyem > rn whiskers, pHmong *a Abeard
(377) *rak > ngjaek > n go against; pMn *-; VN ngc [k D2]
go against
(378) *r[n]> ngin > yn silver; pMn *-; pHM *i n Asilver
(379) *lra > drjang > chng intestines; pMn *d-; pMien *lja A
intestines34
(380) *lek-s > yeH > y easy; pMn*-
(381) *raj-s > ljeH > l reject; VN ry [zi C2] repudiate (oneswife)
(382) *rem > ljem > lin bamboo curtain; VN rm [zm A2] door cur-
tain, bamboo curtain
Old Chinese onsets111
(383) *ra > ljang > ling beam; bridge; VN rng [z A2] beam,
girder
(384) *ra > luX > l salty (sc. land); pMn*l-
(385) *ru > ljuw > li flow (v.); pMn *l-; pMien *ri u C toflow
(386) *r[]m > lim > ln water (v.); pMn *l-; pMien *rm Atowater
(387) *ri-s > ljengH > lng issue a command
In Middle Chinese, words with resonant initials not infrequently have xishng and/or
word-family contacts with words having voiceless obstruent initials. In such cases, we
reconstruct the voiceless obstruents as voiceless resonants in Old Chinese:
Where we have *m()- and *n-, Karlgren often reconstructed *m- and *tn-, respec-
tively; Dng Tngh (1948) substituted *m- for Karlgrens *m-, and the reconstruction
of voiceless resonants was further extended by Li1971.
However, where we reconstruct voiceless resonants like *m()-, Starostin (1989)
and Zhngzhng (2003) instead reconstructed clusters with *s-, like *sm-; and Mei
(2012) reconstructs *s-m-, where *s- is regarded as a prefix. Aside from the fact that
the semantics of the relevant words usually do not match the functions we associate
with the prefix *s-, we prefer the voiceless resonant reconstruction in these cases, since
there is no evidence, internal or external, pointing to *s- in these words in the Old
Chinese period. Some of the Old Chinese voiceless resonants might reflect *s- clusters
at the Sino-Tibetan stage, but the evidence for such Sino-Tibetan *s-clusters comes
from outside Chinese and is thus not directly relevant to Old Chinese. On the other
hand, there is excellent evidence that clusters of OC *s- plus resonants had different
reflexes (see section 4.4.3.4). The matter is discussed in more detail in Sagart and
Baxter (2012).
4.3.5.1 Dialect reflexes of coronal voiceless resonants *n()-, *l ()-, and *r()-
As the examples above show, the coronal voiceless resonants usually have coronal
reflexes in Middle Chinese:
(396) OC MC OC MC
*n- > *n- >
sy-
*l - > *l - > th-
*r- > trh- *r- >
But there is good evidence for an alternative development of these initials to a frica-
tive [x](or perhaps [h]), which became MC x-. This development can be located in the
central and western regions of the country, while the coronal reflexes in (396) were
probably found along the coast. Thus we have alternations likethese:
We have unusually detailed evidence for a dialect pronunciation of tin < then <
*l i[n]with Middle Chinese initial x-:35
One piece of evidence for a pronunciation of tin with initial x- comes from
the Hu Hn Sh , where the character is used to transcribe the first
syllable of Hinduka, an Iranian name for India, in the toponym Tinzh < MC
then-trjuwk. Moreover, as Bodman (1954:28) pointed out, the Sh tin chapter
of the Sh mng (c. 200 ce) by Li X gives sound glosses for two dif-
ferent pronunciations for tin: xin < xenX display, manifest and tn < thanX
level, at ease. By Li Xs time these were perhaps pronounced *xen and *tan,
respectively:
(402)
(Ho
Yxng etal. 1989:1006)
The sky ( tin) is pronounced in Y , S , Yn , and J with
the belly of the tongue:the sky [*xen] is brilliant [*xen]; it is
high and brilliant above. In Qng and X , it is pronounced with
the head of the tongue:the sky [*ten] is flat [*tan]; it is high
and far away, as ifflat.
The approximate positions of the regions mentioned in this passage are mapped in
Figure4.1. The regions with initial x- are to the west of the solid line; those with th- are
to theeast. Thus it appears that OC *l - evolved to MC th- in coastal regions and to x- in
more interior regions.
As Pulleyblank pointed out (19621963:117118), additional evidence for a
central-western pronunciation of tin sky with x- comes from the word xin
< MC xen, used for the Zoroastrian religion and for its primary god Ahura Mazda. On
the basis of Dien (1957), it appears that this word originates as a western pronuncia-
tion of tin sky, heaven; it was also used in Buddhist texts to refer to devas. The
114 Old Chinese
J Qng
Yn
S
Y
FIGURE 4.1 Regions mentioned in the Sh mng's entry for tin sky
character xin is glossed as follows in the Buddhist lexical work Yqi jng yny
by the monk Huln (who died in 820ce):
(403)
[Xin deity]: the first [word] is pronounced [MC
x(enX) + (k)en=xen]. The Ko shng says:the Iranians call
deities [heaven]; nowadays people within the pass [i.e., in modern
Shnx and Gns] call heavenly deities [MC xen].36
It is not surprising that the Bi, located in the southwest, would have borrowed their
forms from a western dialect of Chinese that treated *l - as[x].
There is evidence that the geographical distribution of the two Middle Chinese
reflexes th- and x- of OC *n- was similar to that of the reflexes of *l -. The name of
the Hn river: *nar-s > xanH > Hn, has MC x- from OC *n-, and this river flows
through the central region where *l - is treated asx-.37
On the other hand, for nonpharyngealized *n- and *l - (and, before a front vowel,
for *-), Proto-Mn regularly has *th-, which is also the regular reflex of Old Chinese
nonpharyngealized *t-:this suggests that there may have been an early coastal dialect
with [t] for *n- and *l -, and [t] for nonpharyngealized *n- and*l -.
Additional examples of Old Chinese voiceless resonants:
Old Chinese onsets 115
(405) *ni[] > syin > shn body; self; used in excavated documents as a
phonetic element towrite
rn < nyin<*ni
The nasal initial in shu hand, first proposed by Unger (1995) and Zhngzhng
(1995), accounts for the word-family contacts in(407):
(409) *l j > syij > sh corpse; used in oracle-bone and bronze inscrip-
tions as a loan graphfor
*lj > yij > y foreigner (especially to theeast)
(413) *l u > syuwX > shu head, pMn *th-: Ximn /tsiu 3/
classifier of odes orhymns
(414) *l A > syae > sh trade on credit, pMn *th-:Chozhu /tsia1/
(415) *l k-s > syiH > sh test, try, pMn *th-:Ximn /tsi5/
Note that in Middle Chinese, *l -, *n -, and *s.t- all merge as sy-; but Proto-Mn
regularly has *th- < OC *l - and *n -, contrasting with *t- < OC *s.t- (see section
4.2.1.3).
Old Chinese *r- has a split development in Middle Chinese that is similar to the dia-
lectal split of *qr- (see section 4.3.2):one dialect (Eastern, by analogy to the split in
*qr-) has MC th- for pharyngealized *r- and trh- for nonpharyngealized*r-:
But another dialect, which we suppose is western, has x- for both *r- and *r-; when
nonpharyngealized *r- precedes a front vowel, we have *r- > *x->sy-, as in example
(214) above:
The reflexes of Old Chinese voiceless resonants are summarized in Table4.31. Note
that our Old Chinese voiceless resonants should not be confused with the *mh-, *nh-,
*lh-, etc. reconstructed by Norman (1973) for Proto-Mn; in our reconstruction these
Proto-Mn initials reflect Old Chinese voiced resonants preceded by voiceless presyl-
lables *C.m-, *C.n-, *C.r-, etc. (see section 4.4.5.4).38
In some cases the alternation is between a noun with a voiceless initial and a
stative/intransitive verb with a voiced initial:
(432) kwang > gung light, brightness: hwang > hung yellow
(436) *N-pe > bjieX > b low, short, VN b [ B1] small, little, tiny
(high-registertone)
118 Old Chinese
(437) *N-k-s > gjuwH > ji old, pMn *g-; VN c [ku C2] old
(low-registertone)
(438) *N-[k]en-s > *Ngen-s > *gen-s > henH > xin appear
(439) *N-pret > *Nbret > *bret > bjet > bi be separated (v.i.), pMn*b-
(440) *N-pra[t]-s > *Nbrat-s > *brat-s > baejH > bi suffer defeat
(441) *N-[k]<r>uj-s > *Ngruj-s > *gruj-s > hweajH > hui be destroyed
(442) *N-to[n] > *Ndon > *don > dwanX > dun be cut intwo
(443) *N-tet > *Ndet > *det > dzyet > sh bend (v.i.), pMn*d-
(444) *N-tok > *Ndok > *dok > dzyowk > sh be attachedto
(445) *N-k<r>ep > *Ngrep > *grep > heap > xi narrow, pMn*-
(446) *N-t<r>k > *Ndrk > *drk > drik > zh straight, pMn*d-
(447) *N-ka > *Nga > *ga > hwang > hung yellow, pMn*-
Like Middle Chinese, Proto-Mn does not distinguish Old Chinese singleton voiced
stops (like *b-) from original voiceless stops prefixed with *N- (like *N.p-). As a
rule, when the Proto-Mn reflex of an *N- prefixed form is a stop, that stop belongs to
Normans plain (unaspirated, unsoftened) voiced series:*b-, *d-,etc.
(448) *N-pret > *Nbret > *bret > bjet > bi be separated (intr.), pMn*b-
(449) *N-t<r>o > *Ndro > *dro > drjowngX > zhng heavy,
pMn *d-; reconstructed with *N-t- becauseof
*to > tsyowngX > zhng swell, swollen;tumor
(450) *N-t<r>k > *Ndrk > *drk > drik > zh straight, pMn*d-
(451) *N-tet > dzyet > sh bend (v.i.), pMn*d-
(452) *N-k<r>ep > *Ngrep > *grep > heap > xi narrow; pMn*apD
(453) *N-ka > *Nga > *ga > hwang > hung yellow, pMn
*uoA
(454) *N-kro[n]> gjwen > qun fist (< rolled-up hand), pMn *g-;cf.
*[k](r)o[n] > kjwenX > jun roll(v.)
(455) *N-kr > *Ngr > *gr > gj+nX > jn near, pMn *g-; cf. the
derived transitiveform
*s-N-kr-s > *s-Ngr-s > *s-gr-s > *gr-s > gj+nH > jn be near
to (v.t.); cf. VN gn [n A2], Rc /tk/ (Grard Diffloth, personal
communication)
Old Chinese onsets 119
In (455), the voiceless *k and the nasal preinitial *N- in jn near are visible in
Rc /tk/ near.42 Note the spirantized initial g- [] in Vietnamese, indicating the
presence of a preinitial. The /t/ in the Rc form may correspond regularly to OC
*s- (there is no preinitial /s/ in Rc; see Nguyn Ph Phong etal. 1988). In Chinese,
preinitial *s- is lost before velar initials unless a front vowel follows (see section
4.4.3.1).
It would make sense if the *N- prefix also occurred before voiceless fricatives, not
just before stops and affricates, but we have no clear examples. In the present system,
there are two voiceless fricatives in Old Chinese: *s- and *s-; but we do not know
of any pairs such as s-:z- or s-:dz- in Middle Chinese that would contrast transitive/
dynamic and intransitive/stative verbs. This raises the possibility that *N- did not voice
a following *s-. We also have no clear grounds to reconstruct *N before the Old Chinese
glottal stops *()()-.
Additional examples of Old Chinese voiceless unaspirated obstruents with
preinitial*N-:
(456) *N-k(r)o-s > *Ng(r)o-s > *g(r)o-s > gjowngH > gng
together, all; pMn *g-;cf.
*k(r)o > kjowngX > gng join thehands
(457) *N-k-s > *Ng-s > *g-s > gjuwH > ji old; pMn *g-;
VN c old;cf.
*[k] > kjuwX > ji a longtime
(458)/khi 1/ < *kh- to open (v.t.), from *[k]j > khoj > ki to open
(v.t.);cf.
/gi 1/ < *kh- to open (v.i.) (Downer 1973:1416; Ratliff 2010:208)
Mien initial /g/ here reflects an earlier prenasalized voiceless stop *kh-, suggest-
ing that the Chinese source form also had the *N- prefix. Since in example (458)
Middle Chinese has only a single form khoj corresponding to both the transitive and
intransitive forms in Mien, in our 2010 paper we proposed that *N- had no effect on
a following aspirated stop in Middle Chinese, so that OC *k and *N-k- merged into
MCkh-.
However, we now reconstruct a loosely attached counterpart of *N-, written as
*N- (see section 4.5.1). In Hmong-Mien, *N- and *N- are not distinguished:both are
reflected as prenasalization. But in Middle Chinese, the intervening vowel * of *N-
prevented the nasal from voicing the following stop. Thus we now reconstruct:
(459) *[k]j > khoj > ki to open (v.t.); cf. Mien /khi 1/ < *kh- to
open (v.t.);
*N-[k]j > khoj > ki to open (v.i.); cf. Mien /gi 1/ < *kh- to
open (v.i.)
Given this analysis, we can assume that tightly prefixed *N- did voice a following
aspirated stop or affricate, as might be expected:*N-p- > b-, *N-t- > d-, *N-k- > g-,
etc. Evidence for such onsets is limited but doesexist:
(461) *N-tse > *Ndze > *dze > dzjeng > qng to clear (of weather);
cf. pHM *ntshji clear. Evidently from the same rootas
*tse > tshjeng > qng clear (adj.)
The development is different with *N- plus aspirated uvulars *q- etc. Just as sin-
gleton *q()()- weakened to a fricative, becoming MC x- (probably by way of uvular
[]), OC *N-q(), phonetically [q()], probably weakened to [] and then to [],
ultimately merging with original OC *()- when the velar-uvular contrast was lost.
Old Chinese onsets121
TABLE4.33 Correspondences for Old Chinese voiceless aspirated initials with preinitial*N
OC MC pMn VN pHM
*N.p()- [b-] [*b] [b- H, b- L]
*N.t()- [d-] [*d] [- H, - L]
*N.ts- [dz-] [*dz] [t- H, t- L]
*N.ts- dz- *dz [t- H, t- L] *ntshj-
*N.k- [h-]
*N.k- g- *g [c- H, c- L] *kh-
*N.q()- ng-
This sequence results in transitive vs. intransitive pairs contrasting MC x- vs. ng-,asin
(462) *q<r>ak > xaek > h to frightenvs.
*N-qak (> ak) > ngak > scared
External support for this analysis comes from the word asafoetida, a loan
from Tocharian B akwa into Chinese (Bailey 1946:786, cited in Pulleyblank 1962
1963:99). Our Old Chinese reconstruction for the second character is a good match for
the second part of Tocharian akwa:
(463) *qa[j]> *aj > *a=MC 'a > (slope, riverbank)
*N-quj-s > *N-qwj-s > *wj-s > *wj-s > ngjw+jH > wihigh
The reflexes of *N- with voiceless aspirated initials are summarized in Table4.33;
reflexes that are assumed but unattested are enclosed in square brackets.
The semantics of wi involve the idea of change from an original form to a modified
one, which may then be thought of as made up, false.
122 Old Chinese
(467) *[N-]ru > *Nlu > *lu > yuw > yu float, swim; pMn *z-;cf.
*ru > ljuw > li flow(v.)
(468) *[N.]ru > *Nlu > *lu > yuw > yu pendants of a banner, also read
*[r]u > ljuw;cf.
*[r]u > ljuw > li pendants of a banner orcap
(469) *N-ruj > *N-rwj > *Nlwj > *lwj > dwoj > tu exhausted;cf.
*[r]uj-s > lwojH > li exhausted
(470) *N.r[]m > *Nlm > *lm > yim > yn excess; licentious
*[r][]m > lom > ln tocovet
Old Chinese had several homophonous *m- prefixes (see section 3.3.2.2), one of
them (*m1a-) a prolific verb prefix indicating volitional action. There are also cases
where we must reconstruct a preinitial *m that cannot be identified with any of the
known prefixes; in such cases we write *m. with a following period instead of a
hyphen.
Like *N, tightly attached preinitial *m voices a following obstruent. In Middle
Chinese, *m produces the same reflexes as *N:for example, *N.p()- and *m.p()- both
become MC b-; likewise for aspirates, *N.p()- and *m.p()-, both become MC b-. In
Hmong-Mien, too, preinitial *m- shows up as prenasalization, like preinitial *N. But in
the Mn dialects, *N-p- and *m-p- produce different reflexes:the former yields pMn
*b-, while the latter evolves to pMn *bh-. Likewise, in Vietnamese, *m differs from *N
in that *m produces a spirantized initial, while *N does not. The details are given in the
following sections.
(471) MC gjuwX > ji mothers brother, pMn *g-: Ximn /ku 6/,
Jinyng /kiu 5/, Shbi /kiu1/
MC gjuwX > ji mortar, pMn *gh-: Ximn /ku 6/, Jinyng
/kiu 1/, Shbi /kiu1/
(472) MC bjaeng > png even (adj.), pMn *b-:Ximn /p 2/, Fzhu
/pa2/
MC bjaeng > png make even, pMn *bh-:Ximn /p 2/, Fzhu
/pa2/
Normans solution was to project this aspiration contrast back to Proto-Mn voiced stops
and affricates: for example, *g- *gh- and *b- *bh- (Norman1973).
124 Old Chinese
(473) *[m-p]u > *mbu > *bu > bawX > bo carry in the arms (voli-
tional), pMn *bh-;cf.
*p<r>u > paew > bo wrap, bundle
(474) *m-to > dzyuX > sh plant (v.); place upright (volitional); cf. the
derivednoun
*m-to-s > *mdo-s > *do-s > dzyuH > sh tree, pMn *dh-; cf.
pHM *ntjuH tree. The root is probably the sameasin
*t<r>o > trjuX > zh prop up, support(v.)
*m-t<r>o > drjuX > zh pillar (instrumental), pMn *dh-; cf. pHM
*u Apillar, Proto-Kra *m-tu Apillar (Ostapirat 2000)45
(475) *[m-t]o > *mdo > *do > duw > tu head (body part), pMn
*dh-;cf.
*to > tuw > du helmet,hood
TABLE4.35 OC *N.p- > /p/ (L), *C.b- > /p/ (L), and *m.p- > /p/ (L)in Ximn andFzhu
1 OC *b- *N.p- *C.b- *m.p-
2 voicing after *N and *m N.b- m.b-
3 Nb- > b- b-
4 1st devoicing:b- > p- L p- L p- L
5 C.b > b-, m.b- > b- b- b-
6 2nd devoicing:b- > p- L p- L p- L
7 Ximn, Fzhu p- L p- L p- L p- L
Old Chinese onsets 125
Old Chinese pharyngealized *m.k- becomes Normans pMn *- rather than his
*gh-, but as he notes (Norman 1974a:28), pMn *- shows the same pattern of tonal
reflexes as his voiced aspirates; this suggests that its development was parallel to
them at some stage. Nonpharyngealized *m.k()- becomes Normans pMn *gh-, as in
(478)below.
(476) *m-kre > *mgre > *gre > heaX > xi crab (with animal pre-
fix), pMn *-; cf. VN cy [kai B1] kind of brackish water crab,
fiddler crab, apparently reflecting an unprefixed form OC *kre
(otherwise unattested, as far as weknow).
(477) *m-kop > hop > h come together; bring together (volitional),
pMn *-; cf. VN gp [p D1] to contribute; to pay jointly with
others, gp [p D2] join, integrate, lump together (we suspect
that VN gp is an earlier loan, and gp a later one).Cf.
*kop > kop > g together; put together; combined46
(478) *m-k(r)k-s > *mg(r)k-s > *g(r)k-s > giH > j warn; avoid
(volitional), pMn *gh-:Ximn /ki 6/, Chozhu /ki 6/; cf. the
relatedform
*krk-s > keajH > jiwarn
(479) *m-q(r)a > yoX > y give; for; and, Ximn /h 6/ give <
pMn*-
As noted earlier, in its tonal behavior, pMn *- behaves like the pMn voiced aspirate
initials.
Normans Proto-Mn voiced aspirates like *bh- have other sources as well (e.g.,
onsets of the type *C.b()-; see section 4.4.5.3); we reconstruct them with *m plus
a voiceless initial when there is direct evidence (usually from Hmong-Mien) for a
nasal preinitial; or when we can infer that the root initial was voiceless, as when
there are related forms with voiceless initials (as in the examples above). For
example, we reconstruct sh < dzyuH tree (pMn *dh-) with *m-t- because of
Proto-Hmong-Mien *ntju H tree and because of the related form *t<r>o >
trjuX > zh prop up, support (v.):the *m- prefix originates in the volitional verb
sh < dzyuX < *m-to to plant. We suppose that sh < dzyuH tree origi-
nally referred to planted trees, as opposed to trees in general ( *C.mok >muwk
> m tree, wood).
Because Proto-Mn distinguishes Old Chinese voiceless initials with preinitial *m-
from those with *N-, there are minimal pairs in Mn contrasting the volitional function
of *m and the stative/intransitive function of*N:
126 Old Chinese
(480) *N-te-s > dengH > dng become fixed; settled (v.i.) (intransitiv-
izing *N-):pMn *d-:Jinu /ti 6/ tranquil, quiet, Ximn /ti 6/
steady, motionless
*m-te-s > *mde-s > *de-s > dengH > dng make fixed, settle
(v.t.) (volitional *m-), pMn *dh-: Jinu /tia 6/ fix (a date or
time) in advance (L Rlng and Pn Wishu 1998:192); Ximn
/ti 6/ to take a little food or medicine so as to make ones self feel
somewhat more comfortable, /ti 6 t 3/ (to settle ones gall-
bladder >) to take some slight means of keeping up ones spirits
(Douglas 1899:552)
(482) *m-ta > duX > d belly, pMn *d- (Ximn /t 6/, Chozhu
/tou 4/, Fzhu /tou 6/, Jinu /tu 6/);cf.
*ta > tuX > d belly, stomach
(483) *m-kep > hep > xi grasp, pMn *gap D pinch; pMien *p D
< *nc- to pick up food with chopsticks; VN gp [ap D1] to pick
up with chopsticks, pull out (bullet from wound);cf.
*k<r>ep > keap > ji press between
*C.k<r>ep > kaep > ji chopsticks, also read *kep > kep chop-
sticks;47 cf. Maleng Kha Pong (Ferlus) /tkap/ < Proto-Vietic *t-kap
baguettes griller (spits for grilling); andcf.
*N-k<r>ep > heap > xi narrow, pMn *apD48
In both d < duX belly and xi < hep grasp, a nasal preinitial is indicated by
voicing alternations in word families, and the semantics point to an *m- prefix:but
in that case we would expect Proto-Mn initials *dh- and *- rather than *d- and
*g-. (The Proto-Mn *g- is unusual in a type-A syllable in any case.) We suspect that
these forms are either not Proto-Mn, or not part of the indigenous layer in Proto-
Mn.49 In any case, as the basic word for belly, the Mn dialects tend to retain the
Old Chinese term *p(r)uk > pjuwk > f, or compounds like fd belly +
stomach.
In Vietnamese, preinitial *m plus voiceless unaspirated initials produces spirantized
initials, with high-register tones in earlier loans and low-register tones in later loans.
With high-registertone:
(484) *m-kra-s > gjaengH > jng strive; compete; VN ganh [ai A1]
emulate
Old Chinese onsets 127
With low-registertone:
(486) *qre > 'wea > w and *m-qre > *re > hwea > wfrog
(487) *qruk > 'aewk > xu and *m-qruk > ruk > haewk > xu a
kind ofbird
(488) *m-qur > *ur > ywinX > yn govern; governor; relatedto
*C.qur > kjun > jn lord; ruler (the two words are often written the
same in early documents)
(489) ~ *[m-q]<r>i[n]-s > hwinH > yn harmony, rhyme;50cf.
*C.qi[n]> kjwin > jn even,equal
*C.qi[n]> kjwin > jn potterswheel
*[N-q]i[n]> ywin > yn even, uniform
*s-N-qi[n] > zwin > xn ten-day cycle (circumstantial noun;
see sections 3.3.2.3 and4.6)
(490) *[m-ts]-s > dzojH > zi load on a vehicle (v.t.);cf.
*[ts]-s > tsojH > zi be conveyed in a vehicle
The reflexes of preinitial *m- before voiceless unaspirated initials are summarized
in Table4.36.
These have usually been reconstructed as homophones for Old Chinese as well
(Karlgren: *ngo, Li: *ngagx), but we reconstruct them as *C.a and *[m].qa,
respectively. Sino-Tibetan comparisons strongly suggest that the velar nasal is original
in w < nguX and other words written with this phonetic:
(497) *C.a > nguX > w five, Written Tibetan (WT) lnga, Written
Burmese (WB) (Hung Bfn 1992:267), Lepcha /fo/
(Plaisier 2007), Mizo (Lushai) pa-nga (Lorrain and Savidge
1898:158), Proto-Tibeto-Burman *l/b-a (Matisoff 2003:149)
*a > ngu > w I, my, WT nga, WB a, Lahu , Proto-
Tibeto-Burman *a I, me (Matisoff 2003:487)
Both w and w are frequently used as phonetic elements, and if they really
were homophones in Old Chinese, we would expect that as phonetic elements they
should be more or less interchangeable; but substitutions of one for the other appear
to be rather late.52 While and eventually came to be pronounced the same, the
xishng and word-family connections of are not indicative of a velar nasal root
initial during the Old Chinese period.
First, consider the graphic connections of the character ch < MC tsyhoX
pestle, which includes w as a phonetic, as recognized in the Shuwn (SWGL
Old Chinese onsets 129
(498) *t.qa > *ta > tsyhoX > ch pestle (for this onset see section
4.4.4.2), Jinyng /khy 3/, Shbi /khy3/
*[m].qa > nguX > w seventh earthly branch
The forms below are from Shng-time oracle bones and bronze inscriptions, respec-
tively (J Xshng 2010:1021):
(500) stop
*m-q(r)a > ngjoX > y withstand; hinder;stop
*N-qa-s > huH > h shut in, stopup
*m-qa > huX > h to stop, tocheck
(501) place
*q(r)a > xjoX > x place(n.)
*s-q<r>a > srjoX > su place (n.); thatwhich
*t.qa > tsyhoX > ch beat
*t.qa-s > tsyhoH > ch place(n.)
(502) sound of hewingwood
~ ~ *qa- qa > xuX-xuX > h h sound of
hewingwood
Example (502) occurs in Ode 165.3 of the Shjng (Xio y:F m ), in the
line f m h-h they hew the wood, [sounding like] *qa- qa. The
Mo text writes sound of hewing wood as , for which the Jngdin shwn gives
the pronunciation xuX-xuX (< *qa- qa); some versions of the Jngdin shwn write
the expression instead as xuX-xuX (Hung Zhu 1980:67). The Shuwn glosses
su not as place (its usual meaning) but as the sound of hewing wood (
f m shng y); it notes that the character consists of jn axe with *m-qa > huX
> h door as phonetic element, and quotes the line from Ode 165.3 as
(SWGL 6375b).53
130 Old Chinese
These examples support reconstructing uvular root initials in most words written
with phonetic . It appears that *m.q- had merged with *- at least by the time of the
Shuwn (100 ce), because words with ng- from original *()- appear as sound glosses
in its entry forw:
(503)
[nguX] < *[m].qa] means [nguH < *ak-s] oppose. In the fifth
[ nguX < *C.a] month, the dark q opposes [ ngjaek < *rak] the
bright, covers the earth, and comes out (SWGL 6639b).
It is after this merger that we begin to find and used interchangeably as phonetic
elements.For example, we take the following to be from the sameroot:
In (505), the character contains the phonetic *a > ngu > w I, my, with
original *-; our reconstruction predicts that this should be a late graph, because in
the early script a character with *- should not be used to write a word with the onset
*m-q-. And in fact, the form is clearly older than . The word appears frequently
in Shng oracular inscriptions:the fullest form represents a man with manacles inside
an enclosure; sometimes the figure of the man is abbreviated to ku mouth; and
sometimes we have only the manacles, which corresponds to the standard graph
(GG 8.867):
(506)
But the character , which reflects the late merger of original *m.q- and *-, is not
attested in pre-Qn documents as far as we know (GG 6.153). The Chinese loan to Proto-
Hmong-Mien confirms the nasal preinitial and stop root initial:
(508) *m.b(r)u > bjuw > f float (v.); pMn *bh-; pMien *mbi u A
tofloat
(509) *m-bre > bjaeng > png make even; pMn *bh-:Ximn /pi 2/
to make level, as a piece of ground;cf.
*bre > bjaeng > png even (adj.); PMn *b-:Ximn /pi 2/ level,
even, smooth; pMien *be Alevel
(510) *m-gra-s > haeH > xi descend; pMn *-:Ximn /he 6/ to
lower; pHmong *a B to descend;cf.
*gra > haeX > xi down, pMn *-:Ximn /e 6/ below, Jinyng
/a 5/ ~/ha5/
Because the Mn dialects treat voiced obstruents with and without *m differently, the
modern Mn dialects retain minimal pairs that have merged in Middle Chinese pronun-
ciation, as in (509) and (510). The Proto-Mienic form *be Alevel for the adjective in
(509) (which would reflect pHM *b-; Ratliff 2010:37) gives us confidence that the root
initial was a plain voiced stop in Old Chinese.
The voiced uvular preceded by *m- evolved to MC ng-, like *N.- (section 4.4.1.3). This
is shown by xishng contacts between MC ng- < *m.- and MC y- of uvular origin:for
example, the character y < MC ngae tooth is phonetic in the early graph for y <
yoX give; for; and, which we reconstruct as *m-q(r)a, with a uvular initial, because of
its graphic (and probably etymological) connection to *C.q(r)a > kjoX > j lift, raise.
Examples of early character forms are given in (511) (from GG 2.573, 3.230, 9.672, 6.349).
(511) =
*m-<r>a > *ra > ngae > y tooth
= *s.A > *s.A > *zA > zjae > xi awry; also used in a
placename:
*[r]a-(r)A > lang-yae > Lngy (name of a mountain in
Shndng)
132 Old Chinese
TABLE4.38 Attested correspondences for Old Chinese voiced stops with preinitial*m
OC MC pMn VN pHM
*m.b()- b- *bh *mb-
*m.d- d- *dh
*m.d- dzy- *dh
*m.dz()- (dz-)
*m.g()- h- * *-
*m.g() g- *gh
*m.()- ng- * ng-[]
If the Old Chinese initial of y < MC ngae tooth had been a velar nasal *-,
y would not have been an acceptable phonetic for *m-q(r)a > yoX > y;
thus we reconstruct *m-<r>a with the body-part *m- prefix. Better than a velar
nasal, a uvular stop in *m-<r>a also explains why *s.A > zjae > xi awry
has initial z- in Middle Chinese:as we will show (section 4.4.3.4), *s-- evolves to
MC s-; only before voiced obstruents does Old Chinese preinitial *s evolve to MC
z- (section 4.4.3.3).
Forms related to *m-<r>a > ngae > y tooth occur with initial [] in south-
east Asian languages:for example, Proto-Tai *a (Li 1977:204), VN ng [ B2], and
Bahnar ngla, all meaning tusk, ivory (Norman 1988:19). This could be taken to sup-
port a velar nasal initial in the Chinese word, but we suspect these are ivory trade words
borrowed from Chinese after the change of *m-- to *-, with semantic narrowing,
rather than the other way around.
Additional examples of preinitial *m before voiced obstruents:
(512)
*m-bi[t]-s > bjijH > b nose, pMn *bh-; pHM *mbruiH nose;54cf.
*C-bi[t]-s > bjijH > b smell (v.t.), pMn *-b- (see section 4.5.5.2)
(513) *m-da > dzyangX > shng to put up, pMn *dh-: Ximn
/tsiu 6/ (to cause to ascend, to bring up), to set up, as doors; to store
up, as goods; to sew on, as soles of shoes; to raise, as water; to pro-
duce, as damp, mist, white ants (Douglas 1899:88);cf.
*C.da > dzyangX > shng ascend, pMn *-d-: Ximn /tsiu 6/
above, upon, upper, superior, former... (Douglas 1899:58)
(514) *m-[d]uk-s (?)55 to poison (v.) (volitional), pMn *dhu C:Ximn
/tau 6/, Fzhu /tau 5/ to poison;cf.
*[d]uk > dowk > d poison (n.), pMn*d-
(517) *m.l > zying > chng raised path between fields;cf.
*l -s > syingH > shng overcome; surpass
(518) *m.lut > zywit > sh glutinous millet, pMn *dh-, pHM *mblut
glutinous/sticky
The initial *m in (518) is supported by the intrusive *-b- in Hmong-Mien; the aspirated
initial *dh- in Proto-Mn suggests that the cluster became a prenasalized obstruent in
Proto-Mn:*m.l- > *mdl- > *md-.56 Asimilar exampleis
(519) *m.lru[t]-s > drwijH > zhu fall down; pHmong *mbluei C shed
leaves/drop
Here, too, the Proto-Hmongic intrusive *-b- points to a preinitial *m. The semantics
are inconsistent with the volitional *m- prefix that appears on verbs, so we write
*m. instead of *m-; in this case the *m may have been part of theroot.
The evolution of onset *m.r- into Middle Chinese appears to be dialectally condi-
tioned. In one dialect, *m.r- and *m.r- evolved to y and d, respectively:
This word is reconstructed with *m-r- (with the animal prefix) because the phoneticis
Proto-Mn has *z- in (520) fly (Ximn /sin 2/), the same as the Proto-Mn reflex
of *N.r- and *r- (*z- in swim and salt, (467) and (324), respectively). Proto-Tai
*m.le: Ainsect (Pittayaporn 2009)and Rc mla big fly (Nguyn Ph Phong etal.
1988:20) were presumably borrowed after the change of *m-r- to *m-l-. An example of
*m.r going to MCd-is:
We reconstruct this word with the onset *m-r- because in oracle bones and early bronze
inscriptions the graph is composed of an eye with water falling fromit:
(523)
134 Old Chinese
TABLE4.39 Attested correspondences for Old Chinese voiced resonants with preinitial*m
OC MC pMn VN pHM
*m.n()- m-
*m.()- m-
*m.l y-
*m.l()r- dr-
*m.l- zy- *dh *mbl-
*m.r- d- ~ m-
*m.r- y- ~ m- *z
This suggests that originally represented tears in early Old Chinese, the word later
written as (J Xshng 2010:268):
(524) *m-rp > dop > t reach to (early graph for tears)
*[r][]p-s > (dialect:) *rup-s > lwijH > li tears(n.)
*k-rp > khip > q weep; note the phonetic
*k.rp > lip > l stand(v.)
The early graph for tears would then have been a phonetically appropriate choice to
write a volitional verb *m-rp > dop > t reachto:
(526) *m-ri-s > *mri-s > *mre-s > mjaengH > mng command
(n.);cf.
*ri > ljeng > lng send (a person)
(527) *m-rk > *mrk > meak > mi wheat; the phoneticis
*m.rk > *m.r > loj > li come57
The reflexes of preinitial *m plus voiced resonants are summarized in Table4.39.
The MC reading that for otter can be reconstructed as OC *rat (section 4.3.5).
Ratliff (2010) reconstructs Proto-Hmongic *ntshju a Aotter, which could be related
to the Chinese word. Although we would normally expect Proto-Hmongic tone
*D corresponding to OC *-t, the correspondence of the OC *-t to pHM tone Ahas
parallels:
Sagart and Baxter (2012) discuss the development of onsets with preinitial *s- into
Middle Chinese, as well as the competing proposals in Mei (1989,2012).
Although nonpharyngealized *s.t- generally becomes MC sy-, there are also a few
cases of MC ts- from forms with a root-initial *t-, as if by metathesis, as with pharynge-
alized *s.t-. We have not found a phonological conditioning for this split development
of OC *s.t- to MC sy- and ts-; for the present, we distinguish them by writing sy- <
*s.t- and ts- <*S.t-:
(533) *S-t-s > *ts-s > tsingH > zng boiler for steaming rice (instru-
mental noun; see section 3.3.2.3);cf.
*t > tsying > zhng to steam (v.t.)
(536) *s-kra-s > kaeH > ji marry, Jinu /xa 5/, Hpng /ha 5/; cf. VN
g [ C1] to give (ones daughter) in marriage
(537) *s.[k]raw > kaew > jio teach, Jinyng, Lindncn, Jinu
/xau1/
*s.kraw-s > kaewH > jio teach; instruction
(538) *s.ka[r]> kan > gn liver, Jinyng, Lindncn, Jinu /xue 1/,
Hpng /hon 1/, Yngn /hum 1/ (Central Mn); cf. Rc /tkan/,
VN gan [nA1]
(539) *s.kr[a]m-s > kjaemH > jin sword; cf. Rc /tkm/ sword, VN
gm [m A1] (with tone Afor Chinese qshng)
Reconstructing *s.k- in sword fits well with the fact that the xishng series of
sword includes words that require preinitial *s, including the main phonetic element:
(541) *s-kij > *s-ij > *ij > syij > sh Achillea (?); the phoneticis
*[g]rij > gij > qold
(542) *s-kiw > *s-iw > *iw > syuw > shu collect; harvest; the
phoneticis
*k-riw (dial. > *kriw) > kjiw > ji totwist
(543) *s-qar > *swar > sjwen > xun spread (v.); with the same
phonetic element,cf.
*[]ar > hjwon > yunwall
(544) *s-qi[n] > swinX > sn bamboo sprouts; with the same
phonetic element,cf.
*qi[n]-s > xwenH > xun ornate, decorated
(545) *s.qet > *set > set > xi wedge put in teeth of corpse;cf.
*[k]et-s > khejH > q script notches
(546) *s.q<r>[i]p > tsrip > j crowded together;cf.
*qip > 'jip > y bow (v.), salute59
138 Old Chinese
TABLE4.41 Attested correspondences for Old Chinese voiceless unaspirated stops with
preinitial*s
OC MC pMn VN pHM
*s.p-
*s.t- ts- *ts d- [z] *ts-
*s.t()r- tsr- *ts
*s.t- sy- *t
*s.ts- s-
*k; in NMn:
*s.k- k- g- [] *q-; *sj-F
[x] or [h]
*s.k- k-; sy-F
*s.qr- tsr-
*s.q- s- *s
(547) *s-qrt > *tsrt > tsreat > zh strip (n.), tablet; the phoneticis
*qrt > *rt > 'it > y second heavenlystem
The attested reflexes of voiceless unaspirated stops and affricates with preinitial *s-
are summarized in Table4.41.
(548) *s.tA > syae > sh extravagant, pMn *th- (Ximn /tsia 1/ lav-
ish); the phonetic elementis
*tA > tsyaeX > zh (nominalizing particle)
The phonetic *tA in (548) indicates that the main initial is an alveolar stop rather
than a resonant like *l - or *n-. In the following example we also get a clue from a sound
gloss in the Shuwnjiz:
We reconstruct sh heat as *s-ta > syoX > sh heat; the expected pMn
*th- is preserved in the Zhngpng dialect in the expression /tsi 3 tsi 3/ for
chsh end of heat, the fourteenth solar period (Zhng Zhnxng 1992:38).
Old Chinese onsets 139
For *s-ta > sh Panicum, the Mn dialects mostly have sibilant initials rather than
the expected pMn *th-, perhaps due to stratification in Proto-Mn (Ximn /sue 3/,
Fzhu /s 3/). But some Mandarin dialects have aspirated initials for both *s-ta >
syoX > sh Panicum and *s-ta > syoX > sh heat:Hfi /tu 3/, Yngzhu /tsu 3/
in both cases (Bijng dxu 2003:122). These aberrant forms point to pre-Middle
Chinese substrata.
Hakka also occasionally retains an aspirated affricate for *s.t-:we reconstruct
(550) *s.tu()-s > syuwH > shu (wild) animal, /tsiu 5/ wild ani-
mal in several Hakka dialects (Wngyun , Lchun ,
Gnxin , Chngtng , etc.; from L Rlng and Zhng
Shungqng 1992:249)
The alveolar initial in (550) is supported by graphic variation in early documents. In the
Gudin version of Z y , shu appears on strip 38 where the Shnghi
Museum version (strip 19)and the received version have shu (GD 1998:20, 130
131; SB 1:210), which we reconstruct with *s.t-:
(551) *s-tu > syuwX > shu keep, guard, pMn *t-:Ximn /tsiu 3/ to
guard; to keep carefully (Douglas 1899:56)
(552) *s-tuj > *tsuj > tshwoj > cu urge, repress; derivative with
increased valencyof
*tuj > thwoj > tu pushaway
(553) *s-t<r>or > *tsror > tsrhjweX > chui to measure; to estimate;
also read *tor > tsyhwenX (same meaning).
Preceding *ts-, *s- evolves to MC s-, like *s-ts-; Proto-Mn, as often, loses preinitial
*s- in the word star:
(554) *s-tse > *se > seng > xng star, pMn *tsh-, reflected in eigh-
teen Mn dialects, e.g., Ximn /tsi 1/; see Chn Zhngti and L
Rlng (1991:8).
We reconstruct *s-tse to account for the Mn initial and also for the use of as
a loangraphfor
Under this analysis, the word star is a deverbal noun formed by adding the prefix *s-
to a verb root *tse bright; a related root with nonpharyngealized initial occursin:
A possible example of *s-k- > MC sy- is (557), but the *s- seems to be absent in
many modern dialects:
(557) *s-ke-s > *s-te-s > *e-s > syeH > ch wing; but perhaps there
was also a reading
*ke-s > * te-s (which would give MC tsyheH) > ch, Mandarin /t 5/,
Gungzhu /ti 5/; the phoneticis
*ke > tsye > zh branch (of tree), limb, pMn*kiA
(558) *s-qA > *sA > sjaeX > xi depict; pMien *xja B towrite
*s-qA()-s > *sA-s > sjaeH > to unload
*[m].qa > *a > nguX > w seventh earthly branch
(559) *s-qA > *sA > sjaeX > xi depict; pMien *xja B towrite
*s.qAk > sjek > x slipper, shoe, also read tshjak, as if from *s.qak.
The difference in both initial and vowel between the two Middle Chinese readings of
xi suggests that we are dealing with a dialectal distinction.
With medial *r, developments were, again, affected by dialect divergence. In the
dialect where *qr- > x-, *s.qr- evolved to MCsr-:
(560) *s-q<r>a > srjoX > su place; that which (see discussion in sec-
tion 4.4.2.2)
In the dialect where *qr- > trh-, *s-qr- went to MC tsrh-. An exampleis
(561) *s-qr[]p > tsrhip (JDSW) or tsrhjep > tsrheap > ch gather,
collect. Jngdin shwn also gives the readings khip <
*C.qr[]p, xip < *q(r)[]p, and ngip < *[m-]qr[]p (JDSW
118, 145). The phoneticis
*[m-k-]rp > gip > j reachto.
The correspondences for preinitial *s. followed by voiceless aspirates are summarized
in Table4.42.
Old Chinese onsets 141
(562) *s-a > (*za >) zjang > xing auspicious; the phoneticis
*a > yang > yng sheep
(563) *s-en-s > (*zen-s >) zjwenH > xun whorl of hair on the head;
the rootis
, *<r>en > hjwen > yun round
Clear examples of *s.d- and *s.dz- are difficult to find:a possible example of *s.d-is
(565) *s.[d]a > zjangX > xing elephant, pMn *dzh-, Ximn /tsiu 6/;
Proto-Tai *a: C (Pittayaporn 2009:327), Proto-Lakkia *dza: C
(L-Thongkum 1992:60); pMien *ji B (the initial is unex-
plained), Proto-Vietic *a-a
The Proto-Mn reflex *dzh- is explained if we assume that Proto-Mn still had *s-d-
at the time of the first devoicing in dialects like Ximn (see Table 4.9 in section
4.2.1.1), so the *d- was protected from that devoicing (which gave voiceless unaspi-
rated reflexes); then we have *s-d- > *zd- > *dz- > *dz- > [ts]- by the second
devoicing.
142 Old Chinese
(566) *s-m-tAk > zjek > x mat; circumstantial noun, place for putting
things, pMn *dzh-;63 derived from the same rootas
*t<r>ak > trjak > zhu toplace
The fact that the evolution was not *s.d- > *zd- > zy (as the parallel of *s.t- > *s- >
sy- discussed in section 4.4.3.1 above could lead one to expect) shows that the simplifi-
cation of *zd- to *z- occurred before, and bled, the palatalization of alveolars. However,
the change must have occurred after the first palatalization of velars, since wehave:
(567) *s-gij-s > *zgij-s > *zdij-s > *ij-s > zyijH > sh show (v.);
causativefrom
*gij > *dij > dzyijX > sh look,see
However, in forms with nonfront vowels, where the first velar palatalization did not
occur, *s- simply dropped before velars in Middle Chinese, as in (568), repeated from
(455) above:
(568) *s-N-kr-s > *s-Ngr-s > *s-gr-s > *gr-s > gj+nH > jn be
near to (v.t.); cf. VN gn [n A2], Rc /tk/.
We also have examples of OC *s.b- > dz-, presumably by the route *s.b- > *zb- >
*bz- > dz-. In the Shng oracular inscriptions, the graph , depicting a nose, is used
to write {} b nose as well as {} z to follow; from and {} self (adv.) (X
Zhngsh 1988:378):
(569)
*m-bi[t]-s > bjijH > b nose, pMn *bh-; cf. pHM *mbruiH nose (the
pHM *-r- is unexplained)64
*s.[b]i[t]-s > *zbit-s > *bzit-s > *dzit-s > dzijH > z self (adv.), pMn
*dz- (we would expect pMn*dzh-)
*s.[b]i[t]-s > dzijH > z to follow;from
As the character was commonly used for the second and third words above, a pho-
netic element was added to represent the meaning nose:65
(571) *s.ba-s > *zba-s > *bza-s > *dza-s > dzjangH > jing crafts-
man, pMn *dzh-; we suspect that the phonetic elementis
*pa > pjang > fng container, box (Shuwn); probably connectedto
*C-pa > pjang > fng square
Old Chinese onsets 143
In Shng inscriptions, *pa > pjang > fng has a shape like this (GG 9:1019):
(572)
(573) *pra > paeng > bng side of the temple gate; sacrifice there.
The Shuwn jiz treats jing as a semantic compound of jn axe and fng
box, indicating that the initial was probably already *dz- by that time (100 ce), and
the phonetic role of fng as phonetic was no longer recognized (SWGL 5729b).
The correspondences for preinitial *s with voiced stops and affricates are summa-
rized in Table4.43.
The *s- prefix in *s-ma creates a deverbal noun referring to the circumstances of
an event (section 3.3.2.3):mourning, burial < circumstances associated with dying.66
Another e xampleis:
(575) *s.mi[t]> swit > x eleventh earthly branch; this is the phoneticin
*met > xjwiet > xu extinguish, destroyand
*[m]et > mjiet > mi destroy
TABLE4.43 Attested and predicted correspondences for Old Chinese voiced stops and
affricates with preinitial*s
OC MC pMn VN pHM
*s.b- dz- *dzh
*s.d- z- *dzh
*s.dz-
*s.g- g-; zy-F
*s.()- z-
*s.r- zr-
144 Old Chinese
(576) *s-nu > *su > sjuw > xi shame; pMn *tsh-:Ximn /tsiu 1/,
Fzhu /tsieu 1/; the phonetic element is the same as in
*n<r>u > nrjuwX > ni fastener
(577) *s-lat > *sat > sjet > xi leak, ooze; pMn *tsh-:Ximn /tsua 7/
(in /tsua 7 sai 3/ to have diarrhea); the phoneticis
*l ap-s > *l at-s > syejH > sh generation, also phonetic in and relatedto
, *l[a]p > yep > yleaf
When the onset contains an *r, either as a main initial or as a medial, the result is
MC sr- insteadofs-:
In the following example, sound changes of the Middle Chinese period have added
complications:
(Note that the Mn forms for lo < OC *C.ru old indicate a preinitial consonant,
even though they reflect a pharyngealized *r- rather than plain *r- in the main initial.)
For su, the Gungyn gives only the pronunciation suwX. Normally, MC -uw would
reflect OC *-o, not *-u, but Shjng rhymes clearly indicate that words with this phonetic
rhyme as *-u, not *-o.67 Moreover, in the Jngdin shwn, words with as phonetic
Old Chinese onsets 145
are frequently spelled as MC srjuw (the regular reflex of OC *sru). The pronunciation
suwX results from two sound changes that affected some dialects in the Middle Chinese
period:(1)a general loss of MC -j- (or whatever feature that notation represents) after
initials of the type Tsr- (see section 4.1.1 and Baxter 1992:267269); and (2)a merger of
the Tsr- initials with those of type Ts-:thus we have *s-ru > srjuwX [j-] > sruwX
[-] > suwX [s-].68
A complex but interesting case is(582):
(582) , *s-[l]<r>a > dzrjo > ch hoe (n.); also hoe (v.)69
, *s-[l]<r>a > dzrjo > ch hoe(v.)
*[l]<r>a > drjo > ch remove; get rid of;70 possibly connectedto
*la > yo > y remains; surplus
Normally, the phonetic element in and (a form that occurs for hoe in some texts)
would lead us to reconstruct ch < dzrjo hoe with a dental affricate:
(583) *[ts]A > tshjaeX > qi moreover, also read *tsa > tsjo > j
[final particle]
However, the characters and for ch hoe both appear to be of late origin,
probably originating after *s-lr- would have become an affricate: ch is not in
the Shuwn, which has ch instead, defined as what is used in weeding while
standing (); and neither character appears in pre-Qn documents as far
as we know.71
Now *[l]<r>a > drjo > ch remove is a general verb for removing things, mostly
undesirable onesdisasters, sickness, and so forthbut when used concretely, it can
quite naturally refer to the removing of unwanted plants, as in (584), a well-known pas-
sage from Zu zhun , year 1 of Duke Yn . The advisor Zhi Zhng
warns the duke not to keep giving in to the demands of the dukes mother to extend
favors to the dukes ambitious brother:the image of a creeping grass or vine is used to
describe his ambition.
The various forms for hoe (noun and verb) in Mn dialects are relevant, but com-
plex; it will require further research to work out the details. It is possible to reconstruct
a Proto-Mn form for the verb, with Normans pMn*dh-:
(585)pMn *dhy Ahoe (v.):Fzhu /ty 2/, Ximn /ti 2/, Jinyng/hy2/
146 Old Chinese
These forms although frequently written with the character [] in the dictionar-
ies, cannot be related because of the discrepancy of initial: is a chorng [MC
dzr-] initial word whereas the Miin forms must be related to a dinq [MC d-] or
cherng [MC dr-] initial.
(586) *s.tur > sywijX > shu pMn *tyi B (as if from
water; river OC*tur)
*s-ta > syo > sh write pMn *ty A(as if from OC*ta)
*s-tse > seng > xng star pMn *tsha A(as if from OC
*tse)
*s.tA > syae > sh pMn *thia A(as if from
extravagant OC*tA)
*s-l <r>o > tsrhaewng > pMn *t A(as if from OC
chung window72 *l <r>o)
Given this pattern, the fact that we have Proto-Mn *dh- in hoe (v.) rather than an affri-
cate is less surprising. The Proto-Mn reflex *dhy Awould be regular if {} ch was
actually *s-m-l<r>a, where the *m- prefix is semantically appropriate for a volitional
action. The *s- prefix could either be the prefix that derives oblique deverbal nouns (in
the case of the noun hoe) or the prefix that adds an oblique argument (in the case of
the verb):then perhaps drjo < *m-l<r>a is what one does to weeds, but dzrjo <
*s-m-l<r>a is what one does to a field or a crop:73
But pMn *dhy Ahoe (v.) could simply represent the verb {} ch < drjo remove
without the *s- prefix, if we reconstructed {} ch as *m-l<r>a rather than as *l<r>a.
We reconstruct *s-n- in x west:
There are unanswered questions about the reconstructed rhymes here: {} x < sej
west is reconstructed with the rhyme *-r (section 5.5.5.4), while {} ni then is
reconstructed with the rhyme *- (section 5.4.2.1), so it is unclear why should be
chosen as a phonetic to write {} ni; and the regular reflex of *nr would be MC
nejX, not nojX. Perhaps the apparent confusion of rhymes results from the occurrence
Old Chinese onsets 147
of the adverb then in an unstressed position (as with li come; see section 5.4.2.2).
However that may be, characters ancestral to and are both used already in Shng
oracular inscriptions to write the adverb read ni < nojX (Zho Chng 1988:293), and the
use of x < sej as phonetic to write {} ni < nojX is evidence for the onset *s-n-.
Reconstructing *s-n- in x west also accounts for how x can be phonetic
in shn < syinX smile; MC sy- cannot represent simply *s-, but can regularly
reflect*n-:
According to the Shuwn, MC sej is an alternate graph for x west. The Shuwn
explains the connection between west and nest:
Although there is disagreement about how to analyze the early graphs for x, the idea
that it depicts a nest is accepted by many scholars of Shng inscriptions (Y Xngw and
Yo Xiosu 1996:10291033). Just as the west is etymologically a place for stopping,
symmetrically, dng east is etymologically related to the notion of dng move
(see Sagart2004):
We reconstruct *s-n-in
(593) *s-ni[]-s > sinH > xn truthful; phonetic elements used for this
word in early documents include
*ni[] > nyin > rn (other) person
*ni[] > syin > shn body;self
*s.ni[] > tshen > qin thousand;
148 Old Chinese
We reconstruct *s-r-in
(595) *s-rar > srean > shn mountain, hill, pMn *-; relatedto
*(r)ar() > ngjenX > yn hill (a rhyme in Ode 250.2 indicates
pngshng, not shngshng)
In Ode 250.2, *(r)ar() is part of a sequence of six rhyme words in *-ar (see example
(1035) below); all the other rhyme words are clearly pngshng. We suspect that {}
*(r)ar() is just another form from the same root as *s-rar > srean > shn. The Mo
commentary glosses it as small hill().
The entry for shn < *s-rar in the Sh mng (200), which explains
words by means of sound glosses, is also interesting and supports both the reconstruc-
tion of shn with initial *s-- and the reconstruction of the coda as*-r:
(596) Shmng:
*s-rar mountain is [*rar] river bank; it produces [
[*s-rar] chnshng] things.76
The first item y < ngea evidently reflects a dialect where *-r merged with *-j (see section
5.5.1.4). We suspect that the basic meaning of the related roots *rar (in *s-rar and
*(r)ar()) and *rar (in *rar and *C.rar) is slope, nearly vertical side, applicable
to the side of a mountain, the bank of a river, and the forehead.
We reconstruct *s-r- in(598):
(598) *s-rak > *srak > (srjak >) sraewk > shu first day of the lunar
month;77 the root (and phonetic element)is
(= ) *rak > ngjaek > n go against
The term *s-rak first day of (lunar) month can perhaps be interpreted as an oblique
deverbal noun from the root (= ) *rak go against: the time when the moon
goes the other way (i.e., starts getting larger instead of smaller). The Shuwns entry
includes *s-a > su > s revive as a sound gloss:
Old Chinese onsets 149
(599) [*s-rak]:On the first day [of the lunar month], the moon begins
to revive [s < su < *s-a]. From yu moon, with [n <
ngjaek < rak] as phonetic.78
(603) *s.ni[] > tshen > qin thousand; in early forms of the character
the phoneticis
*ni[] > nyin > rn (other) person
(604), *s-l <r>o > *s-t<r>o > *tsro > tsrhaewng > chung
window;cf.
*l o > *to > thuwng > tng penetrate; possibly from the same
rootis:
*s-l o > tshuwng > cng (penetrating) hear well; intelligent
(605) *s.l ot-s > *s.tot-s > *tsot-s > tshjwejH kerchief, alsoread
*l ot-s > sywejH > shu (same meaning)
We reconstruct *s.r-in
These pairs, first cited in Bi Ypng [Baxter] (1983) and Baxter (1992:205), show
simultaneous alternations along several phonological dimensions and probably need to
be interpreted in dialectalterms.
The attested correspondences for preinitial *s- followed by voiceless resonants are
summarized in Table4.46.
Old Chinese onsets 151
TABLE4.46 Attested correspondences for Old Chinese voiceless resonants with preinitial*s
OC MC pMn VN pHM
*s.m()-
*s.n()- tsh- *tsh
*s.-
*s.l ()- tsh-
*s.l ()r- tsrh- *th
*s.r- tsh-
*s.r- tsrh-
We reconstruct stop preinitials *p, *t, and *k when it is necessary to account for xishng
and/or etymological connections that cannot be explained otherwise. In the case of *t
and *k, we can sometimes identify the preinitial as a prefix (see section 3.3.2); in other
cases, it may have been part of the root. In some cases, comparative evidence from early
loans and Mn dialects can help identify the preinitial, or at least confirm its existence.
When preinitial *p, *t, or *k precede an obstruent initial without an intervening
schwa between them, a consonant cluster is formed that is simplified on the way to
Middle Chinese. Examples:
(609) *[k.]or > *kor > kwan > gun cap (n.), pMn *koi C crest,
comb, e.g., Ximn /ke 5/ cocks comb; headdress (with unex-
pected qshng; Norman 2006:137); cf. pHM *wi n crest/comb
*k.or-s > *kor-s > kwanH > gun cap(v.)
(610) *k-pa > *ka > khjwang > kung square basket (the aspiration
in *p- is unexplained); the phonetic, and a related word,is:
*pa > pjang > fng container, box (Shuwn);cf.
*C-pa > pjang > fng square, VN vung [vuA1]
(611), *[p.k]ap > *pap > pjop > f model, law; the phonetic
appearstobe
*[k](r)ap-s (dial. >) *[k](r)ak-s > khjoH > q depart; also phoneticin
*m-[k]ap > hap > h thatch, cover(v.)
*[k]ap-s > kajH > gi cover (v.); cover(n.)
152 Old Chinese
One notes that in these examples the first consonant is retained in Middle Chinese,
although in (610) kung < khjwang < *k-pa square basket, the second consonant
is also retained, in the form of labialization. More examples of Middle Chinese labiove-
lars from OC *k plus labial consonant are found in the examples below, with *k.m- and
*k.m-. With the cluster *k.m-, dialects appear to have fluctuated between the develop-
ments *k.m- > *km- > *k- > kw- and *k.m->m-.
(612) *k.met (dial. >) *kmet > *ket > kwet > mi sleeve (keeping
preinitial *k; Mandarin mi is based on the other reading)
*k.met-s > *met-s > mjiejH > mi sleeve (with loss of preinitial*k)
(613) *k-mra (dial. >) *kmra > *kra > kjwaengX > jing bright
window; relatedto
*mra > mjaeng > mng bright
(614) *[k-m]a-s > *ka-s > khwangH > kung desolate, waste; pos-
sibly relatedto
*ma > xwang > hung wasteland; uncultivatedland
In most cases where an onset involves two obstruents, one of the twoeither the first
or the secondis an alveolar stop or affricate.81 In these cases the Middle Chinese reflex
is usually the same as that of the corresponding singleton alveolar consonant, regardless
of its position in the cluster, with the voicing and aspiration specifications of the second
consonant of the cluster. Thus OC *t.k- has the same Middle Chinese reflex as OC *t-;
OC *t.g- evolves to Middle Chinese like OC *d-, and OC *k.dz- evolves like OC *dz-.
The evidence of Vietic languages like Rc is crucial:these languages preserve onsets
with preinitial *k, such as *k.t- and *k.dz-, almost unchanged. Vietnamese predictably
shows the clusters second consonant in spirantized form. Although Lakkia is a Kra-Dai
language, its testimony nicely agrees with Rc in indicating a velar preinitial in paper
and bandit. The Lakkia reflex is voiceless aspirated with high tone if the second conso-
nant in the Chinese source form was voiceless, and plain voiceless with low tone if the
same consonant was voiced. Proto-Mn treats voiceless clusters as plain voiceless stops,
and clusters including a voiced obstruent as voiced aspirates (Table4.47):
We now discuss the specific evolutions for each onsettype.
4.4.4.1 Preinitial *p, *t, *k plus voiceless unaspirated obstruents: type *p.k()-
As noted above, we may have *p.k-in:
TABLE4.47 Middle Chinese, Vietic, Lakkia and Proto-Mn evidence for Old Chinese clusters
of two obstruents
examples MC Rc, Sch Vietnamese Lakkia pMn OC
zh paper tsyeX R . /kcay 3/ giy [zi B1] khjei 3< *kt- *t *k.te
zhng seed tsyowngX R. /kco: 3/ ging [zw B1] *t *k.to
dng lamp tong S. /k ten/ *t *k-t
zi bandit dzok R. /kck/ gic [zak D2] kjak 8< *gdz- *dzh *k.dzk
chung bed dzrjang R. /kc 2/ ging [z A2] *dzh *k.dzra
zhn needle tsyim gm [am A1] the:m 1< *tk- *t *t.[k]m
Both (616) and (617) above were written in bronze inscriptions with forms ancestral
to , such as these from the vessel Y dng (GG 8:509):
In both cases, the component in the upper left-hand corner is the form ancestral toq:
(619)
The Shuwn does not say that q is phonetic in f and f (SWGL 4352b), but
all attempts to explain it as anything else seem rather forced, and q clearly is used
as phonetic in a number of other common words with final*-p:
However, the primary readings of q itself are difficult to derive by regular processes
from a syllable with final*-p:
The difficulty is that MC -joH would normally reflect OC *-ak-s, *-a-s, or *-a-s, and MC
-joX would normally reflect *-a. One possibility is that MC khjoH reflects a western dialect
in which labial codas changed unconditionally to velars:*-ap-s > *-ak-s > -joH; this will
account for khjoH but not for khjoX. Another possibility is that since q depart is likely
have appeared frequently without stress, the coda *-p either dropped or weakened to *- in
unstressed position, and then the unstressed variant was restressed, as with li come:
Note also that fi cast aside could be interpreted as make depart, with prefixed *p
and root *kap relatedto:
(623) *[p-k]ap-s > *pap-s > *pat-s > pjojH > fi (make depart:) castaside
The testimony of VN gm [am A1] bamboo or metal needle and pHmong *kj A
needle points to a *k in the Old Chinese onset of MC tsyim > zhn needle;
preinitial *t- is indicated by Proto-Lakkia (L-Thongkum 1992) *the:m 1 needle
(Table4.23):
(626), , *t.[k]m > tsyim > zhn needle, pMn *t-; cf. pHmong
*kj A, VN gm [am A1], Proto-Lakkia *the:m A, Proto-Tai
*qem A(Pittayaporn 2009:340)
Here, we write *t.[k]- with square brackets because *t.q- would also be a possible recon-
struction:the characters and belong to the xishng series of *[g]r[]m > heam
> xin all; everywhere, which includes both velar and uvular material.
The Shng graph for ten is a simple vertical line; in Zhu bronze inscriptions, the
graph generally has a thick spot in the middle, which eventually developed into the
horizontal stroke in the standard character (GG 2:689690):
{} in bronze inscriptions
Qi Xgu (2004) argues that these forms were originally pictograms representing
{} zhn needle; evidently the two words {} zhn needle and {} sh ten
were similar enough in pronunciation to be written with the same graph. We there-
fore reconstruct
(628) *t.[g]p > dzyip > sh ten, pMn *d-;84 cf. pHM *gjup,
Proto-Lakkia *dzep D (L-Thongkum 1992:64)
Thus the character originally representing {} zhn < *t.[k]m came to be used fre-
quently to write a different word {} sh < *t.[g]p, so in order to remove ambiguity,
an additional element was added to the original graph to represent needle.
Old Chinese onsets 155
We find a similar case with the Shng and early Zhu graphs for MC trjuwX
> zhu elbow and MC kjuwX > ji nine. The graph used for nine is believed
to be the original graph for elbow; it looks like the element representing a
hand or arm (629), but with a conspicuous bend to indicate the elbow joint (630).
This character was used as a loan graph to write the numeral nine, because the
two words were similar in pronunciation; then additional elements were added
when necessary to specify the meaning elbow unambiguously, as in (631), where
a line marks the position of the elbow (J Xshng 2010:197, 348, 991; GG 3.374,
4.433, 10.892):
However, the elbow graph with an added line was easily confused with the graph for
cn thumb; the characters for thumb in (632) are from the Warring States period
(J Xshng 2010:235, GG 3.578):
The graphs (631) for elbow and (632) thumb became confused, and the
result is that the standard graph for zhu elbow has the shape , with what
looks like thumb on the right, and with ru flesh added on the left to reduce
ambiguity.
Thus the paleographical evidence makes it clear that {} zhu < trjuwX elbow
and {} ji < kjuwX nine were similar in pronunciation. But in order to account
for the Middle Chinese readings, we must have *t and *r in {} zhu < trjuwX
elbow, and we must have *k in {} ji < kjuwX nine; we reconstruct them as
in(633).
(633) *t-[k]<r>u > trjuwX > zhu elbow, Show /tou 3/ (implying
pMn*t-)
*[k]u > kjuwX > ji nine, pMn*k-
The *t- in elbow may be the inalienable noun prefix (section 3.3.2.4). The square
brackets indicate the possibility of a labiovelar *k- in place of *k-, and of a complex
onset in the case of nine, perhaps something like *t.ku (compare Written Tibetan
dgu nine).85 The words for elbow and nine also have similar pronunciation in sev-
eral Tibeto-Burman languages, for instance in Garo where both aresku.86
156 Old Chinese
(634) or *t-quk > tsyuwk > zhu rice gruel; pMien *tjuok D porridge
*m-quk > yuwk > y nourish; a related rootis
*quk > xjuwk > x nourish
(635) *[t.p] > *[t.p]u > *tu > tsyuwX > zhu broom; cf. Thai
/p:w C1/ (Pittayaporn) to sweep, pMien *u C < pHM
*nc- to sweep87
*m.b > *bu > bjuwX > f woman, wife, pMn *-b-; cf.
pMien *mbu B, VN v [v B2] < Proto-Vietic*-b
Examples of *k.t-:
(636) *k.to > tsyowngX > zhng seed, pMn *t-; cf. VN ging
[zw B1], Cht /kco/
(637) *k.te > tsyeX > zh paper, pMn *t-; cf. VN giy [zi B1],
Rc /kcaj/
The attested reflexes of preinitial stops before voiceless unaspirated stops are sum-
marized in Tables4.48, 4.49, and4.50.
TABLE4.48 Attested correspondences for Old Chinese voiceless stops with preinitial*p
OC MC pMn VN pHM
*p.t()-
*p.ts()-
*p.k()- p-
*p.q()-
*p-()
Old Chinese onsets 157
TABLE4.49 Attested correspondences for Old Chinese voiceless stops with preinitial*t
OC MC pMn VN pHM
*t.p-
*t.p- tsy-
*t.ts()-
*t.k-
*t.k- tsy- *t g- [] H *k-
*t.k()r- tr- *t
*t.q- t-
*t.q- tsy- *tj-
*t-()-
(638) *[p.q]ra > phaeng > png boil (v.); the phoneticis
*qra > xaeng > hng penetrate:alsoread
*qa > xjangX > xing sacrificial offering
In (639), we reconstruct *t-k- (or perhaps *t--) because while Middle Chinese has
tsyh-, Mn shows /k-/:since the vowel is nonfront, ordinary velar palatalization cannot
account for the Middle Chinese initial. As with elbow, *t- may be the inalienable noun
prefix here (section 3.3.2.4).
(639) *t-[k]() or *t- > *t > tsyhiX > ch front teeth; some
Mn dialects reflect pMn *th-, others *kh-, e.g., Ximn, Fzhu
/ki 3/. In the standard script, the phonetic elementis
*t > tsyiX > zh foot; stop (GG2.555)
(640)
The word ch < tsyhae chariot has a variant reading j < kjo as chariot (still
used as the name of the piece in Chinese chess). In the late meaning spinning wheel,
was borrowed by Written Burmese as khya 3.These elements indicate the presence of
a gutturalvelar or uvularin the onset. The word is also presumably cognate with
yo > y vehicle, carriage; carry on shoulders, which excludes a velar. We reconstruct:
(642) *[t.q](r)A > tsyhae > ch chariot; cf. pHmong *tshjua Aspinning
wheel, VN xa [s A1];cf.
*C.q(r)a > kjo > j chariot
*m-q(r)a > yo > y vehicle, carriage; carry on shoulders
The word , kung < khjwang, mentioned above, designates various kinds of
square-shaped containers. Ode 15.2 includes an example of itsuse:
(643) Mo Sh15.2:
, y y chng zh, wi kungjj
She goes to put them in containers; there are square ones [ kung] and
round ones [ j]. The Mo commentarysays:
fng yu kung, yunyuj
If square they are called kung; if round they are calledj.
Old Chinese onsets 159
We suspect that ~ kung itself is related to other words meaning square men-
tioned above (example (571)), and reconstruct:
(644) *k-pa >*ka > khjwang > kung square basket (aspiration
unexplained):from a relatedroot,
*pa > pjang > fng container, box (Shuwn); with the sameroot:
*C-pa > pjang > fng square, VN vung [vuA1]
The modern character is made up of *pa > pjang > fng box, container,
ostensibly as a signific, and *a > hjwang > wng king as phonetic (Zhu
bronzes actually show the character ancestral to *a instead of ). But since
kung and fng evidently both represent square containers and are phoneti-
cally similar, we suspect that they are related words, both originally written as
(as suggested in Hny d cdin 2001, s.v.). The word *pa > pjang > fng box,
container is nearly homophonous with, and presumably contains the same root as,
*C-pa > pjang > fng square. We take it that the word {} kung < khjwang
contains a related root *pa also meaning square, prefixed with *k-, perhaps as
a deriver of deverbal nouns. The consonant cluster in *k-p- simplified to *k- in
early Zhu times, obscuring the relationship of to and triggering the addition
of as a phoneticas early as mid-Western Zhu, judging from the forms in J
Xshng (2010:913).
The color term chng < trhjeng red is written with jng < keng, a velar-initial
phonetic, but preinitial *t- is required to account for the Middle Chinese initial. We
reconstruct:
(645) *t-kre > trhjeng > chng red; possibly connected to Proto-Tai
*dl/rie A(Li 1977:129,274)88
*k.le > keng > jng a vein of water (Shuwn); in many early
forms, the character contains the lateral-initial phonetic
*l e > thengX > tng, an early formof
*l e > thengX > tng stand up straight
The attested correspondences for *p, *t, and *k before voiceless aspirated onsets are
summarized in Tables4.51, 4.52, and4.53.
TABLE4.51 Attested correspondences for voiceless aspirated stops and affricates with
preinitial*p
OC MC pMn VN pHM
*p.t()-
*p.ts()-
*p.k()-
*p.q()- ph-
160 Old Chinese
TABLE4.52 Attested correspondences for voiceless aspirated stops and affricates with
preinitial*t
OC MC pMn VN pHM
*t.p()-
*t.ts()-
*t.k-
*t.k()r- trh-
*t.k- tsyh- *th ~ *kh
*t.q-
*t.q- tsyh- x- [s] pHmong *tshj-
TABLE4.53 Attested correspondences for voiceless aspirated stops and affricates with
preinitial*k
OC MC pMn VN pHM
*k.p()- kh(j)w-
*k.t()-
*k.ts()-
*k.q()-
(646) *k.druj > drwij > chu hammer, pMn *dh-; cf. Pong /ktuj/ mallet
TABLE4.54 Attested correspondences for Old Chinese voiced obstruents with preinitial*t
OC MC pMn VN pHM
*t.b()-
*t.d()-
*t.dz()-
*t.g-
*t.g- dzy- *d
*t.-
TABLE4.55 Attested correspondences for Old Chinese voiced obstruents with preinitial*k
OC MC pMn VN pHM
*k.b()-
*k.d-
*k.d()r- dr- *dh
*k.d- dzy- (~ g-)
*k.dz- dz- *dzh
*k.dzr- dzr- *dzh
*k.g-
*k.-
We take this to be a dialectal development whereby *k.de went to *g.de and ulti-
mately to *ge, the velar prevailing over the alveolar:
The same dialect development may account for the Middle Chinese in (649):
(649) *[k.d]e > *g.de (dial. >) *ge > gjie > q earth spirit
Normally, we would expect *ge > dzye by the usual palatalization of (nonpharyngeal-
ized) velars before front vowels:the failure of the velar to palatalize in (649) argues
for the presence of an intervening segment between the velar and the vowel at the time
of the first palatalization of velars, which occurred during Western Hn, according to
Schuessler (2010).
The attested correspondences for preinitial voiceless stops before voiced obstruents
are summarized in Tables4.54 and4.55.
Developments varied across dialects in mid- and late Hn times, as the entries
for y < ywit b < pit writing brush in the Shuwn jiz (cited above in
chapter3)show:
Another example is the entry for p < bjij ~ pi < bij kind of wild cat in Fng yn
8.2 (Ho Yxng etal. 1989:892):
Such dialectal pronunciations have found their way into Middle Chinese lexica:thus
the old and the modern words for writing brush: *[m-]rut > ywit > y and *p.[r]ut
> *prut > pit > b both have standard Middle Chinese pronunciations although they have
different dialect bases. Note that both the Shuwn and Gu Ps commentary on the Fng
yn indicate loose clusters [*p.[r]ut] and [bij-li < *b.r] in the W area
(southern Jings and northern Zhjing). The regions of Chn (east Hnn) and Ch
(Hbi) and the regions west of the passes (Shnx) either had tight clusters or lost the
preinitial. The Yn region (Hbi) appears to have gone furthest in the direction of
monosyllabization. It is not surprising, then, to find in Middle Chinese doublets likethese:
(656) *p.rak-s > luH > l dew; disclose, pMn *lh-; Proto-Tai *p.rak
to be exposed (Pittayaporn2009)
(657) *p.ra > ljangX > ling a pair; pMn *lh-; the early graph isjust
*pra > pjaengX > bng third heavenly stem written double (Baxter
1992:272).
(658) *k.nuk > nyuwk > ru meat, flesh; pMn *nh-; Pong /kuk 7/
meat,flesh
(659) *k.r[a]n-s > ljenH > lin chick, pMn *lh-; pHM *klan C
younghen
(660) *k.[r]e[n] > len > lin lotus fruit, pMn *lh-; VN sen [n A1]
lotus
(661) *k.ro > luwng > lng cage, coop, pMn *lh-; VN chung
[tu A2] cage, shed, shelter, coop, stable, sty, Proto-Tai *kro A
cage (Li 1977:225)
(662) *k.ra > langX > lng bright; VN sng [ B1] < *kr- bright,
clear, Rc /plj kra/ sunny weather
(663) *k.ram > lam > ln basket; pMn *lh-; the phoneticis
*[k]ram > kaem > jin inspect
Lakkia, as usual, retains the first of two consonants:if the lost second consonant was
a nasal, nasality is retained on thevowel:
(669) *k-a > kang > gng lift high; Lakkia /kha: 3/ to lift; relatedto
*a > ngang > ng high; lifthigh
(670) jng < keng < *k.le vein of water (Shuwn); in the early script,
the top part depicts a loom, and underneath, the phoneticis
164 Old Chinese
In the meanings regulate and norm, *k-le appears to include a prefix *k- (in an
unidentified function) before a root relatedto
(672) *[t.l]aj > ta > du many, pMn *t-; cf. Proto-Tai *hlai A(so Li
1977)or *laj A(Pittayaporn 2009); Proto-Hlai *hl:y (Norquest
2007:464); used as phoneticin
*laj > ye > y move(v.)
(673) *t-lit > tsyit > zh substance, solid part;cf.
*m.li[t] > zyit > sh fruit; full, pMn *-d-; Proto-Tai *m.lec D
grain (Pittayaporn2009)
(674) *k.m(r)a > mjuX > w dance (v.), pMn *mh- (Norman
1991:211); VN ma [mu B1] to dance [ritually, with fan or
sword or veil]; to brandish, twirl, whirl (note high-register tone),
Rc /kuma/ dance
The atttested correspondences for preinitial stops before voiced resonants are sum-
marized in Tables4.56, 4.57, and4.58.
TABLE4.56 Attested correspondences for Old Chinese voiced resonants with preinitial*p
OC MC pMn VN pHM
*p.m()-
*p.n()-
*p.()-
*p.l()-
*p.r()- l-, p- *lh
TABLE4.57 Attested correspondences for Old Chinese voiced resonants with preinitial*t
OC MC pMn VN pHM
*t.m()-
*t.n- t-, n-
*t.n()r- tr-
*t.n- tsy-
*t.()-
*t.l- t- *t
*t.l()r-
*t.l- tsy-
Old Chinese onsets 165
TABLE4.58 Attested correspondences for Old Chinese voiced resonants with preinitial*k
OC MC pMn VN pHM
*k.m()- m-, k- m- [m] H
*k.n-
*k.n- ny- *nh
*k.()- ng-, k-
*k.l()- k- *k
*k.r()- l- *lh s- [] *kl-
(675) (*l o[r]-s? >) *l ot-s > xjwojH > hu snout; topant
*(t-l o[r]-s?) > *t-l ot-s > tsyhwejH > hu snout; to pant (this word
perhaps corresponds to pMn *thyi C mouth; see discussion below);
the phoneticis
*l o[r]-s > thwanH > tun runningpig
Both the onset and the rhyme of (675) deserve comment. The Middle Chinese read-
ing xjwojH reflects the development *l ()- > x- that we ascribe to western dialects (see
the discussion in section 4.3.5). We reconstruct a pronunciation of with *t-l - to
account for MC tsyhwejH, but pMn *th- would also be the regular reflex of OC *l -,
so the Middle Chinese form could reflect a Mn-like development of a singleton initial
*l - rather than a form with preinitial*t-.92
As for the final, MC -jwojH and -jwejH93 would regularly reflect only OC *-ot-s
or *Cat-s; from *-or we would normally expect MC -jwon, -jwen or -jwe. Norman
(1996:22) noticed that the Proto-Mn rhyme of *thyi C does not regularly cor-
respond to MC -jwojH or -jwejH, and therefore questioned the usual association of
pMn *thyi C mouth with {} tsyhwejH. However, the Proto-Mn rhyme *-yi is
consistent with MC -jwe, so the Mn form could be the regular reflex of OC *l or-s or
*t-l or-s (with *-r > *-j as a development in coastal dialects; see section 5.5.1.4). We
suspect that the readings with -jojH and -jwejH may reflect a pre-Qn dialect in which
*-r and/or *-n became *-t, and that it is Middle Chinese rather than Proto-Mn that
is irregular.94
The characters for xn < xjunH instruct and shn < zywinH follow; obey
are frequently used for one another in excavated documents (Bi Yln 2008:343344).
166 Old Chinese
In the present system, only lateral initials can account for contact between MC x- and
zy-. We reconstruct:
(677) *k-ra > khang > kng tranquil; at ease; the phoneticis
*kra > kaeng > gng seventh heavenly stem, which is also
phoneticin
*[N-]ra > dang > tng exaggerate; great (for the onset, see section
4.4.1.4)
The name of the first Shng ruler, usually written in received texts as thang > tng,
is written in pre-Qn documents as or . While the phonetic *kra indicates OC
*r()-, the phonetic *la normally indicates a lateral:
We suggest that the relatively late use of tng to write the first Shng rulers name
reflects the late merger of *r- with *l - (both later becoming *t- in the east, *x- in the
west; see section 4.3.5):
(679), *ra > thang > tng founder of the Shng dynasty; later writ-
ten , as if from*l a
The character q < khip weep includes l < lip < *k.rp stand (v.) as phonetic,
indicating the presence of a rhotic element in the onset. Moreover the verb appears to be
cognate with the noun *[r][]p-s > lwijH > li tear (n.). We reconstruct
(680) *k-rp > *krp > khip > q weep; the phoneticis
*k.rp > *rp > lip > l stand (v.);cf.
*[r][]p-s (dial. >) *rup-s > *rut-s > lwijH > li tears(n.)
The word ko < khawX old; deceased father includes lo < lawX old, an
indication of a rhotic in the onset of . If it had been medial *-r-, we would expect
a division-II final like MC -aew; but since in this case we have MC -aw instead,
the rhotic must be main syllables initial; and since preinitials are never aspirated
in the present framework, we must be dealing with initial *r- preceded by k. We
reconstruct:
Old Chinese onsets 167
(681) *k-ru > khawX > ko old; deceased father; from related roots
*ru and *ru, wehave
*C.ru > lawX > lo old; pMn *lh- (e.g., Jinyng /seu5/)
*s-ru > srjuwX > suwX > su old man (see section 4.4.3.3)
A probable example of *p.r - is the verb pn < phjiengH inquire about (mar-
riage) whose phonetic also occurs in chng < trhjengX gallop. We reconstruct:
The attested correspondences for preinitial stops plus voiceless resonants are sum-
marized in Tables4.59, 4.60, and4.61.
TABLE4.59 Attested correspondences for Old Chinese voiceless resonants with preinitial*p
OC MC pMn VN pHM
*p.m()-
*p.n()-
*p.()-
*p.l ()-
*p.r()- ph-
TABLE4.60 Attested correspondences for Old Chinese voiceless resonants with preinitial*t
OC MC pMn VN pHM
*t.m()-
*t.n()-
*t.()-
*t.l -
*t.l - tsyh- *th?
*t.r-
TABLE4.61 Attested correspondences for Old Chinese voiceless resonants with preinitial*k
OC MC pMn VN pHM
*k.m()- kh-
*k.n()-
*k.()-
*k.l ()-
*k.r()- kh-
168 Old Chinese
In some onsets, a preinitial consonant must be supposed but cannot be identified because
it has been lost in all the pronunciations under consideration.
(683) *C.qra > kjaengX > jng bright; image (also shadow),
*C.qra-s > kjaengH > jng mirror, pMn *k-; VN gng
[ A1];cf.
*qra > 'jaengX > yng shadow(n.)
TABLE4.63 Attested correspondences for Old Chinese voiceless obstruents with unidentified
preinitials
OC MC pMn VN pHM
*C.p()- p- *p v- [v]H
*C.t- t- *t d- [z] H
*C.t()r- tr- *t d- [z] H *tr-
*C.t- tsy- *t gi- [z] H
*C.ts()- ts- *ts gi- [z] H
*C.s()- s-, ts- *s, *tsh r- [z] H
*C.k()- k- *k g- [] H *k-
*C.q()- k- *k g- [] H
(685) *C.saw-s > sawH > zo shout; VN rao [zu A1] shout
(686) *C.saw > sawX > zo dry; VN ro [zu B1]dry
(687) *C.[s]a > sjang > xing box (of a carriage); VN rng
[z A1] box,trunk
In these examples, the occurrence of initial affricates in modern Chinese dialects (for
instance in the Mandarin forms in (685) and (686)) instead of the [s] that would be
expected on the basis of Middle Chinese also supports the reconstruction of a tightly
attached preinitial*C.
Attested correspondences for preinitial *C before plain voiceless obstruents are sum-
marized in Table4.63.
TABLE4.64 Attested correspondences for Old Chinese voiceless aspirated initials with
unidentified preinitials
OC MC pMn VN pHM
*C.p()-
*C.t()-
*C.ts()-
*C.k()-
*C.q()- kh- *kh
*C.qp-s > *C.qt-s > *kt-s > khj+jH > q (inhaled thing:) breath,
air, vapors, pMn *kh-; the rootis
*q(r)p > xip > x inhale, pMn*kh-
Note that the use of *C.qt as phonetic in *C.qp-s (inhaled thing:) breath,
air, vapors already reflects the early change of *-p-s to *-t-s. The word for breath
*C.qp-s is best understood as a deverbal noun with *-s suffix from a *C-prefixed verb
cognate with x inhale.
(690) *C.qar-s > khjwonH > qun encourage; the ultimate phonetic
element is saidtobe
(= , ) *qar > xjwon > xun clamor,shout
(691) *C.g(r)aj > gje > q straddle; ride; pMn *gh-; pHM *ej Atoride
(692) *C.a[n] > hjwon > yun garden; pMn *-; pHmong *wa A
garden
It appears that *C.- has the same reflexes as *C.g-: MC g- and pMn *gh-, as
in(693):
Old Chinese onsets 171
(693) *C.[]r > gj+n > qn cress, pMn *gh-, Ximn /kun 2/; the pho-
netic element usually represents a uvular initial,asin
*qr > xj+n > xn rejoice
But the reflexes of *m-q- are different:MC y- and pMn *-, as in (694) (= (479) in sec-
tion 4.4.2.1 above; see also (511) for paleographical evidence for a uvular):
(694) *m-q(r)a > yoX > y give; for; and (volitional *m-), Ximn
/h 6/ give < pMn *-; relatedto
*C.q(r)a > kjoX > j lift,raise
When we lack specific evidence that the root began in a voiced obstruent, we use
square brackets to indicate uncertainty on the voicing specification of the initial and on
the nature, nasal or not, of the preinitial. Thus the notation *C.[b]- is intended to serve
for both *m.p- and *C.b-. Examples:
The attested correspondences for preinitial *C plus voiced obstruents are summa-
rized in Table4.65.
TABLE4.65 Attested correspondences for Old Chinese voiced stops/affricates with unidentified
preinitials
OC MC pMn VN pHM
*C.b()- b- *bh
*C.d- d- *dh
*C.d- dzy- *dh
*C.dz()- dz- *dzh
*C.g() h- *
*C.g()- g- *gh *-
*C.()- h- *
*C.- g- *gh
*C.- h(j)- * *w-
(702) *C.me[n]-s > mjienH > min face; pMn *mh-, Lfng Hakka
/mian 5/ (high register); pMien *hmien Aface
(703) *(C).m[e]j > mejX > m millet or rice grains, dehusked and pol-
ished; Mixin Hakka /mi 3/ (high register); pMien *hmei B rice,
dehusked
(704), *C.m[r]> mjun > wn mosquito, pMn *mh-; cf. Proto-Vietic
*t.mu:l midge
(705) *C.n[a]m > nyemX > rn to dye, Mixin Hakka /niam 3/; VN
nhum [um B1]tint
(706) *C.ni[] > nen > nin harvest; year, pMn *nh-, Mixin Hakka
/nian 1/; pHM *huHyear
(707) *C.a > nguX > w five, pMn *h-, Mixin Hakka
/ 3/; Proto-Tai (Pittayaporn 2009) *ha C, Proto-Lakkia * C
five (L-Thongkum 1992)
(708) *C.ra[j] > ngwaeX > w roof tile; pMn *h-, Mixin Hakka
/a 3/; VN ngi [i B1] tile; pHM *X tile96
(709) *C.lru > drjuwng > chng insect; pMn*dh-
(710) *C.law > daw > to peach; pMn *dh- pHM *lw Apeach
(711) *C.ru > lawX > lo old; pMn *lh-, Mixin Hakka /lau 3/old
(712) *C.r > liX > l plum; pMn *lh-; pHM *hli Xplum
(713) *C.r[a]t > ljet > li cold, raw; cf. VN rt [zt D1] cold (Bodman
1980:85)
(714) *C.(r)o > nguwX > u lotus root; pMn *hu B, Mixin
Hakka /eu 3/, VN ng [ B1], Thai /au B2/ ~ /hau C1/ (Norman
1991:210)97
(715) *C.rar > ngaen > yn face, forehead; pHmong *hen Aforehead
Old Chinese onsets 173
TABLE4.66 Attested correspondences for Old Chinese voiced resonants with unidentified
preinitials
OC MC pMn VN pHM
*C.m()- m- *mh m- [m]H *hm-
*C.n- n- *nh *hn-, *h-F
*C.n- ny- *nh nh- [] H
*C.()- ng- *h ng- [] H *h-F
*C.l- d- *dh *l-
*C.l()r- dr- *dh pMien *lj-
*C.l-
*C.r()- l- *lh r- [z] H *hl-, *r-
We suppose that MC d- and dr- from Old Chinese lateral sources like *C.l- and
*lr- evolved through [ld] and [ldr]; see Table4.66. We take the curious-looking cluster
*l- of the Proto-Hmong-Mien word for peach in (710) to be an attempt at rendering
[ld] in the donor language:similarly, for *lra (> ldra) > drjang > chng intestines,
Proto-Mienic has *lja Aintestines.
Reflexes of preinitial *C plus voiced resonants are summarized in Table4.66.
In words with Old Chinese stops as main syllable initials, softening in Proto-Mn nor-
mally indicates a loosely attached presyllable. However, as shown in section 4.2.1.1,
not all Northern Mn words with softened initials belong to the inherited layer:some
(among those that have voiced initials in Middle Chinese) appear to be loans from
voicing-preserving dialects; care should be taken to exclude them, or we risk recon-
structing nonexistent preinitials for Old Chinese. When comparisons are available, we
expect Vietnamese to show spirantization corresponding to Northern Mn softening,
although examples are few. Aspecific indication of a loosely attached nasal preini-
tial is when Hmong-Mien shows prenasalization without the initial being voiced in
Middle Chinese.
In words with nasals as main syllable initials, a loosely attached preinitial is
reconstructed when words with Middle Chinese nasals show a preinitial consonant
in a contact language, and Southern Mn denasalizes the nasal. (In a tight cluster,
we would expect the nasal to remain and to propagate nasality rightward; see sec-
tion 4.2.1.2.)
174 Old Chinese
In our earlier work (Sh Jir [Sagart] and Bi Ypng [Baxter] 2010), we did not
reconstruct a loosely attached counterpart of our *N- preinitial; we assumed the
stative intransitive prefix *N- always occurred tightly attached to the following
consonant. But we now reconstruct preinitial *N- besides *N- to account for words
having a prenasalized initial in Hmong-Mien and a softened initial, voiced or voice-
less, in Proto-Mn:
(716) *N.p[u][t]-s > pj+jH > fi boil (v.i.), pMn *-p-, pHM *mpuiH
boil (v.i.)
(717) *N.tsu > tsawX > zo early; pMn *-ts-, pHM *ntsi ouX early
(718) *N-grut > hweat > hu slippery; pMn *-g-, pHM *uat
smooth/slippery
This evidence points to a nasal prefix occurring with stative and intransitive verbs,
which does not voice a voiceless initial in Middle Chinese, but does produce softening
in Mn:this matches the description of a loosely attached form of *N. We write this as
*N in order to underline the parallel with other preinitials.98
(719) *N-[k]j > khoj > ki to open (v.i); Mien /goi 1/ < pHM *kh-
to open, as a flower, theheart
TABLE4.68 Attested correspondences for Old Chinese aspirated stops with preinitial*N
OC MC pMn VN pHM
*N.p()-
*N.t()-
*N.ts()-
*N.k()- kh- *k(h)-
*N.q()-
*N.q()r- trh- *ntsh-
(Our earlier reconstruction for this word was *N-kj; see the discussion in sec-
tion 4.4.1.2.) Compare the transitiveverb
(720) *[k]j > khoj > ki to open (v.t.); Mien /khoi 1/ < pHM *kh- to
open (v.t.)
In such examples we do not expect Mn to be of help, since aspirated stops are not
affected by softening in Mn. Under our new interpretation it is no longer necessary to
assume that *N does not voice an aspirated stop (as assumed in Sagart 2003 and (Sh
Jir [Sagart] and Bi Ypng [Baxter] 2010). We can now use *N- to explain transi-
tivity alternations between voiced and aspirated obstruent initials, or between a nasal
and a voiceless fricative:for instance, a single root *kok can account for both q
< khjowk < *k(r)ok to bend, bent and j < gjowk < *N-k(r)ok bent, curved, and
a single root *qak can explain both h < xaek < *q<r>ak to frighten and <
ngak < *N-qak scared. Such word-family alternations were unexplained under the
hypothesis that *N- does not voice a following aspirate. Additional examples:
(724) *N-grut > hweat > hu slippery, pMn *-g-; pHM *uat
smooth/slippery
See Table4.69.
176 Old Chinese
TABLE4.69 Attested correspondences for Old Chinese voiced stops with preinitial*N
OC MC pMn VN pHM
*N.b()-
*N.d()-
*N.dz()-
*N.g- h- *-g *-
*N.g-
*N.()-
TABLE4.70 Attested correspondences for Old Chinese voiced resonants with preinitial*N
OC MC pMn VN pHM
*N.m()-
*N.n()-
*N.()-
*N.l()-
*N.r()- l- *gr- (?)
(725) *[N-r]ok-s > luwH > lu leak (v.), pHmong *gro Cleak
where Proto-Hmongic *gro C leak provides evidence for a nasal preinitial while sug-
gesting a velar articulation of *N in *N; the pHmong *-g- can be considered a parasitic
segment arising between * and *r. The fact that the preinitial is lost in Middle Chinese
and that the evolution is not to MC y- implies that the preinitial was loosely attached
(from *N.r- we would expect MC y-). The fact that the verb is intransitive and nonvo-
litional also argues for *N against *m. See Table4.70.
(726) *m-tam > tam > dn carry on the shoulder, pMn *-t-; pHM
*ntam carry on the shoulder
Old Chinese onsets 177
TABLE4.71 Attested correspondences for Old Chinese voiceless stops with preinitial*m
OC MC pMn VN pHM
*m.p()- p- *mp-
*m.t- t- *-t *nt
*m.t-
*m.ts()- ts- *nts-
*m.s()-
*m.k()- k- *-k *kj-
*m.qr- - *kl-
*m.-
(727) *m.ka > ku > g mushroom, pMn *-k-; pHM *kju mushroom
(728) *m-tsaw > tsawX > zo bleach; wash; pHM *ntswXwash
(729) *m.pn > pjunX > fn flour; pHM *mpw:n B flour (Wng
and Mo 1995:89,664)
(730) *m-q<r>[u]m-s > 'imH > yn shade; pMien (L-Thongkum 1993)
*glom C < *kl- shade
(731) *m-sa[n]-s > sanH > sn scatter (v.t.); cf. pMien *dzhan C
< pHM *ntsh- to disperse, VN ran [zn A1] to disperse,
propagate
(732) *m-prak > phaek > pi to strike; pHM *mpjk clap (Wng
and Mo1995)
178 Old Chinese
TABLE4.72 Attested correspondences for Old Chinese aspirated stops with preinitial*m
OC MC pMn VN pHM
*m.p()- ph- *mp-
*m.t-
*m.t- tsyh- *nthj-
*m.ts()-
*m.k()-
*m.q()-
(733) *m-ta-s > tsyhangH > chng to lead (in singing); pHM
*twj: Ato sing (Wng and Mo 1995:252, 563, attested in
Mieniconly)
(734) *m-t-s > tsyhingH > chng steelyard; pHM *nthjuH
balance
In the last example, the Kra-Dai language Monn confirms Hmong-Mien pre-
nasalization:/nda 5/ to weigh (Ling Mn 1980:102). Note the failure to aspirate the
initial in two out of three Hmong-Mien examples (see also example748 in 4.5.2.5):per-
haps this reflects a phonotactic constraint on aspiration in early Hmong-Mien. The
attested correspondences are summarized in Table4.72.
(735) *m.b > bjuwX > f woman, wife, pMn *-b-; VN v [v B2]
wife; pMien *mbuB
(736) *m-ba-s > buH > b step, pMn *-b-; VN v [v C2] go on foot,
walk; Yo /bia 6/ < *mb- step, stride
(737) *m-dra > drjoX > zh ramie; flax, pMn *-d-; pHM *nduH
ramie,flax
(738) *m.dz()-s > dziH > z breed, love (v.); character, pMn *-dz-;
*pMien *ndza C word, character
(739) *m-g(r)o > huw > hu monkey, pMn*-g-
TABLE4.73 Attested correspondences for Old Chinese voiced obstruents with preinitial*m
OC MC pMn VN pHM
*m.b()- b- *-b v- [v]L *mb-
*m.d-
*m.d()r- dr- *-d *nd-
*m.dz()- dz- *-dz *ndz-
*m.g- h- *-g
*m.g-
*m.()-
found in classical texts of a word *ma > mju > w being used not as a negation
but as a verbal prefix. Thus in Ode 235.5 the phrase (w nin r z),
which at first sight appears to mean do not remember your ancestors!, is explicitly
said to mean remember your ancestors! in the Mo commentary100 and do not for-
get your ancestors! in r y 3.81.101 We take this *ma as representing
the loosely attached volitional prefix we reconstruct as *m. This prefix attaches to
the verb *nim-s > nemH > nin think of giving it volitional semantics, thereby
licensing its use in an imperative.
An example of *m.r-is
(740) *m.rk > *m. r > loj > li come;102 it is the phonetic elementin
*m-rk (dial.) > *mrk > meak > mi wheat
When words with MC l- have initial *-d- in Proto-Mn, and there is evidence for a
labial nasal in the preinitial, either from the characters xishng series or from contact
languages, we reconstruct *m.r-. Likewise, when words with MC zy- have *-d- in
Proto-Mn, and there is evidence for a labial nasal in the preinitial, we reconstruct
*m.l-. Evidence for a labial preinitial comes from the phonetic series in the case of the
following MC l- words, which show *-d- in Proto-Mn (Lu Jiru [Norman] 2005:4):
TABLE4.74 Attested correspondences for Old Chinese voiced resonants with preinitial*m
OC MC pMn VN pHM
*m.n- n-
*m.n-
*m.()-
*m.r()- l- *-d
*m.l- zy- *-d
TABLE4.75 Attested correspondences for preinitial *m plus Old Chinese voiceless resonants
OC MC pMn VN pHM
*m.m()-
*m.n()-
*m.()-
*m.r()-
*m.l - th- *nt-
*m.l -
(745) *m.lat > zyet > sh tongue, pMn *-d-; pHM *mblet tongue
(746) *m-lk > zyik > sh eat, pMn *-d-; pHmong *mblj C to have
food withrice
(747) *m.li[t] > zyit > sh fruit; full, pMn *-d-; Proto-Tai *m.lec D
grain (Pittayaporn2009)
(748) *m-l ot > thwat > tu peel off; cf. pMien *dut (< *nt-) to peel
off/escape
In (748), *m-l ot was evidently borrowed after the change of *l - > *t- in Chinese,
but while the nasal preinitial was still present. Unaspirated treatment by Hmong-
Mien of Chinese aspirated stops after preinitial *m was noted in section 4.5.2.2. See
Table4.75.
Onsets with preinitial *s- are inferred when a Proto-Mn softened consonant matches a
Middle Chinese initial that is indicative of the phonetic effectof*s.
Old Chinese onsets 181
(749) *s.dur > zywin > chn lip, pMn *-d- (?): (Fzhu /su 2/,
Jinu /y 3/);103 the phoneticis
*[d]r > dzyin > chn fifth earthly branch
The character chn belongs to a xishng series with alveolar stop initials. Old
Chinese singleton *d- should evolve to MC dzy-, so the development here to zy- shows
the influence of a preinitial. Similarly, Proto-Mn initial *-d- is the softened counter-
part of *d-, which reflects Old Chinese nonpharyngealized *d-:by our assumptions it
makes sense that the Old Chinese onset had initial *d- preceded by a loosely attached
preinitial. The effect of the preinitial in Middle Chinese was to turn an affricate into a
fricative. We saw in section 4.4.3 that this is precisely the effect of tightly attached *s
before nonpharyngealized obstruents like *t-, *ts-, *dz-,*k- (with front vowels), and *g-
(with front vowels):thus we saw (in section 4.4.3.3) that the sequence *s.g- > *zg- >
*zd- > zy- explains the Middle Chinese fricative in *s-gij-s > zyijH > sh show
(v.) (567). With OC *s.d-, however, we claimed that after voicing assimilation of the
preinitial on the initial, the resulting cluster [zd] was simplified to z before the alveolar
stop could be changed to a palatal affricate, thus preempting any fricativizing effects
of *s on the following obstruent (section 4.4.3.3). We assume that after these changes
affecting OC *s.d- took place, *s.d- simplified to *s.d-, filling the gap left by original
OC *s.d-; the new *s.d- underwent voicing to *zd, after which palatalization occurred,
giving *zd-, and finally fricativization tozy-.
An example of *s.-is
(750) *s.o > zjowng > sng pine (n.), pMn *-dz-; the phoneticis
*C.qo > kuwng > gng impartial, just; public104
182 Old Chinese
TABLE4.76 Attested correspondences for Old Chinese voiced stops with preinitial*s
OC MC pMn VN pHM
*s.b()-
*s.d-
*s.d- zy- *-d
*s.dz()-
*s.g()-
*s.-
*s.- z- *-dz *-
Compare with the evolution of Old Chinese *s-- to Proto-Mn *dz- in (4.4.3.3).
Sagart and Baxter (2009) showed to be a uvular phonetic, as indicated by, for
example, wng < 'uwngH < *qo-s earthen jar and rng < yowng < *[](r)o
appearance (now written ). Here again we assume that after original *s.- shifted to
*z-, its loosely attached counterpart *s.- filled the gap, shifting to *s.-, undergoing
regressive voicing assimilation to *z-, and ultimately merging with original *s-- as
MC z-. Another exampleis
(751) *s.A > zjae > xi awry, pMn *-dz- (Norman 1996:34); pHmong
(Wng Fsh 1979)*ei A, askew, slanted
(752) *s-lAk-s > zjaeH > xi decline, renounce, pMn *-dz- (Jinyng
/lia 6/); cf. VN gi [z C2] takeleave
We assume that Proto-Mn branched off at a time when the preinitial was still loosely
attached, so that the result was the softened initial *-dz-. The word was loaned to
Vietnamese before the cluster simplified, the preinitial causing spirantization: gi
[z C2] take leave. Even in the absence of comparative evidence we take MC z- in
words with Old Chinese lateral initials as indicating *s.l-. Examples:
(753) *s.loj > zjwe > su follow; with the same phonetic,cf.
*loj > dwaX > du tofall
Old Chinese onsets 183
(755) *s.lu[n]> zwin > xn docile; gradually; with the same phonetic,cf.
*C.lu[n]-s > zywinH > shn follow;obey
*l u[n]-s (dial. >) *xun-s > xjunH > xn instruct
Reconstructing the loosely attached onsets *p, *t, and *k requires two independent
elements:evidence for a specific stop in the preinitial, from a Vietic language, Lakkia,
or a Chinese dialect; and evidence that the preinitial was loosely attached. In words with
obstruent initials, Northern Mn softening provides the evidence for a loosely attached
onsetexcept that Old Chinese aspirated stops are never softened in Mn. (In words
with resonant initials, a Proto-Mn plain resonant *m-, *n-, *l-, etc. can reflect a loosely
attached preinitial stop, but not a tightly attached one, which would produce pMn *mh-,
*nh-, *lh-, etc.) Few words satisfy this double requirement.
TABLE4.78 Attested correspondences for Old Chinese voiced obstruents with preinitial*k
OC MC PM VN pHM
*k.b()-
*k.d- d- *-d d- [z]L
*k.d-
*k.dz()- *-dz
*k.g()-
*k.()-
Vietic provides us with a second example:the word du < duwH neck, which
has softened *-d- in Proto-Mn, has a velar preinitial in Rc:/kadk/ nape (Nguyn Ph
Phong etal. 1988). The corresponding Vietnamese word is dc [zk D2] fleshy leaf-stalk
of certain plants; back of a knife; pipe; stem (of a steelyard), with spirantized d-[z]:
(758) *k.dok-s > duwH > du neck, pMn *-d-; Rc /kadk/ nape
of the neck, VN dc [zk D2] fleshy leaf-stalk of certain plants
[etc.], Proto-Vietic *k-k nape of theneck
(759) *[k.l]u > dawX > do way, pMn *-du B far; pHM *kluX
road/way
with square brackets in recognition of the uncertain nature of the Mn evidence. If this
evidence should turn out to be spurious, the Hmong-Mien evidence by itself would still
point to either *k.lu or *k.lu.
The word nio < newH urine has plain *n- in Proto-Mn. The Jntin and
Lil dialects of Lakkia give /kiu B1/ (L-Thongkum 1992), with nasality on the
first vowel continuing the Old Chinese nasal initial, and showing a preinitial with *k.
The Old Chinese onset cannot have been *k.n-, since that should result in Proto-Mn
*nh-. We reconstruct:
Old Chinese onsets 185
TABLE4.79 Attested correspondences for Old Chinese voiced resonants with preinitial*k
OC MC PM VN pHM
*k.m()-
*k.n- n- *n
*k.n-
*k.()-
*k.r()- l- *l
*k.l- d- *-d *kl-
*k.l-
(760) *k.newk-s > newH > nio urine, pMn *n-; Lakkia /kiu B1/
(Jntin and Lil dialects:L-Thongkum 1992:66)
The word lu < lak fall (v.) has *l- in Proto-Mn (Jinyng /l 8/, Show /lo 6/),
but Southern Mn shows a form with a preceding syllable /ka/:for example, Ximn /
ka-lau 8/ to fall, as things (Douglas 1899:297). This presyllable is matched in the
northern Chinese periphery by preinitial /k-/ in the dialect of Pngyo (Shnx),
currently assigned to the Jn dialect group:/k-l/ to fall in small quantities (Hu
Jngy 1989:200). This velar preinitial moreover makes good xishng sense, since
other words in the same phonetic series, beginning with the head-word, have a velar
stop in their onset. We reconstruct:
(761) *k.rak > lak > lu fall (v.), pMn *l-, Ximn /ka-lau 8/ to
fall, as things (Douglas 1899:297); Pngyo /k-l/ to fall in
small quantities, as earth, ashes (Hu Jngy 1989:200); see Sagart
(1999c:99). With the same phonetic:
, *krak > kaek > g go to (The earliest use of the character
is to represent this word; see Chn Chshng 1987:123.) Also prob-
ably related:
*C.rak-s > luH > l road, pMn *-d- (see (801) in section 4.5.5.3)
4.5.4.3 Preinitial *p, *t, *k plus voiceless resonants: type *k.l ()-
We may have an instance of a voiceless resonant preceded by *k in the word sh <
syep catch, grasp:
(762) *k.nep > syep > sh catch, grasp; cf. Proto-Tai *ni:p to pinch
(Pittayaporn 2009); the phonetic (also used to write {})is
*nrep > nrjep > ni promise; whisper in onesear
We reconstruct *n- in sh < syep < *k.nep because of the xishng connections with
initial nasal, as indicated by the phonetic series (cf. *nrep > nrjep > ni trample etc.)
186 Old Chinese
but Lakkia /khj: p 7/ shows a velar stop preinitial before the nasal, regularly reduced to
nasality on the vowel. Thus we reconstruct:
In many cases, evidence for a loosely attached preinitial is not accompanied by spe-
cific evidence on the nature of the preinitial consonant. In such cases we reconstruct
preinitial*C.
(764) *C.pan > pan > bn move; pMn *-p- (Jinyng /voi9/)
(765) *C.pan > pjonX > fn reverse (v.); pMn *-p- (Jinyng /vai3/)
(766) *C.pat > pjot > f fly forth, send forth; pMn *-p- (Shbi /buai3/)
(767) *C.p[r]> pj+j > fi fly (v.); pMn *-p- (Jinyng /ye9/)
(768) *C.p > pong > bng collapse (v., of a mountain); pMn *-p-
(Jinyng /vai 9/); pHmong *p Atofall
(769) *C.tk-s > tojH > di carry on the head; pMn *-t- (Jinyng
/lue9/)
(770) *C.tar > tan > dn single, simple; pMn *-t- (Jinyng /lue9/)
(771) *C.trok > traewk > zhu chop, cleave; pMn *-t- (Jinyng
/lo3/)
(772) *C.ts[]m > tsom > zn hairpin; pMn *-ts- (Jinyng /la9/)
(773) *C.ts[r]-s > tsenH > jin grass, fodder; pMn *-ts- (Jinyng
/lu 9/) straw mattress105
(774) *C.tsu[t]-s > tswijH > zu drunk (adj.); pMn *-ts- (Jinyng
/ly9/)
(775) *C.kaw > kaw > go lard (n.); pMn *-k- (Jinyng /au9/)
(776) *C.kro > kuwX > gu dog; pMn *-k- (Jinu /e 3/); pHM
*qluwXdog
(777) *C.kot > kjwot > ju bracken (a kind of edible fern); pMn *-k-
(Jinyng /ue 9/); pMien *kjt Dfern
Old Chinese onsets 187
(778) *C.ka > kuX > g male (bovine); pMn *-k- (Jinyng/o3/)
(779) *C.k<r>[j]> kij > j hungry; pMn *-k- (Jinyng /ue9/)
(780) *C.kra > kaeX > ji borrow; false; pMn *-k- (Jinyng /a 3/);
VN g [ B1] (in g ting use someone elsesname)
In b < puX to patch, we have conflicting evidence on the nature of the pre-
initial consonant:Proto-Mn softened *-p- (Shbi /bio 3/) indicates a loosely attached
preinitial; Proto-Hmong-Mien has *mpjaX repair/mend, implying *m-pa; yet Rc
/tpa 3/ to patch points to *t-pa or *s-pa:either we are dealing with two verbs
based on the same root, or there was a single form with complex onset *t-m-pa
or *s-m-pa. We are unable to solve the problem on present evidence and simply
reconstruct
We reconstruct *C.s- when Vietnamese shows high-register r- [z] for MC s-, and
when modern Chinese dialects also give[s]:
(783) *C.sre > srje > shi to strain off wine; VN ry [zi A1] sieve;
tosieve
(784) *C.q(r)aj > 'jeX > y lean on, the origin of later *C.q(r)aj >
'jeX > y chair; VN gh [eB1]
The verbs chng < dzyang taste (v.) and shng < dzyangX ascend were
segmental homophones in Middle Chinese. Both have softened initials in Mn
(Jinu /io 4/, Jinyng /io 5/ ascend; Jinu /io 3/, Jinyng /io 9/ taste).
We reconstruct
For shng we find three forms from the same root:two verbs and a noun. Mn
dialects show different forms for to go up (with pMn *-d-, tone B) and to put up,
cause to ascend (pMn *dh-, tone B), and a third form for the noun top (pMn *d-,
tone C):106
(787) *C-da > dzyangX > shng ascend, pMn *-dio B (Ximn
/tsi 6/, Jiyng /tsi 4/, Fzhu /suo 6/, Jinu /io4/)
*m-da > dzyangX > shng to put up, pMn *dhio B (Ximn
/tsi 6/, Fzhu /tsuo6/)
*da-s > dzyangH > shng top, above (n.), pMn *dio C (Ximn
/tsi 6/, Jiyng /tsi 6/, Jinu /tsio6/)
The forms meaning ascend have a softened initial in Northern Mn, indicating
a preinitial *C- (the first consonant cannot be identified); the causative sense has
the prefix *m-; and the noun appears to have the root *da with a nominalizing
*-s suffix. The distinction between the two verbs has been lost in Middle Chinese,
which preserves only the tonal distinction between the verbs (dzyangX) and the noun
(dzyangH).
We reconstruct *C.- to account for Proto-Mn *-d- in yn < yem eaves and
yng < yangX itch:
(788) *C.am > yem > yn eaves, pMn *-d- (Ximn /tsi 2/, Fzhu
/sie 2/)
Old Chinese onsets 189
(789) *C.a > yangX > yng itch, pMn *-d- (Ximn /tsiu 6/,
Fzhu /suo 6/, Jinyng /io 6/); the phonetic element is
ultimately
*a > yang > yng sheep
We assume that Normans Proto-Mn *-d- (which we suspect was actually *C.d-)
resulted from a change whereby *[j], the post-Old-Chinese outcome of nonpharyngeal-
ized *-, was strengthened to -d- in intervocalic position:*C.- > *C.j- > *C.d-
(cf. Latin major > Italian maggiore). This strengthening change occurred before the
Proto-Mn stage in an area of south China where the ancestor of Wxing was also
located:Wxing has /dzo 3/ itch.
Here are some other examples of Mn softened initials from voiced stops after
preinitial*C:
(790) *C.bo[t]-s > bjojH > fi bark (v.); pMn *-b- (Shbi /by6/)
(791) *C.dar > dan > tn shoot pellets; pMn *-d- (Shbi /duai 2/
pluck a stringed instrument)
(792) *C.[g]i[n] > dzyinX > shn kidney, pMn *-gin B gizzard, e.g.,
Fzhu /kei 6/, Jinyng /i 5/ (Norman 2006:138)
(793) *C.[g](r)o > huwX > hu thick; pMn *-g- (Jinyng /eu 5/),
pMien *u B thick
TABLE4.81 Attested correspondences for Old Chinese voiced stops with unidentified loose
preinitials
OC MC pMn VN pHM
*C.b()- b- *-b
*C.d()- d- *-d
*C.d()r- dr- *-d
*C.d- dzy- *-d
*C.dz- dz- *-dz
*C.dz()r- dzr- *-dz
*C.g- h- *-g *-
*C.g- g- *-g
*C.-
*C.- y- *-d
190 Old Chinese
The argument for reconstructing *C.n- here is as follows. Hakka distinguishes between
two sets of sonorants:in shngshng words, most kinds of Hakka have tone 1 corre-
sponding to Proto-Mn plain sonorants like *n- < OC *n-, and tone 3 corresponding to
Proto-Mn aspirated sonorants like *nh- < *C.n- (Norman 1989). So if ni were from
*C.n-, with a tightly attached voiceless preinitial, we would expect tone 3 in Hakka,
not tone 1.But a voiceless preinitial is required to account for the *- onset in Proto-
Mienic. We account for this by reconstructing a loosely attached preinitial *C.n-.
Old Chinese *C.l()- and *C.r()- have softened reflexes in Mn, so these are easier
to identify; the correspondences are as shown in Table4.82.
Examples of *C.l()-:
(795) *C.lk-s > dojH > di bag; pMn *-d- (Jinyng /lui 6/), pMien
*di Cbag
(796) *C.lot > dwat > du seize; pMn *-d- (Jinyng /lue8/)
(797) *C.lraj > drje > ch pool (n.); pMn *-d- (Shbi /di2/)
(798) *C.lAj > zyae > sh snake; pMn *-d- (Ximn /tsua 2/, Fzhu
/sie 2/; VN x [sA2]
(799) *C.lo[n]> zywen > chun boat; pMn *-d- (Shbi /yi2/)
(800) *C.lAk > zyek > sh hit with bow and arrow; pMn *-d-
(Jinyng /ia8/)
As we saw in section 4.5.2.4, Proto-Mn has *-d- from OC *m.r()- (Norman 2005).
When Proto-Mn has *-d- corresponding to MC l-, and the identity of the preinitial can-
not be ascertained, we reconstruct *C-r()-, as in this example:
(801) *C.rak-s > luH > l road; pMn *-d- (Jinyng /lio 6/, Show
/tio 6/); Lfng Hakka/lu6/
TABLE4.82 Correspondences for Old Chinese laterals after loosely attached preinitials
OC MC pMn VN pHM
*C.l- d- *-d *d-
*C.l()r- dr- *-d
*C.l- zy- *-d x- [s-] L
Old Chinese onsets 191
TABLE4.83 Attested correspondences for Old Chinese voiced resonants with unidentified loose
preinitials
OC MC pMn VN pHM
*C.m()-
*C.n()-
*C.n()r- nr-
*C.()-
*C.l- d- *-d *d-
*C.l()r- dr- *-d
*C.l- zy- *-d x- [s-] L
*C.r()- l- *-d
Here Lfng Hakka /lu 6/ road confirms the absence of a voiceless consonant in
the onset (one would expect tone 5 otherwise). The preinitial consonant was presum-
ably a nasal, but we cannot tell from the semantics whether *m or *N is more
appropriate.
The attested correspondences of preinitial *C before sonorants are summarized in
Table4.83.
The principles outlined in the preceding sections for singleton, tightly attached, and
loosely attached onsets cannot explain all the combinations we find among the pro-
nunciations under consideration:in some cases complex onsets have to be supposed.
Acomplex onset is one that includes two preinitial consonants, sometimes with a schwa.
The first consonant is always a prefix; the second can be a root preinitial consonant.
When two prefixes are involved, they may belong to different time layers, with the outer
prefix being more recent. For that reason the order in which they appear is not strictly
predictable. We have already discussed some instances of complex preinitials (cf. the
discussion of x mat (566) in section 4.4.3.3 and of the verb to hoe in 4.4.3.4).
Below we discuss more examples.
The Middle Chinese reading of xin < zjenH covet, desire appears at first sight
to indicate an Old Chinese onset including preinitial *s followed by a voiced consonant,
either alveolar or uvular (in the present framework, *s- is invisible before *g when a
nonfront vowel follows). Vietnamese ghen [n A1] jealous, envious (with tone Afor
qshng, as frequently) shows that whatever caused the *s- to voice could not have been
an alveolar. At the same time, the upper-register tone shows the initial was voiceless.
We reconstruct
192 Old Chinese
(802) *s-N-qa[r]-s > zjenH > xin covet, desire; VN ghen [n A1] jeal-
ous, envious
to account for the Vietnamese upper-register tone. Here a stative verb with prefixed
*N is being made transitive by means of *s-. The *s- would be responsible for the
Vietnamese spirantized initial gh- []; perhaps Vietnamese borrowed the word before
the initial was voiced by *N, and *N was left out in the Vietnameseform.
In another instance, we appear to have the same two prefixes in reverse order. The
intransitive verb *t > tong > dng ascend is the root in *s-t > tsong > zng
increase (v.), where *s- is valency-increasing. Theword
is derived from zng by means of the stative/intransitive prefix *N-. It is possible that
the string *s-t- had already metathesized to *ts- when *N was prefixed. In that case, our
reconstruction is not a realistic reconstruction for one single period; it does, however,
include the ingredients needed to derive the Middle Chineseform.
In example (645) above (section 4.4.4.2), we showed that jng < keng a vein of
water and jng < keng loom; regulate; norm have the onset *k.l-. In the same pho-
netic series, the word jng < heang stalk (n.) appears to be etymologically related to
tng < deng stalk, stem. We reconstruct:
The function of the first prefix is unclear; perhaps the stem of a plant was thought of as
a bodypart.
The word ln < lanX lazy shows evidence of a velar preinitial in loans to
Siamese:/graan 2/ and to Proto-Hmongic:*gln B lazy. The latter form also indi-
cates prenasalization. Proto-Mn has *-d- (Jinyng /lye 5/, Fzhu /tia 6/), showing
that the preinitial was loosely attached, and, by the time of Proto-Mn, entirely voiced.
We may reconstruct *N-k.r-, evolving to Proto-Mn *g.d-, Normans*-d-:
(805) *[N-k.]ran > lanX > ln lazy, pMn *-d-; pHmong *glnB
Both pHM *kj:m Ahold in the mouth (Wng and Mo 1995)and the Kra-Dai
language Monn (Ling Mn 1980:100) /gam1/ hold in mouth indicate a prenasal-
ized voiceless stop in the onset of hn < hom hold in the mouth. At the same time,
Proto-Mn has *-g- (Jinyng /a 9/):evidently voicing of the initial and softening are
the work of different preinitials. We reconstruct:
The word tshu > c coarse, thick (as hair), would normally be reconstructed
as *s.ra based on the comparison to the related word *sra > srjo > sh wide apart
(see example (607) for parallel examples). However, Proto-Hmongic *ntsha Acoarse,
which shows the aspirated affricate already in place, also indicates a nasal preinitial:on
semantic grounds, presumably *N- or *N-. We reconstruct
Old Chinese onsets 193
The root is *qi[n] with a meaning like cycle, revolution:in xn < zwin, we need
an *N in the preinitial to account for the voicing, and an *s- to account for the sibilant
z-; thus we reconstruct *s-N-qi[n].
This concludes our review of Old Chinese onsets and their developments. In the next
chapter, we turn our attention to Old Chinese rhymes.
{5}
This chapter presents the rhymes of Old Chinese according to our reconstruction and
describes their development to Middle Chinese and modern dialects. We reconstruct
rhymes primarily on the basis of rhymes in Old Chinese poetry, distinctions in Middle
Chinese, and xishng evidence (especially from recently excavated documents). In
principle, we should also systematically include correspondences with Mn and other
modern dialects, and with early loans to Vietic, Hmong-Mien, and Kra-Dai, as we did
in reconstructing syllable onsets in c hapter4. We do use some evidence from these
sources:they support the reconstruction of the coda *-r, for example (see section 5.5.1).
But so far, modern dialects and early loanwords have told us relatively little about
rhymes that we did not already know from other evidence. In part this is because research
on the relevant dialects and languages is still at a rather preliminary stage:achieving
greater clarity and precision about the history of their rhymes is a high priority for future
research (see the discussion in section6.4).
Maximally, Old Chinese rhymes have four components:a main vowel, a coda, a post-
coda *-, and a postcoda *-s. The only obligatory component is thevowel:
Open syllables (those with zero coda *-) still end in vocalic segments in Middle
Chinese and modern dialects, although a final glide sometimes develops, e.g.,
*C > C(w)oj, *Cu > Caw, *Ce>Cej.
The codas *- and *-k remain in Middle Chinese, except that *-i and *-ik usu-
ally change to *-in and *-it, respectively, and before *-s we have *-ik-s > *-it-s
> *-ij-s>-ijH.
The coda *-j remains in most environments, but in Middle Chinese and most modern
dialects, it is lost after *a; e.g., *Caj becomes MCCa.
The coda *-r usually merged with *-n, but in some dialects it became *-j instead.
(Although there is some transcriptional evidence that our *-r was really [r]
phoneticallysee section 5.5.1.3reconstructing this coda as *-l would also
be plausible.)
Final *-n and *-t remain (except that *-t is lost before *-s:*-t-s > *-js>-jH).
The coda *-w normally remains in Middle Chinese; *-wk becomesMC-k.
The labial codas *-m and *-p usually remain in Middle Chinese, except for some
dissimilations, e.g., *prm > pjuwng > fng wind (n.), *C.[](r)m
> hjuwng > xing bear (n.). Final *-p-s changes very early to *-t-s and thus
develops to *-js>-jH.
Not all combinations of vowels and codas are reconstructed; the permissible combina-
tions are listed in Table5.3. There is a slight asymmetry in that we reconstruct *-u as a rhyme
but not *-i, and *-ij but not *-uw.2 The codas *-w and *-wk appear occur only after
unrounded vowels; also, we have found no reason to reconstruct *-w or*-wk.
Following the general scheme developed in Haudricourt (1954a, 1954b), we
reconstruct the postcoda *- as the origin of the Middle Chinese shngshng tonal
category (marked in our Middle Chinese notation with a final -X); it can occur after
all codas (including zero *-) except voiceless stops.3 The hypothesis is that while
196 Old Chinese
the glottal stop was still present, there would have been a tendency for the glottis to
become tenser in anticipation of the closure of the glottis, resulting in a rise in pitch;
subsequently, the glottal stop disappeared (in most dialects, at least), and the rise in
pitch became phonologically distinctive, creating shngshng, the rising or up tone
of Middle Chinese.
Also following Haudricourt, we reconstruct the postcoda *-s as the origin of the Middle
Chinese qshng tonal category (marked in our Middle Chinese notation with a final -H);
in Old Chinese (OC) it could occur after all codas, and after *-. Fairly early in the Old
Chinese period, final *-p-s changed to *-t-s, a process that affects at least some Shjng
rhymes. Eventually, all voiceless stops before *-s were lost, and final *-s was weakened to
[h], eventually leading to a different pitch contour that became phonologically distinctive,
the qshng departing tone of MC (see Table5.4).4
The final sibilants in *-js may have been the last to disappear, because characters
for qshng words of this type are frequently used to transcribe Indic syllables end-
ing in -s, -, or -, as in these examples (from Bailey 1946; Pulleyblank 19621963;
Pulleyblank1973):
Provisionally, we treat all cases of final *-s as morphological suffixes. The sequence
*--s has the same reflexes as *-s; it is assumed on the basis of internal reconstruction to
account for cases where qshng words appear to come from a root with final*-:
(816) Karlgrens *-g corresponds not only to our *-k-s, but also to our*-.
Karlgrens *-og corresponds not only to our *-awk-s, but also to our *-aw.
Karlgrens *-g corresponds not only to our *-uk-s, but also to our *-u.
The details can be found in sections 5.4 through 5.7 below on individual rhyme types,
where there are comparative tables showing how our reconstructions correspond to sev-
eral other systems.
The basic requirements for an adequate reconstruction of the Old Chinese vowel sys-
tem are that it should account for (1)the distinctions observed in Middle Chinese and
in modern dialects, (2) the rhyming distinctions found in the Shjng and other early
poetry, and (3)the way phonetic elements were used in the pre-Qn writing system. In
this section we will sketch the reasoning that led to the reconstruction of a six-vowel
system and show that this reconstruction makes correct predictions about otherwise
unexplained facts. We will begin by seeing what is required to account for the distinc-
tions in Middle Chinese.
It will be convenient to begin with syllables in the Qiyn rhymes traditionally clas-
sified as division I and division IV. Although the terms division I and division
IV are derived from the Sng-dynasty rhyme-table tradition, they correspond to a set
of rhymes that are easily identified from internal evidence in the Qiyn itself:those
in which only the nineteen initial consonants listed in Table 5.6 occur.5 We will use
the capital letters in the leftmost column as cover symbols for the consonants listed to
theirright.
Let us consider now the division-I and division-IV finals ending in MC -ng and
the initial consonant types that occur with them; these are displayed in Table5.7. The
two leftmost columns list the finals in our Middle Chinese notation and in terms of the
traditional terminology. Each entry in the main body of the table stands for a type of
Middle Chinese syllable, regardless of tone:thus Pang represents syllables such as
png < bang side, mng < mang awn, beard of grain,etc.
Notice that the finals -wang, -weng, and -wong occur only with K-type initials. This
distribution suggests that at an earlier period, the syllable structure (ignoring tone cat-
egory) was not as in (817), but as in(818).
(817) *P
*T
*Ts (w) vowel coda
*K
(818) *P
*T
vowel coda
*Ts
*K
*K
The limitation of the MC finals -wang, -weng, and -wong to K-type initials suggests that
the -w- was originally a feature of the initial, and that there is no need to assume that
there was a separate position for *-w- between the initial consonant and the vowel in Old
Chinese. Instead of the skewed distribution of initials and finals that we see in Table5.7,
then, we assume that in Old Chinese, the pattern was as in Table5.8. Old Chinese recon-
structions are in the top row and leftmost column; their Middle Chinese reflexes are in
the body of the table. Each row corresponds to one of the traditional rhyme groups,
named in the second column from theleft.
The only gaps in the table now are those for labialized onsets before rounded vowels,
syllables like *Ko or *Ku. It is possible that such syllable types did exist, but
so far we have found no evidence for reconstructingthem.
The argument so far suggests that in order to account for the syllable types in final
*-, we must reconstruct at least the five vowels *o, *u, *a, *e, and *.6 Would this also
be sufficient to account for syllables ending in MC -n? The Middle Chinese division-I
and division-IV syllable types in final -n are displayed in Table 5.9, comparable to
Table5.7above.
Here the pattern of distribution is more subtle. MC -wen is like the finals -wang,
-weng, and -wong of Table5.7 in that it only occurs with K-type initials. But although
there are no Middle Chinese syllables like Twang or Tswang, we do have syllables
like Twan and Tswan. One way to account for this would be to revert to allowing *w
to occur freely before the main vowel, as in (817) above, as Karlgren had originally
done. But Jaxontov (1960b) suggested an alternative hypothesis:that the -w- in syl-
lables like Twan, Tswan, Twon, and Tswon is a secondary development from an origi-
nal rounded vowel. According to his hypothesis, syllables like MC Twan and Tswan
came from OC *Ton and *Tson, respectively; syllables like Twon and Tswon came
from OC *Tun and *Tsun. Between Old Chinese and Middle Chinese, there was a
diphthongization of the rounded vowels *o and *u before acute codas (in our system,
*-j, *-n, *-t, and *-r, column 2 in Table5.2):e.g., *Ton > Twan, *Tun > Twon. We call
this the rounded-vowel hypothesis, and it is a crucial element of the six-vowel system.
It accounts for the n-final syllable types as shown in Table5.10 (to be revised below),
which should be compared with Table5.8.
Several points should be made about Table5.10. We assume that there were origi-
nally contrasts between *Pan and *Pon, *Kan and *Kon, *Pn and *Pun, and
*Kn and *Kun, but that these pairs merged in Middle Chinese as the result of two
sound changes: (1) rounded vowels diphthongized before acute codas, and (2) *w
became nondistinctive after labial initials.That is, *Kon diphthongized to *Kwan,
merging phonologically with original *Kan (> MC Kwan); *Kun diphthongized to
*Kwn, merging phonologically with original *Kn (> MC Kwon). Then, at some
point, *w became nondistinctive after labial initials:*Pon diphthongized to *Pwan,
but this merged with original *Pan (> MC Pan); *Pun diphthongized to *Pwn,
merging with original *Pn (> MC Pwon). (As with *Ko and *Ku, it is pos-
sible that there were also syllables like *Kon or *Kun, but we have found no reason
to reconstructthem.)
So far, our analysis has been based entirely on the syllable types of Middle
Chinese, and has not taken Old Chinese rhyming into account. Our reconstruction of
the rhymes in final *- is unproblematic from the point of view of Old Chinese rhym-
ing:each reconstructed rhyme corresponds exactly to a traditional rhyme group, as
in Table5.11.
At first glance, Table5.10 would appear to indicate that the same five vowels that
were necessary to account for the syllables in MC -ng will also account for the syl-
lables in MC -n. But the relationship between the traditional rhyme groups and our
reconstructions is more complex with *-n than with *-. This is because, according
to the traditional rhyme analysis, MC -en can come from three different Old Chinese
Old Chineserhymes 201
rhyme groups: Zhn, Yun, and Wn. (Karlgren accounted for this by recon-
structing *-ien, *-ian, and *-in, respectively, as the sources of MC -en in these
cases, but the front-vowel hypothesis proposes a different solution; see the discussion
below.) Examples are given in Table5.12, with references to relevant Shjng rhyme
sequences.
While MC -en corresponds to three different traditional rhyme groups, the syl-
lable types we reconstruct with OC *-an, *-on, and *-en are all assigned in the tra-
ditional analysis to a single rhyme group, Yun; likewise, the syllable types we
reconstruct with *-n and *-un are all assigned to the single group Wn. If we
incorporate the distinctions implied by Old Chinese rhyming into Table5.10, we get
Table5.13 (the rows are numbered this time for convenience; this table, too, will be
revised below).
Now, instead of one row with MC -(w)en, as in Table5.10, we have three (rows 3,
4, and 5), one for each of the three traditional rhyme groups that MC -(w)en can come
from: Yun (corresponding to our *-en), Zhn (for which we reconstruct *-in), and
Wn (whose reconstruction we will discuss shortly). Now instead of five rows, as in
Tables5.8 and 5.10, we have seven rows. Does this mean that we need to reconstruct
seven different vowels? No, because it turns out that rows 5 and 6 are in complementary
TABLE5.13 Origins of MC syllables in final -n, with their traditional OC rhyme groups
(incomplete)
traditional
rhyme *P- *T- *Ts- *K- *K-
rhyme group
1 -(w)an < *-an Pan Tan Tsan Kan Kwan
2 -(w)an < *-on Yun Pan Twan Tswan Kwan
3 -(w)en < *-en Pen Ten Tsen Ken Kwen
4 -(w)en < *-in Zhn Pen Ten Tsen Ken Kwen
5 -(w)en < ? Ten Tsen
6 -(w)on < *-n Wn Pwon Kon Kwon
7 -won < *-un Pwon Twon Tswon Kwon
We account for this situation by reconstructing both rows 5 and 6 of Table5.13 with
*-n, and assuming that * normally became fronted when it was between an acute onset
and an acute coda. Thus *-n becomes -en in the second reading *l n > then, repre-
senting the regular development; and *l n > thon swallow, where *-n becomes
MC -on after an acute initial, is exceptional, possibly due to onomatopoeia. Notice that
for MC thon, Mandarin has tn instead of the tn (a nonexistent syllable) that might
be expected. Although MC -on itself is rare after acute initials, there are a number of
alternations between MC -en and -won after acute initials that suggest a further develop-
ment *-n > -on > -won in some dialects.
Consider,for example,
Ordinarily, we would assign a word with MC -won to row 7 of Table5.13 and recon-
struct it with *-un, but we have good evidence that the correct reconstruction is not
*dzun but *[dz][n], and that MC -won here is an irregular development from *-n.
According to the Shuwn, the phonetic in cn is ci, which is also reconstructed
with*:
The Shuwn says that ci is phonetic in zi also, and uses cn < *[dz][n]as
a sound gloss for zi < *[dz]:
(824) : 7
zi [*[dz]] means [*[dz][n]] exist; it is from t earth
[i.e., it has t as a graphic component] and has ci [*[dz]] as
phonetic. (SWGL6120b)
The reason for the alternation between *- and *-[n] in words that may be etymologi-
cally related is not clear, but in any case, it is likely that analogy to zi < *[dz] played
a role in preventing the regular development *-n > -en in this case; instead we have
perhaps *[dz][n]> dzon > dzwon. Words with cn as phonetic usually have MC
-en as expected, following the normal development of *-n to MC -en after acute initials:
(825) *N-ts[n]-s > dzenH > jin grass, herb (also read dzwonH)
*[dz][n]-s > dzenH > jin to fencein
*[N-ts][n]-s > dzenH > jin twice; a second time, probably related to
*[ts]()-s > tsojH > zi twice; a secondtime
The six-vowel reconstruction incorporates two main hypotheses about Old Chinese vow-
els:the rounded-vowel hypothesis, discussed above, and the front-vowel hypothesis, to
be discussed below. The rounded-vowel hypothesis, due to Jaxontov (1960b), claims that
where Middle Chinese has a contrast between syllables with and without -w- before the
vowel (in traditional terms, a contrast between hku closed-mouth finals and kiku
open-mouthed finals), the -w- has only two possible origins:either the Old Chinese
onset was labialized (e.g., *k-), or the -w- results from the diphthongization of one of the
rounded vowels before acute codas:for example, *-on > *-wan or *-un > *-wn. The rea-
son that Middle Chinese has syllables like Kweng and Kwen, but no syllables like Tweng
or Twen, is that Old Chinese had labiovelar and labiouvular onsets like *k()- and *q()-
but no labialized dental onsets like *t()-; nor was there any rounded vowel that diph-
thongized to produce MC-we-.
The other main hypothesis involved in the six-vowel reconstruction is the front-vowel
hypothesis. The front-vowel hypothesis essentially claims that Old Chinese had no
vowel combinations such as the *ia, *ie, or *i that Karlgren and Li reconstructed in
division-IV rhymes:division-IV finals just come from the front vowels *e and *i (as in
rows 3 and 4 of Table5.14 above), or from the fronting of * between acute onsets and
acute codas (as in row 5 of Table5.14).Table5.15 below shows how the three recon-
structions compare on thispoint.
If combinations like *ia, *ie, or *i are not available in the Old Chinese system, then
it follows that the Middle Chinese division-IV syllables (those whose vowel is written
as -e- in our notation, with no -a-, and no -j- or -y- to the left) must be reconstructed with
one of the front vowels *e or *i (as in rows 1 and 2 of Table5.15, respectively), or else
with a * that becomes fronted between an acute onset and an acute coda (as in row3).8
The other main point of the front-vowel hypothesis involves type-B (nonpharyngeal-
ized) syllables, and has to do with the so-called chngni repeated-initial finals
of Middle Chinese (for discussion, see section 2.1.2.3). Briefly, in our reconstruction
(as in other versions of the six-vowel reconstruction), the syllables traditionally called
division-IV chngni (in our Middle Chinese notation, those that are written with both
prevocalic -j- and -i-) must all be reconstructed with front vowels, OC *i or *e. We will
discuss the chngni problem in more detail in section 5.3.2; for now, we only illustrate
the situation with a couple of examples. Table5.16 shows a three-way minimal contrast
in Middle Chinese, among the syllables we transcribe as bjonH, bjenH, and bjienH, with
our reconstructions and those of Karlgren and Li. In the traditional analysis, all three
words are assigned to the Yun rhymegroup.
Both bjenH cap (row 2)and bjienH comfortable (row 3)are in the Qiyns
Xin rhyme, but they are in different homophone groups, with different fnqi spell-
ings. It is pairs like this that are called chngni repeated initials, because the
Xin rhyme includes two homophone groups with the same initial b-. Since the two
syllables have the same reflexes in virtually all dialects, it is not readily apparent how
they were distinguished phonetically, but in the Sng-dynasty rhyme tables, bjenH
(row 2 in Table5.16) is put in division III, while bjienH (row 3 in Table5.16) is put in
division IV; so bjenH and bjienH are called division-III and division-IV chngni
syllables, respectively.9 In our Middle Chinese notation, we use an arbitrary convention
to distinguish the two types:division-III chngni syllables are written with prevocalic
-j-, or with the vowel -i-, but not both; division-IV chngni syllables are written with
both -j- and-i-.
As for their Old Chinese reconstructions, Karlgren thought that the chngni distinc-
tions were artificial, and that bjenH and bjienH were simply homonyms in both
Middle Chinese and Old Chinese. Li recognized the distinction in Middle Chinese, but
was unable to account for it in his Old Chinese reconstruction system. He used the vowel
combination *ia to distinguish bjenH and bjienH (his *bjianh) from bjonH
(his *bjanh), but had no way to distinguish bjenH and bjienH from each other
(1971:41). In the six-vowel system, Old Chinese prevocalic *-r- plays an important role
in accounting for the necessary distinctions, as will be shown below. But briefly, in the
six-vowel system, syllables like bjonH (row 1) must be reconstructed with a nonfront
vowel; division-III chngni syllables like bjenH (row 2)can be reconstructed with either
front or nonfront vowels; and division-IV chngni syllables like bjienH (row 3)can be
reconstructed only with a frontvowel.
Table5.17 illustrates a parallel contrast in high-vowel syllables, in this case with the
three-way contrast among MC mjut, mit, andmjit. In this case too, Karlgren regarded
mit and mjit as homonyms, even though both the rhyme books and the rhyme
tables clearly distinguish them. Li recognized the distinction, but as with the previous
example, was unable to account for it in his reconstruction; he gave them the same Old
Chinese reconstruction, but with a question mark after mjit (1971:47). In our recon-
struction, they are distinguished by the fact that mit < *mri[t] has a prevocalic *-r-,
while mjit < *mit doesnot.
The main point here is that according to the front-vowel hypothesis, on the basis of
their Middle Chinese readings alone, division-IV chngni syllables like bjienH and
mjit must be reconstructed with front vowels (*e and *i, respectively); the details of
how our system accounts for the chngni distinctions are given later in this chapter,
when individual rhymes are discussed.
We began our discussion by focusing on division-I and division-IV syllables ending
in MC -ng and -n, but it turns out that the six vowels we have reconstructed in order to
account for these syllables are also sufficient to account for division-II and division-III
syllables, and for syllables with other codas, as the later sections of this chapter will
show. It is this line of reasoning that suggests that Old Chinese should be reconstructed
with a system of six vowels.
The arguments for a six-vowel system sketched above were based largely on the
distribution of rhymes and onsets in Middle Chinese. We did rely on Old Chinese rhyme
evidence to distinguish three sources of MC -en (OC *-en, *-in, and in certain environ-
ments, *-n), but the decision to reconstruct *-an, *-on, and *-en with different vowels
was based mainly on the pattern of distribution of phonological elements in Middle
Chinese, not inferred inductively from the corpus of pre-Qn verse. In fact, as we can
see from Table 5.14, there is a mismatch between the phonological rhymes we have
reconstructed with the coda *-n and the traditional rhyme categories:in the traditional
analysis, our *-an, *-on, and *-en are all assigned to the Yun group, and both *-n
and *-un are assigned to the Wn group. Karlgren and Li reconstructed these rhymes
in such a way as to agree with the traditional analysis:see Table5.18, which compares
our reconstructions with theirs.10
Is this disagreement with the traditional analysis an argument against the six-vowel
system? There is no guarantee that rhyming practices in a literary tradition will nec-
essarily correspond in a simple way to the phonological rhymes of any particular
dialect; for example, for some 800 years, Chinese poetsregardless of what kind
of Chinese they actually spokehave been writing the genre called regulated verse
(lsh ) according to a rhyme standard called the Pngshu yn , which
defines 106 rhyme categories (a simplification of the categories of the Qiyn sys-
tem), set out in various reference books. The same kind of verse is even composed by
Japanese poets, so as to rhyme according to this Chinese standard.
However, we have no evidence of prescriptive rhyme books or similar literature
from the pre-Qn period, and our default expectation would be that in the absence of
other constraints, poets would have found it most natural to rhyme according to the
As much work in Chinese philology still relies on the traditional rhyme categories, it
is worthwhile to point out that the six-vowel reconstruction is not just of interest to
historical linguists, but also has consequences for philological work on early Chinese
texts. The Qng scholars who developed the traditional analysis of Old Chinese rhym-
ing were not primarily motivated by an abstract interest in how the pre-Qn language
sounded; rather, they were interested in using their analysis as a philological tool to
understand early texts better. In this section we give some examples to show that our
208 Old Chinese
reconstruction is a better philological tool than the traditional analysis and the recon-
structions basedonit.
Ode106.3
As pointed out in Baxter (1992:387388), there is a textual problem in stanza 3 of Ode
106 of the Shjng (Q fng:Yji ). The Mo version reads as in
(826), with our Middle Chinese notation added for the rhyme words.13
The textual problem occurs in line 5.The Mo commentary has no gloss here, but
the Eastern Hn commentator Zhng Xun (127200) glosses fn as to revert
( f y). The Jngdin shwn has this comment:
So there are two versions of the text here:where the Mo version has fn to revert,
the Hn version (of which only fragments remain such as in quotations like this) had
bin to change. In his Glosses on the Book of odes (19421946, 14:195), Karlgren
gives two translations of the line, one for each possibility:
In the three earlier reconstructions, there is nothing to indicate that the rhyme in line 5
is in any way unusual or suspect.15 However, in our reconstruction, fn must be recon-
structed with *-an, while all the other words must be reconstructed with *-o[n].16 The Hn
Sh reading bin < pjenH change (v.) for the rhyme word in line 5 must be recon-
structed as *pro[n]-s, which fits the other rhyme words perfectly.17 The Hn Sh reading is
therefore preferable on phonological grounds, as pointed out in Baxter (1992:364366).
Further confirmation for the reading {} *pro[n]-s change (v.) in line 5 has recently
come from the bamboo-strip document called Kngz Sh ln (Kngzs
discussion of the Sh), in the Shnghi Museum texts (SB 1.131.41, 1.1211.168).
Line 5 of Ode 106.3 is quoted on strip 22, where the word corresponding to {}
*pro[n]-s is writtenas
(829)
and is often used as a loan character for {} *pro[n]-s change (v.)but not, as far
as we know, for fn. The word *C.[b]ro[n]-s cap itself rhymes as *-o[n]-s in
Ode102.3.
In effect, the six-vowel system predicted that s sh bin x should be the
earlier version of line 5; we would expect that fn < *pan would have been substi-
tuted for {} *pro[n]-s only after the diphthongization *-on > *-wan had occurred (see
section 5.2.1 above), after which it would have seemed to be a satisfactory rhyme with
the other words of the sequence. Baxter (1992) predicted that if an earlier version of line
5 were found, it would be consistent with the reading {} *pro[n]-s rather than with
{} fn < *pan reverse; and the Shnghi strips show that this prediction was correct.
A very similar case is the rhyme sequence from Loz 39, given in Table5.20 (there
is also an internal rhyme sequence in *-e that is not relevant here). In terms of the
traditional categories and reconstructions based on them, there is nothing unusual about
the rhymes; they all belong to the traditional Yu group. But in terms of our recon-
struction, all the rhyme words in Table5.20 are to be reconstructed with *-at except that
210 Old Chinese
system are subject to falsification every time a new early document is discovered; so
far, they have stood the test oftime.
Even though we argue that the traditional Old Chinese rhyme categories are insufficiently
fine-grained, we make frequent reference to them in the rest of this chapter, because they will
be meaningful to some readers, and other readers will need to become familiar with them to
read the literature on the phonology and philology of the pre-Qn period. Some of our recon-
structed rhymes correspond exactly to traditional groups:for example, the set of words we
reconstruct with the rhyme *-e is virtually identical to the set of words traditionally assigned
to the Zh rhyme group; in such cases we will sometimes write *-e= Zh. But in
many cases our reconstructed rhymes correspond to only part of a traditional group:for
example, the rhyme *-en is only part of the traditional Yun rhyme group. In such cases,
we sometimes write *-en Yun, using the mathematical symbol for a proper sub-
set, to remind the reader that the correspondence to the traditional rhyme groups is inexact.
Details will be given in the discussion of individual rhymes in sections 5.4 through5.7.
The main factors affecting the development of rhymes after the Old Chinese period
were (1)whether or not the onset was pharyngealized, and (2)whether or not there was
an *-r- before the vowel. There were also a number of assimilatory and dissimilatory
processes involving onsets, vowels and codas:we will discuss the main trends in this
section, but the details will be found in sections 5.4 through5.7.
But in the Qiyn, shn < *ni[] and rn < *ni[], with originally nonpharyn-
gealized onsets, are in the Zhn rhyme (MC -in), while yun < *[]i[], with
an originally pharyngealized onset, is in the Xin rhyme (MC -en). On this point the
Qiyn agrees with the typical rhyming practice of the sixth century CE and later:
yun < *[]i[] no longer rhymes with shn < *ni[] and rn < *ni[]; instead,
it rhymes with words like qin < dzen < *dzen, which originally had the nonhigh
vowel *e. The general pattern is as in Table5.22.
In general, type-A and type-B syllables that had the same rhyme in Old Chinese are
placed in different rhymes in the Qiyn. The lowering of vowels after pharyngealized
onsets seems to be the best explanation for this. The precise processes probably varied
from dialect to dialect, but the system of the Qiyn indicates the following effects,
which are generally supported by sixth-century rhyming practice. (We omit the possible
MC -w- resulting from labialized initials.)
The effects of pharyngealization differed slightly from rhyme to rhyme, and were
complicated by other factors.For example, the low vowel *a tended to stay low
before velar codas, so Cjang () < *Ca generally continued to rhyme with Cang
() < *Ca; the two rhymes are designated as tngyng in the Gungyn,
and were combined in the Pngshu yn. For details, see the discussion of individual
rhymesbelow.
Prevocalic *-r- is reconstructed to account for three kinds of Middle Chinese sylla-
bles:(1)those that have retroflex sibilant initials (tsr-, tsrh-, dzr-, sr-, zr-) or retroflex
stop initials (tr-, trh-, dr-, nr-) in Middle Chinese (see section 4.1.3); (2)division-II
syllables; and (3) certain division-III syllables (including many of the division-III
chngni syllables). Reconstructing *-r- in division II was proposed (in slightly dif-
ferent form) by Jaxontov (1960a, 1963); reconstructing *-r- in division III was pro-
posed by Pulleyblank (19621963). Li (1971) reconstructed *-r- in cases (1)and (2),
but not (3); the six-vowel reconstructions (including Starostin 1989, Baxter 1992,
Zhngzhng 2003, and the reconstruction presented here) reconstruct *-r- in all three
cases.20 Reconstructing *-r- in cases (2)and (3)involves assuming that prevocalic
*-r- sometimes influenced the quality of the following vowel. Although *-r- was
preserved after coronal initials as a feature of retroflexion, after *K- and *P-type ini-
tials it disappeared; some of the vocalic distinctions introduced by *-r- then became
phonologically distinctive. We will discuss pharyngealized and nonpharyngealized
syllables separately.
*Kren does not become MC Kwean as might be expected, but rather MC Kwaen,
as shown by rhymes, xishng evidence, and word-family connections (see Baxter
1992:382383);for example:21
Similarly, in gu < kuwX dog, Hmong-Mien forms indicate an *-l- before the vowel,
which probably represents an *-r- in the Chinese source, even though it is not apparent
from the Middle Chinese reading kuwX; our reconstruction is *C.kro.24
(847) *C.kro > kuwX > gu dog; cf. pHM *qluwX, Proto-Mienic
*klo B (L-Thongkum 1993:188)
But in general, we cannot distinguish syllables like *Cro from *Co when the onset
is grave.25 However, *Cro and *Crok do have reflexes distinct from *Co and *Cok,
probably because before *- and *-k, *o diphthongized to something like [aw]a
change that did not happen when the coda waszero:
(848) *kro > *kraw > kaewng > jing (Yngz) river; cf. Proto-Monic
*kro large river (Diffloth 1984:132)
(849) *pro > *praw > paewng > bng country
(850) *C.[k]rok > *krawk > kaewk > jio horn, corner
(851) *[p]rok > *prawk > paewk > b cut, flay,peel
However, *-r- also had an effect in syllables that already had front vowels, because there
are also contrasting reflexes in front-vowel syllables with and without a preceding *-r-,
as in examples (864) through(869).
In each pair of examples cited in (864) through (868), we can tell from rhymes or xishng
evidence or both that the two forms had the same main vowel in Old Chinese; yet the
Middle Chinese sources clearly indicate that they had different finals. In examples (864)
and (865), the two forms are in different Qiyn rhymes. In examples (866) through
(868), the forms are in the same rhymes, but are given different fnqi spellings:these
are chngni or repeated-initial pairs, discussed above.26 Furthermore, such pairs
are systematically distinguished in the Sng rhyme tables:in examples (866) through
(868), the first member of each pair is put in division IV while the second is put in divi-
sion III. In example (869), the phonetic elements indicate that both forms originally
had velar initials, but in *kij > tsyij, the *k- is palatalized to MC tsy- before the front
vowel *i; in *[g]rij > gij, palatalization has been blocked by *-r-. (In this example
both words are in division III of the rhyme tables, because tsy- and other palatal initials
are always put in divisionIII.)
The phonetic interpretation of these distinctions is difficult, because there are hardly
any traces of the distinction in modern dialects, and the treatment of the relevant words
in the written sources very likely represents a compromise among two or more dialects.
But there is little doubt that the distinctions were real in some varieties of Chinese,
and there are good arguments that whatever their nature, they can be attributed to OC
Old Chineserhymes 217
prevocalic *-r-.For example, the syllables we reconstruct with prevocalic *-r- on this
basis often have xishng contacts with MC l- < *r()- or other evidence of *-r-, as in
examples (870) through(874).
The hypothesis that such contrasts should be accounted for by reconstructing pre-
vocalic *-r- is sometimes summarized by stating it as a rule that *-r- should be recon-
structed in all division-III chngni syllables; but this is an overgeneralization. The
relationship between OC prevocalic *-r- and division-III chngni syllables is more
indirect:*-r- must be reconstructed to account for some of these syllables, but not all. In
cases like examples (875) and (876), there is a contrast in Middle Chinese between ear-
lier syllables of the form *CaC and *CraC or between *CC and *CrC; in such cases,
*-r- is required to account for the contrast.29
OC MC examples
(875) *Pa Pjang () *pa > pjangX > fng just then, at that
time
*Pra Pjaeng ( ) bng < pjaengX < *pra third heavenly
stem
(876) *P[n] Pjun () *p[n] > pjun > fn divide
*Pr[n] Pin ( ) *[b]r[n] > bin > pn poor
But there are also situations where *Cra(C) has evidently merged with *Ca(C) or
*Cr(C) with *C(C); in such cases there is no contrast in Middle Chinese between the
form with *-r- and the form without, so unless we have other evidence, we do not know
whether there was an *-r- there or not. In such cases, we write *-(r)- to indicate that for
all we know, there could have been an *-r- before the vowel. The notation *-(r)- does
218 Old Chinese
not mean that there is any positive reason to believe that an *-r- was present, only that
we do not know. Here are some examples:
OC MC examples
(877) *Paj
*Praj } Pje ( ) *m(r)aj > mjeX > m fall over
*Praw }
(878) *Paw
Pjew ( ) *p(r)aw > pjewX > bio exterior
*Kr }
(879) *K
Ki () *k(r) > ki > j base (n.)
*Krk }
(880) *Kk
Kik () *k(r)k > kik > j urgently
We can tell from both rhymes and xishng connections that the main vowel in (881)
was *u, but the syllable develops to Middle Chinese as if it had*.30
Final labials *-m and *-p frequently dissimilated to [] and [k], respectively, after
labial and labialized onsets:
(882) *C.[](r)m > *C. > hjuwng > xing bear (n.); cf. Ximn
/him2/
Conversely, when no *-r- intervened, the vowel * was rounded to *u (> MC -juw) after
(nonpharyngealized) *P- and *K-, an assimilation of rounding:
Finally, * and *a were both regularly fronted between an acute onset and an acute coda;
with *, this happened even when the onset was pharyngealized:
(884) *sr > *sn > *sin > sen > xin first31
(885) *[d]an > *den > dzyenX > shn leveledarea
Such dissimilations are discussed in more detail in the sections below on individual
rhymes.
The remainder of this chapter gives the details of our reconstructed rhymes and their
Middle Chinese reflexes under different conditions. Reflexes in Mn and other dialects
Old Chineserhymes 219
are mentioned occasionally, but not systematically; more research needs to be done
before these developments are entirelyclear.
The rhymes are grouped according to the types of codas, as in Table5.24. Within
each section, rhymes with the same main vowel are discussed together:for exam-
ple, in section 5.4, the first subsection 5.4.1 discusses rhymes with *a before the
back codas *-, *-k, and *-; subsection 5.4.2 discusses rhymes with * before the
back codas; and so forth. For efficiency of presentation, most of the information is
presented in tables; general comments and specific problems are discussed in the
accompanyingtext.
As noted above, there is a tendency for vowels to be lowered or to stay low before the
velar codas *-k and *-. Open syllables (those with the zero coda) do not show this
tendency, however. Vowels generally develop as expected, except that original *-ik and
*-i show several different developments, presumably due to dialect differences (see
section 5.4.4 below).
Our reconstruction distinguishes among rhymes of the form *-V-s, *-V-s, and *-Vk-s.
Some other reconstructions do not make this distinction:for example, our *--s, *--s, and
*-k-s are all *-g in Karlgrens reconstruction and *-gh in Lis reconstruction.
For each combination of main vowel and coda type, we include a table like Table5.25
that includes traditional Old Chinese rhyme groups, Middle Chinese readings, and
reconstructions in several systems.32 Only reflexes in type-A (pharyngealized-onset)
syllables with no prevocalic *-r- are included in tables of this type, since these syllables
have a rather simple development that makes them convenient for comparing different
reconstructions. In traditional terms, this means that the syllable types represented in
tables like this are either division I(as here) or divisionIV.33
In tables like Table5.25, the column numbered 1 includes the rhyme we reconstruct
with zero or vocalic codas (here *-a); column 2 has the corresponding rshng rhyme,
that is, the rhyme with a voiceless stop coda (here *-ak). Column 3 (here *-ak-s) gives
the result of adding an *-s suffix to the rshng rhyme in column 2; and column 4 gives
the corresponding rhyme with a nasal coda (here*-a).
220 Old Chinese
The rhymes *-a, *-a-s, and *-a-s all develop as in column 1, except for tone
category. We list *-ak and *-ak-s separately in these tables because the treatment
of *-ak-s differs in different reconstructions; see examples (886) through (888)
below, where our reconstructions are compared with those of Karlgren, Wng L,
and Fang-kueiLi.
Note that the finals in (886b), (887b), and (888b) are all the same in Middle
Chinese (-uH). In Lis reconstruction, they are the same in Old Chinese as well,
namely *-agh. But in our reconstruction, as in Karlgrens and Wng Ls, *-ak-s (in
888b) is distinguished from *-a-s (in 886b) and *-a-s (in 887b). Neither Karlgren,
Wng L, nor Fang-kuei Li distinguished *-a-s (in 886b) from *-a-s (in 887b), how-
ever. The three different Old Chinese finals in (886b), (887b), and (888b) are not
distinguishable from Middle Chinese alone; we reconstruct them on the basis of
xishng and word-family relationships such as those between the (a)and (b)forms
in each e xampleabove.
In assigning words to traditional rhyme categories, there is a similar difference
of treatment: Wng L assigned the rhymes we reconstruct as *-ak and *-ak-s to
the traditional rshng rhyme group Du (putting column 3 with column 2), but
Old Chineserhymes 221
others would assign our *-ak-s to the traditional ynshng rhyme group Y (put-
ting column 3 with column 1). Our treatment here is like that of WngL.
In row 2 of Table5.25, = Y means that our *-a corresponds exactly to the
traditional rhyme group Y (under Wng Ls interpretation of it; see the previ-
ous paragraph); similarly, our *-ak and *-ak-s correspond to the traditional Du
group (in the version of the traditional analysis used by Wng L); and our *-a
(subsuming *-a, *-a-s, and *-a-s, not listed separately) corresponds to the
traditional Yng group. By contrast, in similar tables below we sometimes write
the mathematical symbol , meaning is a proper subset of, instead of =:for
example, under *-an we write Yun (see Table 5.55 below), because the
traditional Yun group does not exactly correspond to our *-an, but also includes
the rhymes *-en and *-on.34
In row 3 of Table 5.25, we give the Middle Chinese reflexes of the Old Chinese
rhymes in each column, followed by the name of the relevant Gungyn rhyme. Some
readers will be more familiar with the names of the Gungyn rhymes than with any
alphabetic representation of them; other readers are advised to become familiar with the
Gungyn rhymes, so that they can understand literature that uses traditional terminol-
ogy, and relate the terminology used there to our reconstruction. In rows 4 through 11,
we give the corresponding Old Chinese reconstructions from eight previous systems,
for comparison.
Ten Old Chinese onset types are distinguished by these cover symbols:five pharyn-
gealized and five nonpharyngealized. When it is unnecessary to specify the position
of articulation, we also use *C- and *C- as cover symbols for any pharyngealized or
nonpharyngealized initial, respectively. Similarly, for Middle Chinese, we use C- as a
cover symbol for an arbitrary initial consonant.
Table 5.27 gives the Middle Chinese reflex types for the finals *-a and *-ra after
different types of Old Chinese initials.For example, the first row gives Cu as the reflex
of *Ca; this means that *-a (but not *-ra) becomes MC -u after any pharyngealized
initial. The next two rows give MC Kwae as the reflex of *Kra and MC Cae as the
reflex of *Cra; this means that *-ra becomes MC -wae after pharyngealized labialized
initials (represented by *K-), but it becomes MC -ae after all other pharyngealized
initials (represented by *C-). (In other words, syllable types like *Kra that have spe-
cial developments are listed first, and the syllable types below with *C- or *C- repre-
sent syllable types not already included above.) The cover symbols for Middle Chinese
initials are those given in Table2.3. The identity of MC C- can be inferred from the
principles given in c hapter4.
The rightmost column gives examples of the developments in that row. The develop-
ment *Ca > Cu is illustrated by *ka > khuX > k bitter.
Tables like Table 5.27 give the reflexes of Old Chinese rhymes after single-
ton initials. The development after more complex syllable onsets can usually be
inferred by analogy to simpler ones: for example, *m-k- later becomes *g-, so
its development is the same as the development after the simple initials included
under the cover symbol *K-. When a presyllable affects the position of articula-
tion of the main consonant, the developments will not match those in our tables
exactly. An e xampleis
where the main initial *q- would fall under the cover symbol *K-, but because of the
change of *s-q- to *sw-, after a certain point the development follows the pattern of
initials of type*Ts-.
For some of the syllable types envisaged in our reconstruction, there are no attested
examples, or no examples that can be clearly assigned to that type; in that case the
table will will be accompanied by a note to that effect. For some Old Chinese syl-
lable types, more than one possible outcome is given; we assume that these cases are
generally due either to dialect mixture or to phonological conditions that have not yet
been identified. Such cases will be discussed in the comments accompanying each
table.For example, in Table5.27, some syllables of the type *Ta develop into MC
Tsyo, while others develop into MC Tsyae:at present we cannot fully explain these
divergent developments, and as an ad hoc notational device, we write *TA instead of
*Ta when the result is MCTsyae.
Developments that need further discussion are identified by numbers in square
brackets to the left of the relevant examples; the explanatory notes are given after the
table. Here, then, is Table5.27.
Old Chineserhymes 223
Notes on Table5.27:
[1] *Ka and *Ka both become MC Ku, so they cannot be distinguished from
Middle Chinese evidence alone; we reconstruct *Ka on the basis of graphic or
word-family connections. For example, we reconstruct *ka > ku > g orphan
with a labialized initial because it is written with the phonetic *kra > kwae >
gu melon, gourd.
[2] In type-B syllables with grave initials, it is impossible to distinguish between
*Ca and *Cra on the basis of Middle Chinese evidence alone:*Ka and *Kra both
become Kjo, *Pa and *Pra both become Pju, and *Ka and *Kra both become Kju.
It is only when we have graphic or word-family evidence that we can reconstruct *-r-
with confidence in such syllables;for example, j < kjoX < *[k]ra round basket
is written with the phonetic *[r]a > ljoX > l spine; pitch-pipe.
[3] The rhymes *-a and *-ra become MC -jo after *K-, but -ju after *K- or *P-,
merging with *K(r)o and *P(r)o, respectively. Most modern dialects have no distinction
corresponding to MC -jo vs. -ju, but the distinction is preserved in Mn and Southern
W dialects (Lu Chngpi 1931; Dng Tngh 1960:1041; Zhu Zm [1943] 1966,
1966; Mei 2001). For example, in Normans reconstruction, we generally have the
Proto-Mn final *-y from OC *C(r)a, but pMn *-io from OC *K(r)a and *P(r)a, as
shown in Table5.28.
[4] In addition to MC -jo and -ju, there is a third Middle Chinese reflex -jae from OC
*-a that we cannot yet fully explain. Thus there are contrasts like the following:
(890) zh < tsyoX (= tsy- + -joX) boil, cook35
zh < tsyoX (= tsy- + -joX) islet
zh < tsyaeX (= tsy- + -jaeX) red earth; red pigment
zh < tsyaeX (= tsy- + -jaeX) (nominalizing particle)
224 Old Chinese
All four words are written with the same phonetic zh, so they should all have the
same main vowel, and they all have initial tsy- in Middle Chinese, which normally rep-
resents OC *t-; it appears that all four should be reconstructed as *ta. Why then do we
have two Middle Chinese reflexes, tsyoX and tsyaeX? For the present, we have no satis-
factory explanation. The contrast is limited to syllables with Middle Chinese initials of
the types Tsy- and Ts-. Given that syllable onsets in Old Chinese are more complex than
in Middle Chinese, and our reconstruction of them is incomplete, it is likely that con-
trasts in the onset play a role in conditioning the Middle Chinese reflexes;for example,
perhaps MC tsyaeX could reflect a form like *C.ta with some preinitial consonant that
keeps the following vowel from developing to MC -jo as it normally would. But for
the present we simply distinguish them by writing OC *-a as *-A when the Middle
Chinese reflex is -jae. Our *A is not intended as a seventh Old Chinese vowel; it is an
explicitly ad hoc notation that basically means a case of OC *-a which for as yet unex-
plained reasons becomes MC -jae instead of MC -jo.36 There is no evidence that *-A
rhymes any differently from ordinary*-a.37
Other common words with MC -jae < OC *-a (written *-A) include:
Notes on Table5.29:
[1] From syllables like *Krak we sometimes have MC Kweak instead of the expected
Kwaek; probably the rhyme books do not distinguish these syllable types reliably:
(896) *qrak > 'waek > w catch (v.), but from the same root wehave
*m-qrak > hweak > hu catch(v.)
[2] Although *-a and *-ra have the same reflexes after type-B grave initials, *-ak
and *-rak have different reflexes:-jak and -jaek, respectively; for grave initials, *Crak
is thus placed in the same Qiyn rhyme as *Crak (> Caek). However, we have no clear
examples of *Krak, which perhaps does merge with*Kak.
[3] Although prevocalic *-r- fronts a following vowel after grave initials, with acute
initials, the *-r- makes the initial retroflex, but the final is -jak, not-jaek.
[4] Syllables of the form Tsraewk from the *-ak rhyme (traditional Du) have
generally been regarded as irregular, but they are evidently the result of the change
Tsrj- > Tsr- that was ongoing in the Middle Chinese period; apparent irregularities in
syllables of this type are probably due to dialect mixture:
(897) *s-rak > *srak > srjak > sraewk > shu first day of month; the
phoneticis
= *rak > ngjaek > n go against
[5] As with the *-a rhyme, there are divergent developments of *-ak and *-ak-s (see
below) after acute initials, which are still not well understood. We write *Cak for forms
that become MC Cjak, and *CAk for forms that become MC Cjek. As with the *-A
226 Old Chinese
notation in the previous section, the *A is not intended as a seventh vowel; *-Ak just
means a case of *-ak that, for unexplained reasons, becomes MC -jek, and similarly
for *-Ak-s below. However, in this case we have evidence that dialect variation may be
responsible:in the Mn dialects, words with *-Ak have pMn *-iok, just like those with
*-ak. Evidently, the change of *-ak to MC -jek after dental and palatal initials was an
innovation in the dialects represented in our Middle Chinese written sources, in which
the Mn dialects did notshare:
(898) *tAk > tsyhek > ch foot (measure), pMn *thiok D: Ximn
/tsio7/, Fzhu /tsuo7/
(899) *dAk > dzyek > sh stone, pMn *diok D:Ximn /tsio8/, Fzhu
/suo8/
(900) *[l ]Ak > syek > sh sting (v.), pMn *thiokD:Ximn /tsio7/
(901) *tAk > tsyek > zh roast, broil, pMn *tiok D: Jinyng /tsio 7/
(Norman 1971:203)
(902) *l Ak > syek > sh wash rice, pMn *thiok D: Ximn /tsio 7/
(Ximn dxu 1982:712)
(903) *C.lAk> zyek > sh hit with bow and arrow; pMn *-diok D,
Fzhu /suo8/
(904) *[ts]Ak > tsjek > ji loan, borrow, pMn *tsiokD:Ximn /tsio7/,
Fzhu /tsuo7/
(905) *s-m-tAk> zjek > x mat, pMn *dzhiokD:Ximn /tshio8/
We reconstruct *-ak-s (rather than *-a-s) in words that have clear etymological or
graphic connections with words in *-ak, as in the following examples:
As examples (909) through (911) show, we write *-Ak-s for those cases of *-ak-s
whose Middle Chinese reflex is -jaeH instead of-joH.
Old Chineserhymes 227
The merger of *-a-s and *-ak-s evidently occurred early enough to affect at least
some Shjng rhyming (see Ode 26.2 for an example).
Notes on Table5.30:
[1] Note that as with *-rak, prevocalic *-r- fronts the vowel in *-a after grave ini-
tials (e.g., *Kra > Kjaeng), but not after acute initials, where the reflex of the *-r- is
simply the retroflexion in the initial:*Tra > Trjang, *Tsra > Tsrjang.
There are a few alternations between *-a and *-a in words of similar meaning.For
example, we reconstruct
But the character is also frequently used the oracle-bone inscriptions and other early docu-
ments where the sense suggests not wng disappear but rather w nothave:
suggesting that *ma disappear and *ma not have are etymologically related.
As Pulleyblank (19621963:232233) pointed out, we have a similar alternation in
these examples:
It seems doubtful that this *- was synchronically a productive suffix in Old Chinese,
but it may reflect an older Sino-Tibetan morpheme. Garo, a Tibeto-Burman language
spoken in Bangladesh and northeastern India, has a suffix -ang /a/ away, added to
verbs of motion to indicate motion away from the speaker,asin
Table 5.31 compares our reconstruction of *-, *-k(-s), and *- with earlier
reconstructions.
The rhymes *-, *-k(-s), and *- are characteristically subject to assimilatory
rounding under the influence of a nonpharyngealized labial or labialized onset, but this
assimilation was evidently blocked by prevocalic *-r-; it also does not occur after pha-
ryngealized onsets.For example:
Apart from syllable types like those in examples (916) through (918), the presence
of *-r- in syllables with nonpharyngealized grave initials is undetectable from Middle
Chinese evidencealone.
5.4.2.1 *- (= traditionalZh)
The Middle Chinese reflexes of *- are summarized in Table5.32.
Notes on Table5.32:
[1] MC Koj can also reflect *Kj; Kwoj can also reflect *Kj or *Kuj; and Pwoj
can also reflect *Pj or *Puj. But since *-j is fronted after acute initials, Toj and Tsoj
can regularly reflect only *T and *Ts, respectively.
[2] After nonpharyngealized *K- and *P-, * is rounded by assimilation, and
becomes MC -juw; but prevocalic *-r- blocks this assimilation:*Kr > Kwij, *Pr >
Pij. Note that these reflexes are in the Qiyns Zh rhyme (-ij); there are no hku or
labial-initial syllables in the Zh rhyme(-i).
[3] As mentioned above, *K and *Kr ( Zh rhyme) cannot be distinguished
from Middle Chinese evidencealone.
Notes on Table5.33:
[1] As with the *- and *- rhymes, nonpharyngealized labial or labialized onsets
cause a following * to become rounded, but this rounding is blocked by prevocalic *-r-,
and does not affect type-A syllables.
[2]As with *K/*Kr and *K/*Kr (see Table5.34 below), we cannot distin-
guish *Kk(-s) from *Krk(-s) on the basis of Middle Chinese alone. But we can recon-
struct *-r- in some cases based on other evidence:
(920) *krk > kik > j thorns; also written in the Hn Sh (JDSW
79)and in the Mwngdu Loz (versionA),as
*krk > kik > j thorns (Go Mng 1996:381), where the phoneticis
*k.rk > lik > l strength, pMn *lht D: Ximn /lat 8/, Jinu
/s6/, Shbi/se1/
We reconstruct *-k-s (rather than *--s) on the basis of etymological or graphic con-
nections with *-k, as in these examples:
(921) *pk > pok > bi north (the direction to ones back; the early graph
depicts two persons back to back)
*pk-s > pwojH > bi theback
(922) *(r)k > 'ik > y remember
*(r)k-s > 'iH > y thought(n.)
(923) *m-lk > zyik > sheat
~ ~ 41 *s-m-lk-s > ziH > s feed(v.)
The common word li come shows an irregular development, perhaps due to the
loss of final *-k in an unstressed form that was later restressed:
In the Shjng, li rhymes as *-k in Odes 168.1, 203.4, 242.2, and 263.6, but it
rhymes as *- in Odes 30.2, 33.3, 66.1, and 91.2. The rhymes with *-k are in sections
generally regarded as older than those with *-. We suppose that the final *-k was lost in
unstressed position, and the form without *-k was then restressed, replacing the original
full form; see the discussion in Baxter (1992:325332). (A similar process was respon-
sible for the loss of the initial /h/ in the Middle English neuter pronoun hit > Modern
Englishit.)
Notes on Table5.34:
[1] Parallel to the other rhymes in this section, the vowel is rounded in *K >
Kjuwng and *P > Pjuwng, but this rounding is blocked by prevocalic *-r- and does not
occur in type-A syllables. There seem to be no clear examples of syllables like Kwing <
*Kr.
[2] As with the other rhymes in this section, we cannot distinguish *K from *Kr
on Middle Chinese evidencealone.
The development of the rhymes *-e, *-ek(-s), and *-e is relatively straightforward.
Certain division-III finals from these rhymes have been regarded as irregular in earlier
reconstructions, but they are easily accounted for in six-vowel reconstructions by recon-
structing prevocalic *-r-. Here are some examples:
(925) *pe > pjie > bi low, humble (division-IV chngni syllable)
*pre > pje > bi pillar (division-III chngni syllable)
232 Old Chinese
Notes on Table5.36:
[1] As explained in section 4.1.2, nonpharyngealized velars tend to palatalize before
front vowels *i and *e unless an *-r- intervenes, but the exact conditions for this pala-
talization are not fully understood.
[2] Original MC Tsrje became Tsrea in some dialects; both types of readings are pre-
served in the Jngdin shwn.For example, commenting on the expression z x
black hair-band (in Y l:Sh gun l ), the Jngdin shwnsays:
(929) :
z x:[pronounced] [sr(ean) + (m)eaX=sreaX]; the older pro-
nunciation is [sr(ean) + (kh)jeX=srjeX] (JDSW143).
Notes on Table5.37:
[1] We know of no cases of velar palatalization before *-ek (or*-e).
[2] Because of the Middle Chinese change Tsrj- > Tsr-, it is difficult to distinguish
pharyngealized *Tsrek(-s) from nonpharyngealized *Tsrek(-s) with confidence. From
*Tsrjek we might expect Tsrjaek > Tsraek, by analogyto
(930) *sre > srjaeng > sraeng > shng bear, be born;live
234 Old Chinese
(see the next section), but in fact, the syllables in the Gungyn with either -jaek or
-aek after Tsr-type initials are either from the *-ak rhyme or are not clearly attested in
pre-Qntexts.
We reconstruct *-ek-s (rather than *-e-s) based on etymological or graphic connec-
tions with*-ek:
But in both Mwngdu versions of the text, the expression is written insteadas
with *-en instead of *-e (Go Mng 1996:356). It is difficult to be sure what happened
to the text here, but at least it is evidence that rng < hjwaeng < *[N-q]re and
hun < hwaen < *C.<r>en had the same mainvowel.
In traditional terms, the fact that *bre has an alternative pronunciation bjien < *ben,
and the fact that one version of the Loz text has *[N-q]re while another has *C.ren,
would be described as cases of Gng-Yun pngzhun , that is, an alterna-
tion between the two traditional nasal-final groups Gng (Lis *-ing) and Yun (Lis
*-an). Notice, however, that Gng-Yun pngzhun is a description of the phenomenon,
Old Chineserhymes 235
not an explanation of it; with a phonetic reconstruction we can propose reasons why such
alternations might occur (that a dialect might have changed *-e to *-en, or perhaps that in
the Loz text *-en was confused with *-e before a following *k-). Secondly, the traditional
terminology overgeneralizes the phenomenon:it is not the entire Yun rhyme group that
shows connections with the Gng group (= our *-e), but only that subset of the Yun
group that we reconstruct with*-en.
The Middle Chinese reflexes of *-e are summarized in Table5.38.
*Ke Kweng *ke > kweng > jing region distant from capital
*k-le > keng > jng regulate; norm
*Ce Ceng *me > meng > mng inscription
*te > tengX > dng top of the head
*[dz]re-[]re > dzreang-hweang > zhngrng
*Kre Kweang
high, precipitous
*k<r>e > keang > gng plow (v.)
*Cre Ceang
*pre-s > peangH > bng drive out
*Ke Kjieng [1][2] *[k]e > khjieng > qng light ( heavy)
*Kre Kjaeng *kre > kjaeng > jng be afraid
*Ke Kjwieng *[k]e > khjwiengX > qng interval, short while
*Kre Kjwaeng [2] *[N-q]re > hjwaeng > rng glory, honor
*Pe Pjieng *C.me > mjieng > mng name
*Pre Pjaeng *bre > bjaeng > png even (adj.)
*Tsre Tsrjaeng ~ Tsraeng [3] *sre > srjaeng > sraeng > shng bear, be born; live
*te-s > tsyengH > zhng correct (adj., v.)
*C(r)e Cjeng *tse > tshjeng > qng clear (adj.)
*tre > trjeng > zhn divine (v.)
Notes on Table5.38:
[1]Note that nonpharyngealized velar initials do not appear to palatalize before
*-e; that is, we know of no cases of *Ke > Tsyeng.
[2] After nonpharyngealized grave initials there is a distinction between -jieng (in
the Gungyns Qng rhyme, placed in division IV of the rhyme tables) and -jaeng
(in the Gungyns Gng rhyme, placed in division III of the rhyme tables). We
account for this by reconstructing *-r- before the vowel in the latter. The result is that
-jaeng has two origins:one from the *-a rhyme and one from the *-e rhyme, as in the
following minimalpair:
(938) *[k]ra > kjaeng > jng hill; capital city; phonetic in *C.ra >
ljang > ling cold
*kre > kjaeng > jng be afraid; cf. Siamese /kre: A1/ to fear
(Manovaibimool 1975:168).
In most earlier reconstructions, MC -jaeng from *-e was unaccounted for and consid-
ered irregular.
236 Old Chinese
[3]The word shng < sraeng has also generally been regarded as irregular, but
the reading sraeng (from the Gungyn: :i.e., sr(joX) + (k)aeng=sraeng) just
reflects the change of Tsrj- to Tsr- that was ongoing in the Middle Chinese period. In
fact, the earlier Wng Rnx Qiyn gives the pronunciation as srjaeng ( :i.e.,
sr(joX) + (k)jaeng=srjaeng), where this change has not takenplace.
We have found no reason to reconstruct an Old Chinese rhyme *-i with zero coda in
contrast with OC *-ij. An alternative analysis would be to reconstruct *-i instead of *-ij
(as Zhngzhng does, 2003:159168), but various developments are somewhat more
conveniently described if we reconstruct*-ij.
However, we do reconstruct *-ik(-s) and *-i. The predominant development is
that *-ik(-s) and *-i merged with either *-it(-s) and *-in or with *-ek(-s) and *-e,
respectively. The rhymes *-ik(-s) and *-i can generally be identified by the fact that
they appear to show connections both with *-it(s) or *-in and with *-ek(-s) or *-e.
Also, in type B we sometimes have *-ik > -ik and *-i > -ing, suggesting a third type
of dialect in which neither of these mergers took place.45 In most earlier reconstruc-
tions, the words we reconstruct with *-ik(-s) and *-i are either reconstructed as if
they were *-it(-s) and *-in (the traditional Zh and Zhn rhyme groups) or as if
they were *-ek(s) and *-e (the traditional X and Gng groups), according to
their Middle Chinese readings, without explaining the connections between *-it(-s)
and *-in on the one hand and *-ek(s) and *-e on the other, and without explaining the
anomalous forms with MC -ik and -ing. Our reconstruction is compared with earlier
ones in Table5.39.
We saw in section 5.2 that we could account for the overall pattern of Middle Chinese
velar-final codas with only five vowels, without including *i, as in Table5.8 (repeated
here as Table5.40).
It was when we turned to syllables in MC -n that it became clear that a sixth vowel *i
was also needed. The apparent absence of the rhymes *-i and *-ik then seems to be a
systematicgap.
However, there is considerable evidence that the rhymes *-i and *-ik did exist. We
reconstruct these in words that show connections with both *-in and *-it on the one hand
and *-e and *-ek on the other; some words of this type also have apparently irregu-
lar reflexes with MC -ing and -ik (which otherwise usually reflect OC *- and *-k,
respectively).
For example, Karlgren mentions three words that have Middle Chinese readings
in final -ng, but which appear to rhyme as *-in in the Shjng. In GSR (762a) he notes
that mng < mjaengH command rhymes several times in the Shjng as if it were
*-in;46 he made similar remarks about lng command(GSR 823a).47 We reconstruct
these words as follows and assume that the apparent irregularities in rhyming are due to
dialect mixture, either in Old Chinese itself, or in the Middle Chinese reading tradition:
(939) *m-ri-s48 (dial.) > *mri-s > *mre-s > mjaengH > mng com-
mand (n.); dialectally also *m-ri-s > *mrin-s (as shown by Shjng
rhymes, e.g., Odes 51.3, 116.3,222.3)
(940) *ri > *re > ljeng > lng issue a command; dialectally also *ri >
*rin (as shown by Shjng rhymes, e.g., Odes 100.2 and126.1)
A slightly different situation obtains with jn < king pity:it appears to rhyme as if it
were *-in, but it has the final -ing in Middle Chinese. Our reconstructionis
(941) *k-ri > king > jn pity (v.); a separate meaning is boastful
Here, the Middle Chinese reading king appears to be anomalous:if *-i had become
*-in in this word we would expect MC kin;49 or if *-i had become *-e, then we would
expect MC kjaeng. In most reconstructions, the only regular source for MC -ing is OC
*-. Our Middle Chinese sources seem to show at least three different treatments of
*-i in nonpharyngealized syllables:
(942) *Ci > *-in > -in dialect where *-i > *-in
*Ci > *-e > -jeng dialect where *-i > *-e
*Ci > -ing conservative dialect that retains *-i
238 Old Chinese
In the case of lng, the Gungyn has no trace of the conservative treatment of *-i,
but the Jngdin shwn gives the reading lingH for the line from the court
they order him (Ode 100.2; JDSW66).
Reconstructing *-i in jn pity (v.) is also supported by paleographic evidence.
The Shuwnsays:
(943)
jn [means] the handle of a spear. It is composed of mo [spear],
with [jn < kim < *[k]r[]m now] as phonetic (SWGL6395b)
But it is implausible that jn < kim < *[k]r[]m now is phonetic, because the final
-m does not match the final -ng of jn < king. Dun Yci pointed out that the origi-
nal phonetic is not jn but lng < *ri (Dun Yci [1815] 1981:719720), sup-
porting both the vowel *i and the prevocalic *r in our reconstruction: jn < king <
*k-ri (see section 4.4.4.4 for the development *k-r- > k-). This has been confirmed
from recently excavated documents. The following passage occurs in the received ver-
sions of Loz30:
(944)
gu rwjn
Accomplish it but do not be boastful
Happily, this passage occurs in both Mwngdu versions of the text (early Western
Hn) and in the Gudin bamboo strips version. Both Mwngdu versions have
(with lng on the right instead of jn). In the Gudin Aversion (GD 3, strip 7), the
character is written thisway:
(945)
This graph is composed of mo on the left and mng on the right. So these three
words, which rhyme as if they were *-in, but have velar codas in Middle Chinese, are
written with the same phonetic element *ri.
In fact, we consider it likely that the {} *k-ri meaning boastful is from the same
root as *m-ri-s; perhaps the intended meaning is something like imperious or
inclined to order people around. The {} *k-ri meaning pity, on the other hand, is
probably related (though perhaps not synchronically)to
(946) *ri > *rin > len > lin love; pity (The character is rather late;
the earliest form of this character in Gwnz gln (GG 8.1065) is
from a Qn stone inscription.)
(948)
which consists of xn heart below with shn body above, as phonetic. Sometimes
we have qin thousand instead of shn (Zhng xn zh do , GD
45, strip8):
(949)
Because of these characters, we can now identify the initial consonant of shn < syin
as *n -:50
The early character for {} qin < tshen thousand itself is composed of the character
rn < nyin with an added stroke (J Xshng 2010:154); for two thousand, a second
stroke is added,etc.:
(951)
We reconstruct
(952) *s.n i[] > tshen > qin thousand
(953) *ni[] > nyin > rn (other) person
To go one step further, {} nin < nen harvest; year can be written with either
*s.n i[] or *ni[] as phonetic; we reconstructitas
(954) *C.ni[] > nen > nin harvest;year
Now although they are all connected, directly or indirectly, with {} rn kind-
ness, which we have decided to reconstruct with *-i, neither shn, qin,
rn, nor nin themselves have Middle Chinese forms that preserve a velar coda; it
is difficult to be sure how far to extend the reconstruction of *-i. Forms like (948)
and (949) might represent a dialect where *-i had already merged with *-in; so the
fact that shn is used as a phonetic to write {} *ni kindness does not necessar-
ily prove that shn itself originally had *-i rather than*-in.
Similarly, there are forms that show a vacillation between the reflexes of *-it and
those of *-ek, and occasional forms with MC -ik. Middle Chinese -ik usually reflects
*-k, in which case it will typically have xishng contacts with MC -ok; but cases of -ik
that instead show contacts with MC -it or -et can be reconstructed with*-ik:
(956) *srik > srit ~ srik > sh louse; cf. Proto-Tibeto-Burman *s-r(y)ik
louse (Matisoff 2003:153). The MC reading srik is found in
Xunyngs Yqi jng yny , jun 17 (Zhu Fgo
1962, #791).51
(957) *[](r)ik > 'ik > y rub, repress
*[]i-s > *[]in-s > 'jinH > ynseal
*[]<r>ik-s > 'ijH > y repress; the phonetic element is also used to
write words with MC -it and-et.
In (957), the alternation between *-k and *- in coda position is unexplained, but the
words seem to be related semantically, and paleographers generally agree that both
repress and seal were originally written the same, with a graph originally depict-
ing a hand ( zho or yu) over a kneeling person (Chn Jin, p.c.; J Xshng
2010:736):
(958)
Such alternations between stops and nasals in coda position are a question for future
research (see section 6.3, Known issues).
Sagart (1999a) also made the case for reconstructing *-ik in xi ~ xu < xwet
blood:
This shows the cluster of apparent etymological and graphic connections to both *-it
and *-ek, and with MC -ik, that lead us to reconstruct *-ik.Cf.
(960) *C.m<r>[i]k > meak > mivein
(961) ~ *m (r)ik > xwik > x channel; moat. It is not clear which way
of writing this word is older, but in either case the use of to write
xwik is an argument for *-ik in *m ik, even if not for*m-.
*m (r)ik > xwik > x still, quiet; also read *m ik-s > *mit-s >
xjwijH>x
A similar case is the common word r < nyit sun; day. There are two indepen-
dent bits of evidence that this word should be reconstructed with *-ik rather than *-it.
First, it appears to be phonetic in the character used to write m < mek, the name
of a river in Hnn. (The M joins with the Lu river to form the Mlu ,
the river in which the poet Q Yun [c. 340278 BCE] is said to have drowned
himself.) The character is clearly composed of shu water and r sun; day.
The Shuwn text as we have it today says that r is an abbreviated form of mng
< meng and that this is the real phonetic element (SWGL 4864b), but this seems highly
implausible; it is much more likely that r itself is the phonetic (GG 9:38). We
can now account for this unusual phonetic by reconstructing m < mek < *m.nik
M river (see section 4.4.2.4) and *C.nik > *C.nit > nyit > r sun; day (with
preinitial *C.to account for high-register tones in some Mn and Hakka dialects; see
Old Chineserhymes 241
Norman 1991:211 and section 4.4.5.4). Since *-ik > *-it is a well-supported dialect
development, these reconstructions can account for the final -t in n < nyit; but if
we reconstructed *-it in n < nyit there would be no way to account for the final -k
in m<mek.
The second argument for a final *-k in r < nyit sun; day is its presence in the
characters and used to write n < nrit close, intimate and as phonetic in the
probably related word *nik > nyit > r a ladys clothes nearest to the body; we
reconstruct these with a final *-k, as *n<r>ik and *nik, respectively. The Jngdin
shwn gives the reading nrik for an occurrence of n in a commentary to Lny
(JDSW 350). In n < nrit, the phonetic n < nrik conceal also
indicates a final *-k. In both and , the element r probably originally served
a phonetic function; attempts to explain it as a semantic element (SWGL 2938a)
seem quite forced.
Table 5.41 compares our reconstruction of *-o, *-ok(-s), and *-o with earlier
reconstructions.
One peculiarity of these finals is that, contrary to the usual pattern, *-r- in some
cases leaves no trace on the Middle Chinese final:there are no division-II reflexes
from the rhyme *-o as there are in others; and based on Middle Chinese evidence
alone, we cannot distinguish *Kro from *Ko (both > Kuw) or *Kro from *Ko
(both > Kjowng). In type-A syllables, however, *o apparently diphthongized before
velar codas, and *-r- does lead to different Middle Chinese reflexes:*Kok > Kuwk,
*Krok (> *Krawk) > Kaewk. The rule seems to be that *-r- had no effect on the
rhyme when the following vowel was rounded at the time *-r- was lost. If a diphthon-
gization made the following vowel unrounded, then *-r- has its usual effects. This
would also explain why (for example) *Pra and *Pa have the same reflexes (see sec-
tion 5.4.1.1):by the time the *-r- was lost, the vowel in such syllables had probably
already changed from *a to a rounded [], [o], or[u].
Notes on Table5.42:
[1] As noted above, in type-A syllables, *-r- seems to have had no effect on the
Middle Chinese reflexes of the *-o rhyme (except in cases where it led to a retroflex
initial); in many cases we can reconstruct *-r- only on the basis of other evidence.For
example, in expressive binomes showing an alternation of *e and *o, both syllables are
generally the same except for the vowel, so from the first syllable of this expression we
can infer that the second syllable also had*-r-:
[2] In syllables like *Tsro (and possibly *Tro also, although clear examples are lack-
ing), it appears that the *-r- produced a retroflex initial in Middle Chinese, but *-o devel-
oped to MC -uw in type Aas expected; in the Qiyn system, the resulting syllables of the
form Tsruw are treated as if they have merged with Tsrjuw. Examples:
Previous reconstructions have been unable to account for examples like (964) and (965),
and simply treat them as irregular.
Notes on Table5.43:
[1] In contrast to the *-o rhyme, in type-A syllables, an *-r- before the *-ok rhyme
(and before *-o as well; see below) produces a distinct Middle Chinese reflex, probably
through a process of diphthongization:
(968) *[p]rok > *prawk > paewk > b flay, peel; compare, with the same
phonetic:
*(p.)rok > luwk > l blessing
[2] Because of the Middle Chinese change Tsrj- > Tsr-, we cannot reliably distin-
guish *Tsrok from *Tsrok. The following are clearly related etymologically, but we
cannot explain the morphology of shu < sraewk at present:
[3]However, in type-B syllables, *Krok and *Prok are still indistinguishable from
*Kok and *Pok on the basis of Middle Chinese evidence alone, and we frequently
write *K(r)ok and *P(r)ok. In some cases direct evidence for an *-r- is available:
(970) *[]rok > ngjowk > y lawsuit; prison; cf. Khmu /grk/ sty, pen; jail.52
Notes on Table5.44:
[1] As with *-ok, *-o and *-ro have different Middle Chinese reflexes after pha-
ryngealized onsets, probably due to a diphthongization process:
(976) *ko > kuwng > gngwork
*kro > *kraw > kaewng > jing (Yngz)river
Old Chineserhymes 245
[2] But after nonpharyngealized grave initials, *-ro and *-o cannot be distin-
guished on the basis of Middle Chinesealone:
(977) ~ ~ *k<r>o > kjowng > gng respectful; reconstructed with
*-r- because often written in early documents with the phonetic:
*[m]-ro > ljowng > lng dragon
In the early script, the element , present in the various forms of {}, originally
depicts two hands offering a vessel of some kind (J Xshng 2010:176):
(978)
We reconstruct
Table 5.45 compares our reconstruction of *-u, *-uk(-s), and *-u with previous
reconstructions.
In type-A syllables, *-u, *-uk(-s), and probably *-u diphthongized through the low-
ering influence of the pharyngealized initial, and no longer rhymed with their type-B
counterparts in Middle Chinese. In type-B syllables of this group, prevocalic *-r- after
grave initials cannot be detected from Middle Chinese evidence alone; we can account
for this if we assume that *-r- did not affect the Middle Chinese final when the following
vowel was rounded at the time *-r- was dropped.
The distinction between syllables like *C(r)u and *C(r)aw (which merge as MC
Caw or Caew) seems to have survived long enough to be reflected in early loans into
Hmong-Mien and Bi, in which we find the correspondences in (980). The Bi forms
are from the Jinchun dialect.
(980) OC Proto-Hmong-Mien Bi Middle Chinese
*Cu *-u /-u/ -aw
*Caw *-w /-a/ -aw
(981) OC *Cu:
*[k.l]u > dawX > do way, Proto-Hmong-Mien *kluX, Bi
/thu1/
*[l]u > dawX > do rice, paddy, Proto-Hmong-Mien *mblu
*u > 'awX > o old woman, Proto-Hmong-Mien *uX elder
sister/wife
*[ts]u > tshawX > co grass, plants, Bi /chu 1/
*qu > xawX > ho good, Bi/xu1/
(982) OC *Caw:
*C.law > daw > to peach, Proto-Hmong-Mien *lw A, Bi /ta 7/
*[C.g]aw > haw > ho call out, Proto-Hmong-Mien *Gw Acry
out/sing
*m-tsaw > tsawX > zo bleach; wash, Proto-Hmong-Mien
*ntswX wash
*C.maw > maw > mo hair, Bi /ma 7/
*C.taw > taw > do knife, Bi/t4/
The fact that OC *Cu and *Caw have separate reflexes in Hmong-Mien and Bi sug-
gests an estimate of the time these words were borrowed:according to Tings study of
rhymes in poetry (1975:238), OC *Cu and *Caw still rhymed separately in Eastern
Hn (25220 CE) but began to rhyme together in the Wi-Jn period (220420 CE); so we
can estimate that these words were borrowed into Hmong-Mien and Bi before about
the middle of the third century CE. Of course, this date is based on overall patterns in
poetic rhyming; the situation must have differed from dialect to dialect, so it is only a
rough estimate.
The Middle Chinese sources are inconsistent in representing the reflexes of *mu,
*muk(-s), and *mu:sometimes we find mjuw, mjuwk, and mjuwng (as expected), and
sometimes we find muw, muwk, and muwng. Normally we would expect muw, muwk,
muwng to reflect *mo, *mok, and *mo, respectively, but there was evidently a com-
mon but not universal change by which (in terms of our notation) MC -j- was lost in the
syllables mjuw, mjuwk, and mjuwng. The situation is similar to the vacillation between
Tsrj- andTsr-.
For example, the following are both mjuw in the Gungyn:
The Jngdin shwn, however, appears to be inconsistent:it has several glosses for
mo, some of which say pronounced like [mu < mjuw] (JDSW 62)and some of
which say [m(jang) + (h)uw=muw] (JDSW 64). The word mo clearly
Old Chineserhymes 247
rhymes in the Shjng as *-u (Odes 133.1, 191.8). Evidently we have *m(r)u > mjuw
as expected, but in some varieties of Middle Chinese there was a further change to
muw. Similarly, MC muwk and muwng sometimes reflect OC *muk and *mu instead
of *mok and*mo.
Notes on Table5.46:
[1] In type-A syllables, the lowering influence of the pharyngealized initial leads to
a diphthongization that causes the reflexes of *-u and *-ru to merge with those of *-aw
and *-raw, respectively, in Middle Chinese:e.g., *Ku > Kaw, merging with original
*Kaw, and *Tsru > *Tsraw > Tsraew, merging with original *Tsraw.
[2] The contrast in this rhyme group between MC -juw and -wij has long been a
puzzle. Our solution, adapted from Li (1971:3132), is to assume a dissimilation of
the rounded vowel in the syllable types *Ku and *Kru under the influence of the
labialized initial:
The resulting syllable types develop like *K and *Kr (see Table5.32 and the
discussion in section 5.4.2.1):a later assimilation rounds the vowel in *K < *Ku,
giving MC Kjuw (the same as if it had originally been *Ku), but this assimilation
is blocked by *-r-, so that *Kr < *Kru becomes MC Kwij, like original *Kr.54
The scenario for the development of MC Kjuw and Kwij from the two rhymes *-u
and *- is summarized in Table5.47.
[3] As noted above, from *m(r)u we often have MC muw instead of the
expectedmjuw:
[4] Reflexes like MC suw result from two changes observed in Middle Chinese
sources:Tsrj- > Tsr- and an early merger of initials of type Tsr- with Ts-; see the discus-
sion at example (581)above.
[5] After nonpharyngealized *K- and *P-, we cannot distinguish *-u from *-ru on
the basis of Middle Chinese reflexes alone, but we sometimes can reconstruct *-ru in
such syllables from etymological or graphic evidence:
(992) *mru-s > mjuwH > muwH > mo to barter; cf., with the same
phonetic element:
*mru > maewX > mo fourth earthly branch
*m-ru > ljuw > li kill (also a surname)
Notes on Table5.48:
[1] We normally cannot distinguish *-ruk from *-uk after (nonpharyngealized) *K-
and *P-. Apossible candidate for the syllable shape *Pruk wouldbe
(993) ?*mruk > mjuwk > m concord; cf., with the same phonetic:
*[r]uk > ljuwk > l land (as opposed to water)
Notes on Table5.49:
[1] It is difficult to find convincing examples of *-u or *-ru with pharyngealized
grave onsets. We generally expect MC -owng as the reflex of *Cu, but some examples
of MC -owng clearly have other origins. The Gungyn gives both kuwng and kowng
as readings for gng attack (as does the Wng Rnx Qiyn manuscript), but the
phonetic gng work clearly has *-o, and gng itself rhymes as *-o in the
Shjng (179.1):
[2] Some cases that look like *-u may come by dissimilation from earlier *-um.For
example,
Example (1002) rhymes as *-u in Odes 14.1, 168.5, 239.2, and 248.4, but words with
the phonetic xing have connections with final *-m; for example, xing is the
phonetic elementin
(1003) *[k]om > khomX > kn (booming sound?)57
For a discussion of this and related words, see Chn Jin (2007).
In fact, Wng L treats Dng (our *-u) and Qn (our *-m, *-im,
and *-um) as a single rhyme group for the Shjng period and before, believ-
ing that the distinction between them arose later. We do not follow him on this;
based on our six-vowel reconstruction, our default assumption would be that
Old Chineserhymes 251
all six vowels would have occurred before all codas, and there is no reason
not to expect that there was a rhyme *-u. There could indeed have been cases where
*-um dissimilated to *-u (see section 5.7 below), but assimilations *-u > *-um are
also possible, and would be one explanation for the scarcity of clear examples of*-u.
The rhymes with acute codas share many common features:for example, there was a
diphthongization of rounded vowels in this environment, probably starting some time in
the late Warring States period, judging from rhymes in texts of that time. The diphthon-
gizations are summarized in Table5.50.
A second change was that some time after these diphthongizations, *-w- became
nondistinctive after labial initials; we cannot be sure of the phonetic details, but we
assume that the developments were more or less as shown in Table5.51 (using *-T as
a cover symbol for any acutecoda).
In the Shjng and Loz, the distinctions between rounded and unrounded vowels are
generally maintained quite strictly in rhyming; there is no sign of the diphthongization.
We have,for example, these sequences with rounded vowels before acutecodas:
But in somewhat later literature, rhymes do occur between original *-oT and
*-aT, and between original *-uT and *-T, suggesting that the diphthongizations
have taken place.For example, in the Ji zhng poems of the Ch c
, we find the following rhymes where original rounded vowels rhyme
with unrounded vowels:
More precise location of this diphthongization in space and time awaits further
research.
In syllables where Baxter (1992) had a two-way contrast between codas *-j and *-n,
our present system follows Starostin (1989) in reconstructing a third coda *-r, so that
there is a three-way contrast in Old Chinese among these codas:*-j *-n *-r. With
Starostin, we assume that the coda *-r was treated differently in different dialects:OC
*-r usually became MC -n, but in some cases, *-r became MC -j (or final zero - from
earlier *-j, in the case of MC -a < *-aj < *-ar and MC -wa < *-waj < *-war < *-or).
Some examples indicate that Proto-Vietic may have had *-l corresponding to OC *-r
in early loanwords; and as we remarked in section 5.1, it would also be possible to
reconstruct the third coda as *-l instead of*-r.
Table 5.52, for the main vowel *. (The Wn and Wi groups also include words we
reconstruct with *u, as discussed in section 5.5.7; we leave these aside for the moment.)
Given the variety of codas that have been reconstructed for the Wi group (*-r,
*-d, *- [= ], *-l, as well as *-i and *-j), it may seem surprising that neither Middle
Chinese nor modern Chinese dialects show any reflex other than [i] for the coda in
these syllables. Karlgrens original reason for reconstructing *-r was the presence of
various kinds of contacts between words of the traditional Wi and Wn rhyme
groups.For example, there are rhymes between the two groups, as in this sequence from
Ode 183.1183.2:
And also this rhyme sequence:
Furthermore, there are many cases where the same phonetic element was used
to write words from both the traditional Wi group and the traditional
Wngroup:
Contacts of this kind were noticed by the Qng philologists; the traditional term for such
contacts between vocalic-final and nasal-final syllables is yn-yng duzhun ,
perhaps to be translated as crossover alternations between vocalic-coda and nasal-coda
254 Old Chinese
jn jn q j
kerchief axe flag how many?
MC kin kj+n gj+j kj+jX
traditional group Wn [ Wn?] [ Wi?] Wi
Karlgren *ki n *ki n *gi r *ki r
Li *kjin *kjn *gjd *kjdx
Baxter (1992) *krjn *kjn *gjj *kjj
Zhngzhng (2003) *krn *kn *gl *kl
Starostin (1989) *krn *kr *gr *kj
Baxter-Sagart *krn *[k]r *C.[]r *kj
Old Chineserhymes 255
Words with *-r form a newly discovered rhyme group, not recognized in the tra-
ditional analysis; *-r can be reconstructed with confidence in words written (at a suf-
ficiently early period) with the following phonetics:
As for rhyme evidence, in addition to Ode 182.3 cited above, we have these rhyme
sequences limited to*-r:
That the tendency of *-r words to rhyme separately from *-j and *-n is not a
random effect of the Shjng sample is suggested by the long rhyme sequence in the
following passage from the Zu zhun (Duke X, year 5). The text is apparently late
enough that original *-ur has already diphthongized to *-wr, but the reconstruction of
*-r in these words is secure:
Words with OC *-r that were borrowed early into other languages sometimes lack
the final -n that we find in Middle Chinese, suggesting either that the donor variety of
Chinese was one that had changed *-r to *-j, or else that the borrowing language had
no [r]. For example, for the cyclical sign chn < *[d]r fifth earthly branch, Li
(1945:336) cites these forms from Tai languages:Ahom shi, L si, Dioi chi.59 Khmu
(a minority Mon-Khmer language spoken primarily in Laos) has /si/ (Damrong and
Lindell 1994:104). If the original coda had been [n], there is no reason it could not have
been borrowed with[n].
256 Old Chinese
(1019) *aj > ngaX > w we, I, Ximn /gua 3/, Fzhu /uai 3/, pMn
*ui B (the upper-register tone in these dialects is unexplained; see
Norman 1973:232)
fn fn bb g
reverse burn martial sing
MC pjonX bjon pa ka
traditional group Yun [ Yun?] [ G?] G
Karlgren (1957) *pi wn *bi wn *pwr *k
Li (1971) *pjanx *pjan *par *kar
Baxter (1992) *pjan *bjan *paj *kaj
Zhngzhng (2003) *pan *ban *paal *kaal
Starostin (1989) *pan *bar *pr *kj
Baxter-Sagart *C.pan *[b]ar *par *[k]aj
Old Chineserhymes 257
(1020) *C.raj > la > lu hamper, basket (n.), Ximn /lua 2/, Fzhu
/lai 2/, Jinyng /sue 2/, pMn *lhi A(Lu Jiru [Norman] 2005:3)
(1021) *paj-s > phaH > p break (v.), Ximn /pua 5/, Fzhu /puai 5/,
pMn *phui C; cf. VN phi [fi C1] to touch or fall upon some-
thing by misfortune, phi tu [fi C1 tau A2] to be shipwrecked
(Rhodes 1651:tu=boat61).
(1022) *aj-s > ngaH > hungry, Wnzhu /ai6/
(1023) *maj > ma > m rub, grind, Ximn /bua 2/, Fzhu /muai 2/,
Show /mai 2/, pMn *mui A; cf. VN mi [mi A2] to file,
sharpen, whet, Korean may grindstone
(1024) *C.mraj > mae > m hemp, Ximn /mua 2/, Fzhu /muai 2/,
Show /mai 7/, pMn *mhuiA
(1025) *C.[k](r)aj-s > kjeH > j entrust to, Ximn /kia 5/, Fzhu /kie5/,
pMn *kii C; cf. VN gi [i C1] ~ gi [i C1] entrust,send
(1026) *C.lAj > zyae > sh snake, Ximn /tsua 2/, Fzhu /sie 2/,
pMn *-diiA
However, along with words of the traditional G rhyme group that are unproblemati-
cally reconstructed with *-aj, there are a number of words that show a vacillation between
the traditional Yun and G rhyme groups, parallel to that discussed above between
Wn and Wi; it is these words that we reconstruct with *-ar. Contacts in rhyming
include the following, listed with their traditional rhyme groups:
Contacts between the Yun and G groups among words written with the
same phonetic element are generally a sign that the words should be reconstructed
with*-ar:
(1030) *C.tar > *C.tan > tan > dn single, simple, Ximn /tu 1/, Shbi
/duai 2/, pMn *-tun A
*[d]ar > *daj > da > tu, alligator; also read *[d]ar > *dan >
dan
*tar > *taj > tsye > zh ritual vessel
(1031) *nar > *nan > nan > nn difficult
*nar > *naj > na > nu expel demons
(1032) *s-qar > *s-qan > sjwen > xun spread (v.)
*qar > *qan > xjwonX > xun to dry in the sun;
also read
*qar > *qaj > xjweX > hu sunlight
*[]ar > *[]an > hjwon > yun wall
*[]ar > *[]an > hwan > hun pillar; martial-looking
The phonetic element xun < sjwen < *swar < *s-[q]ar is used by Zh Lujichn
(?Lokakema the Yuzh ; see Zrcher 2007:35), a Buddhist transla-
tor active in Luyng in the late second century CE, to represent the Indic syllable svar in
(Sanskrit) abhasvara shining; a class of deities (Coblin 1983:244,#68):
The early script distinguishes *-ar from *-an:for example, the phonetic fn
represents syllables of the type *P()ar, contrasting with , which represents *P()an;
similarly, dn represents *T()ar, contrasting with dn, which represents
*T()an. The later standard script no longer maintains these distinctions clearly, prob-
ably because of the change of *-r to *-nand more generally, because sound changes
gradually made the criteria for a phonetic match less strict. So in received texts there
are cases where words of the shape *T()an are written with .For example, Ode
254.1 has this rhyme sequence:
While most of the words can reasonably be reconstructed with *-an, the phonetic
dn in dn illness, toil would seem to indicate *-ar. But the earliest attested
version of the text is the passage quoted in the bamboo strip version of the L j
chapter Z y from Gudin (strip 7); there, the character
corresponding to the Mo versions MC tanX > dn is writtenas
(1039)
that is, as , with shu hand on the left and dn dawn on the right, which
is consistent with *-an (GD 17, 129). In the received version of the Z y, the char-
acter is written as , with the phonetic dn sincere, which itself contains dn
dawn as phonetic. The Jngdin shwn glosses on Ode 254.1 also cite a version of the
Shjng in which the character is written as (JDSW 95). As often, the received ver-
sion of the text of the Shjng, written in the standard script, probably reflects late sound
changes that are irrelevant to reconstructing Old Chinese.
The reconstruction of *-r is also supported by rhyme sequences from texts other than
the Shjng;for example, in the Zhu Y (hexagram B), we have this
rhyme sequence:
There are also relatively clear examples of *-r after *o and *u (see sections 5.5.4.4
and 5.5.7.4 below).
chny < dzyen-hju < (Hn) *dar-wa Xingn ruler (< *[d]ar + *(r)a)
The two syllables individually would be reconstructed for Old Chinese as follows:
which is a close match to Written Mongolian darua governor (also borrowed into
Persian as dra governor, see Doerfer 19631975, 1.3191.323). This does not nec-
essarily mean that the Xingn themselves were Proto-Mongols, of course, since the
word could have been borrowed, either by the Xingn or by the Mongols, one from the
other, or both from some third source.
Khwrazm (also known as Chorasmia) is an oasis region in the lower Amu Darya
valley, in parts of what are now Turkmenistan and Uzbekistan, near the former Aral Sea
(now largely desiccated). Chinese representations of this name are discussed by Pelliot
(1938). The earliest known mention of it in Chinese texts is in the Dyun lizhun
of the Sh j (1982:3173):it is described as a small country
to the west of Dyun (Ferghana), and its name is writtenas
The region is mentioned already around 500 BCE in records of the Old Persian empire; it
appears in Old Persian as Huwrazmi (Mackenzie1983).
The phonetic of the first character hun indicates an *-rcoda:
(1047) (= =) *qar (> *xwar) > xwan > hun rejoice; joyous
The phonetic is (=) *C.qar-s > kwanH > gun heron, whose phonetic in turn
is saidtobe
This same word is written as , , or ; note that these last two have the phonetic
which we identified above as a clear case of *-ar. By Western Hn, *qar would prob-
ably already have changed to *xwar. The second character by itselfis
The name Hn < han < *[g]ar, apart from being the name of a Chinese state of
the Warring States period, was also applied to a state in the Korean peninsula in the Sn
gu zh , completed in 297 CE.67 In Old Japanese, the corresponding
name is Kara. The character hn and graphically related words are among the
best-supported examples of *-ar. The Shuwn gives an earlier form of the character as
, and says that the phonetic element is . The Shuwns textis:
(1051)
[*[g]ar] means wall [ *[]ar] of a well. It is composed of
[wi= wi surround], taking [as a basis] its circumference, with
[gn < kanH < *[k]ar-s sunrise] as phonetic (SWGL2347b)
The gloss wall of a well probably refers to the word usually written gn < kanH <
*[k]ar-s, which does have this meaning. The word yun wall is also reconstructed
with *-ar, and is probably used as a sound gloss here. There is ample evidence that
words with the phonetic gn are to be reconstructed with *-ar.68
In early Western Hn, the Dng H were a non-Chinese group occupying a ter-
ritory to the east of the Xingn; by the late first century BCE, two subgroups of the Dng
262 Old Chinese
H were recognized under the names Xinbi and Whun .69 According to
Pulleyblank (1983:452454), these names can be interpreted as *Srbi and Avar,
respectively. We discuss the Xinbifirst.
Although the name *Srbi is not independently attested, what is apparently
the same name is also found written later as Shwi < MC syit.hjw+j. Pelliot
reasoned that since foreign -r can be represented by MC -n in Hn times, and is
often represented by MC -t in later times, the coda of the first syllable of the foreign
name was *-r in this case; he reconstructed the name as *Serbi, *irbi, or *irvi
(1934:35,n.3).
There are several additional reasons for reconstructing with *-r. First, it rhymes in
the Shjng in the following rhyme sequence:
There are many phonological difficulties with this rhyme sequence, but the mix-
ture of forms with and without the MC -n suggests that the MC -n in sjen is
from*-r.70
The Shuwn (SWGL 5188a) also includes the following character with sjen
as phonetic element, which has two Middle Chinese pronunciations according to the
Gungyn:
We will see below (section 5.5.1.4) that the Eastern Hn commentator Zhng Xun
reported that xin < sjen was pronounced like s < sje < *[s]e this in the
Q and L areas (in the region where we believe *-r had changed early to*-j).
The Whun are the other, more southerly branch of the Dng H mentioned
above. Pulleyblank notes that this would be a possible transcription of a foreign pronun-
ciation like *Awar, and argues that they are the same group who appear in the fourth and
fifth century in the Hephthalite kingdom, in modern Afghanistan, and move west into
Europe when displaced from that region by the Turks. In Byzantine and European texts
they are referred to as Avares or (the Greek beta being pronounced like [v] at
the time).71 The identification is made more plausible by the fact that both Chinese and
Western sources mention the unusual gold- and jewel-encrusted headwear worn by the
groups married women (Pulleyblank 1983:452454).
Old Chineserhymes 263
The Chinese name of this famous outpost on the old Silk Road is evidently based
on its name in Sogdian, an important Iranian language long used in the region.
In Sogdian, written in a consonantal script derived from the Aramaic alphabet,
the name is drwn (the apostrophe stands for the letter aleph, representing an [a]
vowel); in Greek the name is represented as (Pulleyblank 19621963:228).
Here too we have ample evidence for reconstructing *-r in the pronunciation of
dn, because it shows the characteristic alternation of MC -n and -j that we attribute
to the *-rcoda:
Previously, the use of words with MC -n to describe foreign [r] was explained by
saying that the Chinese of the time had no final [r], and that OC *-n was felt to be the
closest equivalent, perhaps having a special [r]-like pronunciation that would make
this more plausible (Pelliot 1934, Pulleyblank 19621963). But it is striking how many
of the words involved are among those that should be reconstructed with *-r under
Starostins hypothesis, for independent reasons. In the examples cited so far, the case
for *-r seems very strong because of the overlapping of several kinds of evidence.
It must be admitted, however, that it is sometimes difficult to know precisely how
wide to throw the net when reconstructing final *-r. Anumber of words in MC -n have
only one or two rhyme or xishng connections with *-r words, and these might reflect
simple irregularities, or dialects where *-r and *-n had merged.For example, it is tempt-
ing to reconstruct *-r in n < 'an peace because of the name nx, attested as
early as the Dyun lizhun of the Sh j as the name
of an Iranian country in the western regions:
The name is evidently based on the name Araka=Arsaces (Parthian rk), the founder
of the Arsacid dynasty of Parthia, which began in about 247 BCE (Bivar 2000:98). To
reconstruct *-r here would entail extending the *-r to a large number of other common
words for which we have little direct evidence for *-r (including the adverbials yn
< hjen and rn <nyen).
Moreover, the same character is used later to write [n]in the name of the Roman
emperor Marcus Aurelius Antoninus (121180 CE) in the chapter X y zhun
(Account of the western regions) in the Hu Hn sh :
Admittedly, this transcription is several centuries later than the others, and probably
reflects a period or dialect where *-r had already become *-n, as strongly implied by
the fact that the second character *tur > *tun > twon > dn solid, thick, with an
264 Old Chinese
*-r coda, is also used to write -ton- in the Latin word. But Pelliot notes that the name
nx might have come to Chinese by way of Sogdian, and that there is some
indication of a tendency within Sogdian to replace -r- with -n- (Pelliot 1938:146,
n.1, citing mile Benveniste), so perhaps n had *-n in Chinese all along. For
this reason, we have followed a somewhat conservative path, reconstructing *-r
only when it is supported by several examples or kinds of evidence; for n <
'an we write *[]a[n], meaning that we are unsure whether the coda is *-n or*-r.72
The following sentence occurs in the Zhng yng , a text from the L j
that was eventually chosen by Zh X (11301200) as one of the Four
Books (S sh ); it relates how King W of Zhu , who overthrew the
Shng dynasty (also referred to by the name of its last capital, Yn ), became ruler of
all under heaven:
(1057)
y rng y r yu tinxi (?)
He united military force against Y (?)and took possession of all
under heaven.
The meaning of the sentence is debated. The expression rngy usually means
armor, so it is commonly understood as James Legge translated it:He once buckled
on his armor, and got possession of the kingdom. But Zhng Xun (127200 CE), who
wrote commentaries on the Zhu l , the Y l , and the L j
, took y clothing as an error for Yn, referring to the Shng dynasty.
His commentis:
(1058)
Y [*(r)j] should be read as Yn [*r]; it is an error in
pronunciation. When the people of Q pronounce , the sound is
like y.
Old Chineserhymes 265
In other words, Zhng Xun says that the substitution of y for Yn is the
kind of error a person from Q would be likely to make, because of their local
pronunciation. Whether Zhng Xuns interpretation of the text is correct or not,
we can probably trust him when he says that in Q, Yn was pronounced like
y, especially since this is confirmed by Go Yu , another Eastern Hn
commentator (see below). But it is not immediately clear what conclusions to draw
from what Zhng Xun says. Did the people of Q make this mistake in just this one
word, or was it something more systematic? In traditional terminology, y < 'j+j
is in the Wi rhyme group, while yn < 'j+n is in the Wn rhyme group.
Does this mean that in the dialect of Q, the Wi and Wn rhyme groups were
not distinguished?
With an improved phonological reconstruction, we believe that we can understand
Zhng Xuns comment more precisely. We reconstruct the two words involved as
follows:
(1059) *(r)j > 'j+j > y clothes (In the Gungyn, there is no syllable 'ij
contrasting with 'j+j, so we cannot be sure that there was no *-r-.)
*r > 'j+n > yn (dynasticname)
The characteristic of the Q dialect that Zhng Xun refers to is the fact that in the
speech of that region, the Old Chinese coda *-r had changed to *-j. There are a number
of such remarks in the commentarial literature about the errors of the speech of Q and
nearby regions, and in each case, there appears to be independent evidence for recon-
structing the words involved with the coda*-r.
The rhyme evidence for reconstructing *-r in *r, and in words with as
phonetic element (Odes 40.1, 192.12, and 257.4), is open to more than one interpreta-
tion; some of the other words involved could be reconstructed with either *-r or *-n.
Although *-r clearly rhymes separately in parts of the Shjng, it may be that some
poems already reflect the change *-r > *-n, so some words are difficult to reconstruct
on the basis of rhyme evidence alone. But on the hypothesis that it is *-r and not *-n
that came to be pronounced as *-j in the Shndng area, Zhng Xuns comment clearly
indicates that Yn should be*-r.
Go Yu on Yn<*r
(1060)
Nowadays the people of Ynzhu all pronounce the family name
Yn [*r] as Y [*(r)j]. (X Wiy 2009:356)
266 Old Chinese
(1061)
......
At the four corners of old post stations, at a distance of 100 paces on
the four sides, pounded earth was built up; there was a building on top,
and on top of the building a pillar came out that was one zhng tall or
more; there were large boards passing through the pillar and coming out
on four sides. This is called a hunbio ... In Chn and Sng
, hun [hwan < *war < *[]ar] is vulgarly pronounced like
[h < hwa < *waj < *[]oj]; today they are still called hbio
.
The word hun < *[]ar is a clear case of *-ar, as shown by its xishng con-
nections (see (1032) above). Originally, h had the rhyme *-oj (it rhymes with
*to[r] > tsyhwe > chu blow (v.) in Ode 85.1), but by Hn times this would have
diphthongized to *-waj. By Hn times the initial voiced uvular stop *- had also
probably become a fricative like []. The point of R Chns remark is that people
in Chn and Sng pronounced *war (or *wan from earlier *war) like
*waj.
Chn and Sng were ancient states slightly south and west of modern
Shndng. Additional evidence that the change of *-r to *-j affected the Chn area is
this rhyme sequence from the Chn fng section of the Shjng, tradition-
ally regarded as containing songs collected from this region:
Although we have good evidence that yun had the rhyme *-ar, this is not true of the
other rhyme words above, which we reconstruct with *-aj:it appears that the rhymes
reflect a dialect in which *-ar has changed to*-aj.
(1063)
xun ringworm [*[s]ar] is x move [*[s]aj]; the stain moves
over a broader area each day; so in Qng and X , they pronounce
[*[s]ar] ringworm like [*[s]aj] move. (Ho Yxng etal.
1989:1101)
The idea of the sound gloss is that *[s]ar ringworm and *[s]aj move sound
similar because ringworm moves on theskin.
What interests us is Li Xs comment that in Qng and X , *[s]ar ring-
worm is pronounced like *[s]aj move, indicating that in those areas *-r had
become *-j. Qng was an administrative region in northern Shndng; X was to
its south along the coast, extending as far as the Yngz.
The map in Figure5.1 gives the approximate positions of the regions for which we
have explicit indications that final *-r became *-j. It includes the Shndng peninsula
and adjacent areas, slightly to the west andsouth.
Finally, we have at least one example of a geographical name in Shndng that shows
the change of *-r to *-j, probably reflecting a local pronunciation. It is Y < ngj+j,
the name of a mountain and river in Shndng; as we saw earlier (section 5.5.1.1), the
phonetic element clearly indicates *-r. Our reconstructionis:
(1066) *[]r > *[]j > ngj+j > Y (mountain and river in Shndng)
268 Old Chinese
Table 5.55 compares our reconstructions of *-aj, *-at(-s), *-an, and *-ar with earlier
reconstructions.
As noted above, Middle Chinese and most modern dialects reflect an innovation by
which *-aj lost its [j]coda, perhaps *Caj > *C > MC Ca, *Caj > *Ce > MC Cje. The
*-j is preserved, however, in Proto-Mn, and sporadically in some other southeastern
dialects. The *-j coda was also lost in *-waj from original*-oj.
Notes on Table5.56:
[1] Although in pngshng and qshng we have *Praj > Pae, we have no good
examples of this development in shngshng, and it appears that the regular develop-
ment in shngshng was *Praj > PeaX, as in these examples (see point [3] below for
further discussion):
(1067) *[b]raj > beaX > b stop,cease
(1068) *mraj > meaX > mibuy
In both (1067) and (1068), the graphic element on top is , depicting a net. The Shuwn
interprets both characters as semantic compounds, with rather forced explanations in
both cases (SWGL 3392b, 2769a). We suspect that the graphic element on top is actu-
ally a phonetic element, standing in these cases not for {} wng < mjangX < *ma
net but rather for {} lu < la < *raj a kind of net (on a handle?). For this reason,
although syllables with MC -eaX would usually be reconstructed as *Cre, we recon-
struct these words with *-aj. This reconstruction is further supported by the fact that
*[b]raj > beaX > b is the phonetic elementin
(1069) *praj > pje > p brown-and-whitebear
270 Old Chinese
which rhymes as *-aj in Odes 189.6, 189.7, and 261.6. (The same reasoning leads us to
reconstruct *praj in example (1069) rather than *paj, which is also a possible source of
MC -je; see below.)
[2]It is difficult to distinguish with confidence between *Tsraj and *Tsraj:based
on the general pattern, we would expect *Tsraj > Tsrae and *Tsraj > Tsrje > Tsrea. But
there is probably a good deal of dialect mixture in our sources:for ch, the Gungyn
has four readings:tsrhje, tsrhae, tsrhea, and tsrheaj.74
[3]After nonpharyngealized grave initials (types *K-, *K-, and *P-), *-aj and *-raj
have the same Middle Chinese reflexes and cannot be distinguished without additional
evidence. But we reconstruct *-r-in
because we suspect that as in examples (1067) through (1069) above, the element
on top stands not for {} wng < mjangX < *ma net, but rather for {} lu < la <
*raj a kind of net (on a handle?), which we take to be the phonetic element:
[4]Similar to the case with *-a and *-ak, we find both -je and -jae as reflexes of *-aj
after nonpharyngealized acute initials. We conjecture that a dialect feature is involved,
and we write *-Aj to represent a case of *-aj that for as yet unexplained reasons
becomes MC -jae instead of MC-je:
(1072) *[t-l ]aj > tsyheX > ch large; also read *[t-l ]Aj > tsyhaeX > ch
large
(1073) *C.lAj > zyae > sh snake; also read ye in *q(r)oj.laj >
'jwe.ye > wiy compliant, complaisant
(1074) *tsAj > tsjae > ji sigh;alas!
The traditional rhyme analysis puts both *-aj and *-oj in a single rhyme group,
G, but the rounded-vowel hypothesis predicts that the two rhymes should be sepa-
rate. Certain syllable types in this group, like MC Ka, Kje, Ta, or Tsye, can come only
from *-aj; while a syllable of the form Twa or Tsywe,for example, can come only
from*-oj:
To resolve such cases, we rely on rhyme evidence and graphic or etymological connec-
tions.For example, we can confidently reconstruct *-oj in h < hwa, and *-aj in
gu < kwa, on the basis of rhyme evidence:
Old Chineserhymes 271
Notes on Table5.57:
[1] is originally type B, having undergone the change Tsrj- > Tsr-; see section4.1.1.
272 Old Chinese
Notes on Table5.58:
[1] The character fi is usually used to represent {} fi < pjojH abandon, but in
early documents this word is written by = *[p.k]ap > pjop > f model, law, which
shows that it was originally *[p]ap-s and not *pat-s, even though it later came to be written
with the phonetic *C.pat > pjot > f, reflecting the change from *-p-s to *-t-s. But for
fi in the meaning great, we reconstruct *pat-s, because it occurs in this meaning in
Ode 204.4, and the Jngdin shwn gives it the pronunciations pjojH and pjot (JDSW83).
[2] The Middle Chinese reading of shi < sreajH < srjejH, like sh < sreat < srjet,
reflects the change Tsrj- >Tsr-.
Normally, the reflexes of *-t-s merge with the reflexes of original *-j-s. However,
because of the loss of the coda in *-aj, the reflexes of *-at-s remain separate from those
of *-aj-s; evidently the change of *-at-s to *-aj-s happened after the coda *-j was lost in
original *-aj, as shown by these examples:
It is because of this change that the finals -ajH, -aejH, -jejH, and -jojH occur only in
qshng in the Qiyn:there were no other finals in -jH for them to mergewith.
In the Mn dialects, however, *-aj did not lose its coda, so *-aj and *-at-s did merge
as pMn*-i:
(1081) *C.ta[t]-s > tajH > di girdle, strap; pMn *ti C belt:Fzhu
/tai 5/, Ximn /tua 5/
Old Chineserhymes 273
The traditional analysis combines our *-at(-s), *-ot(-s), and *-et(-s) in the rhyme
group Yu, but the rhyming distinction among them is established in Baxter (1992).
Certain Middle Chinese syllable types, such as Kat, Tat, and Kjot, can only reflect
*-at(-s), but others are ambiguous:MC pjot,for example, can reflect either *pat or *pot.
In fact we have a near-minimalpair:
Originally it appears that we have two distinct phonetic elements in the writing sys-
tem: for syllables like *Pat, and for syllables like *Pot; in received texts the two
have become confused to a degree,76 but the distinction is well supported by rhyme evi-
dence. The word f < *C.pat fly forth rhymes consistently with unambiguous *-at
words, in Odes 99.2, 149.1, 154.1, 202.5, 204.3, 260.3, and 304.2.77 The word *pot
root; hair (of head) and other words with the phonetic repeatedly rhyme as*-ot:
(1083) *pot > pjot > f root;78 hair (of head) (rhymes with unambiguous
*-ot words in Ode 225.2)
*b<r>ot > beat > b uproot (rhymes with unambiguous *-ot words
in Loz 54)
*bot-s > bajH > bi thinned out (forest) (rhymes with unambiguous
*-ot-s in Odes 237.8 and241.3)
For examples of how we distinguish *-et(-s) from *-at(-s), see section 5.5.3.2below.
Our reconstructions of *e before acute codas are compared with previous reconstruc-
tions in Table5.60.
Syllables of the form *-eT are largely parallel to each other in their development,
following the assumptions of the front-vowel hypothesis. However, the status of the
rhyme *-ej is somewhat unclear.
(1084) *(C).m[e]j > mejX > m millet or rice grains, dehusked and pol-
ished, phonetic in
*me[j] > mjieX > m achieve (as if from *me)
*mij > mej > m go astray (appears to rhyme with *-ij in Odes 191.3
and254.5)
There are other ways of writing {} m < mjieX achieve, however, so the matter may
be more complicated thanthis.
Notes on Table5.61:
[1] Although *r followed by *e after a pharyngealized initial usually produces
MC -ea-, we will see below that the regular reflex of *Kren is Kwaen, not Kwean.
We might expect the development of *Kret to be parallel, but we know of no good
examples. OC *Kret-s becomes KwaejH, not KweajH (see Table5.62).
[2] On the reconstruction *et > *xet > syet > sh set up, see Bi Ypng
[Baxter] (2010); this is the verb from which the noun *et-s > *xet-s > syejH > sh
circumstances, setting is derived by *s-suffixation. The conditions for palatalization of
velars in syllables like *Ket are not well understood.
276 Old Chinese
Notes on Table5.62:
[1] Since the Middle Chinese reflex of *Kren is Kean (see section 5.5.3.3 below),
by analogy we might expect *Kren > Kwean, *Kret > Kweat, and *Kret-s >
KweajH. But instead, we have *Kren > Kwaen and *Kret-s > KwaejH. (We know
of no examples of *Kret.) For example:
This is a phonetic element in several words with MC -wet that must be reconstructed
with*-et:
[2] The conditions for palatalization of velars in syllables like *Ket-s are not well
understood:we might expect *et-s > y cultivation, art, skill to palatalize to MC
nyejH, for example, but instead we have ngjiejH. (In fact there is no such syllable as
nyejH in the Gungyn.) But we do have palatalization in *et-s > *xet-s > syejH
> sh circumstances, setting.
Old Chineserhymes 277
(1086) Gu y 8:Yu y 7 8: 7
*pe[t]-s > pjiejH > b cover (v.)
*[ts]ret > tsrheat > ch examine
*et-s > ngjiejH > y cultivation, art,skill
Notes on Table5.63:
[1] As noted in section 5.5.3.2, *Kren becomes MC Kwaen, not Kwean as we
might expect by analogy to *Kren >Kean.
As we noted in section 5.4.3.3 above, just as *-i usually merges with *-in, we have
occasional examples of *-e merging with, or at least being confused with, *-en, as in
Loz 26, where received versions have the expression
(1087) *[N-q]re *C.qar-s > hjwaeng kwanH > rng gun; variously
translated as imperial palace, walls and watch-towers,etc.
(1088) *C.<r>en *ka[n] > hwaen kwan > hun gun (interpretation
unclear; Go Mng 1996:356)
Whatever the original state of the text, *-e and *-en have become confused, perhaps by
assimilation to the initial consonant of the following syllable (MC k- < *k; *C.q had
probably become *k- by Hn times).
5.5.3.4 *-er> *-en ( traditional Yun) or *-ej (> *-e, traditional Zh?)
Depending on the dialect represented, the Middle Chinese reflexes of *-er are either the
same as those of *-en (as in Table5.63) or the same as those of *-ej, which probably
coincided largely with those of *-e (as in Table5.36). Starostin believed that the coda
*-r did not occur after front vowels in Old Chinese; he argued that pre-Chinese *-er had
already become *-en (1989:341). But based on the same criteria we use for identifying
*-r elsewhere (alternation between the reflexes of OC *-j and *-n), the following seem
likely candidates for reconstruction with*-er:
(1089) *[k]er > ken ~ kej > jin crosspiece of wood on top of pillar
(1090) *pe[r] > penX > bin flat andthin
Starostin compared bin with Mizo (Lushai) pr flat and thin and assumed that *-r
had already changed to *-n after front vowels at the Old Chinese stage. But there is a
phrase shn pin yn good at glib talking that occurs in Shng sh:Qn sh
, but also occurs on strip 8 of the recenty discovered document
Kngz Sh ln (SB 1.20, 1.136), written as , where
pin has been replaced by p. The Middle Chinese readings are as follows:
(1091) pin < bjien insincere words, also bin < bjienX (in the Gungyn);
Jngdin shwn gives MC beanX, phjien, pjienX, phjienH (JDSW
52); all as if from some labial stop plus *-en <*-er?
p slander, MC phjieX, as if from *pe (? < *pej < *per)
So for bin flat and thin, the correct reconstruction may be as *per afterall.
We have another possible *-er word written with bin in the expression
(1092) *be[r]-se[r] > ben-sen > pinxin walk with difficulty (Shuwn)
Old Chineserhymes 279
would appear to have the vowel *e, but it has nn as phonetic, which indicates*-r:
Before acute codas, *o diphthongizes to *wa, probably at some time in the late Warring
States period, to judge from rhyme evidence. In the dialects represented in the Middle
Chinese sources, the *-waj from original *-oj then loses its coda *-j. Our reconstruction
of *o before acute codas is compared with earlier reconstructions in Table5.64.
Notes on Table5.65:
[1] By analogy to *-on and *-ot, we would expect that *-oj would occur after labial
initials:the *-oj would diphthongize to *-waj, and at some point *-w- ceases to be dis-
tinctive after labial initials. Thus, *Poj > *Pwaj > *Paj > Pa, merging with original
*Paj, but we have no clear examples of such syllables.
[2] With *k.roj > lwa > lu and *k.roj (dialect:> *kroj) > kwae > w snail,
we presumably have two different dialect treatments of the same etymon:Usually *k.r-
loses its preinitial and becomes MC l-, but in some dialects *k.r- merged with *kr- (see
section 4.4.4.4).
[3] In the western dialects where *l - > *x-, we have a special development:even
though the original vowel is back and rounded, *l oj is treated in the Middle Chinese
sources as a division-IV chngni syllable:
The status of *-oj as a rhyme separate from *-aj was established in Baxter (1992:413
422). Here is an example from outside the Shjng, not includedthere:
(Continued)
Old Chineserhymes 281
TABLE5.66 (Continued)
OC MC examples
*Crot Cwaet *trot > trwaet > chu iron point at end of whip
*Kot Kjwot *C.kot > kjwot > ju bracken (a kind of edible fern)
*Krot Kjwet [no clear examples]
*Pot Pjot *pot > pjot > f hair (of head)
*Prot Pjet [no clear examples]
*Tot Tywet *l ot > sywet > shu speak, explain
*Trot Trjwet *trot > trjwet > chu stitch
*Tsot Tsjwet *[dz]ot > dzjwet > ju cut off
*Tsrot Tsrjwet > Tsrweat *s-[k]rot > *tsrot > tsrjwet > tsrweat > zhu sprout (v.)
Notes on Table5.68:
[1] The reading srwaenH for twins is probably the result of the change Tsrj- >
Tsr- applied to *[s.r]on-s > srjwenH twins. (The Mandarin reading lun is prob-
ably based on other forms with the same phonetic element, such as *[m].ron > lwan
> lun harness bells.)
Recall from section 5.2.2 that recognizing *-on as a separate rhyme group makes it
possible to resolve a textual problem in Ode 106.3 (see Table5.19).
(1099) *[k]o[r] > *kwar > *kwaj > kwaX > gu fruit; result
*[k]or()-s > *kwar-s > *kwan-s > kwanH > gun pour out
libation
*[g]or > *gwar > *gwan > hwanX > hun turn around (as a
wheel); also read
*[g]<r>or > *grwar > *grwaj > hwaeX > hu turn around (as
a wheel)
*m-k<r>o[r] > *grwar > *grwaj > hwaeX > hu ankle
Old Chineserhymes 283
(1100) *k.ror > *k.rwar > *rwan > lwan egg; also read:
*k.ror > *k.rwar > *rwaj > lwaX > lu egg; cf. pMn *lhonB;
but Jinu has a form /lua 3/ egg of uncertain origin (Bijng dxu
2003:264).
(1101) *[s]or-s > *swar-s > *swan-s > swanH > sun calculate; tally;
cf. Japanese soroban abacus
(1102) *tor > *twar > *twan > twan > dun tip (n.)
*[d]or > *dwar > *dwan > dzywen > chun car with solid wheels
*[d]or-s >*dwar-s > *dwaj-s > dzyweH > ru jade tablet
*[t]or > *twar > *twan > tsyhwenX > chun topant
(1103) *tor > *twar > *twan > twanX > dun short; cf. pMn
*toi B short:Fzhu /ti 3/, Ximn /te3/.
In the Ch-Q dialects of southern Zhjing (Co Zhyn etal. 2000, Akitani
2003), where *-r normally goes to -n, OC *-or is reflected as Proto-Ch-Q *-oi in three
items, including the word short just discussed; see Table5.69.
Rhymes with * before acute codas provide the only cases where the same Old Chinese
rhyme has both division-I and division-IV reflexes in Middle Chinese; this is due to the
regular fronting of * between acute onsets and acute codas:
(1104) *[n]> 'on > n kindness, favor
*N-ts[n]-s > dzenH > jin grass,herb
However, dialect treatments of such rhymes probably differed, because there is one case
of *-n > MC -on after acute onsets, and also cases of *-n > -won as reflexes of OC *-n
or *-r (see discussion in section 5.5.1.1):
(1105) *l n > thon > tn to swallow, pMn *thun A(note also the Mandarin
pronunciation tn, which would normally reflect MC thwon).
This thon is the only case in the Gungyn where the final -on follows an acute initial;
perhaps onomatopoeia is a factor. The Gungyn also gives the reading then for when
it represents a surname, which would be the expected regular development from *l n.
(1106) *C.ts[r]-s > tsenH > jin grass, fodder; pMn *-tsun C straw
mattress,86 probably from the same root as
*N-ts[n]-s > dzenH > jin grass,herb
284 Old Chinese
MC -won as in dzwon normally reflects *-un or *-ur, but further evidence that the
vowel is * comes from this rhyme sequence in Ode93.1:
every one these sequences involves at least some words that the same dictionary assigns
to the Wi group; in half of them, all the rhyme words except ch are assigned to
the Wi group (35.2, 154.2, 162.1, and 300.1). The one apparent exception is 138.1, in
which ch rhymes with j hungry, which the dictionary also assigns to Zh. But
this too is an error: j hungry is transparently related to j famine, which everyone
agrees belongs to the Wigroup:
Once ch and j are correctly reconstructed with *-j (or conceivably *-[j]88), then
all eight of the sequences involving ch are regular.
We must also reconstruct *-j rather than *-ijin
Rhyme evidence clearly indicates that y had *-j,89 and it is well known that the
graph ancestral to sh was used in the early script to represent y:in oracle-bone
286 Old Chinese
texts, {} in the meaning foreigner (or perhaps as the name of a particular foreign
group) takes forms like these (J Xshng 2010:695):
(1111)
These evidently represent a person, seen from the side, with knees bent; at
times the graph is hardly distinguishable from that for rn person (J Xshng
2010:651):90
(1112)
(1113) *[q]ij > *xij> syijX > sh excrement; also read xjij < *[q]ij (with
irregular failure to palatalize)moan
Thus, with evidence from early documents, we can correct the reconstruction of sh
corpse to *l j, accounting for its early use to write {} y < yij < *lj foreigner.91
There do exist some irregular rhymes mixing *-i[j] and *-[j] in the Shjng, but
all the words in Table5.72 (except *tsr[]j > tsreaj > zhi purify oneself, which
does not rhyme) can be confidently reconstructed with *-j (or at least *-[j]). Correctly
making the distinction between *-j and *-ij is important for investigating ancient texts,
reconstructing dialect history, and establishing correspondences between Chinese and
languages that are genetically related toit.92
qshng only. Examples of *-t(-s) after acute initials are difficult to find, so we do not
include a separate table for them; the only relatively clear casesare:
those in Table5.75. However, the traditional analysis puts both *-n and *-un in the
single rhyme group Wn. In fact, the rhyming distinction between *-n and *-un is
unusually clear:the only rhyme that mixes them appears to be in Ode 248.5 (see Baxter
1992:425431).
5.5.5.4 *-r> *-n ( traditional Wn) or > *-j ( traditional Wi), according
to dialect
As after other vowels, *-r is reconstructed after * in words or phonetic series that seem
to combine the reflexes of *-n and *-j; see the discussion and examples in section
5.5.1.1.
Our reconstruction of the rhymes *-ij, *-it(-s), *-in, and *-ir are compared with previous
reconstructions in Table5.76.
These rhymes are noncontroversial for the most part, apart from two issues:(1)we
reconstruct *-ik and *-i in some words that have traditionally been included in the
Old Chineserhymes 289
Zhn rhyme group and reconstructed with *-it and *-in; and (2)we conjecture that there
was a rhyme *-ir, whose reflexes are sometimes like those of *-in and sometimes like
those of *-ij. The first issue was discussed in section 5.4.4 above; the second will be
discussed in section 5.5.6.4below.
Notes on Table5.77:
[1] While ji < keaj all, xi < keaj together, and ji < keaj steps, stairs
all rhyme as *-ij, ji < keaj and ji < heaj, both cold, rhyme as *-j. This is
290 Old Chinese
probably because and are relatively late characters, created at a time when criteria
for a xishng match had loosened:Gwnz gln gives no examples for either that are
earlier than the Shuwn (GG 2.84, 2.142).
[2] There appears to be palatalization of velars before *-ij unless blocked by a pre-
vocalic *-r-, in which case the Middle Chinese reflex of the rhyme is the division-III
chngni final -ij (see the discussion of velar palatalization in section 4.1.2). Depending
on the presence or absence of prevocalic *-r-, we have either division-III chngni -ij
(as in *[g]rij > gij > q old) or division-IV chngni -jij (as in *C.pij > pjijX >
b compare).
Notes on Table5.78:
[1] We would expect *Kit > Tsyit as a theoretical possibility, but we know of no
examples; nor can we explain why *C.qi[t] > kjit > j auspicious fails to palatalize
(unless it has something to do with the fact that the velar k- here comes from the prefixed
uvular *C.q-).
[2] Note that we have both division-IV and division-III syllables in contrast from
this rhyme: *mit > mjit > m honey vs. *mri[t] > mit > m dense.
Old Chineserhymes 291
Although in some cases we have positive evidence for reconstructing *-ik instead of
*-it, the absence of such evidence is not sufficient reason to reconstruct *-it, so we often
write *-i[t](-s).
We reconstruct *-it-s in words that have Middle Chinese reflexes like *-ij-s, but have
etymological or graphical connections with words in*-it:
Notes on Table5.79:
[1] As with *-ij and *-it(-s), depending on the absence or presence of *-r-, we
can get as Middle Chinese reflexes either division-IV chngni -j(w)in or division-III
chngni-(w)in:
(1122) *[N-q]i[n]> ywin > yn even, uniform
*[]ri[n] > hwin > yn rind of bamboo
Notice that here, too, prevocalic *-r- blocks the palatalization of *-, and the resulting
initial is MC hj- insteadofy-.
292 Old Chinese
(1123) *[b]ir > *bin > bjinX > pn female of animals; also read *[b]ir
> *bij >bjijX
(1125) :
[On] :[pronounced] b(jin) + (ny)inX=bjinX; formerly it was said
to be b(ju) + (p)jijX=bjijX; [the Emperor] Jinwn 98 has b(ju)
+ (k)jinX=bjinX. (JDSW356)
So if we are right about the coda *-r, and if we knew when and where this part of the
Loz was composed, it could give us an additional clue about the geographical distribu-
tion of the Old Chinese dialect where *-r became*-j.
Table5.80 summarizes our reconstructions of rhymes with *u before acute codas, and
compares them with previous reconstructions.
Notes on Table5.81:
[1] We seem to have no examples of *-uj after labial (or labialized) initials; if such
syllables did exist, they may have been affected by an early dissimilation of *u in this
environment.
We reconstruct *-w and *-wk after three of the six Old Chinese vowels:*a, *e, and *i. If
these codas appeared after other vowels, we have found no trace of them. Rhymes with
the coda *-wk are parallel in many ways to those with *-w; alternatively, *-wk could be
analyzed as a labiovelar *-k. It is not uncommon for the same phonetic elements to be
used for forms with *-w and *-wk; for example:
It is more common for syllables with vocalic codas and syllables with stop codas to be
kept separate in the writing system, but the contacts between *-w and *-wk may simply
reflect the fact that the coda *-wk is relatively infrequent, so there were fewer appropri-
ate phonetic elements available in the script (see section 3.4). Note also that there is no
parallel nasal coda *-w.
Our reconstructions of *-aw and *-awk are compared with previous reconstructions in
Table5.84.
Notes on Table5.85:
[1] The Gungyn has no syllables of the form Tsrjew. Probably there were
type-B syllables of the form *Tsraw that became first Tsrjew and then Tsraew because
of the tendency of MC -j- to be lost after Tsr-; consequently, we have no good way of
distinguishing between original *Tsraw and original *Tsraw.
[2]In type-B syllables with grave initials, *-aw and *-raw evidently merged, so the
presence or absence of *-r- cannot be detected from Middle Chinese readingsalone.
[3] In order to account for the syllable hjew in chxio < tsyhij.hjew owl,
we must assume that there were Old Chinese syllables like *K(r)aw. Initial MC hj- can
normally reflect only *[]-, which would have lost its labialization through dissimila-
tion from the coda *-w. Other cases of *K()- before *-aw may have occurred, but we
know of no way of identifyingthem.
Old Chineserhymes 297
otherwise:
*[]rawk > ngaewk > yu music
*Crawk Caewk
*trawk > traewk > zhu high; splendid
*Tsrawk Tsrjak > Tsraewk? [2] [no clear examples]
otherwise:
*[](r)awk > ngjak > n cruel
*C(r)awk Cjak [3]
*tawk > tsyhak > chu indulgent, gentle
Notes on Table5.86:
[1] The reflexes of *Cawk are very irregular and unpredictable; they must have
varied from dialect to dialect. Mostly we have MC Cak and Cowk; in a few cases we
haveCuwk.
[2] By analogy to other rhymes, we would expect the developments *Tsrawk >
Tsraewk and *Tsrawk > Tsrjak > Tsraewk, but we have no good examples.
[3] In type-B syllables with initials of other types, we have *C(r)awk > Cjak; the
presence of *-r- cannot be detected in grave-initial syllables.
Our reconstructions of *-ew and *-ewk(-s) are compared with previous reconstructions
in Table5.87.
298 Old Chinese
Notes on Table5.88:
[1] In general it is not possible to distinguish the syllable types *Tsrew and
*Tsrew, because of the change of Tsrj- > Tsr-:*Tsrew > Tsrjew > Tsraew. In one case
the Jngdin shwn gives two readings for sho rations:sraewH and an older
reading, srjewH:
[2] From *Kew and *Pew we get division-IV chngni syllables Kjiew and
Pjiewunless the velar initial palatalizes, which is somewhat unpredictable. We
have palatalization in *[ ]ew > *xew > syew > sho burn, but the glottal stop is
not subject to palatalization, so we get *ew > 'jiew > yo waist.
[3] From syllables like *Krew and *Prew we would expect division-III chngni
syllables Kjew and Pjew, but we know of no good examples of this development.
Notes on Table5.89:
[1] In syllables like *Cewk, *-ewk loses the rounding in the coda and becomes MC
-ek, merging with original *Cek and *Ciwk.
[2] The reflexes of *Tsrewk and *Tsrewk appear to merge as Tsraewk and are thus
indistinguishable from Middle Chinese readingsalone.
[3] Otherwise, *-ewk in type-B syllables becomes -jak; the presence or absence of
prevocalic *-r- after grave initials cannot be detected from Middle Chinese.
Notice in (1139) that the reading 'jiewH has a division-IV chngni final and thus tells
us that the vowelis*e.
Our reconstructions of *-iw and *-iwk(-s) are compared with previous reconstructions
in Table5.90.
Notes on Table5.91:
[1] OC *-iw (after nonpharyngealized initials) is the source of the Middle Chinese
final -jiw ( rhyme), which, with a few exceptions, is limited to syllables with grave
initials. The final -jiw is like a division-IV chngni final in the sense that although it
is from a type-B syllable, it is placed in division IV of the rhyme tables (which is why
we write it as -jiw in our Middle Chinese notation, with both -j- and -i-). The usual
pattern is that where there is a division-IV chngni final, there will be a correspond-
ing division-III chngni final reflecting the same Old Chinese rhyme, preceded by
*-r-. But the graphic evidence indicates that *-riw after grave initials still produced the
division-IV final-jiw:
(1143) *[r]iw-s > ljuwH > li flying high (no pre-Hn text); phonetic in
*m-riw-s > mjiwH > mi lie, error
*k-riw > (dial. > *kriw) > kjiw > ji twist(v.)
[2] The regular type-B reflex of *-iw after acute initials is -juw, as in *tsiw >
tshjuw > qi autumn; crop, implying that at some point there was a change:*-iw >
-juw. But it is likely that some dialects in the Middle Chinese period did not undergo
this change, retaining a distinctive final -iw after acute initials also. The few
acute-initial words with -jiw in the Gungyn probably reflect such a dialect, and the
Jngdin shwn also records some readings like this (for example, li < ljuw <
*[r]iw fine gold spelled as ljiw, JDSW 85). Studies of rhymes from the time of the
Qiyn or before might make it possible to locate this dialect feature geographically.
Notes on Table5.92:
[1] As with *-ewk, in syllables like *Ciwk, *-iwk loses the rounding in the coda
and merges with*-ek.
[2]By analogy to other rhymes, we would expect *Criwk > Caewk, but have found
no clear examples ofthis.
302 Old Chinese
Words with labial codas are significantly less frequent in the early Chinese lexicon than
those with velar or dental codas:as a rough estimate, in our database of lexical items,
about 8 percent have (Middle Chinese) labial codas -p or -m, 24 percent have velar
codas -k or -ng, and 23percent have dental codas -t or -n.108 This would have made it
especially difficult to find appropriate words with labial codas to use as rhymes; and it
also would have made it more difficult to find suitable phonetic elements to write words
with labialcodas.
We can illustrate this point by considering the difficulty of finding a rhyme for
the English word orange. It is often said that no English word rhymes with orange;
in fact there is one obscure word that rhymes with it:sporange, a botanical term
equivalent to sporangium a receptacle containing spores; a spore-case or capsule.109
Now orange has a certain frequency in ordinary text, but its frequency in a corpus of
rhyme words must be much lower, virtually zero, in fact; unless the word sporange
is appropriate to the context, a poet must either relax the usual criteria for rhyming,
or not use orange as a rhyme at all. If there were a few more words that rhymed with
orange, its frequency as a rhyme word might go up, but would still be quite low
compared with its frequency in ordinary text. Similarly, if one were trying to find a
phonetic element to write orange in a Chinese-like script for English, one would face
analogous problems.
The same argument applies to words with labial codas in Old Chinese:rhyme
sequences with labial codas are rather infrequent in the Shjng, and recourse to faute-de-
mieux phonetics (see section 3.4) appears to have been more frequent than with other
types of syllables. For example, it seems to be more common with labial codas for the
same phonetic element to be used for both *-p and*-m.
Thus both rhymes and xishng connections involving labial codas tend to be less
regular than with other syllable types. Since the two main kinds of evidence for recon-
structing Old Chinese rhymesrhymes and phonetic compoundsare less helpful than
with other kinds of syllables, the reconstruction of vowels in individual words with
Old Chineserhymes 303
labial codas is sometimes underdetermined by the evidence, and we often have to use
square brackets to indicate our uncertainty about the vowels.
However, we have good reason to believe that all six Old Chinese vowels did origi-
nally occur before labial codas. The traditional analysis recognized only two rhyme
groups with the coda *-p ( Y and Q) and two with the coda *-m ( Tn and
Qn). These have generally been reconstructed with *a and *, respectively, as in
Tables5.93 and5.94.
However, the hypotheses of the six-vowel reconstruction require us to also
reconstruct front *e and *i before the labial codas *-p and *-m. According to the
front-vowel hypothesis, Middle Chinese division-IV finals and division-IV chngni
finals can only come from Old Chinese rhymes with front vowels.110 There are six
such finals with labial codas in Middle Chinese:-ep, -em, -jiep, -jiem, -jip, and
-jim. In addition, according to the traditional analysis, MC -em and -ep each have
two sources:MC -em can come from either the Tn group or the Qn group,
and MC -ep can come from either the Y group or the Q group, as shown in
Table5.95.
Since our system does not allow combinations like *-ia- or *-i-, we reconstruct front
vowels *e and *i before labial codas to account for examples like the following.
Our reconstructions of *-ep, *-em, *-ip, and *-im are compared with other reconstruc-
tions in Tables5.96 and5.97.
As Tables5.96 and 5.97 show, the rhymes we reconstruct with *-ep, *-em, *-ip, and
*-im have traditionally been reconstructed in such a way as to agree with the traditional
categories:thus both Karlgren and Li reconstructed *-iap, *-iam, *-ip, and *-im, so
they appear to rhyme with *-ap, *-am, *-p, and *-m, respectively.
As a matter of fact, it is difficult to confirm a clear rhyming distinction in the Shjng
between front and nonfront vowels before labial codas:although Baxter (1992) confirmed
the rhyme distinctions predicted by the six-vowel system in syllables with other codas,
the rhyme sequences involving words with labial codas were too few to allow statistically
significant results, and in any case a number of such rhyme sequences are clearly irregular.
In fact, there are no rhyme sequences involving words we reconstruct with *-ep, and only
one ambiguous sequence involving a word with *-em.111
However, the Shjng does have two rhymes that can be confidently reconstructed
with*-im:
(1153) Ode162.5:
*[ts]r[i]m > tsrhim > qn gallop (also read MC tshim, reflecting a
Middle Chinese dialect in which tsr- became ts-, tsrh- became tsh-, etc.)
*n im > syimX > shn remonstrate
(1154) Ode189.6:
*[l]im > demX > din bamboo mat
*[ts]im > tshimX > qn sleep
In (1153), we can reconstruct *n im on the basis of its phonetic element nin <
*nim-s, and in (1154), din < demX must be reconstructed with *-im to account for
the Middle Chinese division-IV final -em. The phonetic element seems to indicate
OC *-im, as we see in both sequences.
There is also a likely example of a rhyme in*-ip:
(1155) Ode5.3:
*s.qrip > tsrip, (dial.) tsip > j cluster together (also read *qip > 'jip
> y bow (v.), salute, showing the *i vowel)
*[d]rip > drip > zh hibernate; cluster
Here j < tsip has the alternative reading 'jip with a division-IV chngni final, which
can only be reconstructed with *-ip; and we reconstruct *-ip in zh < drip because of
division-IV words written with this phonetic:
We have now established that four of the six Old Chinese vowels occur before labial
codas. What about the rounded vowels *o and *u? There is indirect but convincing
evidence that these vowels also occurred before labial codas. Dng Tngh (1948:108
112) identified several phonetic series that resemble (our) *-p and *-m in that they
have division-I reflexes in MC -op or -om, but resemble our *-ap and *-am in that they
have division-III reflexes in MC -jep, -jaep, -jop, or -jem. He included these words in
the traditional Y and Tn groups (with our *-ap, *-ep and *-am, *-em), but he
reconstructed them with a distinctive vowel *. Dng Tnghs *-p and *-m generally
correspond to our *-op and *-om.112
The development of words with final *-p-s often gives clues to the identity of the pre-
ceding vowel.For example, our system predicts that *-op-s should develop as in (1157).
When words in MC -wajH or -jwejH appear to come from the same root as words with
final *-p, we reconstruct them with *-op-s and reconstruct *-op in the unsuffixed root.
An exampleis
As for *-up and *-um, these normally merge with *-p and *-m, so they are also
difficult to identify with confidence. But as with *-op-s, forms in *-p-s give us some
evidence about the main vowel. According to our system, we would expect *-up-s to
develop as in (1159):
When words with these Middle Chinese finals have connections with words in final *-p,
we reconstruct them with *-up-s, and the root with *-up.For example:
Another clue to main vowels before labials is the fact that there was evidently a dia-
lect of Old Chinese in which final labials *-p and *-m simply changed unconditionally
to *-k and *-. Part of the evidence for this is rhyme sequences that mix velar and labial
codas; since vowels are easy to identify before velar codas, these rhymes provide evi-
dence about the vowels in the words that had labial codas. Anumber of Shjng rhymes
seem to show this dialect feature.For example, in Ode 128, we have two different rhyme
sequences of thistype:
(1161) Ode128.2:
*tru > trjuwng > zhng center
*m-srum > tshom > cn team of three horses
Old Chineserhymes 307
(1162) Ode128.3:
*[q](r) > 'ing > yng breast(plate)
*k > kjuwng > gng bow (n.)
*l > dong > tng tie; band
*q(r) > xing > xng lift, rise
*[q](r)m > 'im > yn sound,tone
Here we reconstruct cn < tshom with *-um because of its rhyme with *-u in
128.2, and we reconstruct yn < 'im with *-m because of its rhyme with *-
in128.3.
We conjecture that the unconditional change of final *-p and *-m to *-k and *-
was a western dialect feature; the poems in which such mixed rhymes occur seem
to have mostly a Western Zhu background. We have at least one explicit comment
reflecting this, from Yn sh ji xn by Yn Zhtu, one of the Qiyn
authors:
Since the relevant items with clear reconstructions are comparatively few, we will
treat all the rhymes with high vowels together, then the rhymes with nonhigh vowels.
The Middle Chinese reflexes of rhymes with high vowels before *-p are summarized
in Table5.98.
We reconstruct *-p-s, *-up-s, and *-ip-s in words with etymological or graphical
connections to *-p, *-up, and *-ip, respectively; as a result of the early change of *-p-s
to *-t-s, these suffixed rhymes are predicted to have reflexes like those of *-t-s, *-ut-s,
and *-it-s. Because *-p-s has the same reflexes as *-t-s, a significant number of words
with *-p-s have gone unrecognized, and have previously been reconstructed as if they
had *-t-s (Karlgrens *-d, Lis*-dh).
Notes on Table5.98:
[1] We reconstruct * in , *m-rp > dop > t reach to; and because of the
related forms with *-s suffix:
These words are frequently glossed the same and seem to be related, even if the mor-
phology is not entirely clear. The reading dejH for *m.rp-s, repeatedly given as
an alternate pronunciation in the Jngdin shwn (e.g., JDSW 97, on Ode 257.6), is
most easily explained as resulting from sound changes happening in different orders
in different dialects. In a syllable like *m.rt-s, we would expect the * to be fronted
because it occurs between an acute onset and an acute coda, producing dejH. The read-
ing dojH reflects either a dialect in which this fronting did not occur at all, or a dialect
in which *-p-s changed to *-t-s after the fronting of original *-t-s, so that it escaped
this change.
[2]We reconstruct *u in *n[u]p > nop > n bring or send in and *n[u]p >
nyip > r enter because of the related words with the *-s suffix:116
Here are two more such families of words where we can reconstruct *-[u]p(-s):
[3]For reasons that are not fully understood, the Middle Chinese division-IV
chngni finals -jip, -jim, -jiep, and -jiem occur only with the Middle Chinese initial
'- ( Yng). It could be that the relevant examples with velar initials palatalized:*Kip
> Tsyip etc., although we have no clear examples of this. These division-IV chngni
finals do not occur after labial initials atall.
[4]By analogy to cases like *C.[](r)m > hjuwng > xing bear (n.) and
*prm > pjuwng > fng wind (n.), we would expect that forms like Kjuwk and
Pjuwk could come from *-up, but we know of no such examples.
[5] We reconstruct *i in *[t]ip > tsyip > zh seize because this phonetic element
is used to write division-IV words such as *[t]ip > tep > di (place in Schun)
(see above); the qshng derivatives also develop as if from *-it-s. Finally, the word
*[d]rip > drip > zh hibernate; cluster rhymes as *-ip in Ode 5.3 (example (1155)
above).
The Middle Chinese reflexes of *-m, *-um, and *-im are summarized in
Table5.99.
310 Old Chinese
Notes on Table5.99:
[1] We reconstruct * in *[q](r)m > im > yn sound, tone because it rhymes
with *- in Ode 128.3 (see (1162) above). But although it is a homonym of yn in
Middle Chinese, we reconstruct *q(r)um > 'im > yn dark with *u because it rhymes
with *-u in Ode154.8:
And even though it is written with yn < *[q](r)m as phonetic, we also reconstruct
*-um in *qum-s > 'omH > n dark, because of its likely etymological connection to
*q(r)um. According to Gwnz gln (GG 6.411), the character n is not attested
before Hn times, which makes it too late for the phonetic to be diagnostic for the dis-
tinction between *-m and*-um.
[2] As noted above, *[l]im > demX > din bamboo mat must be reconstructed
with *-im to account for the MC final -em; as seen above (1154), it rhymes in Ode 189.6
with *[ts]im > tshimX > qn sleep, which has the same phonetic as *[ts]r[i]m >
tsrhim > qn gallop, which in turn rhymes with a character whose phonetic is
*nim-s > nemH > nin think of (1153).
[3] As with -jip < *-ip, MC -jim < *-im occurs only with the Middle Chinese initial'-.
[4] Bear and wind, which the traditional analysis recognizes as part of the tra-
ditional Qn group, are to be reconstructed with *-m: we assume that *-m was
rounded to *-um under the influence of the labial or labialized initial, then the final *-m
dissimilated to *-. The graphic connections of *prm > pjuwng > fng wind (n.)
suggest that it had *-r- before the vowel; but at the time *-r- disappeared, the vowel was
Old Chineserhymes 311
already rounded and thus *-r- did not leave a trace on the final (as with original *-ru;
see section 5.4.6.3).
The Middle Chinese reflexes of Old Chinese rhymes with nonhigh vowels before *-p
are summarized in Table5.100.
We reconstruct *-ap-s, *-ep-s, and *-op-s in words that have reflexes like *-at-s, *-et-s,
and *-ot-s, but which have etymological or graphic connections with words in*-p.
Notes on Table5.100:
[1] Generally, MC -ap reflects *Cap and -aep reflects *Crap. We reconstruct
*-ap(-s)in
Note that although the phonetic *[ts]et-s > tsjejH > j sacrifice has *-et-s and prob-
ably represents *-et in *[ts]ret > tsrheat > ch examine, the character j is of
312 Old Chinese
late origin and cannot be relied on to represent either the early vowel or the early coda
(it is not attested in pre-Hn documents according to GG 10.840).
[2] We reconstruct *o in h < hop based on the qshng derivative hu <
hwajH and others:
(1172) *m-kop > hop > h come together; bring together
*N-k<r>[o]p > heap > qi accord with
*m-kop-s > hwajH > hu meeting; have a meeting
*kop-s > kwajH > kui calculate; calculation
*kop-s > kwajH > gu place where two ends of a collar or beltjoin
[3] Generally, MC -jop occurs only after P-type initials, and -jaep only after K-type
initials. Because of their multiple possible origins, it is often difficult to be confident
about which vowel to reconstruct inthem.
[4] Parallel to the other syllables with labial codas, the only Middle Chinese syllable
with the division-IV chngni final -jiep is'jiep.
[5] When labialized initials occurred before rhymes with labial codas, they are usu-
ally difficult to detect because they would have lost their labialization by dissimilation.
But reconstructing *[](r)[a]p > hjep > y shine, gleam is the only way in our sys-
tem to account for the MC initial hj- ( Ysn= Yn); normally MC hj- occurs
only in hku syllables (i.e., accompanied by -w- in our notation); here the -w- has been
lost by dissimilation.
[6] It has long been a problem how to explain the presence (from the earliest
date) of the character q in ~ . Some problems remain, but here is the sce-
nario we suggest:(1)The character q was earliest used for syllables of the form
*Kap (e.g., *k(r)ap > kjaep > ji rob). (2)The word written by ~ has a
labial preinitial: *[p.k]ap > pjop > f model, law. (3)The reading q < khjoH
could represent either the western dialect where *-p in coda position changes to *-k
(*[k](r)ap-s > *-ak-s > *-a-s > khjoH); or the forms with MC -jo could originate in
the restressing of unstressed variants that had reduced or lost the final stop, parallel to
the loss of final *-k in *m.rk > *rk > *r (unstressed form) > loj > li come
(see Baxter 1992:330 and section 5.4.2.2); the MC reading khjoX,for example,
could have evolved from *[k](r)ap > *ka (unstressed form). (4)The word {}
*[p-k]ap-s > *-at-s > pjojH > fi cast aside (written with in early documents)
may somehow be related to the root of *[k](r)ap-s > khjoH > q depart.
[7]The element *[b](r)[o]p > bjop > f lack (v.) seems to be kept distinct from
q as a phonetic element, which would be explained if it had *-op instead of *-ap.
The following example with *-o (a dialect development from *-om) also supports *o
in words with this phonetic element:
(1173) *p(r)om (dial. >) *p(r)o > pjowngX > fng overturn (no pre-Hn
exx.).
[8] As noted above, we reconstruct *-op in *[ts]<r>op > tsrheap > ch pestle
because of the related qshng form with -jwejH:
Notes on Table5.101:
[1] We reconstruct *-om in gn < komH because of its contacts with *-o, pre-
sumably reflecting a dialect that changed final labials to velars:
In its current form, ~ has the phonetic gng < kuwng < *ko officer, but
this is a late addition due in part to graphic confusion; the earliest form of the character
consists of zhng (representing zhng (jade) insignium) and j (a person with
314 Old Chinese
both arms extended), as in this form (from the Western Zhu vessel Gng Yng dng
):117
(1177)
Chn Jin (2007) describes in detail how later forms of the character came to be written
with the phonetic *[k](r)om-s > khjomH > qin yawn and eventually with gng
< kuwng < *ko officer, but the forms with gng are quite late. The association of
~ present (v.) with the ceremonies of the Western Zhu court may have some-
thing to do with its pronunciation as kuwngH < *[k]o-s, with the substitution of *-
for original *-m that appears to be a western dialect feature. The preservation of *-m in
*[k]om-s > komH > gn, the name of the river in Jingx, may be a clue to the region
where *-om was preserved.
[2] qin < khjomH yawn, xin < heamH small pit, and yn < 'jemX
cover are all among the phonetic elements Dng Tngh reconstructed with *-m
because of their connections to both MC -om and MC -jem (see the discussion above,
and Dng Tngh 1948:108109). The words in (1178) are probably all related:
[3] Apart from a few problematic cases, the division-IV chngni final -jiem basi-
cally occurs only with the initial '- ( Yng), as with the other division-IV chngni
finals with labialcodas.
[4] In spite of its phonetic element *[k][a]m > kam > gn sweet, qin < gjem
pincers may have *-em if it is historically related to *m-kep > hep > xi grasp
and other relatedforms:
[7] The phonetic element zhn < *tem can be taken as a reliable indicator that the
rhyme is either *-em or*-ep:
Conclusion
It would be nave to attempt to discuss the question of what kind of language Old
Chinese was without considering the historical context in which the question has
been asked. In the European intellectual world, both China itself and the nature of the
Chinese language have been topics of intense interest for several centuries. The per-
sistence of certain Chinese institutions over several thousand years led some European
observers to conclude that unlike the part of the globe they themselves inhabited, China
had changed very little. In his Philosophy of History, Hegel wrote that, for reasons that
seem rather abstract, both China and India were outside the Worlds history:
With the Empire of China History has to begin, for it is the oldest, as far as his-
tory gives us any information; and its principle has such substantiality, that for the
empire in question it is at once the oldest and the newest. Early do we see China
advancing to the condition in which it is found at this day; for as the contrast
between objective existence and subjective freedom of movement in it, is still
wanting, every change is excluded, and the fixedness of a character which recurs
perpetually, takes the place of what we should call the truly historical. China and
India lie, as it were, still outside the Worlds History, as the mere presupposition
of elements whose combination must be waited for to constitute their vital prog-
ress. (Hegel 1899:116; emphasis in the original)
Chinese is often said to be the human language that has been in continuous use
for the longest time. Of course, if we are referring to spoken languages, this is largely
a matter of terminology:we use different names for Vedic Sanskrit and Hindi, even
though their relationship is comparable to the relationship between Old Chinese and
modern standard Chinese. But as far as scripts are concerned, the Chinese script prob-
ably is the oldest one in continuous use, having been in existence since at least the
thirteenth century bce (thus predating the Phoenician alphabet); and early European
observers had no reason to believe that it had changed much over the previous two or
three millennia either. Given the common failure to distinguish written language from
spoken language, it was easy to draw the conclusion that Old Chinese was just like
modern Chinese, onlyolder.
Conclusion317
Ideas about the early Chinese language have also been affected by
nineteenth-century European views about the nature of language differences. When
modern linguistics was beginning to develop in the nineteenth century, it was com-
monly assumed that human languages would fall into a small number of radically
different structural types, just as the then newly fashionable science of comparative
anatomy divided animals into vertebrata, mollusca, articulata, and radiata on the
basis of their internal structure (see Baxter 2002 for discussion of this connection).
The attempt to classify languages in an analogous manner focused on the ways mor-
phemes were combined into larger expressions. There were many competing pro-
posals, but August Wilhelm von Schlegel (1818:14),for example, divided languages
into three types:(1)those with no grammar at all (such as Chinese); (2)those that
add affixes mechanically to roots, without modification; and (3)inflected languages,
whose elements are combined organically, undergoing internal changes, rather than
mechanically. The Indo-European languages were assigned to this third group, which
(unsurprisingly) was considered the most advanced. Different language families were
thought to be associated with different types of linguistic structure (like the differ-
ences in structure among different groups in the animal kingdom), which in turn were
often assumed to correspond to the specific mentalities of the nations or races that
usedthem.
Furthermore, in the nineteenth century it was still widely believed that the Earth
was no more than a few thousand years old; so it seemed plausible that language had
evolved rather recently, and perhaps differently in different races. Attempts were
made to assess the degree of evolutionary development of particular languages, and
it was common to imagine that Chinese, whatever its other virtues might be, had a
primitive structure, close to what the earliest stages of human language might have
been like. Since it was believed to have the monosyllabic structure that was charac-
teristic of the earliest forms of human speech, it followed that it must have changed
very little over time. Thus William Dwight Whitney could still write, in 1867 (with
curious echoes of Hegel):
Portions of the Chinese literature, it is true, are nearly or quite as old as anything
Indo-European, and the Chinese language...is in some respects more primitive
in its structure than any other human tongue; but what it was at the beginning,
that it has ever since remained, a solitary example of a language almost destitute
of a history. (1867:233234)
One tongue, the Chinese...has never advanced out of its primitive monosyllabic
stage; its words remain even to the present day simple radical syllables, closely
resembling the Indo-European roots, formless, not in themselves parts of speech,
but made such only by their combination into sentences....Yet this scanty and
crippled language has served all the needs of a highly cultivated and literary peo-
ple for thousands of years. (1867:257)
Not very much requires to be said in explanation of the structure and history of
a language so simplea language which might be said to have no grammatical
318 Old Chinese
structure, which possesses neither inflections nor parts of speech, and which has
changed less in four thousand years than most others in four hundred, or than
many another in a single century. (1867:334)
It is unusual nowadays to hear these ideas expressed so clearly, but one still encounters
the idea that Chinese is somehow deeply different from European languages in a way that
is intimately connected with the differences between Eastern and Western thought.
Although the influence of these ideas from nineteenth-century linguistics can still be
detected in sinological discourse, the situation within linguistics has changed. Modern
linguists generally regard the human language faculty as a biological adaptation of our
species, which does not vary significantly from one human group to another. We now
know that the different morphological patterns that were once thought to define essen-
tially different types of languages can change dramatically over time, even within a
single language family.
Consequently, apart from what may be common to all human languages, the members
of a language family do not necessarily have anything in common other than their com-
mon ancestor, and typological similarity is no longer regarded as evidence of a family
relationship. Thus there is no such thing as Indo-European or Sino-Tibetan language
structure.1 The Sino-Tibetan family, once thought to be characterized by tonality and
isolating structure, includes both tonal and nontonal languages, and languages with
and without elaborate inflection. Within Indo-European, there are also tonal languages,
and English itself has lost most of its inflections and moved in the direction of a mor-
phological structure more like that of modern Chinese. Consequently, we have no reason
to assume that any ancient language was typologically similar to the modern languages
descended fromit.
Moreover, human language as a biological adaptation is now believed to be at
least tens of thousands of years old, not a recent cultural invention, as was widely
believed in the nineteenth century. So there is no reason to think that the ancient
languages we can reconstruct were essentially different from or more primitive than
modern languages in any structural sense or that the kinds of changes that affected
them were significantly different from the changes that affect languagestoday.
With these considerations in mind, how can we characterize Old Chinese? We can
begin by saying that it had almost none of the features that are typically thought of as
characteristic of Chinese.
It was not tonal. Tones developed after the Old Chinese period, when consonants
were lost and the accompanying pitch differences became phonologically distinc-
tive:loss of final glottal stop produced shngshng, and the loss of final -[h](from ear-
lier *-s) produced qshng. There are in fact dialects in which the consonantal elements
[] and [h] are still present (e.g., Xioy , in Shnx province; see Sagart 1999b:132
and Gu Jinrng1989).
It was not monosyllabic. Evidence from modern dialects and early loans requires
us to reconstruct word-initial minor syllables in many cases. Tightly attached preinitial
consonants must be reconstructed to account for cases like (1181):
Conclusion319
(1181)
*k.dzra > dzrjang > chung bed, pMn *dzh-; Proto-Vietic
*k- bed, Cht [Sch, Rc] /kc/, Maleng [Br] /kac/;
Maleng [Kha Pong] kc, Vietnamese ging [zA2]
Loosely attached presyllables are reconstructed to account for cases where both Mn and
Vietnamese show intervocalic lenition, as in (1182):
(1182)
*k.dok-s > duwH > du neck, pMn *-d-; Proto-Vietic *k-k
nape of the neck, Rc kadk, VN dc [zk D2] stem of a plant
orpipe
In some cases, the presyllabic material may be a synchronic prefix in Old Chinese
or may have been a prefix at some earlier stage of the language, but we have no
reason to doubt that in some cases it could be part of the root. One possible explana-
tion for the existence of such complex initial consonant clusters is that they could
arise through the reduction of the first syllable of what was originally a full disyl-
labic word. An example of such a process is found in Moroccan Arabic, where origi-
nally disyllabic verb forms have lost their first vowel, producing complex clusters
(Harrell1966):
Although rather unlike that of modern Chinese, the word structure we recon-
struct for Old Chinese is similar to that found in the native vocabulary of modern
Khmer (excluding polysyllabic loanwords): both monosyllabic words and words
with minor presyllables, and considerable derivational morphology. (Of course, this
does not imply that Chinese and Khmer are genetically related, for the reasons out-
lined above.)
It is to be expected that there was considerable dialect diversity during the Old Chinese
period. We have made some progress in identifying a number of Old Chinese dialect
features and locating them geographically:
1. The coda *-r became *-j in dialects in and near the Shndng peninsula (see
section 5.5.1.4). Traces of this development can be seen in the Ch-Q and Mn
dialects, which in a number of words have final [i] corresponding toMC-n.
320 Old Chinese
2. Initial *l ()- and *n()- changed to [t] in coastal areas but into [x]or [h] in areas
more to the west; there was a similar split development of *r ()- (see section
4.3.5).
3. Treatment of presyllables varied significantly from dialect to dialect. The Mn
dialects appear to reflect a variety where preinitial obstruents were lost before
obstruent initials:
(1184) *s-ta > syo > sh write; pMn *ty A(as if from OC*ta)
But obstruent preinitials before resonant initials led to different reflexes in Mn and
Hakka, while in the dialects represented in the Middle Chinese written sources, and
in most modern dialects, such obstruent preinitials were simply dropped (see section
4.4.4.4):
(1185)
*ra > luX > l salty (sc. land), pMn *l-, Mixin (Hakka) /lu 1/
*C.ru > lawX > lo old, pMn *lh-, Mixin /lau 3/
The rich commentarial literature on early texts and the Middle Chinese written
sources probably contain additional clues that may make it possible to further refine our
hypotheses about early dialect differences.
6.3 Knownissues
In the Middle Chinese reflexes of *-a, *-aj, and *-ak, there are contrasts that we are not
yet able to account for:after acute onsets, from *-a we sometimes have -jo and some-
times -jae (section 5.4.1.1); from *-ak we sometimes have -jak and sometimes -jek (sec-
tion 5.4.1.2); and from *-aj we sometimes have -je and sometimes -jae (section 5.5.2.1).
We use ad hoc notations to mark these distinctions, but they simply identify the problem
and do not solve it. In the case of *-a, we have contrasts such as these:
(1187)*-ak > -jak: *tak > tsyak > zhu cut, hack
*-Ak > -jek: *tAk > tsyhek > ch foot (measure)
Conclusion321
The explanation might have to do with dialect differences or prosodic factors; another
possibility is that in some cases the development of the rhyme might be influenced by
the preinitial. For now, it should simply be emphasized that our *A is not a seventh
vowel:it is an explicitly ad hoc notation to flag some unsolved problems.
6.3.2*s- VS.*S-
(1189)
*S-t-s > tsingH > zng boiler for steaming rice, pHM *tsjH;
probably from the same root as
*t > tsying > zhng to rise (of steam); steam (v.t.)
(1190) *s.tan > syen > shn smell of sheep; the phoneticis
*tan > tanX > dn sincere,truly
In these cases, too, we cannot explain the contrast. It is possible that there really were two
kinds of preinitial *s-; or perhaps the difference between *s.C- and *s.C- is involved.
There are some word or syllable structures that we have found no reason to reconstruct
but that may well have existed:for example, we have reconstructed no cases of *s.
before voiceless stops or before nasals. Hypotheses about these missing structures might
provide solutions to other problems (such as the problem of *s.- vs. *S.- just discussed).
We have provisionally assumed that all cases of the postcoda *-s are morphologically
suffixes, and we have reconstructed several well-supported functions for such a suffix,
including (1)deriving nouns from verbs, (2)deriving outwardly directed verbs from
stative verbs or adjectives, and (3)deriving verbs from nouns (see section 3.3.2.7).
However, there are still many attested cases of *-s that are not covered by these expla-
nations. It is quite likely that some of the original morphology of Old Chinese has
322 Old Chinese
become obscured in the reading tradition, so that we cannot always rely on our Middle
Chinese sources to tell us which forms had *-s and which did not. However, the question
deserves further research.
More generally, further and more precise research is needed on each of the kinds of
evidence for reconstructing Old Chinese outlined in Chapter2. This is true of the tra-
ditional trio of Middle Chinese, Old Chinese rhymes, and graphic evidence, and even
more so for evidence from previously neglected sources such as new paleographical
discoveries, modern dialects, and early Chinese loanwords in other languages.
Although much attention has been devoted to the written sources for Middle Chinese,
it has not always been clear what questions we should be trying to answer when examin-
ing them. Now that it is well established that the Qiyn system was not a single dialect,
it would be useful to investigate more thoroughly the dialect diversity represented in
our sources and to try to connect it with what can be learned from direct research on
modern dialects. Our revised reconstruction suggests a number of research questions to
be pursued, such as the geographical distribution of the different reflexes of the coda *-r,
or of the different treatments of preinitials.
So far, the study of Old Chinese rhymes has focused mostly on the Shjng, which
is the largest single corpus of early rhymes, but there are rhymed passages scattered
throughout early texts that are also worthy of study. By using hypotheses about pho-
nological changes and dialect distinctions within the Old Chinese period, it might be
possible to locate texts more accurately in space and time based on the characteristics of
their rhymesor to revise and correct our hypotheses.
Although in principle we have preferred to use evidence from the pre-Qn script
rather than the standard script in developing our reconstruction, the number of avail-
able early documents is steadily increasing, and making full use of them for linguistic
reconstruction will require much time and effort from many people. Here we have tried
to indicate the kind of research that is needed, but much more remains to be done. We
are confident that our reconstruction is a more effective tool for paleography than the
traditional approach now widely used, but additional research is needed to demonstrate
this and to test and possibly correct our hypotheses.
In particular, it would be useful to know with more precision the spatial and temporal
origins of changes in the pre-Qn script. We have made explicit hypotheses about sound
changes that took place during this period and have argued that these changes can explain
certain changes in the scriptsuch as the replacement of hn by mn as a phonetic
element for writing {} wn to hear (see section 3.4). One way of testing and improv-
ing hypotheses like this is to systematically investigate when and where words came to
be written in new ways; this should clarify the chronology and geography of linguistic
changes during the pre-Qn period, which in turn should make it easier to locate particular
texts in place and time based on internal evidence.
Conclusion323
We believe that we have demonstrated the necessity of using data from modern dia-
lects, not just the Middle Chinese of the Qiyn system, in reconstructing Old Chinese.
There is now much better documentation of the Mn dialects than in the past, and
although we have relied on Normans reconstructions, additional precise historical work
on these dialects is needed in order to make full use of the evidence they provide. No
adequate Old Chinese reconstruction can ignore the phenomenon of softened initials in
Northern Mn,for example, but some of the cases where softened initials correspond to
Middle Chinese voiced initials do appear to be of secondary origin (see section 4.2.1.1).
There is still more work to be done on reconstructing both the onsets and the finals
of Proto-Mn in order to properly sort out the historical layers of vocabulary in these
dialects. Also, for reconstructing Old Chinese vocabulary, it is not enough to know that
such-and-such a distinction existed in Proto-Mn; we also need to reconstruct as many
individual words as possible that illustrate the distinction.
It is now also clear that not just the Mn dialects but also Hakka, Wxing, and the
Ch-Q dialects of southern Zhjing preserve features lost in Middle Chinese,
and other dialects can probably be added to the list. All these should be thoroughly
researched from a historical point ofview.
Finally, there is much more to learn about early Chinese loan words in other lan-
guages and probably loans into Chinese from those languages as well. Important histori-
cal work on the history of Kra-Dai, Vietic, and Hmong-Mien is now underway, which
should make it possible to improve our reconstructions in the future.
It is largely evidence of the kinds just mentioned that has led to the main innova-
tions in our new reconstruction of Old Chinese, which mostly involve a more com-
plex system of presyllabic elements. The reconstruction of rhymes has remained
fairly stable since Baxter (1992), apart from the addition of the coda *-r and the
revised reconstructions of individual items. By contrast, our hypotheses about pre-
syllables are based in many cases on still incomplete data and analysis, and although
we are confident that we have made progress, we expect that this part of our recon-
struction will be most likely to require modification as research continues in the
future. As we argued earlier (section 1.2.1), linguistic reconstruction is a continuing
process, and our reconstructions are a basis for further research rather than a final
result.
In elaborating our reconstruction of Old Chinese, we have been careful not to base our
inferences on facts from the Tibeto-Burman languages, not because we think they are
not related to Chinesewe do accept that Sino-Tibetan is a valid groupingbut as a
methodological choice:if Chinese and Tibeto-Burman are the two primary branches of
the Sino-Tibetan family, as some authors believe, it is preferable that reconstruction of
the ancestral language be based on wholly independent reconstructions of Old Chinese
and Proto-Tibeto-Burman.
324 Old Chinese
Our new reconstruction at times brings Old Chinese words closer to existing
Tibeto-Burman forms:the word fly (n.), our *m-r (see section 4.4.2.4), is now
closer to Written Tibetan sbrang (perhaps < *s-mrang) than in any other previous
reconstructions of this word. Our reconstruction does, however, sometimes move some
Old Chinese words away from existing Tibeto-Burman reconstructions:in water,
our *s.tur, final *-r appears to move Old Chinese away from Proto-Tibeto-Burman
*twij water (Benedict 1972), Proto-Sino-Tibetan *tujH water (Peiros and Starostin
1996 2.146). It does, however, lead to more regular sound correspondences. Consider
the word *t.l u[n] > tsyhwen > chun stream, river, where the final consonant
is placed between brackets because it is ambiguous for OC *-n and *-r:suppos-
ing the coda was *-r, the main syllable in this form can be compared with Benedicts
Proto-Tibeto-Burman reconstruction *lwi(y) flow; stream with the same rhyme corre-
spondence:Proto-Tibeto-Burman (PTB) *-wij:OC *-ur, as water. Similarly for egg,
Benedicts Proto-Tibeto-Burman *twiy:the main early word for egg in Chinese,
*k.ror > lwanX > lun egg, is unrelated to the Tibeto-Burman word, but there is a
vulgar word for egg, also used for testicles in southern Chinese dialects:Cantonese
/tn 1/ egg, Hakka /tun1/ eggs of birds, reptiles; roe of fish. This is derivable
from an OC *tu[n] (where [n] is ambiguous between *r and *n). Assuming, again, that
the Old Chinese coda was *-r, this word matches Proto-Tibeto-Burman (PTB) *twij by
the same rhyme correspondence as that for stream and water:
The data in Table6.1 suggest that the Proto-Sino-Tibetan words ended in some kind
of rhotic sound, which went to [j] word-finally in Proto-Tibeto-Burman. The words
for water and egg have different tones in some Tibeto-Burman languages: for
instance Boro /2dy/ water, river versus /1dy/ egg (Bhattacharya 1977:280, 288),
Mizo (Lushai) /tui35/ water versus /tui 55/ egg (Matisoff 2003:195). It is possible
that some of these tonal contrasts, not taken into account in existing reconstructions
of Proto-Tibeto-Burman, might originate in the presence or absence of the glottal stop
inherited by Old Chinese. Compare the Chinese and Proto-Bodo-Garo forms (from
Joseph and Burling 2006) in Table 6.2. The Proto-Bodo-Garo tone 2 comes from final
-, still present in Garo, according to Joseph and Burling (2006:101).
Many comparative questions remain without an answer:What are *s. and *t. in the
Chinese forms for water and stream? Why is initial *l - voiceless in the Chinese
word for stream? Why is the initial stop unaspirated in the Chinese word for water
but aspirated in egg? We expect that progress in Tibeto-Burman reconstruction will
eventually bring some answers; we do not exclude that other answers might come from
comparison of Sino-Tibetan with other language groups of the region.
Although we have avoided using data from outside Chinese to test hypotheses about
Old Chinese, it is still true that an improved understanding of related languages should
eventually give us a better understanding of Old Chinese as well. An example from
Proto-Germanic may serve as a parallel. Some Proto-Germanic roots showed alterna-
tions between *s and *z (surviving in English in was vs. were or lose vs. (for)lorn,
with /r/ from earlier *z) or between * and * (surviving in English seethe and its old
past participle sodden). These alternations can be reconstructed for Proto-Germanic,
but to understand their origin we must look at related languages within the larger
Indo-European family; then we see that they result from differences in placement of the
Proto-Indo-European accent (the pattern known as Verners law). This broader perspec-
tive does not necessarily change our reconstruction of Proto-Germanic, but it gives us
a better understanding of it. Similarly, we expect that some phenomena of Old Chinese
will not be fully understood until the history of the larger Sino-Tibetan family is clari-
fied. More broadly, phenomena that are puzzling from a Sino-Tibetan point of view may
become clearer when examined in a still broader comparative context.
In closing, we hope that, apart from the light thrown on the history of Chinese proper,
our reconstruction will help clarify some aspects of the Sino-Tibetan familys history:in
particular, that it will facilitate the search for early Sino-Tibetan innovations. This, we
believe, is essential to resolving the familys structure and a prerequisite to successful
reconstruction of the protolanguage.
Appendix of reconstructed forms
The list below includes all and only the words reconstructed in this book; a more com-
prehensive set of reconstructions (updated as new evidence becomes available) is avail-
able at http://ocbaxtersagart.lsait.lsa.umich.edu/. Words are listed in alphabetical order
by modern pronunciation in pnyn romanization, then by Middle Chinese transcription.
Modern Mandarin pronunciations are based on Hny d zdin and do not always
match the pronunciations that would be predicted on the basis of Middle Chinese. The
English glosses are for identification only and are not intended as definitive semantic
reconstructions.
b bjijH < *m-bi[t]-s nose: 89, 95, 132, 142, 188, 393n4
b pijX < *pr border town: 229
b pit < *p.[r]ut (dialect:*p.r- > *pr-) writing brush: 42, 43, 89,
162,294
b pjeX < *paj that: 65
b pjijX < *C.pij compare: 289, 290
b bejX < *m-pe femur: 55, 232; see also b < pjieX
b pjieX < *pe femur, haunch: 55; see also b < bejX
b bet < *[b]i[t] fragrant: 290
b bjiek < *[N]-pek law, rule; lawful: 233
b bjieX < *N-pe low, short: 117
b pek < *C.pek house wall: 168, 233
b pjaek < *prak light blue: 225
b pjeH < *por-s ornate (name of a hexagram): 259, 295;
seealso bn < pwon, fn < bjun
b pjiejH < *pe[t]-s cover (v.): 276, 277
b pjijH < *pi[t]-s give: 142, 390n65
b pjit < *pi[t] necessarily: 290
bin pen < *pe[n] side 100, 277
bin pjien < *pe[n] whip (n.) 277, 388n52
bin penX < *pe[r] flat and thin: 278
bin pjemX < *pr[e]m diminish: 313, 403n111
bin bjenH < *C.[b]ro[n]-s cap: 204205, 209
bin bjenX < *[b]ren distinguish: 277
bin bjienH < *[b]e[n]-s comfortable; advantageous: 204205
bin bjienX < *[m-p]e[r] insincere words: 278; see also pin < bjien
bin pjenH < *pro[n]-s change (v.) 208209, 217, 252, 282,
394n13, 395n17
bio pjiw < *p(r)iw long hair: 300
bio pjewX < *p(r)aw exterior: 218, 296
bi bjet < *N-pret be separated (intr.): 88, 116117, 118, 275;
seealso bi < pjet
bi pjet < *pret separate (tr.): 116; see also bi < bjet
bn pjin < *pi[n] guest: 196, 291
bng ping < *p.r (dialect:*p.r- > *pr-) ice: 217, 396n27; see also
lng < ling
bng pjaengX < *pra third heavenly stem: 17, 163, 217, 227
bng pimX < *p.rim (dialect:*p.r- > *pr-) receive: 162
bng pimX < *p.rim (dialect:*p.r- > *pr-) rations: 162; see also ln
<limX
330 Appendix of reconstructedforms
ch tsyhaeX < *[t-l ]Aj large; see also ch < tsyheX: 270, 321
ch tsyhek < *tAk foot (measure): 76, 104, 226, 320, 385n25
ch tsyheX < *[t-l ]aj large: 270, 321; see also ch < tsyhaeX
ch tsyhiX < *t-[k]() or *t- front teeth: 32, 48, 57, 79,
157158
ch syeH < *s-ke-s wing: 140
ch tsyhek < *[t-q](r)Ak red: 103
chng drjuwng < *[d]ru sound of cutting ice: 310
chng drjowng < *[m]-tro repeat; double: 244; see also zhng
<drjowngX
chng drjuwng < *C.lru insect: 109, 172
chng dzrjuwng < *[dz]<r>u exalt, honor: 250
chng trhjowngX < *r o favor, grace: 112
chu trhjuw < *r iw recover: 115
chu drjuw < *[d]riw bind round: 300
chu dzrjuw < *[dz]riw grieved: 74, 300
chu trhjuwX < *n <r>u shackles, handcuffs: 80
chu trhjuwX < *n <r>u shackles, handcuffs: 407
chu tsyhuwH < *t-qu()-s odor, to stink (intransitive)?: 57
ch tsrhjo < *[ts]ra beginning: 223
ch tsyhwit < *t-kut go or come out: 56, 79, 158, 293, 294;
seealso chu < tsyhwijH
ch drjo < *[l]<r>a remove; get rid of 81, 109, 145146, 390n69,
390n70
ch dzrjo < *s-[l]<r>a hoe (n.): 145146, 390n69
ch dzrjo < *s-[l]<r>a hoe (n.) 81, 145146, 390n69
ch tsrhju < *[ts]ro grass for fuel or fodder: 242, 396n25
ch tsrhjoX < *s.r a thorns: 150
ch tsyhoX < *t.qa pestle: 79, 83, 128129
ch tsyhoX < *t.qa be at: 129, 138; see also ch < tsyhoH
ch trhjuwH < *q<r>uk-s domesticated animals: 250; see also
ch <trhjuwk, x < xjuwk, x < xjuwH
ch trhjuwk < *q<r>uk store (v.): 103, 249, 250; see also ch
<trhjuwH, x < xjuwk, x < xjuwH
ch trhwit < *t.k<r>ut expel: 294
ch tsyhoH < *t.qa-s place (n.): 129; see also ch < tsyhoX
chui tsrhjweX < *s-t<r>or to measure; to estimate: 80, 139
chui tsrhwaejH < *[ts]()ro[t]-s bite, eat; see also chui <
tsrhweajH:281
Appendix of reconstructedforms 333
dun dwanX < *N-to[n] be cut in two: 117, 118, 282; see also
dun < twanX
du twoj < *tur (dialect:*-r > *-j) manage, direct: 263, 295;
seealso dn < twon, Dnhung < twon.hwang
du twojH < *[t][u]p-s respond: 309
dn twon < *tur solid, thick: 263, 295:see also du < twoj,
Dnhung < twon.hwang
dn twon, see Dnhung < twon.hwang
Dnhung twon.hwang < *tur.[]a Dnhung (place name): 263, 295
also written ; see also dn < twon, du < twoj
dn dwonH < *lu[n]-s withdraw: 251
du ta < *[t.l]aj many: 1011, 164
du dak < *[d]ak measure (v.): 27, 220, 225, 226, 397n40;
seealso d < duH
du dak < *lak a kind of bell: 27
du dwat < *C.lot seize: 190
du twat < *tot pick, gather: 21
du dwaX < *loj to fall: 182; see also hu < xjwie
du dwaX < *loj lazy: 280
'a < *qa[j] slope, river bank: 121, 271, 399n63; see also
huxun < 'a.hwajH.sjwen
nga < *a[r] goose: 92
ngwa < *m-qaj move; change: 269
'ak < *ak bad, ugly: 59, 226, 226; see also w < 'uH
'eak < *q<r>[i]k part of a yoke: 58
ngaH < *aj-s hungry: 257, 272
ngak < *N-qak scared: 121, 175
ngak < *ak speak frankly: 56
huxun 'a.hwajH.sjwen < E.Hn *a-wajs-swar (< OC *qaj +
*m-kop-s + *s-[q]ar) < Sanskrit bhsvara shining: 258;
see also < 'a, hu < hwajH, xun < sjwen
n 'on < *[n] kindness, favor: 283
r nye < *e child: 77, 108, 133
r nyi < *n and, but: 53, 7273
r nyiX < *C.n ear: 158
r nyeX < *n[][r] you(r): 133
r nyijH < *ni[j]-s two: 110
f pjot < *C.pat fly forth, send forth: 186, 210, 271, 272, 273,
391n82
f bjop < *[b](r)[o]p lack (v.): 311, 312
Appendix of reconstructedforms 337
f pjop < *[p.k]ap model, law: 151, 152153, 272, 311, 312
f pjop < *[p.k]ap model, law: 151, 152153, 272, 311, 312
f pjot < *pot root; hair (of head): 273, 281, 400n78
fn phjon < *par a turn, a time: 259, 399n62; see also bb
<pa-pa
fn pjon < *par hedge, screen, fence: 257, 258
fn pjon < *par hedge, screen, fence: 257, 258
fn bjon < *[b]ar burn, roast: 256, 257
fn bjon < *[b]ar abundant, numerous: 258
fn pjonX < *C.pan reverse (v.): 186, 208209, 256, 259, 274,
394n13, 394n15, 394n16
fn bjomX < *[m-p](r)om bee: 313, 314
fn bjonH < *bo[n]-s cooked rice or millet: 205, 282
fng pjang < *pa container, box (Shuwn): 142143, 151, 159
fng pjang < *C-pa square: 57, 142143, 151, 158159, 227
fng pjangX < *pa just then, at that time: 217
fi pj+j < *pj is not: 285
fi pj+j < *C.p[r] (dialect:*-r > *-j) fly (v.): 86, 186
fi bj+j < *[b][]r fat (adj.): 86
fi bjojH < *C.bo[t]-s bark (v.): 189, 281
fi pj+jH < *N.p[u][t]-s boil (v.): 88, 95, 174
fi pjojH < *pat-s great: 272
fi pjojH < *[p-k]ap-s cast aside: 152153, 154, 272, 312,
391n82
fn pjun < *p[n] divide: 216, 217, 288
fn bjun < *[b]u[n] burn (v.): 251
fn bjun < *[b]ur great, big: 295; see also bn < pwon, b < pjeH
fn pjunX < *m.pn flour: 177
fng phjowng < *p(r)o bee: 104, 244, 314
fng pjuwng < *prm wind (n.): 195, 309, 310
fng bjuwng < *[C.b] Fng (surname): 216, 231
fng pjowngX < *p(r)om (dialect:*-om > *-o) overturn
(nopre-Hn exx.): 312
fng bjowngH < *C.[b](r)o-s seam: 171
fng bjowngX < *m-p(r)o present (v.) with both hands: 128;
seealso fng < phjowngX
fng phjowngX < *p(r)o present (v.) with both hands: 104, 128;
seealso fng < bjowngX
f pju < *pra skin: 223
f bjowk < *[b](r)ok kerchief (Shuwn): 243
338 Appendix of reconstructedforms
gun kwanX < *ka[n] exhausted, helpless: 259; see also gun
<kwanX tube; flute
gun kwanX < *[k]o[n] tube; flute: 282; see also gun < kwanX
exhausted, helpless
gun kwanX < *[k]o[n] tube; flute: 282
gun kwanH < *k.or-s cap (v.): 58, 151, 154; see also gun
<kwan.
gun kwanH < *[k]or()-s pour out libation: 282
gun kwanH < *kon-s pass through the center: 208209, 395n16
gun kwanH < *C.qar-s watchtower: 234, 278; see also
rnggun < hjwaeng.kwanH
gun kwanH < *C.qar-s heron: 261
gung kwang < *ka light, brightness: 81, 117
gung kwangX < *ka wide: 227
gu kjw+j < *[k]j return (v.): 385, 391n80, 401n91
gu kwej < *[k]e jade scepter: 232
gu kwij < *[k]r tortoise: 229, 248, 398n54
gu kjweX < *[k](r)oj perverse: 279
gu kjwijX < *kij tenth heavenly stem: 289
gu kjw+jX < *k-uj ghost: 101, 151, 154, 391n80
gu kwijX < *kru gu ritual vessel: 247, 248, 398n54
gu kwijX < *kru wheel ruts: 218, 247, 248, 397n30, 398n54
gu gwijH < *[g]ruj-s box (n.): 293
gu kjwejH < *k(r)[o][t]-s lift (the dress): 281
gu kjw+jH < *kuj-s precious; expensive: 101, 102, 196, 391n79
gu kwajH < *kop-s place where two ends of a collar or belt
join:312
gu kwak < *kak outer wall: 225
gu kweak < *C.q<r>k severed left ears: 230
gu kwaX < *[k]o[r] fruit; result: 282
gu kwaH < *kaj-s to pass; transgress: 271
hi hajH < *N-kat-s be hurt (v.i.); injury (n.): 197, 272
hi heajX < *[g]r alarmed: 229
hn han < *C.[g]a[n] cold: 274
hn han < *[g]ar (state in the Korean peninsula, Three Kingdoms
period): 261
hn hom < *C-m-k[]m hold in the mouth: 192
hn hanH < *[g]ar (rhymes as *-ar, but MC implies *[g]ar -s)
white (of a horse): 259; see also hn < hanH prop up,
support
Appendix of reconstructedforms 341
hn hanH < *m-kar-s prop up, support: 257, 258; see also hn
<hanH white (of a horse)
hn hanX < *[g]a[r] dry, drought: 106
hn xanH < *n ar-s (W dialect:*n- > x-, *-r > -n) (river name):
112, 114, 387n37
ho xaw < *qu weed (v.): 103
ho xaw < *qu weed (v.): 103, 390n71
ho haw < *[C.g]aw call out: 246
ho xawX < *qu good: 59, 102, 103, 246; see also ho < xawH
ho xawH < *qu-s love, like (v.): 59; see also ho < xawX
h hap < *m-[k]ap thatch, cover (v.): 151, 153, 197, 311
h hop < *m-kop come together; bring together: 125, 127, 311,
312; see also g < kop
h hot < *m-[q]t bite (v.): 287
h hwa < *[]oj harmonious: 266, 270, 271
h hwa < *[]oj growing grain: 279
h haH < *m-kaj-s congratulate: 272
h hak < *[g]awk crane: 297
h hat < *[]at coarse cloth: 44
h howk < *[g]awk high: 297
h xaek < *q<r>ak frighten: 121, 175
h xaek < *qrak red, fiery: 103
hi xok < *m k (dialect:*m- > x-) black: 42
hng xaeng < *qra penetrate: 157; see also xing < xjangX
hng hong < *[g] constant: 231
hng hwaeng < *C.gra crosswise; horizontal: 171
hng huwng < *go pink: 105
hng hweang < *[g]<r> resounding; great: 231
hng hwong < *[] vast: 231
hu huw < *m-g(r)o monkey: 178
hu huw < *[g](r)o feudal lord: 242
hu huwX < *C.[g](r)o thick: 189
hu huwX < *(r)o sovereign; queen: 105
hu huwH, see xihu < heaH.huwH
h xu < *qa call out, shout: 220; see also h < xuH
h xwot < *m ut careless; confused: 111
h hwot < *[g]ut dig out: 158
h xuX < *qra (W dialect:*qr- >r- > x-) tiger: 103
hh xuX.xuX < *qa.qa sound of hewing wood, also written
: 129; see also hh < xuX.xuX
342 Appendix of reconstructedforms
hun hwan < *[]ar pillar; martial-looking 258, 266; see also
Whun < u.hwan
hun hwaenH < *[g]ro[n]-s calamity; distress: 282
hun hwanX < *[g]or turn around (as a wheel): 282; see also hu
< hwaeX
hung xwang < *m a wasteland; uncultivated land: 152
hung hwang, see Dnhung < twon.hwang
hung hwang < *N-ka yellow: 81, 117, 118
hu xjwie < *l oj (W dialect:*l - > x-) destroy: 280; see also du
<dwaX
hu xjw+j < *qr brilliant: 253
hu xwoj < *[r]u[j] (W dialect:*r- > x-) exhausted, weary: 116;
seealso hu < xjw+jX
hu hwoj < *[]j go around: 285
hu xjweX < *qar (dialect:*-r > *-j) sunlight: 258
hu xjw+jX < *r u[j] (W dialect:*r- > x-) sound of thunder: 116;
seealso hu < xwoj
hu hwajH < *m-kop-s meeting; have a meeting: 312, 399n63;
seealso kui < kwajH, huxun < 'a.hwajH.sjwen
hu hwejH < *[]e[t]-s intelligent: 276
hu 'jwojH < *qat-s bad weeds; filth: 272
hu tsyhwejH < *t-l o[r]-s snout; to pant: 33, 57, 165, 393n103;
seealso hu < xjwojH
hu xjwieH < *l oj-s (W dialect:*l - > x-) shred sacrificial meat:
112, 280; see also tu < thwaX
hu xjwojH < *l o[r]-s (W dialect:*l - > x-) snout; to pant: 57, 165;
see also hu < tsyhwejH
hu xwojH < *quj-s wash the face: 101
hu xwojX < *q property, valuables: 44, 229
hn xwon < *m u[n] dusk, dark: 63, 64, 322
hn hwon < *[m.]q[n] spiritual soul: 288
hu hwat < *[g]at to live: 271
hu hweak < *m-qrak catch (v.): 225
hu hwok < *[]k some; or: 3940, 230
j gjaek < *C.[g]rek wooden sandal: 232, 233
j kej < *ke fowl, chicken: 232
j kej < *[k]er crosspiece of wood on top of pillar: 278; see also
jin < ken
j kek < *[k]ewk dam up (water): 299
j ki < *k(r) base (n.): 218, 229
344 Appendix of reconstructedforms
jng kjaengX < *C.qra bright; image: 28, 45, 101, 168
jng tsjengX < *C.tse well (n.): 168
jng dzjengH < *m-tse-s cleanse (v.t.); *N-tse-s clean: 55, 81
jng gjaengH < *m-kra-s ~ C-kra-s strive; compete: 126
jng hengH < *m-ke-s leg, shank: 58, 128
jng kjaengH < *C.qra-s mirror: 101, 168, 385n24
jng kjiengH < *ke-s (no palatalization before *-e?) strong: 78
jing kweng < *ke region distant from capital: 235
jing kjwaengX < *k-mra (dialect:*k.mr- > *kr-) bright
window: 57, 152
ji kjiw < *k-riw (dialect:*k-r- > *kr-) twist (v.): 137
ji kjiw < *k-riw (dialect:*k-r- > *kr-) twist (v.): 301
ji kjiw < *k-riw (dialect:*k-r- > *kr-) tie around, strangle: 57
ji kjuwX < *[k] a long time: 119, 248
ji kjuwX < *[k]u nine: 3132, 80, 155, 247, 248, 397n30,
398n55
ji kjuwX < *s.[k](r)u Allium: 247
ji tsjuwX < *tsu wine: 101, 247
ji gjuwH < *N-k-s old: 118, 119
ji gjuwX < *C.[g] mortar: 123, 229
ji gjuwX < *[g](r)u mothers brother: 106, 123
j kjo < *C.q(r)a chariot: 158, 224; see also ch < tsyhae
j tsjo < *tsa [final particle]: 145; see also qi < tshjaeX
j gjowk < *N-k(r)ok bent, curved: 120, 175
j kjwit < *C.qi[t] well-rope: 82, 290; see also y < ywit
j kwek < *kek shrike: 233
j kjoX < *C.q(r)a lift, raise: 131, 168, 171
j kjoX < *[k]ra round basket: 158, 223
j kjuX < *[k](r)o (a kind of tree): 381n11
j kjuX < *[k](r)a carpenter's square: 381n11
j kjoH < *k(r)a-s saw (n.): 224
j kjuH < *k-ro-s sandal, shoe: 242
j kjuH < *k-ro-s sandal, shoe: 242
j tsjuH < *[ts]ok-s replenish: 244; see also z < tsjowk
jun kjwenX < *[k](r)o[n] roll (v.): 118, 282
jun kjwienH < *[k]en-s a kind of silk stuff: 277
ju dzjwet < *[dz]ot cut off: 281
ju gjut < *[g]ut dig out (earth): 158
ju kaewk < *kruk be aware: 197, 249; see also jio < kaewH
ju kjwak < *C.q(r)ak anxious look: 225
348 Appendix of reconstructedforms
s sreaX < *C.s<r>r (MC srj- > sr-) sprinkle: 187, 396n23;
see also s < sreaH, x < srjeX, x < srjeH
s sreaX < *sre (MC srj- > sr-) hair-band: 232, 233; see also x
< srjeX
sn sam < *s.rum (*sr- > *s-; influenced by *s-?) three: 75
sn sanH < *m-sa[n]-s scatter (v.t.): 177
sng sang > *[s]a mulberry tree: 390n66
sng sang < *s-ma mourning, burial: 56, 143, 390n66; see also
sng < sangH
sng sangH < *s-ma-s lose; destroy: 143; see also sng < sang
so saw < *s-[ts]u scratch (v.): 136
s sok < *[s]k stop up, block (v.): 230
s srik < *s.rk color; countenance: 150
s srik < *s.rk reap: 150
sh srae < *sraj sand: 80, 101, 213
sh sreat < *s<r>at kill: 74, 214, 271, 272, 395n22, 396n23;
seealso shi < sreajH
shi srje < *C.sre to strain off wine: 187
shi sreajH < *s<r>at-s diminish: 272, 395n22, 396n22; see also
sh < sreat
shn sraen < *[s]ra[n] tears flowing: 274
shn srean < *s-rar mountain, hill: 148, 214, 258, 395n23, 399n64
shn syem < *s.tem thatch: 315, 401n86
shn syen < *s.tan smell of sheep: 274
shn syen < *s.tan smell of sheep: 274, 321
shn dzyenX < *[d]an leveled area: 218; see also tn < dan
shn dzyenX < *[g]e[n] good: 77, 78, 277
shng syang < *l a wound: 166
shng syang < *s-ta estimate; business; trader: 56
shng dzyangH < *da-s top, above (n.): 188; see also shng <
dzyangX ascend, shng < dzyangX put up
shng dzyangX < *C-da ascend: 132, 188, 227; see also shng <
dzyangX to put up, shng < dzyangH
shng dzyangX < *m-da to put up: 132, 188; see also shng <
dzyangX ascend, shng < dzyangH
shng syangH < *n a()-s give food; food: 385n23
sho sraew < *[s]()rew bamboo vessel: 298
sho syew < *[]ew burn: 298, 299
sho sraewH < *[s]()rew-s gradually; rations: 298, 395n23
sho sraewH < *[s]rewk-s zone near the capital: 300; see also xio
< sjak and xio < sjewH
Appendix of reconstructedforms 359
sh dzyuX < *m-to plant (v.); place upright: 124, 125; see also
sh < dzyuH
sh srjuH < *s-ro-s number (n.): 80, 144, 243; see also sh <
srjuX, shu < sraewk
sh syoH < *n a-s indulgent: 29
shui srwij < *sruj diminish, decline: 293; see also cu < tsrhjwe
shui srwijH < *s-rut-s lead (v.); commander: 144; see also shui
<srwit
shui srwijH < *s-rut-s leader (of an army): 293
shui srwit < *s-rut follow, go along: 144, 293, 294; see also shui
< srwijH
shui srwit, see xshui < srit.srwit
shung sraewng < *[s]ro a pair: 244
shung srjang < *[s]ra hoar-frost: 196
shu dzywij < *[d]uj who: 293
shu sywijX < *s.tur (dialect:*-r > *-j) water; river: 93, 97, 146,
253, 295, 324
shu sywejH < *l ot-s kerchief: 150; see also shu < tshjwejH
shu tshjwejH < *s.l ot-s kerchief: 150; see also shu < sywejH
shu sywejH < *l ot-s exhort: 281; see also shu < sywet
shn zywinH < *C.lu[n]-s follow; obey: 165166, 183
shu sywet < *l ot speak, explain: 281; see also shu < sywejH
shu sraewk < *s-rak first day of the month: 56, 80, 148, 149,
225, 391n78
shu sraewk < *s-rok frequently: 144, 243; see also sh < srjuX,
sh < srjuH
shu syak < *r ewk (W dialect:*r - > *x-, palatalizing) melt,
infuse:78
s si < *[s] silk: 90
s sij < *[s]j private: 90, 285, 401n91
s sje < *[s]e this: 262; see also sx < sje.sju
s sje < *[s]e split (v.): 232; see also sx < sje.sju
s sje < *s[e]r light rain: 262; see also xin < senH
sx sje.sju < *[s]e.[s]o a short time: 214; see also see also s
<sje
s sijX < *sij die (v.): 90, 289, 292
s sijH < *s.li[j]-s four: 90
s ziH < *s-m-lk-s feed (v.): 230, 398n41
s ziH < *s-m-lk-s feed (v.): 230
362 Appendix of reconstructedforms
s ziH < *s-m-lk-s feed (v.): 230; see also sh < zyik
s ziX < *s.l resemble: 183
s ziX < *s-[] sixth earthly branch: 3031, 382n17, 386n30
s zriX < *s-[]r wait: 141
sng sjuwng < *[s]u high: 250
sng zjowng < *s.o pine (n.): 181
su srjuw < *sru search: 247, 390n67
su suwX < *s-ru (dialect:MC srj- > sr- > s-) old man: 144145,
167, 247, 248, 389n59
su suwX < *s-ru (dialect:MC srj- > sr- > s-) old man: 144, 247
su suwX < *s-ru (dialect:MC srj- > sr- > s-) old man: 144, 247
s su < *s-a revive: 148149, 391n78
s sjuwk < *siwk solemn, severe: 249, 295, 301, 302
s sjuwk < *[s]uk spend the night: 249; see also xi < sjuwH
s suH < *s-ak-s complain, accuse: 56
sun swanH < *[s]or-s calculate; tally: 283
su swij < *s.nuj comfort (v.): 293
su zjwe < *s.loj follow: 182, 252, 271
su sjweX < *s-loj marrow: 279
su swojH < *[s-ts]u[t]-s break: 293
sn swinX < *s-qi[n] bamboo sprouts: 137
sn swinX < *[s]ur hawk, falcon: 253, 295
su sa, see psu < ba.sa
su srjuwk < *[s]ruk shrink: 249
su srjoX < *s-q<r>a place (n.); that which: 129, 140; see also
hh < xuX.xuX, hh < xuX.xuX
t tha < *l aj (E.dialect:*l - > th-) another: 269
t that < *r at otter: 115, 134, 135; see also t < trhaet
t trhaet < *[m-r]at otter: 134, 135; see also t < that
t dop < *m-rp reach to; and: 134, 308
t dop < *m-rp reach to; and: 133, 134, 308
tn than < *n ar (E dialect:*n- > th-) foreshore: 111, 112
tn than < *n ar to sigh: 258, 260
tntn than-than < *tar-tar numerous: 257
tn dom < *N.r[o]m extend, spread: 386n31
tn dan < *C.dar shoot pellets: 189
tn dan < *[d]an altar: 25; see also shn < dzyenX
tn thanX < *[t]a[n] level, at ease: 113
tn thanH < *[t]a[n]-s charcoal, coal: 104
tng thang < *l a hot liquid: 111, 114, 115, 166; see also tng
<thang name of first Shng ruler
Appendix of reconstructedforms 363
tng thang < late OC *l a name of first Shng ruler, from earlier
*r a: 166; see also tng < thang hot liquid
tng thang < *r a name of first Shng ruler: 166; see also kng <
khang
tng thang < *r a name of first Shng ruler: 166; see also tng <
dang
tng dang < *C.la sugar: 109
tng dang < *[N-]ra exaggerate; great: 166; see also tng < thang
toti thaw.thet < *[t]aw.[t]t glutton: 287
to daw < *C.law peach: 109, 172, 246, 387n34
to daw < *law flee: 109
tng dong < *l tie; band: 307
t thej < *l []j stairs: 115
t thek < *l ek cut (v.): 233; see also t < thejH
t dej < *C.le weep, howl: 232
t dej < *C.le weep, howl: 232
t thejX < *r ij body; limbs: 112, 115
t thejH < *l ek-s shave: 234; see also t < thek
tin then < *l n (surname): 202, 283; see also tn < thon
tin then < *l i[n] heaven: 113114, 291; see also xin < xen
tin den < *li field; to hunt: 33, 60, 109
tin denX < *[d][n] cease; destroy: 288
tin themX < *l [i]m to lick, to lap (not in GSR): 113
tio dew < *[l]iw arrange; branch (n.), shoot (n.): 109, 300
tio dewX < *lew provoke: 298
ti thep < *[t]ep submit, peaceful: 315
ti thet, see toti < thaw.thet
tng thengX < *l e stand up straight: 159
tng thengX < *l e good (Shuwn): 159, 164
tng thuwng < *l o penetrate: 56, 81, 150
tng duwng < *[l]o bronze, copper: 36
tng thuwngX < *l o bucket: 36
tu duw < *[m-t]o head: 55, 124
t du < *[d]a butcher (v.): 27
t du < *la path: 26, 27
t thuX < *ta earth: 128, 223, 224, 276, 388n51
t thuX < *ta eject from mouth, spit: 220
t thuH < *ta-s vomit: 220
tun dwan < *C.[d]on round, plenty: 25, 171
tun thwanH < *l o[r]-s running pig: 165
tu thwoj < *tuj push away: 139, 293
364 Appendix of reconstructedforms
xi haeX < *gra down: 105, 131, 197; see also xi < haeH
xi haeX < *[]ra great: 121
xin sen, see pinxin < ben.sen
xin sen < *sr first: 218, 255; see also xin < senH
xin sjem < *s.q[a]m insincere, ingratiating: 137; see also qin
<tshjem
xin sjen < *[s][a]r fresh; good: 262, 400n70; see also Xinbi
<sjen.pjie
xin xen < *l i[n] (W Hn-time dialect:*l - > *x-) heaven:
113114; see also tin < then
Xinbi sjen.pjie < *s[a]r.pe Xinbi: 261262, 399n67; see also
xin < sjen, bi < pjie
xin heam < *C.[g]r[o]m salty: 107
xin heam < *[g]r[]m all; everywhere: 154
xin hean < *m-[k]<r>en spy on, watch: 58
xin hen < *[g]i[n] worthy: 201, 291
xin xjaemX < *qr[a]m precipitous, dangerous: 103
xin xenX < *qen display, manifest: 113
xin heamH < *[]rom-s fall into a pit: 314
xin heamH < *[]rom-s small pit: 314
xin heanX < *[g]r[n] obstacle, limit: 288
xin henH < *m-[k]en-s cause to appear, introduce: 55; see also
xin < henH appear, jin < kenH
xin henH < *N-[k]en-s appear: 54; see also xin < henH cause
to appear, introduce, jin < kenH
xin henH < *N-[k]en-s appear: 116, 118
xin senH < *s[e]r-s sleet: 262
xin senH < *s[e]r-s sleet: 262; see also s < sje
xin senH < *sr-s go first: 201; see also xin < sen
xin xjonH < *qar law; model; rule: 257
xin zjenH < *s-N-qa[r]-s covet, desire: 191192
xing sjang < *C.[s]a box (of a carriage): 169
xing sjang < *s-na remove: 149
xing xjang < *qa fragrance: 102, 403
xing haewng < *m-kru[] (? < *-[u]m) submit: 250
xing haewng < *m-kru[] (? < *-[u]m) submit: 215, 250
xing zjang < *s.a auspicious: 141
xing syang < *n a bring food to: 111
xing xjangX < *qa sacrificial offering: 157; see also hng
<xaeng
368 Appendix of reconstructedforms
yun hjwonX < *C.an far: 67, 252, 259, 394n16, 395n21
yun hjwenH < *ra[n]-s wall around a courtyard: 78, 108, 274
yu 'jak < *[q](r)ewk bind (v.), make an agreement: 299; see also
yo < 'jiewH
yu ngaewk < *[]rawk music: 297; see also l < lak, yo <ngaewH
yu ngjwot < *[]at moon, month: 64, 271
yu ngwaet < *[]<r>at amputate a foot: 271
yu yak < *lewk flute; key; tube: 299
yu ywet < *lot exuviae of insects or reptiles: 109; see also tu
<thwajH
yn hjun < *[][n] cloud: 284, 288
yn hjun < *[][n] (a particle): 3840, 284, 382n26
yn hjun < *[][r] say: 3840, 382n25, 382n27, 389n62
yn hwin < *[]ri[n] rind of bamboo: 216, 291
yn ywin < *[N-q]i[n] even, uniform: 127, 216, 291, 388n50
yn hwinX < *[]r[n] fall down: 288
yn hwinH < *[m-q]<r>i[n]-s harmony; rhyme: 127, 193,
388n50, 393n107
yn 'junH < *un-s anger, angry: 294
z dzop < *[dz][u]p mixed: 309
zi dzojH < *[m-ts]-s load on a vehicle (v.t.): 127; see also zi
< tsojH
zi dzojX < *[dz] be at, be present: 202203, 284, 394n7
zi tsojH < *[ts]-s be conveyed in a vehicle: 127; see also zi
< dzojH
zi tsojH < *[ts]()-s twice; a second time: 203
zn tsom < *C.ts[]m hairpin: 186
zo tsawX < *m-tsaw bleach; wash: 95, 177, 982
zo tsawX < *N.tsu early: 88, 95, 174, 247
zo sawH < *C.saw-s shout: 169
zo sawX < *C.saw dry: 169
z draek < *lrak marsh; moisture: 109
z tsreak < *s-trek demand payment: 80, 98, 136, 233, 234;
seealso zhi < tsreaH
zi dzok < *k.dzk injure; murderer, bandit: 37, 95, 97, 153, 160
zng tsong < *s-t increase: 59, 136, 192, 231
zng tsingH < *S-t-s boiler for steaming rice: 61, 136, 321
zh tsrae < *tsra kind of fruit tree: 223
zh tsreat < *s-qrt strip (n.), tablet: 80, 138, 287
zhi treak < *trek pluck (v.): 233
Appendix of reconstructedforms 375
zhn tsyim < *t.[k]m needle: 37, 71, 79, 97, 153, 154
zhn tsyin < *tr numerous; majestic: 255
zhn tsyin < *ti[n] true, real: 76, 97, 99, 291
zhn trinH < *t<r>i[n]-s press down: 80
zhng tsyeng < *C.te first (month): 97, 98, 168; see also zhng
<tsyengH
zhng tsying < *t to rise (of steam); steam (v.t.): 56, 60, 136, 321
zhngrng dzreang.hweang < *[dz]re.[]re high, precipitous: 235
zhng tsyengX < *te arrange; orderly: 43
zhng tsyengH < *te-s correct (adj., v.): 235; see also zhng < tsyeng
zh trje < *tre know: 232
zh tsye < *ke branch (of tree): 77, 79
zh tsye < *ke branch (of tree), limb: 77, 79, 140, 232
zh tsye < *tar ritual vessel: 258
zh tsyek < *tek single: 98, 99
zh tsyi < *t (3p object pronoun; attributive particle): 53
zh tsyij < *kij fat, grease: 77, 216
zh drik < *N-t<r>k straight: 117, 118, 230
zh tsyip < *[t]ip seize: 308, 309, 131
zh tsyip < *t-nip (dialect:*t-n- > *t-) afraid, to fear: 57, 162;
seealso zh < nep
zhzh drjek.drjowk < *[d]rek.[d]rok stamp the feet: 233, 243
zh tsyeX < *k.te paper: 37, 97, 98, 153, 156, 161
zh tsyijX < *kij fine-tasting: 77, 289
zh tsyijX < *m.kij finger; point: 79
zh tsyijX < *tij whetstone: 289
zh tsyiX < *t foot; stop: 157, 158, 229
zh driH < *lr-s regulate, arrange: 229
zh triH > *trk-s put, place; set upright: 117
zh drjejH < *[d]r[a][t]-s obstruct: 272
zh trijH < *t<r>i[t]-s (cause to arrive):transmit: 291
zh trit < *[t]ri[t] stop up (v.): 290
zh tsrit < *[ts]rik comb: 239
zh tsyaeH < *tAk-s roast, broil: 99; see also zh < tsyek
zh tsyejH < *ke[t]-s mad (dog): 77, 276
zh tsyek < *tAk roast, broil: 226; see also zh < tsyaeH
zh tsyijH < *ti[t]-s arrive: 291
zh tsyit < *t-lit substance, solid part: 164, 290
zhng trjuwng < *tru center: 60, 250, 306; see also zhng <trjuwngH
zhng tsyowng < *to bell: 244
Appendix of reconstructedforms 377
zhun drjwenH < *N-tron-s what has been transmitted: 282; see also
chun < drjwen
zhu trwij < *truj pursue: 293
zhu drwijH < *m.lru[t]-s fall down: 133
zhn tsywinX < *tur water level: 295
zhu traewk < *trawk high; splendid: 99, 297
zhu draewk < *lrewk wash: 299, 300; see also zho < draewH
zhu draewk < *[N-t]rok muddy: 81
zhu traewk < *C.trok chop, cleave: 186
zhu traewk < *m-t<r>ok to peck: 243, 244
zhu trjak < *t<r>ak to place: 61, 99, 142, 225; see also zh < trjoH
zhu trjwejH < *trot-s stitch (n.?): 281
zhu tsrjak < *[ts]rak to cut off: 225
zhu tsrjwet ~ tsrweat < *s-[k]rot sprout (v.): 281
zhu tsyak < *tak cut, hack: 320
z tsij < *[ts]ij property, means of living: 90, 401n91
z tsri < *[ts]r black: 229
z tsijX < *[ts][i]j older sister: 90
z dziH < *m-dz()-s breed, love (v.); character: 88, 90, 178
z dzijH < *s.[b]i[t]-s to follow; from: 142
z dzijH < *s.[b]i[t]-s self (adv.): 90, 142
zng tsowng < *[ts]u ancestral temple: 250
zu tsuwX < *[ts]o run: 242
z dzuwk < *[dz]ok clan: 243
z tsjowk < *[ts]ok sufficient: 244; see also z < tsjowk foot, j
< tsjuH
z tsjowk < *[ts]ok foot: 243; see also z < tsjowk sufficient, j
< tsjuH
z tswit < *[ts]ut finish, die: 293, 294; see also z < tswot
z tswot < *[ts]ut soldier: 294; see also z < tswit
zu tsjweX < *[ts]oj beak: 392n92
zu dzwojX < *[dz]uj crime, offense: 293
zu tswajH < *[ts]ot-s collect; most: 281
zu tswijH < *C.tsu[t]-s drunk (adj.): 186, 293
zn tswin < *[ts]u[n] follow (a road): 294
zn tswon < *[ts]u[n] honor (v.): 294
zu dzwaH < *[dz]o[j]-s seat (n.): 197
zu dzwaX < *[dz]o[j] sit: 197, 279, 394n7
{ NOTES }
Chapter1
1. Although their alphabetic scripts reflect phonology in a relatively direct way, Latin, Greek,
and Sanskrit must also be reconstructed, in a sense. Using any written text presupposes a recon-
struction of what its written symbols stand for, which is not always transparent even with alpha-
betic scripts.
2. We adopt the term Kra-Dai proposed by Ostapirat (2000) in place of the traditional
Tai-Kadai, since to Thai speakers, Tai-Kadai evidently sounds unintentionally funny, mean-
ing something like Tai, or whatever (Montatip Krishnamra,p.c.).
3. For clarity, we will use text to refer to a linguistic entity represented in one or
more physical documents, and document to refer to a physical object that represents
a text. Thus the Shjng is a pre-Qn text, but only fragments of it exist in extant pre-Qn
documents.
4. Karlgren did publish a slightly revised version as Grammata serica recensa (hereaf-
ter:GSR), largely responding to Chao (1941), but there were no essential changes.
5. The use of the character hu < xwaeH < *q<r>aj-s transform as a phonetic
element to write hu < xwae < *qra flower (n.) reflects the Hn-time merger of OC
*-raj and *-ra in syllables with pharyngealized onsets; it illustrates the fact that the standard
script established in Hn sometimes reflects Hn-time phonology rather than Old Chinese
phonology. See the discussion in section 2.3 below. Note that we write hu < xwaeH,
as if the standard Mandarin form hu was a direct descendant of the MC form xwaeH.
Strictly speaking, neither standard Mandarin nor any other spoken variety of Chinese can
be descended from Middle Chinese, because, as discussed below, we do not believe that the
information in Middle Chinese written sources accurately represents one single spoken dia-
lect. In the majority of cases, the Mandarin form can be predicted from the Middle Chinese
form, but even this tendency has many exceptions. However, since our primary focus is Old
Chinese rather than Mandarin or Middle Chinese, for convenience and conciseness we con-
tinue to use the < notation after modern Mandarin forms, even when (as often happens)
the Mandarin reflex is not what would be predicted.
6. Cf. Proto-Tai *m.lec D grain (Pittayaporn 2009), *ml/ret D in Fang-kuei Lis reconstruc-
tion (Li 1977:93, 269). Note:the Old Chinese coda could be either *-t or *-k; we indicate this
uncertainty by writing *m.li[t], with square brackets around the*t.
7. Some users of preliminary versions of our reconstruction have found reconstructions such
as j < *[k](r)aj cumbersome and confusing, and have wished for a simpler, more user-friendly
notation. In our view, the visual awkwardness of our notation is compensated for by its rela-
tive faithfulness in representing our degree of confidence in different elements of reconstructed
forms; and it is a reminder that our knowledge is necessarily incomplete. But for users of our
380Notes
system who deem it appropriate for some purposes or some audiences, a simplified notation
can be generated by (1) removing all parentheses and what is inside them, and (2) omitting
square brackets. In this simplified notation, the reconstruction of j becomes *kaj instead of
*[k](r)aj. For serious philological or comparative research, however, the full, less user-friendly
version is more realistic and is to be preferred.
8. Except in cases where the author is well known under a romanized name in
English-language publications, we write all Chinese and Japanese personal names with sur-
name first, ashere.
Chapter2
1. The main exception is that the position of a syllable in the rhyme tables can sometimes
clarify an ambiguous fnqi spelling found elsewhere. (On fnqi spellings see the discussion in
the text of this chapter.)
2. In the Gungyn, the pngshng section is divided into two parts, shng png upper
png and xi png lower png, but this is simply for convenience, since there are more
words in the png tone than in any of the other tone categories. This has no connection to the
later tonal split between upper png or ynpng (corresponding to tone 1 of Mandarin) and
lower png or yngpng (corresponding to tone2).
3. The method of establishing equivalence classes by linking fnqi spellings together was
first used by Chn L (18101882) in his Qiyn ko([1842] 1995). It was
later used by Karlgren in his reconstruction of Ancient Chinese (our Middle Chinese).
4. Some remarks on notation:we use the McCune-Reischauer romanization for Korean,
with phonetic transcription in the International Phonetic Alphabet; we use the Hepburn
romanization for Japanese. For Vietnamese, we give the standard Quc ng orthography,
with phonetic transcription in the International Phonetic Alphabet. The Vietnamese initial
- (IPA []) is known to result from a sound change [t] > [] within Vietnamese; see Ferlus
(1992:115). Numbers indicate tones in the system devised by Yuen Ren Chao, with 5 repre-
senting the highest pitch and 1 the lowest. For Vietnamese, we represent tone categories in
a conventional system of letters and numbers: in early loanwords from Chinese, which we
cite frequently, the letters A, B, C, and D regularly correspond to the png, shng, q, and
r categories of Middle Chinese respectively, except that Vietnamese tone Asometimes cor-
responds to Middle Chinese qshng. The numerals 1 and 2 after the letters represent
upper- and lower-register tones respectively. (In the Sino-Vietnamese layer of vocabulary,
systematically borrowed during the Tng dynasty [618907], the B and C tone categories are
reversed, so that tone B corresponds to Middle Chinese qshng, and tone C to shngshng.)
The fact that d virtue and d obtain, both MC tok, have different pronunciations
in both Sino-Korean and Sino-Vietnamese probably results from their having been borrowed
from Chinese at differenttimes.
5. Unless otherwise specified, Middle Chinese forms are based on the Gungyn (Y
Niyng2000).
6. Unless otherwise specified, quotations from Chinese classical texts are from the online
Hn j dinz wnxin zliok (Zhngyng ynji yun2013).
7. Jerry Norman and W. South Coblin (1995) have criticized the excessive reliance on
Middle Chinese written sources in Chinese historical linguistics. When it comes to dialectology,
we completely agree with their critique:Middle Chinese cannot be used as a surrogate for the
Notes381
data found in modern dialects, nor is it even a suitable framework, we believe, for investigating
modern dialect history, where what is relevant is how the dialects are related to each other, not
how each individually is related to the Middle Chinese system. But when reconstructing the
earlier history of Chinese, the Middle Chinese sources, although not sufficient in themselves,
provide crucial evidence that is available nowhereelse.
8. We put quotation marks around retroflex and palatal here because the choice of terms
suggests a particular phonetic interpretation or reconstruction, which is probably correct for
some but not all varieties of Middle Chinese. (Southern varieties seem not to have distinguished
MC tr- from MC t-, for example; see Pulleyblank 1984:168.) But like the category division-II,
these terms too can be defined distributionally, without reference to any particular phonetic
interpretation of Middle Chinese or to the rhyme tables. In our notation, the retroflex initials
are those written with an -r- (mnemonic for retroflexion), and the palatals are those written
with a -y- (mnemonic for palatal articulation). These notations can easily be mapped onto other
phonetic interpretations if desired.
9. There is one important exception to this principle:some of the words with Tsr- initials and
division-III finals changed during the Middle Chinese period to division-II finals; e.g., shng
is spelled as MC srjaeng (division III) in the Wng Rnx version of the Qiyn, but later was
spelled as MC sraeng (division II). From the point of view of Old Chinese, words like this are
type B, in spite of the division-II final. See the discussion in section4.1.1.
10. The Gungyn and the Qiyn differ slightly in the names and arrangement of their
rhymes, but it is conventional to use those of the Gungyn. In some respects, however, our
notation is more consistent with the arrangement of the Qiyn than with that of the Gungyn;
for example, the Gungyn puts b wave in the G rhyme as if it had the final -wa, but
the Wng Rnx has only one rhyme G, corresponding to the Gungyns G and G,
and the fnqi spelling of indicates MC pa rather than pwa; so in our notation we write
as pa, following the Qiyn, not pwa as the Gungyn implies.
11. The conditions for the partial merger are that OC *-a merged with *-o after nonpharyngeal-
ized labial or labialized initials:thus f < *p(r)a axe merged with f < *p(r)o reposi-
tory as MC pjuX, and j < *[k](r)a carpenters square merged with j < *[k](r)o
(a kind of tree) as MC kjuX; see section 5.4.1.1.
12. The formula turns out to be this:if a rhyme group contains n words, then there are 2n1
n 1 ways of dividing it into two rhyme groups (assuming that each rhyme group should contain
at least two words).
13. Criteria for deciding whether a given set of rhyme data does or does not support a
given hypothesis are discussed in Baxter (1992:97128). Both the relative frequency of the
two groups being tested and the length of the rhyme sequences involved must be taken into
account.
14. If the basic meaning of wng < *qo-s earthen jar was container, perhaps rng <
*[](r)o contain is etymologically related; another possible reconstruction would be rng
< *N-q(r)o, at least in intransitiveuses.
15. Karlgren simply projected MC sy- back to Old Chinese, reconstructing *i at. Baxter
(1992:786) reconstructed *h(l)jet as a kind of default; in our current notation this would cor-
respond to *l et or *qet. Zhngzhng reconstructs *hljed (2003:458).
16. The example on the left is from the Gudin version of Z y (strip 17, GD
18); the one on the right is from the Shnghi Museum text Kngz Sh ln
(strip 10, SB1:22).
382Notes
17. The examples of j are from the Gudin document Y cng sn(strip
5, GD 97) and Jng jin ni zh (strip 2, SB 5.19); the examples of s
are from the Gudin version of Z y (strip 20, GD 18) and Kngz Sh
ln(strip 5, SB1.17).
18. The notation *<r> indicates that we believe the *-r- is an infix. This infix is often found in
words for things that come in groups of more than one; see section 3.3.2.6below.
19. The Mn word for field is often identified with chng < zying ridge between fields,
but Norman suggests that the etymon may instead be cng < dzong layer, level, so named
because most fields in [Fjin] are terraced (1996:31). We agree:the Mn word reconstructs
as pMn *dzhn A, based on Yngn /ts 2/ and Gizh /ts 2/ (Norman 1981:58; Dng
Xingzhng 2007:370), both central Mn dialects that distinguish between pMn *dzh- and
*dh-. Both the Mn initial *dzh- and the final *-n appear to correspond better to MC dzong
than to MCzying.
20. There is also a significant sub-layer of early loanwords that shows the Vietnamese
ngang-huyn tones (A)for Chinese qshng.
21. As mentioned above in note 2 of Chapter1, we use Kra-Dai, the name suggested by
Ostapirat (2000), in preference to Tai-Kadai, as the name of this family.
22. Li (1977) has Proto-Tai initial *hm- corresponding to Pittayaporns*q.m-.
23. The works of Shn Do, or Shnz , survive only in quotations such as this; for
details see Thompson (1979). The passage occurs in chapter 12 of the Xnz, Fi shr z
.
24. The nasal coda in nong, the Middle Chinese reading of nng, is probably second-
ary:the word rhymes as *- in Ode 220.2 and possibly in Ode 257.10.
25. Although we tentatively reconstruct yn with a coda *-r for Old Chinese because of
rhyme evidence (Odes 192.12, 199.1), it is reasonable to assume that Yng Ling was quoting a
version of Shn Do that was written after this *-r had changed to[n].
26. In other contexts the character is also read yun < MC hjwen in modern Chinese, which
is now the more commonly encountered pronunciation.
27. These include yn mng for yu mng in the biography of the Eastern
Hn scholar Fng Yn in the Hu Hn sh , cited by W Chngyng
([1873] 1956), and two cases of yn b for yu b, literally to have a patch, that
is, to make a contribution to knowledge. At least one form of b had a nasal prefix, as shown
by pHM *mpjaX mend, patch (see section 4.5.5.1). Both examples are from biographies in
the Hu Hn sh by Fn Y (398446), in passages purporting to represent
documents of the Eastern Hn dynasty (25220). These interpretations are from Yng Shd
(1954, cited in Y Mn 1992:68) and the Qng scholar W Chngyng (cited in Y Mn
1992:67).
Chapter3
1. There is one exception:as noted earlier (chapter2, note 9), a sound change was going
on in the Middle Chinese period by which some division-III words with Tsr- initials shifted to
division II:from an Old Chinese point of view, these are type-B syllables.
2. By Northern Mn, we mean the dialects designated as Mnbi by Chn Zhngti
and L Rlng (1991)that is, the dialect group that includes Jinyng , Jinu ,
and other dialects that show a systematic softening contrast in syllable onsets, not found in
Notes383
other varieties of Mn. In an older terminology, the Mn dialects were divided into only two
groups: Northern Mn (Mnbi ), typified by Fzhu, and Southern Mn (Mnnn
), typified by Ximn. In the newer terminology, five subgroups are identified:along the
coast, Fzhu is now assigned to the Eastern Mn group (Mndng ), Ximn still belongs
to Southern Mn, and between them is the P-Xin group (including Ptin and
Xinyu ). In the interior are Northern Mn and Central Mn (Mnzhng ). See L
Rlng and Chn Zhngti (1991) for details.
3. Baxter (1992) used capital letters to mark apparently unpredictable cases of velar
palatalization.
4. Baxter (1992) represented this vowel as a barred-i *, which had the disadvantage
of being both unfamiliar to nonspecialists and easy to confuse with ordinary *i. Starostin
(1989) wrote this vowel as *, as we do here. Zhngzhng (2003 and elsewhere) writes
itas*.
5. Syllabic stops occur in Berber (Dell and Elmedlaoui 1985)and other languages.
6. We write *S- instead of *s- to mark the irregular initial development: from *s-t we
would expect MC sying, as in example (148). For this notation, see section 4.4.3.1.
7. The character was originally not plus , as the Shuwn says, but rather plus
; the elements and were often interchanged (H Lny 1998:410). Karlgren, not recog-
nizing this, treated as a separate phonetic element (GSR 1187a). The Shuwn also records
a gwn ancient character for {}, composed of plus (SWGL 3236a). See Sagart
and Baxter (2009).
Chapter4
1. Zhngzhng cited examples from Drung (Chinese Dlng , ISO 6393 code duu);
Starostin cited examples from Mizo (= Lushai, ISO 6393 code lus). These correspondences
to Chinese types Aand B may well be valid; but even if they are, they do not necessarily mean
that the Old Chinese distinction was also one of vowel length. See our discussion in section 2.7
on the role of Tibeto-Burman evidence in reconstructing Old Chinese.
2. Our notation differs from Normans, however:he indicated pharyngealization by an apos-
trophe before the initial consonant, while we use the IPA symbol (U+02E4, MODIFIER
LETTER SMALL REVERSED GLOTTAL STOP) as the last element of the onset but before
medial *-r- (if present), as in the bottom row of Table4.2.
3. The most widely reported acoustic effects of pharyngealized (emphatic) consonants on
the following vowel in spoken Arabic are raising of F1 and lowering of F2 (Shar and Ingram
2010 and references therein). Raising of F1 is indicative of vowel lowering.
4. Jakobson argued that from a phonological point of view, the Arabic uvular consonants
q- and - [] are pharyngealized velars ([1957] 1971:515518). By Hn times, the original Old
Chinese uvulars would have disappeared; see section4.3.
5. This is the Chnqi passage as quoted in the Gngyng zhun; in the received version
of the Chnqi, the text is the same, except that the dukes consort is referred to as Jng Yng
instead of Qng Xing.
6. In his discussion of these terms, Zhu Zm concludes that ni refers to
hngyn broad sounds and wi to xyn thin sounds, terms that generally corre-
spond to type-A and type-B syllables, respectively. He also cites examples to show that
hunq [spoken with] slow breath and jq [spoken with] fast breath are generally
used in the sameway.
384Notes
7. Since this change happened after the original Old Chinese uvulars had changed to the
various laryngeals huyn of Middle Chinese, there was no merger between the original
uvulars and the ones from original pharyngealized velars: for example, *q- had already become
[] before *k- became[q].
8. In the orthography of Alexandre de Rhodess dictionary (1651), from which modern
Vietnamese orthography developed, orthographic gi- (now pronounced [z] in Hanoi) probably
represented a voiced palatal fricative, following the spelling conventions of Portuguese; nothing
velar is implied by the letterg-.
9. A second much more recent palatalization of velars before palatal vowels occurred in
many modern dialects, including Modern Standard Chinese, e.g., MC kim > jn [tin 1]
gold, MC khwet > qu [ty1] break; defective. This change is outside the scope of our
discussionhere.
10. The failure of *- to palatalize to ny- in *et-s > ngjiejH > y to plant is unexplained,
but it is worth noting that the Gungyn has no such syllable as nyejH, which would be the
expected result.
11. To be sure, there have been some attempts to go beyond this traditional approach; see
footnote 19below.
12. The early Chinese loans in Vietnamese (probably borrowed before the breakup of
Proto-Vietic) should not be confused with the Sino-Vietnamese layer of character readings from
Chinese, which were introduced during the Tng dynasty (618907) and which have little direct
relevance for Old Chinese reconstruction. Unless otherwise indicated, none of our Vietnamese
examples are from the Sino-Vietnamese stratum.
13. Here and below, Jinyng data are from Norman (1971, 1973, 1974a, 1981, 1982,
1986, 1991, and 1996); Jinu data are from Norman (1973, 1981, 1986, and 1996);
Shbi data are from Norman (2000) and Akitani (2004); Hpng data are from
Norman (1995).
14. Jinu tone 3 and Hpng tone 4 are accompanied by glottal constriction, indicated here
by[].
15. We thank Katia Chirkova for valuable information on cluster devoicing in Tibetan dialects.
16. Here and below, Proto-Hmong-Mien (pHM), Proto-Hmongic (pHmong), and Proto-Mienic
(pMien) reconstructions are from Ratliff (2010) unless otherwisenoted.
17. Lindncn data are from Norman (2002); Mixin data are from Norman (1989)
and Bijng dxu (2003).
18. Fzhu data are from Fng izhn (1998); Ximn (Amoy) data are from
Douglas (1899); Wxing data are from W Ynj and Shn Ruqng (2010). The Fzhu
forms in square brackets do not follow the stated correspondences and are presumably literary
pronunciations.
19. The only other systematic attempts we know of to incorporate such distinctions in an Old
Chinese reconstruction are Benedict (1976, 1987)and Starostin (1989). Benedict (1987:46)
proposed to derive Normans pMn *-k- and *-g- from Proto-Chinese (PC) *s-k- and *s-g-;
he drew Normans pMn *gh- from PC *g-, and Normans pMn *g- from PC *C + g-. Starostin
projected parts of Normans Proto-Mn reconstruction back into Old Chinese:he reconstructed
pMn *b- and *bh- as Old Chinese *b- and *bh-, for example; he also projected Normans
voiceless resonants back to Old Chinese (1989:5965). He did not attempt to account for the
Mn softened initials. Where MC sy- shows connections with OC *t-, Starostin regarded MC
sy- as a dialect variant of MC tsy-, the regular reflex of OC *t- (1989:159160).Pulleyblank
Notes385
(1973) suggested that Normans Proto-Mn voiced aspirates could reflect Old Chinese voice-
less aspirates preceded by a prefix *-; thus OC *p- > Normans pMn *bh-, but he offered
no further evidence or arguments.
20. For Kra-Dai, see especially Ostapirat (2000) and Pittayaporn (2009); for Hmong-Mien,
see Ratliff (2010); for Austroasiatic, especially Vietic or Viet-Muong, see Ferlus (1982,
1996,1997).
21. The hypothesis that a presyllabic element was responsible for spirantization in Vietnamese
was first suggested by Haudricourt (1965) who attributed it to an r- prefix, and later expanded at
about the same time by Ferlus (1976) and Thompson (1976).
22. We are grateful to Michel Ferlus for helpful discussions on these points and for extensive
assistance with data from Vietic languages. Unless otherwise noted, our examples from Vietic
languages other than Vietnamese were provided by Ferlus.
23. Unfortunately, very few words with Old Chinese voiceless nasals seem to have been bor-
rowed into Hmong-Mien. Apossible example treating OC *n- as pHM *hn- is *na()-s >
syangH > shng give food; food; cf. pHM *hnraH rice (cooked). But the semantics are not
straightforward.
24. In Sagart and Baxter (2009), we proposed that the shift from uvulars to Middle Chinese
velars was conditioned by loosely attached presyllables. But that view has had to be modified
because of examples like kjaengH > jng mirror. That jng had initial *q- is supported by
the related word *qra > 'jaengX > yng shadow (n.). The spirantized initial in VN gng
[z A1] mirror shows that kjaengH had a preinitial in Old Chinese (see section 4.2.2.1).
But if kjaengH were from OC *C.qra-s, we would expect softened *-k- in Proto-Mn; and in
fact the Proto-Mn initial for mirror is *k-, as shown by Jinyng /kia 5/ and Jinu /kia 5/.
(Proto-Mn *-k- should go to zero in Jinu and possibly in Jinyng). Proto-Mn *k- is, however,
compatible with OC *C.q-: *C.qra-s > kjaengH is thus the only reconstruction compatible
with all the evidence. Similarly, *C.qo > kuwng > gng father; prince, *C.q<r>a-s >
kaeH > ji price and *C.q > kojX > gi change (v.) are reconstructed with a uvular initial
because of xishng and/or etymological evidence, and with a tightly attached cluster because they
have Proto-Mn *k- rather than softened*-k-.
25. VN c [k D2] obtain, get has the expected initial but has an unexplained low-register
tone. The final is also unexpected:in early loans, VN -c [k] generally seems to correspond to
OC *-ak (as in thc [tk D1] meter for *tAk > tsyhek > ch foot (measure)). For *tk
> tok > d virtue (which is homonymous with *tk > tok > d obtain in Old Chinese as far as
we can tell), Vietnamese has c [kD1], with the expected reflexes.
26. Contrary to Karlgren (GSR 561), *l j > syij > sh corpse is not a phonetic in
*[q]ij > *xij > syijX > sh excrement; see section 5.5.5.1.
27. This comparison and the supporting material are drawn from Norman (2006:136).
28. Karlgren assigned yng, qing, and jing to three different phonetic series (GSR
732, 712, and 711), but the Shuwn says that yng is phonetic in both qing and jing
(SWGL 1571b, 5521a).
29. The picture is complicated by the fact that pMn initial *- and *- are difficult to dis-
tinguish from each other in Normans reconstruction. According to Norman, their segmental
reflexes are the same, except that Jinyng sometimes has /h/ instead of zero as the reflex of
*-. The only other way to distinguish them is that *- is supposed to have the same tonal
reflexes as voiced resonants like *m-, while *- has the same tonal reflexes as softened voiced
initials like *-b-. But according to Norman (1974a), the only unambiguous difference in
386Notes
thetonalbehaviorof these two types is that in Jinyng, tone *A syllables with initial *- or
*m- should go to tone 2, while those with initial *- or *-b- should go to tone 9; also, *- in
tone *B sometimes goes to tone 3 in the four dialects discussed (Fzhu, Ximn, Jinyng, and
Yngn), while *- would not (Norman 1974a:2932). Additional research on Mn dialects is
needed to resolve these issues (see section6.4).
30. The case of {} wing is especially complex because of the unusual xishng connec-
tions and possible dialect variation involved; the reconstruction that best accounts for all the
evidence seems to be *rp. The uvular initial is supported by the phonetic *(r)k-s > yiH >
y different, which is also written as (though not in pre-Qn documents, as far as we know),
with the phonetic *(r) > yiX > y cease; already (not distinguished in the early script from
*s-[] > ziX > s sixth earthly branch). But wing is also written as , with the phonetic
*k.rp > lip > l stand (v.), supporting a prevocalic *-r- and suggesting that the final MC -k
was originally *-p. Moreover, in oracle-bone inscriptions, the graph for wing is used consis-
tently (with or without r day added) to write {} y < yuwk next day, later also written
as and (Zho Chng 1988:230231). Here MC yuwk next day could regularly represent
an earlier *(r)uk, but again, the phonetic suggests that there was a prevocalic *-r- and that
the MC -k is from an earlier *-p. Note that in the early script, the graph is used not only for
{} l < lip < *k.rp stand (v.), but also for {} wi < hwijH standing, position, for which
the only viable reconstruction (given the Middle Chinese form and the phonetic ) seems to be
*rp-s. Now we assume that {} wi < hwijH standing, position is etymologically related to
{} l < lip < *k.rp stand (v.), but we know of no regular morphological process that would
connect MC lip and hwijH. Rather, we conjecture that *[]rp-s > hwijH > wi standing,
position is a contraction of something like *-rp-s, where the first syllable * may be
yu < hjuwX < *[] (normally have, exist), and the second syllable is a form of the root of
{} l < lip < *k.rp > lip stand (v.). (The function of yu as a prefix is poorly understood,
but it is found in such expressions as yu Shng for the Shng dynasty.)
If we reconstruct {} y < yik wing as *rp, then the dialect change of final labials
to velars (section 5.7), along with plausible assimilations and dissimilations, could produce
the Middle Chinese forms involved: e.g., *rp > *rp > *rk > yik for {} wing, and
*rp > *rup > *rup > *ruk > yuwk for {} next day. Note that is also phonetic in
*rp > yuwk > y shine, gleam, also read MC hip according to the Gungyn; MC hip
would be the regular reflex of *rp. (It is necessary to reconstruct *- in order to account
for MC h- here, as in *[](r)am > hjem > yn burn, blazing and *C.[](r)m > hjuwng
> xing bear (n.); the development of bear is similar but not identical to that assumed for
*rp > yuwk > y the next day.) In the earliest documents, the phonetic (representing
either {} or {}) is not used for either wing or next day; perhaps it became a suitable
phonetic for *[]rp only after *-rp-s standing, position had contracted to *rp-s.
31. Although the graph for salt supports the reconstruction of *-r-, it occurs rather late,
being first attested in the Qn strips from Shuhd . Earlier forms for salt resemble
*N.r[o]m > dom > tn extend, spread (J Xshng 2010:867); hence our conjecture that the
original vowel maybe*o.
32. OC *r()- thus corresponds to Normans Proto-Mn *l-. However, it is possible that this lat-
eral reflex actually developed after the breakup of the Mn branch as a result of a sound change
spreading from the central regions.
33. For the reconstruction of *-aj in *mraj > meaX > mi buy, see the discussion in sec-
tion 5.5.2.1.
Notes 387
34. Most varieties of Mienic have an initial velar-lateral cluster in this word, e.g., Mun
/kla 2/ (Lnjn dialect), but this does not necessarily reflect a velar onset in Old
Chinese:it could have originated as an attempt to represent an onset like [dl] in the donor
variety of Chinese, from original *lr-; cf. /tla 2/ in the Mien dialect of Xizi town-
ship, Hku Yo autonomous county, Ynnn province (Wng and Mo 1995:349).
Similarly, cf. pHM *lw for *C.law > daw > to peach, Xizi Mien /tlau2/.
35. That there was a dialect development of OC *l ()- > MC x- was originally pointed out by
Pulleyblank (19621963:117118), where his *- corresponds to our *l ()-.
36. The text is quoted from Dien (1957:285), whose translation is adapted here. The Ko
shng mentioned is the Ko shng Qiyn, a work on pronunciation by
Zhng Jin of the Tng dynasty, frequently cited by Huln.
37. The treatment of final *-r as -n in *nar-s > xanH is also characteristic of interior dia-
lects, as opposed to coastal dialects, where we have evidence of *-r > *-j; see section5.5.1.
38. Starostin (1989) did reconstruct Old Chinese *mh-, *nh-, etc., where Norman (1973)
reconstructed them for Proto-Mn, which may be one of the reasons he chose to reconstruct
*sm-, *sn-, etc. where we reconstruct voiceless resonants like *m()- and *n()-.
39. Proto-Hmongic tone *C is the regular reflex of pHM syllables with final *-k in tone *D;
see Downer (1967:590), Ratliff (2010:31).
40. In Ratliffs reconstruction (2010:86106), Proto-Mienic glottalized voiced stops reflect
earlier prenasalized voiceless stops; thus pMien *g- < pHM *k- or *q-. The notation
*NK-- means that the onset could be either *k- or *q-, but there is not enough evidence
to choose betweenthem.
41. Treating these pairs as intransitive roots with voiced initials, devoiced by the causative *s-
prefix (Mei 2012), is not an open option, since we show in section 4.4.3.3 that *s- assimilates in
voicing to a following obstruent (Sagart and Baxter2012).
42. Rc /tk/ evidently reflects the qshng variant gj+nH of jn, because of the huyn
tone (A2) in the corresponding Vietnamese form gn near. Chinese qshng regularly cor-
responds to the ngang (A1) and huyn (A2) tones in one early layer of Chinese loanwords to
Vietnamese.
43. Ratliff compares pHmong *khuw D crooked with *k(r)ok > khjowk > q to bend,
bent, but as she points out, the Hmongic tone in *khuw D crooked points to Proto-Hmong-Mien
final *-p or *-t, not *-k (2010:87n). Perhaps the Proto-Hmongic form represents a form of
q < khjut < *[k]ut bend, subdue:either *N-[k]ut (which would also produce MC khjut) or
a hypothetical Chinese intransitive form *N-[k]ut > gjut bent that is not reflected in Middle
Chinese written sources (as far as weknow).
44. The developments described in Table4.35 are those of Ximn for syllables with OC non-
pharyngealized *b (or *b- < *N.p-) in the onset. In other varieties of Mn, the tonal developments
vary:for example, in Fzhu, syllables with tone *C (qshng) with Normans voiced aspirate
initials (e.g., *bh- < OC *m.p-) have tone 5 (upper qshng), not tone 6 (lower qshng) as might
be expected (Norman 1974a). Also, at some positions of articulation, the segmental reflexes for
pharyngealized and nonpharyngealized onsets may be different in Proto-Mn. For instance, OC
*m.k- becomes Normans pMn *gh-, but OC *m.k- becomes his pMn *-; see examples (476),
(477), and (478). Note that Normans Proto-Mn *b-, *bh-, and *-b- types (as well as his *-p-) all
become voiceless aspirates in Show (Norman 1974b, 1982)and in the closely related Hpng
dialect (Norman 1995). This suggests that these dialects never underwent the first devoicing
(in stage 4 of Table4.35), but only the second devoicing (in stage6).
388Notes
45. The tone *A in Proto-Kra and the tone category in Proto-Hmong-Mien would normally
correspond to Chinese pngshng; we have no explanation for this discrepancy.
46. The Jngdin shwn gives the reading MC kop for in several passages, e.g., in the
expression hjn < kop-kj+nX to drink from cups made from splitting a single gourd, part
of the marriage ceremony described in L j:Hn y (JDSW217).
47. The Gungyns reading kaep is irregular; we would expect keap. The reading kep is from
Jngdin shwn (JDSW164).
48. Another possible Proto-Mn reconstruction for xi narrow is *ap D, with initial *-;
as noted earlier, Normans * and * are often difficult to distinguish, especially if forms are
missing for crucial dialects.
49. Proto-Mn may have overlapped in time with Early Middle Chinese and may contain a
layer of early Middle Chinese loanwords. We have already seen an apparent instance of layering
in the Proto-Mn treatment of OC *l- (section 4.3.4).
50. The standard traditional character for yn rhyme is , which has the reading hjunH
in the Gungyn; if we relied on this evidence alone, we would be led to reconstruct the word
as *[n]-s. But earlier Qiyn manuscripts give instead the reading hwinH, and is a
late character. The word {} was earlier written as , which leads to the reconstruction
*[m-q]<r>i[n]-s; {} is also written as , now used as the simplified form. The reconstruc-
tion *[m-q]<r>i[n]-s makes etymological sense, connecting {} to a root meaning even or
perhaps cyclical (as in the other forms in example (489)), with an *<r> infix indicating repeti-
tion, and an *m- prefix that here can be interpreted as deriving an instrumental noun (section
3.3.2.2):that by which something is made harmonious or even. (We know of no clear cases
of the word being used to mean rhyme as such in pre-Qn documents.) J Xshng says that
the graph originally depicted a potters wheel, {} *C.qi[n] > kjwin > jn potters wheel
(2010:741).
51. In Baxter (1992:755, 793), t < thuX < *ta was erroneously reconstructed with
a lateral *hl- (which would be *l - in the present reconstruction). As pointed out by Sagart
(1993b:256), the dzy- in sh < dzyaeX indicates an Old Chinese alveolar, not a lateral.
52. The earliest such example we know of is that the Gudin bamboo-strip version of the
text Wxng (late fourth century bce) has the character y < ngjoX < *(r)a
(normally speak) for y < ngjoX < *m-q(r)a in a widely quoted line from Ode 260.5:
b wi qing y He does not fear the strong and the oppressive (GD 33, 150). In
quotations of this line in received texts, the word is generally written as *m-[q](r)a > ngjoX
> y ward off; withstand (related to oppressive). From the context and the way the line is
quoted in received texts, it appears that oppressive (which we reconstruct with a uvular) is the
original meaning. If this analysis is correct, then the writing of oppressed with y in the
Gudin W xng suggests that the change of *m-q- > *- had already happened by the late
fourth century in the Charea.
However, several other alleged examples of the interchange of and , cited by Bi Yln
(2008:104105), appear to be flawed.For example, a graphic element sometimes interpreted as
w over yu is actually, according to L Jiho (2004), an early graph for {} *pe[n]>
pjien > bin whip (n.), used as a phonetic element to write {} *[b]ren > bjenX > bin dis-
tinguish, dispute (v.) on strip 9 of the Shnghi Museum text Mn zh fm
(SB 2.25, 2.168) and as a semantic element in early forms of {} *[](r)a-s > ngjoH > y
drive a chariot (which is sometimes written with w as phonetic in the bamboo strips from
the tomb of Zng Hu Y , late fifth century bce). Although {}y < *[](r)a-s drive
Notes 389
a chariot is often written with the character y < *m-[q](r)a in received texts, the words
{} y ward off and {}y drive a chariot are not confused in earlier texts and appear to be
unrelated (Di Jixing 1995, quoted in GG 2.525526).
53. Karlgren (19421946, gloss 1872)also gives examples where is also used for in the
meaning quantity*q(r)a > xjoX > x, further supporting the connection of *[m].qa >
nguX with uvulars.
54. The *r in the Proto-Hmong-Mien form is unexplained; the Middle Chinese division-IV
chngni final -jij normally indicates the absence of prevocalic*-r-.
55. This form is reconstructed on the basis of Mn; there is no reading corresponding to
*m-[d]uk-s in the Middle Chinese written sources. If there were, we would expect it to be
dawH. However, pMn *-u usually corresponds to MC -uw, not-aw.
56. MC zywit for sh reflects a variant form *m.lut; see section 4.5.2.4.
57. For the irregular loss of final *-k in come (probably attributable to the restressing of an
unstressed variant), see Baxter (1992:330) and section 5.4.2.2.
58. One possibility is that our *s.t- > MC sy- was actually *s.t-, while our *S.t- > MC ts- was
actually OC *s.t-. Adialectal distinction is another possibility. This is similar to our use of a
capital *A to mark forms with *a whose reflexes are still unexplained; this *A is not intended as
a seventh vowel (see sections 5.4.1.1, 5.4.1.2, and 5.5.2.1).
59. For , Karlgren gives the Middle Chinese pronunciations tsip and tsrip for to cluster
together, to crowd (Sh), which is the same word as *s.q<r>[i]p > tsrip > j crowded together.
This is based on a single annotation in the Jngdin shwn, for the line
(The wings of the locusts, they are in crowds) in Ode 5.3 (JDSW 54); the only pronuncia-
tion given for in the Gungyn is MC 'jip. The reading tsip probably represents a variety
of Middle Chinese in which tsr-, tsrh-, etc. had merged with ts-, tsh-, etc. (as with *s-ru >
srjuwX > suwX > su old man); see note 68 for a similar example.
60. (SWGL 4026a).
61. The Middle Chinese initial zr- is generally not recognized in traditional phonology,
because in the Gungyn it was not distinguished from dzr-. The fact that there was a contrast
dzr- zr- parallel to dz- z- and dzy- zy- was first revealed by the fnqi spellings in the
manuscript of the Wng Rnx version of the Qiyn, discovered in the Former Palace
Museum in Bijng in the late 1940s; see Baxter (1992:39, 5657).
62. The MC initial in y < hiX (traditional name: Y sn = Yn) is rare except
before a rounded vowel or semivowel (as in *[][r] > hjun > yn say or ~
*<r>en > hjwen > yun round), because the normal source of this initial is OC *-.
Another exception to this pattern is *[]a[n] > hjen > yn (3p locative pronoun), which
like y is a sentence-final particle: when yn occurs sentence-initially, it is read as
*a[n] > 'jen > yn how with an initial glottal stop. In these cases we suspect that the MC
h(j)- initial is the reflex of initial *q- or *- in an unstressed syllable. (Other cases of MC
h(j)- without a following rounded vowel or semivowel are the result of dissimilation, prob-
ably from [] from original *-, as in *[](r)am > *(r)am > *(r)am > hjem > yn
burn, blazing; see section5.7.)
63. Norman (1986:381) reconstructed pMn *dh- rather than *dzh- in x mat, on the
basis of Central Mn dialects that have initial //, but some cases of // in Central Mn, including
this one, appear to be secondary:for example, the Gizh dialect of Shxin has
/tiu 1/ for *tsiw > tshjuw > qi autumn; crop, pMn *tshiu A(Norman 1974a:32, 1981:41;
Dng Xingzhng 2007:362).
390Notes
64. MC bjijH could reflect *-j-s, *-t-s, or *-p-s, but except in the southeast, modern Chinese
forms reflect a form with *-t (which would correspond to MC bjit), excluding the first
possibility.
65. A character with the shape , composed of z and , does occur in Shng inscriptions,
but it is apparently a place name, not the common noun {} b nose (Qi Xgu 1992:9596).
66. Mei (2012:8) has questioned the role of wng as phonetic in sng on the grounds
that in Shng oracle inscriptions, according to Y Xngw (1979:7577), {} sng is written
with the character *[s]a > sang > sng mulberry tree, without the element *ma. While
this is true, *ma does appear in the character for {} in early bronze inscriptions (where
Y Xngw characterizes it as a phonetic element). The fact that *ma is a phonetic element
is further supported by the fact that in the Shnghi Museum texts, the word sng mourning,
burial in the expression sngf mourning clothes is written with di bad on the left,
and mng, MC mang < *ma awn of grain, on the right. di bad is clearly signific, and
*ma can only be the phonetic (SB 4:183). Note that *ma is a type-A word, like . For
details, see Sagart and Baxter (2012).
67. The expression susu soaked rhymes as *-u in Ode 245.7; su (meaning dis-
puted) rhymes as *-u in Ode299.7.
68. The merger of MC Tsr- with Ts- is indicated by many alternate fnqi spellings in the
Jngdin shwn and characterizes many Chinese dialects today; for example, in standard
Mandarin, MC tsy- and tsr- both usually correspond to [t], but some varieties of Mandarin, as
well as Gn, Hakka, and certain Xing dialects regularly have [ts] for MC tsr-, even though
they have a contrasting retroflex [t] for MC tsy- (Sagart 1993a:133134; Coblin 2011:3865).
69. We write square brackets around the *l in , *s-[l]<r>a and *[l]<r>a to indicate
uncertainty about the main syllable initial:it could either be *l or something more complex.For
example, *s-m-l<r>a and *m-l<r>a are also possible, with the *m- prefix typically indicating
volitional action (see the discussion of Mn reflexes below).
70. If the *-r- in *[l]<r>a really is the infix, as we suppose, it suggests that more than one thing
is being removed or that the removal is somehow distributed, and this is indeed frequently the
case with ch in earlytexts.
71. The Shuwns definition (SWGL 6289a) uses the word *qu > xaw > ho weed (v.)
(which occurs as a colloquial term in Mn dialects, e.g., Ximn /kau 1/). Dun Yci explains
the gloss what is used in weeding while standing by saying that in ancient times weeding was
done in a sitting position; the tool used was called a nu [< nuwH < *nok-s]; it had a short
handle ( [nu], ); see Dun Yci ([1815] 1981:707).
72. For the reconstruction of window, see section 4.4.3.5.
73. Some Northern Mn dialects show softened initials for {} hoe (n.):Zhnqin /ty 9/,
Wf /ly 9/ (Norman 1996:34), Shbi /dy 2/ (Akitani 2004:81; Akitanis tone 2 corresponds to
Normans tone 9), implying pMn *-dy A; this would be consistent with Ximn /ti 2/ hoe (n.).
Proto-Mn *-dy A would be the regular reflex of OC *C.l<r>a or *C-m-l<r>a (see section
3.1.4). But when Northern Mn softened initials correspond to Middle Chinese voiced obstru-
ents, they are sometimes secondary (see Table4.14), so further study is needed.
74. Since we do not regard *- as a synchronic suffix in Old Chinese, the root *nr in
*s-nr may be related to, but is not synchronically the same as, the root *nr > nejX > n
to stop. But it is interesting that a similar mismatch occurs in *to > tuwng > dng east,
which we relate to *[C-m-]to > duwngX > dng move but which otherwise has an unex-
pected pngshng tone; see Sagart (2004).
Notes 391
75. The interpretation of the graph for q < tshej wife is disputed, but in any case it is
apparently not a phonetic compound, and its onset is difficult to reconstruct with confidence:one
possibility is *s.l - (see Sagart 1999c:173). The earliest example of the character MC sej used
to write nest recorded by J Xshng is from the Qn site Shuhd (2010:865), so it is
probably a late character introduced fromQn.
76. (Ho Yxng etal. 1989:1015).
77. The unusual correspondence of MC -aewk to OC *-ak, generally regarded as irregular,
results from the change Tsrj- > Tsr-, which affected some varieties of Middle Chinese; see sec-
tions 4.1.1 and 5.4.1.2.
78. [su < *s.a] (SWGL 2995a).
79. MC 'jw+j and 'jw+jH could also reflect initial uvular *q-, but in that case we would expect
to find the words written with the phonetic *kuj-s > kjw+jH > gu precious; expensive,
which seems to be the usual way to write velar- or uvular-initial syllables with OC*-uj.
80. We are grateful to Li Gulng for helpful discussion on this point. We regard
the traditional use of gu < kjw+j < *[k]j return (v.) to gloss gu < kjw+jX < *k-uj
ghost as a sound gloss based on folk etymology, probably from Hn times or later, postdating
the diphthongization of *-uj to *-wj (see section 5.5.7). The connection is not implausible from
a semantic point of view, but since the two words have different rhymes in Old Chinese (*-j
vs. *-uj), they are unlikely to have had a common origin. The earliest examples of this gloss
that we know of appear in the Shuwn (SWGL 4058a, 100 ce) and in Liz , most of
which probably dates from no earlier than the third century ce (Barrett 1993:299301; Graham
19601961).
81. We do not know any examples of onsets with preinitials *p, *t, and *k and main initial
*s()-, i.e., *k.s()-, *p.s()-, *t.s()-; nor do we know any examples of onsets with two alveolar
obstruents such as *t-t, *t-d,*t-dz.
82. The character , with the phonetic *C.pat > pjot > f fly forth, send forth, reflects
the change of *-p-s to *-t-s. We know of no occurrences of this character in pre-Qn documents.
83. Although the reading q < khjoX is absent in recent modern dictionaries, it is in the
Gungyn and is frequently indicated in Jngdin shwn.
84. The expected reflex of OC *t.g- in Normans Proto-Mn would be *dh-; we suspect that
the pronunciation with *d- reflects borrowing from a non-Mn dialect.
85. A form like *t.ku would normally evolve to a softened initial in Mn, giving zero in
Northern Mn dialects such as Jinyng and Jinu, but Jinu has /kiu 3/ for nine, which is
inconsistent with OC *t.kuunless this word, like ten, does not belong to the inherited layer
ofMn.
86. The similarity of nine and elbow in Sino-Tibetan may point to the practice of count-
ing to ten on the hand and arm, where nine sometimes associates with the elbow, as in certain
Papuan languages of New Guinea such as Foe (Rule 1993; Franklin 2001), where elbow and
nine are the sameword.
87. Proto-Mienic *u C < pHM *nc- could reflect *[t.p] broom with a nasal prefix
and a change of tone category.
88. Pittayaporn (2009:160) reconstructs Proto-Tai *C.dwi:, in which he says that the *-w- is
speculative.
89. The Middle Chinese pronunciation bij for is from the Jyn (Dng D [1039] 1985) and
implies OC *br. The annotations to the Fng yn by Gu P say that it is pronounced like
392Notes
p < phij < *pr great, but we suspect that this is a textual error for the place name MC
bij (sometimes written as p), which is given as a homonym of in theJyn.
90. As mentioned in the previous note, the pronunciation MC bij for would normally reflect
OC *br, but we suppose that here Gu Ps notation for the Jingnn pronunciation in
his time was intended to represent disyllabic *b.r rather than *br.r.
91. The hyphen in Ratliffs reconstruction for the last item indicates the presence in
Proto-Hmong-Mien of a loosely attached preinitial. We would expect pHM *kl- rather than
Raliffs *-r. The reason for this irregularity is notknown.
92. Until recently, it was not clear whether the Proto-Mn initial in mouth was *tsh- or *th-,
because Yngn, the Central Mn dialect that Norman largely relied on to distinguish these
positions of articulation, uses a different word for mouth (/tse 3/, the regular reflex of
*[ts]oj > tsjweX > zu beak, which became the colloquial word for mouth in Mandarin and
other dialects). But Dng Xingzhng (2007:363) reports that Gizh , another central Mn
dialect, has /tyi 5/ for mouthful, which points to pMn*th-.
93. Recall that MC tsyhwejH stands by convention for tsyh- plus -jwejH; see section 2.1.2.3.
94. If *-r and/or *-n became *-t in the Ch area,for example, this would explain why {}
*[ts]e[n] > tshjenX > qin shallow is written with a character composed of shu water
plus *[ts]ret > tsrheat > ch examine on strip 46 of the Gudin version of the text W
xng , which also exists in a silk version from Mwngdu (see Qi Xgus com-
ments in GD 154, note63).
95. Recall that Proto-Mn *- also behaves like a voiced aspirate in Normans system.
96. We might expect pHM *h-, but Ratliff does not reconstruct such an onset (2010:30).
97. For u < *C.(r)o lotus root, Bijng dxu (1993) gives /u 3/ as a literary pronun-
ciation and /u 1/ as the colloquial pronunciation in Mixin. But it is possible that the labels
should be reversed:according to the pattern described in Norman (1989:334335), we would
expect that words with initials corresponding to pMn *h- should go to tone 3 and that those
corresponding to pMn *- should go to tone 1; and that the former are more likely to be from
the colloquiallayer.
98. The phonetic realization of preinitial *N is uncertain. As a loosely attached preinitial it
presumably contained a central vowel; the testimony of Hmong-Mien indicates a nasal element,
but whether that element took the form of a consonant or of nasality on the vowel is uncertain; if
a consonant, that consonants place of articulation is also uncertain. Example (725) below may
indicate that *N was phonetically[].
99. We regard the aspiration in the Proto-Hmong-Mien initial as a by-product of the evolution
out of an original fricative, although lack of parallels makes this difficult to verify.
100. .
101. (Ho Yxng etal. 1989:146) r y is a collection of glosses from per-
haps the third centurybce.
102. The verb li come rhymes in some (perhaps early) parts of the Shjng with *-k (e.g.,
Odes 203.4, 242.1, 263.6) but in other parts with *-. Our interpretation is that this common
verb had a variant in unstressed position that lost the coda *-k; the rhymes with *- represent a
restressing of this unstressed form (as modern English it comes from Middle English and Old
English hit). See Baxter (1992:337338) and section 5.4.2.2.
103. Ximn /tun 2/ and Chozhu /tu 2/ lip appear to reflect pMn *d-; the failure of OC
*d- to palatalize in a type-B syllable is unexplained. It is not certain that the Jinu form /y3/
Notes 393
with softened initial is colloquial; the colloquial word for lip in Jinu is /tsy5 py 6/ skin
of the mouth, interpreted in L Rlng and Pn Wishu (1998:35) as , i.e., pMn *thyi C
bhye Amouthskin.
104. The character composed of m and is attested in the late fourth century bce, as part
of a place name in the Boat tally of Q, the lord of (); see GG 5.811812.
105. We think this Jinyng form corresponds to ; Norman (1996:26) compares it instead to
shn straw mat. For discussion see example (1106) in section5.5.5.
106. Data are from Douglas (1899), Norman (1996), Ci Jnmng (1976), and Fng izhn
(1998).
107. For the reconstruction *[m-q]<r>i[n]-s, see example (489) and note 50 in section
4.4.2.1.
Chapter5
1. It is difficult and perhaps of secondary importance to assign priority for the idea of a
six-vowel system. Baxters teacher Nicholas C.Bodman was working on a six-vowel recon-
struction already in the early 1970s, but it was not fully worked out at that time; solutions to
some of the problems were proposed in Baxter (1977). Baxter, Starostin, and Zhngzhng
Shngfng were at first isolated from each other by the Cold War, the Cultural Revolution,
and Sino-Soviet hostility, and they developed their reconstructions independentlyexcept
that they were all strongly influenced by Li (1971) and by Jaxontov and his rounded-vowel
hypothesis. Baxter and Zhngzhng first met in 1982 in Beijing; Baxter and Starostin had
corresponded briefly but first met in Ann Arbor in 1987 (at which time the manuscript for
Starostin 1989 was already complete). All three were astonished at the striking similarities
of their reconstructions.
2. There is nothing essential in this decision, but the phonological changes from Old to
Middle Chinese turn out to be easier to formulate if we write *-ij rather than *-i, while there
is no such consideration for*-u.
3. Baxter (1992:182183) also allowed *- after voiceless stop codas to account for apparent
alternations in Middle Chinese between shngshng words and rshng words with final *-k, but
we exclude this possibilityhere.
4. This describes the main developments to Middle Chinese, but developments were prob-
ably different in different dialects. In some dialects final *-s may have simply been lost, without
the loss of a preceding voiceless stop (possibly after the change of *-p-s to *-t-s). Such a devel-
opment could be responsible for the Mandarin pronunciations of b nose, which is qshng
in Middle Chinese (bjijH) and in some southern dialects, but which imply a rshng pronuncia-
tion like MCbjit.
5. These are the same consonants that Hung Kn (18861935) identified as the nine-
teen initials of ancient pronunciation (g shng shji ni ), which he regarded as
simplest and most basic, based on their distribution in the Middle Chinese system; see Wng L
(1985:348351).
6. Although reconstructing five vowels plus *- seems to us the most natural way to account
for the syllables considered so far, it would be possible to reduce the number of vowels by
increasing the number of types of codas (or onsets). For example, if we recognized *- and
*- as additional codas, we could reduce the number of vowels to two (*a and *) by writing
394Notes
*-a instead of *-o, *- instead of *-u, and *-a instead of *-e. This is essentially the
move made in Pulleyblank (19771978), which uses only the two vowels *a and *. Although
Pulleyblanks reconstruction is not typologically impossible, we see no reason to prefer it, unless
there are other reasons for preferring a small vowel system. Moreover, in his two-vowel system,
in order to account for the rhyme distinctions observed among finals that we reconstruct with
the coda *-n, it would be necessary to add labialized *-n and palatalized *-n to the set of codas,
with *-an corresponding to our *-on, and *-an corresponding to our *-en, and so forth; and
similarly for the codas *-t, *-m, and*-p.
7. The Younger X version of the Shuwn (see Boltz 1993:435436) adds
This is the same rationale as with zu to sit, that is, t earth is a component of zi
for the same reason as in zu, apparently to indicate the notion of location or being at aplace.
8. Incidentally, removing *ia and *i from the vowel system explains a number of distri-
butional gaps in both Karlgrens and Lis reconstructions: in both reconstructions, there are
syllables like *Kian but no syllables like *Kiang, and no explanation of why they should
be absent. In the six-vowel system, Karlgren and Lis *Kian is *Ken > Ken, parallel to the
velar-final *Ke > Keng, and there is no suchgap.
9. The Chinese terms are chngni sndng and chngni sdng.
10. The table is somewhat simplified so as to focus on the most important differences among
the reconstructions; in fact, the reconstructions in a single row do not always correspond to
each other one-to-one. For example, MC Tan is reconstructed as *Tan, *Tn, and *Tan by
Baxter-Sagart, Karlgren, and Li, respectively, in agreement with the first row, But MC Pan is
reconstructed as *Pan or *Pon by Baxter-Sagart, as *Pwn by Karlgren, and as *Pan by Li;
MC Kwan is reconstructed as *Kan or *Kon by Baxter-Sagart, as *Kwn by Karlgren, and as
*KwanbyLi.
11. Actually, the Old Chinese rhymes do not match the Qiyn rhymes exactly; for example,
the Zh rhyme group includes not only the words of the Qiyns Zh rhyme but also some
words of the Zhrhyme.
12. See Baxter (1992) for detailed arguments. In a few cases (the rhymes with labial codas),
there are not enough rhyme examples to yield a statistically significant result, but we have no
reason to think that syllables with labial codas were based on a different vowel system from the
rest of the language.
13. Karlgrens translation (1950:6869) is Lo! how handsome, the clear forehead how beau-
tiful; / when dancing he is in counting [i.e., perfectly follows the rhythm of the music]; / when
shooting he pierces (the target); / his four arrows succeed one another in a regular sequence, /
so as to prevent (disorder:) violation of the rules. This translation of the fifth line follows the
emendation of fn to bin; Karlgren also offers a translation of this line that follows the
Mo text:his four arrows (revert =) all come (one after the other) to the same place.
14. We have not attempted to give reconstructions in Pulleyblanks system (19771978), for
which full reconstructions are not available. However, in his system, all the words, including the
problematic fifth line, would have the rhyme*-an.
15. Nothing, that is, except the fact that fn, a shngshng word, appears to rhyme with
qshng words. But although the Shjng rhymes generally tend to agree with the tone catego-
ries of Middle Chinese, there are many exceptions, and tone mismatches like this are quite
common.
16. The Shjng rhymes involving fn are not all regular, but it rhymes twice with
yun < hjwonX < *C.an far, which itself rhymes repeatedly with words that can only
Notes 395
be reconstructed with *-a[n] (89.1, 158.2, 165.5, 169.3, 223.2, 254.1). The Middle Chinese
finals of ljwenX, sjwenH, and lwanH all point to *-o[n]; 'jwonX rhymes as *-o[n]
in 94.1, 102.3, and 151.4; kwanH rhymes in 199.7 with xn < xjwon < *qo[n] oca-
rina, whose phonetic component indicates a rounded vowel ( xn < xjun < *qu[n] smoke
(v.); vapor, odor).
17. While *pro[n]-s does not rhyme elsewhere in the Shjng, it is written with the same
phonetic element as lin < ljwenX < *[r]on beautiful in line 1, and words with this phonetic
lun < lwan < *m.ro[n] harness bells rhyme consistently as *-o[n] (in 42.2, 102.3, 147.1,
151.4, and 261.6). Finally, *pro[n]-s change (v.) could well be related etymologically to
lun < lwanH < *[r]o[n]-s disorder, rebellion.
18. Translation (adapted from D.C. Lau 1963):Without what makes it clear, heaven might
split; / without what makes it settled, earth might fly away; / without what makes them spiritu-
ally active, the spirits might come to rest; / without what fills them, the valleys might dry up;
/ without what gives them life, the myriad beings might be annihilated; / without what makes
them noble and high, dukes and kings mightfall.
19. The traditional requirement that the Qiyns Yun rhyme (-jon, -jwon) should rhyme
with Hn (-on) and Hun (-won) is our rationale for writing all three with -on in our
Middle Chinese notation. Although the three are separate rhymes in the Qiyn, they are put
adjacent to each other, which probably indicates that they did rhyme in some dialects. They were
subsequently marked as tngyng usable interchangeably, and all three were combined
as the rhyme (number 13, Yun) in the Pngshu yn rhyme standard that
came to be used for regulated verse. In most dialects, however, MC -j(w)on () came to rhyme
with -j(w)en () and -(w)en () as something like [ien] or [yen]; so in order to rhyme correctly,
poets had to simply look up or memorize which cases of [ien] and [yen] in their own speech had
come from the Yun rhyme of the Qiyn and which had come from Xin or Xin. The
Yun rhyme in the Pngshu yn thus acquired the name gis shsn Yun
(the damned number 13 Yun); see Wng L (1988:128).
20. The only innovations regarding *r in the current reconstruction involve the reconstruction
of initial *r- preceded by certain preinitials, and of *-r in coda position; see sections 4.4, 4.5,
and5.5.1.
21. It appears that in the early script, the same phonetic elements were used to write both syl-
lables of the form *en (e.g., , yun < hjwen < *<r>en round) and syllables of the form
*an (e.g., yun < hjwonX < *C.an far); see Qi Xgu (1985). Later, the forms with
m eye on top came to be used for front-vowel syllables like *en, and the forms without
m for syllables like*an.
22. Although the Middle Chinese reading srjet itself is not certainly attested in the Middle
Chinese sources, it can be inferred from the fact that for the qshng form shi < *s<r>at-s,
the Jngdin shwn repeatedly gives both sreajH and srjejH (in some cases calling srjejH the
older reading). It usually does not give a fnqi spelling for the rshng form, instead saying
r z (as the character [is normally read]). The spelling , implying sr(joX)
+ (l)jet=srjet, does occur (pp.74, 214, 366), but where it does it may be a copying error for
sr(joX) + (l)jejH=srjejH.
23. Although the only reading for shn in the Gungyn is srean, the Jngdin shwn
gives srjen as an alternate reading in a gloss to the r y (JDSW 422). Similarly, for
sho gradually, it gives sraewH but says the old reading is srjewH (JDSW 135). The
qshng reading of shi < sreajH < srejH < *s<r>at-s diminish is given as sreajH in the
396Notes
Jngdin shwn, several times with the note the old reading is srjejH (e.g., JDSW 122);
the reading of s < *C. s<r>r sprinkle is given as sreaX, with the note the old read-
ing is srjeX (JDSW 270; the Gungyn gives the readings sreaX, sraeX, and srjejX); and
so forth. Zhngzhng Shngfngs reconstruction does not take the change Tsrj- > Tsr- into
account; for example, he reconstructs shn < srean mountain as *sreen (2003:455),
which would correspond to *sren in our system. But this conflicts with Shjng rhyming, as
in Odes 189.1, 197.8, and 305.6, where the other rhyme words do not have front vowels. (In
fact we reconstruct shn < srean < *s-rar with *-ar rather than *-an; see section 5.5.1.2.)
Similarly, Zhngzhng reconstructs sh < sreat kill as *sreed, which would correspond
to our *sret. Starostin reconstructed shn < srean as *srn (1989:576), which agrees
with the Shjng rhymes but makes the Middle Chinese reading srean irregular; we have not
been able to find his reconstruction ofsh.
24. If the *-r- here was the collective/plural infix *<r>, which is plausible in this case, then
perhaps there were forms with and without *<r>:*C.[k]o and *C.k<r>o; but both would
give MC kuwX. The presyllable *C. is reconstructed to account for the softened initial in
Northern Mn, e.g., Jinu /e 3/, Jinyng /heu 3/; see section 4.5.5.1.
25. In acute-initial syllables like *Tro and *Tsro, it appears that the *-r- produced a retroflex
initial in Middle Chinese, but *-o developed to MC -uw as expected in type A; the resulting
syllables Truw and Tsruw are treated like Trjuw and Tsrjuw in the Qiyn, so that original
type-A *Tsro appears to have merged with original type-B *Tsru. (It is likely that the -j- in our
Middle Chinese notation is somewhat artificial here. Given the tendency of -j- to be lost after
Tsr-type initials, a syllable like tsrjuw was actually probably something like [tuw] phoneti-
cally; so perhaps this was a merger of [tuw] with [tuw].) Examples: *t<r>ok-s > trjuwH
> zhu beak, *[ts]ro-s > tsrjuwH > zhu wrinkles, both from type A, contrasting with
type-B *tro > trju > zh tree root, stem, *[ts]ro > tsrhju > ch grass for fuel or fodder,
with the regular type-B development *-o > -ju. See section 5.4.5.1.
26. In the traditional list of thirty-six zm (initial consonants), MC ywin and hwin are
regarded as having the same initial consonant, namely, Y. As noted in section 2.1.2.2, in
more modern terminology, and in our notation, they are regarded as different initials because the
Qiyns fnqi spellings clearly distinguish them. MC y- is called Y s or Y, while
MC h- in division-III (type B) syllables is called Y sn or Yn.
27. Probably *[](r) > nging > nng freeze, coagulate is somehow related to these, but
the morphology is unclear. In the r y there is a passage where one version has bng where
another version has nng; see the Jngdin shwn (JDSW417).
28. In certain pre-Qn documents, there appears to have been a tendency for the character
mng to be written differently depending on whether it was being used as a noun or a verb.
For example, on strip 7 of Kngz Sh ln (SB 3.19), {} appears six
times:four times with a small double line below, and twice without. The two tokens with-
out are clearly verbs; the four tokens with appear to be nouns. See Ln Sqng (2003).
29. Based on our assumptions, we would expect a contrast between *Kan > MC Kjon and
*Kran > MC Kjen; but in fact, although MC -jon and -jen are in different Qiyn rhymes,
it is difficult to find reliable minimal contrasts between them with velar or laryngeal ini-
tials. For example, the Gungyn has a syllable ngjon but no ngjen, and it has a syllable
kjon but no kjen. Syllables like Kjon and Kjen had merged by the time of the Late Middle
Chinese rhyme tables (Pulleyblank 1984:71) and had probably already merged in one or
more of the dialects on which the Qiyn was based. As a result, we are not confident that
Notes 397
Kjen should always be reconstructed with *-r- and Kjon without *-r-. After labial initials,
however, there are clear minimal contrasts, e.g., *m[o][r] > mjonX > wn late vs.
*mr[o][r] > mjenX > min make an effort; labials became labiodentals before MC -jon
but not before-jen.
30. In Ode 34.2, gu < *kru rhymes with m < muwX (< mjuwX) < *m(r)u male;
and other words with the phonetic ji < kjuwX < *[k]u also rhyme as *u. (We write *[k]
here because we do not know whether to reconstruct *ku or *ku; it is possible that the same
phonetic could have been used to write both.) Karlgren did not account for this phenomenon
in his Archaic Chinese reconstruction:he reconstructed words like gu < *kru as *Ki wg,
corresponding to our *Kr, as if the vowel was*.
31. In this example, three changes have occurred:(1)the change of *-r to *-n, (2)the fronting
of the nonfront vowels between acute onsets and acute codas, and (3)the lowering of high vow-
els in pharyngealized syllables. We do not actually know the chronological sequence of these
changes, which in any case may have varied from dialect to dialect, but the Middle Chinese
result would be the same in anycase.
32. *The *-ks listed in the table below as Starostins reconstruction is actually his recon-
struction of a stage preceding what he calls Old Chinese (drevnekitajskij jazyk); he assumed that
in the stage he calls Old Chinese, earlier *-ks had already changed to *-h, *-ps and *-ts had both
changed to a palatal *-, and *-s after nasals had changed to *-h (1989:332). But what we call
Old Chinese is more comparable to Starostins earlier stage when *-s was still present (see the
discussion in section 1.1), so those are the forms we cite as his reconstructions in this table and
in similar tablesbelow.
33. In general, according to the front-vowel hypothesis, Old Chinese rhymes will have
some Middle Chinese reflexes in either division I(if the vowel is nonfront) or division IV (if
the vowel is front *i or *e) but not both; the only exception is that, because of the fronting
of * between acute onsets and acute codas, the rhymes *-j, *-t(-s), *-n, and *-r have
both division-I reflexes (after grave initials) and division-IV reflexes (after acute initials); see
section5.5.5.
34. Since the coda *-r in *-ar, *-er, and *-or generally becomes MC -n, words with these
rhymes are also usually assigned to the traditional Yun group; see section 5.5.2below.
35. Recall that in our Middle Chinese notation, tsyoX is an abbreviation of tsy- + -joX, and
tsyaeX is an abbreviation of tsy- + -jae; so the two contrasting finals are MC -jo and-jae.
36. We use a similar notation in two other cases:*-Ak (as the source of MC -jek < OC *-ak)
and *-Aj (as the source of MC -jae < OC *-aj); see sections 5.4.1.2 and 5.4.2.2 below. It would be
possible to account for the contrast by adding an additional contrast in the rhyme;for example,
Starostin reconstructed tsyoX as *ta and tsyaeX as *tia (1989:687688); Zhngzhng
Shngfng (2003) reconstructed *tja and *tjaa. But these contrasts are very poorly dis-
tributed in their systems; we prefer to leave the question open fornow.
37. For example, -joX < *-a rhymes with -jaeX < *-A in Odes 173.1, 214.1, and218.4.
38. Ordinarily we would expect *-r- to produce a Middle Chinese retroflex initial of the Tr- or
Tsr- types, but one possible development would be *t-qrA > *t-rA > *t-hA > *tA > tsyhae.
39. These words were erroneously reconstructed with lateral initials in Baxter (1992); see
Sagart (1993b:256257) and note 51 in Chapter4.
40. Note that examples (894) and (895) are unusual in that they show graphic connections to
rshng even though they are shngshng and pngshng, respectively; cf. *s.qAk > sjek >
x slipper, shoe and *[d]ak > dak > du measure (v.). This fact could be a clue to why we
have -jae in such examples rather than-jo.
398Notes
41. The character , with the pngshng character *s-l > si > s superintend as second-
ary phonetic, reflects the change *-k-s > *-s and is probably of late origin; it does not occur in
the Shuwn, and we would be surprised to find it in early documents.
42. In addition to *-e, it is possible that the traditional Zh group includes some words that
should be reconstructed with *-ej; see section 5.5.3.1below.
43. In addition to *-ek(-s), the traditional X group includes some words that we recon-
struct with *-ik(-s); see section 5.4.4below.
44. In addition to *-e, the traditional Gng group includes some words that we reconstruct
with *-i; see section 5.4.4below.
45. It is not yet clear where this conservative dialect was spoken, but a single example found
so far attributes a case of -ik < *-ik to Shndng and Kuij (corresponding to mod-
ern Shoxng in Zhjing province)both areas on the coast. See example (956) and note
51below.
46. See Odes 51.3, 116.3, 222.3, 249.1, 252.8, 261.1, and 262.5, where mng rhymes with
words that are generally reconstructed with *-in. (It may be that some of the other words should
also be reconstructed with *-i, and that the apparent irregularity results from dialect mixture in
our Middle Chinese sources.)
47. In the meaning command (noun or verb), lng ~ lng appears to rhyme as if it were
*-in in Ode 100.2, where the Jngdin shwn says it is to be read lingH (JDSW 66), and in Ode
126.1, where the Jngdin shwn gives the readings ljeng, ljengH, and leng (JDSW69).
48. As is not infrequently the case, the Old Chinese rhymes of mng indicate a different
tone category from that preserved in the Middle Chinese tradition. The Gungyns only read-
ing for mng is mjaengH, but in the Shjng it usually rhymes as if it were in the pngshng
category:i.e., as if it did not have a final *-s. Originally there were presumably forms with and
without *-s. Similarly, most of the rhymes of lng are with pngshngwords.
49. In fact, the character is read as MC gin in the meaning a kind of lance. For what it is
worth, the Mandarin pronunciation jn would regularly reflect MC kin, not MCking.
50. In earlier reconstructions, which depended largely on the standard script, it was difficult
to decide which of the various possible sources of MC sy- should be reconstructed in shn; in
the absence of direct evidence, it had generally been reconstructed with a lateral initial, as a kind
of default, e.g., *hljin in Baxter (1992), which would be *l in in our current system; see the
discussion in section2.3.
51. In both Shndng and Kuij , [ sh] is pronounced like s [MC srik]
(). The text is also found in CBETA (2013), http://tripitaka.cbeta.
org/C057n1163_017, consulted July 14,2013.
52. From Suwilai Premsrirat (2002); data available online at http://sealang.net/
monkhmer/database/; accessed Nov 17,2011.
53. In addition to *-u, the traditional Yu group also includes some words that we recon-
struct with *-iw; see section 5.6.3.1below.
54. Karlgren did not reconstruct a syllable type corresponding to our *Kru but treated them
as if they were all (in our system) *Kr; thus he reconstructed gu < kwijX wheel ruts and
gu < kwijX gu ritual vessel as *ki wg, as if they were in the traditional Zh group,
like gu < kwij < *[k]r tortoise. But gu and gu clearly rhyme as *-u in the Shjng
(34.2, 135.2, 165.2).
55. We write the reconstruction of ji < kjuwX nine as *[k]u, with square brackets, to
express uncertainty about the onset. The use of as a phonetic in examples (987) and (988)
Notes 399
suggests that it could be *ku, a syllable type that our scenario presupposes but that is otherwise
unattested; its connection to zhu < trjuwX elbow (for which is generally considered the
early graph) suggests the possibility of a *t- presyllable.
56. In addition to *-uk(-s), the traditional Ju group also includes some words that we
reconstruct with *-iwk(-s); see section 5.6.3.2below.
57. Ode 165.6 in the Mo Sh has the line *[k]om-[k]om they drum to us;
in the Shuwn entry for kn, this line is quoted as (emended by Dun Yci to
), with *[k]om for the Mo Shs *[k]om (SWGL 2320b).
58. In the Wng Rnx manuscript of the Qiyn, chn has two readings:dzyin, glossed
as zo (early), and zyin, glossed as pngdn (dawn). Our reconstructions are dzyin <
*[d]r early and zyin < *s-[d]r morning; see section 4.5.3.2.
59. Ahom, L, and Dioi are Tai languages. According to Ethnologue (www.ethnologue.com),
Ahom is a nearly extinct Southwestern Tai language of Assam (ISO code: aho); L (ISO
code: khb) is a Southwestern Tai language spoken in southern Ynnn, Laos, Myanmar,
Thailand, and Vietnam; Dioi, also known as Bouyei (Chinese:Zhngji ; ISO code:pcc),
is a Northern Tai language spoken in Guzhu.
60. The change is reflected in Hn-dynasty rhyming when words like ji add < *kraj
began to rhyme with words like ji household < *kra; see Lu Chngpi and Zhu Zm
(1958:2024).
61. The original definitions in Rhodes, in Portuguese and Latin, are phi:tocar ou empeer
em alga cousa por desastre:impingere in aliquid ex infortunio;...phi tu:fazer naufragio:pati
naufragium.
62. The character is also read fn < phjon in other contexts, further supporting the recon-
struction of*-r.
63. Here *qa[j] has already become *a. Note also the use of the qshng syllable
hu < hwajH < *wajs < *m-kwat-s < *m-kot-s < *m-kop-s to represent bhs-, support-
ing the hypothesis that qshng reflects earlier final *-s. Judging from the transcription, the
form of Indic represented is probably not Sanskrit but a Central Asian pronunciation with -bh-
> -w- intervocalically. According to Zrcher (2007:35), Zh Lujichn (MC tsye
luw-kja-tsrhimH) was an Indoscythian, often credited with bringing Mahyna Buddhism to
China. The source of the example is the Taish Tripiaka 224 (Takakusu 19241932:431), a
translation of Aashasrikprajpramit).
64. For additional evidence for an *-r coda in shn, see examples (595), (596), and (597)
in section 4.4.3.4, and the discussionthere.
65. Although the only reading given in the Gungyn for jin admonish is MC kaenH,
the word rhymes twice in the Shjng (Odes 253.5 and, here, 254.1), and in both cases it rhymes
as if it were *kran.
66. Chinese dictionaries consistently give the pronunciation of as chny, but in Western
sinological literature it is usually transcribed as shan-y (in Wade-Giles romanization) or
shanyu (in pnyn), a habit that can probably be traced to Giless Chinese-English dictionary
(1892:1050), which gives the pronunciation shan (which would be shn in pnyn) in
this expression. At the time when Giless dictionary was written, standard Mandarin had not
yet been precisely defined. In modern standard Mandarin there are no morphemes pronounced
shn.
67. The relevant passages are in Wwn, Xinbi, Dng Y zhun
(Account of the Wwn, Xinbi, and Eastern Y), jun 30 of the Wi sh.
400Notes
68. See,for example, the following Shjng rhyme sequences, in which well-established *-ar
words rhyme with each other repeatedly:69.1, 215.3, 244.4, 254.7, 259.1, 259.7, 262.4,263.5.
69. The name is also written as Wwn. Even though hun and wn are pro-
nounced differently in standard Mandarin, in Middle Chinese they are bothhwan.
70. The following explanations of this rhyme are consistent with our hypotheses about
*-r:(1)The poem represents a dialect where *-n, *-r, and *-j are all distinct, and all three words
are from *-r, but the Middle Chinese readings are irregular because of dialect mixture. (2)The
poem represents a dialect where *-r > *-j, and mixes original *-j and *-r; tshjeX and/or
mjieX are from original *-j, while sjen is from original *-r. The Bi state from which the
poem supposedly originates is on the edge of the area where we believe *-r became *-j; see sec-
tion 5.5.1.4 below. Either way, sjen must be reconstructed as *[s]ar or *[s]er.
71. E.g., in Ex historia Menandri Protectoris Excerpta de legationibus barbarorum ad
Romanos, in Niebuhr (1829:281437), containing the Greek original with Latin translation.
72. It is also uncertain whether the MC '- here should be reconstructed as *- or *q-; hence
the brackets around the initial.
73. In addition to *-aj, the traditional G group also includes some words that we recon-
struct with *-oj and (perhaps) *-ej; see sections 5.5.4.1 and 5.5.3.1 below. Words that we recon-
struct with *-ar, *-er, and *-or may also be assigned to traditional G if they follow the dialect
development *-r>*-j.
74. Note that although modern dialects generally reflect the change Tsrj- > Tsr-, the standard
Mandarin pronunciation of cnc uneven, irregular reflects a form like tsrhim.tsrhje,
undoubtedly based on the fnqi in the Jngdin shwn glosses onOde1.
75. In addition to *-at(-s), the traditional Yu group also includes some words that we
reconstruct with *et(s) and *-ot(s); see sections 5.5.3.2 and 5.5.4.2 below. Also, because of the
early change of *-p-s to *-t-s, words that actually had original *-ap-s, *-ep-s, and *-op-s have
often been treated as if they were *-at-s, *-et-s, or *-ot-s and included inYu.
76. A couple of words with the phonetic do rhyme as *-at(-s): *m-pat > bat > b halt
in the open, in Ode 16.1, and *[b][a]t > bat > b sacrifice to the gods of the road in Ode
245.7; neither character is clearly attested in pre-Qn documents, so they are probably of late
origin and reflect the merger of *Pat with*Pot.
77. The only irregularity in all these rhymes is the fact that in Ode 304.2, after six rhyme
words in *-at, there is one in *-et, namely *[dz]et > dzet > ji cut,trim.
78. The Shuwn glosses f as root (, SWGL 3988a), and J Xshng (2010:733)
argues that this is correct; perhaps the word originally referred to small hair-like roots. So it is
possible that it is related to the root in *b<r>ot > beat > b uproot.
79. In addition to *-an, the traditional Yun group also includes some words that we recon-
struct with *-en and *-on (see sections 5.5.3.3 and 5.5.4.3 below); those words that we recon-
struct with *-ar, *-er, and *-or that follow the usual dialect development *-r > *-n are also
traditionally included in Yun.
80. In addition to *-et(-s), the traditional Yu group also includes words that we reconstruct
with *-at(-s) and *-ot(-s) (see sections 5.5.2.2 and 5.5.4.2). Also, because of the early change of
*-p-s to *-t-s, words in original *-ap-s, *-ep-s, and *-op-s are often treated as if they were *-at-s,
*-et-s, or *-ot-s and also included inYu.
81. In addition to *-en, the traditional Yun group also includes words that we reconstruct
with *-an and *-on (see sections 5.5.2.3 and 5.5.4.3). Words that we reconstruct with *-ar, *-er,
and *-or that follow the usual dialect development *-r > *-n are also traditionally included in
Yun.
Notes 401
82. The rhyme evidence for words with qin as phonetic is somewhat ambiguous (Odes 58.2,
200.4, 220.3, 305.6), but the character contains the element *[s][r]-s > sinH > xn fontanel,
which seems to have connections to such *-r words as *s-nr > sej > x west. Also, the
Shuwn gives an ancient character form for qin < tshjen, composed of the elements
shu hand and x west (SWGL 757b), where x could be a phonetic element (though
the vowels do not match).
83. In addition to *-oj, the traditional G group also includes words that we reconstruct
with *-aj (see section 5.5.2.1). Words that we reconstruct with *-ar, *-er, and *-or that follow the
dialect development *-r > *-j are also usually assigned toG.
84. In addition to *-ot(-s), the traditional Yu group also includes words that we reconstruct
with *-at-s and *-et-s (see sections 5.5.2.2 and 5.5.3.2). Also, because of the early change of
*-p-s to *-t-s, words in original *-ap-s, *-ep-s, and *-op-s are often treated as if they were *-at-s,
*-et-s, or *-ot-s, and included inYu.
85. In addition to *-on, the traditional Yun group also includes words that we reconstruct
with *-an and *-en (see sections 5.5.2.3 and 5.5.3.3). Words that we reconstruct with *-ar, *-er,
and *-or that follow the usual dialect development *-r > *-n are also usually assigned to Yun.
86. Norman identifies this Mn form with *s.tem > syem > shn thatch (1996:26), but the
Mn reflexes seem to reconstruct as pMn *-tsun C, except for Fzhu /tsai 5/, which is irregular,
as Norman mentions; the Fzhu form is consistent with pMn *-n C (Norman 1981:58). Both
pMn *-un and *-n could reasonably reflect Old Chinese *-n or *-r. Neither is consistent with
any of the Proto-Mn finals ending in *-m, according to the correspondences in Norman (1981),
which would seem to exclude shn < syem as the etymon in this case, since Middle Chinese
-m usually corresponds to pMn *-m. So we believe that the etymon for pMn *-tsun C thatch is
*C.ts[r]-s > tsenH > jin grass, fodder.
87. In addition to *-j, the traditional Wi group also includes words that we reconstruct
with *-uj (see section 5.5.7.1). Words that we reconstruct with *-r and *-ur that follow the
dialect development *-r > *-j are also usually assigned toWi.
88. In a few cases, we might suspect *-r instead of *-j; in such cases we write *-[j].
89. The rhymes of y < *lj are in Odes 14.3, 90.1, 168.6, 191.5, 257.2, and 284.1 (the last
being an irregular rhyme of *-j with*-uj).
90. In fact, we suspect that the so-called rnfng, often mentioned in Shng inscrip-
tions as one of the groups with whom the Shng fought, were actually none other than the
yfng, also an enemy of the Shng (a view mentioned by Zho Chng 1988:145). The context
would have been sufficient to show that the character represented {} rather than{}.
91. The rhyming of sh itself is ambiguous:it clearly rhymes as *-j in Ode 209.5 (rhyme
words: *l j *[k]j *l<r>[j] *[s]j), but it also rhymes in Ode 254.5, where all the
other rhyme words seem to have *-ij (rhyme words: *[dz][i]j *[b]ij *mij *l j
*[q]ij *gij *[ts]ij *srij). If we reconstruct sh with *-j, then we can explain the
rhyming in 254.5 as the result of the late sound change *- > *-ij after acute initials (although
we suspect there could be a textual problem here); but if sh had *-ij, we have no way of
accounting for its rhyme with *-j in 209.5. So the rhyme evidence also indicates that *l j is
the correct reconstruction.
92. For example, the best comparisons with Proto-Tibeto-Burman *-y (Matisoff 2003:201)
earlier reconstructed as *-iy (Benedict 1972:57, n.188)seem to be with OC *-ij, not with *-j;
see Baxter (1985).
93. In addition to *-t(-s), the traditional W group also includes words that we reconstruct
with *-ut(-s) (see section 5.5.7.2). Also, because of the early change of *-p-s to *-t-s, words
402Notes
in original *-p-s or *-up-s are often treated as if they were *-t-s or *-ut-s, respectively, and
included inW.
94. In addition to *-n, the traditional Wn group also includes words that we reconstruct
with *-un (see section 5.5.7.3). Those words that we reconstruct with *-r and *-ur are also
traditionally included in Wn if they have final -n in Middle Chinese.
95. In addition to *-ij, the traditional Zh group may also include words with *-ir that fol-
low the dialect development *-r > *-j (see section 5.5.6.4).
96. In addition to *-it(-s), the traditional Zh group also includes words with *-ik(-s) that
follow the usual development *-ik > *-it (see section 5.4.4). Also, because of the early change
of *-p-s to *-t-s, words in original *-ip-s are often treated as if they had *-it-s, and also included
inZh.
97. In addition to *-in, the traditional Zhn group also includes words with *-i that follow
the usual development *-i > *-in (see section 5.4.4). Words in *-ir that follow the usual devel-
opment *-r > *-n may also be included in Zhn.
98. The reference is to the emperor Jinwn (503551) of the Ling dynasty, who
wrote a commentary on the Loz, now lost. We are grateful to Wng Hngzh for assis-
tance in interpreting this passage.
99. In addition to *-uj, the traditional Wi group also includes words that we reconstruct
with *-j (see section 5.5.5.1). Words in *-r and *-ur that follow the dialect development *-r >
*-j may also be included inWi.
100. In addition to *-ut(-s), the traditional W group also includes words that we reconstruct
with *-t(-s) (section 5.5.5.2). Also, because of the early change of *-p-s to *t-s, words in origi-
nal *-p-s or *-up-s are often treated as if they had *-t-s or *-ut-s, respectively, and are often
included inW.
101. In addition to *-un, the traditional Wn group also includes words that we reconstruct
with *-n (section 5.5.5.3). Words in *-r and *-ur that follow the usual development *-r > *-n
may also be included inWn.
102. In addition to *-aw, the traditional Xio group also includes words that we reconstruct
with *-ew; see section 5.6.2.1.
103. In addition to *-awk(-s), the traditional Yo group also includes words that we recon-
struct with *-ewk(-s); see section 5.6.2.2.
104. In addition to *-ew, the traditional Xio group also includes words that we reconstruct
with *-aw; see section 5.6.1.1.
105. In addition to *-ewk(-s), the traditional Yo group also includes words that we recon-
struct with *-awk(-s); see section 5.6.1.2.
106. In addition to *-iw, the traditional Yu group also includes words that we reconstruct
with *-u; see section 5.4.6.1.
107. In addition to *-iwk(-s), the traditional Ju group also includes words that we recon-
struct with *-uk(-s); see section 5.4.6.2.
108. The rest have vocalic codas. The estimates are rough because although we have more than
11,000 records in our database, each representing a word written and pronounced a certain way,
a good many words are represented by more than one record.
109. See Are there any words that rhyme with orange?, at <http://oxforddictionaries.com/
words/are-there-any-words-that-rhyme-with-orange>, and Oxford English Dictionary (Second
edition, 1989; online version December 2011), at <http://www.oed.com.proxy.lib.umich.edu/
view/Entry/187451>, consulted 18 December2011.
Notes 403
110. The one systematic exception is that MC -en, -et, and -ej can come from the fronting of *
between an acute onset and an acute coda; see section5.5.5.
111. In Ode 265.3, din < temX < *tem black spot (evidently the same word as din
< temX) rhymes with bin < pjemX diminish, but while pjemX could represent *prem and
thus be a regular rhyme in our system, we cannot rule out the possibility that it was *pram or
*prom.
112. Dng Tngh also reconstructed *-im as the source for MC -em in the Tn group,
recognizing that it tended to be kept separate from (our) *-am in phonetic compounds. His *-im
corresponds to our *-em, and in fact it has few contacts with his*-m.
113. , [lip] [pik], : ,
, [pik] Quoted in Hny
d cdin (2001).
114. The same character is also used to write the unrelated word xing < xjang fragrant,
usually written.
115. Karlgren reconstructed di as *dd, *did; Li has *ddh, which is irregular in his
system (it should become MC dwojH). But Dng Tngh recognized the labial coda:*db.
116. We write *[u] rather than *u here because it is also possible that there was a dialect in
which original *-p simply became *-up and subsequently developed like *-up; the same rea-
soning applies in some other examplesbelow.
117. The vessel is number 2748 in Jnwn wnxin jchng (Li Qngzh
etal. 2004). Our discussion relies on the detailed account in Chn Jin (2007). As Chn Jin
points out, in the later script there came to be a division of labor between gng to present (as
a king to a subject) and gng to present (to a superior), but originally they were a single
word meaning to present in either direction.
Chapter6
1. It is true, of course, that the earliest Indo-European languages, still not too distant in time
from their common ancestor, did share certain structural characteristics.
{ REFERENCES }
Akitani Hiroyuki [Qig Yxng]. 2003. Wy Ch-Q fngyn (xbi pin) gyn
gun . Tky :Kbun Shuppansha
.
Akitani Hiroyuki [Qig Yxng]. 2004. Fjin Shbi fngyn ynx
. Fngyn 2004.7691.
Aurousseau, Lonard. 1923. La premire conqute chinoise des pays annamites (IIIe sicle avant
notre re). Bulletin de lcole franaise dExtrme-Orient 23.136264.
Bi Ypng [William H.Baxter]. 1983. Shngg Hny *sr- de fzhn *sr-
[The development of Old Chinese *sr]. Yyn ynji 1983.2226.
Bi Ypng [William H.Baxter]. 2010. Y, sh, sh dng z de gun h zhngg
sy-
(shm=shn sn) de liyun , , sy-=
. Jinb 5.161178.
Bi Yln . 2008. Jind bsh tngji z zdin . Fzhu
:Fjin Rnmn chbnsh .
Bailey, H. W. 1946. Gndhr. Bulletin of the School of Oriental and African Studies 11.764797.
Barrett, T. H. 1993. Lieh tzu . Early Chinese texts:a bibliographical guide, ed. Michael
Loewe, 298308. Berkeley: Society for the Study of Early China: Institute of East Asian
Studies, University of California, Berkeley.
Baxter, William H. 1977. Old Chinese origins of the Middle Chinese chngni doublets:a study
using multiple character readings. Ph.D.diss., Cornell University.
Baxter, William H. 1985. Tibeto-Burman cognates of Old Chinese *-ij and *-j. Linguistics of
the Sino-Tibetan area:the state of the artpapers presented to Paul K.Benedict for his 71st
birthday (Pacific linguistics, series C, no.87), ed. Graham Thurgood, James A. Matisoff, and
David Bradley, 242263. Canberra:Australian National University.
Baxter, William H. 1992. A handbook of Old Chinese phonology. Berlin:Mouton de Gruyter.
Baxter, William H. 2002. Where does the comparative method come from? The linguists
linguist:a collection of papers in honour of Alexis Manaster Ramer, ed. Fabrice Cavoto,
3352. Mnchen:LINCOM.
Baxter, William H., and Laurent Sagart. 1998. Word formation in Old Chinese. New approaches
to Chinese word formation:morphology, phonology and the lexicon in Modern and Ancient
Chinese, ed. Jerome L. Packard, 3576. Berlin:Mouton de Gruyter.
Bijng dxu Zhnggu yyn wnxu x Yynxu jioynsh
. 1995. Hny fngyn chu . 2nd ed. Bijng :
Wnz gig chbnsh .
Bijng dxu Zhnggu yyn wnxu x Yynxu jioynsh
. 2003. Hny fngyn zhu (d 2 bn chngpi bn)
2 . Bijng :Ywn chbnsh .
Benedict, Paul K. 1972. Sino-Tibetan, a conspectus, ed. James A.Matisoff. Cambridge:Cambridge
UniversityPress.
406 References
Benedict, Paul K. 1976. Sino-Tibetan:another look. Journal of the American Oriental Society
96.167197.
Benedict, Paul K. 1987. Archaic Chinese initials. Wang Li memorial volumes. English volume,
2571. Hong Kong:Joint PublishingCo.
Bhattacharya, Pramod Chandra. 1977. A descriptive analysis of the Boro language.
Gauhati:Department of Publication, Gauhati University.
Bivar, A. D. H. 2000. The political history of Iran under the Arsacids. The Seleucid,
Parthian and Sasanian periods; vol. 3, part 1 of The Cambridge history of Iran, 2199.
Cambridge:Cambridge UniversityPress.
Bodman, Nicholas C. 1954. A linguistic study of the Shih Ming:initials and consonant clusters.
Cambridge:Harvard UniversityPress.
Bodman, Nicholas C. 1980. Proto-Chinese and Sino-Tibetan: data towards establishing the
nature of the relationship. Contributions to historical linguistics:issues and materials, ed.
Frans van Coetsem and Linda R. Waugh, 34199. Leiden:E. J.Brill.
Boltz, William G. 1984. Textual criticism and the Ma wang tui Lao tzu, review of Chinese clas-
sics:Tao te ching, by D. C. Lau. Harvard Journal of Asiatic Studies 44.185224.
Boltz, William G. 1985. The Lao tzu text that Wang Pi and Ho-shang Kung never saw. Bulletin
of the School of Oriental and African Studies 48.493501.
Boltz, William G. 1993. Shuo wen chieh tzu . Early Chinese texts:a bibliographical
guide, ed. Michael Loewe, 429442. Berkeley:Society for the Study of Early China and the
Institute of East Asian Studies, University of California.
Burling, Robbins. 2004. The language of the Modhupur Mandi (Garo), Vol. 1:Grammar. New
Delhi:Promilla.
Ci Jnmng [Choy Chun-ming]. 1976. Choy cdin [A dictionary of the
Chau-chou dialect]. Xinggng :Zhu Fgo.
Co Zhyn , Akitani Hiroyuki [Qig Yxng], ta Itsuku [Titin
Zhi], and Zho Rxn . 2000. Wy Ch-Q fngyn ynji .
Tky :Kbun Shuppansha.
CBETA [Chinese Buddhist Electronic Text Association]. 2013. Hnwn Dzngjng
. http://tripitaka.cbeta.org.
Chao, Yuen Ren [Zho Yunrn ]. 1941. Distinctions within Ancient Chinese. Harvard
Journal of Asiatic Studies 5.203233.
Chn Chshng . 1987. Jnwn chngyng zdin . Xn :Shnx
Rnmn chbnsh .
Chn Jin . 2007. Sh x Zhu jnwn de () z () .
Jig jnwn kosh ln j , 819. Bijng : Xinzhung
shj.
Chn L . [1842] 1995. Qiyn ko ; Qiyn ko wi pin . Shnghi
:Shnghi gj chbnsh .
Chn Xnxing . 1972. Gynxu f wi . Tibi :Jixn shun gngs
wnhu jjnhu .
Chn Zhngti and L Rlng . 1991. Mny ynji . Bijng
:Ywn chbnsh.
Coblin, W. South. 1983. A handbook of Eastern Han sound glosses. Hong Kong: Chinese
UniversityPress.
References 407
Coblin, W. South. 2011. Comparative phonology of the Central Xing dialects. Taipei:Institute
of Linguistics, Academia Sinica.
Coeds, G. 1935. Lorigine du cycle des douze animaux au Cambodge. Toung pao 31.315329.
Di Jixing . 1995. Jnwn d zdin . Shnghi : Xuln
chbnsh.
Damrong Tayanin and Kristina Lindell. 1994. The Khmu cycles of 60 days and 60 years.
Mon-Khmer Studies 23.103118.
Dell, Franois, and Mohamed Elmedlaoui. 1985. Syllabic consonants and syllabification in
Imdlawn Tashlhiyt Berber. Journal of African Languages and Linguistics 7.105130.
Dng Xingzhng . 2007. Mnbi, Mnzhng fngyn yyn ynji
. Ph.D.diss., Ximn dxu.
Dien, Albert E. 1957. A note on Hsien Zoroastrianism. Oriens 10.284288.
Diffloth, Grard. 1984. The Dvaravati Old Mon language and Nyah Kur. Bangkok:Chulalongkorn
University PrintingHouse.
Dng D . [1039] 1985. Jyn . Bijng :Zhnghu shj.
Dng Shngsh and L Rng . 1981. Hny ynyn jingy .
Fngyn 1981.241274.
Doerfer, Gerhard. 19631975. Trkische und mongolische Elemente im Neupersischen; unter
besonderer Bercksichtigung lterer neupersischer Geschichtsquellen, vor allem der
Mongolen- und Timuridenzeit. 4vols. Wiesbaden:F. Steiner.
Dng Tngh . 1948. Shngg ynyn bio go . Zhngyng ynjiyun
lsh yyn ynjisu jkn [Bulletin of the Institute
of History and Philology, Academia Sinica] 18.1249.
Dng Tngh . 1960. Sge Mnnn fngyn . Zhngyng ynjiyun
lsh yyn ynjisu jkn [Bulletin of the Institute
of`History and Philology, Academia Sinica] 30. 7291042.
Douglas, Carstairs. 1899. Chinese-English dictionary of the vernacular or spoken language of
Amoy, with the principal variations of the Chang-chew and Chin-chew dialects. New ed.,
with corrections by the author. London:Presbyterian Church of England.
Downer, Gordon B. 1967. Tone-change and tone-shift in White Miao. Bulletin of the School of
Oriental and African Studies 30.589599.
Downer, Gordon B. 1973. Strata of Chinese loanwords in the Mien dialect of Yao. Asia Major
n.s. 18.133.
Dun Yci . [1815] 1981. Shuwn jiz zh . Shnghi :Shnghi
gj chbnsh .
Dyson, F. W., A. S. Eddington, and C. Davidson. 1920. A determination of the deflection of light
by the suns gravitational field, from observations made at the total eclipse of May 29, 1919.
Philosophical Transactions of the Royal Society of London. Series A, Containing Papers of
a Mathematical or Physical Character 220.291333.
Edmondson, Jerold A., and Quan Yang. 1988. Word-initial preconsonants and the history of
Kam-Sui resonant initials and tones. Comparative Kadai: linguistic studies beyond Tai,
ed.Jerold A. Edmondson and David B. Solnit, 143166. Dallas:Summer Institute of Linguistics.
Evans, Jonathan. 2006a. Origins of vowel pharyngealization in Hongyan Qiang. Linguistics of
the Tibeto-Burman Area 29.95126.
Evans, Jonathan. 2006b. Vowel quality in Hongyan Qiang. Language and Linguistics 7.937960.
Fng izhn . 1998. Fzhu fngyn cdin . Nnjng :Jings
jioy chbnsh .
408 References
Karlgren, Bernhard. 19421946. Glosses on the Book of odes. Bulletin of the Museum of Far
Eastern Antiquities 14.71247, 16.25169, 18.1198.
Karlgren, Bernhard. 1950. The book of odes. Stockholm:Museum of Far Eastern Antiquities.
Karlgren, Bernhard. 1954. Compendium of phonetics in Ancient and Archaic Chinese. Bulletin
of the Museum of Far Eastern Antiquities 26.211367.
Kennedy, George A. 1939. Metrical irregularity in the Shih Ching. Harvard Journal of Asiatic
Studies 4.284296.
L-Thongkum, Therapan. 1992. A preliminary reconstruction of Proto-Lakkja (Cha Shan Yao).
Mon-Khmer Studies 20.5789.
L-Thongkum, Therapan. 1993. A view on Proto-Mjuenic (Yao). Mon-Khmer Studies 22.163230.
Lau, D. C. 1963. Tao te ching. Baltimore:PenguinBooks.
Leong, Sow-Theng. 1997. Migration and ethnicity in Chinese history:Hakkas, Pengmin, and
their neighbors, ed. Tim Wright; introduction and maps by G.William Skinner. Stanford,
CA:Stanford UniversityPress.
Li, Fang-kuei [L Fnggu ]. 1945. Some Old Chinese loan words in the Tai languages.
Harvard Journal of Asiatic Studies 8.333342.
Li, Fang-kuei [L Fnggu ]. 1971. Shngg yn ynji . Tsing Hua Journal
of Chinese Studies n.s. 9.161.
Li, Fang-kuei [L Fnggu ]. 1977. A handbook of comparative Tai. Honolulu:University
Press of Hawaii.
L Jiho . 1979. Sh bin . Gwnz ynji 1.391395.
L Jiho . 2004. Zhngu zhjin Mn zh fm zhng de cibin
. Bijng dxu xubo 41.9698.
L Jnfng . 1999. Byngy ynji . Bijng : Zhngyng mnz
dxu chbnsh .
L Rlng and Chn Zhngti . 1991. Ln Mn fngyn nib de zhyo chy
. Mny ynji , 58138. Bijng : Ywn
chbnsh.
L Rlng and Pn Wishu . 1998. Jinu fngyn cdin .
Nnjng :Jings jioy chbnsh .
L Rlng and Zhng Shungqng . 1992. K Gn fngyn dioch bogo
. Ximn :Ximn dxu chbnsh .
L Rlng etal. 1999. Yux Kji fngyn dioch bogo .
Gungzhu :Jnn dxu chbnsh .
Ling Mn . 1980. Monny jinzh . Bijng : Mnz
chbnsh.
Ln Sqng [Sue-Ching Lin]. 2003. Sh [ln]jin ln Ch jin de yng z tzhng
[Explanation of , with a discussion of the characteris-
tics of the use of graphs in the Chu slips]. Zhngyng ynjiyun lsh yyn ynjisu jkn
[Bulletin of the Institute of History and Philology,
Academia Sinica] 74.293305.
Li Lnxn . 2001. Jingx Kji fngyn gikung . Nnchng
:Jingx Rnmn chbnsh .
Li Qngzh , Dun Zhhng , and Fng Sh . 2004. Jnwn wnxin
jchng . Xinggng : Xinggng Mngsh wnhu guj chbn
yuxin gngs .
References 411
Lng Ychn . 1968. Tng xi qunbn Wng Rnx Knmi bqu Qiyn jio jin
. Xinggng : Xinggng Zhngwn dxu
.
Lorrain, J. Herbert, and Fred W. Savidge. 1898. A grammar and dictionary of the Lushai lan-
guage (Dulien dialect). Shillong: Assam Secretariat Printing Office.
Lu Chngpi . 1931. Qiyn Y Y zh ynzh j q su j fngyn ko
. Zhngyng ynjiyun lsh yyn ynjisu jkn
[Bulletin of the Institute of History and Philology,
Academia Sinica] 2.258385.
Lu Chngpi and Zhu Zm . 1958. Hn Wi-Jn Nn-bi cho ynb
ynbin ynji . Vol. 1. Bijng : Kxu chbnsh
.
Lu Jiru [Jerry Norman]. 2005. Mn fngyn zhng de Li m z h zoq Hny
. Mnz ywn 2005.15.
Mackenzie, D. N. 1983. Khwarazmian language and literature. The Cambridge history of Iran,
volume 3, part2:The Seleucid, Parthian and Sasanian Periods, ed. Ehsan Yarshater, 1244
1249. Cambridge:Cambridge UniversityPress.
Maddieson, Ian, and Richard Wright. 1991. The vowels and consonants of Amis:a preliminary
phonetic report. UCLA Working Papers in Phonetics 91.4563.
Manomaivibool, Prapin. 1975. A study of Sino-Thai lexical correspondences. Ph.D diss.,
University of Washington.
Maspero, Henri. 1912. tudes sur la phontique historique de la langue annamite:les initiales.
Bulletin de lcole franaise dExtrme-Orient 12.1124.
Mathews, R. H. 1943. Mathews Chinese-English dictionary. Revised by Y.R. Chao and M.Y.
Wang. Cambridge, MA:Harvard UniversityPress.
Matisoff, James A. 2003. Handbook of Proto-Tibeto-Burman: system and philosophy of
Sino-Tibetan reconstruction. Berkeley:University of CaliforniaPress.
Mayr, Ernst. 1982. The growth of biological thought: diversity, evolution, and inheritance.
Cambridge, MA:BelknapPress.
Mei, Tsu-lin [Mi Zln ]. 1989. Shngg Hny *s- qinzhu de guc gngyng
*s- [The causative and denominative function of the *s-
prefix in Old Chinese]. Proceedings of the Second International Conference on Sinology
(Section on Linguistics and Paleography), 3352. Tibi : Zhngyng ynjiyun
.
Mei, Tsu-lin [Mi Zln ]. 2001. Xindi Wy h Zh Zh Y Y, gng wi
b yn , . Zhnggu ywn 2001.315.
Mei, Tsu-lin [Mi Zln ]. 2012. The causative *s- and nominalizing *-s in Old Chinese
and related matters in Proto-Sino-Tibetan. Language and Linguistics 13.128.
Michaud, Alexis, Guillaume Jacques, and Robert L. Rankin. 2012. Historical transfer of nasality
between consonantal onset and vowel:from C to V or from V to C? Diachronica 29.2:201230.
Nguyn Ph Phong, Trn Tr Di, and Michel Ferlus. 1988. Lexique vietnamien-rc-franais:par-
ler dune minorit ethnique des montagnes de Qung Bnh, Vietnam, avec notes eth-
nographiques et introduction linguistique. Paris:Universit de Paris VII, Sudestasie.
Niebuhr, Barthold Georg (ed.). 1829. Ex historia Menandri Protectoris Excerpta de legationibus
barbarorum ad Romanos. Corpus scriptorum historiae byzantinae 5:281437. Bonn:Impensis
Ed. Weberi. http://catalog.hathitrust.org/Record/001222061.
412 References
Norman, Jerry. 1971. A guide to the Chien-yang dialect. Washington, DC:Office of Education,
Department of Health, Education, and Welfare.
Norman, Jerry. 1973. Tonal development in Min. Journal of Chinese Linguistics 1.222238.
Norman, Jerry. 1974a. The initials of Proto-Min. Journal of Chinese Linguistics 2.2736.
Norman, Jerry. 1974b. The Shaowu dialect. Orbis:Bulletin international de documentation lin-
guistique 23.328334.
Norman, Jerry. 1979. Chronological strata in the Min dialects. Fngyn 1979.268273.
Norman, Jerry. 1981. The Proto-Min finals. Zhngyng ynjiyun Guj Hnxu huy lnwn
j:Yyn wnz z : , 3573. Tibi :
Zhngyng ynjiyun.
Norman, Jerry. 1982. The classification of the Shaowu dialect. Zhngyng ynjiyun Lsh
yyn ynjisu jkn [Bulletin of the Institute of
History and Philology, Academia Sinica] 53.543583.
Norman, Jerry. 1986. The origin of the Proto-Min softened stops. Contributions to Sino-Tibetan
studies, ed. John McCoy and Timothy Light, 375384. Leiden:E. J.Brill.
Norman, Jerry. 1988. Chinese. Cambridge:Cambridge UniversityPress.
Norman, Jerry. 1989. What is a Kji dialect? Proceedings of 2nd International Conference on
Sinology:Linguistics, vol. 1, 323344. Taipei:Academia Sinica.
Norman, Jerry. 1991. Nasals in Old Southern Chinese. Studies in the historical phonology of
Asian languages, ed. William G. Boltz and Michael C. Shapiro, 205214. Amsterdam:John
Benjamins.
Norman, Jerry. 1994. Pharyngealization in Early Chinese. Journal of the American Oriental
Society 114.397408.
Norman, Jerry. 1995. A glossary of the Herpyng dialect. Yuen Ren Society Treasury of Chinese
Dialect Data 1.107126.
Norman, Jerry. 1996. Tonal development in the Jennchyan dialect. Yuen Ren Society Treasury of
Chinese Dialect Data 2.741.
Norman, Jerry. 2000. Voiced initials in Shyrbei. Yyn binhu y Hny fngyn:L Fnggu
xinsheng jnin lnwnj : [In
memory of Professor Li Fang-Kuei:Essays on linguistic change and the Chinese dialects],
ed. Dng Bngxn [Pang-hsin Ting] and Y iqn [Anne O.Yue], 271280.
Tibi :Zhngyng ynjiyun Yynxu ynjisu chubich
.
Norman, Jerry. 2002. A glossary of the Lianduentsuen dialect. Yuen Ren Society Treasury of
Chinese Dialect Data 3.339394.
Norman, Jerry. 2006. Miin animal body parts. Bulletin of Chinese Linguistics 1.133143.
Norman, Jerry, and W. South Coblin. 1995. A new approach to Chinese historical linguistics.
Journal of the American Oriental Society 115.576584.
Norquest, Peter K. 2007. Aphonological reconstruction of Proto-Hlai. Ph.D.diss., University
of Arizona.
Ostapirat, Weera. 2000. Proto-Kra. Linguistics of the Tibeto-Burman Area 23.1251.
Ostapirat, Weera. 2004. Proto-Hlai sound system and lexicons. Hn-Zng y ynji: Gng
Hungchng xinsheng q zh shu qng lnwn j :
[Studies on Sino-Tibetan languages:papers in honor of Professor Hwang-cherng
Gong on his seventieth birthday], ed. Ln Yngjn [Lin Ying-chin] etal., 121175.
Tibi :Zhngyng ynjiyun Yynxu ynjisu .
References 413
Ostapirat, Weera. 2006. Alternation of tonal series and the reconstruction of Proto-Kam-Sui.
Shn go shu chng: Dng Bngxn xinsheng q zh shu qng lnwn j
:, ed. H Dn [Ho Dah-an], Zhng
Hngnin , Pn Wyn , and W Fxing , 10771121. Tibi
:Zhngyng ynjiyun Yynxu ynjisu .
Ostapirat, Weera. 2011. Linguistic interaction in South China: the case of Chinese, Tai and
Miao-Yao. Paper for the symposium Historical linguistics in the Asia-Pacific region and the
position of Japanese, at the 20th International Conference on Historical Linguistics, Osaka,
June 30,2011.
Pn Wishu . 1998. Jinu fngyn cdin . Nnjng :Jings
jioy chbnsh .
Pn Wyn . 1997. Huyn ko . Mnz ywn 1997.1024.
Peiros, Ilia, and Sergej Starostin. 1996. A comparative vocabulary of five Sino-Tibetan lan-
guages. 6vols. Melbourne:Department of Linguistics and Applied Linguistics, University
of Melbourne.
Pelliot, Paul. 1934. Tokharien et Koutchen. Journal asiatique 224.23106.
Pelliot, Paul. 1938. Le nom du wrizm dans les textes chinois. Toung pao 34.146152.
Pittayaporn, Pittayawat. 2009. The phonology of Proto-Tai. Ph.D.diss., Cornell University.
Plaisier, Heleen. 2007. A Grammar of Lepcha. Tibetan studies library:Languages of the greater
Himalayan region 5.Leiden:E. J.Brill.
Pulleyblank, Edwin G. 19621963. The consonantal system of Old Chinese. Asia Major n.s.
9.58144, 206265.
Pulleyblank, Edwin G. 1973. Some new hypotheses concerning word families in Chinese.
Journal of Chinese Linguistics 1.111125.
Pulleyblank, Edwin G. 19771978. The final consonants of Old Chinese. Monumenta Serica
33.180206.
Pulleyblank, Edwin G. 1983. The Chinese and their neighbors in prehistoric and early his-
toric times. The origins of Chinese civilization, ed. David N. Keightley, 411466.
Berkeley:University of CaliforniaPress.
Pulleyblank, Edwin G. 1984. Middle Chinese: a study in historical phonology.
Vancouver:University of British ColumbiaPress.
Qi Xgu . 1985. Sh Ynx jigwn l de , [] () j yugun zh z
[] . Gwnz ynji 11.8598.
Qi Xgu . 1988. Wnzxu giyo . Bijng : Shngw
ynshgun.
Qi Xgu . 1992. [b] z b sh . Gwnz ln j , 9098.
Bijng :Zhnghu shj.
Qi Xgu . 1998. G wnxin zhng d wi sh de y j q y zh h
zh l . Dngfng wnhu
36.3946. [Dated 1998 but actually published in2002.]
Qi Xgu v. 2000. Chinese writing. Trans. Gilbert Louis Mattos and Jerry Norman.
Berkeley: Society for the Study of Early China and the Institute of East Asian Studies,
University of California. [Translation of Qi Xgu (1988).]
Qi Xgu . 2004. Sh Gudin Zy Ch yn yu , lmn su : jin
shu wi [zhn] zh chwn ,
:. Zhnggu cht gwnxin sh jing
. Shnghi :Fdn dxu chbnsh .
414 References
Starostin, Sergej Anatolevi. 1995. The historical position of Bai. Moskovskij Lingvistieskij
urnal 1.174190.
Sun, Jackson T.-S. 2003. Variegated tonal developments in Tibetan. Language variation:papers
on variation and change in the Sinosphere and in the Indosphere in honour of James
A. Matisoff, ed. David Bradley, Randy J. LaPolla, Boyd Michailovsky, and Graham
Thurgood, 3551. Canberra:Pacific Linguistics.
Suwilai Premsrirat. 2002. Thesaurus of Khmu dialects in Southeast Asia. Salaya, Nakhon
Pathom, Thailand:Mahidol University at Salaya.
SWGL: Dng Fbo (ed.). 1931. Shuwn jiz gln:j by :.
17vols. Tibi :Tiwn Shngw ynshgun .
Takakusu Junjir (ed.). 19241932. Taish shinsh Daizky .
85vols. Tky:Taish Issaiky Kankkai .
Thompson, Laurence C. 1976. Proto-Viet-Muong phonology. Austroasiatic studies, ed. Philip N.
Jenner, Laurence C. Thompson, and Stanley Starosta, 11131203. Honolulu:The University
Press of Hawaii.
Thompson, P. M. 1979. The Shen Tzu fragments. Oxford:Oxford UniversityPress.
Thurgood, Graham. 1988. Notes on the reconstruction of Proto-Kam-Sui. Comparative
Kadai:linguistic studies beyond Tai, ed. Jerold A. Edmondson and David B. Solnit, 179218.
Arlington:Summer Institute of Linguistics:University of Texas at Arlington.
Ting, Pang-hsin. 1975. Chinese phonology of the Wei-Chin period:reconstruction of the finals as
reflected in poetry. Taipei:Institute of History and Philology, Academia Sinica.
Ting, Pang-hsin. 1983. Derivation time of colloquial Min from Archaic Chinese. Bulletin of the
Institute of History and Philology, Academia Sinica 54.114.
Unger, Ulrich. 1995. Finger. Hao-ku 46.131137.
Vendler, Zeno. 1957. Verbs and times. The Philosophical Review 66.143160.
Wng Fsh . 1979. Mioy fngyn shng- ynm bjio . Bijng
:Zhnggu shhu kxu yun Mnz ynjisu .
Wng Fsh . 1982. Hnn Lx Wxing hu yyn . Yyn
ynji 1982.135147.
Wng Fsh and Mo Zngw . 1995. Mio-Yo y gyn gun .
Bijng :Zhnggu shhu kxu chbnsh .
Wng L . [1937] 1958. Shngg ynm xtng ynji . Hny sh
lnwn j , 77156. Bijng :Kxu chbnsh.
Wng L . 1958. Hny sh go . 3 vols. Bijng : Kxu chbnsh
.
Wng L . 1980. Shjng yn d . Shnghi : Shnghi gj chbnsh
.
Wng L . 1985. Hny ynynxu . Vol. 4 of Wng L wnj . Jnn
:Shndng jioy chbnsh .
Wng L . 1988. Hny sh go . Vol. 9 of Wng L wnj . Jnn
:Shndng jioy chbnsh .
Wng L . 1999. Gdi Hny (jiodng chngpi bn) (). 4vols.
Bijng :Zhnghu shj.
Wng L etal. 2000. Wng L G Hny zdin . Bijng :Zhnghu
shj.
Wng Ynzh . [1798] 1956. Jng zhun sh c:f Sn Jngsh b j zi b
. Bijng :Zhnghu shj.
416 References
western and eastern treatments of voiceless expressive binomes, 78, 214, 241
resonants, 78, 112116, 165166, 280, 320, See also alliterative binomes
387n35
diphthongization, Fng yn , 37, 162, 391n89
caused by pharyngealization, 215, 241, 243245, fnqi , 18, 380n3, 381n10
247 and the chngni distinction, 204, 216
induced by voice qualities, according to Ferlus, as spelling technique, 1011
71 initial spellers, 16, 7475, 389n61, 390n68,
of long vowels, according to Pulleyblank, 396n26
6869 faute-de-mieux phonetic elements. See phonetic
of rounded vowels, 25, 200, 204, 209 elements
preceding acute codas, 25, 200, 251252, 255, feet (metrical), 5253, 233, 243
279280, 293 Ferlus, Michel,
preceding velar codas, 215, 241, 243245 explanation of Old Chinese syllable type distinc-
dissimilation, 218 tion, 7072
affecting labial codas, 195, 218, 250251, 310, work on Vietic, 93, 100, 177, 385n21, 385n22
314 final (ynm ), 12, 1620, 198207
affecting preinitial *p, 207 Foe (language of New Guinea), 391n86
affecting rounded vowels, 218, 247248, 293 front-vowel hypothesis,
of *K- to *K- before codas *-w, *-m, and *-p, and rounded-vowel hypothesis, 203205
107, 296, 312, 314, 386n30, 389n62 MC division-IV finals and division-IV chngni
See also assimilation finals from front vowels, 210, 303,
disyllables 397n33
counting for one foot in verse, 53 predicting new rhyming distinctions, 25, 204
presence of in Old Chinese, 50, 318319 reinterpretation of Karlgrens rhyme reconstruc-
See also monosyllables tion based on, 201
divisions (dng ), 16 See also rounded-vowel hypothesis
and pharyngealization, 43, 68, 74 function words,
and prevocalic -r-, 213 as minor syllables 5253
Dng Tngh , Fzhu dialect, 47, 8889, 124, 383n2,
reconstruction of final voiced stops *-g, *-d, and 387n44
*-b, 197, 403n115
reconstruction of the semivowel *-i - in Old gis shsn Yun , 395n19
Chinese, 43 Gn dialects, 87, 124, 390n68
reconstruction of the traditional Y and Go Mng , 115, 230, 234, 278
Tn groups, 306, 314, 403n112 Go Yu , 265266
substitution of *m - for Karlgrens *m-, 112 Garo language, 155, 228, 324
three-way manner distinction projected back Germanic languages, 2, 40, 52, 61, 325
onto Old Chinese, 81 Gngyng zhun , 37, 7273
Drung (Dlng ) language, 383n1 Grammata Serica recensa (GSR), 6, 29, 379n4
Dun Yci , 24, 238, 390n71, 399n57 See also Karlgren, Bernhard
duzhun , 253254 grave initials, 15, 19, 284288
See also pngzhun Greek, 2, 40, 85, 262263
Dlng . See Drung Gungyn , 1620, 221, 380n2, 380n5,
388n50
Early Middle Chinese, 15, 16, 213 correspondences with Old Chinese rhymes in
Edmondson, Jerold A., 37, 96 Qng scholarship, 22, 2425
Einsteins theory of relativity, 5 differences with Qiyn rhyme nomenclature,
r y , 179, 392n101, 397n27 18, 20, 381n10
Evans, Jonathan, 74 failure to distinguish dzr- and zr-, 15, 389n61
excavated documents, 4 loss of MC -j- after retroflex sibilants, 7475,
graphic variation in, 62, 112, 165 145, 236
loan characters in, 63, 65 phonological gaps, 248, 265, 276, 283, 296,
phonetic elements in, 30, 66, 115, 148, 238 384n10, 396n29
See also bamboo strips, Gudin, Mwngdu, surrogate for the Qiyn, 9
Shnghi Museum tngyng rhymes in, 212
422Index
ji nine as the original graph for elbow, lateral initials, 165166, 173, 182183, 190
3132, 80, 155 date of changes affecting, 36
mn replacing hn as phonetic in hear, 63 palatalization in type-B syllables, 70, 77
rn as part of the early graph for qin pharyngealization blocking palatalization of, 75
thousand, 239 retained in Vietnamese, Hmong-Mien,
yu moon and x night not distin- Proto-Mn, Wxing, 109
guished, 64 xishng contacts of, 106
Jyn , 158 lenition
in Fzhu, 47, 89
kiku , 203 in Mn, 8788, 94
See also hku in Vietnamese, 319
Kam-Sui languages, 3637 See also softening, spirantization
Karlgren, Bernhard, Leong, Sow-Theng, 33
and Qng scholarship, 23 Lepcha language, 128
and the chngni distinction, 205 Liz , 391n80
and the front-vowel hypothesis, 204 Li Fang-kuei (L Fnggu ),
and word families, 4 failure to account for the chngni distinction in
Archaic Chinese, 23 Old Chinese, 205
contrast between voiceless and voiced stops in failure to distinguish *-a-s from *-a-s, 220
coda position, 197 reconstruction of division-IV finals, 203, 204
failure to distinguish *-a-s from *-a-s, 220 reconstruction of final voiced stops *-g, *-d, and
medial *-i - ( yod), 43, 6869 *-b, 197
position of the Mn dialects, 84 reconstruction of *-j- for Karlgrens *-i -, 43, 68
reconstruction of consonant clusters, 42, 112 reconstruction of Old Chinese rhymes based on
reconstruction of final *-r, 254 traditional analysis, 22
reconstruction of OC stop series, 81 reconstruction of voiceless resonants, 112
reliance on the standard script, 4 L j , 38, 214, 259, 264, 388n46
See also Grammata Serica recensa L Jiho , 209, 388n52
Kji dialects. See Hakka Li X , 37, 113, 267
Khmer language, 319 loan characters (jiji z ), 26, 32, 63, 109,
Khwrazm, 260 129, 139, 155, 209
Kngz Sh ln . See Shnghi loanwords, 9397
Museum early layers of loanwords as surrogates for late
Kra-Dai languages, 3, 3637, 144, 323 Old Chinese pronunciations, 84
Nnyu kingdom as the earliest source of into Mn dialects, as explanation for irregularity,
Chinese loanwords to, 34 8991
type-A velars reflected as uvulars in, 70 preservation of pre-initial consonants in, 37
Kuki-Chin languages, 61 to Hmong-Mien, 3536, 8182, 9596
to Kra-Dai, 3637
labiodental initials (qngchnyn ), 15, to Vietic, 3435, 91
397n29 See also borrowing
Li Gulng , 391n80 Lokakema (Zh Lujichn ), 258
Lakkia language, lowering of vowels, after pharyngealized initials,
agreement with Vietic and Mn, 152153 211212, 215, 245, 247
rightward spread of nasality, 92 L-Thongkum, Therapan,
simplification of consonant clusters in Chinese work on Lakkia, 37, 96, 154, 172, 184185
loanwords, 71, 94, 9697, 154, 163164, work on Mienic, 94, 96, 177, 215
184186 L Dmng , 3, 9
simplification of indigenous consonant clusters, Lu Chngpi , 399n60
3637 Lushai language. See Mizo (Lushai) language
See also consonant clusters L sh Chnqi , 265
Lakkja language, See Lakkia language
Loz , 209210, 234, 238, 252, 278, Maddieson, Ian, 74
281, 292 Maleng language, 126, 319
Late Middle Chinese, 9, 13, 396n29 See also Vietic
424Index
supposed primitive characteristics of, 317318 See also phonetic elements, xishng
typological characteristics of, 318319 connections
onsets, 68 Pittayaporn, Pittayawat, 3637, 100, 379n6,
inferred from comparative evidence, 9798 382n22, 385n20, 391n88
pharyngealized vs. nonpharyngealized, 43, See also Proto-Tai
6876 Pong language, 126, 160, 163, 319
oracular inscriptions, 12, 4, 29 See also Vietic
See also Shng dynasty graphs postcodas, 194197
Ostapirat, Weera, 3637, 124, 134, 379n2, 385n20 all instances of *-s suffixal, 30, 51, 5859, 197,
321
palatalization in type-B syllables, 7680 as a structural position in Old Chinese main
attributed to preinitial *t, 48, 79, 157 syllables, 51
blocked by *-r-, 76, 78, 99, 102, 291 Indic transcriptional evidence for *-s, 196197
contrast-enhancing, in nonpharyngealized *- as the origin of the Middle Chinese
onsets, 76 shngshng tonal category, 195196
in loanwords, 7677 *-s as the origin of the Middle Chinese qshng
of alveolars, 70, 76, 99, 105, 142 tonal category, 196197
of laterals, 70, 77, 109, 180 pre-Qn documents,
of rhotics, 77 first Shng rulers name written as tng or
of uvulars, 77 kng, 166
of velars before *i or *e, 33, 70, 7779, 105, mng and lng not clearly distinguished,
142, 161, 233 217
velars which do not palatalize before front wi awe-inspiring and wi fear (v.);
vowels, 78, 104, 161, 235, 267 threaten written as , 151
pngzhun , 234 yu moon and x night not
See also duzhun distinguished, 64
Pn Wyn , 28, 44 y proper, refinedand xi great not dis-
parentheses, in Old Chinese reconstructions, 7 tinguished, 121
pharyngealization, 6876 See also pre-Qn script
and prevocalic *-r-, 213215 pre-Qn script,
and type-A syllables, 43, 6876 as evidence on pronunciation, 4, 2632
effect on onsets, 75 institutional and cultural supports not present
effect on vowels, 211212, 245 in, 62
evidence for, 70, 7273 lack of centrally defined standard, 62
regions where attested, 74 learnability, 63
typological artificiality of reconstructing, 7374 nature of, 6267
See also A/B syllable type distinction responsive to linguistic change, 63
phonetic elements, usefulness as evidence, 2832
faute de mieux, 6667, 302, 314 See also pre-Qn documents
in the pre-Qn script 6267 prefixes,
See also phonetic series, xishng connections *k- in nonfinite verb forms, 57
phonetic series, *k- in other verb forms, 57, 159, 164
as evidence for complex onsets, 116193 *m- deriving agentive/instrumental nouns out of
as evidence for rhymes, 261, 306 verbs, 55, 5960, 124
cooccurrence of MC velars and laryngeals in, 28, *m- deriving volitional or causative verbs out
4445, 100101 of nonvolitional verbs, 5455, 123126,
having only the MC '- initial, 44 146, 188
lack of contact between homorganic stops and *m- deriving volitional verbs out of nouns, 55,
fricatives in, 44 132
lateral and uvular sources of MC y- distinguished *m- in names of animals, 5556, 100, 125, 127,
in, 106 132133
lateral-initial series, 106, 109, 166, 184, 276 *m- in names of body-parts, 55, 132, 188, 192
Old Chinese onset distinctions not detectable *s- deriving circumstantial nouns, 56, 127, 136,
in, 84 150, 142
uvular series, 2728, 44, 4546, 66, 100101, *t- in certain intransitive verbs, 5657, 158
128132 *t- in inalienable nouns, 32, 57, 155, 157
426Index
type-A and type B syllables, 16, 43, 6869 rhyme groups (ynb ), traditional, 2225,
Xinbi and Whun as *Srbi and 199202
Avar, 262 rhyme tables (ynt ), 9, 1319, 198, 205,
214216
Qing language, Northern, 74 rightward propagation of nasality,
Qiyn , 3, 910 in Hdng dialect, 158
differences from the Gungyn in rhyme in Lakkia, 3637, 92, 96, 163
nomenclature, 18 in southern Mn, 92, 173
finals occuring only in qshng in the Qiyn, roots,
272 and word families, 5960
information about dialect diversity, 14 doubly prefixed, 54
Mn dialects features not explainable in terms of Old Chinese root structure, 5053
the Qiyn system, 32 related roots, 39, 6061, 108109, 139, 148, 156
not representing a single dialect, 12, 322 See also affixes, infixes, prefixes, suffixes, word
preface, 12 families
type-A and type-B velars spelled with different rounded-vowel hypothesis, 25, 200, 203, 270,
fnqi spellers, 75 393n1
Wng Rnx version, 910, 15, 18, 75, R Chn , 266
236, 388n50 Rc language, 34
Qn (state in ancient China), 43, 162 Chinese preinitial *s reflected as preinitial /t/ or
Qng dynasty philologists, 2, 3, 2225, 3839, /t/, 118119, 136137, 142, 187
207, 253 convergent with Lakkia and Mn, 37, 97,
Qi Xgu , 8, 2930, 154, 392n94 152153, 184
onset correspondences with Middle Chinese and
Ratliff, Martha, Proto-Mn, 95
reconstruction of Proto-Hmong-Mien, 35, 94, onset correspondences with Vietnamese, 94
387n40, 387n43, 392n91 retention of Austroasiatic preinitials, 47, 9394
tightly and loosely attached prenasals in retention of Chinese preinitials, 37, 47, 71, 91,
Proto-Hmong-Mien, 95 96
reconstruction, See also Vietic
goals of, 67 Russian language, 69
internal, 4, 81, 197
methodology, 46 Sch language, 34, 153, 168, 319
regulated verse, 206, 395n19 See also Vietic
related roots. See roots Sagart, Laurent (Sh Jir ), 140, 185
retroflexion, on blood, 152, 240
and loss of MC - j-, 74, 233, 380n9, 382n1 on final laryngeals preserved in a Shnx dialect,
effects on onsets of medial *-r-, 8081, 99, 102, 318
213 on the date of changes affecting laterals, 36, 110
effects on rhymes 213218, 227 on the names of the four directions, 147
Rhodes, Alexandre de, 257, 384n8, 399n61 Salishan languages, 74
rhotics, Sn gu zh , 261
effects of *-r- on vowels, 213219 Sanskrit, 40, 85, 197, 258
medial *-r- leaving no trace in Middle Chinese, Schlegel, August Wilhelm von, 317
214215, 223, 229, 230, 242, 243 Schuessler, Axel, 33, 79, 161
merger with laterals after nasal preinitials, 122 script. See pre-Qn script
nonlateral reflexes of OC *r in Wxing, Mn, Sh Jir . See Sagart
Vietnamese, Hmong-Mien, 109110 Shng dynasty graphs,
rhyme, poetic, b nose and z self (adv.); follow; from,
and date of early loans to Hmong-Mien and Bi, 142
246 ch front teeth, 158
as evidence in Old Chinese reconstruction 2, cn thumb; inch, 155
2026, 194315 fng container, box, 143
rhyme books (ynsh ), nio urine, 286
structure of, 910 rn person, 286
See also Qiyn, Gungyn qin thousand, 239
428Index
Shng dynasty graphs (Cont.) rhyme sequences with labial codas infrequent
sh corpse and y foreigner, 115, in, 302
285286 rhymes not rhyming in modern pronunciation,
sh excrement, 286 21, 24
t reach to, 133134 rhyming of li, 230231, 392n102
wng flee; disappear; die used with the sense textual problems in, 208209
of w not have, 227 Wng Ls analysis of Old Chinese rhyme
wn hear, 63 groups, 23
w seventh earthly branch, 129 Wng Ls treatment of Dng and Qn as a
x west and ni then, 146147 single rhyme group, 23, 250
y rub, repress and seal, 240 words with MC - ng which rhyme as *-in, 237
yu right hand, yu have, exist, words with phonetic rhyming as *-u , 144
zhu elbow, ji nine, 155 Sh mng , 37, 113114, 148, 267
~ y prison, 130 Sh (region in ancient China), 307
zhn needle and sh ten, 154 Shuhd documents, 63, 386n31, 391n75
See also oracular inscriptions Shuwn jiz , 37, 129, 147, 238,
Shnghi Museum documents, 261, 269, 390n71
Kngz Sh ln, 209, 276, 278, anachronistic use of, in reconstructing Old
396n28 Chinese pronunciation, 3
Xng qng ln , 66 dialectal pronunciations in, 4243, 140, 162
Z y , 139 gwn ancient script forms in, 28, 63,
Shng sh , 39, 234, 278, 280 383n7, 401n82
Shn Do , 38 merger of *m.q- with *- reflected in, 130
shngm . See initial sound glosses in, 130, 138, 148149, 150, 203
Sh j , 66, 260, 263 Siamese (Bangkok Thai), 192, 232, 235
Shjng , signific, 159, 390n66
*-r rhyming separately in parts of, 265 Sinitic, 2, 3334, 40
as evidence used by Karlgren in his Archaic Sino-Japanese, 11, 13
Chinese reconstruction, 2 Sino-Korean, 11, 13, 380n4
diphthongization of rounded vowels after acute Sino-Tibetan, 40
initials postdating Shjng rhymes, 251 OC *- and Garo - ang, 228
distinction between the Zh group and the role of evidence from, 323325
Wi group, 284286 *s- clusters in, 112
Hn version, 208 suggesting velar nasal original in w < nguX,
Mo version, 129, 148, 158, 160, 179, 208, 128
259, 399n57 Sino-Vietnamese, 11, 13, 35, 380, 384n12
merger of *-a-s and *-ak-s partially reflected in See also Vietnamese
rhyming, 227 six-vowel system of Old Chinese reconstruction,
merger of *-p-s and *-t-s partially reflected in 198211, 213, 231, 250, 304
rhyming, 196 softening of initial consonants
modern reconstructive work often not based on a as intervocalic lenition, 8889, 94, 173174
direct examination of the rhymes, 3 aspirated stops not subject to, 175, 177178,
need to account for which words do not rhyme 181, 183
with each other, 21 due to preinitials in Vietic, 4748
need to account for which words rhyme with in northern Mn, due to loosely attached
each other, 21, 198 preinitials in Proto-Mn, 4647, 8889
primary rhyme corpus traditionally used in Old in northern Mn literary words, 90
Chinese reconstruction, 2021 See also lenition, spirantization
problems with the traditional analysis of rhymes, Sogdian language, 263264
2425 solar eclipse of May 29, 1919, 5
quoted in the Mwngdu W xng , 160 Solnit, David, 37, 96
received version reflecting late sound changes, 259 Sng (region in ancient China), 266
rhyme sequences involving words with labial spirantization, 385n21
codas, 304307 analogous to Northern Mn softening, 47
rhyme sequences showing dialectal merger of caused by a preinitial lost in Vietnamese, 47,
*-p and *-m to *-k and *-, 306307 168169
Index429
Yqi jng yny (by Xunyng A/B distinction as vowel length, 6970,
), 240 383n1
ynb . See Old Chinese, traditional rhyme nasal initial in shu hand, 115
groups six-vowel reconstruction, 213, 383n4, 393n1
ynm . See final voiceless sonorants reconstructed as clusters of
Y Xngw , 390n66 *s- + sonorant, 112
Zh Lujichn . See Lokakema
Zng Hu Y tomb, 388n52 Zhu vessels,
Zng Ynqin , 26 Y dng , 153
Zhngu period. See Warring States period Gng Yng dng , 314
Zho Chng , 147, 386n30, 401n90 Zhu Zm , 12, 72, 223, 383n6, 399n60
Zho Tu , 34, 36 Zhu Y , 259
Zhng Xun , 37, 76, 208, 262, 264265 Z y . See Gudin, Shnghi
Zhngzhng Shngfng , 236, 381n15, Museum
396n23, 397n36 Zu zhun , 11, 145, 255