See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/269993584

Compounds in English, in French, in Polish, and in General

Article · January 2013

CITATIONS READS

14 934

1 author:

Pius ten Hacken

University of Innsbruck
101 PUBLICATIONS 305 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

PhD candidate View project

Compounding View project

All content following this page was uploaded by Pius ten Hacken on 05 January 2015.

The user has requested enhancement of the downloaded file.

 Compounds in English, in French, in Polish, and in General
Pius ten Hacken

The notion of compound can be taken as a theoretical concept only if it has a precise
definition. In many current discussions it is assumed that such a definition is not
available or not possible. Here, I will show how translation can be used as a heuristic
to determine a concept of compound that is semantically coherent. This concept
includes genitive constructions and constructions of a relational adjective with a
noun, but not prepositional constructions in which the preposition expresses part of
the meaning. An essential component of the use of compounds for naming is shown to
be onomasiological coercion.

Keywords: compound, definition, genitive, relational adjective, preposition,

onomasiological coercion

1. Introduction

It is often claimed that the boundaries of concepts such as compound should be seen in terms
of a continuum between clear compounds and clear members of another class, in this case
derivation or phrase. Here I will argue that this approach is neither necessary nor desirable. I
will start by discussing the nature of theoretical concepts, taking compound as an example
(section 2). Then, I will present some preliminary considerations that have to be taken into
account when discussing compounding as a word formation device (section 3). Section 4 sets
out a methodology by which we can arrive at a set of defining properties for compounding.
Sections 5, 6 and 7 present some results that can be achieved in using this methodology.
Finally, section 8 summarizes the conclusions we can draw from the reasoning in earlier
sections.1

2. Compound as a theoretical concept

The dictionary definition in (1) provides a convenient starting point for the discussion of the
definition of compound.

(1) [Compound:] a word made up of two or more words

This particular definition is from Chambers (1998), but similar definitions can be found in
other dictionaries. The difficulty of applying (1) to a particular expression to determine
whether it is a compound arises from the two occurrences of ‘word’ in (1). Both of them
serve to characterize compound as ‘marked’, less common. The first occurrence refers to the
result. It contrasts with phrase, which is a less marked outcome of the combination of two or
more words. The second occurrence refers to the components. It contrasts with affix in the
sense that complex words are more typically put together from a word and an affix. It is these
two boundaries that Lieber & Štekauer (2009: 4) identify as problematic in (2):

(2) [W]e cannot always make a clean distinction between compound words on the one
hand and derived words or phrases on the other.

97
The fact that (2) is found in the introduction to a Handbook of Compounding shows that
uncertainty about the exact boundaries of the concept is part of many people’s perception of
the current state of research. The formulation in (2) leaves open whether this lack of a ‘clean
distinction’ is due to disagreement between researchers or accepted fuzziness of the concept.
An example of the former situation is the term government. In the framework of
Chomsky’s (1981) Government and Binding framework, a number of different definitions of
government have been proposed. Linguists working with different definitions did not agree
on whether in a particular structure a node  would govern a node . However, no linguist
would claim that it is not clear whether  governs . They would argue for or against a
particular definition, but each definition would decide quite clearly whether  governs  or
not. Newmeyer (1986: 204-205) identifies the question of how government is best defined as
one of the most important theoretical issues in GB-theory at the time.
An indication that in the case of compound the issue of definition is different from the
case of government is Dressler’s (2006: 24) statement in (3):

(3) [U]niversal definitions [of compounding] are not only theory-dependent […] but also
cross-linguistically never watertight—in many languages there are exceptions or
fuzzy transitions to non-compounding[.]

What (3) suggests is that Dressler sees the entire idea of a universal definition of
compounding as unrealistic. His objections are that the definitions that have been proposed
are first that they are theory-dependent and second that they have counterexamples
(‘exceptions’). He seems to suggest that linguists should accept that the transition between
compounding and non-compounding in many (or at least some) languages is fuzzy. This is
reminiscent of the position Bloomfield (1933: 223-224) takes with respect to the boundary of
inflection and derivation when he states that “[t]his distinction cannot always be carried out.”
In psycholinguistic research, it has long been accepted that general-language concepts
such as cup have fuzzy boundaries. In a famous experiment, Labov (1973) asked informants
to classify objects as cups, bowls, and vases. He discovered that the boundaries between these
concepts was fuzzy in the sense that there was no specific point at which all informants
placed the boundary between cup and bowl. Instead, the judgements would vary between
informants and, for the same informant, between instances in different contexts. On the basis
of this type of observations, Rosch (1978) proposed her prototype theory. For concepts like
cup and bowl, each speaker has a prototype, which is encoded as information about the
concept in the mental lexicon. Judgements on categorization are based on the perceived
distance to prototypes. The perception of the distance can be influenced by various factors,
such as which similar items the speaker has recently been asked to categorize.
My claim here is that whereas natural concepts are prototype-based, theoretical
concepts are not. If we accept (3) and do not see (2) as a challenge to change the situation, we
abandon compound as a theoretical concept. This means that we cannot make any testable
claims on compounds, but only use the term in a pre-theoretical sense. Only on the basis of a
proper definition of compound is it possible to evaluate, for instance, whether a particular
theory of compounds can account for all instantiations. It should be noted that if we choose to
abandon compound as a theoretical concept, this does not solve the more general definition
problem. Without compound, we will need other theoretical concepts in order to make
testable claims about them. These concepts also require proper definitions.

98
 In the last paragraph, I changed the focus from definition to proper definition. The
reason for this specification is that some definitions are not suitable for use as a device to
determine the boundaries of a concept. A definition such as (1), while perfectly adequate to
evoke a prototype, is too vague to determine the boundaries of compounding. In Dressler’s
words, it is not ‘watertight’. A proper definition is one that has no exceptions and takes a
decision in borderline cases. Whether a definition has exceptions is in part a matter of how
we use it. If the definition of a theoretical concept clashes with intuitions we have about that
concept, we must accept the categorization that the definition gives. Here a parallel with
zoological taxonomy is enlightening. In many people’s intuition, cat, dog, bird, and fish are
approximately equally specific in designating animal species. Zoologically, however, cat and
dog are quite closely related species (both are carnivores), whereas bird is a class (i.e. at the
same taxonomic level as mammal) and fish constitutes a paraphyletic group, i.e. a group that
does not correspond to any node in the taxonomy. We should not see the intuitions as
counterexamples to zoological taxonomy, but rather accept that theoretical concepts of
zoology are not the same as general-language concepts with the same names.
The fact that a theoretical definition cannot have counterexamples does not mean that
we cannot evaluate it. In fact, my use of proper definition already implies an evaluation. A
definition is not proper if it does not impose a decision in borderline cases. We can also
discuss how good a definition is. However, this type of evaluation is a matter of judgement
rather than one of testing. A good definition of a theoretical concept identifies a useful
concept. In this sense, competing definitions of compounding are no different from
competing definitions of government in Chomskyan syntax or of species and other taxonomic
levels in zoology.
Dressler’s claim in (3) raises two specific issues in relation to definitions of
compounding. One is that definitions are theory-dependent. Although (3) suggests that this is
somehow problematic, it is in fact a general property of definitions, because a theory-free
definition would only rely on unaided primary intuitions. For a theoretical concept it is not
problematic and indeed inevitable that the definition refers to other theoretical concepts. As
mentioned in ten Hacken (2008), in a field such as mathematical linguistics the terms
constitute a network held together by the definitions. This network encodes essential parts of
the knowledge of the field.
The other point Dressler raises in (3) is the cross-linguistic validity of definitions. It
may seem attractive to define compounding on a language-specific basis, because compounds
in the same language have more properties in common. They constitute a more homogeneous
class, so that it is easier to find criteria that distinguish them. However, with any set of
logically independent criteria, there may be borderline cases where some of them are met but
others not. The choice of the most important criteria remains arbitrary when considered in the
space of one language, but taking into account other languages may reduce the arbitrariness
of the selection of criteria. Using cross-linguistic data comparison to decide on language-
specific issues is a technique that is well-known from Chomskyan linguistics, e.g. Chomsky
(1986: 37-38).
Therefore, it is worth looking for a proper definition of compounding. Such a
definition takes priority over any individual judgements or opinions, determines the
boundaries of the concept, uses other theoretical concepts, and is cross-linguistically valid.

99
3. Compounding as a word formation device

In searching to define compounding, it is essential to keep in mind that it is a word formation

process. This statement may seem trivial, but it is sometimes not sufficiently taken into
account when a set of rules for compounding is proposed. Superficially, we could take a
compound such as doghouse to be generated by a rule such as (4):

(4) NNN

Selkirk (1982: 16) proposes an account of compounding that takes a set of rules such as (4) as
its basis. She then notes that compounds in English are right-headed (1982: 20) and that the
non-head specifies the head in some way (1982: 22). However, this gives a very limited sense
of the meaning of a compound such as doghouse. A prototypical doghouse may be an object
such as in Fig. 1.

Figure 1 Doghouse (or kennel in the UK)

The object in Fig. 1 is not a typical house. It is a small structure for a dog. A rule such as (4)
and any supplementary indications on the possible meanings of compounds will not be
sufficient to account for the fact that doghouse refers to an object such as Fig. 1. The point is
even stronger for compounds such as spaceship. A spaceship is not at all a ship.
The explanation why a doghouse is an object such as in Fig. 1 depends on the fact that
compounding is a word formation device. As such it is used to name new concepts. The
reason why doghouse refers to objects such as Fig. 1 and not to ‘any house that is in some
way related to dogs’ is that speakers of English perceived the need to name the concept of
Fig. 1. This concept then contributed essential attributes to the meaning of the compound. We
can call this phenomenon onomasiological coercion.
Lexicalization has often been treated as a circumstantial property of compounds.
Thus, Levi (1978: 10-12) excludes lexicalized compounds from the scope of her theory.
However, it is actually a central property of compounding that it is used to produce
lexicalized expressions. In this sense, the compounding rule (4) is to be distinguished from
the syntactic rule which looks the same. In a language like Dutch, we find the contrast in (5):

(5) a. een wijnfles ‘a wine bottle’

b. flessenwijn ‘bottles wine’, i.e. wine sold in bottles (as opposed to casks)
c. een fles wijn ‘a bottle wine’, i.e. a bottle of wine

100
All three expressions in (5) can be generated by means of (4), but only (5a) and (5b) are
compounds. (5c) is an example of a syntactic construction indicating quantity. The
compounds refer to specific concepts. (5a) is a bottle of a certain shape. (5b) is wine of a
certain (minimal) quality. These properties are added by onomasiological coercion.2

4. Cross-linguistic evidence for a definition

In order to determine which criteria can be used to define compounding, we have to consider
data from different languages. On one hand, it seems straightforward to use translation as a
component of the methodology. On the other hand, we have to be extremely careful in
interpreting the data collected in this way. It is well known that a certain type of equivalence
plays a central role in translation, cf. Munday (2008: 36-54), but this is not necessarily
equivalence at the lexical level. Baker (2011) organizes her entire textbook on the basis of
different levels of equivalence, devoting most of her attention to equivalence at sentential and
textual levels. Functionalist approaches, as described by Nord (1997), play down the role of
equivalence. Thus, in his skopos theory, Vermeer (2000) rates the instructions to the
translator and the coherence of the target text resulting from translation higher than the
correspondence of the target text to the source text.
An example of the problems that can arise in using translation is the triple of English,
French, and Polish in (6):

(6) a. doghouse
b. niche
c. buda dla psa ‘shed for [a/the] dog’

In all trilingual examples, the (a) example is English and the (b) and (c) examples are the
French and Polish translations, respectively. Whereas (6a) is a prototypical example of a
compound, (6b) in French is a simplex word which can also mean ‘recess, alcove’. Polish
(6c) involves the preposition dla (‘for’). In interpreting the Polish translation, one should
keep in mind that Polish does not have articles, so that (6c) can be translated as ‘shed for a
dog’ or ‘shed for the dog’. Obviously, (6b) is not a compound, because it is not possible to
identify two components. For (6c), it is at least not obvious that it is a compound, because
there seem to be three components. We will come back to such prepositional constructions in
section 6.
As will be shown in subsequent sections, however, provided that it is applied with the
appropriate care, translation can be a very useful technique in the identification of properties
of compounds in different languages. The starting point for the data-oriented component of
this research was the list of compounds Levi (1978: 280-284) gives as an appendix to her
monograph on what she calls ‘complex nominals’. In Levi’s theory, complex nominals
include both noun-noun compounds and combinations of relational adjective and noun. In the
first instance, only the 257 noun-noun compounds were considered. They were translated into
French and Polish by native speakers of these languages with a background in translation.
The collection of compounds by Levi (1978) has the advantage of giving a systematic
classification of meanings. However, some of them were quite specific to the geographic,
temporal and intellectual environment Levi was working in and were not understood by the
translators. They were discarded from the analysis. The distribution of the translations among
different constructions is represented in Fig. 2.

101
 Figure 2 Distribution of translations in French (left) and Polish (right)

The most striking feature of the pie charts in Fig. 2 is the low proportion of noun-noun
compounds. In French, such compounds are not readily accepted, although some have
become quite common, e.g. centre-ville ‘centre-town’, i.e. town centre. An example from our
data set is prix plafond ‘price ceiling’, i.e. ceiling price. In Polish, there was only a single
case, płetwonurek ‘flipper-diver’, i.e. frog man, in our data set. As Szymanek (2010: 219)
states, such Polish compounds have a linking element, typically -o-, and are generally quite
rare. The class marked ‘other’ in Fig. 2 includes translations such as niche in (6b), where no
two words can be distinguished, or paraphrases, as in (7) in Polish:

(7) dziura w ubraniu wygryziona przez mole

‘hole in [a] piece_of_clothing chewed by moths’, i.e. moth hole

There are three large classes of translations for English noun-noun compounds in French and
Polish. One is a combination of a noun with a relational adjective, a second combines a noun
with a genitive, and a third connects two nouns by means of a preposition. Together, they
make up around 80% of the translations in French and Polish. They will be considered in
more detail in the following sections.

5. Genitive constructions

An example of genitive constructions in French and Polish corresponding to an English noun-

noun compound is given in (8):

(8) a. car factory

b. usine d’automobiles ‘factory of cars’
c. fabryka samochodów ‘factory carsGEN’

In all three expressions in (8), we can observe a distribution of the meaning among two
components. The genitive markers in French and Polish can be seen as indicators of the

102
construction. Polish has morphological case marking on nouns, so that (8c) has the genitive
plural of samochód ‘car’. In French, morphological case was lost around the 13th century and
the genitive is expressed by means of the preposition de ‘of’. English also has a genitive
marker and it is used in certain types of compounds, as in (9):

(9) a. children’s bible

b. kinderbijbel ‘child bible’

As the Dutch translation (9b) shows, other languages use noun-noun compounds as
equivalents. The form kinder in (9b) is not the same as the singular kind (‘child’) or the plural
kinderen (‘children’), but has an extension -er. In ten Hacken (1994: 258-262) I propose to
analyse kinder as a stem variant of kind, along the lines of Aronoff’s (1994) analysis of the
Latin verbal paradigm. Booij (2002: 22-23) adopts a similar analysis. In a sense, the ’s in
English and de in French can be seen as markers in the same way as Dutch -er in (9b).
Whereas (9a) can be a compound, there are also contexts in which it is clearly not
one. The ambiguity is illustrated in (10):

(10) a. The children’s bible is full of pictures.

b. This book is a children’s bible. It is full of pictures.
c. This bible belongs to the children. It is full of pictures.

Both (10b) and (10c) can be paraphrases of (10a), but they are not synonyms. In (10b),
children’s bible is a single concept, a kind of bible. In (10c), bible and children refer to
separate concepts. The intuition that there is a single concept in one but not in the other is
supported by the possibility of pronominal reference, as illustrated in (11):

(11) a. Anna bought themi a children’s bible, but theyi did not like the pictures.
b. *Anna bought a childreni’s bible, but theyi did not like the pictures.

The pronoun they in (11) cannot refer back to children as part of children’s bible. In (11), no
interpretation of children’s bible along the lines of (10c) is possible, because the article a
does not agree with children. The different interpretations in (10) also correspond to different
structures assigned to the subject NP in (10a), as illustrated in (12):

(12) a. [the [children’s bible]]

b. [[the children’s] bible]

The structure in (12a) corresponds to the paraphrase in (10b) and the one in (12b) to (10c).
This explains that ‘a children’s bible’ in (11) can only have a structure on the pattern of
(12a), because a agrees with bible, but not with children. The structure is also visible in the
placement of adjectives modifying bible, which precede children’s in (12a), but follow it in
(12b).
In French, it is crucial to make a distinction between de with and without the definite
article. An example of this contrast is (13):

(13) a. donneur de sang ‘donor of blood’

b. donneur du sang ‘donor of_the blood’

103
Both expressions in (13) can be translated as blood donor, but (13b) assumes that sang
(‘blood’) has been introduced as an entity in the context in which it is used, whereas (13a)
only introduces a single concept. This contrast is of the same type as illustrated for (10) in
English. The syntactic contrast in (13) suggests that we can use definite articles as a
diagnostic. However, in some contexts the definite article can also be used to indicate
genericity, as in pays des ours ‘land of the bears’, i.e. bear country. For Polish, Szymanek
(2010: 218) considers genitive constructions as syntactic phrases. The absence of articles in
Polish makes it more difficult to distinguish the two readings. Therefore, the interaction of
genitive constructions with articles is more complex and cannot be used in all cases as a
criterion to distinguish compounds from phrasal expressions.
In view of these data, we can take two positions. On one hand, we can consider
genitive constructions as ambiguous between a phrasal and a compounding interpretation.
This requires a more precise characterization of the boundary between these two
interpretations, but, as illustrated in (8) and (9), it increases the cross-linguistic
correspondence between compounds as translational equivalents. Compounds in this
interpretation are primarily naming units and the categorization takes semantic criteria into
account. The semantics of the criteria is above all the reference to a single as opposed to two
distinct entities by means of the genitive construction. On the other hand, we can consider
genitive constructions as phrases. This is the position adopted by Szymanek (2010). In this
perception, compound is a purely formal category. The consequences include an increased
number of phrasal naming units and cross-linguistic mismatches in the sense that compounds
in one languages correspond to phrases in another.

6. Relational adjectives

Relational adjectives (RAs) are adjectives that do not express a property but rather a relation
to a corresponding noun. As shown in Fig. 2, combinations of a noun with a relational
adjective are often used in the translation of English compounds, in particular in Polish. An
example of an English compound with N+RA equivalents in French and Polish is given in
(14):

(14) a. cell division

b. division cellulaire ‘division cellADJ’
c. podział komórkowy ‘division cellADJ’

The suffix -owy, as illustrated in (14c), is the most frequent source of relational adjectives in
Polish. In this case, komórkowy is derived from komórka ‘cell’. It is hardly possible to assign
a specific meaning to such adjectives. Dictionaries often use formulas such as ‘or or relating
to [N]’ (with N the base noun) to characterize their meaning. Kallas (1999: 485-494)
distinguishes 18 different categories of meaning for denominal adjectives and lists -owy as a
suffix for 17 of them. The actual category of meaning depends not only on the adjectival
base, but also on the combination with the noun. In the same way as for N+N compounds,
onomasiological coercion can result in a not fully compositional meaning. Thus, krem
orzechowy ‘peanut butter’ combines krem ‘cream’ and orzech ‘nut, walnut’ to name a
substance that is not really a cream and is based on a very specific type of nut.
In French, cellulaire in (14b) is related to cellule ‘cell’. The status of the suffix -aire
is less obvious than the corresponding suffix -owy in Polish. It corresponds to Latin -arius

104
and there are also adjectives such as pécuniaire ‘moneyADJ’ that do not correspond
morphologically to a French noun, cf. argent ‘money’. Historically, pécuniaire was borrowed
separately from Latin, where it does have a regular relationship to the noun pecunia ‘wealth,
money’. The pair argent – pécuniaire is analogous in its use to cellule – cellulaire, although
from the point of view of its formation it is suppletive.
For English, Levi (1978) makes the case for treating RA+N combinations in the same
way as N+N compounds. The existence of synonym pairs such as atom bomb and atomic
bomb gives such an analysis an initial plausibility. Many of Levi’s arguments are framed in
the terminology of Generative Semantics, so that they cannot be used easily outside that
framework, but a striking syntactic property is coordination as in (15):

(15) a. literary and musical criticism

b. *literary and bitter criticism
c. solar and gas heating
d. *solar and reliable heating

Based on Levi’s (1978: 22-23) examples, (15) gives two contrasting pairs. In (15a-b) we see
that the RA literary can be coordinated with another RA, e.g. musical in (15a), but not with
qualitative adjectives such as bitter in (15b). This means that RAs and regular, qualitative
adjectives belong to different categories for coordination. In (15c-d) it is shown that the RA
solar can be coordinated with the noun gas as the non-head of a compound, but not with the
qualitative adjective reliable. This can be taken as supplementary evidence that RA+N
combinations are in the same category as N+N compounds.
A point that is often mentioned as a counterargument to a compounding analysis of
RA+N combinations is that the RA agrees with the N. This can be seen most clearly in
Polish. An example is the contrast in (16):

(16) a. Bożena lubi krem orzechowy.

‘Bożena likes creamACC nutADJ-ACC’, i.e. Bożena likes peanut butter.
b. Bożena nie lubi kremu orzechowego.
‘Bożena not likes creamGEN nutADJ-GEN’, i.e. Bożena does not like peanut butter.

In Polish negative sentences, the object appears in the genitive instead of the accusative. In
(16b), it is not only the noun krem but also the RA orzechowy which appears in the genitive.
Therefore, the N+RA combination krem orzechowy ‘peanut butter’ is inflected in two places.
For Matthews (1974: 35), this is a reason to discard N+RA combinations from the category
of compounds. He gives Latin tribunus militaris ‘tribune military’, i.e. a particular rank in the
ancient Roman army as an example.
As with the genitive constructions in section 4, we can take two positions as to the
RA+N combinations discussed here. One is to consider them as compounds, the other to
analyse them as phrases. If we take them to be compounds, the notion of compound becomes
more semantically coherent, the link between compounds and naming units is emphasized,
and there is a much larger degree of cross-linguistic correspondence. If we take them to be
phrases, we make compound into a more formally oriented category, thus creating the need
for more phrasal naming units and a considerable set of cross-linguistic mismatches.
Matthews (1974: 35) proposes that Latin N+RA combinations are idioms, treated as units in
lexicography, but not in morphology. In line with such an analysis, Szymanek (2010: 219)
concludes that “the number of nominal compounds is not so spectacular in Polish.” In our

105
data set, there is only one such compound as a translation for the 257 N+N compounds in
Levi’s set.

7. Prepositions

Prepositional constructions occur quite frequently among the translations of English N+N
compounds in our data set, as shown in Fig. 2 above. An example is the Polish translation in
the triple in (17):

(17) a. love song

b. chanson d’amour ‘song of love’
c. piosenka o miłości ‘song about love’

Both the French (17b) and the Polish (17c) involve a preposition, but these prepositions have
a very different status. In French, de is the preposition marking the genitive. In section 4 it
was argued that the function of de can be seen as parallel to morphological case in Polish and
’s in English. Grevisse (1980: 1146-1169) gives a wide range of uses, including one that has
a striking parallel to the Polish genitive in (16b), as illustrated in (18):

(18) a. Charles a une voiture.

‘Charles has a car.’
b. Charles n’a pas de voiture.
‘Charles NEG has not DE car’, i.e. Charles does not have a car.

The indefinite article une of the positive sentence in (18a) is replace by the genitive
preposition de in the negative counterpart (18b).
In (17c), the preposition o is followed by the locative case. As such, it is analogous to
the use in (19):

(19) Dariusz i Ewelina rozmawiają o koncercie.

‘Dariusz and Ewelina talk about concertLOC’
i.e. Dariusz and Ewelina talk about the concert.

Both in (17c) and in (19), o can be translated as ‘about’. This means that o in (17c) has a
different role than de in (17b). The meaning of (17b) is derived from three factors: the
meaning of chanson ‘song’, the meaning of amour ‘love’, and the concept to be named. The
meaning of de does not play a significant role. This makes it parallel to the English
compound in (17a). In (17c), however, the meaning of the whole expression is determined by
the meaning of piosenka ‘song’, the meaning of o ‘about’, and the meaning of miłość ‘love’.
Whereas onomasiological coercion is instrumental in determining the relation between the
two components of the compounds in English (17a) and French (17b), this relation is
expressed by the preposition in Polish (17c). This means that the way (17c) gets its meaning
is the regular one for syntactic expressions.
When we consider the equivalence between the three expressions in (17) in more
detail, we notice that (17c) can be used as a translation of (17a) and (17b), but does not have
exactly the same meaning. The use of love song or chanson d’amour implies that we consider
it a genre, a type of song. This implication does not exist for (17c). Moreover, the

106
specification of the relation in (17c) is more explicit than in (17a) and (17b). Polish speakers I
asked came up with various alternative expressions that highlight different aspects of the
relation between the two components in (17a) and (17b). It seems, then, that the concept is
much less established in Polish than in English and French.
In the data I collected, there were some striking examples of English compounds that
did not match compounds in French and Polish. One of them is (20):

(20) a. abortion vote

b. vote sur l’avortement ‘vote on abortion’
c. głosowanie za aborcją ‘vote for abortion’

Both French (20b) and Polish (20c) can be correct translations of English (20a), but they are
not correct translations of each other. In fact, (20a) is ambiguous in English. In (21), two
contexts are given.

(21) a. The abortion vote is on the agenda for Monday.

b. Senator X was under huge pressure to explain his abortion vote.

In both sentences of (21), the context is that of a Parliament. In (21a), what is on the agenda
must be a vote on a law or motion pertaining to abortion. The outcome of the vote is
generally not known in advance, so the relation can only be one of ‘on’, as in French (20b).
In (21b), however, it is implied that the vote was one approving abortion. The content of the
sentence indicates that Senator X must have voted and if his vote were against abortion, it
would rather be an anti-abortion vote. Therefore, in (21b), Polish (20c) would be an
appropriate translation. This is not to say, of course, that French cannot express the meaning
in (21b) and Polish the one in (21a). It only shows that my informants translating (20a)
assumed different readings. They could not translate (20a) without choosing one and
apparently different readings came to mind more readily. Therefore, the compound (20a)
does not have a translation as a compound in French and Polish, although in any specific
context where it appears there is a translation that is correct in that context.
While o in (17c), sur in (20b), and za in (20c) are clear cases of meaningful
prepositions expressing the relationship between the two nouns, there is an interesting
borderline case in French with the preposition à. Nicoladis (2002) considers both N+de+N
and N+à+N as constructions that can be analysed as compounds and notes that both de and à
can be used to express possession in French (2002: 49). The discussion of à by Grevisse
(1980: 1118-1145) is even longer than the one of de, but it starts by giving a much more
specific range of meanings for à (1980: 1118), translated in (22).

(22) The preposition à is used above all to mark the place, goal, time, means, manner,
property.3

Although Grevisse (1980: 1118-1119) also mentions the possessive reading, it is much more
constrained in its distribution than the possessive meaning of de. In the data set collected for
this study, the French informant used à in 19 cases. In 7 cases the definite article was
included and in 12 it appeared alone. Interestingly, there was a highly significant semantic
bias among these expressions. The examples in (23) are typical.

107
(23) a. ver à soie
worm PREP silk
‘silkworm’
b. fer à vapeur
iron PREP steam
‘steam iron’
c. gâteau aux pommes
cake PREPDEF-PL apples
‘apple cake’

In the type illustrated in (23a), à marks the right-hand element as the product, in (23b) as the
means, and in (23c) as the content. All of the translations with à can be placed in these
categories and the article occurs for all and only the expressions in the type illustrated in
(23c). This suggests that the preposition à is meaningful in a way that de is not. However, it
would be interesting to study this in more detail.

8. Conclusions

The questions addressed in this article are whether compound can be given a precise
definition and if so, how such a definition can be arrived at. As was argued in section 1, the
answer to the first question depends only on whether we want a definition or not. Since
compound is a theoretical notion, we can choose whatever definition we like and impose it on
the data. The problem is that if we take a definition arbitrarily, the resulting notion may not
be of much use. Therefore, the method for arriving at a definition of a useful concept
becomes crucial.
It is possible to approach compound from different perspectives. If we want to
consider it a purely formal notion, we can take as a starting point the syntactic category of the
components and the stress pattern. However, if we do so, it will be quite accidental if there is
any semantic coherence or cross-linguistic correspondence. I would argue that compound is
interesting above all as a category with semantic and syntactic characteristics. This view
matches Jackendoff’s idea that compounding is “a possible protolinguistic ‘fossil’ in English”
(2002: 249).
The method used here takes as a starting point that translation can be a good heuristic
device to identify relevant cross-linguistic correspondences. Assuming that a definition of
compound has at least some semantic aspects, we should expect that translations will tend to
use corresponding realizations. However, we have to be careful to distinguish the influence of
the construction, the influence of the input, and what I called onomasiological coercion. The
distribution of these factors in determining the meaning of the resulting expression should be
similar for translations if they are all considered compounds.
The outcome of applying this method is for a particular construction in a particular
language to suggest whether it belongs to compounding or not. Taking as our basis the list of
compounds given by Levi (1978), we found that there were three strong candidates for such
constructions in French and Polish. Therefore, we considered genitive constructions,
relational adjectives, and prepositional constructions in more detail. Whereas the number of
N+N compounds in French and Polish among translations of English N+N compounds is

108
insignificant, together with the three constructions we studied here, they constitute 80% of
the translations.
If we assume a concept of compound on this basis, the analysis of the three candidate
constructions leads to the conclusion that genitive constructions and RA+N combinations are
compounds, but prepositional constructions are not. In the case of genitive constructions it
was noted that there is a compound variety and a syntactic variety. This contrast is reflected
in a different structure in English, cf. (12). In French, the genitive is expressed by de. As was
noted, de may be accompanied by the definite article. In some cases this expresses genuine
definiteness, but in others genericity. In Polish, the problem is that articles are not expressed
so that the ambiguity involves the equivalent of the entire spectrum of de with or without
definite article in French.
In both French and Polish, we might use the possibility of inserting a demonstrative as
a diagnostic. There are two possible effects, one of which is illustrated in (24):

(24) a. *?usine de ces automobiles ‘factory of these cars’

b. *?fabryka tych samochodów ‘factory theseGEN carsGEN’

The examples in (24) are semantically odd in the same way as the English gloss of (24a) is.
The point seems to be that the genitive cannot be interpreted as sufficiently close to
possession. Another possible effect is illustrated in (25):

(25) a. pays des ours

land of_the bears
bear country
b. pays de ces ours
land of these bears

The contrast in (25) shows that with the demonstrative the interpretation of ours switches
from generic in (25a) to specific in (25b). Demonstratives enforce an interpretation of a
genitive construction as involving two concepts. In a context in which a particular species of
bear has been introduced, (25b) can be used to refer to its habitat. However, (25a) cannot be
used in such a context. Here ours is necessarily generic, so that the expression introduces a
single concept rather than a connection of two separate concepts.
In the case of RA+N constructions, we also find certain borderline cases, but here the
question is how to distinguish relational adjectives from other adjectives. The problem in this
distinction is that many adjectives can have both roles in different contexts. Levi (1978) gives
the well-known example in (26):

(26) a. a nervous applicant

b. a nervous disease

Whereas nervous is a qualitative adjective in (26a) it is a relational adjective in (26b). There

are various diagnostics to distinguish the two types. A well-known test is to make the
adjective predicative, as in (27).

(27) a. the applicant is nervous

b. ?the disease is nervous

109
As suggested by the not quite so ungrammatical (27b), this type of syntactic test has its
limitations. Even nouns that are often left-hand components of compounds are sometimes
reinterpreted as qualitative adjectives, as in (28).

(28) a. It was a fun party.

b. The party was fun.

Nowadays, (28b) is entirely grammatical, with fun as an adjective. The same has happened
with key. Conversely, gradability with very does not exclude only relational adjectives, but
also, e.g., *very superb. Ultimately, only semantic criteria can be used to calibrate the tests.
For prepositional constructions other than with French de, it was concluded that the
preposition assumes a semantic role that is incompatible with the analysis of these
constructions as compounding. Whereas in compounding onomasiological coercion by the
concept to be named plays a crucial role, in N+P+N constructions in French and Polish the
preposition generally constrains the range of possible relations between the two nouns to such
an extent that onomasiological coercion does not have to be appealed to. However, it would
be worth exploring certain prepositions in more detail, such as French à.
In relation to Fig. 2, we can therefore conclude that approximately 61% of French
translations and 66% of Polish translations are compounds. However, in Fig. 2 we only
considered English N+N compounds as a basis. Levi (1978) does not have any examples of
N’s N compounds, but a third of her examples are RA+N combinations. They were excluded
from the original set, because it remained to be established whether they should be classified
as compounds. When we include them, we get the distribution in Fig. 3.

Figure 3 Distribution of translations in French (left) and Polish (right)

As can be inferred from the comparison of Fig. 3 with Fig. 2, the translation of RA+N from
English into French and Polish results more frequently in the use of relational adjectives. As
a result, 68% of French translations and 71% of Polish translations can be classified as
compounds. The remaining translations, i.e. approximately 30% in each language, do not use
compounding to name the same concepts. It should not surprise us that this is a significant
part. After all, concepts do not come with a tag that they want to be named by a compound.

110
 As we saw in some of the examples of N+Prep+N translations, the translational
correspondence is often not complete. These expression are often less strongly lexicalized, so
that several alternative translations with a similar status exist. This effect is even stronger for
paraphrases of the type we saw in (7). In such cases, translational equivalence is not achieved
at word level. As mentioned in section 3, the phenomenon of missing translational
equivalents at word level has been studied a lot in translation theory. Arguably, the reason for
their prominence is that they are not the normal case. Very often we do find good
translational equivalents at word level between different languages.
In conclusion, we have seen that it is useful and possible to define compounding as a
term. It was argued that the definition should be based on the nature of the naming process
rather than the morphophonological properties of the result. If we accept that compounding
takes two lexemes as input and uses onomasiological coercion to determine the relation
between them, we get a cross-linguistically useful class with a high degree of translational
equivalence.

References

ARONOFF, Mark H. 1994. Morphology by Itself: Stems and Inflectional Classes, Cambridge
(Mass.): MIT Press, 1994..

BAKER, Mona. 2011. In Other Words: A coursebook on translation, Second edition, London:
Routledge, 2011.

BLOOMFIELD, Leonard. 1933. Language, London: Allen & Unwin, 1933.

BOOIJ, Geert. 2002. The Morphology of Dutch, Oxford: Oxford University Press, 2002.

CHAMBERS. 1998. The Chambers Dictionary, Edinburgh: Chambers Harrap, 1998.

CHOMSKY, Noam. 1981. Lectures on Government and Binding, Dordrecht: Foris, 1981.

CHOMSKY, Noam. 1986. Knowledge of Language: Its Nature, Origin, and Use, Westport (Conn.):
Praeger, 1986.

DRESSLER, Wolfgang U. 2006. Compound Types. In LIBBEN, G., JAREMA, G. (eds.), The
Representation and Processing of Compound Words, Oxford: Oxford University Press, 2006, pp. 23-
44.

GREVISSE, Maurice. 1980. Le bon usage: Grammaire française avec des Remarques sur la langue
française d’aujourd’hui, onzième édition revue, Paris/Gembloux: Duculot, 1980.

ten HACKEN, Pius. 1994. Defining Morphology: A Principled Approach to Determining the
Boundaries of Compounding, Derivation, and Inflection, Hildesheim: Olms, 1994.

111
ten HACKEN, Pius. 2008. Prototypes and Discreteness in Terminology. In BERNAL GALLÉN, E.,
DECESARIS WARD, J. (eds.), Proceedings of the XIII Euralex International Congress, IULA-UPF,
Barcelona, 2008, pp. 979-987.

JACKENDOFF, Ray. 2002. Foundations of Language: Brain, Meaning, Grammar, Evolution,

Oxford: Oxford University Press.

KALLAS, Krystyna. 1999. Przymiotnik. In GRZEGORCZYKOWA, R., LASKOWSKI, R.,

WRÓBEL, H. (eds.), Gramatyka współczesnego języka polskiego: Morfologia, Warszawa: PWN,
1999, pp. 469-523.

LABOV, William. 1973. The Boundaries of Words and Their Meanings. In BAILEY, C.-J.N., SHUY,
R.W. (eds.), New Ways of Analyzing Variation in English, Washington DC: Georgetown University
Press, 1973, pp. 340-373, repr. in B. AARTS, D. DENISON, E. KEIZER, G. POPOVA (eds.). 2004.
Fuzzy Grammar: A Reader, Oxford: Oxford University Press, 2004, pp. 67-89.

LEVI, Judith N. 1978. The syntax and semantics of complex nominals, New York: Academic Press,
1978.

LIEBER, Rochelle, ŠTEKAUER, Pavol. 2009. Introduction: Status and Definition of Compounding.
In LIEBER, R., ŠTEKAUER, P. (eds.), The Oxford Handbook of Compounding, Oxford: Oxford
University Press, 2009, pp. 3-18.

MATTHEWS, Peter H. 1974. Morphology: An Introduction to the Theory of Word Structure,

Cambridge: Cambridge University Press, 1974.

MUNDAY, Jeremy. 2008. Introducing Translation Studies: Theories and Applications, 2nd Edition,
London: Routledge, 2008.

NEWMEYER, Frederick J. 1986. Linguistic Theory in America, second edition, New York:
Academic Press, 1986.

NICOLADIS, Elena. 2002. When is a Preposition a Linking Element? Bilingual Children’s

Acquisition of French Compound Nouns. Folia Linguistica, 2002, 36:45-63.

NORD, Christiane. 1997. Translating as a Purposeful Activity: Functionalist Approaches Explained,

Manchester: St. Jerome, 1997.

ROSCH, Eleanor. 1978. Principles of Categorization. In ROSCH, E., LLOYD, B.B. (eds.), Cognition
and Categorization, Hillside NJ: Lawrence Erlbaum, 1978, pp. 27-48.

SELKIRK, Elisabeth O. 1982. The Syntax of Words, Cambridge (Mass.): MIT Press, 1982.

SZYMANEK, Bogdan. 2010. A Panorama of Polish Word Formation, Lublin: Wydawnictwo KUL,
2010.

112
 VERMEER, Hans J. 2000. Skopos and Commission in Translational Action (transl. Andrew
Chesterman, orig. 1989). In VENUTI, L. (ed.), The Translation Studies Reader, London: Routledge,
2000, pp. 221-232.

Notes
1
This article is based on a presentation at the Second Conference on Universals and Typology in
Word Formation, Košice 26-28 August 2012. I would like to thank the audience, in particular Pavol
Štekauer and Bogdan Szymanek, for their questions and comments. I would also like to thank Fanny
Messika and Ewelina Kwiatek for their help with the French and Polish translations, respectively.
Needless to say, I am alone to blame for any remaining errors.
2
The fact that (5a) and (5b) are written as one word, but (5c) is not cannot be used as an independent
criterion for the boundary between compounds and phrases, but it does reflect the intuition that only
(5a) and (5b) are single concepts.
3
“La préposition à s’emploie surtout pour marquer le lieu, le but, le temps, le moyen, la manière, la
caracteristique” [my translation, PtH]

Pius ten Hacken

College of Arts and Humanities
Swansea UniversitySingleton Park
Swansea, SA2 8PP
Wales, UK.
[email protected]

In SKASE Journal of Theoretical Linguistics [online]. 2013, vol. 10, no. 1 [cit. 2013-02-04]. Available
on web page <http://www.skase.sk/Volumes/JTL22/pdf_doc/06.pdf>. ISSN 1339-782X.

113

View publication stats