A Study in Prose Styles: Edward Gibbon and Ernest Hemingway

Author(s): CURTIS W. HAYES

Source: Texas Studies in Literature and Language, Vol. 7, No. 4 (Winter 1966), pp. 371-386
Published by: University of Texas Press
Stable URL: http://www.jstor.org/stable/40753876 .
Accessed: 23/09/2014 14:58

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact [email protected].

University of Texas Press is collaborating with JSTOR to digitize, preserve and extend access to Texas Studies
in Literature and Language.

http://www.jstor.org

This content downloaded from 198.102.147.100 on Tue, 23 Sep 2014 14:58:44 PM

All use subject to JSTOR Terms and Conditions
 CURTIS W. HAYES

A StudyinProseStyles:
Edward Gibbonand ErnestHemingway

I. INTRODUCTION

THIS PAPER IS IN PART AN EXERCISE IN THE ANALYSIS OF PROSE STYLE

and in partan abstractofa largerstudythatI madeon theprosestyle
of EdwardGibbon.1In thatstudyI pointedout thatsensitive readers
ofliterature havecertainstylistic "intuitions"thatenablethemto iden-
tifycertainfamiliarauthors.For instance, iftwoextracts fromtheau-
thorschosenforthisstudy,EdwardGibbonand ErnestHemingway,
weregivento thesereaderstheywouldhavelittledifficulty in determin-
ing theirauthorship. This intuitive abilitymaynot, I believe, be ade-
quately accounted forby the more familiar to
approaches stylistic analy-
sis. By "morefamiliar"I mean the schoolof literarycriticism that
attachesimpressionistic labelstoprosestyles. Forexample,Hemingway's
styleinthemorefamiliar stylisticanalysishasbeendescribed as "simple,"
"direct," and sometimes "linear." Gibbon's which
style, is intuitively
morecomplexthan Hemingway's, has been labeledas "grand" and
"majestic."
Theselabels,however, do notdescribe butrathermirror
style, an im-
pression one receives whenreadingextracts fromthesetwoauthors. The
analysismustgo beyondthemeretaggingof impressionistic labelsto
prosestyles. It is theanalyst's job to accountforthesesubjective impres-
sions.I believethatrecentdevelopments inlinguistic
science,particularly
thedevelopment ofthetransformational/generative conceptofsyntax,
is an invaluableaid in formalizing thenotionofwhatone meanswhen
he attachesdescriptive labelstoprosestyles.
Generativegrammarians have in recentpublications statedthe in-
herent advantages ofa generative/transformational modelforrepresent-
ingsyntax.2 In particular theymaintainthata generative grammar fol-
lowing the Chomskian model would have the to
ability generate(a
1 See my
study,"A LinguisticAnalysisof the Prose Styleof Edward Gibbon,"
unpublisheddissertation
(University of Texas, 1964).
2 For example,See Paul Postal, "ConstituentStructure: A Studyof Contem-
poraryModels of SyntacticDescription,"I JAL, XXX, Part 3 (January,1964),
1 ff.

This content downloaded from 198.102.147.100 on Tue, 23 Sep 2014 14:58:44 PM

All use subject to JSTOR Terms and Conditions
372 CURTIS W. HAYES

bettertermwouldbe enumerate)all the well-formed sentences of a

language and would be to
able,further, provide deep a structural de-
scription for each. Since in greatpart the impression that a sensitive
readerperceives froma givenworkrestsuponthetypesofsyntactical
processes thatan authoruses,itwouldseemthatthetheory and methods
ofgenerative grammarians might lend themselves to the explication of
literarystyle. The following study is thusan experiment in the applied
use of generative grammarand reflects the thesis,alreadyimplicitly
stated,thatsucha grammar is a "powerful" and valuabletoolin ana-
lyzingliterary style.3
Forthepurposeofthispaperstylemaybe defined as a characteristic,
habitual,and recurrent use of the transformational apparatusof lan-
guage. Whatever is "characteristic, habitual, and recurrent" mustbe,
moreover, amenabletostatistical measurement. Therearecertaintrans-
formations (statistical attributes) in Gibbon'sstyle,forexample,that
can be measuredand thesetransformations can be comparedtothoseof
anotherwriter.In thissense,thestudyofstyleis a studyof thecom-
plexityofsentences.
The basic unitof description in thispaperis the textualsentence.
Here defined, a textualsentencecorresponds to the"institutionalized"
sentence : thatis,anybodyofmaterialoccurring betweenoneperiodor
question mark and thefollowing one. Textual sentences are notalways
simple or kernel sentences, but are often complexsentences, whichhave
beengenerated fromtwo or more underlying source sentences. It is best
to considertheultimatetextual sentence (the one which appearsafter
thefinaltransformation) a composite oftwotypes of source sentences.
The matrix(independent)sentenceformsthe overallpattern,the
frame,oftheultimatesentence. Thosesentences whichare embedded,
nested,or added to the matrix sentence arethe constituent (dependent)
sentences. Thisviewofmatrixand constituent sentence, RobertB. Lees
says,"makesessential use ofthenotionthatpartofthesyntactic struc-
tureofa sentence is thesetofunderlying, sometimes veryabstract, rep-
resentatives ofthesimplesentences fromwhichit maybe said to be de-
rivedbyexplicit grammatical rulescalledtransformations."4

3 The motivationfor such a studyhas been clearlystated by Richard Oh-

mann, "GenerativeGrammarsand the Concept of LiteraryStyle," Word,XX
(December, 1964), 423-439. See also JamesP. Thorne,"Stylisticsand Genera-
tiveGrammars,"JournalofLinguistics, I (April,1965), 49-59.
4 "The Promiseof Transformational Grammar,"EnglishJournal,LII (1963),
330.

This content downloaded from 198.102.147.100 on Tue, 23 Sep 2014 14:58:44 PM

All use subject to JSTOR Terms and Conditions
 Edward Gibbonand ErnestHemingway 373

Textualsentence : The boywhoiswearingthecap ismybrother.

Matrix: The boyismybrother.
Constituent: The boyiswearingthecap.
The constituent theboyis wearingthecap,is embeddedin the
sentence,
matrixsentenceby the use of an "explicitgrammaticalrule," the
transformation.
relative-clause
H. TRANSFORMATIONS
The followingtransformational rules (one-stringand two-string),
thoughnotcomplete, providethedescriptive framework fordescribing
an author'sstyle.The rulesin pureformare complex,butforthepur-
poseofthispaperI havesimplified them.5
A. The passivetransformation takesan activesentenceand trans-
forms itintothepassivevoice:
Norn+ X-Vtr-Nom'-Y > Norn'+ be + En-Vtr-by-Nom
Nom = Nominal
Vtr= anytransitive verb
En = pastparticiple inflection
X, Y = symbolsincluded to covermaterialwhichmay appear
afterNom,Nom',suchas post-nominal modifiers
> - "rewrite"(convert the left-handsymbolintotheright-hand
of
representation that symbol)
Sentence : MarycalledJoy > JoywascalledbyMary.
The agentbyMarycanbe optionally deleted:
Sentence : JoywascalledbyMary > Joywascalled.
B. Certaintransformations called GeneralizedTransformationsin-
volvethemanipulation oftwoor morestrings(sentences).The Nomi-
nalizingTransformation is justone typeofgeneralized transformation.
Nominalizing transformation Tto (Infinitive)
Nominalizing transformation Ting(Gerund)

5The readermaywishtoexaminetheformalandmoredetailedrulesoftrans-
formational I suggest
grammar. thathe consult
theworksofNoamChomsky, es-
peciallySyntacticStructures("S-Gravenhage,1957) and Aspectsof the Theory
of Syntax(Cambridge, 1965); RobertB. Lees, especiallyhis
Massachusetts,
"Grammarof EnglishNominalizations," I JAL,XXVI, Part 2 (1960; revised,
March,1961); and CarlottaA. Smith,"A Class of ComplexModifiersin Eng-
lish,"Language,XXXVII (July-September, 1961), 342-365and "Determiners
and RelativeClausesin a Generative
Grammar of English,"
Language,XXXX
(January-March, 1964),37-52.

This content downloaded from 198.102.147.100 on Tue, 23 Sep 2014 14:58:44 PM

All use subject to JSTOR Terms and Conditions
374 CURTIS W. HAYES

These rulesconverta sentencehavingthestructure NP + VP intoa

nounphraseoftheformto + VP or ing+ VP. These"crippled"sen-
tencesmaythenbe embeddedintothenominalslotofthematrixsen-
tence.
-ing+ VP (GerundiveNominal)
Textualsentence: Johnenjoysplaying thepiano.
1 2
Matrix: ( Johnenjoys) + ( Nominal: it)
3 4
Constituent:(John+ Tns + Be: ) (-ing+ playthepiano).
: 1+ 2
Transformation
> 1+4 > Johnenjoysplay-
3+ 4 the
ing piano.
to + VP (Infinitival Nominal)
Textualsentence : I askedhimtoplaythepiano.
1
Matrix: ( I askedhim)
2 3
Constituent: (He + Tns) + (playthepiano).
1
Transformation:
> 1 + to-3 > I askedhimto
2+ 3 playthepiano.
The finalNominalizing Transformationto be explicatedis theF ac-
tiveNominal.The ruleswhich generatefactive
nominais willhandlethe
following constructions.
Subject:thatclause
1 2
( Nominal: it) + ( was evident
)
3
(He didit)
Transformation: 1+ 2
> that-3+ 2 > Thathedidit
3 was evident.

Subject: question-wordclause
1 2
(Nominal) + (was a surprise)
3 4
(He went) + (there)

This content downloaded from 198.102.147.100 on Tue, 23 Sep 2014 14:58:44 PM

All use subject to JSTOR Terms and Conditions
 Edward Gibbonand ErnestHemingway 375

1+ 2
Transformation:
> where-3+ 2 > wherehe
3+4 wentwasa surprise.
"Non-Action" VerbObject
1
(I know) + (Nominal: it)
3
(He didit)
: 1+ 2
Transformation
> 1 + that-3 > I knowthat
3 he did it.
A sentence(NP + VP) mayfillan adverbslotand the processis
knownas "adverbialization." Forexample,inthesentence,hekilledthe
man+ (Adv), theAdvslotmaybe filledbyeithera singlelexicalitem,
a phrase,orbya clause(a "crippled"sentence):
He killedtheman+ (quickly)
He killedtheman+ (in thedark)
He killedtheman+ (whilethecityburned)
The constituent sentence whichfillstheAdvslotinsentence
threeis em-
beddedthrough theseoperations:
1 2
Matrix: (He killedtheman) + (Adv)
Constituent: ( thecityburned) .
: 1+ 2
Transformation
> 1 + while-3(Right-branching)
3j
The constituentsentenceand itsslotmaybe optionally to
permuted
sentence-initial
position(left-branching):
Whilethecityburned, hekilledtheman.
Or theconstruction
maybe self-embedded.
He, whilethecityburned,killedtheman.
The RelativeClause Transformation is an adjective-transformation.
Somelinguists
believethatprenominai adjectivaisareultimately
derived
fromthereduction of therelative-clause
construction.The red of the
redhouse,theyposit,canbe derivedinthefollowing way:

This content downloaded from 198.102.147.100 on Tue, 23 Sep 2014 14:58:44 PM

All use subject to JSTOR Terms and Conditions
376 CURTISW. HAYES

thehouseis red > thehousewhichis red > thehousered >

theredhouse.
Otherlinquists, whocombinethetransformational andtagmemic ap-
proaches, wouldnot hold to the notionthat a phrasesuchas thered
housemaybe derivedfromanykindofsentence in whichthewordred
in
appears predicateposition. Archibald A. Hill, forexample,points
out thatthe two approaches - tagmemicand transformational - seem
to be fallingtogether,in thatthe transformational grammarians are
acceptingthenotionofslots formodifierswhich can occurbefore the
noun.Thisavoidsthenecessity ofassuming(at leastinthispaper) that
thered housewas burneddownis anything morethana slotwitha
fillerin it,whereasthehousewas burneddownhas thesameslotbut
hasleftitempty.6 In thispaperadjectiveswillnotbe considered as being
derivedfromthereduction ofrelative
clauses.
The relative-clausetransformation involvestheseprocesses:
1 2
Matrix: (The boy) + (is mybrother)
3 4
Constituent: +
(The boy) (is here)
: 1+ 2
Transformation
> 1 + WH-4 + 2
3+ 4
5 6 7
Result: ( The boy) + ( whois)+ ( is
here mybrother)
Deletion(ellipsis): 5 + 6+7 > 5+7 > The boyhereis my
brother.
The additiveprocess(in essence,a transformational expansion)is
signaledbya coordinate The entireprocessmaybe called
conjunction.
addition:expansionofX slot.Therearevarioustypesofslotexpansions
thata speakerora writer mayperform: hemayconjoinsentences:
51 ]
> S1+ conjunction: + S2
and/but/or
52 ]
For instance,one may
of slotexpansions.
Thereare otherpossibilities
doesand employ
do as Gibbonhabitually transformational
thefollowing
expansions:

6 ProfessorArchibaldHill made thissuggestion

in a paper deliveredbeforethe
Society
Linguistic ofAmerica, December, 1964.

This content downloaded from 198.102.147.100 on Tue, 23 Sep 2014 14:58:44 PM

All use subject to JSTOR Terms and Conditions
 Edward Gibbonand ErnestHemingway 377

N
>N-f conj + N
NJ
NP
>NP + conj+ NP
NPj
V
> V + conj+ V
V
VP
> VP + conj+ VP
VP

PrepPhrase
> PP + conj+ PP
PrepPhrase
If theexpansionincludesonlytwolexicalitems(e.g., Caesareawas
plundered and burntbythelicentious , thentheconstituents
barbarians)
whichfilltheslotarereferred
tobythegenericname,doublet.

m. STATISTICALMETHODS
A linguist whenselectinga corpusoflinguistic
materialforanalysis
takesfromthatcorpusa samplethathe hopeswillbe typicaland repre-
sentative.
He doesthisbecausea corpusis oftentoolargeforindividual
analysis.Samplingis an economicalprocedureand it is particularly
necessaryin thispaper,sinceitwouldbe uneconomicalto analyzeeach
sentencein The Decline and Fall. Each sampleshouldbe testedfor
To testforreliability
reliability. a linguist statisti-
thefollowing
employs
cal formula:7
A. Reliability:StandardError= '/PQ/N
P equalstheproportion (frequency)ofone oftheitemsbeingcounted;
Q equalstheabsence ofthat item.P + Q = 100%. ¿Vis thetotalnum-

7A complete accountofthestatistical
methods in thispapercan
incorporated
be foundin publications
byDavidW. Reed.See his"The HistoryofInflectional
n in EnglishVerbsbefore1500," University
of CaliforniaPublicationsin English,
VII, iv (1950), especially
pp. 172-180,and "A Statistical
Approachto Quanti-
tativeLinguistic
Analysis,"Word,V (December, 1949),235-247.

This content downloaded from 198.102.147.100 on Tue, 23 Sep 2014 14:58:44 PM

All use subject to JSTOR Terms and Conditions
378 CURTIS W. HAYES

berofitemscounted.For example,assumethat70 sentencescontaina

while
device,
givenstylistic 30 do
sentences not.In that
event,

SE~V
100

= 4.5%
If theSE shouldexceed5%, moreitemsmustbe selectedto reducethe
error.
EarlierI positedthatstyleis in parta habitualand characteristicuse
of the transformational apparatusof one's language.To determine
whether a definedstylistic
device(transformation) of
is characteristic
oneauthorin comparison to anotherauthor,a linguist musttestforsta-
In a randomly
tisticalsignificance. chosensamplesupposehe findsthat
in authorX 70 sentences outof 100 containdoublets, whilein author
Y the doubletis foundin only60 sentences. He wantsto determine
whether thesetwoproportions (frequencies)can be said to belongto
the same population(no statistical significance)or to belongto dif-
ferentpopulations(statistical significance).To testforsignificance,a
linguist employsthese formulas :
B. Significance: P = NiPi + N2P2
Ni + N2
forthetwosamplescombined.
The value P is simplythe proportion
Thatis,130ofthe200sentences containdoublets.
P= 100 X. 60+ 100 X. 70
100 + 100
= 130/200
= .65
variance(theStandardError
thesampling
Withtheaboveinformation
between
ofthedifference maybe calculated:
twoproportions)

°Pl ?2~V
N,+ N2
_y/.65 X .35 + .65 X -35~
V 100 + 100
=V-00455
= .675 (StandardErroroftheDifference
between
twopro-
portions)

This content downloaded from 198.102.147.100 on Tue, 23 Sep 2014 14:58:44 PM

All use subject to JSTOR Terms and Conditions
 EdwardGibbonandErnestHemingway 379

The actualdifference
betweenthetwoproportions is .10 (.70 - .60 =
.10).
in StandardErrors= Actualdifference
Significance
SE differencebetweentwo propor-
tions
= .10
.675
= 1.48 StandardErrors
Consultationof a probabilitytable showsthatdeviationsof thissize
(1.48) areexpectedtooccurbychancealonein 13.88% ofthesamples.
The difference
betweentheproportions is,therefore,notsignificant.
A
of
percentage probabilityabove 5.0% is usually takenby statisticians
to
denotea chancedeviation;a factorotherthanchanceis responsible if
thepercentageof probabilityis below5.0%; e.g.,if theSE is 1.96 or
more,thedifference betweentheproportions is significant
at the5.0%
level.Anythinglessthan 1.96 denotesthattheStandardErroris non-
significant.
IV. METHOD OF ANALYSIS

I selectedat randomone hundredsentences fromeach author.These

textualsentences werefirstrewritten intosimplesourcesentences and
the"history" ofeach (a detailedanalysisofthegeneralized transforma-
tionswhichthetextualsentencecouldbe assumedto haveundergone)
was thenshown.Numbersare assignedto theappropriate constituents
forthepurposeofa structural The
description. following sentence from
B.
J. Bury's edition of The Declineand Fall (VI, p. 108,Sentence 7)
willdemonstrate themethodofanalysis :
TextualSentence :
Theirdiscontents weresecretlyfomented byLaetus,theirpraefect,
whofound,whenitwastoolate,thathisnewemperor wouldreward
a servant, butwouldnotbe ruledbya favourite.
SourceSentences:
1. Theirdiscontents weresecretlyfomented byLaetus.*
*Thissentence is a passivetransformoftheactivesentence : Laetus
secretlyfomented theirdiscontents.
A. Laetuswas theirpraefect.
( 1) Laetusfound+ [Nominal]+ ( Adv)
a. It was too late+ (then)
b. Hisnewemperor wouldrewarda servant.

This content downloaded from 198.102.147.100 on Tue, 23 Sep 2014 14:58:44 PM

All use subject to JSTOR Terms and Conditions
380 CURTIS W. HAYES

c. His newemperor wouldnotbe ruledbya favourite.*

*Thissentenceisa passivetransform :
oftheactivesentence
wouldnotrulehisnewemperor.
A favourite
Generalized Transformations: "History"
1. Addition : expansionofVP slot
1
( His new emperorwould rewarda servant
)
2
( Hisnewemperor wouldnotbe ruledbya favourite)
1
Transformation:
> 1 + but+ 2
2
3
Result: (His newemperor wouldrewarda servantbut) +
4 5
(hisnewemperor)+ (wouldnotbe ruledbya favourite)
Deletion:3 + 4 + 5 >3+ 5
2. FactiveNominal:that-clause
6 7 8
(Theirpraefect found)+ ([Nominal]) + ((Adv))
9
(His new emperor wouldrewarda butwouldnotbe ruled
servant,
a
by favourite)
: 6+ 7+ 8
Transformation
>6 + 8 + that-9
9
3. Self-embedding: adverbialclause
10 11
(Their praefectfound)+ ((Adv)) + (that his new emperor
12
wouldrewarda servant, butwouldnotbe ruledbya favourite)
13 14
(It wastoolate) + (then).
10+11 + 12
Transformation:
> 10 + when-(14)-
13+ 14J 13 + 12.
embedding:relativeclause
4. Right-branch
15

This content downloaded from 198.102.147.100 on Tue, 23 Sep 2014 14:58:44 PM

All use subject to JSTOR Terms and Conditions
 Edward Gibbonand ErnestHemingway 381

(Laetuswas theirpraefect)
16 17
+
(Theirpraefect) (found, whenit was toolate,thathisnewem-
peror would reward a but
servant, would not be ruledby a
favourite)
Transformation:
15
> 15 + WH-17
16 + 17
5. Right-branch clauseplusdeletionof
embedding:relative
WH + V.
18
(Theirdiscontents
weresecretly
fomented byLaetus)
19 20
+
(Laetus) (was theirpraefect,who found, whenit was toolate,
thathis new emperorwouldrewarda servant, but wouldnotbe
ruledbya favourite)
Transformation:
18
> 18 + WH-20
19 + 20
21
Result: (Their discontents
weresecretlyfomented by Laetus) +
22 23
(whowas) + (theirpraefect, whofound,whenitwastoo
late, thathis new emperor would rewarda servant,but
wouldnotbe ruledbya favourite) .
Deletion:21+22 + 23 > 21 + 23 = Textualsentence
The followingI foundto be typicalofthetypeofsentencethatGibbon
and Hemingway employ:
Gibbon:
A senseof interestattachedthesemoresettledbarbariansto the
allianceofRome,and a permanent interest
veryfrequentlyripensinto
sincereand usefulfriendship.
The wholeforceofConstantine consisted
ofninety thousandfootand
eight thousandhorse;and, as the defenceof the Rhine requiredan
extraordinaryattentionduringtheabsenceof theemperor, it was not
in hispowerto employabovehalfhistroopsin theItalianexpedition,
unlesshesacrificed
thepublicsafetytohisprivatequarrel.
WhilstRomelamentedthefateofhersovereign, thesavagecoldness

This content downloaded from 198.102.147.100 on Tue, 23 Sep 2014 14:58:44 PM

All use subject to JSTOR Terms and Conditions
382 CURTISW. HAYES

ofhissonwas extolledbytheservilecourtiers as theperfect of

firmness
a heroand a stoic.
The religion ofZoroasterwas abundantly providedwiththeformer,
and possesseda sufficient of
portion the latter.
Hemingway :
I gave themmoneyforplatform ticketsand had themtakemybag-
gage
We lookedat eachotherinthedark
I loweredthevermouth bottleto theothersideofthebed whenshe
camein.
We walkedalongtogether through thetownand I chewedthecoffee.
I lookedbackandsawherstanding onthesteps.

V. RESULTS OF ANALYSIS AND STATISTICAL COMPARISON

Whatfollowsare theresultofthisanalysisand a comparison ofthe
typesof that
transformations Gibbon and Hemingway employ. This is
donein an effortto determinewhether statistical exists
significance in
Gibbon'suse of a specifictransformation comparedto Hemingway's
use of thesame transformation.The attributes (transformations) are
in theleft-hand the
column, frequency of the transformationis in the
is notedin theright-hand
centercolumn,and significance column.
ATTRIBUTE
Gibbon Hemingway StandardErrors
per sentence
Transformations 4.3 1.3
SentencesundergoneGT 98% 60% SE = 6.5
Passive 68% 2% 9.48
Doublet 68% 8% 9.0
Sentencescontainingonlyone doublet 40% 8%
Sentencescontaining2 doublets 18% 0%
Sentencescontaining3 doublets 7% 0%
Sentencescontaining4 doublets 1% 0%
Sentencescontaining5 doublets 2% 0%
Total # of doublets 111 8
# N doublets 64 4
# Adj doublets 33 4
# V doublets 13 0
#Adv doublets 1 0
NP expansions 16% 0%
VP expansions 14% 28% SE = 2.6
PP expansions 7% 2% 1.7
FactiveNominalExpansions 2 0
% of S whichhave doubletexpansions 79% 36% SE = 6.0

This content downloaded from 198.102.147.100 on Tue, 23 Sep 2014 14:58:44 PM

All use subject to JSTOR Terms and Conditions
 Edward Gibbonand ErnestHemingway 383

Gibbon Hemingway StandardErrors

Triplet 14% 4% 2.5
# N expansions(triplet) 4 2
# NP expansions 2 0
# V expansions 2 0
# VP expansions 3 0
# Adj expansions 0 2
# PP expansions 2 0
# FactiveNominalexpansions 1 0
Quadruplet 3 0
# N expansions 1' 0
# NP expansions 1 0
# Factive Nominal expansions 1 0

Nominalizations 49% 22% SE = 4.0

InfinitivalNominal 31% 9% 3.9
Total Number 37 10
S containing1 Infinitival 25 8
S containing2 Infinitivais 6 1
GerundiveNominal 16% 4% .... 2.8
Factive Nominal 19% 13% 1.17

Embedding
# S whichhave embeddedelements 64% 20% 6.3
Total # of embeddedstructures 105 24
# S containing1 embeddedstructure 36 16
# S containing2 embeddedstructures18 4
# S containing3 embeddedstructures 7 0
# S containing4 embeddedstructures 3 0
S having Rei. Cl. structures 51% 8% 7.0
Total # of RC structures 77 10
S having1 RG 30 6
S having2 RG 16 2
S having3 RG 5 0
Types of Branching
Right Branching 26 4
Deleted Right Branching 14 4
Self Embedded 19 0
Deleted Self embedded 17 2
S havingAdv Clause structures 23% 12% 2.0
Total # Adv Glauses 28 14
S containing1 AC 19 10
S containing2 AC 3 2
S containing3 AC 1 0
Typesof branching
Right Branching 9 8

This content downloaded from 198.102.147.100 on Tue, 23 Sep 2014 14:58:44 PM

All use subject to JSTOR Terms and Conditions
384 CURTIS W. HAYES

Gibbon Hemingway StandardErrors

Deleted Right Branching 2 0
Left Branching 12 4
Deleted Left Branching 1 2
Self Embedded 4 2

AdditiveProcess: ExpansionofS Slot

% of S whichare expansions 68% 32% 5.1
S having2 conjoinedS 41 30
S having3 conjoinedS 18 1
S having4 conjoinedS 9 0

The tableindicateswhatwe alreadyknewintuitively, thatthestylesof

Gibbon and Hemingwayare different - thatthe styleof Gibbonis
"grand,""majestic,""complex,"and thatthestyleof Hemingway in
is
comparison "simple." The importance of the tables,however, is that
theyoffer us an objectivemeasureto "capture"thisintuition. In other
words,we mayuse thetableas a toolto showthe reasonsbehind our
intuition. Insteadofbasingour analysison subjectiveimpressions and
usingopaque terminology to describetheseimpressions, we maysay
exactlyhowtwostylesdiffer. Generative grammaris important to the
literary analystforthatreason: it offers hima devicethroughwhich
objectivestatements can be made aboutstyle.For instance, he can say
thatthestyleofGibbonismorecomplexthanthatofHemingway, since
thenumberoftransformations persentence is an indicationof that com-
plexity.This is further indicatedby the numberof sentenceshaving
embeddedstructures (relativeclauses,adverbialclauses),nestedstruc-
tures(nominalizations), and expandedstructures withinthem.Rarely
are theretransformational expansions or embedded structureswithin
Hemingway's sentences.
It is traditional to say thatGibbon'sstyleis characteristically bal-
anced.The tableshowsthedegreeofthisparallelism. Not onlycan we
say thatGibbonusesa highdegreeof parallelstructures, we can also
say something about the kinds of parallelstructures. We can say that
thenumberofdoubletexpansions in Gibbonfarexceedsthosefoundin
Hemingway;and thatthereare othertypesof balancedstructures in
additionto the two-membered balanced entity(forexample, the bal-
ancednounphrases,verbphrases,and prepositional phrases).In sum,
thetableenablesus to make exact descriptions of each author'sstyle.
The following summation
is a statistical of the transformations count-
ed. The listis gradedfrommostsignificant (the passive) to least sig-
nificant (adverbialclause) to nonsignificant.

This content downloaded from 198.102.147.100 on Tue, 23 Sep 2014 14:58:44 PM

All use subject to JSTOR Terms and Conditions
 Edward Gibbonand ErnestHemingway 385

in StandardErrors
Significance
Passive Transformation 9.48
Doublet 9.0
InfinitivalNominal 8.9
Relative-clausetransformation 7.0
S havingundergonetransformation 6.5
% of S havingembeddedstructures 6.3
% of S havingdoubletexpansions 6.0
S whichare additive 5.1
% of S havingNominalizations 4.0
GerundiveNominal 2.8
VP expansion 2.6
S having3 conjoinedS 2.5
AdverbialClause 2.0
Transformationswhichdid notshowstatisticalsignificance
(below 1.96)
PrepositionalPhraseExpansion 1.7
FactiveNominal 1.17

VI. CONCLUSION

Explicitin thispaperhas beenthehypothesis thata generative/trans-

formational grammar based upon theChomskian modelis an extremely
effectiveand usefultoolin capturing thenotionofstyle;and, further,
thatthetransformations whichan authorexploysto expresscontentis
a significantpart of what a readerperceivesas literarystyle.This hy-
pothesis stems from the notion thatspeakers(and thus writers)may
employvariousand different models(sentencetypes)to expressrela-
tivelythesamecontent.That is,an authortheoretically has an option
at each linguisticsituation. Whetheran author,forexample,generates
Johnhittheballor theballwashitbyJohnis a linguistic choice.To be
morespecific, whether an authorhabitually generatessentences in the
passivevoice, as Gibbon does,or whether he prefersthe active voice,as
Hemingway does, is an alternativewhich thelanguageprovides. Notice
thata choicein thissenseis nota lexicalchoicebuta choiceofa trans-
formational rule.
In brief,a transformational/generative grammarhas thepowerto
differentiatedifferent styles.As transformationalists
are fondofsaying,
theirgrammar formalizes thenotionofhowcomplexsentences arepro-
ducedand themannerin whichcomplexsentences arerelatedtosimple
sourcesentences. The amountofcomplexity, it follows,
is an indication
of how an authoremploystherulesofhislanguage;and by breaking
down ("rewriting")sentencesinto theirkernelcomponents and by
specifying the typesof grammatical transformations used to construct

This content downloaded from 198.102.147.100 on Tue, 23 Sep 2014 14:58:44 PM

All use subject to JSTOR Terms and Conditions
386 CURTIS W. HAYES

theultimate textualsentence, onemayarriveat thedegreeofcomplex-

ity,which,I havemaintained, is one indication ofan author'sstyle.
Since complexity is onlyone indicationof style,the complexity in
Gibbon'ssentences may not differ from thecomplexity in the sentences
ofsimilarauthors, saySamuelJohnson and David Hume.8In thatcase,
ifonemayhypothesize, thedifference, ifthereis any,mayexiston other
levels(imagery, metaphor, etc.).
After makingthesestrong claimsfortheefficacy ofa transformational
modelin literary it is
analysis, perhapsself-defeating to pointout that
thismethodis not a panacea forcorrecting the ills of past stylistic
analyses.Withthisin mind,it wouldbe perhapshelpfulto statewhat
thisstudyhas accomplished and whatthelimitations of such a study
are. First,thestudyas presented is an attempttowardcapturingthe
elusivenotionof literary style.Yet, as pointedout,styleexistsat all
not
levels, merely at the syntactical level;and certainly styleexistsbe-
yondthe sentence, in
say the realm of imagery. This study,in other
words,has not exhaustedthe possibilities of discovering the inherent
differencesin Hemingway's and Gibbon'sstyles. Nothing, forexample,
has been said abouttheirrespective vocabularies.Certainly Gibbon's
vocabulary is morelatinatethanHemingway's; and thisis a distinctive
difference.Moreover, a transformational studycouldbe deeperinsome
respects: forexample,inpointing totheposition ofadjectives. Gibbon's
adjectives,which usually occur in attributiveposition, could be com-
to
pared Hemingway's adjectives, which normally occur in predicate
position.To conclude,theimportance ofthistypeofanalysisis thatthe
intuitivelyfelt differences between two differing stylescan at leastbe
explainedby the types(and frequencies) of transformations thateach
authorexploys inconstructing hissentences.
University ofNebraska
Lincoln,Nebraska
8 I am preparinga manuscriptin whichI comparethe prosestylesof Gibbon,
Johnson,and Hume.

This content downloaded from 198.102.147.100 on Tue, 23 Sep 2014 14:58:44 PM

All use subject to JSTOR Terms and Conditions