Constraints on the agentless passive

advertisement
J. UIH(iNRICII
t9 {t()ti.l),2f58. Printedin Great Britain
l. Linguistics,
Anderson.J. & Ewen.C. (forthcoming). principlesof dependency
phonology.
Bing, J. M. (198o).Linguisticrhythm and grammatical'structuie'in
efghan persian.LIn u,
$7-463.
Chomsky,N. (r965).Aspct't.r
of the rhcorl'of .ryntax.cambridge,Mass.:M.l.T. press.
Chomsky,N. & tlalle. M. (r968).The.uruncl
putternof Engtisi.Ncw york: iurp", & Row,
Cooper,w. E.( l98o)'Syntactic-to-phoneticcoding.
In Butteiworrh,B. ted.)tansiiie production,
Vol. L' .speech
und tulk. London: Acadcmicpriss.
Dogil' G. (tqlq). Auto'segmtnlu!
at'utunt oJ'phonologicalemphasis.Edmonton: Linguistic
Research,
Inc.
Faiss'K' (tgttl). Compound,pseudo-compound,
and syntacticgroupespecially
in English.In
Kunsmann'P. & Kuhn, o. (cds.),walt.sprucheEngtisch
in Fischung und Lrihr"; Fests(.hrdt
.liir Kurt Wiichtler.Berlin:Erich Schrnidt.r3z-r5o.
Gicgcrich,H.J.(r98o).on stress-timing
in Englishphonology.
Lingua5r. rg7-zzr.
Giegerich,H-.J' (r98la). Compoundstiessin G".man' tori observations
and an exercise
in
metricalphonology.l|ork in Progre.ss
14,84'-gT.Universiry
of Edinbuigt, o.p"rtr.nt oi
Linguistics.
Gi-egerich,
H. J. (r98r b)' on the natureand scopeof metricalstructure.IndianaUniversity
LinguisticsClub.
Giegerich'H. J. (198rc). Zero syllablesin metrical theory. In Dressler,W.
et a/. (eds,)
Phonologicatg8o" Akten der viertenInternationaten
Phonilogie-Taguig,'Wien,29.Juni- 2.
Juli, rgSo.lnnsbruck:Institutftir Sprachwissenschaft
der Universiiart"nnsuruc[.,53-r59.
Giegerich,H'J. (1982).MetrischePhonologie
und Kompositionsakzent
im Deutschen.papiere
zur Linguistik.
H1lle, M' & Keyser,S' J. (tg?t). Englishrrrerr.' itsform, itsgrowrh,and itsrole in yerse.
New
York: Harper& Row.
Kahn, D. (rg16')-Syllable-based
generalizations
in Englishphonology.Indiana University
LinguisticsClub.
Kiparsky, P. (1966).Uber den deutschenAkzent. studia Grammatica
7.6q-{lg.
Kiparsky,P.(rgll). The rhythmicstructureof Englishverse.LIn e. leg-iai.
Kiparsky,.P.(rglil.Metrical srructureassignment
is cycric.LIn rc, 4ri-4/r'.
Ladefoged,P.(tgZS). A cgursein phonetics.New York: Harcourt Brice Jovanovich.
Langendoen,
T. (t925). Finite-stateparsingof phrase-structure
languages
ir,r statusof
readjustment
rulesin grammar. LIn 6.533-554.
"nJ
Liberman,M. & Prince,A. (rg7:.).on stressano tinguisticrhythm. LIn g. 249136.
Lieberman,P. (rg6z). Intonation,perception,and lanluage.camurlage,Mass.l
ffi.r. pr.rr.
Marchand, H. (rq6g). The categoriesand typq oj
tigiirn n,ori-for^otion: A
- pisent-day
'
synchronic-diachronic
approach.Miinchen: Iieck.
Pike,K. L. (tg+6).Theintonationof AmericanEngtish.AnnArbor: Universityof Michiganpress.
Rischel,l. (t972). compound stressin Danish without a cycle.ARlpuc i. ztr-zzg.
Schmerling,
S. (197r).A stressmess..SLSr. 5z-66.
Selkirk,E. o. (t972). Thephrasephonologyof Engtish
andFrench.Doctoraldissertation,
M.I.T.,
Cambridge,Mass.
Selkirk, E.o. (t978). on prosodic structure and its relation to syntacticstructure.paper
presentedat Conferenceon the Mentat Representation
of Phonoiogy.pistiiuuieo uy itre
Club, l9go.J
_ ldiary UniversityLinguisrics
Selkirk,E. O. (t98o),Th9roleof prosodic.caleggrigs
in Englishword srress.
LIn 1.563-6o5.
Th-ompson,
H' S.(t98o).Slressandsalience
in Eigtish; theorlandpractice.paloeiio,-California:
Xerox.
wollmann, A' (t97r). Stressshift in level-stressed
compoundsand the transformationalcycle.
MilnchnerPapierezur Linguistlr u. r-g.
28
:
i
i
l
I
D
E
Constraints on the agentlesspassive
E .J U D I T H W E I N E R e N DW I L L I A M L A B O V
TempleUniversityand Universityof Pennsylvania
(Received
t8 FebruaryI98I)
Thispaperis a quantitativestudyof the factorsthat determinethe selection
By examining
overactiveonesby Englishspeakers.r
constructions
of passive
a large body of passivesused in spontaneousspeech,togetherwith the
that show an opposingchoice, we are able to throw light on
sentences
the crucial questionof which syntacticand which semanticfeaturesof the
environmentact to constrainthe choiceand whethersyntacticor semantic
factorspredominatein this case.In the courseof the analysis,we will also
havesomethingto say about the socialfactors that have beenreportedto
determinethe useof the passive.
relationship
meanthesamething?The active-passive
Do activeand passive
pairing
forms
productive
of
syntactic
haslong beenrecognizedas a highly
grammar.
of
transformational
andhasplayeda major rdle in earlystatements
In recentyears,limitations of this productivity havebecomea major focus,
and such problems as the difficulty of drawing a sharp line between
verbs have led a number of linguiststo
passivizable
and non-passivizable
excludethe passivefrom the array of movementtransformations(Bresnan,
1978).Differencesin intuitive reactionsto active and passiveforms with
quantifiers,
first notedby Chomsky(1957:Ioo-I), haveled many to argue
that thereis no meaningequivalenceof activeand passive.Thusfor the pair
(t) Everyonelikes someone
(z) Someoneis liked by everyone
'in my speech,though not in that of all speakersof
G. Lakoff statesthat
English,(r) and (z) havedifferentmeanings'(t97o:I4-t5).Katz & Postal
[r] This analysisis the product of the researchproject on Linguistic Changeand Variation,
The original
supportedby the National ScienceFoundationunder NSF SOC75oo245.
frimework for analysiswas developedby W. L. for a study of the passiveby members
Analysis(Linguistics56r) at the Universityof Pennsylvania,
of theclasson Sociolinguistic
who developedthe first body of data referredto below. E.J.W. developedthe present
codingsystemand reanalysedthe initial materialsalongwith other bodiesof speechdrawn
from ihe recordsof the project and ran the variable rule programmes.The final analysis
is the joint work of both authors.We are particularlyindebtedto Ivan Sag for many
detailedcommentsand corrections,though he cannot be consideredresponsiblefor any
failure on our part to closethe gap betweenthe empirical study of the passiveand most
recent thinking in formal syntax. We are also indebted to SusumuKuno, Beatriz
Lavandera,and Anthony Kroch for their critical reactions,which we hope we have
respondedto, at leastin Part.
29
(tg6q: 721) werenot convincedby Chomsky'soriginal
exampleand found
both meaningsin activeand passive;we havethe samereaction
to these.
Yet therecannotbeanydoubtthat somepassive
constructions
do not mean
the sameas theircorresponding
actives.The latestattackon the problemof
explaining
thesedifferences
is thatof McConnell-Ginet
(l9gz),whodealswith
the semanticdifferences
producedby adverbialmodification:
(r) Reluctantly,Joan instructedMarv.
(q) Reluctantly,
Mary wasinstructedby Joan.
McConnell-Ginet
concludes
that thereareseveraldifferenttypesof adverbial
modification
involved,andthatthepassive's
two verbsofferror. possibilities
for such modificationthan the active does. In spite
of thesepotential
differences
in the meaningof activeand passivein particularcontexts,
our
observations
of the useof this option in spontaneous
speech
convinceus that
the choiceof the agentless
passiveand giner alizedactlveis fundamentally
a
syntacticone,and that activeand passivenormallyhave
the samemeaning
in a truth-conditionalsense.Many of the differences
betweenactiveand
passivethat have beenobservedseemto us to
be differences
in focusor
emphasis,characteristic
effectsof the re-orderingof sententialelements.
It
seemsto us preferableto restrict the term 'mianing'
more narrowly to
designatethe couplingof a givensentence
with a givenstateof affairs.z
one strategy
we
might
follow
is
to
say that we are using . rough
-that
semanticequivalence'and admit that a more
irecise senseof the meaning
of activeand passivewould requireus to registereverypossible
inferencethat
one individualor anotherwould be tikelyio makein
interpretingone or the
other form' But it seemsto us that this concession
to an idealisticsemantics
is needlessly
unrealistic.If we isolatewordsfrom their use,we can
showthat
thereis no such thing as a precisesynonym,since
all words have slightly
differentprivilegesof occurrencewhen*r .onrider everypossible
context.But
in practice,the needfor stylisticvariation leadsalt
sjeaters and writers of
Englishto substituteone word for the other with
the expectationthat any
differencesthat might -arisein other contextswill not
atrectinterpretationin
that one. In our study of active and passive,we
are concernedwith the
informationthat is transmittedby this choicein everyday
life. In our data
thereis ampleevidence(seebelow)that thetwo formsare
usedinterchangeably
t2l 'meaning'to
In this respect'Yj: are continuing in the tradition of weinreich,s efforts to restrict
significata
that areshired throughouttt" sp"."t community
it".-rri, review
of osgood et al.,
to evidenceon the
"*. dialects
l95q)..Recent.responses
oi lu"ntinu
haveadmittedthat both meaningscan be found but"";;ii;6iiiri
claimedthat individuals
'prefer'
id^iosyncratically
onereadjngor th! other.In an unpuulisrrrop"pri,: oi.iion"ri*
,l
of the future: A setof parameters
foidescripriu.r.runitr;
weinreichpointed
out that descriptivesemanticsmustdeat*ith rh"'"onrt"]rt,i,o
"pp"u.),
inrti,i"ii"it"iir.a
ortn, t
meaningof signs.. . without denyingthe existence
"-rp".,
of a non-institutionalized
marginof i
meaning'.
i
possibility
;";;";,he
to rererto the samestatesor
"ruirr;;;;;
that thereare somecontextswherethis is not so.
(Labov, I975)
of thechoiceof theget vs.bepassive
An earlierinvestigation
in generalcontexts
that thesefunctionassemanticequivalents
demonstrated
differentiatedin the specialcontextof
eventhough they weresemantically
purposeclauses.Reactionsto He got arrestedandHe wasarrestedin context
showedthe sameproportion of interpretationsof the subjectas the agent
(: 'He got himselfarrested').But He got arrestedto testthelaw gavea much
higherproportion of suchinterpretationsthanHe wasarrestedto testthelaw.
We can infer that there will be contextswhere active and passivelead
of statesof affairsand ultimatelyit will
to differentsemanticinterpretations
but thesewill undoubtedlybe a
these
contexts,
locate
to
be important
not likely to affectour search
and
of
uses
range
the
total
smallsubsetof
vs. passive.
of
active
general
the
choice
on
constraints
for the
'meaning'is consistent
the
term
The moveto restrictratherthan expand
of thestylisticand
studieswhichobtainobjectiveevidence
withsociolinguistic
socialdifferentiationof linguisticforms. At the sametime, we must not
concludethat all linguistic variation subservesthe social functions of
andadjustingto the socialposition
identifyingthe socialpositionof speakers
Variableelementsare found at all levelsof
of the listenersand audience.s
linguisticstructure.Somevariationis the resultof articulatoryconstraintson
somereflectsa variablerecognitionof grammatical
grammaticalprocesses;
boundaries;some appearsto be the residueof historicalprocesseswhich
persists
long after the socialconditionsthat gaveriseto it havedisappeared.
Thereis no reasonto confinethe study of variation to alternativewaysof
sayingthe samething, althoughthis kind of variablehasbeenthe chief focus
has
in the developmentof variablerules. The Heidelbergforschungsproiekl
made great strides in the probabilistic weighting of context-freephrase
structurerulesby speakersof pidgin German(rg16\ and theseareessentially
meaningfulchoices.Studiesof the developmentof phrasestructurein the
earlystagesof languagelearning(Bloom, t97o)inevitablyinvolvevariation
of a meaningful kind. However, it is clear that the sharpestanalytical
conclusionson the conditioningfactorsthat constrainlinguisticchangeand
variationcan be madewhenform variesbut meaningis constant,rather than
whenboth are varying together.Theoretically,it shouldbe possibleto draw
equalprofit from caseswherea singleform is usedwith severalmeanings.
is lessimmediatewith semantic
But the possibilityof accuratemeasurement
variation.We obviouslyhavea muchbetterchanceof gettingintersubjective
in identifying formal variantsthan semanticvariants.
agreement
We thereforeapproachthe passivewith an eyeto a bold simplificationof
the problems of meaning. We will treat active and passiveas truthlll SeeLavandera(rqZ8) and the Issponseof Labov (tgZ8).
I
3o
I
I
i
&
3r
.
A N DW I L L I A ML A B O V
E .J U D T T W
H EINER
conditionallyequivalentand usedon the whole to refer to the samestates
of affairs.
Earlier reporrsof socialdifferencesin the useof activeandpassive.The only
speechis the
previousdata we haveon the useof the passivein spontaneous
'elaborated
Bernstein's
Basil
iesult of its useasoneof thedefiningfeaturesof
'passiveverbs'out of all finiteverbsis saidto
code'. A high proportionof
speakers,
be characteiisticof the elaboratedcodeavailableto middle-class
In
Bernstein's
percentages.a
asopposedto'restrictedcodes'whichshowlower
groupsand three
( r 97r) acomparisonof two middle-class
tateiipresentation
are usedby the former at a
groupsshowsthat more passives
working-class
levelof beyondo.oz(p. tot).
significance
that indicate
Lawton(r968)givesa numberof otherresultson the passive
that there is more variationin this indicatorthan othersusedby these
in the useof passiveverbsby
Therewasno significantdifference
researchers.
r 2-year-oldsbut therewas a differenceamong Is-year-olds(5 per cent for
the workingclass,I2 per cent for the middleclass- p. I25)'
speakers
interpretationof the useof thepassiveby middle-class
Bernstein's
do more
speakers
these
that
fact
of
the
prominent
indication
is
a
this
is that
linguistic
no
is
There
speakers.
class
working
planning
than
syntactic
varioustypesof
discussionaccompanyingthesestudiesthat distinguishes
activeand
between
choice
syntactic
the
isolates
importantly,
more
or
passives,
il
passive.
The notion that the passiveinvolvesmoresyntacticplanningwasperhaps
not as naivewhen it was first formulatedas it appearstoday.The general
is that the
experimentation
that hasemergedfirompsycholinguistic
consensus
with the
coupled
be
easily
cannot
sentences
perception
of
productionand
it is
(Fodor
1974);
al.,
et
grammar
now
available
of
iormal representations
process
that
planning
with
a
evenlessiit.ty that we can makecorrelations
precedesthe production. But in any case,the Bernsteindata does not
representa controllcdstudy of a syntacticchoice:the useof the passiveis
wherethe choicemight havebeenmadebut was
not .orpared to the-'cases
t.
not.
lf we considerall possiblefinite verbs,a largenumberwill be activeswith
two arguments.ln spontaneousspeech,the proportionsof passivesrvith
is vanishinglysmallin all data examinedso far (see
predicates
two-argument
below).The vast majority of passiveconstructionshas only one argument,
that is, no agentis present.The comparisonof passiveverbsto all finite verbs
with
with one argumentto sentences
iTthereforea comparisonof sentences
passives
used
one,two and threearguments.Differencesin the proportion of
may thereforereflectwhat is beingtalked about and the amount
by speakers
of detail being providedmore than any syntacticchoice.
words. This is a very small
t+l The total amount of speechstudiedfor each group was tSoo
body of speechcomparedto thoseusedin sociolinguisticanalysis.
32
PASSIVE
CONSTRAINTS ON THE AGENTLESS
by the formal natureof
studiesare alsocharacterized
The Bernstein-type
to discussthe
were-assembled
,Ltting.In onestudy,groups
theexperimental
in other
is
evidence
There
i97l
(Bernstiin,
).r'
abolitionof capitalpunishment
backgrounds
social
various
fiom
of a differ.niiufreactionof speakers
studies
(Labov, Ig6gb).Van den Broeck's
sort
this
of
situution,
to experimental
s t u d y o f M a a s e i k , B e l g i u m ( I 9 7 7 ) , f o r e x a m p | e , s hformal
o w e dstyle'
thatm
idd|e-class
but
in
speakers
usedmorepuJriu.,than working-class
speakers
that favouredthevernacular'the
whenthe socialcontextwasswitchedto one
in this respect'
relationof the two groupswasrcversed
us to concludethat the experiThe comparisoni-, .lore enoughto allow
m e n t a l t e c h n i q u e s o f B e r n s t e i n d o n o t r e f l e c t d i f f e r e n c e smodes
i n t h e of
availability
but rather different
of particularconstructionsto his subjects,
othercontexts'
that mayemergefreelyin
,rprrrrionof a linguisticcompetence
'restrictcdcode'
portraitof the
Anotherintereitingurp.ri of Bernstein's
a specificreferent:
without
they
pronoun
is the critical focuson the
then the global term
Inasmuchas referentsare not finely differentiated
impliedby
non-specificity
.they'will be adoptedas a generallabel.The
subsequent
thc
and
,they' is a function of the lack of differentiation
concretizingofexperiencewhichcharacterizesarestrictedcodeasawhole.
(rg7t: t to)
'they' and the
the use of
There is no connectiondrawn here between
pronouns
generalized
thoughasit will appearbelow'the useof such
passive,
forrnal
the
that
clear
seems
It
passive'
is a major alternativeto the agentless
by
passive
the
of
use
greater
a
to
did lead
experiments
contextin Bernstein's
speakers'
working-class
the
that
bui he did not realize
speakers
middle-crass
was simply their alternativeway of sayingthe
pronouns
generalized
useof
analysisamong writers with an
samething. Ttre absenceof a tinguisiic
limitations
orientationoften leadJto attributionof conceptual
educational
sPcakers'
to non-standard
asa linguistic
*ill beginwith a descriptionof the passive
unuiyr'i,
Our own
then present
will
we
r
(Scction
)'
variablean4 a iescriptionlf our samplc
thepassive
on
constraints
analysisof theexternal,otiolinguistic
aquantitative
on
constraints
internal
topic,
to our primary
z.r) beforeproceeding
(Section
ihechoiceof activeor passive(Sectionz'z)'
I. DNTINITION OF THE VARIABLE
to
it is necessary
systematically,
In orderto study any variablephenomenon
this
glance,
ftrst
At
with what?
definethe basicalteination:what is varying
alternateswith active' The
passive
passive:
the
seemsa simple matter for
several
subjectsmay be ill at ease'Bernsteinheld
the fact that working-class
tsl Recognizing
"' pi".iit" ,"riiont beforethe recordedsession'
33
LrN l9
E. JUDITH WEINER AND WILLIAM
CONSTRATNTS ON THE AGENTLESS PASSIVE
LABOV
neighbourhood
was settledby Germansalternateswith Germanssettledthe
neighbourhood.
But such agent passivesare rare events.In spontaneous
conversations
that we haveexamined,they make up only r to 2 per centof
the total numberof passivizable
transitiveverbswith subjectand objectnoun
phrases.On the other hand,passives
are not as rare as thesefigureswould
imply. Theyare quitecommonin conversation
and spontaneous
speech,
but
the great majority of them are agentless
passives.Early in this study, the
decisionwasmadeto focuson the agentless
passiveas the primary topic for
variableanalysis.6
The problemof definingthe alternantof agentless
passivesseemsquite
obscureat first. Discussions
of the Englishpassiveusuallydo not presenta
syntacticoptionif the subjectis missing.?
As we examinecolloquialspoken
English,however,we find a very rich alternationof truth-conditionallv
equivalentforms:
Causewe haveboundaries
in thisschool.Like out at like,the w - like you
know, Lower Merion'sallowedto smokein the halls'n' do all that crap,
right?over here,if th - I don't careif they neverallow you to smokein
the halls.
[Ato6r, 16,W, M, WC]t
In this passive,the first sentencewith allowmight havetakenthe form, they
allowLow'erMerion to smokein the halls,andthe second,I don't careif you'ie
never allowed to smoke in the halls. The pron oun they is not a purely
grammaticalformative(i.e.a dummyelement)asin the caseof therein There's
Amanat thedoor,or il in It's a shamehedoesthat. Theyhasa limited semantic
value in that it seemsto excludethe first personand perhapsthe second
singular as well. The choice of a particular generalizedpronoun is nearly
automaticin manycontexts.We wereableto makeonesystematic
small-scale
observationof this when thievesbroke into a lockedclosetin our building
that was usedby the alumni relationsstaff for storingliquor. Weinertook
the occasionto questionpeoplein the vicinity who weremilling aroundthe
brokendoor. In response
to'What happened?'she
receivedfive agentless
passives,
suchas'The liquor closetgot brokeninto' and fiveactiveresponses
of the form 'They broke into the liquor closet.'Three of the nonspecific
subjectswerethey andtwo weresomebody.
Note that thereis no information
[6] Langackerand Munro cometo a similardecisionin their analysisof the passive(rqZS).
of acquisitionof the full passive(r978),it wasn..*r"ry to resortto
Nj In Horgan's.study
experimentaltechniquesto elicit the forms needed,iince so few passiveiwith agents
occurredin spontaneousspeech.Shecites Harwood (r959), who found no instancisof
full passives
in over t2,oooutterancesof 5-year-olds.
This is as true of adulrsas of
children.Brown (rgZl) found no full passives
in zroo"utterances
of parentsstudied.
[7] Excluding,of course,the verbswhereergative-type
constructions
are availableasin The ,
door opened.
t
[8] Quotationsare identifiedby tape number, age, ethnicity (W = white, B : black), sex
(M/F) and socialclass(MC = middleclass,WC: workingclass).
I
34
hereon singularvs. plural, but merelyan exclusionof the speakerfrom the
classof possiblereferents.
We can thereforeinclude they with the classof [-definite] pronouns
people,etc. But a furtherdifferentiationmustbe made.
someone,
somebody,
indicatesthat a referentis not known to the hearer.
pronoun
A I-definite]
in regardto
thereare still two distinctpossibilities
given
situation,
this
Yet
in mind
particular
referent
a
have
may
speaker
the
knowledge:
thespeaker's
('a particularsomebody')or may not ('nobodyin particular');this applies
rquullywell to all of the [-definite] pronouns.The dimensionItdefinite]
abouthearer'sknowledgeof the referent;we
refersto commonassumptions
to refer to the seconddimensionthat bears
specific]
will usethe feature[*
on the speaker'sknowledge.
Given a l-definite] pronoun, [+specific]indicatesthat the speakerhas
of a particularreferentin mind; [-specific]that hedoesnot.The
knowledge
analytictask is thereforeto examineeach[- definite]pronounto seeif it is
'semanticallyempty' pronounsthat are the
[+specific]or [-specific]. The
passivesare [-definite, -specific].Pronounssuchas
of agentless
subjects
they,you, we are[+ definite]if theyhavea specificreferent,but areotherwise
[- definite,- specificl. Youdiffersfrom theyin that it more easilyallowsthe
possibilityof first and secondpersonsand is the closestequivalentto the
dummysubjectsGermanman, formal Frenchon, andformal Englishone.
I think you expecta lot of your children.
[ A 9 o Z , 4 r , W ,F , L W C ]
In rarecases,heor weor perhapseven/can beusedwithout specificreference,
ascan any personalPronoun.
The categoryof [-specific] noun phrasesis usuallyreferredto as general
or impersonal.The major equivalentin Englishof the impersonalmanor on
speechwe did not find a
is the formal one,but in our studyof spontaneous
singleexampleof the impersonalsubjectone.sThe [- specific]nounsformed
with some-) any-, and every-, along with people,have been commonly
recognizedas possiblesubjectsof agentlesspassives,but of the 96t such
we found, therewere only 69 of these,or 7 per cent of the total.
sentences
you and they accountedfor 79 per cent, and this was remarkablyuniform
acrossages and social groups.roBecauseimpersonalyou and they are
of colloquialspeech,their syntacticr6leassubjectsof agentless
characteristic
passives
has beenlargelyignoredin the past and evenstigmatizedas the
closerto colloquial speech
[9] This underlinesthe fact that the interviewswe are studyingare
than other types of speechthat might be analysed.Sinceit will appearthat passive
of passivemay
wittr more formal contexts,a higherpercentage
are associateo
sentences
appearin suchformal productions.
onal we(24 per cent)
tr ol Cinegroup conversationshoweda relativelyhigh frequencyof impers
but most-groupsadheredcloselyto a norm of 85 per cent for you and they.There ate
sizeabledifferencesin the useof you vs. they, apparentlydeterminedby factorsoutside
of the scopeof this studY.
35
2-2
t
E. JUDITH WEINER AND WILLIAM
THE AGENTLESS PASSM
CONSTRAINTS oN
LABOV
r
possibilitit'tt"
to the overallenvelopeof these
'1ll
the
obviously
approximation
are
they
Yet
first
A
referents.rr
not.
without
pronouns
improper use of
astransitiveverbswithl
analysed
be
may
which
agents
without
passives.
We will referto
constructions
mu.lor'.toice for ihe activealternantof agentless
,iru.tur.. it il; ^ l""tt1T:i""' 1H )k
underlying
theii
in
crNEn11'
as
specific]
objects
feature
the
with
used
realized
pronouns
[of
passivetthi
theentirecategory
agentless
passivevariablJ will lead to an
the
PERSoNAL
of
GENERALIZED
as
etc.
you,
they,
pRoNoUNS
occurrence
to
and
rzED
[-specificl
c e t o a n a c t i v e s e n t e n cactive
e w i tand
h a t passive
_ j T . i f rmust
c ] s uremalni
bject..I
PRONOUNS.
'extended n o n - o c c u r e n
the choicebetween
,t,lJyli
the
of
framework
the
in
passives
empiricuf
An
agentless
of
productivtl'*1"
In discussions
'''" i'l*'4
for ,.pr.r.niing the
of any current.formalism
deletionl
and
standardtheory', it has beenpointed out that the lexicalform somebody independent
rn''
or*ntrultion
study
of activ! andpassin..n, in the
the underlyingsubjectform, sincetherearecaseswhere relationship
cannotbeconsidered
This objection o f t h e c o p u l a ( l - a b o v , r y 6 g a ) , * " , u n r e c o g n i z e c l ebe
y t h e s eFormal
tofvariants,
activeswith sorrebodydiffer in meaningfrom the passives.
to a r llocated'
the gru**u' this variationis
in
between
where
say
consistently
distinguish
to
able
we
are
cannot
if
but
resolved
will of coursebe
transformationalmodel
und #;;;;,
arises
'nuiiiro*
of
confusion
ur.
type
Another
somebody.
of
uses
f;
:::t:
arguments
and[- specific]
[+ specific]
of
( C h o m s k y , r g s ! ) w o u l d s u g g e : t a v a r i a b l e pconstraints
a s s i v e t r aon
n s the
form
ation;alater
phrase
when usingg.n.ruiir.d forms which tend to excludesomepossibilities
would ,.luirt setiing
1965)
involve
cases
obvious
(Chomsky,
most
The
particular.
in
version
they
personalresponsibility
More recent
passivewith manner adverbial'
structurerule that biackets
active
negativeacti that the speakermight want to deny,like stealing,and herethe
between
relationship
leavethe
attackson the purrin, transformation
usJ of Theybroke in, or somebodl,broke ln might well seemto excludethe
I9?8),but it is not
(Bresnan,
andpassiv.u **,.,. of ,6unti.;;;tptetation
within the
speakeras a possiblereferent.It is not so clearwith neutralacts,as in They
tt iun be usedto incorporate
exclude
not
does
of
they
use
The
WC].
M,
W,
yetclear to us how such"pprou.t
madeiheirwine[59,Irish,
here'
olryo1,,
described
have
that we
or at any rate no morethan using
grammarthe choiceof activ. nr. fu$ive
towards
first personparticipationherenecessarily,
to-move
prefer
formalisms'we
Avoiding ilrr'inrluuility of these
to
peopleor somepeople.Conversely,the useof the passivein The closetwa
capacity
the
have
'bro'ken
which will
in
a theory wittr mor. ,otia founJations,
into mightbe taken to suggestthe lack of the speaker'sparticipation
language
of
use
the
iu* diawn from
incorporatethe unlimited Outu
just as stronglyas Theybrokeinto the closet'
comparable
is
here
areusingthe term
life. Th;;;rin. u.tirute., *.
everyday
Becausethe generalizedpronouns are not formally recognizedlike the
in sociolinguisticstudies
introduced
il
to the phonological variables
Given
impersonalpronounsof French and German,there is a tendencyfor their
of the selectionof a baseform'
reactions.On the
in introspective
(Labov,1966)ih-u,*.r' ina.p.rrJlnt
same
the
semantic.ontributionsto be exaggerated
saying
possible.ways of
oii*o
in
a mutually e*clusiu, choice
choice'
other hand,educationallyorientatedwriters,drawingon differentkindsof
this
that ietermine
of referentialcontent, thing- activeor passive.:Ilut ur. ,t. iactors
theabsence
of
intuitions,tendto recognizeandstigmatize
understanding
un*"' will illuminateour
speech
sofar as it is detirmined?The
and social
as noted above.However,closestudy of their use in spontaneous
rl""*ledge of their cognitive
the active and passivevariant,"
to will
revealslittle evidencefor the influenceof semanticdistinctionsin the choice
respond
pressuresthey
if ;;;;nO ,t . kinds of linguisiic
place
significance,
specific
of activeor passive,and on the other side,little supportfor the notion that
a
to
of the variation
*oie confiden,;6;Jnt
or
uneducated
of
towaro,
characteristic
are
lead
among
gene
ralized
choice
a
but
ironouns
^
"t:"l1iY1l^.0:ry::l
rulecouldbewritten'
"
passivevs. activewith generalizedsubjectpronoun
in thegrammar.A variabr,'Vnt".iiz
ihe choiceof agentle'Ss
made without a number of
the variousalternativescannJ;intelligentlyagentsspecified'
appearsto beusedby all speakersof Englishastwo alternativewaysof saying
leftdislocations'
activeswith
parallelstuoiesorrelatedrulesof
the samething. For brevily, we will refer to this choiceas agentlesspassive
predicateaai,.tiueconstruction',unasoon.-Hereourfirsttaskistoreport
active,or in this paper,passivevs. active.Our useof the term
vs. generalized
variableat hand'
i" i.n"lig and delimitingthe
. agJntless'
that couldnot havean agentlike They
the decisionsalreadymade
doesnot referto sentences
speechis to
problem in ir.'analysis of spontaneous
and
sentences
to
but
possible
here,
is
fundamental
passive
choice
one
no
since
w,inrhome,
typesthat do not
from intransitiveor otherverb
nerbs
on
information
no
tr"".iti"e
but
implied
is
agent
of
an
discriminat,
existence
contextswherethe
s of have(e.g.Bach,t967),
analyse
*""v
Following
objects.
direct
take
is
exPressed.
hffi or her identitY
with the obvious
".t'er
,f,. 'ultf transitiveverbs'along
from
this
it
does
I
where
and
occurs
variable
exclude
the
where
would
we
ipecify
to
of symmetry
We are no* ubl.
i
as cost and weigh' and verbs
with
casesof verbs of measuresuch
i
symmetrical
whosesubjectsu"
andresemble
marry
as
such
"i'untically
'exophoric'
agentsor patients'
and similar writers havearguedthat the useof
objectsuJOo not qualify as
their
[ t ' ] As noted above,Bernstein
code',
senseto includeverb plus
the'restricted
.
in
of the lack of differentiation
pionoun, like thet,is a consequence
transitiv.l hrr.-in a broad
word
the
use
we
judgedto have
abilityat the higherrlnge:'(t9lt:8t)'
generalizing
nart of the Datternwhich'in-hiUits
pr.poriritn when the clauseswere
plus
verb
somewhat
a
produced
even
have
might
or
pronouns
particle
imp-ersonal
of
rult iung.
6i;;.;'ri;)i;iih;
room+'+Theroom
r;;;;[
, Thry lookedat the
F"t
coder.
ifre
Uy
differentresultin their analysis.
passive
a
37
36
t
E.JUDITH WEINERAND WILLIAM LABOV
LUNSTRAINTS ON THE AGENTLESS PASSIVE
The recognitionof
w'aslookedat. The boundary betweentransitiveverb and intransitiveverb
isnot conceptuallydifferentfrom phonologicalanalysis.
I wentpas'him
oreven
yesterday,
plus prepositionremainsas an area of vaguenessto be investigatedby
aphonological,eio suchas I pass'here
must be
(and
discourse)
empiricalmeans.The proportionof suchcasesthat weredecidedon intuitive
is not a michanical process.The whole sentence
must
meaning
present
past
a
or
groundswas approximatelyt5/, of the total.
of a
first,and ihe appropriateness
understood
passive'
the
of
case
the
As in
Clausescontainingverbswith sententialobjectswerealso systematically bedecidedsubsequent
to semanticprocessing'
excluded.Whilea sentence
like Thevsaythat tintesarehard,withgeneralized theanalystsmusi usetheir full knowledgeof the languageto recognizethe
the1t,could conceivablyhavethe alternateThat timesare hard is said(tobe
of the variable.A formal statementof this procedurewould be
occurrence
thecase),we found in accordance
with our intuitionsthat extrapositionwas
on a completedformal analysisof the grammaras a whole'
dependent
alternantincludesboth
categorical
whenthepassive
alternantwasproduced,andextraposed
sentences
Thedefinitionof thevariableincludedin thepassive
between
like It is saidthat timesare hard involvechangesof surfacestructurethat are
thegetandthe beforms.Thoughthereis a clearsemanticdifference
law (R'
incompatiblewith the constraintsto be considered.
Our examinationof
John gofarrestedto testthe law and John was arrestedto test the
basic
internalconstraintson the passivewill give particularattentionto the
Lakoff, rgl1, experimentsconducted by Labov show that the
equivalent
consequences
of placingsubjector objectin preverbalor postverbalposition
without iutpot. clausesare interpretedassemantically
sentences
gel
remaining
The
in relationto parallelor non-parallelplacementof coreferents
function.
causative
a
had
it
get
where
in preceding
(rgZS).We excluaia
sentences.
once extrapositionhas applied, the underlyingobject of the
were tabulatedseparately,but includedwith the 6e forms as
sentences
agentless
passiveis no longerin initial, preverbalpositionand the effectof
of applicationof the variablerule. The selectionof the passiveas
instances
at this point,
suchparallelstructureis eliminated.
significance
theoutput ofihe variablerule hasno substantive
flormshad
active
generalized
we alsoexcludedthe 'quasi-modals'wantto, beginto, stop,etc.,in clauses
if
the
obtained
and the sameresultswould be
with EQUI-NP deletionapplied,suchas Theybeganto botherMary, though
beenconsideredthe outPut.
white
from working-class
Mary beganto bebotheredisquite
acceptable.
Thechangeof subjects
for both
Webeganour analysiswith a studyof zt speakers
interviews
in
their
philadelphia.
verbsled to a higherprobabilityof semanticdifferentiation.rz
sentences
agentless
The
neighboirhoodsin
It was also necessaryto discriminateforms with -ed or -en which were
contextin an earlierexploratory
haJalreadybeenextractedwith considerable
clearlypassives
currentframeworkto yield825
from homonymousforms which wereadjectives.
our
Thereare
Thesedatawerere-codedwithin
study.rs
a numberof waysto discriminateadjectivesfrom passiveparticiples,most
the centralfiguresfrom our
of
two
tokens.To this initial group we added
and t t speakersfrom
obviously,their compatibility with adverbialsuffixessuchas -ty; happilybut
studiesof working-clasiPhiladelphianeighbourhoods,
We then
items.ra
more
giving
not instructedly.But suchpotentialcompatibilitydoesnot prove that a given
482
twosuburbanmiJdleclassneighbourhoods,
in order
dialect
different
form is an adjectiveand not a homonymouspast participle.Thus distracted
decidedto add a group of speakersfrom a radically
passives
among
agentless
doesaccept-ly but it may not be a predicateadjectivein He wasdistracted
of
to assessthe range of diff.t.nces in the use
group
adolescent
a
black
by this,The intuitivecriterionwhichwe appliedin practiceappearsto bean
Jets,
Englishdialects.We studiedten membersof the
from
sentences
agentless
effectivetest of verbal status: if the subjectof the agentlesspassivecan ba
fr; Harlem (Labov er al., 1968),extractingall
zo3
additional
yielding
an
transformedinto an hctive with a generalizedpronominal subject,it is
as well as individualinterviews,
groupsessions
includedin the variable.We excludefrom the definitionof the variablethose
tokens.
-edor -enforms which do not correspondto an activeverbwith a generalized
Thevariablerule programmeof cedergrenandsankoffwasusedto analyse
choice
theyor you. Conversely,if a pronominalsubjectcan be transformedinto an
the influenceof u nu*b.t of externaland internalconstraintson the
The version of the
agentlesspassivewithout radicalchangeof meaning,it is includedin the
of active or passivein the 1489agentlesssentences.
analysis.
programmeused here is Varbrul II (Sankoff & Labov' 1979),where the
We havestressedthat the quantitativeanalysisof this syntacticvariable
urtiout contributions to the output probability in any given case are
as
represented
PpoPrPzPn
Itzf A readerfor thisjournalpointsout that thereis a potentialambiguityin Mary beganto
l*ui
be hotheredthat is.not fgund in The1,beganto botherMart,.ln tfie first sentence,-began
may be understood
as taking a sentential
subject,'lt beganto botherMary', but thisis
not possiblein the second.Though we excludedthis casefor other reasons,this is one
of a numberof caseswhere the generalizedpronoun doesnot atternatefreily with the
agentlesspassive.Wheneverthe subjectunderstoodis a singlesituation thai might be
referredto by it, thef isnot appropriaie.As notedabove,the testfor the variableexcluded
suchcases.sincethe subjectof the passivecould not be transformedinto an activewith
a generalized
pronominalsubjectwithouta radicalchangein meaning.
38
|- p: GAG:pJG;J
..G-p)'
studyof thespeech
by membersof a classon the
It l] Interviewsand initial codingweredone
whosecontributionwe
of Pennsylvania,
community,Linguistilt;A; at the Univlrsity
gratefullyacknowledge.
carried out by Anne Bower, and the middle-class
trqt ?rr. *"ir.ing-class in-terviewswere
and Variation'
interviewsUv ni"i1" Fuvn",of the projecton LinguisticChange
39
E. JUDITH WEINER AND WILLIAM
CONSTRAINTS ON THE AGENTLESS PASSIVE
LABOV
wliere P is the over-all probability of the rule applying,Po is an input
to theover-allprobability
probability,andp,throughpoarethecontributions
factors.If a factorhasno influenceon the rule,it will
from environmental
show a valueof o.5 (sinceo.s/(t-o.5) - l). Valuesover o.5 favourthe
applicationof the rule and valuesbelowo.5 disfavourit.
2. EXTnRNAL
CONSTRAINTS ON THE CHOTCE OF ACTM
OR PASSM
pronounsarecharacteristic
ofcolloquial
As mentionedabove,thegeneralized
useof the passive
that the excessive
English,and it haslong beenrecognized
is a mark of formal written style,in scientificas well as literarywriting.The
casualand
of our materialsfollowedthedistinctionbetween
stylisticanalysis
caref'ulspeechdevelopedin previousstudies(Labov, t966, 1972:ch. 3;
Cedergren,1973)but with the particularcodingdevelopedin the studyof the
sentences
community.tsThe resultsshowed639agentless
speech
Philadelphia
showedthe
in careful speechand 85o in casual.The over-allpercentages
favouringof activeover passivein the expecteddirection,with 4o per cent
passivein carefulspeech,though the differenceis not a largeone. Table t
from an analysisincludingzz
adds the variablerule probabilities,,taken
environmenta|f@Theprobabi|isticweightsforstyle
percentages:
o.54for carefulspeechand o.46 for casual.
reflEi the-e
Passive
(%)
Style
Careful
Casual
6lg
85o
40
32
at the o'oz
Thuschi-squarefor this stylisticresultis 5.6,which is significant
largeand
shows
coding
stylistic
same
The
levelfor one degreeof freedom.
(DH)
and
(lNG),
as
variables
sociolinguistic
with such
regulardifferences
comBy
sample.
Philadelphia
our
of
subsection
neiativeconcordin every
passiveis not an importantstylisticfactor
fuiiron, the choiceof aciiveand
speech,thoughwe haveno doubt that moreformallanguage
in spontaneous
for the passive'
wilf showa stylisticpreference
variablein English,and
Thechoiceo[activeor passiveis a wellestablished
the sexof thespeaker
that
expect
wewould
if it hadstrongsocialsignificance,
it hasbeen
variables,
stable
of
wouldplay a major r6le.tn all previousstudies
form'
a
stigmatized
of
foundthat womenusemoreof a prestigeform andless
t966;
Riley,
&
wolfram
in carefulspeech(Labov, r966;Shuy,
particularly
do
usea
women
percentages,
When we look at the over-all
Trudgill, r-g72).
analysis
multivariate
more of passivesthan men, but the
snrallpercentage
this is the resultof other distributionalfactors.Table z
that
clearly
shows
variable,
showsthesomewhatsurprisingfactthat sexhasno effectat all on the
likelihood,
log
in
withboth men and womenat exactlyo.5o.The difference
- o.oo.
nothing
precisely
is
factor,
a
as
sex
whenwe remove
VARBRUL
weights
o.54
o.46
Table t
sentences
in agentless
Effectof styleon the choiceof passives
of this typecan be relatedto standardstatisticalmeasures
Smalldifferences
of confidenceby comparingthe total likelihoodof a variablerule analysis
casualfrom careful speechwith a secondanalysisthat
that distinguishes
ignoresthis factor.The differencein the sumsof the log likelihoodsfor each
with degrees
cell,for eachanalysismultipliedby - z, is equalto chi-square,
Ir 5] Thecodingsystcmuscdherediffersfrom that first developedin Labov t 966in that channel
cuesare not utilized.A seriesof discretedecisionsmark as contextslor casualspccch
groupintcractionandtangcntial
movements
of kids'gamcs,
pcrsonulnarrutivcs,
tliscussions
of
oralorical('soapbox')style,and discussions
to questions,
by the spcakcr;responses
lunguagcarc codcd for carclll spcech,as well as the general unmarked body of
convcrsation.
consultativc
4o
of the two
of freedomequal to the differencein the degreesof freedom
analyses.
X' : -z (SUM(LL,) SUM(LLr))'
Tokens
(%)
VARBRUL
weights
759
730
32
39
o.50
o. 50
Passive
z5 males
t9 females
Table z
Effect of sexon the choiceof passivesin agentlesssentences
of
We thenturn to socialclass,wherepreviousreports(particularlythose
to
one
lead
might
Iz5))
I
r8,
t09,
(Lawton,r968:
and his students
Bernstein
passives
than
fewer
use
expectto find that our working-classspeakers
speakers.This is not the case. In Table 3, both the raw
middle-class
and the variablerule programmeoutput indicatethat the white
percentages
uset'hefassivealternantmorethanthemiddlePhiladelphians
workinglslass
from NewYork aremost
classwhitesat pi of o.58,whiletheblackadolesccnts
groups.when we examinethe log likelihoodsas
similarto the mioarc-.tass
at the
are significant
outlinedabove,it appearsthat both of thesedifferences
than
more
o.or level: working-.lutt speakersuse the passivesignificantly
4l
E .J U D I T H W E I N E R A N D W I L L I A M L A B O V
Passive
(%)
Tokens
z3 white workingclassPhiladelphians
t t whitemiddleclassPhiladelphians
ro blackadolescent
New Yorkers
CONSTRAINTS ON THE AGENTLESS PASSIVE
in agentless
sentences.
By comparison,suchstablesociolinguistic
variables
as(DH) or (ING) s
masslve
externa
VARBRUL
weights
variable.
gBz
o.58
213
27
o.46
27
o.45
whichis quitestriking:thechoiceof auxiliary.Table5 showsthedistribution
of get vs. be for adults and teenagers,
male and female.Adults show a
preponderantuse of be, as do femaleteenagersto a lesserextent; male
Auxiliary
Tablej
Effectof classand ethnicityon the choiceof passives
in agentless
sentences
Be*
Sex/agegroup
(%)
Tokens
Get
(%)
.1
\v
r' \
class.Whenwecollapsewhiteworking-class
andmiddle-class
speakers
| lOj.te
(Tu|le ro, Run 6), the chi-square
is 4.g; whenwe collapsewhiteand black
.rr\l
workin8classspeakers
(Tablero, Run 5), thechi-squarc11
9.7.Thoughthese
, ,V\yl
significanteffectsare only moderatein size,they reversethe expectations
"/
|
createdin the literaturereviewedabove.A furtherunderstanding
oithe result
I
I can be obtainedfrom a comparisonwith Van den Broek'sstudy
of class
differences
in Belgium(seebelow).
We can also examineage distributions.Though we have no reasonto
suspectchangein progress,theremay bea differencebetweenthe r9 teenagers
in our sampleand the z3 adults.Table4 showsthat the adultsdo favourthe
t/
t'-,,i
Passive
t9 adolescents
z3 adult s
Tokens
(%)
VARBRUL
weights
418
t o 7r
z8
38
o.46
o.54
-'*t'
265
r7o
77
78
20
I "^"tj
20
, ^oYst A
Adolescents
Female
63
3o
37
Male (total)
87
66
32
Black
25
56
75
White
31
48
52
* The remainingpercentages
are for havepassiveswith numberstoo small
to be consideredfor socialdistribution.
Table 5
Distributionof (get) and (be) passives
by sexand age
Table4
Effectof ageon the choiceof passives
in agentless
sentences
I
I
pu#iu. by the samedifferences
in p, scoresthat we saw in the stylegroup:
o.54 for adults as againsto.46 for adolescents.
But here the ed'ectis not I
significantevenat theo.o5levet:log likelihooddifference
whenageis removed
(Tablero, Run 4), is only z.r.
Thoughageand socialclass
arenoticeable
effects,theydo not indicatethat
externalfactorshavea sizeableinfluenceon the choiceof activeor passive
42
Adults
Female
Male
teenagers
are significantlydifferentfrom all other groupsin their heavieruse
of get,and this tendencyis strongeramongblacksthanwhites.Theseresults
parallelthose of Feaginin her study of white working-classadults and
adolescents
in Anniston,Alabama(tglil.In Anniston,we find a shift from
lessthan 30 per centget for adults to closeto 8o per cent for adolescents
(p.
9Z).Both teenageboys and girls showedthis high useof get. A shift to the
gel passiveappearsto be one of the most activegrammaticalchangestaking
placein English;and at leastin the North, it seemsto be alsoa stigmatized
sociolinguistic
variantwhich is usedmore by malesthan females.
3 . I N T n n N A Lc o N s r R A r N T s
Our attention is then drawn to internal factorsthat may help us to gain a
betterunderstanding
of the linguisticsignificance
of this massivevariation.
43
E. JUDITH WEINER AND WILLIAM
CONSTRAINTS ON THE AGENTLESS PASSIVE
LABOV
children were cued by pictures to remembera sentence,there were more
of
correctrecallswhen t-hepicture focusedattention on the surfacesubject
object
the
on
the activeor passive,but when the picturefocusedattention
therewas a significanttendencyto reversethe voiceof the
of the sentence,
The conceptof ctvENvs. NEwhasbeenput forward as a major explanatory
devicefor the passiveas well as other left and right movementsof major
constituents.It is generallyconsideredthat in the unmarkedcase,crvnN
NEwinformation:
informationprecedes
In the unmarkedcasethe new is, or includes,the final lexicalitem,so that
the unmarkedsequence,
excludinganaphoricelements,
is givenpreceding
new; but the focuscan appearat any point in the informationunit. The
constituentspecifiedas new is that which the speakermarks out for
interpretation
as non-derivable
information,eithercumulativeto or contrastive with what has preceded;the given is offered as recoverable
anaphorically
or situationally.
(Halliday,ry67:zrr)
and in the passive,the logicalobjectbrought to the early subjectposition
crvENinformation:
represents
. . . if the verbof a sentence
is an action-process,
its patientnoun root will
convey new information and its agent noun root old information.
(Chafe, ry7o: ztg)
Neitherof thewriterscitedhereconsiderthechoicebetweenagentless
passive
active.But a choiceis impliedin their consideration
and generalized
of the
'semanticfunctions'of the passive.
Chafepresents
the passiveas a way of
achievingtwo functionssimultaneously.
One is to allow the verb to appear
without an agentor an experiencer.
The other function of the passiveinflectionis to changethe order of
prioritiesfor the distributionof new information.
QgTo:zzo)
Given the possibilitywe are consideringhere of two ways of presenting
transitiveverbswithout agents,it is clearthat Chafe'sanalysiswould show
passivevs.generalized
that thechoiceof agentless
activeis determined
by the
vs.
NEw
GIVEN
statusof the two referents:the actionor experience
referred
to by the verb on the one hand and the referentof the patientnoun on the
other.
Experimentalevidence.There has beena considerableamount of experimentalwork in searchof objectivecorrelates
of thesubtledifferences
between
activeand passivesentences.
Psycholinguists
havenot thoughtin termsof
clvENvs. NEwbut ratherin termsof a wide varietyof cognitiveapproaches
suchas 'conceptualfocus'(Tannenbaum
& Williams,r968),'salience'and
'focusof attention'(Turner
& Rommetveit,
r968),or'importance'(Johnsont-airdl r968).Most of theexperimental
approaches
havecontrasted
activeand
passive
sentences
with agentsexpressed.
(r968)showedthat
Johnson-Laird
sentences
such as ' Blue is followed by Red' tendbdto attribute greater
importanceto their surfacesubjectsthan sentences
such as 'Red follows
Blue',asreflected
by association
of thepassive
sentences
with largerblueareas
in colouredrectangles.
Turner & Rommetveit(1968)showedthat when
44
I
l
ia
l
$
F
F
H
&
sentence.
Tannenbaum& Williams(rq6S)focusedattentionon subjector objectby
a devicewhich is somewhatcloserto the effectsobservedin spontaneous
with preamble
First-yearjunior high-schoolstudentswerepresented
speech.
with the nounphraseof conceptualfocusin consistently
textsof six sentences
and they wereaskedthen to form an activeor a
position,
passive
activeor
iescribinga picturein which that referentwasseenas the
passivesentence
agentor patient of an action. Latencytimesfor the completionof active
actives
between
thedifference
,*t.n.r, werealwayslessthanfor passives;but
on the
focused
preceding
text
the
where
for
cases
maximum
was
andpassives
logical
the
on
focused
the
text
when
greatly
reduced
logical subject,and
to a patientfacilitatedthe
object.Thus a precedingstring of six references
in initial position.
referent
that
with
formation of passivesentences
passivevs. active
of
effect
a
smaller
found
also
Tannenbaum& Williams
reduced
preambles
passive
the
when
preamble,
in the
form of the sentences
The
significantly'
not
but
productions,
the over-all latency of passive
predominance
the
support
to
Tannenbaum& Williamsexperimentappears
contrastwith agents
of the crvnN vs. NEw effect for the active-passive
expressed.
Recent discussionsof crvnN vs. NEw have taken on an increasingly
cIvENasthat whichcanreasonably
character.Chafe(tgl Opresents
subjective
at somepoint in
of the addressee
be assumedto exist in the consciousness
sensethat the
technical
the
in
new
a discourse.Information may then not be
presentin a
be
it
can
that
informationwas objectivelyabsent.Chafeargues
(p. t t z).
time
a
certain
at
memorybut not be in his consciousness
speaker's
assured
be
may
status
Iie points out that the duration of any item's cIvEN
of 'old,
opposition
the
(p. r z9).Kuno (rg7z\ dealswith
for only onesentence
'new, unpredictable
information',but he does
information'and
predictable
not indicateany confidencein objectivecriteria for this distinction.
use
In proceedingwith an objectivestudyof variationin the spontaneous
ways
the
active,we will naturallywant to consider
of passiveand generalized
in which such subjectivefactors might interfere with the validity of an
objectivecodingprocedure.First thereis thematterof how far backwewould
wantto searchior evidenceof crvsNstatus.If the refierentof a givensentence
it can reasonablybe assumed
occurredin an immediatelyprecedingsentence,
that it is availableto a speakeras a'given'. But if we wereto confineour
in
attentionto only this sentencewe would be losing a gteatmany GIVENS
To searchthroughanentireconversation
preceding.
immediately
thesentences
as GIVENwhich are not immediately
elements
*.ny
would assuredlylocate
A decisionto searchfor a fixed
attention.
in the foregroundof the speaker's
45
E. JUDITH WEINER AND WTLLIAM
LABOV
numberof preceding
sentences
is necessarily
arbitrary,but thevalidityof that
decisioncanbeexaminedin a multivariateanalysis
by examiningsuccessively
the effectof beingctvENin sentences
successively
further removed.It seems
clearthat GIVEN
statusis not discretebut gradie'nt,and we would
expectto
find that the effectis strongestfor the imirediately
precedingrrnt.n., una
diminishesgraduallyas we recedeinto the p.rl
of tire preceiingdir.ourse.
Any valid analysisshouldbeableto demonsirate
an effectivevanlshingpoint
wherethe inclusionof all precedingsentences
would makeno differenceat
all.
w,
4w
\)
The other subjective
effectis that the elementmay not be.given, at
all in
an objectivesense,i.e. not occur in the surface
structure,but be presentin
the speaker'sconsciousness
as part of a long chain of inferencefrom the
surfacestructure.we can minimizethis effeciby
not limiting ourselvesto a
searchfor formswith identicalsurfacestructur.l
but acceptinga very broad
notionof coreferential
noun phrase.Two nounphrasesarecoreferent
if their
intendedreferentis thesamedir.ourr. entity.
Somesubtleformsof reference
of inferencewill stiil escapeus. But it seems
to us that this wiil applyequaly
well to the agentless
passiveand the generalizedactive.In some
senses,
the
referents
of 'they' or 'you' may beconsidered
part of a hiddenagenda.There
is no way that we can predictwhat the
effectlf suchsubjectivefactorsmay
be on the whole,but in any casewe will be
cautiousin the interpretationof
our resultsand be sureto rest our caseon
robustand consistentresultsof
objectivecoding.
An objectiveand relevantdefinitionof 'sentence'
woutdseemto becrucial.
The mostabstractextractionof underlying
sentences
would hardlydo, since
many of theseconstructswill have elided
one or the other of the two
constituents
weareinterested
in andwill not berelevantto thechoiceof active
or passive'on the other hand,the notion
of a sentence
in surfacestructure
definedby intonationalcontoursis difficult
to determinereliablyin spontaneousspeechand not necessarilyrelated
to the organizatronof form and
meaningthat we areconcernedwith. The 'sentences'
we will considerassites
for the locationof relevanthounphrases
will be finiteclauses.
Theclausewill
thus provea more usefulunit for our purposes.
we thereforeproceedwith a definitionof
crvnN as follows: the togical
object of an agentlesscrauseis defined
as crvEN if any noun phrase
coreferentialwith it was presentin any one
of the precedingfive crauses,
irrespective
of the terminationof speakerturns. Thus in
the sentence
(s) +r theyaskedya to staywashthe
windows,you washedthe windows
( 5 6 , W ,F , W C ) .
lhe v'indowsis ctvEr'rin the clauseprecedingyou
washed"the
windows.Acrvnx
noun phraseis thereforeone that hasu.or.f.r.ntial
noun phraseanywhere
in.the precedingfive clauses.when we
examinethe over-alldistributionof
this factor amongactiveand passiveagentless
clauses,our first impression
q6
CONSTRAINTS ON THE AGENTLESS PASSIVE
is that it has a stronginfluencein promotingthe passive.When the logical
objectof the clauseis cNrN in the precedingcontext,it occursin initial
positionas subjectof a passiveconstruction4I per centof the time; when
it is unw, or not cIvEN,this happensonly 25 per cent of the time.The other
75 per cent of NEwnoun phrasesare locatedin post-verbalpositionas the
objectsof activeverbs.
BeforeweconcludethatcIvENvs.NEwis thedominantfactorin determining
thechoiceof activeor passive,it would be wiseto considersyntacticfactors
whichintersectwith but are distinctfrom this one.Our earlystudiesof the.
passivesdrew attention to the effectof surfacestructure,whether
agentless
or not the logicalobject referredbackwardto coreferentialnoun phrasesin
to bea powerful
subjectposition.Suchparallelismof surfacestructureseems
passives
arefavoured
passive,
and
factorin determiningthechoiceof activeor
parallel
its
co-referents.
with
position
whenthe logicalobjectmovesinto a
The over-alldistributionof this constraintprovesto be somewhatlarger
than the clvENvs. NEweffect.When coreferentialnoun phraseswerelocated
in subjectpositionin precedingclauses,the logicalobjectof the agentless
clauseappearedin parallelsubjectposition58 per centof the time, that is,
the passivealternantwas chosen.When this was not the case,the passive
choicewas realizedonly 29 per cent of the time.
Thus the clvrN-NEwdistinction produceda t6 per cent differencein the
choiceof a passive;the parallel vs. non-parallelsurfacestructureproduced
comparisonscan be
a 29 per cent difference.But such across-the-board
The
unlessthe data areevenlydistributedamongall possibilities.
misleading
variablerule programmeallows us to comparethe effectof both factorsin
a singleanalysis;Table 6 indicatesthat the weightcontributedby the choice
of passiveby cwnN is o.54 as againsto.46 for Nnw; whereasparallelvs.
derived
non-parallelsurfacestructuregivesus o.6z vs. o.38.The chi-squares
#n:
Tokens
Given
New
Passive VARBRUL
(%)
weights
955
4r
s34
25
ParallelSS
58
350
29
II39
Non-parallelS S
** p < o.ol.
* p<o.os.
o.54
o.46
I
x2
25.27***
')
o.6z
o.38 )
***
7r.6t
*** p < o.ool
Tqble 6
Given vs. new and parallel vs. non-parallelsurfacestructureas constraints
on the passive
47
E. JUDITH WEINER AND WILLIAM
r - r . ; | r S ' l ' R A I N T SO N T H E A G E N T L E S S P A S S M
LABOV
Iiorn log-likelihoodcomparisonsare both highly significant,though the
parallelstructurefigureis at a higherorder of magnitude.
However,this comparisonmay not evaluatethe effectof crveNvs. r.rEwin
noun phrasein the immediatelyprecedingclauses
full, sincea co-reflerential
may havethe strongesteffect.We thereflore
distinguished
eachcrvEx by how
far backwe had to go to find the coreferential
noun phrasewhichestablished
that status:
Given [G]
o not givenat all in the precedingfive clauses
t givenin the immediatelyprecedingclause
2 not given in the precedingclausebut given in the secondpreceding
clause
givenin the two precedingclausesbut givenin the third preceding
not
3
clause
4 not givenin the threeprecedingclausesbut givenin the fourth or fifth
precedingclause
At thesametime,weextendeda similaranalysisto thefactorgroupof parallel
surfacestructure.But parallelismis disruptedby any interveningnon-parallel
clauses,
and a moredetailedanalysisof thisgroup mustconsiderthe number
parallelclausesprecedingthe onein question.In otherwords,
of consecutive
thestrongerparallelstructurecarcsareall subsets
of thebasicconditionsthat
the immediatelyprecedingclausemust containa parallelsubject.
SurfaceSubject[SS]
o is not corcferentialwith subjectof precedingclause
t is coreferentialwith subjectof precedingclause,but not the clause
beforeit
with subjectsof two precedingclausesonly but not the
z is coreferential
preceding
third
clause
3 is coreferentialwith subjectsof three precedingclauses,but not the
fourth precedingclause
4 is coreferentialwiih four or more precedingclauses
Table7 showstheresultof thismoredetailedvariableruleanalysis.
ThecrvEN
vs. NEweffectsshowa fairly regulargradationof weights,but in a direction
oppositeto our first expectation.When the logical object is given in the
immediatelypreceding
clause,thereis no effect.On thecontrary,thiscategory
(Givent ) hasa weightof o.39,slightlylessthan not crvENat all (Giveno).
It is only whenthe ctvrn is locatedin the secondclauseprecedingthat there
is a H2eableeffect,of o.58.A coreferentialnoun phraselocatedin the third
or fourth precedingclauseis only slightly higher,0.6o,and in the fourth
precedingclausethe figurefalls off to o.53.
If theeffectofctvrN statuswerea powerfuldeterminant
of thepassive
form,
it would follow that the largestfigurewould be Given r, and Given 2,3 and
figuresshowthat the
4 would fall offin that order.However,the significance
Given
o
I
534
6r6
2
t46
3
4
SS
o
6r
r32
I
I I39
214
2
57
3
4
37
42
p < o.o5.
Passive
(%\
VARBRUL
weights
25
42
43
47
36
o.40
o.39
o.58
o.6o
o.53
29
5r
7o
67
64
o.2l
** p < o.ol.
x'
25.27***
r8 . 4*r* *
I.I3
o.88
o.45
7t . 6 t * + *
r o . 4t *
o.63
o.87
o.67
o.85
o.56
*** p < o.ool.
Table 7
Surfacestructureconstraintson the passive
t
i
t
II
r
&
48
Tokens
I
t
F
I
&
almostentirelyin Givenz.Thechi-square
effectofcrvsNstatusisconcentrated
of the effectof separatingeach
significance
the
oppositeeachfigure shows
factor above.Thus the effect
the
from
it,
below
fa.ior, along with the factors
with a f of
o is sizeable,
Given
from
2,
t,
all Given
3,4
of disiinguistring
z' When
Given
of
effect
the
on
based
is
part
that
of
25.27;Uut the largest
effect,
very
significant
is
a
there
o,
I
and
from
Given 2, 3, 4 ut. itpuiated
when
Given
significance
additional
3 and
no
is
with Xs oi rA.4r. But there
z.
4 areseparatedfrom Given
Thus we obtain a sizeableimprovementin likelihoodof the analysisby
noun phrase
Giveno and r from caseswherethe coreferential
distinguishing
losses
significant
no
are
there
but
wasin a secondprecedingor earlierclause;
Given
z
and
Given
between
3, of
or gainsfrom makingfurtherdistinctions
Given3 and Given 4.
On the other hand, if the coreferentslocated in the third and fourth
we would
clausehad No effectin encouragingthe choiceof passive,
preceding
out these
we
separated
expectthat the effectof Given z would increasewhen
that
category
by
a
,.i.r. But this doesnot happen:the weightcontributed
we
when
is
o'61;
includescIvENin the znd, 3rd or 4th precedingclauses
remains
but
rise
not
Given 3 and Givenf fto* Given z, Given z does
separate
practicallythe same.It followsthat the effectof ctvnNin the third and fourth
E.JUDITH WEINER AND WILLTAM LABOV
C O N S T R A I N T SO N T H E A G E N T L E S SP A S S I V E
precedingclausesis not to be ignored,but has
about the sameeffectas
Givenz.
We cannotaccountfully for the locationof the crveNeffect
in the second
precedingclause,thoughit may well be connectedwith
the fact that many
of the immediatelyprecedingclauseswere located in
the same surface
sentence
and thepassiveverblocatedin a sententialcomplement.
Thepresent
analysisdoes not distinguisheffectsacrossclauseboundaries
from effects
acrosssurfacesentence
boundaries;the resultof Table6 clearlycallsfor such
a distinctionin further studiesof the passive.
When we turn to the analysisof parallelstructure,we
find a different
picture.The effecton the prrrium of sbr, only
oneprecedingparallelclause,
is considerablygreaterthan SSo,and sSz and s-s3,
*itti iong., parailel
strings,are progressivery
greater.The improvementpeaksat SJ3 and fals
offsomewhat
at SS4.Thechi-square
of 7r.6r indicates
thetotalimirovement
in likelihoodobtainedby distinguishing
all casesof parallelstructurefrom
caseswithoutparallelstructure.Whenwe separate
siringsof two and more
parallelclauses,
thereisa moderate
improvement
of likelihood,
withchi-square
at Io'4t' But if we then separate
out stringsof threeor more,the effectis
not sig.nificant,
(x' : o.8z).It foilows thatltrings of four or
more are arso
not going to showa significanteffect.The Varbiul
weightsfor stringswith
z, 3 and 4 precedingparallelsare all approximately
in the samerange,and
this effectseemsto showthat nothingis
iained in tire way of explanationby
examiningsentences
evenmore remote.
To make a finer comparisonof cIvENvs. NEwand parallel
structurewe
mustconsiderexactlywhat the relationsof thesetwo faciorgroups
are.From
the definitionsjust given,it is apparentthat all the cases parallel
of
structure
correspondto a singlecategoryin the crvENfactor group:
that is, Given r.
Sinceparallelstructurerequiresa coreferential
nounphrasiin theimmediately
Tokens
Gr(P)
Given with
paralleIstructure
Gt(- P;
Given with
non-parallelstructure
Passive
(%)
VARBRUL
weights
36o
58
o.54
277
23
o.36
Table 8
Given with parallelstructurevs. givenwith non-parailer
structureas
constraintson the passive
5o
I
I
I
preceding
clause,theneverycaseof parallelstructurefallsinto the first Given
category.On the other hand, there are casesof coreferentialnoun phrases
in the immediatelyprecedingclausethat are non-parallel:that is, SSo.This
is obviouslythe casewhen the logical objectoccursas the surfaceobject of
theprecedingclause.
Figure r sumsup the interrelationsof thesetwo factor groups:it shows
a four-celltable with the intersectionof cIVENvs. NEw and parallelvs.
non-parallelsurfacestructure.The upper right quadrantis empty: if the
New
ss
1,2,3,4
Parallel
I
2
o
G
GO
.ss
ss0
Non-parallel
Figure L lnterrelationshipof given vs. new and parallelvs. non-parallelstructures.
logicalsubjectis not crvEN,it cannot be parallel.In the upper left quadrant
of SS1,2,3,4 which all co-occurwith given
arethe parallelsubcategories
Gr aswell.In thelowerleftthereisonlySSo,non-parallel,
status:necessarilyall
butherewe canhaveGiven r,2,3,oR 4. Thecrucialcategory,whichpermits
comparisonof the two effects,is Gr. We can examinecasesof Gt wtTH
parallelstructure(theupperhalf of the rectangle)or wnHour (thelowerhalf).
Wethereforeran the variablerule analysisdividing Gl into the two subtypes
shownhere,G(P) and G(- P). within a fixed given status,we can compare
theeffectof parallelstructure.Table 8 showsthe result.First we noticethat
theeffectof parallel structureis sizeable:Given r with parallelstructureis
0.54,and Given r non-parallelis only o.36,at aboutthe levelof thenot-given
category.
Figurer alsoshowsthat a secondclosecomparisoncanbe made:with this
of two kinds of GI, we can makea bettercomparisonof the
differentiation
effectof crvrN status,that is, the entirelowerleft vs. the lowerright quadrant.
Table9 showsthis result.Given r(P) is about the same,o.56.The crucial
comparisonis between Go and the combined category G(- P), which
of G(- P) with Go givesus
includes
Gl(- P) and G2,3,4. Thecomparison
5r
E.JUDITH WEINER AND WILLIAM LABOV
Tokens
Passive
(%)
runs
Factorsmergedin successive
Full
analysis Style
VARBRUL
weights
Class
I
Runno.
o.74
Inputpo
6t6
o.52
33
Givenstatus
o.40
o New
25
o.44
o.39
534
I In preceding
o.58
z znd Preceding
o.56
o.67
360
58
3 3rd Preceding
o.53
4+ 4th Preceding
Table9
SS
Parallel
o.2r
Effectof givenin non-parallelsentences
o Not Parallel
vs. new on choiceof the passive
o.45
Preceding
I
o.63
z Preceding
o.67
3 Preceding
a strictmeasureof the effectof given,and it is in the expected
direction:o.5z
o.56
4+ Preceding
vs.o.44.However,thisis amongthe smallerof theeffectswe havefound,and
Passive
it doesnot supportthe notion that givenvs. newis a powerfuldeterminant Preceding
o.69
Yes
passive
of the choiceof activevs.
agentless
clauses.
o.3I
No
The converseconclusionis that the ordering of surfacesyntax across
'
Style
clausesis the predominantlinguisticinfluenceon this choice.This result
o.54
Careful
reinforcesour earlierdecisionto treatthe choiceof activeand passiveastwo
o.46
Casual
alternativewaysof saying'the samething', sinceit is conditionedby formal,
Sex
syntacticfactorsfar morethantheinfluenceof givenvs.new.Onemightargue
o.50
Male
that parallelstringsarealsocharacteristic
of semantically
significantchoices:
o.50
Female
thereis a cognitivelydeterminedtendencyto keeptalking about the same
Age
-o2
-o3
thing.
o.46
'mixed'
Adolescents
Parallelsurfacestructureis a
category,involvingcoreference
as
*oz
+o3
o.54
Adults
well as surfaceorder. One way of approachingthis questionis to consider
Class/EthnicitY
-03
-04
with
, what would be theeffeetof a purelysyntacticconditionwhichintersects
o.46
White MC
parallel
structure:
whet
surface
+
f
o.58
White WC
+
lbry..d.db" " p4si".
o.45
BlackWC
un
n tne Daslc olvlslons
r42
I84
r48
I38
rzg
zoo
No. of cells
of givenvs. new and parallelstructure.Theseremainunchanged
exceptfor
z
.
r
9.7
o.oo
4.8
.
5.6
Chi square
a fiewhundredths.But the precedingpassiveprovesto be an independent
o.ot
o.o5
.
o.o2
.
p<
and powerfulconditioningfactor: only tz6 caseshad a precedingpassive
(ard thisincludesa precedingp.*iu
-, Factor group eliminated'
.
percentage
passives:
per
this
But
subcategory
showsthe highest
of
cent.
7z
*, Factor mergedwith factor above'
by f/too'
variable
precedigrg
passive
rule
weight
The
contributedby a
is o.69,and the
+i, -f, Greateror lessthan Run t
than o' l '
more
by
t
Run
from
gain in likelihoodproducedby adding this factor to our analysisis very
., Doesnot differ
likelihood')
:
-z(log
large: chi-square 55. Since a precedingpassivecould occur with any
likelihoodr-log
X' :
of the combinationsof given/newor parallel/non-parallel
SS it is underTable ro
standablethat adding this variable factor does not disturb the earlier analyses.
rt--O
full analysisand effectsof
VARBRUL probabilitiesfor the passive:
/
r,/L
52
mergersof externalfactors
successive
alq-
G(- P)
All given non-parallel
Go
New
Gt(P)
Given parallel
W
V-Y
-*[;
Age
LX.
tfvlea*S
*
6?'^/
so.1Ll.
E. JUDITH
WEINER AND WILLIAM
LABOV
The fact that it is a powerful factor reinforcesour conclusionthat the
passiveis conditionedby syntacticconsiderations.
choiceof agentless
Thesefindingscoincide
with otherrecentstudiesthatindicatethelimitations
structure.Linde(tgZ+)did
of the informationfactorin determiningsentence
not find that givenvs. new wasa significantfactorin determiningthe choice
of dummy sentences
in apartmentlay-outs,e.g. Thereis a livingroomon the
right vs.A livingroomis on theright.Poplack(t98o)hascarriedout a massive
variablerule analysisof the informationalconstraintson thedeletionof final
(s)in PuertoRicanSpanish.Although the plural inflection/s/ is most likely
to be found in the first elementof a noun phrase,its presence
or absence
on
succeeding
elementsis determinedby a sequentialeffectsimilarto the effect
of a precedingpassivein this study./s/ tendsto follow lsl, and zero tends
to follow zero. An informational analysiswould have predictedthat a
preceding/s/ would favour the presence
of a followingzero.
The given/newdistinctionhas recentlybeenre-examined
by Prince,with
an eyeto the structureactuallyusedin discourse(rgZg).Sheprovidesa finer
subcategorization
of the possibilities,
with a three-waydivision of rvoxno
(textually or situationally),TNFERRED,
and new (nneND-NEw
or uNusEn).
BRAND-NEw
is in turn categorized
asanchoredor unanchored,
dependingon
whetherthe NP is linked to the rest of the discoursethroughanotherNP,
as with a relativeclause.It seemslikely that the effectof given/newshown
in our resultswould be sharpenedif the data werere-analysedwith a factor
group that registered
thesedistinctions.
Silva-Carvalan's
recentvariablerule analysisof the postposingof subjects
in Mexican-American Spanishshows no effect of given vs. new Ogll\,
However,shereportsa powerful effectof given vs. new on the presence
or
absence
of pronounsubjects.Givenstatusfavouredthe deletionof a pronoun
with a weightingof o.69,ascomparedto o.66for the effectof parallelsubject.
It appearsthat givenvs. new may be a more powerfuldeterminantof the
presence
or absenceof an elementthan the orderingof sentence
constituents.
It is alsopossiblethat the effectwill be strongerif it is exertedon the choice
of placementof two noun phrasesin two-argumentsentences,
ratherthana
singlenoun phrasevs. verb in one-argument
constructions.ro
An extension
of the presentanalysisto the contrastof activeand passivein two-argument
sentences,
with agentsexpressed,
may throw more light on the statusof the
informational component in determiningthe structure of sentences
in
spontaneous
speech.
4 . T H T I N D E P E N D E N C EO F I N T E R N A L A N D E X T E R N A L C O N S T R A I N T S
PASSIVE
CONSTRAINTS ON THE AGENTLESS
runs
Factors mergedin successive
Run no.
Inputpo
Given status
o New
t In preceding
z znd Preceding
3 3rd preceding
4 4th* Preceding
ParallelSS
o Not Parallel
t Preceding
z Preceding
3 Preceding
4+ Preceding
PrecedingPassive
Yes
No
Style
Careful
Casual
Sex
Male
Female
Age
Adolescents
Adults
Class/ethnicitY
White MC
White WC
Black WC
No. cells
Chi square
p<
89lo
-tI
-o4
o.74
+02
+02
+o2
-03
+
rr
-o2
12
-O7
13
-It
14
-23
.
.
.
.
r5
-13
/ ?.2?
o,s,l
++
++
.
o.2l
o.45
o.63
o.67
o.56
-o8
+o5 *o6
+o5 + 15
+o3 +
O'6cD
o .s o
+06 +o4 +o3 +a7 + 13
-o2
+o3 +06 *zt
- o3 +o7 +
-o2
-o2++
-o2
-03
+oz
-o2
o.69
o . 3r
o.54
o.46
o.50
o.50
o.46
o.54
o.46
o.58
o.45
200
, is
164 r y1
o.9
LI
I o9
25
t8
o.oot o.ooI
r 9r
o.9
r78
o.9
::+o3
156 r22
Io
o.o2
I48
55
72
o.oot o.@I
-, Factor grouP eliminated.
+, Factor mergedwith factor above'
+i, -i, Greateror lessthan Run t by i/too'
., bott not differ from Run t by more than o'ot'
-log likelihoodr).
x2, -z(log likelihoodt
Table t t
successive
IARBRUL analysisof the passive:full analysisand effectsof
factors
mergersof internal
Table to also shows the resultsof a systematictesting of each factor for
significanceby examiningdifferencesin log likelihood when factors are
[t6J A point madeby SusumuKuno in the discussionof the oral presentationof this paper.
54
Pre.
-pass
r7
o.4o
o.39
o.58
o.6o
o'53
ParallelSS
Given/new
Full
analysis
55
.
removed.In addition to the full analysis,five columnsshow the resultsof
eliminating,one at a time,the externalvariables:style,sex,age,
successively
race,and socialclass.In thesecolumns,a dashin eachmemberof a factor
group indicatesthat for that run the entirefactor group hasbeenremoved.
The symbol + indicatesthat a particularfactor has beenmergedwith the
one listeddirectlyaboveit. When one of the other factorsin a run showa
from theoriginalRun t of morethano.ot, that difference
difference
is shown;
otherwise,the spaceis left blank.
It is immediatelyapparentthat there are no significantchangesin the
internal flactorsas the result of eliminatingan externalfactor: all of the
perturbationsoccur in the classand age factor groups.Table r r shows
thecorresponding
eliminationsof internalfactors.Herethereis onechangein
an externalfactor: the weightfor black working classadolescents
increases
by o.o3whenthe parallelstructuregroupis removedentirely.All of the other
changesoccurin the internalfactors.Thereare sizeableshifltsin the weights
of the given/newgroup when parallelstructurefactorsare eliminatedand
viceversa.
Thisresultindicates
a substantialindependence
ofthetwo setsofconstraints.
All sectionsof the populationappearto treatthe passive/active
choicein the
sameway, and conversely,the sameconstraintsare found throughoutthe
speechcommunity.This independence
of externaland internalconstraintsis
anempiricalresultthatis quitedifferentfrom theassumption
of independence
of environmentalconstraintswithin the Cedergren/Sankoffprogramme,
whereit functionssimplyasa null hypothesis(Sankoff& Labov, r979).The
separationof the two setsof constraintsconfirmsotherindicationsthat social
factorsoperateprimarily upon surfacepatternsrather than abstractsyntactic
alternatives.Table 5 showedconsiderablesocialdifferentiationof the choice
of thesurfaceformativeget vs.be,incontrastto themoreabstractalternation
passive.
activeand agentless
of generalized
This multivariateanalysisof the passive/active
alternationhasdisengaged
severalintersectingforces"thatstronglydeterminethe choiceof one or the
otherof thesewaysof sayingthe samething.The distributionof information
in discourseis not without influence,but it is a relativelyminor factor
comparedto themoremechanicaltendency
parallelstructure:first
to preserve
in the succession
of passiveconstructions,
secondin the retentionof the same
structuralposition for the samereferentin successive
sentences.
Thereis
undoubtedlya stylisticfactor operating,which would appearmore strongly
if we hsd includedmore formal speechand written materials.All of these
conditionson the selectionof activevs. passiveare generalfeaturesof the
Englishlanguage,
usedin muchthesamewayby theverydifferentsub-sections
communitiesthat we studied.
of the speech
56
I
I
i-_
C t i i i . ; s I R A I N T SO N T H E A G E N T L E S SP A S S M
E . J U D T T HW E I N E R A N D W I L L I A M L A B O V
REFERENCES
Bach,E. (rg6l). Have and6e in Englishsyntax.Lg 43:4!2-411'
(eds'),Studiesin
n.ftin, f"f. (rqiZl. Quantifier-negatiieinteiaction.In R' Fasoldand R' Shuy
variation.Washington,D'C.: GeorgetownU'P'- language
B. (rgzr). class,codesantl control.New York: schocken.
Berns-tein-,
development:
form andfunctionin emerginggrammars'Cambridge:
Bloom,I-. (rqZo).Languag:e
MIT Press.
Brown,R. (rgzl). Afirst language.cambridge: Harvard university Press.
J. & Miller,
l.iiqiSl.A realistiitrinsformationalgrammar.In Halle, M', Bresnan,
Bresnan,
t-59.
Press.
MIT
cambridge:
reality.
psychological
and
theory
G. (eds.), Linguistic
("lgZl). The interpiay of socialand linguisticfactorsin Panama.Unpublished
Cedergren,'H.
CornellUniversitydissertation.
Chicago:Universityof ChicagoPress.
Chafe,W. (r97o). Meaningand thestructureof language.
Lg 5o' I I l-t33'
Chafe,W. (rgZ+).Languageand consciousness.
The Hague:Mouton'
N. (rqSZ).Syntacticstructures.
Chomsky,
Chomsky,N. iriOSl. itprttt of the theoryof syntax.Cambridge:MIT Press.
Feagin,b. trqZgi. Variationand changein Alabama English.Washington,D.C.: Georgetown
UniversityPress.
of laryycge.Ney York: McGraw-Hill'
foOor,J.,Blver,T. & Garrett,M. (I974). Thepsychology
A. K. (1967).Noteson themeand transitilly in English,Part 2. JL 3. t99-244.
HaitiAay,'M.
Hu*o6i, F. (rqSg).Quantitativestudyof the speechof Australianchildren.L&S z. 4Gz7l
JChL 5.65-8o.
of the full passive.
Horgan,D. (lgz8). The development
p. (rg68).The interpretationof the passivevoice.QuarterlyJ. of Experimental
Johrison-LaiiO,
Psych.2o.69-73.
Cambridge:M.l.T.
Kati,J.& postaf,P.(rg6+). An integratedtheoryof linguisticdescriptions.
Press.
grammars:the acquisitionof Germansyntaxby
Klein,W. & Dittmar, N.(rqZg). Developing
foreignworkers.New York: Springer-Verlag.
in NewYorkCity.Washington.D.C.: Center
srroiitrroiionof English
(rq66).The social
t aUou,"W.
for Applied Linguistics.
cpula.Ls lS'
Labov,W. (rqeqal. Contraction,deletion,and inherentvariabilityof the English
7t5-762.
La'bJv,'W.(rg6qb). The logic of non-standardEnglish.GURT 22. r-44.
factl Lisse:Peterde Ridder Press.
-sociolinguistic
Labov,W. iriZSl. Whatis a linguistic
stop? .n t ..potlt to Beatriz
Labov, W.'(i9i'8). Where doJs the
.variable
Austin: SouthwestEducationalDevelopment
Lavandera.Workingpapersin Sociolinguisficl.
Laboratory.
(tg8z)'
.Labov,W., Bower,A., Dayton, E., Hindle, D., Kroch, O', t4nn!S, M' & Scl{[ri1'.O'
U'S'
Philadelphia:
to NSF on SOC75-ooz45.
of soindchange.FinalReport
determinants
Social
RegionalSurveY'
of
Laboi, W., Cohen,P., Robins,C. & Lewis,J. (1968).A studyol the non-standardEnglish
in NewYork City.CooperativeResearchReport3288.Vols'
andpuerto RicanSpeai?,ers
Negro
I and It. Philadelphia:U.S. RegionalSurvey'
and-naturallogic. Ann Arbor: Universityof Michigan.
Lakofr,G. (r97o). Linguisrics
Lakoff,R. (r97r). Passiveresistance'PCLS7' r49-16z'
and their meaning.Lg 52,789-839.
l.ng".Lrt,'R.V. & Munro, P.(rgZS).Passives
Wheredoes-thi sociolinguisticvariable.stop?Zd'S1' zt5--48'
Lavindera,B. (rqZ8).
'sociat
and iducation.London: Routledge,KeganPaul.
class,language
t.r"ton, O. trg'661.
of spatial information. UnpublishedColumbia
linguisii"
Linde, C. tiglq\.itre
"-n.o-aing
Universitydissertation.
S. (1982).Adverbsand logicalform: a linguisticallyrealistictheory.Lg 5t.
McConnell-Ginet,
r4i-r84.
constraints
popiack,
i. (rggo). The notion of the plural in Puerto Rican Spanish:c_o.mryting
'oi
timeandspuce.New York: Academic
ttjdelition.'lnW. Lubou(ed.\, Licatinglanguagein
Press.55-68.
Prince,E. (tqZg).On the given/newdistinction' PCLS t5'
57
E.JUDITH WEINER AND WILLIAM LABOV
Sankoff,D. & Labov,W. (t979).On the usesof variablerules.LrS t. tSgzzz.
in Detroit.Final Report,
Shuy,R., Wolfram,W. & Riley,W. (t966).A studyof socialdialects
Project6-ry+l.Washington,D.C.: Officeof Education.
Silva-Carvalan,
C. (rgltl.A quantitativestudy of subjectdeletionand subjectplacement
in
spokenSpanish.Papergivenat the Winter Meetingof the LSA, Chicago.
Tannenbaum,
P. & Williams,F. (tg68).Generationof activeandpassive
sentences
asa function
of subjector object focus.JVLVB 7. z4Fz5o.
Trudgill,P. (t972).Sex,covertprestigeand linguisticchangein urban British English.li^St.
t79-t95.
R. (tg68).Focusof attentionin recallof activeand passive
sentences,
Turner,E. & Rommetveit,
Jt/LVB 7. 543-548.
Van den Broek, J.(rglll. Classdifferen.eiin syntacticcomplexityin the Ftemishtown of
Maaseik.LtS 6. r49-r82.
Weinreich,U. (t958). Travelsthroughsemanticspace.Reviewof Osgood,C. E., et al., The
nreusurenrcnt
of meaning.Word 14. 374-379.
J. Linguistics19 (r983), 59-78.Printedin Great Britain
The nature of phonological representation:
evidencefrom breaking in Frisianr
PETER TIERSMA
Miami University,Oxford, Ohio
(Received
7 April r98r)
I. INTRoDUCTIoN
A thorough examination of the nature of the phonological portion of the
lexiconmust addressat least two crucial issues.One has traditionally been
termedthe lssrnAcrNEss coNTRovERsy,which concernsitself with the issue
of the level of representation, or whether lexical entries are at a surf,ace
phonetic,phonemic, systematicphonemic, or some other level. The second
issuedeals with the unit of representation,or whether lexical entries are in
the form of morphemes, words, or some of each.
The earlier work in Generative theory (most elegantly formulated for
phonologyin chomsky & Halle, 1968)appearsto presuppose
a rather
abstract,
rnorpheme-based
lexicon.Later contributionsby Chomsky(rgzo)
andAronoff (t976) havearguedthat certaincombinationsof morphemes
shouldbe takenup in the lexiconas units.In addition,variousproposalsby
Kiparskyhaveconstrainedsomewhatthe level of abstractness
allowedin
phonological
representations.
Other phonologistshavegonemuch further,
proposing
what might becalleda concrete,word-based
modelof the lexicon.
Argumentsfor sucha positionmay befound in Vennemanne974)and Leben
& Robinson(t977).Most currentviewslie somewhere
betweenthe abstract
morpheme-based
lexicon and the concreteword-basedmodel. In what
follows, the question of how phonologicalinformation is stored will be
addressed
on the basisof data from the complexmorphophonemicprocess
of breakingin Frisian.
The term 'breaking' has had rather broad applicationin the literatureon
Frisian.In one sensethis refersto the breakingof i and e to iu beforecertain
velar consonantcombinations in Old Frisian. Thus proto-Frisian recht
becameriucht, and singabecamesiunga.This type of breaking,which has
parallels
in Englishand Nordic dialects,will not furtherconcernus here.
Anothertype of breaking,which occurredonly in mostmainlanddialects
of WestFrisian,is of morerecentorigin.Thisprocess,
whichpresumably
took
my gratitudeto JoanBybee,JeffreyElman,AnthonieFeitsma,Orrin
It] I would liketo express
Robinson,SanfordSchane,Wim Zonneveldand othersfor commentson a previousdraft
of this work (aspart of Tiersma,I98ob).My specialthanksgo to Matthe* ihen, who has
beena constantsourceof inspirationto me. None of the abovebearsany responsibility
for the contentsof this paperor necessarily
agreeswith the pointsmade.
58
59
,ii
Download