Memory & Cognition1983, Vol. 11(1),64-76

Intonational marking of given and newinformation: Some consequences for

comprehension

J. KATHRYN BOCK and JOANNE R. MAZZELLAMlchigan StateUniversity, EastLansing, Michigan 48824

The role of intonation in conveying discourse relationships in auditory sentence compre­hension was investigated in two experiments. Using the simple comprehension time paradigm,Experiment 1 found that sentences with accented new information were understood faster thansentences with a neutral intonation contour and that the presence of accent in contextsentences facilitated comprehension of subsequent targets. Both experiments showed fastercomprehension times in conditions in which accent placement was appropriate for the infor­mation structure of the sentence. In Experiment 1, comprehension times were faster when theaccent fell on the information focus than when it fell elsewhere in the sentence. In Experi­ment 2, faster times resulted when new information was accented and given information wasnot, compared to conditions in which this accent pattern was reversed. This effect held forboth active and passive sentences, and whether the new information occurred in the subjector object position.

In contrast with the semantie, syntactic, and lexicalaspects of language, effects of the prosodie features ofutterances on comprehension have received relativelylittle attention. An utterance's intonation can, however,have a profound impact on its interpretation: Given theproper stress pattern, a sentence such as "That was athrilling experience" can mean that the experience wasquite the opposite of thrilling (Cutler, 1976). Intonationalso carries a heavy burden in conveying the informa­tion structure of an utterance, marking different cornpo­nents as given or new (Chafe, 1974;Cutler, 1976;Cutler& Isard, 1980). Appropriate intonation may thereforefacilitate the hearer's understanding of the relationshipbetween a particular sentence and its discourse context.This paper examines certain effects on comprehensionof variations in the stress patterns accompanying alterna­tive arrangements of given and new information withinsentences.

The intonation of sentences reflects a variety ofinfluences (Cutler & Isard, 1980). Some of these arewithin-sentence factors, including the syntax and thedistribution of stressed syllables across the words of the

This research was partially supported by funds from theCollege Scholar Program of the College of Social Science atMichigan State University. Experiment 1 formed a portion of athesis submitted by the second author to Michigan State Univer­sity in partial fulfillment of requirements for the MA in psy­chology. We thank Gordon Wood and Rose Zacks for extensiveadvice on the experiments, David lrwin and DavidEhresrnan forassistance in programming, data analysis, and preparation ofstimulus materials, and Leonard Silvennan for running subjectsin Experiment 2. Requests for reprints should be sent to KathrynBock, Department of Psychology, University of Pennsylvania,3815 Walnut Street, Philadelphia, Pennsylvania 19104.

64

sentence. The prosodie structure that resuIts appears tomake important contributions to the comprehensibilityof individual sentences (Dooling, 1974; Geers, 1978;O'Connell, Turner, & Onuska, 1968; Shields,McHugh, &Martin, 1974; Weener, 1971; Wingfield & Klein, 1971).

Intonation is also substantially affected by a varietyof intersentential or discourse relationships (Chafe,1976; Gunter, 1966). In particular, the placement of theprimary ernphasis or the accent (Bolinger, 1961) withina sentence often seerns to be determined by the locationof the new information that the sentence contains:Information that is novel in the context of prior dis­course is accented, whereas old, given, or backgroundinformation is not. Although there are other ways inwhich given and new information can be marked, intona­tion may be the principal means for indieating thedistinction in English. Compared to certain other lan­guages, English relies more on stress patterns and lesson devices such as word order for information structurecues (MacWhinney & Bates, 1978).

Some evidence for the importance of intonation inconveying information structure comes from the obser­vation that children acquiring English appear to employintonation appropriately very early to mark new orsalient information. Wieman (1976), in a sample of fivechildren between the ages of 21 and 29 months, foundthat when their two-word utterances departed from cer­tain canonical stress patterns, the deviations were attrib­utable to the accenting of the ward in the utterancethat carried new information. In more controlled tasks,children as young as 3 years display adult levels of com­petence in using stress to mark new information (Hornby& Hass, 1970; MacWhinney & Bates, 1978). Stress alsoseems to be children's preferred way of marking infor-

Copyright 1983 Psychonomic Society, Inc.

INTONATIONAL MARKING OF GIVEN AND NEW INFORMATION 65

mation structure (Hornby, 1971; MacWhinney & Bates,1978), although Hornby (1971) observed a decline in itsuse with age for the first- through fifth-grade children inhis study.

In adult 1anguage use, there are a few observationsrelevant to assessing the role of this system in discoursecomprehension. Hornby (1972) presented subjects witha spoken sentence that inaccurately described twopictures and required the subjects to choose which ofthe pictures the sentence was intended to describe.For example, the sentence "The INDIAN is building theig1oo" was accompanied by a picture of an Eskimobuilding an igloo and one of an Indian building a tepee.Subjects tended to select the picture in which theaccented word (INDIAN in the example) mismatchedthe picture, suggesting that the subjects viewed speakersas more like1y to correctly designate given or presup­posed information than new or asserted information. A1ater experiment (Hornby, 1974) showed that listenersperform as though this were true in the immediateverification of sentences as weil. Subjects heard eleft andpseudocleft sentences (e.g., "It's the boy that the girl iskicking"; "The one the girl is kicking is the boy"), fo1­lowed by a 50 msec exposure to a picture that mis­matched either the agent or the object of the action ofthe sentence. Subjects made significantly fewer errorswhen the picture mismatched the clefted or pseudo­clefted element. Since such sentences are naturallyspoken with the accent within the elefted ("It's the BOYthat the girl is kicking") or pseudoelefted ("The one thegirl is kicking is the BOY") noun phrase (see the findingsof Allen and O'Shaughnessy cited by Cutler, 1976),this effect may be partially attributable to a tendencyto attend to and verify the emphasized constituentfirst. Although similar effects are obtained with visualpresentation of sentences (Carpenter & Just , 1977a;Langford & Holmes, 1979), implying that this maynot be the entire explanation, such an interpretationis supported by the results of other experiments thatexplicitly varied accent in a sentence-picture verifi­cation task (Hornsby, Carr, & Bock, Note 1).

The effects of given and new information on under­standing 1anguagehave been examined most frequently ininvestigations of sentence comprehension in discoursecontexts. Many of these studies were motivated by Clarkand Havi1and's (1977) sketch of the procedures requiredto interpret contextualized sentences, in which thegiven-new distinction plays a central role. These proce­dures, collective1y called the given-new strategy, specifythree steps that must be carried out for a sentence to beunderstood: (1) The given and new infonnation mustbe identified, (2) any given information must be relatedto its antecedent, and (3) the new information mustbe incorporated into memory. Problems in performingany of these operations should disrupt comprehension.

The second step of the given-new strategy has beenthe most thoroughly examined of the three, with a num­ber of experiments showing that a sentence is morereadily understood when the immediately preceding

text provides appropriate antecedents for its given infor­mation (Clark & Sengul, 1979; Garrod & Sanford, 1977;Haviland & Clark, 1974; Yekovich & Walker, 1978;Yekovich, Walker, & Blackman, 1979). However, somesupport for the first step, identifying the given and newinformation in the sentence, comes from experimentsthat have shown facilitation of comprehension byelear,accurate marking of what is given and new. Thus, eleftedand pseudoclefted noun phrases are elearly marked asnew information; if this marking is appropriate, com­prehension is easier than if it is not (Carpenter & Just,1977b). Likewise, givenness tends to be associated withthe surface subject position of sentences, whereas newinformation more often occurs in object position(Halliday, 1967), and this arrangement also appears tobe more comprehensible than the opposite ordering(Yekovich et al., 1979).

Although such syntactic devices as clefting, pseudo­elefting, and given-new ordering seem to be effectiveindicators of given and new information, the develop­mental evidence reviewed earlier suggests that they maybe subsidiary to intonational marking, particularly inspoken English. Since the sentences in the experimentson comprehension in discourse contexts were presentedvisually and read silently by the subjects, the effective­ness of intonation as a marker for information structureand its influence on comprehension are unclear. Ifintonation is a fundamental system for conveying infor­mation structure in English, it should be possible tofacilitate auditory sentence comprehension simply byaccenting new information or to disrupt it by accentinggiven information. The two experiments reported belowexamined several variations of this hypothesis, using thesimple comprehension time paradigm employed in pre­vious reading comprehension experiments. The partici­pants listened to pairs of sentences in which the initialsentence (the context sentence) established some mini­mal background information, and the second sentence,the target, was related to the context in some way.Participants indicated when they feit they had under­stood each senten ce by pressing a key (Clark & Sengul,1979; Haviland & Clark, 1974; Yekovich & Walker,1978; Yekovich et al., 1979).

Both experiments explored the basic question offacilitation vs. disruption of comprehension by appro­priate vs. inappropriate use of intonation to markinformation structure. In addition, Experiment 1examined the effect of presence vs. absence of accentin the context sentence on comprehension of the target,and the effect of omitting accent in the target sentence.Experiment 2 varied the voice of the target sentence(active vs. passive) and the position of the accent in thetarget sentence (surface subject vs. surface object).

EXPERIMENT 1

What is given or new in a sentence depends on therelationship of the information in the sentence tomaterial in prior utterances or text. These relationships

66 BOCKAND MAZZELLA

may be established by the presuppositions of the pre­vious discourse. A particularly effective way to createthese presuppositions is with a prior "wh-" question(Chomsky, 1971; van Dijk, 1977). For example, the newinformation in a simple sentence such as " Mary went tothe store" is different when preceded by the question''Where did Mary go?" than when preceded by ''Whowent to the store?" In production tasks, the differ­ences in information structure created by these kinds ofvariations in prior questions result in different syntacticarrangements of the components of the same basicproposition (Bock, 1977; Bock & Irwin, 1980). Cutlerand Fodor (1979) found that comprehension wasaffected when prior questions were used to manipulatesentence focus (the location of the asserted or mostinformative material). They preceded a sentence such as"The man on the corner was wearing the blue hat"either with the question ''Which man was wearing thehat?", so that the word "corner" was the focus, or withthe question ''Which hat was the man wearing?", sothat the word "blue" was the focus. Cutler and Fodorexamined phoneme-monitoring reaction times to initialphonemes of such words both when they were and whenthey were not focused, using the same recording of eachsentence in all conditions. They found faster reactiontimes when the word bearing the target phoneme wasfocused than when it was not.

Experiment 1 employed a related method of establish­ing discourse presuppositions, one that relied on varyingthe intonation pattern of a context sentence. Thus, thequestion ''Who fixed the radio?" and a sentence such as"ARNOLD didn't fix the radio" both establish the pre­supposition that someone flxed the radio. This presup­position can be changed by varying the location of theaccent, as well as by omitting it: "Amold didn't FIX theradio" presupposes that Arnold did something else tothe radio, whereas a relatively neutral rendition of thesentence conveys neither of these presuppositions,although it is compatible with both states of affairs. Theaccent delineates what the focal information in subse­quent discourse should be, the actual radio repairer, orwhat Amold really did to the radio.

Variations in intonation like these were used in thecontext sentences in Experiment 1, coupled with targetsentences whose stress patterns marked focal informa­tion appropriately or inappropriately. Thus, in "DORISfixed the radio," the accent falls appropriately on focalinformation when the sentence ispreceded by "ARNOLD

didn't fIX the radio," but when preceded by "Arnolddidn't FIX the radio," the accent falls inappropriatelyon nonfocal information. If intonation effectively indi­cates the information focus in sentences, appropriateaccent should produce faster comprehension times thaninappropriate accent. When the stress pattern of the con­text sentence is neutral, with no clear accent, compre­hension times for the targets should be intermediate,since in these cases additional processing may be requiredto identify the discourse presuppositions or antecedents,although the accent in the target is appropriate.

Since it would be possible for inappropriate accent todisrupt comprehension without appropriate accentnecessarily facilitating it, Experiment 1 also included acontrol condition in which the intonation patterns ofboth the context and the target sentences were neutral.If information structure is important in comprehension,and intonation is an effective marker for it, this conditionshould result in slower comprehension times than thecondition in which the neutral context sentence is fol­lowed by an accented target.

MethodParticipants. The participants were 40 native speakers of

English who served in the experiment for extra credit in anintroductory psychology course at Michigan State University.

Materials. The auditory stimulus materials were constructedfrom 20 sets of four sentence pairs each, examples of which arepresented in Table I. All 20 sets are listed in Appendix A. Thefour pairs in each set were identical except for differences in thelocation of the accent or prirnary stress within the two sentences.Each pair of sentences consisted of a context sentence and arelated target sentence. The context sentence in all pairs was asimple negative declarative sentence. The structure of the targetsentence was parallel to that of the context, except that it waspositive and replaced a content word from the context sentencewith a different word.

Four types of accent relations were represented by the sen­tence pairs in each set. In the fust type (appropriate targetaccent; see Table 1), the content word in the context sentencethat was most prominently stressed corresponded to the changedand accented word in the target. In the second type (inappro­priate target accent), the same target sentence waspreceded by acontext sentence in which a different word was accented,thereby creating apresupposition for which the target stresspattern was inappropriate, In the third type (no context accent),none of the words in the context sentence received specialemphasis. These three pairs employed identical target sentences.The fourth type, a control pair, omitted accent in both thecontext and the target.

The word accented in the target sentence was either thesubject, verb, object, adjective, or object of the preposition ofthe sentence. Equal numbers of sentence sets instantiated eachof these stress assignments.

Condition

Appropriate target accentInappropriate target accentNo context accentControl

Table IA Sentence Set from Experiment I

Context Sentences

ARNOLD didn't fix the radio.Arnold didn't FIX the radio.Arnold didn't fIX the radio.Arnold didn't fIX the radio.

Target Sentences

DORIS fixed the radio.DORIS fixed the radio.DORIS fixed the radio.Doris fixed the radio.

INTONATIONAL MARKING OF GIVEN AND NEW INFORMATION 67

-------------_._-- ----

"Equivalcnt to 14 msec before the end of the target sentence.

partieipant respondcd. When a responsc was made before asenteneo was eomplcted, however, the tape continued.

The recognition mcrnory test was administered after thepresentation list was cornpleted. Participants were asked toindieate whether each sentenee was one that had just beenheard by writing "yes" or "no" after each sentenee on thetest, Participants werc not informed of the rnernory test at thebeginning of the experiment.

All of the participants in Experiment I performed a relatedexperimental task du ring the same session. Half completed therclated task before Experiment I, and half following Experi­ment I.

Design. The fuur experimental eonditions represent fourlevels of the aeeent faetor (appropriate target aeeent, inappro­priate target aeeent, no eontext aeeent, and eontro\). Eaeh of the40 partieipants heard a different set of five iterns (sentenee pairs)in cach of these four eonditions. Different participants heard thefour different versions of eaeh item corresponding to the fourexperimental eonditions.

ResultsThe comprehension times used in the analyses were

measured relative to the ends of the sentences in orderthat the times would not reflect the durations of thesentences. Times for responses that occurred prior to theends of sentences were treated as negative; a responsethat occurred 15 msec before the end of a sentence thusyielded, for purposes of the analyses, a comprehensiontime of -15 msec.

The mean comprehension times for target sentencesin the four conditions of the experiment are shown inTable 2. The mean target comprehension times for eachparticipant and item in each condition were submittedto analyses of variance that treated both participants anditems as random factors (Clark, 1973; Coleman, 1979).The alpha level for significant effects, unless otherwiseindicated, was set at .05.

Thc overall analysis showed that the four condi­tions differed significantly: with participants random[F(3,ll4) =9.02], with items random [F(3,57) =6.38],and with both random [min F'(3,l33) = 3.74]. Plannedcomparisons revealed significant differences betweenthe appropriate target accent condition and each of theother three conditions (the halfwidths of the 95% confi­dence intervals were 54 msec in the analysis with parti­cipants random, and 67 msec in the analysis with itemsrandorn). The no context accent condition differed sig­nificantly from the control condition in both analyses.The 50-msec difference between the inappropriatetarget accent condition and the no context accent condi­tion did not quite reach significance in either analysis,

Table 2Experiment I : Mean Target Sentence Comprehension

Times (in MiUiseconds)

-14*75

125143

Compre­hension TimeCondition

Appropriate Target AeeentNo Context AeeentInappropriatc Target AecentCuntrol

Four presentation Iists of 20 sentenee pairs eaeh wcrc con­strueted. Every list contained five sentenee pairs representingeach of the four possible eonditions, with no list containingmore than one sentence pair from any sentenee set. Aeross alllists, eaeh of the 80 possible sentencc pairs was used only onee.The same random order of experimental eonditions was used forall presentation lists, with sentences having similar strcss assign­ments (subjeet, verb, etc.), but from different sets, appcaringin the same positions aeross lists. Three additional sen tenee pairs,similar in form to the experimental pairs, were addcd to thebeginning of eaeh list to serve as praetiee items. The same prac­tice pairs were used for all presentation lists.

Each unique sentenee in the eomplete stimulus sct wasreeorded onee by a female speaker on audiotape using aRevox A-700 tape recorder and an AKG D120E mierophone.Within eaeh set of sentenee pairs, the unique sentences eonsistedof three versions of the eontext senteneo (eaeh with a differentstress pattern) and two versions of the target senteneo (one withand one without aeeent). The stress patterns of thc rccordcdsentences were eheeked by a seeond person to insure that theloeation of the aeeent was eorreet in the aeeented sentcncesand that no words were prominently stressed in thc neutral,unaeeented sentences. All sentences whose intonation did notclearly eonform to the intended patterns were rerecorded.

These reeordings were used to eonstruet the four differenteontext/target eombinations. The reeorded sentenccs weredigitized and stored on magnetic dise using a PDP 11/40 corn­puter interfaeed with analog-tc-digital and digital-to-analog con­verters (Model DAC-16, 3 Rivers Computer Corporation). Theanalog-to-digital converter sampled the auditory input at a rateof 10,000 times/see; at this rate, the analog eonversions of thedigitized stimuli were indistinguishable, under normal listeningeonditions, from the original recordings. The beginnings andendings of the digitized records of the sentences were edited byvisual inspection of a graphie display of eomponents of theauditory waveform for each sentence, accompanied by auditoryplayback of the analog eonversion of the sentence , to ensurerelatively uniform onset and offset of all sentences.

The four presentation lists were constructed by playing sen­tences back in the order predetermined for each list and record­ing them on one channel of an audiotape. Recording was con­trolled by a program that placed a timing tone at the editedbeginning and end of eaeh sentence on another tape ehannel.The program inserted a 5-sec silent interval between sentencepairs and a l-sec interval between the two sentences in eaehpair. The purpose of this spacing was to emphasize the relation­ship between the paired sentences.

For each presentation list, a yes-no recognition test was con­strueted that included all of the experimental target sentencesfrom the list plus an equal number of distraetor sentences simi­lar in strueture to the targets, but employing different words.The 40 sentences for eaeh test were typed in random order ona single page, with eaeh sentence fol1owed by a blank for theparticipant to indieate his or her response.

Procedure. Participants were tested individually. TIICY wereseated in front of a response panel and listened to the binaural1yprcscnted stimuli through headphones (Koss Pro/4AA). Theywere instrueted to listen earefully to each sentence and to pressthe rightmost lever on the panel as soon as they understoodwhat the sentenee meant. They were not explicitly told thatthe sentences were related to each other or that they oeeurredin pairs. Al1 23 sentence pairs (3 praetiee plus 20 experimentalpairs) in a single presentation list were presented withou t abreak.

Stimulus presentation and response monitoring and tim­ing were control1ed by a PDP 8/i computer interfaeed witha TEAC A3340S tape recorder. Sinee it was possible for parti­eipants to respond before the end of a sentence, compre­hension times were measured from both onset and offset ofthe sentences. If no response had occurred by the time a sen­tencc cnded , the tape stopped and did not restart until thc

68 BOCKAND MAZZELLA

although it was marginal (p< .10) in the participants'analysis. The 18-msec difference between the controland inappropriate target accent conditions was not sig­nificant.

Because heightened stress is correlated with increasedduration (Lehiste, 1970), the lengths of the accentedand unaccented sentences were measured and compared.The mean duration of accented sentences (2,125 msec)was significantly greater than that of unaccented sen­tences (1,861 msec) [F(1 ,99) =12.8]. The comprehen­sion times for accented vs. unaccented context sentenceswere analyzed to investigate possible effects of thisdifference on comprehension. However, the compre­hension times for accented context sentences, with amean of 261 msec, did not differ significantly fromthose for unaccented context sentences, with a mean of281 msec (all Fs < 1).

The order in which participants served in Experi­ment 1, relative to thc related experiment, was includedas a factor in the overall analysis. There was a trendtoward faster comprehension times when Experiment 1was second, but the interaction between the order factorand the accent factor was not significant in any of theanalyses.

Recognition test performance was 90% correct over­all, indicating that participants were attending to thesentences while performing the task.

DiscussionThese results demonstrate that intonation, as a device

for conveying the information structure of sentences,makes a substantial contribution to their comprehen­sion. The comprehension times for target sentences weresignificantly shorter when new, focal information wasaccented (appropriate target accent) than when nonfocalinformation was accented (inappropriate target accent),suggesting that appropriate intonational marking of newinformation facilitated comprehension. This conclusionis strengthened by the contrast between the no contextaccent condition and the control condition. The onlydifference between these two conditions was in thepresence of accent within the target sentence; compre­hension times were significantly shorter when the tar­get was accented. These findings complement andextend earlier observations (Carpenter & Just, 1977b;Yekovich et al., 1979) that the syntactic marking ofgiven and new information influences the comprehen­sion of written sentences. In the current experiment,only intonation served to distinguish new information.Nonetheless, in line with claims for the role of intona­tion in marking information structure (Chafe, 1974;Gunter, 1966), its appropriate use speeded the compre­hension of simple active sentences.

The significant difference between the appropriatetarget accent condition and the no context accentcondition shows that the specification of a discoursepresupposition by the context sentence's accent patternalso contributed to comprehension of the target sen-

tence. This finding is in line with the results of Yekovichet a1. (1979), who found that the placement ofinforma­tion in context sentences influenced target cornprehen­sion: Target comprehension times were faster whenpresupposed information occurred in subject positionand when focal information occurred in object positionwithin context sentences than when this order wasreversed. Since there is a general relationship betweenpresupposed vs. focal information and subject vs. objectposition, this placement appears to have aided identifica­tion of the context's information structure, which inturn supported target comprehension. In the presentexperiment, a context sentence such as "SYLVIA didn'tpet the monkey" specified two pieces of informationnecessary for the understanding of the subsequent con­trastive sentence (Chafe, 1976): A particular unfilledrole (the one who petted the monkey) and certainbackground information (that the monkey had in factbeen petted). In comparison, the unaccented contextsentence "Sylvia didn't pet the monkey" is ambiguouswith respect to which role is unfilled and what informa­tion is presupposed. In terms of the given-new strategy(Clark & Haviland, 1977), the accented version of thesentence may have promoted comprehension of afollowing target sentence, for example, "DAVE pettedthe monkey," in either of two ways: Its presuppositionprovided a clear antecedent for the given information inthe target sentence, and its specification of an unfilledrole supplied an explicit memory address for the newinformation.

The regular difference in duration between accentedand unaccented sentences dictates that any conc1usionsbased on this contrast be treated cautiously. So, althoughcomprehension times for targets were shorter when tar­gets were accented than when they were not, and whencontexts were appropriately accented than when theywere unaccented, the presence of accent is naturally con­founded with greater length. As a result, the fasterresponses for accented targets may be due in part to theavailability of more time for processing. Although theabsence of a significant effect of accent on the compre­hension times for context sentences makes a simpleprocessing time explanation somewhat less probable, the20·msec difference between the two was neverthelessin the direction predicted by such a hypothesis.

EXPERIMENT 2

In addition to intonation, there are various syntacticand lexical devices that are used to distinguish givenfrom new information. One frequently discussed syntac­tic marker is placement of given information in thesurface subject and new Information in the predicate ofthe sentence (Chafe, 1970; Firbas, 1966;Halliday, 1967).In visual sentence processing, Yekovich et al. (1979)have shown that such an arrangement results in fastercomprehension times than the opposite ordering.

Very little is known about the relationship between

INTONATIONAL MARKING OF GIVEN AND NEW INFORMATION 69

such syntactic deviees and intonation. There do seem tobe grass associations between syntaetic and intonationalindieators of inforrnation strueture: The accent in eleftand pseudocleft sentences naturally falls on the cleftedand pseudoclefted noun phrases (Allen & O'Shaughnessy,eited by Cutler, 1976), which have been taken to repre­sent the information foeus (Homby, 1974), and primarystress in the so-called neutral intonation contour faUsat the end of the sentenee (Cutler & Isard, 1980), theeanonical position for the information focus. Such corre­lations may be attributable to the intluence of intona­tional preferenees on syntax, with syntax being adjustedto conform to prosodie patterns. Alternatively, intona­tion and syntax may constitute independent, eomple­mentary systems for marking inforrnation structure.

Experiment 2 was designed to explore the contribu­tions of intonational and syntaetic markers of informa­tion strueture to sentence comprehension. Using thesimple comprehension time paradigm, accent and sub­jeet vs. objeet placement of given and new infonnationwere varied factoriaUy in target sentences, with the stresspattern of the context sentences contralIed . The com­plementary systems hypothesis suggests that aeeent andlocation of given and new infonnation should makeindependent eontributions to comprehension time,whereas the intonational primacy hypothesis, in itsstrongest form, predicts that intonation alone shouldaffeet comprehension.

Both aetive and passive target sentences were usedin Experiment 2. Sinee passive sentences seem to markinformation structure more strongly than simple actives(Anisfeld & Klenbort, 1973; Hornby, 1974), the appro­priate positioning of given and new infonnation maymake a greater eontribution to the comprehension ofpassives than actives. However, if the effects of subjectvs. objeet placement are general, they should not differaeross the two sentence types.

MethodParticipants. Forty-eight undergraduates at Michigan State

University served in the experiment as an extra-credit option inan introductory psychology course. All were native Englishspeakers, and none had participated in the previous experiment.

Materials. The auditory stimulus materials were constructedfrorn 24 sets of eight pairs of context and target sentences. Oneof these sets is presented in Table 3, and all of the sets are listedin Appendix B. Each set was composed of all possible pairs oftwo different context sentences preceding four different targetsentences. The context sentences were simple actives that werethe same except for their surface subject and surface object

noun phrases. Insofar as possible, no obvious stress was placedon any part of the context sentences, although they were spokenas naturally as possible within this constraint. The four targetsentences consisted of two versions of both the active and pas­sive variants of a single simple sentence. Once version accentedthe surface subject, and the other version accented the surfaceobject. The surface subject of one of the context sentencesmatched the surface subject of the active target and the surfaceobject of the passive target. The surface object of a secondcontex t sentence matched the surface object of the activetarget and the surface subject of the passive target.

The eight experimental conditions that resulted from pairingthe two context sentences with the four versions of the targetwere defined by (I) whether the target was active or passive,(2) whether the surface subject or surface object constituentin the target was accented, and (3) whether the subject andobject constituents of the target positioned the given and newinformation appropriately, given the accent pattern (e.g., whenthe subject was accented, new information was located in thesubject), or positioned it inappropriately , given the accent pat­tern (e.g., when the subject was accented, new information waslocated in the object).

Two different presuppositional relationships, each used in12 sentence sets, were employed in writing the sentence pairs.Each relationship was distinctively marked in the target sen­tence by an adverb, "too" or "either" (see Haviland & Clark,1974, Experiment 3, for a discussion of the presuppositionsassociatcd with these adverbs). All of the sentences in "too"pairs were positive, and all those in "either" pairs were negative.

In addition to the two context sentences used in the experi­mental sentence pairs, an unrelated context sentence was pairedwith each of the four target sentences in each sentence set. Forexarnple, the context sentence "Lila kissed Michael" was pairedwith each of the target sentences in the set in Table 3. The pur­pose of these unrelated context sentences was to discourageparticipants from attempting to predict the object of the verbin the target sentences. In the experimental sentence pairs,whenever the subject of the target sentence differed from boththe subject and object of the context sentence, an active targetsentence's object matched the object of the context sentence,and the object in a passive target sentence matched the contextsentence's subject. A pilot experiment suggested the possibilitythat this predictability would influence participants' responses.

Four presentation lists were constructed, each composed ofthree blocks of 24 items, for a total of 72 items in every list.Each list contained 48 experimental sentence pairs, 6 from eachof the eight experimental conditions, and 24 sentence pairs inwhich the context and target were unre1ated. Within Iists, threesentence pairs from each sentence set were employed, represent­ing the possible com binations of one version of the target sen­tence with three different context sentences (including theunre1ated context), Each of these was assigned to a differentblock. Across lists, any given sentence pair was used only once.The same random order of items was used for all presentationlists, so that sentence pairs from the same set with the same con­text sentence appeared in similar positions in their respectivelists. At least one and no rnore than four items in each conditionappeared in each block. Four practice items similar in form to

Condition

Active /S ubject/AppropriateActive/Subject/InappropriateActive/Object/AppropriateActive/Object/InappropriatePassive/Subject/AppropriatePassive/Subject/InappropriatePassive/Object/AppropriatePassive/Object/Inappropriate

Table 3A Sentence Set from Experiment 2

Context Sentences

Evelyn kissed J eremy.Rhonda kissed Jason.Rhonda kissed Jason.Evelyn kissed Jeremy.Rhonda kissed Jason.Evelyn kissed Jeremy.Evelyn kissed Jererny.Rhonda kissed Jason.

Target Sentences

RHONDA kissed Jeremy, too.RHONDA kissed Jererny, too.Rhonda kissed JEREMY, too.Rhonda kissed JEREMY, too.JEREMY was kissed by Rhonda, too.JEREMY was kissed by Rhonda, too.Jererny was kissed by RHONDA, too.Jererny was kissed by RHONDA, too.

70 BOCKAND MAZZELLA

the experimental sentence pairs were placed at the beginning ofthe first block in each list. The same practice items were used forall four presentation lists.

The unique sentences in each set of sentence pairs (the threecontext sentences and the four target sentence versions) were pre­pared and assembled into lists using the procedures described inExperiment 1.

To check accent placement in the recorded sentences, fournaive raters listened to the four tapes containing each presenta­tion list and indicated on a written transcript where the heavieststress fell in each context and target sentence. Each tape wasevaluated by one rater. For the target sentences, each of whichoccurred three times within a given list, the percentages of agree­ment between the raters' accent assignments and the intendedassignments were 100%, 100%, 98.6%, and 93.1 %. Three of thefive disagreements that produced the lowest percentage occurredfor only one of the three occurrences of three different sen­tences, so that in each case the rater agreed with the intendedassignments on two of the three occurrences. The remainingtwo disagreements occurred on the same sentence, although therater agreed with the intended assignment on the third occur­rence. For the context sentences, which were the same on allfour tapes, there was a trend toward indicating the subject as thelocus of heaviest stress, with the raters marking it on 62% of thesentences overall. The object, at 30%, was next most frequent,and different components of the verb and auxiliary made upthe remaining 8%. There were large individual differences inthese evaluations, however: For the four raters, the percentagesof subject selection were 100%,35%, 71%, and 40%. The stresspatterns of the context sentences were thus relatively neutralin comparison to the targets.

For each presentation list, a yes-no recognition test was con­structed that included a representative half of the target sen­tences and 12 distractor sentences, which used the verbs fromthose target sentences not included in the test, but with new sub­ject and object noun phrases, The syntax of the distractors wassimilar to that of the targets. The 24 sentences for each test weretyped in random order on a single page, with each sentence fol­lowed by a blank for the participant to indicate his or herresponse.

Procedure. The procedure followed was the same as thatdescribed for Experiment 1, except that participants weregiven a short break between blocks.

Design. Each of the 48 subjects contributed data to the eightcells of the factorial formed by crossing the two levels of thetarget sentence type factor (active vs. passive), the two levels ofthe stress factor (surface subject vs. surface object), and the twolevels of the location factor (appropriate vs. inappropriate). Eachsubject received three versions of each item, each in a differentexperimental or unrelated context condition, and versions ofsix different items in every condition.

ResultsThe comprehension times used in the analyses, as in

the previous experiment, were measured relative to theends of the sentences. Mean comprehension times forthe target sentences were computed for each participantand item in every condition. The 61 comprehensiontimes exceeding 3,000 msec were discarded (10,13, and38 in the appropriate, inappropriate, and unrelated con­ditions, respectively, or 1.8% of all responses). The over­all condition means, inc1uding the unre1ated contextconditions, are shown in Table 4.

The data were submitted to analyses of variance treat­ing both participants and items as random effects; thelevel required for significance was set at .05 unless other­wise noted. These analyses exc1uded the data from theunrelated context conditions.

Table 4Experiment 2: Mean Target Sentence Comprehension

Times (in Milliseconds)

Placement*Accent

Voice Placement A U

Active Subject 207 221 406Active Object 187 232 356Passive Subject 239 326 402Passive Object 198 276 364

Note-A =appropriate, 1= inappropriate, U =unrelated."Placement ofaccent relative to given and new information.

The analyses showed that the difference between theappropriate and inappropriate conditions, with fastercomprehension times in the appropriate condition, wassignificant in all analyses [F(l ,47) =5.78 for partici­pants; F(1,23) =26.63 for items; min F'(1,64) =4.75].The inter action between accent placement and appro­priateness, which tests the hypothesis that syntac­tic marking of information structure influences com­prehension, was not significant in either analysis[F(l ,47) < 1 for participants and F( 1,23) < 1 foritems]. The three-way interaction between voice,accent placement, and appropriateness was also notsignificant [F(l ,47) < 1 for participants and F(l ,23) < 1for items] , indicating that there was no reliable differ­ence between passives and actives in the importance ofthe syntactic marking of information structure. Subjectvs. object placement had no overall effect (all Fs< 1).

The main effect of voice was significant in both theparticipants and the items analyses [F(1,47) =5.96 andF(1 ,23) =5.35], reflecting generally faster cornprehen­sion times for active than for passive sentences. Thiseffect was, however, marginal in the joint analysis[rnin F'(1 ,59) =2.82, P < .10]. The interaction betweenvoice and appropriateness was significant in the analysiswith items random [F(l,23) =5.16], although not inthe analysis with participants randorn [F(1,47) = 2.64],or in the joint analysis [min F'(1,70) =1.75]. Theinteraction in the iterns analysis is attributable to the1arger difference between passives and actives in theinappropriate than in the appropriate condition: The75-msec difference between inappropriate passives andactives was weIl beyond the 34-msec halfwidth of the95% confidence interval for items, whereas the differ­ence between appropriate passives and actives was anonsignificant 21 msec.

Changes in performance over the three blocks of theexperiment were evaluated in an analysis of variance thatincluded block as a factor. The means are given inTable 5. The analysis was perfarmed only with partici­pants as a randorn effect, because each itern did notoccur in every condition in every block. There was asignificant main effect of block [F(2,94) =27.03] dueto decreases in comprehension times in successiveblocks: The mean comprehension tirnes in Blocks 1, 2,and 3 were 344, 202, and 134 msec, respectively. Themain effects of appropriateness [F(l ,47) = 8.81] and

INTONATIONALMARKING OF GIVEN AND NEW INFORMATION 71

Table 5Experiment 2: Mean Target Sentence Comprehension

Times Over Blocks 1,2, and 3 (in Milliseconds)

Placement*Accent -----_.

Voice Placement Block A-_._- ._._.._--

1 309 254Active Subject 2 158 239

3 172 114

1 293 375Active Object 2 212 161

3 130 159

1 368 424Passive Subject 2 116 275

3 60 200

1 345 381Passive Object 2 223 230

3 95 260

Notc-A = appropriate, 1 = inappropriate."Placement ofaccent relative to given and new information.

voice [F(1 ,47) =7.01] were again significant. The inter­action between appropriateness and voice was alsosignificant [F(1 ,47) = 5.31], suggesting that variabilityattributable to the block factor was responsible for thefailure of this interaction to reach significance in theprimary analysis with participants random.

Two three-way interactions involving block were sig­nificant. First, the interaction between block, appropri­ateness, and accent location was significant [F(2,94) ==

4.46]: The appropriateness effect wasreliable, with a 95%confidence interval halfwidth of 79 msec, for subject­accented sentences only in the second block, and forobject-accented sentences only in the third block. How­ever, the direction of the effect was consistently in favorof faster comprehension times in the appropriate condi­tion for both subject- and object-accented sentences inevery block except the second, in which there was a 22­msec reversal for object-accented sentences. Secend. theinteraction between block, accent location, and voicewas significant [F(2,94) =3.37]; subject-accented activeswere significantly faster than subject-accented passives inthe first block, whereas object-accented actives were sig­nificantly faster than object-accented passives in thethird block, with a 95% confidence interval halfwidth of66 msec. These interactions suggest some inconsistencyin performance over blocks, which is most probably theresult of the small number of trials in each conditionwithin each block. No systematic trends across blocks,other than the overall decrease in comprehension times,were apparent.

Examination of the means in the unrelated conditionrevealed a general tendency for sentences with objectaccent to be responded to more rapidly than sentenceswith subject accent: The mean for subject-accentedunrelated sentences was 404 msec, whereas for object­accented unrelated sentences it was 360 rnsec. This44-msec difference was tested and found to exceedthe 95% confidence intervals for both participants(40 msec) and items (35 rnsec). A similar effect appeared

in the appropriate and inappropriate conditions ina trend toward faster comprehension times for object­accented than for subject-accented passives, althoughthe interaction between accent placement and voicewas not significant [F(1,47) == 1.99 for participants andF(1 ,23) < 1 for items]. Object accent thus seems tohave been associated with faster comprehension timesin the more difficult conditions of the experiment.

The mean of the scores on the recognition test was93% correct.

DiscussionThese results show again that the appropriate use of

intonation to mark given and new information facilitatesthe comprehension of sentences. The data thereforereplicate the results of Experiment 1 regarding appro­priate vs. inappropriate use of intonation and extendthem to different locations within sentences and to pas­sivesas weil as actives.

The absence of an effect of placing given informa­tion in the subject and new information in the objectsuggests that the syntactic marking of informationstructure may play a much smaller role in auditorysentence comprehension than intonational marking.This suggestion is strengthened by the absence of adifference between passives and actives in the effect ofsyntactic marking, because such marking is more explicitin passives (Anisfeld & Klenbort, 1973; Hornby, 1974).Since previous experiments using the same paradigrn, butwith a reading rather than a Iistening task (Yekovichet al., 1979), have found that such syntactic cues pro­duce faster comprehension times, these results point toan important difference between visual and auditory sen­tence comprehension.

The effect of voice, with passive sentences requiringmore time to understand than actives, is consistent withthe extensive literature on the problems of processingthe more complex syntax of passives (e.g., Forster &Olbrei, 1973; Gough, 1965). However, the interactionbetween appropriateness and voice indicates that theseproblems can be offset by appropriate intonationalcuing of information structure. When such roles as agentand object are established in one sentence, and theseroles or their perpetrators serve as antecedents forexpressions in a subsequent sentence, clear markingof the relationships between the constituents of the twosentences may make it possible to determine the under­lying conceptual structure of the second sentence with­out doing a fuil syntactic analysis. The interaction maythus be similar to that found for reversible vs. non­reversible actives and passives (Slobin, 1966).

The faster comprehension times for object-accentedthan for subject-accented unrelated sentences, and thetrend toward such a difference for passive sentences,suggests that sentences with accent at the end may begenerally easier to understand than sentences in whichthe accent is closer to the beginning. The increase inthe duration of the final part of the sentence that wouldnormally accompany object stress provides additional

72 BOCK AND MAZZELLA

time to integrate the sentences and relate them to theircontexts, activities that seem generally to be carriedout at the ends of clauses or sentences (Aaronson &Scarborough, 1976; Just & Carpenter, 1978, 1980).This extra time might be expected to benefit primarilysentences for which comprehension is impeded by suchfactors as incoherent discourse relations (as for theunrelated sentences) or difficult syntactic structure(as for passives). Alternatively, the difference maysimply reflect a bias toward object-accented sentences:Object accent may be somewhat more natural thanaccent elsewhere in sentences (Cutler & Isard, 1980;however, see Bolinger, 1972, and Schmerling, 1974).

GENERAL DISCUSSION

The two experiments reported above demonstratethat the appropriate use of intonation to mark informa­tion structure facilitatcs the comprehension of sentences.In Experiment 1, accenting focal information producedfaster comprehension times than accenting nonfocalinformation. In Experiment 2, comprehension timeswere faster when new information was accented andgiven information was not, compared to comprehensiontimes for sentences in which given information wasaccented. This pattern of effects conforms to hypothesesderived from linguistic observations of the discoursecorrelates of variations in the intonation patterns ofsentences (Chafe, 1974, 1976; Cutler & Isard, 1980;Halliday, 1967): The information focus of a sentence,or the part that carries information new to the discourse,is also the part that is normally accented.

Experiment 1 suggested that the appropriate accent­ing of a sentence facilitates comprehension not onlyrelative to inappropriate accenting, but also relative tosentences with a neutral intonation contour. This wastrue both for context sentences and for target sentences,for different but related reasons: The presence of accentin context sentences served to c1early mark a discoursepresupposition necessary for the comprehension of sub­sequent targets, whereas appropriate intonation patternsin targets indicated the information focus. Comprehen­sion times for target sentences were faster when con­texts were appropriately accented than when they hadno accent, and when the targets themselves were appro­priately accented than when they were not. These find­ings, together with those for appropriate vs. inappro­priate accenting of the information in target sentences,are consistent with previous experiments that haveshown that both the availability of contextual antece­dents and the marking of information structure areimportant to the processing of sentences in connecteddiscourse (Carpenter & Just, 1977a, 1977b; Garrod &Sanford, 1977; Haviland & Clark, 1974; Yekovich &Walker, 1978; Yekovich et al., 1979).

A major problem in comparing effects for sentencescontaining accent with those that do not is that the per­ception of increased stress tends to be associated with anincrease in the duration of a sentence (Lehiste, 1970),

and this was true in Experiment 1. However, presenceof accent did not significantly reduce comprehensiontimes for the context sentences themselves, indicatingthat the combination of accent and greater durationdoes not reliably affect the time required to understandan isolated sentence. Obviously, however, furtherresearch will be required to determine the contributionof time to the effect of accent on comprehension.

Experiment 2 examined the effects of intonational vs.syntactie marking of information structure on compre­hension time. Given information tends to occur towardthe beginning of a sentence, and new informationtoward the end (Firbas, 1966), perhaps optimally, withgiven information in the subject and new information inthe object (Halliday, 1970). Consistent with such apattern, Yekovich et al. (1979) found that placing giveninformation in the surface subject, and new informationin the object, facilitated comprehension relative to theopposite arrangement. Experiment 2, however, found nosignificant effect of the location of given and new infor­mation within target sentences, for either active or pas­sive sentences: Comprehension times for both positionsdepended primarily on whether their intonation wasappropriate for the status (given or new) of the informa­tion they carried. It therefore appears that the positionof given and new information may play a much lessimportant role than their intonation in auditory sen­tence comprehension: Since spoken language has prosodiestress available as a means to mark information structure,listeners may rely less on positional cues. This is consis­tent with evidence that intonation is the primary indiea­tor of information structure in English (MacWhinney &Bates, 1978).

These results are particularly interesting in light ofspeculations about the effect on reading comprehensionof the absence or relative paucity of explicit cues tointended intonation in visual language (Olson, 1977;Schallert, Kleiman, & Rubin, Note 2). To compensate,location may perform a function in writing that it doesnot ordinarily serve in spoken language, or that it servesinadvertently as a byproduct of a bias toward placingthe accent in sentences near the end (Cutler & Isard,1980). Botinger (1957) and Schubiger (1964) have notedthe possibility of using certain syntactie structures tomaneuver new information to the ends of sentences ; pro­ficient writers apparently do this routinely (Smith,1971). Beginning readers may need to learn to useposition as a cue to information structure, as a partialreplacement for the intonational cues provided inspoken language.

The interpretation of the present experiments rests inpart on the assumption that contrastive information isprocessed similarly to new information, since the sen­tences employed were in many cases paradigmatie exam­ples of contrastiveness. There is a recurrent debateamong linguists conceming the nature of the distinctionbetween newness and contrastiveness. According to cri­teria proposed by Chafe (1976), what separates con­trastive from new information is that, in the case of

INTONATIONALMARKING OF GIVEN AND NEW INFORMATION 73

contrastiveness, the speaker believes the hearer to havc avery limited nurnber of eandidates in mind for a partieu­lar role. Thus, the word "artichokes" in a scntenee suehas "He ate some ARTICHOKES" would most probablybe new following a statement sueh as "Hetman atesomething for dinner he's never had before ," but eon­trastive following the statement "It wasn't anehoviesthat Herman ate." Moreover, Chafe maintains that eon­trastive aeeent diffcrs in prominenec from the aeeentthat aeeompanies new information. Bolinger (1961,1978), on the other hand, argues that the accent rnark­ing eontrastiveness does not differ from that markingnew information, and that the eontrastive funetion isnot qualitatively different from that of eonveying ncwinformation.

It seems clear that the diseourse presuppositions forsentenees viewed as being eontrastive are more stringentthan those for sentenees that simply earry new informa­tion; the issue appears to be whether this restrietivenessis only the end of a eontinuum or a unique status. Thesame is true for aeeent. The problem for the presentexperiments, given that the diseourse presuppositions andintonation patterns used were contrastive, is whether thelistener proeesses sueh information differently than newinformation.

In terms of the given-new strategy proposed by Clarkand Haviland (1977), it is reasonable to assume thatthe proeessing of new and eontrastive infonnation issimilar. The foeus of a eontrast in a sentenee is clearlythe most informative part; the remaining informationtends to be given. The given inforrnation must have aunique antecedent in mernory, eonsisting of the back­ground information (Chafe, 1976) or presuppositionestablished by the eontext. Finally, the eontrastiveinformation, like new information, must be integratedinto memory for comprehension to be eomplete.Although additional steps may be required in the ease ofa eontrast (e.g., as in a denial, it may be neeessary forcontrastive information to displace previously storedinformation as part of its integration into memory,ef. Clark & Clark, 1977), the basie operations should bethe same. The overall eonvergence between the eurrentresults and those of previous experiments on the proeess­ing of given and new information supports this claim.

In summary, the experiments reported here provideevidenee for a substantial eontribution of intonation toconveying information structure in auditory sentenceeomprehension. The appropriate use of aceent to markthe new information in a sentenee facilitates eomprehcn­sion relative to eonditions in which aeeent is absent orused inappropriately. As a eue to information strueturein listening, intonation also appears to playa greater rolethan the syntaetic device of surface position (subjeet vs.object). These effeets point to the fundamental impor.tance of intonation in conveying information structurein spoken English.

REFERENCE NOTES

I. Hornsby, M. E., Carr, T. H., & Bock, J. K. Prosodie

stress directs attention and can 't be ignored in verifylng descrip­tlons. Paper presented at the meeting of the Western PsychologicalAssociation, San Francisco, April 1978.

2. Schallert, D. L., Kleiman, G. M., & Rubin, A. D. Analysesofdifferenees between written and oral language (Tech. Rep. 29).Urbana-Champaign, Ill: University of Illinois, Center for theStudy of Reading, 1977.

REFERENCES

AARONSON, D., & SCARBOROUGH, H. S. Performance theoriesfor sentenee eoding: Some quantitative evidence. Journal ofExperimental Psychology: Human Perception and Performance,1976,2,56-70.

ANISFELD, M., & KLENBORT, I. On the funetions of strueturalparaphrase: The view from the passive voice, PsychologicalBulletin, 1973,79,117-126.

BOCK, J. K. The effeet of a pragmatic presupposition on syntac­tie strueture in question answering. Journal of Verbal Learningand Verbal Behavior, 1977,16,723-734.

BOCK, J. K., & IRWIN, D. E. Syntactic effeets of informationavailability in sentenee production. Journalof Verbal Learningand Verbal Behavior, 1980,19,467-484.

BOLINGER, D. Maneuvering for aeeent and position. CollegeComposition and Communication, 1957,8,234-238.

BOLINGER, D. Contrastive aeeent and eontrastive stress. Lan­guage, 1961,37,83-96.

BOLINGER, D. Aeeent is predietab1e (if you're a rnind-reader).Language, 1972,48, 633-644.

BOLINOER, D. Intonation across languages. In J. H. Greenberg(Ed.), Universals of human language (Vol. 2): Phonology.Stanford: Stanford University Press, 1978.

CARPENTER, P. A., & JUST, M. A. Integrative processes in eom­prehension. In D. LaBerge & S. J. Samuels (Eds.), Basic pro­cesses in reading: Perception and comprehension. Hillsdale,N.J: Erlbaum, 1977. (a)

CARPENTER, P. A., & JUST, M. A. Reading eomprehension aseyes see it. In M. Just & P. Carpenter (Eds.), Cognitive pro­cesses in comprehension. Hillsdale, N.J: Erlbaum, 1977. (b)

CHAFE, W. L. Meaning and the structure of language. Chicago:University of Chicago Press, 1970.

CHAFE, W. L. Language and eonsciousness. Language, 1974,50,111-133.

CHAFE, W. L. Givenness, eontrastiveness, definiteness, subjects,topics, and point of view. In C. N. Li (Ed.), Subject and topic.New York: Academic Press, 1976.

CHOMSKY, N. Deep strueture, surfaee structure, and semantieinterpretation, In D. D. Steinberg & L. A. Jakobovits (Eds.),Semanties: An interdisciplinary reader in philosophy, linguisticsand psychology. Cambridge: Cambridge University Press, 1971.

CLARK, H. H. The language-as-fixed-effect fallacy: A eritique oflanguage statistics in psyehological research. Journal of VerbalLearning and Verbal Behavior, 1973,12,335-359.

CLARK, H. H., & CLARK, E. V. Psychology and language.New York: Hareourt Braee Jovanovich, 1977.

CLARK, H. H., & HAVILAND, S. E. Comprehension and the given­new eontraet. In R. O. Freedle (Ed.), Discourse production andeomprehension (Vol. 1). Norwood, N.J: Ablex, 1977.

CLARK, H. H., & SENGUL, C. J. In seareh of referents fornouns and pronouns. Memory & Cognition, 1979, 7, 35-41.

COLEMAN, E. B. Generalization effeets vs. random effects: Isoh a souree of Type 1 or Type 2 error? Journal of VerbatLearning and Verbal Behavior, 1979,18, 243-256,

CUTLEIl, A. Beyond parsing and lexical look-up: An enricheddescription of auditory sentenee cornprehension. In R. J. Wales& E. C. T. Walker (Eds.), New approaches to ianguage mecha­nisms. Amsterdam: North-Holland, 1976.

CUTLER, A., & FODOR, J. A. Semantic foeus and sentence eom­prehension. Cognition, 1979,7,49-59.

CUTLER, A., & ISARD, S. D. The production of prosody. In B.Butterworth (Ed.), Language production (Vol. 1). London:Aeademie Press, 1980.

74 BOCK AND MAZZELLA

DOOLING, D. J. Rhythm and syntax in sentence perception,Journal 0/ Verbal Leamlng and Verbal Behavior, 1974, 13,2SS-264.

FIRBAS, J. On defining the theme in functional sentence analysis.Travaux Linguistiques de Prague, 1966,1, 267-280.

FORSTER, K. 1., & OLBREI, I. Semantic heuristics and syntacticanalysis. Cognition, 1973,2,319-347.

GARROD, S., & SANFORD, A. Interpreting anaphoric relations:The integration of semantic information while reading, Journal0/ VerbalLearning and Verbal Behavior, 1977,16,77-90.

GEERS, A. E. Intonation contour and syntactic structure as pre­dictors of apparent segmentation. Journal0/Experimental Psy­chology: HumanPerception andPerformance, 1978,4,273-283.

GOUGH, P. B. Grammatical transformations and speed of under­standing. Journal 0/ Verbal Learning and Verbal Behavior,1965,4,107-111.

GUNTER, R. On the placement of accent in dialogue: A featureof contextgrammar. Journal 0/ Linguistics, 1966, 2, IS9-179.

HALLlDAY, M. A. K. Notes on transitivity and theme in English,Part 2. Journal0/Linguistics, 1967,3,199-244.

HALLlDAY, M. A. K. Langnage structure and language function.In J. Lyons (Ed.), New horizons in linguistics. Baltimore:Penguin, 1970.

HAVILAND, S. E., & CLARK, H. H. What's new? Acquiringnew information as a process in comprehension. Journal 0/Verbal Learning and Verbal Behavior, 1974,13, S12-S21.

HORNBY, P. A. Surface structure and the topic-comment distinc­. tion: A developmental study, Child Development, 1971, 42,

1975-1988.HORNBY, P. A. The psychological subject and predicate, Cogni­

tivePsychology, 1972,3,632-642.HORNBY, P. A. Surface structure and presupposition, Journal0/

Verbal Learning and Verbal Behavior, 1974,13, S30-S38.HOUBY, P. A., " HAll, W: A. Use of contrutive stress by

prescbool chßdren. Jouf1Ul1 0/ $pe«h and H«Irln, R_rch,1970,13,39S-399.

JUST, M. A., & CARPENTER, P. A. Inference processes duringreading: Reflections from eye fixations. In J. W. Senders,D. F. Fisher, & R. A. Monty (Bds.), Eye movements and thehigherpsychological functions. Hillsdale, N.J: Erlbaum, 1978.

JUST, M. A., & CARPENTER, P. A. A theory of reading: Fromeye fixations to comprehension, Psychological Review, 1980,17, 329-3S4.

LANGFORD, J., & HOLMES, V. M. Syntactic presupposition insentencecomprehension, Cognition, 1979,7,363-383.

LEHISTa, I. Suprasegmentals. Cambridge, Mass: MIT Press,1970.

MACWHINNEY, B., & BATEs, E. Sentenrialdevicesfor conveyinggivenness and newness: A cross-cultural developmental study.Journal 0/ Verbal Learning and Verbal Behavior, 1978, 17,S39-SS8.

O'CONNELL, D. C., TURNER, E. A., & ONUSKA, L. A. Intonation,grammatical structure, and contextual association in immediaterecall. Journal0/ Verbal Learning and Verbal Behavlor, 1968,7,110-116.

OLSON, D. From utterances to text: The bias of language in speechand writing. HarvardEducational Review, 1977, 47, 2S7-281.

ScHMERLING, S. F. A re-examinationof normal stress,Language,1974,SO, 66-73.

ScHUBIGER, M. The interplay and cooperation of word order andintonation in Englisb. In D. Abercrombie, D. B. Fry, P. A. D.MacCarthy, N. C. Scott, & J. L. M. Trim (Eds.), In honour0/ Daniel Iones: Papers contrlbuted on the occasion 0/ hiseightieth birthday. London: Longmans, Green, 1964.

SHIELDS, J. L., McHuGH, A., & MARTIN, J. G. Reaction timeto phoneme targets as a function of rhythmic cues in con­tinuous speech,Journal0/Experimental Psychology, 1974,102,2So-2SS.

SLOBIN, D. I. Grammatical transformations and sentence compre­hensionin childhood and adulthood. Journal0/ Verbal Learn­ingand Verbal Behavior, 1966,5,219-227.

SMITH, C. Sentences in discourse. Journal 0/ Linguistics, 1971,7,213-23S.

VAN DIJK, T. A. Text and context:Explorations in the semanticsandpragmaticso/discourse.London: Longman, 1977.

WEENER, P. Language structure and the freerecallof verbal mes­sages by children, Developmental Psychology, 1971, 5,237-243.

WIEMAN, L. A. Stress patterns of early child language. Journalo/Child Language, 1976,3,283-286.

WINGFIELD, A., & KLEIN, J. F. Syntactic structure and acousticpattern in speech perception, Perception cl Psychophysics,1971,9,23-2S.

YEKOVICH, F. R., & WALKER, C. H. Identifying and usingreferents in sentence comprehension, Journal 0/ Verbal Leamingand Verbal Behavior, 1978,17, 26S-277.

YIIKOVICH, F. R., WALKII" C. H., & BLACKIIAN, H. 8. The roleof presupJ)OMd ud focallnformation In Intearatina sentencea.Journal 0/ Verbal Learning and Verbal Behavior, 1979, 18,S3S-S48.

Appendix AExperiment 1 Sentence Pairs

GeorgeA didn't kill Billl. Alfred killed Bill.LynnA didn't go to Brazili. Joan went to Brazil.AmoldA didn't fixI the radio. Doris fixed the radio.DennisA wasn't drinking warmI milk. Danwas drinking wann milk.Henry doesn't have an olderl sisterA. Henry has an older brother.Frank isn't growings a beardA. Frank is growing a moustache.Ray didn't play footballA in collegel. Ray played baseball in college.Jackl isn't reading a newspaperA. Jack is reading a magazine.lillI doesn't take dancingA lessons.Jill takes singing lessons.Marciaisn't wearingl a redA coat. Marciais wearing a blue coat.Betty didn't buy a newA car from Mikel' Betty bought a used car from Mike.The plasticA bottlei didn't break. The glass bottle broke.Harry doesn't livel near Chicago A. Harry livesnear Detroit.The hockeYI game isn't on TuesdaYA' The hockey gam~ is on Friday.The officerl wasn't in the Army A. The officer was in the Navy.John didn't get a jobl in Michigan A. John got a job in Wisconsin.Marvinl doesn't ownA the building. Marvinrents the building.Alice didn't buyA the bananal bread. Alicebaked the banana bread.Ralph didn't direetA the moviel' Ralph produced the movie.Jane didn't get marriedA in Junel. Jane got divorced in June.

Note-Italic type indicates possible accent positions. In context sentences, subseripted A and I mark aeeent loeations in the appro­priateand inappropriate conditions, respectively.

INTONATIONAL MARKING OF GIVEN AND NEW INFORMATION 75

Appendix BExperiment 2 Context and Target Sentences

Contexts

10 Evelyn kissed Jeremy.IS Rhonda kissed Jason.IV Lila kissed Michael.20 Carolyn planted daffodils.2S Jennifer planted marigolds.2U losephine planted tulips,30 Vanessa fed the ducks.3S Margaret fed the pigeons.3U Colleen fed the sparrows.40 Jack paid the carpenter.4S Mary paid the electrician.4U Denise paid the housekeeper.50 Jane slapped Edgar.5S Vivian slapped Carlos.5U Jill slapped Irving.60 The jeweler seils watches.6S The peddler seils cuffIinks.6U The grocer seils apples.70 Brenda hit the Cadillac.7S Mildred hit the stop sign.7U Geraldine hit the lamppost.80 Marilyn shot the burglar.8S Harvey shot the policeman.8U Darlene shot the fireman.90 Gertrude saw the airplane.9S Patrick saw the helicopter.9U Helen saw the sailboat.100 George petted the dog.lOS Harry petted the cat.10U Sylvia petted the monkey.110 Jim kieked the trashean.11S Bob kicked the footstool.11U Brian kicked the television.120 Emily wrecked the truck.12S Ronald wrecked the car.12U Linda wrecked the bicycle.130 Laura didn't bake the eookies.13S Hannah didn't bake the cake.13U Roxanne didn 't bake the bread.140 Barney didn't train the lion.14S Frank didn't train the tiger.14U Dave didn't train the walrus.150 Maxine didn't write the novel.15S Deborah didn't write the poem.15U Patricia didn't write the article.160 Sylvester didn 't build the house.l6S Reginald didn 't build the garage.16U Albert didn't build the tool shed.170 Ted didn't paint the mailbox.17S Jerry didn't paint the beneh.17U Charlie didn 't paint the fence.180 lessiea didn't recommend Walter.18S Janet didn't recommend Elizabeth.18U Allee didn't recommend Melvin.190 Louise didn 't kill Daniel.19S Phyllis didn't kill Howard.19U Susan didn't kill Herbert,200 Mabel didn't steal the canary.20S Agncs didn't steal thc parakeet.20U Oliver didn't steal the penguin.210 Virgil didn't compliment Andrea.21S Roscoe didn't compliment Gloria.21U Anthony didri't compliment Gina.220 Fred didn't play the trumpet.22S Carl didn 't play the clarinet.22U Mathew didn't play the tuba.

Targets

lA Rhonda kissed Jeremy too.IP Jeremy was kisscd by Rhonda too.

2A Jennifer plan ted daffodils too.2P Daffudils were plan ted by Jennifer too.

3A Margaret fed the ducks too.3P The ducks were fed by Margaret too.

4A Mary paid the carpenter too.4P The carpenter was paid by Mary too.

5A Vivian slapped Edgar too.5P Edgarwas slapped by Vivian too.

6A The peddler sells watches too.6P Watches are sold by the peddler too.

7A Mildred hit the Cadillac too.7P The Cadil/ae was hit by Mildred too.

8A Harvey shot the burglar too.8P The burglarwas shot by Harvey too.

9A Patrick saw the airplane too.9P The airplane was seen by Patrick too.

10A Harry petted the dog too.lOP The dog was petted by Harry too.

llA Bob kicked the trashcan too.11P The trashcan was kicked by Bob too.

l2A Ronald wrecked the truck too.l2P The truck was wrecked by Ronald too.

BA Hannah didn't bake the cookies either.13P The eookies weren't baked by Hannah either.

14A Frank didn 't train the !ion either.14P The !ion wasn't trained by Frank either.

15A Deborah didn 't write the novel either.15P The novel wasn't written by Deborah either.

16A Reginald didn 't build the house either.16P The house wasn 't built by Reginald either.

17A Jerry didn 't paint the mailbox either.17P The mailbox wasn 't painted by Jerry either.

18A Janet didn 't recommend Waltereither.18P Walter wasn 't recomrnended by Janet either.

19 A Phyllis didn 't kill Daniel either.19P Daniel wasn 't killed by Phyllis either.

20A Agnes didn 't steal the canary either.20P The canary wasn't stolen by Agnes either.

21 A Roscoe didn't compliment Andrea either.21P Andrea wasn't complimented by Roseoe either.

22A Carldidn't play the trumpet either.22P The trumpet wasn 't played by Carleither.

76 BOCK AND MAZZELLA

Contexts

230 Molly didn't he1pBill.23S Joan didn't he1pBruce.23U Karen didn't he1pMark.240 The principa1didn't fix the projector.24S The teacher didn't fix the camera.24U The coach didn't fix the net.

Targets

23A Joan didn't he1pBill either.23P Bill wasn't he1pedby Joan either.

24A The teacher didn't fix the proiector either.24P The projector wasn't fixed by the teacher either.

Note-Contexts: 0 = matches active object and passive subject; S = matches active subject and passive object; U =unrelated. Targets:A = active; P = passive. Italic type indicates possible accent position. In subject-accented conditions, the head noun 0/ the subjectnoun phrase was accented. In object-accented conditions, the head noun of the object noun phrase was accented.

(Received for publication November 24, 1981;revision accepted Ju1y22, 1982.)