Sentence Comprehension in Adolescents With Down Syndrome and Typically Developing Children: Role of Sentence Voice, Visual Context, and Auditory-Verbal Short-Term Memory

172 Journal of Speech, Language, and Hearing Research • Vol. 48 • 172–188 • February 2005 Journal of Speech, Language, and Hearing Research • Vol. 48 • 172–188 • February 2005 • ©American Speech-Language-Hearing Association1092-4388/05/4801-0172

Giuliana MioloUniversity of Wisconsin—Whitewater

and Waisman Center,University of Wisconsin—Madison

Robin S. ChapmanHeidi A. Sindberg

Waisman Center,University of Wisconsin—Madison

Sentence Comprehension inAdolescents With Down Syndromeand Typically Developing Children:Role of Sentence Voice, VisualContext, and Auditory-VerbalShort-Term Memory

The authors evaluated the roles of auditory-verbal short-term memory, visualshort-term memory, and group membership in predicting language comprehen-sion, as measured by an experimental sentence comprehension task (SCT) andthe Test for Auditory Comprehension of Language—Third Edition (TACL–3; E.Carrow-Woolfolk, 1999) in 38 participants: 19 with Down syndrome (DS), age12 to 21 years, and 19 typically developing (TD) children, age 3 to 5 years,matched on syntax comprehension, as measured by TACL–3 Subtests II and III. Ofthe 5 dependent measures of comprehension, auditory-verbal short-term memoryaccounted for significant amounts of variance in 4; group membership, 1(semantic role assignment); and visual short-term memory, 0. In the group withDS, hearing status predicted variation in Grammatical Morphemes (TACL–3Subtest II). Using the SCT, the authors also investigated the effects of varyingsentence voice and supporting visual context on sentence comprehension. SCTperformance was significantly poorer in terms of (a) referent selection andsemantic role assignment, for passive (vs. active) sentences in both groups, and(b) semantic role assignment in all sentences for the group with DS (vs. the TDgroup). Vocabulary strengths in the group with DS were found with the PeabodyPicture Vocabulary Test—Third Edition (L. M. Dunn & L. M. Dunn, 1997) but notthe TACL–3 Vocabulary subtest.

KEY WORDS: language comprehension, Down syndrome, memory, preschoolchildren, language functions and disorders

Behavioral Phenotype of Cognitionand Language in Down Syndrome

Down syndrome, arising from trisomy of chromosome 21 and thesubsequent cascade of genetic effects and physical sequelae (Lott& McCoy, 1992), exerts a phenotypically distinct influence on a

child’s cognitive and language development (Abbeduto et al., 2001;Chapman & Hesketh, 2000; Miller, 1995). Down syndrome is typically

Miolo et al.: Sentence Comprehension 173

accompanied by mild to moderate delays in the develop-ment of nonverbal cognitive skills and additional defi-cits in short-term memory, with a greater deficit in theshort-term memory of auditory than visuospatial infor-mation (Chapman, 1995; Marcell, 1995).

Linguistic features of the behavioral profile includea disproportionate delay in language production rela-tive to nonverbal cognition and language comprehen-sion (Chapman, Hesketh, & Kistler, 2002; Chapman,Schwartz, & Kay-Raining Bird, 1991; Chapman, Seung,Schwartz, & Kay-Raining Bird, 1998; Dykens, Hodapp,& Evans, 1994; Kernan & Sabsay, 1996; Miller, 1995;Rosin, Swift, Bless & Vetter, 1988), with the delay inlanguage production appearing to affect syntax to agreater degree than vocabulary (Chapman et al., 1998).In contrast, syntactic comprehension has been reportedas commensurate with, or delayed relative to, nonver-bal cognitive levels, depending on the measures of cog-nition used (Chapman, Hesketh, & Kistler, 2002;Chapman, Schwartz, & Kay-Raining Bird, 1991; Rosinet al., 1988). Vocabulary comprehension is commensu-rate with—or, in adolescence, in advance of—nonverbalcognition (Chapman, Hesketh, & Kistler, 2002;Chapman, Schwartz, & Kay-Raining Bird, 1991; Rosinet al., 1988). Therefore, asynchronous development ofvocabulary and syntax documented within the domainof language production also appears to occur within lan-guage comprehension (Chapman et al., 1991; Miller,1988; Rosin et al., 1988).

Lexical–Syntactic Discrepancyin Language Comprehension

Vocabulary comprehension, as measured by thePeabody Picture Vocabulary Test—Revised (PPVT–R;Dunn & Dunn, 1981), is associated more strongly withchronological age than mental age in individuals withcognitive disabilities, including those with Down syn-drome (Facon, Facon-Bollengier, & Grubar, 2002; Facon,Grubar, & Gardez, 1998). Facon et al. (2002) showedthat for individuals with cognitive disabilities, vocabu-lary comprehension, as measured by the PPVT–R, wasassociated with chronological age even when the contri-bution of cognitive ability was controlled. They concludedthat vocabulary comprehension was linked to the par-ticipants’ life experience.

Mental age in individuals with cognitive disabili-ties is more strongly related to vocabulary comprehen-sion, as measured by the Boehm Test of Basic Concepts(BTBC; Boehm, 1971) than by the PPVT–R (Fazio,Johnston, & Brandl, 1993). Each test evaluates a differ-ent type of vocabulary knowledge. The BTBC evaluatesknowledge of abstract relational terms, such as between,

separated, and equal. In contrast, the PPVT–R evalu-ates object- or event-related vocabulary, selected andorganized hierarchically according to its frequency ofuse. For example, the word bed occurs early in the testand rhombus, toward the end. Fazio et al. suggested thatage-related experience, rather than cognitive ability,might influence object- and event-related vocabularyknowledge to a greater degree than abstract, concep-tual vocabulary knowledge. The stronger influence of atest taker’s age and consequent life experience on suc-cessful performance on the PPVT, compared with con-ceptually based vocabulary comprehension tests andmeasures of syntax comprehension, might explain thediscrepancy between lexical and syntactic comprehen-sion reported in individuals with Down syndrome. Totest this hypothesis, we compared performance, in termsof z scores, between (a) the Peabody Picture VocabularyTest—Third Edition (PPVT–3), a frequency-based mea-sure of object- and event-related vocabulary, and (b) themore conceptually based Vocabulary subtest of the Testfor Auditory Comprehension of Language—Third Edi-tion (TACL–3; Carrow-Woolfolk, 1999), in a group of ado-lescents and young adults with Down syndrome.

Factors Influencing LanguageComprehension in IndividualsWith Down SyndromeAuditory-Verbal Short-Term MemoryPrevious Research

Several researchers have suggested that specific dif-ferences in auditory-verbal short-term memory may ex-plain the characteristic expressive language profile ofindividuals with Down syndrome (Chapman, 1995; Hulme& Mackenzie, 1992; Jarrold & Baddeley, 1997; Jarrold,Baddeley, & Phillips, 2002; Kay-Raining Bird &Chapman, 1994; Laws, 1998; Seung & Chapman, 2000).Laws has reported significant relationships, in childrenand adolescents with Down syndrome, between auditory-verbal short-term memory, as measured by the nonwordrepetition task (Gathercole, Willis, Baddeley, & Emslie,1994), and (a) word span, (b) reading, and (c) languagecomprehension, as measured by the Test for Reception ofGrammar (TROG; Bishop, 1983).

Longitudinal research reported by Chapman et al.(2002) has shown that syntax comprehension in adoles-cents and young adults with Down syndrome, as mea-sured by the Test of the Auditory Comprehension ofLanguage—Revised (TACL–R; Carrow-Woolfolk, 1985),is best predicted by not only (a) auditory-verbal short-term memory, as measured by the Auditory Memorysubtest of the Illinois Test of Psycholinguistic Abilities(Kirk, McCarthy, & Kirk, 1968); but also (b) visual

174 Journal of Speech, Language, and Hearing Research • Vol. 48 • 172–188 • February 2005

short-term memory, as measured by the Bead Memorysubtest of the Stanford–Binet Intelligence Scale—FourthEdition (SB-BM; Thorndike, Hagen, & Sattler, 1986);and (c) chronological age.

Role Within Working Memory ModelUsing Baddeley and Hitch’s (1974) three-component

model of working memory, several researchers have sug-gested that the auditory-verbal short-term memory defi-cit in individuals with Down syndrome may be associ-ated with impaired functioning of the phonological loop,the component responsible for the short-term storageand processing of phonological information (Hulme &Mackenzie, 1992; Jarrold et al., 2002; Kay-Raining Bird& Chapman, 1994; Laws, 1998). The other componentsin the model are the visuospatial sketchpad (later namedthe visuospatial scratchpad), responsible for storage andprocessing of visual and spatial information; and a cen-tral executive component, which allocates resources. Inthe model’s current formulation, Baddeley (2001) hasadded another component, the episodic memory buffer,and has acknowledged the role of long-term memory inthe functioning of the phonological loop and thevisuospatial scratchpad.

Gathercole and Baddeley (1993) have hypothesizedthat the short-term storage function of the phonologicalloop serves an important role in the comprehension ofambiguous or syntactically complex sentences. Theyhave suggested that the phonological representation ofa complex sentence is stored temporarily in the phono-logical loop and is repeatedly referred to, during offlinelinguistic analysis of the sentence. If phonological loopfunctioning is impaired in individuals with Down syn-drome, their comprehension of more complex linguisticforms, such as passive sentences (which requires set-ting aside usual word-order-based comprehension strat-egies), might be compromised to a greater degree thanthat of their typically developing counterparts.

Understanding the effect of impaired phonologicalloop functioning on language comprehension, and theextent of this impairment in adolescents and youngadults with Down syndrome, has important clinical im-plications. Evidence for such a relationship would sug-gest that structures particularly vulnerable to auditory-verbal short-term memory limitations, such assyntactically complex and ambiguous sentences, wouldneed to be carefully evaluated and considered as poten-tial targets for intervention.

Visual Context andVisual Short-Term Memory

Research has shown that visual context (VC) ap-pears to support complex sentence comprehension in

adults without Down syndrome (Tanenhaus, Spivey-Knowlton, Eberhard, & Sedivy, 1995). Tanenhaus et al.monitored adults’ saccadic eye movements as they fol-lowed complex instructions to manipulate objects andfound that their eyes moved to objects in the visual dis-play just milliseconds after hearing the referential ex-pressions in the instructions. This finding led Tanenhauset al. (1995) to conclude that “visual context influencedspoken word recognition and mediated syntactic process-ing, even during the earliest moments of language pro-cessing” (p. 1632). A similar facilitatory effect of VC onlanguage comprehension in typically developing chil-dren and in individuals with Down syndrome is plau-sible, particularly for those aspects of comprehensionmore dependent on visual imagery, like referent selec-tion. The presentation of sentence referents in bothvisual and auditory modes may permit their simulta-neous representation in auditory-verbal and visualshort-term memory. In this situation, visual short-termmemory might support complex sentence comprehen-sion in a manner similar to auditory-verbal short-termmemory. Therefore, the role of visual short-term memoryin sentence comprehension is of interest, particularlygiven Chapman et al.’s (2002) finding that visual short-term memory, as measured by the SB-BM, was one ofthe best predictors of syntax comprehension, as mea-sured by the TACL–R, in their longitudinal study of lan-guage development in adolescents and young adults withDown syndrome.

Evaluating Predictorsof Language ComprehensionSelecting Measures and Matching Variables

We evaluated the claim that sentence comprehen-sion is potentially influenced by visual short-termmemory in all participants and disproportionately af-fected in individuals with Down syndrome because oftheir auditory-verbal short-term memory deficits, bydoing the following:

1. We measured auditory-verbal short-term memoryusing a number recall task and a nonword repeti-tion task. Each task measured the storage capacityof the phonological loop but varied in temporal du-ration of the stimuli used and the degree to which itwas influenced by prior phonotactic and lexicalknowledge.

2. We measured language comprehension using twotasks: a traditional picture-pointing test of syntaxcomprehension and an experimental sentence com-prehension task (SCT), designed to maximize theinvolvement of auditory-verbal and visual short-term memory.


3. We matched the group with Down syndrome to atypically developing control group on syntax com-prehension (as measured by TACL–3 Subtests II andIII, Grammatical Morphemes and ElaboratedPhrases and Sentences) rather than lexical compre-hension, mean length of utterance (MLU), or men-tal age. Matching on syntax comprehension consti-tuted a strong test of the claim that increasedauditory-verbal short-term memory demands in theSCT would lead to poorer performance in the groupwith Down syndrome.

Experimental SentenceComprehension Task

Response ModeTraditional tests of syntax comprehension (e.g.,

TACL–3 or TROG) require examinees to listen to a testword or sentence and then point to the correspondingpicture from a set of alternatives. Correct picture se-lection might be influenced, not only by a participant’sability to remember the test item but also by the possi-bility of a chance response (Kramer, Koff, & Fowles,1980), the plausibility of the depicted event (Bishop,1987), or the comprehension of one or few words orgrammatical morphemes in the test sentence. Further,the pointing response does not provide informationabout the participant’s sentence interpretation. Giventhese limitations, we decided to develop an object ma-nipulation task to evaluate sentence comprehension inthis study.

Sentence StimuliFor the SCT, participants listen to, then act out,

information-dense sentence stimuli, varying in sentencevoice (e.g., “The yellow cow chases the white chicken”and “The red horse is kicked by the white sheep”). Suchsentences have been used in comprehension measures,such as the Token Test for Children (DiSimoni, 1978)and an experimental task developed by Montgomery(2000) to evaluate the influence of verbal workingmemory on sentence comprehension in children withspecific language impairment.

We designed the test sentences so that correct se-lection of referents and their enactment requiredmemory of each sentence component, thus increasingthe involvement of auditory-verbal short-termmemory. To stress auditory-verbal short-term memoryeven further, we included passive-voice test sentences,for which correct semantic role assignment requiredmemory of the entire sentence, presumably storedtemporarily in the phonological loop during sentencereinterpretation.

Quantification of SentenceComprehension

We designed the task so that we could quantify par-ticipants’ responses in terms of their ability to accurately(a) enact the verb (verb enactment), (b) select the exactreferent (referent selection), and (c) assign the referentsto the role of actor or patient specified in the sentence(semantic role assignment).

Supporting Visual ContextWe presented visual referents in the SCT in two

conditions: the context-present condition, in which visualreferents were presented simultaneously with the testsentences, permitting representation of the referents invisual short-term memory; and the context-absent con-dition, in which the participant was able to see the ref-erents only after the auditory presentation of the testsentence.

Research Questionsand Predicted Outcomes

In this study, we asked four main questions andpredicted their outcomes as follows:

1. Would vocabulary comprehension performance (andconsequent intralinguistic profile) in the group withDown syndrome vary as a function of the vocabu-lary measure used? If age-related experience influ-enced object- and event-related, but not abstract–conceptual, vocabulary knowledge, then we wouldexpect the estimate of vocabulary comprehensionobtained using the PPVT–3 to be significantlygreater than that obtained from the Vocabularysubtest of TACL–3 for the chronologically oldergroup with Down syndrome, but not for the typi-cally developing group.

2. Would group membership (Down syndrome vs. typi-cally developing), visual short-term memory, andauditory-verbal short-term memory, as character-ized by performance on a nonsense word repetitiontask and a number recall task, account significantlyfor variation in each subcomponent (verb enactment,referent selection, and semantic role assignment)of the SCT, a high-information-load sentence com-prehension task? Given the hypothesized role of thephonological loop in complex sentence comprehen-sion, we would expect auditory-verbal short-termmemory to predict variation in semantic role assign-ment, which was more dependent on one’s ability tostore and process complex information, and refer-ent selection, but not verb enactment. In addition,if phonological loop functioning was impaired in


individuals with Down syndrome, then we wouldexpect group membership also to predict semanticrole assignment. Similarly, visual short-termmemory would predict a greater amount of varia-tion in referent selection (potentially facilitated bythe stored visual representation of the referents)than in semantic role assignment or verb enactment.

3. Do the same predictor variables account for varia-tion in performance on two subtests of the TACL–3,a traditional picture-pointing measure of syntaxcomprehension? Extending the logic from the SCTto the TACL–3, we would expect auditory-verbalshort-term memory to account for variation in thecomprehension of Grammatical Morphemes andElaborated Phrases and Sentences. However, be-cause the groups were matched on their averagedperformance on these subtests, we would not expectgroup membership to play a role in their variation.

4. How do group membership (Down syndrome vs. typi-cally developing), sentence voice (active vs. passive),and supporting VC (presence vs. absence), individu-ally or in combination, affect each subcomponent ofthe SCT? We would expect (a) the comprehension ofpassive sentences for all participants to be poorerthan active sentences, in terms of semantic role as-signment and referent selection, but not verb en-actment; (b) supporting VC to influence referentselection for all participants (because of its depen-dency on the visual image), but not verb enactmentand semantic role assignment; and (c) the compre-hension of passive sentences to be poorer in thegroup with Down syndrome, compared with the typi-cally developing group, in terms of semantic role as-signment (because of its dependency on bufferlikefunctioning of the phonological loop, which is im-paired in individuals with Down syndrome), but notreferent selection or verb enactment.

MethodParticipants

Nineteen adolescents and young adults with Downsyndrome (Trisomy 21) and 19 typically developing chil-dren, who were participants in a larger study investi-gating language and cognition in individuals with Downsyndrome, participated in this study. The group withDown syndrome consisted of 9 female and 10 male par-ticipants who ranged in chronological age from 12 years9 months to 21 years. The participants with Down syn-drome, recruited from Wisconsin and northern Illinois,met the following inclusionary criteria: (a) monolingualuse of spoken English at home and school; (b) MLU, froma narrative language sample, of at least 2.0 morphemes;

(c) pure-tone average of less than 40 dB in at least oneear; (d) chronological age between 12 and 21 years ofage; and (e) as confirmed via parental report, no otherdevelopmental disorder or history of severe behavior,attention, or psychiatric disorder.

The typically developing participants, 12 girls and7 boys, ranging in age from 3 years to 5 years 8 months,were recruited from Madison, WI, and surrounding ar-eas. The participants met the first three criteria listedabove, and their parents reported no history of language,behavior, attention, psychiatric, or developmental dis-order. All families were paid for their participation inthe study.

The groups were pairwise matched on syntactic lan-guage comprehension, as characterized by averaging theage-equivalency scores from Grammatical Morphemesand Elaborated Phrases and Sentences. Table 1 sum-marizes each group’s characteristics in terms of age-equivalency scores.

Hearing, Cognitive, and LanguageComprehension Measures

Hearing status was quantified by summing the num-ber of passes at 20 dB across the frequencies of 500,1000, 2000, and 4000 Hz for each ear, resulting in a scorewithin the range of 1 to 8. We used performance (in termsof raw scores) on the Pattern Analysis subtest (SB-PA)and the Bead Memory subtest (SB-BM) of the Stanford–Binet Intelligence Scale—Fourth Edition (Thorndike etal., 1986) to quantify visual nonverbal cognition and vi-sual short-term memory, respectively. Auditory-verbalshort-term memory was indexed in two ways: the rawscore obtained on the Number Recall task of theKaufman Assessment Battery for Children (K–ABC;Kaufman & Kaufman, 1983) and the number of pho-nemes correctly produced out of 96 on the Nonword Rep-etition Test (NRT; Dollaghan & Campbell, 1998). Vocabu-lary comprehension was quantified by performance (interms of raw scores) on the PPVT–3 and the TACL–3Vocabulary subtest, and syntax comprehension by wasquantified by performance (in terms of raw scores) onTACL–3 Subtests II and III, Grammatical Morphemesand Elaborated Phrases and Sentences and on the SCT.

Sentence Comprehension TaskSentence Stimuli

In the SCT, adapted from Miolo (1998), participantsenacted eight active and eight passive, semantically re-versible, subject, verb, object sentences (see Appendix),for example, “The yellow horse climbs over the blackcow,” using an array of 16 small toy animals as the vi-sual referents. We selected four familiar farm animals


(i.e., cow, sheep, horse, and chicken) and four colors (i.e.,red, black, yellow, and white), creating 16 (4 colors × 4animals) different adjective–noun referents. We assignedthe 16 referents to the test sentences so that each refer-ent occurred twice across all sentences, once as the firstnoun phrase referent and once as the second. We madethese assignments with the constraint that neither colornor animal was repeated in one sentence. We selectedeight transitive verbs that when acted out, were rela-tively distinguishable from each other: push over, jumpover, kiss, climb on, chase, ride on, stand on, and kick.Each verb occurred once in each set of active and pas-sive sentences.

In an acoustically isolated booth, we recorded the16 sentence stimuli using a digital audio recorder(Tascam BP D1) as they were spoken at a normal speak-ing rate by a female, adult, Standard American Englishspeaker who wore a head-mounted microphone (AKG C410) to ensure consistent mouth-to-microphone distance.We downloaded the digitized sentences to a CD-ROM

after they were compiled randomly into four differentsets, using CSpeech (Milenkovic, 1996), a computerizedspeech analysis program.

Visual Display and BarriersThe visual display consisted of the 16 animals po-

sitioned in a V shape on a 30-in. by 20-in. foam coreboard. To control for position effects, the examinerplaced the animals on the board in a randomized orderfor each participant. To vary the participants’ abilityto see the animals, the examiner placed either a trans-parent Plexiglas barrier or an opaque foam core bar-rier in front of the visual display, just before each sen-tence was presented. Thus, while listening to thesentence, the participant was able to see the displaythrough the transparent barrier (VC present) or not(opaque foam core barrier; VC absent). The examinerremoved these barriers after the presentation of eachsentence, so the participants could manipulate theanimals.

Table 1. Characteristics of each group.

Down syndromea Typically developingb

Characteristic M SD M SD

Chronological age 15.96b 2.05 4.21b 0.67Maternal education levelc 15.21b 2.10 17.21b 1.90Syntax comprehensiond 5.12 1.04 5.28 0.95Lexical comprehension

TACL–3 Vocabularye 5.25 1.46 5.64 1.50PPVT–3f 6.21 1.73 5.51 1.33

Nonverbal visual cognition (SB-PA)g 5.16b 1.37 4.02b 0.56Visual short-term memory (SB-BM)h 4.68 1.21 4.40 1.07Auditory-verbal short-term memory

Number Recall (K–ABC)i 4.42 1.71 4.88 1.45Nonword repetition (NRT)j 55.68b 13.84 67.58b 11.31

Mean length of utterancek 4.58b 1.54 6.35b 1.43Hearingl 6.37a 2.45 7.76a 0.75

Note. Means in the same row having the subscript a differ significantly at p < .05 in a one-way analysis ofvariance (ANOVA). Means in the same row having the subscript b differ significantly at p < .01 in a one-wayANOVA.an = 19; 10 male and 9 female participants. bn = 19; 7 male and 12 female participants, for all variables exceptKaufman Assessment Battery for Children (K–ABC; n = 15) and Hearing (n = 17). cNo. years mother spent informal schooling. dAge-equivalent score average from Grammatical Morphemes and Elaborated Phrases andSentences subtests of Test for Auditory Comprehension of Language—Third Edition (TACL–3; Carrow-Woolfolk,1999). eAge-equivalent score from Vocabulary subtest of TACL–3. fAge-equivalent score from the Peabody PictureVocabulary Test—Third Edition (Dunn & Dunn, 1997). gAge-equivalent score from the Pattern Analysis subtest andhage-equivalent score from Bead Memory subtest of the Stanford–Binet Intelligence Scale (4th ed.; Thorndike,Hagen, & Sattler, 1986). iAge-equivalent score from Number Recall subtest of the K–ABC (Kaufman & Kaufman,1983). Four typically developing (TD) participants were excluded because testing stopped before ceiling wasreached. jNumber of phonemes correctly produced out of 96 on Nonword Repetition Test (Dollaghan & Campbell,1998). kMean length of utterance in interview with “yes/no” responses removed. lNumber of passes at 20 dBacross frequencies 500, 1000, 2000, 4000 Hz for each ear. Two TD participants were excluded because thehearing test could not be completed.


Assignment of VisualContext Conditions to Trials

Eight test sentences (four active and four passive)were presented in each VC condition (VC present vs.VC absent), creating a total of 16 test trials and fourexperimental conditions (i.e., active/VC present, active/VC absent, passive/VC present, and passive/VC absent).To control for order effects, we assigned the VC condi-tions to the four 16-sentence sets in four different ways,creating 16 VC-assignment/sentence-set combinations.We assigned one of these sets to each participant a priori,so that each VC-assignment/16-sentence set combina-tion occurred no more than three times across all par-ticipants.

Recording EquipmentThe experimental protocol was videotaped using a

Panasonic video camera (WV3260) positioned in the cor-ner of the room. The audio signal for the videotape wasrecorded via a Crown Sound Grabber II microphone,placed on the table between the examiner and the par-ticipant. All recorded directions and test stimuli for theexperimental tasks were presented under Koss R/30Sheadphones at a comfortable listening level using aPanasonic compact disc player (SL-S260). The partici-pants’ responses during experimental tasks were re-corded using a Crown PZM-185 microphone, placed onthe table between the examiner and the participant, anda Marantz PMD-185 audiotape recorder.

ProcedureTesting Sessions

Testing was completed during two sessions, the firstand the sixth in a series of testing sessions conductedas part of the larger study in which these participantswere involved. At the start of the first testing session, alicensed audiologist administered a hearing screeningat 20 dB across the frequencies of 500, 100, 2000, and4000 Hz. Threshold testing followed the screening forthose participants who did not respond at 20 dB for anyfrequency. Only those participants who met the hearingcriteria of a pure-tone average of less than 40 dB in atleast one ear could continue with the experimental pro-tocol. During the remainder of the first session (approxi-mately 3 hr long), certified speech-language pathologistsor graduate students in speech-language pathology ad-ministered the NRT and the standardized measures de-scribed above, in an order counterbalanced across par-ticipants. The SCT was administered during a separate90-min session, which occurred, on average, about 33days later for the participants with Down syndrome and41 days later for the typically developing participants.

To minimize the effects of boredom or fatigue in thetypically developing children, we used reinforcement,such as stickers and opportunities to play computergames, during the experimental sessions. For those par-ticipants who seemed to be more affected by sessionlength during prior sessions, we administered the SCTearlier in the testing protocol for that day.

Standardized TestsEach standardized test was administered and scored

according to the procedures described in the examiner’smanual, with the exception of the K–ABC. The NumberRecall subtest of the K–ABC consists of units of two orthree test items. Stopping points (always the last item ina unit) are designated a priori according to the examinee’schronological age. The manual specifies that the exam-iner should stop testing (a) at the designated stoppingpoint or (b) if the examinee fails every item in one unitbefore the designated stopping point. If the examineepasses all the items in the last unit specified for theexaminee’s age, then testing is continued beyond the stop-ping point until one more item (not unit) is failed. Testadministration followed this procedure for all typicallydeveloping participants except 4, for whom the examinerinadvertently stopped testing before either stopping cri-terion had been met. Because of this, we excluded these4 participants from any analyses that involved K–ABCscores. For the participants with Down syndrome, wedesignated the stopping point as a function of perfor-mance rather than chronological age. The examinerstopped testing when participants could not completeall three items in one unit or, in two cases, when partici-pants became extremely frustrated because they couldnot complete two consecutive items in one unit.

Nonword Repetition TaskAdministration. In the NRT (Dollaghan & Campbell,

1998), participants listened to and then repeated 16nonwords, 4 at each of 4 syllable lengths. First, the par-ticipants listened to prerecorded directions, asking themto repeat each “word” they heard. Then the examinerpaused the CD player and presented, via live voice, twopractice items. After the participant was able to com-plete the practice repetition task, the examiner pre-sented each prerecorded nonword once only, pausing theCD after each item. The examiner transcribed onlineand audiotaped the participants’ repetitions.

Scoring. We scored the participants’ audiotapednonword repetitions according to the rules establishedby Dollaghan and Campbell (1998): (a) Phoneme addi-tions were excluded from the analysis; (b) each conso-nant and vowel in a nonword was scored as correct orincorrect in relation to the target phoneme; (c) distor-tions of phonemes were scored as correct, and phoneme


substitutions and omissions, even if consistent articu-lation errors, were scored as incorrect; and (d) for thosenonword repetitions in which syllables were added oromitted by the participant, “individual scoring proceededafter aligning the syllable sequence produced by thesubject as nearly as possible to that of the target, usingvowels repeated as syllable anchors to maximize thesubject’s score” (Dollaghan & Campbell, 1998, p. 1139).

Sentence Comprehension TaskPretesting. The examiner asked the participant,

seated at the adjacent side of a small table, to point toeach of the four colors and four animals in the arrayand then act out each of the eight actions, used in thetest sentences, with two of the animals. All participantswere able to identify the colors and animals correctly,and approximately 40% of the participants performedall actions accurately at their first attempt. For the re-mainder of the participants whose first attempts at anyaction were inaccurate, or could be easily confused withanother target action (most often the pairs, kick and pushover and climb over and jump over), the examiner dem-onstrated the target action or explained how to manipu-late the animals so that one action was distinguishablefrom another. For example, to differentiate kick frompush over, the examiner explained and demonstratedthat the animal being kicked should not fall over, or todifferentiate jump over from climb over, the animal be-ing jumped over should not be touched. The examinerthen asked the participant to demonstrate the actionagain. The examiner provided positive verbal feedbackif the action was performed correctly or provided onemore demonstration if it was incorrect. All participantswere able to perform all actions correctly after no morethan two demonstrations.

Next, the examiner presented four pairs of practicesentences: the first three pairs, via live voice, and thefourth, from the CD, via headphones. For the first sen-tence pair (both active) and the second pair (both pas-sive), the examiner, using the animals from the array,demonstrated the first sentence in the pair and thenasked the participant to act out the second sentence.The examiner presented the third and fourth pairs (eachcontaining one active and one passive sentence) with-out demonstration. The examiner presented the firstsentence in the third and fourth pairs while holding theopaque barrier in front of the array and the second sen-tence in each pair while holding the transparent bar-rier. After each presentation, the participant acted outthe test sentence with the animals.

Administration. Immediately after the pretesting,the examiner placed the appropriate barrier in front ofthe animal array and presented the first test sentencevia headphones at a comfortable listening level. After

presenting each test sentence, the examiner removedthe barrier so the participant could manipulate the ani-mals. To control the amount of time the animal arraywas in view, the examiner placed the opaque barrier infront of the array as soon as the participant finishedmanipulating the animals. After doing so, the examinerwould place the animals back into their appropriatepositions before presenting the next sentence.

Scoring. The examiner recorded online, in writingand via videotape, the participants’ responses to (i.e.,enactment of) each SCT test sentence. A transcriber in-dependently interpreted, transcribed, and scored thevideotaped responses and, if the videotaped action wasnot clearly visible, referred to a list of target sentencesand the examiner’s written record of the responses. Ontwo occasions, when the videotape recording of part ofthe SCT administration failed, the transcriber used theexaminer’s written descriptions for scoring. Occasion-ally, participants would complete, or nearly complete,an action and then self-correct. In those cases, the tran-scriber scored the self-corrected action.

The transcriber assigned verb enactment scores toeach response as follows: 1 if participants enacted thetarget verb correctly or 0 if they did not enact a verb orenacted the wrong verb. For referent selection, 2 wasassigned if both referents selected were an exact matchto the target referents, in both color and animal, irre-spective of which semantic role they played; 1 if onlyone referent was an exact match; or 0 if neither referentwas an exact match. For semantic role assignment, 2was assigned if both referents selected were an exactmatch, in animal (but not necessarily color) and seman-tic role (actor or patient), to those in the target sentence;1 if only one referent was an exact match; or 0 if neitherreferent was an exact match.

For example, consider two responses to the targetsentence, “The white sheep is jumped over by the redcow.” One participant responds by making the whitesheep jump over the red cow, and another by makingthe white cow climb over the red sheep. To the firstresponse, in which the participant enacts the verb cor-rectly and selects the correct referents but reversestheir semantic roles, one would assign verb enactment,referent selection, and semantic role assignment scoresof 1, 2, and 0, respectively. To the second response, onewould assign scores of 0, 0, and 2, respectively, becausethe participant enacted the wrong verb but assignedthe correct semantic roles to the selected referents,matched to the target referents in terms of animal butnot color.

These two examples show that a maximum scorefor referent selection does not necessarily predict a maxi-mum score for semantic role assignment and vice versa.Also, certain patterns of scores can provide evidence for


sentence interpretation strategies. For example, scoresof 2 and 0 for referent selection and semantic role as-signment, respectively, as described above in the firstresponse to the passive test sentence, indicate semanticrole reversal and the possible use of a word-order inter-pretation strategy.

Reliability. A second transcriber independently in-terpreted, transcribed, and scored the videotaped en-actments for each SCT test sentence for 16 randomlyselected participants (42%), under the same conditionsas the first transcriber. The first and second transcrib-ers obtained the following percentages of agreement fortheir interpretations of the enactments by the 16 par-ticipants: (a) 93% for verb enactment, (b) 100% for ref-erent identification, and (c) 97.7% for semantic roleassignment. Transcribers resolved any interpretationdifferences by consensus while watching the tape. Afterthis, the second transcriber scored the responses to eachsentence. The transcribers obtained 100% agreement onthe scoring of responses.

ResultsLanguage and Cognitive Profile

To determine whether the group with Down syn-drome possessed the phenotypic profile of language andcognitive abilities typical of groups from previous stud-ies, we computed (using standardized raw scores calcu-lated across all participants) the difference between thenonverbal cognition measure (SB-PA) and measures ofeach of the following skills: language production (MLU),syntax comprehension (TACL–3 Grammatical Mor-phemes and Elaborated Phrases and Sentences), audi-tory short-term memory (K–ABC and NRT), and visualshort-term memory (SB-BM). A one-way between-sub-jects multivariate analysis of variance (MANOVA), us-ing the five discrepancy scores as dependent variables,was statistically significant using Wilks’s criterion, F(5,28) = 4.69, p = .003, partial η2 = .46. Univariate testsrevealed significantly larger discrepancies for the groupwith Down syndrome, compared with the typically de-veloping group, between nonverbal cognition and (a)language production, F(1, 32) = 18.05, p < .001, partialη2 = .36; (b) syntax comprehension, F(1, 32) = 6.33, p =.017, partial η2 = .17; (c) auditory short-term memory,as measured by the NRT, F(1, 32) = 20.54, p < .001, par-tial η2 = .39; and the K–ABC, F(1, 32) = 6.43, p = .016,partial η2 = .17; but not (d) visual short-term memory,F(1, 32) = 4.01, p = .054.

Vocabulary ComprehensionWe compared estimates of vocabulary comprehen-

sion, in terms of z scores, obtained using the PPVT–3

and the Vocabulary subtest of the TACL–3, to determinewhether the vocabulary comprehension performance foreach group varied as a function of the measure used toderive the estimate. The difference in mean z scores wasstatistically significant at the .05 level for the group withDown syndrome, with means of .22 and –.15 for thePPVT–3 and the TACL–3, respectively, t(18) = 2.40, p =.03; but not for the typically developing group, with meansof –.22 and .15 for the PPVT–3 and the TACL–3, respec-tively, t(18) = –1.85, p = .08. Thus, for the group withDown syndrome, the estimate of vocabulary comprehen-sion derived using the PPVT–3 was higher than that de-rived using the Vocabulary subtest of the TACL–3.

Predictors of Sentence Comprehension:Group Membership, Auditory-Verbaland Visual Short-Term Memory andHearing Status

Analyses and Data ScreeningWe used multiple correlation analyses to evaluate,

for all participants, the degree to which visual short-term memory (SB-BM), auditory-verbal short-termmemory (NRT and K–ABC) and group membership, firstas a set and then individually, might account for varia-tion in sentence comprehension, as measured by thethree SCT components (verb enactment, referent selec-tion, and semantic role assignment) and two TACL–3subtests (Grammatical Morphemes and ElaboratedPhrases and Sentences). In addition, we conducted mul-tiple correlation analyses for the group with Down syn-drome, with hearing status replacing group member-ship in the set of predictor variables, to evaluate thepotential effect of their significantly poorer hearing sta-tus on sentence comprehension.

Each of the dependent and predictor variables wasquantified using total raw scores (see Table 2). We con-ducted a preliminary screening of the data by examin-ing the residuals for unusual patterns or influential datapoints. The data met the assumptions of normality, lin-earity, homoscedasticity, and independence of residuals;and using Mahalanobis distances and Cook’s distances(see Tabachnick & Fidell, 2001), we did not detect anymultivariate outliers in the data set.

Relationships Among Predictor VariablesBefore conducting the multiple correlation analyses,

we examined the degree of association among the predic-tor variables (group, SB-BM, K–ABC, and NRT, and hear-ing status for the group with Down syndrome) by com-puting their bivariate correlations (see Table 3). We alsoincluded the measure of nonverbal cognition (SB-PA) inthe correlation matrix to examine its potential influence


on the predictor variables. Nonverbal cognition was sig-nificantly associated with visual short-term memoryand group. The significant correlations between groupand both nonverbal cognition and nonword repetitionwere expected given that the groups’ performance inthese areas was significantly different (see Table 1). In-terestingly, the K–ABC and NRT, both used to measureauditory-verbal short-term memory, were correlated toone another, but not significantly so (r = .28). Failureto find a significant correlation between the two vari-ables may be explained by the relatively small samplesize, by the greater articulatory demands of the NRTversus the Number Recall task, or by potential differ-ences in the aspects of auditory-verbal short-termmemory represented by each measure: the NRT reflect-ing, to a greater degree, the contribution of long-termphonotactic knowledge; and the K–ABC, the demandsof short-term store.

Variable Set as a PredictorTo determine the degree to which the set of predic-

tor variables (group, K–ABC, NRT, and SB-BM)

accounted for variation in each measure of comprehen-sion, we calculated the adjusted squared multiple cor-relation coefficients (adjusted R2) between the predictorset and each dependent variable. Because the ratio ofpredictor variables to sample size was relatively low(approximately 1:9), we evaluated the strength of therelationships between the set of predictor variables anddependent variables using the adjusted R2, which rep-resents a less biased estimate of the population R2 thandoes the sample R2.

To control for inflated Type I error, we madeBonferroni-type adjustments in which we assigned analpha to each dependent variable “so that the alpha forthe set of dependent variables does not exceed some criti-cal value [in this case, .05, and] more important depen-dent variables can be given more liberal alphas”(Tabachnick & Fidell, 2001, p. 349). We divided the al-pha of .05 for the SCT among the three significance tests,assigning an alpha of .02 to referent selection and se-mantic role assignment, because of their centrality toour hypotheses, and a more conservative alpha of .01 toverb enactment. The adjusted R2 values for the SCT were

Table 2. Dependent and predictor variables for each group.

Down syndromea Typically developingb

Variable M SD Range M SD Range

Dependent variables

TACL–3Grammatical Morphemes 18.47 9.38 2–35 20.63 6.45 9–32Elaborated Phrases and Sentences 20.11 8.06 9–40 21.05 9.22 8–40

Sentence comprehension taskVerb enactmentc 8.58 3.11 1–13 9.95 3.17 4–15Referent selectiond 16.11 6.26 6–31 17.16 5.57 9–26Semantic role assignmentd 12.84b 4.88 5–23 18.00b 6.77 3–31

Predictor variables

Visual short-term memory (SB-BM) 10.26 4.43 5–19 9.32 3.90 4–18Auditory-verbal short-term memory

Number Recall (K–ABC) 5.42 2.36 0–11 6.27 2.46 0–10Nonword repetition (NRT) 55.68b 13.84 31–75 67.58b 11.31 41–87

Hearinge 6.37a 2.45 1–8 7.76a 0.75 5–8

Note. Values represent raw scores. TACL–3 = Test for Auditory Comprehension of Language—Third Edition(Carrow-Woolfolk, 1999); SB-BM = Bead Memory subtest of Stanford–Binet Intelligence Scale (4th ed.; Thorndike,Hagen, & Sattler, 1986); K–ABC = Kaufman Assessment Battery for Children (Kaufman & Kaufman, 1983); NRT =Nonword Repetition Test (Dollaghan & Campbell, 1998). Means in the same row having the subscript a differsignificantly at p < .05 in a one-way analysis of variance (ANOVA). Means in the same row having the subscriptb differ significantly at p < .01 in a one-way ANOVA.an = 19. bn = 19 for all variables except K–ABC and Hearing. For K–ABC, n = 15. Four typically developing (TD)participants were excluded because K–ABC testing stopped before ceiling was reached. For hearing, n = 17. TwoTD participants were excluded because hearing test could not be completed. cMaximum score possible = 16.dMaximum score possible = 32. eNumber of passes at 20 dB across frequencies 500, 1000, 2000, 4000 Hz foreach ear. Hearing was not included as a predictor variable for the TD group because of minimal variation inscores; 15 of the 17 TD participants passed all eight screening items.


statistically significant for referent selection (.24), F(4,29) = 3.59, p = .017, and semantic role assignment (.48),F(4, 29) = 8.66, p < .001, but not verb enactment (.04),F(4, 29) = 1.32, p = .29.

For the TACL–3, we split the alpha of .05 betweenthe significance tests for each syntactic subtest. Bothadjusted R2 values were statistically significant: (a) .24for the Grammatical Morphemes subtest, F(4, 29) = 3.55,p = .018; and (b) .27 for the Elaborated Phrases andSentences subtest, F(4, 29) = 4.09, p = .010.

Individual Variables as PredictorsNext, we determined how much variation in each

measure of comprehension was accounted for by eachpredictor variable (group, SB-BM, K-ABC, and NRT)while holding constant the influence of the remainingvariables in the set. To control for an inflated Type Ierror rate, we conducted further significance testing onlyif the adjusted R2 between the dependent variable andthe entire set of predictor variables was statistically sig-nificant (Cohen, Cohen, West, & Aiken, 2003). As a re-sult, we tested the significance of the semipartial corre-lations between each predictor variable and alldependent variables, except SCT verb enactment.

Group membership. Group membership accountedfor a significant amount of variation in SCT semanticrole assignment only, individually accounting for 11%of its variance, t(29) = 2.65, p = .013. Therefore, we evalu-ated the contributions of auditory-verbal and visualshort-term memory to semantic role assignment for eachgroup separately.

Auditory-verbal short-term memory. For all partici-pants, auditory-verbal short-term memory, as measuredby the NRT and K–ABC, contributed significantamounts of unique variance to all dependent measuresof comprehension analyzed: (a) 19% to SCT referent se-lection, F(2, 29) = 4.80, p = .016; (b) 24% to TACL–3Grammatical Morphemes, F(2, 29) = 6.16, p = .006; and(c) 27% to TACL–3 Elaborated Phrases and Sentences,F(2, 29) = 7.92, p = .002. Before evaluating the contribu-tion of auditory-verbal short-term memory to semanticrole assignment in each group, we divided the originalalpha of .02 from the omnibus test by 2, assigning analpha level of .01 to each significance test. Auditory-ver-bal short-term memory accounted for 48% of the vari-ance in semantic role assignment for the group withDown syndrome, F(2, 15) = 9.84, p = .002, but none ofvariance for the typically developing group, F(2, 11) =0.17, p =.84.

Individually, the K–ABC made statistically signifi-cant contributions to all measures, except SCT referentselection: (a) 27% to TACL–3 Grammatical Morphemes,t(29) = 3.42, p = .002; (b) 13% to TACL–3 ElaboratedPhrases and Sentences, t(29) = 2.43, p = .021; and (c)14% to SCT semantic role assignment for the group withDown syndrome, t(15) = 2.19, p = .044. The NRT ac-counted for significant amounts of unique variance inall measures, except TACL–3 Grammatical Morphemes:(a) 15% for TACL–3 Elaborated Phrases and Sentences,t(29) = 2.58, p = .015; (b) 14% for SCT referent selection,t(29) = 2.48, p = .019; and (c) 25% for SCT semantic roleassignment in the group with Down syndrome, t(15) =2.93, p = .01.

Table 3. Intercorrelations among predictor variables and nonverbal cognition for all participants.

Variable 2 3 4 5 6

1. Nonverbal cognition (SB-PA) –.50* .48* –.22 –.09 –.302. Group membership –.12 .18 .44* —c

3. Visual short-term memory (SB-BM) .06 .24 –.184. Number Recall (K–ABC)a .28 .145. Nonword repetition (NRT) –.056. Hearingb (group with DS only)

Note. n = 38, except for K–ABC and hearing. n = 34 for K–ABC. Four typically developing participants wereexcluded because of missing data. n = 19 for hearing status because correlations were computed for group withDS only. SB-PA = Pattern Analysis subtest of Stanford–Binet Intelligence Scale (4th ed.; SB; Thorndike, Hagen, &Sattler, 1986); SB-BM = Bead Memory subtest of SB; K–ABC = Kaufman Assessment Battery for Children(Kaufman & Kaufman, 1983); NRT = Nonword Repetition Test (Dollaghan & Campbell, 1998); DS = Downsyndrome.an = 34 for K–ABC. Four typically developing participants were excluded because of missing data. bNumber ofpasses at 20 dB across frequencies 500, 1000, 2000, 4000 Hz for each ear. cHearing was included as apredictor variable for DS group only (n = 19); therefore, no correlation is reported between group membershipand hearing.

*p < .01, one-tailed.


Visual short-term memory. The semipartial correla-tions between visual short-term memory (SB-BM) andeach dependent measure were not statistically signifi-cant, indicating that visual short-term memory did notpredict variation in any aspect of comprehension mea-sured in this study.

Hearing status. We evaluated the degree to whichhearing status might account for variation in languagecomprehension for the group with Down syndrome byconducting multiple correlation analyses with hearingstatus replacing group membership in the set of predic-tor variables. Hearing status made a significant contri-bution to variation in TACL–3 Grammatical Morphemesonly, individually accounting for 23% of its variance, t(14)= 2.69, p = .018.

Summary of Findings:Correlation Analyses

Of the variables selected a priori as potential pre-dictors of sentence comprehension, group membershippredicted variation in only one measure, SCT semanticrole assignment; and visual short-term memory, none.Auditory-verbal short-term memory (as measured by theNRT and K–ABC) was the most influential of the pre-dictor variables, accounting for, in all participants, ap-proximately one fifth of the variation in SCT referentselection and one fourth of the variation in TACL–3Grammatical Morphemes and TACL–3 ElaboratedPhrases and Sentences. In the group with Down syn-drome, auditory-verbal short-term memory accountedfor almost half of the variation in semantic role assign-ment, indicating that sentence comprehension, in terms

of semantic role assignment, placed substantial de-mands on auditory-verbal short-term memory in thisgroup.

The NRT alone predicted SCT referent selection, andthe K–ABC alone predicted performance on TACL–3 Gram-matical Morphemes. They each contributed a uniqueamount of variance to TACL–3 Elaborated Phrases andSentences in all participants and SCT semantic role as-signment in the group with Down syndrome. Their uniquecontributions to the former dependent measure were com-parable; but to the latter, the contribution from the NRTwas almost twice that of the K–ABC.

Effects of Group, Sentence Voice, andVisual Context on Sentence Comprehension

Each of the dependent variables of verb enactment,referent selection, and semantic role assignment wasanalyzed in a 2 × 2 × 2 (Group × Sentence Voice × VC)mixed measures ANOVA, with group as the between-subjects factor and sentence voice and VC as repeatedmeasures factors. For the same reasons described ear-lier, we made a Bonferroni-type adjustment to the TypeI error rate for the ANOVA models for each dependentvariable: .02 for referent selection, .02 for semantic roleassignment, and .01 for verb enactment. Summary datafor each dependent measure as a function of conditionare displayed in Table 4.

Verb EnactmentThe ANOVA calculated for verb enactment revealed

no significant main effects or interactions, indicatingthat variation in verb enactment scores was not

Table 4. Subcomponents of sentence comprehension task as a function of condition for each group.

Verb enactment a Referent selectionb Semantic role assignment b

Condition M SD M SD M SD

Active voiceContext present

Down syndromec 2.26 0.93 4.21 1.90 3.95 2.15Typically developingc 2.47 1.17 4.79 1.96 4.58 2.65

Context absentDown syndrome 2.00 1.25 4.05 1.84 4.42 2.32Typically developing 2.58 1.12 5.00 1.89 5.16 2.03

Passive voiceContext present

Down syndrome 2.05 0.91 4.05 2.09 2.00 1.49Typically developing 2.74 1.05 4.11 1.56 4.26 2.51

Context absentDown syndrome 2.26 1.10 3.79 1.99 2.47 1.87Typically developing 2.16 1.07 3.26 1 .91 4.00 2.11

Note. Values represent raw scores.aMaximum score possible = 4. bMaximum score possible = 8. cn = 19.


influenced by group membership, sentence voice, or thepresence or absence of VC.

Referent SelectionThe ANOVA calculated for referent selection re-

vealed a significant main effect for sentence voice, F(1,36) = 8.34, p = .007, partial η2 = .19, with referent selec-tion more accurate in active sentences (M = 4.51) thanpassive sentences (M = 3.80) for all participants. Theother main effects and interactions were nonsignificant.

Semantic Role AssignmentThe ANOVA for semantic role assignment revealed

significant main effects of sentence voice, F(1, 36) =12.47, p = .001, partial η2 = .26; and group, F(1, 36) =7.25, p = .011, partial η2 = .17. The main effect of VC andall interactions were nonsignificant. The assignment ofsemantic roles was more accurate for (a) active sentences(vs. passive) sentences for all participants (Ms = 4.53and 3.18, respectively) and (b) all sentences for the typi-cally developing group compared with the group withDown syndrome (Ms = 4.50 and 3.21, respectively). Theseresults indicated that the effects of sentence voice andgroup membership, but not VC, accounted for signifi-cant amounts of variation in semantic role assignmentscores (26% and 17%, respectively).

In addition, we calculated the frequency with whichthe participants assigned correctly selected animals totheir correct semantic roles. The typically developinggroup assigned 73% of the animals to correct roles inactive sentences and 64% in passive sentences, whereasthe group with Down syndrome assigned 61% of the ani-mals to correct roles in active sentences but only 34%for the passive sentences. In other words, the partici-pants with Down syndrome reversed semantic roles forcorrectly selected animals (27% of the time more oftenin passive than active sentences, 66% vs. 39%, respec-tively), suggesting the use of a word-order interpreta-tion strategy for passive sentences in the group withDown syndrome.

Summary of Findings:Analyses of Variance

The accuracy of verb enactment appears to be unaf-fected by variation in sentence voice and VC and was com-parable for both groups. Variation in sentence voice, butnot VC, influenced referent selection and semantic roleassignment, with their accuracy reduced in passive (vs.active) sentences for both groups. The only between-groupdifference was in the accuracy of semantic role assign-ment across all sentences, with the group with Down syn-drome performing more poorly than their typically de-veloping counterparts matched on syntax comprehension.

DiscussionLanguage and Cognitive Profile

The language and short-term memory profile of ado-lescents with Down syndrome, reviewed in the intro-duction, is referenced to nonverbal cognitive ability(Chapman & Hesketh, 2000). In contrast, the adoles-cents with Down syndrome in this study were matchedto typically developing participants, not on nonverbalcognition, but on syntactic comprehension, a skill thatslows or declines in adolescents with Down syndrome(Chapman et al., 2002). Therefore, to determine whetherthe group with Down syndrome exhibited the charac-teristic profile of language and short-term memory abili-ties, we calculated discrepancy scores, expressed in stan-dardized raw scores, between nonverbal cognition andfive variables representing language and short-termmemory skills (auditory-verbal short-term memory, vi-sual short-term memory, MLU, vocabulary comprehen-sion, and syntax comprehension). The discrepancy scoresfor the group with Down syndrome were significantlygreater than those for the typically developing group,confirming the description of the phenotype gatheredfrom previous research.

Vocabulary ComprehensionWe found that the level of vocabulary comprehen-

sion in the group with Down syndrome, but not the typi-cally developing group, was higher when measured us-ing the PPVT–3 than the TACL–3 Vocabulary subtest.We interpreted our findings to mean that the life expe-rience of the chronologically older participants withDown syndrome influenced the comprehension of fre-quency-based vocabulary items in the PPVT–3, but notthe conceptually based vocabulary items in the TACL–3 Vocabulary subtest. These findings might also explainvocabulary strengths documented in adolescents withintellectual disability generally (Facon & Facon-Bollengier, 1997; Facon, Facon-Bollengier, & Grubar,2002; Facon, Grubar, & Gardez, 1998). In future work,we plan to analyze the type of vocabulary items passedand failed on the two tests by each group.

Predictors of Language ComprehensionWe evaluated the relative contributions of group

membership, visual and auditory-verbal short-termmemory (and the contribution of hearing status in thegroup with Down syndrome) to sentence comprehension,as measured by: (a) the Grammatical Morphemes andElaborated Phrases and Sentences subtests of theTACL–3, a traditional picture-pointing task; and (b) theSCT, an experimental sentence comprehension task


more dependent on the short-term storage of auditoryinformation than the picture-pointing task. The SCTstressed auditory-verbal and visual short-term memorycapacity by manipulating linguistic complexity and sup-porting VC and allowed us to evaluate comprehensionperformance in terms of its constituent components(verb enactment, referent selection, and semantic roleassignment).

Group Membership andAuditory-Verbal Short-Term Memory

In accordance with our predictions, group member-ship contributed significantly to variation in semantic roleassignment (11%), but not the TACL–3 subtests. Audi-tory-verbal short-term memory significantly influencedperformance on the TACL–3 syntactic subtests and theSCT, with its contributions ranging from 19% for SCTreferent selection in all participants to 48% for SCT se-mantic role assignment in the group with Down syn-drome. The substantial contribution of auditory-verbalshort-term memory to variation in semantic role assign-ment for the group with Down syndrome implies a sig-nificant role for auditory-verbal short-term memory inthe comprehension of information-dense sentences byadolescents and young adults with Down syndrome. Fur-ther evidence of its involvement in complex sentence com-prehension in this group is the higher rate of semanticrole reversal in passive sentences for the participantswith Down syndrome, suggesting their probable use ofa sentence comprehension strategy that is heavily de-pendent on the auditory memory of word order.

Unexpectedly, the measures of auditory-verbalshort-term memory were not significantly associatedwith variation in semantic role assignment for the typi-cally developing group. The typically developing group’smore accurate assignment of semantic roles in the testsentences, while low enough to be attributable to chanceperformance for some participants, may suggest a moreimportant role for the central executive, in the syntac-tic reinterpretation of passive sentences for this group,than auditory-verbal short-term memory. In future work,we will include measures reflecting central executivefunction to evaluate the role of working memory in thecomprehension of active and passive sentences in typi-cally developing children.

The NRT and K–ABC contributed significantamounts of variance, jointly and independently, toTACL–3 Elaborated Phrases and Sentences in all par-ticipants and SCT semantic role assignment in the groupwith Down syndrome. This finding suggests that per-formance on these dependent measures (representingthe more memory-dependent aspects of sentence com-prehension) might involve a broader range of aspects ofphonological working memory, such as long-term

phonotactic knowledge and short-term storage (as mea-sured by the NRT and K–ABC, respectively) and otheruntapped aspects, and activate them to a greater degree,than the comprehension of other linguistic forms (e.g.,referential expressions or grammatical morphemes). Thesignificant contribution of the NRT and K–ABC (and thedifferent aspects of phonological memory they access) tovariation in semantic role assignment in the group withDown syndrome, but not the typically developing group,might suggest also that the involvement of each compo-nent of working memory (phonological loop, the centralexecutive, and even the visuospatial scratchpad) in sen-tence comprehension varies according to not only the lin-guistic form to be understood but also the requirementsof the comprehension task (picture pointing vs. actingout) and the listeners’ linguistic levels, including theiruse of comprehension strategies.

Visual Short-Term MemoryWe had expected that visual short-term memory

might have played a role in the participants’ ability toselect correctly the referents in the SCT; however, wefound no significant contribution of visual short-termmemory to any of the aspects of sentence comprehen-sion measured in this study. The complexity of the vi-sual search task may have contributed to this finding.The large number of foils in the visual array may havecompromised the participants’ ability to locate the firstreferent in the visual array, before hearing the second,therefore inhibiting the visual representation of refer-ents in short-term memory.

Hearing StatusBecause of the significantly poorer hearing levels in

the group with Down syndrome, we conducted anotherset of multiple correlation analyses for this group, withhearing status replacing group membership in the set ofpredictor variables. Hearing status uniquely accountedfor 23% of the variation in performance on TACL–3 Gram-matical Morphemes. This relationship makes sense giventhat the comprehension of grammatical morphemes isheavily dependent on one’s ability to discriminate high-frequency sounds, such as /s/ and /z/. Children with Downsyndrome often experience varying degrees of permanentand fluctuating hearing loss, explaining why grammati-cal morphology is one of the areas of language produc-tion most affected and most variable (Chapman et al.,1998) in children with Down syndrome.

Effects of Group, Sentence Voice, andVisual Context on Sentence Comprehension

Using the SCT, we evaluated the effects of varyingsentence voice and VC on sentence comprehension in


terms of its constituent components (verb enactment,referent selection, and semantic role assignment). Inaccordance with our predictions, the comprehension ofpassive sentences was significantly poorer than the com-prehension of active sentences for all participants, interms of referent selection and semantic role assignment,but not verb enactment.

We expected also that for all participants, the si-multaneous presentation of the visual array with thetest sentences would benefit referent selection, but notsemantic role assignment or verb enactment. However,we were unable to detect a significant effect of VC onreferent selection, which may be attributable, once again,to the complexity of the visual search task.

The group with Down syndrome performed morepoorly than the typically developing group in the assign-ment of semantic roles in all sentences, despite thegroups’ match in syntax comprehension. This findingsupports our hypothesis that when those aspects of sen-tence comprehension particularly dependent on audi-tory-verbal short-term memory are stressed, they canbe compromised to a greater degree in the group withDown syndrome.

Limitations of StudyIn this study, the participants with Down syndrome

ranged in age from 12 to 21 years; therefore, the rel-evance and implications of our findings are restricted toindividuals with Down syndrome within that develop-mental period. Although the range of tasks (picturechoice, sentence enactment) is broader than usually usedin comprehension assessment, these tasks do not assessthe temporal and pragmatic demands of ongoing com-prehension in conversational contexts. The complexityof the visual array may have contributed to our inabil-ity to detect an effect of visual context (or visual short-term memory) on any comprehension measure. There-fore the sensitivity of the task may be improved byreducing the number of foils in the visual array.

Finally, this study is limited in the relatively smallsamples studied (19 in each group). The size of thesample limits our ability to make claims regarding theexternal validity of our study, that is, the degree to whichour findings can be generalized to the whole populationof adolescents and young adults with Down syndrome.A smaller sample size also reduces the probability ofdetecting possible between-group differences in perfor-mance for the variables of interest.

Clinical ImplicationsThe results of our study showed that vocabulary

comprehension in adolescents and young adults with

Down syndrome is characterized differently when mea-sured by the PPVT–3 and the TACL–3 Vocabularysubtest, emphasizing the importance of evaluating thenature of vocabulary items as a source of difficulty inlexical comprehension. The findings also imply that theintralinguistic profiles of individuals with Down syn-drome may vary depending on the test used to charac-terize lexical comprehension, which in turn may influ-ence intervention decisions.

Our findings also support the utility of the objectmanipulation task as a measure of sentence comprehen-sion. The task we used allowed for the scoring of differ-ent aspects of comprehension: verb enactment, referentselection, and semantic role assignment. Had we used abinary scoring of performance on this task, many of thedifferences in comprehension performance reported inthis study would have been masked.

The K–ABC and the NRT accounted for variation,either separately or jointly, in most of the comprehen-sion measures, perhaps reflecting the unique challengesof each task. Our findings would suggest both the use ofthese tasks to measure auditory-verbal short-termmemory and the use of developmentally appropriatemeasures of central executive functioning in researchstudies and clinical practice.

This study provides further evidence of a signifi-cant relationship between auditory-verbal short-termmemory and language comprehension and, consequently,the potentially limiting effect auditory-verbal short-termmemory deficits might have on language functioning inindividuals with Down syndrome. Our findings suggestthat intervention programs aimed at improving lan-guage comprehension in individuals with Down syn-drome should focus not only on a general, developmen-tally sequenced set of goals but also on structuresparticularly vulnerable to auditory-verbal short-termmemory limitations, such as syntactically complex andambiguous sentences. Furthermore, adolescents andyoung adults with Down syndrome might benefit fromthe complementary teaching of comprehension monitor-ing and repair strategies for communication situationsin which these structures occur. The influence of hear-ing status on the comprehension of grammatical mor-phology in the group with Down syndrome underscoresthe need for ongoing monitoring and management ofhearing and the importance of aural rehabilitation ineducational and vocational settings.

AcknowledgmentsThis research was supported by National Institutes of

Health Grant R01 HD23353 to the second author, withadditional support from the National Down Syndrome Society.We are grateful to the parents, children, adolescents, andyoung adults who participated in our study. We thank


Cynthia Bridge, Stephen LeMire, Mary Lindstrom, SallyMiles, Peggy Rosin, and the research team for their assis-tance, and we thank the reviewers for their helpfulcomments.

ReferencesAbbeduto, L., Pavetto, M., Kesin, E., Weissman, M.,

Karadottir, S., & O’Brient, A. (2001). The linguistic andcognitive profile of Down syndrome: Evidence from acomparison with fragile X syndrome. Down SyndromeResearch and Practice, 7, 9–16.

Baddeley, A. D. (1986). Working memory. Oxford, England:Clarendon Press.

Baddeley, A. D. (2001). Is working memory still working?American Psychologist, 56, 849–864.

Baddeley, A. D., & Hitch, G. J. (1974). Working memory.In G. Bower (Ed.), The psychology of learning andmotivation (Vol. 8, pp. 47–90). New York: Academic Press.

Bishop, D. V. M. (1983). The Test for Reception of Grammar.Manchester, England: Author. (Available from Age andCognitive Performance Research Centre, University ofManchester, Manchester, England M13 9PL)

Bishop, D. V. M. (1987, April). The concept of comprehen-sion in language disorder. Paper presented at the FirstInternational Symposium of Specific Speech and Lan-guage Disorders in Children, University of Reading,Reading, England.

Boehm, A. (1971). The Boehm Test of Basic Concepts. NewYork: Psychological Corporation.

Carrow-Woolfolk, E. (1985). The Test for Auditory Compre-hension of Language—Revised. Allen, TX: DLM TeachingResources.

Carrow-Woolfolk, E. (1999). Test for Auditory Comprehen-sion of Language—Third Edition. Circle Pines, MN: AGS.

Chapman, R. S. (1995). Language development in childrenand adolescents with Down syndrome. In P. Fletcher & B.MacWhinney (Eds.), Handbook of child language (pp. 641–663). Oxford, England: Blackwell.

Chapman, R. S., & Hesketh, L. J. (2000). The behavioralphenotype of Down syndrome. Mental Retardation andDevelopmental Disabilities Research Review, 6, 84–95.

Chapman, R. S., Hesketh, L. J., & Kistler, D. J. (2002).Predicting longitudinal change in language productionand comprehension in individuals with Down syndrome:Hierarchical linear modeling. Journal of Speech, Lan-guage, and Hearing Research, 45, 902–915.

Chapman, R. S., Schwartz, S. E., & Kay-Raining Bird,E. (1991). Language skills of children and adolescentswith Down syndrome: I. Comprehension. Journal ofSpeech and Hearing Research, 34, 1106–1120.

Chapman, R. S., Seung, H.-K., Schwartz, S. E., & Kay-Raining Bird, E. (1998). Language skills of children andadolescents with Down syndrome: II. Production deficits.Journal of Speech, Language, and Hearing Research, 41,861–873.

Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. (2003).Applied multiple regression/correlation analysis for thebehavioral sciences (3rd ed.). Mahwah, NJ: Erlbaum.

DiSimoni, F. (1978). The Token Test for Children. Allen, TX:DLM Teaching Resources.

Dollaghan, C., & Campbell, T. (1998). Nonword repetitionand child language impairment. Journal of Speech,Language, and Hearing Research, 41, 1136–1146.

Dunn, L. M., & Dunn, L. M. (1981). The Peabody PictureVocabulary Test—Revised. Circle Pines, MN: AGS.

Dunn, L. M., & Dunn, L. M. (1997). The Peabody PictureVocabulary Test—Third Edition. Circle Pines, MN: AGS.

Dykens, E. M., Hodapp, R. M., & Evans, D. W. (1994).Profiles and development of adaptive behavior in childrenwith Down syndrome. American Journal of MentalRetardation, 98, 580–587.

Facon, B., & Facon-Bollengier, T. (1997). Chronologicalage and Peabody Picture Vocabulary Test performance ofpersons with mental retardation: New data. PsychologicalReports, 81, 1232–1234.

Facon, B., Facon-Bollengier, T., & Grubar, J.-C. (2002).Chronological age, receptive vocabulary, and syntaxcomprehension in children and adolescents with mentalretardation. American Journal on Mental Retardation,107, 91–98.

Facon, B., Grubar, J.-C., & Gardez, C. (1998). Chronologi-cal age and receptive vocabulary of persons with Downsyndrome. Psychological Reports, 82, 723–726.

Fazio, B. B., Johnston, J. R., & Brandl, L. (1993).Relation between mental age and vocabulary developmentamong children with mild mental retardation. AmericanJournal on Mental Retardation, 97, 541–546.

Gathercole, S. E., & Baddeley, A. D. (1993). Workingmemory and language. Hove, England: Erlbaum.

Gathercole, S. E., Willis, C., Baddeley, A., & Emslie, H.(1994). The Children’s Test of Nonword Repetition: A testof phonological working memory. Memory, 2, 103–127.

Hulme, C., & Mackenzie, S. (1992). Working memory andsevere learning difficulties. Hove, England: Erlbaum.

Jarrold, C., & Baddeley, A. D. (1997). Short-term memoryfor verbal and visuospatial information in Down’s syn-drome. Cognitive Neuropsychiatry, 2, 101–122.

Jarrold, C., Baddeley, A. D., & Phillips, C. (2002). Verbalshort-term memory in Down syndrome: A problem ofmemory, audition, or speech? Journal of Speech, Lan-guage, and Hearing Research, 45, 531–544.

Kaufman, K., & Kaufman, N. (1983). Kaufman AssessmentBattery for Children. Circle Pines, MN: AGS.

Kay-Raining Bird, E., & Chapman, R. S. (1994). Sequen-tial recall in individuals with Down syndrome. Journal ofSpeech and Hearing Research, 37, 1369–1380.

Kernan, K. T., & Sabsay, S. (1996). Linguistic and cogni-tive ability of adults with Down syndrome and mentalretardation of unknown etiology. Journal of Communica-tion Disorders, 29, 401–421.

Kirk, S. A., McCarthy, J., & Kirk, W. (1968). The IllinoisTest of Psycholinguistic Abilities. Urbana: University ofIllinois Press.

Kramer, P. E., Koff, E., & Fowles, B. (1980). Enactmentvs. picture choice tasks in studies of early languagecomprehension: When a picture is not worth a 1,000words. Psychological Reports, 46, 803–806.


Laws, G. (1998). The use of nonword repetition as a test ofphonological memory in children with Down syndrome.Journal of Child Psychology and Psychiatry, 39,1119–1130.

Lott, I. T., & McCoy, E. E. (Eds.). (1992). Down syndrome:Advances in medical care. New York: Wiley-Liss.

Marcell, M. M. (1995). Relationships between hearing andauditory cognition in Down’s syndrome youth. DownSyndrome Research and Practice, 3, 75–91.

Milenkovic, P. (1996). CSpeech (Version 4) [Computerprogram]. Madison: University of Wisconsin—Madison,Department of Electrical Engineering.

Miller, J. F. (1988). The developmental asynchrony oflanguage development in children with Down syndrome.In L. Nadel (Ed.), The psychobiology of Down syndrome(pp. 167–198). Cambridge, MA: MIT Press.

Miller, J. F. (1995). Individual differences in vocabularyacquisition in children with Down syndrome. Progress inClinical and Biological Research, 393, 93–103.

Miolo, G. (1998). Sentence comprehension in young children:Effects of phonological working memory, visual context andlinguistic complexity. Unpublished doctoral dissertation.University of Wisconsin—Madison.

Montgomery, J. W. (2000). Verbal working memory andsentence comprehension in children with specific languageimpairment. Journal of Speech, Language, and HearingResearch, 43, 293–308.

Rosin, M. M., Swift, E., Bless, D., & Vetter, D. K. (1988).Communication profiles of adolescents with Downsyndrome. Journal of Childhood Communication Disor-ders, 12, 49–64.

Seung, H.-K., & Chapman, R. S. (2000). Digit span inindividuals with Down syndrome and typically developingchildren: Temporal aspects. Journal of Speech, Language,and Hearing Research, 43, 609–620.

Tabachnick, B. G., & Fidell, L. S. (2001). Using multivari-ate statistics (4th ed.). Boston: Allyn & Bacon.

Tanenhaus, M. K., Spivey-Knowlton, M. J., Eberhard,K. M., & Sedivy, J. C. (1995). Integration of visual andlinguistic information in spoken language comprehension.Science, 268, 1632–1634.

Thorndike, R. L., Hagen, E. P., & Sattler, J. M. (1986).Stanford–Binet Intelligence Scale (4th ed.). Chicago:Riverside.

Received December 14, 2002

Revision received August 28, 2003

Accepted June 29, 2004

DOI: 10.1044/1092-4388(2005/013)

Contact author: Giuliana Miolo, Department of Communica-tive Disorders, University of Wisconsin—Whitewater,Whitewater, Wisconsin 53190. E-mail: [email protected]

Appendix. Sentence stimuli.

The yellow sheep kicks the black horse.The white chicken jumps over the black sheep.The red sheep kisses the yellow chicken.The white cow rides on the yellow sheep.The yellow cow chases the white chicken.The yellow chicken pushes over the white horse.The black cow stands on the red chicken.The yellow horse climbs over the black cow.The black sheep is kissed by the red horse.The red chicken is ridden on by the yellow horse.The red horse is kicked by the white sheep.The red cow is chased by the black chicken.The white horse is pushed over by the red sheep.The white sheep is jumped over by the red cow.The black horse is climbed over by the white cow.The black chicken is stood on by the yellow cow.

Sentence Comprehension in Adolescents With Down Syndrome and Typically Developing Children: Role of Sentence Voice, Visual Context, and Auditory-Verbal Short-Term Memory

Documents