Grammar in Discrimination,Imitation

 

Normal and Scrambled Grammar in Discrimination, Imitation, and Comprehensio Comprehension n Author(s): John Neil Bohannon III Source: Child Development, Vol. 47, No. 3 (Sep., 1976), pp. 669-681 Published by: Wiley on behalf of the Society for Research in Child Development Stable URL: http://www.jstor.org/stable/1128181 Accessed: 10-06-2015 02:47 UTC

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/  info/about/policies/terms.jsp  JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Wiley and Society for Research in Child Development  are collaborating with JSTOR to digitize, preserve and extend access to   Child Development. http://www.jstor.org

This content downloaded from 181.118.153.129 on Wed, 10 Jun 2015 02:47:46 UTC All use subject to JSTOR Terms and Conditions

 

Grammar n Discrimination, Normal and Scrambled ScrambledGrammar Imitation, and Comprehension John Neil BohannonIII EmoryUniversity Emory University

Grammar n Discrimination, mitation, and ScrambledGrammar BOHANNON,OHNNEIL, III. Normal and Scrambled CHILD 669-681. Previousstudies Previous studies suggested the possibilDEVELOPMENT, 976, 47, Comprehension. notbe etweennormal nd ndscrambled be ableto ableto discriminate etweennormal scrambledyntax.This This ity that5-7-year-old hildrenmaynot examinedhe examined he ndother other etween discriminationnd discrimination killswith kills with50 50 children study relationship syntax anguage each in kindergarten,irst, and secondgrade.In addition o the discriminationask, the children childrenwere were asked o imitate imitateand andshow showcomprehensionselecta criterionpicture rom roman an array fterhearinga stimulus normaland and scrambledgrammarentences. The resultsshowed resultsshowedthat thatthe entences.The the percentage f dissentence)of normal criminatorsthosewho whomade made ess than than5 errorson on the discriminationask) 5 errors ask)changed changed ignificantly ver age from he minority n n o the second 22%) kindergarten minority22%) 78%) grade.Whileminimal ifferences majority78%) majority existed between discriminators nd nondiscriminatorsn imitatingor comprehendingcrambled enchildrenwho who discriminated discriminatedetween randomand grammaticallyorrect entenceswere etweenrandom entenceswere superior tences, children for both imitation imitationand and comprehension f normal entences.

As unlikely as it may seem, 5- and 6-year-old children may be unable to tell the difference between normal and scrambled syntax. Previous studies have shown that this age child (a) has failed to express a listening preference for normal over scrambled stories (Bohannon & Friedlander 1973; Rileigh 1973), (b) has produced only grammatical prepositional phrases (to the store) even when the use of ungrammatical phrases (the store to) was modeled (Liebert, Odom, Hill, & Huff 1969; Odom, Liebert, & Hill 1968), (c) has imitated both normal and scrambled sentences to an equally poor degree (Frasure & Entwisle 1973; Weener 1971; McNeill, Note 1; Weener & Wright, Note

Fortunately, recent work (Bohannon 1975) has laid the artifact hypothesis to rest. This study assessed both syntax discrimination and sentence imitation, a measure gaining increasing acceptance for testing emerging language skills (Slobin & Welsh, Note 3). The results from first-, second-, and fifth-grade children (18 per grade) indicated that (a) some children did not discriminate normal from scrambled syntax in a successive discrimination task, (b) the ability to discriminate was a function of age (40% of first-grade, 75% of secondgrade, and 95% of fifth-grade children could learn the discrimination task with five errors or less), and (c) the ability to discriminate was related to

2), and (d) has failed to recognize grammaticalsengrammaticalsentences when presented in a list containing both normal and scrambled items (Vasta & Liebert 1973).

normal sentence imitations (r = .63, p< .01). The results further suggested that at least two variables were associated with the ability to imitate sentences: age and the ability to discriminate. Weener (1971) has suggested that increased age may provide increased memory capacity (such capacity was reflected in an increased ability to imitate scrambled sentences). Bohannon's data did not provide the opportunity to examine the relative effects of these variables because only four of his sample of 18 second graders failed to discriminate. Only by assessing a large number of second

The reason why the above mass of data has been ignored may be that the results were considered methodological artifacts. That is, 5- and 6-year-old children may have failed to distinguish the grammatical from the ungrammatical in certain contexts; but it was no more than a parlor trick, essentially unrelated to the mainstream of language development.

This study was done in partial ulfillmentof the requirements f the Ph.D. degree at the State Universityof University of New Yorkat Yorkat Stony Brook.The author authorwishes wishes to thank Dr. Gabriel GabrielMassaro, Massaro,assistant assistant Harborfieldschool EastSide School; oan choolDistrict;Dr. DonaldWhite, principal f EastSide superintendent f the Harborfields Kolombotovich, rincipal f Broadway chool;MauriceBeulieau,principal f WashingtonDrive School; andthe and the teachers,parents,andchildren andchildrenof ofthese heirassistance ndcooperationn the courseof these schools or heirassistance this study. Thanks Thanksare are also due Drs. FrancisPalmer,HerbertKaye, RogerSchwaneveldt, nd Aaron Lipton for their help in the preparation f the manuscript nd MeredithPublishing PublishingCorp., Corp., for their houldbe be sent to Dr. J. N. Bohannon, sentto permissiono use the pictures n LanguageLotto.Reprint equests hould of 30322. Department Psychology,Emory Emory University,Atlanta,Georgia

[Child Development, 1976, 47, 669-681. @ by the Society for Research in Child Development, Inc. All rights reserved.]

This content downloaded from 181.118.153.129 on Wed, 10 Jun 2015 02:47:46 UTC All use subject to JSTOR Terms and Conditions  

670 Child Development graders and identifying a substantial sample of nondiscriminatorsat nondiscriminators at that age level could the relationship between age and the ability to discriminate syntax be examined. At the same time, a substantial population of kindergarten children could be examined, thus extending the results to a

criminator and nondiscriminator children) x 2 (random and normal sentences) for multiple measures (sentence imitation and comprehension). Grades, ability to discriminate, and stimulus lists were between-subject factors, and syntax was a within-subject variable.

younger age group.

Subjects.--Children were from kindergarten, first, and second grades (50 children per grade) in two suburban Long Island school systems, for a total of 150 children (77 females, 73 males). Mean ages of the children were 5-9 (SD = 3.4 months), 6-9 (SD = 3.5 months), and 7-9 (SD = 3.7 months), kindergarten through second grade, respectively.

It was also the case that Bohannon used only sentence imitation as a measure of language development. Although some investigators argue that it is a valid assessment of production (McNeill 1970; Menyuk 1969; Slobin & Welsh, Note 3) and comprehension (Carroll 1972; McNeill 1970), imitation remains a measure of productive speech. If comprehension and production are different linguistic processes (Miller & Ervin 1964), their relationship to the ability to discriminate should be examined independently. A technique similar to the one used by Fraser, Bellugi, and Brown (1963) would provide an independent measure of comprehension. In this procedure the child listens to a sentence and selects one picture from an array as the best representative of the meaning of the sentence. Since the child need not speak to respond correctly, a measure of comprehension is achieved relatively independent of productive speech processes. Furthermore, if the stimulus sentences were both normal and scrambled, the comprehension of each type could be compared. The present study examined the relationships between the ability of children in kindergarten, and first, second grade to and discriminate between random and normal the ability to imsyntax itate and comprehend both random and normal sentence constructions. From previous work it was assumed that the ability to discriminate and age are positively related and that a large sample of children were needed to provide substantial nondiscriminatorsat at groups of discriminators and nondiscriminators each age level. Given those samples and independent measures of imitation and comprehension, it is hypothesized that (a) the proportion of children who could discriminate syntax should increase from ages 5 to 7, (b) there should be no differences between discriminators and nondiscriminators in or in scrambled senimitating and (c) discriminators comprehending should be superior tences, to nondiscriminators in the imitation and com-

Materials.-The stimuli for the discrimination task were identical with those used by Bohannon (1975). They consistecd of 24 different sentences, five to 15 words in length, and requiring 3-9 sec of presentation time. Half the sentences were in normal word order and half in scrambled word order. The sentences were recorded on tape cassettes (male narrator) narrator)by by a Fisher RC70 tape deck. The imitation stimuli were identical with those used by Weener and Wright (Note 2). Two lists of six items were generated, half the items with correct grammar and the other half with scrambled word order. Each item consisted of two five-word sentences. The normal grammar fiveword sentences had an adjective-plural nountransitive verb-adjective-plural noun structure (e.g., Good girls wear pretty coats, Strong lions climb tall trees). Sentences in the scrambled syntax condition had the word order randomized within sentences (e.g., Girls pretty good wear coats, Climb lions trees tall strong). Sentences were assigned to the lists so that no two versions of the same question occurred in the same list (see table 1). Thus no child heard both versions of the same sentence. The sentences were recorded by a male narratoron narratoron tape cassettes with a Fisher RC70 narrated with a lively tape deck. All stimuli were narratedwith intonation pattern and had durations of between 7 and 8 sec. The comprehension stimuli were: (1) 20 sentences, five to 15 words in length; (2) each sentence with a normal normaland and scrambled version; version;and and (3) 20 arrays of four pictures. Each array contained

prehension of grammatical sentences.

Method Design.--The design of this study was a 3 (kindergarten, first, and second grade) x 2 (dis?

only one picture which corresponded to the meaning of the stimulus sentence (see fig. 1). Pictures for the arrays were simple line drawings taken from Gotkin's Language Lotto.1 Twenty sentences, half normal and half scrambled, were randomly assigned to one of two lists with the stipula-

1966 by the Meredith Publishing Corp., New York. Used by permission.

This content downloaded from 181.118.153.129 on Wed, 10 Jun 2015 02:47:46 UTC All use subject to JSTOR Terms and Conditions  

John Neil Bohannon II

671

TABLE 1 SENTENCE IMITATION STIMULI

List 1 In ............. 2r ............. 3n ............. 4r ............. 5n ............. 6r .............

climb tall trees. Good girls wear prettycoats. Strong ions climbtall Soldiersbad Horses heavy big pull wagons. Soldiers bad brave men fight. Cool waterswash waterswash dry sand. Hungrybearseat sweet apples. Childrenhappy games playfunny. Toolsfresh fresh meat sharpcut. play funny. Tools Little deerjump high fences. Nice peoplehelp sick boys. Fish small ducks younglike. Hair long wind warm blows. List 2

In .............. 2r ............. 3n ............. 4r ........... 5n ............. 6r .............

toolscut cut fresh meat. Happy childrenplay funny games. Sharptools Girls prettygood wear coats. Climb ions treestall treestall strong. Big horsespull heavy wagons. Brave soldiers ight bad men. Watersdry cool sand wash. Eat apples sweet bearshungry. like small fish. wind blows long hair. duckslike fish. Warm Warmwind Young ducks Fencesjump high deer little. Nice help boys peoplesick.

tion that no two versions of the same sentence occur in the same list. Procedure. The order of presentation of the three tasks (discrimination, imitation, and com-

prehension) was randomized across children. All subjects were tested in a single test session lasting about 25 min by the same male examiner. Only one child left the testing situation prior to completion of all tasks. He was dropped from the study.

CIDQ2

FIG. 1.-Example of picture ersionof the stimulus entence: Niceboys picturearray array or comprehensionest. Normal ersionof beat louddrums beatloud drumsand andother other boys shake shakebells. bells. The random yntaxversionwas, Loud Loudbeat beatdrumsboys nice bellsand bells and othershake. other shake. Meredith boys Copyright PublishingCorporation, ll rightsreserved.

This content downloaded 181.118.153.129 on and Wed, 10 Jun 2015 02:47:46 UTC All use from subject to JSTOR Terms Conditions  

672 Child Development Discrimination procedure.--Children were tested individually and instructed in a conversational manner:

During testing, the imitation stimuli were delivered by a Sharp cassette tape deck, and the subjects' responses were recorded on a Fisher

We'regoingto playa listeninganda guessing guessinggame gameand talkdifferdifferI'lltell I'll about t. You Youknow know hat peopletalk tell you all about Not do sound but some different, ently. only people or sentencesto sentences to use different wordsor words say the people may samething. I have twofriends,Norman nd Ralph,who talkdifferently. went out and wrote downsome down some of the wentout things these guys have said and I put them on tape. [Experimenteroints [Experimenter oints o the tape recorder.] 'mgoing o play these thingsfor you, one by one, and you have to makethe the game easier, I have an guess who said it. To make Normanhas has said and someexampleof something hat Norman that has said. Listen thing Ralph closely. Here is Norman. [Experimenterplays the recordedexamplesentence, Mother old you to washyour hands. ]Did you yourhands. hear that? That was Norman.Now here's how Ralph wouldsay would saythe the same samething. thing. Here Hereis is how howRalph Ralphwould wouldsay say that. [Examinerplays the recorded sentence, Told mother you hands to your wash. ]Did you hear that? Thatwas That was Ralph.Now let's play the game.

RC70 cassette tape deck for later transcription transcriptionand and scoring. Stimulus delivery depended upon the child's responses in that subsequent items were not presented until the child finished responding to the previous stimulus. Responses were transcribed and scored using the Weener (1971) method. Subjects' scores consisted of the number of words imitated regardless of word order.

Children were presented the stimulus sentences one by one via a Sharp cassette tape deck and were required to respond Norman or Ralph to each stimulus. All subjects received appropriate feedback ( right or wrong ) after each response. The children were tested until they gave eight consecutive correct responses or until they completed 24 trials. They were classified as discriminator if they achieved eight within 24 trials consecutive or made less trialsor responses than five errors. Imitation procedure.-Subjects were seated in front of an acoustical speaker and a microphone. Instructions were delivered in a conversational manner:

Comprehension procedure.--Children were seated in front of the experimenter and were instructed in a conversational manner: We'regoingto playa picture pictureguessing guessinggame. game. I'm going to say a sentence and I wantyou to pointto the picture sentenceand that best goes with that sentence. Now look at these pictures the firstpractice rray onsistedof pictures f a coat,a puppetshow,a potholder,anda stringof beads]. What picturewould you pick if I said, Where s the coat. After he child responded, he experimenter evealed he secondpractice rray onsisting f picturesof a doll, some bells, some blocksand blocksandsome cookies.]Now what picturewould you pick if I said, Everyone ikes knowhow how to playthe cookies. Now, do you thinkyou know game?Good. Let's begin. No child needed extra instruction or failed to perform the practice items. Therefore, upon completion of practice, each child began the comprehension test of 20 items. Half the children received stimulus sentence list 1 and half received list 2. Prior to sentence stimulus delivery, children were allowed to view the appropriate picture array. After stimulus narration, subjects were allowed as much time as they needed to respond. If a child stated that he/she did not know the answer, he was encouraged to guess. Regardless of the response, the experimenter responded positively

We'regoing to playa listeningand a talking anda talkinggame. game.The to to is listen here way play [experimenterpoints to into here [experimenter oints to talkinto speaker]and to talk the microphone].Whenyou hearsomething, wantyou to say the same thing. Thatis the game. I wantyou to listen [experimenter gain points to the speaker]and gainpoints then to saythe samething [experimenter gainpoints gainpoints o the microphone].Don'tworry f you can't remember ll can'tremember of it. Just Justdo do the best you can. Do you think thinkyou you know how to play?Good. Let's practice.What if you heard dog. What would you do? Depending upon the child's response, the experimenter appropriately instructed the child to speak louder or talk into the microphone. Pretraining continued with the delivery of two other practice stimuli, the big dog and big dogs chase little cats. When the experimenter was satisfied with the child's knowledge of the procedure, testing began. Half the children received list 1 and half list 2.

and warmly. A correct response was defined as the selection of the picture corresponding to the meaning of the normal syntax version of the stimulus sentence, regardless of delivered stimulus word order. Subject scores consisted of the number of correct picture selections.

Results Discrimination.-On the basis of five errors or less to criterion performance, the kindergarten children yielded 11 discriminators (22%)versus 39 nondiscriminators; first grade provided 29 discriminators (58%) versus 21 nondiscriminators; and second grade provided 39 discriminators (78%) versus 1ii nondiscriminators. Age and the ability to discriminate syntax varied positively and in the predicted direction, X2(2) = 31.12, p < .001.

This content downloaded from 181.118.153.129 on Wed, 10 Jun 2015 02:47:46 UTC All use subject to JSTOR Terms and Conditions  

Bohannon III 673 John Neil BohannonIII The relationship between the ability to discriminate syntax and age on the imitation and comprehension of normal and scrambled sentences is examined below. imitation scores were Imitation. -Sentence Imitation.-Sentence analyzed by two separate 3 (grades) x 2 (discrimination) x 2 (lists) x 2 (syntax) ANOVAs. Grade, discrimination, and lists were between-subject effects and syntax was a within-subject effect. The first ANOVA, a least-squares analysis, yielded several negative sum squares because of the abrupt increase in the proportion of discriminators over age. The grade x discriminator (G x D) interaction was influenced most, with the result that all factors depending upon the G x D factor (G x D x syntax, G x D x list and G x D x list x syntax) were rendered meaningless. The error terms were inflated as well and made the analysis fairly conservative. It should be noted that the sum squares of the other factors in the analysis not factorsin dependent on the G x D interaction were unaffected. Therefore, as a check upon the significance levels of the affected factors another unweighted means ANOVA was computed on the means of the G x D x L x S interaction. This analysis treated each of the 24 means as a single score, thus overestimating the four-way interaction and slightly underestimating the lowerorder factors. For the means for the example, grade main effect in the least-squares analysis are kindergarten = 3.88 words per sentence, first

3. Discriminator children imitate better than nondiscriminatorchildren nondiscriminator children (means = 5.02 vs. 4.08 words per sentence). 4. List 1 was easier to imitate than list 2 (means = 4.84 vs. 3.86 words per sentence). The significant grade x syntax interaction, F(2,138) = 14.00, p < .01, subsequent t-test comparisons revealed: 1. There were no significant increases over age in the random syntax imitations (means = 3.27, 3.26, 3.83 words per sentence, for kindergarten, first, and second grade, respectively). 2. There was a significant (p < .01) increase in normal syntax imitations between kindergarten and first grade. First and second graders were not significantly different (means = 4.49, 5.98, 6.63 words per sentence for kindergarten, first, second grade, respectively). The significant discriminator x syntax interaction, F(1,138) = 24.40, p < .01, and subsequent t-test comparisons revealed: Discriminatorchildren 1. Discriminator children were superior (p < .01) to nondiscriminators on normal syntax imitations (means = 6.46 vs. 4.85 words per sentence, respectively). 2. But there were no significant differences between discriminators and nondiscriminators when imitating scrambled sentences (means 3.57 vs. 3.32 words per sentence). There were no

grade = 4.62, and second grade = 5.23. The grade means in the unweighted means analysis are kindergarten = 4.04 words per sentence, first grade = 4.56, and second grade = 5.12. Even with this reduction of variance, the pattern of significant effects is the same for both the leastsquares analysis and the unweighted means analysis. Therefore, only the F ratios of the more appropriate least squares will be reported. The least-squares analysis revealed significant main effects for grades, F(2,138) = 11.89, p < .01; discrimination, F(1,138) = 16.82, p < .01; lists, F(1,138) = 5.32, p < .05; and syntax, F(1,138) = 266.59, p < .01. The results of individual t-test comparisons show: 1. Second graders imitate all sentences better (p < .01) than first-grade and kindergarten children (means = 5.23, 4.62, 3.88 words per sentence, respectively). First graders were also superior (p < .01) to kindergarten children. 2. All children imitate normal syntax sentences better than scrambled sentences (means = 5.80 vs. 3.45 words per sentence).

other significant effects. Comprehension. The comprehension scores were analyzed by two 3 (grades) x 2 (discrimination) x 2 (lists) x 2 (syntax) ANOVAs, with grade, discrimination, and lists the between-subjects factors and syntax the within-subject factor. The effect of the changing number of discriminators and nondiscriminators over age on the sum squares of the grade x discrimination interaction in the least-squares analysis was similar to that in the imitation The was dealt with in the same analysis. problem the spurious sum manner by checking squares by means of an unweighted means analysis. The pattern of significant effects was identical over both analyses; therefore the more appropriate least-squares F ratios will be reported. The analysis revealed significant main effects for grades, F(2,138) = 26.27, p < .01; discrimination, F(1,138) = 78.54, p < .01; lists, F(1,138) = 9.99, p < .01; and syntax, F(1,138) = 242.29, p < .01. There were no other significant effects. Further examination of the main effects using t-test comparisons revealed:

This content downloaded from 181.118.153.129 on Wed, 10 Jun 2015 02:47:46 UTC All use subject to JSTOR Terms and Conditions  

674 Child Development 1. Second-grade children (mean = 8.26 items) comprehended significantly (p < .01) more items than both first-grade (mean = 7.73 items) and kindergarten children (mean = 6.81 items), and first-grade subjects were significantly (p < .01) better than kindergarten subjects. 2. Discriminator children (mean = 8.29 items) comprehended more items than nondiscriminator children (mean = 6.83 items). 3. The children's performance on list 2 (mean = 7.86) was slightly superior to their performance on list 1 (mean = 7.34 items). 4. All children comprehended more of the normal syntax items (mean = 8.47 items) than random syntax items (mean = 6.74 items). Discussion Discrimination and imitation.--Few 5-yearolds and most 7-year-olds discriminate between normal and scrambled grammar under the conditions of this study. The age at which half the children can discriminate appears to be 6 (first (firstgrade). grade). This finding validates the author's previous work (Bohannon 1975). Indeed, the comparative data

occurred with age among nondiscriminators. This is clearly not the case (see fig. 2). One possibility does exist for the original hypothesis. Some children may be able to discriminate syntax but fail to respond correctly on the discrimination measure for other reasons. The reverse is unlikely as chance acts against a true nondiscriminator'smisclassification discriminator's misclassification as a discriminator. Thus, if a percentage of those classified as nondiscriminators may have differentiated normal from scrambled syntax but responded incorrectly, the increase in normal syntax imitations of nondiscriminatorsover criminators over age seen in figure 2 could be exSeveral imitation scores misclasplained. highwould not greatly byinfluence sified discriminators the mean score for kindergarten nondiscriminators (N = 39). But a few such scores in the smaller sample of nondiscriminators in the first (N = 21) and second grade (N = 11) could inflate those means to produce results similar to figure 2. To explore the possibility that some children may have been misclassified as nondiscriminators, and to assess the effect of more explicit instructions on the ability to discriminate, additional subjects were exposed to a modification of the original

are highly consistent. The first study, conducted during the first half of the school year, found that 40% and 71% of the first and second grades, respectively, could discriminate. This study conducted with children who were on the average 4 months older than the children of the first study found 58% of first graders and 78% 58%of of second grad78%of ers could discriminate syntax. These combined data reflected the behavior of 204 children in three different schools. The children were predominantly middle class, and the schools were almost certainly above average in their educational program. Whether SES or the caliber of school systems would alter the percentages of discriminators by gtade is, of course, an empirical question which must be tested by further research.

instructions on the discrimination task. These children were given the original measure of discrimination until 10 were identified as nondiscriminators by the original criteria. Eight were kindergarteners and one each was in the first and second grades. After these subjects had completed the original measure and were unable to discrimimaterials but nate, they were rerun with the same materialsbut with different instructions. They were told, Norman talks correctly like your teacher, and you and I do; he makes sense [and the tape of Norman was played again]. Ralph talks funny, mixes up his words and does not talk like your teacher [and the tape of Ralph was played again].

Pertinent to the relationship between the ability to discriminate and sentence imitation performance, figure 2 emphasizes a finding of consequence: the ability to discriminate syntax relates only to the imitation of grammatical sentences. Discriminators were no better at imitating scrambled sentences than nondiscriminating children.

10 children (the second grader and two kindergarteners) reached the original criterion of discriminating, but seven (the first grader and six kindergarteners) did hot. Those results suggest (1) that some children may have distinguished word order but responded incorrectly, and were misclassified as nondiscriminators, and (2) that the inability to discriminate found with the original measure seems relatively impervious to more explicit instructions.

However, the grade x syntax interaction shows that other variables operate in the ability to imitate. This finding is inconsistent with predictions by the author that the imitation of normal syntax would be exclusively a function of the ability to discriminate. Had that been found, no increase in imitations of normal syntax would have

With the additional instructions, three of the

Thus, the increased ability of nondiscriminators to imitate normal syntax with age may be an artifact. Some children might have distinguished word order but responded incorrectly, with the result that they were misclassified. Imitation and comprehension tests were not given to the addi-

This content downloaded from 181.118.153.129 on Wed, 10 Jun 2015 02:47:46 UTC All use subject to JSTOR Terms and Conditions  

John Neil BohannonIII 675 Discriminator

100

O Nondiscriminator

90 w

80 70 Normal Syntax

7) 0

-) *

60

8

50

o

40

0

Scrambled Syntax

P

0

20

10 1

Kg

2

GRADES FIG. 2.-Imitation scores fordiscriminator for discriminatornd nd nondiscriminatorhildren hildrenover over grades data do not additionaldata tional sample of 10, so these additional about possible differences beinformationabout provide information tween the seven and three with respect to their sentence imitation ability. The inability of the seven to discriminate even with instruction appears to validate the original measure and to speak to the strength of the phenomena. It is reasonable to assume that some children are on the threshold of being able to discriminate and that, while on that threshold, they are more aware of word-order differences than those who have not approached it. But for seven to be uninfluenced by the instruction implies that the inability to discriminate is a more fundamental phenomena than may have been previously supposed. discussing the resultsComprehension.--Before of the comprehension test some caveats should be noted. The measurement tool was chosen for its independence from assessments of

productive speech. An identical comprehension procedure has been shown to relate to other aspects of language development like spontaneous speech (Fraser et al. 1963), but language comprehension is certainly more complex than what this measure alone reflects. Spearritt (1962) suggests listening comprehension involves attentional, motivational, auditory, memory, and semantic factors. The pick-apicture method may only involve a subset of these, such as attentional, auditory, and semantic factors. Syntax may not even have been a necessary factor for responding correctly to all items. Correct performance on the sentence, Big pianos make nice music, could have been achieved by comprehending the word piano and matching that with the picture of a piano in the array. Other sentences, like Mother says to put cups on top of saucers, require comprehension of a grammatical relationship rather than single words, particularly

This content downloaded from 181.118.153.129 on Wed, 10 Jun 2015 02:47:46 UTC All use subject to JSTOR Terms and Conditions  

676 Child Development since the picture array arrayprovided provided pictures of both a cup on a saucer and a saucer on a cup. Syntax was presumably more critical for correct interpretation of the second sentence. The pick-a-picture method may elicit responses dependent upon different levels of comprehension-some grammatical, some lexical. With this qualification on the measure, the results of the comprehension procedures can be examined. The measure of comprehension indicated that children who discriminate syntax comprehend both normal and scrambled sentences better than nondiscriminators (see fig. 3). Were it not for the

ity in the imitation of scrambled sentences. Would not brighter or more attentive children reproduce more words from random word strings? The answer may be in the comprehension test itself. An analysis of discriminators' and nondiscriminators' performance on the easy and difficult items of the test was conducted to determine whether discriminator children performed better on both. If they are brighter or more attentive, should they not perform better on both easy and difficult test items? Those items upon which children made 40 or more errors were classified classifiedas as difficult. Those with 31 or less errors were classified as

easy.

scrambled sentence imitations one might suggest that discriminators are simply brighter or more attentive, and the ability to discriminate is a function of these general factors. For that hypothesis to hold, however, a reasonable explanation must be found for the discriminators'nonsuperior-

For difficult items (see fig. 4) the comprehension results are very similar to the results of the imitation test (fig. 1). Discriminators are superior to nondiscriminators only on the normal syntax items (means = 84% vs. 61.5% items correct,

100 Normal

90

Syntax

80

0)

70

0 U

60

ct

50

Scrambled Syntax

H 4J

0)

40'

30'

Discriminator Nondiscriminator

20

10

K9

1

2

GRADES FIG. 3.--Comprehension cores fordiscriminators for discriminatorsnd nd nondiscriminatorsver grades or all 20 items

This content downloaded from 181.118.153.129 on Wed, 10 Jun 2015 02:47:46 UTC All use subject to JSTOR Terms and Conditions  

John Neil BohannonIII 677 100 90 80

Normal Syntax

70

U 0)

u0 U U)

1

60 50

Scrambled Syntax

H

U) U

40

NJ

30 Discriminator Nondiscriminator

20 1

2

1

Kg

GRADES FIG. 4.-Mean comprehension cores coresfor for discriminator nd nondiscriminatorhildrenover gradesfor eight difficult tems.

t(148) = 6.99, p < .001). No significant differences existed for the comprehension of difficult scrambled items (means = 48.5% vs. 45% items correct for discriminators and nondiscriminators respectively, t(148) = .88, N.S.). It was only the easy scrambled items (fig. 5) on which discriminators were superior and which provided the difference between groups when all items were analyzed together. Presumably, discriminators should have been superior on both easy and difficult items if brightness or attentiveness was the cause. Figure 5 reveals a ceiling effect. Discriminator children responded correctly almost 90%-97% of the time in both normal and scrambled easy items. Only the nondiscriminators' performance on the scrambled items was under 80%. Performance on difficult (fig. 4) and easy (fig.

discriminators' re5) items was similar except for discriminators'responding to random syntax. Discriminators comprehended more easy normal syntax items (means = 97% vs. 85% items correct, t(148) = 6.68, p < .001) and more easy scrambled syntax items (89.5% vs. 72% items correct, t(148) = 6.68, p < .001) than nondiscriminators. The last finding is in contrast to the discriminators' performance on difficult scrambled items, where they were no different from nondiscriminators. An examination of the imitation (fig. 2) and comprehension of difficult items (fig. 4) reveals three levels of performance. Poorest was the imitation and comprehension of scrambled sentences discriminators and nondiscriminators. Inby both discriminatorsand termediate was the performance of nondiscriminators on normal syntax sentences. The highest level was attained by discriminator children in com-

This content downloaded from 181.118.153.129 on Wed, 10 Jun 2015 02:47:46 UTC All use subject to JSTOR Terms and Conditions  

678 Child Development 100

Normal Syntax

90

Scrambled Syntax

80

44)

70'

0

60

U

50

Discx imina tox Nondiscriminator

H

40

30 20

10 1

Kg

2

GRADES for discriminators nd nondiscriminatorsver grades or the easy items coresfor discriminatorsnd FIG. 5.--Comprehension cores prehending and imitating grammatical sentences. Three different processing levels or strategies may be involved. The first and poorest strategy is exhibited by nondiscriminator children on all scrambled sentences and by discriminators on difficult scrambled sentences. The chil children dren seemed to use isolated words as the sole basis for response. For the comprehension task, some individual words were recalled and applied to the picture arrays. Occasionally this was correct, which accounts for responses above the level of chance. Support for this is found in the imitation test where children would often invent sentences, only one or two words of which came from the imitation stimulus. The second process was used by nondiscriminators when processing grammatical sentences. This may be similar to the first strategy with the addition of extra cues supplied by the structure of the sentence. These cues may serve

only as aids to lexical memory, since nondiscriminators seem to gain little from sentence structure and cannot detect the absence of grammatical cues on the discrimination test. Finally, discriminator children processing grammatical constructions use their knowledge of sentence structure to comprehend. sentences far better than simple memory alone can provide. It is the author's opinion that they alone are able efficiently to combine linguistic skills with memory for better performance. The discriminators' performance on the easy scrambled sentences may have been a variation of the third strategy. These children, when confronted with scrambled sentences, are aware of the sentences' ungrammatical nature. They actively attempt to recover the structure of the mixed up lexical elements. When the sentence is normalversion version complex, the meaning of long or its normal the original sentence is not attained. Thus, their

This content downloaded from 181.118.153.129 on Wed, 10 Jun 2015 02:47:46 UTC All use subject to JSTOR Terms and Conditions  

III 679 BohannonIII John Neil Bohannon imitation and comprehension performance on the long scrambled sentences was no better than nondiscriminators'. But when the sentences were shorter and grammatically simple as in the easy comprehension items, this strategy regained most of the meaning of the sentences. Since nondischildren seem to be unaware of the uncriminatorchildren criminator grammaticality of the scrambled sentences, they may not have tried to restructure the scrambled items, and thus their performance was relatively poor. The comprehension strategy hypothesis conforms to the data and is preferable to a general factor explanation. If discriminator children were brighter or more attentive their comprehension performance on the difficult scrambled items should have been superior to that of nondiscriminator children. Indeed, the similarity between the pattern of imitation (fig. 2) and comprehension of difficult items (fig. 4) is striking. It must be concluded that discriminators are distinin from nondiscriminators guishable primarily their dealings with structured language. Theoretical discussion.-Why do 6-year-old children (nondiscriminators), considered language sophisticates by some (Lennenberg 1967; McNeill 1970), fail to notice anything peculiar about sentences like, Told mother you hands to your wash? There are at least two hypotheses that would correctly identify such sentences during the Norman-Ralph test: (1) Label ungrammaticalsentences with Ralph. (2) Label incomprehensible sentences with Ralph. Posttest questions about the nature of the discriminative stimuli revealed that discriminator children selected either hypothesis (1) or (2) to direct their responding. But it must be noted that random responding does not exclude the possibility that nondiscriminators tried these hypotheses reimmaturereduring testing. It is conceivable that immature ceptive grammars would be flexible enough to permit considerable input distortion (a point to be dealt with later). It is also possible that partial comprehension is subjectively more similar to fragmentary comprehension (nondiscriminator comprehension scores, fig. 3) than full comprehension versus partial comprehension (discriminapartialcomprehension tor scores, fig. 3). If either of the above is true, it would have rendered the normal and scrambled sentences sufficiently similar for the nondiscriminators to respond randomly. Yet what is really puzzling is the relationship between this linguistic awareness of word-order differences and the ability to imitate and comprehend structured grammar. Relevant suggestions do exist in the litera-

ture. There are some (DeVilliers & DeVilliers 1972; Gleitman & Gleitman 1970; Gleitman, Gleitman, & Shipley 1972; Shipley, Smith, & Gleitman 1969) who state that for 2- and 3-year-old children just such a relationship between recognition of grammaticality and other language skills does exist. They agree with generative grammarians (e.g., Chomsky 1965) in assuming that the child's recognition of grammaticality in his own and others' speech is essential for language acquisition. It is not enough for the child to follow the rules of grammar; the child must also be able to reflect on those rules (Gleitman et al. 1972). A corollary to the above would predict that a child who recognizes grammaticality is at a higher level of language development than one who does not recognize this distinction. On this Gleitman et al. (1972) and the present study agree. But these authors insist children as young as 2 and 3 years can distinguish grammaticality in sentences they hear. If the Gleitman et al. (1972) position is strictly interpreted, the nondiscriminators in this study should have acquired only the rudiments of language. This is clearly not the case, as nondiscriminator children did perform significantly better on normal syntax sentences than scrambled sentences. Perhaps children can acquire grammar in productive speech without being aware of the rules of grammar they use. Bloom (1974) has suggested that children may practice with grammatical structures in productive speech before complete mastery of those structures. This implies a dual (receptive and productive) rule system of language. It accounts for the nondiscriminators'ability to use structure in productive speech yet remain unaware of the grammatical/ungrammatical distinction in receptive speech. Thus, a child's ability to discriminate syntax may not be a determining factor in language acquisition but may be a result of attaining a certain level of grammatical skill. A similar suggestion has been made for concept acquisition. Vygotsky (1965) suggests a preintellective verbal stage in cognitive development when the child acquires and uses concepts before being aware of the rules determining them. He argues that a concept can become conscious only it has become of a [after] part system .... .. [and] leads...to becoming conscious of our operations their mastery (p. 92). In this way, Vygotsky sees awareness as both a symptom and determinant. After the child has used concepts he becomes aware of the rules he used in their formation. He will subsequently use that awareness in the formation of new concepts and in understanding old

This content downloaded from 181.118.153.129 on Wed, 10 Jun 2015 02:47:46 UTC All use subject to JSTOR Terms and Conditions  

680 Child Development ones better. Vygotsky argues further that the school experience facilitates focusing on the rules of concept formation and accelerates awareness of those rules. If the rules of syntax are equated to concepts, Vygotsky's position about concept acquisition appears relevant. The age discrimination (or to Vygotsky) is 6 of awareness years, when the subjects of the present study entered their firstyear of serious language instruction. The research on the acquisition of linguistic skills has to date largely emphasized the productive speech of the- preschool child (Brown & Fraser 1964; Lennenberg 1967). Only a few investigators (Fodor, Bever, & Garrett 1974) have suggested that the comprehension process may be different from processes used in productive speech. Fewer still (Bloom 1974) have suggested that productive skills may precede receptive skills. The data reported here lend substance to the argument for separate processes. A single-process theory must account for these facts: (1) children who produce grammatical sentences may not be able to discriminate between the presence or absence of grammar n sentences they hear, (2) children who can discriminate derive more information from grammatical constructions in both comprehension and imitation, and (3) the age at which most children become discriminators exceeds considerably that at which most researchers claim the rules of grammar are acquired. A dual- or multiple-process theory would appear to conform better to those facts. A wealth of data from Brown (1973) and others shows that children readily utilize many grammatical sentence constructions in productive speech. Data from the present study suggest that in the receptive mode of language the rules of syntax do not completely control sentence comprehension. Two grammatical processes, reception and production, seem to be required. A theory which restricts the child's understanding to grammatical structures already mastered for productive speech must be questioned. If this were the case, the holophrastic child could not comprehend structures exceeding one word in length. There must be some heuristic process which allows the child to obtain meaning from complex adult utterances about whose structure the child knows but little. Figure 3 shows that both discriminator and

sion must have depended at least in part on some process other than that of productive syntax-for example, the child's knowledge of semantic relationships and general world knowledge. Fodor et al. (1974) have argued that there is more to sentence comprehension in adults than can be explained by the constructs of generative (productive) grammar. If true for mature speaker/ listeners, it is probably more true for children. The argument for separate processes for production and listening seems more deserving of serious consideration than has been the case until now. In summary, this study investigated the correlates of the 5-7-year-old child's ability to recognize grammaticality in the receptive mode of language. Discriminator children, those who readily distinguished between the grammatical and ungrammatical, were consistently superior to nondiscriminatorsin discriminators in the imitation and comprehension of normal syntax. Yet there were minimal differences between these groups when asked to imitate and comprehend randomword random word strings. The results suggest an important change in language development that occurs around the age of 6 years. This change seems to enable discriminators to imitate and comprehend roughly twice as much structured grammar (relative to baseline performance on scrambled grammar) as nondiscriminator children.

Reference Notes 1. McNeill, D. Developmentof the semantic ystem. Unpublishedmanuscript, enter orCognitiveStud1965. ies, Harvard University, 2. Weener, P., & Wright,P. The effects effectsof of linguistic structureon structure on the recall recallof of verbalmessages messagesby by children. Paper at the of the Paperpresented presented meeting Society for Research n Child Development, Development,Philadelphia, Philadelphia, April 1973. 3. Slobin, D., & Welsh, C. Elicited mitationas a research oolin ool in developmental sycholinguistics. npublishedmanuscript,Department Departmentof Psychology, of University California,Berkeley, Berkeley,1967.

References Bloom, L. Talking, understanding nd thinking. In R. Schiefelbusch& Schiefelbusch& L. Lloyd (Eds.), Language

perspectives:cquisition, etardation, nd perspectives:cquisition, ndintervenintervention. Baltimore:University Park, 1974. UniversityPark, Bohannon,J. J. N. The relationship etween syntaxdiscriminationand crimination and sentence imitation in children. Child Development,1975,46, 1975, 46, 444-451. & B. Z. The effect of Friedlander,B. Bohannon,J. Bohannon, J. N.,