This article was downloaded by: [University of New Hampshire] On: 17 February 2015, At: 17:26 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

The Journal of Psychology: Interdisciplinary and Applied Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/vjrl20

Inference of Word Meaning from Syntax Structure by Normal Children and Retarded Adolescents a

Carolyn W. Lieber & Herman H. Spitz

a

a

Edward R. Johnstone Training and Research Center, Bordentown , New Jersey, USA Published online: 02 Jul 2010.

To cite this article: Carolyn W. Lieber & Herman H. Spitz (1976) Inference of Word Meaning from Syntax Structure by Normal Children and Retarded Adolescents, The Journal of Psychology: Interdisciplinary and Applied, 93:1, 3-12, DOI: 10.1080/00223980.1976.9921368 To link to this article: http://dx.doi.org/10.1080/00223980.1976.9921368

PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or

indirectly in connection with, in relation to or arising out of the use of the Content.

Downloaded by [University of New Hampshire] at 17:26 17 February 2015

This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions

Published as a separate and in The Journal of Psychology, 1976 93, 3-12

INFERENCE OF WORD MEANING FROM SYNTAX STRUCTURE BY NORMAL CHILDREN AND RETARDED ADOLESCENTS*

Downloaded by [University of New Hampshire] at 17:26 17 February 2015

Edward R . Johnstone Training and Research Center, Bordentown, New Jersey

Children from five to 12 years of age ( N = 136 boys and girls) and institutionalized retarded adolescents ( N = 30 boys and girls) were given a task developed by Brown to test ability to infer nonsense word meanings from syntactic clues. Response accuracy for the normal children was consistently low from five years up to eight years of age, after which it increased steeply with increasing age. However, accuracy levels were markedly below those reporl.ed by Brown and raise questions concerning his original finding that young children are largely successful in using the grammatical context of new words to guess their meaning. Performance of the retarded adolescents did not parallel that of the nonretarded children, and on two of the three parts of speech their performance fell below mental age expectations, a result attributed to their deficient generalization ability.

One of the best known experimental studies of the linguistic competence of young children is Brown's (3) investigation of the semantic implications of part-of-speech membership. Having determined that certain form classes have definite semantic implications in child language (for example, in children's speech verbs usually name observable actions while nouns refer to concrete objects), and that nursery school children consistently make grammatical distinctions between these various form classes in their spontanecus speech, Brown devised an ingenious technique for investigating whether these distinctions are used by the child in the acquisition of semantic concepts.

* Received in the Editorial Office on January 27, 1976, and published immediately at Provincetown, Massachusetts. Copyright by The Journal Press. 1 The authors are most grateful to John Hydock, Principal of the Columbus Elementary School, Columbus, New Jersey, for his gracious assistance in providing space and arranging for the participation of students; and to Mary Anne Brzoska for successfully executing the difficult task of drawing pictures of unknown objects. Requests for reprints should be sent to Herman H. Spitz at the address shown at the end of this articlc.

Downloaded by [University of New Hampshire] at 17:26 17 February 2015

4

JOURNAL OF PSYCHOLOGY

As Brown noted, a new word is commonly introduced to the young child not in isolation but in the context of a sentence or phrase which makes explicit its part-of-speech membership. Brown suggested that the concepts "dog" and "running" might be introduced by parents as follows: "Look at the dog. See him running." If the attentive child has grasped the semantic inlplications of the noun class and the verb class, he or she would be likely to realize, upon hearing "dog" in the context of the first sentence, that it probably denotes a concrete object of defined shape and size. I n the second sentence, the grammatical framework in which "running" is presented would indicate that the word refers to some motion or observable action. Does the young child actually use grammatical cues in such a manner? In exploring this question, Rrown used nonsense word stems which were assigned specific part-of-speech membership (verb, mass noun, or count noun) by virtue of granlmatical treatment in an introductory sentence. The task of the child was to guess the meaning of the nonsense word by pointing to a picture showing either an action, a mass substance, or a particular object. Rrown reported that the 3- to 5-year-old children in his exljeriment were largely successful in this task. The results may be interpreted as indicating that the a filtering mechpart-of-speech membership of an unknown word operates anism which directs the child's attention to relevant aspects of the immediate environment. If the young child does in fact use syntactic cues to make inferences about word meaning, this ability would seem to be a valuable tool in the rapid acquisition of language during the preschool years. Brown's study has been widely cited in both professional (e. g., 4, 8 ) and popular ( e . g., 5) accounts of the linguistic competence of the young normal child. The language development of the mentally retarded child is characterized by many writers (e. g., 6 ) as a process that follows the same course as that of normal children, although at a slower rate. Retarded children are typically reported to have smaller vocabularies and more primitive grammatic constructions than normal children of the same chronological age: however, no structures are found in the language of retarded people that do nnt also occur in the course of the language development of intellectually normal children [see Yoder and Miller (10) for a review]. I t seems possible, on the basis of related research, that a basic deficit in the ability to generalize about rules plays a role in the slower language development of the retarded. Previous studies ( 7 , 9 ) using Berko's ( 1) morphology test found that educable mentally retarded (ERIR) Ss were less able than normal children to generalize rules for word inflection from familiar to nonsense words.

=

CAROLYN W. LIEBER AND HERMAN H . SPIT2

5

Downloaded by [University of New Hampshire] at 17:26 17 February 2015

Brown's (3) experiment with nonsense words provides another format for the investigation of one aspect of rule generalization: i. e., the ability to theorize about the semantic implications of part-of-speech membership. The purpose of the present study was to compare the relative abilities of E M R and intellectually normal children of various ages to make inferences about the meaning of unfamiliar words from the context in which the words are introduced.

1. Subjects Testing of a pilot group of middle-class preschool children, mean CA = 4.64 years, resulted in such uniformly poor performance that it was decided to begin testing a t the kindergarten level. The normal Ss consisted of 136 children enrolled in grades K to 6 in a semirural elementary school in central New Jersey. Since children had been assigned to classes on a random basis, we te:jted one entire class a t each grade level. The 71 girls and 65 boys (all white) ranged in CA from 5.16 to 12.75 years. They were divided, by age, into four approximately equal-sized groups: Group 1, 17 girls and 17 boys, mean age = 6.23 years, S D = .62: Group 2, 17 girls and 16 boys, mean age =I 8.19 years, S D = .49; Group 3, 19 girls and 16 boys, mean age = 9.84 years, S D = .47: and Group 4, 18 girls and 16 boys, mean age = 11.45 years, S D = .46. The retarded group was composed of 30 E M R adolescents (17 white, 13 black) from a state residential institution (Johnstone). There were 15 girls and 15 boys, mean CA = 17.11 years, S D = .29, mean IQ (from individually administered Wechsler Intelligence Scales) = 57, S D = 8, and mean MA = 9.18 years, S D = 1.29. 2.

Materials

Materials consisted of three sets of pictures, each set containing four individual pictures drawn in watercolor on separate 8.5 X 11 in. (21.59 X 27.94 cm) heavyweight illustration board. Since the original pictures used by Brown were unavailable (personal conlmunication from Brown), two of our sets l'ollowed closely the descriptions provided in his 1957 article, while one differed slightly. They were designed to be unusual in appearance so that Ss would have no familiar labels for the various items. One of the four pictures of each set was a key picture which contained all three referents: the mass noun, the count noun, and the verb. For example,

Downloaded by [University of New Hampshire] at 17:26 17 February 2015

6

JOURNAL OF PSYCHOLOGY

the key picture of set 1 showed a pair of hands engaged in a kneading activity (verb) with a quantity of small, red, four-sided polygons (mass noun) which were heaped into and spilling out of an odd-shaped blue container (count noun). T h e three remaining pictures of the set reproduced one of the three important features of the key pictures. For example, in set 1 the picture depicting the mass noun consisted only of a pile of the red, four-sided polygons shown in the key picture. The picture depicting the count noun depicted only the odd-shaped blue container shown in the key picture. Since the picture showing the verb could not be depicted clearly without showing the material and the container in addition to the hands, the red polygons were changed to yellow circular and oblong objects and the blue container was changed to a different shaped purple container. The only thing that remained exactly the same as in the key picture were the hands, which were again engaged in a kneading activity. I n each of sets 2 and 3, the key picture represented a different action, mass substance, and specific object, while the remaining pictures showed each feature individually. T h e key picture of set 2 showed a pair of hands shaking a mass of rice-like material through a lumpy green colander-like object; while the set 3 key picture displayed a pair of hands pouring some yellow slushy liquid out of a distorted horn-shaped container. 3 . Procedure a. Experimental task. The Ss, tested individually, were presented with the three sets one a t a time. Each set was paired with one of three nonsense word stems (latt, niss, kebb) and one of three grammatical categories (verb, mass noun, count noun). Prior to the test, 10 presentation conditions were constructed, each condition randomly varying the combinations of set presentation order, word stem, and part of speech. T o control for position preferences, the action, mass noun, and count noun pictures were arranged in a different left to right sequence for each set. Presentation conditions were assigned to Ss on a random basis. Each S was introduced to the task as follows: "We're going to play a game. I'm going to say some new words, and you show me what they mean." The key picture of one of the sets was presented on an upright stand while the relevant nonsense word was introduced in a sentence which made explicit its part-of-speech membership. If the word was to be a count noun, the E said: "Do you know what a latt is? I n this picture (indicating the key picture) you can see a latt" (or, for another set, a kebb or a niss). The other

Downloaded by [University of New Hampshire] at 17:26 17 February 2015

CAROLYN W. LIEBER AND HERMAN H. SPITZ

7

three pictures of the set were then placed flat on the table in a horizontal row just beneath the key picture (which remained in view), and the S was asked to ". . . point to another picture of a latt (kebb, niss) ." If he word stem was to function as a mass noun, the S was asked: "Have you ever seen any latt? In this picture you can see some latt. Now point to another picture of latt." For presentation of a word as a verb, the E asked, "Do you know what it means to latt? In this picture you can see latting. Now point to another picture of latting." I t was expected that--if the S made use of context-when the new word was presented as a count noun, the S would point to the picture of the particular object alone; when it was presented as a mass noun, the picture of just the material would be chosen; and that the illustration depicting the motion or action would be selected when the word stem functioned as a verb. The above procedure followed that described by Brown ( 3 ) , who used the word stems latt, niss, and sib in his study. The nonsense word kebb was introduced in this experiment as a substitute for sib, which is an actual word. b. Control task. When an S completed the experimental task, the sets were reintroduced with the following explanation: "Now I'm going to say some other words, and you show me what they mean." The experimental procedure was replicated, except that real words were substituted for nonsense terms. Words were chosen on the basis of some resemblance to the object:; and activities portrayed, and on their estimated familiarity to the Ss. Although the items had been drawn so as to avoid depicting any reid-life objects, there was enough superficial resemblance to allow the labeling of the pictured objects with commonly used words. Since the terms were associated with specific items, words were fixed to each individual set instead of being assigned randomly. For set 1, the words mixing, paper, and bowl were paired respectively with the action, mass substance, and particular object pictures. The words used in set 2 were shaking, dirt, and pot, and in set 3 they were pourin,:, lemonade, and horn. If the S originally had been asked to indicate latting (the verb) in set 1, he or she was now requested to point to the picture showing mixing, and so on, following the particular presentation condition. The purpose of the control task was to investigate to what extent inability to perform the experimental task accurately might have been due to the inherent difficulty of the matching task for Ss of particular ages, rather than tc a presumed problem in making semantic inferences about new words on the basis of syntactic clues.

J O U R N A L OF PSYCHOLOGY

Downloaded by [University of New Hampshire] at 17:26 17 February 2015

1.

Expcvimentnl Task

I n computing accuracy scores for the parts of speech, a tendency was noted, especially among the younger normals and the retarded Ss, to choose the same part-of-speech picture for all three presentations. T o control for response bias, accuracy scores on each picture were corrected for guessing according to the following formula: Number of actual true positives divided by the number of possible true positives, minus the number of actual false positives divided by the number of possible false positives. According to this formula, chance equals zero, and Ss can score below chance, since subtraction is algebraic (e. g., 0 - .33 = - . 3 3 ) . Figure 1 presents the performance of the normal and retarded Ss for each part of speech. It shows no change from mean ages 6.23 to 8 years, but a pronounced improvement in task accuracy from mean ages 8.19 to 11.45 years. A two factor mixed design analysis of variance for the normal Ss indi-

C H R O N O L O G I C A L AGE

FIGURE 1

PLHFOR~~.ASCES OF T H E S GROUPSON

T H E E X P E K I M L N T A ITASK .

Downloaded by [University of New Hampshire] at 17:26 17 February 2015

CAROLYN W. LIEBER AND HERMAN H. SPITZ

9

cated that the improvement with age was a reliable effect [F (3, 132) = 5.67, p < .OD51 but that the part-of-speech variable was not. None of the part-of-speech scores for the youngest group (Group 1) was significantly above chance, as measured by single mean t tests, and this was true also for the verb accuracy score of Group 2. Chance levels of performance were exceeded by Group 2 for mass noun and count noun ( p < .05) and by the two older normal groups for each part of speech ( p at least < .05). Although females are reputed to excel on verbal tasks, they reliably exceeded males only on the verb part of speech in Group 3 [t (33) = 2.09, p < .O!i]. One reliable difference out of 12 could, of course, readily occur by chance. Ana1,ysis of the performance of the EMR sample on the experimental task reveals a part-of-speech accuracy order in exact reverse of that found with most of the normals; i. e., the retarded Ss had most success with count nouns, followec-l by verbs, and then by mass nouns. The differences between count noun and mass noun scores were statistically significant [ t (29) = 2.20, p < .OC;]. A visual comparison of the normal and EMR performance shown in Figure 1 indicates that only on the count noun presentation did the retarded ,group achieve a level of accuracy expected on the basis of their mean MA of 9.18. For nonsense words presented as verbs, EMR accuracy approximated that of Group 2 (CA = 8.19), while performance with mass nouns fell below that of any age group. In other words, on the verb and mass noun parts of speecth, retarded adolescents performed at least one year below their MA, when the MAS of normal children were used as criteria. Race was not a statistically significant variable among the retarded adolescents, and, as with the normal children of Group 3, sex was significant only on verb:;, where females were superior [t (28) = 2.16, p < .05]. 2.

Control Task

Performance of all groups on the control (meaningful) words was significaptly above chance level, and in all but three instances scores on the control words were significantly higher than on the experimental (nonsense) words ( p a t least < .01). The three cases where control performance did not exceed experimental performance were for the verb and mass noun in Group 1, and the mass noun in Group 3. The difficulty encountered by Group 3 with mass nouns during the control task may be attributable to difficulty in perceiving as "paper" the red, poly-

10

JOURNAL OF PSYCHOLOGY

Downloaded by [University of New Hampshire] at 17:26 17 February 2015

gon-shaped material in set 1. Younger children (and the retarded) had little difficulty making this association, possibly because the substance bore a resemblance to pieces of red construction paper often used for various projects in the early grades of school. Comparison of the three meaningful words used for each part of speech showed no significant differences within Group 1, Group 4, and the E h l R group. For Group 2 , performance on the mass nouns "lemonade" and "dirt" exceeded ( p < .001) accuracy for "paper." In Group 3, Ss performed significantly ( p < .01) better on the verb "mixing" compared to "pouring."

The results of this study indicate a much lower level of task accuracy than reported by Brown in 1957. This finding is especially noteworthy when one considers that Brown's sample of 16 Ss (CAs 3-5 years) was younger than the youngest group in the present study. Direct comparison of accuracy scores is difficult because of the fact that we used adjusted scores (corrected for response bias) while Brown reported only uncorrected scores. However, even use of uncorrected scores reveals a substantial superiority of Brown's Ss over the Ss studied here. Only in = 11.45 years) did uncorrected accuracy percentages for all Group 4 parts of speech resemble those obtained by Brown with his preschoolers. I t should be emphasized again that our pictures were different from Brown's. However, if the variable under investigation is responsive only to a particular set of pictures, it loses its potency. I t is conceivable that cultural and class differences may partially account for the discrepancy in results. The children tested by Brown were attending a preschool at Harvard and perhaps were exposed to a more highly verbal environment than were the present sample, whose backgrounds were predominantly middle and lower-middle class. Results obtained from the latter group would be more applicable to the population as a whole than would findings obtained from children of a highly educated minority. The retarded group showed a different order of response accuracy and a generally lower level of accuracy than normal children of equivalent MA. Since the test involves the ability to infer knowledge about syntactic/semantic correlates to novel situations, it seems reasonable to hypothesize that the conlparatively poorer showing of the EAlRs is a reflection of a basic deficiency in generalization ability (e. g., 2 ) . Also, in contrast to the normal sample, the E h l R s displayed a response bias toward the count noun pictures

(a

Downloaded by [University of New Hampshire] at 17:26 17 February 2015

CAROLYN W. LIEBER AND HERMAN H . SPITZ

11

of each set, which was the only part of speech in which the EMR group's accuriicy approached MA expectations. I t is difficult to account for this behavior. One possible interpretation is that-since the count noun picture in each :jet was visually the least complex--choice of this picture in an uncertain situation reflects a preference by the retarded for the least complicated option. Behavior of the normals in a similar situation of uncertainty resulted in a general preference for the most complex alternative, the verb picture, a more logical guess, since the verb picture was the only one which depicted all three original referents, albeit the count noun and mass noun referents had been substantially altered. Lircitations of the test must be considered in interpreting the results. The younger children in the present study often appeared attracted more by the colors in the pictures than by the linguistic context of the nonsense words. The fact that the performance of Group 1 on the control presentations of verb and mass noun was not significantly higher than on the experimental presentations suggests that for these Ss the inherent nature of the matching task presented difficulty. Informal questioning of Group 1 Ss who gave incorrect answers indicated that, although they could point to the correct partof-speech referent in the key picture of a set, they were unable to transfer this response to the duplicate referent shown in isolation on a separate card. I t is likely that the younger Ss would have been more successful in making these distinctions had they been presented with a less abstract task involving, for exirmple, actual manipulation of objects and imitation of body actions. In any case, results of the present study, which was based on Brown's original test procedure, call into question his widely disseminated findings that young, normal, language-learning children are able, with a high degree of success, to use syntactic clues in making semantic inferences about unfamiliar words in this experimental task.

BERKO, J . The child's learning of English morphology. Word, 1958, 14, 150-177 BII.SKY,L., EVANS,R. A., & GILBERT,L. Generalization of associative clustering tendencies in mentally retarded adolescents: Effects of novel stimuli. Amer. J. Aienl. Defic., 1972, 77, 77-84. B R ~ W NR., W. Linguistic determinism and the part of speech. J. Abn. 6 Sor. Fsychol., 1957, 55, 1-5. CAPROLL, J . B. Language development. In C. Harris (Ed.), Enryrlopedia of Educ~ztionalResearch (3rd e d . ) . Amer. Educ. Res. Assoc., 1960. F A I ~ BP., Word Play. New York: Knopf, 1974. LEIVNEBERC, E. H. Biological Foundations of Language. New York: Wiley, 1967.

12

J O U R N A L O F PSYCHOLOGY

LOVELL,K.,& BRADBURY, B. The learning of English morphology in educationally subnormal special school children. Amer. J. Ment. Defic., 1967, 71, 609-615. 8. ~ I C N E I L LD. , The Acquisition of Language: The Study of Developmental Psycholinguistics. New York: Harper & Row, 1970. 9. NEWFIELD,M . U., & SCHLANGER, B. B. The acquisition of English morphology by normal and educable mentally retarded children. /. Speech & Hearing Res., 1968, 11, 693-706. 10. YODER,D. E., & MILLER,J . F. LVhat we may know and what we can d o : Input toward a system. In J. E. McLearn, D. E . Yoder, & R. L. Schiefelbusch (Eds.), Language Intervention with the Retarded. Baltimore: Univ. Park Press, 1972. 7.

Downloaded by [University of New Hampshire] at 17:26 17 February 2015

E . R, Johnstonc Training and Research Center Bordentown, New Jersey 08505

Inference of word meaning from syntax structure by normal children and retarded adolescents.

Children from five to 12 years of age (N = 136 boys and girls) and institutionalized retarded adolescents (N = 30 boys and girls) were given a task de...
498KB Sizes 0 Downloads 0 Views