Children's understanding and production of sentences

Children's understanding and production of sentences

JOURNAL OF VERBAL LEARNING AND VERBAL BEHAVIOR 8, 2 8 9 - 2 9 4 (1969) Children's Understanding and Production of $entences~ ELEANOR P. GAER Newark,...

419KB Sizes 32 Downloads 52 Views

JOURNAL OF VERBAL LEARNING AND VERBAL BEHAVIOR 8, 2 8 9 - 2 9 4

(1969)

Children's Understanding and Production of $entences~ ELEANOR P. GAER Newark, Delaware

Children, ages 3 to 6, and adults were tested on their understanding and production of sentences varying on two dimensions, type and complexity. The order of understanding sentence types, from easiest to hardest, was active, question, passive, negative. The order of production of sentence types was active, negative, question, passive. Simple sentences were understood significantly better than center-embedding and double-embedding sentences. Single-embedding sentences were understood as well as simple sentences but were produced more poorly than simple sentences for all ages. Center and double-embedding sentences were produced infrequently by all ages. The effect of age on understanding and production is described. This study is concerned with the development of the language capabilities of the child. It tests the abilities of children ages 3 to 6 to understand and produce sentences varying on two grammatical dimensions: complexity and transformational type. In recent years there has been a great deal of discussion on how a child learns a language. Much of the work in this area consists of theoretical discussions or naturalistic observation (Bellugi and Brown, 1964; Smith and Miller, 1966). While naturalistic observation is useful, it is limited both in the number of children who can be intensively observed and the age at which a complete record of a child's language can be recorded (much of the work has been with 2-year-olds). It also cannot handle the problem of understanding and its relation to production. In an attempt to overcome the shortcomings cited above, the present study deals experimentally with the age at which the child learns certain grammatical constructions. The sentences presented to the children were based on Chomsky's (1957, 1965) generative

grammar; they were varied as to sentence type and level of complexity. The question to be tested was: how well do differences in syntax correspond with differences in actual production and understanding of sentences. The sentence types were active, passive, negative, and question. The passive, negative, and question have "Pass," " Q , " and " N e g " in their base structures and are the result of the application of specific transformational rules. The simple active sentence should be the first learned, since it does not involve additional transformational rules. The question and negative sentences have certain similarities in their transformational rules (the insertion of the appropriate form of do and the same organization of the auxiliary constituent), which should encourage their being learned at approximately the same time. I f there is any difference, the question may be learned slightly later than the negative, since it involves transposing the subject noun phrase and the first member of the auxiliary. The passive transformation involves the most operations and should be learned the latest. The second dimension tested in this 1 This study is based on a doctoral dissertation experiment is sentence complexity. The more submitted to the University of Illinois. The author thanks Dr. Charles Osgood for his help and guidance. constituent sentences embedded in a matrix 289 10

290

CAER

sentence, t h e m o r e c o m p l e x is the sentence. The m o r e sentences which are e m b e d d e d , the m o r e t r a n s f o r m a t i o n a l o p e r a t i o n s have to t a k e place. The sentence with the fewest e m b e d d i n g s s h o u l d be the easiest for c h i l d r e n to u n d e r s t a n d a n d p r o d u c e , with difficulty increasing as c o m p l e x i t y increases. G r a m m a t i c a l l y , center e m b e d d i n g s are n o t d e r i v a t i o n a l l y m o r e c o m p l e x t h a n o t h e r sorts o f e m b e d d i n g s , b u t studies with a d u l t s have shown t h a t c e n t e r - e m b e d d e d sentences are h a r d e r to u n d e r s t a n d t h a n sentences with the s a m e c o m p o n e n t s t h a t are n o t center-emb e d d e d (Miller a n d l s a r d , 1964). A d u l t s c o u l d n o t learn sentences with three or f o u r selfe m b e d d e d relative clauses w i t h o u t e n c o u n t e r ing difficulty. I f this is true for adults, children s h o u l d have difficulty with sentences c o n t a i n ing one c e n t e r - e m b e d d e d relative clause. Such sentences s h o u l d be h a r d e r for t h e m to u n d e r s t a n d a n d p r o d u c e t h a n sentences with one clause which is n o t c e n t e r - e m b e d d e d . Sentences with t w o e m b e d d i n g s require m o r e o p e r a t i o n s to go f r o m their deep structure to their surface structure t h a n sentences with one c e n t e r - e m b e d d i n g , a n d should thus be h a r d e s t to u n d e r s t a n d . W h i l e it is c o m m o n l y claimed t h a t p r o d u c t i o n lags b e h i n d u n d e r s t a n d i n g , there have been few studies e x a m i n i n g this e x p e r i m e n t a l l y (Fraser, Bellugi, a n d Brown, 1963). I t is expected t h a t p r o d u c t i o n will lag b e h i n d u n d e r s t a n d i n g with correct p r o d u c t i o n occurring at a later age t h a n u n d e r s t a n d i n g for the same constructions. Even t h o u g h p r o d u c t i o n lags b e h i n d understanding, a child w h o u n d e r s t a n d s a simpler c o n s t r u c t i o n should be able to p r o d u c e it s o o n e r t h a n a c o n s t r u c t i o n which is m o r e difficult for h i m to u n d e r s t a n d . I t is therefore p r e d i c t e d t h a t success in p r o d u c t i o n o f specific c o n s t r u c t i o n s will occur in the same o r d e r as success in u n d e r s t a n d i n g them. METHOD The Ss were 24 children at each age level (3, 4, 5, and 6) and 24 adults. Adults were used to find the upper

limits of performance on the tasks. The 3- and 4-yearolds attended nursery schools. The 5- and 6-year-olds were kindergarten and first-grade pupils. The adults were University of Illinois students. The four sentence types were active, negative, passive, and question. An example of an active (A) construction is The boy sings the song; negative (N) is The boy does not sing the song; passive (P) is The song is sung by the boy; question (Q) is Does the boy sing the song? There were four levels of sentence complexity. The first level is a sentence which has no other sentences embedded in its underlying phrase marker (P-marker); the second is a sentence with one sentence embedded in its underlying P-marker; the third is again a sentence with one sentence embedded in its underlying P-marker, but in the surface structure it is center-embedded; the fourth is a sentence with two sentences embedded in its underlying P-marker, but neither center-embedded. Examples illustrating the levels of complexity are: I. (simple sentence) The boy is throwing the ball down the hill. II. (single embedding) The girl sees the boy who is kicking the ball. III. (center embedding) The cat the girl sees is playing with a ball. IV. (double embedding) The man is watching the boy catching the ball the girl is throwing. Subjects at each age level were randomly divided into four groups, with each group getting a different rotation of vocabulary over sentence type. Each child heard a sequence of 16 out of the possible 64 sentences. The pictures used to illustrate the sentences were simple ink drawings, 5 in × 5 in square. Two pictures, a positive instance and a negative instance, were used to illustrate all four transformations of a sentence with a specific vocabulary. For example, The boy is throwing the ball down the hill; The boy is not throwing the ball down the hill; Is the boy throwing the ball down the hill? and The ball is being thrown down the hill by the boy were either illustrated by a picture of a boy throwing a ball down a hill (positive instance) or a boy standing on a hill with a ball at his feet (negative instance). An example of the pictures is shown in Fig. 1. Each child was tested individually. He was told to say "yes" if the picture showed what he had heard, "no" if it did not. Every time he was right, he was told he was right and given a piece of candy. The sentences were tape-recorded. All Ss were presented with three practice sentences. When the understanding phase was completed, the Ss were presented with the production sequence of sentences, which consisted of the same sequence of sentences that he heard in the first part of the experiment. This time, five random numbers were read immediately after the sentence, approximately 1 see apart. The child repeated each number aloud immediately after he heard it. After repeating the fifth number, the child was shown the same picture he had

CHILDREN'S SENTENCE PERFORMANCE

291

to the girl eating near her mother and said "yes." A more complete test of understanding the entire sentence would involve additional pictures with girls eating and not eating, etc. The adult group received the same stimulus materials as the children. They were tested in groups of three, and they wrote their responses. Instead of repeating the numbers in the production sequence, the adults were told to add the number seven to each number and to write down the sum. This was done to prevent automatic repetition of the numbers by the adults while they also repeated the sentence to themselves. RESULTS

FIG. 1. Complexity level 0. Sentence 1. The boy is throwing the ball down the hill. seen in the first part of the experiment and was asked to recall the sentence. If he indicated that'he did not remember, E said, "What do you think she said?" If the child could not respond to this, E said, "Tell me what you seein the picture." In some cases the child could then repeat the sentence correctly. Although a correct response to a sentence like The boy sees the girl eating a cookie her mother baked

involved only a visual contrast between the boy watching and not watching, the Ss showed evidence of attending to the entire sentence. For example, if the picture showed the boy not watching, the correct response would be "no." However, many Ss pointed 10"

A n analysis o f variance o f the understanding data (see Table l) p r o d u c e d a significant Sentence Type by Age interaction, F(9, 36) = 2 . 6 2 0 , p < . 0 5 . A D u n c a n Range test at p < .05 showed that for all children except the 3-year-olds the A was understood significantly better than at least one other sentence type. Performance on the Q sentence became as g o o d as the A at 5 years. F o r the 4-year-olds Ps were p o o r e r than Qs, while for the 5- and 6-year-olds the difference between Ps and Qs was not significant. A n analysis o f variance o f the p r o d u c t i o n data (see Table 2) also resulted in a significant Sentence Type by Age interaction, F(9, 3~6)= 2.081, p < . 0 0 5 . A D u n c a n Range test at p < .05 showed that at 5 the Q sentence does not differ f r o m the A, but at 6 it is p r o d u c e d significantly better than the A construction. The N construction becomes superior to A at 5. The P is consistently poorest in production, and at 6 years it becomes significantly p o o r e r than any other construction. The adults were significantly better than the 6-year-olds in b o t h understanding, F(1, 6) = 88.759, p < .001, and production, F(1, 6) = 3 9 . 1 5 8 , p < .001 (see Tables 1 and 2). There was a significant Type by Age interaction in production, F(3,18) = 4.457, p < .025. The P sentences were significantly p o o r e r than the Q and A sentences even for adults ( D u n c a n Range test, p < .05). The N c o n s t r u c t i o n was poorest o f all sentences in understanding. In production, however, it was better than the Q construction

292

GAER TABLE 1 PERCENTAGE OF SENTENCES CORRECTLY UNDERSTOOD

Complexity Age 3 Years 4 Years 5 Years 6 Years Total Adults

Type

0

1-E

C-E

2-E

Total

65 69 69 76 70 97

55 63 67 78 66 92

50 66 59 70 61 96

58 56 63 70 62 99

57 64 65 73 65 96

Active Question Negative Passive 58 79 75 82 74 96

61 73 77 85 74 96

51 46 38 50 46 94

57 56 69 76 56 97

TABLE 2 PERCENTAGE OF SENTENCES CORRECTLY PRODUCED

Complexity Age 3 Years 4 Years 5 Years 6 Years Total Adults

Type

0

1-E

C-E

2-E

Total

25 44 51 60 45 93

7 17 25 31 20 97

1 0 4 7 3 96

0 8 13 13 8 79

8 17 23 28 19 91

for the 3-, 4-, and 5-year-olds. The poor showing of the N may not have been due to syntax. Understanding involved answering "Yes" or " N o " to a picture, and this is a semantic or truth-value problem, not a syntactic one. The plausibility of the sentence may have also affected the results. The more implausible the sentence the easier is the negation (Wason, 1965). Success in production of sentence types occurred in the same order as success in understanding, except for the N sentences. A and Q sentences were produced and understood better than P. Analysis of variance for type in understanding yields F(3, 36) = 34.105, p < .001 ; in production, F(3, 36) = 31.477, p < .001, Duncan Range test a t p < .05. There were not as many changes in understanding and production of levels of complexity

Active Question Negative Passive 13 24 23 24 21 99

5 11 22 40 20 92

14 29 34 35 28 93

1 4 14 11 8 81

over ages as changes in sentence types. An analysis of variance showed that levels of complexity C-E and 2-E (center-embedding and double-embedding) were understood less well than simple sentences (0) by all children, F(3, 36) = 4.890, p < .01, Duncan Range test a t p < .05 (see Table 1). An analysis of variance showed that the overall order of production, from easiest to hardest, was 0, l-E, 2-E, C-E, /7(3, 3 6 ) = 125.826, p < . 0 0 1 , Duncan Range test at p < .05. See Table 2. The children show increasing control over single-embedding sentences (l-E) from ages 3 to 4 to 5, as indicated by the significant Age by Complexity interaction at p < .001, F(9, 3 6 ) = 2 . 9 8 1 , Duncan Range test at p < . 0 5 . At age 3, single-embedding sentences (l-E) do not differ significantly from center-embedding

CHILDREN'S SENTENCE PERFORMANCE

293

down the hill. Confusion may arise between answering "Yes" (he is not throwing it) and "No" (he is not throwing it) which may result in the children giving the wrong answer consistently. This difficulty appears to be overcome by the time the child reaches 6, but the negative still confuses him enough so that he cannot perform significantly better than chance. As Wason (1965) pointed out, the plausibility of the sentence is also important in negation. If the sentence had been unusual in content (The boy is not throwing the cow down the hill), it may have been easier for the children to say "no" to a picture of a boy tossing a cow. Throwing and not throwing balls probably occur with nearly equal frequency in the lives of children; one class is not subsidiary to the other. If a construction is syntactically complex, it does not necessarily follow that it will be psychologically complex. However, there is evidence that syntactic complexity and psychological complexity are related. Sentences more complex syntactically are more difficult to produce, although they may be understood as well as less complex sentences. Level 1-E is understood as well as 0, but I-E is more difficult to produce. DISCUSSION The passive construction is the last learned The fact that the negative was superior to in understanding and provides a great deal of the question and passive for certain ages in difficulty in production. The passive interacts production but was poorer in the understand- with the levels of complexity. The passive ing phase suggests that there are factors other transformation was the only one which made than syntax operating here. The negative single-embedding sentences (l-E) as difficult sentences were never understood significantly to produce as center-embedding (C-E) and better than chance at any age level, except double-embedding (2-E) sentences. As senadult (binomial test). As a matter of fact, the tences become more complex, in terms of negative sentences became significantly poorer number of embeddings, more operations must than chance in one instance for the 4-year-olds be performed to change the sentence from and in two instances for the 5-year-olds. The its deep structure to its surface structure. 6-year-olds return to the chance performance More information on the sentence he has of the 3-year-olds. Getting significantly fewer heard must be stored by the S who has to right than would be predicted by chance produce it. The passive, hardest to learn, suggests that an incorrect interpretation may makes this task more difficult. The two transhave been dominant for these sentences. The formations which are the last learned (queschild hears The boy is not throwing the ball tion and passive) are the two which interact (C-E) and double-embedding (2-E) sentences. At 4 years, level 1-E is significantly better than C-E, but not 2-E, and at 5 and 6 years level 1-E is significantly better than both C-E and 2-E. The single-embedding sentence (l-E) is always poorer in production than the simple sentence (0). The biggest jump in production of simple sentences (0) is from ages 3 to 4. Levels C-E and 2-E do not differ significantly at any age level. For adults (see Table 1) there was no significant difference among levels of complexity in understanding. However, in production (see Table 2) double-embedding sentences are significantly poorer than the other three levels, as shown by the Complexity by Age interaction significant at p < .001, F(3, 18)= 28.814, Duncan Range test atp < .05. In understanding, simple sentences (0) and single-embedding sentences (l-E) did not differ significantly. Single-embedding sentences, however, are produced significantly more poorly than simple sentences for all age groups. Center-embedding (C-E) and double-embedding (2-E) sentences were understood as well as single-embedding sentences, but they were produced very infrequently.

2 94

GAER

with double-embedding sentences to produce poorer performance in adults. Although center-embedded sentences probably are less frequent than noncenterembedded sentences in the language which adults speak, adults perform better on them than on sentences with two embeddings. Sentences with two embeddings require more linguistic operations to change the sentence from its deep structure to its surface structure. These findings indicate they are psychologically more complex as well. The surface structure.of center-embedding (C-E) sentences apparently interferes enough to make them more difficult to understand and produce than single-embedding sentences (l-E), although they both have the same number of embeddings. This is true for children but not for adults. Adults have no more difficulty with C-E than 1-E. perhaps the children do not "know" the grammar well enough to recognize that these sentences both have the same.deep structure. Limitation

of memory may also be involved here, since even adults cannot remember sentences with more than two center ernbeddings (Miller and Isard, 1964). REFERENCES BELLUGI, U., AND BROWN~R. (Eds.) The acquisition of language. Monogr. Soc. for Res. in Child Devel., 1964, 29, No. 1. CrIOMSKY, N. Syntactic structures. The Hague, The Netherlands: Mouton and Co., 1957. CnOMSKY, N. Aspects of the theory of syntax. Cambridge: The M. I. T. Press, 1965. FRASER, C., BELLUIG,U., AND BROWN,R. Control of

grammar in imitation, comprehension, and production. J. verb. Learn. verb. Behav., 1963, 2, 121-135. MILLER, G., AND ISARD, S. Some perceptual consequences of linguistic rules. J. verb. Learn. verb. Behav., 1963, 2, 217-228. SMITH, F., AND MILLER, G. The genesis of language: A psycholinguistic approach. Cambridge: The

M. I. T. Press, 1966. WASON,P. C. The~contextsof plausible denial. J. verb. Learn. verb. Behav., 1965, 4, 7-11. (First received FGbruary2, 1968)