Language acquisition following hemidecortication: Linguistic superiority of the left over the right hemisphere

BRAIN

AND

LANGUAGE

3, 404-433

(1976)

Language Acquisition following Hemidecortication: Linguistic Superiority of the Left over the Right Hemisphere MAUREEN Department

DENNIS AND HARRY A. WHITAKER

of Psychology, Hospital Department of Psychology,

for Sick Children, Toronto; University of Rochester

and

The language development of three 9- and lo-year-old children possessing only a right or a left hemisphere was studied. Surgical removal of one brain half antedated the beginning of speech, so each child has acquired speech and language with only one hemisphere. Different configurations of language skill have developed in the two isolated hemispheres: phonemic and semantic abilities are similarly developed but syntactic competence has been asymmetrically acquired. In relation to the left, the right brain half is deficient in understanding auditory language, especially when meaning is conveyed by syntactic diversity; detecting and correcting errors of surface syntactic structure; repeating stylistically permuted sentences; producing tag questions which match the grammatical features of a heard statement; determining sentence implication; integrating semantic and syntactic information to replace missing pronouns; and performing judgments of word interrelationships in sentences. Language development in an isolated right hemisphere, even under seizure-free conditions, results in incomplete language acquisition.

Language use and understanding are typically localized in the left half of the brain, but early cerebral injury may necessitate a right hemisphere control of linguistic functions. Either hemisphere in this event can acquire gross expressive speech and the skills measured by Verbal IQ (McFie, 1961; Carlson, Netley, Hendrick, & Prichard, 1968); but one linguistic function does not develop as well in a right hemisphere as it does in a left brain half. The enforced development of speech and language in the right hemisphere after early left hemisphere insult causes an impairment in the understanding of language structure in adulthood. In individuals with one hemisphere surgically removed for infantile cerebral injury, the level of syntactic comprehension is higher in those hemidecorticates possessing This research was supported by Ontario Mental Health Foundation Hendrick for permission to test his the phoneme production data. We CA to play our word games. Address reprint requests to Dr. Children, 555 University Avenue,

a Hospital for Sick Children Research Scholarship to the first author. We patients, and Dr. C. A. Wiegel-Grump acknowledge the unlimited willingness Maureen Toronto,

Dennis, Research Associate, Ont., Canada M5G 1X8.

404 CopyrIght All rights

0 1976 by Academic Press. Inc. ofreproductiun in any kvorm 1ese8ved.

Fellowship and an thank Dr. E. B. for her analysis of of MW, SM, and Hospital

for Sick

LANGUAGE

ACQUISITION

BY

ONE

HEMISPHERE

405

the left brain half (Dennis & Kohn, 1975). Discrimination of syntactic forms in the residual right hemisphere is inferior in two ways to that of the left: the adult rate of proficiency is lower, and the age at which nonrandom performance is attained is higher. Both the adult level of this language function and the rate of its acquisition are different in right and left hemidecorticate infantile hemiplegics. It is not known whether verbal capacities other than syntactic discrimination are acquired at different rates by the left and right hemispheres. The present study therefore documents the acquisition of a variety of language skills in three children possessing only a right or a left hemisphere. Surgical removal of one brain half antedated the beginning of speech, so each child has acquired speech and language with only one hemisphere. None of the children has had seizures during the period of language acquisition, so relatively normal educational progression has been possible. The reported observations, then, reflect how well each isolated but proficiently functioning hemisphere can acquire language. SUBJECTS

The three children were diagnosed soon after birth as cases of Sturge- Weber- Dimitri syndrome. In each, a brief period of seizures was followed by surgical removal of one hemisphere before 5 months of age. MW (RH,

Right Hemidecorticate

Male)

MW was born with a port wine naevus of the upper face. At 3% months of age, he had a left-sided seizure which proved unresponsive to barbiturates, Paraldehyde, Dilantin, or vitamin B6 but which was finally controlled with ether. Both generalized and left-sided seizures recurred during the next 2 weeks. On physical examination, his head was normal in shape; there were no cranial bruits heard. The head was kept turned to the right much of the time. Pupils were large, symmetrical, and reacted sluggishly to light. A visual field defect was suspected but could not be confirmed. No facial asymmetry was commented upon, but there was dn obvious left hemiparesis and some posturing movements of the left hand. The whole left side was hypotonic. The EEG record was abnormal and reflected considerable disturbance of function over the entire right hemisphere. A right hemispherectomy was done at the age of 41%months. Exposure of the brain showed an extensive capillary angioma over the posterior frontal and parietal lobes. There were large angiomatous veins lying in the Sylvian fissure, and the draining veins into the sag&al sinus were also tortuous and dilated. The whole brain appeared soft. The right cortex and basal ganglia were removed in stages. Histological examination of the

406

DENNIS

AND

WHITAKER

resected tissue showed the small pathological changes characteristic of Sturge- Weber- Dimitri syndrome. There were microscopic foci of calcification in the cortex itself and a diffuse angiomatosis of the meninges of the parietal, occipital, and temporal lobes. Immediately after the hemispherectomy, MW could move his left leg slightly. Over the next few days he gained more control of movement and began to move his left arm and leg, although there was still a significant degree of hemiparesis. During the first 2 postoperative weeks he did have some seizures but these had stopped by the time of his discharge from hospital at 5% months of age. MW has been seizure-free in the 10 years since his hemispherectomy except for one episode involving a IO-min generalized seizure when he was 9. The EEG at this time showed a seizure focus in the left midhead regions. On reexamination 6 months later, the EEG was normal and there has been no recurrence of the seizures. A neurological examination at age 9 noted that the hemiparesis involved the upper more than the lower extremities. Sensation (2-point discrimination, proprioception and stereognosis) was disturbed on the left. The left homonymous hemianopsia was confirmed. Position sense was absent at the toes and fingers. There was no deficit to pain and light touch. He had full ocular movements. There appeared to be a conductive hearing loss on the left side, although the right ear was normal. MW has managed to maintain his grade level at school, despite multiple hospital admissions for orthopedic and cosmetic surgery. His cognitive functions are at a normal level (WISC: Verbal IQ 96, Performance IQ 92, Full Scale IQ 93). SM (LH, Left Hemidecorticate

Male)

SM was born with port wine staining of the left face and body and a congenital glaucoma of the left eye. Immediately after birth he developed severe convulsive right-sided seizures which could not be controlled with medication. Physical examination at 4 months of age revealed a right facial weakness and a right-sided hemiparesis. The right arm and hand were smaller than the left and there was decreased movement on the right side. Deep tendon reflexes on the right were increased. The eyes deviated to the left. No abnormalities were apparent on the routine laboratory tests. The X-ray revealed a skull asymmetry. An air encephalogram suggested localized cerebral atrophy in the left parieto-occipital area. A persistent asymmetry was evident on the EE G. The findings were felt to be consistent with an atrophic left hemisphere abnormality. A left hemispherectomy was performed at 4% months of age. The exposed cortex was red in appearance with prominent capillaries and unusually large draining veins. The left hemisphere was removed in stages. The basal ganglia were spared since it was possible to delineate them. The histological features of the removed cortex were those of the Sturge-

LANGUAGE

ACQUISITION

BY

ONE

HEMISPHERE

407

Weber- Dimitri syndrome. Areas of demyelination, gliosis and calcification were observed in the cerebral tissue. SM developed right-sided twitching a few days after surgery. This was readily controlled by anticonvulsive medication. He has been seizure-free in the 9 years since the hemispherectomy. A right hemiplegia and right facial weakness were noted on examination at age 8. Most of the right-sided reflexes were increased, and stereognosis was poor on the right. SM had normal hearing bilaterally to pure tones, and his speech discrimination was also good bilaterally with no ear asymmetry in speech discrimination scores. Despite the considerable blocks of time he has spent in hospital for surgical tattooing of the face and orthopedic corrections of his leg length discrepancy, SM is maintaining good standing at school. His intellectual abilities are normal (WISC: Verbal IQ 94, Performance IQ 87, Full Scale IQ 90). CA (LH,

Left Hemidecorticate

Female)

CA was born with a marked port wine stain on the left side of the face and over the distribution of the trigeminal nerve on the right. Twitching of the right arm and leg was observed immediately afterbirth, and she soon began to have right-sided seizures. On admission to hospital at 6 days of age, CA’s right arm twitching was evident. The skull on X-ray proved to be within normal limits in size and contour, and there was no evidence of increased intracranial pressure. The EEG record was significantly asymmetrical with a depression of voltage over the left hemisphere. A left hemispherectomy was carried out when CA was 28 days of age. The exposed cortex was plum-colored over its surface, with an arachnoid which was dense purple in some areas. The entire left hemisphere was removed in stages. The basal ganglia were spared. The brain tissue was much softer than normal on gross examination. The pathological diagnosis was that of haemangioma of the meninges and calcification and gliosis of the subadjacent cortex, i.e., Sturge-WeberDimitri syndrome. In all lobes of the cortex, there were focal areas of calcification, mostly confined to the white matter but also scattered throughout various parts of the grey matter as well. The parieto-temporal lobe showed some alteration in the normal pattern of the cortical grey matter. Some questionable right-sided twitching was observed in CA postoperatively, and so she was placed on anticonvulsive medication. This could be discontinued after 10 days. She has been free of seizures in the 9 years since the hemispherectomy. A repeat EEG after the operation had revealed no sign of seizures on the right side. CA now presents with a spastic hemiplegia. She is at an age-appropriate level at school. Her intellectual abilities are normal (WISC: Verbal IQ 91, Performance IQ 108, Full Scale IQ 99).

408

DENNIS

AND WHITAKER

EXPERIMENTAL

INVESTIGATIONS

Two types of language tests were used. The first were standardized tasks on which performance could be described in terms of normal childrens’ scores or of the percentage of correct responses in excess of chance. Tests were chosen to sample the general level of language development, phoneme discrimination and production, word discrimination and production, syntactic discrimination and the production of syntactic forms, and the comprehension of complex verbal material. Tests of the second kind were either devised for the present investigation or adapted for children from adult neurolinguistic tests designed by one of us (Whitaker, 1971). The aim with these tests was to explore areas of language function which had seemed difficult for the children on standard tests, to evaluate linguistic abilities not commonly tested in brain-damaged patients, or to explore some issue in neurolinguistic the0ry.l Standardized

Illinois

Test of Psycholinguistic

Abilities

Tests

(Kirk, McCarthy,

& Kirk, 1968)

This test provides separate measures of different language processes (comprehension, association, expression) and modalities (auditory, manual, visual). The scores on the test are expressed in age equivalences, so performance can be expressed normatively. The configuration of subtest performance is different in the three hemidecorticates, but their Psycholinguistic Ages (Table 1) are comparable. Each child shows considerable within-test variation. MW’s auditory comprehension is above the age norms for the test (i.e., above 10 yr 4 mo), while his visual receptive skill is below his chronological age. Since the auditory score is at ceiling, the scale score difference fails to reach statistical significance. For the associative language tests, the difference between MW’s auditory and visual skills is dramatic: both age and scale scores for the two channels are significantly different. The second striking dissociation of modalities MW shows is between his grammatical and visual completion. His acquisition of automatic habits for handling syntax and grammatical inflections is slightly above his actual age, but visual closure is at a low level of development. Auditory memory is poorer than visual memory, but not significantly so. Manual and verbal expressive skills are at comparable levels of development. The dissociation of modalities in both comprehension and associative processes is not shown by either of the left hemidecorticates. Auditory and visual reception, and auditory and visual association, are similarly developed. SM is more proficient at grammatical than at visual closure but 1 The children’s ages at testing on the standardized tests are indicated in the tables; they were 9 (SM, CA) and 10 (MW) during the second experimental tests.

LANGUAGE

ACQUISITION TABLE

ILLINOIS

TEST

409

BY ONE HEMISPHERE 1

OF PSYCHOLINGUISTIC

ABILITIES

MW (RH)

SM (LH)

CA (LH)

Age at test (years-months) Subtest Auditory reception

Age ss* Age ss As ss Age ss Age ss Age ss

Visual reception Auditory association Visual association Verbal expression Manual expression Grammatic

closure

As

ss Age ss Age ss Age ss

Visual closure Auditory memory Visual memory Composite psycholinguistic Mean SS

age

9-5

9-4

9-5

AN** 31 7-9 31 lo- 1 39 5-3 20 7-8 32 8-4 34 9-8 37 4-8 6-O 29 8-4 35

S-10 34 8-10 35 7-11 29 7-7 31 6-10 29 6-9 29 8-10 34 7-o 27 6-10 31 l-10 34

8-4 32 8-10 35 7-8 28 6-6 27 6-O 26 7-2 30 7-o 22 6-6 25 7-7 33 8-4 35

7-8

7-9

7-3

30.9

31.3

29.3

15

* Scaled Score. ** Above the norms for the test, i.e., performance above 10-4.

CA is not. Neither left hemidecorticate shows any difference between the two modalities of language expression or memory. Some of the low scores on the test perhaps reflect the limiting physical disability suffered by all three children. Their evident embarrassment at performing bilateral motor movement with a hemiparesis probably caused a general lowering of their manual expression scores. Nevertheless, all scores are not depressed, and the children differ on tests of visual closure where a hemianopsia might be expected to lower scores uniformly. The level of development of some language functions appears to depend on the laterality of the hemidecortication. MW is superior to SM and CA in the auditory channels for comprehension and association; SM and CA exceed MW in the ability to understand and make associations to visual stimuli. Some of the ITPA visual tests, it should be noted, are linguistic

410

DENNIS

AND

WHITAKER

only by a stipulative definition. Visual closure, for example, is a classical test of right hemisphere function (e.g., DeRenzi and Spinnler, 1966). Thus, the total ITPA score measures more than purely linguistic functions. The configuration of ITPA performance of the three children reflects a superiority of the isolated right hemisphere for visuo-spatial tasks (Kohn and Dennis, 1974) as well as a greater left hemisphere proficiency at auditory language. Auditory

Discrimination

Test (Wepman,

1958)

This test assesses the ability to identify differences between English phonemes. The examiner reads pairs of words to the subject, who must then indicate whether the words read were the same (a single word repeated) or different (two different words). The matches are made within phonetic categories, i.e., phonemes in the articulatory category of simple stops are matched only with other phonemes within that category. Discrimination is thereby based on the auditory features of the word and not on articulatory position. Scores are expressed as the percentage of correct same-different judgments. Goldman -Fristoe - Woodcock man, Fristoe, & Woodcock,

Test of Auditory 1970)

Discrimination

(Gold-

The ability to discriminate English phonemes is also assessed by this test. A word is read to the child who then selects the pictorial representation of what was heard from a 4-choice display. In the second part of the test, the auditory discrimination must be performed with background noise 9 dB less intense than the signal. The existing normative tables were not used since the test tape was given througha speakerand not through earphones. Results are presented as the percentage of correct responses made under the quiet and noise conditions. Goldman-Fristoe

Test of Articulation

(Goldman

& Fristoe,

1969)

This measures the ability to produce the phonemes used in English. The subject is required to name pictures of familiar objects while the examiner records articulation of the major speech sounds. The score is the percentage of the 73 sounds correctly produced. Acquisition of phoneme discrimination is remarkably similar in the right and left hemidecorticates (Table 2). The ability to make same-different judgments to pairs of phonemes and to match phonemically similar words against pictures is similarly proficient inall three subjects. The requirement of making a phonemic discrimination against background noise depressed performance (as with any subject) but the extent of the interference is comparable in the three children. The success rate on this latter condition

LANGUAGE

ACQUISITION TABLE

PHONEME

DISCRIMINATION

411

BY ONE HEMISPHERE 2

AND PRODUCTION(PERCENTAGE

MW (RH)

CORRECT)

SM (LH)

CA (LH)

Age at test (years-months) ADT TAD, TA

Test

10-6 10-11

Auditory Discrimination Test (Wepman, 1958) Test of Auditory Discrimination (Goldman, Fristoe & Woodcock, 1970) Quiet subtest Noise subtest Test of Articulation (Goldman & Fristoe, 1972) Sounds in words subtest

9-o 9-6

9-2 9-7

92.5

92.5

97.5

87.0 73.3

87.0 73.3

87.0 63.3

100.0

100.0

100.0

(63-73) shows that the similar phonemic discrimination performance is not a ceiling effect. The error pattern, also, was comparable in the right and left hemidecorticates. Voiced sounds (plosives, continuants, or nasals) were more difficult to discriminate than unvoiced sounds (plosives and continuants) by a ratio of 6: 1. Continuants were slightly more difficult than plosives or nasals. Each child on the first trial was able to produce all the phonemes and blends of English: no articulation defect was apparent to the speech pathologist who performed a blind analysis of the phoneme production data. Peabody Picture

Vocabulary

Test (Dunn, 1965)

This test measures comprehension vocabulary. The subject must identify a heard word from a4-choice display. The scores are expressed as a test mental age and test IQ. Word Discrimination

Test (Goodglass

& Kaplan,

1972)

The comprehension of words in six semantic categories (objects, geometric forms, activities, letters, colors, and numbers) is assessed. Responses are rated for both correctness and latency. The score is the percentage of the total possible score on the test. Visual Confrontation

Naming

(Goodglass

& Kaplan,

1972)

This measures the production of words in the six semantic categories of the previous test. The stimulus material in the Word Discrimination Test

412

DENNIS

AND WHITAKER

must be named. Both correctness and response latency are measured. Scores are the percentage of the total possible test score. Responsive

Naming

(Goodglass

& Kaplan,

1972)

Auditorily cued word retrieval is assessed here. Correctness and latency comprise the score for each response. The score is the percentage of the total possible score. Naming

Fluency (Goodglass

& Kaplan,

1972)

The subject must name as many animals as possible in 90 sec. The starting word “Dog” is given to help the subject initiate a response. The score is the number of animal names produced in the most fluent consecutive 60-set period. The results of the word discrimination and production tests are in Table 3. The performances of the three hemidecorticates are similar. The PPVT scores are all within the normal range. The difference between MW and the two left-operates is not a significant one, i.e., it is of a magnitude which could occur on repeated testing of one subject. Word discrimination is at ceiling level. Word production also appears to be well developed in each of the children. With both visual and auditory cues they are able to generate TABLE

3

WORD DISCRIMINATION AND PRODUCTION MW (RH)

SM (LH)

CA (LH)

Age at test (years-months) Test

PPVT Other tests

Peabody Picture Vocabulary (Dunn, 1965) PPVT mental age PPVT IQ Word Discrimination (Goodglass & Kaplan, 1972) Percentage correct Visual Confrontation Naming (Goodglass & Kaplan, 1972) Percentage correct Responsive Naming (Goodglass & Kaplan, 1972) Percentage correct Naming Fluency (Goodglass & Kaplan, 1972) Number of names in 60 set

9-5 IO-11

7-11 9-6

7-11 9-7

9- 10 106

7-3 93

7-5 95

94.4

98.6

98.6

100.0

98.0

99.0

100.0

83.3

90.0

14

16

10

LANGUAGE

ACQUISITION

BY ONE

413

HEMISPHERE

names for actions, objects, colors, numbers, letters, and forms. Naming was not tested beyond the difficulty level of the Boston Diagnostic Aphasia Examination so it does remain possible that the subjects would differ on more demanding naming or word retrieval tests. Naming fluency, a task with no ceiling, is at a comparable and normal level in the three children: 10 year olds typically produce about 12 animal names in 60 set (Goodglass and Kaplan, 1972). Test of Syntactic

Comprehension

(Parisi & Pizzamiglio,

1970)

The child’s ability to understand a variety of syntactic forms is measured by this test. Sentences are read to the subject who must point to that section of a bipartite picture which shows what was heard. The sentences sample 20 different syntactic contrasts, from simple (on vs. under) to complex (direct-indirect object; between vs. beside). The score is the percentage of the 80 items correctly discriminated. Active-Passive

Test (Dennis & Kohn, 1975)

This test measures the ability to discriminate the meaning of heard statements in four different voices: active affirmative, passive affirmative, active negative, and passive negative. After hearing a statement the subject sees a picture which illustrates the subject-object relationship of the statement, in one segment, and an alternative subject-object relationship, in the other. The task is to indicate the segment which shows what was heard. The scores are the percentages of correct syntactic discriminations for the 32 trials of each voice. Story Completion

Test (Goodglass,

Gleason, Bernholtz,

& Hyde, 1972)

The productive control of syntactic forms is measured here. Sentences of a simple story situation are presented orally to the subject who must add a highly predictable final “target” sentence or phrase. The constructions tested range from simple imperative intransitives to embeddings. Each of the 46 responses was scored as: + , correct target revealing productive control of the grammatical construction; -, ungrammatical response violating the rule being tested (e.g., Cue: “I sold her a small car. The car was red. In other words I sold her. . . ?” Response: “A red small car”); NT nontarget response, clearly not directed towards the same ends as those envisaged by the examiner (e.g., Cue: “Mr. Jones wants to hear the news. The radio is off. What happens?” Response: “He goes out and buys a newspaper.“); G, grammatical response to the statement which does not indicate whether the subject can produce the rule being tested (e.g., Cue: “Peter likes Jimmy but Jimmy doesn’t like. . . .” Response: “Peter.” When it is the structure of language, rather than language content, which

414

DENNIS

AND WHITAKER TABLE

4

SYNTAX DISCRIMINATION AND PRODUCTION MW (RH)

SM (LH)

CA (LH)

Age at test (years-months)

Tests

TSC AP SCT

Test of Syntactic Comprehension (Parisi & Pizzamiglio, 1970) Percentage correct Active-Passive Test (Dennis & Kohn, 1975) Percentage correct Active affirmatives Passive affirmatives Active negatives Passive negatives Story Completion Test (Goodglass, Gleason, Bernholtz, & Hyde, 1972) Number of + Number of Number of G Number of NT

9-4 9-4 10-5

7-11 8-O 9-2

9-6 9-6 9-1

94.4

90.1

93.4

100.0 100.0 96.9 100.0

100.0 50.0 90.6 50.0

90.6 50.0 84.4 71.9

39 2 4 1

36 1 7 2

36 2 7 1

must be processed, the right and left hemidecorticates are at different developmental levels (Table 4). All three perform well on the Test of Syntactic Comprehension. The interpretation of this result is complicated by several factors: children by the age of 6 are quite proficient at this task (Parisi, 1971); there is a wide range of difficulty in the syntactic contrasts sampled (very young children, for example, discriminate on vs. under or up vs. down); and there are only four samples of each syntactic type so that the probability of getting a contrast correct by chance is high. Nevertheless, analysis of the error patterns revealed differences between the left and right hemidecorticates. All three were perfect in their identification of differences between simple locative prepositions (behind/in front of, on/under, near/far, in/out, up/down); all three made some errors on the more difficult locative contrasts (to/from, behind/beside, beside/between). All subjects found “between vs. beside” and “present vs. past tense” most difficult; these two items rank 9th and 17th (/20) in difficulty, respectively, for normal 3- to 6-year-old children. On items requiring analysis of verb tense, MW scored only 75% while SM and CA scored 85%. The discrimination between present and future tense accounted for these differences. On those items requiring the processing of permutations of word order (passives, direct and indirect object constructions) MW scored

LANGUAGE

ACQUISITION

BY

ONE

HEMISPHERE

415

lOO%, while SM and CA scored 88%. MW, in fact, was the only child to make consistently correct choices on items where meaning was revealed by word order (actives, passives, direct-indirect object, from/to, subordinate phrase). MW’s greater capacity to attend to syntactic forms permuting word order is shown in the children’s performance on the Active Passive Test. Only MW is able to maintain a proficient level of performance throughout passive affirmative and passive negative structures. These syntactic forms usually debilitate the discrimination of SM and CA to chance levels. The productive control of the grammatical forms tested is similar. Each child is successful at producing the various grammatical forms in the Story Completion Test: there were few errors and few nontarget responses. Token Test (De Renzi & Vignolo,

1962)

This test measures the ability to execute a variety of complex verbal commands varying in information and syntactic complexity. In parts 1 to 4, the adjectival content of the noun phrase increases systematically but the commands are syntactically the same. In Part 5 a variety of different verbs and noun phrase structures are introduced into the predicates, with the effect of complicating the syntactic features of the commands. Scores are given for the percentage of correct responses on each of the five parts of the test. All hemidecorticate subjects deal well with the informationally complex commands of Parts 1 to 4, but only MW maintains competent performance in the face of the increased syntactic demands of Part 5 (Table 5). A global inability to handle complex verbal material or immediate memory problems could be expected to debilitate performance most markedly on those commands with the greatest information load, i.e., those of Part 4. But it is syntactic rather than informational complexity which degrades the scores of both the left hemidecorticates. When language acquisition is assessed by such gross indices of function as the WISC Verbal IQ or the ITPA Psycholinguistic Age, the right and left hemispheres appear to be at similar developmental levels by the ages of 9 and 10. Yet these global measures do not reveal the widely discrepant development of some language functions. In these cases, the isolated left hemisphere is much more sensitive than the right to the grammatical structure of English. Experimental

Semantic

Anomalies

and Syntactic

Language

Tests

Errors

These tests measure the ability to detect and to correct anomalies in meaning and errors of syntactic structure. The subject is asked to decide

416

DENNIS

AND WHITAKER TABLE

COMPREHENSIONOF

5

COMPLEX

COMMANDS

MW (RH)

SM (LH)

CA (LH)

Age at test (years-months) Test Token Test (DeRenzi & Vignolo, 1962) Percentage correct Part 1 (Touch the red circle) Part 2 (Touch the small red circle) Part 3 (Touch the yellow circle and the red rectangle) Part 4 (Touch the small blue circle and the large green circle) Part 5 (Touch the yellow circle with the blue rectangle, etc.) Total score

9-4

8-O

8-O

100.0 100.0

100.0 90.0

100.0 100.0

100.0

100.0

100.0

90.0

70.0

100.0

90.9

63.6

77.3

95.2

80.6

91.9

whether a heard statement is acceptable or not, and then to correct any statement judged “unacceptable.” The Semantic Anomalies Test comprises 12 acceptable and 12 anomalous sentences, randomly mixed. Most test for common knowledge of the properties of familiar things such as cars, trains, foodstuffs, and household objects (e.g., *“My favourite breakfast is radios with cream”); two require attending to word order (e.g., *“Cash shouldn’t send people through the mail since it might be stolen”). The Syntactic Errors Test also comprises 12 grammatical and 12 ungrammatical sentences, randomly mixed. The syntactic forms tested include number agreement, complementizer choice, personal pronoun, preposition, adverb order, subordinating conjunction, some-any alteration, and determiner choice. The results are in Table 6. No difficulty with the Semantic Anomalies was apparent. MW and SM made no erroneous judgments and CA made only two (one of which was the word-order problem mentioned above). However, both left hemidecorticate subjects performed poorly on the Syntactic Errors Test. MW’s performance was perfect both in identification and correction. The fact that SM and CA made errors of both types, judging ungrammatical sentences to be correct and grammatical sentences to be incorrect, clearly indicates that they were not adopting a strategy of replying “O.K.” to each stimulus sentence. They were attempting to make syntactic judgments but were doing so erroneously. As would be expected from the high number of errors, CA and SM had eight errors of failure to detect ungrammaticality in common. None of their five errors of

LANGUAGE

ACQUISITION TABLE

BY ONE

417

HEMISPHERE

6

DETECTION OF ANOMALY AND ERROR (N ERRORS) MW

Semantic anomalies Anomalous sentence judged acceptable Acceptable sentence judged anomalous Syntactic errors Ungrammatical sentence judged grammatical Grammatical sentence judged ungrammatical

(RH)

SM (LH)

CA (LH)

0

0

2

0

0

0

0

9

11

0

3

2

erroneously judging a grammatical sentence to be ungrammatical was the same. Following the test of Syntactic Errors, both SM and CA were asked to correct the ungrammatical sentences; they were told, “Fix up the sentence in your own words so that it will sound better.” SM was reluctant to attempt this, responding that he didn’t know or wasn’t sure in most of the sentences. For the ungrammatical sentences he identified and for which he provided his corrected version, it was not the syntactic error that he addressed but rather a factual, real-world aspect of the sentence content. Thus, given these ungrammatical sentences, t

* The best cars in Canada is a Ford and some Datsun. * The quickly sun sets in the winter.

his corrected versions were, a-SM “Wrong, it’s a Mercury.” b-SM “Wrong, if the snow falls down, it melts.” SM also applied the same strategy when he erroneously judged grammatical sentences to be ungrammatical. To the following stimulus sentences, I wouldn’t buy just any old house since some them are leaky. ifi The best trains in Canada are the CN and CP. his responses were, c-SM

“Wrong, if the rain comes down, water comes into the house.” d-SM “No, it should be the GO-train” (a commuter train local to the Toronto area). CA did venture to repeat many of the ungrammatical sentences on our

418

DENNIS

prompting. grammatical

AND

WHITAKER

In so doing, she spontaneously corrected errors she had failed to identify. Thus,

(repeating (repeating

some of the

(a)): “The best cars in Canada is Fords and Datsuns”. (b)): “The sun sets quickly in the winter.”

There is somewhat of a dissociation between her ability to make me&linguistic judgments and some (although demonstrably not all) of her spontaneous utterances. The results of the Semantic Anomalies and Syntactic Errors tests show that the two hemispheres differ in what they will consider an acceptable English utterance. A sentence is acceptable for the left hemidecorticates if it is consistent with real-world states of affairs, if the properties of objects are what they should be. Acceptability for the right hemidecorticate involves an additional factor: the utterance must be consistent with the grammatical structure of the language. Sentence Repetition

Test

In order to assess the way different grammatical forms are recalled, a sentence repetition test was constructed. Immediately after hearing them, the subjects were asked to repeat sentences of various types: active affirmative (“The angry old lion bit the frisky young tiger”), active negative (“The kind nurse didn’t tell the fat new doctor”), passive affirmative (“The black cat was chased by the brown dog”), passive negative (“The tall guard wasn’t shot by the armed robber”), question (“Did the black horse lead the fat brown cow?“), negative question (“Didn’t the blue car bump the shiny red train?“), negative passive question (“Wasn’t the bank robber shot by the tall guard?“), indirectdirect object (“The library gave the books to the new school”), clefted subject (“It’s the thin man that paid the blond lady”), clefted object (“It’s the tall guard that the bank robber shot”), pseudoclefted object (“What the big stone broke was the kitchen window”), pseudoclefted subject (“What fell on the red house was the tree”), pseudoclefted indirect object (“What the robber stole money from was the bank”), pseudoclefted passive object (“What the Army was given planes by was the Navy”), and pseudoclefted passive indirect object (“What was lent money by the bank was the Church”). The subject-object relationship was reversible in some sentences, nonreversible in others. Verbatim recall was stressed. The 41 sentences were composed of common, high-frequency nouns, verbs, and adjectives. Each sentence had 9 words, except for the pseudoclefted passives, which had 10. The variation in syntactic structure revealed important differences in the repetition capacities of the subjects. MW repeated 36 of the 41 sentences verbatim, while SM and CA correctly repeated only 25 and 6, respectively.

LANGUAGE

ACQUISITION TABLE

SENTENCE

REPETITION:

7

ERROR

Error type

419

BY ONE HEMISPHERE

(N ERRORS)

ANALYSIS

MW (RH)

Substitution of lexical word Substitution of function word Reordering, grammatical result Reordering, ungrammatical result Omission, grammatical result Omission, ungrammatical result Change in polarity Major syntactic error No response

SM (LH)

CA (LH) II 26 2 I 4 5 5 I7 0

0 I 0

SM constantly asked for a second or even third trial for the sentences. He was scored on his best production; his score would be considerably lower had he been rated only on first production. CA tended to repeat the stimulus sentences the first time she heard them. This difference in strategy may account for the difference in total score between the two left hemidecorticates. The analysis of the repetition errors is shown in Table 7; Table 8 presents the errors the three subjects made on some syntactic forms. MW is proficient at the test; in only two sentences did his response violate a syntactic rule or produce an ungrammatical result. SM and CA, by contrast, often gave ungrammatical responses. They made major syntactic errors and even their omissions resulted in ungrammatical utterances. The passive structure and the pseudocleft are clearly quite difficult for SM and CA to repeat: CA in addition had some difficulties with cleft structures. If one considers the relationship between these syntactic forms and the standard subject-verb-object (SVO) word order, a further detail of this pattern emerges. All the passives reverse the normal SVO word order; clefting on the object noun phrase and pseudoclefting on the subject noun phrase also alter the normal SVO order. All pseudoclefted passive sentences have an uncharacteristic word order. If we consider only the errors on sentences which do not have the SVO order, we find that MW TABLE SENTENCE

REPETITION:

THE EFFECT

8 OF SYNTACTIC

Syntactic form

MW (RH)

SM (LH)

Passives (14) Clefts (8) Pseudoclefts (13)

I 0

0

2

9

7

FORM

(N

ERRORS)

CA (LH) 13 7 II

420

DENNIS

AND

WHITAKER

erred in repeating only one, a passive clefted on the indirect object (“What was lent money by the bank was the Church”). MW’s response was nevertheless a pseudoclefted passive. Both SM and CA missed this sentence. SM did not respond at all and CA’s response was not a correct pseudocleft (“What was lent money from the bank from the Church”). The one correct cleft sentence and the two correct pseudocleft sentences by CA were all in the normal SVO order. She failed to repeat accurately any of the non-SVO ordered sentences. SM repeated correctly all the cleft sentences. Of his four correct pseudocleft sentences, two were in a non-SVO order. Otherwise, he found non-SVO sentences difficult to repeat. The three subjects had performed equally well on the WISC Digit Span subtest, and all three produced nine and ten word sentences using the same vocabulary items as in the Sentence Repetition Test. The principal determinant of repetition accuracy was not sentence length, but the variation in syntactic structure, primarily on passive and pseudoclefted sentences and to some extent on clefts. The left hemidecorticate children repeated these sentence types less well than MW. The deviation from normal SVO word order is probably one factor affecting performance for these two subjects. It is also reasonable to suppose that the additional function words, as well as their unusual juxtaposition in the pseudoclefts, may have contributed to the decrement in performance. Sentence Recall

Test (based on Newcombe

& Marshall,

1%7)

This test was also one of immediate sentence repetition, stressing verbatim recall. The 40 sentences include the 20 original sentences from Newcombe and Marshall (1%7) and 20 more constructed to parallel them: the same number of words, the same syntactic structure, and the same level of vocabulary difficulty. The sentences were of five types: (1) fully grammatical and meaningful sentences of from five to nine words varying in syntactic complexity, (2) grammatically well-formed sentences which violated various semantic selectional constraints and were thus anomalous, (3) reversible and nonreversible passive sentences which were grammatical and semantically well-formed, (4) sentences containing four adjectives modifying a single noun, in the correct and incorrect order, and (5) random word strings. Both CA and SM made more errors than MW on this test. SM incorrectly repeated 24, CA 29 of the 40 sentences. The nature and number of errors for the various sentence types are shown in Table 9. Most of MW’s errors were either omissions or semantically related substitutions of adjectives and adverbs. Neither his omissions nor his substitutions resulted in changing the syntactic structure of the sentence. Only two of his errors could be described as normalizing a deviant string. One involved changing

LANGUAGE

ACQUISITION TABLE

SENTENCE

RECALL

BY ONE HEMISPHERE

421

9

TEST: ERROR AND SENTENCE

TYPE

Subject MW (RH) Sentence type Four adjectives in wrong order Four adjectives in correct word order Random word string Random word string Semantic anomaly Semantic anomaly Semantic anomaly Semantic anomaly

Error type

N

Changed to correct order Omitted one or more adjectives

I 4

Omitted one or more words No response Phoneme substitution resulting in different word Omitted an adjective or an adverb Changed to nonanomalous sentence Changed relative pronoun (which -+ that)

2 2

Subject SM (LH) Sentence type Four adjectives in wrong order Four adjectives in wrong order Four adjectives in correct word order Random word string Random word string Semantic anomaly Semantic anomaly Well-formed sentence Well-formed sentence Well-formed sentence Well-formed sentence

Error type Omitted one or more adjectives; ungrammatical response Began but didn’t finish Omitted one or more adjectives Omitted one or more words Began but didn’t finish Began but didn’t finish; ungrammatical fragment No response Omitted words but structure retained as stimulus Substituted words; ungrammatical response Began but didn’t finish; ungrammatical fragment Substituted words; grammatical response

N 1 I 4 2 2 3 I I 3 5 I

Subject CA (LH) Sentence type Four adjectives in wrong order Four adjectives in correct word order Random string Random string Random string Semantic anomaly

Error type Omitted one or more adjectives; grammatical response Began but didn’t finish; ungrammatical fragment Began but didn’t finish Omitted and substituted words No response Began but didn’t finish; ungrammatical fragment

N

422

DENNIS

AND WHITAKER

TABLE

9 (Continued)

Subject CA (LH) Sentence type Semantic anomaly Semantic anomaly Semantic anomaly Well-formed sentence Well-formed sentence Well-formed sentence Well-formed sentence

Error type

N

Substituted and omitted words; ungrammatical response Spoonerism Substituted words; grammatical response Substituted words; grammatical response Substituted and omitted words: ungrammatical response Phoneme substituted; nonword resulted No response (passive sentence)

3 1 2 4 9 1 1

an incorrect adjective order to a correct one and the other switched subject and object nouns in an anomalous sentence to make it well-formed. With one exception, neither SM nor CA normalized deviant sentences. In one sentence, CA omitted some of the adjectives in a wrongly ordered adjective string and her resultant utterance was in fact grammatical. Rather than normalizing, SM and CA tended to produce ungrammatical utterances, not only when the stimulus sentence was ungrammatical but also when it was fully grammatical. Eight and eleven, respectively, of their repetitions of well-formed sentences were ungrammatical. Both SM and CA made a number of incomplete responses, i.e., they would begin to repeat the sentence and then stop and fail to complete it. These elliptical fragments were also ungrammatical. Two of the test sentences are indicative of the nature of the problems SM and CA had with well-formed sentences. MW repeated both of them correctly. e Wasn’t the stone wall built by the kind husband‘? f Wasn’t the poor cousin helped by the old lady? The responses are as follows (. . . indicates a hesitation child spoke the first sentence as a question. e-SM e-CA f-SM f-CA

pause). Neither

“Wasn’t by the. . . .” “Weren’t the stone . . . built a house.” “Wasn’t by the cousin helped by the old lady?” “Wasn’t the poor cousin . . . helped the old lady?”

SM and CA’s repetitions preserve the lexical content of the stimulus sentences. Function words, typically, are not accurately repeated. For example, in sentences 26 and 40 CA does not repeat “by” at all; SM is

LANGUAGE

ACQUISITION

BY ONE

HEMISPHERE

423

uncertain of its location in the sentence and so puts it in twice. Neither left hemidecorticate is properly processing the function words which signal surface structure syntax. By Test

This is a test of the various usages of the function word “by.” The word is polysemous, marking a temporal, locative, or passive agentive adverbial phrase. The test required our subjects to repeat sentences using “by” in each of its three senses. Our interest in how the children handled “by” was motivated both by the left hemispherectomies’ deficient passive voice discrimination and by their difficulty in correctly placing the passive agentive “by” in the Sentence Recall Test. MW, the right hemidecorticate, repeated all the “by” sentences correctly. SM repeated all the passive and locative “by” sentences correctly, but changed all the temporal uses to either “in” or “at” (e.g., “by ten o’clock” was repeated as “at ten o’clock”). CA seemed to have great difficulty with this test, only repeating one sentence verbatim out of nine. Her error on one of the passive sentences was a change in word order switching the locative and agentive phrases; her error on one of the locative sentences was the substitution of another noun in the locative phrase. Her other six repetitions were ungrammatical, in addition to the substitution of different words for “by.” “From” and “near” were substituted for the locative uses of “by,” and “from” and “for” were substituted for the temporal “by” use. The results of this test extend previous findings that certain function words are difficult for SM and CA to repeat. Their repetitions reveal deficient processing rather than impaired memory for the statement. That SM and CA have great difficulty with temporal use of “by” is interesting. Young children, also, link causal clauses by means of causal rather than temporal connectives (Katz & Brent, 1968). SM and CA repeated the passive sentences in the “by” test with relative ease (in contrast with their difficulty on the longer and/or semantically more complex Sentence Recall sentences). Since SM and CA do not comprehend the meaning of passive sentences of the same length and vocabulary difficulty as those in the ‘By’ Test, it is evident that the ability to repeat a syntactic form need not evidence grammatical understanding of that form. The significance of the relationship between repetition and comprehension is discussed more fully below. Several of the previous tests suggested that some of the functional information in a statement is extracted by MW but not by SM or CA. It seemed important, then, to explore more systematically just what each child did manage to derive from a heard statement. This was the purpose of the next set of tests.

424

DENNIS

Tag Questions

AND WHITAKER

Test

Tag questions were elicited from the subjects in this test in order to determine which elements of a cue statement they can identify. The original idea of eliciting tag questions by role-playing is from Langendoen (1970). The concept of a tag is introduced by two puppets, with the experimenter and subject taking the roles of a lion and a hippo, respectively. In our clinical experience, it is easier to induce a child to adopt the role of having said a statement in fact uttered by someone else if puppets are used to establish the alternate roles. The experimenter’s puppet makes a statement (e.g., “Jill is very smart,“). The child, in the role of the hippo, is required to add the tag question to the experimenter’s statement (e.g., “Isn’t she?“), as though he or she had actually made the original statement. This test assesses the subject’s ability to identify the person, number and gender of the sentence subject and to select the corresponding personal pronoun; to establish whether the sentence is affirmative or negative and to reverse the polarity of the tag; to determine if the sentence has an auxiliary verb or not; to use either the auxiliary verb or a do-support in the tag; and to make the do-support agree in tense and number with the main verb of the sentence. Correct tag questions are based upon all of these features of the sentence. Errors are shown in Table 10. The children differed in their ability to reverse the polarity of the tag. CA and MW’s tags reliably (but erroneously) retained the polarity of the cue but SM correctly reversed it. Aside from polarity, MW extracted more information from the statement. He made only three errors of person and one of number. The 11 errors SM made are distributed over all the measured categories. CA made 24 errors of person, number, use of the auxiliary verb, and tense. The semantic and syntactic features required for the construction of tag TABLE TAG

QUESTIONS

Error type Number Gender Auxiliary verb Polarity Tense Do-support Inversion auxiliary and pronoun There

TEST

ERRORS

MW (RH)

10 (N

ERRORS/38

SENTENCES)

SM (LH)

CA (LH)

3 1 0 0 33 0 0

9 4 0 9 37 2 0

0 0

0 0

LANGUAGE

ACQUISITION

BY ONE

HEMISPHERE

425

questions are part of those aspects of surface structure English associated with the late rules of the grammar: person, number and gender agreement, auxiliary do-support, and word-order constraints. CA, especially, does not appear to have acquired these grammatical rules. Implication

and Presupposition

Test (Kelleher,

1974)

This test measures the subjects’ ability to determine the presupposition and implication of four implicative verbs, remember, forget, manage, and careful, in sentences where the implicative verb and the complement verb could be affirmative or negative. The presupposition depends on whether the complement sentence is affirmative or negative, e.g., for “Max was careful not to eat an apple” the presupposition is that he was not supposed to eat it; and for “John remembered to close the window” the presupposition is that he should have done so. The implication depends on the interaction of affirmativity and negativity of the implicative verb and its complement. When these are both affirmative (“John managed to cut his toenail”) or both negative (“Mary didn’t manage not to get wet”) the implication is that the action occurred; when the polarity differs between the implicative verb and its complement, the implication is that the action did not occur. “Forget,” of course, follows this pattern in reverse, since it is inherently negative. There were 16 tests of presupposition and 16 of implication. All three children were proficient in identifying the presupposition (MW 15/16, SM 14/16, and CA 14/16). The ability to determine the implication, however, differed in the three children (MW 14/16, SM 8/16, and CA 9/16). The task of deciding that a complement verb is positive or negative is one that can be done by all the children (comprehension of affirmative and negative active statements on the Active-Passive Test was also equally proficient in each). But when part of the meaning of a statement depends on assimilating the polarity features of the two predicates, only MW performs correctly; judgments of the two left hemidecorticates are at chance levels. Presupposition and implication are important components of the conversational use of language. Clearly, all of what we understand of a speaker’s utterances is not literally expressed in what is stated; utterances refer to contexts that are implicitly understood by both speaker and hearer. Our results indicate, therefore, that despite the overt normalcy of CA and SM’s informal discourse, there may be deficiencies in how they understand conversation that are not obvious from their global interactive behavior. Pronoun

Substitution

Test (Whitaker,

1971)

Performance on this test requires the integration of both semantic (the features of person, number and gender) and syntactic (the grammatical category features of pronoun and case agreement) information. The

426

DENNIS

AND

WHITAKER

subject and examiner read together a short narrative passage. The subject’s task is to replace underlined noun phrases with the appropriate pronouns. The test items involve instances of masculine and feminine personal pronouns in subjective and possessive cases, plural personal pronouns in subjective and objective cases, the neutral pronoun “it” and the indefinite animate pronoun “one.” The test is not begun until the subject demonstrates recognition by reading aloud the story without error. Since the entire story is on one page, it is always in the subject’s view while he or she is doing the task. These procedures reduce the load on linguistic memory and, also, maximize the subject’s ability to use contextual cues. MW is linguistically more sophisticated than either SM or CA with respect to being able to recognize relationships between pronouns and noun phrases. He produced all tested pronouns except “one,” and even in that instance his response resulted in a grammatical sentence. Both left hemidecorticates also failed to do the one-pronominalization: SM was unable to produce any response and CA responded with an ungrammatical sentence. They made other errors as well. SM did not produce any plural pronouns and CA showed some difficulty with subjective-objective contrasts. Neither SM nor CA had any difficulty with possessive or with masculine-feminine contrasts. A test like this bears a special relationship to linguistic abilities and language acquisition, since all three subjects produce the target pronouns in their conversational speech and appear to comprehend the pronouns when spoken by others in conversational speech. Clearly, however, conversational control over a linguistic form does not entail manipulativeanalytic competence with that form. Hierarchical

Clusters

Test (Zurif,

Caramaua,

& Myerson,

1972)

How do MW, SM, and CA think the words in a sentence are interrelated? What is the basis, for example, on which they link function words with the other words in the sentence? The children were asked to judge how the words in a written sentence went together. They were given all the three-word left-to-right triads (both adjacent and nonadjacent words) from six five-word sentences (“the dog chases a cat,” “the train hits a car, ” “the teacher sees a child,” “Ann is paid by Joe,” “John is pushed by Mary, ” “policemen are threatened by robbers”), and asked to decide on the two words in each triad which went best together and the two which were least related. The written sentence was always in view during the test, and the examiner repeated it at intervals. A word-relatedness matrix was tabulated and an algorithmic scaling procedure used to induce a phrase structure tree (Zurif, Caramazza, & Myerson, 1972; Martin, 1970). The tree graphs from the minimum distance matrix are shown in Fig. 1. The children did attempt to make the required metalinguistic judgments:

LANGUAGE

ACQUISITION

BY ONE HEMISPHERE

427

FIG. I. Word relatedness judgments (top line: active sentences; bottom line: passive sentences).

they did not cluster words randomly or on the basis of spatial proximity. MW and SM sort active sentences in a reasonably normal fashion (Zurif & Caramazza, 1975). Both choose to organize the sentence as ((SV)O) rather than (S(V0)). One of the determiners is not integrated into the rest of the sentence for MW, but SM establishes a tight association between the nouns and their determiners. CA has a tight determiner-noun bond for the first noun, although not for the second. Her verb is not placed in the sentence. The determiner-noun association is the strongest relationship in the sentence for SM and CA; but for MW the link between the agent and verb is the most salient feature. The passive sentence clusters are different in the right and left hemidecorticates. MW recognises the preposition which marks the sentence agent and he associates it with the agent. SM and CA do not. SM is unable to relate “by” to any sentence element, suggesting that he is uncertain of its role in the sentence. CA links “by” with the first noun. It is possible that she recognises the fact that “by” marks the sentence agent but, being unable to process the syntactic permutations of the sentence, erroneously associates “by” with the first noun phrase that could semantically act as agent. The three children, it is evident, differ in their capacity to make metalinguistic judgments. On active statements, they have a somewhat different organization. The difference in determiner-noun and noun-verb clustering can be seen as consistent with the interpretation that SM and CA are making their judgments on the basis of surface structure information but that MW sorts more often on the basis of deep structure/semantic relationships. The meaning of the passive sentence depends on the lexical role of the preposition “by.” Significantly, MW has formed a strong link between the noun and the passive marker which signals its semantic role,

428

DENNIS

AND

WHITAKER

but SM and CA have not. The two left hemidecorticates, it appears, have a knowledge of the role of function words in a passive statement which is no better than their discrimination of such statements’ meanings. DISCUSSION

A less complete language capacity has developed in these subjects by the end of the first decade of life in the right hemisphere than in the left, even when language has been acquired under seizure-free and educationally normal conditions. The impairments of right hemisphere linguistic processing do not reflect a global language breakdown nor a deficiency of verbal cognitive capacity; and they are evident only when particular linguistic functions are taxed. Different configurations of language skills have developed in the two isolated hemispheres: phonemic and semantic abilities are similarly developed, syntactic competence has been asymmetrically acquired. Phonemic discriminations are made accurately by all three children, and each has normal articulation. It is interesting to note that phonemic production is disordered with a similar frequency after left and right hemisphere damage in infancy. Articulatory disorders are equally probable in right and left infantile hemiplegics (Bishop, 1965). The two perinatal hemispheres appear to be equally good substrates for the acquisition of phonemic competence, just as they are similarly at risk for phonemic disorder. Either hemisphere, furthermore, is an adequate substrate for semantic skills. The three children are equally good at producing and discriminating words. Even relatively subtle semantic features can be identified. Verb presuppositions, for example, are discriminated well. Each hemisphere, it appears, has developed a normal lexicon and has no gross defect in retrieving the words contained in it. Syntactic abilities have not been acquired as well by the isolated right hemisphere as by the left. In relation to the left, the right brain half is deficient in understanding auditory language, especially when meaning is conveyed by syntactic diversity; detecting and correcting errors of surface syntactic structure; repeating stylistically permuted sentences; producing tag questions which match the grammatical features of a heard statement; determining sentence implication; integrating semantic and syntactic information to replace missing pronouns; and performing judgments of word interrelationships in sentences. What is the characteristic of the right hemisphere defect? It appears to be an organizational, analytical, syntactic, and hierarchic problem rather than a difficulty with the conceptual or semantic aspects of language. As such, it is evident on tests which require sensitivity to sentence structure or to function words. The right hemisphere impairment can be demonstrated in

LANGUAGE

ACQUISITION

BY ONE

HEMISPHERE

429

many contexts: production, repetition, comprehension, awareness of anomaly, and word-relatedness judgments. The diversity of the test procedures allows us to trace the same language function throughout different situations. Such an analysis was made for passives. The right hemisphere does not accurately comprehend the meaning of passive statements. This defect is associated with a failure to identify the structural role in passives of the function word “by.” The right hemisphere’s judgment of grammaticality is correspondingly inaccurate. The children were asked to assess the acceptability of the following three sentences: g h i

* I paid the money by the man. * I was paid the money to the lady. I was paid the money by the boy.

They responded, g-MW g-SM g--CA h-MW h-SM h-CA i-MW i-SM i-CA

“That’s silly . . . because you can’t pay money by a man” (MD: “Can you fix it up?“) “I paid the money to the man.” “It’s alhight.” “No . . . you shouldn’t pay money to a stranger.” “That’s silly because it’s not a good way of saying it” (MD: “Can you fix it up?“) “I was paid the money by the lady.” “That’s allright.” “O.K.” “It’s O.K.” “It’s allright.” “O.K.”

Both isolated right hemispheres are insensitive to those sentence characteristics which establish agent and object. For this syntactic dimension, the left hemisphere has both implicit knowledge and overtly understood control. So far, the analysis of passive sentences is consistent. Deficient comprehension is mirrored in impaired perception of sentence structure and inaccurate judgments about examples of the grammatical form. However, all three children did produce passive statements in response to the eliciting cues of the Story Completion Test. It is difficult to accept this as a dissociation between production and comprehension. The passives elicited by the Story Completion Test are GET-passives which are related to the form get + predicate adjective (get sick, get hit). The GET-passive is acquired earliest in ontogeny, followed by passives without the agent, followed finally by agentive passives. In addition, the Story Completion Test passives are nonreversible (man hit by train, girl bitten by dog) while the Active-Passive Test sentences are reversible (dog chased by cat, car bumped by train). Only in the latter passives is there a syntactic problem in

430

DENNIS

AND

WHITAKER

establishing agent and object. MW is the only child to have produced, spontaneously, a nontruncated standard reversible passive (sentence h- MW above). SM and CA sometimes repeated passives whose meaning they did not understand. This accurate repetition may be accomplished by a word-list strategy, which would not require processing of the syntactic structure. When the complexity of the passive sentence is increased beyond the capacity of the word-list strategy, SM and CA make repetition errors. The fact that they then produce ungrammatical repetitions suggests that they do not analyze meaning and structure. It is likely that MW reliably processes structure. Even when he could not give verbatim recall, his utterances maintained the sentence meaning. The tests did provide two clear instances of a production-comprehension dissociation. CA’s spontaneous restatements of ungrammatical utterances were in two cases grammatical, even though her judgments of the grammaticality’ of those statements were wrong. Both SM and CA produced some target pronouns in their conversational speech but not in the Pronoun Substitution Test. In this context, it is interesting that children use the /az/ plural correctly before they are able to generalize it to new situations (Berko, 1958); and also that the correct spontaneous use of connectives in speech precedes the overtly understood manipulative linguistic control of these forms (Katz and Brent, 1968). The ability to produce or repeat a grammatical structure is not itself a guarantee of understanding, correct grammatical judgment, manipulative control or metalinguistic knowledge of that structure. The described limitations of right hemisphere language do not find a simple parallel in other clinical groups. The differences between SM and CA, on the one hand, and developmental dysphasics, childhood aphasics, or adult aphasics, on the other, seem more salient than the similarities. The right hemisphere language defect is not a general developmental retardation, and it is selective. Phonemic and semantic functions are well developed in the right hemispheres; syntactic skills are not. As an instance of this, the impairment of SM and CA on the Token Test was one of syntax rather than content. Developmental dysphasics, by contrast, are impaired on the Token Test in proportion to the informational complexity of the command; their performance actually improves when syntactic complexity is increased but information level is reduced (Tallal, 1975). Nor are the described defects comparable to childhood aphasia. In the latter condition, it is the expressive aspects of language which are primarily affected (Guttman, 1942; Alajouanine & Lhermitte, 1965). Expressive language was not selectively affected in the isolated right hemispheres; indeed, in some instances, expressive skills were better than comprehension. The word-relatedness judgments of the right hemispheres are unlike those of Broca’s aphasics. SM and CA responded to the determiner-noun

LANGUAGE

ACQUISITION

BY ONE

HEMISPHERE

431

relationship in the sentences but were unsure of the role of the syntactic marker “by.” Broca’s aphasics have a reverse pattern of performance: they are more sensitive to lexical contrasts like “to vs. by” than to determiners (Zurif & Caramazza, 1975). Since the lexical information in “by” gives information not otherwise available in the sentence, Broca’s aphasics appear to be responsive to that structural feature necessary to convey meaning. SM and CA are not. Sentence repetition is different in the isolated right hemispheres and adult aphasics. The latter subjects preserve the syntactic and semantic structure of the sentences with well-formed material and normalize the structure when the sentence is deviant (Newcombe & Marshall, 1%7). The impaired Sentence Recall of SM and CA, it is evident, is not of this kind. Their errors did not serve to normalize deviant forms and their utterances clearly revealed that they did not know how the sentences were structured. The described repetition defects, then, appear to differ in kind from those of adult aphasics. One reason why the right hemisphere impairment is difficult to label in terms of existing clinical syndromes is that the limitations of right hemisphere language processing are not entirely uniform. A number of the characteristics of the linguistic competence of SM and CA are different. CA was somewhat above average on the WISC Digit Span, while MW and SM were at normal levels. Her processing of repetition material was impaired and her output was frequently ungrammatical, but she always responded immediately to the stimulus sentences. SM, by contrast, often asked for several repetitions of the stimulus before he would respond. His processing, too, was defective but his output was less often ungrammatical than CA’s. The divergent language strategies that the two right hemispheres have developed can be seen clearly in their processing of passive affirmative and passive negative statements. Each child has a different subjective organization for passive sentences: SM does not relate the agentive “by” to either of the two noun phrases; CA links “by” to the surface subject. Both show random discrimination of passive affirmative sentences, but for CA and SM, passive negatives are identified better than passive affirmatives. The results for passive sentence comprehension and word-relatedness judgments, considered together, suggest that SM cannot systematically process the relationship the noun phrases and verb have with each other, while CA in difficult syntactic forms assigns the agentive role to the first noun phrase. The two right hemispheres have developed different strategies for processing sentences in which the agent is not lodged in the high frequency, first noun phrase position. The possibility must then be considered that the right hemisphere, in relation to the left, models language not only less well but also less uniformly. The reported results reflect on the deficiencies of right hemisphere language in a special way. The observations which document the claim that

432

DENNIS

AND WHITAKER

the right hemisphere has less measurable language than the left have been made in organisms whose language development involved two brain halves, whether healthy or sick. It has not so far been shown that language acquisition by an isolated right hemisphere, under seizure-free conditions, results in slow or incomplete linguistic development. This is clearly demonstrated by the present results. The two cerebral hemispheres are not ontogenetically equal for language at the close of the first decade of life. Whether they have become so by the end of the second decade is the subject of a report to follow. REFERENCES Alajouanine, T., & Lhermitte, F. 1%5. Acquired aphasia in children. Bruin, g&653-662. Berko, J. 1958. The child’s learning of English morphology. Word, 14, 150- 177: Bishop, N. 1%7. Speech in the hemiplegic child. Proceedings of8th Medical Educationnl Conference of Australian Cerebral Palsy Association, pp. 141- 153. Carlson, J., Netley, C., Hendrick, E. B., & Prichard, J. S. 1968. A reexamination of intellectual disabilities in hemispherectomized patients. Transactions of the American Neurological Association, 93, 198-201. Dennis, M., & Kohn, B. 1975. Comprehension of syntax in infantile hemiplegics after cerebral hemidecortication: left hemisphere superiority. Brain and Language, 2, 472-482. DeRenzi, E., & Spinnler, H. 1966. Visual recognition in patients with unilateral cerebral disease. Journal of Nervous and Mental Disease, 142, 515-525. De Renzi, E., & Vignolo, L. A. 1962. The Token Test: A sensitive test to detect receptive disturbances in aphasics. Bruin, 85, 665-678. Dunn, L. M. 1%5. Expanded manualfor the Peabody picture vocabulary test. Circle Pines, Minnesota: American Guidance Service. Goldman, R., & Fristoe, M. 1969. Goldman-Fristoe Test of Articulation. Circle Pines, Minnesota: American Guidance Service. Goldman, R., Fristoe, M., & Woodcock, R. W. (1970). Goldman-Fristoe-Woodcock Test of Auditory Discrimination. Circle Pines, Minnesota: American Guidance Service. Goodglass, H., Gleason, J. B., Bemholtz, N. A. & Hyde, M. R. 1972. Some linguistic structures in the speech of a Broca’s aphasic. Cortex, 8, 191-212. Goodglass, H., & Kaplan, E. 1972. The assessment of aphasia and related disorders. Philadelphia: Lea and Febiger. Guttmann, E. 1942. Aphasia in children. Bruin, 65, 205-219. Katz, E. W. & Brent, S. B. 1968. Understanding connectives. Journal of Verbal Learning and Verbal Behavior, 7, 501-509. Kelleher, T. R. 1974. Young children’s understanding of presupposition and implication. Unpublished doctoral dissertation, University of Rochester. Kirk, S. A., McCarthy, J. J., & Kirk, W. D. 1%8. Examiner’s manual: Illinois Test of Psycholinguistic Abilities. Urbana, Illinois: University of Illinois Press. Kohn, B., & Dennis, M. 1974. Selective impairments of visuo-spatial abilities in infantile hemiplegics after right hemidecortication. Neuropsychologia, 12, 505-512. Langendoen, D. T. 1970. Essentials of English grammar. New York: Holt, Rinehart and Winston. Martin, E. 1970. Toward an analysis of subjective phrase structure. Psychological Bulletin, 74, 153-166. McFie, J. l%l. The effects of hemispherectomy on intellectual functioning in cases of infantile hemiplegia. Journal of Neurology, Neurosurgery and Psychiatry, 24, 240249.

LANGUAGE

ACQUISITION

BY ONE HEMISPHERE

433

Newcombe, F., & Marshall, J. C. 1%7. Immediate recall of “sentences” by subjects with unilateral cerebral lesions. Neuropsychologia, 5, 329-334. Parisi, D. 1971. Development of syntactic comprehension in preschool children as a function of socioeconomic level. Developmental Psychology, 5, 186-189. Parisi, D., & Pizzamiglio, L. 1970. Syntactic comprehension inaphasia. Cortex, 6,204-215. Tallal, P. 1975. Perceptual and linguistic factors in the language impairment of developmental dysphasics: An experimental investigation with the Token test. Cortex, 11, 196-205. Wepman, J. M. 1958. Auditory discrimination test. Los Angeles: Western Psychological Services. Whitaker, H. A. 1971. On the representation of language in the human,brain. Edmonton: Linguistic Research. Zurif, E. B. & Caramazza, A. 1975. Psycholinguistic structures in aphasia: Studies in syntax and semantics. In H. Avakian-Whitaker and H. A. Whitaker (Eds.), Studies in neurolinguistics. New York: Academic Press. Pp. 261-292. Zurif, E. B., Caramazza, A., & Myerson, R. 1972. Grammatical judgments of agrammatic aphasics. Neuropsychologia, 10,40.5-417.