Facilitating sentence formulation: A case study

Facilitating sentence formulation: A case study

JOURNAL OF COMMUNICATION DISORDERS 9 (1976), 191-197 FACILITATING SENTENCE FORMULATION: A CASE STUDY CYNTHIA M. SHEWAN Queens College of The City ...

473KB Sizes 0 Downloads 59 Views

JOURNAL

OF COMMUNICATION

DISORDERS

9 (1976), 191-197

FACILITATING SENTENCE FORMULATION: A CASE STUDY CYNTHIA M. SHEWAN Queens College of The City University

of New York*

A training program was designed to facilitate sentence formulation in a Broca’s aphasic patient. Two training methods were compared and generalization to untrained stimuli was measured. Training, using either method, resulted in increased production of correct linguistic constituents and appropriate lexical items. Gains tended to occur on the trained items and generalization to untrained items was limited.

Introduction Broca’s aphasics demonstrate impaired oral communication and their expressive language impairment has often been characterized as exhibiting agrammatism (Goodglass, 1968; Goodglass et al., 1972). Inflections required in English may be lacking; auxiliary verb forms may be omitted; and certain linguistic constituents in a sentence, such as a subject noun phrase, may be absent. Since rehabilitation in aphasia can never encompass all the sentences that a patient will encounter in communication situations, the concept of generalization is important. Holland and Levy (1971) found that generalization of one syntactic form (an active declarative sentence) to several transformations of that form was limited. Generalization occurred for the interrogative but not for the negative or passive transformations of the trained declarative sentence. Generalization also occurred for sentences of the same syntactic form that contained different vocabulary, although the gain was not statistically significant. Therefore, training did improve the production of a specific sentence and results suggested improvement for sentences of the same syntactic structure but different vocabulary. In view of these generalization data and the agrammatic difficulties of Broca’s aphasics, the present study was designed to determine the effects of training on the sentence formulation abilities of a Broca’s aphasic patient. Two training methods were compared to assess their usefulness for facilitating production of declarative sentences. Provisions were also made to assess generalization from trained to untrained items. Assessments of the appropriateness of a sentence’s syntactic form and its semantic content were also made. * Now at Program in Communicative Ontario, Canada N6A 5C2. o American

Elsevier Publishing

Company,

Disorders,

Inc.,

The University

1976

of Western

Ontario,

London,

191

192

Methods

CYNTHIA

M. SHEWAN

and Procedures

Subject The subject, SJ, was 64-year-old male who suffered a CVA 21 months prior to the study. Etiology was reported as atherosclerosis-ischemic. Six weeks after the stroke he was thought to have a repeat pulmonary embolism. He was left with a right hemiplegia, more severe in the arm than in the leg, right facial weakness, and moderately severe aphasia. SJ has received speech and language training since the CVA. When the study began he demonstrated moderate aphasia of the Broca’s type. His expressive language was characterized by grammatical errors, word retrieval problems, and verbal dyspraxia. He understood general conversation adequately but had more difficulty when the complexity and length of discourse increased. He could read short paragraphs with good comprehension but did not use writing as a means of expression. His spontaneous speech was characterized by a generally short phrase length. He could produce some complete sentences, but more frequently used single words or short phrases. He had more difficulty describing visual picture stimuli and could not seem to integrate the various aspects of a picture, frequently limiting his responses to naming items or to short phrases describing one aspect of the picture. He demonstrated word retrieval problems; sometimes he substituted a semantically related word (for example, mother/wife), other times he could not generate a word and would say “No, I don’t know.” His repetition skills were somewhat better than spontaneous speech. He could repeat simple sentences without much difficulty. His speech was characterized by verbal dyspraxic errors that were present at a single word level and in sentences. Familiarity with the material appeared to aid him somewhat since he repeated high probability sentences more accurately than low probability ones and he had little difficulty with high frequency polysyllabic words. He was generally aware of dyspraxic errors and his multiple attempts at difficult items generally resulted in a production closer to the target. The prosodic aspects of spontaneous speech were impaired. Frequent pauses and repetitions were noted. Intonation was preserved and stress patterns were generally appropriate during repetition tasks. Procedure Because of SJ’s sentence formulation difficulties in spontaneous conversation and in describing pictures and because he could repeat simple sentences, a goal in therapy was to improve his sentence formulation skills and use. The sentence type selected for training was that of the simple active affirmative

FACILITATING

SENTENCE

FORMULATION:

A CASE

STUDY

193

declarative (SAAD), which could have the surface forms of subject (S)-verb (V), S-V-object (0), or S-V-prepositional phrase (PP). This type was chosen for several reasons. SJ had produced at least some of these in the past and he could repeat them. This form was relatively easy linguistically since it required no permutation or embedding operations. It contained linguistic constructions that showed a high rate frequency of usage by normal adult speakers (Ludlow, 1973) and that appeared early in language recovery in Broca’s aphasics (Ludlow, 1973). To establish baseline data, 45 pictures were presented in succession and SJ was asked to formulate a simple sentence about each. That each picture could be described with an SAAD sentence was established by selecting pictures that three normal adult speakers had described without difficulty using an SAAD form. SJ’s responses to the picture stimuli were recorded verbatim and his last response was used. It was scored as pass if syntactically correct and semantically appropriate, or fail, if syntactically incorrect, semantically inappropriate, or both. Verbal dyspraxic errors were not counted as failures. From the items that were scored as fail responses, 30 pictures were selected at random to be used for the study. They were assigned randomly to each of three groups with the exception that the actions (verbs) depicted in Group 1 stimuli were different from those in Group 2. This was done to avoid confounding the two training method results. SJ received the following training in sentence formulation for Group 1. A picture was presented accompanied by six cards on which were written linguistic constituents: two subjects (NP), two verbs (VP),’ and two objects (NP) or two prepositional phrases (PP). From these cards SJ selected the three appropriate for the picture stimulus, sequenced them to make a sentence, and then orally read the sentence. If he made errors, the clinician provided prompts until he corrected his errors. After reading the sentence the graphic cues (cards) were removed and SJ attempted an unaided description. Either a repetition of the written sentence or a spontaneously formulated sentence was accepted as correct. If his unaided sentence was correct, the next item was presented; if incorrect, the clinician provided a correct model. Training on the 10 Group 2 stimuli consisted of providing both the auditory and written form of the verb (VP) as cues. SJ was instructed to make up a sentence about the picture using, for example, “is reading.” The graphic cue was then removed and he attempted an unaided response. If it were correct, the next stimulus was presented; if incorrect, the clinician provided a correct model. The Group 3 stimuli served as control items and SJ received no training on them. Following baseline, Group 1 and 2 stimuli were presented for six successive training sessions. Order of presentation was counterbalanced to avoid practice ‘The

VP consisted

of Auxiliary

+ Verb.

194

CYNTHIA

M. SHEWAN

and/or fatigue effects. In the seventh, a test session, all 30 stimuli were presented in a random order with no training and SJ was asked to formulate a sentence for each. Each response was scored for both semantic and syntactic aspects and for the number of correct linguistic constituents and appropriate lexical items included. If the linguistic constituents used were correct, for example, NP-VP-NP, the sentence was scored as syntactically correct, regardless of the semantic content. If the lexical items used were appropriate to the visual stimulus, the sentence was considered semantically correct. Results The responses for the 30 stimulus pictures obtained during baseline were compared with those obtained after training. When syntactic correctness alone was examined SJ produced over twice as many grammatically correct sentences after training, 16 as compared with 7 at baseline. This increase was statistically significant Cp < 0.01) when subjected to az-test for correlated proportions. Since this comparison included all stimulus items, further analysis separated trained from untrained items to determine where the improvement had occurred. Difference scores, obtained by subtracting the baseline scores from the posttraining scores, were used for the comparison in performance between trained and untrained items. As previously, only correct syntactic form was considered in this comparison. (Semantic substitutions were not considered as errors and were analyzed separately.) Difference scores rather than raw scores were used to account for possible differences in baseline data between trained and untrained items. As indicated in Table 1, after training SJ formulated more grammatically correct sentences for trained items, while his performance declined slightly on those items for which he received no training. This difference, significant at the 0.01 level of confidence, indicated that the positive effects of the training did not generalize to untrained items. The previous analysis dichotomized responses into correct or incorrect categories but gave no credit to partially correct responses. Since SJ produced portions of some sentences using appropriate grammatical forms, although the

Number of Grammatically

TABLE 1 Correct Sentences at Baseline and Posttraining Scores for Trained and Untrained Items Number of grammatically Baseline

Trained items Untrained items

3 4

and the Difference

correct sentences

Posttraining 13 3

Difference + 10 -1

FACILITATING

SENTENCE

FORMULATION:

A CASE STUDY

195

entire sentence was not correct, his performance was analysed using linguistic constituents as the unit scored rather than the entire sentence. The constituents used were the noun phrase (NP), the verb phrase (VP), and the prepositional phrase (PP). It can be seen from Table 2 that SJ made gains in the proportion of correct responses for each of the linguistic constituents. Thus, after the training period he produced more appropriate linguistic constituents necessary to construct simple sentences. These gains were statistically significant for each of the constituents. The training did not differentially affect one linguistic constituent more than the others since there were no statistically significant differences among the proportion gains for NP, VP, and PP. The training, designed to facilitate sentence construction, was equally effective for the three linguistic constituents produced. After training, for both methods, SJ produced an increased proportion of correct linguistic constituents. Gains occurred on all three constituents and the data for them were pooled. Whether both methods were equally effective in producing the gains was examined by comparing the gain proportions for Group 1 and Group 2 items. Group 2, which provided an auditory and a graphic VP cue, resulted in a greater proportion gain (0.430), but this was not significantly greater than the gain for Group 1 (0.298), which provided graphic cues for the linguistic constituents. This suggested that the nature of the cue provided was not as important for SJ as the presence of some cue. The training procedures emphasized production of appropriate sentences. Of course for a sentence to be meaningfully interpreted, the constituents produced must be semantically appropriate to the stimulus. To determine if there were any change in semantic appropriateness a comparison was made between baseline and posttraining scores for semantically correct sentences. The gains were small, from 3 at baseline to 5 at posttraining, and not significant statistically @ > 0.05). To obtain a more refined measure, the number of appropriate lexical items in the responses was compared at baseline and posttraining. With this scoring method, SJ was given credit for each appropriate lexical item in his responses. The proportion of appropriate items increased at posttraining and the gain (0.150) was significant Cp < 0.025). Both trained items and untrained items revealed increases but they were not significantly different from one another.

Proportion

of Correct Linguistic

TABLE 2 Constituents at Baseline and Posttraining Proportions Proportion

Linguistic

constituent

NP VP PP

*p < 0.05. **p < 0.005.

of correct responses

Baseline

Posttraining

Gain

0.630 0.300 0.546

0.872 0.700 0.889

0.242** 0.400** 0.343*

and the Gain

196

CYNTHIA

M. SHEWAN

Examination of SJ’s errors during testing revealed that half of them could be attributed to perseveration, semantic substitutions, and word retrieval problems. The other half were either grammatical errors or no response. Therefore, problems with selection, of correct lexical items appeared to interfere as frequently as problems with syntactic form in SJ’s sentence formulations. Discussion Since sentence production is an important part of language recovery for the aphasic patient, it was important to determine if training with sentence construction would effectively increase a subject’s ability to produce appropriate sentences. Findings indicated that training did result in an increased number of grammatically correct sentences, although the increases occurred on trained stimuli only. This finding was in contrast to Holland and Levy’s (197 1) report that generalization in correct sentence production did occur within the sentence type trained. They did not report individual subject data so it can not be determined if all of their subjects generalized. That SJ did not demonstrate syntactic generalization could be explained in several ways. He may have abstracted a strategy or (strategies) but was not able to apply it to unfamiliar material. He may have needed more training before he had firmly established a strategy or rule. Since he had attained only 65% accuracy on the trained items at posttraining, a higher asymptote may have been needed before generalization to untrained items would occur. Another possibility, consistent with Goldstein’s (1948) viewpoint, was that SJ could not abstract a strategy at all. Consequently, the training would aid trained items by providing practice and repetition, but no improvement would be anticipated for untrained items. Although there was no marked increase in the number of semantically appropriate sentences, there was a significant increase in the proportion of appropriate lexical items in the responses. Gains occurred on both trained and untrained items suggesting that the training influence generalized at least to some extent. The appropriate lexical items were presented in Group 1 stimuli and Group 2 presented the correct lexical verb. In addition, all incorrect responses were followed by a correct sentence formulated by the examiner. Therefore, exposure to appropriate lexical items during training resulted in better selection during testing. This was consistent with general clinical findings that auditory and visual stimulation can facilitate word retrieval. In view of the failure for the grammatical sentence construction to generalize, it would appear that difference processes are involved in recovery of these aspects of language. One of the problems that reduced semantic appropriateness was SJ’s perseveration. Although he would start his response sentence with an NP, it was frequently an NP from the previous response. Semantic substitutions and word retrieval problems also reduced semantic appropriateness of responses. Techniques to

FACILITATING

SENTENCE

FORMULATION:

A CASE STUDY

197

reduce these error types would be helpful in improving the semantic appropriateness of SJ’s productions. The results of the study confirm the advantages of using a more refined scoring system than a simple correct-incorrect dichotomy. There were many occasions on which SJ could not generate a totally correct sentence, but could produce some correct linguistic constitutents and/or semantically appropriate lexical items. At posttraining SJ formulated s/ome response for each of the 30 stimuli while at baseline he could not generate any response for nine pictures. A correct-incorrect scoring system would not have reflected these changes. A single subject case study has obvious limitations, but suggests where future research might be directed. Another study might lengthen the training period to determine whether generalization would occur to untrained stimuli and additional subjects should be used to test the generality of these findings. The author wishes to thank Ellen Spund for her assistance in carrying out the training program and in collecting the data.

References Goldstein, K. Language and language disturbances. New York: Grune & Stratton, 1948. Goodglass, H. Studies on the grammer of aphasics. In S. Rosenberg and J. H. Koplin (Eds.), Developments in applied psycholinguistics research. New York The Macmillan Company, 1968. Goodglass, H., Gleason, J. B., Bemholtz, N. A., Hyde, M. R. Some linguistic structures in the speech of a Broca’s aphasic. Cortex, 1972, 8, 191-212. Holland, A. L., Levy, C. B. Syntactic generalization in aphasics as a function of relearning an active sentence. Acta Symbolica, 1971, 2, 34-41. Ludlow, C. The recovery of syntax in aphasia: an analysis of syntactic structures used in connected speech during the initial recovery period. Doctoral dissertation, New York University. Ann Arbor, Mich.: University Microfilms, 1973, No. 74-12, 847.