Two hundred word samples were typed with every fifth .... unequivocal diagnosis of schizophrenia; (ii) no ECT for at least six weeks prior to testing; (iii) no known.
Er. 1. SOC. d i n . Psychol. (1977), 16, 337-34s
Printed in Greut Britain
337
The influence of context on language perception in schizophrenia W. P. de Silva and D. R. Hemsley The present experiment employed the 'Cloze ' procedure to investigate language perception in schizophrenia. The stimulus immediacy hypothesis of Salzinger, Portnoy & Feldman (1966) predicts that with increasing context, schizophrenic performance, unlike that of normals, will fail to improve. A comparison of groups of schizophrenics and normals on such a task, with varying levels of context, not only showed that schizophrenics failed to improve with greater context but that their performance deteriorated. A comparison of acute and chronic schizophrenics indicated that the deterioration was most prominent in acute patients, the performance of chronics remaining relatively stable. An interpretation of the data in terms of narrowed attention in chronic schizophrenia appears plausible, and consistent with Broen's (1968) theory. Such an interpretation may be 'tested by an extension of the range of context conditions.
The abnormal nature of verbal behaviour in schizophrenic patients has long been recognized (e.g. Woods, 1938). Investigations of such disturbance have frequently employed the Cloze procedure introduced by Taylor (1953). In this, every nth word of a passage is deleted for which a blank is substituted; subjects are required to fill in the blanks by guessing the deleted words. It has been used both in studies of language production and language perception. Thus, language samples produced by various subject groups have been mutilated and given to raters for completion, and measures of the communicability of such language samples been obtained (e.g. Silverman, 1972). Equally, various subject groups have been given mutilated passages of normal prose and their scores are interpreted to show their ability or deficit in language perception (e.g. Deckner & Blanton, 1969). The most systematic Cloze procedure studies of schizophrenic language production are those of the Salzinger group (Salzinger, Portnoy & Feldman, 1964, 1966; Salzinger, Portnoy, Pisoni & Feldman, 1970). In their first study they used speech samples taken from monologues of 13 schizophrenics and from a control group of physically ill patients. Two hundred word samples were typed with every fifth word deleted, and given to college students for completion; comprehensibility was greater for the non-schizophrenic samples. Similar findings were reported in a later paper (Salzinger et al. 1966). The third major study by Salzinger and his colleagues (1970) employed a modified version of the Cloze technique. This was an attempt to test the stimulus immediacy hypothesis that Salzinger himself had developed (Salzinger et al. 1966), according to which schizophrenic speech, and indeed schizophrenic behaviour in general, is governed, to a greater extent than that of normals, more by immediate stimuli than by remote ones. For this investigation they obtained speech samples from ten schizophrenic and ten physically ill patients, matched for age and education. From the speech samples, segments were selected; the middle word was deleted from each segment and the segments differed in length so that a blank had a context of 2, 4, 16 or 28 words surrounding it, half on either side. These segments were then given to normal subjects for completion. The immediacy hypothesis suggests that greater context should lead to progressively more correct scores for the normal speech segments, but not for those from schizophrenics. The overall trend of the data clearly indicated support for this prediction. A later paper by Silverman (1972) compared speech samples of schizophrenic and non-schizophrenic patients, with both fourth and fifth word deletion. Normal raters performed significantly less accurately on the schizophrenic samples, but in contrast to the study by Salzinger et al. (1970), the fourth word deletion yielded significantly lower scores than the fifth,
338 W . P. de Silva and D. R . Hemsley even for the schizophrenic samples. However, only four subjects filled in the blanks, and the same subjects completed both fourth and fifth word deletion versions of the samples. Despite the two-week gap between occasions of testing, such a procedure could seriously have affected the results. More recently, Rutter, Wishner & Callaghan (1975) found limited support for the hypothesis that schizophrenic material is less predictable, but compared only two schizophrenic and two orthopaedic patients. Quite unexpected, and throwing doubt on the results of the study in general, was the finding that normals guessed normal material significantly better at the fourth word than the fifth word deletion. Taken together, the studies clearly indicate that schizophrenic speech is less comprehensible than that of normals; it is characterized by greater unpredictability. There is also a growing literature reporting Cloze procedure studies of language perception in schizophrenia. An early study by Honigfeld (1963) indicated that in predicting missing words of Cloze passages, schizophrenics perform less well than normals. Subsequent studies have generally confirmed this. Deckner & Blanton (1969) employed passages deleted by every tenth, eighth or fifth word in the three versions of the text. Schizophrenics were significantly poorer overall; however, while less context led to lower scores for both groups, there was no differential effect on the groups. This fails to support the stimulus immediacy hypothesis, which would predict that the degree of schizophrenic impairment would increase with greater context, although the study was not designed with this hypothesis in mind. It must, however, be noted that the scoring system employed by Deckner & Blanton was unconventional, the responses being rated, rather than verbatim answers being required for a correct response. Blaney (1974) has attempted to test the immediacy hypothesis with the Cloze technique similar to that employed by Salzinger et al. (1970) in their attempt to test the same hypothesis in relation to language production. That is, the texts used were language segments rather than continuous passages, with the middle word deleted. As in the Salzinger et al. (1970) study, the length of the segments varied so that the blanks had contexts of 2, 4, 8 or 16 words, half on each side. The subjects were schizophrenics, normals, and non-schizophrenic psychiatric controls. The prediction was that the schizophrenics would not benefit from added context, whereas the other groups would do progressively better as the context provided increased. The schizophrenics’ performance was significantly poorer overall; they also benefited little from added context, in support of the stimulus immediacy hypothesis. It is therefore clear that schizophrenic patients are far less able than normals to predict deleted words in passages. That is, they are less able to benefit from the redundancies of normal language in their processing of linguistic input. This is consistent with studies investigating the perception and recall of passages of varying degrees of contextual constraint by schizophrenics (e.g. Lawson, McGhie & Chapman, 1964; Truscott, 1970), although Rattin & Chapman (196) have recently argued that such results are an artifact of the use of unmatched tasks. The stimulus immediacy hypothesis of Salzinger has been supported by his own study of schizophrenics’ language production (Salzinger ef af. 1970), and by Blaney’s (1974) investigation of language perception. In employing the construct of ‘breadth of attention’, Salzinger’s hypothesis may be linked to the considerable literature on abnormal attentional processes in schizophrenia. Maher (1972) has also argued for an interpretation of abnormalities of schizophrenic language behaviour in terms of a disturbance of selective attention. Aims of the study The first aim of the study to be reported below was to test the stimulus immediacy hypothesis in relation to schizophrenics’ language perception, using the Cloze procedure, manipulating the degree of context provided to investigate schizophrenics’ ability to benefit from increasing context. Blaney’s (1974) work, which specifically investigated this problem, used only language
Context and language perception in schizophrenia 339 segments, and it was felt that it would be more useful to employ continuous and meaningful prose passages, more typical of language perception situations in normal life. The Deckner & Blanton (1969) study, comparable in form to the present investigation although its aim was different, used an unconventional scoring system. The comparison by Rutter et al. (1975) of fourth and fifth word deletions would not be expected to show marked differences even if differential context effects were present. The stimulus immediacy hypothesis therefore predicts that normals will progressively benefit from increasing context while schizophrenics perform at the same level in all context conditions. The second aim of the study was to explore possible differences between acute and chronic schizophrenics. Salzinger’s immediacy hypothesis centres around the concept of attention and it is with regard to attention that some of the most interesting evidence on acute-chronic differences has been reported (e.g. Broen, 1968; Feeney, 1971; Broen & Nakamura, 1972). According to Broen, the fundamental deficit in schizophrenia is a greater equivalence of dominant and competing responses, leading to increased response interference. This is seen as an aversive condition and the acute schizophrenic is considered to learn gradually to adjust to this deficit by a reduction in the range of stimuli scanned. The problem of defective attention in schizophrenia is a complex one (cf. Hemsley, 1979, but recent studies of schizophrenic language behaviour have provided support for Broen’s (1968) position. In investigations of dominant response bias in schizophrenia, noted by Chapman, Chapman & Miller (l964), Williams, Hemsley & Denning-Duke (1976) and Strauss (1975) have found it to be more prominent in chronic than acute schizophrenics. The findings were interpreted as supporting Broen’s ( 1968) suggestion that the narrowed observation of alternate meanings of words is learned over time as an attempt to cope with response disorganization. Broen’s theory would predict that chronics should show less ability than acutes to use context in a language perception task, since they are considered to scan a narrower range of the context provided. Method
Materials Three simple prose passages of approximately equal length (155-165 words) were used as the instruments for the Cloze task. The passages were of approximately equal difficulty, as determined by a pilot study. In this, normal volunteers guessed deleted words in a pool of seven simple prose passages; fifth word deletion was employed. The three passages closest in mean scores were chosen for the experiment. They were prepared for the study by deleting every fourth, seventh and tenth words. Thus there were altogether nine Cloze passages, consisting of three source passages mutilated in three different ways so that each passage yielded three different context conditions. If the word to be deleted was a proper noun or number it was retained and the next word omitted. All punctuation marks and capital letters, except those at the beginning of names, were omitted. The test unit used for any given subject consisted of three passages, one for each contextual condition and representing all three source passages, i.e. no subject received the same source passage twice. A balanced design was employed so that the three source passages were used in equal proportion at the different levels of context in the main subject groups. The order of presentation in each test used was predetermined to prevent bias from order effects. Two shorter passages were prepared by deleting every fifth word, to serve as demonstration and ‘ warm-up ’ passages.
Subjects The experimental subjects were 30 hospitalized schizophrenics. The criteria for selection were: (i) unequivocal diagnosis of schizophrenia; (ii) no ECT for at least six weeks prior to testing; (iii) no known organic damage; (iv) WAIS vocabulary subtest scaled score of at least 7; (v) age 18-60. Drugs could not be controlled; all but four were receiving phenothiazine medication. An analysis of responses did not reveal any differencesin performance of off-drugsubjects from others. Investigations of the effects of phenothiazines on cognitive functioning in schizophrenia have generally shown either no change or an improvement (Baker, 1968; Goldberg, 1972).
340 W. P. de Silva and D. R . Hemsley
Table 1. Characteristics of subject groups ~~
Normals (n = 20)
Schizophrenics (n = 30)
Mean
Mean
S.D.
~
~
P
t
S.D.
~
Age WAIS vocabulary scaled score
30.8 11.9
5.6
34.4 10.3
5.3
17.9 2.7
0.872 1.441
~~
> 0.05 > 0.05
The normal controls were 20 volunteers chosen with the same age and vocabulary criteria as the schizophrenics. Descriptive characteristics of the subject groups are shown in Table I ; they do not differ significantly on age or vocabulary. From the total schizophrenic sample, two subsamples of acutes and chronics were chosen for further comparison. Of the three chronicity variables possible -time since first admission, total time in hospital, number of admissions -the first was used in assigning subjects to the subsamples. It was considered plausible to assume that the typical adaptive strategy of restriction of attention, that Broen (1%8) claims characterizes chronics, might be most directly related to the time since onset of the disturbance. In any case, all three chronicity measures were highly positively correlated. The characteristics of the acute and chronic subsamples are shown in Table 2. There is no significant difference in vocabulary score, but the chronics are, not surprisingly, older.
Table 2. Characteristics of acute and chronic subsamples Acutes (n = 10)
Chronics ( I I = 10)
Mean
S.D.
Mean
S.D.
t
8.5 3.4
42.3 9.4 219.0
7.8 2.4 71.8
4.198 1,145 4.932
Age 27.0 Vocabulary 10.9 Time since first 19.0 admission (months)
18.0
P 60401
n.s. 6 0401
Procedure The WAIS vocabulary subtest was first administered. The experimental task was then explained in the following words: ‘You will be given some passages from which certain words have been deleted. Your task is to guess the missing words and to write them in the blank spaces provided. Each blank is for one word only and the missing words are not proper nouns or numbers. When a passage is given to you read it over first, and then write in the words you guess are missing, Try your best to fill in the blanks, please. And try to do it quickly.’ The demonstration passage was then shown and completed by the experimenter. The warm-up passage was next given, and help was given if required. After answering any procedural inquiries, the experimental passages were administered, one at a time. There was no time limit. Only two subjects, both acute schizophrenics, failed to complete the task and had to be excluded. The scoring was carried out in the standard way (Taylor, 1953, 1956). A response was marked as correct if it matched the target word verbatim. The correct responses were added, and this, taken as a percentage of the total blanks in the passage, was taken as the score (cf. Rutter et al. 1975). Thus each subject had a score for each of three contextual conditions.
Context and language perception in schizophrenia
341
Results
Normals vs. Schizophrenics The scores for the two groups, under varying context conditions are shown in Table 3. The scores were examined by analysis of variance. Schizophrenics were, as expected, poorer in performance (F=25.78, d.f. = 1,48, P < 0401). There was no overall difference across Table 3. Mean Cloze scores of normal and schizophrenic subjects c4
Normals Schizophrenics
c 10
c7
Mean
S.D.
Mean
S.D.
Mean
S.D.
42.9 36.2
11.8 14.7
46.8 30.4
11.1 16.2
58.0 27.3
15.9 15.8
conditions (F=0.94, d.f. = 2,96, n.s.), but a significant interaction between groups and conditions (F=18.00, d.f. 2, 96, P < 0401). It can be seen from Table 3 that the normal group improved with increasing context, whereas the performance of the schizophrenic group deteriorated. The differences between the groups became increasingly marked with greater context. The nature of the increasing difference was examined, using t tests for correlated samples, by comparing C4 and C10 scores within the two groups separately. Both showed significant changes from C4 to C10, in opposite directions (P
