Context-dependent information processing in patients with schizophrenia

Abstract

Thirty schizophrenic patients fulfilling the Diagnostic and Statistical Manual of Mental Disorders IV criteria for schizophrenia and 30 control participants were shown a set of incomplete sentences, and were asked to complete them with the first word(s) that came to mind. Target sentences included an ambiguous word, the ambiguity of which was not resolved within the clause. However, completion necessarily required participants to select one specific meaning. Each target sentence was preceded by another sentence playing the role of context, which was designed to prime the less frequent meaning of the ambiguous word. The results showed that schizophrenic patients, especially those with thought disorder [on the basis of their TLC scores (Thought, Language and Communication Scale; Andreasen, N.C., 1979. Thought, language and communication disorders. Clinical assessment, definition of terms and evaluation of their reliability. Diagnostic significance. Arch. Gen. Psychiatry 39, 778–782)], used the most common meaning of the ambiguous word more frequently than controls, thus revealing a specific deficit in context use. The deficit was observed whether or not the relation between context and target sentences was made explicit. These results are in line with the cognitive models of schizophrenia that postulate a decreased ability to use context information. However, when considered in the light of prior studies (e.g., Bazin, N., Perruchet, P., 1996. Implicit and explicit memory in patients with schizophrenia. Schizophr. Res. 22, 241–248), they suggest that the deficit in processing contextual information is limited to what Baddeley (Baddeley, A.D., 1982. Domains of recollection. Psychol. Rev. 98, 708–729) called the interactive context (which affects the meaning, or the interpretation, of the target event) in contrast to the independent context (which does not interfere with the meaning-based interpretation of the target event).

Introduction

Deficits in schizophrenic patients do not extend to all cognitive abilities, but tend to be limited to specific tasks. However, it is still difficult to interpret the pattern of impaired and spared abilities within an integrative framework. Although different integrative models have been proposed (e.g. Frith, 1987), several authors have claimed that most of the cognitive deficits observed in schizophrenia could be dependent on a specific impairment in the processing of contextual information. By contextual information, we mean the background or surrounding circumstances that can mediate a response to a target event, except the information conveyed by the target event itself. This may include, for instance, the general instructions given prior to the task, and the stimulus or the participant's response immediately preceding the target event. Cohen and Servan-Schreiber (1992) posit that many of the cognitive deficits in schizophrenia can be accounted for by the idea that schizophrenic patients have a degraded ability to construct and maintain internal representations of context. Other authors have reached similar conclusions on the basis of different paradigms. For instance, the fact that latent inhibition, which is commonly exhibited in normal subjects, is absent in schizophrenic patients (Lubow et al., 1987), has also been accounted for by the idea that contextual information processing is impaired in those patients (Hemsley, 1992). Along the same lines, Hardy-Bayle postulates that the inability to take account of contextual data is the source of other cognitive impairments (Hardy-Bayle, 1992, Widlocher and Hardy-Bayle, 1989).

In order to examine the role of context processing in schizophrenics' memory deficits, Bazin and Perruchet (1996) used an implicit memory task in which the use of context may improve performance. We were especially interested in implicit rather than explicit memory, because several authors have posited that only the controlled processes are impaired in schizophrenia, while more automatic processes are spared (e.g., Hardy-Bayle, 1994). As support for this hypothesis in the memory field, it has been shown that performance on recall and recognition tests is often altered, whereas performance in implicit memory tasks is the same as in controls (Clare et al., 1993, Danion et al., 1992, Goldberg et al., 1993a, Goldberg et al., 1993b, Schmand et al., 1992, Schwartz et al., 1993). Implicit memory tests tap the automatic effects of a previously studied stimulus in subsequent identification or production tasks, the term ‘automatic’ here being intended to mean that the effects are not mediated by the explicit retrieval of these stimuli.

The paradigm involved in the Bazin and Perruchet (1996) study was adapted from that used by Graf and Schacter (1985), who investigated implicit memory for newly acquired associations (referred to as implicit associative memory below). The study context is a semantic one represented by words. The subjects were first presented with a set of pairs of normatively unrelated words. Within each pair, the first word formed the context for the second, target word. Because the associative memory effect appears only, or at least most clearly (Schacter and Graf, 1986), when a meaningful relation between the two words is formed during the encoding phase, the subjects either had to read a sentence including the two words or generate such a sentence themselves. The subjects were then given a word-stem completion test involving the target word of each pair. Some stems appeared in the same context as they had during the study phase, whereas other stems appeared in a different context. The result was that subjects completed stems more often when they appeared in the same context than when they appeared in a different context. However, contrary to the authors' hypothesis, normal subjects and schizophrenic patients were equally sensitive to the congruence of the context target relationship between the study phase and the test phase. In a second experiment, Bazin and Perruchet replicated these results, and extended their conclusions to an explicit associative memory task.

The discrepancy between Bazin and Perruchet's (1996) findings and the earlier results exhibiting impaired context processing highlights the need to explore the differences between our paradigm and earlier ones. Although the situations differ in several features, the specific role of context in dealing with target information may constitute a particularly pertinent difference. Baddeley (1982) distinguished between two categories of context. One, the interactive context, is defined by the fact that it affects the meaning, or the interpretation, of the target event. For instance, a semantic context which makes it possible to disambiguate the meaning of a homophone is undoubtedly interactive in nature. In contrast, independent context does not interfere with the meaning-based interpretation of the target event. A well-investigated independent context in studies on list memory is whether the words studied are in the context of `on land' or `under water' (e.g. Godden and Baddeley, 1975). In this case, the environmental context is encoded independently of and in parallel with the target information. Note that Baddeley's distinction does not match the distinction between verbal and physical contexts, as the above examples might suggest. For instance, the ground of a picture can serve as an interactive context for the interpretation of an ambiguous figure. Similarly, the pairing of semantically unrelated context/target words instantiates independent context.

Baddeley's distinction appears relevant in several research domains. For instance, Ackerman (1986) showed that the allocation of attention varied in context-interactive and context-independent encoding situations, and that adults and children differ in the way they attend to context information when context is interactive. Likewise, a change in context has been shown to impair performance in a lexical decision task only when this change alters the perceived meaning of the target word (Bainbridge and Lewandowsky, 1993). The distinction between interactive and independent context can help to account for the discrepancy between studies involving schizophrenic patients. Indeed, the context used by Chapman et al., 1964, Cohen and Servan-Schreiber, 1992 was interactive, insofar as it dispelled the ambiguity of the meaning of the target word, whereas the context in Bazin and Perruchet's (1996) experiments was not.

The present study was intended to test the hypothesis that the deficit in context processing in schizophrenic patients can be revealed in a priming task in which context is interactive, irrespective of whether context processing is implicit or explicit. Participants were shown a set of incomplete clauses and were asked to complete them with the first word(s) that came to mind in order to form meaningful sentences. Half of the clauses for completion included a word with multiple meanings (an ambiguous word), the ambiguity of which was not removed within the clause. However, completion of the sentence necessarily required participants to select one specific meaning. For instance, one ambiguous clause was: Chez eux, il s'assied, prend sa serviette et commenceà … (At their house, he sits down, takes his serviette and begins to …; The word serviette in French is ambiguous: its common meaning is napkin, its less frequent meaning is briefcase). Each ambiguous clause was preceded by a context clause that primed the less frequent meaning of the word with multiple meanings of the ambiguous clause. The context clause that primed the less frequent meaning of serviette (briefcase) at the expense of its more frequent meaning (napkin) was: Ce médecin … va chez ses patients avec ses ordonnances. (This doctor … goes to his patients' homes with his prescriptions). In order to examine the role of intentionality in the exploitation of the context, the context clause and the ambiguous clause were either presented as two separate sentences or linked with “and” within a same sentence: Ce médecin consciencieux va chez ses patients avec ses ordonnances et chez eux, il s'assied, prend sa serviette et commenceà … (This conscientious doctor goes to his patients' homes with his prescriptions and at their house, he sits down, takes his serviette and begins to …). Because participants were asked to generate meaningful sentences, this procedure meant that the use of context was implicit in the former case, and explicit in the latter.

We expected schizophrenic patients to complete the ambiguous sentence using the more frequent meaning of the ambiguous word more often than normal subjects, thus providing evidence that they fail to use contextual information, irrespective of the fact that use of contextual information was a part of the explicit task demands.

Because of the well-known heterogeneity of schizophrenic patients, we studied several subtypes of patients individually depending on their TLC score (Thought, Language and Communication Scale, Andreasen, 1979). This scale seems to be the one best suited for discriminating between schizophrenic patient groups (Besche et al., 1997, Passerieux et al., 1997, Sarfati et al., 1997).