286 The Need to Relate Cognitive Deficits to Specific Behavioral Referents of Schizophrenia

The implicit rationale for many cognitive studies of schizophrenia hinges on the recognition that verbal communication generated by patients with this disorder is often elusive or difficult to comprehend. This obser-vation has led to the inference that a cognitive dysfunction, which mediates the production of discourse failure, is characteristic of schizo-phrenia. Unfortunately, most investi-gators have chosen to examine this type of hypothesis by comparing heterogeneous groups of schizo-phrenic patients (without regard to whether they exhibit verbal communication impairment) with various control samples; they have not studied the association between cognitive processes and specific schizophrenic symptoms. Data are presented from two studies indicating that such relationships, even when highly plausible, cannot simply be presumed. In both studies, one with adults suffering from schizophrenia and the other with children at risk, a laboratory measure of referential communication failed to be strongly related to language disorder. These data suggest that future investigators should specify the features of schizo-phrenia that are expected to correlate with their laboratory measures and empirically evaluate these relations. Several methodological problems have received attention in the recent literature regarding cognitive factors in schizophrenia. Greater care is being taken in the process of patient diagnosis; many studies include comparison groups of patients from other specific diagnostic categories; and longitudinal designs are being used to examine the temporal relationship between cognitive variables and the course of the disorder (thus addressing the issue of directionality). Unfortunately, a conceptual issue that has important methodological implications has been largely ignored. The problem involves the failure to examine the manner in which cognitive variables may be related to specific overt features of schizophrenia. This issue has been raised previously (Banniste