Empirical evaluation of language disorder in schizophrenia

Abstract

Objective: Studies of the content of speech and of verbal hallucinations in schizophrenia point to dysfunction at multiple levels of language. In this study, we empirically evaluated language processes.

Methods: We examined the performance of 22 schizophrenia patients and 11 healthy control subjects with procedures designed to explore the sublexical, lexical, semantic, syntactic and discourse levels of language processing.

Results: Schizophrenia patients exhibit impairment in the recognition of incorrect, but not correct, linguistic stimuli at all but the sublexical level of language processing. The patients were not impaired in the recognition of nonlinguistic stimuli.

Conclusion: This language-specific differential impairment could explain speech abnormalities in schizophrenia. The nonrecognition of incorrect linguistic information would prevent patients from correcting the abnormal speech they may occasionally produce. A model of decreased power of linguistic computations (reduced number of operations) adequately accounts for this differential impairment.

Fig. 1: Outline of the linguistic task.

Fig. 2: Average proportion of correct responses for each group, condition and stimulus type. The error bars represent the standard error of the mean. The ordinate is in an angular scale. Overlapping errors have been omitted for clarity.

Fig. 3: Average latency of correct responses for each group, condition and stimulus type. The error bars represent the standard error of the mean. Overlapping errors have been omitted for clarity.

Fig. 4: Illustration of the effect of diminished computational power on the recognition of correct and incorrect linguistic stimuli (P = property, uncrossed box = correct property, crossed box = incorrect property). In the first scenario, correct and incorrect stimuli require an equal number of computations. In control subjects, n properties are computed. Therefore, correct and incorrect stimuli will be correctly classified. In patients, n – p properties are computed. Thus correct stimuli will be classified correctly; however, some incorrect stimuli will be classified in the correct category. In the second scenario, incorrect stimuli require a higher number of computations than do correct stimuli (n + m). With n computations, all correct stimuli will be recognized in patients and control subjects. In control subjects, n + m computations are done. Therefore, all correct and incorrect stimuli will be recognized. In patients, n + m – q computations are done. Therefore, some incorrect stimuli will be classified in the correct category.