Hospitalized chronic and acute schizophrenics were compared with age- and social-status-matched controls on word perception tests to measure the effect of context on recognition thresholds. In Experiment 1 the method of J. Morton (1964, British Journal of Psychology, 55, 165-180) was adopted to see how related and unrelated contexts influenced word recognition thresholds when both context and stimulus word were presented visually. Experiment 2 was an auditory analog of Experiment 1 and in addition the chronic schizophrenics were selected on the basis of a presence or absence of auditory hallucinations. The influence of context on perceptual thresholds was quite normal in all schizophrenics. However, analysis of the incorrect responses elicited showed that some schizophrenic subjects do produce bizarre or irrelevant responses or perseveration to a greater extent than their controls. These results are explained in terms of automatic and strategic modes of information processing. The automatic processes responsible for the adjustment of perceptual thresholds operate normally in schizophrenics but response production which demands strategic selection and editing of responses is prone to malfunction which results in the observed deviations in the normal use of language.