Differences between acute (N = 26) and chronic (N = 27) schizophrenics diagnosed by Research Diagnostic Criteria and normal controls (N = 53) were examined in a task measuring latent inhibition, i.e., the retardation of learning that normally occurs when a subject forms an association to a stimulus previously repeatedly presented without consequence. In rats, latent inhibition is disrupted by amphetamine and restored by neuroleptics. It was predicted that latent inhibition would be similarly disrupted in acute schizophrenics (presumably in a hyperdopaminergic state) but not in chronic, medicated schizophrenics. Each group was subdivided and assigned randomly to two experimental conditions, preexposure or non-preexposure. Preexposed subjects first heard 30 bursts of white noise through headphones while monitoring a list of nonsense syllables; non-preexposed subjects listened to the nonsense syllables without the white noise. Subjects in both conditions were then given the opportunity to learn that the noise signalled increments in a visually displayed number. Preexposed normals and chronic schizophrenics learned this association more slowly than non-preexposed subjects (latent inhibition); as predicted, acute schizophrenics failed to display this effect. After 6 to 7 weeks, 11 acute and 13 chronic patients were retested; both groups now showed latent inhibition. These results are discussed in relation to the dopamine hypothesis of schizophrenia.