Contingent negative variation (CNV) is supposed to be a psychophysiological indicator of attention and arousal. Both have been reported to be deteriorated in schizophrenic and depressed patients. Thirty-four patients with major depression, 43 patients with schizophrenia and 49 healthy subjects were investigated during acute illness with a complex three-stimulus go/no-go task which requires different states of attention: trials consisted of three complex figures that were tachistoscopically presented. Three identical figures had to be confirmed by pressing a button (target condition). CNV was measured: (1) after the first figure waiting for the second (baseline condition), (2) after two identical figures waiting for the third (response-relevant condition), (3) after two different figures waiting for the third (response-irrelevant condition). The response-relevant condition compared to baseline significantly intensified CNV in healthy controls and to a minor extent in depressed patients but not in schizophrenics. In the response-relevant conditions in healthy controls, CNV was significantly reduced compared to the response-relevant condition. This clear discrimination between response-relevant and response-irrelevant conditions was not observed in either group of patients. Thus, the applied CNV paradigm was able to discriminate schizophrenic and depressed patients from healthy controls. Furthermore, subtle differences between schizophrenic and depressed patients were detected, reflected by the different CNV development across experimental conditions.