Several forms of confabulation have been identified recently in schizophrenic patients, but it has not yet been investigated whether these forms are specific to schizophrenia. Furthermore, the origin of confabulation is unclear. The present study investigated recall and recognition confabulation and their relations with symptomatology, cognitive domains (abstraction and flexibility, verbal fluency, verbal memory, motor activity, and visual-motor processing/attention), computed tomographic (CT) measures (ventricular, cerebral, and Sylvian fissure size), and auditory event-related potentials (amplitudes and latencies of peak components in oddball paradigms) in 33 schizophrenic patients, 35 bipolar I patients, eight schizoaffective patients, and seven patients with other psychotic disorders. We found that neither type of confabulation was specific of any diagnostic group. Recall confabulation was mainly predicted by the predominance of positive symptoms, while recognition confabulation was predicted by a delay in P300 latency and the doses of antipsychotics used. Our results suggest two different mechanisms for both types of confabulation based on interference with the adequate retrieval of information and slowness in early stimulus detection.