To identify which improvements in cognitive function are associated with symptom resolution in schizophrenic patients treated with atypical antipsychotics. Design: a prospective open trial with atypical neuroleptics (risperidone, clozapine, quetiapine). Setting: Inpatient and outpatient units, Institute of Psychiatry. Patients: Thirty-nine patients with schizophrenia according to DSM-IV criteria were included. Clinical and cognitive assessment were done at baseline (T0) and again after six months of treatment (T2). Twenty-five patients completed the trial. Interventions: New-generation antipsychotics during six months. Patients were considered as responders if their PANSS score decreased at least 20% (n = 15) and non-responders if it did not (n = 10). Outcome measures: a computerized cognitive assessment comprised tests of short-term-memory (digit span), explicit long-term memory (word pair learning), divided attention, selective attention and verbal fluency (orthographic and semantic). Clinical assessment included PANSS and ESRS. Results: A discriminant function analysis was performed to determine which changes in cognitive performance predicted symptomatic response status. Semantic fluency and orthographic fluency were significant predictors. Together they correctly predicted responder status in 88% of cases. Memory was not a significant predictor of symptomatic response. Conclusion: Verbal fluency discriminated the responder from the non-responder group during a pharmacological treatment.