While many acutely ill schizophrenic patients suffer from depressive symptoms, most studies on the efficacy of antipsychotic drugs focus on positive and negative symptoms. Dimensional models of schizophrenic symptoms, based on confirmatory factor analysis (CFA) using structural equation modelling, offer a methodological alternative to compare antipsychotics on empirically justified latent factors. The present report is a refined analysis of a published double-blind study on the D2/D3-selective antagonist amisulpride (ASP) versus the mixed D1-5/5-HT2 antagonist flupentixol (FPX). CFA was applied to Brief Psychiatric Rating Scale, Scale for the Assessment of Negative Symptoms, Bech-Rafaelsen Melancholia Scale and Simpson-Angus Scale subscores to examine differential effects of high doses of ASP and FPX on negative and depressive symptom dimensions in 126 acutely ill schizophrenic patients. A four-factor model comprising the full spectrum of acute symptomatology and a three-factor model ('negative', 'anhedonia-apathy', 'depressive') restricted to negative and depressive symptoms were yielded with an identical 'depressive' dimension in both models. Analyses of CFA-derived factor scores showed that ASP was significantly superior to FPX regarding the latent 'depressive' dimension, independent of baseline scores, dosage and changes in akinesia. Neither the negative' dimension nor 'anhedonia-apathy' showed significantly different treatment effects. CFA-based analyses appear to be suitable for psychotropic drug evaluation when more refined and data-related information on drug efficacy profiles are required.