Heterogeneity in schizophrenia is an old concept, already evoked by Kraepelin and Bleuler. During the XXth century, this conception has led to a multiplication of diagnostic systems, which allow to define similar groups of patients from a country to another. However, these systems cannot select homogeneous groups of patients, since they include "polythetic" diagnostic criteria. Moreover, comparison of groups selected by different diagnostic systems shows that they are very heterogeneous too. It is why a polydiagnostic approach could be a relevant strategy: this consists of using many diagnostic systems at the same time in order not to select an arbitrary subgroup of schizophrenic patients over another. Therefore, several ways are possible to isolate the most homogeneous groups of patients from the heterogeneous population obtained with the polydiagnostic approach, depending on the aim of the research. A first possibility is a dimensional approach, in individualizing syndromic dimensions from psychopathological scales using principal component analyses. This approach might be useful to test the efficiency of an antipsychotic on a syndromic dimension. A second possibility is to identify groups of patients characterized by the largest homogeneity inside each group and the largest differences between groups. One way to perform it is the cluster analysis. Authors present a study using such a method: 5 different subgroups are identified in 138 schizophrenic patients. This approach is useful to study a subgroup of patients in comparison with a control group. A third possibility is represented by Carpenter et al. approach (7) with the definition of the deficit syndrome; a very homogeneous group of patients defined by standardized criteria can be individualized and characterized by physiopathologic anomalies.