Multivariate meta-analysis was performed on 39 studies of ventricular size in schizophrenia which used the ventricle:brain ratio (VBR). The size of the VBR was dependent both upon the date when studies were carried out (more recent studies showing a reduction in the difference between schizophrenics and controls), and upon the diagnostic criterion used in the studies. Methodological factors in study design seemed more important than the characteristics of the schizophrenic subjects, in determining the VBR. Our analysis suggests that there is a difference in VBR between schizophrenics and controls which would seem to be an indisputable characteristic of schizophrenia. However, the difference is smaller than has previously been thought, so that, although of undoubted theoretical interest in accounting for the aetiology of schizophrenia, it is probably too small to be of practical significance in diagnosis, or in the differentiation of subtypes.