Although more than 20 factor analytic studies have been published on the Beck Depression Inventory (BDI), only Steer and co-workers (1987) have used a sample composed exclusively of patients diagnosed with depression. The component structure found in their study of depressed patients differs in important respects from the structure summarised in several reviews. The main aim of the present study was to investigate whether this structure could be confirmed with the BDI responses of an independent sample of 139 patients diagnosed with DSM-III Major Depressive Episode. Three principal components were extracted and rotated to maximum congruence with a target based on the results of Steer et al. (1987). The significance of the fit to this target was then evaluated by rotating the same matrix of loadings to 5000 random permutations of the target. The fit was found to be highly significant, though some possible improvements could be identified ad hoc. An alternative factor structure for the BDI, derived from covariance structure analysis by Tanaka and Huba (1984), was also tested but could not be confirmed.