Sutherland's two stage theory of discrimination learning is briefly described. Three experiments on overtraining and its effects on reversal and non-reversal shifts which were carried out by Mackintosh (1962, 1963a, b) and which are held to support this theoretical approach are discussed. The interpretation of these results is criticised in terms of possible differences between the nominal stimulus and the functional stimulus. Mackintosh assumes the identity of these two categories in his interpretations which are claimed to support Sutherland's formulations and to provide difficulties for other theoretical approaches. Control studies of a training and transfer type which might have indicated the effective stimulus were not carried out and, so it is argued, the experiments fail to provide unambiguous support for Sutherland's approach to discrimination learning problems. The identity of the mechanisms mediating transfer along a continuum and the effects of overtraining on reversal and non-reversal shifts is questioned. The lack of compelling evidence to support the notion of sensory dimensional analysers is suggested. Reinterpretation is made in terms of generalizations from a range of learning experiments, and differences between the implications of this approach and that of Sutherland are indicated.