Empirical papers published between 1997-2001 from four health psychology journals which tested or applied one or more social cognition model (TRA, TPB, HBM and PMT, n=47) were scrutinised for their pragmatic and conceptual basis. This assessment indicated that in terms of their pragmatic basis these four models were useful for guiding research. The analysis of their conceptual basis was less positive. First, it is argued that these models do not enable the generation of hypotheses as their constructs are unspecific. They therefore cannot be tested. Secondly, they focus on analytic truths rather than synthetic ones and the conclusions resulting from their application are often true by definition rather than by observation. Finally, they may create and change both cognitions and behaviour rather than describe them.