Testing the models? New data and commentary on Stanhope & Cohen (1993)

Burton & Bruce (1992) proposed an extension of the interactive activation model of face recognition (Burton, Bruce & Johnston, 1990), in which the difficulty of retrieving proper names was explained without requiring an extra stage specific to proper name retrieval as suggested in the model of Bruce & Young (1986). Recently, Stanhope & Cohen (1993) have published data which it is claimed cast doubt upon both the Burton & Bruce (1992) and the Bruce & Young (1986) models of name retrieval. In this paper we first provide experimental data from a similar methodological ‘test’ of the Burton & Bruce model, which appears to support its prediction that information which is shared by several people should be easier to retrieve than information which is unique. We show that subjects are better at learning novel names or novel occupations that are shared by many faces than those that are unique. These data appear at odds with those of Stanhope & Cohen (1993). However, in our discussion we note a number of difficulties in interpreting the Stanhope & Cohen experiments. Moreover, we critically examine the use of paired associate learning experiments, such as those presented here and those of Stanhope & Cohen (1993), to test any model of the ‘steady state’ of a system, and make suggestions about the kind of future work which could indeed provide critical tests of these models.