Extending the Rescorla-Wagner theory to account for transswitching

This article is a brief response to commentaries made by Bond and Siddle (1988) and Lovibond (1988) to our article (Kimmel & Lachnit, 1988) outlining some of the difficulties encountered by the Rescorla-Wagner theory in predicting the outcome of transswitching experiments. We show that Bond and Siddle's assertion that increasing the value of beta would "bring the model into line with the empirical data" (p. 186) is incorrect. Increasing beta in simulated transswitching experiments causes the theory to predict a more rapid increase in associative values, to higher levels, but does not change the degree of predicted phasic switching. We agree with Lovibond's observation that the Rescorla-Wagner theory can predict phasic switching more effectively if it is assumed that an unobserved additional stimulus is present whenever the phasic stimulus occurs in positive tonic segments and a different unobserved additional stimulus is present whenever the phasic stimulus occurs in negative tonic segments. But this kind of usage of the "unique cue hypothesis" does little more then immunize a theory against problematic empirical findings. When the salience of the unique cue is increased relative to that of the stimuli that are actually physically present, simulated transswitching experiments show greater and greater phasic switching. But, unless there is a good a priori reason to postulate such unique cues, or unless the unique cue hypothesis leads to empirically testable new hypotheses, it does not provide a scientifically useful solution to the theory's problem with transswitching.