Simulation of associative learning with the replaced elements model

Associative learning theories can be categorized according to whether they treat the representation of stimulus compounds in an elemental or a configural manner. Since it is clear that a simple elemental approach to stimulus representation is inadequate, there have been several attempts to produce more elaborate elemental models. One recent approach, the replaced elements model (Wagner, 2003), reproduces many results that have until recently been uniquely predicted by Pearce's configural theory (Pearce, 1994). Although it is possible to simulate the replaced elements model using "standard" simulation programs, the generation of the correct stimulus representation is complex. The present article describes a method for simulation of the replaced elements model and presents the results of two example simulations that show differential predictions of replaced elements and Pearce's configural theory.