Carpentier, Smeets, & Barnes-Holmes, 2005

The authors, researchers at the national Universityof Ireland and Leiden University, based their study on previous research that had shown that training of match-to-sample tasks with AB compounds as samples and unitary C stimuli as comparisons (AB-C) leads to novel tasks with interchangeable stimulus elements: C-AB, AC-B, and BC-A (e.g., Markham & Dougher, 1993). These performances remained intact when, after subsequent C-D training, the C stimuli were replaced by novel D stimuli (D-AB, AD-B, BD-A). The present study was conducted to investigate if these performances will also be evident if an element of the ABC triad (A) is replaced by another triad element (B) rather than by a fourth stimulus (D). In Experiment 1, adults were trained on multiple AB-C tasks and tested for corresponding C-AB and AC-B performances. At that point, some subjects received A-B training followed by C-BB and BC-B probes. These probes were the same as the C-AB and BC-A probes except that the A element was replaced by an equivalent B element. Other subjects received A-D training followed by C-DB and BC-D probes. These probes were the same as the C-AB and BC-A probes except that the A stimuli were replaced by corresponding D stimuli. The C-BB and BC-B probes (A-B subjects) led to a massive performance breakdown whereas the C-DB and BC-D probes (A-D subjects) did not. The CBB and BC-B performances improved when, in Experiment 2, these probes were preceded by AB-BB probes. Experiment 3 demonstrated that A-B training generated class-consistent responding during AB-BB, C-BB, BC-B probes and subsequent probes with novel X stimuli (AB-XX, C-XX, XC-X ), but only when the AB-BB, C-BB, and BC-B probes were presented first. These findings indicate that elements of previously established sample-comparison configurations not only can be replaced by novel stimuli but also by same-class elements of the same or of different combinations.