Evaluating the impact of differences in the definition of overlap training on target acquisition in non-square matrices

Abstract

Matrix Training is a way of training stimulus relations which allows untrained targets to be acquired through recombinative generalization by training a specific sequence of targets in the array. Because, given a constant size of an array, a training sequence with fewer targets results in more targets which can be acquired through recombinative generalization, it is important to identify the training sequence with the fewest targets which still results in learners acquiring all the recombinative generalization targets as this results in the most efficient training. This study evaluates two different definitions of overlap training, a form of matrix training, to see which definition results in the most efficient training sequence in a non-square matrix. This study found that for one participant, the training sequence using the definition put forth in Curiel et al. (2021) resulted in 91.67 percent correct responding for all targets in the probes, and 100 percent correct responding for targets in the exclusionary responding row, making training more targets so that the training sequence met the definition of overlap training put forth in Pauwels et al. (2015) was unnecessary. The other participant's correct responding remained low through both training sequences, indicating the difference in definitions did not impact correct responding.