Document Type

Dissertation

College

College of Arts and Sciences

Department

Psychology

Degree

PhD in Behavior Analysis

Dissertation Defense Date

2025-01-24

First Committee Member

Dickson, Chata A

Second Committee Member

Ahearn, William H

Third Committee Member

Thompson, Rachel H

Abstract

This study examined the recombinative generalization (RG) outcomes in three variations of matrix training conditions: overlapping of training components (MET), non-overlapping of training components (Control), and non-overlapping of training components with a mediating response (MeR). Participants were first taught to translate three different sets of Malay words into English on a computer; each training set was associated with a training condition. Following training, participants completed RG probes with written or multiple-choice paper assessments. In Experiment 1, a multielement design with all conditions was implemented with four participants. In Experiment 2, a reversal and alternating-treatments design was implemented with six other participants to evaluate the RG outcomes of each condition when trained in isolation. When the mediating response was available, all ten participants demonstrated either partial or complete generalization during RG probes. Although findings from Experiment 1 suggested that training with non-overlapping matrices (Control and MeR) was the most effective and efficient, findings from Experiment 2 demonstrated that training with overlapping matrix (MET) was most effective in producing positive RG outcomes, and carryover effects from MET in Experiment 1 may have led to the positive generalization outcomes attributed to the other conditions. The limitations of each training condition were assessed with a generalization error analysis, offering important insights for educators and practitioners to identify the most suitable training strategies that optimize generalization outcomes. We suggest areas for future research, including the evaluation of an alternative overlapping matrix planner (MET).

Share

COinS