Responding in the Presence of Nontrained Stimuli: Implications of Generalization Error Patterns

Abstract

The technology of applied behavior analysis has documented impressive advances in procedures for changing behavior. Behavior change is often functional, however, only if it endures over time and is performed in nontrained situations. One approach to building a technology that meets these standards has been systematic analysis of generalization “successes.” The present paper extends this approach through an analysis of generalization failures (i.e., the patterns of generalization errors). A format is provided for categorizing the errors made in generalization situations, and linking these errors back to stimulus characteristics present during training. Stimulus control serves as the pivotal concept that directs this error analysis, and provides the bridge between infrahuman research on generalization and the broader applied issue of obtaining performance across nontrained situations. Implications of the error analysis are defined for building a technology of generalization, and for developing an effective research methodology for studying generalized responding in applied settings.