Associative concept learning in animals

Abstract

Nonhuman animals show evidence for three types of concept learning: perceptual or similarity-based in which objects/stimuli are categorized based on physical similarity; relational in which one object/stimulus is categorized relative to another (e.g., same/different); and associative in which arbitrary stimuli become interchangeable with one another by virtue of a common association with another stimulus, outcome, or response. In this article, we focus on various methods for establishing associative concepts in nonhuman animals and evaluate data documenting the development of associative classes of stimuli. We also examine the nature of the common within-class representation of samples that have been associated with the same reinforced comparison response (i.e., many-to-one matching) by describing manipulations for distinguishing possible representations. Associative concepts provide one foundation for human language such that spoken and written words and the objects they represent become members of a class of interchangeable stimuli. The mechanisms of associative concept learning and the behavioral flexibility it allows, however, are also evident in the adaptive behaviors of animals lacking language.

Keywords: associative concepts; equivalence; many-to-one matching; within-class representation.

Fig. 1

Transfer of training for individual pigeons demonstrating the class consistent responding developed during many-to-one training (adapted from Urcuioli, Zentall, Jackson-Smith, & Steirn, 1989).

Fig. 2

Retention functions following matching training in which hue samples and line-orientation samples were associated with different comparisons (one-to-one matching) or common comparisons (many-to-one matching) (adapted from Zentall, Urcuioli, Jagielo, & Jackson-Smith, 1989).

Fig. 3

Following many-to-one matching training pigeons were transferred to a successive discrimination in which responding to samples associated with the same comparison stimulus during training were either both reinforced or not reinforced (consistent) or responding to one sample was reinforced but not the other (inconsistent) (adapted from Kaiser, Sherburne, Steirn, & Zentall. 1997).

Fig. 4

Following many-to-one matching training, sample–comparison associations were reversed for only the hue samples, only the line samples, or for all of the samples (adapted from Zentall, Sherburne, Steirn, Randall, Roper, & Urcuioli, 1992).

Fig. 5

Retention functions following many-to-one matching training in which red and food samples were associated with choice of a vertical-line comparison and green and no-food samples were associated with choice of a horizontal-line comparison (adapted from Zentall, Sherburne, & Urcuioli, 1995).

Fig. 6

Comparison-response rates (in pecks/sec) on arbitrary matching baseline (Training) trials and on nonreinforced symmetry (Probe) trials during the first two test sessions following successive matching training involving hue samples and form comparisons (arbitrary matching) and hue and form identity training. Data are from two pigeons (EXT2 and EXT7) run in Urcuioli (2008a) and one pigeon (PRF4) run in Urcuioli (2008b).