Academic journal article The Psychological Record

Varieties of Stimulus Control in Matching-to-Sample: A Kernel Analysis

Academic journal article The Psychological Record

Varieties of Stimulus Control in Matching-to-Sample: A Kernel Analysis

Article excerpt

**********

The matching-to-sample (MTS) procedure has been used to study a wide range of complex psychological phenomena. These include the establishment of relations of sameness (Cumming & Berryman, 1961, 1965; Carter & Werner, 1978; Mackay, 1991; Wright, Cook, Rivera, Sands, & Delius, 1988; Wright, Santiago, & Sands, 1984; Wright, Santiago, Urcioli, & Sands, 1983), difference (Pepperberg, 1988; Stromer & Stromer, 1990), and opposite (Dymond, Roche, Forsyth, Whalen, & Rhoden, 2008); arbitrary relations between meaningful terms or objects and referential symbols (Cerutti & Rumbaugh, 1993; Gisiner & Schusterman, 1992; Kennedy, Itkonen, & Lindquist, 1994; Lynch & Cuvo, 1995; Pepperberg, 1981; Schusterman & Gisiner, 1988); the formation of equivalence classes (Fields & Nevin, 1993; Schusterman & Kastak, 1993; Sidman, 1994); object permanence (Pepperberg, 1986); and the assessment of working memory (Santi & Roberts, 1985), among others.

In these studies, the control exerted by the relation between the stimuli was indexed by the percentage of trials that occasioned the selection of comparison stimuli that were related to a sample stimulus. When the trial-based percentage approximates 100% correct responding, one can reasonably conclude that the conditional relations have been established between sample and related comparison stimuli, and the phenomenon under study has been demonstrated.

When the trial-based percentages are below 100% accuracy or above 0% accuracy, however, the sources of stimulus control that are the determinants of responding are subject to a wide range of interpretations. One is that the conditional discrimination is only partially formed; another is that behavior is being controlled by some features or relations among the stimuli in the trials other than the conditional relation between the sample and positive comparison (Iversen, 1993, 1997; Iversen, Sidman, & Carrigan, 1986; McIlvane, Serna, Dube, & Stromer, 2000; McIlvane, Withstandley, & Stoddard, 1984; Sidman, 1992; Stromer & Osborne, 1982; Tomanari, Sidman, Rubio, & Dube, 2006). Each of these forms of stimulus control is referred to as a stimulus control topography, or STC (McIlvane & Dube, 1992). Thus, the underlying phenomenon cannot be evaluated because of the interfering effects of the other forms of stimulus control, or interfering STCs. A similar point was noted by Harlow (1949) in his presentation of error factor theory, which posited that errors were not mistakes but, rather, control of behavior by aspects of a stimulus array other than that defined by the experimenter.

Sidman (1978, 1980) illustrated the problem of interpreting trial-based percentage measures of accuracy by considering trials that contain one of two samples, Al and A2, and two comparisons, Bl and B2, that can be presented on each side of the sample. A trial-based accuracy of 50% can be produced in a number of ways, three of which follow. First, 50% accuracy can be generated by the selection of comparison stimuli on a random basis. Second, 50% accuracy can be generated by the selection of the left-side comparison on all trials, regardless of the comparison stimulus in that position, which would indicate a position preference. Third, 50% accuracy can be generated by the selection of a given comparison on all trials, regardless of the comparison position of the sample stimulus, which would indicate a stimulus preference. Thus, the same trial-based percentage correct can be engendered by many different stimulus control topographies.

A similar argument was made for other trial-based percentage-correct measures. For example, 75% accuracy could be produced because the Bl. comparison is selected in the presence of Al on 75% of the trials and the B2 comparison is also selected in the presence of A2 on 75% of the trials. Thus, an overall trial-based accuracy might reflect similar levels of stimulus control exerted by the Al and A2 samples. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.