A Computer-Based Laboratory Project for the Study of Stimulus Generalization and Peak Shift.

Abstract

One of the difficulties faced by instructors of courses on the psychology of learning is to develop laboratory projects that effectively recreate the methods used in actual research. The traditional steady-state research methodology used with pigeons or rats does not readily lend itself to classroom investigation because of the extensive timeframe involved, the difficulty in procuring the necessary equipment, and problems with finding an appropriate housing arrangement for the subjects. Zuriff (2005) recommended that instructors use laboratory exercises that recreate the "golden oldies" of past behavior-analytic research, that is, classic experiments that illustrate basic learning processes. Among the numerous advantages of this approach is that laboratory exercises of this kind can be conducted with commercially available computer simulations of animal behavior if live animals are unavailable (e.g., Graf, 1995; Venneman & Knowles, 2005). The laboratory project described here is intended to help instructors create a laboratory project involving original data collection from live subjects. In brief, human participants are used in lieu of nonhumans. Also, the project does not require extensive investigations of single individuals; rather comparisons can be between-groups. The topic of the laboratory project is stimulus generalization and peak shift, an area that qualifies as both a "golden oldie" (cf. Ghirlanda & Enquist, 2003; Honig & Urcuioli, 1981) and the subject of ongoing research and theory development (e.g., Blanco, Santamaria, Chamizo, & Rodrigo, 2006; Derenne, Breitstein, & Cicha, 2008; Ghirlanda & Enquist, 2007; Lynn, Cnaani, & Papaj, 2005; Martindale, 2006; McLaren & Mackintosh, 2002; Spetch, Cheng, & Clifford, 2004). In the traditional method of measuring stimulus generalization, participants first receive generalization training followed by a generalization test. During training one stimulus (S+) is presented repeatedly and responses in its presence are reinforced. During testing, a variety of related stimuli are shown and stimulus generalization is measured as responses to stimuli other than S+. The graphical depiction of response frequency to each stimulus constitutes a generalization gradient. The generalization gradient is usually symmetrical in shape with the modal response or "peak" of the gradient directed to the S+. If participants receive discrimination training instead, responses to the S+ are reinforced while responses to a second stimulus (S-) are not. This procedure often produces an asymmetrical generalization gradient, with the gradient displaced away from S+ and towards stimuli on the opposite end of the stimulus dimension from S-. Peak shift describes displacement in the modal response; if the mean response is displaced but the modal response is not, then the term area shift may be used (Rilling, 1977). Research on stimulus generalizationn and peak shift has used both humans and nonhumans. When nonhumans are used, the reinforcer is typically food and subjects are trained with response-contingent food in the presence of one stimulus (S+) but not the other (S-). When humans are used, the "reinforcer" is typically feedback indicating that the response was correct or incorrect. The generalization test also varies. For nonhumans, responses to the test stimuli (including S+) are extinguished to prevent additional learning. For humans, meaningful feedback about response accuracy is withheld (subjects may be told that a response was recorded, but not whether the response was correct or incorrect; cf. Derenne, 2006). Despite these important differences in procedure, many of the variables investigated with humans and nonhumans have been shown to have a similar effect (e.g., Wills & Mackintosh, 1998; Thomas, 1993). Keith (2002) has also recommended peak shift as a topic of in-class laboratory investigations. Keith, however, suggested that the project should be conducted with materials that the students create. …