The effects of stimulus similarity on different types of behavioral contrast

Abstract

Pigeons were trained on multiple schedules with component stimuli of different degrees of similarity. In Experiment 1, a two-component schedule was used in which the two stimuli were either two line orientations or a line orientation versus a diffuse color. Reinforcement rate was varied in one component to determine the effects of stimulus similarity on different aspects of behavioral contrast. Contrast in terms of average response rates (molar contrast) was larger with less similar stimuli. Local contrast effects at the beginning of the component were larger for more similar stimuli, but these effects were more variable and did not attain statistical significance. Independent of the level of molar contrast, the local pattern of schedule interaction differed for the two levels of similarity: with more similar stimuli, the maximum degree of interaction occurred at the beginning of the components and then decreased; with less similar stimuli, the degree of interaction increased throughout the components and was at its maximum near their end. In Experiment 2, the same three stimuli were used while reinforcement rate in the middle component of a three-component sequence was varied; this isolated the effects of the preceding schedule from those of the following schedule. Contrast effects were generally greater in the target component preceding the variable schedule, and these were enhanced by less similar stimuli. Contrast in the target component following the variable schedule was manifested primarily in terms of the behavior at the beginning of the component, and these effects were inconsistently related to stimulus similarity. The functional separation of the effects of stimulus similarity on the different locations of contrast suggest that "anticipatory contrast" and "local contrast" depend upon different mechanisms, thus excluding any account of contrast solely in terms of relative rate of reinforcement.An important distinction in the study of contrast effects in free-operant schedules is that between local versus molar contrast. Molar contrast refers to the average change in response rate as a function of the change in reinforcement rate in the alternative component(s); local contrast refers to the larger than average rate changes in the early portions of the target component (immediately following the transition from the alternative schedule).

Overall and local contrast in multiple schedules: Effects of stimulus similarity and discrimination performance

“…However, a is a highly derived value that combines information about rates in both multiple-schedule components. Furthermore, as seen in Williams (1988), local effects based on a differ from those based on rates during the unchanged component. This point is important because White, Pipe, and Mclean (1984) found that a increased with increasing dissimilarity among signaling stimuli; moreover, this effect was most pronounced at component onset.…”

“…Although these effects were not strong enough to yield within-eomponent ratios significantly different from 1.0, baseline was not an effective reference condition. Thus, as in Williams's (1988) study, local contrast in the present report is defined in terms of patterns obtained within experimental conditions. A second source of confusion occurs when within-eomponent and between-eomponent definitions are poorly distinguished.…”

“…An additional issue in measurement concerns the use of the parameter a, which describes the sensitivity of response ratios to reinforcement ratios (Williams, 1988). This measure makes intuitive sense, because the dependence between the two ratios accounts for overall contrast (Hermstein, 1970).…”

“…In his review, Williams (1983) pointed to the transitory nature of local contrast as a major feature that distinguishes local from overall contrast. In this and subsequent papers (Williams, 1988; This research was supported by Grant BNS 80-25515 from the National Science Foundation. I thank StephenR.…”

“…Malone (1976) explored this issue in an experiment that used multiple line-tilt stimuli and concluded that stimulus similarity diminished withincomponent contrast. In his comparison between intra-and interdimensional stimulus pairs, Williams (1988) did not obtain reliable within-eomponent local contrast.…”