Stimulus Equivalencies Through Discrimination Reversals

The sensory systems of advanced animals frequently input more stimulus infonna-tion into the nervous system than the motor system can possibly output as behavior patterns. This bottleneck demands a drastic infonnation reduction. Two types of reduction can be distinguished: selective attention involving behavioral context-dependent infonnation censoring (as when during sexual behavior food stimuli are usually not reacted to); and categorization, a pooling into fewer response outputs (as when different foods all elicit the same ingestive response). Here we are concerned with processes underlying this latter kind of infonnation reduction, that is, how the nervous system manages to classify stimuli so that they result in a restricted number of behaviors. In human psychology, such infonnation pooling has been much studied under the heading of concept fonnation, where the responses of interest have been words, which through suitable experience, come to correspond semantically to collections of stimuli (Sloman and Rips, 1998). There are several theories about the formation of concepts but no single one has emerged as being uniquely correct (Fodor, 1998). The deviousness of the human mind at conceiving named categories exceeds any succinct hypothesis that can be put to paper. Consider the tenns male and female as applied to organisms from algae to humans and to objects such as the sun and the moon. "Le soleil" and "la lune," "die Sonne" and "der Mond" are conversely male and female to franco-and germanophone people. Most theories assume that the stimuli that come to be categorized together do so on the basis of perceptual similarities, either by simply being neighbors along a physical dimension such as size or wavelength, or by sharing some physical features such as feathers or legs. The contribution of less immediate similarities of stimuli or items, such as the ability to fly or to kill, or whatever confers genders to the sun and the moon, have also been considered theoretically but hav~ hardly been examined empirically. A stimulus categorization by these kind of functional attributes interests us here, though not so much in humans as in pigeons. The ability of pigeons to learn to distinguish sets of stimuli that are physically similar within sets and physically different between sets (for example, slides containing humans or not containing humans) in a directly perceivable way as belonging to different categories (as being worth or not worth pecking for food returns, for example), and the competence to then spontaneously generalize this …