Discriminability and Distinctiveness in Complex Arrays of Simple Elements
Werner K. Honig, Dalhousie University, Halifax, Nova Scotia
The procedures used to assess the discriminability of stimuli in animal psychophysics require differential reinforcement; for example, correct identifications of a stimulus are rewarded, while false alarms or "misses" are not ( Blough & Blough, 1977). Such differential reinforcement may well enhance the discriminability of the stimuli, as their differential association with reward would make them distinctive. Indeed, there is a history of research on acquired distinctiveness of cues ( Lawrence, 1949; see Osgood, 1953, for a review). In the area of animal psychophysics, the issue of acquired distinctiveness of cues needs to be addressed. Ideally, it should be possible to separate the discriminability of stimuli from their associative value, in order to equate their distinctiveness. In practice, this is difficult, because the training procedures normally involve differential reinforcement. Even in cases where this can be avoided, as in a conditional discrimination (e.g., if the stimuli are the same, go left; if different, go right), the different responses associated with the stimuli may make them more distinctive and thus enhance their discriminability.
In addition to the specific effects of discrimination training on distinctiveness, there may be more general effects of an "attentional" nature ( Mackintosh, 1977). In a well-known study, for example, Jenkins and Harrison ( 1960) showed that the dimension of tonal frequency gained stimulus control following interdimensional training, in which pigeons were first taught to discriminate the presence of a tone from its absence. Such general effects of discrimination training on the distinctiveness of stimuli again makes it very hard to "isolate" discriminability of the stimuli as a determinant of stimulus control.