In response to and in defense of Mitchell and Revusky: An analysis of nonassociative effects

Abstract

The existence of factors and their contribution to both traditional and contemporary analyses of learning have recently generated considerable interest and discussion (Bitterman, 1975; Lol.ordo, 1978; Mackintosh, 1974). The recent exchange of comments between Mitchell (1977) and Revusky (1977) reflects that food aversions have not escaped a similar concern. The difficulty in attempting to elucidate the possible role of factors in the avoidance of previously poisoned foods, however, is made clear from the discussion between Mitchell and Revusky. The problem arises from the fact that, while presenting opposing views, Mitchell and Revusky clearly are both correct. Mitchell asserts that controls have, in general, been omitted in the analysis of the determinants of food aversions. Mitchell is correct. Revusky, on the other hand, argues that the experimental literature on food aversions has sufficiently analyzed the role of factors. In these analyses appropriate controls were run. Revusky, too, is correct. The issue becomes a bit clearer when one defines nonassociative. It is readily apparent that this concept has a number of meanings. The literature on food aversions in the late 50s and early 60s did assess the possible role of factors in food aversions (see Riley & Clarke, 1977; effects). In these early assessments, nonassociative apparently meant the contribution of factors other than the control exerted by the CS-US contingency, i.e., the taste-illness pairing. Such control groups as taste alone, illness alone, and taste and illness specifically unpaired were all examined. The results were unequivocal. Unless the taste and illness were paired, there was no aversion to the test solution. It is important to note exactly what these controls were assessing. The controls were assessing the contribution of factors that were independent of the contingency between the CS and US. As controls for this purpose, these groups were appropriate and adequate. Revusky is correct. A second definition of effects and a totally separate approach to their measurement were introduced by Chitty (1954), Barnett (1963), and Rzoska (1954), and have recently been thoroughly analyzed (Best & Batson, 1977; Carroll, Dine, Levy, & Smith, 1975;Domjan, 1977). This second approach redefines with the addition of neophobia-the rat's innate tendency to avoid consumption of novel foods and solutions. Neophobia, which is accentuated by illness or stress, could be a factor that contributes to or totally mediates what has been assumed to be conditioned, i.e., the avoidance of previously poisoned foods. The aforementioned controls that tested for the effects of taste alone, illness alone, or unpaired taste and illness presentations are no longer necessarily adequate. The effects of these manipulations must be assessed on novel test solutions and within a design in which neophobia can be demonstrated. It is important to note that the controls that originally tested for effects seldom allowed for the assessment of neophobic effects. There does appear to be a paucity of appropriate controls. Mitchell is correct. That neophobia exists and controls the rat's consumption of novel solutions is not in question. Its role in food aversions is. To criticize past research for its lack of precognition, however, is hardly appropriate (Bitterman, 1975, 1976, 1977; Mitchell, 1977, 1978). To criticize current research for attempting to investigate thoroughly a phenomenon is also inappropriate (Garcia, 1978; Revusky, 1978; Smith, 1978). The proper assessment of both associative and factors in the avoidance of previously poisoned foods is necessary and correct.