Overcoming Response Bias Using Symbolic Representations of Number by Chimpanzees (Pan troglodytes)

We previously reported that chimpanzees were unable to optimally select the smaller of two candy arrays in order to receive a larger reward. When Arabic numerals were substituted for the candy arrays, animals who had had prior training with numerical symbols showed an immediate and significant improvement in performance and were able to select reliably the smaller numeric representation in order to obtain a larger reward. Poor performance with candy arrays was interpreted as reflecting a response bias toward the intrinsic incentive and/or perceptual features of the larger array. In contrast, the Arabic numerals represent numerosity symbolically and appear to promote response choice on the basis of abstract processing of numerosity, with minimal interference from the inherent properties of the choice stimuli. The present study tested the hypothesis that, for mixed symbol-candy choice pairs, the requisite processing of the abstract numeral may foster a mode of numerical judgment that diminishes the interfering incentive/perceptual effects of the candy stimuli. The results were consistent with this hypothesis. Whereas performance on candy-candy arrays was significantly below chance levels, performance on numeral-candy choice pairs was significantly above chance and comparable with performance on numeral-numeral pairs.numeral may foster a mode of numerical judgment that diminishes the interfering incentive/perceptual effects of the candy stimuli. The results were consistent with this hypothesis. Whereas performance on candy-candy arrays was significantly below chance levels, performance on numeral-candy choice pairs was significantly above chance and comparable with performance on numeral-numeral pairs. Chimpanzees have been shown to demonstrate remarkable cognitive capabilities that are often strikingly reminiscent of human reasoning abilities and conceptual skills (see, e.g., Boysen, Berntson, Shreyer, & Hannan, 1995; Fujita & Matsuzawa, 1986; Gillan, Premack, & Woodruff, 1981; Matsuzawa, 1985a, 1985b; Premack, 1986; Savage-Rumbaugh, 1986). For example, they have been shown to have impressive capabilities in numerical processing, symbolic representation of number, and even numerical reasoning that does not require the support of language, as typically defined (see, e.g., Boysen, 1997; Boysen & Berntson, 1989, 1995; Matsuzawa, 1985b; Murofushi, 1997). Behavioral and cognitive differences nonetheless exist between chimpanzees and humans, and those differences can sometimes be more informative than similarities for the comparative understanding of cognitive function (Boysen, Berntson, Hannan, & Cacioppo, 1996). Several years ago, we encountered a rather striking training failure with two chimpanzees (Boysen & Berntson, 1995). The task required the seemingly simple choice between two different-sized candy arrays. A reversed reinforcement contingency was imposed, so that the candies in the selected array were removed and the animal received the remaining, nonchosen array as a reward. In order to maximize payoff, it was in the best interest of the chimpanzee subject to choose the dish with the smaller number of candies, in order to reap the larger remainder. Both of the animals in the preliminary study failed to do so, and even after hundreds of training trials, they persisted in selecting the larger candy array and, thus, consistently received the smaller amount of reward. To further explore their seeming inability to acquire more optimal performance under a relatively simple task contingency, we capitalized on the previously acquired skills of one of the animals (Sheba) in counting and in the use of Arabic numerals as representations for quantities (Boysen, 1993; Boysen & Berntson, 1989). All the task rules and features were identical, except that Arabic numerals were substituted for the candy arrays. The same contingency applied, and the subject was rewarded with the number of candies represented by the nonselected numerical stimulus. Without further training, Sheba immediately invoked the optimal selection rule. As long as Arabic numerals served as stimuli, she consistently selected the smaller numeral and earned the greater number of candies. When candies were reintroduced as stimuli, performance dramatically fell but immediately returned to more optimal levels when Arabic numerals were again substituted for the candy stimuli. On the basis of these data, we suggested that there might be two incompatible response dispositions operative in this task. An associative disposition to select the smaller of the two arrays would be expected to develop on the basis of the instrumental reward contingency. The fact that such a disposition had developed was apparent when Arabic numerals were presented, resulting in more optimal performance despite the fact that this implicit knowledge did not manifest when candy arrays served as stimuli. We hypothesized that the poor performance with candy arrays was attributable to an interfering nonassociative disposition toward the direct perceptual and/or incentive features of the larger array. This may have been fostered by numerosity judgments based on subitizing or magnitude estimationprocesses that depend on the immediate perceptual features of the task stimuli (see, e.g., Gallistel, 1993). We further speculated that performance with the numerals was more optimal because these stimuli promote an abstract enumeration process that is not as closely tied to the direct perceptual/incentive features of the task stimuli (Berntson, Boysen, & Cacioppo, 1993; Boysen, 1997; Cacioppo & Berntson, 1994). By promoting an abstract mode of processing, symbolic stimuli may direct attention to relevant aspects of their referents (numerosity), while minimizing the impact of other aspects (incentive or perceptual features). What remains unsettled is the extent to which a symbolic representation, as an essential element in a cognitive task, might foster an alternative mode of cognitive processing of other nonsymbolic (candy array) elements. To explore this issue, in the present study we examined the performance of chimpanzees on the interference task as described above, with an added condition in which one of the choice stimuli was a candy array and the other was an Arabic numeral. We hypothesized that the need to process the numerical value of the symbol abstractly, in a less perceptually based mode, might foster a similar mode of processing of the paired candy array and, thereby, minimize the perceptual or incentivebased interference typically associated with such arrays. An alternative hypothesis predicted just the opposite‒‒that the incentive features of the candy array, when compared with the lack of inherent incentive features of the Arabic numerals, would lead to an even more suboptimal performance. A secondary goal of the study was to evaluate the stability of the task interference effect over the approximate 6-year period since the original study.