Appendix
Results
EXPERIMENT 1.
To assess whether the performance of the four maximizing capuchins depended on the numerical ratio of the food quantities represented by token offers, we compared the percentage of correct choices between conditions by t-tests for dependent samples (Figure 2a). As expected, for three out of the four maximizing subjects, performance was higher (although not significantly) in the condition 1B vs. 1A (ratio: 33.3%) than in the condition 1B vs. 5A(ratio: 60%) (Gal: t39 = 1.68, p = 0.10, NS; Paprica: t39 = -1.26, p = 0.21, NS; Pepe: t39 = -1.73, p = 0.09, NS), whereas the fourth subject (Sandokan) showed an opposite pattern, being his performance higher in the condition 1B vs. 5A than in the condition 1B vs. 1A (t39 = 2.50, p < 0.05). Similarly, for all subjects, performance was higher in the condition 1B vs. 5A (ratio: 60%) than in the condition 1B vs. 2A (ratio: 66.7%) (Gal: t39 = 2.01, p = 0.05; Paprica: t39 = -1.69, p = 0.10, NS; Pepe: t39 = 0.45, p = 0.65, NS; Sandokan: t39 = 4.96, p < 0.0001). However, when comparing the conditions 1B vs. 2A (ratio: 66.7%) and 1B vs. 4A (ratio: 75%), three out of the four maximizing capuchins showed an opposite pattern to that expected on the basis of the ratio between quantities, i.e. their performance was higher in the condition 1B vs. 4A than in the condition 1B vs. 2A (Paprica: t39 = 0.65, p = 0.52, NS; Pepe: t39 = 0.56, p = 0.58, NS; Sandokan: t39 = 2.47, p < 0.05). Only Gal’s performance was higher (although not significantly) in the condition 1B vs. 2A than in the condition 1B vs. 4A: t39 = -0.10, p = 0.92, NS (Figure 2a).
Other four capuchins (Virginia, Pippi, Robinia, and Cammello) preferred token B regardless of condition (Table 1), although their performance showed some differences between conditions. For two out of these four subjects, performance was higher in the condition 1B vs. 2A than in the condition 1B vs. 3A (Virginia: t39 = 2.62, p < 0.05; Pippi: t39 = 1.61, p = 0.11, NS), whereas one subject (Cammello) showed the opposite pattern (t39 = -0.89, p = 0.38, NS), and the fourth one (Robinia) performed similarly in both conditions. For three out of the four subjects, performance was higher in the condition 1B vs. 3A than in the condition 1B vs. 4A (Virginia: t39 = 4.03, p < 0.0001; Pippi: t39 = 3.02, p < 0.01; Cammello: t39 = 0.73, p = 0.47, NS), whereas the fourth subject (Robinia) showed the opposite pattern (t39 = -1.27, p = 0.21, NS). Finally, for three out of the four subjects, performance was higher in the condition 1B vs. 4A than in the condition 1B vs. 5A (Virginia: t39 = 3.22, p < 0.01; Cammello: t39 = 0.98, p = 0.33, NS; Robinia: t39 = 2.50, p < 0.05), whereas the fourth subject (Pippi) showed the opposite pattern (t39 = -0.47, p = 0.64, NS).
Finally, two capuchins (Robot and Carlotta) significantly preferred token B only when presented against one token A; in all other conditions (except for Carlotta in 1B vs. 2A), they significantly preferred tokens A (Table 1). To assess whether the performance of these two capuchins was merely based on the number of tokens A presented, we compared the percentage of correct choices between conditions by t-tests for dependent samples. Since it was not possible to identify a “correct” choice in the condition 1B vs. 3A (because the two offers yield to the same amount of food), we limited our analysis to the conditions 1B vs. 2A, 1B vs. 4A, and 1B vs. 5A. As expected, capuchins’ performance increased with the number of tokens A presented and their performance was higher (i) in condition 1B vs. 5A than in condition 1B vs. 4A (Robot: t39 = 2.08, p < 0.05; Carlotta: t39 = 0.15, p = 0.88, NS), and (ii) in condition 1B vs. 4A than in condition 1B vs. 2A (Robot: t39 = -19.88, p < 0.0001; Carlotta: t39 = 7.62, p < 0.0001).
EXPERIMENT 2
To assess whether the performance of these individuals depended on the numerical ratio of the food quantities represented by token offers, we compared the percentage of correct choices between conditions by t-tests for dependent samples. As predicted on the basis of the ratio between quantities, the performance of the two maximizing subjects was significantly higher in the condition 1B vs. 5A (ratio: 60%) than in the condition 2B vs. 4A (ratio: 66.7%) (Carlotta: t19 = 2.18, p < 0.05; Sandokan: t19 = -2.52, p < 0.05). Similarly, their performance was higher in the condition 1B vs. 4A (ratio: 75%) than in the condition 2B vs. 5A (ratio: 83.3%) (Carlotta: t19 = -4.34, p < 0.001; Sandokan: t19 = -1.71, p = 0.10, NS). However, in contrast to what expected, Carlotta’s performance was significantly higher in the condition 1B vs. 4A (ratio: 75%) than in the condition 2B vs. 4A (ratio: 66.7%) (t19 = -2.44, p < 0.05), whereas Sandokan’s performance was similar in both conditions (Figure 2b).
3