Anim Cogn (2007) 10:387–396 DOI 10.1007/s10071-007-0076-5
O RI G I NAL PAPE R
Mangabeys (Cercocebus torquatus lunulatus) solve the reverse contingency task without a modiWed procedure Anna Albiach-Serrano · Federico Guillén-Salazar · Josep Call
Received: 8 May 2006 / Revised: 12 September 2006 / Accepted: 19 January 2007 / Published online: 22 February 2007 © Springer-Verlag 2007
Abstract Problem solving often relies on generating new responses while inhibiting others, particularly prepotent ones. A paradigm to study inhibitory abilities is the reverse contingency task (Boysen and Berntson in J Exp Psychol Anim Behav Process 21:82–86, 1995), in which two diVerent quantities of food are oVered to an individual who receives the array he did not choose. Therefore, mastery of the task demands selecting the smaller quantity to obtain the larger one. Several nonhuman primates have been tested in the reverse contingency task. To date, only great apes and rhesus monkeys (Macaca mulatta) have succeeded in the original task, with no need of procedural modiWcations as the large-or-none contingency, correction trials or symbolic stimuli substituting for actual food quantities. Here, four mangabeys were presented with two stimulus arrays of one and four raisins in the context of the reverse contingency task. Three of them learned to perform the task well above chance without a modiWed procedure. They also reached above-chance performance when presented with two stimulus arrays of zero and four raisins, despite the initial diYculty of choosing a null quantity. After a period of 7–10 months, in which the animals were not tested on any task, all three subjects continued to perform well, even when presented with novel quantity pairs. A. Albiach-Serrano (&) · F. Guillén-Salazar Unidad de Etología y Bienestar Animal, Universidad Cardenal Herrera, 46113 Moncada (Valencia), Spain e-mail: [email protected]
J. Call Max Planck Institute for Evolutionary Anthropology, Deutscher Platz 6, 04103 Leipzig, Germany
Keywords Inhibitory control · Reverse contingency task · Monkeys · Quantity discrimination
Introduction Much of problem solving consists of inhibiting certain prepotent responses in favor of more indirect means of obtaining a desired goal. Some examples in which, individuals inhibit their natural tendency to directly approach food that is located outside of the reach behind a fence include taking a detour around the fence, grabbing and using a tool, or requesting help from a partner (Gómez 1990; Guillaume and Meyerson 1930; Köhler 1925). Selecting the larger of two food quantities is another prepotent response found in many species. One task that has investigated the ability to inhibit reaching for the larger quantity is the so-called reverse contingency task. In this task, individuals are presented with two food quantities and they are required to select the smaller quantity to net the larger one, whereas if they select the larger quantity, they receive the smaller one. Boysen and Berntson (1995) presented two chimpanzees (Pan troglodytes) with pairs of quantities ranging from 1 to 6. Unexpectedly, the chimpanzees failed to solve the task after 384 and 96 trials, respectively, but one of them immediately succeeded when the actual items were substituted by Arabic numerals. Follow-up experiments have conWrmed this result (Boysen et al. 1996, 1999, 2001) and shown that subjects with numerical skills may solve the task with numerals yet fail with actual items. Silberberg and Fujita (1996) also tested reverse contingency performance in Japanese macaques (Macaca
fuscata) but using only the 1:4 pair of quantities. Macaques were unable to master the task after 200 trials. However, they succeeded with a modiWed procedure that consisted of giving the subjects four raisins for choosing one but giving them no raisins (instead of one) for choosing four, the so-called large-or-none contingency (Anderson et al. 2000). This procedure was combined with another modiWcation of the task, correction trials, according to which the position left/ right of the quantities within a session, rather than being randomly assigned, remained the same until there was a successful trial. Similar results have been reported for squirrel monkeys, Saimiri sciureus, (Anderson et al. 2000, 2004), cotton-top tamarins, Saguinus oedipus, (Kralik et al. 2002; Kralik 2005), and lemurs, Eulemur fulvus and E. macaco, (Genty et al. 2004). After an initial failure with the original version of the reverse contingency task, these species succeeded with the large-or-none procedure, in most cases combined with other modiWcations of the task as the above-mentioned correction trials or the use of substitute stimuli for the quantities. In this respect, Kralik et al. (2002) found that presenting colored dishes representing the quantities, as opposed to the quantities themselves, improved the performance of cotton-top tamarins. Vlamings et al. (2006) also found that item visibility aVected the performance of the great apes because they performed better with the colored dishes than with the actual items. These results support the idea suggested by Boysen and colleagues (Boysen and Berntson 1995; Boysen et al. 1996, 1999, 2001) that substituting items for some representation of quantity improves performance. There are also some studies that have shown successful performance on the reverse contingency task without procedural modiWcations. Shumaker et al. (2001) tested two orangutans, Pongo pygmaeus, with the same procedure originally used by Boysen and Berntson (1995) involving multiple pairs of quantities. They found that both orangutans succeeded within an average of 210 trials (140 and 280 trials). However, those orangutans did not show an initial preference for the larger quantity and, therefore, one could argue that they did not have to inhibit their responses to the larger quantities. Murray et al. (2005) tested rhesus macaques with the 1:4 pair and found that on average they mastered the task in 1,087 trials (ranging from 340 to 2,700 trials), again without modifying the procedure. Vlamings et al. (2006) also found that some apes, including chimpanzees, were able to solve the 1:4 pair with an unmodiWed procedure. It took these subjects an average of 323 trials (ranging from 160 to 480 trials) to achieve above chance performance, which is comparable to
Anim Cogn (2007) 10:387–396
orangutans and substantially faster than rhesus macaques. Restricting the analysis to the Wrst 480 trials, which was the maximum number of trials given to the apes, showed that only one out of six rhesus macaques (17%) had solved the task compared to seven out of 18 apes (39%). Vlamings also tested the 0:4 pair, a pair that had not been mastered in previous studies with chimpanzees (Boysen et al. 1996, 1999), probably because pointing to zero to obtain four is harder than pointing to one (Vlamings et al. 2006). Again, some apes also succeeded with this pair although their performance was not as good as with the 1:4 pair. Some studies have demonstrated long-term retention of the reverse contingency. Anderson et al. (2004) found that squirrel monkeys still performed above chance after 8 months without testing. Similarly, Genty et al. (2004) found that lemurs also performed well 3 and even 18 months after their initial study, although their performance was poorer. Additionally, some studies have found good transfer of the reverse contingency ability to new pairs of stimuli. Anderson et al. (2000, 2004) found that squirrel monkeys still performed well when presented with novel array pairs diVerent from the 1:4 with which they were Wrst trained on the reverse contingency. Genty et al. (2004) and Kralik (2005) found similar results for lemurs and cotton-top tamarins, respectively. In sum, there are several studies on the reverse contingency task in several primate species including rhesus and Japanese macaques, squirrel monkeys, cottontop tamarins, lemurs, chimpanzees, orangutans, gorillas (Gorilla gorilla), and bonobos (Pan paniscus). The main results are as follows. Reverse contingency is a task that is hard to master. Most species that have been tested have required several procedural modiWcations such as substituting the quantities by Arabic numerals or colored dishes, and/or the implementation of special reinforcement regimes. Only rhesus macaques and apes have solved the task without such procedural modiWcations and it appears that apes can solve the task faster than rhesus macaques. Finally, quantities involving zero are particularly diYcult. The goal of this study was to test mangabeys on the reverse contingency task in the absence of a modiWed procedure. Since rhesus are the only monkeys to have been shown to master the original version of the task, it is important to determine whether this Wnding is speciWc to rhesus monkeys or whether it would include other Old World monkeys such as mangabeys on which little is known. We selected the 1:4 pair because it has been used in most previous studies thus facilitating comparative analyses. Moreover, we included the 0:4 pair because pairs with zero have proven particularly
Anim Cogn (2007) 10:387–396
diYcult and, therefore, mastery of the task in this condition may indicate strong assimilation of the reverse contingency. Finally, we tested long-term retention of the reverse contingency after a period of several months and transfer to new quantities, to compare with previous studies. Transfer to new quantities is especially important to know if subjects had learned a Wxed rule (e.g., avoid the four raisins) or they had instead developed a more generalized ability to solve the reverse contingency task (e.g., choose the smaller quantity).
Methods Subjects Four male white-crowned mangabeys aged 6-, 8-, 9and 16-years old, and designated S1 to S4 respectively, participated in the present study. However, S4 took a long time to habituate to the testing situation and even then often refused to participate in the experimental sessions so it had to be dropped from the study. Subjects were housed at the Valencia Zoo (Spain), as a group in an outdoor cage during the day, in pairs in two indoor cages at night. The experiment was held in one of these indoor cages. We could easily move the subjects between the indoor cages (Wrst with the help of peanuts, until we established a routine) so they could be isolated when necessary. None of the subjects had taken part in any experimental study before. They were fed twice a day on a diet of fresh fruits, vegetables,
monkey chow, and yogurt. Dried fruits and nuts, normally used as environmental enrichment, were removed from their diet in order to keep the nutritional balance while the study took place. Subjects were not water-deprived during testing. Materials The apparatus (Fig. 1) was attached to the mesh of one of the indoor cages. It consisted of a Plexiglas panel (90 cm £ 40 cm) with three holes (3.6 cm in diameter) at the bottom forming a straight line separated by 30 cm (from center to center). A wooden box (90 cm £ 40 cm £ 25 cm) Xush against the mesh just below the Plexiglas panel was used as a table-drawer. White plastic dishes (11 cm in diameter) containing food quantities were placed on a sliding wooden platform (90 cm £ 40 cm) resting on the box. We used raisins (a favorite food of the subjects) as rewards. Subjects could introduce their Wngers through the panel holes to select alternatives and receive the rewards. We prevented subjects from observing the baiting process by using an articulated wooden visual barrier (151 cm £ 40 cm). The apparatus was illuminated from above while the rest of the room remained unlit. Procedure Subjects were tested individually every morning, before feeding and release to the outdoor cage. In order to avoid cuing the animals, the experimenter sat
Fig. 1 Test apparatus and sequence of steps in the presentation procedure: baiting process (a), visual barrier removal (b), dishes positioning (c), oVer by pushing the platform forward (d), subject reward after choice (e). See text for details
straight in front of the test apparatus, trying to keep symmetry in body position and hand movements, she kept a neutral expression and remained silent throughout the experiment. Testing was divided into three phases: pretest, test, and retention and transfer. The pretest assessed whether subjects had an initial preference for the larger food quantity. This assessment was a critical pre-condition to advance to the test phase because if subjects do not show any preference for larger quantities the reverse contingency task is moot—there is no preference to be overcome. There were two steps in the pretest: Pretest pairs We administered the following pairs of quantities: 0:1, 0:4 and 1:4 in 12-trial sessions (4 trials per pair). The quantities position (left/right) was counterbalanced and trials were presented in random order with the provision that the larger of the two quantities could not appear more than three consecutive times in the same position. The experimenter baited the dishes behind the visual barrier, placing them in the middle of the platform (Fig. 1a). Once the visual barrier was removed (Fig. 1b) and the subject had looked at the dishes, the experimenter pulled them in opposite directions until they were aligned with the left and right holes of the Plexiglas panel (Fig. 1c) and then pushed the platform forward (Fig. 1d). When the subject had chosen a food quantity by reaching or sticking its Wngers in one of the two panel holes, the experimenter simultaneously pushed the chosen dish forward and pulled the other one backward (Fig. 1e) and gave the subject the content of the chosen dish, through its respective hole. If the subject failed to choose within 5 min the session was terminated. A session was considered successful when the subject selected the larger quantity in at least 10 out of 12 trials. Three consecutive correct sessions were needed to advance to the next step. Pretest trios We used the same quantities as presented before, in trios (i.e., 0:1:4). This allowed us to assess whether subjects preferred the largest of all the quantities when presented together. The presentation procedure was similar to the one used in the previous step except that the three dishes were placed in front of each panel hole from the beginning, so that there was no need to pull the dishes apart after the visual barrier removal. Sessions also consisted of 12 trials and the criterion of success was the same as before.
Anim Cogn (2007) 10:387–396
Upon completing the pretest, subjects received the test phase, in which we examined subjects’ ability to select the smaller quantity in a pair to obtain the larger one. In the test phase, we used the same basic presentation procedure as in the pretest phase except that after the subject’s choice, the experimenter pushed forward the non-selected dish and pulled backward the selected one in order to reward the subject with the content of the non-selected dish. We used the 0:4 and 1:4 pairs in the following three steps: Test 1 We administered both the 0:4 and 1:4 pairs within a session. Each session consisted of 16 trials (8 trials per pair). Choosing the smaller quantity in at least 14 out of 16 trials in a session in three consecutive sessions was considered a success. S1 and S2 received 12 sessions whereas S3 received 11 sessions. Test 2 We administered the 1:4 pair alone. Each session consisted of 16 trials. Subjects received testing sessions until they selected the smaller quantity in at least 14 trials out of 16 in three consecutive sessions. Test 3 Same as Test 1 except for S3, that for logistical reasons unrelated to the current study only completed two consecutive sessions above chance before going to the next phase. Upon completing the test, subjects received the retention and transfer phase in which we tested whether subjects (1) remembered the reverse contingency task after a long delay (7–10 months) and (2) generalized to novel pairs. We re-tested the subjects again on the reverse contingency task, 10 months after the end of test 3 for S1 and 7 months for both S2 and S3. No other experiments were conducted in the time between the test phase and the retention and transfer phase. The presentation procedure was the same as in the test phase but we used the following pairs: 1:4, 2:8 and 4:7. Sessions consisted of 12 trials (4 trials per pair). A session was considered successful when the subject chose the smaller of the two quantities in at least 10 trials out of 12. Three consecutive successful sessions were needed to reach the criterion for task mastery. Data scoring and analysis We videotaped all trials and scored the dish selected by the subject, deWned as reaching, introducing the Wngers into a hole toward a dish or orienting and moving their face toward the hole. We randomly selected 20% of the trials in each session and a second coder scored them to assess inter-observer reliability. The inter-observer reliability was excellent (Cohen’s Kappa = 0.94).
Anim Cogn (2007) 10:387–396
We analyzed the percent of correct responses in each session of the pretest, test and retention and transfer phases. Correct responses in the pretest phase consisted of selecting the larger quantity whereas correct responses in the test and retention and transfer phases consisted of selecting the smaller quantity. Due to the small number of subjects we analyzed the data at the individual level by means of two-tailed non-parametric statistics. We used the binomial test to assess whether subjects’ choices deviated from chance (P = 0.50) both overall and within sessions and Fisher’s exact test to examine the diVerences between conditions. Additionally, we used Spearman correlations to assess the changes over sessions. We only considered those sessions that subjects completed in their entirety. For instance, if subjects only completed Wve trials in one session, the session was cancelled and the data from those Wve trials were discarded. We used this procedure to ensure that subjects received the same number of trials per session and that they were suYciently motivated to work in all sessions. This meant that we discarded two sessions for S2 out of 60 total sessions [one in test 1 (6 trials) and one in test 2 (1 trial)] and one session for S3 out of 62 total sessions [test 2 (12 trials)]. S1 missed no sessions in 41 total sessions. Since the number of trials per condition varied depending on the testing phase, in some occasions we combined sessions to make the inter-phase analyses more comparable. For instance, the pair 1:4 was administered 8 and 16 times per session in test 1 and test 2, respectively. Therefore, we added the last two sessions of test 1 to compare them to the last session of test 2 and to chance.
and S2 missed only one trial. Moreover, all subjects selected the correct alternative in each of the Wrst two trials. Acquisition of the reverse contingency Test 1 (0:4, 1:4) None of the subjects mastered this stage. However, there were signiWcant diVerences between the 0:4 and the 1:4 pairs for all subjects. Overall, subjects performed better in the 1:4 pair than the 0:4 pair (Fisher test: P < 0.001 in all cases). Figure 2 shows the diVerential progress in each condition. Although subjects’ initial performance in both quantities was below chance, indicating that subjects preferred the larger quantity, they progressed to chance level in the last two sessions of the 1:4 pair combined (Binomial test: P > 0.05 in all cases). Indeed, there was a steady progress for the 1:4 pair across sessions for S1 (r = 0.78, P = 0.003, n = 12) and S2 (r = 0.73, P = 0.011, n = 11) but not for S3 (r = 0.43, P = 0.19, n = 11). In contrast, there was no progress in the 0:4 pair as none of the subjects solved a single 0:4 trial during this stage. Test 2 (1:4)
Figure 2 presents the percentage of correct trials for subjects S1 (Fig. 2a), S2 (Fig. 2b), and S3 (Fig. 2c) during the various testing phases.
All subjects mastered this stage. Figure 2 shows that subjects continued to improve their performance on 1:4 pairs in this phase (S1: r = 0.81, P = 0.001, n = 12; S2: r = 0.84, P < 0.001, n = 30; S3: r = 0.92, P = 0.001, n = 32). In fact, all subjects reached above chance performance (Binomial test: P < 0.005 in all cases), thus choosing the smaller quantity, at some point during this stage. However, subjects diVered in the number of sessions required to reach the criterion for successful performance. S1, S2, and S3 required 8, 21, and 28 sessions, respectively. Adding the number of 1:4 trials from the previous stage to the current one meant that the 1:4 pair was mastered on an average of 395 trials (number of trials: S1 = 224, S2 = 424, S3 = 536).
Test 3 (0:4, 1:4)
Overall, all subjects in the pretest preferred the larger quantity both when they were presented in pairs (Binomial tests: P < 0.001, N = 36 in all cases) or trios (Binomial tests: P < 0.001, N = 36 in all cases). All subjects were above chance for all pairs, i.e., 0:1, 0:4, and 1:4 (Binomial test: P < 0.006 in all cases). Such preference was clear from the Wrst session in which all subjects selected the correct alternative above chance (Binomial test: P < 0.01). S1 and S3 got all the trials correct
All subjects mastered this stage. The high performance on 1:4 pairs observed in the previous stage continued here for all subjects from the outset (Binomial test: P < 0.001, in all cases). Moreover, subjects in this phase also mastered the 0:4 pair, although at diVerent rates. S1 reached above chance performance after combining sessions 3 and 4 (Binomial test: P = 0.004), S2 after combining sessions 4 and 5 (Binomial test: P = 0.001), and S3 after combining sessions 1 and 2 (Binomial test:
Anim Cogn (2007) 10:387–396
Fig. 2 Percent of correct responses in sessions for subjects S1 (a), S2 (b) and S3 (c), during the pretest, test and retention and transfer (R&T) phases. Correct responses in the pretest phase represent trials in which subjects selected the larger quantity (D: direct contingency task) whereas correct responses in the test and retention and transfer phases represent trials in which subjects selected the smaller quantity (R: reverse contingency task). Quantity pairs are shown in brackets. Arrows in each phase depict the session in which subjects mastered the task by reaching the criterion. Note that in c Test 3 the arrow is missing. We were unable to administer the complete testing schedule to subject S3 due to logistical reasons, and therefore no more than two consecutive sessions were performed above chance before going to the next stage
P = 0.001). Adding the number of 0:4 trials from test 1 to the current test 3 meant that the 0:4 pair was mastered on an average of 120 trials (number of trials: S1 = 128, S2 = 128, S3 = 104). Such a relatively quick acquisition of the reverse contingency for the 0:4 pair contrasts with the poor performance demonstrated during the initial testing stage in which subjects did not solve a single 0:4 trial. Indeed, comparing the last 2 sessions of test 1 with the Wrst 2 sessions of test 3 indicated that S2 and S3 suddenly and immediately improved their performance (Fisher’s test: P < 0.001 in both cases, see Fig. 2b, c). This suggests that there was some transfer from the mastery of the 1:4 pair to the 0:4.
Retention and transfer All subjects mastered this phase. Overall, all subjects performed above chance in this phase (Binomial test: P < 0.001, n = 36, in all cases). All subjects were above chance in all types of pairs (Binomial test: P < 0.05 in all cases), except S3 in the 4:7 pair (score = 75%, Binomial test: P = 0.15).
Discussion Three white-crowned mangabeys learned to select the smaller of two quantities to receive the larger one in a
Anim Cogn (2007) 10:387–396
reverse contingency task, thus overcoming their spontaneous tendency to select the larger of the two quantities. This included solving the problem for the 0:4 pair as well as the 1:4 pair, although subjects mastered the 1:4 pair earlier than the 0:4 pair. Moreover, they were able to perform the task successfully after a delay of several months, and they were able to transfer this ability to novel array pairs. Thus, this study adds to the growing literature on the ability of nonhuman primates to solve the reverse contingency task (Anderson et al. 2000, 2004; Boysen and Berntson 1995; Boysen et al. 1996, 1999, 2001; Genty et al. 2004; Kralik 2005; Kralik et al. 2002; Murray et al. 2005; Shumaker et al. 2001; Silberberg and Fujita 1996; Vlamings et al. 2006). Interestingly, unlike most published studies, the present study showed that mangabeys were able to solve the task without modiWed procedures or the aid of symbolic devices representing the quantities. Comparing the current results to those of other studies that did not use modiWed procedures indicated that the performance of mangabeys was faster than that of rhesus macaques but slower than that of the apes that performed the task (Murray et al. 2005; Shumaker et al. 2001; Vlamings et al. 2006). So rhesus macaques are not the only monkeys that have shown to be able to master the original version of the reverse contingency task; there is at least another Old World monkey species that shares this ability, the white-crowned mangabey. Particularly interesting was the mastery of the 0:4 pair, given that until the current study the mastery of a pair involving 0 had only been reported in great apes (Vlamings et al. 2006). The 0:4 pair is especially challenging because it involves selecting an empty dish to get the four raisins. Although the diYculty of choosing 0 raisins might be compensated by the high cost of failing a reverse contingency trial in this condition (given that selecting 4 raisins means being rewarded with no raisin at all), our current results show that 0:4 is harder to master than 1:4. This conWrms previous studies that have indicated that the larger the disparity between the quantities in a pair, the lower is the performance in the reverse contingency task (Boysen and Berntson 1995; Boysen et al. 1996, 1999; Genty et al. 2004; Vlamings et al. 2006). Another possible explanation for the diYculty of the pairs involving zero could be the lack of experience of the subjects in selecting a null quantity, especially when the other option was food. This may have been especially diYcult for experimentally naïve individuals who might have not seen an empty dish as being a viable response option, given that they were faced with only one quantity of food. The special diYculty of the
0:4 pair compared to the 1:4 pair became clearly apparent during testing. Mangabeys progressed to chance level for the 1:4 pair but did not solve a single 0:4 trial during the initial stage. After about 12 sessions in this condition the subjects’ initial good disposition to participate in the study declined alarmingly. We suspected that the 0:4 pair was responsible for this because subjects often delayed in coming back for the next trial after failing in a 0:4 trial and because sometimes when they approached the dishes and saw the 0:4 pair they declined to participate. Therefore, we opted for ending Test 1 and starting Test 2, with only the 1:4 pair. Nevertheless, subjects were able to master the 0:4 pair after they had fully mastered the 1:4 pair. Indeed, subjects in Test 3, when the 0:4 pair was re-introduced, continued to perform at high levels in the 1:4 pair and also quickly mastered the 0:4 pair. This quick acquisition of the reverse contingency for the 0:4 pair suggests that there was some transfer from the mastery of the 1:4 pair to the 0:4 pair. In the retention and transfer phase, the last phase of the study, subjects continued to perform at high levels on the familiar pairs despite the fact that 7–10 months had passed since the end of the testing phase. Such good long-term retention has also been described in previous studies (Anderson et al. 2004; Genty et al. 2004). More importantly, mangabeys transferred their good performance to novel quantities with the same ratio of one of the original pairs (i.e., 2:8) and to a pair in which the quantity that had to be selected used to be the quantity to be avoided (i.e., 4:7). Since subjects showed this good performance from the Wrst session both on familiar and novel pairs, we can rule out rapid re-learning (Anderson et al. 2004). Thus, our results suggest that subjects had acquired a rule that went beyond learning to respond in certain ways to particular pairs. Other similar studies have shown that squirrel monkeys, lemurs, and cotton-top tamarins also transfer to novel array pairs (Anderson et al. 2000, 2004; Genty et al. 2004; Kralik 2005). In the case of squirrel monkeys, they showed transfer also with a novel pair (4:8) that included a smaller quantity that used to be the larger in previous phases (1:4). These transfer data are remarkable because subjects in those studies were initially trained with only one or two pairs. Other studies that have used multiple pairs in the learning phase have obtained poorer results (e.g., Boysen and Berntson 1995; Boysen et al. 1996), although this could be due to other procedural factors as noted below. Although Shumaker et al. (2001) found that orangutans learned the reverse contingency rule with multiple pairs during training, these results are diYcult to interpret because subjects did not show an initial preference
for the larger quantity. Future studies should systematically investigate how the number of training pairs aVects both initial acquisition and transfer to novel pairs. When interpreting these results, one could argue that subjects were not indicating one of the quantities to net the alternative but that they had learned to perform an action on a certain location, far from the desired quantity, to obtain the larger quantity. Thus, they did not point to the smaller to receive the larger but they produced a behavior that produced the larger quantity. However, this still would entail refraining from directing behavior to the larger quantity as they had displayed in the initial phase. Moreover, the diYculties in mastering the 0:4 pair in Test 1 while showing improvement with the 1:4 pair indicate that mangabeys were not pointing at a certain panel hole without considering the food quantity of the corresponding dish. Nonetheless, future studies should consider Boysen and Berntson (1995) experimental procedure for the reverse contingency task, in which the two food quantities were presented in randomly varying positions among four holes. This would ensure subjects selecting food amounts and not following any other rule involving position to choose the smaller quantity. Murray et al. (2005) suggested a hypothesis to explain the underlying mechanisms for the reverse contingency task in which the values of the food amounts as stimuli change due to the values of the obtained rewards. The process would have three steps. On a reverse contingency with the 1:4 pair, in the Wrst step (step 1) 4 would have more value than 1 as a stimulus, though reaching for the larger quantity would be a prepotent response. But given that selecting four results in a lower reward than selecting one, its value would decrease at the same time that the value of one would increase. That would result in arriving to equilibrium (step 2), with results at chance level, due to the diYculty of assigning values to the diVerent food amounts. Persistence in detection of errors would lead to a change of problemsolving strategy that might permit mastering the task (step 3). An example of a new strategy would be considering the whole array pair as a single stimulus and solving the problem by learning conditional associations between each stimulus, the potential responses and their outcomes (Murray et al. 2005). Murray et al. (2005) note that rhesus monkeys are adept at solving such conditional associations. Our results generally support the three steps identiWed by Murray et al. in learning to select one raisin to obtain four raisins. First, subjects selected the smaller quantity below chance, then they progressed to chance performance, and Wnally they selected it above chance
Anim Cogn (2007) 10:387–396
levels. It is not clear why there was a diVerence in acquisition between the 1:4 and 0:4 pairs. If the quantity 4 is gradually loosing strength when paired with the quantity 1, it is unclear why that does not also happen in the 0:4 pair, as both quantities appeared in the same session. Note that there is a sharp discrepancy in performance between the two pairs in the Wrst phase of testing. As stated above, a possible explanation could be that at the beginning of the experiment our subjects might have not seen the situation as a choice, given that they were faced with only one quantity of food. Only after a training phase with the 1:4 pair would they have extrapolated the rule. A discrepancy with the Murray et al. hypothesis is that in the second step mangabeys did not perform randomly in the 1:4 pair but showed a side bias, which has been also observed in previous studies (Anderson et al. 2000; Shumaker et al. 2001; Kralik et al. 2002; Genty et al. 2004). This strategy of mainly selecting the left or the right dish resulted in a 50% of correct responses, which was a better result than previous ones but would explain chance responding. Thus, our results suggest that mangabeys passed from step 1 to step 3 of the Murray et al.’s hypothesis (at least no step 2 with random performance was detected), but followed the side-bias strategy (which gives similar results than random performance) before Wnding the one that permitted them mastering the task. Currently, the literature shows marked diVerences in the ability of various primate species to solve reverse contingency tasks. Shumaker et al. (2001) speculated that species’ social systems might explain these diVerences in performance, and suggested that orangutans had a “distinct level of mental Xexibility” greater than chimpanzees due to their less solitary nature. However, subsequent studies have not found diVerences between chimpanzees and orangutans (Vlamings et al. 2006). Therefore, it is unclear to what extent the diVerences between studies reXect inter-speciWc diVerences in inhibitory skills or can be explained in other ways. It is conceivable that at least some of these diVerences might be due to diVerences in the experimental procedures more than to interspeciWc diVerences (Silberberg and Fujita 1996; Kralik et al. 2002; Genty et al. 2004; Kralik 2005; Murray et al. 2005). For instance, in previous studies the average number of trials needed by successful orangutans (Shumaker et al. 2001), apes (Vlamings et al. 2006), mangabeys and rhesus (Murray et al. 2005) was 210, 323, 485 and 1,087 respectively, whereas in other studies the original reverse contingency task was substituted for modiWed versions of it after only 200 or 300 trials (Silberberg and Fujita 1996; Anderson et al. 2000;
Anim Cogn (2007) 10:387–396
Kralik et al. 2002; Genty et al. 2004). It is conceivable that if members of these species had been tested for longer periods, some could have performed well in the original reverse contingency task. Future studies should take into account the number of sessions needed by the subjects that have previously mastered the task in the design of new testing schedules. Small sample sizes paired with strong individual variation may also be an alternative explanation for the observed inter-speciWc diVerences. Previous studies had detected large individual diVerences in performance in the reverse contingency task (Anderson et al. 2000; Genty et al. 2004; Kralik et al. 2002; Silberberg and Fujita 1996; Vlamings et al. 2006). The current study conWrmed this result. There were large individual diVerences in the speed of acquisition, although the learning curves were similar for all mangabeys. However, the speciWc factors underlying such diVerences remain unclear. Genty et al. (2004) pointed out that age may contribute to the individual diVerences in inhibitory control as their best-performing lemur was the youngest. Unlike Boysen et al. (1999), our results Wt well with this idea because speed of acquisition was directly related to age such that younger individuals performed better than older ones. Genty et al. (2004) also argued that rank might inXuence self-control, as low-ranking individuals are more likely to express self-inhibition during food or reproductive competition. However, this hypothesis awaits empirical veriWcation. Other variables such as personality (Shumaker et al. 2001), previous experience (Shumaker et al. 2001; Hauser et al. 2002; Kralik et al. 2002; Vlamings et al. 2006), or innovation capability may also aVect performance. In this vein, Kralik (Kralik et al. 2002; Kralik 2005) noted that failing to solve the reverse contingency task may not necessarily reXect inhibition problems but it may indicate an inability to generate alternative responses. In other words, whereas producing an innovative response may indicate the inhibition of a prepotent response, the lack of innovation may not necessarily imply poor inhibitory control. In conclusion, this study demonstrated the ability of mangabeys to inhibit their prepotent tendency to select the larger of two quantities, retain it for a long period, and generalize it to novel pairs of quantities. Such performance was obtained without modiWed procedures in subjects without any prior experimental history on cognitive tasks. Future studies on several species with larger samples will be needed to test the predictive power of each of the hypotheses presented to account for the diVerences between studies.
395 Acknowledgments This study was supported by a granting-aid for ScientiWc Research (PRUCH 04/27) from the Universidad Cardenal Herrera (Valencia, Spain) and a grant from the Spanish Science and Education Ministry as part of the Spanish–German integrated actions program (Reference: HA2005-0010). We thank Valencia Zoo, and especially the zookeepers, for their collaboration in the study and Lluís Ros-Martí for his help with the construction of the test apparatus and the design of Fig. 1. Finally, we would like to thank four anonymous reviewers for providing their insightful comments on an earlier version of this manuscript. The animals used in this research were treated in accordance with Spanish law (Real Decreto 1201/2005).
References Anderson JR, Awazu S, Fujita K (2000) Can squirrel monkeys (Saimiri sciureus) learn self-control? A study using food array selection tests and reversed reward contingency. J Exp Psychol Anim Behav Process 26:87–97 Anderson JR, Awazu S, Fujita K (2004) Squirrel monkeys (Saimiri sciureus) choose smaller food arrays: long-term retention, choice with nonpreferred food, and transposition. J Comp Psychol 118:58–64 Boysen ST, Berntson GG (1995) Responses to quantity: perceptual versus cognitive mechanisms in chimpanzees (Pan troglodytes). J Exp Psychol Anim Behav Process 21:82–86 Boysen ST, Berntson GG, Hannan MB, Cacioppo JT (1996) Quantity-based interference and symbolic representations in chimpanzees (Pan troglodytes). J Exp Psychol Anim Behav Process 22:76–86 Boysen ST, Mukobi KL, Berntson GG (1999) Overcoming response bias using symbolic representations of number by chimpanzees (Pan troglodytes). Anim Learn Behav 27:229– 235 Boysen ST, Berntson GG, Mukobi KL (2001) Size matters: impact of size and quantity on array choice by chimpanzees (Pan troglodytes). J Comp Psychol 115:106–110 Genty E, Palmier C, Roeder JJ (2004) Learning to suppress responses to the larger of two rewards in two species of lemurs, Eulemur fulvus and E. macaco. Anim Behav 67:925–932 Gómez JC (1990) The emergence of intentional communication as a problem solving strategy in the gorilla. In: Parker ST, Gibson KR (eds) “Language” and intelligence in monkeys and apes: comparative developmental perspectives. Cambridge University Press, Cambridge, pp 333–355 Guillaume P, Meyerson I (1930) Recherches sur l’usage de l’instrument chez les singes. I. Le probleme du detour. J Psychol Normale Pathol 27:177–236 Hauser MD, Santos LR, Spaepen GM, Pearson HE (2002) Problem solving, inhibition and domain-speciWc experience: experiments on cottontop tamarins, Saguinus oedipus. Anim Behav 64:387–396 Köhler W (1925) The mentality of apes. Vintage Books, New York Kralik JD (2005) Inhibitory control and response selection in problem solving: how cotton-top tamarins (Saguinus oedipus) overcome a bias for selecting the larger quantity of food. J Comp Psychol 119:78–89 Kralik JD, Hauser MD, Zimlicki R (2002) The relationship between problem solving and inhibitory control: cotton-top tamarin (Saguinus oedipus) performance on a reversed contingency task. J Comp Psychol 116:39–50
396 Murray EA, Kralik JD, Wise SP (2005) Learning to inhibit prepotent responses: successful performance by rhesus macaques, Macaca mulatta, on the reversed-contingency task. Anim Behav 69:991–998 Shumaker RW, Palkovitch AM, Beck BB, Guagnano GA, Morowitz H (2001) Spontaneous use of magnitude discrimination and ordination by the orangutans (Pongo pygmaeus). J Comp Psychol 15:385–391
Anim Cogn (2007) 10:387–396 Silberberg A, Fujita K (1996) Pointing at smaller food amounts in an analogue of Boysen and Berntson’s (1995) procedure. J Exp Anal Behav 66:143–147 Vlamings PHJM, Uher J, Call J (2006) How the Great apes (Pan troglodytes, Pongo pygmaeus, Pan paniscus, and Gorilla gorilla) perform on the reversed contingency task: the eVects of food quantity and food visibility. J Exp Psychol Anim Behav Process 32(1):60–70