Edited by: David A. Washburn, Georgia State University, USA
Reviewed by: Gisela-Kaplan, University of New England and University of Queensland, Australia; Jorg J. M. Massen, University of Vienna, Austria
*Correspondence: Mathias Osvath, Department of Cognitive Science, Lund University, Lundagård Kungshuset, SE-222 22 Lund, Sweden e-mail:
This article was submitted to Comparative Psychology, a section of the journal Frontiers in Psychology.
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
The topic of cognitive foresight in non-human animals has received considerable attention in the last decade. The main questions concern whether the animals can prepare for upcoming situations which are, to various degrees, contextually or sensorially detached from the situation in which the preparations are made. Studies on great apes have focused on tool-related tasks, e.g., the ability to select a tool which is functional only in the future. Dufour and Sterck (
The last decade has seen a number of studies on episodic-like memory and foresight in animals, primarily on corvids and great apes (e.g., Clayton and Dickinson,
Regardless of whether there is an episodic component associated with the future directed behaviors of the animals in question, it remains important to study such behaviors in detail. Future directed behaviors exhibited in such studies appear difficult to explain by merely associative learning of key stimuli, or by rigid mechanisms, such as fixed action patterns (e.g., Raby and Clayton,
Current methods in the research on animal cognitive foresight are influenced by views forwarded by Wolfgang Köhler in the 1920s. Köhler studied the cognition of chimpanzees, and described his observations of chimpanzees anticipating events that were “planned acts of the animal itself” (Köhler,
Today, there exist only a few studies on the abilities of great apes to act toward future goals where the goals are outside the animal's current sensory scope. (Mulcahy and Call,
The two studies solely based on tool use showed that chimpanzees, bonobos and orangutans are capable of selecting an appropriate tool well in advance of its use. Bonobos and orangutans could keep the tool overnight (14 h per trial) (Mulcahy and Call,
The study by Dufour and Sterck (
The first reason is that the results of Dufour and Sterck (
Secondly, the results might indicate that the task of deferred exchange represents a domain where future-directed cognition in chimpanzees is restricted. The authors of the study speculate that it might be a result of the cooperative nature of the task (“I give you what you want, and you give me what I want”), which is a context that in general has been suggested to be more cognitively demanding for chimpanzees than competitive contexts (e.g., Hare and Tomasello,
It is also important to mention that a recent study on brown capuchins (
Exchange tasks as such, when not deferred, usually pose few problems for great apes. The exchange of items with a human for food rewards typically develops spontaneously in chimpanzees (e.g., Hyatt and Hopkins,
In order to discern whether the ability for deferred exchange with humans is outside the cognitive scope of great apes, the current study aimed to replicate the main experiment of Dufour and Sterck (
Condition (1), used in Experiment 1, followed the item selection procedure used in a study on planning for future tool-use by Osvath and Osvath (
Condition (2), used in Experiment 2, was similar to the selection procedure used in Dufour and Sterck (
This division of conditions was used because of the possibility that the different procedures might influence the results. That is, if the apes would succeed in condition (1) but not in condition (2), then the negative results in Dufour and Sterck (
Before training was undertaken, subjects were tested in a selection procedure for their potential spontaneous preferences for the different items. This was done to make sure that the subject actually learned the value of the exchangeable item in the training phase. If the subject would have a spontaneous preference for the item designated for exchange, then it could superficially pass the learning criterion (see below). That is, the reason for the selection of the correct item could be the result of a spontaneous preference and not a response to training.
All procedures were performed in compliance with relevant laws and institutional guidelines. Participation was voluntary and testing was approved by Uppsala regional ethics committee (approval no. C356/9). The Swedish Agricultural board (No. 31-2599/09) has approved Furuvik Zoo as a cognitive research facility on chimpanzees and orangutans.
Two chimpanzees (
Subjects were tested at Lund University Primate Research Station in Furuvik Zoo. At the time of testing, the participants had experienced few previous experimental tests, and none requiring object exchange. One chimpanzee, Maria-Magdalena, and one orangutan, Naong, had previously taken part in a planning experiment involving selection of items from a tray (Osvath and Osvath,
The individuals were tested in their caretaking compartments, as well as in larger indoor areas. No public visitors were present at the time of testing. No changes in feeding procedures were made and access to water was continuous. Some changes in indoor housing routines were made to minimize disturbance from group members.
Four different objects were used in the selection procedures. A piece of blue plastic rope, a piece of jute cloth, a wooden rod, and a strip of bent metal (see Figure
Each subject received 15 presentations of the tray baited with the four objects. The tray was slid toward the subject and was then retracted as soon as a selection had been made, thereby allowing the subject to choose only one item. A choice was scored when the animal touched or grabbed one of the objects. If the tray was retracted before the subject had managed to grasp the object, the one touched by the subject was handed over. The criterion for a “refusal” to select was when the subject entered the selection situation, looked at the items but did not attempt to touch any of them before leaving.
All subjects but the orangutan female had previous experience of this type of procedure. The order of the items was pseudo-randomized between trials. The experimenter handling the tray did not gaze at the items but slightly above or at the face of the ape.
Naong selected the piece of rope 3 times, the wooden rod 4 times, the jute cloth 1 time, and the piece of metal 3 times; in 4 trials he refused to select. The other orangutan, Dunja, selected the wood 2 times and refused to select in the rest of the trials. Manda selected the wood 12 times, the rope 2 times, and the jute 1 time. Maria-Magdalena selected the wood 12 times, the rope 2 times and refused to select once. No spontaneous preference thus existed for the metal strip at the onset of training for any of the subjects.
Subjects and materials were identical to the preference testing described above. The four subjects were trained until able to reliably exchange the target item (the metal strip) in a direct setting with no time delay. On the initial trials only the target item was placed in the enclosure and experimenter pointed to this and requested the “grunka” (Swedish for “the thingamajig”). Later all items were placed simultaneously into the enclosure and, pointing if needed, the experimenter asked for the “grunka.” On successful exchanges the subject was rewarded with verbal praise and a food item consisting of approximately a fifth of a banana. If the subject handed the wrong item back, the experimenter again pointed to the target item and asked for the “grunka.” In later learning trials pointing was phased out when the subject collected the target item without coaxing.
Successful learning was corroborated in tests of learning, in which 4 out of 5 trials had to be correct, which was followed by a second test (a test of retention) also requiring 4 out of 5 successful exchanges. Tests of learning and first test of retention were given on two consecutive days (except for Dunja who, due to practical reasons related to housing, received her first test of retention later the same day as she met the learning criteria).
Additional retention tests, which functioned as warm ups, were given for each new day of participation in the experimental conditions. These did not always amount to a full 5 trials before testing started, depending on how well the subject performed and/or on how motivated the animal appeared on the particular day.
In the tests of learning and retention the subject was not cued by pointing or gaze toward a particular choice object, instead the experimenter looked directly at the ape or at a point beyond it. The items were presented in a different area than the exchange. Tests of learning and tests of retention, all took place by presenting the items onto the cage floor, i.e., they were not given on a selection tray. The exception was for the female orangutan subject (Dunja) who had little previous experience of choosing from a tray, and appeared to be wary toward the tray. Thus, she received additional training trials using the selection tray. By using a test of learning and a retention test with a set criterion (80% success), we could make sure that all the subjects could perform the task in an immediate setting.
All subjects swiftly learned to hand out the target item at the expense of the other items in exchange for food (and verbal praise). The average number of trials required before we decided to test their learning against the criteria was 8.5 ± 3.9 (5–14 trials). All subjects met, and exceeded, the criteria in the subsequent test of learning: 5 out of 5 trials. All subjects also met, and exceeded, the criteria in their first test of retention: 5 out of 5 trials.
One chimpanzee, Manda, and the two orangutans participated in this experiment. The other chimpanzee refused to take part in the experiment. The chimpanzee and one orangutan, Naong, received this experiment as their first test in the study, while the other orangutan had already been tested in Experiment 2. The chimpanzee that refused Experiment 1 had previously completed Experiment 2. Manda was tested in this experiment at the age of 5. Dunja was tested at the age of 21. Naong received his first 4 trials at the age of 20, and the other 9 trials at the age of 22. This split in time was due to circumstances unrelated to the study. We judged it to be more conservative to resume with 9 trials instead of completing a new set of 13 trials at this later occasion, as the first 4 trials could be seen as unwarranted training on the deferred task.
Naong did not receive additional training after 2 years before the 9 trials, only a test of retention, which he passed (5 out of 5). Continued good performance after a 2-year hiatus is arguably a testament to a profound ability to solve the task.
The selectable items in this experiment were identical to the ones used in the preference testing and training. The metal strip was the target item used in the deferred exchange for food. The food reward was approximately a fifth of a normal sized banana.
The procedure was similar to the one described in Preference testing. However, the experimenter (E1) operating the tray during the selection phase differed from the experimenter (E2) who performed the deferred exchange. Both experimenters were known to the subjects, but E1 had not been involved in the prior exchange training. Furthermore, the deferred exchange was made at a different location, out of sight, from the location of the selection. These additions were made to preclude sensory feedback from elements pertaining to the goal situation, other than the exchangeable item and the general presence of a human. The delay between the selection and the exchange was 15 min. The time of delay was chosen for practical reasons. The exact length of the delay plays a minor role in these types of tests as long as the selection and the exchange is parted in space and by a time span exceeding the period of storage in the working or short term memory (compare to e.g., performances of episodically amnesic patients, e.g., Tulving,
One chimpanzee, Manda, selected the correct object in 9 out of 13 trials (exact binomial test,
The male orangutan, Naong, selected the correct object in 12 out of 13 trials (exact binomial test,
The female orangutan, Dunja, selected the correct item in 11 trials out of 14 (exact binomial test,
Manda ( |
9 |
4 | 9 |
0 | 12 |
8 | 6 |
0 |
Maria-Magdalena ( |
Did not participate | 7 n.s. | 11 | 7 |
0 | |||
Naong ( |
12 |
1 | 11 |
1 | 7 |
1 | 7 |
1 |
Dunja ( |
11 |
3 | 0 n.s. | 4 | Did not enter selection room | 0 | 2 |
In this experiment all four subjects took part. One chimpanzee, Manda, was 7 years old, and one of the orangutans, Naong, was 22 years old, when they participated in the experiment. One of the chimpanzees, Maria-Magdalena, at the age of 11, and one of the orangutans, Dunja, at the age of 21, received this experiment as their first experiment in the study. As 2 years had transpired since Manda received Experiment 1, she was given a new set of refreshment training consisting of 5 trials including only the target item. She handed it out in all trials, as she did in the subsequent retention test.
The selectable items and the food reward in this experiment were identical to the ones described above. In this condition no selection tray was used, as the items were placed on the floor in a room, similar to the condition described in Dufour and Sterck (
The four selectable items were placed on the floor of a closed compartment in the everyday enclosure. A hatch was then opened and the subject was given access to enter the compartment. The hatch was left open for a 15-min selection phase with no humans present. The one handling the hatch was different from the one who later performed the exchange. The closing was followed by a 15-min delay after which the experimenter asked the subject for the exchangeable item at a different location. The exchange procedure followed the one described in Experiment 1. The behavior of the subject in the compartment baited with the items was video recorded. The video was then analyzed for the order in which the items were touched and for what items were transported out of the room.
One chimpanzee, Manda, brought the correct item out of the selection compartment in 12 out of 12 trials. Additionally she brought the rope 4 times (trial number 2, 3, 4, and 10), the wood 3 times (trial number 1, 2, 7) and the jute 1 time (trial number 1). This means that she brought the exchangeable item alone, together with no additional objects, in 5 trials. Distractor objects were never taken without also additional objects, and importantly the target object. She exchanged the correct object in 6 trials.
As the subject could select more than one item in this set-up, a standard binomial could not be used. Instead, we used Monte Carlo simulation (Robert and Casella,
The other chimpanzee, Maria-Magdalena, brought the correct object out of the selection room in 7 out of 12 trials (trial number 1, 3, 4, 5, 6, 7, and 9). Additionally she brought the rope 5 times (trial number 1, 8, 10, 11, and 12), the wood 4 times (trial number 1, 2, 8, and 10) and the jute 2 times (trial number 1 and 2). This means that she brought the exchangeable item alone, with no additional objects, in 6 trials, and that the rope was the only other object that was ever taken with no additional object. This was done on 2 occasions. According to the same method as described above, the number of selections of the exchangeable item was not significantly above chance (
One orangutan, Naong, only entered the compartment in 8 trials (trial numbers 1, 2, 3, 8, 9, 10, 11, and 12). In 7 of these trials he brought the exchangeable item. In one trial he brought the rope and the wood piece (see below). This is significantly above chance (
The other orangutan, Dunja, refused to enter the compartment in all trials, and refused to come to the location of exchange in all trials but 4. In 2 of these cases she tried to exchange, once with a piece of faeces and the other time with a sponge (enrichment material from the enclosure). To test whether Dunja actually understood that the compartment contained exchangeable items, three additional trials were run. In these trials she was coaxed into the room with food that was placed next to the choice items on the floor. In all cases she took the food and ignored any other items. See Table
This study did not attempt to directly study future planning, understood as an action taken in the current situation with the intention to reach a future goal, unrelated to the current psychological or perceptual state. It did, however, investigate behaviors that in effect become future oriented, and as such might constitute elements in planning acts. The goal was to discern whether great apes indeed are incapable of using an arbitrary but reinforced object to make exchanges for food with humans after a considerable delay, as has been suggested by the results in Dufour and Sterck (
This study, however, cannot in itself unambiguously distinguish what strategies the apes used to solve the tasks. There is a possibility that the selection of the item and the later use of it for exchange were, in a sense, cognitively separated. Given the amount of training, the selection of the item could in this case be the result of its status as a heavily reinforced stimulus. The exchange, in turn, could be a result of a combination of the reinforced action of handing out the target item to a human, and a memory for whether the item had been selected and were present in the enclosure or not. This would hold true also for the significant results that seemingly demanded coupling of correct selection with correct exchange [see e.g., (Dickinson,
Regardless of whether one assumes merely associative learning, episodic planning, their combination, or something else, when explaining the results, they must still be contrasted to the negative ones of Dufour and Sterck (
It is obvious that learning plays a pivotal role when it comes to couple objects with valuable exchangeability, and the situations in which such exchanges can be made. One could speculate that, apart from learning the value of the key stimuli, it might be necessary to learn more abstract relationships. For example that humans are reliable exchange partners over time. To be able to perform deferred exchange the ape might need more experiences, in several contexts, of humans as stable bartering partners. Acting future orientedly toward static objects in the environment, e.g., a nut tree or a reward apparatus, even if they are out of sight, could be a simpler task as such things are less changing in comparison to the dynamic behaviors of others.
No systematic differences between the results of Experiment 1 and 2 could be found. As the conditions differ in setup and require different statistics, direct comparisons are not possible. A comparison would only be useful if, across subjects, one condition would yield significant results and the other would not. Even then, given the few number of subjects of this study, it would still only be an indication. Nevertheless, it was shown that apes could successfully manage both conditions, and, despite the small number of subjects, the study thus answers the question of whether deferred exchange is possible for great apes in the affirmative.
Two subjects in this study, Maria-Magdalena and Naong, also participated in a study on planning for tool use (Osvath and Osvath,
When it comes to the poor performance of the female orangutan, Dunja, one can only speculate. There could be several reasons for her results. She did indeed select the correct item in Experiment 1 significantly above chance, but did not get the chance to show whether she would exchange it or not in the future, as she after selection always returned to the male who took the item. One could of course reason that such behavior would be a sign of poor maximation of future rewards, but it could also be that the company of the male was more rewarding than a fifth of a banana. Interestingly, she did not ever attempt to take the item in Experiment 2, not even when coaxed into the room containing the objects. It can be that the mere presence of a human—a close up and potentially interactive partner—was the relevant factor for success, but we cannot know. There is one conspicuous difference in background between Dunja and the rest of the subjects. Dunja arrived at the zoo only 1 year before the study, while the rest of the subjects had been at the zoo for many years. At Furuvik Zoo the apes routinely exchange objects for food with the caretakers. This is a skill the zoo upholds in order to make the apes pick up objects from inside the compound that should not be there. When it comes to Dunja the background is less clear. Her previous caretakers reported that she did spontaneously and frequently engage in exchanging objects (mainly faeces) for food rewards at her former zoo from a young age. Without targeted studies, however, we do not know if there are systematic differences in the exchange behaviors between the two zoo environments. There might be a possibility that a difference in long-term experiences could explain the results. Relating to this issue, it would be interesting to know the background experiences of the chimpanzees in the study of Dufour and Sterck (
In any case, this study shows that it is within the capacity of apes to succeed in deferred exchange. More studies are needed to understand more precisely what type of future oriented tasks apes can solve and why, including the possible effect of individual background experience.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
The authors wish to acknowledge Helena Osvath and Miriam Sima and the orangutan and chimpanzee caretakers at Furuvik Zoo—Linda-Marie Hjelm, Elin Karvonen, Torsten Lönn, Natalie Magnusson, Ing-Marie Persson, and Janet Thelberg—for assistance with testing. Richard Andersson, Elainie Madsen, Rasmus Bååth, and Christian Balkenius contributed with helpful analysis and comments on the manuscript. Gerd Grätz and Susan Sayehli gave us valuable information. The study was funded by The Swedish Research Council (grant no. 2009–2332) and was made possible by additional donations from The Crafoord Foundation, The Krapperup Foundation, The Magn Bergvall Foundation, Elizabeth Rausing's Memorial Fund, and The Foundation Landshövding Per Westling's Memorial Fund.