Cooperative problem solving in African grey parrots (Psittacus erithacus)
- First Online:
- Cite this article as:
- Péron, F., Rat-Fischer, L., Lalot, M. et al. Anim Cogn (2011) 14: 545. doi:10.1007/s10071-011-0389-2
- 2.2k Views
One of the main characteristics of human societies is the extensive degree of cooperation among individuals. Cooperation is an elaborate phenomenon, also found in non-human primates during laboratory studies and field observations of animal hunting behaviour, among other things. Some authors suggest that the pressures assumed to have favoured the emergence of social intelligence in primates are similar to those that may have permitted the emergence of complex cognitive abilities in some bird species such as corvids and psittacids. In the wild, parrots show cooperative behaviours such as bi-parental care and mobbing. In this study, we tested cooperative problem solving in African grey parrots (Psittacus erithacus). Our birds were tested using several experimental setups to explore the different levels of behavioural organisation between participants, differing in temporal and spatial complexity. In our experiments, African grey parrots were able to act simultaneously but mostly failed during the delay task, maybe because of a lack of inhibitory motor response. Confronted with the possibility to adapt their behaviour to the presence or absence of a partner, they showed that they were able to coordinate their actions. They also collaborated, acting complementarily in order to solve tasks, but they were not able to place themselves in the partner’s role.
KeywordsAfrican grey parrotsCooperationSynchronyCoordinationCollaborationSocial cognition
Cooperation is a widespread phenomenon occurring between individuals of the same (and even different) species, even if human cooperation exceeds that of all other species with regard to the scale and range of cooperative behaviours (for a review see; Bergmüller et al. 2007; Melis and Semmann 2010). In non-human animals, cooperative behaviours such as cooperative hunting, group defence and cooperative breeding. are observed. However, the complexity of the cognitive mechanisms involved in these cooperation events is often difficult to measure, as the relative contributions of genetic predetermination, learned behavioural strategy or higher cognition are not known (Brosnan and Bshary 2010). Laboratory studies and cross-species comparisons using similar paradigms are useful in studying the proximal mechanisms that can be shared by different species and to discover convergent evolutionary processes. As many definitions of cooperation can be found, in this paper, we use Noë’s (2006) definition, i.e. all interactions or series of interactions that, as a rule (or ‘on average’), result in net gain for all participants.
Studies conducted in psittacids have highlighted cognitive abilities as complex as those observed in primates (Pepperberg 2006) and similarities in their neurobiology (e.g. ratio of brain/body size; Iwaniuk et al. 2005) and socio-ecology (e.g. fission–fusion dynamics, long lifetime, long juvenile period, etc.). The social brain hypothesis suggests that individuals living in social groups with complex interactions have bigger brains in order to manage social relationships (Joly 1966; Humphrey 1976; Byrne and Whiten 1988; Dunbar 1998). The ‘relationship intelligence hypothesis’ (Emery et al. 2007) predicts that complex social life and long-term monogamous partnership lead to elaborated socio-cognitive abilities, similar to those encountered in primates. Indeed, it seems that behavioural coordination and synchrony necessary to maintain stable pair-bonded relationships represent cognitive demands that could explain the relative size of the brain in monogamous species (Shultz and Dunbar 2007; Dunbar and Shultz 2007). Shultz and Dunbar (2010) found that relative brain size in birds is strongly related to bi-parental care, pairbonding, stable social relationships and altricial development. Thus, psittacids species could represent a good model for social cognition studies. African grey parrots (Psittacus erithacus) form stable monogamous couples over breeding seasons and both parents take care of the chicks, although females invest more effort (Cockburn 2006). These birds show high levels of social tolerance (May 2004) and exchange affinitive behaviours such as preening and regurgitation (Skeate 1984, Harrison 1994). Mobbing events have been reported against red kites (Milvus milvus) on Principe Island (Jones and Tye 2006). All these elements taken together support the hypothesis that African grey parrots could have the cognitive abilities necessary to manage complex social relations and to cooperate.
The experiments conducted here do not have direct ecological validity, since grey parrots do not cooperate to obtain food in the wild (as they mainly eat fruits, nuts and seeds). Nevertheless, they display cooperative actions in the wild such as bi-parental care, mobbing, and display altruistic behaviours (allopreening and regurgitation). They are highly social and tolerant; they develop strong pairbonds and are able to learn contingencies. All these characteristics represent favourable conditions for cooperation skills to develop (de Waal and Suchak 2010).
In our study, we used the ‘loose string paradigm’ used by Hirata and Fuwa (2007) with chimpanzees (Pan troglodytes) in which two individuals were faced with a tray with food rewards placed out of reach. Both subjects had to pull a string simultaneously in order to get access to the reward. Several studies have highlighted the fact that tolerance has a direct impact on the outcome, as the more tolerant a subject is, the better that individual performs (capuchins: Mendres and de Waal 2000; chimpanzees: Melis et al. 2006b; bonobos: Hare et al. 2007; rooks: Seed et al. 2008). A recent study has also highlighted that rooks’ temperament influences the success of the dyad insofar as bolder individuals appeared to be more willing to solve the task, whereas shyer birds dared to approach the apparatus only with a partner already present (Scheid and Noë 2010).
The aim of this experimental series was to test our parrots in tasks that required different levels of spatial and temporal complexity. Firstly, we evaluated if the birds were able to perform identical actions (Goal test). Studies on capuchin monkeys revealed that they were able to solve such a cooperative task, but without understanding their partners’ role (Chalmeau et al.1997; Visalberghi et al.2000). In a second experiment, we tested their ability to relate their actions in time (‘Delayed partner arrival’ test). As emphasised by Noë (2006), simultaneity is an important proximate mechanism needed to accomplish cooperation. In a study conducted with brown capuchins, an increase in gazing at a partner was interpreted as a sign of an understanding of the partner’s role as the individuals visually monitored their partners and adjusted pulling behaviour according to their partner’s presence (Mendres and de Waal 2000). Tested in a ‘delayed partner arrival’ task with a ‘loose string’ paradigm, chimpanzees were able to wait before starting to pull (Melis et al.2006b). Additionally, when the chimpanzees tested could decide to allow a partner to enter, they did this significantly more often when they were facing an apparatus needing cooperation than when facing an experimental setup that they could solve alone. Seed et al. (2008) also conducted a ‘delayed partner arrival’ task with rooks but, contrary to chimpanzees, rooks were not able to wait for a partner. We tested whether the parrots can assess their partners’ role in a cooperative task (Apparatus choice test) and thus used the same protocol with our parrots as Seed et al. (2008) used with rooks. In this experiment, the individual faced two different options, one apparatus that could yield a reward when handled alone (‘Solo’) and another baited with twice as much food (per bird) but which required cooperation to obtain it (‘Duo’). The subjects were tested in two separate situations: alone or with a partner. We expected that the parrots would choose the ‘Solo’ apparatus in the first situation, whilst trying the other apparatus when a partner was present in order to get more food. Rooks did not seem to understand the task as only two out of six individuals showed a tendency to prefer the single platform when tested alone, and no cooperation was observed at all during the experiment. Recent studies revealed that chimpanzees (Melis et al. 2009) and hyenas (Drea and Carter 2009) are able to coordinate their actions in order to solve a similar task. This duplication of the apparatus increased the challenge of the task, as successful actions required not only to take into account the presence or absence of a partner but also the choice of the apparatus. Thus, successful trials were less likely to occur by chance.
Finally, in order to test whether the birds could carry out different but complementary actions (Complementary actions test), we designed a more complex apparatus in which one individual had to climb on a perch in order to release the tray that was pulled by the second bird in order to test. In this experiment, both birds needed to handle the device, contrary to previous studies conducted with keas in which a seesaw apparatus was used on which only one individual had to work (Tebbich et al. 1996; Huber et al. 2008).
Proximate mechanisms contributing to cooperation include both non-cognitive (emotion and/or personality) and cognitive factors (individual recognition, problem solving, etc.). In the wild, animals cooperate but the degree of cognitive processes involved in the achievement of the task is unknown. Indeed, the underlying cognitive abilities necessary to cooperate are not clear as most of the cooperation per se does not require advanced cognition (cooperation being ubiquitous in nature, existing from single-cell organism to plants, invertebrates and humans; see Brosnan et al. 2010 for a review). Nevertheless, cognition may help to make coordination more efficient and also help decision-making concerning the best behavioural option in a given situation (Brosnan et al. 2010). The primary cognitive aspect involved in a cooperative action is the comprehension of the partner’s role (Noë 2006), which can explain the difference in performances observed during non-synchronised actions and actions performed in unison, as in the second situation individuals have to monitor their partner’s actions. Collaboration could be assessed if individuals are able to exchange their roles and thus show evidence of partner’s role comprehension. Such ‘true collaboration’ has been observed in hunting species (see Brosnan et al. 2010 for a review).
We also expected that the quality of relationships would influence the cooperative attempts through avoidance/affiliation or dominance/subordination behaviours (Noë 2006). Indeed, underlying mechanisms such as emotional factors (fear, frustration, etc. See Brosnan and de Waal 2002), social preferences (long-term social bonds; Silk 2002; Brosnan et al. 2010), temperament (Bergmüller et al. 2010; Scheid and Noë 2010) or cognitive processes (problem solving, partner discrimination, etc., see Shettleworth 2009, 2010) could impact on the performances.
Materials and Methods
We tested three hand-reared African grey parrots, two males (Shango and Léo, 4 and 6 years old, respectively) and one female (Zoé, 6 years old). They hatched in captivity and arrived at the laboratory at three months of age. They were housed together in an aviary (340 cm × 330 cm × 300 cm) with three tables (120 × 60 × 75 cm), two large perch structures and many toys, at a constant temperature of 25°C and with a 12/12 h light–dark cycle. The parrots were tested in their aviary. During a test session, subjects that were not tested were placed in a cage in the corridor with water, food and toys available. Parrots were fed daily with fresh fruits and vegetables in the morning and parrot formula (Nutribird A21) in the evening. Water and parrot pellets (Harrison’s high potency coarse) were available ad libitum, and vitamins (Muta-Vit Versele-Laga) were given twice a week. Léo was dominant over Shango, Shango over Zoé, but no dominance was observed between Léo and Zoé.
A flat rectangular cardboard tray (31 × 17 × 4 cm) baited with food (parrot formula and seeds) was placed in a cage (54 × 28 × 36 cm), impeding direct food access. A piece of string was threaded through metal loops placed on the tray so that both ends of the string extended out of the cage by 20 cm. The lower part of the cage had a gap that enabled the birds to pull the tray out of the cage. Birds could move the tray only by pulling both ends of the string simultaneously (‘loose string’ paradigm). The cage was placed on two tables separated by 15 cm from each other. The string was not attached and each end was on a separate table. Two experimenters conducted all the test sessions.
Habituation phase and training
African grey parrots are neophobic birds, so they were previously familiarised individually with each new apparatus or new element of it for 1 week before each experiment. Birds were trained to pull the string first with both ends of the string attached so that an individual alone could succeed. Then with free ends, and in order to maintain motivation even when one of the two birds was gone, the experimenter pulled the string with the remaining bird. We stopped the training when all birds were able to stay in front of the cage and pulled the string (rarely simultaneously at this point). We completed two sessions (with a variable number of trials) per day, lasting about 30 min for each session over 2 weeks.
Experiment 1: goal test
Experiment 2: ‘delayed partner arrival’ test
Experiment 3: apparatus choice test
Experiment 4: complementary actions test
Scoring and data analysis
We report the number of single actions (string-pulling) before and after the arrival of the partner, simultaneous actions, the latency times before each parrot pulled the string for the first time, the time spent on the perch (experiment 4) and finally the outcome (access to food or not) for each trial. A trial ended when (1) at least one individual reached the reward, (2) the string was out of reach for the partner or (3) at least one or the other individual lacked motivation (more than 90 s without any behaviour directed towards the apparatus). For the first experiment, we determined the success rate for each session of 20 trials as the number of successful cooperative events over the total number of attempts.
Spearman correlations were used to assess any within-individual changes in behaviour across the testing period. We ran binomial tests in order to evaluate the choices of the birds facing different experimental setups. The significance level was set at α = 0.05. When multiple comparisons were made, we used a Bonferroni adjustment (α’ = α/c where α = 0.05 and c corresponds to the number of comparisons).
The birds required between 6 and 9 sessions to solve the task with a partner with a success rate higher than 90%. The number of solitary actions decreased (Spearman rank order correlation; Zoé-Léo: N = 9, rs = −0.702, P = 0.0101; Zoé-Shango: N = 7, rs = −0.861, P = 0.00609; Léo-Shango: N = 6, rs = −0.853, P = 0.0333), and at the same time, simultaneous pulling increased. Indeed, the number of successful cooperative actions improved over time for each dyad (Spearman rank order correlation; Zoé-Léo: N = 9, rs = 0.685, P = 0.0186; Zoé-Shango: N = 7, rs = 0.991, P < 0.001; Léo-Shango: N = 6, rs = 0.853, P = 0.0333).
‘Delayed partner arrival’ test
Birds cooperated successfully in 76% of the trials. We noticed that all the subjects pulled more when a partner was available than when they were alone (Wilcoxon: Zoé: W = 610.5, N = 50, P < 0.001; Léo: W = 989.5, N = 50, P < 0.001; Shango: W = 367.5, N = 50, P < 0.001; see Fig. 3). Two out of three individuals showed no significant change in the number of times they grabbed the string with their beak before the partner’s arrival (Spearman: Zoé: rs = −0.196, N = 50, P = 0.230 and Léo: rs = 0.292, N = 50, P = 0.104), while Shango showed a significant decrease (rs = −0.498, N = 50, P < 0.05). The latency time (before the first pulling) did not increase significantly across the experiment for Léo and Zoé (Spearman: Léo: rs = −0.070, N = 50, P = 0.551. Zoé: rs = −0.201, N = 50, P = 0.137) but increased for Shango (rs = 0.339, N = 50, P = 0.017).
Apparatus choice test
Zoé and Léo pulled less when they were waiting for a partner (Wilcoxon; Zoé: W = 50, N = 50, P < 0.001; Léo: W = 91.5, N = 50, P < 0.005) compared to the situation where no partner was present. Moreover, they improved their performance between the two experiments. Indeed, the mean number of pulling occurrences in each condition (alone and with a partner) decreased across the study (Comparing ‘Delayed partner arrival’ test and ‘Apparatus choice test’ when alone; Wilcoxon: Zoé: W = 848.5, N = , P < 0.001; Léo: W = 1016, N = 50, P < 0.005. Comparing ‘Delay’ and ‘Solo Duo’ tasks when the partner was present; Wilcoxon: Zoé: W = 546.5, N = 50, P = 0.34; Léo: W = 985, N = 50, P < 0.01; see Fig. 5).
Complementary action test
Success rate: number of successful cooperative action out of the total number of attempts
No of trials
Success rate (%)
Failure perching bird (%)
Failure pulling bird (%)
No attempt (%)
Pulling before (%)
In experiment 1 (Goal test), our three birds had the same goal and solved the task doing the same action (pulling the string). The number of simultaneous pulling occurrences increased as their identical actions were completed at the same time. This type of cooperation has also been observed in primates such as cotton-top tamarins (Cronin et al. 2005), marmosets (Werdenich and Huber 2002), capuchins (Mendres and de Waal 2000), bonobos (Hare et al.2007) and chimpanzees (Melis et al. 2006a), and more recently in non-primate species such as rooks (Seed et al.2008), wolves (Möslinger et al.2009) and hyenas (Drea and Carter 2009). These results show that the subjects’ success rate increased, and this was enough to conclude that they could solve the task (goal similarity), but not decide whether or not they understood partner’s role. Indeed, individuals could have been simply attracted by the reward and succeed as a by-product of individuals’ independent but identical simultaneous actions (Visalberghi et al. 2000).
In experiment 2 (‘Delayed partner arrival’ test), one of our subjects (Shango) adjusted his behaviour according to the presence of a partner and delay the pulling of the string until the other subject arrived in front of the cage. On the contrary, Zoé and Léo, like rooks (Seed et al.2008; Scheid and Noë 2010), seemed unable to wait for the partner. The three parrots were able to take into account the presence or absence of a partner, since they all pulled more when a partner was present. However, this could be due to instrumental learning rather than to a real understanding of the task.
In experiment 3 (Apparatus choice test), our African grey parrots were able to coordinate their actions like chimpanzees (Melis et al.2009) and hyenas (Drea and Carter 2009) and to discriminate the two apparatus. One subject always chose the ‘Solo’ apparatus, whereas two of the three parrots preferred the ‘Duo’ apparatus when tested with a partner. In the study by Seed et al. (2008), only two rooks (out of six) showed a preference in the second half of the trials and this preference was for the single apparatus when they were alone. Nevertheless, we must specify that our birds were placed directly on the table and had fewer alternative options compared to the rooks that were in an adjacent aviary.
Zoé made a clear distinction between conspecifics, and the decision for entering in a cooperation process was dependant on who the partner was. While she chose to always cooperate with Léo when he was present, she always went for the ‘Solo’ apparatus when Shango was the partner. She probably understood the advantages of cooperation, but also realised when it was better not to do so. Indeed, Shango was dominant over Zoé and they often had antagonistic interactions. Furthermore, she could have kept in mind that in this experiment he never cooperated (when he was the tested bird). Clearly, tolerance and social preferences could explain the results as the birds could seek proximity with a specific partner.
Shango adopted a different behaviour, always choosing the ‘Solo’ apparatus whatever the situation was (tested alone or with a partner, dominant or subordinate). Whilst Léo showed a clear preference for the ‘Duo’ apparatus when tested with a partner, he behaved in a random fashion when alone. This could be because he did not really understand the task and the partner (when present) represented an attractive stimulus. In previous experiments, our birds showed that they could rely on cues from human (Giret et al.2009a) and conspecific (Giret et al.2009b) agents to find hidden food. Thus, it is possible that Léo simply relied on the presence of a conspecific in order to make his decision and did not differentiate between partners because he could obtain food from both of them.
In the Apparatus choice test, the parrots seemed to improve their self-control and waited for their partner before pulling the string. However, in the complementary actions test, Zoé pulled the string before the arrival of Léo more when partner entry was delayed. Shango did not wait and most of the time preferred to leave the table rather than to wait for his partner. He was also the subject that showed the worst performance in a self-control experiment (Vick et al. 2009). Thus, personality (impulsivity) and emotional state (frustration) could explain Shango’s behaviour.
In experiment 4 (Complementary actions test), all three birds learned to act complementarily in order to reach the tray. However, after we exchanged their positions, they acted appropriately only three times (pulling or perching), but it was always unsuccessful because the partner did not produce the complementary action. Even if they were trained at performing both actions at the beginning, it seems that they specialised in their primary role and then were not able to adapt their behaviour to the new situation. The inability of the parrots to swap roles was probably due to their inability to understand the general setting and the necessity for the two subjects to engage in complementary actions simultaneously. This is probably different from the hunting behaviour observed in several species (see Brosnan et al. 2010 for a review) that are able to collaborate in the wild with a more or less lengthy learning period according to the species. However, the fact that our parrots at least tried collaborating once could also mean that they had some understanding of the actions needed. For example, Léo stayed on the perch until a partner retrieved the tray entirely, and thus, he seemed to realise that his partner could not retrieve the tray if he was not on the perch. An overall laziness and lack of motivation could also explain our results, since the parrots would not even walk a few more steps in order to go to their usual places to solve the task.
In this study, we were not able to draw conclusions regarding their intentions. None of our parrots made any recruitment attempt (or even emitted vocalisations); however, no observations from the wild or from captive parrots have shown the presence of this behaviour in the African grey species. During our studies, the birds oriented their heads towards their partner most of the time when partner’s entry was delayed. Because of the side-ways position of their eyes, it is hard to draw any conclusions based on their gaze direction (Dawkins 2002). Nevertheless, we cannot exclude a visual coordination as it has been observed in capuchins (Mendres and de Waal 2000), chimpanzees (Melis et al. 2006b) or hyenas (Drea and Carter 2009).
Our positive results suggest the need to study different groups of grey parrots with different devices in order to evaluate how they deal with their physical and social environment. Indeed, studies conducted in keas (Tebbich et al. 1996; Huber et al. 2008) and macaws (Spitzhorn 2009) reveal that according to the social organisation of the group and the experimental setup, different strategies are used to solve a cooperation task. During the experiments, we observed that instrumental cooperation emerged as a learned behavioural strategy. All three parrots were able to synchronise their actions. One bird (Shango) clearly showed that he was able to wait for the partner and another subject (Zoé) discriminated between the ‘Solo’ and the ‘Duo’ apparatus according to the presence or absence of a partner and thus was able to access the larger resource. Although our subjects seemed to understand the need of a partner to solve certain tasks, the same effect could probably, as emphasised by Noë (2006), have been achieved by combining the opportunity to gain a reward with a red light or any other cue. However, this ability to learn such contingencies could probably allow parrots to cooperate more efficiently (Melis and Semmann 2010). We observed that social preferences and tolerance had an impact on the probability of cooperation to occur and also on a dyad’s efficiency (Péron et al. submitted). There is no doubt that in nature parrots have to face complex social situations during which they must cooperate, for instance during bi-parental care.
We wish to thank Hélène Normand for her help in conducting the experiments. We also thank Philippe Groué, Colette Désaleux, Nicolas Giret, Alexandre Lerch and Marie Monbureau for taking care of the parrots. Lauriane Rat-Fischer drew the illustrations. We thank the anonymous reviewers and Ronald Noë for constructive comments on an earlier version. This experiment was setup within the framework of the ‘Integrating Cooperation Research Across Europe’ project. The experiments comply with the French laws concerning animal care (agreement n° 92–380).