ANIMAL BEHAVIOUR, 2000, 60, 195–202 doi:10.1006/anbe.2000.1457, available online at http://www.idealibrary.com on

True imitation in marmosets

BERNHARD VOELKL & LUDWIG HUBER Institute of Zoology, University of Vienna

(Received 20 September 1999; initial acceptance 29 October 1999; final acceptance 1 April 2000; MS. number: 6359R)

Marmosets, Callithrix jacchus, observed a demonstrator removing the lids from a series of plastic canisters to obtain a mealworm. When subsequently allowed access to the canisters, marmosets that observed a demonstrator using its hands to remove the lids used only their hands. In contrast, marmosets that observed a demonstrator using its mouth also used their mouth to remove the lids. Since hand and mouth demonstrators brought about identical changes in the canisters, the differential test behaviour of the observer groups suggests that they learned about the demonstrator’s behaviour. Furthermore, marmosets that had not been given the opportunity to observe a demonstrator prior to testing had a low probability of mouth opening, even if the canisters were previously opened by a mouth-opening demonstrator in an olfactory control experiment. Corroborating Bugnyar & Huber’s (1997, Animal Behaviour, 54, 817–831) earlier findings, our results provide further evidence that marmosets can imitate.  2000 The Association for the Study of Animal Behaviour

Over the last decade, the adoption of a cognitive agreement among researchers on the conclusion that re- approach to the study of animal behaviour has consider- cent evidence has not supported the existence of imitation ably improved our understanding of the processes under- in monkeys’ (Custance et al. 1999, page 14). To our knowl- lying social learning. Researchers have investigated to edge, Bugnyar & Huber (1997) were the first to claim true which aspects of a conspecific’s instrumental behaviour movement imitation by monkeys, using marmosets as ex- animals attend, and which they understand, in particular perimental subjects. However, their procedure had weak- contexts (reviews in Zentall & Galef 1988; Whiten & nesses (see below) and only two of the five observers tested Ham 1992; Heyes 1994; Heyes & Galef 1996, 1998). In matched the action sequence of the demonstrator. Further- addition, a comparative approach has widened our more, it has been suggested that these animals may not knowledge of the taxonomic distribution of social learn- be able to learn anything more demanding from a demon- ing processes. For example, some researchers have sug- strator’s behaviour than to attend to either the relevant gested that great apes are the only animals capable of location (local enhancement, Thorpe 1963) or the rele- learning a behaviour by seeing it done (‘true imitation’), vant stimulus (stimulus enhancement, Spence 1937; see at least ‘in the limited sense of copying for its own sake Visalberghi & Fragaszy 1990; Whiten & Ham 1992; divorced from normal behaviour’ (Byrne & Tomasello Fragaszy & Visalberghi 1996; Moore 1996). 1995, page 1419; see also Byrne & Russon 1998). How- Only recently have methods been designed that control ever, others have suggested that imitation is widely scat- for nonimitative social learning, irrespective of the cog- tered among species (see, for instance, comments by nitive complexity of these processes (Akins & Zentall Fishbein, Heimann, Huber, Roitblat, Vogt & Carey and 1996, 1998; Kaiser et al. 1997; Campbell et al. 1999). Early Zentall on Byrne & Russon 1998). Although humans and experiments on imitation in animals used a nonexposed great apes are the most prolific imitators, dolphins (Tayler control (Heyes et al. 1992) and compared the perform- & Saayman 1973), some birds such as parrots, Psittacus ance of animals allowed to observe a demonstrator erithacus (Moore 1992), pigeons, Columba livia (Zentall et peforming an action on an object for food (observers) al. 1996; Kaiser et al. 1997) and Japanese quail, Coturnix with that of animals never given this opportunity (non- japonica (Akins & Zentall 1996, 1998) show some evi- observers). If the observers acquired the target response dence of imitative learning as well. faster or more efficiently than the nonobservers (Zentall Nevertheless, although improved procedures have ‘indi- et al. 1996), it was concluded that they had imitated the cated that a range of species can imitate’ (Heyes & Galef demonstrator’s behaviour. 1998, page 74), ‘there has been a relatively high amount of To control for nonimitative factors such as moti- Correspondence: L. Huber, Department of Theoretical Biology, Insti- vational, perceptual and species-specific effects on the tute of Zoology, University of Vienna, Biocenter, Althanstrasse 14, observer, experimentalists have used or recommended A-1090 Vienna, Austria (email: [email protected]). the two-action procedure (Dawson & Foss 1965; Galef 0003–3472/00/080195+08 $35.00/0 195  2000 The Association for the Study of Animal Behaviour 196 ANIMAL BEHAVIOUR, 60, 2

et al. 1986; Galef 1988; Whiten & Ham 1992). In this strator’s manipulandum, contagion may be sufficient to procedure, groups of animals are compared after explain demonstrator-consistent responding in observers they observed a conspecific demonstrator performing one (Campbell et al. 1999). of two or more alternative actions on a single manipulan- Huber (1998) argued that novelty should be used more dum. For example, Dawson & Foss (1965) allowed budg- carefully as a means of identifying cases of true imitation erigars, Melopsittacus undulatus, to observe demonstrators for two reasons. First, there is little agreement as to how removing the lid from a bowl by either pushing or novelty should be defined. Second, it is difficult to see twisting it off with the beak, or twisting it off with a foot. how the emergence of novel behaviours could be ident- True imitation may be said to have occurred if the ified. However, Thorpe’s (1963, page 135) definition of observer groups differ significantly in the proportion of imitation ‘as the copying of a novel or otherwise improb- the two actions performed during their initial attempts to able act or utterance’ makes these problems more man- manipulate the test object; that is, if they show a bias ageable. The probability that an action will be performed towards the demonstrator’s response (see Galef et al. and could be determined by using a group of nonobservers Tomasello in Heyes & Galef 1996). and measuring the frequency of occurrence of the target The two-action procedure may be contrasted with the action. Using this method, Bugnyar & Huber (1997) two-object procedure (Campbell et al. 1999), in which the showed that the exact reproduction of a sequence of performances of two observer groups are compared after actions used by a demonstrator to open a pendulum door they witnessed a demonstrator manipulating one of two was improbable. Therefore, the exact reproduction of the different objects or parts of an object. In this procedure, the demonstrator’s sequence by two of the observers was information acquired by the observers may consist of interpreted as evidence of imitation. knowing what parts of the physical environment are rel- We permitted two groups of marmosets, Callithrix jac- evant for a task or where to begin manipulating an object. chus, to observe a demonstrator using one of two alterna- A clear separation of the two types of learning is not tive techniques and compared their initial test responses easy to achieve. Campbell et al. (1999) allowed starlings, with one another and with a third group of marmosets Sturnus vulgaris, to observe demonstrators removing a that were never given the opportunity to observe a plug from a hole in the lid of a food container either by demonstrator. The particular requirement of our exper- pulling on a loop of string attached to the plug or by iment was that the two opening techniques (pulling with pushing the plug down. As the authors admitted, these a hand versus pulling with the mouth) have the same two actions may have rendered different regions of the effect on the environment, that is, the lid to be opened plug attractive (Lefebvre et al. 1997). The two groups of would ‘behave’ identically if moved automatically. Fur- observers may also have learned different affordances of thermore, one technique consisted of a behavioural the plug (i.e. dynamic properties, potential uses; Gibson ‘peculiarity’ (mouth opening), that is, mouth opening 1966; Tomasello et al. 1987), for example, that the plug was neither common in the animals under investigation moved up and out of the food canister when lifted, and nor necessary for lid removal. This requirement ensured down and into the canister when pushed. that if the technique was performed by the observers, Therefore, to determine whether animals learn by then they were most probably influenced by what they observation about responses or about changes of state in witnessed (Whiten & Custance 1996; Whiten et al. 1996; the environment, a two-action/one-outcome procedure is Custance et al. 1999). Mouth opening was occasionally preferable. Zentall and colleagues ruled out observed found in one female marmoset out of five animals and differences in the movement of a manipulandum by occurred as a by-product of a previous experiment on comparing naïve pigeons (Zentall et al. 1996) and quail associative learning. We expected that observation of a (Akins & Zentall 1996) that had observed a demonstrator mouth-opening demonstrator would affect the prob- either pecking at or stepping on a treadle for food. In both ability of other monkeys showing this behaviour. On the studies, a significant correlation was found between the other hand, we predicted that observation of hand open- observers’ performance and the demonstrated behaviour ing would have no effect on the probability of mouth even though only the demonstrators’ response topogra- opening by the monkeys. We tested this by making the phies (pecking versus stepping) differed but not the removal of the lid more difficult (and perhaps mouth effects on the environment. opening more efficient). We also used a nonexposed A further methodological complication emerges when control group to control for nonsocial factors and to one attempts to determine whether the copied action was assess the baseline probability of the mouth-opening elicited by nonimitative processes such as contagion response. Finally, given the current concern in the litera- (Thorpe 1963), rather than acquired as a result of learning ture with the potential role of olfactory cues in promoting a novel or improbable motor act (Thorpe 1956, 1963; demonstrator-consistent responding, we included an Meltzoff 1988; Zentall 1988; Whiten & Custance 1996). olfactory control group. For example, if a specific response is part of the animal’s behavioural repertoire for approaching an object, such as pecking is in pigeon or quail, the observation of a pecking METHODS demonstrator may have acted as an innate releaser of Subjects pecking in the observer (Akins & Zentall 1996, 1998; Zentall et al. 1996). In combination with stimulus The subjects were 36 adult common marmosets, of enhancement, or increased attention towards the demon- which two were demonstrators and 11 were observers. We VOELKL & HUBER: TRUE IMITATION IN MARMOSETS 197 used 11 subjects (including the two models) for the nonexposed control and 14 for the olfactory control. The animals were maintained in four family groups at the Institute of Zoology of Vienna University, Austria, two family groups at the Konrad Lorenz Institute for Evol- ution and Cognition-Research, Altenberg, Austria, and three family groups at the Instituto Superiore di Sanita, Rome, Italy. All animals were born in captivity. At the Institute of Zoology two groups lived in indoor cages (250
250
250 cm) that were attached to outdoor cages of the same size and two groups lived in indoor cages (120
80 cm and 180 cm high and 260
260 cm and 250 cm high, respectively). At the Konrad Lorenz Institute both groups lived in indoor cages (200
350 cm and 300 cm high). At the Instituto Superiore di Sanita the groups lived in indoor cages (150
50 cm and 220 cm high). All cages were equipped with branches, ropes and living plants. All animals were fed fruits, vegetables, monkey pellets and protein supplements. The tempera- ture was 26–30C during the day and 21–23C at night. The humidity was ca. 70–80%. In summer daylight was the main source of lighting, while in winter additional UV-fluorescent tubes were used to maintain a 12:12 h light:dark cycle.

Apparatus The apparatus consisted of a wooden board (10
33 cm) with five black plastic film canisters for 35-mm Kodak slide films attached to it with screws. The board and the lids of the film canisters were painted blue. In each session the subjects were given three identical boards (with a total of 15 canisters). The canisters were each filled with one mealworm as a reward. In the Figure 1. Schematic representation of the two opening techniques performed by (a) the mouth-opening demonstrator and (b) the observer sessions and in the first test session the lids of hand-opening demonstrator. the canisters were half shut (only one half of the lid was firmly attached to the canister, whereas the other half did not click into place and therefore left a gap between the lid and the upper end of the canister) while in the second successive sessions and needing less than 2 min for each test session the lids were completely shut. At the labora- session), we gave BI two more sessions with half-shut tories in Altenberg and Rome the experiments were con- canisters. After we confirmed that BI continued with this ducted in the home cage of the subjects, while the technique (only mouth opening in two sessions), we used subjects in Vienna were tested in an experimental cage it as demonstrator for six subjects (observers) of its neigh- (120
70 cm and 260 cm high) adjoined to the home bouring group (group Mouth). During the observation cage. sessions, the other animals of BI’s group were locked out and we placed the three boards 25 cm from the wire-mesh partition that separated BI’s cage from the observers’ cage. Procedure The demonstrator (BI) was then allowed to open the canisters while all the observers had the opportunity to Group Mouth observe it. In a previous experiment, female BI was the only one of We conducted two observation sessions each day. The five subjects that opened the canisters by removing the sessions were terminated as soon as the demonstrator had lids with its mouth (Fig. 1a). This animal was selected as opened all the canisters. Immediately after the fifth obser- the mouth-opening demonstrator. The other subjects in vation session all the observers were tested individually in this previous experiment were not used in the present two successive test sessions. For this purpose we removed one. In six training sessions BI was separated from the all but one subjects of the observer group from the other group members and we confirmed its original observers’ cage and placed the boards into that cage. In preference for the mouth-opening technique by offering the first test session the canisters were half shut and the completely shut canisters, which could only be opened subjects had a maximum of 10 min to open them. If by this technique. After the subject performed the the subjects needed less than 10 min to open the 15 task well (making only correct responses within three canisters, the session was terminated earlier. After the first 198 ANIMAL BEHAVIOUR, 60, 2

test session the subjects immediately had a second the subjects were tested individually in their home cages. test session, in which the canisters were completely shut. The sessions were terminated after 15 min. In this exper- The session was again limited to 10 min. iment we gave all the subject two boards with a total of 10 canisters in each session. Group Hand A male subject (GR) was separated from its group and Data Analysis confronted with the apparatus with half-shut canisters. All sessions were videotaped with a DV camcorder GR spontaneously started to open the canisters using its (Sony VX 1000). We classified the behaviour at the hand (Fig. 1b). To confirm that GR opened the canisters apparatus as either mouth opening or hand opening. An in this manner we repeatedly offered only half-shut action was called mouth opening when the subject bit canisters. Unlike the first demonstrator (BI), GR was not into the lid and raised its head while still keeping a firm given completely shut canisters, but it had seven succes- hold on the lid with its teeth (Fig. 1a). An act was called sive training sessions with half-shut canisters. During hand opening when a subject inserted one or more these training sessions the experimental cage was visually fingers at the edge of the lid and lifted the lid by isolated from the home cage by a cardboard partition. stretching its forearm (Fig. 1b). As in studies with pigeons After GR had shown only correct responses within the last and quail in which stepping and pecking responses were three sessions, it was used as a demonstrator. We removed compared (Akins & Zentall 1996, 1998; Zentall et al. the cardboard screen and placed the demonstrator in the 1996; Kaiser et al. 1997), the two response topographies experimental cage while the other group members (group contrasted here (mouth versus hand opening) were easily Hand) could observe it. Again we conducted five sessions. distinguished by the experimenter. Here, too, the pos- The sessions were terminated as soon as the demonstrator tures and body parts involved in the two responses were had opened all the canisters. After these observation different (compare Fig. 1a and b). sessions every observer was separately tested in the Sometimes a subject opened the lid only half way by experimental cage as described above. hand or mouth, then put its head into the gap between the lid and the canister, which caused the lid to open. If Nonexposed control group such a sequence was started by one of the two actions To assess the frequency of spontaneous mouth opening (mouth opening or hand opening) we accordingly classi- we conducted a control experiment. The control group fied it as mouth opening or hand opening. If a subject did consisted of 11 individuals that had access to the boards not open the canister with the first movement and without having the opportunity to observe a social repeatedly used the same technique several times in model. As in the previous experiments, the subjects were succession, we counted this as one action. If a subject did tested individually in their home cages. The sessions were not open the canister with the first movement (i.e. hand terminated after 15 min. If the subjects did not succeed in opening) and switched to the other opening technique opening at least three canisters in one session, further (i.e. mouth opening) and opened the lid, we recorded this sessions were added on subsequent days. as mouth opening, because only full actions entered into the analysis. We defined a full action as an opening Olfactory control group action where the lid was removed completely from the Odour cues may have influenced our marmosets’ behav- canister. The subjects could reach the reward only if the iour. The mouth-opening demonstrator (BI) may have left lid was completely removed. As every canister was filled saliva containing food particles on the lids of the canisters, with a mealworm, each full action was rewarded. making its observers more likely to sniff at the lids than the The entire database was coded from videotape by B.V. marmosets that observed the hand-opening demonstrator. To measure observer reliability we selected two short If this is correct, scent-mediated local enhancement instead video sequences: one with 10 opening acts of an observer of, or in addition to, response learning by observation of group Hand and one with seven opening acts of an might have been responsible for demonstrator-consistent observer of group Mouth. We asked 11 independent responding in the mouth-opening observers. To determine persons to classify the opening acts as hand opening’, whether differences in the opening techniques used by the ‘mouth opening’ or ‘not visible’ (when the opening act two groups of observers resulted from a tendency for was not clearly visible on the videotape). The persons did mouth-opening observers to approach the lid with their not know that one of the subjects had a mouth-opening head and/or mouth, we analysed videotapes of these demonstrator and the other a hand-opening demon-   approaches in detail. strator. They coded 99.4 2.2% (X SD) of the 17 actions We also conducted a second control experiment to in the same way that B.V. did. This indicates that these determine more directly whether olfactory or visual cues, actions were clear and easy to distinguish and therefore that is saliva or scratches, that the previous mouth- the coding bias was negligible. opening animal might have left on the canisters’ lids, influence the subjects’ behaviour. This control group RESULTS consisted of 14 subjects that had access to canisters that Observation Phase were previously opened by another animal by mouth. The subjects could not see the other animal opening the While being observed, both demonstrators showed per- canisters, so this was again a nonexposed control. Again, fect discrimination between the two opening techniques. VOELKL & HUBER: TRUE IMITATION IN MARMOSETS 199

group Mouth while none of the five subjects in group Hand opening Mouth opening Hand opened a canister with its mouth (Fig. 2). Despite the small group size this difference was statistically significant (Fisher’s exact test: N=11, P=0.045). 8 To compare the performance of the two groups we NS calculated a discrimination ratio for each subject by 6 dividing the number of hand-opening acts performed * during the first test session by the total number of opening acts during that session. The mean discrimin- 4 ation ratioSD was 0.640.34 (N=6) for group Mouth, 

Number of subjects and 1.00 0(N=5) for group Hand (Table 1). We com- puted the standardized deviates (Sokal & Rohlf 1995) 2 separately for each subject’s discrimination ratio, pooled the resulting deviates and tested them for normality by a Kolmogorov–Smirnov test for goodness of fit. The Mouth Hand Mouth Hand observed data did not depart significantly from a normal Group distribution (Z=0.754, P=0.62) and heterogeneous vari- ances were assumed. Therefore we used Welch’s t test Figure 2. Number of observers that opened the canisters at least (Sokal & Rohlf 1995) to compare the means of the two once by hand (Hand opening) or by mouth (Mouth opening) during the first test session. Six observers saw a mouth-opening demon- groups. The discrimination ratios for groups Hand and strator (group Mouth) and five saw a hand-opening demonstrator Mouth were significantly different (t4.5=2.33, P=0.033, (group Hand). *P<0.05. one tailed).

Both were presented with 75 canisters. The demonstrator Test Phase: Second Test Session for group Mouth (BI) opened all 75 canisters (100%) using its mouth. The demonstrator for group Hand (GR) In the second test session the canisters were completely opened 74 out of 75 canisters (98.7%) with its hand. One shut and could only be opened by mouth. This session lid accidentally opened when GR jumped on to the was conducted to investigate if the subjects that used board. GR opened the lid only half way by hand 12 times only their hands in the first session were also able to open and removed the lid by putting its head into the canister. canisters with the mouth. From the two subjects from group Mouth that did not use the mouth in the first Test Phase: First Test Session session, one immediately began to open the canisters with its mouth, whereas none of the subjects from group In both groups, all the subjects opened at least one Hand managed to open a canister with its mouth. Thus in canister with their hand, but opening the canisters with the second test session five out of six subjects from group the mouth was shown by four out of six subjects from Mouth opened the canisters with their mouth and one

Table 1. Opening behaviour of the observers and control subjects during the first and second test sessions Session 1 Session 2 Nose-near-lid Mouth Hand Opened Discrimination Opened Group Subject approaches opening opening canisters ratio canisters

Mouth DV 3 0 15 15 1.00 0 MO 2 6 8 14 0.57 15 WI 4 13 2 15 0.13 15 SQ 5 11 4 15 0.27 15 NI 7 2 13 15 0.87 6 MA 0 0 15 15 1.00 15 Mean 3.5 5.3 9.5 14.8 0.64 11 Hand KL 2 0 15 15 1.00 0 SU 2 0 15 15 1.00 0 VI 3 0 15 15 1.00 0 GI 5 0 15 15 1.00 0 SH 5 0 14 14 1.00 0 Mean 3.4 0 14.8 14.8 1.00 0 Control Mean 1.0 10.9 13.0 0.92 1.6

The total numbers of nose-near-lid approaches, mouth-opening and hand-opening actions and opened canisters and the discrimination ratio (the number of hand-opened canisters divided by the total number of opened canisters) are shown for session 1. For session 2 only the total number of opened canisters is shown, as all canisters were opened by mouth. For the nonexposed control group (N=11) only the mean values are shown. 200 ANIMAL BEHAVIOUR, 60, 2

subject failed to open at least one canister within 10 min, mouth opening. The mean discrimination ratioSD while none of the five subjects from group Hand suc- was 0.850.34. The subjects opened the canisters ceeded in opening a canister (Fisher’s exact test: N=11, significantly more often with their hands (Wilcoxon P=0.013). signed-ranks test: Z=2.128, P=0.033, two tailed).

Olfactory Exploration Olfactory Control Group

Animals of both groups typically sniffed at the canis- This control group consisted of 14 subjects. Five did not ters, especially during the first test session. This behaviour open any canister within the 15 min of the test session. was clearly part of their olfactory exploration at the onset Of the remaining nine subjects, eight opened the canis- of the test. We counted all approaches where an animal’s ters with their hands and only one opened them with its nose touched the lid of a canister, or came very close (to mouth. The mean discrimination ratioSD was 0.89 within 5 mm) to the lid (nose-near-lid). The mean 0.33. The subjects opened the canisters significantly number of nose-near-lid bouts until the first canister was more often with their hands (Wilcoxon signed-ranks test: openedSD was 3.52.4 for group Mouth and 3.41.5 Z=2.433, P=0.015, two tailed). A comparison of the mean for group Hand and the number of nose-near-lid bouts discrimination ratio with the nonexposed control group during the entire first session was 13.75.6 for group showed no significant difference (Mann–Whitney U test:  Mouth and 13.2 3.3 for group Hand. The mean latency U=41.5, N1=11, N2=14, P=0.652) whereas a comparison to the first nose-near-lid boutSD was 45.844.2 s for of the mean discrimination ratio with group Mouth was  group Mouth and 21.0 2.6 s for group Hand. significant (Mann–Whitney U test: U=10.0, N1=6, N2=14, Again we used the standardized values of both groups P=0.016) even when we applied a Bonferroni method to test them for normality by a Kolmogorov–Smirnov test (Sokal & Rohlf 1995) to evaluate the experimentwise error for goodness of fit. The observed data did not depart rate, which reduced the critical P value to 0.025. Thus the significantly from a normal distribution for the number subjects from group Mouth opened the canisters signifi- of bouts until the first canister was opened (Z=0.463, cantly more often with their mouth than the subjects N=11, P=0.983), for the number of bouts during the first from the olfactory control group, which were given can- session (Z=0.537, N=11, P=0.935) and for the latency to isters previously mouth-opened by another monkey. On the first bout (Z=0.754, N=11, P=0.621). Therefore we the other hand no difference could be found between the used a two-tailed t test to compare the number of bouts olfactory control group and the nonexposed control until the first canister was opened and the number of group, where the subjects were given canisters that were bouts during the first session. To compare the latencies to not previously mouth-opened. Therefore visual or olfac- the first bout a Welch’s t test was used after Levene’s test tory cues on the lids of the canisters were not sufficient to for equality of variances found heterogeneous variances promote mouth opening in naïve monkeys.

between the two groups (F5,4=5.998, P=0.037). No sig- nificant differences were found between group Hand and group Mouth in the number of nose-near-lid bouts until DISCUSSION the first canister was opened (t test: t =0.08, P=0.938), 9 Common marmosets copied the response topography of the number of nose-near-lid bouts during the entire first a conspecific demonstrator to open a Kodak film canister test session (t test: t =0.16, P=0.874) and the latency to 9 and obtain a desired food item. More specifically, marmo- the first nose-near-lid bout (Welch’s t test: t =1.37, 4.5 sets that had observed a demonstrator using its mouth to P=0.228). remove the lid from a film canister were more likely to These data cannot exclude the possibility that the use this technique during a subsequent test session than observers smelled some deposits on the lid without being marmosets that had observed a demonstrator using exclu- influenced in the way they approached the canisters; the sively its hand. In fact, mouth opening was rare in 30 same is true for visual cues such as saliva droplets on subjects that had never seen a mouth-opening demon- the lid. But the argument that such cues promoted the stration or any opening demonstration before. These approach with the head towards the lid and thus facili- results are consistent with a previous study of imitative tated the probability of mouth opening is not supported learning in marmosets (Bugnyar & Huber 1997) and raise by our data. doubts about the claim that the ability to imitate is unique to great apes, including humans (Tomasello & Nonexposed Control Group Call 1997; Byrne & Russon 1998). This effect cannot be accounted for by either enhance- The nonexposed control group consisted of 11 subjects, ment effects or object movement re-enactment (learning including BI and GR, the two models. Two subjects did how an object moves; Heyes 1998), since the lid of the not open even one canister within the first test session canisters moved in the same way for both observer and two opened fewer than three canisters. However, groups. Contagion is also an unlikely explanation, since these four subjects successfully opened canisters in the the observers were tested in the absence of a demon- second test session. To assess the discrimination ratio of strator. Finally, it is also unlikely that olfactory cues these animals, we pooled the data from the first and influenced the direction and strength of the observers’ second test sessions. Only two of 11 subjects showed approach behaviour. A control experiment showed that VOELKL & HUBER: TRUE IMITATION IN MARMOSETS 201 deposits left by the demonstrators on the lids of the opening technique shown by a single female during a canisters did not increase the probability of mouth open- pilot study. We further profited from the fact that this ing. Observers of hand-opening demonstrators also behavioural peculiarity involved an action that differed sniffed the lid as frequently and closely as observers of on a large scale from the default action in that it involved mouth-opening demonstrators. a different body part. Thus, we avoided a problem com- After decades of refining both methods and terminol- mon to studies of social learning in great apes (see ogy, the future of social learning will inevitably involve Tomasello 1996 and Whiten & Custance 1996 for attempts to extend the range of species comparisons. Not reviews), in which a complex object or a tool is manipu- only will questions of how animals become proficient lated with the hands in one of two or more different imitators achieve heuristic value, but also questions about ways, only one of which is demonstrated to a subject. what conditions promote social learning in contrast, or in A further advantage of our two-action/one-peculiarity addition, to individual learning. As suggested by Caldwell approach is that copying of a peculiar behaviour qualifies et al. (1999), the all-or-none approach to imitation is as being truly imitative. According to Thorpe (1963), one being superseded by an approach that recognizes the should consider only motor acts that do not pre-exist in the complex concatenations of the processes and circum- animal’s behavioural repertoire at the onset of the exper- stances of social learning. The repeated failure to find iment, otherwise socially mediated behaviour (contagion) robust imitative effects in animals other than apes may be would be sufficient to result in behavioural matching. We due to either an overestimation of the species’ manipu- were able to show that the probability of spontaneous lative abilities or an underestimation of the species’ other mouth-opening behaviour was low; in the two non- capacities that render the task more manageable. For observer control groups mouth opening was rare and in the example, marmosets experienced difficulty in turning a hand-opening observer group it never occurred. Although handle in order to open a plastic box with a hinged lid mouth opening may not be completely novel at the level of (Caldwell et al. 1999) and olfactory cues played a greater motoric execution, it is a rather novel technique for open- role than visual cues in the rat’s exploratory behaviour ing plastic film canisters (see Akins & Zentall 1996 for a (Mitchell et al. 1999). similar argument on ‘novelty’). To identify the conditions under which robust imitative To conclude, our findings support the suggestion that effects occur, we need some understanding of the species’ basic forms of imitative learning occur not only in ecological niche and the natural context in which move- humans and great apes but also in monkeys. Regardless of ment imitation appears. In humans, and probably in some the mechanism responsible for behavioural matching, it great apes and dolphins, faithful copying of a demon- seems plausible that social learning allows the experience strator’s behaviour is part of a social game. Learning to of one individual to be acquired by others more effi- exploit hidden food sources may be well established in ciently than would be the case with trial-and-error learn- highly explorative and manipulative mammals (e.g. chim- ing (Boyd & Richerson 1988). As we saw here, and also in panzees, Pan troglodytes: Whiten et al. 1996) and birds (e.g. Bugnyar & Huber (1997), even faithful copying in the ravens, Corvus corax: Fritz & Kotrschal 1999; keas, Nestor absence of insight may help to solve a task that otherwise notabilis: L. Huber, S. Rechberger & M. Taborsky, unpub- could not be solved. Those monkeys in our study that did lished data). Vocal mimicry is widespread among birds and not have the opportunity to observe a mouth-opening common in humans, but occurs rarely, if ever, in great apes demonstrator failed to open canisters, and hence to and other mammals (Moore 1996). Furthermore, the occur- obtain a highly nutritious food, when the canisters were rence of social-learning mechanisms may also be influ- completely closed (second test session). This is especially enced by the species’ group size (Boyd & Richerson 1996) interesting in the case of the hand-opening observers, and social tolerance (Coussi-Korbel & Fragaszy 1995). In which had extensive experience with the canisters. Thus, tolerant primate groups, such as the marmoset family, our findings also provide some support for the hypothesis individuals often follow one another, exploring the forest that social learning may be adaptive. canopy and approaching others busy with interesting objects. In these situations, copying behaviour is likely to Acknowledgments occur, especially if an individual cannot understand the causal structure of the task. Recently, Huber (1998) argued This experiment was conducted by B.V. in partial fulfill- that copying fidelity would be especially high if it was not ment of the requirements for the Master degree in Zool- affected by an insight into the causal relationships of the ogy at the University of Vienna. Ruth Sonnweber built task (programme imitation; Byrne & Byrne 1993), the af- the apparatus and conducted a pilot study. Augusto fordances of the objects (emulation or affordance learning; Vitale, Armelle Queyras and Michela Zanzony helped Tomasello 1996) or the goal and intentions of the demon- conduct the control experiment. Financial support was strator (goal emulation; Whiten & Ham 1992). If all of provided by the Austrian Science Foundation (FWF-grant these types of information are beyond the cognitive P 12011-BIO). We thank Celia Heyes for comments and capacity of the species, then slavish copying is a valuable Fiona Campbell for improving the English. alternative. Behavioural peculiarities, unnecessary extensions and References even errors in performance can be used to determine the Akins, C. K. & Zentall, T. R. 1996. Imitative learning in male degree and type of behavioural matching. In the present Japanese quail (Coturnix japonica) using the two-action method. experiment, we were able to benefit from a peculiar Journal of Comparative Psychology, 110, 316–320. 202 ANIMAL BEHAVIOUR, 60, 2

Akins, C. K. & Zentall, T. R. 1998. Imitation in Japanese quail: the Huber, L. 1998. Movement imitation as faithful copying in the role of reinforcement of demonstrator responding. Psychonomic absence of insight (comment on Byrne & Russon). Behavioral and Bulletin and Review, 5, 694–697. Brain Sciences, 21, 694. Boyd, R. & Richerson, P. J. 1988. An evolutionary model of social Kaiser, D. H., Zentall, T. R. & Galef, B. G. Jr. 1997. Can imitation in learning: the effect of spatial and temporal variation. In: Social pigeons be explained by local enhancement together with trial- Learning: Psychological and Biological Perspectives (Ed.byT.R. and-error learning? Psychological Science, 8, 459–460. Zentall & B. G. Galef, Jr), pp. 29–48. Hillsdale, New Jersey: L. Lefebvre, L., Templeton, J., Brown, K. & Koelle, M. 1997. Carib Erlbaum. crackles imitate conspecific and Zenaida dove tutors. Behaviour, Boyd, R. & Richerson, P. J. 1996. Why culture is common, but 134, 1003–1017. cultural evolution is rare. In: Evolution of Social Behaviour Patterns Meltzoff, A. N. 1988. Imitation, objects, tools, and the rudiments of in Primates and Man (Ed. by W. G. Runciman, J. Maynard Smith & language in human ontogeny. Human Evolution, 3, 45–64. R. I. M. Dunbar), pp. 77–93. Oxford: Oxford University Press. Mitchell, C. J., Heyes, C. M., Gardner, M. R. & Dawson, G. R. Bugnyar, T. & Huber, L. 1997. Push or pull: an experimental study 1999. Limitations of a bidirectional control procedure for the on imitation in marmosets. Animal Behaviour, 54, 817–831. investigation of imitation in rats: odour cues on the manipu- Byrne, R. W. & Byrne, J. M. E. 1993. Complex leaf-gathering skills of landum. Quarterly Journal of Experimental Psychology, 52B, 193– mountain gorillas (Gorilla g. beringei): variability and standardiz- 202. ation. American Journal of Primatology, 31, 241–261. Moore, B. R. 1992. Avian movement imitation and a new form of Byrne, R. W. & Russon, A. E. 1998. Learning by imitation: a mimicry: tracing the evolution of a complex form of learning. hierarchical approach. Behavioral and Brain Sciences, 21, 667–721. Behaviour, 122, 231–263. Byrne, R. W. & Tomasello, M. 1995. Do rats ape? Animal Behaviour, Moore, B. R. 1996. The evolution of imitative learning. In: Social 50, 1417–1420. Learning in Animals: the Roots of Culture (Ed.byC.M.Heyes& Caldwell, C. A., Whiten, A. & Morris, K. D. 1999. Observational B. G. Galef, Jr), pp. 245–265. San Diego: Academic Press. learning in the marmoset monkey (Callithrix jacchus). In: AISB Sokal, R. R. & Rohlf, F. J. 1995. Biometry. New York: W. H. Freeman. Convention: Symposium on Imitation in Animals and Artifacts (Ed. by Spence, K. W. 1937. Experimental studies of learning and higher K. Dautenhahn & C. Nehaniv), pp. 27–31. Edinburgh: The Society mental processes in infra-human primates. Psychological Bulletin, for the Study of Artificial Intelligence and the Simulation of 34, 806–850. Behaviour. Tayler, C. K. & Saayman, G. S. 1973. Imitative behaviour by Indian Campbell, F. M., Heyes, C. M. & Goldsmith, A. R. 1999. Stimulus Ocean bottlenose dolphins (Tursiops aduncus) in captivity. Behav- learning and response learning by observation in the European iour, 44, 286–298. starling, in a two-object/two-action test. Animal Behaviour, 58, Thorpe, W. H. 1956. Learning and Instinct in Animals. London: 151–158. Methuen. Coussi-Korbel, S. & Fragaszy, D. M. 1995. On the relation between Thorpe, W. H. 1963. Learning and Instinct in Animals. 2nd edn. social dynamics and social learning. Animal Behaviour, 50, 1441– London: Methuen. 1453. Tomasello, M. 1996. Do apes ape? In: Social Learning in Animals: Custance, D., Whiten, A. & Fredman, T. 1999. Social learning of an the Roots of Culture (Ed. by C. M. Heyes & B. G. Galef, Jr), artificial fruit task in capuchin monkeys (Cebus apella). Journal of pp. 319–346. San Diego: Academic Press. Comparative Psychology, 113, 13–23. Tomasello, M. & Call, J. 1997. Primate Cognition. New York: Oxford Dawson, B. V. & Foss, B. M. 1965. Observational learning in University Press. budgerigars. Animal Behaviour, 13, 470–474. Tomasello, M., Davis-Dasilva, M., Camak, L. & Bard, K. 1987. Fragaszy, D. M. & Visalberghi, E. 1996. Social learning in monkeys: Observational learning of tool-use by young chimpanzees. Human primate ‘primacy’ reconsidered. In: Social Learning in Animals: the Evolution, 2, 175–183. Roots of Culture (Ed. by C. M. Heyes & B. G. Galef, Jr), pp. 65–84. Visalberghi, E. & Fragaszy, D. M. 1990. Do monkeys ape? In: San Diego: Academic Press. ‘Language’ and Intelligence in Monkeys and Apes. Comparative Fritz, J. & Kotrschal, K. 1999. Social learning in common ravens, Developmental Perspectives (Ed. by S. T. Parker & K. R. Gibson), Corvus corax. Animal Behaviour, 57, 785–793. pp. 247–273. Cambridge: Cambridge University Press. Galef, B. G. Jr. 1988. Imitation in animals: history, definition and Whiten, A. & Custance, D. 1996. Studies in imitation in chimpan- interpretation of data from the psychology laboratory. In: Social zees and children. In: Social Learning in Animals: the Roots of Learning: Psychological and Biological Perspectives (Ed.byT.R. Culture (Ed. by C. M. Heyes & B. G. Galef, Jr), pp. 291–318. San Zentall & B. G. Galef, Jr), pp. 207–223. Hillsdale, New Jersey: Diego: Academic Press. L. Erlbaum. Whiten, A. & Ham, R. 1992. On the nature and evolution of Galef, B. G. Jr, Manzig, L. A. & Field, R. M. 1986. Imitation learning imitation in the animal kingdom: reappraisal of a century of in budgerigars: Dawson and Foss (1965) revisited. Behavioural research. In: Advances in the Study of Behavior (Ed. by P. J. B. Slater, Processes, 13, 191–202. J. S. Rosenblatt, C. Beer & M. Milinski), pp. 239–283. New York: Gibson, J. J. 1966. The Senses Considered as Perceptual Systems. Academic Press. Boston: Houghton Mifflin. Whiten, A., Custance, D. M., Gomez, J.-C., Teixidor, P. & Bard, Heyes, C. M. 1994. Social learning in animals: categories and K. A. 1996. Imitative learning of artificial fruit processing in chil- mechanisms. Biological Review, 69, 207–231. dren (Homo sapiens) and chimpanzees (Pan troglodytes). Journal Heyes, C. M. 1998. Theory of mind in nonhuman animals. Behav- of Comparative Psychology, 110, 3–14. ioral and Brain Sciences, 21, 101–148. Zentall, T. R. 1988. Experimentally manipulated imitative behavior Heyes, C. M. & Galef, B. G. Jr (Eds) 1996. Social Learning in Animals: in rats and pigeons. In: Social Learning: Psychological and Biological the Roots of Culture. San Diego: Academic Press. Perspectives (Ed. by T. R. Zentall & B. G. Galef, Jr), pp. 191–206. Heyes, C. M. & Galef, B. G. Jr. 1998. The Napoli social learn- Hillsdale, New Jersey: L. Erlbaum. ing conference. Introduction. International Journal of Comparative Zentall, T. R. & Galef, B. G. Jr (Eds) 1988. Social learning: Psychology, 11, 73–92. Psychological and Biological Perspectives. Hillsdale, New Jersey: Heyes, C. M., Dawson, G. R. & Nokes, T. 1992. Imitation in rats: L. Erlbaum. initial responding and transfer evidence. Quarterly Journal of Exper- Zentall, T. R., Sutton, J. E. & Sherburne, L. M. 1996. True imitative imental Psychology, 45B, 229–240. learning in pigeons. Psychological Science, 7, 343–346.