Question 1: the extent of imitation in the two species
In total, 3794 observations were made, of which 58% (n = 2211) were visitor actions and 42% (n = 1579) were chimpanzee actions (for additional results on interaction rates see Online Resource 2). In four additional cases it was unclear whether the agent was a human or a chimpanzee. Inter-observer agreement based on 521 additional observations conducted simultaneously by the two observers was very high: k = 0.843, SE = 0.013 (see Online Resource 2 for details). The 3794 actions constituted 974 episodes, of which 36% (n = 354) were bidirectional, involving actions performed by both species, in a turn-taking manner. Fifty-six episodes could not be classified with certainty as either uni- or bidirectional. Imitative actions were attested in both species in 37% (n = 132) of all bidirectional episodes, and in 11% (n = 6) of the episodes that could not be classified with respect to bi-directionality, for a total of 356 occurrences (repetitions excluded). To compare cross-species differences with respect to imitation rate and imitative precision, z-tests of proportions were employed. There was no difference between the species in terms of the extent of cross-species imitation (Z = 0.04, P = 0.97), as 9.37% (n = 148) of all actions performed by the chimpanzees and 9.41% (n = 208) of all actions performed by the visitors were classified as imitation.
Question 2: the facilitating effect of imitation on social interaction
To determine if imitation had a facilitating effect on cross-species interaction by contributing to maintaining longer and more engaged interactions, episode length (i.e. number of recorded actions) was compared between episodes with and without imitation using the Mann–Whitney U-test. All unidirectional and unclassifiable episodes (n = 620) were excluded from this analysis. Significantly more actions (U = 7877.50, Z = −7.33, P < 0.001; median 1 = 7, median 2 = 4) were found in bidirectional episodes that included imitation (n = 132), as opposed to those that did not include imitation (n = 222).
Question 3: the precision of imitation in the two species
With respect to imitative precision, exact imitation (as opposed to partial imitation) was observed significantly more often (Z = 2.93, P = 0.003) in humans (73% of all imitative actions, n = 151) than in chimpanzees (58% of all imitative actions, n = 86). There were primarily two types of actions that gave rise to partial imitations: manual actions on windows (hit, knock, press, stroke), and kiss-like gestures (knocking or pressing window with lips).
Question 4: variations in the extent, quality and repertoires of imitation across individual chimpanzees and visitor categories
Four of the five chimpanzees produced imitative actions, the exception being SF1, who hardly interacted with visitors. A 2 × 4 χ2-test was conducted to compare the imitation rates of the four imitating individuals. This revealed a significant relationship between the presence of imitation and imitator identity (χ2 = 17.25, df = 3, P < 0.001; see also Table 1 for more details). Separate comparisons based on the z-test for proportions showed significantly higher imitation rates (all Ps < 0.001) for AM (10.6%, n = 114) and SF2 (15.2%, n = 16) compared to AF (2.8% n = 5). The latter also imitated less than JF (6.4%, n = 13), but this difference was non significant (Z = 1.66, P = 0.097). In turn, the imitation rate of JF was significantly lower than that of SF2 (Z = 2.52, P = 0.012), and marginally lower than that of AM (Z = 1.85, P = 0.06). Partial imitation occurred significantly more often (Z = 0, P < 0.001) in the younger (79%, n = 23) than in the adult chimpanzees (33%, n = 44).
Table 1 Number of actions, imitation rates, and imitative precision in the outdoor and indoor environments for the individual chimpanzees and visitor categories
To determine if imitation rates differed among visitor categories (see Table 1), a 2 × 8 χ2-test was conducted on a data subset that included all data collected for the eight sex-age categories, excluding the collective category Crowd. The result of this test was non-significant (χ2 = 9.14, df = 7, P = 0.243). Imitative precision, on the other hand, was significantly related to age (χ2 = 22.41, df = 3, P < 0.001), with partial imitation occurring more often among toddlers (68% of all imitative actions, n = 15) than among juveniles (20%, n = 11), subadults (29%, n = 6) or adults (20%, n = 17).
Overall, the imitative repertoire of the chimpanzees included 14 distinct actions (Table 2). AM exhibited most flexibility, and his imitation repertoire (13 distinct actions) almost equated the cumulative repertoire of the group. The females imitated two to five action types each. The cumulated imitative repertoire of the visitors (Table 2) was larger, including—besides the 14 actions imitated by the chimpanzees—nine additional behaviours. Almost all of these additional behaviours were postures (e.g. body hugging, thumb sucking), or acts with a physiological function (e.g. yawning, scratching). Among visitors, the imitative repertoire of toddlers (five distinct actions of which three were shown by males and all five by females) was lower than that of adults (15 actions; 12 of which were shown by males and 11 by females), subadults (11 actions; seven of which were shown by males and six by females) and juveniles (14 actions; all 14 of which were shown by males and seven by females).
Table 2 Imitative repertoires for visitors and chimpanzees
Only a few actions were frequently imitated by either species. These coincided across species and included hand clapping and actions on the windows (see Table 2). Both species imitated primarily actions that were already in their behavioural repertoire.
Question 5: the influence of interactional proximity on imitation
To assess if presence and precision of imitation was influenced by proximity to the interaction partner, and whether these potential influences were differentially expressed across species, two log linear analyses were conducted. Interactional proximity was defined by two values: distal interaction, i.e. when the species interacted across the moat or high wall (outdoors), and proximal interaction when the interaction took place through glass (indoors).
The first analysis focused on determining if interactional proximity affected the extent of imitation in the two species, and included the following factors: proximity (outdoors, indoors); presence of imitation (present, absent); and species (Homo sapiens, Pan troglodytes). The analysis produced a final model that retained the highest level effect, indicating a three-way significant interaction between presence of imitation × species × proximity (χ2 = 11.840, P = 0.001). The likelihood ratio for this model was χ2(0) = 0, P = 1, suggesting it to be a perfect fit for the observed data. To break down this interaction effect, χ2 analyses were conducted separately for each species which revealed a significant relationship between proximity and presence of imitation. Both species produced more imitative actions indoors than outdoors (chimpanzees, χ2 = 40.72, df = 1, P < 0.001; visitors, χ2 = 6.31, df = 1, P = 0.012; see also Table 1).
The second log linear analysis aimed at establishing interactional proximity proximity affected imitative precision, and included the following factors: imitative precision (exact imitation, partial imitation); species (H. sapiens, P. troglodytes), and proximity (outdoors, indoors). This analysis produced a model that retained, at the highest level, two two-way interaction effects: proximity × species (χ2 = 4.99, P = 0.025), and proximity × imitative precision (χ2 = 97.76, P < 0.001). A likelihood ratio test showed that this model was an acceptably good fit for the data (χ2 = 5.08, df = 2, P = 0.08). The first interaction has been discussed in the previous paragraph, whereby imitation by both species was found to be more frequent in indoor than outdoor locations. χ2-tests conducted to follow up on the second interaction revealed that partial imitation by either species was significantly more frequent indoors (chimpanzees, χ2 = 35.57, df = 1, P < 0.001; visitors, χ2 = 43.14, df = 1, P < 0.001). Indoors, the distribution of partial and exact imitation was not significantly different between chimpanzees and humans (χ2 = 2.92, df = 1, P = 0.08).
Among the chimpanzees, only AM engaged in imitative interactions both indoors and outdoors. The females imitated (and predominantly acted) either exclusively outdoors (AF) or indoors (SF2, JF). Likewise, the youngest visitor category (toddlers) imitated (and predominantly acted) indoors (Table 1).
Question 6: chimpanzees’ responses to being imitated
Of the 1579 actions produced by the chimpanzees, 196 were imitated by humans. Chimpanzees’ responses to being imitated included 71 instances of returned imitations, and 37 instances of non-imitative actions. In four additional cases it was not possible to establish if a chimpanzee responded with returned imitation or another action. Most responses to being imitated were single actions, as opposed to action sequences (of typically two, but up to seven actions), which were found in 10.95% of instances. Whether the chimpanzees responded or not, was not related to the type of imitation (exact or partial) to which they were exposed (χ2 = 0.51, df = 1, P = 0.52).
In 42 imitative exchanges the chimpanzees performed the minimal two turns required by the imitation game criterion, and all four imitating chimpanzees engaged at least once in such imitative games. Of these occasions, 11 were extended imitative exchanges, involving at least three—but up to ten—turns by the chimpanzee. Besides these 42 exchanges, we recorded an additional episode, which was difficult to fit to our restrictive contingency criterion, but which unfolded as a prolonged imitation game lasting as long as 10 min. This interaction consisted primarily of a sequence of mutual imitations of hand clapping and body rocking with arms slightly lifted in front of the body (the chimpanzee) or at shoulder height (the visitor). For the chimpanzee, play signals (play face, ground slaps) were observed to co-occur with imitation in the course of this interaction. Since the visitor appeared to be a tourist from abroad, it is unlikely that an interactional history existed between her and the chimpanzee.
No response was recorded in 84 cases (see Online Resource 2 for likely causes of a lack of response). Interestingly, chimpanzees appeared unlikely to respond when visitors imitated bodily postures or actions with physiological functions, such as yawning, scratching, etc. (two potential responses in 19 cases; P < 0.001, binomial test).