Experience-based human perception of facial expressions in Barbary macaques (Macaca sylvanus)

Barbary macaque tourism as a case study for interspecies interactions

Barbary macaques are currently present in Morocco, Algeria and Gibraltar, where they are a popular tourist attraction (Perez & Bensusan, 2005; Majolo et al., 2013). Interactions between Barbary macaques and tourists/locals are a major concern in Gibraltar (Perez & Bensusan, 2005) and are a growing issue in many tourist destinations in Morocco (Maréchal et al., 2011; Maréchal et al., 2016a; Maréchal et al., 2016b). The species was classified in 2008 as Endangered by the IUCN (Butynski et al., 2008), and primate tourism is thought to have the potential to be beneficial for the conservation of the Barbary macaque by increasing public awareness, providing protection for the species and their habitats, and contributing to the local and/or national economy (Russon & Wallis, 2014). However, to be successful, primate tourism must ensure the safety of the tourists and minimise the costs to animal welfare. Therefore, it is important to explore people’s ability to assess Barbary macaques’ emotional states correctly based on their facial expressions. Such research is crucial to help inform policy-making and guide the regulation of primate tourism as well as, inform management decisions in tourist destinations, for example, to assess whether and how close tourists might approach macaques to reduce the risk of injury.

Participants

In order to test our hypotheses, we collected data on how three groups of participants (N = 124) rated the emotional state of a series of macaque facial expressions: (1) naïve participants (N = 34, 10 men and 24 women); (2) exposed participants (N = 46, 19 men and 27 women); (3) experts (N = 44, 12 men and 32 women). Participants were assigned to one of these three groups on the basis of the following criteria: (1) naïve participants were those who had never worked with nonhuman primates, but might have encountered them on rare occasions when visiting zoos, animal parks or in their natural habitat; (2) Exposed participants were naïve participants who we exposed to different Barbary macaques’ facial expressions during the experiment (see section on exposure procedure below); (3) Experts were participants with knowledge about macaque behaviour acquired while working with macaques over a prolonged period of time (at least two months) either in the wild or in captivity (e.g., researchers). Participants were grouped into four categories (18–25, 26–40, 41–60 and +61 years old), with the majority of participants being in the 26–40 years old category (naïve and expert: 72.7%; exposed: 46.6%). All age categories were collapsed together for the purpose of analysis. From a total of 285 participants, we excluded 161 participants from the analyses who: (1) did not fully complete the questionnaire (N = 102), (2) had less than two month experience with Barbary macaques (N = 11), (3) did not work with Barbary macaques but worked with other nonhuman primate species, as they could not be classified as either naïve or expert participants (N = 48). Naïve and exposed participants were recruited via social media, and were randomly assigned to each condition without demographic selection. Expert participants were recruited via social media and professional networks. The selection of participation inclusion was done after data collection, which explains the high rate of participants excluded from the analyses presented below.

Stimuli

Each picture showed a face of an adult male or female Barbary macaque. A total of 22 pictures were used in the study, including 16 pictures used for the questionnaire (i.e., four pictures for each emotional state) and presented to every participant, and a series of six additional pictures that were presented only to exposed participants (see exposure procedure, Fig. 1). The pictures used in the present study were high resolution 2D pictures, and came from different sources who all gave permission to use their pictures. All pictures were independently labelled by two experts on Barbary macaques (LM and BM). Of the 22 pictures used, the facial expression and underlying emotion depicted in 16 pictures was independently confirmed by two experts on Barbary macaques (LM and Prof. Julia Fischer). Also, in order to have four pictures in each category, we added six pictures taken by professional photographers (Roger Eritja and Andrew Forsyth), for which the context of the picture was not known. These pictures were selected only if they contained clearly recognisable emotions, i.e., aggressive and neutral, and presented no ambiguity in their interpretation by the experts (LM and BM).

We selected six facial expressions associated with four common emotional states described below. The detailed description of the affiliative, submissive and threat faces can be found in Hesler & Fischer (2008).

(A–B) ‘Aggressive’ or ‘Threat’ face: In the first picture (A), the eyebrows are raised, the animal stares intently and the mouth is open showing the teeth. In the second picture (B), the eyebrows are raised, the animal stares intently and the lips are protruded to form a round mouth;

(C–D) ‘Distressed’ or ‘Submissive’ face: In the first picture (C), the mouth is widely open, and the animal is yawning. Yawing can be related to distress and anxiety in primates (Maestripieri et al., 1992). In the second picture (D), the corners of the lips are fully retracted and the upper and lower teeth are shown;

(E) ‘Friendly’ or ‘Affiliative’ face: In picture (E), the mouth is half open and the lips slightly protruded. This expression involves a chewing movement and clicking or smacking of the tongue and lips;

(F) ‘Neutral’ face: In picture (F), the mouth is closed and the overall face is relaxed.

Each picture of the Barbary macaques’ facial expressions was chosen to show both eyes, the nose and the mouth. This was done so that participants had access to the same amount of information on each image to be able to determine the macaque’s emotional state. Head orientation (horizontally or vertically) could not vary by more than 90° (e.g., Fig. 1), and all pictures were taken in natural daylight and with natural background slightly visible (see details in Supplemental Information 3).

Procedure

Statistical analysis

We used a generalised linear mixed model (GLMM) to investigate whether participants’ ability to recognise each macaque’s facial expression was predicted by their experience (i.e., between subjects: expert, exposed, naïve) or the type of emotional states (i.e., within subjects: aggressive, distressed, friendly or neutral). The dependent variable, participants’ ability to recognise facial expressions, was binary (i.e., yes or no). The independent variables, i.e., experience and emotion, were categorical. We included as random factors the participants’ and pictures’ identification numbers. The GLMM was run using R 3.1.3 (R Development Core Team, 2015), using the function glmer of the R-package lme4, family = “binomial” for binomial mixed-effects models (Bates & Maechler, 2010). The significance of the individual independent variables was determined based on the z- and p-values provided by lmer. The significance of the full model was compared to the corresponding null model, containing only the dependent variable and the two random factors, using a likelihood ratio test (R function ANOVA, Bolker et al., 2009). We checked that the model did not violate any assumptions of collinearity, with all variance inflation factors <2. We tested the over dispersion of the data and we accepted the model if the result was equal to 1.

We then calculated the percentage of correct choices for each participant across the four emotions. These data were not normally distributed (i.e., Kolmogorov–Smirnov tests were all p < 0.05); therefore, we used non-parametric Mann–Whitney U tests. We ran a series of pairwise comparisons between the three categories (naïve, exposed and experts) for the percentage of correct choices and for each type of emotions using SPSS v22 (©IBM).

Finally we described the choice made by the participants for the aggressive and distressed pictures using a confusion matrix. We ran a series of Chi Squared tests and Mann–Whitney U tests to assess whether participants confused aggressive and/or distressed pictures with either neutral or friendly faces based on their experience. These analyses were run using SPSS v22 (©IBM).

Applied perspective: implications for primate tourism

Even though it is commonly prohibited, interacting closely with wild animals, such as feeding and physical contact, is a very popular activity in wildlife tourism, and can also occur in zoos or animal parks despite the enclosure. However, such interactions present a potential risk of injury for both tourists and the animals involved (Zhao & Deng, 1992; Paterson & Wallis, 2005; Maréchal et al., 2016b). Therefore, the ability to recognise the emotional state of an animal people interact with, can reduce such risk. This is particularly important for aggressive or distressed faces, which, if ignored, are more likely to trigger inappropriate interactions and hence potential injuries.

Short exposure to pictures of Barbary macaques’ faces increases the ability of the participants to recognise monkey emotion. However, the performance rates of exposed participants were under 50% chance levels when assessing aggressive, distressed and friendly faces. There was also a significant difference between experts and exposed participants in discriminating aggressive faces. Taken together, these findings indicate that simply exposing people to Barbary macaque facial expressions and their corresponding emotional states using pictures on a board as is routinely done in a wildlife tourism or zoo setting, might not be a sufficient educational tool to reduce misinterpretation of an animal’s emotional state. The presentation of different facial orientation might make the recognition more difficult. Therefore, people should be exposed to similar stimuli taken from different angles to evaluate whether increased experience on different tilted faces improve their recognition abilities. As previously stated, learning might also be improved by showing a sequence of behaviours either live or on video, as the movement or context in which facial expressions are displayed can give additional clues that can help people to discriminate macaques’ emotional states.

Participants confused aggressive faces with ‘distressed,’ neutral and friendly faces; this was particularly true for exposed and naïve participants. Experts made 20.2% mistakes in interpreting aggressive facial expressions, whereas exposed and naïve people were even more likely to misinterpret aggressive faces (exposed: 51.4%, naïve: 60.2%) which might lead to situations where inappropriate behaviour towards the macaques is shown, increasing the risk of injury. Anecdotally, one of us (LM) often heard tourists in Morocco saying that the monkey seemed to blow them a kiss when they actually displayed a threatening face (Fig. 1B). The tourists often responded by imitating the monkey’s facial expression, which generally ended by either an aggression given by the monkey towards the tourists or the monkey leaving the interaction. Such a misunderstanding of Barbary macaque emotion can be related to an interpretative anthropomorphism, when people attribute intentions and emotions to nonhuman animals based on their behaviour (Fisher, 1991). This is particularly common for species that are phylogenetically close to humans, such as nonhuman primates (Urquiza-Haas & Kotrschal, 2015).

All participants confused distressed faces with other emotional states. In all the distressed faces presented, the teeth were visible with the lips fully retracted, with or without an open mouth. Showing teeth with the lips fully retracted is often considered as threatening in a number of animal species such as dogs (Tami & Gallagher, 2009), rodents (Zenuto, Vassallo & Busch, 2001), or as angry in humans (Pinkham et al., 2010). Therefore, this might explain why the participants confused distressed faces for aggressive faces. In addition, all participants have also mistaken distressed faces with neutral or friendly faces. We included in the questionnaire a picture of a yawning macaque, which we classified as distressed face. Variations in rates of yawning as well as other self-directed behaviours, such as self-scratching, have been found to reflect animal emotional state such as anxiety or frustration, generally displayed during stress-related situations (Maestripieri et al., 1992). Pharmacological and behavioural studies lend strong support to this idea (Schino et al., 1996; Castles, Whiten & Aureli, 1999; Troisi, 2002). For instance, increasing rates of self-scratching were positively related to interacting with tourists in wild male Barbary macaques, suggesting that male macaques had higher levels of anxiety when interacting with tourists (Maréchal et al., 2011; Maréchal et al., 2016a). However, the participants might not have been aware of the association between yawning and self-directed behaviours, and distressed emotional state in animals, which would explain this result. They might also have mistaken the showing teeth for a “smile” as displayed in humans (Preuschoft, 1992; Preuschoft & Van Hooff, 1997). This mistake is also observed in children who mistake a dog’s display of teeth as a happy and smiling facial expression (Meints, Racca & Hickey, 2011). However, being able to recognise such distressed faces or behaviour in animals, as well as knowing in which contexts tourists might elicit such facial expressions, are important in order to propose appropriate measures that reduce the costs for animal welfare, while improving tourist experience. In humans, when confronted with fearful faces, participants were more likely to approach than avoid the stimuli (Marsh, Ambady & Kleck, 2005; Adams et al., 2006). Only one study has so far been conducted using dog faces as stimuli (Meints, Racca & Hickey, 2011), but none using nonhuman primates. However, there is anecdotal evidence that shows tourists sometimes initiate an interaction with wild animals because they perceive them in distress, and this may lead to a worse situation for the animals. Future research should explore human responses to animals perceived as distressed or fearful, and educational measures should be initiated at tourist sites to ensure that people do not approach such animals in order to reduce the risk of injury.

Overall, the present study found that experience helps in correctly decoding Barbary macaque facial signalling, but a simple exposure to such stimuli using pictures is not efficient enough to significantly improve the performance of exposed and naïve people above chance (50%) for aggressive, distressed and friendly faces. This is an important issue, because these people are likely to be the tourists who encounter or interact with Barbary macaques in different wildlife tourism settings. Therefore, we propose some measures that aim to mitigate the potential risks for the tourists who have low understanding of animals’ behaviour and/or facial expressions. Actual training of tourists or guided visits by experienced professionals may be required to acquire a recognition rate that would effectively reduce misinterpretation of primate emotion based on facial expression, and thus decrease the risk of injury. These guided visits by local professionals can also benefit the local economy and increase public awareness, leading to a better conservation of the species involved. A successful example of the positive impact of guided visits by local experts can be found in gorilla tourism (Williamson & Macfie, 2010; Nielsen & Spenceley, 2012).

It would also be interesting to test in future research whether a more interactive training such as video or participative games increase the ability of inexperienced people to assess Barbary macaque emotion. Extensive information of an animal’s emotional state can also be conveyed by posture and vocalisations (Seyfarth & Cheney, 2003; Tami & Gallagher, 2009); therefore, including other sources of information in addition to facial clues might increase the performance of inexperienced people. Interactive training would be useful particularly for zoo settings. However, such a recommendation is unlikely to be feasible in a number of wildlife tourism settings; therefore, we suggest in addition to guided visits, regulating the distance between the tourists and wild animals as keeping a safe distance will reduce such risks. In addition to increased safety for tourists, these recommendations can improve animal welfare. Indeed, it has previously been found in a number of species that the presence, proximity and/or interactions with tourists can negatively affect animal welfare by increasing their physiological stress levels (e.g., Magellanic penguin, Spheniscus magellanicus: Fowler, 1999; Barbary macaque: Maréchal et al., 2011; western lowland gorilla, Gorilla gorilla gorilla: Shutt et al., 2014), risks of injuries related to higher aggression and conflicts with conspecifics (e.g., stingray, Dasyatis americana: Semeniuk & Rothley, 2008; Barbary macaques: Majolo et al., 2013) and tourists (e.g., Tibetan macaques, Macaca Thibetana: Zhao & Deng, 1992; stingray: Lewis & Newsome, 2003). Therefore, to reduce human-animal conflicts, preventing inappropriate behaviour from tourists towards animals is crucial and this improves animal welfare and tourist experience simultaneously. Understanding how people perceive animal emotion and behaviour, and how they respond to them will be key in future wildlife and zoo tourism research.