Monkeying around: Non-human primate behavioural responses to humans reproducing their facial expressions

People are often observed mimicking animals’ facial expressions in an attempt to communicate with them. However, to date, there is limited understanding of how animals respond to humans reproducing their facial displays, or if this type of human behaviour presents a risk for either human safety or animal welfare. In the present study, we investigated how Barbary macaques (Macaca sylvanus) responded to pictures of humans and conspecifics displaying their facial expressions. Macaques viewed pictures of neutral, mildly threatening or highly threatening human or macaque faces. We recorded aggressive, submissive, and self-directed behaviours exhibited by individuals while in front of each stimulus. Macaques displayed more self-scratching toward human neutral face compared to the corresponding macaque face, and vice versa for the mild threat. They also exhibited more aggressive behaviours toward macaque neutral and mildly threatening stimuli compared to the human stimuli. However, macaques did not display any more submissive behaviour between human and macaque for any facial stimuli. There was also no significant difference in behavioural responses for highly threatening faces between species. These findings suggest that while the reproduced facial expression by humans might carry meaning for macaques, their responses vary between species. Therefore, these results highlight some potential issues for facial signalling (mis)communication between species, which has implications for animal welfare and


Introduction
Facial expressions provide crucial information allowing individuals to navigate social interactions (Fridlund, 1994;Waller et al., 2017;Waller and Micheletta, 2013). Therefore, correctly interpreting facial expressions is key for interspecific communication and interaction. Previous research in human-animal interactions showed that both humans and animals can use facial expressions as communication cues between species to adapt their behavioural responses (Langford et al., 2010;Gavin et al., 2015). For instance, some non-human animals are able to distinguish the valence of human facial expressions (e.g. dog, Canis lupus familiaris: Albuquerque et al., 2016;horse, Equus caballus: Smith et al., 2016;Giant Panda, Ailuropoda melanoleuca: Zhang et al., 2017). Animal facial expression information is often used as an indicator of animal welfare by humans (Langford et al., 2010) and can be used in our decision to approach or interact with animals. For example, humans are more likely to approach dogs displaying non-threatening facial displays than when displaying threatening faces (Gavin et al., 2015). However, previous research has shown that people are often poor at judging animal perceived emotional states based on their facial expressions and consequently, might not be able to accurately predict their subsequent behaviour Meints et al., 2018). For instance, non-experienced humans often misinterpret Barbary macaques' mildly threatening facial expressions as friendly . This raises serious concerns for human safety and animal welfare as this poor judgement can lead to aggressive interactions between species (Maréchal, personal observations).
In interspecific interactions, humans seem to reproduce animal behaviour (i.e., gesture and facial expressions) in an attempt to communicate and foster a connection with the animals they encounter.
For example, Persson et al. (2018) reported that visitors imitated chimpanzees' behaviours (Pan troglodytes), mainly reproducing manual actions on windows (e.g., hit, knock, press, and kiss-like gestures). Moreover, tourists have been observed mimicking macaques' facial expressions during human-macaque interactions at a tourist site in Morocco . Tourists were seen mimicking facial threats (e.g., open-mouth faces) towards macaques thinking they were reproducing friendly expressions to initiate a friendly interaction with macaques . This misunderstanding in facial signalling, and its reproduction can result in an escalation of aggressive exchanges between species which can have dire consequences for both parties, such as physical injury and pathogen transmission (Wallis and Rick Lee, 1999;Carne et al., 2017). However, to date, little is known about how animals respond to humans reproducing animal facial displays.
Facial expressions are the result of a set of movements of the facial muscles. Based on previous research, facial muscle morphology is similar among several species in the primate order, which enables humans to reproduce other primates' facial expressions such as macaques (Burrows, 2008). Therefore, previous research suggested that macaques view humans reproducing macaques' facial expressions the same way they view other macaques making those same expressions (Paukner et al., 2007;Goossens et al., 2007). Indeed, these studies used humans reproducing macaques' facial expressions as stimuli to assess gaze following in pigtailed macaques, Macaca nemestrina, and longtailed macaques, Macaca fascicularis, assuming macaques would respond similarly as if it was a conspecific (Paukner et al., 2007;Goossens et al., 2007). It was found that, for instance, human gaze accompanied by silent bared-teeth, a facial expression often leading to affiliative contact, elicited stronger gaze following responses than neutral gaze as expected with macaque stimuli (Goossens et al., 2007). The authors concluded that, although the macaques responded to stimuli, and so, the facial movements might carry meaning for them, their responses might have differed if the stimuli were conspecifics rather than humans. However, to date, no study has explored whether primates might respond differently between humans and conspecifics displaying the same facial movements.
The aim of the current study was to explore primate behavioural responses to humans recreating their facial expressions. We investigated whether Barbary macaques displayed aggressive, submissive, selfdirected or look behaviour as well as time spent in proximity of the stimuli when presented with a picture of a human or macaque face, either neutral, or threatening (mild or high) for up to 5 min. It was predicted that similarity in facial expression between macaques and human faces would elicit similar rates of behavioural responses following the gradient of threat displayed in the stimuli (Whittaker and Knight, 1998;. Therefore it was predicted that (1) neutral faces might cause a low level of anxiety despite not being threatening as all stimuli were males, which has been shown to elicit an increase in self-scratchinga behavioural indicator of anxiety-in primates (Maestripieri et al., 1992;Troisi, 2002), (2) mild threating faces would elicit higher rates of aggressive or/and submissive behaviours (Preuschoft et al., 1998), and (3) highly threatening faces would elicit more submissive behaviours including avoidance (Whittaker and Knight, 1998). In addition, it is predicted that there will be no significant difference in behavioural responses between species for all three stimuli. Encounters with wild primates such as Barbary macaques are a growing tourist activity. Therefore, as monkey bites are the second most common animal bite risk to travellers after dog bites (WHO, 2018), and tourist behaviours can negatively affect primates' behaviours, physiological stress and anxiety (Maréchal et al., 2011, understanding how different species perceive communication clues is essential for both human safety and animal welfare.

Study site and subjects
The project was conducted in the wildlife park "Terre de singes" in Lumigny-Nesles-Ormeaux, France, where 45 free-ranging Barbary macaques lived in an enclosure of 15 ha of open grass and woodland. The park is part of the European Association of Zoo and Aquaria and follows the European animal welfare regulations for keeping non-domestic animals. The macaques originated from four groups corresponding to their zoos of origin: group α, coming from Apenheul zoo (Netherlands); group β, from Beauval (France); group ε from Erfurt (Germany); group, µ, from Montpellier (France), the latter composed of animals rescued from animal traffic. Animals were all fed in the morning at 9 am, and in the evening at 5:30 pm, with fruits and vegetables distributed to avoid competition and fights between groups. Throughout the day, they also received some seeds and primate pellets. In addition, there were a maximum of six presentations to tourists throughout the day, at 11 am, 11:45 am, 2 pm, 2:30 pm, 3:15 pm and 4 pm, one presentation per group during which one bucket of fruits was distributed to the macaques. Data were collected on 40 individuals (18 adult males, 11 adult females, 6 sub-adult males, 4 sub-adult females and 1 one-year-old male infant) from the four groups, 5 individuals did not participate in the experiment because they never looked in the direction of the stimuli.

Ethics
The study was approved by the University of Lincoln's Ethics Committee (reference: LEAS430). The study followed the Animal Behaviour ethical guidelines for animal research (Behaviour, 2018). Due to welfare guidelines and as we aimed to test macaques who experienced regular and varied human encounters, we choose a free-ranging group exposed to tourists. For these reasons, the macaques could not be isolated to participate in the experiment, and thus spontaneously participated in the experiment while in their normal social setting. For safety reasons, pictures of neutral and threatening stimuli were used instead of using live human facial reproduction.

Apparatus
The experimental set-up consisted of presenting one picture of a macaque or human face (i.e., stimulus) at a time to an individual macaque. The stimulus was placed in a wooden structure closed on the sides (4 wooden planks 1.8 cm thick, and measuring 50 cm in length, 30 cm in width and 30 cm in height) and open on the top and front (Fig. 1). The stimulus was protected between two clear acrylic sheets of A4 size. This structure (hereafter, box) made it possible to limit the field of vision to the stimulus and therefore to place the image more precisely in front of a target individual and then know whether the target was looking at the image. It also prevented macaque who might not be in front of the image from seeing it.

Stimuli
The stimuli were composed of photographs of three adult male macaques (unknown to the macaque participants), one neutral face, one open-mouth threat (mild-threat), and one scream-fight threat (highthreat), and a human (adult male unknown to the macaque participants) recreating these expressions (Fig. 2). For the human stimuli, a series of photographs was taken of 5 participants (4 men and one woman) with no experience with Barbary macaques. They were asked to imitate the facial expressions of Barbary macaques from the photographs described above . All humans were photographed by the researcher, from the front, eyes looking at the camera. Photographs were taken with a Canon SX540 HS PowerShot camera. Then, each photograph was coded using FACS (Human: Ekman and Friesen, 1978; Barbary macaque: Julles-Danière et al., 2015). The Facial Action Coding System (Ekman and Friesen, 1978) is an anatomically-based, objective methodology used to study facial expressions based on the description of muscle movements, called Action Units (AUs). It was first developed for humans Friesen, 1978, Ekman et al., 2002). The FACS for macaques (MaqFACS) was created in 2010 (Parr et al., 2010) for rhesus macaques, Macaca mulatta, but has since been validated for Barbary macaques (Julle-Danière et al., 2015). To reduce facial feature variations between the participants presented (Clark et al., 2020), we selected the photographs of only one male participant whose expressions were the most similar to the macaques (neutral, NE: both UA 0; open-mouth threat, OM: human: UA 18 +25 +26 and macaque: UA1 +2 +8 +18ii +25 +26; scream-fight threat, SF: human: AU6 +10 +12 +16 +25 +27 and macaque: AU6 +9 +10 +12 +16 +25 +27). For the macaque stimuli, due to limited stimuli available for each facial display, we selected the images with the highest quality from Brian Gomila (Monkey talk Gibraltar) and images freely available on Google Image. The images of the selected macaques are either front-facing or at an angle of less than 45 • . The photographs were then edited with Adobe Photoshop software (version 20) to contour the face, neck and shoulders with a white background, and for the human pictures, adjust the colour of the clothing to black. All images were in A5 (148 × 210 mm) size, with a length:width ratio of 4:3, centred on a white A4 (210 × 297 mm) background. Stimuli were laminated after printing.

Procedure
The data collection was carried out from 25th February to 18th May 2019. The box was placed 5 m in front of the macaque targeted for testing. From the moment the tested macaque could see the stimulus, the  experimenter recorded the interactions for a maximum of 5 min with a GoPro Hero 5 session camera. If a macaque left the area (over 10 m from the stimulus or could not see the stimulus) before 5 min, the experimenter continued recording, in case the macaque came back. However, only behaviours occurring within 10 m of, and in front of, the stimuli were reported. The camera was held one meter behind the box at onemeter height, held by the researcher who had to stay in a sideway position and never look directly to the macaque to reduce the researcher's impact on the macaque behaviour (Fig. 1). Prior to exposing the macaques to the stimuli, the macaques were habituated to the researcher and to the box, first empty, then with pictures of an object (drawing of a bike) and fruit (strawberries and bananas), over a two-week period. Due to time constraints, we chose to first present the OM and NE stimuli in a randomised order to the 40 individuals tested. This ensured that all stimuli were presented to all individuals tested. After all OM and NE stimuli presentation, as time permitted, we presented the SF stimuli to 20 individuals randomly selected, and who first completed the OM and NE tests. The presentation order of the human and macaque SF stimuli was randomly selected for each macaque. Macaques were not exposed to more than 2 stimuli per day. In fact, only 4 out of 40 individuals were exposed to two images on the same day (two hours apart minimum), the rest were only exposed to one stimulus in any particular day. The testing was done in open space, in the absence of rain.

Behavioural data
The videos were used to measure behavioural responses to each stimulus and were obtained with the GoPro were 960 × 1280 pixels in size with 960p video resolution. Video analysis was done using BORIS software (Friard and Gamba, 2016). For each video, we recorded any behaviour when the monkeys were within 10 m of, and in front of, the stimuli (see close variable in Table 1) to ensure that these behaviours were directed at the stimuli, or produced in response to the stimuli. The behaviours recorded were as follow (Table 1): (1) aggressive behaviour, e.g., facial threats, lunge; (2) submissive behaviour, e.g., teeth-chattering, retreat; (3) self-directed behaviours, i.e., self-scratching behaviours, as an indicator of anxiety (Troisi, 2002). We also measured the duration the macaques stayed within 10 m of, and in front of the stimuli, (i.e., Close variable see Table 1) and the time spent looking at it. In addition, to control for factors which might influence the behavioural response, we recorded through direct observation during each test the following information: whether any humans (i.e., staff or tourist) were in proximity (within 10 m), any macaques were present within 10 m and their identity, as well as whether food distributed by the staff was in proximity (within 5 m). An inter-reliability check was conducted on 40 out of 200 videos (20%) between the two researchers who extracted the behavioural data (Cohen's Kappa 0.875).

Assessing macaque rank
We assessed the dominance rank of the macaques to be included as a control variable when looking at the influence of the proximity of other macaques during testing. To determine their dominance rank, we calculated the hierarchical ranks of each macaque using the Elo-Rating method (Neumann et al., 2011). Elo-Rating uses the succession of victories or defeats over time and assigns a score to each individual after each conflict. We collected agonistic behaviours when a conflict occurred between two individuals to determine the 'winner' and the 'loser' of the dyadic encounter. These data were collected using 10 min continuous focal and ad libitum samplings throughout the study period over 160 h (Altmann, 1974).

Data analysis
First, we calculated the frequency of each behaviour (i.e., aggression, submissive and self-directed) by dividing the number of behaviours observed by the duration the monkeys spent within 10 m, and in front of the stimuli. As it was not possible to run statistical comparisons due to the high risk of type II error associated with the low frequencies of behaviour displayed and the large number of stimuli categories to be compared (Bonferroni correction issues), we transformed the frequency as percentage to present the descriptive data. Descriptive data were used to evaluate whether the behavioural responses (i.e., self-scratching, aggressive and submissive behaviours) followed a gradient of threats (i.e., neutral, mild threat and high threat).
We used a generalised linear mixed model (GLMM, Bolker et al., 2009) to investigate whether macaques' behavioural responses were predicted by the stimulus shown, while controlling for different factors that could influence these behavioural responses.
All the analyses were run using R-studio with R version 3.6.0 (RStudio Team, 2020). As the frequency of each behaviour (aggressive, submissive, and self-scratching behaviours) was low and some variables had high number of zero, we transformed the dependent variables as binomial such as > 0 = yes and 0 = no. We used the function glmer of the R-package lm4, family= 'binomial'. For the continuous dependent variables (durations spent within 10 m of the stimulus 'close' and staring at the stimulus 'look') we used the function lme of the R-package nlme, method: 'ML'. As fixed effects, we included the main predictor (type of stimulus) and added some control variables in our models such as macaque sex and rank (categorical variables), the presence or absence of food, other macaques, humans, and dominant macaque (binomial variables: yes or no) and order of appearance of the stimulus and age of the macaque (discrete variables). From the full model, including all the fixed effects, we selected the best fit model using the backward stepwise selection methods, the one with the lowest AIC and BIC values, and retaining the stimuli as predictor by default. The backward stepwise selection approach begins with a full model (including all variables) and at each step gradually eliminates variables from the model to find a reduced model that best explains the data. In our best fit model, only human presence, order of the stimuli and food presence were retained as control variables. All other variables were not included in the best fit model. The significance of the best model was compared to the null model, containing only the dependant variable tested, using an ANOVA test (R function ANOVA, Bolker et al., 2009), and none of the models Table 1 Ethogram used for the video analysis on Boris. Aggressive, submissive and selfdirected behaviours were recorded as events (i.e., an event concluded when the behaviour stopped), and attention was recorded as state (i.e., duration).

Category Behaviour Description
Aggressive Lunge Abrupt movement of the head, sometimes also the whole chest, towards the stimuli and raising of the bow line Open-mouth Mild-threat facial display: round and open mouth, and brow line raised Scream face Highly threat facial display: wide opened mouth, teeth are exposed, usually followed by a scream Submissive Bared-teeth Teeth exposed by open-lips and closed jaws Teethchattering Quick clapping of the teeth, snubbed lips

Retreat
Slowly Moving away, e.g., walking, from the stimuli after seeing it Flight Rapidly moving away, e.g., running, from the stimuli after seeing it Crouch Macaque's whole body is against the ground, in front of the stimulus (not resting) Self-directed behaviour

Selfscratching
Macaque rubs its own body with its nails

Attention
Looking Time spent with the head and opened eyes positioned in the direction of the stimulus Close Time spent in proximity of the stimulus (10 m), in a zone where the picture is visible by the macaque. The behaviour started from the first look at the stimulus violated any assumptions of collinearity, with all VIF < 10.

Descriptive comparison of behavioural responses to different threat intensity stimuli
Following the prediction, macaques exhibited more self-scratching (47.0%) in front of a human neutral stimulus and compared to any human threat stimuli (Fig. 3). However, this was the opposite for the macaque stimuli, with higher self-scratching rates in front of the highly threatening macaque stimulus (56%) compared to mild and neutral stimuli. Macaques displayed aggression (ranging between 42.2% and 49.9%) for any human face stimuli. However, macaques had a higher percentage of aggressive behaviours towards neutral and mildly threatening stimuli (over 40%) compared to highly threatening macaque stimulus (20.9%). Furthermore, macaques displayed more submissive behaviour towards human mild and high threats compared to the neutral face; whereas macaques exhibited more submissive behaviour toward the neutral macaque stimuli compared to the other stimuli (Fig. 3). For additional details about mean behavioural response frequency, see Table 2.

Comparison of behavioural responses to human versus macaque facial stimuli
There were no significant differences in behaviours displayed by Barbary macaques (i.e., aggression, submission and self-scratching) in response to human and macaque high threat faces (Table 3). In addition, there were no significant differences in submissive behaviour displayed between any facial expressions of human and macaque presented (χ df = 6, χ 2 = 4.908, P = 0.555). Furthermore, the different facial expressions in both species did not predict how long macaques looked at the stimuli (L Ratio = 9.289, P = 0.0981), or spent time in front of them (Table 3).
The Barbary macaques were more likely to display aggression when presented with a macaque face either neutral or mildly threatening, compared to human faces with the corresponding facial expression. They were also more likely to self-scratch when in front of a mildly threatening macaque face compared to a human mildly threatening recreated face. However, Barbary macaques were more likely to display more self-scratching when presented with a human neutral face compared to a macaque neutral face.

Discussion
This study aimed to explore primate behavioural responses to humans reproducing their facial expressions. Overall, macaques do not seem to respond using a behavioural gradient related to the threat intensity of the facial display presented. As such, the descriptive data showed that macaques exhibited higher percentage of self-scratching toward human neutral stimulus and toward highly threatening macaque faces, while they exhibited lower percentage of aggression toward high threat macaque face. Macaques also showed lower percentage of submissive behaviours towards human neutral face and macaque mild threat face. When comparing corresponding facial expressions between human and macaque stimuli, macaques were more likely to display aggressive behaviours in front of macaque neutral and mild threat stimuli. They were also more likely to display more scratching behaviours in front of mild threat macaque stimuli and human neutral stimuli. However, there was no significant difference in submissive behaviours, nor the time spent in front of or looking at the stimuli. Therefore, the results suggest that reproducing facial expressions might carry meaning for primates, but their responses differed between conspecific and human stimuli.
Animals exhibit different behavioural responses when exposed to  different stressors, which can be categorised from avoidance, aggression, habituation to being attracted to them (Whittaker and Knight, 1998;. In addition, previous study showed that in response to visitors' threat, Formosan macaques displayed more submissive and avoidant behaviours than aggressive behaviours (Hsu et al., 2009). In the present study, there was no clear evidence that macaques responded following the gradient of threat displayed in the stimuli as predicted. As such, while human neutral faces presented higher percentage of self-scratching, percentage of aggression were similar for all human stimuli, and percentage of submissive behaviours were lower for human neutral faces than mild and highly threatening faces. However, it was the opposite for the macaque stimuli. There were higher percentage of self-scratching and lower aggression for highly threatening stimuli compared to neutral and mild macaque threats, and higher percentage of submissive behaviours toward the neutral macaque face compared to the more threatening faces. These results suggest that macaques respond differently between human and macaque faces. Although these descriptive data must be taken with cautious as no statistical analysis was conducted, these findings provide support that further research is needed to better understand how primates might respond to humans reproducing their facial displays. It also highlights the importance to be cautious when using human as proxy to other primate stimuli in cognitive experiments (Paukner et al., 2007;Goossens et al., 2007), especially as human reproduction of other primates' facial movements might not be accurate as tested via maqFACS in our participant poll.
In the present study, macaques displayed higher rates of selfscratching behaviours but lower aggression in front of a human neutral face compared to the corresponding macaque facial expression. Animals exhibit higher rates of self-scratching behaviours as a behavioural coping mechanism to deal with stressors (Maestripieri et al., 1992). For example, primates use self-scratching behaviour to cope with the uncertainty associated with tourist interactions and proximity (Maréchal et al., 2011. Evidence also suggests that primates use aggressive behaviour as a short-term coping mechanism in response to mild stressors (Gustison et al., 2012). Therefore, if we follow the behavioural response gradient to threats (e.g., self--scratching<aggression<avoidance), the findings suggest that human neutral faces might be perceived by the macaques as less stressful than a neutral macaque face. As the maqFACS scoring was identical between both neutral stimuli, we suggest that the difference in behavioural responses is associated with either species or stimulus characteristics. Future research should include more variations into the stimuli presented, including for example, different age/sex classes to (1) assess the generalisation of these findings, and (2) be able to test whether species or stimulus characteristics influence animal behavioural responses. Macaques had higher rates of self-scratching and aggression in front of a mildly threatening macaque stimulus compared to a mildly threatening human stimulus. Similar to the neutral stimulus, macaque mildly threatening stimulus might be perceived as more stressful than the human stimulus. This difference might also be due to species or stimulus characteristics. However, the difference might also be explained by the lower maqFACS accuracy in the facial reproduction compared to the other stimuli (i.e., neutral and high threat). Indeed, the mildly threatening facial expression in Barbary macaque, or open-mouth threat, is often misinterpreted by humans as a non-aggressive facial display . This might explain why none of our models were able to accurately mimic the open-mouth expression of the Barbary macaque, particularly around the eye area. To be representative of possible tourist macaque mimicking observed in the field , we have selected the stimulus to be the closest possible of the maqFACS scoring (Julle-Danière et al., 2015), although this stimulus presented some differences in the muscles solicited around the eye areas. Therefore, it is possible that the macaques perceived these differences in the human stimulus compared to the corresponding macaque face. Primates have been showed to present higher gaze fixation when viewing human faces, with higher percentage of fixations around the eyes, followed by the nose and the mouth (Guo et al., 2019). This might explain the difference in macaque behavioural responses to the human stimulus in the present study. Future research should include variations in animal facial reproduction to better understand whether the accuracy of the reproduction has an impact on animal behavioural responses.
There was no difference in primate behaviour responses between human and macaque highly threatening stimuli. Human scream face reproduction appears to be perceived by the macaques similarly as the macaque scream face, suggesting possible shared facial signalling for interspecies communication in primates. However, the eye gaze of highly threatening macaque stimulus was not fully facing forward compared to the other stimuli, which could have reduced the threat perceived from the stimulus. Previous studies have found that young rhesus macaques respond to human eye contact with treats, compared to no-eye contact for which macaques tend to 'freeze' (Kalin and Shelton, 1989). Other studies also suggested that individual factors such as temperament, sex, age and context might influence macaques' responses to human eye gaze, and so, to different levels of threats (Hamel et al., 2017). However, in our study, the dominance status, sex, and age of the macaques were not selected in the best fit models, meaning that these factors were not among the main predictors influencing macaques' behavioural responses. All these factors (eye gaze and individual attributes) might present some confounding variables, which not all were possible to control for in this study. Therefore, our results must be taken Table 3 GLMMs results of the relationships between macaques' behavioural responses and animal recreation stimuli. with caution, and further testing controlling for additional individual factors should be conducted to ensure the generalisation of these findings.
Overall, the findings suggest that human facial reproduction might carry some meanings for the macaques as the human stimuli elicited some behavioural responses. However, it is unclear how macaques perceived these stimuli as their responses seem to differ between human and conspecifics. Therefore, the reproduction of animal facial signalling might be interpreted differently from their own species facial expression, either because of the species difference or stimulus characteristics. However, these findings must be interpreted with caution because only one human male model was used for this preliminary study. Previous observations have shown that adult male primates were more likely to display aggression towards women compared to men (Fuentes and Gamerl, 2005). Therefore, other individual characteristics such as facial features, age and gender might influence primate responses. Further research using different human models is needed to assess whether stimulus characteristics might affect primate behavioural responses similarly to conspecific characteristics (Maestripieri, 2005;Micheletta et al., 2013). In addition, the study used pictures to explore how primates responded to human and macaque' s facial displays for practical (better control over facial expression accuracy displayed using maq-FACS) and ethical reasons. However, future research might consider using conspecifics and additional experimenters trained to mimic the facial expressions with different levels of accuracy as tourists would, or, to use naturally occurring situations where tourists mimicked the facial expressions of the primates to explore whether such behavioural responses are also found.
In this study, macaque displayed over 40% of aggressive behaviour towards all human stimuli, highlighting the concerns for human safety when mimicking animal facial expressions. This is particularly problematic as humans often misinterpret facial expressions , which might escalate aggressive interactions and result in physical aggression such as scratches or bites. Here, affiliative facial recreation was not presented. However, as previous studies have suggested, affiliative imitation within the same species might promote positive behavioural responses in nonhuman primates (Sclafani et al., 2015;Persson et al., 2018). This could be equally problematic, as close physical proximity between humans and macaques can increase the risk of pathogen transmission between both species (Wallis and Rick Lee 1999;Carne et al., 2017). Therefore, further research should also explore the use of affiliative animal facial reproduction to explore potential positive interspecific communication.
Overall, the findings suggest that such research has important implications for human safety and animal welfare, and informing the public about human-animal miscommunication, such as involuntarily mimicking animal threats might improve these issues. Finally, to date, this preliminary study is the first research exploring how animals respond to human reproducing their facial displays, and it provides support that further research into this topic is needed to better understand human-animal communication, including in companion, farm and other wild animals.

Ethics approval
All applicable institutional and/or national guidelines for the care and use of animals were followed. The study was approved by the University of Lincoln's Ethics Committee (reference: LEAS430). The study followed the Animal Behaviour ethical guidelines for animal research (Behaviour, 2018).

Approval of the submitted version of the manuscript
Please check this box to confirm that all co-authors have read and approved the version of the manuscript that is submitted. Signatures are not required.

Funding
The University of Lincoln provided funding for the equipment and apparatus.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.