Monkeys integrate facial expressions and direct gaze to modulate gaze-following behavior

11 For humans, being looked at directly can boost our readiness to follow another’s gaze, but can monkeys 12 invite an observer to engage with them in the same way? We trained three rhesus macaques on a head 13 gaze-following task in which the portrait of a demonstrator monkey would face the viewer before 14 turning to look at a distinct spatial target. The demonstrator could look at the viewer with his eyes 15 opened or closed and display different facial expressions. Unlike in humans, we found that direct gaze 16 alone, devoid of an accompanying spe cific expression failed to influence the latency of the subsequent 17 gaze-following response. However, when combined with threat , direct gaze significan tly accelerated 18 gaze-following. In a second experiment, we show that once turned away the expression associated with 19 prior direct gaze no longer mattered; instead a submissive facial expression accompanying the gaze shift 20 delayed gaze-following. Direct comparison of both experiments reveals that expressions accompanying 21 direct gaze trigger earlier gaze-following responses than the same expressions joining gaze aversion. 22 These results document the pronounced behavioral importance of the valuation of expressions signaling 23 danger, arguably creating heightened alertness in monkey observers, thereby priming their ga ze-24 following to allow for immediate conflict resolution. 25 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 21, 2025. ; https://doi.org/10.1101/2025.05.18.654731doi: bioRxiv preprint 1. Introduction 26 Humans are able to rely on many forms of nonverbal communication, which include facial expressions, 27 gestures, body poses or gaze to communicate their desires and intentions. When the eyes of the other 28 are gazing directly at an observer – a dyadic interaction usually referred to as “direct gaze” – they signal 29 the other´s intention to approach or interact with the observer (1,2). However, di rect gaze may also 30 serve as a means to intimidate and exert dominance, thereby maintaining distance and discouraging 31 approach. Hence, direct gaze can elicit strong arousal, and depending on the social context, may evoke 32 pleasurable or discomforting experie nces (3,4). An important potential consequence of direct gaze, 33 drawing the observer´s attention to the eyes of the other is to follow his/her gaze, in case the other may 34 decide to attend to a novel object of interest. By following the other´s gaze the observer becomes able 35 to identify this object and to establish a joint attentional focus on it (5–7). In humans, it has been found 36 that prior direct gaze from a human avatar elicits earlier subsequent gaze-following responses (8). This 37 effect is apparent even though the face avatar had neither identity nor expressions and came without 38 semantic context. Hence, in humans, direct gaze alone is powerful enough to prime us to follow the gaze 39 of the other more readily. 40 Rhesus macaques are also capable of gaze-following, which is recognized to be a fast, accurate, and 41 highly reflexive process that can be subject to cognitive control, sharing many similarities with human 42 gaze-following (9,10). However, the manner in which they put this into practice is still poorly 43 understood. It is unknown if macaques can use direct gaze to initiate ga ze-following like in humans. 44 However, as a relatively despotic species rhesus macaques may shy away from prolonged gaze as it can 45 signal threat and aggression (11). Meanwhile, regardless of clear species -dependent differences in the 46 semantics of facial expression s, nonhuman primates too rely on the meaning of facial expressions in 47 order to guide their interactions with conspecifics. For instance, facial expressions can facilitate gaze -48 following in Barbary macaques, which exhibit a ‘commenting e xpression’. This is produced when 49 observing others and has been found to promote gaze -following, informing an observing monkey of 50 further interactions taking place in the vicinity (12). This partially answers how a bout of gaze -following 51 may be kick-started but is still reliant on an observer witnessing the commenting gest ure by chance. 52 Moreover, rhesus macaques do not possess a commenting gesture. Other work in long-tailed macaques 53 reported that fear expression s accompanying a gaze -shift can boost gaze -following (13). Yet, the 54 problem of this work and others is that it studied monkey observers reacting to human demonstrators 55 (14,15), and monkeys have been found to not treat human and conspecific facial information in the 56 same way (16). 57 Champ and colleagues (17) explored behavioral reactions by having a rhesus macaque observer watch 58 two videos displayed side -by-side, showing a dominant and a submissive monkey respecti vely, in an 59 attempt to simulate two monkeys interacting. They reported an increase in the frequency of gaze -60 following responses in the interval following the meeting of the eyes. Furthermore, it seemed that 61 submissive expressions of the subordinate monkey increased the number of exploratory saccades made 62 by the observing monkey that landed on the dominant monkey. This suggests that submissive facial 63 expressions can potentially encourage gaze-following behavior due to the need to seek out the source of 64 threat. However, because the observer passively watched these interactions, it is unlikely that they felt 65 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 21, 2025. ; https://doi.org/10.1101/2025.05.18.654731doi: bioRxiv preprint part of the interaction, and therefore it is hard to predict the consequences of any coincidental direct 66 gaze. Finally, these direct gaze events were sporadically induced by the experimenters, and it is 67 unknown if they were mixed with other social cues . We do not believe rhesus macaques would leave 68 gaze-following to chance – so what would prompt a rhesus macaque to follow the gaze of another? We 69 suspected they may be able to resort to using direct gaze to initiate gaze -following like humans. 70 Moreover, it seems that macaques are able to distinguish between facial expressions directed towards 71 them versus those that are averted, although this has never been explicitly tested. 72 Great strides have been made in studying monkey social interactions in natural settings in a quantitative 73 manner. However, monitoring eye movements with sufficient reliability and precision has remained 74 difficult under such circumstances compared to the higher degree of quantification offered by 75 laboratory environments. Studying the interaction of facial expressions and gaze-following – the interest 76 of our study – in unconstrained settings is particularly challenging. The fragile and in mo st cases 77 unfavorable geometric relationship between the camera and interacting monkeys makes it 78 tremendously difficult to collect high resolution data on interesting combinations of expressions and 79 gaze behavior of the agents. Moreover, much of what we currently know about the effects of direct gaze 80 and facial expressions on gaze -following has been performed in humans, and the few studies done in 81 monkeys are little controlled field studies , exploring these events taking place by chance, or suffer from 82 the limitation that human demonstrators interacted with monkeys. Therefore, in an attempt to explore 83 the interdependence of direct gaze, facial expressions and gaze-following in monkeys in a more rigorous 84 manner and to avoid the pitfalls of unnatural, learnt interspecies interactions, we studied gaze-following 85 responses of monkeys looking at portraits of conspecifics presenting distinct facial expressions while 86 staring at the observer, followed by gaze shifts towards objects. Our results provide clear evidence o f 87 the behavioral impact of evaluating danger -related expressions in direct gaze, likely enhancing the 88 observer’s vigilance and preparing their gaze-following response for conflict avoidance. 89 2. Results 90 2.1 Behavioral Paradigm 91 The behavioral paradigms used in the present study are adapted from the gaze-following paradigm 92 which had previously been used in the discovery, exploration, and causal manipulation of the gaze -93 following patch (GFP) in macaques (10,18,19). In the first part of this study (Experiment 1), we were 94 interested in the effects of direct gaze and facial expression on gaze-following in macaques (Fig. 1A). The 95 averted head of a conspecific (averted gazer) was presented on a mon itor in front of our observing 96 experimental monkeys, and for each trial the observer was required to follow the head gaze of the 97 conspecific towards one of four possible targets presented in a row in front of the conspecific, and make 98 a saccade towards the target singled out by the conspecific’s head gaze. These averted portraits carried 99 a neutral expression. Importantly, prior to the gaze -following portion of the trial, the portrait of the 100 same monkey individual was presented in the center of the screen, i n the same location as the averted 101 gazer but here the monkey´s portrait (direct gazer) faced the monkey observer. The direct gazer either 102 looked straight ahead towards the observer (portraying direct gaze) with the eyes open or, alternatively, 103 closed. This portrait could also display a neutral, threatening (open mouth), or fear -grin (silent bared -104 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 21, 2025. ; https://doi.org/10.1101/2025.05.18.654731doi: bioRxiv preprint teeth) expression. All portraits, whether direct or averted, were prepared from photos of the same 105 three monkeys participating in this study. More details of their preparation are included in the Methods 106 section. Additional categories also used as the direct gazer included inverted and scrambled versions of 107 these stimuli (Fig. 1B). These two categories were added as a means of experiment al control to disrupt 108 normal face-processing. While the former may elicit the face inversion effect (which will be detailed in 109 section 2.3), the latter preserves low-level visual features but destroys holistic face-processing. 110 In the second part of our study (Experiment 2), we aimed to assess if facial expressions oriented towards 111 the targets had any effect on the latency of gaze -following. The gaze-following portion of the paradigm 112 was modified so that the averted gazing head of the conspecific did not just display neutral expressions, 113 but also threat, fear -grin, or lip-smacking expressions (Fig. 1D). At the time Experiment 2 was prepared 114 we were only able to elicit lip -smacking in one monkey (monkey C) serving as model for the portraits, 115 which is why all lip-smacking demonstrator portraits are based on this particular monkey. For the other 116 expressions (neutral, threat, and fear -grin) all 3 monkeys that took part in this study were able to serve 117 as models. The demands of the task were the same as in Experiment 1, and the observing subject was 118 required to use the head gaze of the demonstrator to single out the correct target (Fig. 1C). Considering 119 the findings obtained in Experiment 1, we restricted the direct gazer type in Experiment 2 to only 120 upright neutral and threat expressions, portraying open or closed eyes, and waived the scrambled 121 versions of the various direct gazer categories. 122 Trials in both Experiment 1 and 2 commenced with fixation of a central white colored fixation point, and 123 500ms after trial onset the direct gazer was displayed, centered behind a red fixation point. As described 124 before, the direct gazer consisted of a forward orient ed monkey portrait, with his eyes open or closed, 125 and presenting a neutral, threatening, or fear -grin expressio n. The direct gazer could be presented 126 upright or inverted, and every category of stimuli also had a scrambled equivalent, which was produced 127 via shuffling of the pixels of the portrait in a random manner to disrupt the holistic facial image while 128 preserving low level visual features. The direct gazer in Experiment 1 was presented for 100, 200, 300, 129 400, or 800ms, while in Experiment 2 only two dur ations were used, 200 or 5 00ms, after which the 130 direct gazer disappeared along with the red fixation point, and was replaced by a demonstrator monkey 131 gazing towards one of four spatial targets. For Experiment 1, the direct gazer categories and their 132 different presentation durations were completely randomized (Fig. 1A). In Experiment 2 only one direct 133 gazer category was tested per experimental session (neutral or threatening), as there were four 134 demonstrator gaze directions (i.e. the portrait gazing at one o f four targets) each associated with four 135 (neutral, threatening, fear -grin, and lip -smack) facial expressions (Fig. 1C). As control, we used a black 136 background, which was presented for the same durations as that of the portraits mentioned above. 137 Unlike the gaze-following paradigm used in prior studies, we opted to not use a go cue as it may conceal 138 a potential influence of the direct gazer on the latency of the subsequent gaze -following response. 139 Instead, the appearance of the demonstrator monkey served as a go cue to initiate gaze -following, and 140 the experimental monkey could initiate his gaze -following response whenever he was ready. 141 Identification of the correct spatial target that the demonstrator monkey was looking at was followed by 142 a juice reward. The identity of the demonstrator monkey used as a spatial cue was always consistent 143 with the identity of the direct gazer. 144 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 21, 2025. ; https://doi.org/10.1101/2025.05.18.654731doi: bioRxiv preprint 2.2 Direct gaze facilitates gaze-following when accompanied with a threat expression 145 In Experiment 1, we explored whether the expression associated with direct gaze influenced the onset 146 of gaze -following responses. We separated all gaze -following trials according to their direct gazer 147 categories, and identified the average response latency of the gaze-following saccade. The latencies for 148 the control condition (black background) were subtracted from the latencies of the gaze -following 149 saccades of each of the direct gazer categories (and for each viewing duration) to standardize the 150 latency measures. 151 Figures 2A-I summarize the effects of direct gaze with eyes open (red) and eyes closed (blue) on the 152 latency of the gaze-following response for each type of facial expression (Fig. 2A-C for monkey L, D-F for 153 monkey J, and G -I for monkey C). A 7 x 2 x 5 ANOVA (direct gazer type x eye visibility [direct gaze vs 154 closed] x presentation duration) revealed an interaction between eye visibility and direct gazer type for 155 two of the monkeys (Monkey L: F=2.34, df=6, p=0.0294; Monkey C: F=2.76, df=6, p=0.011) , while in a 156 third monkey an interaction between eye visibility, direct gazer type, and presentation duration was 157 found (Monkey J: F=2.94, df=24, p<0.0001). For all three monkeys, direct gaze from the eyes had no 158 significant effect on gaze -following respon ses when the direct gazers were displaying neutral 159 expressions, for any presentation duration (Fig. 2A, D, and G, posthoc t -tests not significant). This result 160 is contrary to findings in humans, where direct gaze coupled with a neutral expression is suffic ient to 161 trigger an earlier gaze-following response (8). However, when direct gaze with eyes open was paired 162 with the threat expression, gaze -following response times were significantly earlier, compared to if the 163 threat expression was associated w ith closed eyes (Fig. 2B, E, and H). This effect was significant when 164 the threatening direct gazer was displayed for 200ms in monkeys L and J (posthoc t -tests, p=0.0028 and 165 p=0.0065 respectively, using the Benjamini-Hochberg procedure at a false discovery rate of 5%), and for 166 400ms in monkey C (p=0.0078). We did not find effects for other presentation times for the monkeys, 167 but that is likely because shorter presentation times do not allow sufficient time for the stimulus to be 168 processed, while longer presentation times lead to habituation of the stimulus. 169 In combination with a fear-grin expression gaze-following response times were delayed (Fig. 2F, posthoc 170 t-tests p=0.0174 and p=0.0081 for presentation times 300 and 400ms respectively). However, this effec t 171 of delaying gaze-following with a directly gazing fear-grin expression was only observed in monkey J. Our 172

suggest that not only are monkeys sensitive to the direct gaze of a conspecific and use it to 173 accelerate gaze -following, but they also expe rience it in a manner different from humans and require 174 reinforcement from facial expressions. 175 2.3 Gaze-following in macaques is affected by face inversion 176 The face inversion effect, characterized by more rapid recognition and discrimination of upright as 177 compared to inverted faces is a phenomenon observed in humans and chimpanzees (20). There have 178 been many conflicting reports as to whether macaques experience the face inversion effect, because 179 there is no agreed way to test for it in monkeys (21–24). To the best of our knowledge the effect of face 180 inversion on gaze -following is untested in both humans and non -human primates. In Experiment 1, we 181 introduced inverted faces that included various facial expr essions, with or without direct gaze, and 182 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 21, 2025. ; https://doi.org/10.1101/2025.05.18.654731doi: bioRxiv preprint assessed the effects of these stimuli on the latency of gaze-following. If inverted and upright direct 183 gazers were processed in the same way, we would have expected identical gaze-following profiles. Here 184 we show that is not the case. While the upright directly gazing threat expression had a fairly consistent 185 effect across all our monkeys, each of our monkeys had a unique response profile associated with 186 inverted directly gazing faces. For instance, the gaze-following responses of monkey L were not affected 187 by direct gaze at all when presented in the inverted configuration (Supp. Fig. 1A -C), whilst monkey J 188 displayed earlier gaze-following responses when confronted with inverted expressions with their eyes 189 closed (Su pp. Fig.1E and F). Finally, Monkey C even showed earlier gaze-following responses when 190 confronted with inverted neutral faces with their eyes closed (Supp. Fig.1G). 191 Whatever the reason for these inconsistencies may be, the clear deviation from responses to upright 192 faces suggests that in macaques inverted faces are processed differently from upright faces. Otherwise, 193 the impact of direct gaze from the eyes combined with the threat expression would have been 194 preserved, which we do not observe. On the other ha nd, we also recognize that inverted faces are 195 artificial configurations; even more uncanny when they rotate back to the upright orientation of the 196 demonstrator, and is unlikely to be a familiar experience for rhesus macaques. This could have also 197 contributed to the profound inter -individual differences in our findings. Finally, direct gaze had no 198 significant effect for scrambled direct gazers, which is what we expected (posthoc t-tests not significant). 199 2.4 Gaze-following performance is not affected by direct gazer category 200 Given that direct gaze and threat expressions can accelerate gaze-following, we wondered if threat and 201 more generally the particular expression associated with direct gaze also mattered for the accuracy of 202 subsequent gaze-following. Hence, we calculated the gaze -following performance for each of the direct 203 gazer categories independently of the presentation duration and asked if certain direct gazer categories 204 had an impact on the gaze -following efficiency as gauged by the percen tage of cor rect target choices. 205 However, we were unable to uncover any such effects in our monkeys (posthoc t-tests not significant). 206 2.5 Affiliative expressions towards spatial targets delay gaze-following responses 207 In humans, facial expressions convey different meanings depending on whether they are directed 208 towards the observer or another location of interest (25–27), and we suspected that this may also hold 209 true for macaque monkeys although this has never been explicitly tested. We maintained our original 210 question of whether or not direct gaze could affect gaze -following, but now asked how expressions 211 added to the spatial cue might modulate gaze-following responses (Fig. 1C and D). Given our finding that 212 gaze-following could be influenced by prior direct gaze, we speculated that subsequent facial 213 expressions could have an add -on ef fect to gaze -following. If we take what is reported in the human 214 literature to also hold for rhesus macaques, one might expect that a fear -grin expression following a 215 threatening direct gaze might entail extremely early gaze-following responses. 216 The experiments testing this assumption were divided into two groups; sessions that displayed neutral 217 faces as the direct gazer, and sessions that made use of threatening faces as the direct gazer. This 218 reduced the number of possible facial expression combinations w ith the spatial cue expressions, i.e. the 219 expressions accompanying the oriented faces. To sum up, in a neutral -based session a neutral face 220 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 21, 2025. ; https://doi.org/10.1101/2025.05.18.654731doi: bioRxiv preprint direct gazer (with eyes showing direct gaze or eyes closed) was followed by a spatial cue portraying a 221 monkey avert ing gaze in association with one of four expressions (neutral, threat, fear -grin, or lip -222 smacking), and likewise for a threat -based session. We also simplified the direct gazer presentation 223 durations to short (200ms) or long (500ms), also for the purpose o f lowering the number of potential 224 combinations of direct gazer expression type, gaze type, presentation duration, and spatial cue 225 expression. The reduction of the types of combinations also afforded more repetitions within each 226 session, and therefore could decrease the number of experimental sessions in Experiment 2 (Fig. 1C). 227 We conducted a 2 x 2 x 4 ANOVA (eye visibility [direct vs closed] x duration (200 vs 500ms] x spatial cue 228 expression [neutral, threat, fear, lip -smack]) for neutral and threat direct gazer sessions. Regardless of 229 which direct gazer type we used, we no longer saw any influence of the direct gazer on the gaze -230 following response of any of the observing monkeys, nor any consistent interaction between eye 231 visibility and the other factors. This is an unexpected outcome as one might have expected that a 232 threatening direct gazer with his eyes open would continue to facilitate gaze -following. However, there 233 was no difference between gaze -following responses that started with threatening eyes op en or 234 threatening closed eyes direct gazers, even when followed by a neutral face gazing towards one of the 235 four targets. It is possible that the addition of the second facial expression in the demonstrator 236 overshadows the information provided by the eyes in the direct gazer, leading to the loss of the latter’s 237 effect on gaze-following. 238 Under the assumption that the eye visibility of the direct gazer does not matter when followed by an 239 expressive demonstrator, we pooled the data over all eye visibility con ditions for neutral and 240 threatening direct gazers and separated the resulting set s according to only direct gazer presentation 241 duration and demonstrator expression s. This also allowed us to double our sample size as we would 242 effectively only be performing a 2 x 4 ANOVA (duration x spatial cue expression) with the same 243 experimental results. We identified an interaction between the averted gaze duration and expression 244 (Monkey L: F=19.63, df=3, p<0.0001; Monkey J: F=4.98, df=3, p=0.0019; Monkey C: F=10.27, df= 3, 245 p<0.0001) when the direct gazer was neutral. Importantly, we also detected significant differences 246 between gaze-following responses within the facial expression category (p<0.001 for all 3 monkeys). 247 Despite the simplification of the paradigm and reduction of the number of configurations we had to 248 compare, it was difficult to conclude how our subjects were reacting to the different presentation 249 durations, direct gazers, and the different expressions on the spatial cues, especially since we were also 250 comparing across three monkeys. Our statistical tests informed us that the facial expression category of 251 the spatial cue caused significant differences in gaze -following latencies, but at first glance it was 252 difficult to draw a meaningful conclusion as to how t he expressions were affecting gaze -following. We 253 decided to normalize and pool all gaze -following responses as a first step towards interpreting our data. 254 Fig. 3A and B show results collected from sessions that utilized neutral direct gazers and threat dir ect 255 gazers respectively. We found that in sessions that began with a neutral direct gazer (Fig. 3A), gaze -256 following responses resulting from spatial cue expressions that were affiliative (fear -grin and lip-smack) 257 were significantly later than those that we re neutral or antagonistic (threat). This result was the same 258 for both direct gazer presentation durations. 259 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 21, 2025. ; https://doi.org/10.1101/2025.05.18.654731doi: bioRxiv preprint In the sessions that were initiated with a threatening direct gazer (Fig. 3B), the spatial cue with the 260 threat expression facilitated gaze -following responses the most (left). Unlike the result with neutral 261 direct gazers where the gaze -following responses were divided between affiliative or antagonistic 262 expressions, now the response for the threatening demonstrator stands out the most among the four 263 expressions. It is possible that its consistency with the expression of the direct gazer prevents the 264 dilution of the social salience of the cue, whereas in the other conditions an expression change occurs, 265 i.e. from threat to lip -smacking, which leads to t he influx of new expression information. What is 266 apparent is that in both direct gazer conditions affiliative expressions consistently lead to later gaze-267 following responses. This effect is not restricted to specific presentation durations, and this is lik ely due 268 to the appearance of the second expression in the spatial cue overriding all effects of the direct gazer. 269 Taken together, the results of Figures 2 and 3 suggest that while direct eye gaze can be a powerful cue in 270 expediting gaze -following when comb ined with a threatening expression, the presentation of new 271 information will take priority and determine the gaze-following response, irrespective of the direct gazer 272 features. Where an antagonistic face is looking at seems to take precedent over where an affiliative face 273 is attending to. 274 2.6 Direct expressions lead to earlier gaze-following expressions than averted expressions 275 In this final part, we wanted to compare gaze-following responses from Experiments 1 and 2. Previously, 276 we had separately explored the effects of facial expressions when they were directed towards the 277 observer and when they were averted and directed towards spatial targets. This presented a unique 278 opportunity to assess within the same experimental monkeys if expressions were perceive d differently 279 depending on where they were directed, using their gaze -following responses as a readout. Past 280 research had shown that monkeys may be more likely to follow gaze if an averted face was expressing 281 fear (13), but it is unknown how that same monkey would have reacted if the fear expression had been 282 directed towards himself prior to gaze -following. By combining the results of Experiment 1 and 2, we 283 explored the impact of different expression transitions from direct to target oriented gaze (direct gazer 284 to demonstrator) on the observer´s gaze -following. We only used trials that had direct gazer 285 presentation durations of 200ms because that was the only timing shared between both experiments. 286 The results for the three monkey individuals are summarized in Fig. 4A-C. As a reference, we used trials 287 from both Experiments 1 and 2 in which the neutral expression was maintained from the direct gazer to 288 the spatial cue (column 1 in each plot). The figures show a clear and significant delay in gaze -following 289 whenever there was an averted expression. In two of our three monkeys this effect was present for both 290 threat and fear expressions (two sampled t -test, monkey L: p=0.00221 and p<0.001 respectively; 291 monkey C: p<0.001 for both), whereas the third monkey (Fig.4B) only showed a significant effect for the 292 fear expression (p<0.001). One interpretation is that the observing monkeys place less importance on 293 expressions that are not directed towards them, and therefore follow gaze after a longer pause . 294 Alternatively, the demonstrator expressions (which appear later in the course of each trial) may exert an 295 increased informational load on the observer, causing delayed gaze-following responses. 296 3. Discussion 297 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 21, 2025. ; https://doi.org/10.1101/2025.05.18.654731doi: bioRxiv preprint When the other´s eyes meet our own, we experience direct gaze, a potent visual cue indicating that we 298 have become the focus of the other’s attention. This is a vital step towards establishing mutual gaze, 299 allowing both parties to silently acknowledge their attention to each other . If the other subsequently 300 looks towards a third object, the observer may be compelled to follow the other’s gaze to the same 301 location, thereby establishing joint (object-related) attention. Gaze-following may also occur without 302 prior gaze, yet at least in humans it may be promoted by preceding direct gaze (8). Whether monkey 303 gaze-following also benefits from direct gaze in the same manner was unclear. Hence the primary aim of 304 the present study was to test whether monkeys are able to use direct gaze to inform their gaze -305 following decisions and, moreover, to determine the impact of accompanying facial expressions. 306 The first experiment ’s results clearly indicate that direct gaze can indeed enhance gaze-following 307 responses in monkeys, yet only when accompanied by an expression of threat. However, the synergistic 308 effect of direct gaze and threat is not independent of the demonstrator´s expression when shifting gaze 309 to a target. Our second experiment showed that whenever a distinct facial expression accompanied the 310 demonstrator´s gaze aversion , neither the eyes nor the exp ression of the direct gazer mattered 311 anymore. When the expression accompanying gaze aversion indicated submission, the gaze -following 312 response was delayed compared to both neutral and threatening expressions. Finally, direct comparison 313 of both experiments revealed that facial expressions accompanying direct gaze trigger ed earlier gaze-314 following responses than the same expressions accompanying the demonstrator´s gaze shifted towards 315 a target (summarized in Table 1). 316 Table 1 317 Expressions Direct Gazer Effect on gaze-following (Experiment 1) Averted Gazer Effect on gaze-following (Experiment 2) Neutral No effect No effect Threat Earlier, with eyes No effect Fear-grin No effect Delayed, independently of eyes Lip-smacking Untested Delayed, independently of eyes 318 In order to avoid the many shortcomings of studying social interactions in natural settings outlined in 319 the introduction, we resorted to a study of the interaction between gaze -following and facial 320 expressions under laboratory conditions guaranteeing stable high resolution recordings of the 321 observer´s eye movements prompted by exposure to demonstrator portraits offering combinations of 322 highly standardized facial expressions and gaze orientations from an extensive library. In an attempt to 323 standardize the demonstrator stimuli as much as possible we used static stimuli rather than video clips. 324 The ease of interpreting our results in terms of ethological validity promotes trust that our paradigm 325 indeed reflects natural social behaviour of macaques. If these views are truly justified we must await 326 ground truth studies with comparable data quality that may one day become feasible also under natural 327 conditions. 328 In the wild, facial displays of threat and aggression accompanying direct gaze are frequently employed 329 by dominant males to assert their status within the ir group and to maintain control over resources 330 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 21, 2025. ; https://doi.org/10.1101/2025.05.18.654731doi: bioRxiv preprint (11,28). It would be extremely beneficial for others in the group to pay attention to these signals in 331 order to get ready for the potential consequences of a dominant male´s upcoming move as this could be 332 hostile and potentially dangerous. As these moves are typically initiated by gaze shifts to third parties or 333 objects of mutual interest , the observer – no matter if competitor or subordinate – would be well 334 advised to follow gaze as quickly a s possible. Findings by Champ et al. (17) showed that appeasing and 335 submissive behavior attracted more gaze -following from an observer, which is at odds with what our 336

show. The reason for th is discrepan cy might be that in our study each gaze shift of the 337 demonstrator was preceded by a n exchange of direct gaze, a strict sequence that was lacking in the 338 Champ study. 339 The existence of a face inversion effect in monkeys has been subject to much debate: some studies have 340 reported no change in face recognition upon inversion of faces (21,22), whilst others have reported an 341 inversion effect in the sense of an impaired ability to recognize inverted faces (24). Another study even 342 reported an inversion effect for human but not for monkey faces (23). If inverted faces are not 343 processed in the same way as upright faces (i.e. inversion effect present) , one might expect threat 344 expressions combined with direct gaze to have a similar effect on gaze -following no matter of the 345 orientation of the face. Our data shows that this is not the case, as inverting the direct gazer exhibiting a 346 threat expression saw the loss of effect of direct gaze on the subsequent gaze -following response that 347 differed between individual s. The lack of consistency may reflect individually differing perceptual 348 interpretations of inverted faces. After all, instances in which a monkey might encounter a fully inverted 349 conspecific exhibiting direct gaze in association with meaningful expressions must be rare, if not 350 completely absent. In other words, such configurations are simply unnatural and if presented in an 351 experiment may prompt highly idiosyncratic interpretations. 352 In Experiment 2, the facilitatory influence of threat accompanying direct gaze on gaze-following was 353 cancelled as soon as a new averted facial expression appeared. The reason may be that t he appearance 354 of two expressions in quick succession create d a redundancy of social signals, leading to the 355 prioritization of the more recent expression and therefore the loss of influence of direct eye gaze. It is in 356 the best interest of m onkeys to respond quickly to social cues and changes in their environment for 357 survival, which is why the later expression with the gaze shift may take precedence. Recency bias is a 358 documented phenomenon in macaques (29), although it has not been explored in the con text of social 359 cue retention . Finally, also restrictions of processing capacities seem conceivable, entailing an 360 attentional focus on the second facial expression, especially when this is necessary in our paradigm for 361 reward. 362 In monkey groups, gaze-following aids in the determination of rank with the dominating individual 363 receiving the most attention , often associated with signs of submission . Following the gaze of group 364 members expressing submission helps others to shift attention to this dominating and potentially 365 dangerous individual (30). As described above, the results of Champ et al. (17) support the view that this 366 shift is facilitated by concomitant fear-grin as an indicator of submission to a higher ranking individual. 367 Similarly also the work by Goossens et al. (13) showed more gaze-following responses when a gaze-shift 368 was accompanied by fear-grin. Against the backdrop of this work, one might have expected that also in 369 our study the demonstrators exhibiting fear-grin in conjunction with their shifting gaze would elicit 370 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 21, 2025. ; https://doi.org/10.1101/2025.05.18.654731doi: bioRxiv preprint earlier gaze-following responses. Instead, the responses to the demonstrator´s averted gaze, associated 371 with fearful or lip-smacking expressions actually occurred later (Fig. 3A and B ) in our study. Hence, how 372 can we reconcile this difference? 373 In humans, averted expressions of fear have been found to be difficult to disengage from, leading to 374 weaker gaze -cueing effects , especially if the subjects were prone to anxiety (31–33). Surprisingly, 375 averted rather than direct gaze has been found to elicit increased arousal in monkeys (34). A directly 376 gazing face indicates the attention of the other is on the observer, and while this is socially significant for 377 macaques, they may have adapted to downre gulate their level of arousal, for instance by breaking eye 378 contact. However, an avertedly gazing face points to a yet unknown target of interest which needs to be 379 identified, adding an element of unpredictability which demands vigilance and m ay contribute to the 380 increase in physiological arousal. Combined with the inherent ambiguity of the fea r-grin expression 381 which may indicate anxiety/fear rather than submission, we conclude that the need to closely analyze 382 arousing averted expressions conveying important information on potential danger may be the cause of 383 the delayed attentional disengagement and therefore later gaze-following responses. 384 The fact that delayed gaze-following responses occur only when facial expressions are accompanying 385 the conspecific´s gaze shift demonstrates that rhesus macaques can flexibly interpret expressions based 386 on context analogous to humans. Modulation of gaze -following and the gaze -cueing effect via changes 387 in context provided by facial expressio ns has been recognized in humans, although the diversity of 388 paradigms used makes it difficult to pinpoint what exactly the interaction is between gaze direction, eye 389 visibility, and facial expressions (35). By contrast, the way the perception of an expression can change 390 based on gaze is a more consistently reported phenomenon in humans. According to the shared signal 391 hypothesis gaze direction influences the perception of facial expression s when they communicate 392 similar intent (36). For insta nce expressions such as joy or anger are perceived as approach -type 393 expressions and are therefore enhanced when viewed directly, whereas avoidance-type expressions like 394 sadness and fear are promoted by averted gaze (25,26). If we break down both of our experiments in 395 terms of events that might occur in the wild, it seems there is more urgency for an observ ing monkey to 396 respond to aggressive intent from a conspecific than to identify what the conspecific may be threatening 397 in the environment . In the same vein, monkeys seem to prioritize interpreting another’s fear in them 398 rather than discovering what may have frightened the other. 399 While humans can demonstrate enhanced gaze-following in response to direct gaze alone (8), we were 400 unable to observe a comparable result in any of our macaques. A factor contributing to this discrepancy 401 may be the differential processing of social signals that hinges on our available social cognitive 402 capabilities and how our social groups are structured. The cognitive processes involved in perceiving and 403 responding to facial expressions, particularly in the context of gaze -following, could be shaped by 404 species-specific adaptations. Unlike humans who have evolved a variety of ways to communicate our 405 intentions, monkeys have a much smaller social arsenal to draw upon, and may have adapted responses 406 that prioritize the detection and response to potential threats, and to avoid meaningless conflict (37–407 39). 408 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 21, 2025. ; https://doi.org/10.1101/2025.05.18.654731doi: bioRxiv preprint Sensitivity towards direct gaze is a phylogenetically ancient feat not confined to primates, and has been 409 discovered in other mammals, and even some species of bird and fish (40). Because the direct gaze of 410 predators represents danger to the observer, it was beneficial for many animals to develop the ability to 411 detect direct gaze. Rhesus macaques have been rated high in their tenden cy to use direct gaze as a 412 threat signal and are reluctant to make prolonged eye contact (11). In other words direct gaze is 413 generally not regarded as an affiliative gesture. By contrast, gaze -following is recognized as a more 414 evolved cooperative behavior to affo rd living in a social group (2), arguably derived from the ability to 415 exploit direct gaze. T he ability to produce and read facial expressions appeared much later, and for the 416 most part seems to be restricted to the mammalian lineage (41,42). Finally i n species prone to 417 aggression, t he ability to bridge from a provocative stare into cooperative communication m ay have 418 been an important step towards establishing social harmony. Our demonstration that rhesus macaques 419 can use direct gaze and the integration of facial expressions to shape gaze -following behavior and 420 modulate its exigency is congruent with this proposal. It would be interesting to examine whether or not 421 possession of a more sophisticated facial muscle control architecture that allows for a wider repertoire 422 of facial movements drives the emergence of more pros ocial and affiliative behav iors in other non -423 human primates and mammals. 424 425 Funding 426 This work is supported by the Deutsche Forschungsgemeinschaft (DFG) [TH 425/12-2, to P.T.] and the 427 Werner Reichardt Centre of Neuroscience [EXC 307, to P.T.]. 428 Author Contributions 429 I.C. and P.T. designed research; I.C. performed research and analyzed the data; I.C. and P.T. wrote the 430 paper and contributed to discussing analysis. 431 432 4. Materials and Methods 433 4.1 Animals and Surgery 434 Three male rhesus macaques ( Macaca mulatta) were involved in the present study. They had all been 435 used before in unrelated electrophysiologcal work. The three – Monkey L (18 years old, 11kg), Monkey J 436 (21 years old, 16kg), and Monkey C (16 years old, 11kg) – were all housed separately and alone. They all 437 had originally played the dominant roles in their previous pairings (all male). Monkey L and J had lost 438 their partners due to natural causes, while monkey C lost his dominant status in his pairing and 439 subsequently had to be separated from his partner to avoid injury . They had never undergone trial 440 pairings nor had they been involved in skirmishes with each other , but had occasional visual contact 441 with each other in the experimental setups and animal facility. While we cannot rule out their familiarity 442 with each other’s faces, we do not believe the monkeys had an established hierarchy as they d id not 443 have prolonged contact with one another. 444 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 21, 2025. ; https://doi.org/10.1101/2025.05.18.654731doi: bioRxiv preprint All three monkeys had been implanted with a titanium head-post for restraining the monkey's head in 445 order to allow precise measurement of eye movements and chambers for electrophysio logical 446 recordings from the superior temporal sulcus , the latter not related to the present study. Being able to 447 head fix the monkeys was essential for the quantitative characterization of gaze -following. Surgeries 448 were carried out under combination anesthesia with isoflurane (1.3%) and remifentanil (1 – 2 μg per kg 449 per minute), with meticulous monitoring and control of body temperature, heart rate, blood oxygen 450 saturation, and blood pressure. Opioid analgesics (buprenorphine) were administered until the monkeys 451 showed no signs of residual pain and were given ample time for full recovery before entering the 452 behavioral training for the experiments that preceded the behavioral study at issue here . All 453 experimental preparations and procedures were sanctioned by the local animal care committee 454 (Regierungspräsidium Tübingen, Abteilung Tierschutz), fully complying with German and European law 455 and the National Institutes of Health's Guide for the Care and Use of Laboratory Animals. 456 Table 2 summarizes some potentially expedient behavioral observations made on the three monkeys 457 that participated in this study. These observations were collected by the experimenters and the primary 458 animal caretaker at the start of the study. 459 Table 2 460 Monkey Movement Personality Feeding Response to net (threat) Experimenter tolerance Speed of entering setup Sensitivity to new stimuli Willingness to work Lenny (monkey L, 18 yrs , 11kg) Fast, active Calm, alert Normal Normal (Threat +retreat) Indifferent Fast (if properly deprived Low High Joris (monkey J, 21yrs, 16kg) Fast, active Anxious Normal- Fast Normal (Threat +retreat) Engaging Fast (regardless of deprivation) High If properly deprived Chris (monkey C, 16yrs, 11kg) Slow (but not sluggish) Calm, indifferent Slow High aggression Indifferent Slow (regardless of deprivation) High If properly deprived 461 4.2 Behavioral Paradigms 462 For the present study, the three monkeys were trained on two variants of a gaze-following paradigm 463 that both required them to use head gaze information provided by the portrait of a conspecific 464 demonstrator. Of the three monkeys used in this study, only Monkey L had previously par ticipated in a 465 gaze-following study and therefore was already familiar with the demands of the paradigm. Monkey J 466 and L were new to the gaze -following paradigm. In the final behavioral paradigm, e ach individual trial 467 began with a white fixation point on a black background, and the experimental monkey was required to 468 fixate on the point for 500ms. Subsequently, the white fixation point was replaced by a red fixation 469 point centered on a portrait gazing at the observer, a configuration we refer to as the “direct gazer” 470 stimulus. In Experiment 1 the direct gazer was presented for 100, 200, 300, 400, or 800ms, while in 471 Experiment 2, the direct gazer was seen for 200 or 500ms. The direct gazer was followed by the 472 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 21, 2025. ; https://doi.org/10.1101/2025.05.18.654731doi: bioRxiv preprint presentation of the “demonstrator”, a photograph of the same individual as before, but now with his 473 head gaze turned to a distinct spatial target, randomly chosen from a set of 4 dots arranged on the 474 horizontal axis. Although both the demonstrator and all potential targets remained present until the end 475 of the trial , the observer was encouraged to make a saccade towards the target singled out by the 476 demonstrator’s head gaze as a speedy gaze -following response allowing earlier access to reward . If the 477 observer had successfully initiated the trial by fixating the white central fixation point and correctly 478 followed the demonstrator’s gaze to the correct spatial target, he received a drop of juice as reward. If 479 central fixation failed or the observer´s gaze did not meet the target within 300ms after the appearance 480 of the demonstrator, the trial was terminated and the monkey received no reward. The presentation 481 duration and direct gazer category was randomized from trial to trial. In Experiment 1, 32 sessions were 482 collected in monkey L, 40 in monkey J, and 66 in monkey C. In Experiment 2, 10 sessions were collected 483 in monkey L, 10 in monkey J, and 11 in monkey C. Each session for both Experiments 1 and 2 contained 484 an average of 800 gaze -following saccades. Data for Experi ment 1 was collected in approximately 3 485 months with a pause in between for analysis, before starting Experiment 2 which required around 2 486 months. 487 The size of the direct gazer and demonstrator portraits were 5.6° by 5.6°, while the white fixation point 488 and the red spatial targets had a diameter of 0.8° each. The spatial targets were arranged in a virtual 489 horizontal row 1° below the center of the demonstrator portrait. With respect to the observing monkey 490 subject, the horizontal eccentricities of the targets were -10°, -5°, 5°, and 10° (or -40°, -20°, 20°, and 40° 491 with respect to the demonstrator). Tracking of eye movements was accomplished using an infrared 492 camera (iViewX, SensoMotoric Instruments, spatial resolution of <0.3° at 50 Hz). 493 4.3 Stimuli preparation 494 The demonstrator portraits we used in our paradigms for both the direct gazer and the demonstrator 495 were stills taken of the same three monkeys (L, J, and C) that participated in the study. Video-496 photography was captured while our monkey subjects were seated in their chairs, but were not head 497 fixed to maximize comfort, ease of moving their mouths as required in 3 of the 4 expressions, and for 498 turning their heads. In Experiment 1, the direct gazers could exhibit a neutral, threat, or fear-grin 499 expression, while in Experiment 2 lip-smacking was added. The neutral expression consisted of a closed 500 and relaxed mouth. It is the default expression of monkeys, present during rest or indifference , not 501 prompting specific behavioral reactions . The threat expression is also known as the open -mouthed 502 threat, in which the mouth is opened and the canines are exposed. Such a display may be used to 503 intimidate other members of a social group, and when displayed between two similar ly ranked 504 individuals may be interpreted as a physical challenge. The fear expression is often referred to as the 505 fear-grin, exposing all the teeth while the jaws remain closed and the lips are retracted. Such an 506 expression is not only used to indicate submission, but can also be used as a friendly gesture to signal 507 affiliation and appeasement . Finally, the lip- smacking expression consists of a closed mouth and 508 puckered lips, and like the fear-grin it is categorized as a n appeasing expression used to alleviate 509 tension. However the lip -smack is more affiliative, and can be used to signal friendliness rather than 510 subordination. 511 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 21, 2025. ; https://doi.org/10.1101/2025.05.18.654731doi: bioRxiv preprint The expressions documented by the portraits were prompted by the caregivers performing in front of 512 the monkeys in appropriate manners. All video -photography were captured while our monkey subjects 513 were seated in their chairs, but were not head fixed to maximize comfort, ease of moving their mouths 514 as required in 3 of the 4 expressions, and for turning their heads to document averted gaze in 515 conjugation with distinct expression. Neutral expressions were the easiest to capture as they were the 516 default expressions of our monkeys easily produced in calm, non -stimulating situations. Threat 517 expressions were more likely to b e produced if a human experimenter approached them too closely, or 518 made abrupt movements. Bringing the video camera close to the monkey was also an easy way to cause 519 them to react with the threat expression, possibly as a proximity warning towards the expe rimenter. 520 The fear-grin was a rare occurrence in our setup, but the use of a handheld mirror to show the monkeys 521 their reflection, or special face masks with prints of monkey mouths could sometimes induce the fear -522 grin expression. Finally, the lip -smacking expression could be induced by exposure of our chaired 523 monkey to unfamiliar and unrestrained monkeys in the animal facility. In this setting, the fear -grin 524 expression could sometimes be observed. Because the monkeys habituated quickly to the presence of 525 other monkeys or the various props that we showed them, it was important that all videos were taken 526 before the monkeys got bored and regressed to exhibiting neutral expressions. All monkeys produced 527 the neutral, threatening, and fear-grin expressions, while only one monkey (monkey C) produced the lip-528 smacking expression. After the videos were taken, we isolated frames in which the monkey was 529 conveying the desired expression with direct or averted gaze and cropped out the background , their 530 bodies, and their implants. 531 The paradigms used in Experiments 1 and 2 only differ ed in the types of stimuli used and the 532 presentation duration of the direct gazer stimuli ; the structure of each trial was essentially the same. 533 Table 3 summarizes the different stimuli involved in E xperiments 1 and 2, as well as the presentation 534 durations used. In some trials, t he orientation of the direct gazer portraits could also be inverted. 535 Moreover, scrambled direct gazers were used for additional control in Experiment 1, and as expected 536 there was no effect of eye visibility in scrambled stimuli . Generation of scrambled stimuli was achieved 537 by taking the upright form of the direct gazer, dividing it into evenly sized squares with sides 9 pixels in 538 length and rearranging the squares randomly. This disrupted the facial information whilst retaining the 539 low level visual features of the portraits. 540 Table 3 541 Direct gazer types Direct gazer duration (ms) Demonstrator expression Experiment 1 1) Neutral (direct and eyes closed) 2) Threat (direct and eyes closed) 3) Fear-grin (direct and eyes closed) Inverted forms of 1-3 Scrambled forms of 1-3 100, 200, 300, 400, 800 Neutral Experiment 2 Neutral (direct and eyes closed) OR 200, 500 Neutral Threat .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 21, 2025. ; https://doi.org/10.1101/2025.05.18.654731doi: bioRxiv preprint Threat (direct and eyes closed) Fearful Lip-smacking 542 4.4 Statistical Analysis 543 Each daily session of behavioral testing in Experiment 1 consisted of hundreds of gaze-following 544 saccades, but given the number of conditions that were being tested ( direct gazer, eye visibilit y, and 545 presentation duration), the number of saccades made for each condition was few. Therefore, we pooled 546 all the saccades made for each condition across all sessions for each individual monkey for analysis (for 547 example in Monkey L, approx. 12 saccades per category for 75 total stimulus categories totals 900 trials 548 per session, 32 sessions results in 384 trials per stimulus category (gaze-following responses represented 549 by saccadic latencies)). For each trial, we extracted the onset of the gaze-following saccade with respect 550 to the appearance of the spatial cue, and determined the average saccade onset across all sessions for 551 that specific condition (eg. direct gazer: upright threat; eyes: direct; duration: 200ms). Gaze-following 552 saccade onsets were then standardized relative to the average saccade onset produced towards a black 553

in place of a direct gazer before the appearance of the demonstrator. 554 To investigate if one of the main variables we manipulated in Experiment 1 or any of their interactions 555 had an impact on the gaze-following saccade latency we resorted to a 7 x 2 x 5 ANOVA (direct gazer type 556 x eye visibility [direct vs closed] x presentation duration) , separately for each monkey . Post hoc t -tests 557 were then performed between direct gaze and closed eyes conditions , for all presentation durations, 558 and within each facial expression type. This approach allowed us to uncover which combination of facial 559 expression, presentation duration, and gaze type could modulate the gaze -following response. The 560 Benjamini-Hochberg procedure was applied to control the false discovery rate (0.05). 561 Since we did not find any effect of direct gaze on the subsequent gaze-following saccade latencies in 562 Experiment 2, regardless of the expression accompanying direct gaze being neutral or threatening or the 563 eyes open or closed, we decided to ignore the eye visibility distinction in order to increase the statistical 564 power of the analysis of the impact of the facial expression of the averted gazer on gaze-following. In 565 other words, in Experiment 2 trials were grouped according to the facial expression of the direct gazer 566 (neutral or threat), presentation duration (200 or 500ms), and the facial expression of the averted gazer 567 (neutral, threat, fear-grin, or lip-smacking). Furthermore, in order to be able to pool data from all three 568 monkeys, we normalized their gaze-following between a range of 0 and 1 . A 2 x 4 ANOVA (presentation 569 duration x expression) was then performed on this pooled data for the normalized gaze -following 570 response times, separately for sessions involving a neutral or threatening direct gazer. Posthoc t-tests 571 were performed on groups that were signific antly different from each other, and again the Benjamini-572 Hochberg procedure was applied to control the false discovery rate (0.05). 573 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 21, 2025. ; https://doi.org/10.1101/2025.05.18.654731doi: bioRxiv preprint 574 Figure 1. Behavioral Paradigms and Stimuli. A. Sequence of events in each trial of the head gaze-575 following paradigm used in Experiment 1. B. Direct gazer stimuli used in Experiment 1. In this visual, we 576 show all the variants of one direct gazer identity, which include his three facial expressions (neutral, 577 threat, and fear -grin) and eyes open and closed conditions, as well as examples of inver ted and 578 scrambled stimuli of the same direct gazer in the lower right corner. Variants of the other two identities 579 of the direct gazers are not shown here. C. Sequence of events in each trial of the head gaze -following 580 paradigm used in Experiment 2. D. Averted demonstrator stimuli harboring facial expressions of one of 581 the monkey identities used in Experiment 2. From left to right, we have a neutral, open -mouth threat, 582 fear-grin, and lip -smacking expression. All four examples shown here are oriented 40° rig htward from 583 the demonstrator’s perspective (or 10° leftward for the observer). 584 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 21, 2025. ; https://doi.org/10.1101/2025.05.18.654731doi: bioRxiv preprint 585 Figure 2. Effect of eye visibility and facial expressions of the direct gazer on gaze-following responses in 586 Experiment 1. Columns depict facial expression types of the direc t gazer (from left to right: neutral, 587 threatening, fear-grin). Red and blue lines represent eyes open and eyes closed conditions of the direct 588 gazer respectively. Mean gaze -following responses ± SEM are shown, and posthoc t -tests were 589 performed, ** P<0.01. Benjamini-Hochberg procedure was applied to control for the false discovery rate 590 (0.05). A-C. Results for Monkey L: 2B shows significantly faster gaze -following responses for a 591 threatening direct gazer with eyes open compared to the same expression with e yes closed (200ms 592 display time). D-F. Results for Monkey J: 2E shows significance of eye visibility when a threatening direct 593 gazer is displayed for 200ms. G-I. Results for Monkey C: 2H shows significance of eye visibility when a 594 threatening direct gazer is displayed for 400ms. 595 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 21, 2025. ; https://doi.org/10.1101/2025.05.18.654731doi: bioRxiv preprint 596 Figure 3. Effect of demonstrator expression type on gaze-following responses in Experiment 2. Because 597 eye visibility no longer mattered in Experiment 2, we pooled together all trials portraying neutral (A) and 598 threatening expressions (B) regardless of eye visibility respectively. Gaze-following responses per subject 599 for each demonstrator expression category were then normalized between 0 and 1, and then pooled 600 together. The solid black line within the box represents the median normalized response, and the 601 whisker lengths are set as 1.5x the interquartile range. Red crosses represent 1.5x outliers. A. Left and 602 right plots show normalized responses where the direct gazer expression was neutral and presented for 603 200 and 500ms respectively. Gaze -following of affiliative expressions is significantly slower than for 604 antagonistic expressions, regardless of direct gazer presentation duration. B. Left and right plots show 605 responses where the direct gazer expression was threatening and presented for 200 and 500ms 606 respectively. Gaze-following of the threat expression is faster than the following of the other facial 607 expressions, but only when the prior threat direct gazer was presented for 200ms. 608 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 21, 2025. ; https://doi.org/10.1101/2025.05.18.654731doi: bioRxiv preprint Figure 4. Comparing the effects of facial expressions on gaze-following depending on whether the 609 expression was directed towards the observer or the spatial target. The solid black line within the box 610 represents the median, and the whisker lengths are set as 1.5x t he interquartile range. Red crosses 611 represent 1.5x outliers. A-C. Results from monkeys L, J, and C respectively. Two sample t -tests were 612 performed, ** P < 0.01, *** P < 0.001. Abbreviations: Neutral (Neu), Threat (Thrt). Neu -Thrt signifies a 613 transition fro m a neutral direct gazer to a threatening demonstrator, and vice versa for the other 614 notations. Expressions that were portrayed in the demonstrator significantly reduced gaze -following 615 reaction times compared to the same expression that was directed toward s the observer as a direct 616 gazer; for A and C both threat and fear, and only fear in B. 617 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 21, 2025. ; https://doi.org/10.1101/2025.05.18.654731doi: bioRxiv preprint 618 Supplementary Figure 1. Effects of face inversion of the direct gazer on gaze-following responses. 619 Columns depict facial expression types of the direct gazer (from l eft to right: neutral, threatening, fear -620 grin, in their inverted orientations). Red and blue lines represent eyes open and eyes closed conditions 621 of the inverted direct gazer respectively. Mean gaze -following responses ± SEM are shown, and posthoc 622 t-tests were performed, ** P<0.01, *** P < 0.001. Benjamini -Hochberg procedure was applied to 623 control for the false discovery rate (0.05). A-C. Results for Monkey L, no significance of eye visibility for 624 any expression. D-F. Results for Monkey J: faster gaze -following responses when confronted with 625 inverted threat and fear expressions with their eyes closed (E and F). G-I. Results for Monkey C: faster 626 gaze-following responses when confronted with inverted neutral faces with their eyes closed (G). 627

628 1. Pelphrey KA, Viola RJ, McCarthy G. When Strangers Pass. Psychol Sci. 2004;15(9):598 –603. 629 2. Shepherd S V. Following gaze: Gaze-following behavior as a window into social cognition. Front 630 Integr Neurosci. 2010;4(MARCH 2010):1–13. 631 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 21, 2025. ; https://doi.org/10.1101/2025.05.18.654731doi: bioRxiv preprint 3. Hamilton AF de C. Gazing at me: the importance of social meaning in understanding direct-gaze 632 cues. Philos Trans R Soc B Biol Sci [Internet]. 2016 Jan 19;371(1686):20150080. Available from: 633 https://royalsocietypublishing.org/doi/10.1098/rstb.2015.0080 634 4. Kleinke CL. Gaze and eye contact: A research review. Psychol Bull [Internet]. 1986;100(1):78 –100. 635 Available from: https://doi.apa.org/doi/10.1037/0033-2909.100.1.78 636 5. Emery NJ. The eyes have it: The neuroethology, function and evolution of social gaze. Neurosci 637 Biobehav Rev. 2000;24(6):581–604. 638 6. Langton SRH, Watt RJ, Bruce V. Do the eyes have it? Cues to the direction of social attention. 639 Trends Cogn Sci. 2000;4(2):50–9. 640 7. Baron-Cohen S. How to build a baby that can read minds. Curr Psychol Cogn. 1994;13:513 –52. 641 8. Dalmaso M, Alessi G, Castelli L, Galfano G. Eye contact boosts the reflexive component of overt 642 gaze following. Sci Rep. 2020;10(1):1–10. 643 9. Marciniak K, Dicke PW, Thier P. Monkeys head-gaze following is fast, precise and not fully 644 suppressible. Proc R Soc B Biol Sci. 2015;282(1816). 645 10. Chong I, Ramezanpour H, Thier P. Causal manipulation of gaze-following in the macaque 646 temporal cortex. Prog Neurobiol [Internet]. 2023 Jul;226(December 2022):102466. Available 647 from: https://doi.org/10.1016/j.pneurobio.2023.102466 648 11. Harrod EG, Coe CL, Niedenthal PM. Social Structure Predicts Eye Contact Tolerance in Nonhuman 649 Primates: Evidence from a Crowd-Sourcing Approach. Sci Rep [Internet]. 2020;10(1):1–9. 650 Available from: http://dx.doi.org/10.1038/s41598-020-63884-x 651 12. Teufel C, Gutmann A, Pirow R, Fischer J. Facial expressions modulate the ontogenetic trajectory 652 of gaze-following among monkeys. Dev Sci. 2010;13(6):913–22. 653 13. Goossens BMA, Dekleva M, Reader SM, Sterck EHM, Bolhuis JJ. Gaze following in monkeys is 654 modulated by observed facial expressions. Anim Behav. 2008;75(5):1673–81. 655 14. Bettle R, Rosati AG. Flexible gaze-following in rhesus monkeys. Anim Cogn [Internet]. 656 2019;22(5):673–86. Available from: https://doi.org/10.1007/s10071-019-01263-4 657 15. Paukner A, Anderson JR, Fogassi L, Ferrari PF. Do facial gestures, visibility or speed of movement 658 influence gaze following responses in pigtail macaques? Primates. 2007;48(3):241–4. 659 16. Luisi B, Micheletta J, Julle-Danière E, Catinaud J, Maréchal L. Monkeying around: Non-human 660 primate behavioural responses to humans reproducing their facial expressions. Appl Anim Behav 661 Sci. 2023;265(February):1–8. 662 17. Champ TM, Lee SH, Martin AB, Bolles CM, Kim SW, Gothard KM. Social engagement revealed by 663 gaze following in third-party observers of simulated social conflict. Front Psychol. 664 2022;13(December):1–16. 665 18. Ramezanpour H, Thier P. Decoding of the other’s focus of attention by a temporal cortex module. 666 Proc Natl Acad Sci U S A. 2020;117(5):2663–70. 667 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 21, 2025. ; https://doi.org/10.1101/2025.05.18.654731doi: bioRxiv preprint 19. Marciniak K, Atabaki A, Dicke PW, Thier P. Disparate substrates for head gaze following and face 668 perception in the monkey superior temporal sulcus. Elife. 2014;3(July2014):1–18. 669 20. Parr LA, Dove T, Hopkins WD. Why faces may be special: Evidence of the inversion effect in 670 chimpanzees. J Cogn Neurosci. 1998;10(5):615–22. 671 21. Bruce C. Face recognition by monkeys: Absence of an inversion effect. Neuropsychologia. 672 1982;20(5):515–21. 673 22. Gothard KM, Erickson CA, Amaral DG. How do rhesus monkeys (Macaca mulatta) scan faces in a 674 visual paired comparison task? Anim Cogn. 2004;7(1):25–36. 675 23. Wright AA, Roberts WA. Monkey and human face perception: Inversion effects for human faces 676 but not for monkey faces or scenes. J Cogn Neurosci. 1996;8(3):278–90. 677 24. Keating CF, Keating EG. Monkeys and mug shots: cues used by rhesus monkeys (Macaca mulatta) 678 to recognize a human face. J Comp Psychol. 1993;107(2):131–9. 679 25. Rigato S, Menon E, Farroni T, Johnson MH. The shared signal hypothesis: Effects of emotion-gaze 680 congruency in infant and adult visual preferences. Br J Dev Psychol. 2013;31(1):15–29. 681 26. Adams RB, Kleck RE. Effects of direct and averted gaze on the perception of facially 682 communicated emotion. Emotion. 2005;5(1):3–11. 683 27. Ricciardelli P, Lugli L, Pellicano A, Iani C, Nicoletti R. Interactive effects between gaze directio n 684 and facial expression on attentional resources deployment: The task instruction and context 685 matter. Sci Rep. 2016;6(August 2015):1–12. 686 28. Thierry B. Unity in diversity: Lessons from macaque societies. Evol Anthropol. 2007;16(6):224 –38. 687 29. Brunet N, Jagadeesh B. Familiarity with visual stimuli boosts recency bias in macaques. PeerJ. 688 2019;2019(11):1–20. 689 30. Chance MRA. Attention Structure as the Basis of Primate Rank Orders. Man [Internet]. 1967 690 Dec;2(4):503. Available from: https://www.jstor.org/stable/2799336?origin=crossref 691 31. Fox E, Mathews A, Calder AJ, Yiend J. Anxiety and Sensitivity to Gaze Direction in Emotionally 692 Expressive Faces. Emotion. 2007;7(3):478–86. 693 32. Georgiou GA, Bleakley C, Hayward J, Russo R, Dutton K, Eltiti S, et al. Focusing on fear: 694 Attentional disengagement from emotional faces. Vis cogn. 2005;12(1):145–58. 695 33. Holmes A, Richards A, Green S. Anxiety and sensitivity to eye gaze in emotional faces. Brain Cogn. 696 2006;60(3):282–94. 697 34. Hoffman KL, Gothard KM, Schmid MCC, Logothetis NK. Facial-Expression and Gaze-Selective 698 Responses in the Monkey Amygdala. Curr Biol. 2007;17(9):766–72. 699 35. Dalmaso M, Castelli L, Galfano G. Social modulators of gaze-mediated orienting of attention: A 700 review. Psychon Bull Rev. 2020;27(5):833–55. 701 36. Lobmaier JS, Tiddeman BP, Perrett DI. Emotional Expression Modulates Perceived Gaze Direction. 702 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 21, 2025. ; https://doi.org/10.1101/2025.05.18.654731doi: bioRxiv preprint Emotion. 2008;8(4):573–7. 703 37. Tomasello M. Origins of Human Communication [Internet]. The MIT Press; 2008. Available from: 704 https://direct.mit.edu/books/book/3292/Origins-of-Human-Communication 705 38. Fischer J. Primate Vocal Communication and the Evolution of Speech. Curr Dir Psychol Sci. 706 2021;30(1):55–60. 707 39. Judge PG, De Waal FBM. Conflict avoidance among rhesus monkeys: Coping with short-term 708 crowding. Vol. 46, Animal Behaviour. 1993. p. 221–32. 709 40. Davidson GL, Clayton NS. New perspectives in gaze sensitivity research. Learn Behav. 710 2016;44(1):9–17. 711 41. Waller BM, Micheletta J. Facial expression in nonhuman animals. Emot Rev. 2013;5(1):54 –9. 712 42. Parr LA, Waller BM, Fugate J. Emotional communication in primates: implications for 713 neurobiology. Curr Opin Neurobiol [Internet]. 2005 Dec;15(6):716–20. Available from: 714 http://www.ncbi.nlm.nih.gov/pubmed/16274983 715 716 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 21, 2025. ; https://doi.org/10.1101/2025.05.18.654731doi: bioRxiv preprint