Emotional states following grooming in female mandrills

Grooming is a common cooperative behavior whose exact costs and benefits are still to be fully elucidated. In this study, we evaluated the emotional consequences of giving and receiving grooming in mandrills (Mandrillus sphinx), and how these may change along time after the termination of grooming. We used scratching as a behavioral indicator of anxiety‐like emotions. Groomees showed increased scratching immediately after the termination of grooming, while in the subsequent minutes scratching decreased below baseline. The initial increase was larger after longer grooming events, suggesting it represented a case of postinhibitory rebound. The subsequent decline in scratching rates was larger after grooming received by a kin, suggesting interactions with kin are particularly relaxing. Scratching rates shown by groomers were unaffected by grooming interactions. These results highlight that the emotional states following grooming can have a complex time course, and may contribute to explain the inconsistencies found in the previous literature.


| INTRODUCTION
Among cooperative behaviors (i.e., behaviors that provide benefits to the recipient, irrespective of their fitness consequences for the actor), allogrooming is one of the most common among both mammals and birds, and play a significant role in regulating social relatioships.During allogrooming (grooming, hereafter) an animal removes ectoparasites and dead skin from the body of a partner (Clayton et al., 2010;Zamma, 2002).Given the significant negative consequences that ectoparasites have on animal health (Akinyi et al., 2013), the reduction in ectoparasite load associated with receiving grooming can have important fitness consequences (Brown et al., 1995;Lehmann, 1993).In turn, the benefits of receiving grooming suggest that natural selection should make receiving grooming pleasurable.Furthermore, since the effectiveness of grooming is maximized if the recipient remains motionless, natural selection is expected to have rendered the receipt of grooming both pleasurable and relaxing.Indeed, human observers often subjectively ascribe to animals being groomed a feeling of extreme relaxation.
Attempts to obtain a more objective test that receiving grooming is associated to relaxation and positive emotional states have been based on both physiological and behavioral measures.
Physiologically, receiving grooming has been shown to cause a reduction in heart rate and the release of beta-endorphins (Aureli et al., 1999 in Macaca mulatta; Keverne et al., 1989 in Miopithecus talapoin).Behavioral tests have traditionally been based on measuring displacement activities, that is, self-directed behaviors such as scratching that are shown under conditions of tension and motivational conflict.Displacement activities are considered a good measure of anxiety-like emotions (Maestripieri et al., 1992;Schino et al., 1996).Several studies examined the rate of displacement activities following the receipt of grooming and compared it with control conditions.The results, however, have been mixed, with different studies reporting decreases or increases in displacement activities following the receipt of grooming (Schino et al., 1988 Yates et al., 2022 in Macaca silenus).Some studies also examined the effect of giving, rather than receiving grooming, although it is theoretically less clear why giving grooming should have any emotional consequence.The results of these studies were again conflicting (e.g., Aureli & Yates, 2009 in Macaca nigra; Schino & Alessandrini, 2015 in M. fuscata).
Whatever the effect of grooming, it is clear that this effect is expected to wane progressively after the termination of grooming, although it is difficult to say how rapidly it should disappear.
However, no previous study has properly analyzed how the rate of displacement activities changes with time after the termination of grooming.This is all the more surprising, considering that monkeys have often been observed to scratch themselves right at the end of receiving grooming, so that the immediate effect of grooming seems to be an increase rather than a decrease of displacement activities.This immediate increase has been hypothesized to derive either from postinhibitory rebound (since self-directed behaviors are rare during grooming) or from frustration due to the termination of a pleasant situation (Kennedy, 1985;Schino et al., 1988).
Based on the information reviewed above, it could be hypothesized that receiving grooming should be followed by an immediate increase in the rate of scratching (the most common self-directed behavior used as an indicator of anxiety-like emotions), while in the subsequent minutes, the rate of scratching should fall rapidly below the baseline.Most previous studies have simply compared 5-15 min postgrooming (PG) observations with control observations, and the overall effect they detected may have derived from the relative prevalence of the two conflicting phenomena described above (the immediate increase or the subsequent decline; Schino et al., 1988;Ueno et al., 2015), thus explaining the inconsistent results found in the literature.
Furthermore, the effects of grooming may depend on the identity of the partner, as interactions with different partners are supposed to be more or less stressful or relaxing depending on the state of the existing social relationship.Interactions with kin are supposed to be more relaxing (Ueno et al., 2015), while interactions with higher-ranking animals are supposed to be more stressful because of the inherent risk of aggression (Schino & Alessandrini, 2015).
In this study, we examined the effect of grooming on the scratching rates shown by both the groomee and the groomer, and how these rates change with time after the termination of grooming.
For the groomee, we tested the following hypotheses and associated predictions: • Grooming has different effects immediately after its termination and in the subsequent minutes.
− Prediction: There will be a significant interaction between the effects of grooming and of time.
• The termination of grooming causes an immediate increase in scratching due to postinhibitory rebound or to frustration due to the termination of a pleasant situation.
− Prediction: Immediately after the termination of grooming, the rate of scratching will be higher than the baseline.
• In the subsequent minutes, the relaxing effect of grooming will prevail.
− Prediction: Toward the end of the 10-min observation session, the rate of scratching will be lower than the baseline.
• The immediate increase in scratching is due to postinhibitory rebound.
− Prediction: The rate of scratching immediately following grooming will be positively correlated to the duration of grooming, because the longer the inhibition, the stronger the rebound.
• Alternatively, the immediate increase in scratching is due to frustration due to the termination of a pleasant situation.
− Prediction: The rate of scratching immediately following grooming will be negatively correlated to the duration of grooming, because an earlier termination is presumably more frustrating.
• The relaxing effect of grooming received will depend on the relationship between the groomer and groomee.
− Prediction 1: The rate of scratching in the last minutes of PG observations will be lower after receiving grooming from a kin because interactions with kin are presumably more relaxing.
− Prediction 2: The rate of scratching in the last minutes of PG observations will be higher after receiving grooming from a higher-ranking group mate because interactions with higherranking monkeys are presumably more stressful.
Analyses of the effect of grooming on the behavior of the groomer were more exploratory, because it is difficult to formulate specific predictions.
• We tested whether scratching rates after giving grooming differed from control observations, and if the effect of grooming interacted with that of time.
• Given that we detected no interaction effect (see Section 3), we tested whether scratching rates in the whole PG observation varied in relation to the duration of the initial grooming episode and to kinship and rank difference between groomer and groomee.
Mandrills form very large groups living in African primary tropical forests.Females are phylopatric, while the presence of males varies seasonally (Abernethy et al., 2002).Grooming is common and is often reciprocated and directed up the dominance hierarchy (Schino & Lasio, 2018;Schino & Pellegrini, 2009).

| Subjects and housing
Subjects of this study were the nine female mandrills (Mandrillus sphinx) living in a social group that also included one adult male.The group lived in the Rome zoo (Bioparco) in a 240 m 2 outdoor enclosure connected with indoor rooms.Monkeys were fed twice a day with fresh fruit, vegetables, and monkey chow.
Degrees of maternal relatedness were derived from the genealogical records of the zoo.

| Data collection
We conducted PG and matched control (MC) observations between May 2018 and December 2019.PG observations began on the termination of a grooming event, focused either on the groomer or on the groomee, and lasted 10 min or until the focal subject entered the indoor rooms.For each PG observation, we recorded the identities of the original groomer and groomee and the duration of the grooming event (with a minimum accepted duration of 15 s).
During PG observations, we recorded the timing and the individuals involved in all episodes of scratching, grooming, aggression, approaches, and leavings (within 1 m, and only between the original groomer and groomee).MC observations were made on the next possible day and followed an identical procedure, focusing on the same subject of the corresponding PG observation but in the absence of previous grooming.We conducted a total of 248 and 254 PG-MC pairs of observations on the groomer and the groomee, respectively.

| Data analysis
All analyses were run in Stata 17.0 (StataCorp, 2021).We used survival analysis (Cleves et al., 2010) to analyze the time between the beginning of the observation and the scratching events.Since scratching can occur repeatedly during a single observation, we relied on multiple-failure survival analysis following Andersen and Gill (1982) and Cleves (1999).We considered observations to be censored at the end of the observation session or at the first occurrence of an event of aggression or grooming involving the focal subject.We tested the model assumptions following Pregibon (1980).
In the first analysis, we entered as an independent variable the type of observation (PG or MC) and allowed its effect to vary linearly along time (equivalent to an interaction effect with time).In a second analysis, we included only PG observations and entered the duration of the initial grooming event, the degree of kinship between groomer and groomed, and their rank difference as independent variables.Both analyses included subject identity as a fixed effect to obtain within-subject analyses and avoid pseudoreplication, and proximity between groomer and groomee as a time-varying control variable.

| RESULTS
Below, we report only the results useful to test our predictions.
Complete regression tables are presented in the Supporting Information.

| Behavior of groomee
Survival analysis showed that in PG observations the rate of scratching shown by the groomee was overall higher than in control observations (hazard ratio = 1.420, z = 2.55, N = 2209, p = 0.011).The hazard of scratching, however, decreased with time after the termination of grooming, so that a significant interaction between observation type (PG or control) and time from the beginning of the observation was observed (hazard ratio = 0.915, z = −3.25,N = 2209, p = 0.001). 1 showed that the rate of scratching was higher than the baseline for about 1.5 min (97 s) immediately after the termination of grooming and was lower than the baseline for about 1.5 min (92 s) at the end of the observation session.

Inspection of Figure
Limiting analysis to the first 97 s of PG observations, we found that longer durations of the grooming event received before the beginning of the observation were associated to increased rates of scratching (hazard ratio = 1.065, z = 2.50, N = 533, p = 0.012).
Limiting analysis to the last 92 s of PG observations, we found that grooming events received from a kin were associated to decreased rates of scratching (hazard ratio = 0.060, z = −2.16,N = 104, p = 0.030; Figure 2), while the difference in dominance rank between the groomer and the groomee did not affect the latter's rate of scratching (hazard ratio = 0.862, z = −1.56,N = 533, p = 0.119).

| Behavior of groomer
Analysis of the behavior of the groomer showed that giving grooming was not associated with overall changes in the rate of scratching (hazard ratio = 0.925, z = −0.50,N = 2071, p = 0.620), and that this effect did not change with time after the termination of grooming (interaction effect between observation type and time: hazard ratio = 0.970, z = −0.77,N = 2071, p = 0.443).
None of the factors that we analyzed had any significant influence on the rates of scratching shown by the groomer during PG observations (see Supporting Information: Table S5).

| DISCUSSION
The results of this study have to be interpreted with caution, given the relatively small sample size of our study.The extent to which they can be generalized to other species and age/sex classes is also to be determined.A second note of caution regards the interpretation of scratching (and of displacement activities in general) as an indicator of emotions.Displacement activities are, by definition, shown in response to situations involving frustration or motivational conflict.
Physiological and pharmacological evidence supports the use of displacement activities as indicators of anxiety-like emotions (Maestripieri et al., 1992;Schino et al., 1996) and they have proven particularly useful in understanding the emotional consequences of negative social interactions (e.g., Aureli et al., 1989).Although displacement activities have also been repeatedly used to explore the emotional consequences of positive social interactions such as grooming, their use has been less successful (see Section 1).The extent to which positive emotions such as pleasure and relaxation can be equated to a reduction in negative emotions and thus indexed by a reduction in displacement activities remains unclear.Indeed, finding indicators of positive emotions has proven difficult (Schino et al., 2016).
With these provisos, our results suggest that the effects of receiving grooming on a behavioral indicator of emotions changed along time after the termination of grooming even on a time scale of a few minutes, and were also modulated by characteristics of the grooming event itself (its duration) and of the relationship between the interacting individuals (their kinship).In contrast, we could not detect any emotional consequence of giving grooming.Our results add to a contradictory literature and may contribute to explain its inconsistencies.They suggest receiving grooming does have positive emotional consequences, although this effect is initially masked by postinhibitory rebound.Given that self-directed behaviors show such rapid changes even along the limited time span usually investigated, it is clear that analyses that do not take into account these changes risk to lump together heterogeneous phenomena.
We and others (Manson & Perry, 2000;Schino et al., 1988;Ueno et al., 2015) demonstrated that immediately after the termination of grooming received, monkeys show elevated rates of scratching that have been hypothesized to derive either from postinhibitory rebound or from frustration owing to the termination of a pleasant situation.
Our observation that this immediate increase is particularly marked after longer grooming events supports the first of these alternative explanations.In the subsequent minutes, scratching rates shown by mandrills decreased and fell below baseline in the last part of the 10min observation window.This result suggests that the relaxing effect of receiving grooming prevails after a few minutes, an interpretation confirmed by our finding that this supposed relaxing effect is particularly marked after interactions with kin.The contrasting effects observed within a few minutes of the end of grooming may explain why previous studies reported such inconsistent results.In the absence of a fine-grained analysis of temporal variations, the observed effect will depend on the relative prevalence of the initial decrease or the subsequent decline in self-directed activities.Other attempts at measuring the emotional consequences of grooming using different methods have been equally unsatisfactory (Schino et al., 2016) and may have been plagued by similar problems.
We observed a particularly marked reduction in displacement activities after interactions with kin, suggesting interactions with kin F I G U R E 2 Rates of scratching (means and standard errors) in the last part of postgrooming observations following grooming received by a kin or by a nonkin.Note that groomers were categorized as kin or nonkin only for illustrative purposes, while statistical analyses included the degree of kinship as a continuous variable.See Section 2 for details of the analyses.are particularly relaxing.Kin are generally characterized by stronger social bonds, but our sample size do not allow us to disentangle the independent effects of prolonged familiarity and actual social bonds (see, e.g., Ueno et al., 2015, for a more detailed analysis).In contrast, we did not detect any effect of dominance rank differences between groomer and groomee on the subsequent emotional state, as indexed by scratching rates.While interactions with higher-ranking animals are inherently risky and may be more stressful, monkeys also strive to interact with high-ranking group mates (see Schino & Lasio, 2018, for data on mandrills), so that conflicting processes are likely to accompany interactions between disparately ranking individuals.
Analyzing subtler aspects of the interactions, such as signals exchanged or the role of the two individuals in approaching and initiating the interaction, may allow to better comprehend their emotional consequences.Few comparable data on the factors modulating the emotional response to grooming exist in the literature.Radford (2012) found that green woodhoopoe showed larger PG decreases in self-grooming in subordinate than in dominant groomees, while Semple et al. (2013) observed an increase in macaque scratching following grooming by both dominant and subordinate group mates, but no direct comparison was made.Both Ueno et al. (2015) and Yates et al. (2022) found no effect of social bond on PG scratching.
While our results may provide an explanation for the inconsistencies observed in the effects of receiving grooming, we cannot offer a similar explanation for the analogous heterogeneity that affects the published literature on the effects of giving grooming.
Other, perhaps subtler, aspects of the grooming interaction may affect its consequences for the individual initiating grooming.
In conclusion, we urge researchers to pay more attention to the short-term temporal variations in the emotional consequences of social interactions.This seems a necessary step to reach a better understanding of the costs and benefits of social interactions in animals.

F
I G U R E 1 Time course of the rate of scratching during postgrooming observations (blue line, with 95% confidence intervals in the shaded area), and comparison with the baseline calculated in the absence of previous grooming (red line).