Introduction

Living in groups can increase interspecific competition over food, space, mates, and other resources (Arnold et al., 2011; Sterck et al., 1997; Wrangham, 1980). Consequently, recurring, intragroup aggressive behaviors such as threats, displays, and physical confrontations, lead to the establishment of dominance hierarchies (de Waal, 1986; de Waal & Luttrell, 1985; Drews, 1993; Lewis, 2019). These hierarchies are an emergent phenomenon due to the individual’s consistent competitiveness for resources (Tibbetts et al., 2022; Wittig & Boesch, 2003). In addition to engaging in dominance behaviors, individuals living in groups spend time participating in affiliative behaviors to enhance social bonds, where bondedness is associated with higher fitness (Campos et al., 2020; Ellis et al., 2019; Feldblum et al., 2021; McFarland et al., 2017; Schulke et al., 2010; Silk et al., 2003, 2009). This creates a tradeoff between competing with an individual for resources and enhancing the social bond with that individual through affiliative behaviors such as grooming and spending time in contact.

Given this tradeoff, conflict mediation is critical for maintaining strong social bonds when conflicts over resources occur. Reconciliation is one such mechanism that dyads can use to directly mediate any negative consequences of aggression on their relationship (Aureli & van Schaik, 1991; Das et al., 1998; Kutsukake & Castles, 2001; Silk, 2002; Sussman et al., 2005). Reconciliation is defined as affiliation between former opponents that takes place soon after a conflict to restore tolerance and reduce chances of future conflict (Cords, 1992). Determining what qualifies as ‘soon enough’ for reconciliation after a conflict requires a thorough understanding of the typical social dynamics of a species, or even more specifically of a group or dyad, in the absence of conflict. Given the interspecies variability in sociality, simple time-based definitions for reconciliation are insufficient (Sussman et al., 2005). One solution to identify reconciliation behavior is to compare the latency to affiliate after a conflict with the latency to affiliate during control periods without confict (Arnold et al., 2011).

A prominent hypothesis for the evolutionary function of reconciliation is the valuable relationship hypothesis, which suggests that reconciliation serves to settle conflicts between combatants whose relationship generates mutual fitness benefits (de Waal & Aureli, 1997). Thus, according to the hypothesis, reconciliation is expected to occur more frequently between close affiliative partners. This prediction has been supported in a variety of species, especially in catarrhine monkeys (as reviwed in Arnold et al., 2011). Many reconciliation studies have focused on Macaca, a genus known for its interspecific variation in the style of dominance hierarchies. Macaques exhibit a range of dominance hierarchies from highly despotic to highly tolerant (Balasubramaniam et al., 2012; Thierry, 2000). In groups with highly despotic hierarchies, social bonding opportunities are limited due to the high selectivity for kin in affiliative social interactions (Amici et al., 2020; Aureli et al., 1997), leading to relatively weak social bonds and low social tolerance. If strong despotism limits social bonding, and social bonding is dependent on repair from reconciliation, it follows that despotic species should have lower rates of reconciliation. Indeed, macaque species that are considered more tolerant tend to reconcile more than despotic species (Berman et al., 2004; Cooper & Bernstein, 2002; de Waal & Luttrell, 1985; de Waal & Ren, 1988; Demaria & Thierry, 2001; Thierry, 2000).

Due to the differences in primate societies across species and the ubiquity of affiliative and agonistic behaviors, understanding the evolutionary function of reconciliation requires investigations of the behavior across a wide range of taxa, social systems, and ecological conditions. However, studies of reconciliation are not evenly distributed across the order Primates. Reconciliation behavior has been studied extensively in catarrhine monkeys and other haplorhine primates (as reviewed in Arnold et al., 2011). Haplorhines tend to engage in affiliative behaviors, such as playing, handholding, grooming, and initiating physical contact, sooner after a conflict than in control periods (Arnold et al., 2011). Various demographic factors (such as kinship, rank, and sex) and ecological factors (such as seasonality) influence the probability of reconciliation, sometimes in species-specific ways (Kin: Aureli et al., 1997; Castles & Whiten, 1998; Judge, 1991; Kutsukake & Castles, 2001; Majolo et al., 2009; Palagi & Norscia, 2011; Rank: Aureli et al., 1993; Judge, 1991; Sex: Cordoni et al., 2006; Leca et al., 2002; Palagi et al., 2005; Penate et al., 2009; Schino et al., 1998; Season: Majolo & Koyama, 2006). Across species, dyads that are close social affiliates outside of the conflict are also more likely to reconcile (Castles et al., 1996; de Waal & Yoshihara, 1983; Kappeler & van Schaik, 1992), which supports the valuable relationship hypothesis (de Waal & Aureli, 1997).

Investigations of potential reconciliation behavior are much rarer in strepsirrhine primates (as reviewed in Arnold et al., 2011; Fig. 1). Most of the published studies focused on Lemuridae (20/22 studied groups) or groups living in captivity (13/22 studied groups). The presence and characteristics of reconciliation appear to vary within and across species of strepsirrhines, and do not cluster in the same clade (Fig. 1). For example, in Lemur catta, reconciliation was observed in some circumstances (Palagi et al., 2005; Palagi & Norscia, 2015; Rolland & Roeder, 2000), but not all (Kappeler, 1993a). Furthermore, in closely related Eulemur species, some species show reconciliation behavior (Kappeler, 1993a; Norscia & Palagi, 2011; Roeder et al., 2002), and others do not (Roeder et al., 2002). Factors that explain this apparent variation in reconciliation in lemurs may be connected to aspects of the dominance hierarchy, group demographics, seasonality, and intensity of fights. While the current literature does not suggest that phylogenetic relatedness among species predicts the presence or absence of reconciliation behavior, the lack of phylogenetic diversity in the currently published studies (Fig. 1) makes it difficult to fully understand how these patterns are organized across strepsirrhine species.

Fig. 1
figure 1

Phylogenetic tree showing the results of reconciliation studies in Lemuridae and Indriidae (Eulemur fulvus: Roeder et al., 2002; Eulemur macaco: Roeder et al., 2002; Eulemur fulvus rufus: Kappeler, 1993a; Palagi & Norscia, 2011; Lemur catta: Kappeler, 1993a, Rolland & Roeder, 2000, Palagi & Norscia, 2015, Palagi et al., 2005; Propithecus verreauxi: Lewis, 2019; Palagi et al., 2008). Phylogenetic data from 10K Trees v3 (Arnold et al., 2010).

Full size image

One challenge of the lack of phylogenetic diversity in the study of reconciliation across primate species is that strepsirrhines and haplorrhines fundamentally differ in several aspects of behavior and biology (Simmen et al., 2021) that may be salient to reconciliation behavior. For example, the social behavior of strepsirrhines and haplorrhines differs (van Schaik & Kappeler, 1996). Specifically, strepsirrhines tend to have low levels of affiliative contact (Jolly, 1998; Sussman et al., 2005), and grooming plays a less crucial role in their social networks in comparision to non-grooming behaviors (Kulachi et al., 2015, 2018; Lewis, 2019 — although see Lewis, 2010). In contrast, haplorrhines tend to affiliate by physical contact, making grooming and time spent in contact good measures for reconciliation after a conflict (McFarland & Majolo, 2013). The generally low levels of affiliative contact in strepsirrhines combined with their generally small group sizes compared to haplorrhines (Wright, 1999) makes triadic interactions (Jolly, 1966) such as polyadic conflicts (Kappeler, 1993b), coalitions, or alliances less common. In other words, individual strepsirrhines in smaller groups with fewer affiliative interactions have fewer social relationships to manage, without many chances for triadic or more complex relationships to form. This could impact reconciliation by reducing the necessity of and opportunities for complex social strategies to repair relationships after conflicts. Furthermore, strepsirrhines rely more on olfactory communication than haplorrhines (Fichtel & Kappeler, 2022). In fact, sizes of the accessory olfactory bulb, which processes social information and pheromones (Keverne, 1999), systematically vary with the social system in strepsirrhines (Barton, 2006). This may influence reconciliation behavior because individuals can assess conspecifics without any physical contact, lessening the chance of conflicts over space. Lastly, lemurs are seasonal breeders, and most show female dominance (Kappeler et al., 2022; van Schaik & Kappeler, 1996; Wright, 1999). These two concepts together can limit the importance of some social bond types and can also increase competition during the short mating season (Wright, 1999). Therefore, it is possible that conflicts during the mating season are too important to reconcile. In fact, reconciliation in Lemur catta is less common during the mating season than in the non-mating season (Schino et al., 1998; Palagi & Norscia, 2015). Thus, due to these fundamental differences between strepsirrhines and haplorrhines, fully understanding the purpose and structure of reconciliation behavior requires studying it across taxa of diverse social systems, including strepsirrhines.

Propithecus diadema and Eulemur fulvus live in sympatry in the central-eastern rainforests of Madagascar yet show notable contrasts in their respective dominance hierarchies (Kappeler et al., 2022). P. diadema lives in stable groups that are usually between two and 11 individuals and include one or more adult males and one or more adult females (Lutz et al., 2019; Powzyk & Mowry, 2003; Rasolonjatovo & Irwin, 2020). Similarly, E. fulvus lives in multi-male/multi-female social groups that can range between three and ten individuals (Ganzhorn, 1988). While Propithecus species generally show female dominance (Irwin, 2006; Pochron et al., 2003; Ramanamisata et al., 2014; Richard & Nicoll, 1987), with a higher tendency for female aggression towards males (Rasolonjatovo & Irwin, 2020), in Eulemur fulvus, females and males share near equal status (Kappeler, 1993a; Kappeler et al., 2022), and sex does not have a significant effect on the distribution of grooming and agonistic interactions (Kappeler, 1993a; Pereira et al., 1990). In fact, both male and female E. fulvus can obtain low and high dominance ranks when resources are limited (Roeder & Fornasieri, 1995). Reconciliation has been observed in captive E. fulvus (Roeder et al., 2002), but to our knowledge, it has not been studied in wild P. diadema or E. fulvus. Thus, in the present study, we investigate whether these two lemur species show reconciliation behavior in the wild.

Using five years of observational data that we collected on three Propithecus diadema groups and two Eulemur fulvus groups, we aimed to provide a more comprehensive understanding of reconciliation behavior, particularly by examining intraspecies variation across groups in strepsirrhines. To that end, we analyzed the dominance hierarchies and reconciliatory tendencies of both species. By combining the valuable relationship hypothesis and the differences between strepsirrhines and haplorrhines, we hypothesize that differences in dominance structures will correlate with reconciliatory tendencies. Specifically, we predict that the species with steeper dominance hierarchies, as a proxy for the despotism present in a group, will exhibit lower reconciliatory tendencies. Lastly, given that reconciliation in haplorrhines is often dependent on demographic attributes, we investigate whether our two strepsirrhine species follow similar demographic patterns of sex and rank impacting reconciliation rates.

Methods

Study Site

We collected all data in the Maromizaha Protected Area (S18°58.6'; E48°27.9'). This 21.5 km2 protected rainforest is in the corridor between Ankeniheny and Zahamena (CAZ; Andasibe commune, Alaotra-Mangoro region) in Madagascar’s eastern mountains and is managed by Le Groupe d’Etudes et de Recherche sur les Primates de Madagascar (GERP).

Study Groups

We collected observational data on three groups of Propithecus diadema and two groups of Eulemur fulvus across 14 months between 2015 and 2020 (Table I; full subject table available as Table S1 in the electronic supplementary material). Previous researchers habituated the P. diadema groups in 2010, and we habituated the E. fulvus groups in 2019 prior to starting data collection.

Table I Summary of group size for all study groups (Propithecus diadema and Eulemur fulvus) in the Maromizaha Protected Area, Madagascar, 2015-2020.
Full size table

Behavioral Data Collection

We collected observational data on affiliative and agonistic behavior using continuous recording focal sampling (Altmann, 1974). All group members served as focal animals, and the observational order was chosen according to a pre-randomized list. Our focal samples varied in length over the course of the study, since we originally collected the observational data used in this study for other related purposes. While focal samples from 2015 and 2016 were 30 minutes long, we increased focal sample length to 60 minutes starting in 2018. The last author collected data from 2015–2020 and was aided in 2019–2020 by a team of seven additional, trained observers. From 2015–2018, a team of two to three people followed a single study group. From 2019–2020, three observation teams (of two to three people each) rotated across the four study groups to ensure relatively even sampling across study groups.

We used a consistent ethogram across time when collecting the original data that expanded over time. From 2015–2016, we collected data on only affiliative (social play, contact, and allogrooming) and aggressive behaviors (threat). Beginning in 2018, we expanded the set of aggressive behaviors to include biting, chasing, displacing, hitting, and lunging. Full ethogram definitions of all behaviors are available in the electronic supplementary material (Table S2). For each social behavior in which the focal animal was involved, we recorded the start time, end time, initiator, and recipient. The final data set included 1422.9 hours of observations on Propithecus diadema and 631 hours on Eulemur fulvus (Table I).

Post Conflict Matching

Following methodology by Palagi et al. (2008) for studying reconciliation in Propithecus verreauxi, we defined the post-conflict (PC) period as the 15 minutes after a dyadic aggressive behavior (i.e., a conflict) took place. In post-conflict studies, it is common to collect matched control data (MC) 24 hours following the aggressive behavior (e.g., Palagi et al., 2008). This was not possible here, since the data we used were not originally collected for this purpose. Rather, we assigned 15-min matched control periods from collected focal observations using the following two criteria: (1) same focal animal and (2) within 3 hours of the time of day that the conflict happened in order to control for diel cycles. We picked the closest date that matched these conditions (median number of days between PC and MC — 2 days, interquartile range = 4 days, 20.2% (17 out of 84) of intervals were longer than 7 days). If an MC was not possible within 3 hours we eliminated the PC from the data (median interval between PC and MC time = 60.2 minutes, interquartile range = 60.3 minutes). We chose the exact start time of the MC such that the latency to conflict from the start of the focal in the PC was the same as the latency to MC start. Importantly, this procedure ensured that the PC and MC were additionally matched on the duration of observation. Twenty of the 84 PC/MC pairs were less than 15 minutes in duration because the focal observation time ended before the 15 minutes were reached (mean duration = 13.2 minutes, standard deviation = 3.9 minutes).

Statistical Analysis

We performed all statistical analyses for this study using R (R Core Team, 2023), and all code and data are available on FigShare (https://doi.org/10.6084/m9.figshare.23293043).

We analyzed dominance hierarchies using Elo ratings (Neumann et al., 2011). The Elo method is useful in cases where aggression and submission data are limited and when group membership changes over time (Neumann et al., 2011), both of which occurred in our study groups. Each individual in a group was assigned an Elo rating based on their observed aggressive and submissive interactions with other individuals. The Elo rating of an individual was then updated after each dominance interaction involving the individual. We measured the strictness of the dominance hierarchy by calculating the steepness of the dominance hierarchies using a Bayesian framework (Neumann & Fischer, 2023), which offers several advantages over the more commonly used David’s scores method, which defines steepness as the absolute slope of the straight line fitted to the normalized David’s scores (De Vries et al., 2006). The Bayesian procedure is similar to the original in that steepness is defined as the slope of a best-fit line; however, the line regresses the summed winning probability on the mean ordinal rank across iterations (Neumann & Fischer, 2023), which keeps the new value on the same scale as the original measure. As it is a Bayesian procedure, it generates distributions of plausible steepness values, rather than point estimates (Neumann & Fischer, 2023). This Bayesian procedure also eliminates the need to assign arbitrary start values and a k value (a scaling factor for the amount to which Elo ratings change after an interaction), but rather implements them as distributions that are influenced by the data itself (Neumann & Fischer, 2023). Importantly, the scores are robust to the density of interactions (Neumann & Fischer, 2023), meaning that large or small datasets are not more likely to have a particular steepness value. This allows the analysis of smaller datasets than is typical with David’s scores steepness. We used the ‘EloSteepness’ package in R to calculate steepness values (Neumann, 2022).

Many analyses of reconciliation have used latency to affiliation as a metric (e.g., de Waal & Yoshihara, 1983; Kutsukake & Castles, 2001; Majolo & Koyama, 2006). However, due to the sparsity of our data, we opted to analyze the occurrence of any affiliation as a binary variable. We modeled the occurrence of affiliation in an observation period with a logit link function as a multi-membership model implemented in brms (Bürkner, 2017, 2018). This type of model allowed us to account for repeated observations on dyads while recognizing that individuals are also a part of multiple dyads. Our predictor of interest was the interaction between the type of observation period (PC vs MC) and species. We also included an offset of the natural logarithm of observation time in hours. We included random intercepts for study group and dyad ID as well as the conflict ID, to account for the matching between PC and MC. Lastly, to control for the fact that some animals engaged in repeated conflicts across the dataset, we also included a random intercept for the dyad members. This was implemented using a multi-membership approach rather than two separate random intercepts for the aggressor and victim, because the multi-membership approach accounts for the fact that animal A is the same animal A whether they were the aggressor or victim of a particular dyadic interaction. Two separate random intercepts would not treat animal A as the same individual when they were in different roles in the interaction. We used default priors for the random effects, and the analyses presented were robust to making these priors less informative.

Additionally, we wanted to understand the demographic factors which predicted reconciliation bouts. We coded the post-conflict interval of the PC/MC pair as reconciled if affiliation occurred in the post-conflict period but not in the matched-control period. We used a multi-membership model with a binary outcome and logit link function to analyze the odds of reconciliation. Our fixed-effect predictors included the sex combination of the dyad (male–male, male–female, or female–female) and the absolute difference in rank between the individuals, as measured from their mean rank order from the Bayesian Elo analysis. We also included a random intercept for group, dyad, and a multi-membership term for the aggressor and victim, as in the previous analysis. We did not include an offset in this model, as by design, the members of the PC/MC pair were of the same duration. We would not expect affiliation to be earlier or later in the post-conflict interval compared to control depending on the observation length, meaning an offset was unnecessary for this model.

For all analyses, we also ran the analyses excluding data prior to 2018, due to the change of ethogram in 2018. We compared nested models with and without a fixed effect of interest using Bayes factors to evaluate the weight of evidence for including the fixed effect. We use Kass & Raftery’s (1995) scale for the interpretation of the Bayes factors.

Ethical Statement

Research adhered to the legal requirements of Madagascar and was approved by Madagascar’s CAFF/CORE committee. Research was conducted as educational training and under permit numbers N°127/16/MEEF/SG/DGF/DAPT/SCBT.Re, N°68/18/MEEF/SG/DGF/DSAP/SCB, N°105/19/MEDD/SG/DGF/DSAP/SCB.Re, and N°58/19/MEDD/SG/DGEF/DGRNE. The authors declare that they have no conflict of interest.

Data Availability

The datasets analyzed during the current study and code used for analysis are available on FigShare, (https://doi.org/10.6084/m9.figshare.23293043).

Results

Dominance Hierarchies

Across the entire dataset, we observed 559 dominance interactions, 494 in Propithecus diadema (105, 172, and 217 interactions for groups 1, 2, and 3 respectively) and 65 in Eulemur fulvus (19 and 46 interactions for groups 2 and 3 respectively). Aggression was initiated by a female in 76% of the interactions (212 out of 278 aggressive interactions), by a male 23% of the interactions (65 out of 278), and by an infant of undetermined sex 0.4% of the interactions (one out of 278). Contrastingly, submission was initiated by a female in 32% of interactions (91 out of 281 submissive interactions), a male in 66% of interactions (185 out of 281), and an infant of undetermined sex in 2% of interactions (five out of 281).

For Propithecus diadema, the top-ranked animal was always the dominant female (Fig. S1 available in the electronic supplementary material). In P. diadema groups, all adult females outranked males, except for group 1 from 2015–2016, where the second adult female ranked third behind the top-ranked female and top-ranked male (who was a subadult). The identities of the lowest ranked animals varied, and included a subadult female nearing dispersal age (group 1 2015–2016), juveniles (group 1 2018), a dispersing subadult female (group 2), and an adult male (group 3). Hierarchies for Eulemur fulvus were less clear (Fig. S1 available in the electronic supplementary material). In one group (group 2), the most dominant animal was an adult male, followed closely by an adult female. Younger males fell at the low end of this hierarchy. Contrastingly, in the other group (group 3), there was a clear dominant female, a clear lowest ranked male, and the other individuals were relatively close to each other in the middle.

Dominance hierarchy steepness was generally higher for Propithecus diadema than for Eulemur fulvus (Fig. 2). The median steepness value for P. diadema over the years was between 0.70 and 0.83. For E. fulvus the median steepness value was between 0.60 and 0.66.

Fig. 2
figure 2

Distribution of dominance hierarchy steepness values for each Propithecus diadema and Eulemur fulvus group in the Maromizaha Protected Area, Madagascar in various data collection periods between 2015 and 2020. Data for Propithecus diadema group 1 are split between 2015–2016 and 2018 due to our expansion of the ethogram in 2018. Dashed vertical lines represent the most frequent steepness value for the study group.

Full size image

Conflict Behavior

We recorded a total of 84 paired PC/MC observations across species. Of the 84 conflicts, 68 were observed in the Propithecus diadema groups (approximately one conflict every 20.9 hours), and 16 were observed in the Eulemur fulvus groups (approximately one conflict every 39.4 hours). The conflicts for P. diadema were most common in group 1 (31 conflicts) and more evenly spread across the remaining groups (18 and 19 conflicts in groups 2 and 3 respectively), but the E. fulvus conflicts were highly skewed towards a single group (2 and 14 pairs respectively in groups 2 and 3). Most dyads did not engage in conflicts. Within our sample, 35% of P. diadema dyads (43 out of 123 possible dyads) and 20% of E. fulvus dyads (11 out of 56 possible dyads) engaged in a conflict. Between dyads that had a conflict, there was a median of one conflict per dyad (Interquartile range = 1, range = 1–4).

Post-Conflict Affiliation

We found very strong evidence that affiliation patterns differed by species (Bayes factor in favor of model accounting for species = 673.79; see full model results in Table S3). For Propithecus diadema, 95% of the generated posterior distribution for the log odds ratio (\(\text{log}(\frac{Odds\; of\; affliation\; PC}{Odds\; of\; affliation\; MC})\), where log odds ratio of 0 indicates equal probabilities of affiliation in post conflict and matched control) fell between 0.76 and 4.49 (median estimate for log odds ratio = 2.35; Fig. 3), providing support for the prediction that the odds of affiliation were higher in post-conflict periods compared to control periods for P. diadema. However, we did not find a difference in the log odds of affiliation for Eulemur fulvus between post-conflict periods compared to control periods (median estimate for log odds ratio = −0.85; 95% credible interval: [−2.65, 0.92]; Fig. 3). These results did not change when we excluded data gathered before 2018 due to the change in data recording (full results in Table S4).

Fig. 3
figure 3

Posterior distribution of the log odds ratio (\(\text{log}(\frac{Odds\;of\;affliation\;PC}{Odds\;of\;affliation\;MC})\)) for Propithecus diadema and Eulemur fulvus, Maromizaha Protected Area, Madagascar (2015–2020). The black, vertical dotted lines represent the median log odds ratios for the two species. The red, vertical dashed line at x = 0 represents the value if the odds of affiliation were the same in PC and MC. Posterior distributions that overlap less with 0 offer more support for the prediction that the odds of affiliation differ in PC and MC. Based on 95% credible intervals, the odds of affiliation for P. diadema were higher in post-conflict periods compared to matched control; however, we detected no difference between the conditions for E. fulvus.

Full size image

Which Propithecus diadema Dyads Reconcile?

We only analyzed demographic trends in reconciliation in Propithecus diadema, since Eulemur fulvus did not show reconciliation behavior. Out of a total of 43 P. diadema dyads who interacted, ten dyads showed reconciliation behavior. We found very strong support that sex combinations varied in how frequently they reconciled (Bayes factor comparing models with and without sex combination = 5773.07; see full model results in Table S5). Specifically, male–male dyads were more likely to reconcile than other sex combinations (Fig. 4left; 95% CI for log odds ratio of reconciliation given no difference in rank for: male–male = [0.06, 10.25]; male–female = [−12.91, −1.27]; female–female = [−9.84, 2.99]). Additionally, we found very strong evidence that the odds of reconciliation varied by absolute rank difference (Bayes factor comparing models with and without rank difference = 5903.13). Specifically, dyads where the two individuals were further apart in ranking had higher log odds of reconciliation (Fig. 4right; 95% CI for log odds ratio of reconciliation for rank difference for a male–female dyad = [−0.11, 2.38]). These results were generally consistent when we analyzed data only from 2018 and beyond (Table S6), except that male-male dyads reconciled less frequently in the smaller dataset.

Fig. 4
figure 4

Posterior distribution of the log odds ratio for reconciliation in Propithecus diadema, Maromizaha Protected Area, Madagascar (2015–2020) broken down by sex classes (left) and rank difference (right). The black, vertical dotted lines represent the median log odds ratios for each distribution. The red, vertical dashed line at x = 0 represents the value of no effect of the demographic variable on the log odds ratio of reconciliation. Based on 95% credible intervals, the odds of reconciliation were higher for male–male dyads than other sex classes and for dyads where the difference in rank was greater.

Full size image

Discussion

Propithecus diadema had more frequent paired post-conflict/matched-control intervals than Eulemur fulvus, happening approximately twice as frequently. The dominant female in P. diadema groups consistently held the highest rank, with adult females outranking adult males. However, we observed one exception in a case where the second adult female ranked third behind the top-ranked female and male. In E. fulvus, dominance hierarchies were less clear, with one group showing a dominant adult male followed closely by an adult female, while in another group, only a dominant female and a lowest-ranked male were evident. We observed reconciliation behavior only in P. diadema, in which the odds of affiliation were higher in post-conflict periods compared to matched control. We detected no difference between the conditions for E. fulvus. While 40% of P. diadema dyads who interacted reconciled (23 out of 58 dyads who interacted), various demographic classes did not reconcile at the same rates. Male–male dyads reconciled more than male–female or female–female dyads. Dyads of more disparate ranks also reconciled more.

Differences in Dominance and Post-Conflict Behavior Across Species

Our finding of steep and well-defined hierarchies with alpha females in Propithecus diadema and less defined hierarchies not based on sex in Eulemur fulvus is consistent with other findings in these species and their congeners. For example, P. diadema in Tsinjoarivo generally show female dominance hierarchies as measured by female leadership of group movements and displacements of males during feeding events (Rasolonjatovo & Irwin, 2020). Although data from additional P. diadema populations is lacking, both rainforest-living and dry forest-living Propithecus show similar patterns (Pochron et al., 2003; Ramanamisata et al., 2014; Richard & Nicoll, 1987). Data on dominance patterns are more varied among Eulemur species, both in captivity and the wild (Kappeler et al., 2022). How dominance is expressed (Erhart & Overdorff, 1999; Jacobs et al., 2008) and which sex is dominant (Digby & Kahlenberg, 2002; Marolf et al., 2007) can both vary among Eulemur species. This species-level variability in dominance patterns can trickle down within populations to group-level differences (Marolf et al., 2007; Ostner & Kappeler, 1999; Pereira & McGlynn, 1997; Roeder & Fornasieri, 1995), which aligns with our finding of group-specific hierarchies.

Our post-conflict reconciliation results across species did not support our original prediction generated by the valuable relationships hypothesis and differences between strepsirrhines and haplorrhines. While we predicted that steeper dominance hierarchies would be associated with decreased reconciliation of conflicts, we observed the exact opposite. Propithecus diadema (whose groups showed steeper dominance hierarchies) showed evidence of reconciliation while Eulemur fulvus (whose groups showed less steep (i.e., more shallow) dominance hierarches) did not show evidence of reconciliation. These results diverged from previous work in catarrhine monkeys, particularly Macaca, that suggests that despotic hierarchies are associated with lower reconciliation tendencies (Berman et al., 2004; Cooper & Bernstein, 2002; de Waal & Luttrell, 1985; de Waal & Ren, 1988; Demaria & Thierry, 2001; Thierry, 2000). Without further data on additional strepsirrhine species and populations, it is difficult to pinpoint a specific cause leading to this contrast between the haplorrhine literature and our results on two strepsirrhine species. While it could reflect a true difference in the social behavior of the two primate suborders, it is also possible that there are more nuanced relationships between reconciliation behavior, despotism, and hierarchy steepness that have yet to be explored.

In our case, it may be that dominance hierarchies control behavior less in Eulemur fulvus, particularly due to the low rate of aggression. Thus, there would be less selective pressure on the evolution of reconciliation behavior. In species with a higher rate of aggression, reconciliation behavior may evolve as a mechanism to reduce the cost of aggression. Without frequent aggression, there may not be enough of a selective pressure on individuals to evolve reconciliation. Further investigation into the rate of aggression, rather than the steepness of the hierarchy alone, is needed to fully understand the lack of evidence of reconciliation observed in the study.

Our results for the two species were partially consistent with the existing literature on reconciliation in Propithecus and Eulemur. While reconciliation behavior has not been previously reported in Propithecus diadema to our knowledge, it aligns with results in other congener species (Propithecus verreauxi: Lewis, 2019; Palagi et al., 2008). For example, P. verreauxi living in southeastern Madagascar reconciled their conflicts, particularly those of low intensity or in non-feeding contexts (Palagi et al., 2008). Similarly, P. verreauxi living in western Madagascar engaged in affiliative interactions within 15 minutes of a conflict (Lewis, 2019). Our results, however, diverged from past research on E. fulvus, where affiliative contacts occurred sooner in PC periods than MC periods (Roeder et al., 2002). Lemurs in these previous studies lived in captivity, which has been proposed as a source of increased reconciliation due to the space constraints and inability for individuals to move away from others (Judge, 2000). However, reviews of the literature in this area have suggested that this is not a consistent driving factor in the presence or absence of reconciliation behavior in a particular group of primates (Aureli et al., 2002; Colmenares, 2006). Thus, it seems unlikely that the difference in captivity versus wild is the sole explaining factor in the difference of results. Other research on Eulemur has shown mixed results across species, including reports of reconciliation and reports of no reconciliation (Kappeler, 1993a; Palagi & Norscia, 2011; Roeder et al., 2002), so further investigation into their social dynamics within and across species is warranted.

Differences Across Dyads in Reconciliation Rates in Propithecus d iadema

Within our observations, less than one-fourth of the Propithecus diadema dyads who engaged in dominance interactions reconciled, with uneven distribution across demographic variables. In P. diadema we found that male–male dyads reconciled more than any dyads involving females. In related species, such as P. verreauxi, rates of male–male grooming are higher than male–female or female–female grooming, and males do not discourage multimale groups from forming (Lewis, 2008). While the benefits of living in a multimale group for males are not well understood, if they exist at all (Ostner & Schulke, 2014; Port et al., 2012), further research should be conducted to identify possible aspects of value in the relationship (for example, coalitionary support: Leimberger & Lewis, 2017; Arnold et al., 2011), as would be suggested by the valuable relationship hypothesis.

Interestingly, in terms of Bayes factors, the effect of rank difference was approximately equal to the effect of sex in our dataset, supporting the inclusion of the parameter. While this corroborates much of the literature on rank difference and reconciliation patterns where individuals of disparate ranks are more likely to reconcile (Judge, 1991; Silk et al., 1996; Palagi et al., 2008), the parity in the size of effects for rank and sex in Propithecus diadema may reflect the female dominant social structure that manifests as sex-biased aggression (Rasolojavato & Irwin, 2020). An interesting consideration for future work is the effect of rank difference and individual rank on reconciliation behavior. For example, the relationship value may not be the same between a low- and mid-ranking animal versus a mid- and high-ranking animal, despite having similar rank differences due to the resource holding potential of top-ranked animals. The valuable-relationships hypothesis would then predict differences in the probability of reconciliation based on dominance rank alone regardless of rank difference. In this case, investigating these concepts in a species that lives in larger groups with frequent conflicts may be necessary to generate the data richness necessary to differentiate between ordinal rank and rank difference.

Caveats and Ideas for Further Study

While interpreting these results, it is important to consider that there is currently no consensus on what duration should be used for PC/MC observations. A common duration is between 5 and 15 minutes (for examples see: Arnold, 1997; Castles & Whiten, 1998; Kutsukake & Castles, 2001; Leca et al., 2002; Matsumura, 1996; Norscia & Palagi, 2011; Palagi & Norscia, 2011), but some studies have used durations up to 70 minutes resulting in conflicting results (Rolland & Roeder, 2000). For this study, we decided to use 15-minute intervals to align with previous research on Propithecus and Eulemur (Norscia & Palagi, 2011; Palagi et al., 2008; Palagi & Norscia, 2011). Given that E. fulvus has long periods of inactivity during the day due to its cathemeral activity pattern (Razanaparany & Sato, 2020), it is possible that this species reconciles but just on a much longer time-scale that our analysis fails to capture.

Additionally, the data used in this study, while not originally collected with the specific aim of investigating lemur reconciliation, provided a valuable resource for addressing these research questions. One notable advantage of this dataset was the possible inclusion of matched controls that preceded the conflicts, which is typically not feasible in traditionally collected data where conflicts cannot be anticipated. Thus, the dataset presented opportunities to analyze control data without any influence of the conflict, even if it is long-lasting.

Several other sociodemographic factors beyond the scope of the study are important in structuring reconciliation patterns across dyads, and may have impacted the results reported here. For example, across primate species, reconciliation is generally more common in close genetic kin than non-kin (as reviewed in Silk, 2002; Arnold et al., 2011). At this time, the relatedness among group members in our population is unknown, so we were unable to directly test this hypothesis. Interestingly, among eastern sifakas (Propithecus diadema, P. edwardsi, P. candidus, P. perrieri, P. tattersalli), both sexes regularly disperse (Irwin, 2006; Irwin et al., 2019; King et al., 2009), meaning that both males and females regularly live in groups that vary in relatedness to them (including groups where a focal member has close genetic kin and groups where a focal member is relatively unrelated to the rest of the group). This forms an interesting system for evaluating how reconciliation functions dependent on the kin relations present in the group, and deserves further study combining genetic relatedness estimates with behavioral data. Additionally, social bonding is also known to influence reconciliation patterns across species (Castles et al., 1996; Cooper et al., 2005; Pereira et al., 2000; Romero et al., 2009; Schino et al., 1998), including in Propithecus (Palagi et al., 2008). This aspect supports the valuable relationships hypothesis (de Waal & Aureli, 1997), which suggests that reconciliation serves to repair relationships that were damaged by conflict. Although beyond the scope of this manuscript, previous work in the population has suggested that sifaka play partner choices that are correlated with grooming partner choice (Lutz et al., 2019). Sifaka tended to play with like-aged, like-ranked, and same sex animals, so these preferences in social bonding may also influence the distribution of reconciliation behavior within the group.

Temporal differences in reconciliation may be expected, especially considering lemurs’ strict seasonal breeding (Wright, 1999). These temporal variations could be due to changes in resource availability (Verbeek & de Waal, 1997) and/or reproductive seasonality (Majolo & Koyama, 2006; Palagi et al., 2005; Palagi & Norscia, 2015; Schino et al., 1998; Silk, 2002; but see Aureli et al., 1993). While changes in resource availability are known to alter many aspects of primates’ behavior, this may not drive seasonal patterns of reconciliation patterns, as feeding-related conflicts are not reconciled as frequently (Aureli et al., 2002; Castles & Whitten, 1998; Schaffner et al., 2005; Westlund et al., 2000). However, in terms of the seasonality in mating, our data are aggregated across multiple seasons and did not include data from the gestation period. This means that it may not fully capture the nuances of time-sensitive variables that influence reconciliation behavior such as seasonality in mating. Interestingly, Lemur catta reconcile the least during the mating season, a season that is represented in our dataset (Palagi & Norscia, 2015). Therefore, further investigation is warranted to examine the potential seasonal distribution and its impact on the observed patterns.

We chose to focus on dominance hierarchies as a proxy of the differences between species, given the pervasiveness of female dominance in the lineage and the relationships between dominance, power, conflict, and affiliation. Dominance hierarchies can influence who in the group has access to food resources and mating opportunities, with dominant individuals having priority access. In lemurs, for example, dominant females have been found to have greater access to food resources (Overdorff et al., 2005; Rasamimanana, 1999; Rasolonjatovo & Irwin, 2020; White et al., 2007), which may be an avenue that leads to higher reproductive success for these dominant animals (Gouzoules et al., 1982; Pusey et al., 1997). Additionally, dominance can influence patterns of aggression and conflict, with dominant individuals being more likely to win fights and maintain social order (Kappeler et al., 2022; Seex et al., 2022). Lemur social affiliation patterns are also governed by differences in dominance rank in many species (Erhart & Overdorff, 2008; Jolly, 1998; Lewis, 2010; Lutz et al., 2019; Port et al., 2009; Waeber & Hemelrijk, 2003). However, the studied species differ in other aspects, even though they were studied sympatrically. Specifically, factors such as group size, diel period, mating competition, diet, and territoriality can differ between these genera. For example, mating competition and reproductive strategies can shape dominance interactions (Gachot-Neveu et al., 1999; Sauther, 1993). Furthermore, dietary variations stemming from differences in resource distribution can alter social dynamics (Gemmill & Gould, 2008; Janson, 1985). Without further data on additional populations of these species and/or more longitudinal data on this population, it is difficult to tease apart these effects from differences in dominance hierarchies.

Conclusion

We had evidence of reconciliation in Propithecus diadema, but not in Eulemur fulvus. This finding did not support our hypothesis that strict dominance hierarchies would be associated with lower rates of reconciliation, and may suggest that frequent aggressive bouts are needed to support the evolution of reconciliation. We found that within P. diadema, male–male dyads were more likely to reconcile, and that dyads further apart in rank tended to reconcile more. Further studies on reconciliation in P. diadema are needed to clarify whether our results establish a species-specific pattern or whether they are site- or group-specific. Additional studies on other strepsirrhine species, closely related or not, could lead to further understanding of the evolutionary forces driving reconciliation behavior and how the behavior may or may not function differently in strepsirrhine and haplorrhine societies.