Abstract

Adolescence is a sensitive period for the maturation of neural circuits governing goal-directed social behaviors and stress regulation. Disruption of stable social relationships during adolescence can alter neuropeptide and dopaminergic systems that shape adult social behaviors. We investigated the behavioral and neurobiological consequences of adolescent social instability stress (SIS) in male prairie voles (Microtus ochrogaster), a species that forms selective social bonds between peers, mating partners, and parents and their offspring. During adolescence, SIS subjects experienced repeated reshuffling of cage mates to disrupt stable peer bonds, while control (CTL) subjects remained in fixed pairs. Home cage observations after and right before each reshuffling revealed that SIS subjects exhibited reduced affiliative contact and sustained social investigation compared to CTL subjects, despite no group differences in body weight throughout adolescence. Moreover, SIS and CTL groups did not differ in social zone duration or latency to approach a novel conspecific during the social approach test (SAT). Stress phenotypes were classified by assessing the duration of social zone occupancy during the SAT under baseline and stimulus-present conditions. Remarkably, all SIS subjects expressed a consistent stress resilient phenotype in contrast to CTL subjects whose responses were more variable, spanning both stress resilient and susceptible phenotypes. Gene receptor expression analyses revealed no group differences in oxytocin (Oxtr), arginine vasopressin (Avpr1a), and dopamine (Drd1 and Drd2) gene expression within the lateral septum (LS), nucleus accumbens (NAc), or anterior cingulate cortex (ACC), brain regions important for modulating goal-directed social behaviors and stress responses. However, correlation analyses indicated distinct relationships between gene receptor expression and social behaviors across groups, including a negative association between LS- Avpr1a expression and the latency to approach a novel conspecific in only CTL subjects. Additionally, associations between ACC-Drd2 expression and the latency to approach a stimulus were in opposing directions between groups. Correlation analyses solely between gene receptor expression revealed the loss of oxytocin-dopamine receptor coupling in the LS and ACC of SIS but not CTL subjects. Together, these findings suggest that adolescent SIS does not globally suppress social behavior but instead may reorganize social reward circuitry to promote behavioral flexibility and stress resilience.

Introduction

The integrity of adult brain function and behavior is profoundly shaped by experiences during sensitive developmental periods. Adolescence, a transitional stage characterized by rapid social and neurobiological maturation, represents a critical window of vulnerability to environmental stressors (Andersen and Teicher, 2008; Tottenham and Galvan, 2016). Social adversity encountered during this period can profoundly alter neural circuitry, producing enduring effects on emotional regulation and social functioning. Such experiences are linked to long-lasting impairments in adult social behaviors, such as reduced affiliation and heightened aggression, as well an increased risk for mood disorders, including anxiety and depression.

The Social Instability Stress (SIS) paradigm, which is based in principles of neuroethology for its ecological relevance, has been used to effectively model the complexity of social stress and has been used to understand the effects of adverse social environments on brain and behavior (Mccormick et al., 2015; Burke et al., 2017; Koert et al., 2021). In this paradigm, repeated reorganization of cage-mates disrupts the formation and maintenance of stable social relationships, effectively modeling the instability of a dynamic social environment. In principle, the SIS paradigm maps onto the everchanging social dynamics of social turnover of group living animals, and SIS is considered ecologically relevant to humans because it mirrors experiences faced by children and adolescents exposed to residential instability, unstable family structures, or peer rejection (Zehrung et al., 2024; Fomby and Sennott, 2013). These forms of social stress disrupt critical social bonds, alter brain development, and can potentially lead to reduced social motivation.

The present study adapts this robust SIS paradigm to the prairie vole (Microtus ochrogaster), a socially monogamous rodent that exhibits strong and enduring sex-naïve peer bonds and mating-induced pair bonds (Getz et al., 1981; Getz et al., 1993; Lee and Beery, 2021). The prairie vole offers a valuable opportunity to investigate the neurobiology of social attachment, because its socially selective behaviors can more closely parallel human social relationships than those of traditional laboratory rodents. Social behaviors in prairie voles are orchestrated by the coordinated actions of oxytocin (OXT), arginine vasopressin (AVP), and dopamine (DA) signaling within limbic and cortical circuits associated with social bonding and stress regulation (Aragona and Wang, 2009; McGraw and Young, 2010). Specifically, oxytocin receptor (Oxtr) and arginine vasopressin receptor 1a (Avpr1a) gene expression and/or the protein receptors they encode in the lateral septum (LS), nucleus accumbens (NAc), and anterior cingulate cortex (ACC) modulate social recognition, pair bonding, and empathy-related behaviors (Lim and Young, 2006; Prounis and Ophir, 2020). As a central node of the Social Decision-Making Network (O’connell and Hofmann, 2012), the LS integrates social and emotional signals, and it contributes to social recognition, social behaviors, and anxiety regulation (Singewald et al., 2003; Sheehan et al., 2004). The NAc, a central component of the mesolimbic reward pathway, mediates motivational aspects of affiliative and partner-directed behaviors through dopaminergic signaling (Liu and Wang, 2003; Aragona et al., 2006). Within the NAc, the dopamine receptor subtypes D1R (Drd1) and D2R (Drd2) play opposing roles, where D2R activation facilitates pair bond formation and D1R activation is associated with the maintenance of selective affiliation and the exclusion of novel conspecifics (Aragona et al., 2006). The ACC contributes to the evaluation of social context, empathy-like processes, and adaptive responses to social stress (Allsop et al., 2018; Prounis and Ophir, 2020).

Extensive research focusing on the physiological and behavioral impacts of social instability stress with non-monogamous rodents has revealed profound effects on stress systems, emotional behaviors, and cognitive functions (Mccormick and Green, 2013; Koert et al., 2021; Mccormick et al., 2015). However, the longstanding effects of adolescent social instability stress on sub-adult social behaviors in a species capable of forming selective attachment to conspecifics remain poorly understood. This gap limits our understanding of how early-life social adversity influences the neural systems that govern experience-dependent social motivation, affiliative responses, and stress susceptibility. To bridge this gap, the present study examined the behavioral and neurobiological consequences of adolescent SIS in male prairie voles. We hypothesized that exposure to SIS during adolescence would disrupt affiliative and investigative interactions necessary for the formation of peer bonds and lead to persistent deficits in sub-adult social approach and enhance stress susceptible phenotypes. Furthermore, we predicted that these behavioral outcomes would be associated with dysregulation of Oxtr, Avpr1a, Drd1, and Drd2 mRNA expression within the LS, NAc, and ACC.

Materials and methodsAnimals

Prairie voles (N = 20) used in this experiment were F1–F2 generation descendants bred from wild-caught voles originally trapped in Champaign-Urbana, IL and raised by both parents. All animals were housed in transparent polycarbonate rodent cages (46.5 × 25 × 15.5 cm) lined with Sani-chip bedding (P. J. Murphy Forest Products, Montville, NJ, United States). All animals were given wooden chew blocks and cotton nesting material for environmental enrichment, and ad libitum access to food (Laboratory Rodent Diet 5001, LabDiet, St. Louis, MO, United States) and water. Sex was determined based on external genitalia. At weaning on postnatal day (PND) 21, male subjects were ear-tagged and loosely fitted with zip-ties around their necks for identification. From PND21 until the start of testing on PND31, male subjects were pair-housed with a male sibling. Handling was limited to weekly cage changing days. The experimental procedures followed the guidelines of the National Institutes of Health and were approved by the Cornell University Institutional Animal Care and Use Committee (protocol # 2013-0102).

Social instability stress paradigm

At weaning, paired subject littermates were randomly assigned to either the social instability stress (SIS; 10 males) group or the control (CTL; 10 males) group. Subjects in the SIS group were paired with a different SIS individual every 2 days in a new cage over the course of 10 days, with each subject experiencing five different social encounters (i.e., “Pairing Periods”) from PND31-41 (Figure 1). To minimize the number of animals needed for this study, the 10 males used in the SIS condition were used to create each of the five unique pairings for each male. For each male, the series of pairing periods was unique to that male. Thus, both males in each SIS pair over the course of the study served as a focal subject; however, the series of pairings (i.e., social encounters) were unique to each subject (see statistical analyses below). CTL subjects were placed in a new cage with the same littermate every 2 days across all pairing periods, which controlled for any stress-induced effects by frequent cage changes.

Experimental design. During the social instability stress (SIS) paradigm, subjects (purple) were re-paired with a novel conspecific (also serving as an SIS subject) every 2 days from PND31 to PND41 for a total of 5 Pairing Periods. CTL subjects (green) remained pair-housed with the same littermate throughout PND31-41 but were placed in a clean cage every 48 h. Social interactions were recorded in the first 5 min (Introductory Period 1–5, IP1–5) of subjects being placed in a new home cage at the beginning of each Pairing Period. The last 5 min of the 2-day cohabitations were recorded (Familiarity Period 1–5, FP1–5) for each Pairing Period. Change within each pairing period (ΔPPn) was calculated as the difference between FPn and IPn (FPn - IPn). Sociability of the SIS and CTL groups was assessed in the Social Approach Test on PND41 (also see Figure 2). Brains were extracted after subjects were sacrificed (within 10 min after the SAT).

Experimental setup for the social approach test (SAT). (A) The subject (green vole) was placed in the main chamber for 10 min of habituation in the absence of a stimulus animal. (B) Subjects were briefly removed from the main chamber and a presentation chamber containing a stimulus animal (black vole) was attached to the main chamber. The subject was returned to the main chamber for another 10 min to assess social approach behaviors. The social zone was defined by the distance from the wall attached to the stimulus box in which an average prairie vole (3 cm length) could fit.

Home cage observations during the SIS paradigm

Home cage observations between subjects in the same group were conducted to characterize spontaneous, naturalistic, and physical social behaviors. We recorded the first 5 min when subjects were introduced to either a novel conspecific (in the SIS condition) or the same littermate (in the CTL condition) during the five home cage pairing periods on PND31, 33, 35, 37, and 39 (Figure 1). These initial social encounters were intended to capture the introductory phase of the pairing period. We also recorded the last 5 min of the 2-day pairing period on PND33, 35, 37, 39, and 41 to capture the social encounters after familiarity was established. We refer to these two recording phases as the Introductory Period (IP) and the Familiarity Period (FP), respectively (Figure 1). For each pair, the total time engaged in social behaviors between conspecifics was scored because zip tie colors were blocked by fur during substantial portions of the recordings and the video resolution was not optimal making it difficult to discriminate between cagemates in video recordings. Home cage observation sessions were scored using BORIS (version 8.23) behavioral coding software (Friard and Gamba, 2016) for huddling, allo-grooming, social investigation (flank sniffing, anogenital sniffing, following, and nose-to-nose sniffing), and aggression (chasing and wrestling). The first familiarity period (FP1) between animals in CTL3 and CTL4 had to be excluded due to technical difficulties.

Body weight measurement

The body weight of each subject was recorded on each cage-change day (PND31, 33, 35, 37, 39, and 41). The overall body weight gain was calculated (PND41 body weight—PND31 body weight) and used as a measure of physical development in response to social instability stress.

Social approach test

All subjects underwent a Social Approach Test (SAT) after the FP5 on PND41 (Figures 1, 2). The SAT enabled us to evaluate social approach under conditions that preclude direct physical contact and therefore to isolate approach motivation independent of reciprocal social interaction or aggression and stress phenotypes (Sailer et al., 2022b; Golden et al., 2011). The social approach apparatus consisted of a main testing chamber (20 × 40 × 28 cm) with a doorway that can be blocked or serve as an attachment point for a presentation chamber (10.06 cm3) with a perforated plexiglass wall between the chambers. The perforated wall allows olfactory, visual, and auditory interactions between the subject and stimulus animal, while preventing physical contact. Subjects were habituated to the main testing chamber and testing room for 10 min; no presentation chamber was attached, and the doorway was blocked by a 10.06 cm2 plexiglass barrier (Figure 2A). At the same time, a novel sex-matched and age-matched stimulus animal was habituated in the presentation chamber. Subjects were briefly removed from the main testing chamber, the barrier was removed, and the presentation chamber containing the stimulus animal was attached. The SAT began when subjects were returned to the main testing chamber (Figure 2B). Interactions with the stimulus were recorded for 10 min. The section of the testing chamber closest to the stimulus animal, measured as one body length of a vole (within 3 cm of the stimulus chamber), was used to define the “social zone.” The rest of the testing chamber was considered the “non-social zone.”

The videos from the SAT were scored and analyzed using Noldus Observer XT13 video scoring software (Noldus, 1991) to measure time spent in the social and non-social zones, latency to approach a stimulus, Social Investigation (SI) ratio (time spent in the social zone with the stimulus present divided by the time spent in the social zone with the stimulus absent), frequency of visits into each zone, distance moved, and velocity. The SI ratio was used to define stress susceptible and stress resilient phenotypes (Sailer et al., 2022b; Golden et al., 2011). Subjects with SI ratios < 1.0 were defined as displaying stress susceptibility and subjects with SI ratios > 1.0 were defined as displaying stress resiliency. One SIS animal had to be excluded from the SI ratio analysis because it did not enter the social zone when the stimulus animal was absent and its SI ratio was undefined.

Tissue processing

Subjects were euthanized following recommended ethical and regulatory guidelines: with CO2 asphyxiation (displaced between 30 and 70% of the chamber volume per minute) before rapid decapitation on PND41 immediately after the SAT. Brain tissue was immediately extracted and frozen on powdered dry ice before being stored at −80°C. Coronal sections (200 μm) were selected that anatomically matched Plates 14–34 for the lateral septum (LS), Plates 15–25 for the nucleus accumbens (NAc) and cingulate cortex (ACC) from Paxinos and Watson’s rat brain atlas (Paxinos and Watson, 2007). Tissue punches (1 mm diameter) were collected bilaterally from the LS, NAc, and ACC for each subject, and stored at −80°C until further processing for total RNA extraction using TRI-reagent according to the manufacturer’s protocol (Molecular Research Center) and as previously described (Wang et al., 2013; Sailer et al., 2019; Sailer et al., 2022b).

Gene receptor expression analysis

Total RNA (200 ng) was reverse-transcribed with the LunaScriptTM RT SuperMix Kit (New England Biolabs, E2010) to examine the mRNA expression for Oxtr, Avpr1a, Drd1, and Drd2 RT-qPCR in triplicates (see Supplementary Table S1 for primer sequences) for each subject. Primer specificity was verified by the melt curve analysis. For each primer pair, amplified cDNA was normalized to nicotinamide adenine dinucleotide dehydrogenase (Nadh) (Wang et al., 2013; Sailer et al., 2019; Sailer et al., 2022b). All data were included in the analyses unless statistically defined as an outlier (> 2 standard deviations from the mean).

Statistical analyses

The data from the home cage observations, body weight measurements, the social approach test, and RT-qPCRs were analyzed using the Grubbs test to detect and remove significant outliers. Home cage observations of behavior (huddling, allo-grooming, social investigation, and aggression) were analyzed with R software (version 4.5.1). Because we were unable to individually identify each animal within a pair during the home cage observations (i.e., pairing periods), the nature of our data was interdependent but the sequence of pairings for each subject was unique. Thus, we analyzed our data using a linear mixed-model (LMM) framework with the packages lme4 (Bates et al., 2015), with fixed effects of Group (CTL vs. SIS) and Pairing Period, and with random effects of subject ID and pair ID. Although the data give the visual impression of pseudoreplication, leveraging an LMM with these random effects included account for this apparent non-independence across the repeated measures, and correctly estimates the variance contributed at each level (see Brown, 2021). Individual analyses were used to compare these behaviors during the Introductory Periods (IP1–5), the Familiarity Periods (FP1–5), and to compare the change within each pairing period (ΔPPn = FPn - IPn). Because subject within the dyads were indistinguishable, subject ID and dyad ID were included as random effects in the LMM for home cage observations. Significant interactions or significant main effects (α = 0.05) were followed by the R package emmeans (Lenth et al., 2022). Body weight, SAT, and RT-qPCR data were tested for normality (Shapiro–Wilk test) and equal variance to compare the CTL and SIS groups by using Prism (version 10.6.1, GraphPad Software, San Diego California United States). When data were found to be normally distributed, an unpaired t-test (two-tailed) was performed. If data were not normally distributed, a Mann-Whitney test was performed. Body weight and body weight change were analyzed using an LMM to compare the effects of Group (CTL vs. SIS) and Postnatal Day (PND31 vs. PND41). Behaviors from the social approach test were analyzed to compare the effects of Group (CTL vs. SIS) on social zone duration, latency to approach a stimulus, SI ratio, zone frequency, distance moved, and velocity. The Wilcoxon signed-rank test was used to evaluate SI ratio medians to the hypothetical mean of 1, which is used to define stress susceptible (< 1) and stress resilient (> 1) phenotypes. Figures 3–8 and Supplementary Figures S1, S2 were created using Prism (version 10.6.1, GraphPad Software, San Diego California United States) and all data are presented as the means ± standard error of mean (SEM). We performed Pearson correlations between mRNA gene receptor expression within each group and p-values were adjusted for multiple comparisons with False Discovery Rate (FDR) (Benjamini and Hochberg, 1995). The Hmisc and corrplot packages in R were used to visualize the correlograms in Figure 9 and Supplementary Figures S3–S8 (Wei, 2024; Harrell and Dupont, 2025).

Duration of huddling during home cage observation. Introductory Periods 1–5 (A) occurred during postnatal days 31, 33, 35, 37, and 39, respectively. Familiarity Periods 1–5 (B) occurred during postnatal days 33, 35, 37, 39, and 41, respectively. (C) Change in huddling within Pairing Period (ΔPPn = FPn - IPn). Data are presented as mean ± SEM and dots represent individual data (n = 8–10/group); CTL group shown in green dots and bars; SIS group shown in purple dots and bars; *p < 0.05; **p < 0.01; ***p < 0.001.

Duration of social investigation during home cage observation. Introductory Periods 1–5 (A) occurred during postnatal days 31, 33, 35, 37, and 39, respectively. Familiarity Periods 1–5 (B) occurred during postnatal days 33, 35, 37, 39, and 41, respectively. (C) Change in social investigation within Pairing Period (ΔPPn = FPn - IPn). Note that the y-axes differ between panels AC. Data are presented as mean ± SEM and dots represent individual data (n = 8–10/group); CTL group shown in green dots and bars; SIS group shown in purple dots and bars; **p < 0.01; ***p < 0.001.

Duration of aggression during home cage observation. Introductory Periods 1–5 (A) occurred during postnatal days 31, 33, 35, 37, and 39, respectively. Familiarity Periods 1–5 (B) occurred during postnatal days 33, 35, 37, 39, and 41, respectively. (C) Change in aggression within Pairing Period (ΔPPn = FPn - IPn). Data are presented as mean ± SEM and dots represent individual data (n = 8–10/group); CTL group shown in green dots and bars; SIS group shown in purple dots and bars.

Social instability stress does not impact physical development during adolescence. (A) Body weight in grams for subjects at every cage switch day of the SIS paradigm. (B) Body weight gain from PND31 to PND41 (= PND41-PND31). Data are presented as mean ± SEM and dots represent individual data (n = 10/group); CTL group shown in green dots and bars; SIS group shown in purple dots and bars.

Social instability stress promotes stress resilience. (A) Subjects in the CTL and SIS groups spent similar durations in the social zone of the social approach test. (B) SIS subjects tended to display a shorter latency to approach a novel sex- and age-matched stimulus. (C) The SI ratios between CTL and SIS groups were significantly different and only the SI ratios of the SIS group significantly differed from the theoretical median of 1. Data are presented as mean ± SEM and dots represent individual data (n = 8–10/group); CTL group shown in green dots and bars; SIS group shown in purple dots and bars. Comparison between groups: #p < 0.05; group theoretical median comparison to theoretical median: **p < 0.01.

Effects of social instability stress on gene expression in the lateral septum (LS), nucleus accumbens (NAc), and anterior cingulate cortex (ACC). (A–D) LS mRNA expression of Oxtr, Avpr1a, Drd1, and Drd2. (E–H) NAc mRNA expression of Oxtr, Avpr1a, Drd1, and Drd2. (I–L) ACC mRNA expression of Oxtr, Avpr1a, Drd1, and Drd2. Data are presented as mean ± SEM and dots represent individual data (n = 9–10/group); CTL group shown in green dots and bars; SIS group shown in purple dots and bars.

Correlograms separated by group and brain region. Pair-wise Pearson correlations with false discovery rate were calculated for mRNA expression by group in the LS (A,B), NAc (C,D), and ACC (E,F). Positive correlations are displayed in blue and negative correlations in red color. Color intensity and the size of the circle are proportional to the correlation coefficients. In the right side of each correlogram, the legend color shows the correlation coefficients and the corresponding colors. Statistically non-significant correlations are left blank. The FDR adjusted alpha was α < 0.025 for LS-CTL, α < 0.017 for ACC-CTL, and α < 0.008 for LS-SIS, NAc-CTL, NAc-SIS, and ACC-SIS (Benjamini and Hochberg, 1995).

ResultsHome cage observations during the SIS paradigm

We assessed home cage interactions between dyads throughout the SIS paradigm (Figure 1) to determine how huddling, allo-grooming, social investigation, and aggressive behavior would change over time. Home cage observations were recorded during the first 5 min of introducing subjects to one another (Introductory Periods 1–5, IP1–5) and during the last 5 min of the 2-day cohabitation (Familiarity Periods 1–5, FP1–5), right before each dyad was re-paired (Figure 1). CTL subjects were re-paired with the same litter mate throughout the experimental paradigm. We analyzed IP and FP data separately to determine whether SIS impacted social behavior upon first meetings (IP) or after familiarity had stabilized (FP). We also compared the change in duration for huddling, allo-grooming, social investigation, and aggressive behaviors between the first 5 min of being introduced to a novel conspecific (SIS) or being re-paired with the same conspecific (CTL) and the FP (last 5 min of cohabitating for 2 days with the same conspecific) to determine if the interactions between dyads differed between groups (CTL vs. SIS) and within each pairing period (ΔPPn = FPn - IPn).

Huddling duration

Huddling duration during the IP significantly differed by Group [SIS v CTL: F(1, 94) = 31.50, p = 2.01e-7; Figure 3A], but neither Pairing Period [F(1, 94) = 0.00, p = 0.10] nor the interaction between Group and Pairing Period [F(1, 94) = 0.16, p = 0.69] were significant. Post hoc comparisons showed that the CTL group significantly spent more time huddling than the SIS group during each introductory period (IP1: t94 = 3.50, estimate = 35.0, s.e. = 10.00, p = 0.0007; IP2: t94 = 4.73, estimate = 33.4, s.e. = 7.05, p < 0.0001; IP3: t94 = 5.62, estimate = 31.7, s.e. = 5.65, p < 0.0001; IP4: t94 = 4.40, estimate = 30.1, s.e. = 6.84, p < 0.0001; IP5: t94 = 2.94, estimate = 28.5, s.e. = 9.71, p = 0.004). Huddling duration during final interactions during the FP did not differ by Group [F(1, 94) = 2.01, p = 0.16] or Pairing Period [F(1, 94) = 0.31, p = 0.58]. However, we found a significant interaction between Group and Pairing Period [Figure 3B, F(1, 94) = 5.02, p = 0.03] for huddling duration. Post hoc comparisons showed that the CTL group significantly spent more time huddling than the SIS group during the familiarity periods 4 and 5 (FP4: t94 = 2.45, estimate = 34.94, s.e. = 14.3, p = 0.02; FP5: t94 = 2.66, estimate = 53.83, s.e. = 20.3, p = 0.009), but not during FP1–3 (FP1: t94 = −1.04, estimate = −21.71, s.e. = 20.9, p = 0.30; FP2: t94 = −0.19, estimate = −2.83, s.e. = 14.7, p = 0.85; FP3: t94 = 1.36, estimate = 16.06, s.e. = 11.8, p = 0.18). Lastly, we found that change in huddling duration was not significant for either Group [F(1, 94) = 1.56, p = 0.21] or Pairing Period [F(1, 94) = 0.30, p = 0.58]. However, the Group by Pairing Period interaction was significant [F(1, 94) = 5.68, p = 0.02; Figure 3C]. Overall, subjects in both groups tended to spend more time huddling with their conspecific toward the end of the pairing period (once they became familiar), producing positive mean ΔPP values. Our post hoc comparisons showed that the SIS group increased huddling durations more than the CTL group during the first pairing period (t94 = −2.66, estimate = −56.7, s.e. = 21.3, p = 0.009) and second pairing period (t94 = −2.41, estimate = −36.2, s.e. = 15.0, p = 0.018). Huddling duration did not differ between SIS and CTL for pairing periods 3–5 (ΔPP3: t94 = −-1.30, estimate = −15.7, s.e. = 12.0, p = 0.20; ΔPP4: t94 = 0.33, estimate = 4.80, s.e. = 14.6, p = 0.74; ΔPP5: t94 = 1.22, estimate = 25.3, s.e. = 20.7, p = 0.22).

Taken together, these data indicate that SIS animals engaged in less huddling overall, particularly during the introductory periods, which is to be expected for animals that are less familiar with each other. Notably, the difference in huddling between SIS and CTL males at the end of a pairing period (when familiarity should have been established) was only different in the fourth and final paring periods, suggesting that SIS males demonstrate less forms of physical attachment only after several bouts of instability. Similarly, the change in huddling within a pairing period, after 2 days of co-habituation with a new conspecific, was greatest in SIS males but only within the first two pairing periods, consistent with the idea that repeated rounds of social instability eliminated this increase huddling over a pairing period.

Allo-grooming duration

Like huddling duration, allo-grooming during the IP significantly differed by Group [F(1, 94) = 7.59, p = 0.007; Supplementary Figure S1A], but neither Pairing Period [F(1, 94) = 2.54, p = 0.11] nor the interaction between Group and Pairing Period [F(1, 94) = 2.61, p = 0.11] were significantly different. Post hoc comparisons showed that the CTL group significantly spent more time allo-grooming than the SIS group during introductory periods 1 and 2 (IP1: t94 = 2.91, estimate = 9.61, s.e. = 3.30, p = 0.004; IP2: t94 = 3.21, estimate = 7.46, s.e. = 2.32, p = 0.002). The SIS groups significantly spent more time allo-grooming than the CTL group during the Introductory Period 3 (IP3: t94 = 2.85, estimate = 5.31, s.e. = 1.86, p = 0.005). The CTL and SIS groups did not differ in time spent allo-grooming during introductory periods 4 and 5 (IP4: t94 = 1.40, estimate = 3.16, s.e. = 2.26, p = 0.17; IP5: t94 = 0.31, estimate = 1.00, s.e. = 3.20, p = 0.75). Allo-grooming duration during the Familiarity Period did not differ by Group [Supplementary Figure S1B; F(1, 94) = 1.09, p = 0.30], Pairing Periods [F(1, 94) = 0.31, p = 0.58], or an interaction between Group and Pairing Periods [F(1, 94) = 0.23, p = 0.63]. Lastly, the duration of allo-grooming did not change over each pairing period resulting in values close to zero for both groups across all pairing periods (Supplementary Figure S1C). Our analysis demonstrated that change in allo-grooming duration did not significantly differ by Group [F(1, 94) = 0.31, p = 0.58], Pairing Period [F(1, 94) = 1.62, p = 0.21] nor the Group × Pairing Period interaction [F(1, 94) = 1.50, p = 0.22]. Altogether, these data indicate that the SIS group engaged in less allo-grooming only during the first two Introductory Periods. Only 1 SIS pair and no CTL animals exhibited allo-grooming during IP3. The difference in allo-grooming between SIS and CTL males during IP4, IP5, and all Familiarity Periods did not differ, suggesting that SIS males demonstrate less forms of affiliation only after initial exposure to social instability.