Better together than alone: sociality mitigates the negative impact of early family disruption on adult problem-solving performance in zebra finches

Abstract

Adolescent cognitive development, characterized by social learning and innovation, is sensitive to social disruptions. We investigated how family disruption during emancipation affects adult problem-solving performance in social contexts. We manipulated the social environment of 22 zebra finch (Taeniopygia guttata) broods during emancipation by replacing the biological father with an unfamiliar male or by leaving the family intact. In adulthood, offspring underwent a social task, solving a novel problem with their siblings. Then, they had to complete the same task on their own to evaluate the persistence of the treatment effects after the social task. Experimental individuals were less successful than controls in the social task. Although average brood performance was similar, experimental broods showed greater heterogeneity than controls. In the isolated trial, the differences between experimental and control individuals disappeared. Our findings suggest that family disruption during emancipation affects individual problem-solving skills in social settings. However, the success of a few siblings in experimental broods, who likely shared their discoveries, compensated for this impairment and improved overall brood performance. These individuals likely provided social information during the social trial, and this probably benefitted their siblings during the isolated trial. We emphasize the importance of studying animal cognition in social settings and the role of social learning in mitigating cognitive impairments.

1. Introduction

Cognition is the process by which animals collect, retain and use information to make decisions [1], and it is influenced, at least partially, by the social environment experienced throughout life [2–4]. From prenatal stages to adulthood, the social environment shapes cognitive abilities by affecting cognitive development, neural structure, function and plasticity [5,6]. In vertebrates, early social disruptions (i.e. ‘social stress’ sensu [4,7]) can have short- and long-lasting detrimental effects on cognitive development [2,3,8–11]. Therefore, such disruptions can permanently affect both individual and social cognitive performance [12–16] and the use of personal and social information [13,17]. For instance, in numerous vertebrate species, individuals raised without parents exhibit cognitive impairments, such as poorer emotion regulation, poorer social skills, poorer learning abilities [8,18–20] and a reduced capacity to cope with environmental challenges [21–23].

Conversely, living in groups or kin-based networks can act as social buffers during or after stressful events [4,24–27], and it can even enhance cognitive development by strengthening social bonds and memory of past interactions (the social intelligence hypothesis [2,28–30], but see [30,31]). From birds to primates, substantial evidence shows that growing in enriched social environments improves communication skills, social and cooperative behaviours, courtship abilities or escape responses through imitation, social learning and cultural transmission [29,32–34]. In rodents, evidence suggests that social enrichment later in life can, to some extent, rescue cognitive deficits from isolation during development [35,36]. However, the potential of such social environments to act as a social buffer during or after a stressful event remains unexplored in other taxa, especially apart from early life disruptions.

While the prenatal and early postnatal phases are the most sensitive periods for cognitive development, adolescence is also a critical stage [37]. This phase is characterized by emancipation and nutritional independence, and it involves complex social interactions, innovation and risk-taking behaviours. It allows individuals to reach maturity while continuing to develop social skills [38,39] and cognitive control [40,41] that are essential in adulthood [9]. In birds and mammals, juveniles acquire social information through the observation of their parents, siblings or unfamiliar individuals, while gaining personal information through a trial-and-error process [9,13,33]. However, in social species, adolescence is also a period of vulnerability, particularly when individuals face social adversity such as defeat, assault, neglect or family disruption due to divorce or widowhood, which can impair cognitive, physiological and behavioural development (i.e. rodents, primates and birds [4,9,20,22,42,43]). For instance, chronic stress during adolescence has been shown to impair spatial memory (mice, Mus musculus [44]), and pair disruption within the family structure can alter adolescent responses to novelty or danger (rhesus macaques, Macaca mulatta [45]; naked mole rats, Heterocephalus glaber [46]; zebra finches, Taeniopygia guttata [47]).

Among these social stressors, family disruptions are understudied but represent potentially challenging events during emancipation. In many species of birds and mammals, including humans, family structures are frequently altered by divorce or high adult mortality [48–53]. When family disruption occurs while offspring still need care [54,55], they may experience a wide range of outcomes from aggression, rejection and persecution, to tolerance or even adoption by the substitute caregiver [55–58]. Such instability within the family unit may impact an offspring’s learning capacity, modify its stress responses and ultimately impair its ability to cope with new challenges later in life. Although much is known about the effects of social disruption on cognitive and neural development at early stages in life [11], we still know little about how pair disruption experienced during emancipation may have long-lasting effects on the ability to solve new problems.

In group-living species, adult innovative performance can strongly depend on motivational, social and ecological conditions [59]. For instance, individuals with a stronger need for resources are more likely to engage in novel foraging tasks than less motivated individuals [60]. High perseverance increases both the opportunity to gather information about the new task and the chances of solving it through a trial-and-error process [60]. In social contexts, group members may obtain food benefits from exploratory and/or skilled individuals through non-aggressive [61,62] or aggressive scrounging [63]. Such interactions may limit the group members’ opportunity to learn new information, especially when interacting with the task is mandatory to solve the problems on their own. Alternatively, group members may learn from innovators by observing them and copying their behavioural technique, allowing them to exploit food sources on their own [64–66]. Therefore, beyond acting as a social buffer during development, group living can help mitigate the lasting effects of cognitive impairments in adulthood while also affecting problem-solving performance of skilled individuals [24,35,36,67,68].

In this study, we used a group-living, biparental avian species, the zebra finch (T. guttata), to investigate whether pair disruption experienced during emancipation alters adulthood problem-solving performance in a social context. In this species, juveniles leave the nest around day 17, and they still rely on their parents to develop social skills until they reach nutritional independence by day 40 [25,69–71]. This transitional phase, called emancipation, bridges dependency and early adolescent behaviours. Then, adolescence starts at nutritional independence and ends around day 100, when individuals reach sexual maturity and adulthood [71]. This species serves as an effective model to investigate this question as re-pairing frequently occurs during development [47] and offspring imitate their adult tutors’ behaviour to solve new problems [15]. As adults, zebra finches are gregarious and are also capable of innovation and social learning [15,72,73]. Here, we used 22 reproductive pairs and experimentally manipulated the social environment of their offspring during emancipation, either by replacing the biological father with an unfamiliar male (n = 11 experimental families) or letting the biological father raise his offspring (n = 11 control families).

In adulthood, offspring had to solve a novel problem (a lid-sliding task) with their siblings (in a social setting) and to complete the same task on their own (in an isolated setting). We assessed the problem-solving skills at both individual and brood levels and accounted for non-cognitive biases such as motivation, neophobia and perseverance. Altogether, we expected that pair disruption during emancipation would impair adulthood problem-solving performance in the social setting, and such impairment would persist in the isolated setting, even after gaining experience (through a trial-and-error process or social information use) in the social task. Specifically, we predicted that control individuals would outperform experimental individuals in the social foraging tasks. Although the performance of some birds might mitigate the effect of pair disruption on adult problem-solving performance at the brood level, we expected that experimental broods would have lower mean performance (i.e. fewer wells opened by siblings) and lower performance rate (i.e. fewer siblings opening fewer wells) compared to control broods.

2. Methods

(a). Housing and social environment conditions

We used a breeding colony of 44 zebra finches. Males and females were kept physically, visually and acoustically separated for nine weeks. Before experiments, we housed 22 breeding pairs in reproductive cages (47.5 cm × 38 cm × 51 cm) in a 14 : 10 light : dark cycle, with access to a nest box (12 cm × 13 cm × 16 cm), nest materials (coconut fibre), millet seeds mixtures, water, minerals and grit. Every morning, we marked each new egg, kept it on cotton wool at room temperature and replaced it with a false clay egg. After 2 consecutive days without egg laying, we returned the clutch (i.e. mean ± s.d.: 5 ± 1 eggs) to the nest for incubation to synchronize hatching and standardize the brood development conditions. After 10 days of incubation, we selected three viable eggs and let the parents incubate them until hatching. At hatching, because brood size can affect nestling development [74], we only kept families composed of two or three nestlings (removing two single offspring broods from the study). From days 0 to 15 post-hatching, all offspring were raised by their biological parents and from days 15 (fledging) to 39−40 post-hatching (after full nutritional independence), each brood was randomly assigned either to: (i) the experimental group (n = 11 pairs) where the biological father was replaced by an unfamiliar male; or (ii) the control group (n = 11 pairs) where the biological father was captured for 5 min and then released back into the enclosure (figure 1). At the end of the treatment (days 39−40), we removed the parents and the nest box from the cages and determined the sex of each juvenile based on plumage characteristics. The juveniles were left together in their home cage until problem-solving testing (days 200−206).

Figure 1.

Timeline of the experiment. The social disruption treatment is represented in hatched light grey and the cognitive trials in orange, with the schematic representation of the 1 h social—(a) familiarization, (b) training and (c) test trials, and the 30 min isolated—(d) familiarization, (e) training and (f) test trials. The grey bird in the experimental group represents the new unfamiliar male. The test trials are represented in bold.

(b). Problem-solving tasks

When juveniles reached day 200 post-hatching (24 experimental individuals from 10 broods and 24 control individuals from 10 broods), we carried out a social lid-sliding task [72] (one 1 h social trial per day over 3 days; figure 1a–c), in which siblings had to learn how to remove obstacles to gain access to food rewards. The three subsequent days, we carried out the same lid-sliding task but in isolation (one 30 min isolated trial per day; figure 1d–f), where each individual had to solve the same problem alone, as a follow-up test. All familiarization and experimental phases took place in an experimental aviary covered by opaque panels (100 cm × 70 cm × 120 cm) and equipped with two Logitech webcams and an LED strip light. To conduct the problem-solving tasks, we used a wooden board composed of nine wells covered by white plastic lids that could be slid open (electronic supplementary material, figure S1). Birds were food deprived for 4 h prior to the tests to ensure they were motivated to find food. During the experiments, the subjects were visually but not acoustically isolated from the other birds. The experimenter remained outside the aviary and could observe the birds via a live stream on a computer. All trials were videotaped and analysed afterwards using Boris (v. 8.20.4).

(i). Social lid-sliding task

Each brood went through a three-step lid-sliding task procedure: one familiarization trial, one training trial and one test trial (figure 1a–c), with one trial per day over 3 consecutive days. At each trial, siblings were captured from their home cage and released together for 1 h into the experimental aviary. The familiarization, training and test boards were composed of open, half-covered and covered wells, respectively, each filled with 10 millet seeds. In the familiarization and the training trials, the food was visible within pecking distance, while in the test trial, the food was hidden under the opaque white lids. Before each trial, the experimenter filled and covered the wells depending on the trial. Then, the experimenter captured and released all siblings from a given brood in the aviary. The trial started after the experimenter turned on the camera and left the room. After each trial, the broods were placed back into their cages.

We analysed the videotapes of the familiarization and training trials to check whether each brood member successfully foraged in a well. From the videotapes of the test trial, we noted the latency of each sibling to go on the board (i.e. as a proxy of neophobia), whether the individuals opened a well (i.e. individual success), the latency to first open a well, the number of wells opened by each sibling (i.e. individual performance) and the number of times they touched a well (i.e. perseverance). As defensive and sharing behaviours could have biased the problem-solving measures during the test trial, we also noted the occurrence of agonistic and sharing behaviours (individual opening then sharing a reward) between siblings to control for their effects in the subsequent analyses. Since the behaviour and the success of one individual depended on the behaviours and success of its siblings, we assessed problem-solving skills at both the individual and brood levels. At the individual level, we used the success of each sibling to open a well, the latency of resolving the task (i.e. latency to first open a well minus latency to go on the board) and the individual performance during the test trial (electronic supplementary material, table S1). Because the duration of each trial was approximately 1 h, we used the amount of perseverance, agonistic and sharing behaviours per min (i.e. perseverance, agonistic and sharing behaviours rate) as covariables. At the brood level, we used the brood performance as the average number of wells opened by siblings within the brood, and the brood performance rate as the number of wells opened divided by the number of siblings who opened at least one well in the test trial. We also used the mean and the standard deviation of agonistic and sharing behaviour rates as covariates.

(ii). Isolated lid-sliding task

After the social task, all the birds had solved the task or had seen a sibling opening a well. Thus, we performed the same lid-sliding task but in isolation to test whether such experience eliminated any differences observed in the social task. We used the same three-step procedure as in the social task: one familiarization trial, one training trial and one test trial (figure 1d–f), and conducted one trial per day. At each trial, each bird was captured and released alone for 30 min into the experimental aviary. The familiarization, training and test trials were composed of open, half-covered and covered wells, respectively, each filled with five millet seeds. Using the same video analysis procedure, we assessed the success of each individual to open a well, its latency of solving the task, its perseverance and the total number of wells it opened (individual performance) during the test trial (electronic supplementary material, table S1).

(c). Statistical analysis

All statistical analyses were conducted in R (v. 4.0.2; R CoreTeam 2020). From the package ‘lme4’, we used the ‘glmer’ function to perform generalized linear mixed models (GLMMs), ‘lmer’ function to perform linear mixed models (LMMs) and ‘lm’ function for linear models (LMs). Collinearity and validation were checked for all models performed (suggested cut-off VIF > 3; DHARMa package). We compared the models with and without interactions and selected the model with the lowest AICc (electronic supplementary material, table S2).

Familiarization, training trials and covariables

To check if the experimental and control individuals differed in their success on the familiarization and training trials (figure 1a,b), we first performed chi-squared tests. We also performed LMMs to test whether the covariables (i.e. log-transformed latency to go on the board, square-root transformed rates of agonistic and sharing behaviours and perseverance) were affected by the treatment in the test trial. We used each covariable as a response variable, the treatment, the sex, their interaction and the brood size as fixed effects and the brood identity as a random effect.

Individual success, latency to solve and performance in the test trial

We tested whether experimental and control individuals performed similarly in the social test trial. To do so, we performed a GLMM with a binomial distribution including the individual success as a response variable, a LMM including the latency to resolve the task as a response variable and a LMM including the performance of individuals as a response variable. For all three models, we used the treatment, the sex of individuals and their interaction as fixed effects. We added the brood size (i.e. two or three nestlings) as a categorical variable, the latency to go on the board and the rate of agonistic behaviours as covariables and included the brood identity as a random effect. Because more perseverant individuals increase their chances to resolve the task by simply interacting with it, we also included the perseverance rate of individuals as a fixed effect.

Mean brood performance and brood performance rate

We tested whether experimental and control broods differed in their mean brood performance and brood performance rate. First, we used a LM including the mean brood performance as a response variable, and as fixed effects, the brood size and interactions between the treatment and the mean sharing behaviours, and between the treatment and the mean agonistic behaviours. We included these interactions to understand whether performing in the social test trial increased siblings’ propensity to share their discovery or to be aggressive toward each other. Second, we used a LM, including the brood performance rate as a response variable, and as fixed effects, the brood size and interactions between the treatment and the standard deviation of sharing, and between the treatment and the standard deviation of agonistic behaviour rate. We included these interactions to investigate whether the performance heterogeneity at the brood level was associated with the variance of sharing and agonistic behaviours rate within broods. For both LMs, we reported the results from best-fitted models after a model selection procedure (lowest AICc).

Familiarization, training trials and covariables

We performed chi-squared tests to check whether the experimental and control individuals differed in their success during the familiarization and training trials (figure 1d,e). We then performed LMMs to test whether the covariables (i.e. log-transformed latency to go on the board and square-root transformed perseverance) in the test trial were affected by the treatment. We used each covariable as a response variable, the treatment, the sex, their interaction and the brood size as fixed effects and the brood identity as a random effect.

Individual success, latency to solve and performance in the test trial

We performed three models to investigate whether any differences observed in the social task disappeared in the isolated task. First, we used a GLMM with a binomial distribution, including the success of individuals in the isolated test trial as a response variable. Second, we used a LMM including the latency to solve the task as a response variable. Third, we used a LMM including the performance of individuals as a response variable. For all three models, we used the treatment, the sex of individuals, their interaction, the perseverance and the latency to go on the board in the isolated test trial and the brood size (i.e. two or three nestlings) as fixed effects. Because individuals within and between broods could have acquired different experiences in resolving the task during the social phase, we also included the individual performance in the social test trial as a confounding variable. Brood identity was included as a random effect.

3. Results

(a). Social lid-sliding task

(i). Familiarization, training trials and covariables

In the familiarization and the training trials, 91.7% and 96.2% of the birds succeeded in foraging in wells, respectively. The experimental treatment did not alter individual success during the familiarization (χ² = 0.91, p = 0.763) and the training trial (χ² = 0.220, p = 0.639). We did not find any effect of the treatment on the covariables: control and experimental groups did not differ in their latency to go on the board (LMM: χ² = 0.728, p = 0.394), their aggressiveness (LMM: χ² = 0.458, p = 0.499), their sharing behaviour rates (LMM: χ² = 0.292, p = 0.589) and their level of perseverance (LMM: χ² = 0.181, p = 0.671).

(ii). Individual success, latency to solve and performance in the test trial

The experimental treatment had a detrimental effect on the individual success and the latency to solve the test trial, but it did not affect the individual performance (table 1). First, individuals from the control group were more likely to succeed than individuals from the experimental group (table 1A; figure 2A), and more perseverant individuals had higher chances to be successful than less perseverant ones (table 1A). Second, the experimental treatment had a sex-dependent effect on the latency to succeed (table 1B; figure 2C). Experimental males were slower to succeed than control males (Tukey’s test: estimates = −18.79, s.e. = 8.94, p = 0.045) and experimental females (test: estimates = −18.19, s.e. = 8.65, p = 0.043), while females from control and experimental groups did not differ in their latency to succeed (Tukey’s test: estimates = 4.23, s.e. = 7.38, p = 0.572; figure 2C). In addition, more perseverant individuals were faster to succeed (table 1B). Third, the experimental treatment did not impair individual performance (table 1C; figure 2B) while more perseverant individuals performed better than less perseverant ones (table 1C). We also found that the latency to go on the board, the aggressiveness of individuals and the brood size did not have any significant effect on the three problem-solving metrics (table 1).

Table 1.

Results for social disruption treatment as predictors of variation in the individual (A) success (i.e. whether the individuals opened a well; GLMM), (B) latency to solve (LMM) and (C) performance (i.e. the number of wells each sibling opened; LMM) in the social test trial. Models with and without the interaction between the treatment and the sex of individuals were compared, and the model with the lowest AICc was selected (electronic supplementary material, table S2).

problem-solving measures at the individual level	fixed effects	χ2	d.f.	p
success in the social test trial (n = 48)	treatment	4.752	1	0.029
	sex	3.608	1	0.058
	brood size	1.742	1	0.187
	latency on the board	2.333	1	0.127
	agonistic rate	0.312	1	0.576
	perseverance rate	4.162	1	0.041
latency to resolve the social test trial (n = 48)	treatment	0.885	1	0.347
	sex	1.124	1	0.289
	brood size	2.067	1	0.151
	latency on the board	3.229	1	0.072
	perseverance rate	5.340	1	0.021
	agonistic rate	0.263	1	0.608
	treatment × sex	4.141	1	0.042
performance in the social test trial (n = 48)	treatment	0.098	1	0.754
	sex	1.147	1	0.284
	brood size	0.290	1	0.590
	latency on the board	0.133	1	0.715
	agonistic rate	0.026	1	0.871
	perseverance rate	14.222	1	<0.001

Significant p-values are indicated in bold. The brood identity is a random effect.

Figure 2.

Effects of social disruption treatment on the individual (A) success, (B) performance and (C) latency to solve the social test trial according to the sex of individuals. The barplots represent the mean values and the error bars represent the standard errors of the mean. Control and experimental individuals are represented in dark and light grey, respectively. Significant differences between groups are indicated with different letters.

(iii). Mean brood performance and brood performance rate

The experimental treatment did not affect the mean brood performance as the treatment was not retained in the model after model selection (table 2A; figure 3A; electronic supplementary material, table S3A). The mean brood performance was not associated with the mean agonistic behaviours, but broods that performed better in the test trial shared their discoveries more compared to less well performing broods (table 2A). The broods composed of three siblings also performed better on average than the broods composed of two siblings (table 2A). Regarding the brood performance rate (i.e. the number of wells opened by the siblings divided by the number of siblings that opened wells in the social test trial), the performance of individuals within experimental broods was more heterogeneous than within control broods (figure 3B). Specifically, we found a significant interaction between the treatment and the variance in sharing behaviours within broods (LM: sum sq = 9.830, p = 0.001; table 2B; figure 3C; electronic supplementary material, table S3B). This result indicates that the positive relationship between the performance heterogeneity and the variance in sharing behaviours within broods was stronger for experimental than control broods (i.e. treatment × s.d. in sharing behaviours; table 2B; figure 3C). No such result was found for the variance in agonistic behaviours (electronic supplementary material, table S3B).

Table 2.

Linear model’s (LM) results for social disruption treatment as predictors of variation in the (A) mean brood performance (the averaged number of wells opened by siblings within the brood) and (B) brood performance rate (the number of wells opened by the siblings divided by the number of siblings that opened wells) in the social test trial. From the full models (electronic supplementary material, table S3), we performed model selections and kept the models with the lowest AICc.

problem-solving measures at the brood level	fixed effects	sum sq	d.f.	p
mean brood performance in the social test trial (n = 20)	brood size	0.006	1	0.002
	mean sharing behaviour	0.011	1	<0.001
	mean agonistic behaviour	0.002	1	0.070
brood performance rate in the social test trial (n = 20)	treatment	0.437	1	0.412
	s.d. sharing behaviour	45.584	1	<0.001
	treatment × s.d. sharing behaviour	9.830	1	0.001

Significant p-values are indicated in bold.

Figure 3.

Effects of social disruption treatment on the (A) mean brood performance and (B) brood performance rate in the social test trial. (C) The relationship between the brood performance rate and the standard error of the mean sharing behaviours in the social test trial. The barplots represent the mean values and the error bars the standard errors around the mean. Control and experimental broods are represented in dark and light grey, respectively. Significant differences between groups are indicated with different letters.

(b). Isolated lid-sliding task

(i). Familiarization, training trials and covariables

When individuals were isolated in the familiarization and the training trials, 83.67% and 89.80% of individuals succeeded in foraging in wells, respectively. The experimental treatment did not affect individual success in the familiarization (X² = 0.1, p = 0.752) and the training trials (X² = 0.209, p = 0.647). Control and experimental groups did not differ in their latency to go on the board (LMM: χ² = 0.017, p = 0.895), and their level of perseverance (LMM: χ² = 0.044, p = 0.834) in the isolated test trial.

(ii). Individual success, latency to solve and performance in the test trial

We found no effect of the experimental treatment on individual success, latency to go on the board or performance (table 3; electronic supplementary material, figure S2). The birds from smaller broods and the less perseverant individuals went onto the board sooner compared to the birds from larger broods and more perseverant individuals (table 3B). In addition, the birds performing well in the social test trial performed better in the isolated test trial than the others (table 3C). We also found that more perseverant individuals performed better in the isolated test trial (table 3C). No effect was found for the other covariables included in the models (p > 0.081; table 3A–C).

Table 3.

Results for social disruption treatment as predictors of variation in the individual (A) success (GLMM), (B) the log-transformed latency to solve and (C) performance (LMM) in the isolated test trial. Models with and without the interaction between the treatment and the sex of individuals were compared, and the models without the interaction had the lowest AICc and were selected (electronic supplementary material, table S2).

problem-solving measures at the individual level	fixed effects	χ2	d.f.	p
success in the isolated test trial (n = 48)	treatment	0.660	1	0.417
	sex	1.613	1	0.204
	brood size	3.051	1	0.081
	performance in the social test trial	0.828	1	0.363
	latency on the board	0.473	1	0.492
	perseverance rate	2.896	1	0.089
latency in the isolated test trial (n = 48)	treatment	0.135	1	0.713
	sex	0.958	1	0.328
	brood size	4.077	1	0.043
	performance in the social test trial	0.552	1	0.457
	latency on the board	0.760	1	0.383
	perseverance rate	14.682	1	<0.001
	treatment × sex	2.575		0.109
performance in the isolated test trial (n = 48)	treatment	1.929	1	0.165
	sex	0.117	1	0.732
	brood size	1.919	1	0.166
	performance in the social test trial	7.675	1	0.006
	latency on the board	2.796		0.095
	perseverance rate	55.841		<0.001

Significant p-values are indicated in bold. Brood identity was included as a random effect.

4. Discussion

Since adolescence is a critical period for both social and cognitive development [25,47,75], we hypothesized that pair disruption during emancipation would impair adult zebra finches’ problem-solving performance in social contexts. Our results partially support this hypothesis. In the social setting, experimental individuals were less successful than controls. Experimental males were slower to open wells compared to control males, but control and experimental birds did not differ in the number of wells they opened individually. Although both experimental and control broods had similar average brood performance, we found greater variability in performance within experimental broods compared to control ones. These differences suggest that pair disruption had variable effects on individuals within experimental broods. Finally, the reported differences between experimental and control individuals in the social setting disappeared in the isolated setting. Altogether, these findings suggest that pair disruption during emancipation has a long-lasting effect on individuals’ ability to solve new problems in social settings but does not appear to affect their ability to replicate the solution once discovered. Unexpectedly, the detrimental effects on social problem solving were not uniform within experimental broods, suggesting that the success of a few experimental siblings might compensate for the lack of success of others, thereby maintaining overall brood performance. The presence of some siblings who successfully solved the new problem and shared their discoveries may have provided both food rewards and social information about task solutions during the social trial, which was valuable during the subsequent isolated test trial.

(a). Cognitive impairment on the social problem-solving task

Extensive research has shown that social disruption can impact cognitive and neural development, though much of this work focuses on the prenatal, early postnatal or adult phases in mammals [4,11]. Our results further suggest that pair disruption during emancipation also impairs cognitive development in a passerine bird species, and more particularly, problem-solving skills later in life [6]. Specifically, experimental individuals were less successful than controls in solving the novel social foraging problem, despite similar success in familiarization and training trials. Potential confounding variables, such as motivation (i.e. the latency to go on the board [60]), aggressiveness, sharing behaviours and perseverance—traits that could bias problem-solving performance in social tasks [76]—were not affected by the treatment in our study. Although we cannot exclude the possibility that problem-solving measures could have been biased by other non-cognitive factors sensitive to pair disruption, our results suggest that experimental individuals were neither less motivated nor less perseverant than controls, but rather less capable of solving the new social foraging problem. While the treatment altered individuals’ success in the social trial, it did not affect individuals' performance (i.e. the number of wells individuals opened) during the social trial. This suggests that pair disruption may impair the ability to solve a new problem but not the capacity to replicate the technique once discovered. Additionally, experimental males were slower to solve the social task than control males or experimental females, whereas no sex differences were found in any covariables. In many species, males and females differ considerably in their ecology and can face different challenges. In birds, however, evidence for sex differences in problem-solving performances remains scarce and inconsistent. For instance, in string-pulling and box-opening tasks, females skuas (Stercorarius skua) outperformed males [77], whereas male and female great tits (Parus major) performed similarly [78]. While sex-dependent effects on problem solving seem to vary across species and types of tasks, further investigation is needed to determine whether the observed differences stem from intrinsic sex differences in problem-solving skills [77,79,80] or from a sex-specific effect of pair disruption during development.

(b). Higher performance heterogeneity within experimental broods

Although individual success during the social trial was higher for control than experimental individuals, the mean brood performance was similar across treatment groups. This result suggests that a few successful individuals within experimental broods compensated for the lack of success of their siblings, resulting therefore in greater performance heterogeneity within experimental broods compared to control ones (i.e. brood performance rate). Three non-mutually exclusive explanations may account for this performance heterogeneity.

Firstly, social disruption may have impacted siblings differently within experimental broods, with some siblings receiving more aggressions or socially buffering than others during the treatment [81]. These social microenvironmental differences could have led to divergent cognitive development trajectories within the same brood, leading to different problem-solving abilities in adulthood [82,83] (but see [84]). Unfortunately, we lacked fine-scale social interaction data to investigate this hypothesis in the current study [85].

Secondly, agonistic interactions during the social test could have contributed to such performance heterogeneity. For instance, non-innovative siblings in experimental broods may have benefitted from the success of innovative siblings through aggressive scrounging behaviours [63,85], obtaining food rewards without solving the problem themselves. Alternatively, dominant innovators may have guarded their discoveries, limiting learning opportunities for other siblings [6,86]. However, these explanations seem less likely, as we found no relationship between problem-solving measures and agonistic interactions during the social trial, both at the individual and the brood levels.

Finally, pair disruption may have increased affiliative behaviours among experimental siblings during adolescence, leading them to rely more on each other when facing new challenges [87,88]. In our experiment, non-aggressive scrounging may have enabled most experimental individuals to obtain food benefits through the success of a few siblings [62,89]. In contrast, most control siblings may have succeeded on their own, without relying on scrounging behaviour. Supporting this, we found a stronger positive relationship between brood performance heterogeneity and the variance in sharing behaviours within experimental broods. This result suggests that a few successful individuals in these broods may have systematically shared rewards, while control siblings mostly gained rewards through their own discoveries. This idea is also consistent with the fact that larger broods (i.e. three siblings) performed better than smaller broods (i.e. two siblings), probably because they had higher chance to have a sibling able to resolve the task and demonstrating the technique.

(c). Problem-solving differences in the social task disappeared in the isolated task

Despite lower success of experimental individuals compared to control individual in the social trial, these problem-solving differences disappeared when individuals were tested alone. Similarly, differences observed between experimental and control males in the latency to solve the task disappeared in the isolated test trial. This may indicate that experimental individuals relied on social information gathered during the social trial to solve the task when tested alone [90–92]. In animals, social learning and imitation are well-known processes allowing individuals to learn new techniques from the experience and mistakes of their conspecifics, while reducing the time and effort invested in the task [93–97]. Our findings suggest that pair disruption may have impaired individual ability to solve problems in the social setting. It also suggests that such impairment may have been mitigated through social learning in the isolated setting. Although we could not directly test this hypothesis in the current study, future experiments with reversed trial orders (isolated first, then social trials) could help better understand the reversibility of pair disruption on cognitive development when individual face social and isolated cognitive challenges in adulthood. Using a balanced experimental design—a group performing the social than the isolated test and another group performing the reverse—should be the next step to disentangle the effect of pair disruption on social and individual problem-solving skills.

5. Conclusion

In support to studies showing detrimental and long-lasting effects of social disruption during early phases of cognitive development [3,9–11,98], our results demonstrated that the social environment during emancipation can have detrimental effects on adult problem-solving abilities in a social setting. However, such cognitive impairment might be compensated in group-living species by the success of a few individuals, who may transfer their discoveries or learning techniques to less successful individuals through social information. This phenomenon may be particularly relevant in bird species exhibiting social transmission of novel foraging techniques or the use of new food resources in the wild [99]. These findings offer new insights into how social stability during early adolescence may foster cognitive traits essential for both individual survival and social dynamics in group-living species. We also highlight the critical, yet often overlooked role of social compensation in adult birds [24,35,36,67,68], which can mitigate the detrimental effects of early social environments on problem-solving skills. Our study emphasizes the importance of studying animal cognition in social settings to better understand the lasting effects of social disruption on cognitive development and the potential for cognitive compensation through social learning in adulthood.

Ethics

The experiments were conducted at the Centre d’Etudes Biologiques de Chizé (CNRS) and all procedures were approved by the regional committee of ethics on animal use (COMETHEA, permit no. 34178-20211201041162). In line with ethical protocols, hatchlings eggs removed during the experiments were immediately frozen at −20°C. To limit the stress of individuals during the cognitive tests, we observed the birds from a computer through live streaming, remaining unseen by the focal bird (outside the aviary behind a curtain). We also let each individual recover from the capture events before the behavioural tests. To minimize the restraint duration and limit the risk of injury, all experiments were performed by an experienced researcher.

Data accessibility

The data, code and meta-data are available in the electronic supplementary materials [100].

Declaration of AI use

We have not used AI-assisted technologies in creating this article.

Authors’ contributions

M.B.-D.: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, visualization, writing—original draft; C.V.: data curation, formal analysis, methodology; F.A.: conceptualization, funding acquisition, investigation, methodology, resources, supervision, validation, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

This work, as well as C.V. and M.B.-D., was financially supported by the Fyssen Fondation awarded to M.B.-D. and by the ANR (ANR-20-CE34-008) awarded to F.A.