Introduction

In many animal species, the juvenile period is critical for acquiring skills necessary for survival, including knowledge about diet [1,2], foraging techniques [3], predator avoidance [4,5], migration routes [6,7], and how to socially interact with conspecifics [8]. Social learning, i.e., learning that is influenced by observation of or interaction with conspecifics [9], can greatly increase the efficiency with which individuals acquire novel information or behavior. For efficient acquisition of behavior, reliance on social learning is therefore often particularly pronounced during the juvenile period [10]. The adaptive benefit of social learning during the juvenile period has been argued to lead to selection on family living and parental care, with the cultural inheritance of knowledge from parent to offspring in turn leading to selection for extended juvenile periods and cognitive evolution [11,12].

Thus, many species with prolonged periods of parental care rely extensively on so-called vertical learning between parent and offspring, e.g., [13], leading to patterns of behavioral similarities within family units (e.g., Bornean orang-utans [1], vervet monkeys [14], and dolphins [3,15]). This then in turn leads to the intergenerational cultural transmission of knowledge, promoting stable multigenerational cultures with adaptive significance and ecological implications [16]. Yet, social learning is not restricted to species with prolonged dependence periods. This is particularly true in many bird species who are typically characterized by short durations of parental care [17]. The learning strategies of developing individuals in species with limited periods of dependency are less well understood, and information on how such strategies might change during transition to independence is lacking (but see [18,19]).

The vertical transmission of behavior from parent to offspring may not always be adaptive, especially in rapidly changing environments or in environments that differ over fine spatial scales, where dispersing juveniles may experience different conditions to their parents. Here, young individuals may need to be strategic in when to rely on social learning. The value of parental social information is expected to rapidly decrease as young individuals gain independence and experience new environments, in which they may benefit from other role models who may be more familiar with the local environment [20,21]. Furthermore, individuals also need to be strategic in who to choose as role models, as conspecifics may differ in the value of information they hold [22–24], and parents may already possess outdated knowledge. Finally, as transmission of information or behavior can only occur among individuals who are in close enough proximity for learning to occur, a species’ social system and social network structure will strongly influence the social learning strategies that developing individuals employ, with limited parental care potentially favoring learning from siblings or other peers. However, the effect of such social learning strategies on the potential for multigenerational cultures is not well understood.

Here, we conduct a comprehensive investigation into the ontogeny of social learning in a species with a comparatively short period of dependence, the European great tit (Parus major). Great tits have become a model species for social learning and animal culture, with extensive evidence that both foraging behaviors and song can be socially learned [25,26], with new foraging behaviors spreading rapidly through social networks to establish as multigeneration cultural traditions [27]. This body of evidence suggests that individuals can learn foraging behaviors throughout life via oblique or horizontal (peer to peer) learning from flock-mates and exhibit adaptive social learning strategies when doing so. For example, they may exhibit a conformist learning bias when learning new behaviors [27], but shift to a pay-off learning bias when moving into new environments with different foraging behaviors [28]. However, almost all studies on foraging behavior have been conducted in the winter flocking period when first-year juveniles are already fully independent. We currently lack information on how social learning is expressed during ontogeny, a time marked by dynamic changes in association patterns with parents, siblings, peers and other adults [29], and with it, shifting availability of potential role models for social learning.

Great tits are seasonal cavity breeders, with parents raising their offspring in the nest for approximately 22 days [30]. After fledging, offspring spend between 10 and 32 days in their family group, during which they are fed by parents [31]. With the onset of independent feeding, they reach full independence from parents and integrate into local flocks, but may continue to prefer to associate with siblings and peers over adults [29]. Previous cross-fostering experiments, where eggs were swapped between great tits and blue tits (Cyanistes caeruleus), found that for behaviors related to mate choice, foraging behavior and song, birds showed more similarity to their foster parent species than their biological parent species, indicating the these behaviors are learned rather than genetically determined [32–35]. However, the mechanisms of transmission remain unresolved. On the one hand, it is plausible that offspring could have learned these behaviors directly from their foster parents through vertical learning. Alternatively, however, developing individuals could have imprinted on the “wrong” species during a sensitive period and subsequently copied behavior from other individuals in their foster species other than their foster parents after reaching independence.

To examine the ontogeny of social learning in wild great tits, we conducted a social learning experiment using two-option foraging puzzles with a sliding door that have been previously used in social learning experiments in this species, leading to multigenerational cultures [27,36]. To obtain a mealworm reward, birds could slide a door either to the left or the right [37]. We introduced and diffused knowledge of these puzzles in two populations of breeding adults and then used a selective access mechanism [37] to constrain individuals in their side options. This resulted in four experimental conditions for breeding pairs: (1) parents with the same knowledge state (both left or both right), (2) parents with conflicting knowledge states (left and right), (3) parents with only one knowledgeable adult (left or right), and (4) parents without any knowledge. We then examined the acquisition of knowledge by their offspring in the immediate period after they fledged from the nest. First, to explore to what extent offspring behavior is predicted by vertical transmission from the parents, we asked whether parental solving frequency during dependence—irrespective of side choices—increased the likelihood of offspring learning to solve the puzzle. Second, we investigated the detailed pathways through which juveniles learn to solve in a “network-based diffusion analysis” (NBDA) [38–40], to establish the relative importance of vertical, oblique, and horizontal learning. Finally, we investigated the decision-making strategies juveniles employed when choosing between the two options (solving right or left), the influence of parental solving behavior on these strategies, and how these strategies change with increasing personal experience.

Results

Both adult great tits and nestlings in two breeding populations (MI and GU) were equipped with unique passive integrated transponder (PIT) tags for remote identification using RFID. Across both populations, we monitored 306 nest boxes, from which 51 breeding pairs successfully fledged 229 chicks.

For the social learning experiment, we deployed seven automated puzzle boxes with a sliding door and diffused solutions through population for two weeks (Fig 1A). Solvers were identified via RFID antennae built into the landing perches. Of the 51 breeding pairs, 26 were both knowledgeable in solving: 13 were restricted to solve on the same side (eight right, seven left) and 13 to opposite sides. Two pairs had one knowledgeable partner each (both right), and 23 pairs had no knowledge. The numbers of parents with different knowledge states reflect both our experimental design, which prioritized exposing juveniles to solving parents, and natural variation in survival and site use, which limited which breeding pairs were ultimately included in the analyses.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Social learning pathways of learning to solve a puzzle box.

A) Fully automated puzzle box. Birds can access a mealworm tray by either sliding the door “left” or “right”. B) Illustration of social network at MI depicting different potential pathways of learning. Adults are shown as squares, juveniles as circles with knowledgeable individuals in red and naïve individuals in blue. The data underlying Fig 1B can be found at https://doi.org/10.5281/zenodo.16930533.

https://doi.org/10.1371/journal.pbio.3003401.g001

After fledglings emerged from their nest boxes, puzzles were deployed for an additional eight weeks on four days per week, recording identity and side choices of both solving adults and juveniles. Juveniles were considered knowledgeable if they produced at least 10 solves.

On two days per week, we additionally deployed seven RFID-equipped feeder columns (“network feeders”) to track associations based on cofeeding events. In total, 231 tagged great tits—of which 103 were juveniles—were recorded (MI: 170 [Fig 1B]; GU: 61). Of the juveniles, 53 were knowledgeable of the puzzle box, producing 10,668 solves across the experiment.

Establishing social learning pathways of solving

To estimate the strength of social transmission and investigate pathways of the diffusion of solving behavior, we ran a multinetwork NBDA (S2 Table) [39] with four different social networks reflecting different pathways (Fig 1B): (1) a parent-offspring network to allow for vertical learning; (2) an oblique network allowing for social transmission from nonparental adults to offspring; (3) a horizontal network allowing transmission among siblings; and (4) a horizontal network allowing transmission among nonsibling peers. All networks were weighed by the dyadic association strengths based on cofeeding events at network feeders (see methods). We then modeled how closely the diffusion data—here the time at which juveniles learned to solve the box—followed the different networks, or combinations thereof. We included two individual-level variables (ILVs) with potential effect on the juveniles’ learning rate [39], namely the number of parental solves produced during dependence, as well as the number of scrounges the juvenile produced up to the time of acquisition, as scrounging might facilitate future learning [42].

We found most support for social models with vertical transmission from parents, oblique transmission from adults, and horizontal transmission among siblings (summed Akaike weights ; S1 Fig). In the best performing model (), the social transmission rate was highest among siblings with 42.1% [95% CI: 41.5%–43.9%] of acquisition events, followed by oblique learning with 37.8% [95% CI: 32.2%–38.3%], and vertical learning with 19.8% [lower 95% CI: 14.0%] occurring as a result of the connections within the respective network (Fig 2A). Note that upper confidence intervals for the vertical social transmission parameter could not be obtained due to convergence issues in the best-performing model (likely due to a small number of acquisition events at some sites). However, the lower bound of the confidence intervals provides a reasonable lower limit for the social transmission effects in this case. The asocial learning rate in the best model was estimated at 0.3%, suggesting that no juveniles learned to solve the puzzle box through individual exploration.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Probability of learning through different pathways.

A) Estimated percentage of acquisition events occurring as a result of vertical, oblique, horizontal, and individual social learning. Social transmission was estimated to be strongest among siblings, followed by social transmission from nonparental adults (oblique) and parents (vertical), with no evidence for transmission among peers or asocial learning. B) Estimated probability of learning through different pathways for first and subsequent learners within a sibling cohort. First siblings learned primarily from parents and nonparent adults, while for subsequent siblings within a cohort, social learning was most likely to occur from the already knowledgeable first sibling. The data underlying this figure can be found at https://doi.org/10.5281/zenodo.16930533.

https://doi.org/10.1371/journal.pbio.3003401.g002

While there was variation in who juveniles learned the solving behavior from, there was a distinct pattern in role models when comparing the first learners of a sibling cohort versus subsequent sibling learners. Of the first learners of each sibling cohort, 74.6% [median] were estimated to have learned through oblique social learning from nonparent adults and 23.6% vertically from parents. Of the subsequent learners in each sibling cohort, 93.9% were estimated to have learned from their siblings, 3.6% from nonparent adults and 1.6% from parents (Fig 2B). Neither of the ILVs tested (number of parental solves, number of scrounges produced by juveniles) had any effect on the social or asocial learning rate (all ; S4 Table).

Side choices

In a final set of models, we investigated the decision-making strategies that influenced juveniles’ side choices (left or right), and how these strategies changed with increasing experience. As a proxy for social information available to juveniles, for each solve, we extracted the proportion of solves occurring on the right side in the five minutes immediately prior to solving. While there was evidence for local biases of solves occurring either on the left or right, juveniles were exposed to a wide range of social information on side choices (S2 Fig). We first tested the predictors of side choices on the first day of solving, when juveniles produced an average of 4 solves (range 1–51), totaling 211 solves. Our results showed that parental side choice during dependence did not predict juveniles’ side choices on their first solving day (S5 Table; Fig 3A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Predictors of solving behavior of juveniles on their first solving day (A, B) and across all solving days (C, D), including 95% credible intervals.

Note that A,B show raw data (side choices on first day) as points. C,D are restricted to solves where juveniles have both social and personal data available (considering solves across all days). A) Influence of parental solving behavior during dependence on first-day side choices: The proportion of right solves produced by juveniles on their first solving day was not predicted by the type of information they received from parents during dependence, regardless of whether it was conflicting or nonconflicting. B) Influence of social environment on first-day side choices: Side choices by juveniles were predicted by side choices observed within 5 min prior to each solve. Juveniles were more likely to pick the side that was preferred by the majority of prior solving birds. C) Influence of personal experience on decision-making across all solves: Juveniles increased their reliance on personal information with increasing experience (measured in cumulative number of solves produced). D) Social influence on decision-making across all solves: While there was moderate evidence for an overall reliance on social information for side choices, juveniles reduced their reliance on social information with increasing experience. The data underlying this figure can be found at https://doi.org/10.5281/zenodo.16930533.

https://doi.org/10.1371/journal.pbio.3003401.g003

Instead, on the first solving day, the side choices of juveniles were strongly predicted by the side choices expressed by other birds immediately prior to solving, with the likelihood of producing a right solve increasing by a factor of 19.50 [95% CI: 5.84–68.62] when all solves by other birds occurred on the right compared to all occurring on the left. This was independent of the number of solves juveniles observed in the five minutes prior to solving (OR: 2.39 [95%: 0.72–8.62] per solve; S6 Table; Fig 3B).

Finally, to investigate how reliance on social and personal information might change with increasing experience, we considered 8,041 solves by 53 juveniles who had both social and personal information available at the time of solving. Overall, side choice was strongly predicted by personal experience (OR: 7.30 [95% CI: 5.07–10.49]), and reliance on personal information increased with increasing experience (OR: 1.22 [95% CI: 1.04–1.44] per 100 solves increase in experience; S7 Table; Fig 3C). We found a sizeable effect of social information influencing overall side choice (OR: 4.15 [95% CI: 2.89–5.98]), but juveniles decreased their reliance on social information with increasing experience (OR: 0.76 [95% CI: 0.67–0.86] per 100 solves increase in experience; S7 Table; Fig 3D).

Discussion

Using fully automated foraging puzzles that allowed us to manipulate the knowledge state of breeding adults, we conducted a comprehensive investigation into the ontogeny of social learning in juvenile great tits, disentangling emergent patterns from detailed mechanisms and learning pathways. We found that the knowledge state of juveniles was strongly predicted by that of their parents, with offspring of parents that produced more solves during dependence being significantly more likely to learn to solve. Meanwhile, parental scrounging behavior during dependence did not predict the offspring probability of learning to solve, indicating that local enhancement to the puzzle site alone is not sufficient to acquire knowledge but instead suggestive of an effect of observational learning.

Yet, our detailed network-based diffusion analyses revealed that the acquisition of knowledge did not primarily occur through vertical transmission from parents, but rather from a mix of role models, including parents, nonparent adults, and siblings. Strikingly, there was a difference in role models when comparing first and subsequent learners of each sibling cohort. While the first learners in each sibling cohort learned primarily from nonparental adults, subsequent learners within the sibling cohort almost exclusively learned from their already knowledgeable siblings. Meanwhile, only few seemed to have learned directly from parents. Our results provide a dramatic contrast from most previous studied systems exhibiting multigenerational cultures, where cultural transmission has been shown or assumed to overwhelmingly occur from parent to offspring, and give insights into the more variable transmission routes might occur across social learning species.

First, our results provide insights into potential alternative underlying mechanisms to vertical transmission that can lead to diet similarities observed between parents and offspring in passerines [34,35]. Our study suggests that early exposure to the foraging behavior of parents during dependence may facilitate learning, likely through observational learning of general motor patterns. Yet, full acquisition of foraging techniques may largely occur after juveniles have reached independence from their parents but are still associating with siblings [29]. This extends findings on diet similarities between cross-fostered passerine chicks and their foster parents [34,35], suggesting that these similarities are likely only in part a result of direct copying of parents, and social transmission among siblings is perhaps a more important driver of behavioral similarities within family units. Limited cultural inheritance of behavior from parents is further supported by the fact that for solving, the parental side choices during dependence did not influence juveniles’ learning of side choices, regardless of parental experimental condition. In addition, the number of parental solves produced during dependence did not increase the speed with which juveniles learned to solve. The latter finding is consistent with a similar study on juvenile hihi, where juveniles exposed to experimental feeders during dependence by parents were not quicker to start to use them [18].

Second, our results highlight the importance of siblings as potential sources for social information in species with multiple offspring. This has been repeatedly demonstrated in humans, e.g., [43,44]. Among nonhuman animals, however, only few studies have investigated the role of siblings as sources for social information during ontogeny [45–47]. In line with our results, a study on captive juvenile ravens found that during experimental demonstrator–observer trials, siblings spent significantly more time in close proximity compared to nonsiblings and were much more likely to copy siblings’ behavior [47]. Yet, to our knowledge, no studies have explicitly demonstrated the importance of siblings as role models in the wild in their natural social environment. Previous work at our site has found that great tit juveniles tend to stay in close association with their siblings for some time even after reaching independence from parents [29]. Our current study extends these findings, suggesting that the close relationships among siblings early in life are crucial for obtaining social information. Interestingly, in contrast with these results, there was little evidence for social learning of behaviors from nonsibling peers. The reason for this needs more investigation, but may be related to a preference or variation in tolerance between siblings and nonsiblings, or to preferences towards copying individuals most similar in developmental stage.

Finally, our findings provide insights into the dynamics of decision-making strategies across development. On their first solving day, side choices of juveniles were strongly predicted by the side choices of other birds immediately prior to solving themselves, disproportionally copying the majority [22]. Such conformist tendencies have previously been reported in great tits [27,48], as well as other passerines [18]. Yet this initial reliance on social information decreased rapidly with increasing experience, with juveniles showing a strong overall preference for personal over social information. These findings align with results from a study on captive great tits that found that after initial conformist social learning, individuals preferred personal over social information until experiencing changes in their social or physical environment [28,36]. This observed pattern is consistent with a strategy to learn socially when uncertain [22], and is a commonly found strategy in nonhuman animals (e.g., Norway rats [49], chimpanzees [50] or bumblebees [51]). Our findings on a strong reliance on personal information in juvenile great tits stand in contrast to a study on juvenile hihis, where initial side choices of entering a feeder were random, but subsequent visits appeared to be socially mediated with juveniles updating their preferences to match their peers’ side choices through conformity [18].

Overall, our study provides experimental insights into the ontogeny of social learning and cultural inheritance in a species with limited parental care. We found that parental knowledge state strongly predicted knowledge acquisition of offspring, but that exact foraging techniques were shaped by social information available at the point of learning and then crystallized over time with personal experience. More specifically, we demonstrate that offspring rely on a mix of role models, with our models highlighting the importance of siblings as important sources for social information. This suggests that in species that exhibit a short dependency period, cultural inheritance of information is first shaped by parents and then refined by later social interactions. This has implications for the persistence and patterning of cultures in such species, because if cultural knowledge is spread via multiple pathways, including oblique and horizontal learning, then cultural behaviors may be less vulnerable to extinction [52] compared to species with extended and exclusive parental care, where cultures might be restricted to specific maternal lines [53,54].

Future studies on learning during ontogeny should continue to investigate the various extrinsic and intrinsic factors that have been found to affect social learning and decision-making strategies in other species, including environmental factors (e.g., blue jays [55]), species-specific dispersal patterns (e.g., vervet monkeys [56]), personality traits (e.g., guppies: [57], barnacle geese: [58]), or physiological stress experienced during dependence (e.g., zebra finches: [59]).

Materials and Methods

Supporting information