Introduction 

Nest construction is a widespread behavior among many bird species (Hansell 2000). Nests serve to create a suitable microclimate during incubation of eggs and for developing offspring and provide shelter to minimize predation risk (Mainwaring et al. 2014). As “vessels of reproduction” and determinants of reproductive success (Järvinen and Brommer 2020), nest-building behavior, including nest site selection and construction, is under strong selection (Mainwaring et al. 2014). Yet, contrary to historical beliefs that nest construction was largely based on a fixed genetic template (e.g., Nickell 1958), it is now well understood that nest building requires behavioral plasticity to optimize conditions in which offspring develop (Britt and Deeming 2011; Deeming et al. 2012; Mainwaring et al. 2012; Healy et al. 2015; O’Neill et al. 2018).

Learning is a vital form of plasticity, allowing individuals to adapt to their local environment (Snell-Rood 2013). Research from the past two decades has shown that learning plays a significant role in a variety of behavior related to nest building (reviewed in Breen et al. 2016). Thereby, individuals may refine their nests based on personal breeding experiences. For example, birds of several species were found to re-use or avoid breeding sites based on previous breeding success (e.g., Suryan and Irons 2001; Fisher and Wiebe 2006). Furthermore, in captive zebra finches (Taeniopygia guttata), males that successfully fledged chicks continued to prefer one of two provided colors as nest material, while males that failed to fledge chicks switched color preference when building a subsequent nest (Muth and Healy 2011).

Alternatively, individuals can use social cues to guide their decisions on nest-site selection and building behavior, thereby incorporating the experience of others (Laland 2004; Kendal et al. 2005). In the context of nest site selection, there is strong evidence that several bird species use social information when choosing breeding locations (Breen et al. 2016). For example, migratory flycatchers (Muscicapidae), when returning to breeding sites, were found to use the breeding success of the resident tits (Paridae) to assess habitat quality and select nest sites (Forsman and Thomson 2008; Forsman and Seppänen 2011; Jaakkonen et al. 2015). Several studies have also suggested an influence of social information use in the context of nest construction (reviewed in Breen et al. 2016; Breen 2021), through observation and copying of material choice. The most comprehensive evidence for social information use in material choice comes from experiments on captive zebra finches (Guillette et al. 2016; Breen et al. 2019, 2020). Individuals with no prior experience in nest building switched their initial color preference of nest material after watching an experienced individual build a nest with their less preferred color material, but only if the demonstrator was a familiar individual (Guillette et al. 2016). Even when presented with only a completed nest of their non-preferred color without a demonstrator present, first-time builders lost their initial color preference and picked material colors at random, indicative that a nest alone without a demonstrator present may be sufficient to influence a future preference for nest material (Breen et al. 2019).

Meanwhile, evidence from the wild for social information use in material choice remains both scarce and more mixed (Breen 2021). The first evidence comes from as early as 1924, when Williams provided colored yarn to nest-building birds in his backyard (Williams 1934). Over the years, orioles (Icteridae) seemed to follow the “fashion of the season” when choosing yarn, with birds converging to choose only white yarn after a conspecific had built its nest using only white yarn, indicative of a social influence on color choice (Williams 1934). More recent, indirect evidence comes from a study on blue tits (Cyanistes caeruleus) that reported local clustering in preferences for particular plants used for nest lining. These preferences did not appear to be predicted by local availability, suggesting the potential for social transmission of plant preferences (Mennerat et al. 2009). Contrasting results were found in a study providing colored wool to breeding tits, where females opportunistically incorporated all materials close to the nest site (Surgey et al. 2012). Finally, a cross-fostering study between blue and great tits (Parus major) did not find any evidence of cultural inheritance of nest lining material (Aasen and Slagsvold 2020). Here, the authors compared the proportion of feathers (naturally preferred by blue tits) and fur (naturally preferred by great tits) (Perrins 1979) in the nests of cross-fostered and control young but found no effect of foster parents’ preference on the offspring’s material choice (Aasen and Slagsvold 2020).

Here, we experimentally test the importance of social information use in finding and choosing nest materials in a wild mixed-species community of tits (Paridae sp.) marked with Passive Integrated Transponder (PIT) tags. In tits, females build a nest consisting of a base layer made mostly of moss and a lining layer of various soft materials such as hair, feathers, fur, and wool for insulation of eggs and chicks (Perrins 1979). These soft materials often represent ephemeral and potentially limited resources. For example, fur may be obtained from a rabbit or badger carcass, or even plucked from live mammals (Pollock et al. 2021). We provided PIT-tag reading dispensers containing two different colors of wool across five replicate areas and recorded both visits to dispensers and presence of wool in nests.

First, we investigated whether females used social information to locate lining materials by using the time of arrival to dispensers in a “network-based diffusion analysis” (NBDA) (Hasenjager et al. 2020). We considered two different networks, a foraging association and a spatial breeding network, which allowed to distinguish between two possible pathways of information transmission. As tits are known to rely on social information from both individuals of their own as well as other species to locate food resources (Aplin et al. 2012; Farine et al. 2015), it is plausible that breeding females may also obtain social information from foraging associates about other types of resources, including the location of nest lining material. Alternatively, birds may gather social information about the nest site and construction material by inspecting other birds’ next boxes (Forsman and Thomson 2008; Loukola et al. 2012; Schlicht et al. 2015; Szymkowiak et al. 2017). We therefore hypothesized that tits may potentially develop a search image for the provided lining material through inspection of nests of other females breeding in close proximity or obtain indirect social information about the location of the lining material through directional cues when observing a neighboring female entering her nest with the provided colored wool. As we could not measure prospecting directly, we used spatial proximity as a proxy network for this pathway.

Second, we investigated whether females socially acquired preferences for particular lining material colors. In five replicate areas, we initially created local preferences for one of the two provided colors by blocking access to the second color before allowing access to both colors. We then monitored the specific wool colors females first incorporated into their nests in each area, expecting to find similarities between the seeded color and the color first incorporated into nests if color preferences were socially influenced.

Material and methods

a) Field methods

We conducted our experiment between 23 February and 11 May 2021 in a study population around the Max Planck Institute of Animal Behavior in Radolfzell, Germany, where 207 nest boxes (Schwegler type 1B, 2 M, 3SV) have been provided for the use of breeding tits. As part of a long-term project on this community, we caught blue tits, great tits, and marsh tits (Poecile palustris) in mist nests or trapped them in nest boxes as nestlings or adults. All birds were equipped with a metal leg ring (EURING ‘Radolfzell Germania’) and a plastic leg ring containing a PIT-tag (Eccel Technology Ltd). Adults were aged and sexed based on plumage (Svensson 1992).

Over three periods of 48 h (6 full days in total) between 23 February and 11 March 2021, immediately prior to the breeding season, we recorded visits by PIT-tagged birds to six bird feeders filled with a mix of kibbled peanuts and sunflower seed. We spaced these feeders around the woodland in an approximate grid to cover the entire study area (Fig. 1C). Access points to feeders were equipped with radio-frequency identification (RFID) antennae (NatureCounters Ltd), with data loggers (Priority1 Design, Australia) recording visits of PIT-tagged birds. We then used the spatiotemporal patterning of these visits to build a foraging association network (see below).

Fig. 1
figure 1

A Illustration of a wool dispenser. Two transparent buckets presented felting wool that could be pulled through mesh at the bottom. Wool colors were either paired as orange and pink or blue and purple, with RFID antennae registering visiting birds’ PIT tags. B Foraging association network of females breeding in the study area. Each node represents an individual bird with shapes indicating the species it belongs to (circle = great tit; square = blue tit; triangle = marsh tit). Edge thickness is proportional to the association strength among birds. C Map of the study area. Dots: nest boxes with colors indicating the color of incorporated wool (no fill = empty; grey = occupied, no wool; pink = pink wool; purple = purple wool; orange = orange wool; blue = blue wool; if two colors in nest: initial color at center, second color as outline). Nests of demonstrator birds that started building when only one color was accessible are additionally marked by a black bar across. Squares: locations of wool dispensers: the first seeded color with a solid outline, the second color with a dashed line. Dispenser areas are marked by a circle with 200 m radius. Black stars: locations of RFID feeders for collecting foraging association data

Full size image

On 26 March 2021, we deployed five dispensers containing felting merino wool, each offering two colors of similar hues ad libitum (Fig. 1A). We used a balanced experimental design with three dispensers providing orange and pink wool, and two containing blue and purple wool (Fig. 1C). These were spaced approximately 300 m apart to avoid too much overlap between dispenser areas, based on Surgey et al. (2012) that suggested that tits are unlikely to travel much further than 200 m to collect wool. Each dispenser was equipped with RFID antennae and data loggers to record visits to the wool by PIT-tagged birds. We initially created an artificial local preference for one of two wool colors in each area by blocking access to the other color with a transparent plastic sheet. Once provisioned material was observed by experimenters in one or more nests in each dispenser area, we allowed access to both colors (Table S1). To test for any potential innate color preferences, we additionally deployed four dispensers in two separate control areas (not shown in Fig. 1A), with unlimited access to the two-color pairs (orange/pink and blue/purple) at all times.

Throughout the experiment, we monitored all nest boxes across the woodlands every 2 − 4 days, recording nest stage and the presence (yes/no), color, and amount (categories 1 − 4, see Table S2) of any provisioned lining material. We identified PIT-tagged breeding females by deploying “faceplate loggers” (NatureCounters Ltd) with an RFID antenna around the nest box entrance for 48 h as nests reached completion, i.e., as females started laying eggs (supplementary Information: SI). This further allowed us to identify which females were not PIT-tagged and needed trapping in nest boxes for ringing.

b) Statistical analysis

  1. i)

    Constructing networks

To investigate whether tits relied on social information to locate lining resources (irrespective of color choice), we used NBDA (Franz and Nunn 2009; Hoppitt et al. 2010; Hasenjager et al. 2020). NBDA tracks the spread of a behavior through a network and infers that it is socially learned if the diffusion of the behavior follows the network connections (Hasenjager et al. 2020). Here, we compared two alternative pathways of information spread about locations of lining resources in a multi-network NBDA (Farine et al. 2015; Wild et al. 2019) by including two different networks: (i) the foraging association network and (ii) the spatial breeding network.

To construct the foraging association network among females, we recorded the identity of all visiting birds with PIT-tags to the RFID feeders that we had deployed around the woodland between 23 February and 11 March 2021 (see field methods; Fig. 1C). From these visits, we identified groups using a Gaussian mixture model to detect clusters in the data stream (Psorakis et al. 2012). We then built social networks using a gambit of the group approach and calculated edge weights using the simple ratio index which ranges from 0 (never observed together in a group) to 1 (always observed in the same group) (Cairns and Schwager 1987; Farine 2013; Farine and Whitehead 2015; Hoppitt and Farine 2018). Since the foraging network was highly stable across the 3 weeks of data collection (week 1 to week 2, Mantel R = 0.634, p = 0.001; week 2 to week 3, Mantel R = 0.650, p = 0.001), we used a static network in NBDA based on all the association data (Hasenjager et al. 2020).

To construct the spatial breeding network, we calculated the Euclidean distances between nest boxes using their GPS locations. We then used the inverted square root of distances between nests to account for the fact that space use is non-linear and that females breeding closer together had higher values. As tits occupy relatively small home ranges during the breeding season (Naef-Daenzer 1994), we the entries between females that were nesting more than 50 m apart to 0. This network served as a proxy for prospecting opportunities (e.g., Schlicht et al. 2015), as well as a proxy for opportunities to observe other females carrying wool, which are expected to occur at higher frequency among birds nesting in close proximity. Recording of prospecting events on faceplate loggers were insufficient to create a prospecting network directly (SI).

The maximum known distance travelled between a nest box and dispenser in our study area was 184 m. We therefore subset both networks to only include birds that were breeding in nest boxes within a 200 m radius of each dispenser and considered dispenser areas as independent replicates (Fig. 1C). This meant that a minority of birds was assigned to more than one dispenser area (Fig. 1C; Table S1).

  1. ii)

    Locating lining resources

We used the “time of acquisition diffusion analysis” (TADA) variant of NBDA v0.9.6 (Hasenjager et al. 2020) and used the time of first arrival of each female at the wool dispensers as diffusion data. We restricted our analyses to visits by PIT-tagged females. If no information on sex was available, we assigned sex either by a process of exclusion if sex was known for the breeding partner, or as female if the bird visited the wool dispensers at least three times, assuming that males would be unlikely to re-visit a resource with lining material. Additionally, we included 10 females that had visited the dispenser but whose breeding location was unknown and set their connections in the breeding network to 0.

We additionally included three individual-level variables (ILVs) that could potentially influence social and asocial learning rates: First, we included species as an ILV as “great tit” or “other.” Great tits naturally prefer wool and fur-like material for lining their nests, while blue tits show a preference for feathers if available (Perrins 1979), which may lead to differences in the rates of social or asocial discovery of the provisioned wool. Second, we included age of individuals as “adult” or “first-year” to account for age-biased learning (e.g., Aplin et al. 2013), as first-time builders may rely more strongly on social transmission of information compared to experienced nest builders. Finally, we controlled for the distance of each female’s nest to the nearest wool dispenser, as females in close proximity to a dispenser are also expected to be more likely to locate the resource. For females whose breeding location was unknown, we assigned them the average distance between nest boxes and dispenser within the respective area. We used the standardized square root of distances for better model fitting (Hasenjager et al. 2020). We built unconstrained models, which allow ILVs to influence the social and asocial learning rate independently (Hoppitt and Laland 2013). We created models in all possible combinations of the two networks and three ILVs (Hasenjager et al. 2020), resulting in 200 different models, and used the Akaike Information Criterion corrected for sample size (AICc) to infer model support (Burnham and Anderson 2002). For each supported variable (with summed Akaike weights \(\sum {w}_{i}\) >0.5), we extracted model averaged estimates as weighted medians and extracted profile likelihood confidence intervals based on the best performing model in which the respective parameter occurred (Morgan 2008). All statistical analyses were conducted in R v4.1.2 (R Core Team 2022).

  1. iii)

    Selecting lining material

To investigate whether tits acquired the color preferences initially seeded in each area, we compared the first color birds incorporated, i.e., the first color of provided wool to appear in their nests, with the color seeded in each dispenser area using a Fisher’s exact test. We additionally accounted for any potential local environmental influence of bucket placement (e.g., in case one bucket was more easily accessible). We did this by extracting the number of reads of each visiting PIT-tagged female on both antennae after access was granted to both colors, expecting females to use both antennae for perching if equally accessible.

Results

Locating lining resources

A total of 46 PIT-tagged females, 31 great tits, 11 blue tits and 4 marsh tits, were recorded in the foraging association network and subsequently observed breeding in the study area. Thirty-six of those were recorded in nest boxes, while 10 were recorded only on the wool dispensers, and were presumably nesting in natural tree cavities (Fig. 1B). Of those 46 females, 21 visited the dispensers, 13 great tits, 4 blue tits, and 4 marsh tits, between 1 and 64 times, with an average of 12 visits (Table S1; Supplementary file 1). With only one tagged learner, dispenser area 4 was excluded from the NBDA analysis (Table S1). Overall, the foraging association and spatial breeding network were not significantly correlated (Mantel test, p = 0.060; r = 0.055) and were therefore both included in NBDA models.

NBDA strongly supported models that included social transmission of information about the location of the lining dispensers through the foraging association network (summed Akaike weights (\(\sum {w}_{i}\)) = 0.64). This was followed by purely asocial models (\(\sum {w}_{i}\) = 0.17), models that included transmission through both foraging association and breeding network (\(\sum {w}_{i}\) = 0.12), and models that included transmission through the breeding network alone (\(\sum {w}_{i}\) = 0.07). In the best performing model (Table S3), 39.9% [95% CI 1.5 − 61.5%] of females were estimated to have used social information to locate lining dispensers, with the rest learning socially. This asocial learning rate was highest in birds close to the dispenser and decreased by a factor of 0.84 [0.60 − 0.99] per m increase in distance between a female’s nest and the dispenser (\(\sum {w}_{i}\) = 0.60; Table S4). Age and species had no influence on either social or asocial learning rate (all \(\sum {w}_{i}\) <0.5; Table S4).

Selecting lining material

We found provisioned wool in 26 out of 68 occupied nest boxes by 19 tagged and seven untagged females (Fig. 1C). Of those 26, eight nests were built while access was restricted to a single color. In the remaining 18 nests, 13 birds used only one color in their nests. Of the 18 females, 1 first incorporated blue, 1 orange, 12 pink, and 4 purple wool. We found significant non-random spatial color clustering, with 10 out of the 18 females preferring the initially seeded color in their dispenser area as their first choice (Fisher’s exact test: N = 18; p = 0.025).

With only 18 nests (15 great tits, 1 blue tit, 2 of unknown species; 8 adults, 5 first-years, 5 of unknown age), sample size was too small to investigate the effect of species and age on color preference (Table S5). By contrast, we recorded eight females in the control areas using provisioned wool. Data on the first incorporated wool color was only available for two out of the eight nests (one pink, one yellow), but females of all eight nests incorporated both colors.

Of the 23 tagged females that visited the dispensers after access was granted to both colors, 16 were registered on both antennae, while seven were only registered on one antenna (Fig. S1). This indicates that both antennae were used for perching and that RFID data were not a reliable indicator of the color choice (see also Supplementary file 1).

Discussion

Our study provides experimental evidence that tits use social information when locating sources of nest lining material and that social information also influences their choice of lining material color. Such quantitative evidence for social information use in finding and using sources of nest lining material in wild birds has been lacking thus far. Given that great and blue tits are well known to rely on social information in a variety of behavioral contexts, including finding new food sources (Aplin et al. 2012; Farine et al. 2015), selecting nest sites (Parejo et al. 2007; Slagsvold et al. 2013), and acquiring foraging behavior (Slagsvold and Wiebe 2011; Aplin et al. 2013, 2015), our results are perhaps unsurprising. Yet, they add to our understanding of the extensive influence of social information on the life history and ecology of these species and highlight the importance of social information use as a potential mechanism for behavioral plasticity in nest-building birds.

Our results from NBDA analyses suggest that social information about the location of lining material was acquired from intra- and interspecific foraging associates, in line with previous work showing that tits rely on social information for locating food resources from both con- as well as heterospecifics (Farine et al. 2015). Meanwhile, we found little evidence for transmission through the spatial breeding network. This suggests that females did not obtain information about the location of the lining material through prospecting or through directional cues when observing neighboring females returning to their nest with colored wool, but rather when in foraging flocks. The lack of a correlation between foraging network and spatial breeding suggests that even though foraging associations shape spatial breeding decisions in great tits (Firth and Sheldon 2016), they may not necessarily predict the identity of birds breeding in direct vicinity.

Females breeding closer to a lining dispenser were more likely to discover its location socially. Given the tits’ territories are concentrated around the nest site during breeding (Naef-Daenzer 1994), it appears plausible that they would be more likely to discover resources that are in close proximity to the nest site. In fact, these results are in line with a previous study providing colored wool to breeding tits, finding that the proportion of birds that used the provided material declined with increasing distance between the source of the material and the nest site (Surgey et al. 2012).

We furthermore observed significant non-random spatial clustering of the provisioned wool colors in each dispenser area, with females preferentially lining their nests with the color that was first seeded in the respective area. In addition, the majority of females in the experimental area (13/18) only incorporated one color, while females in the control area all incorporated both provided colors. Taken together, these results are suggestive of a potential social influence when initially selecting wool color. Our findings are consistent with the early observations of Williams who found that nest-building birds appeared to undergo “fashions” across the years when selecting colored nest material (Williams 1934), as well as the previously documented local preferences for particular aromatic plants in wild blue tits that appeared to be unrelated to the local abundance of those plants (Mennerat et al. 2009). Yet, our results are in contrast to Surgey et al.’s (2012) study, in which breeding tits did not appear to have any preference for a particular wool color. This may, however, be explained by differences in experimental setup. Surgey et al. (2012) initially provided wool in four different colors dispersed throughout the woods during six breeding seasons, before placing four dispensers together in one location to ascertain whether birds had preferences for a particular color. Therefore, at the time of investigating color preferences, females may have already collected extensive personal information and could therefore have been less reliant on social information when choosing colors.

Our study is also in contrast to Aasen and Slagsvold (2020) cross-fostering study between great and blue tits that found no evidence for cultural inheritance of the choice of nest lining material. However, their study investigated preferences at the level of nest material (Aasen and Slagsvold 2020) rather than material color. It seems quite likely that higher-level preferences for lining material, for example, wool or feathers, are a relatively fixed species-specific trait (Perrins 1979; Britt and Deeming 2011), while behavioral flexibility acts within these preferences. Alternatively, it is possible that transmission of social information about lining material does not occur during early life from parent to offspring but is transmitted later in life from peer to peer when birds build their own first nest (Aasen and Slagsvold 2020).

It should be noted that our sample size for both NBDA analyses and investigating color preferences were limited. That said, in NBDA, low sample size, e.g., through missing observations due to untagged individuals, reduces power to detect social transmission and increases uncertainty about the strength of a social transmission effect (Wild and Hoppitt 2018). As such, our NBDA results provide a conservative estimate for the plausibility of social transmission. However, the strength of the effect size should be interpreted with caution, with confidence intervals providing a plausible range (Hoppitt 2017). We were also unable to establish the pathways of transmission of color preferences due to small sample size. Future studies including larger sample size should therefore aim to investigate how females may socially acquire preferences for certain colors, i.e., whether this occurs through prospecting, or whether they observe other females carrying wool while entering their nest boxes or at dispensers or get cues from wool that may be scattered around the dispensers. Studies including larger sample sizes could also investigate whether females rely more strongly on social information for locating lining resources from con- over heterospecifics, as has been demonstrated in a foraging context (Farine et al. 2015). Furthermore, future studies should examine a possible effect of age or species on whether females showed a preference for the initially seeded color. This would give insights into potential learning strategies that nest-building females employ, such as a transmission bias from more experienced to first-time builders (e.g., Guillette et al. 2016; Breen et al. 2019, 2020), or a preference for con- over heterospecific demonstrators (e.g., Farine et al. 2015; Jaakkonen et al. 2015).

Taken together, our study provides quantitative evidence for a role of social information use in both finding and selecting nest-lining material in wild birds and shows that this social transmission occurs horizontally through observation of other breeding females. It demonstrates that while many aspects on material choices in nest building may be based on an innate template (Perrins 1979; Britt and Deeming 2011), social information use can constitute an important factor towards plasticity in nest building. Repeated social transmission of information or behavior can lead to the establishment of local cultural traditions (Fragaszy and Perry 2003), as has previously been demonstrated in wild birds in a foraging context (e.g. Aplin et al. 2015; Klump et al. 2021). We therefore concur with Breen (2021) conclusions that avian nest construction provides a promising avenue for studying animal cultural phenomena and suggest that future studies should aim to investigate experimentally whether such local preferences for nest lining material can persist across generations.