Fighting ability, personality and melanin signalling in free-living Eurasian tree sparrows (Passer montanus)

Study site and period

The study was carried out in the Botanical Garden and the Central Campus of the University of Debrecen (Debrecen, Hungary) during three winters (i.e., between mid-October and mid-March), in 2013–2016. Birds were trapped, marked and tested for exploratory behaviour during all three years of the study, while data on agonistic interactions were collected during the last two years (see also Fülöp et al., 2019). The study site is an open area with scattered trees and bushes, also containing some buildings of various sizes forming a heterogeneous semi-urban landscape mosaic. Further details about the study site are given in Barta, Liker & Mónus (2004) and Fülöp et al. (2019).

Both the bird trapping procedures and the observations on the social behaviour of individuals were performed on three feeders installed at our study site. Feeders consisted of wooden platforms made of oriented strand board (120  × 120 cm) placed on the ground (Fülöp et al., 2019; Fig. 1). During all three winters, we provided ad libitum food (sunflower seeds) for the birds on a daily basis, on all the feeders throughout the whole winter period, except periods when recordings of social behaviour were made (see below; Fülöp et al., 2019).

Field procedures

Field procedures were the same as described in Fülöp et al. (2019). We captured tree sparrows with mist nets (Ecotone, Poland; https://www.ecotone.com.pl/) at bird feeders each winter between mid-October and mid-January, except the first winter (2013/14), when birds were captured during the whole winter (i.e., between mid-October and mid-March). At capture, we fitted each individual with a standard ornithological metal ring, issued by the Hungarian Bird Ringing Centre, which was supplemented with a unique combination of three plastic colour rings, enabling the identification of the marked individuals from distance. Furthermore, we measured body mass (± 0.1 g with a Pesola spring balance), tarsus length (± 0.01 mm with a digital caliper) and wing length (± 0.5 mm with a ruler); and because sexes are alike in terms of appearance and size in tree sparrows (Mónus et al., 2011), we also took a blood sample (∼50–150 µl) from the brachial vein of the birds to carry out molecular sex determination of individuals (for details, see Fülöp et al., 2019; Fülöp et al., 2021). To quantify bib size (i.e., area; mm²), we photographed the black bib of the birds on the field and measured bib area subsequently from the photographs with the ImageJ software (ver 1.51i ran on a Linux operating system; https://imagej.nih.gov/ij/index.html). While preparing the photographs, we held the focal individual in a standardized position: the head of the bird facing the camera so that the axis of the beak was perpendicular to the axis of the body and the camera sensor plane. We also included on each photograph a standard reference for length (i.e., ruler or millimetre paper), which was necessary to calibrate unit length during measurements. Measurements were completed in three steps for every individual: first, we calibrated unit length on the photograph using the “set scale” function of ImageJ, then we traced the outline of the black bib patch with the highest accuracy possible using the “freehand selection” tool, and lastly, we obtained bib area using the “measure” function. All bib size measurements were carried out by the same person (AF). To increase measurement precision of our data, we measured bib size for each individual from the same photo twice and averaged the values. Repeatability of the two measurements was high (intraclass correlation coefficient, performed with the “ICC” package for R (Wolak, Fairbairn & Paulsen, 2012; R Core Team, 2021): ICC = 0.991, 95% CI = 0.988–0.994, N = 175). Finally, each individual was tested for exploration (see below). After completing all the procedures, individuals were released at the site of their capture. Between 2013 and 2016, we individually marked 194 tree sparrows in total (80 males and 114 females) (Fülöp et al., 2019).

Quantifying exploratory behaviour

We measured exploratory behaviour of individuals using the standard open-field test (Dingemanse et al., 2002). The tests and further statistical analyses of the data were carried out as detailed in Fülöp et al. (2019). Briefly, we recorded the behaviour of each individual for 10 min in a novel environment (i.e., an empty room with four artificial trees; room size: 3.25 × 2.55  × 1.95 m (L × W × H)) using a handheld video camera (Panasonic HC-V510) through a one-way mirror window. On any day, recordings were made by one of two observers (AF or ZN). Then, from each video, we coded a series of events, as follows: the time of the first landing of the focal individual after its release into the test room, the time of its first movement after the first landing, the number and type of movements performed, and its propensity to explore the test room. We distinguished three different movement types: hops (short distance movements performed either on the ground or on the trees without the bird using its wings), flight-assisted hops (short distance movements combining hops that are assisted with usually one wing beat), and continuous flights (longer distance movements, lasting usually a few seconds, where the bird performs typical flight). We characterized the individuals’ propensity of exploration of the test room by dividing the test room into eight equal-sized imaginary compartments and counting the number of occurrences of the focal individual in the different compartments of the room. Compartment limits of the test room were marked by reference lines painted on the walls of the room. All videos were analysed by the same person (TB) with a video analysis software (“mwrap”; Bán et al., 2017).

Using the events coded from the videos, we first calculated the variables “movement latency” (s; i.e., the time elapsed between the first landing of the individual in the test room and its first movement), cumulative “number of movements” performed during the test (i.e., sum of the three movement types), “number of different compartments visited” (i.e., ranging 1–8) and “total number of position switches between compartments” of the test room (i.e., sum of the position switches between compartments). Then, using these four variables, we performed a principal component analysis (PCA) and extracted predicted values for the first principal component (PC1). PC1 strongly correlated with three of the four behavioural variables measured, the exception being “movement latency”; and explained 65.95% variance, having the following loadings for “number of movements” during the test, “number of different compartments visited” and “total number of position switches between compartments” of the test room: 0.556, 0.534 and 0.518, respectively. Therefore, we used these predicted values from PC1 to characterize exploratory behaviour of individuals. Note that higher PC1 scores indicate a higher exploratory tendency of individuals. In total, between 2013 and 2016 we tested 189 individuals (78 males and 111 females) for exploration.

In order to confirm that exploratory behaviour is a personality trait in tree sparrows, we have tested a subset of individuals (N = 48 individuals, 20 males and 28 females) multiple times for exploration (45 individuals twice and three individuals three times). Exploratory behaviour was significantly repeatable (adjusted repeatability calculated with a linear mixed-effects model with normal distribution, R package “rptR” (Stoffel, Nakagawa & Schielzeth, 2017): r = 0.455, SE = 0.105, 95% CI = 0.253 to 0.669, P < 0.001, P_perm = 0.002; see Fülöp et al., 2019 for details). For repeatedly measured individuals, we used the exploration score of their first exploration test in the subsequent statistical analyses.

Assessment of fighting ability

In order to determine individual fighting ability we recorded agonistic social interactions between sparrows while foraging socially on feeders during the last two winters of the study (2014/15 and 2015/16) at two of the feeders (for details see Fülöp et al., 2019). During the winter of 2014/15, recordings were made on 17 different days, between 7 January and 11 February 2015, while during the winter of 2015/16, on 29 different days, between 21 January and 4 March 2016. Recording sessions were performed during daytime (between 8 AM and 5 PM; GMT +1), overlapping with the activity period of the tree sparrows during the winter. During the recording sessions, we restricted individuals’ access to food to stimulate the emergence of various forms of social interactions between them (i.e., agonistic interactions and producer-scrounger foraging tactic use; Fülöp et al., 2019). To do this, we placed a second layer of oriented strand board of matching size on the feeders which contained 121 holes (diameter: 24 mm, depth: 18 mm, distance between adjacent holes: 10 cm) arranged in a 11 by 11 rectangular grid (Fig. 1). Before the start of the video recording sessions we first removed all the food from the feeder, then filled 10 randomly selected holes with one teaspoon of corn grit. During a recording session we typically recorded multiple foraging bouts. A foraging bout started when the first individual arrived at the feeder and ended when the last bird from the flock flew away from it. The recordings were made by one of three observers at a time (AF, ZN and ZB) using a video camera (Panasonic HC-V510) without disturbing birds (for details see Fülöp et al., 2019).

Agonistic interactions between pairs of individuals (i.e., dyads) were analysed subsequently from the videos by the same person (GC). Agonistic interactions included various forms of behaviour exhibited by one or both of the participants, such as displacement, displaying, pecking and/or fighting. For each dyadic interaction in which at least one colour marked individual was involved, we noted the identity of the initiator, the winner and the loser. Besides, for each interaction we have calculated the density of the foraging individuals on the feeder by dividing the number of tree sparrow individuals with the area of the feeder visible in the moment of the interaction from the same video frame (see Fülöp et al., 2019). In total, we have recorded a total of N = 581 fights, out of which N = 74 were fights where no clear winner and loser could be recognized. These latter fights were excluded from the statistical analysis. We were unable to calculate the density of the individuals on the feeder for N = 25 fights because the visible area of the feeder could not be determined from the video frame accurately (e.g., the feeder was covered with snow). Hence, these fights were also discarded from the data. Finally, since no exploration data was available for two marked individuals, another N = 16 fights were dropped from the data set. Our final data set thus contains N = 466 agonistic interactions with an unambiguous outcome from 32 different individuals, of which 14 were males and 18 were females (mean number of fights per individual = 16.19, SD = 21.72, median = 6.00, range = 1–106).

We characterised the fighting ability of individuals using the probability of winning as an index. Instead of calculating dominance rank for individuals, we used winning probability because for the majority of fights (N = 509 from the total of 581 fights, 87.61%) the identity of only one participant is known and calculating dominance ranks using a data set with a similar structure is not possible (Sánchez-Tójar, Schroeder & Farine, 2018a). However, fighting ability (i.e., RHP) is an important individual attribute determining dominance status (Tibbetts, Pardo-Sanchez & Weise, 2022), and indeed many previous studies (e.g., Liker & Barta, 2001; Dingemanse & de Goede, 2004; Cole & Quinn, 2012) have shown that fighting ability of individuals is correlated with their dominance status.

Statistical analyses

All calculations and statistical analyses were performed in the R statistical environment version 4.2.0 (R Core Team, 2022) run in R Studio version 2022.02.2.485 (RStudio Team, 2022). We have analysed the relationship between individual phenotypic traits and fighting ability with a Bayesian generalized linear mixed-effects model with a binomial response distribution and logit link function, as implemented in the R package “brms” (Bürkner, 2017), an interface for Stan (Stan Development Team, 2015a; Stan Development Team, 2015b). The model contained fight outcome as a dependent variable (coded with a binary value, 0 or 1, depending whether the individual lost or won the given fight, respectively), observation year (2014/15 and 2015/16) and sex (male or female) as categorical predictors, and the following individual traits as continuous predictors: body size, exploration score and bib size. Agonistic behaviour of individuals, and hence chances of winning fights, can show strong temporal variations on different time scales, as follows. Since tree sparrows usually arrive sequentially on the feeder within a foraging bout, competition between individuals can change (presumably increase) with the time spent on the feeder. Hence, the probability of winning can change within a foraging bout. Furthermore, behaviour of individuals can also vary during the day, due to the daily variation in individuals’ energy reserves, and can also change over the season, owing to changes in environmental conditions (e.g., temperature). Therefore, to control for these possible effects, we entered into the model the time of the agonistic event within the foraging bout (expressed in seconds passed from the start of the foraging bout) and time of the foraging bout (expressed as minutes spent from midnight), as continuous predictors. To account for the similarity in conditions within each observation day, instead of including date as an additional continuous effect in the model, we entered date (the unique identity for each observation day) as a random effect. We also accounted for the effects of intraspecific competition and environmental harshness by entering in the model the density of the foraging individuals and the minimum temperature of the day (°C; source of data: https://www.ncei.noaa.gov/access/past-weather/debrecen). In addition to the above listed main effects (i.e., categorical and continuous effects), we have tested in the model all the second order interactions of observation year and sex, with the continuous predictors. Individual identity and foraging bout identity were entered in the model as two additional crossed random effects, to control for repeated measurements from the same individuals and potential temporal autocorrelation in behaviour within the same foraging bout, respectively. Although we carried out our observations on two different feeders, we did not include this effect in our final analysis, since entering feeder identity as a categorical predictor and its’ interactions with the continuous variables in the model resulted in convergence difficulties in a preliminary phase of the analysis, due to the moderate sample sizes in terms of number of individuals (N = 32 individuals), and due to the uneven distribution of sexes across years and feeders, respectively. For the same reason we omitted also the interaction between observation year and sex from the model. Nevertheless, a priori we are not expecting a strong effect of feeder location on the relationship between individual traits and fighting ability, since the two feeders were located close to each other (aprox. 400 m) in the same habitat. Body size was expressed using predicted values of the first principal component (PC1) extracted from a PCA (“prcomp” function in R, package “ade4”; Dray & Dufour, 2007) that included body mass, tarsus length and wing length of individuals, all measured at capture (see above). Only PC1, which explained 62.97% of the total variation, had an eigenvalue >1 and it had the following loadings for body mass, tarsus length and wing length, respectively: 0.639, 0.572 and 0.514. Prior to entering into the model, bib size values and time of the agonistic event within the foraging bout were log-transformed to handle the skewness of these variables, and all the continuous predictors were Z-transformed (mean = 0, SD = 1) using the “scale” function in R to facilitate model convergence. We ran the model with default priors (i.e., relatively uninformative priors), four Hamiltonian Monte Carlo chains, each chain for 12,000 iterations, and using a warm-up period of 2000 iterations. We checked model convergence using trace plots and Rhat values (all Rhat = 1). Post-hoc comparisons between different slopes of continuous predictors involved in interactions with a 95% credible interval not overlapping 0 (i.e., effects considered “significant” in a frequentist sense) were performed using the R package “emmeans” (function “emtrends” in R; Lenth, 2020). Plots were made using the R package “ggplot2” (function “ggplot” in R; Wickham, 2016). We report posterior means for all estimated parameter coefficients ± 95% credibility intervals (CrI) or 95% highest posterior density intervals (HPDI).

Ethical note

Tree sparrows were ringed under a license from the Hungarian Bird Ringing Center (license nr. 390 accredited to ZB) and permission for the study was granted by the Hajdú-Bihar County Governmental Office, District Office of Debrecen - Department of Environmental and Nature Protection (permit nr. HB/10-KTF/00487-1/2016). The study complies with the European laws regarding animal welfare, and adheres to the ASAB/ABS guidelines for the use of animals in behavioural research.

Exploratory behaviour and fighting ability

Some of the previous studies concluded that individual personality traits and fighting ability might be related to each other (e.g., Cole & Quinn, 2012; Rudin & Briffa, 2012; Lane & Briffa, 2017), while a handful of others found no support for similar relationships (e.g., Courtene-Jones & Briffa, 2014; O’Shea, Serrano-Davies & Quinn, 2017; Kaiser, Merckx & Van Dyck, 2019). Our results are in line with the latter ones, as we found no association between probability of winning a fight and exploratory behaviour in any of the sexes. This could occur, for instance, if the relationship is context-dependent. Findings supporting context-dependency of this relationship are available from great tits. For example, Verbeek, Boon & Drent (1996) and Verbeek et al. (1999) found that the sign of relationship between exploratory behaviour and fighting ability has changed (i.e., positive vs. negative) depending on the social context (i.e., individuals tested in small vs. large groups). Similarly, Dingemanse & de Goede (2004) found that the sign of relationship between exploratory behaviour and dominance, which was positively correlated with the proportion of fights an individual won, varied from positive to negative, depending on the territorial context (i.e., territory holding adult vs. non-territorial juvenile individuals). In tree sparrows social context, determined especially by groups’ phenotypic composition, might be the most relevant factor influencing trait correlations during the winter. In our study system, group composition seems to be variable both in the short term (i.e., within-season) and in the long term (i.e., between years). In the short term, group composition can vary because tree sparrow flocks likely exhibit a fission–fusion type of group dynamics (pers. obs.). In the long term (i.e., between years), flock membership can change due to emigration/immigration and natural mortality. In both cases flocks’ composition might change, potentially reshaping dominance structure of groups and/or determinants of status. Since our study was carried out on free-living individuals, part of them unmarked, we cannot rule out potential confounding effects originating from group dynamics that could influence our results.

A further explanation for not detecting an association between exploratory behaviour and probability of winning may be that the relationship between the two is shaped by a variable we did not measure. Individuals included in the present study can form a mixed sample regarding factors outside the scope of this study, for instance, age, which cannot be determined on the basis of plumage characteristics or other physical traits in tree sparrows. If any such unmeasured trait is important in determining the relationship between exploratory behaviour and fighting ability (see e.g., Dingemanse & de Goede, 2004; Couchoux et al., 2021), an association between the two could remain masked. Overall, the fact that we have not found a significant relationship between fighting ability and exploratory behaviour does not exclude that a correlation still exists between the two under certain scenarios. In order to clarify these aspects, further studies are needed.

Sex-specific melanin signalling of fighting ability

The role of conspicuous melanin-based ornaments serving as badges of status has been the focus of multiple studies (reviewed by Senar, 2006; Santos, Scheck & Nakagawa, 2011), and apparently, the signalling function of various badges of status is far more complex than it was initially expected. This phenomenon is well exemplified by the case of the house sparrow (Passer domesticus), where the black bib of males was considered one of the classic examples of status signalling, but the signalling role of the badge has been recently questioned by a meta-analysis (Sánchez-Tójar et al., 2018b). Our findings on a closely related, but much less studied species, the tree sparrow, are similarly intriguing. We found that probability of winning a fight was sex-dependently associated with bib size. Moreover, while in a previous study in the same population, performed 15 years earlier (i.e., winter of 1999/2000), Mónus et al. (2017) found that bib size was related to the proportion of fights won in males, but not females, here, we found that bib size was associated with a higher probability of winning a fight in females, but not in males. Our results suggest, for the first time, that bib size might have a status signalling function also in females in this species.

The opposite pattern we found regarding the status signalling function of the bib, compared to Mónus et al. (2017; see also Torda, Liker & Barta, 2004), might be the consequence of the marked changes in environmental factors occurring between the times of the two studies through their effect on or independently of group dynamics (e.g., size, composition, density). Environmental conditions are capable to influence social dynamics for free-living individuals, for instance, through food availability (e.g., quantity, distribution), which can have social implications in the short-term (i.e., within-season) by influencing fission–fusion group dynamics (see e.g., del Mar Delgado et al., 2021). In the long-term (i.e., between years), environmental changes can potentially shape population structure (e.g., sex ratio, age structure, size), for instance, through regulating annual survival (e.g., Santisteban et al., 2012; Gullett et al., 2014). Here, we found that minimum temperature of the day had no effect on the probability of winning, suggesting that individual fighting ability is not influenced by environmental harshness during the winter. In contrast, we found a negative effect of density on the probability of winning in females. This suggests that intraspecific competition can affect fighting ability, at least in one of the sexes. The fact that density apparently acts differently on the fighting ability of individuals in males and females suggests that the RHP of males depends primarily on factors which are independent of their environment (e.g., internal factors), while in females RHP might dependent on internal and/or external factors as well. Nevertheless, the sex-specific effect of density on individual’s RHP has to be further studied.

Similarly to exploratory behaviour, another potential explanation for the differences between the two studies can be that one or more confounding effect(s) acting on the association between bib size and fighting ability which could differ between the two studies were not quantified. For instance, body condition, immunity and stress level, or parasite load are among the factors which can be listed as potential candidates. These effects have been shown to be associated with melanin-based colouration (e.g., Fitze & Richner, 2002; Wiebe & Vitousek, 2015; Fülöp et al., 2021; reviewed by San-Jose & Roulin, 2018). Although bib size of individuals is acquired during the autumn (i.e., between July and October), when tree sparrows perform their complete annual moult, these factors can still influence the relationship between bib size and fighting ability later in winter, for instance, if bib size is an indicator of individual quality.

Finally, an additional aspect to consider, which is applicable for both studies (Mónus et al., 2017 study and this study), and which can influence the association between fighting ability and plumage traits both in the short-term (i.e., within-season) and in the long-term (i.e., between years), is familiarity between individuals. Familiarity can be an important factor influencing the information value of plumage signals (Chaine et al., 2018). Thus, changes in the social environment can shape the correlation between fighting ability and bib size. We need to point out that for a large fraction of the fights only one of the interacting individuals’ identity is known in both studies (see Methods here and in Mónus et al., 2017). Hence, limited information is available about the familiarity between opponents in both studies.

It remains an interesting question what mechanism can explain the apparent sex-specific differences in the signalling role of this mutual ornament, as found in multiple instances in tree sparrows (e.g., Matsui et al., 2017; Mónus et al., 2017; Fülöp et al., 2021, this study), and what are the ultimate consequences of it. Overall, the causes and implications of this phenomenon are poorly understood, and hence, the sex-specific signalling function of mutual colour traits needs to be further investigated.

Is body size an indicator of fighting ability?

According to theoretical studies (Parker, 1974; Maynard Smith & Parker, 1976), which are supported also by a large body of empirical evidence (e.g., Richner, 1989; Lindström, 1992; Briffa, 2008; Chamorro-Florescano, Favila & Macías-Ordóñez, 2011; Rudin & Briffa, 2011; Couchoux et al., 2021), larger body size confers a higher RHP to individuals. Therefore, in our analysis we have accounted for the effect of body size, being a potential confounding factor in determining fighting ability of individuals. In contrast with previous expectations, however, body size was not related to the probability of winning a fight in tree sparrows, in any of the sexes.

Interestingly, our results differ again from those of Mónus et al. (2017). They found that a sex-specific relationship exists between measures of body size (i.e., body weight and wing length) and proportion of fights won. Specifically, Mónus et al. (2017) showed that in males smaller individuals, while in females larger individuals won more fights. This apparent difference between the two studies might be also the consequence of the marked changes in environmental factors occurring between the times of the two studies and its effect on social dynamics (see above). Similar dynamics occurring in the social environment of individuals might ultimately influence correlational patterns between morphological traits, like body size, and fighting ability, similarly to bib size.