Quantitative genetics of the use of conspecific and heterospecific social cues for breeding site choice

Abstract Social information use for decision‐making is common and affects ecological and evolutionary processes, including social aggregation, species coexistence, and cultural evolution. Despite increasing ecological knowledge on social information use, very little is known about its genetic basis and therefore its evolutionary potential. Genetic variation in a trait affecting an individual's social and nonsocial environment may have important implications for population dynamics, interspecific interactions, and, for expression of other, environmentally plastic traits. We estimated repeatability, additive genetic variance, and heritability of the use of conspecific and heterospecific social cues (abundance and breeding success) for breeding site choice in a population of wild collared flycatchers Ficedula albicollis. Repeatability was found for two social cues: previous year conspecific breeding success and previous year heterospecific abundance. Yet, additive genetic variances for these two social cues, and thus heritabilities, were low. This suggests that most of the phenotypic variation in the use of social cues and resulting conspecific and heterospecific social environment experienced by individuals in this population stems from phenotypic plasticity. Given the important role of social information use on ecological and evolutionary processes, more studies on genetic versus environmental determinism of social information use are needed.

. Characteristics of the 15 forest patches included in the study. Average minimum distance between nest boxes is the average distance between each nest box and its closest neighboring box.

Section B. Derivation of the response variables
Derivation of the response variables describing the use of social cues in breeding site choice was identical to that in Kivelä et al. (2014). Because we were unable to determine the exact dates of breeding site choice (initiation of nest building), we assumed that it took place five days before egg laying started, which is approximately the average time interval observed recently (pers. obs.). Kivelä et al. (2014) tested for the impact of this assumption by repeating analyses using two other time intervals between nest initiation and egg laying (two and eight days); the results were robust to variation in the time interval used.
Response  E[O(x, α, k, t, T -Δ)] was calculated as a weighted average, each nest box being weighted by its probability of occupancy by collared flycatchers during the period 1990-2000. We used the period 1990-2000 to avoid having the same individuals in both the data set used for estimating the general preference of collared flycatchers for certain nest boxes and for the data set used in the repeatability and heritability analyses. In other words, we wanted an independent sample of individuals to estimate the preference for certain nest boxes. Years earlier than 1990 were not included, because they may not have reflected habitat quality at the sites later on, here between 2005 and 2010, due to forestry management and forest succession To define breeding site choice in relation to previous year breeding success of conspecifics and great tits, we derived the effective average number of fledglings F (x, α, i, k, t, T-1) in the neighborhood of the focal nest box i produced by species x in the previous breeding season T -1 as where ( , , , , -1) is the number of fledglings produced by species x in the nest box j of the forest patch k. The difference variable D F was calculated as where E [F(x, α, k, t, T -1)] is the expected average number of fledglings produced by species x in the previous year in the neighborhood of the available nest boxes in forest patch k on day t in year T. As for F(x, α, k, t, T -1)] as a weighted average incorporating collared flycatcher occupancy probabilities in 1990-2000.
The four D O (x, α, i, k, t, T -Δ) and two D F (x, α, i, k, t, T -1) variables were normalized (see main text) for analyses. These variables were then used as the response variables in the quantitative genetic (animal model) analyses.
Section C. Summary of the use of the social cues Figure S2. Summary of the use of (a) conspecific abundance in the current year, (b The predicted estimates for the use of the social cues for breeding site choice for the different collared flycatcher pair combinations (combinations of the female and male "Status" variables), based on the repeatability models (see Tables S4), were similar to those found in Kivelä et al. (2014) (Figure S2). For all five social cues, most of the pair combinations showed significant attraction or avoidance, but varied in the timing of the use of the cues. Breeding sites with high current conspecific abundance were avoided by late settling pairs consisting of at least one young or immigrant pair member ( Figure G=list(G1=list(V=diag(c(0.95*Vp,(0.05/8) We tested the effect of prior specification using the maximum models fitted in the original analyses (i.e. full fixed effects structure, and including the cross-sex additive genetic covariance in the quantitative genetic models). Estimates of the variance components were qualitatively and quantitatively similar irrespective of the prior specification (Tables S2 and S3). with the accuracy of two significant digits) using the five alternative priors are presented for each variable.  Table S3. Comparison of the variance component estimates in the quantitative genetic models (full models including the cross-sex additive genetic covariance) with varying prior specifications. Median estimates using the original prior and the range of median estimates (in parentheses; with the accuracy of two significant digits) using the nine alternative priors are presented for each variable.

Repeatability analyses
Models with the full fixed effects structure Table S4. Parameter estimates (posterior medians) and their 95% credibility intervals in the univariate GLMM estimating repeatability for the use of five social cues for breeding site choice in collared flycatcher: conspecific reproductive success in the previous year (n = 1395 breeding pairs), great tit abundance in the previous year (n = 1446), conspecific abundance in the current year (n = 1432), conspecific abundance in the previous year (n = 1430), and great tit abundance in the current year (n = 1430). The fixed effects include the "Status" (combination of dispersal status and age) for both females and males, the date of nest site choice "NestDate" and all their interactions (denoted with ":"). V PI♀ and V PI♂ are the female and male permanent individual variances, V PATCH is the spatial variance across forest patches, V BOX is the variance between nest boxes and V R is the residual variance. Also the derived metrics total phenotypic variance V P , total permanent individual variance V PI total and the repeatabilities for females R ♀ and males R ♂ and the total repeatability R total are reported. N eff is the effective MCMC sample size. Models with only the intercept as the fixed effect and great tit abundance in the current year (n = 1430). Only the intercept is included as a fixed effect. V PI♀ and V PI♂ are the female and male permanent individual variances, V PATCH is the spatial variance across forest patches, V BOX is the variance between nest boxes and V R is the residual variance. Also the derived metrics total phenotypic variance V P , total permanent individual variance V PI total and the repeatabilities for females R ♀ and males R ♂ and the total repeatability R total are reported. N eff is the effective MCMC sample size. Quantitative genetic analyses Figure S3. Estimates of female (red circles), male (blue triangles) and total (black squares) additive genetic variances (median ± 95% CI) in the use of (c) conspecific success in the previous year and (e) great tit abundance in the previous year as social cues for breeding site choice by collared flycatchers, based on the models with the full fixed effects structure, but excluding the cross-sex additive genetic covariance (see Table S7).  Models with the full fixed effects structure, but excluding the cross-sex additive genetic covariance Table S7. Parameter estimates (posterior medians) and their 95% credibility intervals in the univariate GLMM estimating additive genetic variance and heritability for the use of two social cues for breeding site choice in collared flycatcher: conspecific reproductive success in the previous year (n = 1395 breeding pairs), great tit abundance in the previous year (n = 1446). Compared to the model in Table S6, this model excludes the additive genetic covariance between females and males. The fixed effects include the "Status" (combination of dispersal status and age) for both females and males, the date of nest site choice "NestDate" and all their interactions (denoted with ":"). V A♀ and V A♂ are the female and male additive genetic variances, V DOM♀ and V DOM♂ are the female and male dominance genetic variances, V PI♀ and V PI♂ are the female and male permanent individual variances, V PATCH is the spatial variance across forest patches, V BOX is the variance between nest boxes and V R is the residual variance. Also the derived metrics total additive genetic variance V A total , total phenotypic variance V P and female ℎ ♀ 2 , male ℎ ♂ 2 and total heritabilities T 2 are reported. N eff is the effective MCMC sample size. Models with the full fixed effects structure, but excluding the cross-sex additive genetic covariance and the dominance genetic random effects  Table S6, this model excludes the additive genetic covariance between females and males and the dominance genetic effects for both sexes. The fixed effects include the "Status"

Response variable
(combination of dispersal status and age) for both females and males, the date of nest site choice "NestDate" and all their interactions (denoted with ":").V A♀ and V A♂ are the female and male additive genetic variances, V PI♀ and V PI♂ are the female and male permanent individual variances, V PATCH is the spatial variance across forest patches, V BOX is the variance between nest boxes and V R is the residual variance. Also the derived metrics total additive genetic variance V A total , total phenotypic variance V P and female ℎ ♀ 2 , male ℎ ♂ 2 and total heritabilities T 2 are reported. N eff is the effective MCMC sample size. Models with only the intercept as the fixed effect and excluding the cross-sex additive genetic covariance  Table S7, this model includes only the intercept as a fixed effect.

Response variable
V A♀ and V A♂ are the female and male additive genetic variances, V DOM♀ and V DOM♂ are the female and male dominance genetic variances, V PI♀ and V PI♂ are the female and male permanent individual variances, V PATCH is the spatial variance across forest patches, V BOX is the variance between nest boxes and V R is the residual variance. Also the derived metrics total additive genetic variance V A total , total phenotypic variance V P and female ℎ ♀ 2 , male ℎ ♂ 2 and total heritabilities T 2 are reported. N eff is the effective MCMC sample size. Comparison of repeatability estimates between the repeatability and the quantitative genetic models

Response variable
To illustrate that the quantitative genetic models (with full fixed effects structure and including the dominance genetic variances but excluding the cross-sex additive genetic covariance, see Table S7) give similar repeatability estimates as the more simple repeatability models (full fixed effects structure, see Table   S4), we compared the repeatability estimates derived from these models in Table S10. Similar repeatability estimates suggest that the additive genetic, dominance genetic and (other) permanent individual variance component estimates in the quantitative genetic models are not biased low, and thus show that the data is adequate for fitting the quantitative genetic models. For the quantitative genetic models, the permanent individual variance was derived as the sum of additive genetic, dominance genetic and (other) permanent individual variance components.