Aim: Nest building is widespread among animals. Nests may provide receptacles for eggs, developing offspring and the parents, and protect them from adverse environmental conditions. Nests may also indicate the quality of the territory and its owner and can be considered as an extended phenotype of its builder(s). Nests may, thus, function as a sexual and social signal. Here, we examined ecological and abiotic factors—temperature, nest predation and interspecific information utilization—shaping geographical variation in a specific nest structure—hair and feather cover of eggs— and its function as an extended phenotype before incubation in great (Parus major) and blue tits (Cyanistes caeruleus) across Europe. We also tested whether egg covering is associated with reproductive success of great tits.

Location: Fourteen different study sites and 28 populations across Europe.

Taxon: Parus major, Cyanistes caeruleus.

Methods: We recorded clutch coverage estimates and collected egg covering nest material from the tit nests. We also measured nest specific breeding parameters and phenotypic measurements on adults. We tested whether mean spring temperatures, nest predation rates and flycatcher (Ficedula spp) densities in the study areas explain the large-scale geographical variation of clutch coverage and reproductive success of tits.

Results: The degree of egg coverage of great tits increased with lower mean spring temperature, higher nest predation rate and higher flycatcher density. We did not find egg covering of blue tits to be associated with any of the ecological or abiotic factors. Moreover, egg covering of great tits was not associated with reproductive success in our cross-sectional data, yet a rigorous assessment of fitness effects would require long-term data.

Main conclusions: Our findings suggest that, in great tits, egg covering may simultaneously provide thermal insulation against cold temperatures during egg-laying in spring and also represent a counter-adaptation to reduce information parasitism by flycatchers and nest predation. Hence, geographical variation in interspecific interactions, and consequently in co-evolutionary processes, may affect the evolution of nest characteristics besides environmental conditions.

Study areas. The great tit data for this study were collected in spring 2013 from 10 different countries, 14 study areas (Figure 1, Table S1) and 28 populations in Europe. The blue tit data were collected in the same year from 6 of the 10 countries, 8 of the 14 study areas and 22 of the 28 populations. All study populations breed in nest boxes. Research was carried out in accordance with legislation of each country.

Field procedure. Nest building state and the beginning of the egg-laying (laydate) were checked during regular field observations. Use of egg cover tends to increase during the first day of the laying stage (Haftorn & Slagsvold, 1995). During the egg-laying stage, when tits had laid their fourth to eighth egg, (a) the nest was photographed to get a measurement of the extent of the clutch coverage, that is, the proportion of the visible clutch surface (%) and (b) all the hair and other material that covered tit eggs and nest cup was removed to expose the eggs and placed in a zip lock bag for later measurement of hair mass and the nest was photographed again. After photographing the nest, the removed material was replaced by same quantity of sheep hair. The onset of incubation was determined by observing the presence of female on the nest and touching eggs to determine whether the eggs were cold or warm. Nest specific breeding parameters (number of hatched eggs and fledglings) and phenotypic measurements on adults (Table 1) were also collected. We recorded clutch coverage estimates and mean spring temperatures (from the nearest available meteorological stations to each of the study area) from 476 great tit nests and 123 bluetit nests and nest predation rates from 345 great tit nests and 74 blue tit nests. Flycatcher (either Ficedula hypoleuca or Ficedula albicollis) density was measured in the end of breeding season as the proportion of nest boxes occupied by flycatchers in the study population.

Measurement of nest characteristics. The clutch coverage rate was measured by comparing the proportions of the visible clutch surfaces from the digital photographs taken from the nest before and after cover removals using ImageJ software (US National Institutes of Health, http://imagej.nih.gov/ij). The clutch surface was measured using freehand tracing and area calculator tools. Clutch surfaces were measured twice from each picture to minimize measurement error and average values were used in the analyses. Masses of the collected hair samples were weighed to the nearest 0.0001 g by using an Ohaus AS120S analytical balance. Phenotypic measurements on adult tits were obtained when they were captured during food provisioning. Age was classified in the field as 1-year-old (second calendar year) or older (at least third calendar year) (Jenni & Winkler, 1994). Adult and young birds were handled under the ringing licenses of the authors. Hence, our study complied with the national legislation of Belgium, Czech Republic, Denmark, Estonia, Finland, Hungary, Italy, Spain, Sweden and Switzerland concerning handling wild animals. Variables used in statistical analyses are listed in Table 1.

Statistical methods. The distribution of clutch covering rate (proportion of covered eggs) was slightly U-shaped with a high peak at one (all eggs covered), which is problematic for analysis. Therefore, we measured clutch coverage by combining clutch covering rate and the mass of material used to cover the eggs, because these variables together measure the investment of the tit parents in covering their clutch. For this purpose, we ran principal component analysis for the data on egg coverage and the mass of the covering material and used the first principal component (‘clutch coverage’ hereafter, explains 72.3% of the variance, eigenvalue = 1.0) as a response variable when analysing variation in clutch covering behaviour. Clutch coverage variable was symmetrically (approximately normally) distributed, and positively correlated with both clutch covering rate and mass of the covering material, higher values, thus, indicating higher investment in clutch covering (Figure S1). All statistical analyses were conducted with R version 3.4.3 (R Core Team, 2017). Linear mixed-effects models (LMMs; function lme in package ‘nlme’ (Pinheiro et al., 2017)) were used to analyse variation in the clutch covering of great (Model sets 1–4) and blue tits (Model set 5). Generalized linear mixed-effects models (function glmer in package ‘lme4’; Bates, Mächler, Bolker, & Walker, 2014) with Poisson distribution and a logarithmic link function were used to analyse variation in the number of hatched eggs and fledglings of the great tits (Model sets 6–9). In model sets 6–9, we standardized all continuous explanatory variables of the model. Standardization makes the quantitative interpretation of model parameters less intuitive, which is the reason why standardization was used only when it was really needed for aiding/facilitating model convergence. We used multi-model inference; effects of analysed variables were summarized by model averaging (Burnham & Anderson, 2002) (function model.avg in package ‘MuMIn’; Barton, 2009). We derived 10 model sets. Model set 1 (Table 2a) tested if the alternative hypotheses (i.e. insulation, nest predation, information parasitism), nest floor surface area or forest type (dominant tree genus) in the study site explain variation in clutch coverage in great tits. This model set was fitted to data (Nobservations = 341) including observations from all the study sites, also including sites where the flycatcher density was zero. Mean spring temperature, nest predation rate, flycatcher density, nest floor surface area, first principal component of geographical variables (altitude, latitude and longitude) and dominant tree genus in the study site were set as fixed effects, and population as a random effect in the global model. Time of the year (Laydate) was not included in the analysis because it is strongly negatively correlated with the first principal component of geographical variables (Pearson's correlation, r = −.61, t = −3.41, df = 343, p < .001) (see Laydate in Table S3). Nest predation rate positively correlates with mean spring temperature (r = .67) but both of these variables were retained in all models because of their importance for assessing the study hypotheses. No interactions were included in any model. The set of all meaningful models simpler than the global model was derived with the function ‘dredge’ (package ‘MuMIn’; Bartón & Barton, 2017) for model averaging, the global model being included in model averaging (see Table S2a for the set of averaged models). Bivariate correlations between the study variables are provided in Table S3. Model set 2 (Table 2b) was derived otherwise similarly to model set 1, but all possible interactions among the nest predation rate, mean spring temperature and flycatcher density were included in the global model (see Table S2b for the set of averaged models). Model set 3 (Table 3a) was derived otherwise similarly to model set 1, but these models were fitted to data including only areas where the flycatchers were present (flycatcher density > 0, Nobservations = 169, see Table S4a for the set of averaged models). This was done to reliably estimate the effect of flycatcher density. In the full data the high number of zeros (flycatchers not present) might confound the estimation (underestimation) of flycatcher density effect on great tit egg covering behaviour. Model set 4 (Table 3b) was derived otherwise similarly to model set 3, but all two-way interactions among the nest predation rate, mean spring temperature and flycatcher density were included in the global model (the three-way interaction was ignored as the model-fitting failed when it was included in the global model; see Table S4b for the set of averaged models). Model set 5 (Nobservations = 74; Table S5) was otherwise similar to model set 1, but it focused on variation in clutch coverage in blue tits (see Table S6 for the set of averaged models). Because of the low number of blue tit observations, we did not conduct any further analyses for this species. Model sets 6 and 7 tested whether the clutch coverage of great tits or the ecological and abiotic environment explain the number of hatched eggs of great tits by using all great tit data. In model set 6 (Table S7), number of hatched eggs was used as a dependent variable and clutch size was added as a covariate in the global model in both model sets to take the effect of clutch size variation into account. In addition to clutch size, fixed effects of the global models included also clutch coverage, mean spring temperature, nest predation rate, flycatcher density, as well as female age and its interaction with clutch size, because female age affects clutch size (Perrins & Mccleery, 1985). Population was set as a random effect. Except clutch size (a non-negative integer), continuous variables were standardized (by subtracting mean from each observation and dividing this difference by standard deviation, see Table S8 for the set of averaged models) to aid model convergence. Model set 7 (Table S9) was derived otherwise similarly to model set 6, but all possible interactions among the nest predation rate, mean spring temperature and flycatcher density were included in the global model (see Table S10 for the set of averaged models). Model sets 8 (Table S11; see Table S12, for the set of averaged models) and 9 (Table S13; see Table S14, for the set of averaged models)) tested whether the clutch coverage of tits or the ecological and abiotic environment explain the number of fledged offspring of great tits and were derived otherwise similarly to model sets 6 and 7, respectively, but number of fledged offspring was used as a dependent variable. Population was set as a random effect in all models. Model set 10 (Table S15; see Table S16, for the set of averaged models) was otherwise similar to model set 1, but flycatcher density was replaced with flycatcher presence (binary variable; flycatcher present or not) for checking whether the results are sensitive to the way how flycatcher density is handled (continuous vs. presence/ absence).