Table S1 – Male Mass
Details of 4 competing models investigating the effects of dominance status and time (expressed as a two level factor denoting the first and last half of the breeding season) on the mass of male white-browed sparrow weavers. The most complex model contained only an effect of time (positive coefficient for the 2nd half of season – i.e. males increased condition as the breeding season progresses). Using the nesting rule of Richards (2008), only the top model was retained as all other models within Δ6 AICcwere more complex versions of the top model. ‘Seasonhalf’ – 2 level factor denoting the first and last half of the breeding season; ‘Dom. Status’: 2 level factor classifying individuals as either dominant or subordinate; ‘Season*Status’: an interaction term for the ‘Seasonhalf’ and ‘Dom. Status’ variables.
Table S2 – Female Mass
Details of 4 competing models investigating the effects of dominance status and time (expressed as a two level factor denoting the first and last half of the breeding season) on the mass of female white-browed sparrow weavers. The most complex model containing an interaction between dominance status and time received overwhelming support in the data, with an Akaike weight of 0.99. ‘Seasonhalf’ – 2 level factor denoting the first and last half of the breeding season; ‘Dom. Status’: 2 level factor classifying individuals as either dominant or subordinate; ‘Season*Status’: an interaction term for the ‘Seasonhalf’ and ‘Dom. Status’ variables.
Table S3 – Clutch Size
Intercept / Group Size / Unrelated Males Present / k / logLik / AICc / ΔAICc / Akaike weight0.7191 / -0.006395 / 3 / -4.543 / 15.4 / 0 / 0.74
0.7158 / -0.003502 / + / 4 / -4.491 / 17.5 / 2.1 / 0.26
Details of 4 competing models investigating the effect of group size and the presence of unrelated males on the total clutch size. Model estimates are from a generalized linear mixed model with a Poisson response and log link. When applying the nesting rule (Richards 2008), only the top model was retained as all models with weaker AICc support were more complex version of that top model.
Table S4
Details of 4 competing models investigating the effect of group size and the presence of unrelated males on the hatching probability of eggs. Model estimates are from a generalized linear mixed model with a binomial response and logit link. When applying the nesting rule (Richards 2008), only the top model was retained as all models with weaker AICc support were more complex version of that top model.
Building Candidate Parent Sets
Our candidate parent sets were compiled separately for each clutch in a two step process. First, we identified a lay date window for every clutch in the analysis (a range of dates within which the eggs were laid), either using the exact range of lay dates (where known) or by calculating the earliest possible lay date by backdating 19 days from the earliest possible hatch date and the latest possible lay date by backdating 14 days from the latest possible hatch date, when hatching information was more accurate than laying information (14 to 19 days being the lag from laying to hatching of the first egg). For the few offspring of unknown origin (first discovered as fledglings), we calculated their latest possible lay date as their date when first seen minus 30 days and their early possible lay date as their date when first seen minus 180 days (reflecting the widest age span in which fledgling characteristics may be shown). Second, we compiled conservative inclusive sets of candidate parents for every clutch, by including all individuals who could have been alive (and within the offspring’s natal group for candidate mothers) and over 30 days of age on any day of the clutch’s lay date window. To achieve this we calculated a candidate window for every individual in the population, to identify those days on which it could have been available to mate (within the offspring’s group in the case of candidate females), with the philosophy that this window should be as inclusive as possible (without allowing for time travel) to avoid the false exclusion of true parents by candidate set restrictions. As candidate fathers were drawn from across the entire study population, just one candidate window was calculated for each male, reflecting his period of presence in the population. As candidate mothers are drawn from within the offspring’s own natal group, the candidate windows of females were group specific; females who moved between study groups were therefore given a candidate window specific to each such group.
Candidate window start date (CWS): Birds first observed as adults at any point in the study were assumed to have been present in the population as adults since before the start of the study. As such, if male, their CWS was set to before the start of the study. Whereas, if female, the CWS relating to the group in which they were first seen was set to the first date on which they could have been in that group (before the start of the study if that group had never been previously observed). The CWS for individuals first observed as nestlings was set as the date at which they reached 30 days of age (by which age they had invariably fledged; chosen in the absence of age at fertility information). Any offspring that died in the nest were therefore never considered as candidate parents. The CWS for individuals first observed as fledglings was set to their first date of observation minus 180 days (the oldest they could have been at that time) and females first found as fledglings were assumed to have started life in that group. For any females that moved groups during the study, the CWS for their time in the new group was set to the first date on which our observations suggest they could have been present.
Candidate window end date (CWE): The CWEs for males were set to: (i) the date on which they were last seen in the population plus 365 days (to allow for undetected survival as floaters in the population and for sperm storage by females); or (ii) beyond the end of the study if they were still thought to be alive at the end of this study period. The CWE for a given female in a given group was set to: (i) the date on which she was last seen in that group (plus 365 days if she wasn’t known to have become resident in another group, to allow for an undetected floating association with her source group); or (ii) beyond the end of the study, if she was still alive within that group at the end of this study period.
Exclusion-Based Analysis of Extra-Group Paternity
Of the 292 offspring considered, 63 had genotypes that mismatched the genotypes of every potential within-group father at one or more loci (after accounting for the mother’s genotype), yielding an estimate of 21.6% (63/292) extra-group paternity (EGP). This figure is likely to overestimate EGP as genotyping error or mutation could be yielding the false exclusion of within-group fathers. By comparison, 31 offspring had genotypes that mismatched the genotypes of every potential within-group father at more than one locus, yielding an estimate of 10.6% (31/292) EGP. This figure is doubtless conservative, as some of these within-group father-offspring pairs permitted with single mismatches will reflect genuine extra-group paternities. Our exclusion analyses therefore suggest that the true figure of EGP lies somewhere between 10.6 and 21.6% of offspring (12.2 to 24.4% of clutches).