SEARCH

SEARCH BY CITATION

Keywords:

  • birds;
  • comparative methods;
  • mating system evolution;
  • parental care;
  • parent–offspring interactions;
  • parent–parent interactions;
  • sexual size dimorphism

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

1. Determining the factors that shape the unequal division of parental effort between the sexes is key to understanding the evolution of mating systems and sexual cooperation and conflict. The range of possible care strategies may be constrained, however, by the duration of the care period. Thus, comparative analyses of parental effort should consider both these dimensions of care; here, we present a method for quantifying parental care that does this.

2. We test three models of the relationship between care duration and the division of parental effort. Using detailed information on parental effort for 330 bird species from 13 families, we use Linear Mixed Effects models and Phylogenetically Weighted Generalized Least Squares models to analyse the relationship between the two dimensions of care. These models provide a starting point for more detailed comparative analyses of the role of covariates, such as sexual size dimorphism on determining the division of parental effort.

3. Within families, although males generally provide less care than females, their relative role often increases with the duration of the care period. Across all species, care strategies vary more in species with short development times.

4. Understanding the evolution of parental care strategies requires consideration of the total duration of care in addition to the division of parental effort. Our method provides a simple means to incorporate both dimensions of care into more extensive comparative analyses.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

Determining the ecological and evolutionary factors that influence unequal division of parental effort by males and females, and termination of care by one or other sex before the offspring are independent (i.e. offspring desertion) is key to understanding variation in mating systems (Trivers 1972; Clutton-Brock 1991; Székely et al. 1996; Bennett & Owens 2002). Patterns of parental care provision also provide a model system with which to investigate cooperation and conflict between the sexes more generally (Lessells 1999; Houston, Székely, & McNamara 2005; Kokko & Jennions 2008). This is because parental care is costly (Clutton-Brock 1991; Bennett & Owens 2002; Houston, Székely, & McNamara 2005), as each parent pays the cost individually (time, energy, survival) but the benefit (the offspring) is shared (Smiseth & Moore 2004). The best interest of both parents then is often to desert early and leave care provisioning to their mate (Parker, Royle, & Hartley 2002; Houston, Székely, & McNamara 2005; Szentirmai, Székely, & Komdeur 2007). Various developmental and ecological factors may influence the extent to which parents are able to desert, including developmental mode and ambient temperature (Bennett & Owens 2002; Kokko & Jennions 2008; Olson et al. 2008). Constraints on the range of care strategies available to a species will in turn limit other aspects of ecology and life history.

Although detailed studies of individual species provide critical information regarding these factors (e.g. Hinde & Kilner 2007; Szentirmai, Székely, & Komdeur 2007), comparative analyses offer a complementary and powerful means to explore the causes and consequences of variation in parental care strategies across large numbers of species. A necessary first step to any such analysis is to quantify the relative reproductive investments of males and females, but it is not immediately clear how in a comparative context one should characterize the particular strategy observed in any one species.

One possibility is a simple linear relationship between male and female care provision, such that a reduction in care by males is compensated by a corresponding increase in female care. This means that the discrepancy between male parental effort and female parental effort, for example in terms of time spent on caring (e.g. Owens & Hartley 1998; Liker & Székely 2005; Olson et al. 2008), can be used as an index of care disparity. Such an index might be used, for example to suggest a continuum between low levels of parental conflict (i.e. sexual conflict between parents over care) in species with an equitable division of care to high levels of parental conflict in species with uniparental care (Székely et al. 2007a).

There are good reasons to believe that this simple index of care provision may not account for all facets of parental behaviour. The discrepancy between male and female effort neglects the total duration of the care period (e.g. Sunde 2008). A highly asymmetric division of care between the sexes is likely to have very different ecological and evolutionary implications in a species in which the care period is very brief (such that multiple broods are possible within a single season) compared with a species in which care extends over an entire breeding season. Improving our understanding of care division also has ramifications for our interpretation of the relationship between parental care and sexual selection. Traditionally, this relationship has been assumed to be unidirectional: following Trivers (1972) sexual selection is seen as an emergent property of differential parental care (defined by Trivers as parental investment) by the male and the female. However, because care (and the decision to desert the offspring) are partly driven by mating opportunity and sexual selection (Székely et al. 1996; Alonzo & Warner 2000; Székely, Webb, & Cuthill 2000), there are in fact feedbacks between parental care, sexual selection and mating opportunities (Queller 1997; Kokko & Jennions 2008; Alonzo 2010) which appear to explain breeding system evolution in shorebirds and cichlid fishes (Thomas & Székely 2005; Gonzalez-Voyer, Fitzpatrick, & Kolm 2008). It is likely that both the division of parental effort and the duration of care will play a key role in these feedback relationships, and so a comprehensive picture of parental care requires that we consider simultaneously both the inter-sexual difference in care allocation and the total amount of care provided.

Here, we consider the problem of how to characterize in a comparative context the range of parental care strategies observed within a group of related species across these two axes of care provision. We focus on birds given they have some of the best-studied parental care systems, although the concept should be applicable to other organisms. Figure 1 outlines a simple theoretical framework and three broad scenarios.

image

Figure 1.  Three possible relationships between total care duration and proportional male care. (a) Males provide a constant proportion of care regardless of care duration, leading to a slope of 0 (with intercepts between 0 and 1 indicating the extent of male involvement). (b) Males provide little care when total care duration is short, but care is divided more equitably at longer total care durations. (c) A full range of potential strategies in rapidly developing species is constrained to a strategy of biparental care when total care duration is extended. We highlight in identical positions on each plot two hypothetical species. In the species represented by the solid symbol, the female provides most care, whereas in the species represented by the open symbol both sexes provide equal care. Assessing the importance of these discrepancies is possible only in the context of the underlying relationship between total care and proportional male care observed within the group of interest. The bold arrow on the left of each panel indicates the total range of variation in proportional male care; the arrow on the right of the panel indicates the range in residual variation around the relationship between proportional male care and care duration. In (a), these two measures are identical; in (b) residual variation is substantially less than total variation; and in (c), residual variation equals total variation but varies systematically with total care duration.

Download figure to PowerPoint

According to the first scenario, males provide a constant share of care regardless of the total care required (Fig. 1a). This may occur if an equal division of care is always observed, or if the parents divide care provisioning in a specific way (‘multi-dimensional care’, Houston, Székely, & McNamara (2005)) and the relative durations of the different tasks remain constant. The degree of scatter around this relationship will indicate the extent to which care strategies are constrained; in this case, the total variation in proportional male care alone provides a good indication of the variation in care strategies observed within the group.

Secondly, it may be that early desertion or low parental effort is only observed in one sex (males in Fig. 1b), and then only at low care durations: as the period of care required by offspring increases, so the opportunities for that sex to reduce its relative effort decrease. The result is a more equitable division of care as total care duration increases. Here, the total variation in proportional male care provides a misleading picture of the observed range of care strategies. Rather, potential care strategies are constrained by the total duration of care, and the strength of this constraint is better measured as the residual variation around the relationship between the two measures of care.

Thirdly (Fig. 1c), proportional male care may be highly variable when the offspring need little care, allowing either parent to compensate for desertion by its mate by providing for the young largely or exclusively on its own. As the demands of offspring grow through longer periods of dependency, it becomes harder for a single parent to succeed and so this variation reduces with proportional male care converging to a (non-zero) constant. Here, neither total nor residual variation in proportional male care fully reveals the observed constraints on care division. Rather, it is the trend in residual variation with total care duration that is of interest.

These three scenarios illustrate what we consider to be the most plausible relationships between proportional male care and total care duration, with the division of parental effort either entirely constrained within a taxon (Fig. 1a), or else the opportunities for one or other parent to desert are reduced when the total care period is long. Of course, many variations on these themes possible, including the converse of Fig. 1b (proportional male care declining from >50% as care duration increases) or Fig. 1c (more variation at long care durations, perhaps if offspring have a long period of post-fledging dependence on only a single parent), as well as nonlinear (threshold-type) relationships. All would be simple to represent under the same framework, by deriving expectations in terms of the slope and pattern of variation for a given scenario.

In this article, we propose a comparative approach to test these three models of care distribution using a new database of parental care provision across 330 bird species drawn from 13 families with contrasting life-histories and ecologies. We use the scenarios illustrated in Fig. 1 to classify the relationship between proportional male care and care duration both within and across families. Our results suggest that interpretation of variation in parental roles requires consideration of other factors, particularly the duration of the period of care, and we discuss how to reconcile measures of parental care discrepancy based solely on parental roles (Székely et al. 2007a; Olson et al. 2008) with constraints imposed by the demands of offspring.

We illustrate the use of this method in a comparative context by considering the influence of sexual size dimorphism (SSD) on the division of parental effort, and that of body size on total care duration. In birds, SSD is a strong correlate of the intensity of sexual selection (e.g. Andersson 1994; Bennett & Owens 2002; Thomas, Székely, & Reynolds 2007), although not always in a straightforward way (Székely, Freckleton, & Reynolds 2004), and body size correlates with a range of other life-history traits (e.g. clutch size, egg mass), which may influence the duration of parental care (Bennett & Owens 2002). For the 13 families in our care dataset, then, we compare the results of simple models linking SSD to proportional male care, to those obtained from models in which the total care duration is also taken into account, and of models linking total care duration to body size, both with and without proportional male care included as a covariate. These analyses suggest that using both dimensions of care is likely to improve the identification of robust correlates of parental care strategies.

Materials and methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

Components of care: data and scores

Testing the scenarios outlined in Fig. 1 requires quantitative data on the duration of care provided by males and females across multiple species in a range of taxa. Birds provide the ideal group to illustrate our approach, as they display the full range of parental care strategies, including male-only, female-only and biparental care. Moreover across birds, the total duration of the care period varies by at least a factor of 50 (Bennett & Owens 2002). We collected extensive data on parental care provision by males and females across 330 species of birds, with between 6 and 66 species from each of 13 families representing four orders (following the classification of Sibley & Ahlquist 1990). Data were extracted from standard texts covering the Palaearctic (BWPi 2006), North America (Poole & Gill 2002) and Southern Africa (Hockey et al. 2005). We divided the period of parental care into the three stages typically described in such texts: incubation, chick rearing and post-fledging care. The total duration (d) of each of these stages was recorded. The proportional contribution to each stage by males was determined by scoring male and female contributions to each of a series of activities within each stage (Table 1). Contributions to each activity were scored quantitatively where possible (e.g. number of feeds per hour by males and by females), otherwise proportional contributions were considered (e.g. ‘males provide 65% of food’). Where only qualitative information was available, we used a binary score (0 or 1) for a given activity (e.g. ‘only males feed the chicks’ scores 1 for males, 0 for females; ‘both parents feed chicks’ scores 1 for males and 1 for females). The aim was to generate as full a picture as possible of the relative contributions of males and females across the three critical stages of care provision. We paired male and female scores for as many activities as possible within each of the three broad stages to calculate proportional male care, P(male) as

  • image
Table 1.   Summary of the data collected on parental care
Care stageActivitySubdivision
  1. Data were collected at the finest scale possible for each stage (i.e. subdivision if possible, otherwise activity, otherwise an overall score for the stage), and quantitative data (e.g. percent activity hours, food items delivered) were preferred over relative data (e.g. ‘male provides 65% of food’), which in turn were preferred over simple binary scores (e.g. ‘male provides all food’, ‘only female broods’).

IncubationIncubationBy day
By night
Chick rearingBroodingBy day
By night
Feeding 
Guarding 
Post-fledgingFeeding 
Guarding 

We then combined the three scores into a single value of proportional male care. This was simply the mean of the three stage scores, weighted by the duration of each stage, and could take any value between 0 (female-only care) and 1 (male-only care). Note that alternative weighting schemes could be adopted to examine in more detail particular key stages of the chick rearing period, or to emphasize certain activities (e.g. food provision) over others (e.g. brooding).

Analysis

All analyses were conducted within families. The family is a pragmatic choice for analysis where complete species-level phylogenies are not available, as species within a family are biologically and ecologically rather similar, yet families contain both adequate numbers of species and sufficient variation in traits, such as parental care strategy to enable meaningful analyses. We use the families defined by Sibley & Ahlquist (1990) for consistency with previous analyses, and because the Sibley & Ahlquist phylogeny uses objective criteria for defining higher taxonomic levels (i.e. family is defined on the basis of specific DNA hybridization distance values; Sibley & Ahlquist 1990). Phylogenetic relationships within families may of course be important, and we discuss this issue further below.

For each family, we modelled the relationship between proportional male care and the log of total care duration. Any significant difference from a slope of 0 indicates that proportional male care varies systematically with the total duration of care. A slope of 0 could indicate either that males always provide a fixed proportion of care, or that proportional male care varies but not predictably with total care duration. To distinguish these two scenarios, we define total male effort and total female effort as follows. In species with an equal division of care, both total male and total female effort are equal to total care duration (in our dataset there is no species in which parents exchange care duties midway through the care period, e.g. situations, such as a female deserting and a male appearing after incubation were not observed). In species in which one sex provided all of the care, the total effort of this sex is equal to the total care duration, and that provided by the other sex is 0. In species in which females provided >50% and <100% care, female effort is equal to total duration and

  • image

Finally, where males provided >50% and <100% care, male effort is total duration and

  • image

Thus, male and female effort sum to give the total number of ‘parent days’ of care provided, which ranges between total care duration and 2 × total care duration. We then tested the correlation between male and female effort [both log(+ 1)-transformed], with a significant positive correlation indicative of males providing a constant proportion of care regardless of total care duration.

To account for phylogenetic relationships between species within families, we adopt two methods. The first method is appropriate for taxonomic groups for which a fully-resolved species-level phylogeny is available. For the three shorebird families in our dataset (Charadriidae, Laridae, Scolopacidae), which together comprise 139 species, we used the phylogeny presented by Thomas, Wills, & Székely (2004) and employed the method of Pagel (1999) to fit phylogenetically weighted Generalized Least Squares models (PGLMs; see, Freckleton, Harvey, & Pagel 2002 for details). The second method is appropriate for families for which a species-level phylogeny is lacking. For these families, we fitted linear mixed effects models (Gelman & Hill 2007) of proportional male care on log(total care duration) with genus included as a random factor. This implementation of the nested taxonomic model (Clutton-Brock & Harvey 1977) fits a common slope across all species within a family, but allows the intercept to vary between genera, and is fully consistent with phylogenetic mixed models (Hadfield & Nakagawa 2010).

As a simple measure of systematic changes in the variance in care strategies with total care duration, we use the correlation between total care duration and the absolute (unsigned) residuals from the (PGLM or linear mixed effects) model of proportional male care on log(total care duration). This is a standard diagnostic used for identifying systematic trends in residuals in linear models (Faraway 2004), and does not constitute misusing residuals as data (sensuFreckleton 2002). This measure allows us to analyse whether there is more (or less) flexibility in care strategies of species with short (or long) development times.

Finally, to measure the scatter of the relationship in each family (i.e. the degree to which individual species tend to vary from the family-level strategy), we used the standard deviation of the residuals from the models described above. We compare this with the 95% CI of the distribution of proportional male care values for each family.

We also performed a cross-species analysis to assess the relationship between proportional male care and total care duration across all species in our data set. We used linear regression to characterize this relationship, and the correlation between total care duration and the absolute (unsigned) residuals from the regression to look for broad-scale evidence across birds for care duration constraining the range of care strategies possible. We further analysed this constraint using quantile regression (Koenker 2008) to estimate slopes centred at the 5th and 95th percentiles of the data. These analyses across all species are intended as a general qualitative illustration of the pattern observed across birds, and we thus make no attempt to account for the pronounced phylogenetic structure in these data, employing Ordinary Least Squares regression.

To illustrate the use of our method in a comparative context, we combined with our care data base on SSD using data on adult male and female mass (g) taken from Székely, Lislevand, & Figuerola (2007b). SSD was estimated as log(male mass/female mass), such that positive values indicate male-biased SSD and negative values female-biased SSD. SSD estimates were available for 247 of the 330 species (75%) used in the care analyses. For each family with sufficient SSD data, we then compared the results of simple models of proportional male care on SSD, with the estimated effect of SSD obtained by modelling proportional male care as a function of both log(total care duration) and SSD. We performed similar analyses modelling log(total care duration) as a function of adult female mass (g, log-transformed), first as a univariate model and then including proportional male care as a covariate. All of these models took account of the phylogenetic relationships between species within families either by employing PGLMs (for the shorebird families) or linear mixed effects models with genus included as a random factor (all other families).

All data manipulation and statistical analysis were carried out using R 2·9·2 (R Development Core Team 2008), with PGLMs implemented using the package CAIC (Orme, Freckleton, & Thomas 2008) and linear mixed effects models implemented using the package lme4 (Bates & Maechler 2009).

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

Parental roles

Across all families, mean proportional male care was 0·40 (95% CI: 0·34–0·47), significantly less than 0·5 (t = 3·16, d.f. = 12, = 0·0083). Within families, males provided <50% of care on average in all families except the Charadriidae, Laridae, Podicipedidae, Rallidae and Scolopacidae (Table 2). 95% Confidence intervals of the proportional male care values in each family, however, reveal substantial differences between families in the range of behaviours observed, from the Podicipedidae where males always provided 50% of the care, to the Scolopacidae where values were between 0% and 100% (Table 2).

Table 2.   Summary of the avian family-level results
FamilyNumber of speciesMean P(Male) (95% CI)P(Male) on log(total care) slope (95% CI)Correlation [|residuals|, log(total care)]SD (residuals)Correlation between female and male efforts
  1. For each family, we give the number of species for which we have data, the mean proportion of care provided by males and its 95% CI. The slope (and 95% CI) of relationships between proportional male care and total care duration (log-transformed) is shown, with slopes differing significantly from 0 highlighted in bold. The correlation between the absolute values of the residuals from this model and total care duration indicates trends in variance in care strategy with total effort. The standard deviation of these residuals is a simple index of variability in proportional male care within a family around the underlying pattern. These models are all linear mixed effects models, with genus included as a covariate, except for the Charadriidae, Laridae and Scolopacidae, where results from phylogenetically weighted generalized least squares models are shown (see text for details). Finally, we give the correlation of the relationship between male and female effort.

  2. aStatistics for the Rallidae exclude the striped crake Aenigmatolimnas marginalis, which is the only species in a monotypic genus and is a conspicuous outlier in this family with the male providing all care (Fig. 2). Including it as an 18th species in this family has little effect on the mean P(Male) (0·53, 95% CI: 0·45–0·82) but it removes significance from the relationship between male and female effort (= 0·32). Because the model of P(Male) on total care duration includes genus as a random factor, including this species does not significantly affect the slope (−0·01, 95% CI: −0·09 to 0·08), the correlation of the residuals (= −0·20) or the standard deviation of residuals (0·02). +< 0·1, *< 0·05, ***< 0·0001.

Accipitridae570·37 (0·22–0·5)0·10 (0·06–0·14)−0·070·050·88***
Charadriidae310·52 (0·44–0·75)−0·01 (−0·06 to 0·05)−0·060·100·73***
Corvidae100·44 (0·36–0·5)0·06 (0·03–0·09)0·170·030·99***
Fringillidae270·33 (0·21–0·46)0·10 (0·07–0·13)−0·030·020·97***
Hirundinidae160·37 (0·20–0·53)0·00 (−0·36 to 0·36)−0·51+0·090·56*
Laridae660·50 (0·42–0·73)0·01 (−0·02 to 0·03)−0·28*0·060·92***
Muscicapidae240·38 (0·26–0·49)0·09 (0·05–0·12)−0·100·030·98***
Paridae120·32 (0·25–0·39)0·14 (0·07–0·20)0·050·020·91***
Phasianidae150·15 (0–0·46)0·11 (−0·05 to 0·27)0·340·110·50
Podicipedidae70·5 (–)0 (–)01***
Rallidaea170·50 (0·45–0·57)−0·01 (−0·07 to 0·04)−0·180·030·95***
Scolopacidae420·50 (0–1)−0·47 (−0·97 to 0·04)−0·180·27−0·15
Sittidae60·33 (0·25–0·45)0·14 (0·11–0·17)−0·470·020·99***

Relationship between proportional male care and care duration

The proportion of care provided by males significantly increases with total care duration in 6 of 13 families (Accipitridae, Corvidae, Fringillidae, Muscicapidae, Paridae and Sittidae) with a trend in the same direction in the Phasianidae (Table 2, Fig. 2). There was evidence for a significant reduction in the variability of male care within increasing care duration in the Laridae (= 0·0207), and a similar trend in the Hirundinidae (= 0·0539; Table 2). This correlation was negative in 9 of 12 families in which proportional male care varied, suggesting that a longer duration of care may indeed constrain possible care strategies within a family (Table 2, Fig. 2).

image

Figure 2.  Proportional male care against total care duration (natural log-transformed) for each family. The same range of total care durations is illustrated on each plot, but for clarity the scale on the proportional male care axis changes (although the interval remains the same). The dashed line indicates equal care, and a fitted linear model is shown as a solid line if the relationship is significant, a dotted line otherwise (note that these are ordinary least squares linear models, to illustrate broad trends; all analyses used either linear mixed effects models or phylogenetically weighted generalized least squares to account for relationships between species within families). We do not illustrate the relationship for the grebes (Podicipedidae), where total care duration ranges from 46 to 119·5 days (3·8 to 4·8 on the log scale) but in which males always provide 50% of care.

Download figure to PowerPoint

The standard deviations of residuals from these family-specific models ranged from 0 in the Podicipedidae to 0·27 in the Scolopacidae (Table 2), largely reflecting the overall levels of variation in proportional male care present in each family (correlation between 95% CI in proportional male care and SD in residuals = 0·95, d.f. = 11, < 5 × 10−5).

Relationship between male and female care

Male and female care duration are significantly positively correlated across 10 of 13 families (Table 2); of the exceptions (the Phasianidae, Rallidae and Scolopacidae), the relationship approaches significance in the Phasianidae (= 0·0557) and is highly significant in the Rallidae if the conspicuous outlier, the striped crake Aenigmatolimnas marginalis in which the male provides all care, is excluded (Table 2). Thus, in the seven families in which there is no significant relationship between total care duration and proportional male care, this generally arises because male effort is a constant proportion of total effort, rather than due to variation in the division of care at random with respect to total care duration.

The role of phylogeny

Across all 330 species, and using the family-level phylogeny of Sibley & Ahlquist (1990) with branch lengths within families scaled to include information on genus, we find that closely related species tend to display similar parental strategies, both in terms of total care duration and proportional male care. The statistic λ introduced by Pagel (1999) with tests outlined in Freckleton, Harvey, & Pagel (2002) is indistinguishable from one across our data set, indicating concordance with a constant variance Brownian model of evolution.

For the three shorebird families for which we had a species-level phylogeny, all PGLMs of proportional male care on log(total care) resulted in an optimized value of λ significantly greater than 0, with λ = 1 for Charadriidae and Laridae and λ = 0·69 in Scolopacidae. This suggests that a considerable degree of phylogenetic signal in parental care remains present within families. For instance, in the Laridae distinct subgroups within the family show relationships which are not apparent across the entire family (Fig. 3).

image

Figure 3.  Proportional male care against log(total care duration) for three groups within the Laridae, Auks (15 species of the genera Aethia, Alca, Alle, Brachyramphus, Cepphus, Cyclorrhynchis, Fratercula, Ptychoramphus, Synthliboramphus and Uria), Gulls (20 species of the genera Larus, Rissa and Xema) and Terns (25 species of the genera Anous, Chlidonias, Gygis and Sterna). The dashed line represents equal care. For clarity, we omit six further species of Laridae for which we have care data (four species of Skua and two species of Skimmer), although these are included in all analyses. Correlations between proportional male care and log(total care) are positive in the auks (= 0·77, d.f. = 13, = 0·0009) and the terns (= 0·35, d.f. = 23, = 0·091) and negative in the gulls (r = −0·75, d.f. = 18, = 0·0001), and there are significant differences in the variances of proportional male care between the three groups (Bartlett’s test, K2 = 58·3, d.f. = 2, < 0·0001), with standard deviations of 0·11 in the auks, 0·01 in the gulls and 0·04 in the terns.

Download figure to PowerPoint

Patterns of variation in proportional male care across all birds

Across all 330 species, there is no significant relationship between proportional male care and log(total care duration) (Fig. 4; slope = 0·02, 95% CI: −0·01 to 0·04). However, the residual variance in proportional male care does decline significantly with increasing care duration (Fig. 4; correlation between unsigned residuals from the model and log(total care duration), = −0·33, d.f. = 328, < 0·00001; correlation of standard deviations within 10% total care duration quantiles and the midpoint of the quantiles, = −0·69, d.f. = 8, = 0·271). This results from a reduction in frequency of both predominantly male and predominantly female care, reflected in a significantly negative slope centred on the 95th percentile of the data (slope = −0·09, 95% CI; −0·13 to −0·08) and a significantly positive 5th percentile slope (slope = 0·16, 95% CI: 0·06–0·18; Fig. 4).

image

Figure 4.  Proportional male care against log(total care duration) for 330 bird species. Individual species are shown in grey, and means ± SD across species within each 10% quantile of log(total care duration) are shown as solid symbols. The dashed horizontal line is at 0·5 (i.e. equal male and female care) and the fine solid line is the fitted linear regression, which has a slope not significantly different from 0. However, variance around the relationship declines significantly with increasing care duration: the bold solid lines are from a quantile regression analysis, and show a significant negative trend for the 95th percentile and a significant positive trend for the fifth percentile.

Download figure to PowerPoint

Relationships between parental care, SSD and body size

The coefficients associating proportional male care with SSD were negative in 7 of 11 families for which sufficient SSD data were available, and significantly so for the Hirundinidae and Scolopacidae (Table 3). Including log(total care duration) as a covariate finds similar patterns: 8 of 11 SSD coefficients are negative, the two significant relationships from the simple models remain significant, but they are joined by Fringillidae, in which the significant negative relationship between proportional male care and SSD is only apparent after controlling for variation in total care duration.

Table 3.   The relationship between proportional male care and SSD, or between log(total care duration) and log(female mass, g) within each family in our database, given as both the slope (and 95% CI) of a simple univariate model of the care variable on the size variable and as the partial regression coefficient linking the care and size variables in a model which also included the other care variable as a covariate
FamilyNSSD coefficientLog(female size) coefficient
Univariate regressionControlling for total care durationUnivariate regressionControlling for proportional male care
  1. N is the number of species in the family for which we had both care data and an estimate of sexual size dimorphism (SSD). We exclude Sittidae from these analyses because of a low number of species with data on SSD (= 4), and Podicipedidae due to no variation in proportional male care. Models are linear mixed effects models with genus as a covariate except in the Charadriidae, Laridae and Scolopacidae, where phylogenetically weighted generalized least squares models were used, and in the Phasianidae where ordinary least squares is used, as only one genus is represented by more than one species. Coefficients which differ significantly from 0 are highlighted in bold.

Accipitridae350·11 (−0·07 to 0·30)0·03 (−0·15 to 0·22)0·24 (0·15–0·34)0·22 (0·13–0·31)
Charadriidae23−0·56 (−1·16 to 0·04)−0·55 (−1·16 to 0·06)0·29 (−0·07 to 0·64)0·27 (−0·10 to 0·63)
Corvidae90·32 (−0·05 to 0·69)−0·12 (−0·38 to 0·15)0·33 (−0·02 to 0·68)0·11 (−0·17 to 0·38)
Fringillidae23−0·13 (−0·40 to 0·14)−0·17 (−0·33 to 0·02)0·09 (−0·46 to 0·64)0·10 (−0·15 to 0·34)
Hirundinidae13−1·47 (−2·65 to −0·28)−1·48 (−2·72 to −0·24)0·05 (−0·24 to 0·33)0·04 (−0·26 to 0·34)
Laridae430·12 (−0·05 to 0·29)0·10 (−0·08 to 0·29)0·31 (0·16–0·46)0·30 (0·14–0·45)
Muscicapidae14−0·15 (−0·74 to 0·45)−0·07 (−0·62 to 0·48)−0·07 (−0·38 to 0·25)0·02 (−0·30 to 0·35)
Paridae100·17 (−0·46 to 0·79)0·21 (−0·24 to 0·67)0·46 (0·18–0·74)0·42 (0·13–0·72)
Phasianidae10−0·01 (−0·01 to 0·00)−0·13 (−0·39 to 0·13)−0·17 (−0·59 to 0·25)−0·09 (−0·35 to 0·18)
Rallidae14−0·04 (−0·16 to 0·08)−0·04 (−0·18 to 0·10)0·24 (0·08–0·41)0·24 (0·06–0·43)
Scolopacidae42−0·93 (−1·34 to −0·51)−0·90 (−1·32 to −0·49)0·15 (0·09–0·22)0·18 (0·13–0·22)

In models of log(total care duration) on log(female body mass), relationships were positive in 9 of 11 families, and significantly so in the Accipitridae, Laridae, Phasianidae, Rallidae and Scolopacidae (Table 3). Controlling for proportional male care does not change these conclusions (Table 3), although some coefficients do change quite markedly (e.g. in the Corvidae and Phasianidae).

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

Across the birds in our dataset, males typically provide less than 50% of parental care, supporting the view that male-desertion is a common form of offspring desertion in birds (Székely et al. 1996; Bennett & Owens 2002). In those families where there is no relationship between proportional male care and total care duration, male and female efforts were generally highly significantly positively correlated. This suggests that male effort is not entirely random with respect to care duration; rather, in these families males typically provide a constant proportion of the total parental effort regardless of care duration. Variation in male care does differ substantially between families, however. For instance, in the grebes (Podicipedidae) no variation is observed and care is always equally divided between the parents; whereas in the sandpipers and allies (Scolopacidae), the full range of strategies (male-only care, female-only care and all points in between) is observed (Fig. 2).

In 6 of the 13 families, however, there exists a strong positive relationship between proportional male care and care duration, suggesting a rather common trend for males to contribute relatively more care as the duration of the care period increases. For these families, residual variation around this relationship is substantially less than total variation in male care (Fig. 2), and variation in care strategy is therefore better quantified as the standard deviation of residuals from the fitted model, rather than as total variation in proportional male care. In other words, the right-hand arrow on Fig. 1a and b better reflects the realized variation in male parental care than does the left-hand arrow.

Within families, then, we find strong support for two of the scenarios outlined in the ‘Introduction’ (Fig. 1a,b), but no evidence for a reduction in variance in care strategies with increasing total duration (Fig. 1c). In other words, at the family level, there is no evidence for a situation in which early desertion by either sex is possible only in more rapidly developing species. This is somewhat surprising, given that previous work on shorebirds has shown that opportunities for one parent to desert are greater when young are less demanding (Thomas & Székely 2005). However, if we consider a broader taxonomic scope than our within-family analyses, such a reduction is clearly apparent (Fig. 4). Thus, it appears that the variety of care strategies possible within a family may be constrained by the range of development times observed in the family. Typically, a greater variety of strategies is possible in families in which development is more rapid, consistent with the more detailed analysis of shorebirds (Thomas & Székely 2005). These general patterns of development time across families are likely to be determined by various environmental, ecological and life-history factors including climate, body size and the intensity of sexual selection (Ricklefs 1968, 1976, 1979; Redfern 1989; Eppley 1996; Queller 1997; Bennett & Owens 2002; Gonzalez-Voyer, Fitzpatrick, & Kolm 2008; Kokko & Jennions 2008).

While finer-scale data may reveal negative relationships within species (i.e. one parent compensates for lack of effort from the other; Hinde & Kilner 2007; Szentirmai, Székely, & Komdeur 2007), at this broad comparative scale, we can state therefore that as the duration of female care increases, so too does the duration of male care, either as a constant or an increasing proportion of the total care provided. Thus, across a diverse range of avian families the duration of the care period, which may be determined by developmental constraints (Temrin & Sillén-Tullberg 1995; Starck & Ricklefs 1998) and modified through parent–offspring conflict (e.g. Sunde 2008), itself represents a constraint on the range of care strategies that can be adopted or on the absolute amount of care provided by each parent.

Our approach suggests several routes to understanding the distribution of different parental care strategies in a comparative context. At a family level, residual variation in proportional male care could be used to identify those families with a particularly large or small range of observed strategies. For instance, although swallows (Hirundinidae) and nuthatches (Sittidae) display a similar total range in proportional male care (0·17–0·55 and 0·24–0·46, respectively), the observed strategy in nuthatches appears to be much more constrained by total care duration (SD of residual variation: 0·02) than does that in swallows (0·09). Seeking robust behavioural, developmental or ecological correlates of such family-level differences would be a natural extension of this approach. In addition, given the existence of marked differences across families in the relationship between proportional male care and care duration, we emphasize that it is especially critical to account for these differences in any comparative analysis of parental care including species drawn from multiple families. For instance, fixed-slope, random-intercept mixed models would not be appropriate for analyses conducted across families where slopes clearly vary substantially (Fig. 2, Table 2).

Within families, the parental strategy of a focal species is also best understood in terms of its residual distance from the regression of proportional male care on log(total care duration) across its relatives. Thus, the hypothetical species in Fig. 1 would be characterized by their divergence from the solid lines, rather than from the dashed lines which assume no relationship between relative parental effort and care duration. For the purposes of comparative analyses, care duration can simply be entered as a covariate in models with proportional male care as the response, as in our analyses of SSD. This approach recognizes that differences in care strategy can only be sensibly interpreted in an interspecific comparative context. For example, equal male and female care would be interpreted differently in a species whose close relatives all display male-only care, compared with one across whose relatives female-only care is the norm. Implementing this approach requires quantitative data on care duration, and information on male and female contributions to various stages of offspring rearing. Fortunately, such data are readily obtainable for large numbers of bird species from standard texts (e.g. Poole & Gill 2002; Hockey et al. 2005; BWPi 2006). An interesting development would be to break down the single score of proportional male care that we use, to consider how the male contribution to different stages of offspring development (e.g. incubation vs. feeding of chicks) varies with care duration. Placing different weights on different activities (e.g. food provision vs. nest-based care) may also be a useful approach in a more detailed treatment of specific families in which both parents care, but have distinct roles.

A logical next step would be to investigate the environmental, life history and phylogenetic correlates of the patterns of variation that we have documented. We have shown how it is simple to include life-history covariates into an analysis, using the examples of female body size and SSD. Although including total care duration in models associating proportional male care with SSD did not produce striking differences in SSD coefficient estimates compared with simple models of proportional male care on SSD (Table 3), it does facilitate interpretation. The general negative relationship between proportional male care and SSD suggests that males typically contribute less care in species with high male-biased SSD, but it is useful to know that this is independent of the duration of the care period. In addition, in one family (Fringillidae), it is only when total care duration is taken into account that the negative relationship between proportional male care and SSD becomes apparent. Similarly, relationships between care duration and female size are generally not markedly affected by the inclusion of proportional male care as a covariate, but coefficients do vary considerably in some families. Including both dimensions of care allows us better to identify those families in which a real effect may be present, and to interpret such an effect independently from any correlation between body size and care duration. In general, then, the bivariate description of the division of care clarifies relationships found using a univariate approach, as well as minimizing the risks of either missing important relationships, or concluding that a relationship is important when in fact it is not.

We have also shown that substantial phylogenetic structure in care strategies can persist within families (e.g. Fig. 3). It seems likely that phylogenetic constraints operate at higher taxonomic levels too. For instance, we note that significant positive relationships between proportional male care and total care duration are observed in five of the six passerine families in our data set, but in only one of the seven non-passerine families. It would be interesting to examine patterns of life-history variation which may help to explain this. For example, the relatively high prevalence of precocial development in several of the non-passerine families may present opportunities for variable parental care strategies to evolve (e.g. early desertion of clutches by either sex; Temrin & Sillén-Tullberg 1995; Thomas & Székely 2005; Olson et al. 2008).

Conclusions

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

Quantifying relative male and female contributions to parental care has important implications for the general understanding of breeding systems (Reynolds 1996; Bennett & Owens 2002), and especially for the investigation of cooperation and conflict between the sexes over the rearing of offspring (Houston, Székely, & McNamara 2005). Previous comparative analyses have typically only considered discrepancy in care on a single dimension (e.g. differences in the proportion of care provided by males and females). We have shown that the discrepancy in care is not independent of the duration of care across diverse avian families, and that this can have consequences for the conclusions drawn from comparative analyses. The simple method we introduce allows variation in both of these dimensions to be characterized, providing a solid basis for extensive future comparative analyses aiming to uncover the ultimate biological and environmental causes of breeding system variation in birds.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

This research was funded by NERC (NE/C004167/1). T.J.W. and R.P.F. are Royal Society University Research Fellows.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  • Alonzo, S.H. (2010) Social and coevolutionary feedbacks between mating and parental investment. Trends in Ecology & Evolution, 25, 99108.
  • Alonzo, S.H. & Warner, R.R. (2000) Dynamic games and field experiments examining intra- and inter-sexual conflict: explaining counter-intuitive mating behavior in a Mediterranean wrasse, Symphodus ocellatus. Behavioural Ecology, 11, 5670.
  • Andersson, M. (1994) Sexual Selection. Princeton University Press, Princeton.
  • Bates, D. & Maechler, M. (2009) lme4: linear mixed-effects models using S4 classes. R package version 0.999375-31. Available at: http://CRAN.R-project.org/package=lme4
  • Bennett, P.M. & Owens, I.P.F. (2002) Evolutionary Ecology of Birds. Oxford University Press, Oxford.
  • BWPi (2006) Birds of the Western Palearctic 1.1. Available at: http://www.birdguides.com. Oxford University Press, Oxford.
  • Clutton-Brock, T.H. (1991) The Evolution of Parental Care. Princeton University Press, Princeton, NJ.
  • Clutton-Brock, T.H. & Harvey, P.H. (1977) Primate ecology and social-organization. Journal of Zoology, 183, 139.
  • Eppley, Z.A. (1996) Charadriiform birds in Antarctica: behavioral, morphological, and physiological adjustments conserving reproductive success. Physiological Zoology, 69, 15021554.
  • Faraway, J. (2004) Linear Models With R. Chapman & Hall/CRC, Boca Raton, FL.
  • Freckleton, R.P. (2002) On the misuse of residuals in ecology: regression of residuals. multiple regression. Journal of Animal Ecology, 71, 542545.
  • Freckleton, R.P., Harvey, P.H. & Pagel, M. (2002) Phylogenetic analysis and comparative data: a test and review of evidence. American Naturalist, 160, 712726.
  • Gelman, A. & Hill, J. (2007) Data Analysis Using Regression and Multilevel/Hierarchical Models. Cambridge University Press, Cambridge.
  • Gonzalez-Voyer, A., Fitzpatrick, J.L. & Kolm, N. (2008) Sexual selection determines parental care patterns in cichlid fishes. Evolution, 62–8, 20152026.
  • Hadfield, J.D. & Nakagawa, S. (2010) General quantitative genetic methods for comparative biology: phylogenies, taxonomies and multi-trait models for continuous and categorical characters. Journal of Evolutionary Biology, 23, 494508.
  • Hinde, C.A. & Kilner, R.M. (2007) Negotiations within the family over the supply of parental care. Proceedings of the Royal Society of London. Series B, 274, 5360.
  • Hockey, P.A.R., Dean, W.R.J. & Ryan, P.G. (eds). (2005) Roberts–Birds of Southern Africa. John Voelcker Bird Book Fund, Cape Town, RSA.
  • Houston, A.I., Székely, T. & McNamara, J.M. (2005) Conflict between parents over care. Trends in Ecology & Evolution, 20, 3338.
  • Koenker, R. (2008). quantreg: Quantile Regression. R Package Version 4.26. Available at: http://www.r-project.org.
  • Kokko, H. & Jennions, M.D. (2008) Parental investment, sexual selection and sex ratios. Journal of Evolutionary Biology, 21, 919948.
  • Lessells, C.M. (1999) Sexual conflict in animals. Levels of Selection in Evolution (ed. L.Keller), pp. 7599. Princeton University Press, Princeton, NJ.
  • Liker, A. & Székely, T. (2005) Mortality costs of sexual selection and parental care in natural populations of birds. Evolution, 59, 890897.
  • Olson, V., Liker, A., Freckleton, R.P. & Székely, T. (2008) Parental conflict in birds: comparative analysis of offspring development, ecology and mating opportunities. Proceedings of the Royal Society of London B, 275, 301307.
  • Orme, D., Freckleton, R. & Thomas, G. (2008). CAIC: Comparative Analyses Using Independent Contrasts. R package version 1.0.4-83.
  • Owens, I.P.F. & Hartley, I.R. (1998) Sexual dimorphism in birds: why are there so many different forms of dimorphism? Proceedings of the Royal Society of London B, 265, 397407.
  • Pagel, M. (1999) Inferring the historical patterns of biological evolution. Nature, 401, 877884.
  • Parker, G.A., Royle, N.J. & Hartley, I.R. (2002) Intrafamilial conflict and parental investment: a synthesis. Philosophical Transactions of the Royal Society of London B, 357, 295307.
  • Poole, A. & Gill, F. (eds). (2002) The Birds of North America. The Academy of Natural Sciences/The American Ornithologists’ Union, Philadelphia/Washington, DC.
  • Queller, D.C. (1997) Why do females care more than males? Proceedings of the Royal Society of London. Series B, 264, 15551557.
  • R Development Core Team. (2008). R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. Available at: http://www.R-project.org.
  • Redfern, C.P.F. (1989) Rates of primary-feather growth in nestling birds. Ornis Scandinavica, 20, 5964.
  • Reynolds, J.D. (1996) Animal breeding systems. Trends in Ecology & Evolution, 11, 6872.
  • Ricklefs, R.E. (1968) Patterns of growth in birds. Ibis, 110, 419451.
  • Ricklefs, R.E. (1976) Growth rates of birds in the humid New World tropics. Ibis, 118, 179207.
  • Ricklefs, R.E. (1979) Patterns of growth in birds. V. A comparative study of development in the starling, common tern, and Japanese quail. The Auk, 96, 1030.
  • Sibley, C.G. & Ahlquist, J.E. (1990) Phylogeny and Classification of Birds: A Study in Molecular Evolution. Yale University Press, New Haven, CT.
  • Smiseth, P.T. & Moore, A.J. (2004) Behavioral dynamics between caring males and females in a beetle with facultative biparental care. Behavioural Ecology, 15, 621628.
  • Starck, J.M. & Ricklefs, R.E. (1998) Avian Growth and Development: Evolution Within the Altricial-Precocial Spectrum. Oxford University Press, Oxford.
  • Sunde, P. (2008) Parent-offspring conflict over duration of parental care and its consequences in tawny owls Strix aluco. Journal of Avian Biology, 29, 242246.
  • Székely, T., Freckleton, R.P. & Reynolds, J.D. (2004) Sexual selection explains Rensch’s rule of size dimorphism in shorebirds. Proceedings of the National Academy of Sciences, USA, 101, 1222412227.
  • Székely, T., Lislevand, T. & Figuerola, J. (2007b) Sexual size dimorphism in birds. Sex, Size and Gender Roles (eds D.J.Fairbairn, W.U.Blanckenhorn & T.Székely), pp. 2737. Oxford University Press, New York.
  • Székely, T., Webb, J.N. & Cuthill, I.C. (2000) Mating patterns, sexual selection and parental care: an integrative approach. Vertebrate Mating Systems (eds M.Apollonio, M.Festa-Bianchet & D.Mainardi), pp. 194223. World Science Press, London.
  • Székely, T., Webb, J.N., Houston, A.I. & McNamara, J.M. (1996) An evolutionary approach to offspring desertion in birds. Current Ornithology (eds V.NolanJr & E.D.Ketterson) Vol. 13, Chapter 6, pp. 271330. Plenum Publisher, New York.
  • Székely, T., Kosztolányi, A., Küpper, C. & Thomas, G.H. (2007a) Sexual conflict over parental care: a case study of shorebirds. Journal of Ornithology, 148 (Suppl. 2), S211S217.
  • Szentirmai, I., Székely, T. & Komdeur, J. (2007) Sexual conflict over care: antagonistic effects of clutch desertion on reproductive success of male and female penduline tits. Journal of Evolutionary Biology, 20, 17391744.
  • Temrin, H. & Sillén-Tullberg, B. (1995) A phylogenetic analysis of the evolution of avian mating systems in relation to altricial and precocial young. Behavioural Ecology, 6, 296307.
  • Thomas, G.H. & Székely, T. (2005) Evolutionary pathways in shorebird breeding systems: sexual conflict, parental care, and chick development. Evolution, 59, 22222230.
  • Thomas, G.H., Székely, T. & Reynolds, J.D. (2007) Sexual conflict and the evolution of breeding systems in shorebirds. Advances in the Study of Behavior, 37, 277340.
  • Thomas, G.H., Wills, M.A. & Székely, T. (2004) A supertree approach to shorebird phylogeny. BMC Evolutionary Biology, 4, 28. doi:10.1186/1471-2148-4-28.
  • Trivers, R.L. (1972) Parental investment and sexual selection. Sexual Selection and the Descent of man (ed. B.Campbell), pp. 136179. Aldine, Chicago, IL.