The evolution of flexible parenting

See allHide authors and affiliations

Science  15 Aug 2014:
Vol. 345, Issue 6198, pp. 776-781
DOI: 10.1126/science.1253294


Parenting behaviors, such as the provisioning of food by parents to offspring, are known to be highly responsive to changes in environment. However, we currently know little about how such flexibility affects the ways in which parenting is adapted and evolves in response to environmental variation. This is because few studies quantify how individuals vary in their response to changing environments, especially social environments created by other individuals with which parents interact. Social environmental factors differ from nonsocial factors, such as food availability, because parents and offspring both contribute and respond to the social environment they experience. This interdependence leads to the coevolution of flexible behaviors involved in parenting, which could, paradoxically, constrain the ability of individuals to rapidly adapt to changes in their nonsocial environment.

Parents often have to deal with multiple, competing demands simultaneously, such as feeding and defending offspring. The costs and benefits of parental decisions at any given moment in time will be sensitive to a range of environmental factors. These include not only climatic (e.g., temperature and rainfall) and ecological (e.g., predation pressure, pathogen load, and food availability) factors, but also social factors, in the form of partner and/or offspring behavior. When individuals modify their behavior in response to environmental factors, this is known as plasticity of behavior (14). If plasticity improves the fitness of individuals, then it is adaptive (57). To understand how parenting evolves, it is essential to determine when and how individuals vary in their plastic responses to environments (8). However, individual variation in plasticity of parenting behaviors has rarely been quantified (3, 8).

Here, we identify the gaps in our understanding and provide directions for future research. We will (i) outline the diversity of parenting behavior in animals, (ii) review the current evidence for adaptive plasticity, (iii) demonstrate the need for individual-based studies, and (iv) illustrate how evolving social environments can help us understand how plasticity contributes to adaptive evolution of parenting behaviors. To do this, we combine approaches from behavioral ecology and quantitative genetics. Our intention is to highlight how incorporating individual-level variation in parenting response to environments, particularly social environments, is essential for understanding the evolution of parenting. We demonstrate that parenting provides a particularly fertile context in which to investigate the role of plasticity in adaptation and evolutionary processes more generally.

The diversity of parental care in animals

The extent of parenting, defined here as behavioral interactions directed toward improving the growth or survival of offspring after birth or hatching, varies across the animal kingdom (Fig. 1). For example, feeding offspring (hereafter referred to as “provisioning”) occurs in only ~1% of insect species [all ants and some bees, wasps, termites, and beetles (9)] but is ubiquitous in mammals and nearly so in birds (10). Provisioning is not the only form of care, however. Parents also protect offspring from predators, and in many vertebrates they clean, carry, and provide warmth to offspring. Additionally, parental care occurs in ectothermic vertebrates: Various forms of parental care are found in fish (~30% of families), amphibians (e.g., 6 to 15% of anuran species, ~20% of salamander species), and reptiles (all crocodilians, ~1% of lizards, and 3% of snakes provide some form of care) (10). In ectothermic vertebrates, most forms of care are provided before offspring hatch or are born (e.g., egg guarding), but postnatal care also occurs (Fig. 1). Finally, care can be provided by the mother or father alone (uniparental care), both parents (biparental care), or parent(s) plus nonparents (cooperative care), with several species across different taxonomic groups showing more than one mode of care within a population [e.g., burying beetles (Nicrophorus vespilloides), acorn woodpecker (Melanerpes formicivorus), Galilee St. Peter’s fish (Sarotherodon galilaeus), and gray wolf (Canis lupus) (9, 10)].

Fig. 1 Parenting behavior is found in diverse forms across a broad taxonomic spread.

Parent(s) commonly have to decide when and how much to allocate to distinct forms of behavior directed at or for offspring, including: defending and protecting (A to D), escorting and/or carrying (E and F), and provisioning (G to I) young. Parenting can be provided by a single parent, either the male (E) or, more commonly, the female (A, D, F, G, I); pairs (B, C, H); or cooperative groups (C and D). Not uncommonly, there is variation in parenting mode in the same species (C). (A) Female common earwig (Forficula auricularia) guarding nymphs [photo by Jöel Meunier]. (B) Male Gambel’s quail (Callipepla gambelii) with chicks [photo © Alex Badyaev/]. (C) White’s skink (Egernia whitii) pair with offspring [photo by Geoffrey While]. (D) African elephants (Loxodonta africana) [photo by Nick Royle]. (E) Male Hyloxalus nexipus frog carrying tadpoles [photo by Jason Brown]. (F) Female red kangaroo (Macropus rufus) with joey [photo by Andy Russell]. (G) Stegodyphus lineatus spiderlings eating their mother (matriphagy) [photo by Trine Bilde]. (H) Wattled starling (Creatophora cinerea) feeding fledged young [photo by Nick Royle]. (I) Grey rhebok (Pelea capreolus) suckling young [photo by Nick Royle].

Parenting is complex and responsive to environmental factors

A parent spending more time foraging to provision offspring will have less time for offspring defense. Parents have to balance these competing demands when deciding how to allocate time to each activity. Parenting is therefore multidimensional [i.e., it is a multivariate trait (Fig. 2)]. The allocation of resources to these competing requirements is known to be sensitive to a host of environmental factors, both abiotic (e.g., rainfall and temperature) and biotic, with the latter further separable into nonsocial (e.g., food, predators, and pathogens) and social (e.g., offspring begging and partner contributions) categories. Figure 2 illustrates this idea for the simple case of two competing behaviors in two contrasting environments. For example, a parent experiencing an increased risk of nest predation would maximize fitness (reproductive success) by spending more time on nest defense, even though this must come at a cost to offspring provisioning. Consistent with this prediction, a recent study of 10 temperate and subtropical bird species from the Americas showed that parents provisioned offspring less when the perceived threat of nest depredation was raised by playing vocalizations of nest predators [Stellar’s jays (Cyanocitta stelleri) in North America and plush-capped jays (Cyanocorax chrysops) in South America]. Parents were more responsive to the threat in those species with a higher natural risk of nest predation (11). Thus, whereas flexibility in behavior was found in all 10 species studied, the most responsive species were those with an evolutionary history of higher predation risk, providing evidence that behavioral plasticity has itself evolved under selection.

Fig. 2 Parental fitness as a function of nest defense and offspring provisioning.

Darker shading denotes trait combinations yielding higher fitness. (A) Imagine a parent whose bivariate behavioral phenotype (blue circle) is optimal under low predation. (B) An increase in predation risk alters the fitness surface and also induces a plastic response (arrow) to a new phenotype (red circle) in which more time is spent in defense and less in provisioning. In this example, the plasticity is adaptive because fitness is greater than it would have been with no behavioral response. However, it is not optimal because higher-fitness trait combinations are possible. All else being equal, selection in the population should favor plastic responses that track the phenotypic optimum across environments.

A defining feature of parenting behavior is that it involves repeated interactions with other individuals of the same species: The behavior of a mother blackbird feeding a brood of hungry nestlings will be affected by the behavior of her partner (e.g., how much food he provides to the nestlings or how much time he spends defending the nest) and that of her offspring (e.g., how much they beg). Consequently, one of the most important forms of environment that parents experience is the social environment provided by family members. Experimental studies have clearly established that individuals respond to changes in their social environment. For example, a recent meta-analysis of 54 bird studies showed that individuals generally responded by working harder when their partner’s contributions were experimentally reduced (12). The average response differed between parental behaviors, with parents responding by increasing provisioning more than incubation and/or brooding. Additionally, females were more responsive than males to changes in partner contributions. It is currently unclear whether these results arise because individuals respond directly to changes in the contribution of their partner (13) or indirectly due to changes in offspring behavior (14). Importantly, however, these studies show that the social environment can induce changes in behavior and that such changes vary with the form of parental care, sex, and species involved.

Studies of plasticity at the individual level are needed

As highlighted above, many studies of parental care have shown changes in mean levels of behavior across environments, suggesting that adaptive phenotypic plasticity has evolved. However, the question of how much plasticity varies among individuals within populations has been almost entirely overlooked in the context of parenting behavior (8). If we hope to understand the contribution of plasticity to parental care evolution, then this is an important omission, because selection acts on among-individual variation (2, 8). Additionally, if this variation in plasticity is heritable, then plastic responses will themselves evolve (2, 7). This evolution of phenotypic plasticity will contribute to the adaptation of parenting. Quantifying among-individual variation in plasticity requires repeated-measures studies, in which the behavior of individually identifiable animals is recorded multiple times across varying environments. Such repeated-measures data are routinely collected in both lab and field studies of parenting (see Box 1 for an example). It is therefore surprising that little is currently known about among-individual variation in the plasticity of parenting or plasticity in behavioral traits, more generally (3, 4, 15). Nonetheless, successful studies of individual plasticity in other fields of evolution and ecology provide a useful template to follow. For example, in a study of lay date in wild great tits (Parus major) (16), analysis of repeated-measures data showed that plastic responses to spring temperature were both under directional selection and genetically variable. This leads to the prediction that plasticity will evolve, contributing to an adaptive advance in lay date as spring temperatures increase as a consequence of climate change (16).

Box 1

Individual plasticity in parental care in burying beetles.

Nicrophorus vespilloides burying beetles breed on the carcasses of small vertebrates (such as mice) that they process and bury underground as a source of food for their offspring (40). Parental care is complex and involves direct feeding of begging larvae and maintenance of the carcass to reduce competition from microbes (25, 26, 36, 40). Both sexes care for their young, either as a pair or alone, but there is considerable variation in the amount of parental care provided by individuals (36, 40). In the data presented below, individuals of both sexes reared offspring alone and in a pair in a repeated-measures, crossover-design experiment (n = 40 individuals of each sex), with the amount of time spent providing parental care over a 30-min period as the response variable. (A) On average, females (red) provide more care than males (blue), regardless of which social environment they experience (error bars denote SEs on sex- and environment-specific means). Yet both sexes provide, on average, less time caring in a biparental context. Analyzed with a simple linear model, the main effects of sex (P < 0.001) and social environment (P < 0.004) are both significant. However, there is also a significant interaction between them, meaning that the average reaction norm slope differs between sexes (P = 0.026).

Embedded Image

In fact, single-sex analysis shows that the mean reduction of female behavior associated with a switch between uniparental and biparental care is not significantly greater than zero. We might therefore conclude that, on average, males show plasticity but females do not. We know that the shift in mean male behavior must be due to plasticity, because the same individuals were observed in both social environments. However, we must also remember that population-level patterns only tell us what is happening on average. (B) Plotting individual reaction norms for the males reveals considerable among-individual variation around the average response. The correlation of care levels across environments for males is r = 0.032 (95% confidence interval = –0.284, 0.348). The correlation coefficient is <1, which provides evidence of IxE (i.e., there is variation in reaction norms of individual males). The photograph shows in (A) wild female N. vespilloides feeding offspring [credit: Paul Hopwood].

Behavioral reaction norms: A framework for studying individual-level plasticity

Variation in developmental plasticity has long been studied using the “reaction norm” approach (6, 17). More recently, this approach has also been adopted by researchers studying flexible life history traits, such as body size (1) and behaviors (3). In the latter context, behavioral reaction norms (BRNs) can be used to describe an individual’s behavior (e.g., the amount of food a parent provides to offspring) across an environment gradient, which may be nonsocial (e.g., food abundance) or social [e.g., offspring begging rate (18)]. Typically, though perhaps not always justifiably, a linear form of the BRN is assumed, which can be described by two parameters: a slope and an intercept (Fig. 3). The former describes plasticity, whereas mean-centering the x axis ensures that the intercept, or elevation, of the line describes behavior in an average environment (Fig. 3A). For empiricists, a useful feature of the BRN concept is that reaction norms are readily incorporated into linear mixed-effect models to test for among-individual variance in plasticity [commonly referred to as an “individual-by-environment” interaction (IxE) (2) (Fig. 3B)]. Furthermore, if relationships among individuals in the population are known, then pedigree-based quantitative genetic methods can be used to test for and estimate any heritable variation in plasticity [that is, “genotype-by-environment” interaction (GxE)] (2). Although there are some important pitfalls of interpretation to avoid, there is now a mature literature on how to apply these statistical methods to evolutionary and ecological studies of phenotypic plasticity (2, 19, 20).

Fig. 3 The BRN concept.

(A) Typically, BRN studies model an individual’s behavior (y) as a straight-line function of environment (x) such that plasticity is determined as the slope of a reaction norm. (B) Although many studies of parenting have shown that population-average behavior changes with environment (gray line), few have quantified among-individual variation in plasticity (IxE) or tested for a genetic component to it (GxE). Such individual-level studies are required to understand the contribution of plasticity to adaptive parenting and its evolution.

Examples of BRN use in the study of parenting

Although the BRN approach has yet to be widely applied to studies of parenting, its utility has been demonstrated recently by Westneat et al. (8) in a study of provisioning behavior in house sparrows (Passer domesticus). In this study, Westneat et al. used a suite of different environmental variables to estimate population-level plasticity (that is, mean reaction norm slope across individuals) and quantify among-individual variation in reaction norm intercepts and slopes for provisioning [IxE (see Fig. 3B)]. The number of provisioning trips per hour made by a parent was significantly repeatable across days (brood age 2 to 14 days)—with differences among individuals accounting for ~11% of the observed variation—and did not differ between the sexes overall. IxE was present for the environmental variables of offspring age and partner visit rate, but not for brood size. There was a population-level response to brood size (foraging rates increased when there were more nestlings), but the lack of IxE meant that individuals all responded in a similar way to a change in brood size. In addition, males and females differed in their average plasticity of provisioning behavior: Females were less responsive than males to differences in brood size, but more responsive to changes in nestling age.

The above example clearly illustrates that parenting behaviors can be plastic with respect to multiple environmental variables, both social and nonsocial. However, responses to one variable will also depend on other variables. Figure 4 illustrates this for a hypothetical scenario in which mothers always forage for food to nourish offspring, but where fathers may or may not help [e.g., red-wing blackbirds (Agelaius phoenicius)]. From a mother’s point of view, uniparental (maternal only) and biparental care represent two social environments, in each of which she needs to optimize foraging effort in relation to food abundance [i.e., a second, nonsocial environmental variable (Fig. 4)]. With no help from males, females will, on average, tend to decrease foraging effort as food abundance increases [solid gray line with slope β1 < 0 (Fig. 4A)], but there is also some variation among individuals in their BRN elevations and slopes (that is, differences in plasticity). Under biparental care, the picture is different: On average, females forage less and have a weaker plastic response to food abundance [shown by μ2 < μ1 (where μ is the mean phenotype in the environment) and |β2| < |β1| (Fig. 4B)]. This is because mothers and fathers now share the effort of finding food, as well as the savings that can be made when food is abundant.

Fig. 4 Parental plasticity across nonsocial and social environments.

Individual (blue dashed lines) and population-average (solid gray lines) behavioral reaction norms for female foraging effort (y) as a function of food abundance (x). (A) In the absence of a male partner, the behavior of a female i at food abundance x (yix) is specified by an elevation (c1) and slope (b1) expressed by convention as deviations from the population averages (μ1 and β1, respectively). (B) With male partner j present, female behavior will also depend on plastic response (di) to his foraging effort (yjx). Except in the case that di and yjx are both invariant across individuals, this leads to an expectation of more variable female BRNs under biparental care (see text for discussion).

However, it is also apparent that female BRNs are more variable under the biparental care scenario. We would expect this if females modify their behavior not only according to food abundance and the presence of a male partner, but also according to the level of foraging effort made by her partner. Variation in female BRNs will increase when mothers vary in their responses to a given male effort and/or when there is variation in effort among fathers partnered with different females. Of course, just as female behavior responds plastically to male effort, males typically alter their own behavior in response to female effort (12). Thus, how much care is provided, and by whom, will depend on reciprocal feedback between social partners (13). However, models of BRN evolution show that there is a cost to excessive plasticity in feedback between individuals (21). This is because unlimited feedback could lead to behavioral instability of the reaction norm, with individuals overreacting to behavioral change by their partners. Selection is therefore expected to favor reaction norms that allow behaviors to return to normal if there has been a temporary disturbance [for instance, when there has been an overreaction of one individual to another (21)].

Integrating the social environment

More generally, the scenario depicted in Fig. 4 shows that variance in female behavior and its plasticity (for example, IxE of foraging effort with respect to food abundance) can depend on the social environment provided by the male phenotype. In an evolutionary context, social environments differ from nonsocial environments in one vital respect: they are able to evolve. This is because social environments are composed of the phenotypes (and, thus, genotypes) of interacting conspecifics, an idea that has led to the development of indirect genetic effect (IGE) theory (22, 23). IGEs occur when characteristics (phenotype) of a focal individual (in this case, the parenting behavior of the female) are causally influenced by genes being expressed in one or more interacting individuals (e.g., the parental effort provided by her partner). These are therefore distinct from the direct genetic effects (DGEs) of an individual’s genes on its own phenotype. Thus, if there is genetic variation for female BRN elevations and slopes (i.e., plasticities), then during biparental care this will result from both DGEs (her own genetic makeup) and IGEs (arising from the genetic makeup of her partner).

The general importance of IGEs arising from social interactions has now been widely shown in laboratory and field studies (2327), although field-based investigations of IGEs on parenting behaviors per se are limited. Nonetheless, in a study of wild common gulls (Larus canus), Brommer and Rattiste showed that variation in female lay date was subject to IGEs arising from their partners (28). In the same way that males and females can be a source of IGEs on each other, they will also be a source of IGEs on the offspring if parenting behavior is heritable. Maternal genetic effects on offspring development, morphology, and/or behavior have been widely demonstrated in captive animal populations (29) and a number of species in the wild, including red squirrels (Tamiasciurus hudsonicus) and soay sheep (Ovis aries) (30). Clearly, the scope for complexity is potentially enormous when individuals interact and respond plastically to each other’s behavior. Exactly how IGEs will affect the evolutionary dynamics of the parenting BRN will depend on both the genetic covariance structure between DGEs and IGEs and the nature of selection acting, potentially antagonistically, through males and females (or parents and their offspring). Nonetheless, a key consequence of IGEs is that the evolution of parenting behaviors, and those phenotypes expressed by social partners to which they respond, will become coupled (that is, these behaviors will coevolve) (31, 32). Thus, incorporating IGEs into the reaction norm framework outlined above [as, for example, in (33)] is a necessary step to model the coevolutionary dynamics of plasticity in parenting.

Could coadaptation actually limit plasticity in parenting?

The presence of IGEs means that plasticity of parenting behaviors is expected to coevolve with traits expressed by partners (and their offspring) in their social environments. When selection favors particular combinations of parent and partner or parent and offspring behaviors, coadaptation will follow (32). Coadaptation occurs when the genetics underlying specific behaviors or other traits covaries between, for example, parents and offspring, and when this association is maintained over time via mechanisms that reinforce the associations and/or limit recombination between interrelated traits (32). As such, coadaptation is an evolutionary compromise in which conflicting fitness interests of parents and offspring impose limitations on the evolution of each other’s behavior. The best evidence for such coadaptation of traits comes from cross-fostering studies of parent-offspring interactions, where offspring have higher fitness (are more likely to survive and/or reproduce) when matched with their genetic parents (usually the mother) than when mismatched with foster parents (32). Positive relationships between maternal feeding and offspring begging behaviors have been found for numerous species, including great tits (P. major) (31) and burying beetles (N. vespilloides) (25): Mothers that are good providers have offspring that beg most strongly. In contrast, negative correlations have been found for rhesus macaques (Macacca mulatta) (34) and burrower bugs (Sehirus cinctus) (35): Mothers that are the best providers have offspring that beg the least. In most species, males provide less care than females, and their reduced contact might limit the potential for coadaptation between fathers and offspring (36). In the only studies where this has been quantified, no coadaptation between fathers and their offspring was found [great tits (P. major) (31) and burying beetles (N. vespilloides) (26)]. Although collectively these studies reveal coadaptation in parent and offspring behaviors, they do not address the role of plasticity in the evolution of coadapted behaviors.

Coadaptation models that allow the slopes of the behavioral reaction norms to evolve suggest that the responsiveness of parents and offspring to each other leads to stability in the amount of parental care provided (37). This is because there are multiple parent-offspring behavior combinations that provide solutions to the same problem [obtaining food if you are an offspring and providing food if you are a parent (37)]. Although this means that individuals are highly responsive to changes in the social environment, we hypothesize that coadaptation will actually limit the plasticity of parenting behaviors in response to nonsocial environments (for example, the threat of nest predation). This is because traits are genetically interdependent both within and between parents and offspring, meaning that behaviors are not free to evolve in the direction that always best serves the interests of the individuals involved.

Evolutionary interdependence of behaviors involved in social interactions stems from the presence of IGEs. Although IGEs can sometimes facilitate rapid evolutionary change (because the social environment can respond to behavioral change of individuals), they can also impose constraints on behavioral responses to other environments (22, 23). If correlations between DGEs and IGEs on fitness are positive, more rapid evolutionary responses to selection are expected to occur; if correlations are negative, the opposite will be true. Empirical work on IGEs in parental care shows that DGEs are most often (strongly) negatively correlated with IGEs, regardless of whether the care is provided by mothers or fathers (25, 26). This will limit the response of parenting behavior to selection on offspring and vice versa. As a result, IGEs and coadaptation traits involved in parenting are likely to be dynamic and responsive to changes in the social environment in which they have evolved, but this interdependence may limit adaptation to changes in nonsocial environments. An analogy can be drawn to organizations such as the United Nations (UN) responding to global crises, where the ability of the UN to respond effectively will be limited to some extent by compromises that have to be reached as a consequence of the often competing interests of member states: When behavior depends on others, the ability to respond to wider problems may be constrained.

Suggestions for the future

Generating simple predictions about the contribution of plasticity to adaptive evolution of parenting behavior is difficult at this stage, primarily due to the lack of suitably analyzed data. However, studies of parenting to date, conducted primarily to test between sex or among-population differences, provide a robust foundation for the next step. For many species, we know both the key parenting behaviors that determine fitness and the environmental factors (abiotic and biotic) to which these behaviors should respond adaptively. By collecting repeated-measures data on individuals in many studies of parenting, we have learned that parents respond plastically to changes in their environment. We now need to adopt an individual-based analytical approach to analyzing these data (2). Only by doing so can we address the issue of among-individual variation in plastic responses to different environments, particularly social ones. In fact, many researchers will already have the necessary data to adopt the BRN approach to quantifying IxE that we outline (see Box 1 for an example). Additionally, investigating genetic variation for plasticity (i.e., GxE) currently requires breeding experiments [in the lab; for example, selection lines (36)] or building a pedigree [in wild populations (30)], although the growing feasibility of obtaining high-density single nucleotide polymorphisms means that pedigree-free quantitative genetic analyses (38) may soon dominate for nonmodel organisms. Regardless of methodology, the key point is that genetic studies must consider both direct and indirect genetic sources of variation in parenting behaviors. By integrating IGEs into the existing BRN framework, it will be possible to more realistically model the evolution of parenting behaviors.

The rewards for doing so are likely to be great, as such work will not only provide insights into the plasticity of parenting and the robustness of responses of parents and offspring to rapid changes in their environment, but may also provide insights into the evolution of between-individual characteristics (personality) and complex societies. If plasticity is costly (3, 4), theoretical models demonstrate that this can favor the coexistence of plastic and nonplastic individuals within populations through frequency-dependent selection. This is because plastic individuals are good at responding to variation in their environment, which has the effect of reducing environmental variation experienced by others; nonplastic individuals can then exploit this environmental condition (39). This can lead to selection for between-individual differences in plasticity, which in turn, may have positive effects on population-level responses to environmental change because not all individuals will respond in the same way, maintaining variation in the population as a whole (39). Given the current rate of human-induced environmental change, we believe that it is important to address how plasticity, IGEs, and coadaptation interact to influence population responses to such perturbations. Consequently, to understand behavioral plasticity, we need to take a multidimensional view of both environmental and response variables. New theory and empirical tests of these questions and others we have highlighted are required if we are to understand variation in and the importance of plasticity in parenting and, more generally, the evolution of behavior.

References and Notes

  1. Acknowledgments: We thank C. Hinde, M. Kölliker, H. Klug, A. Moore, and two anonymous reviewers for comments on a previous version of the manuscript. We also thank H. Fitzjohn for collecting the data presented in Box 1, which have been deposited in the Open Research Exeter (ORE) data repository and are freely available at N.J.R. was funded by Natural Environment Research Council (NERC) grant NE/C002199/1, A.F.R. by NERC grant NE/K005766/1, and A.J.W. by a Biotechnology and Biological Sciences Research Council David Phillips Fellowship (BB/G022976/1).
View Abstract

Navigate This Article