Add Journal to My Library
Behavioral Ecology
, Volume 29 (1) – Jan 1, 2018

11 pages

/lp/ou_press/boldness-aggression-syndromes-can-reduce-population-density-behavior-719MBp3N9w

- Publisher
- Oxford University Press
- Copyright
- © The Author(s) 2017. Published by Oxford University Press on behalf of the International Society for Behavioral Ecology. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com
- ISSN
- 1045-2249
- eISSN
- 1465-7279
- D.O.I.
- 10.1093/beheco/arx068
- Publisher site
- See Article on Publisher Site

Abstract Behavioral syndromes are widely recognized as important for ecology and evolution, but most predictions about ecological impacts are based on conceptual models and are therefore imprecise. Borrowing insights from the theory of demographic heterogeneity, we derived insights about the population-dynamic effects of behavioral syndromes. If some individuals are consistently more aggressive than others, not just in interspecific contests, but also in foraging, mating, and antipredator behavior, then population dynamics could be affected by the resulting heterogeneity in demographic rates. We modeled a population with a boldness–aggressiveness syndrome (with the individual’s trait constant through life), showing that the mortality cost of boldness causes aggressive individuals to die earlier, on average, than their nonaggressive siblings. The equilibrium frequency of the aggressive type is strongly affected by the mortality cost of boldness, but not directly by the reproductive benefit of aggressiveness. Introducing aggressive types into a homogeneous nonaggressive population increases the average per-capita mortality rate at equilibrium; under many conditions, this reduces the equilibrium density. One such condition is that the reproductive benefit of aggression is frequency dependent and the population has evolved to equalize the expected fitness of the two types. Finally, if the intensity of aggressiveness can evolve, then the population is likely to evolve to an evolutionarily stable trait value under biologically reasonable assumptions. This analysis shows how a formal model can predict both how a syndrome affects population dynamics and how the population processes constrain evolution of the trait. INTRODUCTION Individual behavioral variation is now recognized as important for ecology and evolution (e.g., Dall et al. 2012). Of particular interest are systematic behavioral differences among individuals that are consistently manifested through time; these are often called animal personalities (Wolf and Weissing 2012). Often species also manifest behavioral syndromes, in which 2 or more behavioral traits covary across individuals (e.g., Sih et al. 2004); these have been observed in a range of taxa, including mammals, birds, fish, and a variety of invertebrates (e.g., Riechert and Hedrick 1993; Gosling 2001; Sih and Watters 2005). In a species with a behavioral syndrome, an individual is classified as having a particular behavioral type (BT), often quantified along axes such as aggression, activity, sociability, or fearfulness. In a population expressing a behavioral syndrome, individuals appear to manifest maladaptive behavior in some ecological contexts: for example, while an aggressive individual may be more effective at mating, its associated boldness makes it more vulnerable to predation (Sih et al. 2004). Proximally, these behavioral correlations can be understood in terms of the architecture of behavior (Wolf and Weissing 2010); in particular, they might be caused by genetic correlations, which in turn may arise from pleiotropy, linkage disequilibrium caused by a past history of correlated selection, or other molecular mechanisms such as physically linked regulatory regions (Dochtermann and Dingemanse 2013). Additive genetic variance can lead to BTs being heritable (e.g., Dingemanse et al. 2002), such that BT frequency may change in response to selection. A number of studies (e.g., McElreath and Strimling 2006; Wolf et al. 2007; Wolf et al. 2008; McNamara et al. 2009; Botero et al. 2010; Dubois et al. 2010; Houston 2010; Luttbeg and Sih 2010; Mathot and Dall 2013) have considered the kinds of underlying mechanisms that might lead to the evolution of BTs; building on these studies, the present work assumes that distinct BTs exist within a population, and explores the demographic consequences of having multiple BTs in a population. Sih et al. (2012) and Wolf and Weissing (2012) hypothesized a variety of ecological consequences of behavioral syndromes, including effects on population dynamics. Aggressive behavior might stabilize density-dependent equilibria, through a shift from scramble to interference competition (Sih et al. 2012); a similar effect may emerge if aggressive individuals cannibalize less aggressive individuals (Andersson et al. 2007; Pruitt et al. 2008). In contrast, increasing frequencies of aggressive individuals may destabilize an equilibrium if they have a faster “pace of life” (Réale et al. 2010), if temporal behavioral correlations introduce time lags in the population’s response to fluctuating environmental conditions (Sih et al. 2012), or if frequency-dependence leads to coupled oscillations of BT frequency and population abundance (Sinervo and Calsbeek 2006). Finally, behavioral syndromes may affect the intensity of density dependence, with potential impacts on equilibrium density: interaction rates may increase if aggressive individuals are more active (Pintor et al. 2009) and hence encounter conspecifics more frequently (Sih et al. 2012), but they may decrease if the various BTs use different resources and habitats (Wolf and Weissing 2012). Intriguing as these hypotheses are, few are based on formal population models (a notable exception is the study by Fogarty et al. (2011), showing how heterogeneity in a sociality syndrome could affect invasion speed), and thus we do not know the range of conditions over which they hold. There is some empirical evidence suggesting that behavioral syndromes may modify demographic rates such as birth and mortality rates. For example, a review by Biro and Stamps (2008) found that aggressiveness and boldness were consistently associated with increased birth rate, and a meta-analysis by Smith and Blumstein (2008) found that aggression was positively associated with birth rate and negatively associated with survival. These studies involved small and idiosyncratic sets of species, so we cannot draw strong conclusions about the generality of the results. However, when such effects occur, behavioral syndromes will lead to among-individual variation in demographic rates, which has come to be called “demographic heterogeneity” in population dynamics (e.g., Fox et al. 2006). Theoretical studies linking behavioral and life-history heterogeneity have mostly examined how the latter can generate the former. For example, Wolf et al. (2007) describe a model in which individuals may choose to delay reproduction in hope of finding a more favorable environment situation. If the fitness functions can maintain a stable polymorphism of life-history strategies, the heterogeneity in life history selects for correlated behaviors across individuals. In a conceptual model, Stamps (2007) suggests that personality and or behavioral syndromes might create or maintain among-individual heterogeneity in the position along a growth-mortality tradeoff, with heterogeneity among individuals in their “preferred” growth rate being driven by nonbehavioral factors. However, the position on the tradeoff is assumed to be fitness-neutral, so potential impacts of behavioral variation on population dynamics or evolution are not explored. There is a body of quantitative theory in population ecology showing that, depending on within-population correlation structure (Engen et al. 1998) and the underlying stochastic process, demographic heterogeneity can change the variance in both demographic outcomes and in the population growth rate due to demographic stochasticity (Kendall and Fox 2003; Vindenes et al. 2008). Furthermore, persistent survival heterogeneity (i.e., phenotypic variation that creates lifelong differences in instantaneous or annual mortality risk) in long-lived organisms can increase the population’s density-independent growth rate (Kendall et al. 2011), increase its equilibrium density (Stover et al. 2012), and reduce its extinction risk (Conner and White 1999). In contrast, persistent birth rate heterogeneity alone has more limited effects on dynamics (Kendall et al. 2011; Stover et al. 2012). The differential effects of heterogeneity in the two types of vital rates can be understood by recognizing that differences in survival accumulate multiplicatively with age: as a cohort ages, the more “frail” individuals tend to die off and the mean survival of the cohort increases. This “cohort selection” (Vaupel and Yashin 1985) means that the expected survival (averaged across all age classes in the population) is greater than would be found in a population with the same baseline value but no heterogeneity in survival. Differences in birth rates, however, accumulate only additively, and cohort selection on birth rate will only occur if birth rate is correlated with annual survival rate. A number of studies use formal theoretical models to investigate the conditions needed for the evolution of BT polymorphisms. Variation in metabolic rate, thought to be the likely basis for BT polymorphisms in some early studies, has been shown to have a complex relationship with BTs (Houston 2010; Mathot and Dall 2013). Luttbeg and Sih (2010) asked what sorts of ecological strategies might lead to coexistence of multiple BTs and showed that one plausible scenario was state-dependent safety (i.e., individuals with superior resources can best cope with predation risk, a positive feedback). Alternatives involving negative feedbacks (caution by those with more resources because they have more to lose, and boldness by those with fewer resources, to avoid starvation) cannot, they found, lead to stable coexistence of multiple BTs. Still other studies have considered conditions necessary for the evolution of behavioral consistency (McElreath and Strimling 2006; Wolf et al. 2008; Wolf et al. 2011; Wolf and McNamara 2012), how variation in individual quality can lead to BT polymorphisms (Botero et al. 2010), and how particular frequency-dependent scenarios can lead to BT polymorphisms (Dubois etal. 2010). We build on these results by asking a different question: given that BTs have evolved—by whatever mechanism—what are the demographic consequences for the population as a whole? Models of demographic heterogeneity lead us to expect that any behavioral syndrome that introduces persistent heterogeneity in survival will have impacts on the low-density population growth rate and on equilibrium abundance, in ways not addressed by Sih et al. (2012) or Wolf and Weissing (2012). Here, we develop a population model that incorporates a syndrome in which aggressive individuals are more successful at reproducing, but experience greater mortality (e.g., because of energetic costs of aggression, or because of greater exposure to predation); this creates a within-population pace-of-life syndrome (POLS) (Réale et al. 2010) in which an individual’s behavioral trait is associated with its expected “speed” of life history. The notion that POLS might vary within populations is not novel (Réale et al. 2010); however, we present the first quantitative model of population dynamics to examine how variation in POLS within a population can affect ecological dynamics and evolutionary outcomes. Using this model, we show four quite general results. First, the equilibrium BT frequency is directly controlled by the mortality cost of aggressiveness but is affected only indirectly by the reproductive benefit of aggression (via a parent-offspring correlation). Second, a population with a polymorphism for BTs will typically have a different density-dependent equilibrium than one made up entirely of the nonaggressive BT, and the polymorphic equilibrium is most often lower. Such effects on density could affect the species’ local extinction risk and influence on the ecological community. Third, if parents evolve to produce an offspring BT distribution that equalizes the expected fitness of both types (as found in a social spider; Pruitt and Goodnight 2014), then the equilibrium population abundance will always be lower than that for a monomorphic, nonaggressive population. Finally, we show that selection on the strength of the aggression trait may lead to a stable evolutionarily singular value (stable evolutionarily singular strategy [ESS]); while the resulting level of aggressiveness depends on details of model functions, the existence and stability of the ESS are nearly guaranteed if the mortality cost is an accelerating function of aggressiveness. By establishing quantitative links between population dynamics and behavioral syndromes, we hope to open up new realms of empirical inquiry in both fields. MODEL DESCRIPTION For simplicity of exposition, we modeled a population with two BTs, using the subscripts a and n to identify aggressive and nonaggressive individuals, respectively. As has been documented in various species (e.g., 3-spined stickleback; Huntingford 1976), aggressive individuals can monopolize mates or good territories, and thus have a higher birth rate, β: all else being equal, βa>βn. The other component of the syndrome is that aggressive individuals are bolder in contexts that may increase their mortality risk; thus, we model nonaggressives as having death rate μ and aggressives as having death rate (1+γ)μ, where γ>0 is the additional risk born by the aggressive BT. Thus, aggressive individuals have a fitness advantage over nonaggressive individuals if βa−(1+γ)μ>βn−μ (1) or βa−βn>γμ. (2) Table 1 provides a reference for all symbols used in the paper. Table 1 Symbols used in this paper Symbol Definition Variables, parameters, and indices in model N Population abundance wa Frequency of aggressive BTa in population π Frequency of aggressive BT among newborns β Birth rate μ Death rate γ Proportional increase in death rate experienced by aggressive individuals α Intensity of aggressive trait a Subscript to indicate the aggressive BT n Subscript to indicate the nonaggressive BT Quantities and symbols used in model analysis x* Asterisk: Superscript indicating that quantity x is evaluated at demographic equilibrium (dN⁄dt = 0) x¯ Overbar indicates the mean of quantity x across the population N0* Equilibrium abundance of monomorphic nonaggressive population wa˜ Frequency of aggressive BT in population at demographic equilibrium when average birth rates equal average death rates wa^ Frequency of aggressive BT in population at demographic equilibrium when differences between birth rates equal differences between death rates Quantities and symbols used in ESS analysis R Subscript indicating resident population I Subscript indicating invader population B Birth rate of invader in environment created by the resident population sx(y) Invasion exponent: per-capita growth rate of invader with phenotype y in environment created by resident with phenotype x Symbol Definition Variables, parameters, and indices in model N Population abundance wa Frequency of aggressive BTa in population π Frequency of aggressive BT among newborns β Birth rate μ Death rate γ Proportional increase in death rate experienced by aggressive individuals α Intensity of aggressive trait a Subscript to indicate the aggressive BT n Subscript to indicate the nonaggressive BT Quantities and symbols used in model analysis x* Asterisk: Superscript indicating that quantity x is evaluated at demographic equilibrium (dN⁄dt = 0) x¯ Overbar indicates the mean of quantity x across the population N0* Equilibrium abundance of monomorphic nonaggressive population wa˜ Frequency of aggressive BT in population at demographic equilibrium when average birth rates equal average death rates wa^ Frequency of aggressive BT in population at demographic equilibrium when differences between birth rates equal differences between death rates Quantities and symbols used in ESS analysis R Subscript indicating resident population I Subscript indicating invader population B Birth rate of invader in environment created by the resident population sx(y) Invasion exponent: per-capita growth rate of invader with phenotype y in environment created by resident with phenotype x aBehavioral type. View Large If aggressive individuals always hold a fitness advantage, then, if there is an additive genetic component to the syndrome, we would expect the aggressive BT to become fixed. Thus, to model a population that maintains multiple BTs (as is often observed in natural populations), we must either invoke a genetic mechanism such as heterozygote advantage (which has not been demonstrated empirically), assume there is no heritable component to the syndrome (which contradicts empirical evidence), or assume that the fitness difference between the BTs varies with density and/or frequency. We adopt the latter assumption in our analysis. In particular, we adopt the plausible assumption that aggressive individuals lose reproductive fitness by interacting with one another (e.g., Pruitt and Riechert 2009; Lichtenstein and Pruitt 2015). Thus, defining the frequency of aggressive BTs in the population as wa=NaNa+Nn, (3) the birth rate of both BTs declines with wa: ∂βi∂wa<0, (4) but that of the aggressive BT does so faster than that of the nonaggressive BT: −∂βa∂wa>−∂βn∂wa. (5) Finally, we assume that the birth rate of both BTs declines with density in the same way: ∂βa∂N=∂βn∂N<0, (6) where N=Na+Nb is the total population size. An example of birth rate functions displaying these qualitative features is illustrated in Figure 1. Figure 1 View largeDownload slide Example of density- and frequency-dependent birth rates that could arise from an aggression syndrome. Birth rates of both aggressive (heavy lines) and nonaggressive (thin lines) individuals decline with the frequency of aggressives in the population, but the relative advantage of aggression declines with increasing frequency of the aggressive BT. Birth rates of both behavioral types also decline with overall density (low density shown in solid lines, high density with dashed lines). Figure 1 View largeDownload slide Example of density- and frequency-dependent birth rates that could arise from an aggression syndrome. Birth rates of both aggressive (heavy lines) and nonaggressive (thin lines) individuals decline with the frequency of aggressives in the population, but the relative advantage of aggression declines with increasing frequency of the aggressive BT. Birth rates of both behavioral types also decline with overall density (low density shown in solid lines, high density with dashed lines). It is, of course, biologically plausible that frequency or density-dependence could instead (or in addition) occur in the death rate, or that the density-dependence is frequency dependent. We chose these particular assumptions to better draw upon the insights provided by the models in Stover et al. (2012). However, we do not expect that alternate assumptions about density- and frequency-dependence will qualitatively change our conclusions. For the model to be explicitly about the boldness–aggressiveness behavioral syndrome, we must specify constraints and tradeoffs on the various functions and parameters (for notational simplicity, we assume a perfect correlation between boldness and aggression). We characterize the phenotype of the aggressive BT with the parameter α (which we call aggressiveness), and we assume that it evolves slowly (allowing us to treat it as fixed in analyses of population dynamics). We let the birth rates depend on α as well as on the frequency of the aggressive BT, wa, and density, N, and assume that increasing the aggressiveness parameter increases the birth rate difference between the two BTs at a given frequency and density: ∂∂α[βa(α,wa,N)−βn(α,wa,N)]>0. (7) Furthermore, we expect that mortality is a function of α, so that increasing the aggressiveness parameter will also increase boldness, resulting in a greater mortality penalty: ∂∂αγ(α)>0. (8) This tradeoff is needed to prevent runaway selection on the aggressiveness parameter. We need one more component to build the population model: the frequency of each BT among the newborns. Behavioral syndromes have been demonstrated to be heritable (Dingemanse et al. 2002), but the underlying mechanisms have not been described. Therefore, we assume simply that a certain fraction, πa, of an aggressive individual’s offspring are aggressive; 1−πa of them are nonaggressive. Likewise, a fraction πnof a nonaggressive individual’s offspring are nonaggressive ( πaand πn need not have the same value). The explicit functional forms of πaand πndepend on the details of the inheritance mechanisms, and even with simple two-sex genetic models (e.g., one locus and two alleles with dominance, or quantitative variation in an underlying latent trait) the functions will be quite complex (e.g., if the behavior is controlled by an underlying continuous latent trait, πaand πnwill depend on the frequency of aggressives in the population; Falconer 1989). However, under a wide range of genetic mechanisms and mating systems, it is reasonable to assume that the rate at which a BT reproduces itself increases with the frequency of the BT; only when inheritance is near-perfect (e.g., parthenogenic with mutation or strong assortative mating) will the rates be independent of BT frequency (Crow and Kimura 1970; O’Donald 1980; Holsinger 1991; Hartl and Clark 1997). Thus, we assume ∂∂waπa≥0 (9) ∂∂waπn≤0 (10) We can now write the population model: dNadt=πa(wa)Naβa(α,wa,N)+[1−πn(wa)]Nnβn(α,wa,N)−μ[1+γ(α)]NadNndt=πn(wa)Nnβn(α,wa,N)+[1−πa(wa)]Naβa(α,wa,N)−μNn (11) It is sometimes convenient to first analyze a model in which the fraction of newborns that are aggressive is always a fixed value, which we call π.While our general analysis does not require this assumption, there may be cases in which this is a valid description of the biology, as when an individual’s BT is a plastic response to the developmental environment (including controls imposed by the parents). Even in this case, the frequency of each BT among an individual’s offspring ( π) and the magnitude of aggression in the aggressive BT ( α) may still be under selection. Under this simplifying assumption, πa(wa)=1−πn(wa)≡π, and Equation 11 can be written dNadt=π[Naβa(α,wa,N)+Nnβn(α,wa,N)]−μ(1+γ)Na (12) dNndt=(1−π)[Naβa(α,wa,N)+Nnβn(α,wa,N)]−μNn, (13) where π is the fraction of offspring with the aggressive BT. These models differ structurally from the model of Stover et al. (2012) in 3 important ways: the flexible function for the fraction of newborns in each BT, which allows us to include both BT heritability and adaptive control of newborn BT frequency; birth rate functions that are both more flexible (the Stover model assumed linear density dependence in which the heterogeneity parameter affected both the slope and intercept) and allow for frequency dependence; and association of the “baseline” death rate with the nonaggressive BT rather than with the average across BTs. Nevertheless, many insights and analysis techniques can be carried over from Stover et al. (2012). Note that some of inequalities (1–2) and (4–8) might be relaxed (become equalities) under special circumstances such as low density. However, it is reasonable to assume that they apply when the population is near its equilibrium density, which is where we conduct our analysis. MODEL ANALYSIS There are a number of questions we want to answer about the model. First, for a given level of aggression, α, with associated boldness cost γ(α), what is the equilibrium frequency of the aggressive BT ( wa*)? Second, is the associated equilibrium abundance ( N*) greater or less than the equilibrium abundance that would be found in a population made up only of nonaggressive individuals ( N0*)? Third, is there an ESS for the fraction of newborns that have the aggressive BT ( π)? The value of the ESS approach here is that it describes a predicted state to which populations tend in the long run, by its definition: an ESS is a strategy that, if fixed in a population, cannot be invaded by an alternative strategy that begins at low frequency. Finally, given a tradeoff between the benefits and costs of aggression, is there an ESS for α? As written, Equation 11 is too general to explicitly solve for N* and wa*.Even with the simplest form of the birth rate functions (linear dependence on density and frequency), the formulas for the equilibrium are too complex to provide much insight. However, we can get some qualitative (if sometimes vague) answers to these questions by looking at the model from different perspectives. For example, rather than attempting to examine four-dimensional figures (birth rates as a function of total population size, the fraction of aggressives in the population, and the measure of aggressiveness), in Figure 1, we examine a 2-dimensional slice: for fixed α, we focus on how wa, the frequency of aggressives, might affect birth rates at 2 population densities. Much of the analysis below uses a similar heuristic approach. Equilibrium frequency of the aggressive BT We can find the equilibrium BT frequency through a judicious manipulation of Equations 12 and 13. We divide both sides of Equation 12 by π and both sides of Equation 13 by 1−π, and then subtract the second resulting equation from the first. This gives 11−πdNndt−1πdNadt=μ[(1+γ)Naπ−Nn1−π]. (14) At equilibrium, both dNndt and dNadt are zero; applying this to Equation 14 and rearranging gives wa*≡Na*Na*+Nn*=ππ+(1+γ)(1−π), (15) where the stars indicate that the model is being evaluated at equilibrium. Thus, the aggressive BT’s equilibrium frequency depends only on its birth frequency and the mortality cost of boldness; increasing γ reduces the frequency of aggressives in the population relative to their frequency at birth (Figure 2). This result is a consequence of cohort selection (Kendall et al. 2011): as a cohort of newborns ages, the aggressive BTs die faster, and so their frequency in the cohort declines. At equilibrium, the population growth rate is zero, and so the age and BT distributions of the population match the life table of a cohort. Thus, the aggressive BT frequency in the population is found by averaging the frequency over all ages in a cohort, weighting by the fraction surviving to a given age, which will be less than the frequency at birth. Figure 2 View largeDownload slide Equilibrium frequency of the aggressive BT in the population ( wa*) as a function of the mortality cost of aggression ( γ). The curves are for different values of the aggressive BT frequency at birth ( π), at equal intervals from 0.1 to 0.9. Figure 2 View largeDownload slide Equilibrium frequency of the aggressive BT in the population ( wa*) as a function of the mortality cost of aggression ( γ). The curves are for different values of the aggressive BT frequency at birth ( π), at equal intervals from 0.1 to 0.9. Note that if the population is growing, then the age structure will tend to be biased towards younger individuals, relative to the equilibrium population. Younger cohorts have a higher frequency of aggressive BTs, since they have not been subject to so much cohort selection, and so a growing population will tend to have a higher aggressive BT frequency than will be found at equilibrium. By a similar argument, a population that is declining from above the equilibrium will have a lower aggressive BT frequency than the population will have at equilibrium. When the BT is heritable, we can still analyze the model at the demographic equilibrium, where πa(wa*) and πn(wa*) are constant. Here, we can write π*=πa(wa*)Na*βa*+[1−πn(wa*)]Nn*βn*Na*βa*+Nn*βn*=πa(wa*)wa*βa*+[1−πn(wa*)](1−wa*)βn*wa*βa*+(1−wa*)βn*, (16) where βi* are the birth rates evaluated at wa* and N*. Inserting Equation 16 into Equation 15 and solving for wa* will give the equilibrium frequency. Unfortunately, for most inheritance functions this will not be analytically tractable, but it will still be true that the frequency of aggressive BTs will be lower in the population as a whole than among the newborns. In fact we can be more specific: at equilibrium, the average death rate in the population is the harmonic mean of the newborn death rates, as was shown by Stover et al. (2012). Note that if a second gender carries the genes for the behavioral syndrome but does not express them, that gender will not be subject to cohort selection. Thus, the nonexpressing gender will have a genotype frequency that matches that of newborns (and differs from that of the expressing gender). This substantially complicates the expression for the inheritance functions, but does not qualitatively change the fundamental result above. Aggression’s effect on the equilibrium population density As a point of reference, we take the equilibrium density of a population made up of only nonaggressive individuals (which we call N0*), and we ask whether a population with both BTs has a population equilibrium that is larger or smaller than this reference. N0* is defined as the density at which the nonaggressive birth rate matches its death rate: βn(0,N0*)=μ (in this section we are holding α constant so we suppress it for notational simplicity). If, near wa=0, the aggressive BT’s birth rate is lower than its death rate, then the aggressive BT cannot invade the population, and so we focus on the situation where βa(0,N0*)>μ(1+γ). Increasing the aggressive BT frequency, wa, while holding N=N0* constant leads to declines in the birth rates of both BTs, but does not affect the death rates of either BT (because birth rates are density-dependent but death rates are not; Figure 3). We can also define average birth and death rates: Figure 3 View largeDownload slide Birth and mortality rates of the aggressive BT (heavy lines), the nonaggressive BT (thin lines), and the population average (grey lines), as a function of wa, the frequency of aggressive BTs in the population. The overall population density is at the demographic equilibrium for a population made up of only nonaggressive BTs (at wa=0, nonaggressive birth rate equals nonaggressive mortality rate). The vertical line indicates wa˜, the positive frequency of aggressives that would reach demographic equilibrium at the same density (average birth rate equals average mortality rate). If 0<wa<wa˜ then the average birth rate exceeds the average mortality rate and the population would grow until it reaches demographic equilibrium at a higher density (lowering the birth rate curves; see Figure 1). If wa>wa˜ then the average birth rate is less than the average mortality rate and the population would decline until it reaches demographic equilibrium at a lower density. Figure 3 View largeDownload slide Birth and mortality rates of the aggressive BT (heavy lines), the nonaggressive BT (thin lines), and the population average (grey lines), as a function of wa, the frequency of aggressive BTs in the population. The overall population density is at the demographic equilibrium for a population made up of only nonaggressive BTs (at wa=0, nonaggressive birth rate equals nonaggressive mortality rate). The vertical line indicates wa˜, the positive frequency of aggressives that would reach demographic equilibrium at the same density (average birth rate equals average mortality rate). If 0<wa<wa˜ then the average birth rate exceeds the average mortality rate and the population would grow until it reaches demographic equilibrium at a higher density (lowering the birth rate curves; see Figure 1). If wa>wa˜ then the average birth rate is less than the average mortality rate and the population would decline until it reaches demographic equilibrium at a lower density. β¯(wa,N)=waβa(wa,N)+(1−wa)βn(wa,N) (17) μ¯(wa)=waμ(1+γ)+(1−wa)μ=μ(1+waγ). (18) While the death rates of each BT are constant, the average death rate increases linearly with wa because the aggressive type has a greater death rate. The average birth rate may show more complex patterns, because birth rates can also be density-dependent. In general, given constraints (5) and (6), the average birth rate will be maximized for a positive value of wa and its derivative with respect to wa will be greatest at wa=0, as shown in Figure 3. At N=N0*, the average birth and death rates are equal at wa=0. If the aggressive BT enters the population at low frequency (so wa>0), then, given the model assumptions there are three general cases we might see. Case 1 First, the average birth rate might be greater than the average death rate for all values of wa. This would require that the average birth rate increase quite rapidly with wa. In particular, inspection of Figure 3 reveals that this case requires that the birth rate increase faster than the death rate when the aggressive BT is rare and that the aggressive BT’s birth rate exceeds its death rate even when wa=1. This can be shown to require: −∂βn∂wa|wa=0<βa(0,N0*)−βn(0,N0*)+μγ (19) βa(1,N0*)>μ(1+γ). (20) Case 2 The average birth rate might be less than the average death rate for all values of wa. This would require that the nonaggressive birth rate decline sufficiently rapidly with wa when the aggressive BT is rare. In particular: −∂βn∂wa|wa=0>βa(0,N0*)−βn(0,N0*)+μγ. (21) Case 3 The average birth rate might be larger than the average death rate for small wa and be less than the average death rate for large wa as is illustrated in Figure 3. The average birth and death rates are equal at an intermediate value of wa which we call wa˜. In all 3 cases, if the population is found at BT frequency wa and total abundance N0*, then abundance will increase if β¯(wa,N0*)>μ¯(wa) and decrease if β¯(wa,N0*)<μ¯(wa) (of course, wa will then change dynamically as well). Now, we know that at equilibrium, the equilibrium BT frequency is wa*, defined by Equation 15. We also know that at equilibrium, where N=N*, the average birth and death rates must be equal. Therefore, it is only possible for N*=N0* under the conditions of case 3 and when wa*=wa˜. If β¯(wa*,N0*)>μ¯(wa*), then N*>N0*; likewise, if β¯(wa*,N0*)<μ¯(wa*), then N*<N0*. Thus, the introduction of the aggressive behavioral syndrome into a naive population will increase the equilibrium abundance only if the aggressive BT has a very strong fitness advantage (taking into account the boldness cost) even at high frequencies (case 1) or if the equilibrium frequency of the aggressive BT is relatively low (case 3). Under case 2 or most circumstances of case 3 the syndrome will reduce the equilibrium abundance. Evolution of the birth frequency of the aggressive BT For a given set of demographic parameters and inheritance functions, (N*,wa*) is the demographic equilibrium of the population model. However, it will not, in general, eliminate the fitness differences between the two BTs. To see this, suppose that case 3 applies and that wa*=wa˜, so that the demographic equilibrium is at (N0*,wa˜), as shown in Figure 3. At this point, the average birth rate equals the average death rate. However, the birth and death rates are not equal for either of the BTs. In particular, the nonaggressive BT must have negative net fitness, because its net fitness when N=N0* is zero only when wa=0 and declines with increasing aggressive BT frequency. To achieve zero net fitness at the population level, the aggressive BT must have positive fitness. At the demographic equilibrium, the population growth rate is zero, and the reproductive value at birth for each BT is its birth rate divided by its death rate. At the equilibrium in Figure 3, these are unequal. Just as Fisher (1930) showed with regard to the evolution of primary sex ratios, there will be selection on parents to increase the frequency of the type with the higher reproductive value (here, the aggressive BT) among their offspring. If the inheritance function is purely genetic, with a fixed genetic architecture, then there is no way to respond to selection at this particular demographic equilibrium. However, if there are environmental influences (expressed directly or via epigenetic mechanisms) on an individual’s BT, then parents may be able to increase the frequency of the aggressive BT among their offspring, effectively changing the inheritance function (e.g., androgen levels in egg yolk can influence offspring behavior, although this has not yet been explicitly linked to a behavioral syndrome; Ruuskanen and Laaksonen 2010). Increasing the frequency of aggressives among newborns π will, in turn, increase the demographic equilibrium wa* to a value greater than wa˜. As shown in the previous section, this will lead to an equilibrium density that is less than N0*. As long as the birth rates of the two BTs respond in the same way to density, then at this new equilibrium, under the conditions of case 3, there will be a wa>wa˜, which we call wa^, that satisfies Equation 1—that is, the differences between the birth rates match the differences between the death rates. At the associated demographic equilibrium then not only do births match deaths for the population as a whole but also for each of the BTs (Figure 4). Further increases in walead to a fitness advantage for the nonaggressive BT, so wa^ is the evolutionarily stable BT frequency. An analogous argument applies in case 2. Figure 4 View largeDownload slide Birth and mortality rates of the aggressive BT (heavy lines), the nonaggressive BT (thin lines), and the population average (grey lines), as a function of wa, the frequency of aggressive BTs in the population. The overall density is at the demographic equilibrium for wa=wa^, the frequency of aggressives at which the two BTs have equal relative fitness (the difference in fecundities equals the difference in mortalities). Because wa^>wa˜, this demographic equilibrium is at a density lower than that of the nonaggressive-only equilibrium, allowing the frequency-dependent fecundities to be elevated. Figure 4 View largeDownload slide Birth and mortality rates of the aggressive BT (heavy lines), the nonaggressive BT (thin lines), and the population average (grey lines), as a function of wa, the frequency of aggressive BTs in the population. The overall density is at the demographic equilibrium for wa=wa^, the frequency of aggressives at which the two BTs have equal relative fitness (the difference in fecundities equals the difference in mortalities). Because wa^>wa˜, this demographic equilibrium is at a density lower than that of the nonaggressive-only equilibrium, allowing the frequency-dependent fecundities to be elevated. In contrast, in case 1, where βa(1,N0*)>μ(1+γ), there is no frequency at which the 2 BTs have equal fitness at a demographic equilibrium (mathematically, wa^>1), so the evolutionarily stable BT frequency is 1 (fixation of the aggressive BT). Only in this last situation (at which we would no longer recognize a syndrome, as there is no behavioral variation) would evolution to a stable BT frequency result in a demographic equilibrium density that is larger than the density of a nonaggressive population (N0*). ESS for aggressiveness In addition to the BT frequency, the strength of the aggressiveness trait ( α) might itself be subject to selection. Incorporating an explicit genetic model for α would add a great deal of complexity to the model, so we instead take an adaptive dynamics approach, and look for an ESS for aggressiveness (Geritz et al. 1998). In particular, we focus on the conditions allowing a resident population (with a given aggressiveness parameter, αR) that is at demographic equilibrium to be invaded by a population with a different aggressiveness parameter (αI). We start by stating the results, and then give the mathematical derivation. To our existing model, we need to add one new quantity: the birth rate of an aggressive individual with trait αI in the presence of a given abundance and BT frequency of a resident population with trait αR. We call this Ba(αI,αR,wa*,N*); note that it will not be the same as the resident birth rate, βa(αR,wa*,N*). We also assume that the resident evolves to a birth rate frequency that equalizes the fitnesses of the two BTs, so that wa*=wa^. The evolutionarily singular strategy, α*, is the value of aggressiveness that satisfies the condition ∂Ba∂αI=μ∂γ∂α (22) when evaluated at αI=αR=α. In other words, from the invader’s perspective, the birth rate benefits of increased aggression are exactly matched by the death rate costs when the invader and resident traits are identical. The ESS is “ESS-stable,” meaning that a resident population that is at the ESS cannot be invaded (Geritz et al. 1998), if ∂Ba2∂αI2<μ∂2γ∂α2. (23) In particular, under the biologically reasonable assumption of diminishing returns to reproduction from increased aggression (making the left hand side negative), this condition will always be met if the mortality cost is linear or accelerating in the aggression trait. However, ESS-stability does not guarantee that the ESS can be reached through successive mutations of a resident population that is not at the ESS. This requires an additional property, called “convergence stability” (Geritz et al. 1998), and the ESS is called a convergence stable strategy (CSS; Diekmann 2004). Unfortunately, the formal conditions for the ESS to be convergence stable in this model require additional information, such as the relative sensitivity of the invader and resident birth rates to changes in the resident BT frequency and the inheritance mechanisms that determine the BT frequency in the invader population. While the condition can easily be calculated if functional forms are assumed, the general expression is sufficiently complex as to be noninformative. However, it seems clear that, if α=0 is not an ESS (i.e., a mutant with slight aggressiveness can invade a population with none), then the existence of one or more ESSs at positive values of α will ensure that at least one of them is convergence-stable. Mathematical derivations The analysis of an ESS focuses on the invasion exponent sx(y), which is the low-density per-capita growth rate of an invader with trait value yin the environment created by a resident population at equilibrium and having trait value x (the notation here follows Geritz et al. 1998; Diekmann 2004). In the present context, x=αR and y=αI. The analysis proceeds by looking at derivatives of s evaluated at y=x. In particular, the condition for x to be an ESS is c2≡∂∂ysx(y)|y=x=0, (24) and the condition for ESS-stability is c22≡∂2∂y2sx(y)|y=x<0. (25) To calculate the invasion exponent we write out the dynamics of the invader population. Since the invader is rare, we assume that only the resident population N* impacts the invader’s reproduction and that individuals of both populations are unaffected by the invader’s aggressiveness or BT frequency: dNa(I)dt=πa(wa(I))Na(I)Ba(αI,αR,wa*,N*)+[1−πn(wa(I))]Nn(I)Bn(αI,αR,wa*,N*)−μ[1+γI]Na(I) (26) dNn(I)dt=πn(wa(I))Nn(I)Bn(αI,αR,wa*,N*)+[1−πa(wa(I))]Na(I)Ba(αI,αR,wa*,N*)−μNn(I), (27) where wa(I) is the aggressive BT frequency among invaders (which might affect the BT frequency at birth) and γI≡γ(αI) is the death rate penalty of the invader aggressives. Adding these together and dividing by N(I)=Na(I)+Nn(I) gives sx(y)=wa(I)[Ba−μ(1+γI)]+(1−wa(I))[Bn−μ]. (28) It is quite reasonable to assume that the invader nonaggressive BT is identical to that of the resident, so that Bn=βn. Furthermore, if we assume the resident is at the fitness equalizing frequency, such that βn=μ,then the second term is zero, leaving sx(y)=wa(I)[Ba−μ(1+γI)] (29) The first derivative is ∂∂ysx(y)=∂wa(I)∂αI[Ba−μ(1+γI)]+wa(I)[∂∂αIBa−∂∂αIμ(1+γI)]. (30) When αI=αR, it is quite reasonable to assume that Ba=βa and γI=γR. Again, if we assume the resident is at the fitness equalizing frequency then βa=μ(1+γR) and the first term is zero. As long as wa(I)>0, this leads directly to the ESS condition in Equation 22. The second derivative is ∂2∂y2sx(y)=∂2wa(I)∂αI2[Ba−μ(1+γI)]+2∂wa(I)∂αI[∂∂αIBa−∂∂αIμ(1+γI)]+wa(I)[∂2∂αI2Ba−∂2∂αI2μ(1+γI)]. (31) As above, when αI=αR,then the first term is zero. Furthermore, if the resident is at an ESS, then the second term is also zero (the quantity in brackets is just c2). As long as wa(I)>0, this leads directly to the ESS-stability condition in Equation 23. DISCUSSION We have developed a model that links behavioral and population processes by noting that the fitness associated with a particular behavior translates into demography—birth and death rates—at the population level. Applied to behavioral syndromes, this allows us to draw on existing theoretical frameworks (demographic heterogeneity and adaptive dynamics) to make a number of predictions about ecological and evolutionary outcomes, including the factors that control the frequency of the BTs in the population, the effect of the syndrome on equilibrium abundance, and how selection should drive the evolution of both the distribution of the BTs among offspring and the overall intensity of the trait in each BT. We applied the model to the boldness–aggression syndrome, but the general approach should apply to any syndrome with identifiable fitness consequences. In the boldness–aggression syndrome, the BTs differ in both their birth rates and their death rates. To a scientist not steeped in demographic theory, it may come as a surprise that the effects these two sources of heterogeneity are not commensurate. For example, we showed that when the population growth rate is zero, the frequency of the two BTs depends only on the frequency of the types among newborns and the relative death rates of the types—not on the differences in birth rates (note that if there are genetically, epigenetically (e.g., Francis et al. 1999; Weaver et al. 2004) or environmentally induced correlations between the BTs of parents and their offspring, then the differences in birth rate will have an indirect effect via their effect on the distribution of newborn types). This occurs because mortality heterogeneity results in cohort selection, in which the composition of a cohort changes as the cohort ages (Vaupel and Yashin 1985), whereas birth rate heterogeneity does not. One way of developing some intuition about this is to think about lifetime reproductive success (LRS) in a simple life history in which a BT’s birth ( β) and death ( μ) rates are both age-independent. Here, for an individual with BT i, the expected LRS is simply βiLi, where Li is the BT’s expected longevity. In a population with heterogeneous birth rates, the mean LRS can be found by using the mean of the BT-specific birth rates, and is unaffected by the amount of heterogeneity. Likewise, in a population with heterogeneous death rates, the mean LRS can be found by using the mean of the phenotypic-specific longevities. However, an individual’s expected longevity is an inverse function of its mortality rate: Li=1/μi. This nonlinear relationship means that the average longevity in the population will not be the same as the longevity of an individual with an “average” mortality rate; Jensen’s inequality (Zens and Peart 2003) tells us that heterogeneity in μ will cause the population mean longevity to be larger than the longevity with the mean death rate, with the discrepancy increasing as the magnitude of the heterogeneity gets larger. This translates directly into effects on mean LRS. This fundamental distinction between birth and mortality heterogeneity persists even with age- and environment-dependent vital rates, as long as there is some degree of within-individual auto-correlation in vital rates, as would be caused by consistent expression of a given behavior. We found that the boldness–aggression syndrome often leads to a reduction in equilibrium abundance relative to a population made up of only nonaggressive types. This occurs because of the intersection of 2 factors. First, as the frequency of the aggressive BT increases, the mean death rate in the population also increases, reflecting the higher risk associated with boldness. Second, increasing the aggressive BT frequency increases the mean birth rate (for a fixed abundance) when the aggressive BT is rare (because the aggressives have higher birth rates), but the frequency-dependent depression of the birth rate drives down the mean birth rate when the aggressive type is more common. Once the frequency is high enough that the mean birth rate (at the nonaggressive equilibrium density) is below the mean death rate, then an equilibrium can only be reached if the density-dependence in the birth rate is relaxed by settling to a lower abundance. Note that, while our model incorporates density- and frequency-dependence in the birth rate only, qualitatively similar results would obtain if one or both dependencies were in mortality. This effect on equilibrium density means that the behavioral syndrome may increase the population’s local extinction risk, with implications for conservation and metapopulation dynamics. It may also reduce the species’ trophic and facilitative impacts on other members of the ecological community. Even at the demographic equilibrium, the two BTs will not necessarily have equal fitness. When they do not, there will be selection to increase the proportion of the more fit BT. If there is a response to this selection (whether through plasticity or evolutionary change), the BT frequency will move towards a value where both BTs have the same fitness. We have shown that this frequency equilibrium will always be at a value that results in a reduced equilibrium abundance, relative to a purely nonaggressive population. In density-dependent populations, selection maximizes the equilibrium abundance (often thought of as the carrying capacity, K), if fitness is not frequency dependent (Charlesworth 1980). But because of the frequency-dependence of fitness of each BT in our model, behavioral evolution reduces abundance and thus increases the risk of population-level extinction due to stochastic fluctuations or exclusion by a competitor that can persist at lower resource densities (Webb 2003). How does aggressiveness ( α) evolve? Our results show that there may be an evolutionarily stable value, but predicting where that will occur requires an understanding of the fitness of an invader with one aggressiveness level in a population of residents with a different aggressiveness. It may be reasonable that the outcome of the interaction between 2 aggressive individuals with different levels of aggressiveness only depends on the difference between the 2 α’s. Thus, the left hand side of Equation 22—the derivative of the invader birth rate with respect to the invader aggressiveness, evaluated where the invader and resident have the same aggressiveness—will be independent of the resident aggressiveness level. However, changing the resident aggressiveness level will also, in general, change the resident’s equilibrium density and BT frequency; by analogy to the analysis comparing polymorphic and monomorphic populations, we might expect that increasing aggressiveness will decrease N* and increase w*. We cannot say much in general about how these will impact the invader birth rate, especially as the predicted changes are likely to have opposite effects. If the effects balance out, then the left hand side of Equation 22 will be a constant, and an ESS will only exist if the right hand side is nonconstant—that is, the death rate cost is a nonlinear function of aggressiveness. Turning to the stability condition, it is reasonable to assume that the invader birth rate has diminishing returns to aggressiveness, making the right-hand side of Equation 23 negative. Thus, if the mortality cost is an accelerating function of aggressiveness, then both the existence of the ESS and its stability will be guaranteed. We have modeled the behavioral syndrome as a dichotomous trait, in large part for ease of analysis and exposition, but a trait such as aggressiveness may take on a continuous range of values. In simple models of demographic heterogeneity (without density-dependence, frequency-dependence, or inheritance), a continuous distribution of death rates has been shown to have virtually identical effects on population dynamics as a dichotomous trait, the key value being the variance of death rates (Kendall et al. 2011). In models with density dependence, a key value for 2-type models is the harmonic mean death rate among individuals in a newborn cohort (Stover et al. 2012); one might expect this to generalize to continuous trait distributions, but this has not been tested. Frequency-dependence in continuous traits has to be handled with care; the simplest approach is to assume a strict hierarchy, so that the fitness of an individual with aggressiveness trait αi only depends on the frequency of individuals with traits greater than αi. Some qualitative predictions that derive from our model are described in Table 2. Nevertheless, applying this theoretical framework to particular species will require explicit functional forms for the density- and frequency-dependence in vital rates, as well as behavioral effects on these functions. These can be estimated empirically in focal populations, but further generalization will require a more mechanistic understanding of the underlying physiological bases of behavior (which generates the behavioral correlations) and the factors that link behaviors to fitness. A predictive theory also needs a mechanistic basis for the newborn BT distribution—for example, what are the roles of genetics, parental effects, and plasticity? Where the genetic mechanisms are known, we would need to develop explicit inheritance models using population genetics (Charlesworth 1980) or quantitative genetics (Barfield et al. 2011); the latter would probably be most appropriate when the behavioral trait is continuously distributed. These are all areas where empirical research needs to guide model development. Such models could be used to understand the causes and consequences of phenomena such as the reduced fitness of heterotypic matings across a boldness syndrome in guppies (Ariyomo and Watt 2013). Table 2 Qualitative predictions about populations exhibiting a boldness-aggression tradeoff that derive from the theory developed in this paper Prediction Rationale Compared with a population at a BT frequency that equalizes fitness between the types, an artificial population made up solely of nonaggressive individuals will, under identical environmental conditions, grow to a larger equilibrium abundance. Corollary of the result that introducing an aggressive BT to a nonaggressive population will lower equilibrium abundance at the fitness-equalizing BT frequency. The aggressive BT in populations growing in low-predation environments will exhibit stronger aggressive tendencies than the aggressive BT in high-predation environments.a Reduced boldness cost at a given aggressiveness level will allow evolution to higher aggressiveness. Assumes local evolution of aggressiveness trait. Populations in low-predation environments will have a higher aggressive BT frequency and may, depending on the strength of the boldness cost and the contribution of predation to overall mortality, have a lower equilibrium abundance than high-predation populations.a Low predation reduces μ, reducing mortality difference between BTs; this moves the fitness-equalizing frequency to a greater fraction of aggressives, where the birth-rate difference is also lower. For a given density, this reduces mean birth rate; if baseline mortality is not much affected by predation then reduced birth rate may exceed reduced mean death rate, requiring reduced density to attain demographic equilibrium. Assumes local adaption to fitness-equalizing BT frequency. Populations in which aggression is under sexual selection will have an increased BT frequency and reduced equilibrium abundance relative to otherwise identical populations without sexual selection. Sexual selection increases difference in birth rates without affecting difference in death rate; thus fitness-equalizing BT frequency will be a greater fraction of aggressives. Effect on equilibrium density follows directly from this. Assumes no differences in aggressiveness trait. Prediction Rationale Compared with a population at a BT frequency that equalizes fitness between the types, an artificial population made up solely of nonaggressive individuals will, under identical environmental conditions, grow to a larger equilibrium abundance. Corollary of the result that introducing an aggressive BT to a nonaggressive population will lower equilibrium abundance at the fitness-equalizing BT frequency. The aggressive BT in populations growing in low-predation environments will exhibit stronger aggressive tendencies than the aggressive BT in high-predation environments.a Reduced boldness cost at a given aggressiveness level will allow evolution to higher aggressiveness. Assumes local evolution of aggressiveness trait. Populations in low-predation environments will have a higher aggressive BT frequency and may, depending on the strength of the boldness cost and the contribution of predation to overall mortality, have a lower equilibrium abundance than high-predation populations.a Low predation reduces μ, reducing mortality difference between BTs; this moves the fitness-equalizing frequency to a greater fraction of aggressives, where the birth-rate difference is also lower. For a given density, this reduces mean birth rate; if baseline mortality is not much affected by predation then reduced birth rate may exceed reduced mean death rate, requiring reduced density to attain demographic equilibrium. Assumes local adaption to fitness-equalizing BT frequency. Populations in which aggression is under sexual selection will have an increased BT frequency and reduced equilibrium abundance relative to otherwise identical populations without sexual selection. Sexual selection increases difference in birth rates without affecting difference in death rate; thus fitness-equalizing BT frequency will be a greater fraction of aggressives. Effect on equilibrium density follows directly from this. Assumes no differences in aggressiveness trait. Quantitative predictions will require a model that is tailored to the empirical system being studied. Deviations from the model’s qualitative assumptions about density- and frequency-dependence will require new analysis to confirm that the predictions apply. aAssumes that the boldness cost is a consequence of heightened vulnerability to predation. View Large The population model used here is very simple; in particular, it does not have age- or size-structure, and assumes that environmental conditions are constant. These factors can generate time-lags in feedback loops (because of the time to reach maturity) and fluctuating selection (e.g., fluctuations in predator populations, or in populations of alternate prey, which might lead to fluctuations in the relative cost of boldness as well as in overall mortality rates), respectively, and so may prevent the population from settling down to an ecological or evolutionary equilibrium. However, except in long-lived species, these are likely to be second-order effects that primarily affect quantitative rather than qualitative predictions. Of greater import, the model does not account for differences between sexes. Some species exhibit sexual dimorphism in behavioral traits (e.g., Pruitt et al. 2011; Han et al. 2015), and while individuals of the sex that does not express the syndrome do not experience the direct fitness effects, they may still influence the BTs of their offspring via genes or parental effects. If the frequency of aggressive BTs is below the level where both types have equal fitness and there is a genetic contribution to behavior, then individuals of the nonexpressing sex could increase their fitness by preferentially mating with aggressive individuals. This would put the syndrome under sexual selection, and would further increase the birth rate advantage of aggression without necessarily increasing the boldness cost (Logue et al. 2009). A recent essay on “data-free papers” in the behavioral syndromes literature (DiRienzo and Montiglio 2015) suggests that such papers (encompassing syntheses of older theories as well as novel conceptual frameworks, terminologies, or statistical approaches) are contributing relatively little to our understanding of the subject (although Davis et al. (2015) argue that data-free papers sometimes contribute substantially to the conceptual framework of the field). Notably absent from the critique by DiRienzo and Montiglio (2015) are formal models; this is perhaps due to their relative paucity (we have found only one model linking behavioral syndromes to population dynamics; Fogarty et al. 2011). However, DiRienzo and Montiglio (2015) suggest (correctly, in our view) that formal models are an important avenue (along with empirical study) to effectively study the speculative links suggested by verbal conceptual frameworks. The work here represents such a contribution. In particular, our findings that the equilibrium frequency of the aggressive BT depends strongly on the mortality cost of boldness and that the equilibrium population abundance is negatively related to the frequency of the aggressive BT could only have been derived from a model that translates the fitness consequences to the individual into birth and death rates, and included dynamic feedbacks via density-dependence and frequency-dependence. Indeed, these feedbacks help shape the fitness landscape in which the behavioral syndrome evolves in such a way that understanding the optimal balance between the value of individual behavioral traits and the BT frequencies within the population requires explicit modeling of the species’ population ecology. Making quantitative predictions about specific systems will require tailoring the model to those systems; such species-specific models can then be used, for example, to predict and explain patterns observed in common garden experiments with animals drawn from different selective environments. We hope that this paper will stimulate studies that integrate empirical observation and formal modeling. Funding This work was supported by the National Science Foundation (grant numbers DEB-1120865 to G.A.F., DEB-1120330 to B.E.K.). The authors thank Jonathan Pruitt and two anonymous reviewers for comments on the manuscript. REFERENCES Andersson J Byström P Claessen D Persson L de Roos AM. 2007. Stabilization of population fluctuations due to cannibalism promotes resource polymorphism in fish. Am Nat . 169: 820– 829. Google Scholar CrossRef Search ADS PubMed Ariyomo TO Watt PJ. 2013. Disassortative mating for boldness decreases reproductive success in the guppy. Behav Ecol . 24: 1320– 1326. Google Scholar CrossRef Search ADS Barfield M Holt RD Gomulkiewicz R. 2011. Evolution in stage-structured populations. Am Nat . 177: 397– 409. Google Scholar CrossRef Search ADS PubMed Biro PA Stamps JA. 2008. Are animal personality traits linked to life-history productivity? Trends Ecol Evol . 23: 361– 368. Google Scholar CrossRef Search ADS PubMed Botero CA Pen I Komdeur J Weissing FJ. 2010. The evolution of individual variation in communication strategies. Evolution . 64: 3123– 3133. Google Scholar CrossRef Search ADS PubMed Charlesworth B. 1980. Evolution in age-structured populations . Cambridge (UK): Cambridge University Press. Conner MM White GC. 1999. Effects of individual heterogeneity in estimating the persistence of small populations. Nat Resour Model . 12: 109– 127. Google Scholar CrossRef Search ADS Crow JF Kimura M. 1970. An introduction to population genetics theory . New York (NY): Harper & Row. Dall SR Bell AM Bolnick DI Ratnieks FL. 2012. An evolutionary ecology of individual differences. Ecol Lett . 15: 1189– 1198. Google Scholar CrossRef Search ADS PubMed Davis GH Payne E Sih A. 2015. Commentary: four ways in which data-free papers on animal personality fail to be impactful. Front Ecol Evol . 3: 102. Diekmann O. 2004. A beginner’s guide to adaptive dynamics. In Rudnicki R, editor. Mathematical modelling of population dynamics . Warsaw: Polish Academy of Sciences. p. 47– 86. Dingemanse NJ Both C Drent PJ van Oers K van Noordwijk AJ. 2002. Repeatability and heritability of exploratory behaviour in great tits from the wild. Anim Behav . 64: 929– 938. Google Scholar CrossRef Search ADS DiRienzo N Montiglio PO. 2015. Four ways in which data-free papers on animal personality fail to be impactful. Front Ecol Evol . 3: 23. Dochtermann NA Dingemanse NJ. 2013. Behavioral syndromes as evolutionary constraints. Behav Ecol . 24: 806– 811. Google Scholar CrossRef Search ADS Dubois F Morand-Ferron J Giraldeau LA. 2010. Learning in a game context: strategy choice by some keeps learning from evolving in others. Proc R Soc B . 277: 3609– 3616. Google Scholar CrossRef Search ADS PubMed Engen S Bakke Ø Islam A. 1998. Demographic and environmental stochasticity – Concepts and definitions. Biometrics . 54: 840– 846. Google Scholar CrossRef Search ADS Falconer DS. 1989. Introduction to quantitative genetics . 3rd ed. Harlow (UK): Longman Scientific & Technical. Fisher RA. 1930. The genetical theory of natural selection . Oxford (UK): Clarendon Press. Google Scholar CrossRef Search ADS Fogarty S Cote J Sih A. 2011. Social personality polymorphism and the spread of invasive species: a model. Am Nat . 177: 273– 287. Google Scholar CrossRef Search ADS PubMed Fox GA Kendall BE Fitzpatrick JW Woolfenden GE. 2006. Consequences of heterogeneity in survival probability in a population of Florida scrub-jays. J Anim Ecol . 75: 921– 927. Google Scholar CrossRef Search ADS PubMed Francis D Diorio J Liu D Meaney MJ. 1999. Nongenomic transmission across generations of maternal behavior and stress responses in the rat. Science . 286: 1155– 1158. Google Scholar CrossRef Search ADS PubMed Geritz SAH Kisdi É Meszéna G Metz JAJ. 1998. Evolutionarily singular strategies and the adaptive growth and branching of the evolutionary tree. Evol Ecol . 12: 35– 57. Google Scholar CrossRef Search ADS Gosling SD. 2001. From mice to men: what can we learn about personality from animal research? Psychol Bull . 127: 45– 86. Google Scholar CrossRef Search ADS PubMed Han CS Jablonski PG Brooks RC. 2015. Intimidating courtship and sex differences in predation risk lead to sex-specific behavioural syndromes. Anim Behav . 109: 177– 185. Google Scholar CrossRef Search ADS Hartl DL Clark AG. 1997. Principles of population genetics . Sunderland (MA): Sinauer. Holsinger KE. 1991. Mass-action models of plant mating systems: the evolutionary stability of mixed mating systems. Am Nat . 138: 606– 622. Google Scholar CrossRef Search ADS Houston AI. 2010. Evolutionary models of metabolism, behaviour and personality. Philos Trans R Soc Lond B . 365: 3969– 3975. Google Scholar CrossRef Search ADS Huntingford FA. 1976. The relationship between anti-predator behaviour and aggression among conspecifics in the three-spined stickleback, Gasterosteus aculeatus. Anim Behav . 24: 245– 260. Google Scholar CrossRef Search ADS Kendall BE Fox GA. 2003. Unstructured individual variation and demographic stochasticity. Conserv Biol . 17: 1170– 1172. Google Scholar CrossRef Search ADS Kendall BE Fox GA Fujiwara M Nogeire TM. 2011. Demographic heterogeneity, cohort selection, and population growth. Ecology . 92: 1985– 1993. Google Scholar CrossRef Search ADS PubMed Lichtenstein JL Pruitt JN. 2015. Similar patterns of frequency-dependent selection on animal personalities emerge in three species of social spiders. J Evol Biol . 28: 1248– 1256. Google Scholar CrossRef Search ADS PubMed Logue DM Mishra S McCaffrey D Ball D Cade WH. 2009. A behavioral syndrome linking courtship behavior toward males and females predicts reproductive success from a single mating in the hissing cockroach, Gromphadorhina portentosa. Behav Ecol . 20: 781– 788. Google Scholar CrossRef Search ADS Luttbeg B Sih A. 2010. Risk, resources and state-dependent adaptive behavioural syndromes. Philos Trans R Soc Lond B . 365: 3977– 3990. Google Scholar CrossRef Search ADS Mathot KJ Dall SRX. 2013. Metabolic rates can drive individual differences in information and insurance use under the risk of starvation. Am Nat . 182: 611– 620. Google Scholar CrossRef Search ADS PubMed McElreath R Strimling P. 2006. How noisy information and individual asymmetries can make ‘personality’ an adaptation: a simple model. Anim Behav . 72: 1135– 1139. Google Scholar CrossRef Search ADS McNamara JM Stephens PA Dall SRX Houston AI. 2009. Evolution of trust and trustworthiness: social awareness favours personality differences. Proc R Soc B . 276: 605– 613. Google Scholar CrossRef Search ADS PubMed O’Donald P. 1980. A general analysis of genetic models with frequency-dependent mating. Heredity . 44: 309– 320. Google Scholar CrossRef Search ADS Pintor LM Sih A Kerby JL. 2009. Behavioural correlations provide a mechanism for explaining high invader densities and increased impacts on native prey. Ecology . 90: 581– 587. Google Scholar CrossRef Search ADS PubMed Pruitt JN Goodnight CJ. 2014. Site-specific group selection drives locally adapted group compositions. Nature . 514: 359– 362. Google Scholar CrossRef Search ADS PubMed Pruitt JN Riechert SE. 2009. Frequency-dependent success of cheaters during foraging bouts might limit their spread within colonies of a socially polymorphic spider. Evolution . 63: 2966– 2973. Google Scholar CrossRef Search ADS PubMed Pruitt JN Riechert SE Harris DJ. 2011. Reproductive consequences of male body mass and aggressiveness depend on females’ behavioral types. Behav Ecol Sociobiol . 65: 1957– 1966. Google Scholar CrossRef Search ADS Pruitt JN Riechert SE Jones TC. 2008. Behavioural syndromes and their fitness consequences in a socially polymorphic spider, Anelosimus studiosus. Anim Behav . 76: 871– 879. Google Scholar CrossRef Search ADS Réale D Garant D Humphries MM Bergeron P Careau V Montiglio PO. 2010. Personality and the emergence of the pace-of-life syndrome concept at the population level. Philos Trans R Soc Lond B . 365: 4051– 4063. Google Scholar CrossRef Search ADS Riechert SE Hedrick AV. 1993. A test for correlations among fitness-linked behavioural traits in the spider Agelenopsis aperta (Araneae, Agelenidae). Anim Behav . 46: 669– 675. Google Scholar CrossRef Search ADS Ruuskanen S Laaksonen T. 2010. Yolk hormones have sex-specific long-term effects on behavior in the pied flycatcher (Ficedula hypoleuca). Horm Behav . 57: 119– 127. Google Scholar CrossRef Search ADS PubMed Sih A Bell A Johnson JC. 2004. Behavioral syndromes: an ecological and evolutionary overview. Trends Ecol Evol . 19: 372– 378. Google Scholar CrossRef Search ADS PubMed Sih A Cote J Evans M Fogarty S Pruitt J. 2012. Ecological implications of behavioural syndromes. Ecol Lett . 15: 278– 289. Google Scholar CrossRef Search ADS PubMed Sih A Watters JV. 2005. The mix matters: behavioural types and group dynamics in water striders. Behaviour . 142: 1417– 1431. Google Scholar CrossRef Search ADS Sinervo B Calsbeek R. 2006. The developmental, physiological, neural, and genetical causes and consequences of frequency-dependent selection in the wild. Ann Rev Ecol Evol Syst . 37: 581– 610. Google Scholar CrossRef Search ADS Smith BR Blumstein DT. 2008. Fitness consequences of personality: a meta-analysis. Behav Ecol . 19: 448– 455. Google Scholar CrossRef Search ADS Stamps JA. 2007. Growth-mortality tradeoffs and ‘personality traits’ in animals. Ecol Lett . 10: 355– 363. Google Scholar CrossRef Search ADS PubMed Stover JP Kendall BE Fox GA. 2012. Demographic heterogeneity impacts density-dependent population dynamics. Theor Ecol . 5: 297– 309. Google Scholar CrossRef Search ADS Vaupel JW Yashin AI. 1985. Heterogeneity’s ruses: some surprising effects of selection on population dynamics. Am Stat . 39: 176– 185. Google Scholar PubMed Vindenes Y Engen S Sæther B. 2008. Individual heterogeneity in vital parameters and demographic stochasticity. Am Nat . 171: 455– 467. Google Scholar CrossRef Search ADS PubMed Weaver ICG Cervoni N Champagne FA D’Alessio AC Sharma S Seckl JR Dymov S Szyf M Meaney MJ. 2004. Epigenetic programming by maternal behavior. Nat Neurosci . 7: 847– 854. Google Scholar CrossRef Search ADS PubMed Webb C. 2003. A complete classification of Darwinian extinction in ecological interactions. Am Nat . 161: 181– 205. Google Scholar CrossRef Search ADS PubMed Wolf M McNamara JM. 2012. On the evolution of personalities via frequency-dependent selection. Am Nat . 179: 679– 692. Google Scholar CrossRef Search ADS PubMed Wolf M Van Doorn GS Leimar O Weissing FJ. 2007. Life-history trade-offs favour the evolution of animal personalities. Nature . 447: 581– 584. Google Scholar CrossRef Search ADS PubMed Wolf M Van Doorn GS Weissing FJ. 2008. Evolutionary emergence of responsive and unresponsive personalities. Proc Natl Acad Sci USA . 105: 15825– 15830. Google Scholar CrossRef Search ADS PubMed Wolf M Van Doorn GS Weissing FJ. 2011. On the coevolution of social responsiveness and behavioural consistency. Proc R Soc B . 278: 440– 448. Google Scholar CrossRef Search ADS PubMed Wolf M Weissing FJ. 2010. An explanatory framework for adaptive personality differences. Philos Trans R Soc Lond B . 365: 3959– 3968. Google Scholar CrossRef Search ADS Wolf M Weissing FJ. 2012. Animal personalities: consequences for ecology and evolution. Trends Ecol Evol . 27: 452– 461. Google Scholar CrossRef Search ADS PubMed Zens MS Peart DR. 2003. Dealing with death data: individual hazards, mortality and bias. Trends Ecol Evol . 18: 366– 373. Google Scholar CrossRef Search ADS © The Author(s) 2017. Published by Oxford University Press on behalf of the International Society for Behavioral Ecology. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com

Behavioral Ecology – Oxford University Press

**Published: ** Jan 1, 2018

Loading...

personal research library

It’s your single place to instantly

**discover** and **read** the research

that matters to you.

Enjoy **affordable access** to

over 18 million articles from more than

**15,000 peer-reviewed journals**.

All for just $49/month

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Read from thousands of the leading scholarly journals from *SpringerNature*, *Elsevier*, *Wiley-Blackwell*, *Oxford University Press* and more.

All the latest content is available, no embargo periods.

## “Hi guys, I cannot tell you how much I love this resource. Incredible. I really believe you've hit the nail on the head with this site in regards to solving the research-purchase issue.”

Daniel C.

## “Whoa! It’s like Spotify but for academic articles.”

@Phil_Robichaud

## “I must say, @deepdyve is a fabulous solution to the independent researcher's problem of #access to #information.”

@deepthiw

## “My last article couldn't be possible without the platform @deepdyve that makes journal papers cheaper.”

@JoseServera

DeepDyve ## Freelancer | DeepDyve ## Pro | |
---|---|---|

Price | FREE | $49/month |

Save searches from | ||

Create lists to | ||

Export lists, citations | ||

Read DeepDyve articles | Abstract access only | Unlimited access to over |

20 pages / month | ||

PDF Discount | 20% off | |

Read and print from thousands of top scholarly journals.

System error. Please try again!

or

By signing up, you agree to DeepDyve’s Terms of Service and Privacy Policy.

Already have an account? Log in

Bookmark this article. You can see your Bookmarks on your DeepDyve Library.

To save an article, **log in** first, or **sign up** for a DeepDyve account if you don’t already have one.

All DeepDyve websites use cookies to improve your online experience. They were placed on your computer when you launched this website. You can change your cookie settings through your browser.

ok to continue