Current address: National Center for Ecological Analysis & Synthesis, University of California, 735 State St. Suite 300, Santa Barbara, California 93101.


Numerous integrated traits contribute to any aspect of organismal performance, but favorable trait combinations are difficult to maintain in the face of genetic recombination. We investigated the role of maternal effects in promoting integration of alternative reproductive strategies (∼throat colors) with antipredator traits (escape behaviors and dorsal patterns) in the side-blotched lizard (Uta stansburiana). Previously, we reported that maternally derived estradiol adaptively pairs dorsal patterns with progeny throat colors. Here we show adaptive maternal effects on escape behaviors within each throat color morph. Specifically, yellow-throated females and all females mated to yellow-throated sires lay larger eggs. Larger eggs produce stockier offspring, who remain stockier throughout life. Stockiness promotes evasive escape behaviors (e.g., reversals), which are adaptive in barred, yellow-throated offspring. Orange-throated females lay smaller eggs, producing leaner hatchlings who perform vertical escape behaviors (e.g., jumping). Vertical behaviors enhance survival in striped, orange-throated progeny. Escape behavior was not heritable, but was organized by natural or experimental egg size variation. Maternal effects on adaptive phenotypic integration are likely common in polymorphic species, because recombination otherwise breaks apart beneficial trait combinations. Furthermore, our results provide insight into the role of body shape in organizing (and constraining) evolution of integrated reproductive and antipredator strategies.

Adaptive phenotypic integration is the achievement of adaptive patterns of covariation among fitness-related traits; it involves orchestration of numerous genetic, structural, and environmental influences in the development of a complex phenotype that adaptively performs multiple functions across varying environmental contexts (Arnold 1983; Cheverud 1996; Pigliucci and Preston 2004). Phenotypic integration is often studied by comparing genetic, environmental, developmental, and phenotypic variation across populations or species to determine how interactions among these factors can evolve (Roff 2004). However, polymorphic species also offer a unique opportunity to study the mechanisms promoting phenotypic integration (Wright 1978; Lande and Arnold 1983; Brodie 1992; Ahnesjö and Forsman 2003; Forsman et al. 2008), because they exhibit alternative adaptations within populations, involving alternative patterns of variation and covariation in many traits that are often comparable to differences between populations or species (West-Eberhard 1986). In contrast to divergent suites of traits among populations, the maintenance of phenotypic integration of alternative types across generations is opposed by recombination, which tends to dissolve suites of coadapted traits. As a result, the mechanisms contributing to phenotypic integration in polymorphic populations are available for direct examination within each generation (i.e., adaptive phenotypic integration must be reachieved after each bout of recombination, thus the process can be studied directly).

Common polymorphisms in animals involve alternative antipredator phenotypes (Ahnesjö and Forsman 2003; Bond 2007), resource utilization tactics (Smith and Skúlason 1996), or reproductive strategies (Shuster and Wade 2003). Because alternative strategies represent alternative ways of achieving high fitness, they necessarily involve differences in a number of traits, such as behaviors, morphologies, and physiological capabilities (Sinervo and Svensson 2002; Sinervo and Calsbeek 2006; Mills et al. 2008). Each morphotype (i.e., polymorphic phenotype) can also express differing interactions among environmental, genetic, and developmental causes of trait formation (Lancaster et al. 2008). As such, different morphotypes represent alternative patterns of phenotypic integration, which are maintained in the face of interbreeding.

The susceptibility of well-integrated suites of traits to the destructive force of recombination suggests that polymorphic species may commonly achieve phenotypic integration through plasticity or indirect-genetic means. One mechanism by which suites of traits could be reliably integrated within each generation is through maternal effects. In general, maternal effects represent contributions of the maternal phenotype to offspring phenotypes, acting in addition to the direct inheritance of genetic material (Kirkpatrick and Lande 1989; Mousseau and Fox 1998; Wolf and Wade 2009). These causal effects may be direct, such as when maternal provisioning behavior affects offspring growth and development, or they may be indirect, such as when the environment in which a female chooses to lay her eggs has effects on offspring traits. Maternal effects may or may not interact with offspring genotypes in determining the outcome for offspring phenotypes.

Previous treatments of maternal effects have considered their adaptive potential when the maternal environment predicts the offspring environment: environmental cues experienced by the female trigger maternal effects that cause development of offspring phenotypes well suited to the (future) environment that they will experience (Dingle 1996; Galloway and Etterson 2007). Here we extend this traditional view of adaptive maternal effects to consider maternal responses to environmental cues that predict genes offspring are likely to inherit. We hypothesize that in conspicuously polymorphic animals, a female may adjust patterns of resource allocation to offspring in response to both her own morphotype (a form of maternal–offspring coadaptation; Wolf and Brodie 1998) as well as the perceived morphotype of her mate, which could be used to generate adaptive combinations of traits in offspring. In other words, we hypothesize that females plastically alter offspring traits in anticipation of genes for morphotype that offspring will likely inherit. These maternally affected traits are then beneficial to offspring, but only in combination with the inherited morphotype. Trait integration through maternal effect plasticity could adaptively reduce genetic load arising from recombination and promote alternative patterns of phenotypic integration favored by correlational selection. This represents a novel approach to the study of adaptive maternal effects.


We investigated alternative, high-fitness reproductive and antipredator traits, and maternal effects promoting alternative patterns of phenotypic integration of these traits in the common side-blotched lizard, Uta stansburiana. This species exhibits two discrete kinds of color polymorphism. One is a throat color polymorphism that signals alternative reproductive strategies in each sex. In males, socially dominant territorial males are orange-throated, mate-guarding males are blue-throated, and female mimicking sneaker males are yellow-throated (Sinervo and Lively 1996; Zamudio and Sinervo 2000). In females, orange-throated lizards lay large clutches of small eggs (Sinervo et al. 2000b) and reproduce at higher rates when under stress and facing a survival versus reproduction trade-off (Lancaster et al. 2008). Yellow- and blue-throated females lay smaller clutches of larger eggs, and yellow females forego reproduction in favor of survival when stressed. Linkage mapping (Sinervo et al. 2006), theory (Sinervo 2001), and laboratory crosses (Sinervo et al. 2001) indicate that throat color expression has a strictly genetic basis and is controlled by three codominant alleles at the OBY locus: o, b, and y. Throat color is also highly correlated with other aspects of reproductive strategy, including territoriality (Zamudio and Sinervo 2000), hormones (Sinervo et al. 2000a; Comendant et al. 2003), and immune function (Svensson et al. 2001). Previous studies show that correlational selection (Sinervo and Svensson 2002; Sinervo and Clobert 2003) and other forms of fitness epistasis (Sinervo et al. 2008) maintain these strategy-specific trait associations.

Uta stansburiana also exhibits a polymorphism in patterning on the dorsum. These patterns are highly cryptic (L. T. Lancaster, A. G. McAdam, and B. Sinervo, pers. obs.; see also Lancaster et al. 2007Fig. 1 for photographic depictions). Dorsal patterning exhibits continuous variation in the shapes of lightly colored pattern elements in two dimensions. A baseline, “chevron” pattern resembles two columns of inverted chevrons extending from the neck, down the back, to the base of the tail. In some lizards, this pattern is extended in the cranial–caudal dimension, resulting in cranial–caudal stripes or an intermediate pattern of dashes. Alternatively, the chevron pattern can be extended in the medal–lateral direction, resulting in two rows of laterally flattened chevrons or bars (for photographic depictions, see Fig. 1 in Lancaster et al. 2007). Dorsal patterning is heritable but not genetically correlated with throat color (in the context of a laboratory cross in which the two traits were experimentally decoupled in parents; Lancaster et al. 2007). However, we previously measured correlational selection acting to link dorsal pattern with throat color in adult lizards in the wild (Lancaster et al. 2007, 2009). Specifically, barred dorsal patterns were favored in yellow-throated lizards of each sex, whereas stripes were favored in orange-throated males (Lancaster et al. 2007).

Figure 1.

Alternative high-fitness multitrait phenotypes involving throat color, dorsal patterning, and escape behavior. (A) Orange-throated offspring who were both striped and performed vertical behaviors survived well. (B) Yellow-throated offspring who were barred and performed evasive behaviors also survived well. In (A), the graphical depiction was achieved by representing only sibling-groups estimated to be carrying at least one o allele. In (B), depiction is limited to sibling-groups carrying at least one y allele. Therefore, neither of these graphs depicts the full dataset upon which the reported results are based (which would result in a four-dimensional surface).

Dorsal patterning is also a target of correlational sexual selection in this species, with females preferring to mate with barred males, but only if males are yellow-throated (Lancaster et al. 2009). Mating with sires according to this preference enhances offspring survival via “good genes” sexual selection (Lancaster et al. 2009). In addition, females can induce appropriate dorsal patterning in offspring of each throat color genotype and sex via modulating concentrations of maternally derived yolk estradiol. Elevated yolk estradiol induces barring in yellow-throated offspring, and induces striping in orange sons and in nonorange daughters (Lancaster et al. 2007). Female choice (exerting correlational sexual selection) and yolk estradiol (maternal effect plasticity) represent alternative means by which females can adaptively promote the integration of throat color with dorsal patterning in this species. The hypothesized agent of selection favoring specific correlations between throat color and dorsal patterning is predation, the primary cause of mortality in U. stansburiana. Specifically, our sightings of the coachwhip snake (Masticophis flagellum), a highly visually oriented predator, are associated with a drop in expected survival time of lizards from 2 months to 5 days (B. Sinervo, unpubl. data). In the course of fieldwork, we have also witnessed many coachwhip attacks, but only one attack by a different (kestrel) predator (L. T. Lancaster and B. Sinervo, pers. obs.).


We hypothesized that the previously measured, strong correlational selection on throat color and dorsal pattern may be in part explained by differences in antipredator or escape behaviors among the different throat color morphs. Previous studies have shown that escape behavior and color pattern can synergistically affect survival against predation (Brodie 1989, 1992; Forsman and Appelqvist 1998; Carretero et al. 2006). Escape behavior may also be linked with reproductive strategy morphotype via expression of similar behaviors across social and antipredator contexts (personality; Sih et al. 2004a,b) or habitat preferences, in which different escape behaviors work best in each morph's preferred microhabitat. To investigate the role of escape behavior in shaping high-fitness throat color/dorsal pattern combinations, we considered a fitness model testing for effects of interactions among escape behavior, throat color morph, and dorsal pattern on survivorship. Specifically, based on our previous fitness results for throat color paired with dorsal patterns (Lancaster et al. 2007, 2009 and described above), we hypothesized that yellow-throated lizards would experience high fitness in combination with barring and some escape behaviors, whereas orange-throated lizards would experience high fitness in combination with striping and alternative escape behaviors.

We then hypothesized two, distinct maternal effect mechanisms as putative adaptations for integrating escape behavior with throat color and dorsal patterning. First, maternally derived yolk estradiol, which adaptively induces dorsal patterning to match offspring throat colors (Lancaster et al. 2007), may also affect offspring escape behavior in addition to its effect on patterning. A second possible maternal effect on these trait combinations could arise through variation in egg size. The female throat color morphs differ in the average mass of their eggs (Sinervo et al. 2000b), and egg mass variation has allometric effects on hatchling mass, sprint speed, stamina, limb morphology, and body proportions (Sinervo 1990; Sinervo and Huey 1990). Allometric morphological variation across species and populations is known to functionally affect escape behavior and locomotion in other study systems (Vanhooydonck et al. 2002; Fitzpatrick et al. 2003; Schulte et al. 2004; Husak and Rouse 2006; Dial et al. 2008; Bergmann et al. 2009), and so we hypothesized that allometric morphological variation within our population, due to divergent, morph-specific maternal egg size effects, could adaptively direct escape behavior to match reproductive strategy type.

We also predicted that such maternal effects would be adaptive when the maternal (social) environment predicts which genes offspring will likely inherit. Maternal investment of estradiol into yolks is known to vary as a function of the sire's throat color genotype (Lancaster et al. 2007). We also hypothesized that female egg mass allocation decisions might respond to the sire's and female's own reproductive strategy type to promote phenotypic integration within morphotypes.

To test our hypotheses, we bred lizards in the laboratory, with matings randomized with respect to parental throat color genotypes. We measured natural variation in mated females’ egg masses and yolk estradiol concentrations. We also experimentally manipulated egg mass of two eggs within each clutch (by sampling yolk for estradiol assay, we reduced the mass of eggs, resulting in smaller hatchlings; Sinervo 1990). This experiment was designed to distinguish direct egg-size effects (which are purely maternal) from any confounding direct genetic effects that may correlate with maternal genes for egg size. We then measured escape behavior, morphology, and dorsal patterning of progeny before releasing progeny to the wild to assess survival to adulthood (i.e., to identify any beneficial patterns of trait integration).



We conducted two laboratory crosses, one in 2004 and one in 2006, using field-captured lizards obtained in early spring before reproduction was likely to have occurred. Lizards were captured in Merced Co., CA, from rock outcrops adjacent to long-term study plots (Sinervo et al. 2001; Lancaster et al. 2007). Throat color, dorsal patterning, mass, and snout-to-vent length (SVL) were recorded following capture. In each cross, n= 71 (2004) and n= 56 (2006) sires were each paired with three dams. Sires were randomly paired with dams, with an attempt to balance throat color alleles and dorsal patterning within and between breeding enclosures. In 2006, we also balanced female preferences for sires among enclosures (Lancaster et al. 2009). Breeding enclosures were each supplied with a sand substrate, 75-W basking lamp and a UVB fluorescent bulb (both from Zoo Med Laboratories, Inc., San Luis Obispo, CA). Lizards were fed an ad libitum supply of crickets daily, dusted with calcium supplement (Rep-Cal; Los Gatos, CA). Once each week, we scored breeding lizards for throat color, and females were scored for reproductive stage via abdominal palpation. We removed gravid females to individual ovipositoria: containers with a deep layer of moistened peat moss/sand mixture, a basking rock, and a 40-W light bulb. Gravid females were checked daily for laying. Eggs were weighed individually and incubated at 28°C in individual cups of moistened vermiculite. Two eggs per clutch were randomly selected for yolk sampling, in which 20% of the yolk was removed with a syringe and frozen for later hormone analysis. Sampled eggs were reweighed following yolk removal and incubated as usual. Hatchlings born from sampled eggs exhibit reduced body mass, and are therefore referred to as “miniaturized” hatchlings (Sinervo and Huey 1990). We also use the term “miniaturization” to describe the experimental procedure of aspirating a yolk sample to produce smaller hatchlings. Previous studies have shown that the effect of sampling eggs per se (insertion of a syringe without aspirating yolk) does not affect hatchling size or performance but slightly increases egg mortality (Sinervo 1990). Upon hatching, progeny were individually toe-clipped, and we recorded sex (males exhibit enlarged postanal scales), mass, SVL, and dorsal patterning. Following behavioral trials (see below), we released hatchlings to the same outcrop from which parents had been captured, but specific release locations were randomized with respect to siblingship to control for effects of maternal territory quality. The next year, we recaptured lizards as adults to estimate survival (Sinervo et al. 2001; Lancaster et al. 2007). Laboratory cross protocols are also described in Sinervo et al. (2001) and Lancaster et al. (2007, 2009). Yolk samples were assayed for estradiol using radioimmunoassay (Schwabl 1993; Lancaster et al. 2007).


Prior to release, we observed hatchling escape behavior. Escape behavior trials were performed in 2006 on a circular track constructed of concentric rings of sheet metal. The diameters of the outer and inner rings were 87.5 cm and 58.5 cm, respectively; making the width of the track 29 cm. Distance run was measured in 5 cm increments along the outer ring of the track, which had a total circumference of 275 cm. The floor of the track was sprayed with a textured coating for traction and sprinkled with a light coating of sand. We warmed hatchlings to activity body temperature with a thermal gradient (75-W basking bulb) prior to trials. For each trial, a hatchling was placed on the track and allowed to acclimate for several seconds. We then approached the hatchling with a short, rapid poke of the hand, and recorded the following behaviors: (1) Total distance run. We recorded the starting and stopping locations of the lizard, number of times around the track (usually they ran less than one lap, and never more than two laps) as well as the positions of all reversals to calculate total distance (including back-tracked distance). (2) Number of pauses during the run. (3) Number of times the lizard jumped up to escape. (4) Number of attempts to climb either the inner or outer wall of the track. (5) Whether the lizard ran in a zigzag pattern to escape. (6) Number of reversals. (7) Number of attempts to crouch during the run (usually at the end of the run, when the lizard stopped to crouch against the inner or outer wall of the track). (8) Whether the lizard attempted to bury in the sand during or at the end of the run. When we determined that the lizard had stopped running, we repeated the hand approach. Each lizard was approached 10 times in succession during the escape behavior trial.


To assess morphological effects, we measured allometric variation in body shape, calculated as residuals of ln (body mass) regressed on ln (SVL). This calculation is commonly used to quantify body condition in reptile studies (Jakob et al. 1996; Schulte-Hostedde et al. 2005), but may also reflect inherited structural differences among individuals involving size of bones, muscles, and organs in addition to fat deposits (see Discussion). For this reason we use the more general term “body shape” to describe this measure, to avoid implying that variation in this trait mainly tracks short-term changes in resource acquisition. Body shape was calculated for dams and sires at capture (i.e., before females had yolking follicles, which bias body shape estimates), for progeny at birth, and for progeny when recaptured as adults in the spring of the following year. Because miniaturization affects body shape (see results), we excluded miniaturized hatchlings from calculations of within-dam or within-sire averages of progeny body shape.

Throat color in U. stansburiana was used as a proxy for social strategy. Throat color is expressed only during the breeding season in adulthood. For dams and sires, throat color was scored weekly to get an accurate scoring. For progeny recaptured as adults, scoring may be less accurate, depending on when and how many times progeny were captured as adults. We began to recapture lizards soon after they emerged from winter brumation sites, and during that time throat color is not yet expressed. Color expression is codominant, such that a lizard with a full, orange throat is scored as oo, whereas a lizard with blue and orange on its throat is bo, etc. Genotypes were determined after evaluating all recorded observations of throat color phenotype for each lizard. Effects of throat color were statistically analyzed on three axes: orange, blue and yellow, according to how many copies (0, 1, 2) of each allele a lizard possessed (Sinervo et al. 2001). For analyses of correlations between social strategy and escape behavior, and for analyses of natural selection acting on throat color, we estimated progeny throat color by using mid-parent values (Lancaster et al. 2007). This was done because most mortality occurs during maturation, prior to the onset of throat color expression in progeny. However, we used observed throat colors (as adults) of surviving progeny for genetic correlation calculations.

Dorsal patterning was scored continuously along its two axes of variation, stripedness and barredness. A lizard with a chevron back pattern received a “0” on both axes. A dashed pattern received a “1” on the stripedness axis, whereas a striped lizard received a “2.” Striped and dashed lizards were scored as 0 for barredness. On the barredness axis, a laterally flattened chevron received a 1 for barring, whereas a completely barred lizard received 2. Intermediate scores of 0.5 and 1.5 were also used on both axes (see also Lancaster et al. 2007).

Escape behavior was partitioned into categories by the following method: within individual approach attempts, the number of jumps and climbs performed by the lizard were summed to form the variable “vertical,” indicating a vertically oriented escape behavior. Similarly, zigzags, reversals, and pauses were summed to form “evade,” indicating evasive escape behaviors. Crouches and burying attempts were summed to form “hide.” To approximate an average number of behaviors within each synthetic class of behaviors, we divided the summed values obtained for vertical and hide by two (because each resulted from summing two distinct behaviors—e.g., jumps + climb attempts) and values for evade were divided by three (because obtained by summing over three kinds of behavior). Distance run was analyzed without transformation. For all analyses of escape behavior except repeatability, we characterized an individual's escape phenotype (within each category) by summing behavior across the 10 approaches per trial. Distance run was averaged across the 10 approaches.


Statistical analyses were carried out in JMP version 6.0.2 and version 7.0 (SAS Institute, Inc., Cary, NC) using REML estimates. Analysis of fitness was performed on progeny born in 2006 (the year escape behavior trials were performed). In analyzing fitness, we were primarily concerned with testing effects of specific linear combinations of traits on fitness (evidencing alternative patterns of phenotypic integration). We were not specifically interested in predicting a response to selection: heritability of dorsal patterning is complicated (Lancaster et al. 2007) and escape behavior is not heritable (see below), so response to selection could not be predicted in any case. We also were not interested in measuring disruptive or stabilizing selection on individual traits. Because of these reasons, and primarily because the number of trait interactions we wanted to analyze was so large (interactions up the three way among two throat colors, two dorsal patterns, and four escape behaviors), precluding the power to calculate traditional selection gradients, we did not formally estimate selection gradients in assessing fitness. A canonical rotation analysis of correlational selection would reduce the number of parameters, but would not allow tests of directed hypotheses about specific types of trait combinations hypothesized to improve fitness (Blows and Brooks 2003). Instead, we started with a fitness model including all linear combinations (up the three-way) among y alleles, barredness, and each measured escape behavior, and in the same model, interactions between o alleles, stripedness, and each behavior. Backwards, stepwise elimination was performed to arrive at the final fitness model. Our final, reported fitness model was then compared with alternative fitness models (including models estimating traditional directional or correlational selection gradients) using AIC.

Predictor variables for the fitness analysis were averaged within dams and standardized to standard deviation (SD) units. Our response variable was the relative fitness of progeny, averaged within dams (her proportion of surviving offspring divided by the mean proportion of surviving offspring). Offspring fitness was averaged within dams to avoid pseudoreplication of offspring sharing the same dam, and therefore expressing the same maternal effect. Furthermore, by averaging offspring survivorship within dams, the resulting fitness estimates can also be ascribed to dams as the “offspring quality” episode of selection on females for their maternal effect traits. However, conclusions about the strength of selection based on ascribing offspring fitness to dams should be interpreted with caution (Wolf and Wade 2001). Because fitness was nonnormally distributed, we used the delete-one jackknife method to estimate standard errors and significance (Mitchell-Olds and Shaw 1987).

Heritability and repeatability of escape behaviors were calculated using REML estimates of variance components. We evaluated statistical significance of these components by evaluating the 95% confidence interval (C.I.), approximated in JMP following Satterthwaite (1946). The C.I.s of some variance components include negative values. Although variance is always positive, if the measured offspring behaviors are sometimes very different within factors (i.e., sires), this can lead to negative estimates of variance components. To calculate the heritability of escape behaviors, a full-sibling, half-sibling nested analysis was performed following Falconer and Mackay (1996), excluding miniaturized hatchlings. For repeatability of escape behavior, we counted each hand approach as a separate observation (10 observations per hatchling), and modeled escape behavior categories by Hatchling ID (random effect). We included miniaturized hatchlings in this analysis. Repeatability of body shape was calculated by running a regression of adult body shape on the body shape of the same individual as a neonate. Heritability of body shape was calculated as twice the slope of sire body shape regressed on egg size-corrected offspring body shape. Genetic correlation between body shape and throat color was calculated through sires as


(Falconer and Mackay 1996), with the standard error calculated using heritabilites derived from sire–offspring regression (Falconer and Mackay 1996). Repeatability and heritability of body shape, effects of egg mass and miniaturization on body shape, and sire effects on egg mass and body shape were analyzed using the combined 2004 and 2006 dataset. For all analyses of progeny averages within dams, the female's enclosure was included in the model as a random factor, to control for potential effects of shared sires or common social environment. We analyzed effects of sire and social environment effects using within-sire averages. Analyses of effects of experimental variation (miniaturization) on body shape were conducted following analysis of natural variation; effects were hypothesized to be in the same direction as in the natural variation, thus hypotheses for effects of miniaturization were one-tailed. The miniaturization experiment verifies that egg mass effects on body shape (in the natural variation) represent a true maternal effect and not genetic correlations with egg size genes inherited through dams. To further corroborate this, we tested the hypothesis that VM > 0 for offspring body shape (i.e., the maternal environmental component of offspring phenotypic variance is significantly positive) in SAS version 9.2 (© 2002–2008) using the procedure outlined by Fry 2004, p. 17–19, which compares the σ2D and σ2S given by the data to a model in which σ2D2S and therefore maternal effects and/or dominance are lacking. Although VM calculated this way cannot be unequivocally attributed to maternal effects, a significant VM would provide evidence for a maternal effect in conjunction with our other results, particularly the results of our experimental manipulation of egg mass (which confirms causation).

Reported error terms are standard errors. Models were checked to meet assumptions of normality and homoscedasticity of residuals, and noncollinearity of independent variables (by checking variance inflation factors). All lower-order terms are retained in models in which only higher-order effects are reported below. Estradiol concentrations were not normally distributed and were log-transformed for analysis. Reported P-values are based on two-tailed tests except where indicated.



We measured two alternative patterns of selection acting on escape behaviors, throat color, and dorsal patterning (Fig. 1). Female fitness arising from offspring survivorship was significantly positively affected by an interaction among vertical escape behaviors, o alleles, and a striped dorsal patterning in offspring (effect of standardized vertical behavior × mid-parent number of o alleles × stripedness = 0.36 ± 0.14, t= 2.63, n= 78, P= 0.01; Fig. 1A). Conversely, fitness was also significantly improved for females whose offspring exhibited evasive escape behaviors, y alleles, and barredness (effect of standardized evasive behavior × mid-parent number of y alleles × barredness = 0.68 ± 0.32, t= 2.16, n= 78, P= 0.03; Fig. 1B).

In the same model, we also detected selection against hiding behaviors in interaction with the o allele (effect of hiding behaviors ×o alleles on fitness =−0.67 ± 0.22, t=−3.11, P= 0.003). Similarly, we detected marginally positive selection for total distance run in combination with a striped dorsal pattern (effect of total distance run × stripedness = 0.34 ± 0.17, t= 1.98, P= 0.05).

To ensure that the fitness model we obtained via backwards elimination was a better fit than simpler or alternative models, we compared the AIC from our model (AIC = 38.28; lower AIC is better) to a model that included only second-order interaction terms (also no quadratic terms; AIC = 51.30), a model containing only linear terms (65.52), and a model containing first-order, quadratic, and second-order interaction terms (i.e., a traditional model for estimating correlational selection gradients; AIC = 73.64). From this comparison, we were able to determine that the third-order interactions reported above represent better predictors of offspring fitness than do lower-order trait interactions, traditional correlational selection gradients, or directional selection gradients for offspring traits.


Throat color and dorsal patterning are both heritable (Sinervo et al. 2001; Lancaster et al. 2007). Escape behavior was repeatable within hatchlings, but not heritable. Within-hatchling repeatability (r) was low but significant for all measured behaviors (vertical escape behavior: r= 0.12 ± 0.01, P < 0.05; evasive behavior: r= 0.19 ± 0.02, P < 0.05; hiding behavior: r= 0.16 ± 0.02, P < 0.05; total distance run: r= 0.22 ± 0.02, P < 0.05). Lack of heritability was indicated by a nonsignificant between-sire component of variance for all behaviors (values represent 95% C.I. for the within-sire variance, followed by the total phenotypic variance [in brackets]: vertical behaviors, σ2sires=−0.022 to 0.12 [0.76]; evasive behaviors, σ2sires=−0.24 to 0.02 [1.72]; hiding behaviors, σ2sires=−0.01 to 0.30 [1.39]; distance run, σ2sires=−129.6 to 53.15 [803.98]). This pattern of repeatability of behaviors exhibited at birth, in the absence of heritability or opportunity to learn, suggests that environmental/maternal developmental processes organize escape behavior phenotypes (see Results below).


Body shape variation affected expression of hatchling escape behaviors. In the natural variation, vertical escape behaviors were negatively correlated with body shape (effect of within-dam average progeny body shape on vertical behaviors: estimate =−1.77 ± 0.86, F1,75= 4.29, P= 0.04; Fig. 2A); progeny with long, narrow body shapes tended to perform more vertical escape behaviors. Because observed vertical behaviors were potentially drawn from a nonnormal distribution, we reran this test using nonparametric measures and obtained a marginally significant Spearman ρ=−0.20, P= 0.07 for the relationship between body shape and vertical behaviors. Evasive escape behaviors were positively correlated with body shape (estimate = 3.42 ± 1.39, F1,74= 6.08, P= 0.02; Fig. 2B); progeny with a fuller body shape performed more evasive escape behaviors.

Figure 2.

Effects of natural and experimental body shape variation on escape behavior. (A) A long, lean body shape is associated with vertical escape behaviors such as jumping and climbing. (B) A heavy-bodied, stocky shape is associated with evasive maneuvers such as reversals, pauses, and zigzag movements. Miniaturized hatchlings excluded from (A) and (B). (C) After experimental egg mass manipulations, miniaturized eggs produced progeny more likely to perform vertical behaviors and less likely to perform evasive behaviors. Average body shape is 0, with stocky hatchlings >0 and lean hatchlings <0.

Results of experimental manipulations confirmed that these correlations are directly due to allometric, functional effects of morphology on behavior, rather than genetic effects on trait correlations such as linkage or pleiotropy. A repeated measures multivariate analysis of variance on evasive versus vertical behaviors with individual hatchlings as subjects revealed that miniaturized hatchlings (this manipulation results in hatchlings with a leaner shape, see below) were more likely to perform vertical behaviors, and less likely to perform evasive behaviors (within-subject effect: F1,225= 3.22, N= 352, Pone-tailed= 0.04; Fig. 2C). The dam's ID was included as a random factor to adjust for maternal effects, such as egg size, that could affect the relative magnitude of effect of miniaturization on body shape, but was not significant (F95,255= 1.14, P= 0.21).

No escape behaviors were correlated with yolk estradiol, dorsal patterning, or midparent values for throat color alleles, which were eliminated from the final models. Nonsignificant effects of yolk estradiol on escape behaviors are as follows: effect of log(average estradiol within females) on within-dam average total distance run (excluding miniaturized hatchlings) = 1.84 ± 2.06, P= 0.38, n= 62. Effect on evasive behaviors: 0.07 ± 0.14, P= 0.59. Effect on hiding: −0.06 ± 0.11, P= 0.55. Effect on vertical behaviors: −0.05 ± 0.06, P= 0.39. Distance run and hiding behaviors did not correlate significantly with any measured predictor variables.


Progeny neonatal body shape (nonminiaturized progeny averaged within dams) was positively correlated with the female's average egg size, a maternal effect (estimate = 0.44 ± 0.13, F1,85= 11.88, P= 0.0009; Fig. 3A); hatchlings from larger eggs tended to be more full-bodied than hatchlings from smaller eggs. This direct, functional effect of egg size on body shape was confirmed with our experimental data. Hatchlings from experimentally miniaturized eggs were leaner than nonminiaturized hatchlings (mean body shape of miniaturized hatchlings is −0.04 ± 0.005; mean for nonminiaturized = 0.02 ± 0.004; t= 9.48, N= 1364, Pone-tailed < 0.0001; Fig. 3B), confirming that the effect of egg size (a maternal effect) on hatchling body shape is causative. As further corroboration that maternal effects direct body shape variation, we identified that VM (phenotypic variance due to maternal effects) contributing to body shape of nonminiaturized hatchlings is significantly positive (VM= 35.7% of the total variance. Likelihood ratio test for VM > 0: χ2= 5.54, df = 1, P= 0.02).

Figure 3.

Egg size effect on body shape. (A) A female's egg mass allocation decision affected body shape at birth of her (nonminiaturized) hatchlings. (B) Experimental variation in egg mass also affects neonatal body mass, with miniaturized eggs producing leaner lizards. (C, D) Body shape is repeatable throughout life for nonminiaturized (C) and miniaturized (D) lizards. Average body shape is 0, with stocky hatchlings >0 and lean hatchlings <0.

Moreover, body shape is repeatable throughout life. Body shape repeatability for nonminiaturized lizards was 0.77 ± 0.21 (F1,70= 13.47, P= 0.0005; Fig. 3C). Miniaturized lizards, which are leaner at birth, are also leaner than nonminiaturized lizards at adulthood (mean adult body shape of miniaturized lizards =−0.045 ± 0.026; mean for nonminiaturized = 0.021 ± 0.018; t= 2.13, N= 108, Pone-tailed= 0.02; Fig. 3D). This suggests that maternal egg size effects on offspring body shape persist into adulthood and are not attenuated by compensatory growth during maturation.

We also detected significant sire effects on offspring body shape, both directly and indirectly, via sire effects inducing females to adjust their maternal effect allocation decisions. Females lay differently sized eggs according to their reproductive strategy type; yellow females lay large eggs, orange females lay smaller eggs (Sinervo et al. 2000b). Here we found that females also plastically adjust their egg sizes according to the throat color of the male with whom they have mated (effect of sire y alleles on within-sire average egg size: estimate = 0.02 ± 0.007, t= 2.32, N= 110, P= 0.02; Fig. 4A). In addition to evaluating sire throat color, females also assess sire body shape in making egg size allocation decisions (effect of sire body shape on within-sire average egg size in the same model: estimate = 0.18 ± 0.06, t= 3.00, P= 0.003; Fig. 4B). We found significant effects of sire body shape on offspring neonatal body shape (effect of sire body shape: estimate = 0.21 ± 0.06, F1,111= 13.31, P= 0.0004), which were partly explained by indirect effects mediated by maternal adjustment of egg sizes in response to the sire's appearance. However, we also calculated a significant heritability of body shape from sires to progeny after correcting progeny body shape for egg size effects (using residuals of within-sire offspring body shape from a regression on within-sire egg size of hatched progeny): h2sires-body  shape= 0.29 ± 0.11, F1,110= 7.56, P= 0.007. Throat color (y alleles) in sires affects body shape only indirectly because the two traits do not appear to be genetically correlated (rA= 0.07 ± 0.33).

Figure 4.

Sire phenotypes as environmental cues for maternal egg mass allocation decisions. (A) Females increase egg mass in response to increasing amount of yellow on the sire's throat. (B) Females also increase egg mass in response to a stockier sire body shape.



We found evidence for two distinct patterns of adaptive phenotypic integration in our polymorphic population of U. stansburiana. Progeny produced in our laboratory crosses were more likely to survive to adulthood when they possessed specific combinations of throat color (which indicates social strategy), dorsal patterning, and escape behavior trait values (Fig. 1).

Specifically, orange-throated progeny experienced high fitness when striped (see also Lancaster et al. 2007 for similar results from 2004) and if they performed vertically oriented escape behaviors (Fig. 1A). In other reptile species, striping patterns commonly correlate with running long distances or escaping in a straight line along the ground (Jackson et al. 1976; Brodie 1989), and this is thought to occur because striping impairs the ability of predators to judge rates of prey acceleration and speed (Jackson et al. 1976). As in these previous studies, we detected selection favoring striped lizards that ran farthest. A similar mechanism may explain the adaptive combination of striping and jumping/climbing reported here if striping also causes predators to fail to anticipate the rate of ascension. Orange males usually control the highest quality territories, which are predominantly rocky with numerous elevated crevasses (Calsbeek and Sinervo 2002a,b). In this microhabitat, vertical escape behaviors are likely highly effective against coachwhip attack, which usually occurs from along the ground, below the height of these elevated refuges (Jones and Whitford 1989; L. T. Lancaster and B. Sinervo, pers. obs.).

Yellow-throated progeny were selected for barring (see also Lancaster et al. 2007) in combination with evasive maneuvers (Fig. 1B). A barred color pattern in combination with evasive maneuvers is also adaptive in other reptile species (Brodie 1989), and this may be attributable to enhanced crypsis of barredness in grass (Brodie 1989). Yellow-throated males are nonterritorial and spend most of their time sneaking through the grass or basking on small, isolated stones (Calsbeek and Sinervo 2002a). Yellow-throated lizards might be predicted to adopt a strategy of evasion and crypsis, as they often lack access to refuges and are highly unlikely to outrun their much faster, visually acute, coachwhip predators (Jones and Whitford 1989). Whether differences in territoriality and microhabitat usage between orange and yellow-throated females are related to alternative escape strategies (Vanhooydonck et al. 2002; Schulte et al. 2004) is currently under investigation.

Additionally, yellow- and orange-throated lizards may have adopted these particular alternative escape strategies due to constraints on body shape in their respective, strategy-specific reproductive functions (see below). Most likely, as described below, social strategies and antipredatory strategies are tightly coevolved in this system. Hiding behaviors measured in captivity were universally selected against in the wild and were particularly detrimental in orange-throated lizards. The reasons for this are not clear, but may be explained by the artificial aspects of the circular track, which offered no refuges. In general, hiding behaviors such as crouching may draw a predator's attention to the lizard's location by performing the extra movements involved in hiding behavior, when simply freezing in place is more effective (Ioannou and Krause 2009).


Escape behavior differences among sibling-groups directly resulted from differences in body shape, with leaner lizards more likely to jump or climb to escape a predator (Fig. 2A), and heavier-bodied lizards more likely to perform evasive maneuvers such as pausing, zigzagging, and reversing (Fig. 2B). The fact that this effect of body shape on behavior could be induced through the experimental manipulation of body shape (Fig. 2C) indicates that functional morphology rather than genetic correlation is responsible for this phenotypic correlation. Many previous studies have noted that differences in body shape in reptiles among species, or among populations within a species are often correlated with differences in locomotion (Vanhooydonck et al. 2002; Husak and Rouse 2006; Goodman 2007) and/or escape behaviors (Schulte et al. 2004; Goodman 2007; Gifford et al. 2008). However, direct effects of morphology on behavior, whereas nearly universally assumed, have rarely been confirmed with experimental manipulations of body shape (such as our miniaturization effects). Effects of within-population variation in body shape on alternative escape behaviors have rarely been reported for terrestrial vertebrates (but see Bulova 1994), but studies on within-population covariation of body shape and sprint speed (Calsbeek and Irschick 2007) suggest that corresponding effects of within-population variation in morphology on alternative escape behaviors may also be common. The reasons why morphology directly affects locomotor and escape behavior are largely unknown, but may relate to allometric effects of stockiness on development of motor neurons or key muscles and skeletal features functioning in these behaviors (Johnson et al. 1993). No escape behavior measured in our study correlated directly to throat color, suggesting that functional morphology rather than a pleiotropic or otherwise correlated effect of alternative social behaviors on escape behavior (Sih et al. 2004a,b) is primarily responsible for interindividual variation in escape behavior in our study species.


Natural and experimental variation in body shape were both repeatable across life stages (from birth through adulthood), suggesting that our measure of body shape as residuals of mass on SVL reflects underlying structural differences—some individuals are inherently stockier than others (Fig. 3C,D). In many reptile studies, body mass residuals or other statistical indices of body shape are used as a proxy for an animal's condition, which is expected to reflect current resource availability, the intraspecific competitive environment, and the organism's potential to win those resources (Jakob et al. 1996; Green 2001; Schulte-Hostedde et al. 2005). However, our results of lifetime repeatability were consistent when measured within two different cohorts of lizards (those born in 2004 and maturing in 2005, and those born in 2006 and maturing in 2007). In addition, we calculated significant sire–offspring heritability for body shape (after correcting for sire effects on egg size). Together, these results indicate that body mass residuals (or other statistical measures of body shape) in part reflect genetic variation in body shape rather than condition, which is a trait more likely to track short-term environmental changes. Body shape is known to be highly evolutionarily labile in lizards, with differences in structural “stockiness” evolving rapidly within genera (Bergmann et al. 2009), further supporting the idea that genetic variation for stockiness should exist within populations and species. This suggests that genetic variation in and lifetime repeatability of body shape should be accounted for when using mass residuals or other body shape indices as a proxy for condition.

In addition to a heritable component to body shape, egg size (a maternal effect) has allometric effects on body shape (Fig. 3A,B). Smaller eggs result in skinnier hatchlings, which remain leaner throughout life, whereas larger eggs result in stockier offspring. Experimental manipulation of egg size, and calculation of a significant VM for body shape both support the conclusion that offspring body shape is directly affected by a maternal egg size effect. Egg size is determined by female reproductive strategy (Sinervo et al. 2000b), but we have shown here that it is also plastically affected by females’ perceptions of sire phenotypes. Sires with y alleles and stocky sires (heritable traits advantageous in progeny hatching from larger eggs) induced females to increase investment into egg mass (Fig. 4). Females cannot change the throat color of their offspring, but they can visually identify this trait in sires, and alter offspring morphology (and thus escape behaviors) to match inherited y alleles from themselves and their sire.


Other, previously studied maternal effects operating in this population impact the patterns of maternal effects on phenotypic integration reported here in two ways (Fig. 5). First, maternally derived estradiol causes the appropriate dorsal patternings to develop in orange- versus yellow-throated progeny, which can override potentially inappropriate heritable dorsal patterns and ensure that progeny express high fitness throat color/dorsal pattern combinations (Lancaster et al. 2007). Females modulate yolk estradiol in response to social cues, including being crowded by orange-throated neighbors and/or being mated to yellow-throated sires (Lancaster et al. 2007). Second, female U. stansburiana adjust egg size in response to elevations in the stress hormone corticosterone (Lancaster et al. 2008), and these adjustments improve survival of offspring. Specifically, yellow-throated females are more likely to increase offspring size when under stress, and larger size increases survival when dams are yellow-throated (reported in Lancaster et al. 2008), likely in part due to effects on antipredator behaviors described here.

Figure 5.

Summarization of results. Sire traits (Column 1) and dam traits (column 2) influence maternal effect traits (boxed traits in column 2). Maternal effects promote expression of offspring traits (behavior and dorsal pattern; column 3) that work well with inherited offspring traits (throat color and dorsal pattern; column 3). Hollow arrows indicate connections involving adaptive plasticity. Solid arrows represent heritability (or survivorship in the case of arrows leading to fitness). Black arrows indicate results of this study, whereas gray arrows indicate previously reported results (see text for details).


We conclude that maternal effects adaptively facilitate alternative patterns of phenotypic integration (i.e., reliable formation of complex, alternative strategies) when sexual signals predict traits offspring are likely to inherit, and when recombination between alternative, well-integrated phenotypes would otherwise break apart advantageous combinations of traits. When females experience sire traits as environmental cues, they can translate these cues into maternal effects on offspring traits that are adaptive in combination with traits inherited from herself and the sire.

Our results also putatively explain why particular mating strategies (expressed in males) are genetically correlated with particular life-history strategies (expressed in females). In U. stansburiana, inheritance of the y allele is associated with a “sneaker” strategy in males, and with a “K-strategist” life-history strategy in females. A similar pattern of laying large eggs being associated with nonterritorial males and females was reported in skinks by Stapley and Keogh (2005), suggesting that this association between sneakers and K-strategists may also exist in other lizard taxa. There is no a priori reason to hypothesize an association between the sneaky male-mating strategy and the large-progeny producing female life-history strategy. However, our results tentatively suggest that these disparate strategies may commonly necessarily be correlated due to the direct, functional relationships between egg size and behavior. Morphological constraints and predator regimes likely play a strong role in which social strategies can develop in each sex, and how social strategies on one sex may constrain or define strategy spaces available to the other sex (Forsman and Shine 1995; Rice and Chippindale 2001). Future studies on the role of body shape in alternative male and female social behaviors will be illuminating in this regard.

In general, the selective forces maintaining alternative, polymorphic phenotypes within populations can be various and complex. This is because organisms function in many contexts. Most aspects of organismal performance in a single context require the integration of multiple traits, some of which also function in other contexts. As a result, selection operating on an integrated suite of traits in one context will also affect function in other contexts (Roff and Fairbairn 2007; Wolf et al. 2007; Sinervo et al. 2008). Even when a polymorphism could be maintained by disruptive or frequency-dependent selection in one context, this may not be sufficient to swamp directional selection (favoring a single, most fit type) on associated traits that function in other contexts. Theoretical treatments, with their highly simplified conditions, have commonly modeled the maintenance of polymorphisms without considering these multiple contexts of organismal performance. In contrast, it is likely that in most observed natural polymorphic populations, a complex pattern of correlational and disruptive selection gradients acting across contexts is required to maintain alternative types.

In this case, U. stansburiana have commonly been studied for their alternative social strategies, which are maintained in the long term by frequency- and density-dependent selection (Sinervo and Lively 1996; Sinervo et al. 2000b). Here we show that these social types also exhibit alternative, holistic antipredator phenotypes involving dorsal patterning, throat color, and escape behavior. Selection for alternative antipredator phenotypes caused by predator switching behavior (Bond 2007) or differential survival in varying microhabitats (i.e., rock outcrops vs. grass) has also been shown to maintain stable polymorphisms within populations (Hedrick 2006). Therefore, it would be difficult to conclude whether frequency-dependent selection on social strategy or selection for alternative antipredator strategies is primarily responsible for the origin of alternative types in U. stansburiana. It is possible that both forms of selection, occurring across contexts, are actually required for the stable maintenance of polymorphism in this and other species of real organisms inhabiting a complicated world.

Associate Editor: J. Wolf


We thank C. Hipsley, L. Hazard, S. Mills, D. Haisten, and numerous undergraduates for assistance with the lizards. We thank J. Wingfield and L. Erckmann for assistance with hormone assays and providing radioimmunoassay reagents and supplies. We also thank M. Waldman, M. Moy, C. Preston, and I. McFadden for assistance with behavioral trials; B. Bastiaans, A. Corl, A. Davis, C. Hipsley, B. Lyon, J. Mueller, S. Nunes, and J. Thompson for comments on the manuscript; the Arbeilbedee family for permission to work on their land; and an NSF grant to BS and AM.