Transgressive aggression in Sceloporus hybrids confers fitness through advantages in male agonistic encounters




1. We investigated agonistic behaviour and associated characteristics of Sceloporus woodi (Florida scrub lizard), Sceloporus undulatus (Eastern fence lizard) and their hybrids using staged territorial encounters.

2. These Sceloporus hybrids exhibit transgressive aggression and transgressive head-girth relative to the parental species and the transgressive aggression was specifically associated with an advantage in agonistic encounters. Our results suggest a hybrid advantage in natural habitats when defending and invading territories against either parental species.

3. We further analysed general advantages in agonistic encounters across the entire three-group system to elucidate characteristics that may be advantageous under specific circumstances. Individuals with larger body size (SVL) and greater aggression had an overall advantage in agonistic encounters; however, smaller individuals could win when slightly more aggressive and fatter, and less aggressive individuals could win when slightly larger, especially with greater head-girth.

4. The extreme hybrid phenotypes likely occurred through transgressive segregation, which has been implicated as a process through which homoploid, hybrid speciation can occur. Some form of ecological divergence is necessary, however, to impede parental gene flow. Our data suggest that ecological divergence could manifest in territorial species through transgressive aggression.


Although hybridization is commonly thought to produce individuals with intermediate phenotypes and low fitness, hybrids often exhibit phenotypes that resemble one parental species more than the other or are extreme relative to both parental species (Smith & Riechert 1984; DeVicente & Tanksley 1993; Rieseberg & Ellstrand 1993; Cosse et al. 1995; Mallet 2007; Scarelli-Santos et al. 2007), and can exhibit higher fitness (Rieseberg et al. 2003; Kirk, Vrieling & Klinkhamer 2005; Rhode & Cruzan 2005). In fact, hybridization resulting in extreme phenotypes may be the general rule rather than the exception (Rieseberg, Archer & Wayne 1999). Extreme phenotypes are termed transgressive and generally occur through the segregation of complementary alleles. These alleles are often not found together in either parental species because of divergent selection for suites of alleles in mono or polygenic traits, but are then found together after recombination in hybrids (Rieseberg et al. 1999; Burke & Arnold 2001; Mallet 2007).

Transgressive segregation can produce new and unexpected phenotypes that often confer fitness benefits for hybrid organisms (heterosis or hybrid vigour) and have evolutionary consequences (Schwarz et al. 2007). Hybrid vigour can permit hybrid lineages to colonize new niche space or allow speciation through parallel selection pressures and divergent natural selection (Gross & Rieseberg 2005). For hybrid speciation to result, that divergent natural selection must persist within the hybrid lineage through reproductive isolation resulting from geographical or ecological processes. Alternatively, hybrid vigour can result in hybrid invasion of the existing niches of either parental species causing extirpation or even extinction through introgression (Rhymer & Simberloff 1996). Transgressive segregation is most common in intraspecific crosses of plants and in domesticated, inbred lineages of both plants and animals (Fernandez-martinez et al. 1986; Rieseberg & Ellstrand 1993; Clarke et al. 1995; Maluf, Pereira & Figueira 1997; Rieseberg et al. 1999). The phenomenon has been demonstrated less frequently in studies of wild animals (Economidis & Sinis 1988; Harrison & Hall 1993; Walker, Taylor & Cordes 1994; Giessler 1997; Reichert, Singer & Jones 2001) and rarely in regard to behavioural traits (Rieseberg et al. 1999).

We investigated morphological and agonistic behavioural traits exhibited by hybrids of Sceloporus lizards (Sceloporus undulatus Bosc and Daudin subspecies undulatus and Sceloporus woodi Stejneger) from an established, viable hybrid zone (Jackson 1973). Preliminary genetic analyses, using six previously published microsatellites for S. woodi (Ernst et al. 2004), indicated 95% of 44 individuals from the hybrid zone were consistent with hybrid lineage and suggested a hybrid swarm that included high levels of backcrossing to S. undulatus.

We focused on male agonistic encounters. Lizards of the genus Sceloporus are excellent study organisms in which to examine agonistic encounters because they are relatively easy to care for in the laboratory, much is known about agonistic behaviours of lizards during territorial encounters (Carpenter 1962; Vinegar 1975; Robson & Miles 2000; Perry et al. 2004), and investigating agonistic encounters using paired trials is relatively simple with common protocols (Perry et al. 2004; Jenssen, Decourcy & Congdon 2005). Furthermore, success in these agonistic contests over resources in lizards, especially in territoriality, has fitness consequences because dominant territorial males receive greater access to females by maintaining larger and/or better territories (Anderson & Vitt 1990; Hews 1990; Olsson 1992; Abell 1997; Haenel, Smith & John-Alder 2003a, b). Male territoriality in lizards also influences sexual dimorphism as a result of male–male combat (Anderson & Vitt 1990; Wikelski & Trillmich 1997; Cox, Skelly & John-Alder 2003). Thus, establishing and defending territories is paramount for male reproductive success in lizard populations. Vigour in the territorial behaviour of Sceloporus hybrids could facilitate hybrid persistence through sexual selection, increasing the potential for introgression into either parental species or hybrid speciation through ecological isolation.

Staging agonistic encounters between resident and intruding males, we investigated the display behaviours, associated morphologies, and potential success of territory defence and intrusion. Across the three-group system (S. undulatus, S. woodi, and hybrids), we also examined circumstances under which specific characteristics were advantageous. We were specifically interested in comparing Sceloporus hybrids to their parental species to determine whether hybrid traits were intermediate or transgressive, and when transgressive traits were found, we examined whether these traits were associated with advantages in agonistic encounters and, by proxy, fitness. We asked the following four main questions. (i) How well can hybrids defend and invade territories against the parental species?; (ii) Which hybrid phenotypes, if any, influence the outcome of agonistic encounters against the parental species?; (iii) Over the entire three group system – not using species as a partitioning factor to increase the variation and therefore inference across a larger Sceloporus system – which morphological and behavioural phenotypes, if any, are advantageous during agonistic encounters?; and (iv) Are there circumstances under which certain characteristics, or ranges in trait values, are more advantageous than others?

Materials and methods

Study animals

The Florida scrub lizard, S. woodi, lives in open scrub habitats on remnant Pliocene and Pleistocene sand ridges in central Florida. Open scrub habitats consist of sparse sand pines, oak shrubs and extensive bare ground. Sceloporus woodi occurs in disjunct, genetically divergent populations, commonly metapopulations, along the Florida ridge (Clark, Bowen & Branch 1999; McCoy, Hartmann & Mushinsky 2004) and is a rare species, although locally abundant (McCoy & Mushinsky 1992). The eastern fence lizard, S. undulatus, is common in the south-eastern USA (Conant 1975) and abundant in sandhill habitats of central Florida. Sandhill habitats consist of long-leaf pines, turkey oaks and ground cover of wiregrass and fallen pine needles (Myers 1992). The hybrids zones, many of which occur in Florida (Jackson 1973), are generally narrow ecotonal regions between relatively extensive sandhill habitats and patchy scrub habitats.

Collection and housing

Adult male lizards (= 57) were collected from their respective habitats between September 2005 and August 2006. Average snout–vent length (SVL) of each parental species (see Results) was consistent with adult male sizes of S. undulatus and S. woodi in Florida, 57·0 mm and 47·5 mm respectively (Jackson 1972; Mobley 1998), confirming that adults were used in our experiment. Individuals of S. undulatus (= 18) were collected in sandhill habitat (N 29°02′18′, W 81°33′35′), individuals of S. woodi (= 19) were collected in open sand-pine scrub habitat (N 29°06′29′, W 81°48′34′; as described in Myers 1992) and hybrids (= 20) were collected in a known hybrid zone (Jackson 1973) along one ecotonal region in the Ocala National Forest, 3·2 km south-west of Alexander Springs (N 29°03′58′, W 81°35′00′). The hybrid zone was c. 4 km long and 1·5 km wide (see Jackson 1973 for map). Lizards were captured with a noose and taken to the laboratory. We measured SVL and vent–tail length (VTL) with a ruler to the nearest 0·5 mm, mass with an electronic balance to the nearest 0·01 g and head-girth across the widest point of the head with digital callipers to the nearest 0·01 mm. Each lizard was given a unique toe clip for identification (Waichman 1992), housed individually, and provided fresh water and crickets daily. Containers (30 × 17 × 12 cm) included a sand substrate, one water dish, and a refuge, on top of which lizards could also bask. Heat lamps maintained temperature gradients within containers that averaged 31 °C during the daytime portion of a 12/12 h day/night cycle.

Staged agonistic encounters

Males from both parental species and hybrid males were observed in various pairings during staged agonistic encounters. Members of each pairing were chosen at random from the individuals housed in the laboratory. We conducted intraspecific trials between S. undulatus males (= 33 UU trials) and between S. woodi males (= 23 WW trials), and interspecific trials between males of the two parental species (= 26 UW trials with S. undulatus as the focal species and = 30 WU trials with S. woodi as the focal species; see below). We considered these trials’ baseline references to which we could compare trials between hybrids and parental species and as indicators of potential observer effects on behaviours. After establishing a baseline, we conducted trials between hybrid males and males of the two parental species (= 39 HU trials with S. undulatus as the parental species and = 34 HW trials with S. woodi as the parental species) to examine the relative strength of hybrids as territorial residents and intruders. Hybrid males were the focal individuals in these final two sets of trials. Within each of the six sets of trials, individuals of the focal group were observed once each as a territorial resident and once each as a territorial intruder. Some trials were excluded because the opponent’s tail had broken prior to the encounter. Our use of focal individuals means that the outcome of a trial was judged in only one direction. Avoiding reciprocal pairings, and, consequently, a round robin design, allowed us to consider each pairing (= 185 total trials) as an experimental unit and to employ standard anova models (Malloy & Albright 2001). When opponents were used more than once in trials between hybrids and parental species, we compared the proportion of wins between opponents used once and those used more than once with a binomial test to examine individual bias.

Trials occurred within 5 weeks of capture for all lizards and were conducted between 17.00 and 20.00 hours, during the daytime of our light/dark cycle. Males were given a minimum of 24 h between trials. Trial enclosures (36 × 60 × 51 cm glass terraria) had a 3-cm sand substrate, two pieces of pine bark situated similarly on each side, a visual divider in the centre and heat lamps positioned at each end that maintained temperature gradients in the enclosures averaging 31 °C. Males were considered established residents after a 1-week acclimation period in one side of the trial enclosure (Jenssen et al. 2005). The resident side of the trial enclosure was rotated to avoid side bias. During the week of acclimation, individuals were cared for as described previously for general housing. Intruder males were taken directly from general housing, placed in an unfamiliar side of the trial enclosure adjacent to the resident males and given a 30-s acclimation period before the visual divider was removed between the intruder and resident. Trials were conducted in a dark room with the only lights (heat lamps) positioned over the trial enclosure, and observations were made from 3 m away to minimize observer influence on behaviour. The territorial behaviours exhibited during intraspecific encounters were those that were expected from previous studies (see Table 1, Figs 1 and 2; Carpenter 1962; Vinegar 1975; Robson & Miles 2000; Perry et al. 2004); so, we have confidence that observer influence was minimized during encounters between hybrids and parental species.

Table 1.   Species-specific and hybrid advantages in agonistic encounters
EffectHU (= 39)HW (= 34)UU (= 33)UW (= 56)WW (= 23)
  1. Results are shown for the hierarchical loglinear analyses among winners from each trial type. Statistics are for parameter estimates resulting from the models. Effects are reported with respect to the winner of the encounter. R/I is for resident/intruder, size for snout–vent length being either ‘equal to and smaller than’ or ‘larger than’ the opponent, and group refers to the identity being a parental species or hybrid depending on trial type tested. Trial type is labelled at the top of the columns with sample size in parentheses. The WU and UW trial types were pooled as UW because there was not a focal individual in regard to morphological characteristics. Bold text represents influential effects.

R/I status × size × group  0·720·471−0·260·798  0·240·809
R/I status × size−0·060·956−0·260·7980·170·868  0·240·8090·340·735
R/I status × group−1·150·249  0·050·958−0·270·785
Size × group−0·790·428  0·760·448  0·840·403
R/I status  2·250·024  0·050·9581·420·156−0·270·7851·390·164
Size  1·890·059  0·760·4482·390·0170·840·4631·760·079
Group  2·050·040  3·72<0·0014·38<0·001
Figure 1.

 Estimated marginal mean scores from mancova on specific behaviours during agonistic encounters among winners and losers of parental species and hybrids. Resident/intruder status and group were factors. Covariates included were differences between the focal individual and opponent in snout–vent length (SVL) and in mass, vent–tail length, and head-girth adjusted for difference in SVL. These means were used to calculate the proportion that each behaviour made up of the total behavioural repertoire for each species and hybrids. Grey bars (bsl00036) represent hybrids, black bars (bsl00001) Sceloporus undulatus and white bars (□) Sceloporus woodi.

Figure 2.

 Aggression measured as behavioural scores for residents and intruders within agonistic trial types. Points are estimated marginal mean scores from ancova (Table 3) and error bars represent 2 SE. Trial types are listed as focal group vs. opponent (U = undulatus, W = woodi and H = hybrid). Means are estimated for the focal group. Open circles (○) represent intruders and filled circles (•) residents. Intruders are shifted to the left to eliminate overlapping points and error bars.

Each trial was scored visually for 10 min, during which aggression was measured only for the focal lizard as a total behavioural score. The opponent characterized the treatment applied to the focal group and was not used otherwise in calculating the behavioural scores for the focal individuals. Positive values were awarded for aggressive displays (1 = bite, 1 = charge, ¼ = pushup, 1 = vibrate) and negative values were awarded for submissive behaviours (1 = fleeing, 1 = jumping away, 1 = flattening out, 1 = burrowing) (Carpenter 1962). Pushups were scored lower because they are aggressive displays of high frequency but low intensity (Vinegar 1975; Perry et al. 2004). The total behavioural score for each focal male was the sum of aggressive and submissive values. One focal male was scored per trial, but general aggression and/or submission was recorded for both individuals to directly deem the loser of each encounter as the lizard that ended the encounter by burrowing or fleeing without returning. We then examined the influence of aggression and body size on this independent determinant of outcome.

Data analyses

Differences between parental species

To examine species-specific advantages, we used hierarchical loglinear contingency table analyses. For interspecific trials, we used a three-way analysis, tabulating winners contingent upon the factors of R/I status (resident or intruder), relative size in SVL (‘equal size and smaller’ or ‘larger’) and group identity (S. undulatus or S. woodi). The size categories included ‘equal size and smaller’ or ‘larger’ because we were interested in group-specific advantages. If an individual that was equal size or smaller won, it did so without the advantage of being larger, which means that R/I status or group influenced the outcome, not size difference. Species-specific advantages cannot occur in the intraspecific trials, so we used two-way analyses, tabulating winners contingent upon only R/I status and relative size in SVL.

Morphological characters that were compared between species included SVL (mm), mass (g), VTL (mm) and head-girth (mm). These comparisons were not specific to trials, but merely comparisons between groups. No transformations were necessary for the morphological data. Differences in SVL were examined using anova with group as a factor. To examine differences in mass (body condition), VTL and head-girth, mancova was used with SVL as a covariate and group as a factor. Tails that were broken and/or re-grown (= 11) were not used when comparing tail lengths.

Agonistic behaviours were analysed at two levels, as repertoires of individual behaviours and as total behavioural scores. Behavioural repertoires were calculated by summing mean scores of specific behaviours, separately for winners and losers, and standardizing them to sum to one. Behavioural repertoires were compared between groups using Kolmogorov–Smirnov two-sample tests to examine whether particular behaviours disproportionately accounted for total behavioural scores among groups. The mean scores of specific behaviours were estimated marginal means from mancova with each behaviour as a dependent variable and R/I status and group as factors. Covariates included in the model were differences between the focal individual and opponent in SVL and in mass, VTL and head-girth adjusted for difference in SVL (residuals from regression of each on difference in SVL). The mancova was used only to estimate mean values for the specific behaviours, not to test for differences between groups, because the behaviours are not necessarily independent.

Total behavioural scores, which measured aggression, were also compared between groups. We used total behavioural scores instead of factor scores from a principal components analysis (PCA) of specific behavioural scores for three reasons. (i) We did not find any particular behaviour to disproportionately explain any species (or hybrid) total score among winners or losers (see Results). (ii) Factor scores from the first axis of a PCA were significantly correlated with total behavioural scores (P < 0·001 with a correlation coefficient of 0·88). (iii) Total behavioural scores are easier to interpret than factor scores. Aggression was examined using ancova with total behavioural score as the dependent variable and trial type (group vs. group) and R/I status as factors. Covariates included in the model were the date of capture, to account for any seasonal difference in behaviour, and differences between the focal individual and opponent in SVL and in mass, VTL and head-girth adjusted for difference in SVL (residuals from regression of each on difference in SVL). Factors and covariates were chosen because they can affect communicative displays in lizards (Robson & Miles 2000; Perry et al. 2004).

Differences between hybrids and parental species

We compared hybrids against each of the parental species to examine advantages in agonistic encounters and the associated characteristics by employing the same analyses as those described above for the parental species. A generalized linear model (GLM) was then used to examine which specific characteristics were contributing most to winning. The GLM used win/loss as the binary-dependent variable and the logit link function to link the dependent and independent components (Quinn & Keough 2002).

To specifically link any transgressive hybrid phenotypes to any hybrid advantage, trials including hybrid winners and losers must be examined. Trials in which only hybrids won (i.e. HW trials) cannot be used because the effects of phenotypes cannot be separated from the effect of group. Therefore, a GLM was used within the HU trials to examine whether any transgressive hybrid phenotypes influenced the outcome of agonistic encounters. We used R/I status as a categorical factor, and as continuous factors we used those variables that were found to be different among the parental species and hybrids, which included aggression and differences between the focal individual and opponent in SVL, and in mass and head-girth adjusted for differences in SVL (residuals from the regression of each on difference in SVL). Aggression was adjusted to eliminate the influence of other variables by using residuals from an ancova model similar to that used to address differences among the species and hybrids (Table 3), except it was across all groups (trial type was not used as a factor). Adjusted behavioural scores represent an individual’s inherent level of aggression.

Table 3.   Characteristics that influence the aggression of male, Sceloporus lizards during agonistic encounters
  1. Aggression was measured as total behavioural scores. Two-factor ancova results are presented with covariates denoted by (cov). Covariates that were first adjusted by difference in snout–vent length (SVL) between the focal lizard and its opponent are denoted by (SVL adj. cov). R/I is for resident/intruder and trial type refers to the six types of pairings among specific groups. Bold text represents covariates and factors that explain behavioural scores.

Difference in SVL (cov)128985·0161·36<0·001
Difference in mass (SVL adj. cov)11702·033·600·059
Difference in VTL (SVL adj. cov)1128·710·270·602
Difference in head-girth (SVL adj. cov)11269·842·690·103
Date of capture (cov)14427·319·370·003
R/I status17682·6516·26<0·001
Trial type58661·9318·34<0·001
R/I status × trial type5295·460·630·681

Advantages of specific characteristics under certain circumstances

To examine what influenced winning agonistic encounters over the whole three-group system, we used a GLM with win/loss as a binary-dependent variable and factors including the categorical variable of R/I status and continuous variables of aggression and the differences in SVL, mass and head-girth between the focal individual and its opponent (Maynard Smith 1982; Jenssen et al. 2005). Differences in mass and head-girth were residuals from a regression of each on difference in SVL. Aggression was again adjusted to eliminate the influence of other variables and represents an individual’s inherent level of aggression. Appropriate interactions between variables were also included in our model of the three-group system, which resulted in 11 predictor variables (see Results). The full model was tested for the goodness-of-fit to the data using the Hosmer–Lemmeshow goodness-of-fit test. Akaike’s information criterion (AIC) values were calculated for all possible subsets of our model (2048 subsets) to find the best subset of variables (Quinn & Keough 2002). AICc, which is an AIC value corrected for small sample size relative to the number of parameters in the model, was also calculated and compared with the AIC values to ensure that our sample size was sufficient. The difference between the best AICc value and those of subsequent models (Δi) was calculated along with Akaike’s weight (wi) and the evidence ratio to determine which models adequately explained the data (Burhnam & Anderson 2002). To run the GLMs and choose the best subset, spss and statistica were used in concert (SPSS Inc. 2006; StatSoft, Inc. 2005).

We used a decision tree as a post hoc exploratory test to better define the level of influence each factor had on winning agonistic encounters and, more specifically, the ranges of trait values that were most influenced by these factors. The decision-tree analysis was used in a similar fashion to post hoc multiple comparison tests after finding significant differences using an anova (in our case, we used a GLM). We did not use a Bonferroni adjustment for branch significance in this post hoc exploratory analysis because the factors were previously shown to significantly influence the winning of territorial encounters. The main effects that persisted in the best subset GLM of the whole three-group system were used to build the decision-tree model (= 176 trials). Trials with missing data for any of the main effects were not used in the decision-tree analysis. The Chi-squared Automatic Interaction Detector (CHAID) method (Kass 1980; as cited in SPSS Inc. 2001) was used to grow the tree with the maximum tree depth set at 7 and the minimum number of cases for the parent and child nodes set at 7 and 3 respectively. The CHAID method merges statistically homogeneous values with respect to the response variable while maintaining the heterogeneous values within each factor and then selects the best factor as the next branch of the tree. In branch succession, this process occurs at each node until significant merges no longer exist or the maximum tree depth or minimum number of cases has been reached. The CHAID method uses F-tests or chi-squared tests for continuous or categorical response variables respectively (SPSS Inc. 2001). In our case, chi-squared tests were used because our response variable was win/loss. The decision tree was constructed using spss AnswerTree v.3.0 (SPSS Inc. 2001). For all other statistical analyses, spss v.15 was used, except where noted otherwise (SPSS Inc. 2006). Statistical significance was associated with P-values <0·05.


Differences between parental species

Our contingency table analyses showed a species-specific advantage, in that S. undulatus always defeated S. woodi (Table 1). Notably, larger S. woodi lost when paired with smaller S. undulatus (= 13). Sceloporus undulatus was longer in SVL and had more mass relative to SVL (greater body condition) than S. woodi (Table 2). VTL was similar, however, between parental species. Because no specific agonistic behaviour disproportionately influenced the total behavioural scores (Kolmogorov–Smirnov values for the comparison of behavioural repertoires: U vs. W, > 0·60; Fig. 1), total behavioural scores were used to measure aggression. Overall, aggression was greater in S. undulatus than S. woodi (Table 3, Fig. 2) in trials against similar opponents. Both S. undulatus and S. woodi had greater scores when opposing S. woodi (UW and WW) than when opposing S. undulatus (UU and WU; Fig. 2), resulting in intraspecific S. undulatus trials (UU) having similar scores to intraspecific S. woodi trials (WW). These results indicate that, within the overall trend, the opponent species plays a role in determining the aggression of the focal species, with S. woodi eliciting greater aggression from both species. Aggression of the focal lizard was also influenced by the difference in SVL, the difference in mass relative to SVL, R/I status and date of capture (Table 3, Fig. 2).

Table 2.   Differences in morphological characters measured among Sceloporus undulatus, Sceloporus woodi and hybrids
HybridS. undulatusS. woodiMSF2,42P
  1. Estimated marginal means are shown with results from the analyses of variance. anova was used to analyse snout–vent length (SVL) among groups and mancova was used to analyse mass, vent–tail length (VTL), and head-girth using SVL as a covariate. Significant differences are highlighted in bold. Similar superscript letters denote statistically similar groups.

SVL (mm)52· 0A52·5A47·0B178·4 9·51<0·001
Mass (g) 4·79AB 5·06A 4·42B  1·2 4·62  0·016
VTL (mm)70·0A64·0A69·5A136·7 1·66  0·204
Head-girth (mm)12·5A11·7B11·5B  3·7111·75<0·001

Differences between hybrids and parental species

A hybrid advantage in male agonistic encounters was seen in our contingency table analyses while accounting for R/I status and difference in SVL (Table 1). Hybrids defeated S. woodi in every encounter and defeated S. undulatus more than expected (26 of 39 total encounters). Notably, larger S. woodi lost when paired with smaller hybrids (= 8) just as they did when paired with smaller S. undulatus.

Overall, S. undulatus won 33% of the HU trials, with one individual winning 66% and others never more than 50% of the trials. Binomial tests indicated that an individual S. undulatus would need to win four of four trials to have won significantly (< 0·05) more than 33% of the HU trials. Also, the percentage of wins (33%) among S. undulatus individuals used in only one trial (= 6) was nearly identical to that of individuals used in multiple trials (35%; = 11). Sceloporus woodi won 0% of the HW trials, so individual bias was not relevant. Using some opponents more than once did not result in any obvious individual bias.

Both morphological and behavioural differences were detected between parental species and hybrids, including transgressive hybrid phenotypes. The transgressive morphological phenotype found in hybrids was head-girth. Hybrids had the widest heads relative to SVL (Table 2). Hybrids had similar length and mass to S. undulatus, but were longer in SVL and had a greater mass relative to SVL (greater body condition) than S. woodi (Table 2). No differences existed in VTL among parental species and hybrids. For behaviour, in our case aggression, the overall trend from highest to lowest was hybrids > S. undulatus > S. woodi, suggesting transgressive aggression in hybrids. Again, no specific behaviours were found that disproportionately influenced the overall behavioural scores (Kolmogorov–Smirnov values for the comparisons of behavioural repertoires: H vs. U, H vs. W, both > 0·60; Fig. 1), so transgressive aggression was found throughout all hybrid agonistic behaviours. The greater hybrid aggression had a pattern similar to that found in the parental species, which was dependent on the opponent species. Hybrids exhibited greater aggression against S. woodi than against S. undulatus and hybrids opposing S. undulatus (HU) had similar scores to S. undulatus that were opposing S. woodi (UW; Fig. 2). Again, these results indicate S. woodi eliciting greater aggression than S. undulatus. Similar to the aggression of the parental species, the aggression of hybrids was influenced by the difference in SVL, the difference in mass relative to SVL, R/I status and date of capture (Table 3, Fig. 2), reflecting the importance of addressing these factors when inherent species differences are the focus.

The transgressive hybrid aggression did explain the outcome of agonistic encounters (GLM Type III, < 0·001) against S. undulatus; however, the other transgressive phenotype, greater head-girth relative to SVL, did not. The factors of R/I status and difference in SVL also explained the outcome (P < 0·001 for both) of the HU trials, but differences in mass relative to differences in SVL (body condition) did not. Body condition was similar between hybrids and S. undulatus, which is likely why it did not affect the outcome.

Advantages of specific characteristics under certain circumstances

The outcomes of male agonistic encounters in the three-group system were explained by R/I status, aggression, difference in SVL, and difference in mass and head-girth relative to SVL. Note that, at this level of analysis, these characteristics were analysed across the Sceloporus system without any species grouping factor. Our full model with 11 factors explained the system well (Hosmer–Lemeshow, χ2 = 0·129, d.f. = 8, = 1). The model subset with the lowest AIC value, however, had nine factors (Tables 4 and 5) – the five main factors and four interactions – that significantly influenced who won an agonistic encounter among individuals in the three-group Sceloporus system. Models 1–4 (Table 5) were within a Δi of 2, and thus by convention meaningful differences were not evident among these models. The values for Δi in Table 5 were calculated using AICc (corrected AIC for small sample sizes), which was calculated to compare with AIC values. The similarities between AIC and AICc values suggest adequate sample sizes with respect to the number of parameters in the models (Burhnam & Anderson 2002). The low Akaike weight (wi) for the best model and low evidence ratios for all four models further suggest a weak argument for a single best model or a difference between the best four models respectively (Burhnam & Anderson 2002). The five main effects and the interaction between difference in SVL and aggression were found in all four best models, however, which suggests that these factors have the greatest influence on winning an agonistic encounter (Table 5).

Table 4.   Characteristics associated with winning agonistic encounters in the entire three-group Sceloporus system (Sceloporus undulatus, Sceloporus woodi and hybrids)
  1. Results of a generalized linear model with win/loss as the dependent variable (= 176). This is model 1 in Table 5 with the lowest Akaike’s information criterion value. Factors that were adjusted by differences in snout–vent length (SVL) between the focal lizard and its opponent are noted with (SVL adj.). R/I is for resident/intruder. Aggression was adjusted as described in the text.

R/I status30·61<0·001
Difference in SVL133·39<0·001
Aggression (adj.)145·61<0·001
Difference in mass (SVL adj.)28·08<0·001
Difference in head-girth (SVL adj.)24·56<0·001
R/I status × aggression5·800·016
R/I status × difference in head-girth (SVL adj.)7·170·007
Difference in SVL × aggression (adj.)11·900·001
R/I status × difference in SVL × aggression (adj.)6·360·012
Table 5.   Resulting Akaike’s information criterion (AIC) values for the best subsets of models describing characteristics associated with winning agonistic encounters in the entire three-group Sceloporus system (Sceloporus undulatus, Sceloporus woodi and hybrids)
  1. Win/loss was the dependent variable (= 176) in the generalized linear models. An X denotes inclusion of that variable in the model. Factors that were adjusted by differences in snout–vent length (SVL) between the focal lizard and its opponent are noted with (SVL adj.). R/I is for resident/intruder. Aggression was adjusted as described in the text. The difference between the best AICc value and those of subsequent models (Δi) was calculated along with Akaike’s weight (wi) and the evidence ratio.

R/I statusXXXXX
Difference in SVLXXXXX
Aggression (adj.)XXXXX
Difference in mass (SVL adj.)XXXXX
Difference in head-girth (SVL adj.)XXXXX
R/I status × difference in SVL XXXX
R/I status × aggression (adj.)X XXX
R/I status × difference in mass (SVL adj.)  X X
R/I status × difference in head-girth (SVL adj.)XX XX
Difference in SVL × aggression (adj.)XXXXX
R/I status × difference in SVL × aggression (adj.)XX XX
Evidence ratio0·641·031·611·99 

In this Sceloporus system, we found many circumstances under which specific characteristics were advantageous. Most main effects and some interactions that were found in the best experiment-wide GLM models appear in the decision-tree model (Figs 3 and 4); however, some do not. CHAID successively finds the most significant split in the dependent variable, which means that other significant splits in the data are overlooked at each branching of the tree, in effect ranking the factors by influence. From most to least influential, the ranking was SVL > aggression > mass and head-girth on the same level. Within this system, individuals that are more than 3 mm smaller than their opponent almost certainly lose, unless they are very aggressive and have greater mass relative to SVL (Fig. 3, Node 12). If an individual is between 3 mm smaller and 0·5 mm larger than its opponent, then it must be more aggressive to win, or at least balance aggression with the size difference (Fig. 3, Node 2). Individuals that are at least 0·5 mm larger generally win unless they are very submissive, in which case they can still win if they have greater head-girth relative to SVL (Fig. 4, Nodes 15 and 17).

Figure 3.

 Part of the decision tree depicting circumstances under which certain characteristics were advantageous during agonistic encounters within the entire three-group Sceloporus system (Sceloporus undulatus, Sceloporus woodi and hybrids). P-values were derived from chi-squared analyses of the Chi-squared Automatic Interaction Detector procedure. Node numbers are included for reference purposes. At each node, the per cent of the total experiment wide trials (starting at 100% at node 0) is in parentheses. Underneath each factor is the significance of its split at that node, and below each horizontal branch are those trait ranges that make up the homogeneous subsets with respect to the win/loss-dependent variable. See text for how factors were adjusted.

Figure 4.

 Part of the decision tree depicting circumstances under which certain characteristics were advantageous during agonistic encounters within the entire three-group Sceloporus system (Sceloporus undulatus, Sceloporus woodi and hybrids). P-values were derived from chi-squared analyses of the Chi-squared Automatic Interaction Detector procedure. Node numbers are included for reference purposes. At each node, the percent of the total experiment wide trials (starting at 100% at node 0) is in parentheses. Underneath each factor is the significance of its split at that node, and below each horizontal branch are those trait ranges that make up the homogeneous subsets with respect to the win/loss-dependent variable. See text for how factors were adjusted.


Our results show that hybrids between S. undulatus and S. woodi possess a clear advantage in agonistic encounters. The full hierarchy of advantage was hybrid > S. undulatus > S. woodi. Hybrid males were shown to possess two transgressive phenotypes, wider head-girths and greater aggression. Only the hybrid’s greater aggression was associated with the hybrid advantage, however. Furthermore, the greater aggression was not associated with a specific behaviour; rather hybrids exhibited more intense aggressive behaviours and less intense submissive behaviours than either parental species.

In the overall system, body size (SVL) and aggression contributed most to winning agonistic encounters. Specifically, larger and/or more aggressive individuals had greatest success. The ‘bigger is better’ notion has been found in a number of other studies on agonistic encounters between male lizards (reviewed in Stamps 1983). Larger S. woodi, however, were never successful, whether they were a resident or intruder, against smaller S. undulatus or hybrids, suggesting that the greater aggression of S. undulatus and hybrids trumped the effect of body size.

The evolutionary balance between aggression and body size may explain why S. undulatus is more aggressive than S. woodi, but not the transgressive aggression of hybrids. Both characters are linked to fitness, in that larger and/or more aggressive individuals have more access to females; therefore the larger size and greater aggression observed in S. undulatus may experience correlational selection within the species. The greater aggression of Sceloporus hybrids, however, is more likely explained by transgressive segregation because of the relatively recent hybrid history. Furthermore, hybrids have similar body sizes to S. undulatus, so strong selection on hybrid aggression to compensate for smaller body size is unlikely, and strong selection in the recent past would be needed to explain the large difference in aggression between S. undulatus and hybrids.

Lizards that were established residents and likely defending resources against intruders also had greater chances of winning the encounter. The greater success of defenders is consistent with previous studies (Maynard Smith 1982; McMann 1993), and associated with more aggressive displays (Tables 1 and 2, Fig. 2). The level of aggression exhibited by S. woodi, however, was insufficient even in a larger resident to defeat S. undulatus or hybrids.

Our examination of the variation across all three groups showed that certain morphological characteristics influenced the outcome of agonistic encounters within the SVL/aggression framework (Figs 3 and 4), but only under specific circumstances. For instance, having a wider head-girth was beneficial, but only for larger individuals that were less aggressive (Fig. 4, Node 3). Because larger head-girth has been linked to greater bite pressure (Gier 2003; Perry et al. 2004), less aggressive individuals with relatively wide heads may have won by effectively influencing their opponent when they did bite (e.g. Huyghe et al. 2005). Incidentally, biting was a component of the behavioural repertoire for winners and losers of both parental species and hybrids, although much less frequent in losers. The larger head-girth that hybrids exhibited did not directly influence winning nor did it influence aggression within the HU trials. This finding suggests that the magnitude of wider head-girth, relative to body size, detected in hybrids does not influence the focal individual or the opponent during initial posturing because focal individuals did not show more aggression just because they had a wider head-girth.

The hybrid advantage in agonistic encounters increases the potential for hybrid speciation and/or introgression into either of the parental species. Specifically, the transgressive aggression exhibited by hybrid males should result in a greater ability to both defend and invade territories in their natural habitat, which should provide hybrid males a greater number of mating opportunities. Furthermore, hybrid female lizards from the ecotonal region show a shift in female choice away from male S. woodi, which is what females of both parental species tend to choose, and toward male S. undulatus (Jackson 1972). If this shift in female choice occurs in hybrid females because of the greater size and aggression of S. undulatus, then hybrid males would have an advantage by hybrid female selection over both species. Mating opportunities of hybrids would therefore increase because hybrid males could establish and defend larger and/or better territories plus hybrid females would select more aggressive hybrid males. The hybrid swarm might persist within the hybrid zone only because ecological and sexual selection are creating isolation through competitive advantage. Therefore, geographic isolation may not be necessary for hybrid persistence if hybrid advantage creates ecological isolation. A hybrid advantage in territorial encounters alone, however, may increase fitness enough to explain hybrid persistence. Nonetheless, if greater reproductive output is realized by hybrids through male dominance, the likelihood of introgression or speciation increases in this hybrid system.

Introgression of hybrid genes into S. woodi is a concern because of the conservation status of this rare, endemic lizard. Sceloporus woodi is restricted geographically to a system of small, disjunct populations in central Florida and specializes in scrub habitats (McCoy et al. 2004). Because of urbanization, expanding agriculture, and mining, scrub habitat has been disappearing at an alarming rate (Fogarty 1978; Enge, Bentzien & Percival 1986), and 75% of S. woodi populations are in serious danger of extinction (Enge et al. 1986; DeMarco 1992). Compounding this effect is the inherent fluctuation of population sizes in Sceloporus (Ferner 1976; Parker 1994; Smith & Ballinger 1994), and S. woodi’s limited dispersal ability (Tiebout & Anderson 1997; Clark et al. 1999; Hokit, Stith & Branch 1999). Mating success of these Sceloporus hybrids might have implications on the conservation of S. woodi in its remaining stable populations.

Our study provides evidence that transgressive morphological and behavioural traits are present in these Sceloporus hybrids, and the transgressive behaviour (aggression) is associated with a hybrid advantage in agonistic, and likely territorial, encounters. Therefore, S. undulatus ×S. woodi hybrids are a rare example of viable outbreeding in wild animals that resulted in transgressive phenotypes and, through transgressive behaviour, potentially greater fitness. Because hybrid persistence alongside S. woodi populations may be threatening the genetic integrity of this already threatened species, management schemes should take the hybrid threat into account when S. undulatus populations are known to be adjacent. Broader implications of our results include the link between transgressive segregation and homoploid, hybrid speciation. Transgressive segregation has been implicated as a process through which homoploid, hybrid speciation can occur, but some form of ecological divergence is necessary to impede parental gene flow (Rieseberg et al. 1999). That ecological divergence could manifest in territorial species through transgressive aggression. Future research should examine the ecological and sexual selection of hybrid systems along with the underlying genetics of the hybrid zones to better understand how hybridization can influence animal evolution.


We thank Stephen Karl and Kelvin Gorospe for help with genetic data analysis and 14 undergraduate research assistants for caring for lizards in the laboratory. Thanks are extended to the many students of Stonesoup School in Crescent City, Florida, especially Brandon Rallo, for assisting with lizard collection in the field. Thanks also go to Mark Barrett, Sidney Pierce, Elodie Vercken, Mike Boots and one anonymous reviewer for comments that greatly improved this manuscript. Funding (in part) for this project was awarded to TRR by the American Museum of Natural History – Theodore Roosevelt Memorial Fund, and the USF Hearl Foundation, and to LES by the USF Office of Undergraduate Research. Lizards were collected under collection permit WX05107 issued by the State of Florida Fish and Wildlife Conservation Commission. All protocols were reviewed and accepted by the USF Institutional Animal Care and Use Committee, IACUC file no. 2778.