Diversification in sexual signals is often taken as evidence for the importance of sexual selection in speciation. However, in order for sexual selection to generate reproductive isolation between populations, both signals and mate preferences must diverge together. Furthermore, assortative mating may result from multiple behavioural mechanisms, including female mate preferences, male mate preferences and male-male competition; yet their relative contributions are rarely evaluated. Here, we explored the role of mate preferences and male competitive ability as potential barriers to gene flow between two divergent lineages of the tawny dragon lizard, Ctenophorus decresii, which differ in male throat coloration. We found stronger behavioural barriers to pairings between southern lineage males and northern lineage females than between northern males and southern females, indicating incomplete and asymmetric behavioural isolating barriers. These results were driven by both male and female mate preferences rather than lineage differences in male competitive ability. Intrasexual selection is therefore unlikely to drive the outcome of secondary contact in C. decresii, despite its widely acknowledged importance in lizards. Our results are consistent with the emerging view that although both male and female mate preferences can diverge alongside sexual signals, speciation is rarely driven by divergent sexual selection alone.

Study species and husbandry

We used 90 adult lizards (>65mm snout-vent length; SVL) comprising 21 male and 24 female northern lineage C. decresii from Caroona Creek Conservation Park, South Australia (-33.4114°S, 139.0945°E), and 21 male and 24 female southern lineage C. decresii from private properties around Palmer, South Australia (-34.8223°S, 139.1621°E). Lizards were collected in September in 2015 and 2016, and subsequently kept in captivity at The University of Melbourne, Victoria, Australia, where they were housed individually in 55 × 34 × 38cm (length × width × height) opaque plastic enclosures containing a layer of sand and a crevice between two ceramic tiles for shelter. Housing was maintained at temperatures and lighting cycles that mimicked natural seasonal variation, with UV lights (ZooMed T8 ReptiSun® 10.0 UVB) above each enclosure (30cm), emitting both UVA and UVB radiation. A heat lamp was provided to generate a thermal gradient and allow the lizards to attain their preferred body temperatures (approx. 36°C). Lizards were misted with water for hydration and fed live crickets dusted with multi-vitamins three times per week. All behavioural trials were conducted during the breeding seasons (August–December) in 2016 and 2017. Research methods used in this study were reviewed and approved by the Animal Ethics Committee of The University of Melbourne (1413220.3) and the South Australian Wildlife Ethics Committee (25/2015).

Female-male behavioural trials

Females are receptive to mating approximately 2–3 weeks after emergence from hibernation, and after laying their first or second clutch. We conducted mate preference trials during these known receptive periods, when females were in good body condition (average mass of 16.7g ± 2.9g), though receptivity cannot be determined with certainty a priori. Each female was paired with both a southern and a northern lineage male, with half of the females paired with a southern male first and the other half with a northern male first. Females were placed into the first male’s enclosure for a period of 24 hours, and then into the second male’s enclosure for the subsequent 24 hours. Both encounters were monitored and recorded using a Swann DVR8-1525 8 channel 960H digital video recorder with a PRO-615 camera attached. We conducted a total of 147 trials, with individual females paired with one southern and one northern male per reproductive cycle, in up to 2 reproductive cycles (average of 3.34 trials, with a range of 2–4 trials, per female).

Videos were analysed using Behavioural Observation Research Interactive Software (BORIS) version 4.1.5 and both female and male behaviour was scored. For females, we recorded the number of head-bobs (pronounced nodding movement of the head), and combined the number of aggressive behaviours (biting and chasing) and times the female fled from the male as a measure of “rejection”. For males, we also recorded the number of head-bobs (courtship behaviour) as well as the number of attempts to copulate, and whether or not copulation was successful. We did not analyse the number of successful copulations as copulation was observed in only 7 of the 147 trials (although more may have taken place under the tile). Lizards were not paired for long enough to ensure mating; rather, we were interested in behaviour during initial contact as an indicator of mate preference.

We tested whether female lineage, male lineage, or their interaction predicted: 1) number of copulation attempts, 2) number of male head-bobs, 3) number of female head-bobs and 4) number of female rejection behaviours using generalised linear mixed models (lme4 package, R). Female ID, male ID and pairing number (female’s first or second trial) were included as random factors in all models to account for repeated use of individuals, and response variables were log transformed to meet model assumptions of normality. We performed pairwise comparisons by calculating least squares means and confidence intervals using the Satterthwaite’s approximation for degrees of freedom (lmerTest package, R).

Male-male behavioural trials

A previous study investigating aggression levels among morphs of the northern lineage found that orange-throated males were significantly more aggressive towards territory intruders than yellow, orange-yellow or grey-throated males. Therefore, we categorised males into three behavioural groups based on lineage and throat colour morph:  southern, northern high aggression (orange), or northern low aggression (yellow, orange-yellow, grey). We designed trials such that each focal male was matched with three others, representing each of the behavioural groups, in random order. Pairs were size-matched to minimize the effect of body size on contest outcome, with an average difference of 1.59mm ± 1.16mm snout vent length (SVL) between competing males.

Contest trials were conducted in a neutral 120 × 30 × 60cm (length × width × height) enclosure (i.e. not the home enclosure of either male). An opaque divider initially separated the enclosure into two equally sized holding areas, each containing a layer of sand, ceramic tile and heat lamp. Just prior to the trial, males were weighed to obtain a measure of body condition as the residuals of a linear model of mass and SVL. The designated “focal” and “opponent” males were then placed into the separate holding areas and allowed to acclimatise for 48 hours to establish residency. At the commencement of the trial, the divider was removed and the interaction was recorded from two different angles using Panasonic HC-V770M video cameras. Trials were conducted for a maximum of 25 minutes and monitored to ensure there was no risk of injury to animals (as required under the Animal Ethics permit). Consequently, we did not record contest outcome (i.e. winner, loser) as some trials were stopped before a winner was established. To minimize stress and the potential influence of previous contest outcomes, males were not used in a subsequent trial for at least 48 hours. We conducted a total of 120 trials (involving 42 males), 26 of which were excluded due to no interaction, resulting in 94 trials which were used in the statistical analysis.

We scored focal male behaviour from the video footage using BORIS. C. decresii males perform energetic displays during territory defence prior to engaging in physical aggression. Therefore, we recorded the number of head-bobs, tail flicks and push-ups performed by the focal male as a measure of “display behaviour”, and combined the duration of chasing and wrestling (involving biting) as a measure of “physical aggression”. We also recorded the time between the start of the trial and the focal male’s emergence from beneath the tile (“latency”), as this is an indicator of individual boldness. Display behaviour and physical aggression were divided by the total trial duration (minus latency) to account for differences in trial lengths.

We tested whether behavioural group or body condition predicted: 1) focal male latency to emerge, 2) focal male display behaviour and 3) focal male physical aggression using generalised linear mixed models. We included focal male behavioural group, opponent male behavioural group and their interaction, as well as focal male body condition and opponent male body condition as predictor variables in the models. Additionally, focal male ID and focal male trial number were included as random factors in all models to account for repeated use of individuals. For models 2 (display behaviour) and 3 (physical aggression), the response variables were log transformed to meet model assumptions of normality, and we performed post hoc pairwise comparisons as detailed above.

Includes excel spreadsheets of data collected from female-male behavioural trials (female_male_data.xlsx) and male-male behavioural trials (male_male_data.xlsx), and an R Markdown file (Analysis.rmd) and output document (Analysis.html) with all statistical analyses performed in this study.