Reinforcement occurs when selection against hybrid offspring strengthens behavioral isolation between parental species and may be an important factor in speciation. Theoretical models and experimental evidence indicate that both female and male preferences can be strengthened upon secondary contact via reinforcement. However, the question remains whether this process is more likely to affect the preferences of one sex or the other. Males of polygynous species are often predicted to exhibit weaker preferences than females, potentially limiting the ability for reinforcement to shape male preferences. Yet, in darters (Percidae: Etheostoma), male preference for conspecific mates appears to arise before female preferences during the early stages of allopatric speciation, and research suggests that male, but not female, preferences become reinforced upon secondary contact. In the current study, we aimed to determine whether the geographically widespread darter species Etheostoma zonale exhibits a signature of reinforcement, by comparing the strength of preference for conspecific mates between populations that are sympatric and allopatric with respect to a close congener, E. barrenense. We examined the strength of preference for conspecifics for males and females separately to determine if the preferences of one or both sexes have been strengthened by reinforcement. Our results show that both sexes of E. zonale from sympatric populations exhibit stronger conspecific preferences than E. zonale from allopatric populations, but that female preferences appear to be more strongly reinforced than male preferences. Results therefore suggest that reinforcement of female preferences may promote behavioral isolation upon secondary contact, even in a genus that is characterized by pervasive male mate choice.
To estimate the strength of preference for conspecific mates, we conducted two types of behavioral experiments. Dichotomous mate choice trials gave individual fish a choice between a conspecific or heterospecific of the opposite sex and did not allow fish to physically interact. Artificial stream assays simulated natural conditions and allowed multiple fish of both sexes and species to freely interact. Details of the experimental designs are provided below. Both trial types were conducted previously for a single sympatric population (Roberts & Mendelson, 2017; Williams & Mendelson, 2010). For the current study, we conducted trials with two allopatric populations and an additional sympatric population of E. zonale with respect to E. barrenense. We combined data from these populations with those from the previously studied sympatric population to test for a signature of reinforcement.
Dichotomous choice assays
Dichotomous choice assays followed previously published methods from Roberts & Mendelson (2017) and Williams & Mendelson (2010). Briefly, a 37.9-litre glass ‘focal’ tank (50L x 25W x 30H cm) was positioned between two 9.6-litre glass ‘stimulus’ tanks (30L x 15W x 20H cm) so that the long sides of the stimulus tanks were flush against the short side of the focal tank. The focal tank was marked with two 5 cm ‘association zones’ at either end closest to the stimulus tanks. Prior to a trial, an opaque partition was placed between each stimulus tank and the focal tanks, then one E. barrenense was placed into a stimulus tank, one E. zonale of the same sex was placed in the other, and a focal E. zonale (opposite sex of stimulus fish) was introduced into the test tank. Once the focal fish began free-swimming activity, the opaque partitions were removed and acclimation began. Acclimation was complete after the focal fish entered both association zones and subsequently entered the ‘neutral zone’ (i.e. was not in either association zone). Following acclimation, a 15-minute observation period began, and time spent in each association zone was recorded using JWatcherTM V1.0 (Blumstein et al. 2000).
We tested the preferences of male and female E. zonale from the Middle Fork Red River (MF, allopatric), French Creek (FC, allopatric) and Line Creek (LC, sympatric) populations (N = 18 for each sex and population). Stimulus fish for the allopatric MF and FC populations were a conspecific individual from the same site as the focal fish and a heterospecific individual (E. barrenense) from the EF population. Stimulus fish for the sympatric LC population were from the same site: a conspecific individual from LC and a heterospecific individual (E. barrenense) from LC. Stimulus individuals were size matched within 15% of their standard length (snout to caudal peduncle).
Artificial stream assays
Stream assays took place in a Living Stream artificial stream system; model LSW-700 (Frigid Units Inc., Toledo, OH, USA). The stream unit (213 L x 61 W x 56 H cm) contained approximately 530 l of flowing water maintained at a constant temperature of 12° C and matched to the water quality of the aquarium housing. We began a six-hour observation period during which we tallied five types of behavior for both E. zonale and E. barrenense: (1) spawning events, defined by the rapid and synchronized quivering of a male-female pair, followed by a brief contact of the genital regions to the substrate (indicating the release of gametes); (2) successful male solicitation, defined by a male’s approach toward a stationary female followed by his periodic body twitching alongside the female, but where the release of gametes was not observed by either sex; (3) unsuccessful male solicitation, defined as a male’s approach towards a stationary female and the female’s subsequent fleeing; (4) chases between males, defined as a male’s approach toward another male resulting in accelerated pursuit and fleeing by the respective males; and (5) male chases of females, defined as a male changing his direction or speed in pursuit of a swimming female. Trials began between 900 and 1100 h. Both the acclimation time and entire trial time were recorded by two Panasonic HX-A1 Action Cam’s (Panasonic Corp., Osaka, Japan), each situated to record one half of the artificial stream unit. Video of each trial was used to score behaviors by one researcher (NSR).
Strength of preference (SOP) was calculated for each individual using the equation:
- SOP = (TC – TH) / (TC + TH)
where TC is the total time spent in the conspecific association zone and TH is the total time spent in the heterospecific association zone over the entire 15-minute trial. SOP can range from +1 to –1 indicating a complete preference for conspecific or heterospecific individuals respectively, whereas a score of 0 indicates no preference.
We used a general linear model to determine which factors, or interactions of factors, explain variation in SOP in dichotomous trials. The main effects we examined were context (i.e. sympatry or allopatry), population (i.e. EF, LC, MF, or FC), sex, and interactions of these terms. Population was nested within context in all models since these two factors exhibit hierarchical data structure (i.e., population is a smaller spatial scale encompassed by context). Model selection started with a full model, including all possible terms and interactions. We then sequentially removed non-significant variables from the full model to reduce model parameters. Our final model included all single variables (i.e., context, sex, and population) and the interaction of sex and population as explanatory variables for SOP. Post hoc analyses compared strength of preference between context and populations using least squared means with a Bonferroni adjustment for multiple comparisons and Mann-Whitney U tests to analyze data from male and female E. zonale independently.
For the artificial stream trials, isolation indices (I) were calculated separately for both E. zonale and E. barrenense after Stalker (1942) for each of the five behaviors recorded (spawning, successful solicitation, unsuccessful solicitation, male-male chase, male-female chase) as
I = (mean no. conspecific interactions across replicates – mean no. heterospecific interactions across replicates) / (mean no. conspecific interactions across replicates + mean no. heterospecific interactions across replicates)
Additionally, a total isolation index, taking into account all five behaviors together, was calculated for the artificial stream assays in the current study. For each replicate, count data for the five behaviors were summed, for each species separately, to yield the total number of conspecific-directed behaviors and the total number of heterospecific-directed behaviors in each replicate. We took the average of these summed totals across the three replicates per population and applied these values to equation 2 to calculate a total isolation index for each species and population.
We used a generalised least squared (GLS) model to determine the effect of context (sympatry or allopatry), species, and population on the total isolation index calculated for the artificial stream assays for the LC, MF, and FC populations. Using a GLS allowed us to account for heteroskedastic data (Zurr et al., 2009). All E. barrenense were coded as sympatric with respect to E. zonale, while E. zonale were coded as sympatric or allopatric accordingly. Model selection started with a full model, including all possible terms and interactions, and included trial replicate as a random effect. We then sequentially removed non-significant terms to select the fixed effect structure of the final model. The final model included only context (sympatry or allopatry) as explanatory variables for total isolation. All analyses were conducted in R (ver. 3.5.0; R Core Development Team 2015). Post-hoc comparisons were calculated using the package ‘lsmeans’.
Data from the East Fork population for females (SOP) and for East Fork popilation artificial stream assays (I) are from Williams & Mendelson (2010) and data from the East Fork population for males (SOP) is from Roberts & Mendelson (2017).