Female preference is widely described in various taxa, and the underlying mechanisms shaping preferences remain a major focus of sexual selection studies, particularly in species where males contribute minimally to offspring. Female preference is associated with maintaining male secondary sexual traits (SST). However, how male SST impact female preference is less understood. We hypothesized the strength of female preference should scale with the expression of male SST. To test this prediction, we compared female preference for male body size (an easily quantifiable trait that scales with other SST) in three species of Limia (Poeciliidae) varying in secondary sexual traits: L. perugiae, L. dominicensis, and L. zonata. The degree of SST was assessed based on the amount of ornamentation and the presence of courtship in the species. Limia perugiae, L. dominicensis, and L. zonata were designated as possessing high, intermediate, and low male SST, respectively. Female preference was quantified as the relative amount of time females associated with males of various size classes: small, intermediate, and large. Therefore, we predicted because L. perugiae males have the most SST, females would associate more strongly with large males. Limia perugiae females were the only species to display female preference in relation to male body size, but they preferred small males. Although preference was observed, the direction of preference was unexpected. Moreover, the lack of preference for large male size and thereby other SST in the species suggests pre-copulatory female preference is unimpacted by male SST. We suggest cryptic female choice (i.e., preference enacted during or after copulation) may maintain costly male traits. However, future work remains necessary. The present study provides foundational behavioural work on Limia and examines the ubiquity of the evolution of female preference in poeciliids.

Collection and Size Classification

For this study, we used three species of closely related Caribbean livebearing fishes: Limia perugiae, L. dominicensis, and L. zonata (Weaver et al., 2016). The fishes we used were descendants of wild-caught populations from the Dominican Republic on Hispaniola. Limia perugiae were collected in 2014 in a small ditch near the south shore of Lake Enriquillo (18°24'4.61"N, 71°34'16.61"W). Limia dominicensis were also caught in 2014 in a ditch east of Polo, Barahona Province (18°19'6.93"N, 71°34'14.24"W). Limia zonata were caught in 2012 in the shallows of the Río Yuna near Bonao (18°57'33.5"N, 70°24'32.1"W). All specimens were transported to a greenhouse (now the International Stock Center for Livebearing Fishes) at the Aquatic Research Facility at the University of Oklahoma, where they were kept under common garden conditions. The L. perugiae and L. dominicensis populations were kept in 1000-l flow-through stock tanks. Limia zonata were housed in a similar 500-l tank due to their smaller population size compared to L. perugiae and L. dominicensis.

We haphazardly collected 40 individuals (20 females and 20 males) of each species (N = 120) from these stock tanks, moved them into an indoor fish room, divided them by sex, and placed them in 37-l holding tanks. All individuals were acclimated to the laboratory environment for 14 days prior to any subsequent handling. All individuals were kept in a climate-controlled room with 26ºC (±3º) on a 12-hour day-night cycle. The fishes were fed ad libitum, a mixture of frozen brine shrimp naupliae, Daphnia, bloodworms (mosquito larvae), and Tetra Min flakes twice daily before and after behavioural assays.

A laminated grid was used to measure the standard length of each individual (to the nearest 0.1 mm). Fishes were then photographed (Nikon D5200 camera with a Nikon AF-S DX NIKKOR 18-200mm f/3.5-5.6G ED VR II Standard Zoom Lens). Fish were gradually cooled to anesthetize them for photographs; they were held in ice water until they became still enough to be photographed (Klontz and Smith,1968; Collymore, et al. 2014). We elected to cool fishes because preliminary studies of L. dominicensis found that mortality increased when alternative methods of anaesthesia, such as tricaine methanesulfonate (MS-222), were used (Spikes Obs). Immediately afterward, individuals were placed in a recovery tank and were given two days to recover before beginning the behavioural tests. No mortality was associated with this procedure. Females of each species were randomly assigned an ID number and then placed in individual 5-l tanks under identical conditions in a climate-controlled fish room. To ensure individuals were reproductively mature, we used only fish with a standard length (measured from the tip of the snout to the end of the spinal column) greater than 16 mm (Arriaga and Schlupp, 2013). All females less than 16 mm were classified as juveniles and returned to their respective stock populations. Males with fully developed gonopodia, the intromitting organ typical for the family, were determined to be reproductively mature.

We used male size as a likely female-preferred trait in Limia because female preference for large males is well-documented in other Poeciliid species (Ríos-Cardenas and Morris, 2011). Within each species, we sorted males into discrete size categories based on the measured population distribution of male standard length. Using median standard length as an anchor, we divided the size distribution into equal quartiles. We classified males as small if they were in the lowest quartile (five males), intermediate if they were within the two middle quartiles (10 males total), and large if they fell into the upper quartile (five males). Limia perugiae size classes were defined as small (14 mm – 20 mm), medium (21 mm – 26 mm), and large (27 mm <). Similarly, L. dominicensis size classes were defined as small (14 mm – 23 mm), medium (24 mm – 26 mm), and large (27 mm <). L. zonata size classes were defined as small (14 mm – 20 mm), medium (21 mm – 24 mm), and large (25 mm <). The males of each species were then placed into three 37-l aquaria (9 aquaria total), according to size classifications.

Experimental Setup

We used an absolute preference function assay to measure female preference in all three species. Assays were conducted in a 76-l tank that was divided lengthwise into three equal zones (Figure 2). Two vertical lines were drawn on the front pane of the tank to designate zones. The centre zone was deemed the neutral zone, with the two outer zones designated preference and non-preference zones. In each zone, we placed a clear Plexiglas tube (8.5 cm x 8.5 cm rectangular prism), which was used to restrict a male within its assigned zone and to reduce the transmission of mechanosensory and chemical signals (Figure 2). In the preference zone, one male fish was placed in a Plexiglas tube, and in the non-preference zone, there was no male present (i.e., an empty clear Plexiglas tube to control for any bias the female might show toward the tube). The preference and non-preference zones were randomized to control for side bias.

During absolute preference function assays, females of each species (N = 20 per species) were sequentially presented with three conspecific males of varying size (small, medium, and large) in a randomized order. Females were randomly selected by ID number and placed in a clear Plexiglas tube in the centre of the neutral zone after her assigned male was first placed into the preference zone. Both individuals were then given 300 seconds to acclimate to the environment. To begin the trial, we released the female from the tube, and after she was swimming freely around the tank (i.e., showing no signs of stress-like behaviours), we recorded for 300 seconds the amount of time (s) the female spent in the preference or non-preference zone. Females were defined to be in the preference or non-preference zone if their entire head passed the vertical line on the front pane of the tank. After the 300 seconds passed, we gently placed the female back in the tube in the centre of the neutral zone, and the male was removed. Then a male of another size class was presented to the focal female. The process was repeated until the female was exposed to one male from each of the three size classes. Once a female was exposed to three males, she was returned to her home tank, and males were placed in size-specific recovery tanks labelled ‘used.’ Males remained in recovery tanks for 3 days before haphazardly being used in another trial with a different conspecific female. After a trial, the testing tank received a 50% partial water change to reduce any lingering chemical signals from affecting future trials. All trials for any given species were conducted over the course of 30 days. Trials began in November 2017 and were concluded in January 2018.

Data Analysis

All analyses were conducted in R using the nlme and BayesFactor packages (R, v.3.0.2 R Core Team, 2019; Pinheiro et al., 2019; Morey and Rouder, 2018). A linear mixed-effects model was used to determine if the intensity of male secondary sexual traits influenced female preference. Specifically, we loaded species and male size class as explanatory variables and association time as the response variable. Presentation order, female size, and female identity were considered as random factors to control for order effects and prevent pseudo-replication in the analyses, respectively. However, to preserve non-singularity, both presentation and order female size were removed from the model. If a significant effect was observed, we conducted a post-hoc Tukey test to determine the direction of preference. After running the full model, including all species, we used independent species-specific linear mixed-effects models on each species to compare female association time as it relates to male size within species, loading the same factors excluding species.

Because Limia has previously demonstrated atypical behaviour as compared to other poeciliids (Spikes & Schlupp, 2021), we elected to conduct a Bayes factor post-hoc analysis. The Bayes factor analysis uses Bayesian inference to computes an integer known as a Bayes factor (BF), which can be interpreted using the cut-offs posited by Morey et al. (2016). The Bayes factor - like the p-value - is then indicative of how much support one has for the hypothesis. Unlike the p-value, however, the Bayes factor allows for interpretation of the support for the null hypothesis. Hence the analysis is particularly advantageous when interpreting insignificant results. Particularly if a study yields negative data that requires the rejection of the hypothesis, Bayes factors will indicate if the results are inconclusive or if the null hypothesis should be accepted.

See methods for data variable details.