Sexually attractive traits predict predation-threat sensitivity of male alternative mating tactics

Although visual sexual signals, such as ornamental colors and courtship displays, and large body size in males are attractive to females in numerous species, they may also inadvertently attract the attention of eavesdropping predators and thus may be costly in terms of increasing individual risk of mortality to predation. Theoretically, more color-ornamented and larger males should be more predation threat sensitive and suppress their sexual signaling and(or) mating effort relatively more than their less color-ornamented and smaller counterparts when under predation hazard. Here, we experimentally tested this hypothesis by quantifying concurrently the rates of alternative mating tactics (courtship displays, sneak mating attempts) expressed by male Trinidadian guppies (Poecilia reticulata) varying in color ornamentation and body size under a staged immediate threat of predation. Males suppressed their overall mating effort in response to the perceived predation threat, decreasing the frequency of their (presumably more conspicuous) courtship displays significantly more on average than the frequency of their sneak mating behavior. Statistically controlling for body length, more color-ornamented males were more threat-sensitive in their courtship displays, but not sneak mating attempts, under predation hazard than drabber males. Controlling for body coloration, larger males exhibited lower courtship and sneak mating efforts than smaller males in both predation treatments, but body length only influenced threat sensitivity in sneak mating behavior. These results are consistent with both the threat-sensitive hypothesis and asset protection principle and highlight the phenotype-dependency and adaptive plasticity of alternative mating tactics in male guppies under varying predation risk.

Methods
In this experimental hypothesis-driven laboratory study, individual male Trinidadian guppies were exposed to a matched pair of predation-risk treatments (fish predator absent, fish predator present) in random order on the same day across the test subjects, and the frequencies of their courtship displays and sneak mating attempts directed towards two adult female guppies were recorded over 15 minutes.  A total of 53 wild-caught males, varying naturally in body length and colour ornamentation, were similarly tested for the threat sensitivity of their two alternative mating tactics. The total body length (mm) and body colour ornamentation (proportion of body covered with pigmented and iridescent colours) of each test male were measured and quantified, respectively. 

All data were analyzed in the R statistical framework and all statistical tests were two-tailed.  We used the R packages "tidyverse", "readxl", "ggbeeswarm" and "patchwork" for data wrangling and visualisation. Moreover, we used the R packages "glmmTMB" for analyses, "DHARMa" for diagnostic checks, and "emmeans" for extracting the coefficients of trend lines (along with their associated 95% confidence intervals and contrast) for each predator treatment. Depending on the biological hypothesis being tested, we used generalised linear mixed models (GLMM) or zero-inflated GLMMs to fit the observed frequencies of the two mating tactics, courtship displays and sneak mating attempts, as response variables. In each model, predator treatment, male body length and male body coloration, and their interactions, were included as fixed effect predictor variables, predator treatment order as a fixed covariable, and maleID as a random effect. For each of the final models, we separately plotted the raw data and predicted relationship (with its estimated slope and associated 95% CI) between the particular behavioural variable of interest (courtship display or sneak mating attempt rate) and either male color ornamentation or body length for each of the two predation treatments separately, while holding the other variables constant.  We tested the difference in slopes of the predicted relationships for the paired predator-absent and predator-present treatments, respectively, using the R package "emmeans".  We considered the difference in paired slopes being compared a measure of the predator threat sensitivity of the particular fitted behavioural variable.