https://doi.org/10.5061/dryad.k3j9kd5fv

Description of the data and file structure

The analysis can be re-run using the R code called “Chung et al. mosquitofish mate choice”. For each response variable (time females spent with each male; total distance males swam; total time males spent inspecting females; proportion of time males spent with the solitary female), we used mixed models including male ID as a random factor, to account for repeated measurements of the same male. We included reproductive treatment (“naïve”, “mating only”, “mating and ejaculation”), treatment duration (8 weeks, 16 weeks) and their interaction in initial models. A non-significant interaction was removed from the final model to interpret the main effects. Removing non-significant interactions did not significantly alter the model fit, as determined by log-likelihood ratio tests. Full model outputs (for both initial and final models) are provided in the supplementary material. Wald chi-square tests were performed to calculate the P value using the Anova function in the car package (R studio v1.3.1093 with R v4.0.5). We used type III sums of squares for models with the interaction term and type II sums of squares for models without the interaction. Results are presented as mean ± SE. The significance level is set at alpha = 0.05 (two tailed). We ran Tukey’s post hoc pairwise tests (emmeans package) if the treatment effect was significant. Any deviations from this approach are specified below.

(i)             Male attractiveness

We employed a two-step analysis to investigate male attractiveness. First, we tested whether females preferred to spend time with males rather than alone by comparing the proportion of time spent near the empty compartment to that expected by chance (= 0.25) using separate one sample t-tests for trials at weeks 8 and 16. The proportion of time was power-transformed to meet the assumption of normality (determined via Shapiro-Wilk tests). Second, we tested whether male reproductive history, treatment duration and their interaction affected how much time a female spent with a male using a generalized linear mixed model (GLMM). Quasi-Poisson error was employed to ensure that the data variance conformed to the model assumption, which was assessed using a dispersion test in the DHARMa package. The total time females spent in the association zone of each male was the response variable, with female ID as a random factor to account for three males being tested in the same trial.

We analyzed the first step using the file called “relative_time_empty”, which contains data on the time females spent within four association zones when tested with males at two different age groups: when males were young (Week8: W8) and old (Week 16: W16). The columns include “female_id”, “male_age”, “male1_time” (i.e., time spent with the first male)”, “male2_time” (i.e., time spent with the second male), “male3_time” (i.e., time spent with the third male), and “empty_time” (i.e., time spent in the empty corner). Finally, we calculated the relative time females spent in the empty corner (“relative_time_empty”) by dividing the “empty_time” by the sum of “male1_time”, “male2_time”, “male3_time”, and “empty_time”. The time was recorded in seconds.

We analyzed the second step using the file called ‘male_attractiveness,’ which contains data on the time females spent with each treatment male: namely, “Naive” (i.e., males without reproductive history), “Mating_only” (i.e., males that could mate but could not ejaculate), and “Mating_and_ejaculation” (i.e., males that could mate and ejaculate). In this file, each treatment male has an individual row (i.e., “male_id”) with their testing age (i.e., “age”), either Week 8 (W8) or Week 16 (W16). Each row also reveals which females chose them (“female_id”) and the time females spent with them (“time” in seconds). Finally, we report whether or not the focal male was located in the middle corner (“in_middle_corner”: yes or no), meaning the corner located between the other two corners that contained a focal male.

(ii)           Male mate choice

We ran linear mixed models (LMMs) to test the effects of male reproductive history, treatment duration and their interaction on male: (a) distance swum and (b) time spent inspecting females. We analyzed the proportion of time males spent with the solitary female compared to the other female using a GLMM (binomial error, cbind function of absolute time with each female), where trial ID was considered as a random effect to account for overdispersion. Pair ID was included as a random factor to account for repeated usage of a combination of stimulus fish (max 3 trials per combination; one per treatment).

We further tested whether the proportion of time males of each type at 8 and 16 weeks respectively spent with a solitary female differed from the null expectation (= 50%). To do this, we ran six separate intercept-only GLM models (quasi-binomial error, cbind function of absolute time with each female). An intercept of 0 corresponds to males spending 50% of the time with each female (ln(p/[1 − p]), where p = proportion of time with the solitary female). If the intercept is significantly greater than zero, this indicates that males spent significantly more time with the solitary female.

The analysis was conducted using the file named “male_mate_choice”. Each focal male has its individual row (“male_id”) with the assigned “treatment” (Naive, Mating_only, Mating_and_ejaculation; see above for details) and “age” (Week 8 = W8; Week 16 = W16). Male information was all reported, including:

w8_survival: Whether the male survived until Week 8 (1 means alive; 0 means died). This is only shown for males tested at Week 8 (i.e., age = W8).

w16_survival: Whether the male survived until Week 16 (1 means alive; 0 means died). This is only shown for males tested at Week 16 (i.e., age = W16).

pre_body_length: Male body size before treatment. For an age of Week 8, this means the size at Week 0. For an age of Week 16, this means the size at Week 8 (in millimeters).

post_body_length: Male body size after treatment. For an age of Week 8, this means the size at Week 8. For an age of Week 16, this means the size at Week 16 (in millimeters).

pre_gonopo_length: Male’s gonopodium length (i.e., intromittent organ) before treatment. For an age of Week 8, this means the gonopodium size at Week 0. For an age of Week 16, this means the gonopodium size at Week 8 (in millimeters).

post_gonopo_length: Male’s gonopodium length after treatment. For an age of Week 8, this means the gonopodium size at Week 8. For an age of Week 16, this means the gonopodium size at Week 16 (in millimeters).

Total_distance_swum: The total distance males swam during the mate choice trial (in centimeters).

Total_inspection_time: The total time males spent with both females (in seconds).

Fish_pair_id: The identity for each combination of two females and a model male in each trial.

time_with_female_with_rival: The time males spent with the female surrounded by another male (in seconds).

time_with_solitary_female: The time males spent with the solitary female (in seconds).

body_length_female_with_rival: The body length of the female surrounded by another male (in millimeters).

body_length_solitary_female: The body length of the solitary female (in millimeters).

body_length_rival: The body length of the model male (in millimeters).

trial_id: The identity of each individual trial.

Proportion_time: The relative time males spent with the solitary female (i.e., dividing “time_with_solitary_female” by the sum of “time_with_female_with_rival” and “time_with_solitary_female”).

Note: “na” indicates non-available data (i.e., the fish died before measurement).

Finally, we have updated a raw data file that includes the overall data for both (i) male attractiveness and (ii) male mate choice (note: “na” means missing values where the fish had died before measurement).

Sharing/Access information

Our experimental and analysis plans were registered online before data collection (osf.io/swdv7).