Data from: Does losing reduce the tendency to engage with rivals to reach mates? An experimental test
Data files
Apr 26, 2024 version files 65.87 KB
Abstract
Male-male contests for access to females or breeding resources is critical in determining male reproductive success. Larger males and those with more effective weaponry are more likely to win fights. However, even after controlling for such predictors of fighting ability, studies have reported a winner-loser effect: previous winners are more likely to win subsequent contests, while losers often suffer repeated defeats. While the effect of winning-losing is well-documented for the outcome of future fights, its effect on other behaviors (e.g., mating) remains poorly investigated. Here, we test whether a winning versus losing experience influenced subsequent behaviors of male mosquitofish (Gambusia holbrooki) towards rivals and potential mates. We housed focal males with either a smaller or larger opponent for 24 hours to manipulate their fighting experience to become winners or losers, respectively. The focal males then underwent tests that required them to enter and swim through a narrow corridor to reach females, bypassing a cylinder that contained either a larger rival male (competitive scenario), a juvenile or was empty (non-competitive scenarios). The tests were repeated after one week. Winners were more likely to leave the start area and to reach the females, but only when a larger rival was presented, indicating higher levels of risk-taking behavior in aggressive interactions. This winner-loser effect persisted for at least one week. We suggest that male mosquitofish adjust their assessment of their own and/or their rival’s fighting ability following contests in ways whose detection by researchers depends on the social context.
README: Data from: Does losing reduce the tendency to engage with rivals to reach mates? An experimental test
https://doi.org/10.5061/dryad.s4mw6m9dv
Description of the data and file structure
The analysis details can be found in the corresponding R Markdown (RMD) files. Specifically, in the RMD file “Zang_C_winner_loser_effect_initiation,” we conducted paired t-tests on body size, comparing focal males in size-matched pairs, focal winners and their competitors, and focal losers and their competitors. This analysis utilized data from the file called “Fish_size”, which contains the body length of focal fish and model males used in contests and encounter tests. The columns include “Pair ID” (i.e. ID of each size matched pair of a winner and a loser), “ID” (i.e. focal fish ID), “length” (i.e. standard length of focal fish), “winner_loser” (i.e. the contest outcome: winner/loser), “length_of_competitor” (i.e. the standard length of the stimulus fish in experimntally manipulated contests), and “length_of_rival” (i.e. standard length of the larger rival in the encounter tests). The length was recorded in millimeter.
Regarding the winner-loser effect, a two-step analysis was performed for each recorded behavior using data from the file called “Winner_loser_effect”. The file contains the time taken in three anticipated behaviors (i.e. leaving a start area, staying the encounter area and entering the reward area) by focal males in different treatment and tests. The columns include “PairID”, “ID”, “tankID” (i.e. the ID of two test tanks were used in encounter tests), “winner_loser”, “Timing_of_encounter_test” (i.e. the time taken for experiencing encounter tests after contests, with values that 0 represents that fish were tested on the same day of contests and 7 represents fish were tested 1 week after contests), “test_order” (i.e. Test I, Test II, or Test III), “treatment” (i.e. transparent cylinders in the corridor that contains either a larger male, a juvenile or was empty), “Time_taken_to_enter_reward_area”, (i.e. time to reach females in seconds with 600 indicating never reaching females), “Time_taken_in_encounter_area”, (i.e. time within 3cm of cylinder in corridor), “Time_taken_to_leave_start_area”, (i.e. time spent in the start area in seconds with 600 indicating never leaving), “Whether_fish_left_start_area” (i.e. whether the fish left the start area, with 0 = no and 1 = yes), “Whether_fish_entered_encounter_area” (i.e. whether the fish entered the encounter area, with 0 = no and 1 = yes), and “Whether_fish_entered_reward_area” (i.e. whether the fish entered the reward area, with 0 = no and 1 = yes). In this file, data marked as NA indicates that it is not available, typically due to data loss during the experiment. Any data marked as NA will not be included in the subsequent analyses.
The winner-loser effect on three responses was analyzed in separate RMD files titled: (1) “Zang_C_winner_loser_effect_initiation,” (2) “Zang_C_winner_loser_effect_encounter,” and (3) “Zang_C_winner_loser_effect_reward.” For each analysis, first we examined the likelihood that the male: (1) left the start area, (2) entered the encounter area, and (3) entered the reward area. We ran separate generalized linear mixed models (GLMMs) with binomial error (glmmTMB package). Second, for those males who exhibited the above behaviors, we ran individual GLMMs (with Gaussian error) to analyze the time spent on each behavior. We log-transformed the time data to fulfil the model assumptions. Model residuals of all GLMMs were checked using the DHARMa package (Hartig, 2020). Contest outcome (winner, loser), timing relative to contest/trial experience (Test I, Test II, Test III) and encounter type (male rival, juvenile, empty cylinder) were treated as fixed factors, and their three-way and two-way interactions were included in the initial model. Male identity was treated as a random factor to control for repeated measures from the same male. We also treated the size-matched winner/loser pair’s identity as a random factor to account for both males facing the same stimulus fish, on the same day, with the same encounter type order.
We disentangled the effect of test order (first or second test) and timing (time elapsed since the 24h contest treatment) using contrast matrices derived from the summary results of the GLMMs. For each response variable, we first investigated whether test order altered the effect of contest outcome on the response to different encounter types (i.e., the three-way interaction) by only considering data from Test I (first test immediately after contest experience) and Test III (second test immediately after contest experience). As there were no significant differences between Test I and III, we combined the data to test whether the immediate effect of contest outcome depended on the encounter type (i.e., a significant two-way interaction). If the two-way interaction was significant for a given response variable, we reported the effect of contest outcome for each encounter type (empty, rival, juvenile); otherwise we reported the main effect of contest outcome. If there was a significant effect of contest outcome for a given response variable, we further compared data from Test II (second test, one week after contest experience) to Test III (second test, immediately after contest experience) using the same approach to test for the persistence of the observed winner/loser effect. These comparisons are detailed in the RMD files “Zang_C_persistence_of_effect_initiation” and “Zang_C_persistence_of_effect_reward”. P values and effect estimates were calculated using generalized linear hypothesis tests (GLHTs) in multicomp package (Hothorn et al. 2023).
Although it was not the focus of our study, we also tested the repeatability of boldness between Tests I and II, as outlined in the RMD file “Zang_C_repeatability_test.” These tests utilized data from the “Winner_loser_effect” dataset. Sheltering behavior in a neutral environment is commonly used to measure boldness, so we ran repeatability tests on whether males left the start area for the non-agonistic/control treatments (i.e., an empty cylinder and a cylinder with a juvenile). We treated the response as binary (i.e., left/did not leave the start area) and male ID as a random factor in the repeatability tests (using the rptR package) (Stoffel et al. 2017).
All analyses were conducted in R version 4.2.1. Results are presented as mean ± s.e., with the significance level set at alpha = 0.05 (two tailed).
Sharing/Access information
Supplementary materials and raw data have been deposited at Mendeley Data (https://data.mendeley.com/preview/jkbxgdj23d.
Methods
Here, we examined whether winners/losers changed their behavior with the time that had elapsed since their previous contest experience. To do this, each male was tested twice in encounter tests: once immediately after the contest period ended (Test I), and then again one week later (Test II). To control for any potential effects of time and/or familiarity with the aquarium setup in Test II (Fig. 2), we ran an additional pair of encounter tests with another set of males who did not initially experience staged contests. After their first encounter test, focal males were isolated in individual 1L tanks for a week. Then they experienced contests for 24 hours (as above), after which we recorded these males’ behavior during their second encounter test (Test III).
The encounter test contained a start area at one end which was connected by a corridor to a reward area at the opposite end. The focal male was introduced into a mesh cylinder in the start area. Four large females were housed in the reward area. There was a transparent plastic cylinder at the center of the corridor that housed either a rival male, a juvenile, or was left empty. We recorded for each focal male: (i) initiation time: the time taken to enter the corridor; (ii) encounter time: the time spent < 3 cm (equivalent to 1 SL of the male) from the cylinder in the corridor; and (iii) total trial time: the time taken from release until reaching the reward area (see Supplementary Material for details). The trial concluded when the male entered the reward area, or when 10 minutes had elapsed.
A two-step analysis was conducted for each of the recorded behaviors. First, we examined the likelihood that the male: (1) left the start area, (2) entered the encounter area, and (3) entered the reward area. We ran separate generalized linear mixed models (GLMMs) with binomial error (glmmTMB package). Second, for those males who exhibited the above behaviors, we ran individual GLMMs (with Gaussian error) to analyze the time spent on each behavior. We log-transformed the time data to fulfil the model assumptions. Contest outcome (winner, loser), timing relative to contest/trial experience (Test I, Test II, Test III) and encounter type (male rival, juvenile, empty cylinder) were treated as fixed factors, and their three-way and two-way interactions were included in the initial model. Male identity and size-matched winner/loser pair’s identity was treated as a random factor to control for repeated measures from the same male. If there was a significant effect of contest outcome for a given response variable, we further compared data from Test II (second test, one week after contest experience) to Test III (second test, immediately after contest experience) using the same approach to test for the persistence of the observed winner/loser effect.