The effect of turbidity on female mate choice in the guppy, Poecilia reticulata

Animals often moderate behaviours based on environmental and social cues to enhance their chances of survival and reproductive success. This requires reliable cues that re ﬂ ect changes in ambient conditions. Environmental pollutants can increase communication costs and distort the accuracy of received information, leading to negative effects on the outcome of behavioural adjustments. Consequently, selection may reduce behavioural plasticity to avoid errors or promote more prudent decision making when certain cues become unreliable. In aquatic ecosystems, water turbidity shows high levels of natural variation, but human activities amplify its magnitude and extent by increasing algae and suspended solids in the water. Here, we investigated how elevated levels of water turbidity affected mating decisions made by female guppies in a laboratory setting. Guppies are a widespread species with a broad tolerance of polluted environments, and females rely heavily on visual cues when making mate choice decisions. We manipulated turbidity by adding suspended solids to water and tested how this affected female preference for males with more orange coloration. We found that females spent more time near males with a greater area of orange coloration in low turbidity water, but that this preference disappeared in both the mid and high turbidity water treatments. Moreover, female preferences were repeatable in low turbidity water and high turbidity water but not in mid turbidity water. Together, our results suggest that water turbidity impairs the ability of females to discern male coloration and affects the consistency of mate preferences. This could have consequences for the

With a limited reproductive budget, individuals select higherquality mates to ensure survival of their progeny and to contribute more robust genes to future generations (Puurtinen et al., 2009).Mate choice is typically performed by females due to their greater reproductive investment and limited egg number which increases the 'time out' of the mating pool (Rosenthal, 2017).To accurately evaluate mates, females rely on traits that signal male body condition, good genes and/or parenting ability (i.e.sexually selected traits).This selects for increased sexual signalling which allows males to outcompete their rivals for access to females (McCullough & Simmons, 2016a;McCullough et al., 2016b;Lüpold et al., 2019;del Sol et al., 2021), including elaborate courtship displays (e.g. the jump-snaps of golden-collared manakins, Manacus vitellinus, or the gular pouches of frigatebirds, Fregata magnificens; Chastel et al., 2005;Schlinger & Chiver, 2021) and/or weapons (e.g. the antlers of white-tailed deer, Odocoileus virginianus, or the claws of fiddler crabs, Uca pugilator; Allent & Levinton, 2007;Gomes et al., 2023).These signal elements are eventually delivered through the physical environment to females which use a combination of sensory modalities to assess the overall quality of potential mates (Candolin, 2003;Santac a et al., 2021;Shohet & Watt, 2004).Alterations to the physical environment may interfere with mate choice processes (reviewed in Candolin & Wong, 2019), leading to increased costs of processing signal information (Vasconcelos et al., 2007) and/or reduced efficiency in mate searching (Bird & Parker, 2014).Consequently, the disruption of sexual signalling due to altered environments can severely impact an individual's reproductive success.This affects not only males, which may struggle to attract a mate, but also females, which may face challenges in selecting an optimal partner (Candolin, 2019;Heubel & Schlupp, 2006).
In aquatic ecosystems, turbidity causes a decrease in light availability underwater (Anthony et al., 2004;Davies-Colley & Nagels, 2008).The reduced availability of visual cues is likely to moderate the direction and strength of sexual selection.For example, when turbidity levels are high, male three-spined sticklebacks, Gasterosteus aculeatus, increase their mating effort to attract females, which strengthens sexual selection on male courtship displays (Engstr€ om-€ Ost & Candolin, 2007).On the other hand, female sticklebacks in turbid water rely more on olfactory, rather than visual, cues to assess males (Heuschele et al., 2009).Likewise, a weakened mate preference for visual sexual signals has been demonstrated in females of Lake Victoria cichlids, Pundamilia nyererei, (Maan et al., 2010) and the sand goby, Pomatoschistus minutus, (J€ arvenp€ a€ a & Lindstr€ om, 2004).In three-spined sticklebacks, this has been shown to relax sexual selection on male nuptial coloration and courtship display, with a subsequent reduction in offspring viability (Candolin et al., 2016), highlighting the role of visual signals in allowing females to select males that provide greater benefits.Notably, however, Candolin et al. (2016) also revealed greater hatching success and faster offspring growth under turbid conditions, indicating the mixed effect of turbidity on population dynamics.More research is needed to understand how females adjust their mate preference in turbid water, and whether such adjustments could enhance or relax the selective pressures on male sexually selected traits.
Guppies have been used as a model organism for answering questions in both visual ecology (Endler, 1980;Sandkam et al., 2016) and sexual selection (Brooks & Endler, 2001;Farr, 1977;Kodric-Brown, 1983).The nature of the guppy mating system makes them especially apt for experiments on mate choice.Males have conspicuous colour patterns on their body, with orange, black and iridescent patches which the males show off to females during 'sigmoid displays' courtship (Houde, 2019).Females are known to prefer more colourful males (Kodric-Brown, 1983;Godin and Dugatkin (1995); Locatello et al., 2006;Karino et al., 2010;Houde, 2019), and this preference varies depending on the extrinsic environment, including light environment (Gamble et al., 2003), visual background (Endler et al., 2022) and predation risk (Godin & Briggs, 1996).Receptive females are more responsive to males in environments where their visual contrast is greater, and courting guppy pairs orient themselves to maximize the visual contrast of the male (Cole & Endler, 2016;Endler et al., 2022).These findings imply that visibility of the male plays an important role during guppy courtship.Thus, it is plausible that female guppies may adjust their mate preference in turbid water due to the increased difficulty of detecting visual cues and the lower reliability of received visual signals due to reduced contrast with the environment.However, while evidence shows that males increase mating effort in turbid water (Ehlman et al., 2018), and that males reared in turbid water possess more intense coloration (Camargo-dos-Santos et al., 2021), the effect of water turbidity on mate choice of female guppies remains poorly studied.
Here, we conducted mate choice trials to investigate whether female guppies adjust their mate preference according to ambient levels of water turbidity when given a chance to choose between males with relatively high and low amounts of orange coloration on their bodies (i.e.proportion of body area that is orange).We also examined the repeatability of the mate preference by running the trial twice under the same turbid conditions.Given previous studies suggesting that water turbidity weakens female preference for visual sexual signals (Candolin et al., 2016;J€ arvenp€ a€ a & Lindstr€ om, 2004;Maan et al., 2010), we first predicted that when choosing between males that differed in their amount of orange coloration, females would prefer the more colourful male in low turbidity water, but this preference would weaken as the level of turbidity increased.As repeatability could reflect the accuracy of detecting the same preferred males, we further predicted lower repeatability with increasing water turbidity (Karino et al., 2010) because water turbidity should increase the difficulty of accurately recognizing the preferred colourful males.Alternatively, if females relied more on other cues (e.g.olfactory) in turbid water, there would be neither an effect of turbidity on female mate preference nor a reduction in repeatability with increasing water turbidity.This is because other cues should remain unaffected by the levels of turbidity and should enable females to accurately and repeatedly recognize preferred males.

Fish Origin and Maintenance
All guppies used in this experiment were descendants of fish collected from two independent laboratory stocks which were originally captured from Alligator Creek near Townsville, Queensland, Australia (19.45 S, 146.97 E;Lindholm et al., 2014;Cole & Endler, 2016).Our stock population has been maintained in aquarium facilities at the Australian National University since 2019.Males and females used in this study were virgin individuals that were separated into single-sex stock tanks before reaching maturity (identified by visible eggs for females and an elongated anal fin for males).Aquaria were kept at 26 ± 1 C with a light:dark cycle of 14:10 h.Fish in stock tanks were fed with flake food in the morning and Artemia sp.nauplii in the afternoon.Experimental fish housed in individual 1-litre tanks were fed with Artemia sp.nauplii twice daily.

Male Selection and Pairing
Given the known effect of orange coloration on male attractiveness (Endler, 1983;Evans et al., 2004a;Grether et al., 2005;Houde, 1987;Karino et al., 2010), we paired males that differed in the proportion of orange coloration on their body to determine how varying water turbidity affects female choice.For pairing, we randomly selected virgin males (N ¼ 180) from stock tanks and anaesthetized them using Aqui-S (0.0075% v/v) for 10e15 s (https:// www.aqui-s.com/).Males were then placed on their right side and photographed using a digital camera Tough TG-6 (Olympus, Japan) against a predominantly white background under constant lighting.Male standard length (snout tip to base of caudal fin), body area (mm 2 ) and the area of orange spots (mm 2 ) were later measured from the photographs using ImageJ (Abr amoff et al., 2004).We paired males with a relatively large proportion of their body covered in orange (0.131 ± 0.002) with males that had a relatively small proportion of their body covered in orange (0.0783 ± 0.002; Fig. A1), hereafter referred to as colourful and noncolourful, respectively, to produce 90 pairs of males for use in mate choice trials (described below).The males in these pairs differed significantly in the proportion of body area that was coloured orange (paired t test: t 179 ¼ 47.974, P < 0.001), but they did not differ in standard length (paired t test: t 179 ¼ À0.271, P ¼ 0.787).After the proportion of orange was measured, males were housed in individual 1-litre tanks prior to, and in between, mate choice trials.We randomly assigned pairs of males to one of the three treatment tanks (i.e.low, mid or high turbidity; N ¼ 30 pairs per treatment).

Manipulating Turbidity
We generated varying levels of water turbidity using bentonite clay (Earths Purities, Australia, https://earthspurities.com.au), an additive commonly used for this purpose in studies of fish behaviour (Allibhai et al., 2023;Ehlman et al., 2018;Hartman & Abrahams, 2000).We first determined the required quantity of clay using a standard curve of the bentonite concentration against turbidity in nephelometric turbidity units (NTUs; Warzecha et al., 2018).We measured the absorbance at 750 nm at five separate locations within each tank using the FLUOstar Omega microplate reader (BMG LabTech, Ortenberg, Germany), and the obtained values were then converted to NTUs using the equation: 0.191 þ 926.1942Â A 750 (Goodner, 2009).Three treatments were created: low turbidity (0 NTU: no bentonite clay was added), mid turbidity (ca. 25 NTU: 1.539 g of bentonite clay was added to 17 litres of water), high turbidity (ca.50 NTU: 3.078 g of bentonite clay was added to 17 litres of water; Fig. A2).The turbidity values were chosen based on the average NTU range of 40e60 NTU in turbid Trinidadian streams (Luyten & Liley, 1985).We agitated the water using a pump and airstone to prevent the clay from settling before each trial.

Mate Choice Trials
We conducted two-choice trials to investigate whether water turbidity affected female preference for males with a greater proportion of orange coloration.For each treatment, 30 females chose between a colourful and noncolourful pair of males (total number of females ¼ 90).Trials were conducted in a 17-litre tank (50 Â 20 cm and 20 cm high), where we placed a mesh barrier and a removable opaque barrier at each end of the tank to create two end compartments and a central compartment (Fig. 1).Prior to the mate choice trial, a colourful and a noncolourful male from each pair were introduced separately into each of the end compartments, with a virgin female introduced in the central compartment.After a 10 min acclimation period, we removed the opaque barriers to allow the female to inspect and approach the males for 10 min.Mesh barriers between compartments prevented physical contact between individuals but allowed females to perceive olfactory and visual cues from each male.We recorded the time females spent in each association zone (<8 cm from the mesh barrier).
To assess whether individual female choices were repeatable, we performed a second mate choice trial using the same maleefemale combinations 7e14 days following the first trial.The period of 7e14 days ensured that the mate preference was not impacted by the female's memory from the previous trial (Griffiths & Magurran, 1997).We also alternated the end compartment (in which the colourful male was placed) between trials to prevent the possibility of side biases impacting the results.All trials were performed in a controlled temperature room (26 ± 1 C) between 0900 and 1600 with ambient light provided by overhead fluorescent lights.

Ethical Note
All work conducted for this experiment was approved under the Australian National University Animal Ethics permit A2021/04.Fish were kept in temperature-controlled rooms with alarms that notify when the rooms malfunction.Between experimental trials, fish were housed individually in 1-litre tanks with a gravel substrate and plants to allow them to hide.These 1-litre tanks were transparent, and we ensured that fish had visual access to other individuals when housed alone.The high surface area to volume ratio of water, the low density of fish in tanks and the short duration of time spent in these tanks meant that it was not necessary to provide aeration or filtration.Fish were observed twice daily in these tanks to ensure they swam and foraged normally.
We regularly tested water quality (i.e.pH, ammonia, nitrites, nitrates and hardness), and did half-water changes fortnightly for stock tanks.The water was primed and aged prior to adding it to the tanks.Individual tanks were topped up with aged water from stock tanks when necessary.We used food-grade bentonite clay to manipulate water turbidity.This substance is nontoxic, and only harmful to fish at much higher concentrations than what we used here (i.e.oxygen consumption rates are reduced at concentrations over 3500 NTU ;Horkel & Pearson, 1976).After the study, the fish were returned to the stock tanks.

Statistical Analyses
All analyses were conducted in R version 2023.03.0þ386 using RStudio 4.2.3 (R Core Team, 2020).Results are presented as mean ± SE.Significant differences were considered as P < 0.05 (two-tailed).

Preference for more colourful males
To determine the effect of water turbidity on female choice for more colourful males, we conducted a two-step analysis.First, we tested whether water turbidity and/or trial number affected the likelihood of a female making a choice (i.e.spending time with either of the males) using a generalized linear mixed model (GLMM) with binomial error (the glmmTMB function in the 'glmmTMB' package 1.1.3).Second, of those females that did inspect males, we analysed the proportion of time a female spent with the colourful male using a GLMM (beta regression).In this second analysis we specified a 'beta_family' as our error distribution and when necessary substituted values of 0 and 1 with 0.00001 and 0.99999, respectively.Trials where females did not spend time near either male were excluded from the second analysis.Thus, the final sample size of the second analysis consisted of 84 individuals for the first trial (low turbidity: N ¼ 26; mid turbidity: N ¼ 30; high turbidity: N ¼ 28) and 81 for the second trial (low turbidity: N ¼ 27; mid turbidity: N ¼ 30; high turbidity: N ¼ 24).Notably, some females only spent time in one of the association zones, including 37 females in the first trial (low turbidity: N ¼ 13; mid turbidity: N ¼ 10; high turbidity: N ¼ 14) and 31 females in the second trial (low turbidity: N ¼ 12; mid turbidity: N ¼ 12; high turbidity: N ¼ 7).These females might have either exhibited a very strong mate preference or simply not noticed one of the males throughout the trial.As we were unable to rule out these possibilities, we further ran an analysis that included only females that

Mesh barrier
Opaque barrier 20 cm 50 cm 8 cm

Preference zone
Figure 1.Layout of the tank.A virgin female was placed in the centre section, with a male (either colourful or noncolourful) placed at either end.Our experimental design consisted of 90 females (30 in each turbidity treatment) choosing twice between a pair of colourful and noncolourful males.
spent time in both association to examine the robustness of our results.
In both analyses, water turbidity (three levels: low, mid, high) and trial number (two levels: first, second) were included as fixed factors (we did not model the interaction between these fixed factors), and female identity was included as a random factor to account for repeated measurements of each female.We conducted dispersion tests (testDispersion function in the 'DHARMa' package 0.4.6) to confirm the data variance was not overdispersed.Wald chi-square tests (type II sums of squares) were performed to calculate the P value using the Anova function in the car package 3.1.0.If the effect of turbidity was significant, we ran Tukey's pairwise comparison tests (emmeans function in the 'emmeans' package 1.7.4.1) to determine how mate preferences between water conditions differed from each other.Additionally, we ran onesample t tests (t.test function) to examine whether the mate preference (i.e. the proportion of time spent with the more colourful male) differed from that expected by chance (i.e.0.5).

Repeatability of female mate choice
To determine whether water turbidity affected the repeatability of female mate choice, we calculated repeatability estimates using the rpt function (datatype ¼ Proportions, link ¼ logit) from the 'rptR' package 0.9.22 (Nakagawa & Schielzeth, 2010).To control for fixed effects in the model, we calculated adjusted repeatability across all the treatments and for each treatment separately (Nakagawa & Schielzeth, 2010;Stoffel et al., 2017).Female identity was used as the grouping factor.Confidence intervals (CIs) for repeatability estimates calculated from parametric bootstraps (N ¼ 1000) that create distributions for likelihood ratios are reported.We tested statistical significance of the repeatability using likelihood ratio tests, and repeatability estimates for each treatment were compared using the overlapping CI method (O'Neill et al., 2018;Schuster et al., 2017).Depending on the ratio of standard errors, CIs between 83 and 85% were calculated.A significant difference between repeatability estimates were considered if there was no overlap in their CIs (Payton et al., 2003).

RESULTS
There was no effect of turbidity treatment or trial number on whether, or not, females spent time with either male (binomial error: turbidity: c 2 2 ¼ 0.031, P ¼ 0.985; trial number: c 2 1 ¼ 1.772, P ¼ 0.183).

Effect of Water Turbidity on Female Mate Choice
When considering all trials in which females made a choice (i.e.associated with at least one male) water turbidity significantly affected female preference for colourful males (c 2 2 ¼ 13.343, P ¼ 0.001; Fig. 2a; Table A1).Females spent significantly more time near the colourful male in low turbidity water, compared to both the mid turbidity (Tukey's test: P ¼ 0.006) and high turbidity conditions (Tukey's test: P ¼ 0.012), while there was no difference between mid and high turbidity treatments (Tukey's test: P ¼ 0.980).The stronger preference in low turbidity water (73.7%) was significantly different from 50% (one-sample t test: t 52 ¼ 5.521, P < 0.001).However, the preference in mid turbidity (48.8%) and high turbidity water (49.8%) did not differ from that expected by chance (one-sample t test: mid turbidity: t 59 ¼ À0.268, P ¼ 0.790; high turbidity: t 51 ¼ À0.031, P ¼ 0.976).
When considering only females that inspected both males the effect of turbidity on female choice became nonsignificant (c 2 2 ¼ 3.320, P ¼ 0.190; Fig. A3, Table A2).This appears to be due to a weaker preference for colourful males in low turbidity water (60.9%),rather than a change in preference in mid turbidity (53.3%) or high turbidity water (48.1%), as preferences in the latter two treatments were similar to those from the analysis when considering all females (see above).Despite this nonsignificant effect of water turbidity, the mate choice of females (that spent time with both males) in low turbidity water remained significantly different from 50% (one-sample t tests: t 27 ¼ 2.478, P ¼ 0.020), and preferences in both mid turbidity (t 37 ¼ 0.845, P ¼ 0.403) and high turbidity conditions (t 30 ¼ À0.380, P ¼ 0.707) remained random.

Effect of Trial Number on Female Mate Choice
When considering all trials in which females made a choice (i.e.associated with at least one male) trial number did not significantly affect female preference for colourful males (c 2 1 ¼ 0.103, P ¼ 0.748; Fig. 2b, Table A1).Despite the lack of difference between the first and second trial, however, females across all water conditions exhibited a significant preference for colourful males in their second trial (59.6%; one-sample t test: t 80 ¼ 2.429, P ¼ 0.017) but not in their first trial (54.7%; one-sample t test: t 83 ¼ 1.121, P ¼ 0.265; Fig. 2b).When only trials where females inspected both males the effect of trial number on female choice remained nonsignificant (c 2 1 ¼ 0.785, P ¼ 0.376; Fig. A4, Table A2) and the percentage of time spent with colourful males in both the first trial (50.9%) and second trial (56.6%) was not significantly different from 50% (one-sample t test: first trial: t 46 ¼ 0.247, P ¼ 0.806; second trial: t 49 ¼ 1.900, P ¼ 0.063).

Effect of Water Turbidity on Female Mate Choice Repeatability
When examining the female preference for colourful males in each water condition separately, we found that repeatability estimates were only significant in low turbidity and high turbidity water, but not in mid turbidity water (Table 1).Similarly, our pairwise comparisons showed that repeatability did not differ between the low and high turbidity treatments, but both were significantly different from that of the mid turbidity treatment (Table 2).

DISCUSSION
In this study, we explored the effects of turbidity on female mate choice in the guppy.Understanding turbidity-induced behavioural changes during sexual interactions is important because of its potential to impact future generations and to shape the evolution of traits used in mate recognition.However, the impact of turbidity on female mate choice in guppies has not been well studied, despite the fact that their natural habitats vary considerably in turbidity (Luyten & Liley, 1985).Here, we manipulated water turbidity and looked at how this affected female preference for colourful males.We predicted that based on the natural preference of female guppies towards males with more orange coloration (Houde, 2019;Karino et al., 2010), females would prefer colourful males in low turbidity water.Further, we predicted this preference for colourful males would become weaker and less consistent (i.e.show lower repeatability) as water turbidity increased due to reduced visual contrast between male coloration and the environment in more turbid conditions.
Our results confirmed our hypothesis that perturbation of the visual environment affects female mate choice in guppies.We found that turbidity did not influence the likelihood of a female inspecting males, but female preference for colourful males was significantly weaker in mid and high turbidity water compared to low turbidity water.While females preferred more colourful males in low turbidity water, females in the mid and high turbidity water treatments did not exhibit a preference.Interestingly, there was no difference in preference between mid and high turbidity conditions.A decrease in mate preference for visual sexual signals in turbid water has also been reported in three-spined sticklebacks (Candolin et al., 2016;Engstr€ om-€ Ost & Candolin, 2007;Heuschele et al., 2009).These studies found a significant effect at much lower levels of turbidity (6e15 NTU) than those used in the current experiment (25 and 50 NTU).Together our results suggest that female mate choice for coloration in fish can be disrupted at low turbidity levels that are well within the range of those experienced naturally.As such, water turbidity may be an important factor driving population variation in male coloration and female preferences for this trait (Brooks, 2002).
Effects on the ability of females to locate, assess and eventually choose more colourful males could have population level consequences.More colourful males are generally expected to make higher quality mates because they offer females both direct and indirect benefits (Herdegen-Radwan et al., 2021;Przesmycka et al., 2023).Females may benefit by mating with more orange males because these males are less likely to harbour parasites that can be transmitted to the female during mating, and the offspring may inherit parasite resistance genes (Houde & Torio, 1992;Stephenson et al., 2020).Females may also benefit indirectly by obtaining genes for their offspring that code for increased foraging efficiency that is associated with greater coloration (Karino et al., 2007).In addition, it has been shown that the offspring of males with larger orange spots are more successful at evading predators than those of males with smaller orange spots (Evans et al., 2004b).Taken together, these findings demonstrate how female guppies can significantly improve their own and their offspring's fitness by choosing mates with more orange coloration.The potential consequences of water turbidity on fitness that may result from inhibited mate choice could be alleviated if females are able to switch sensory modalities that are less affected by water turbidity.For instance, a study on sticklebacks showed that females switched from relying on visual cues to olfactory cues for mate choice when in turbid water (Heuschele et al., 2009).In our study, mesh barriers were employed to separate males and females, enabling the use of olfactory cues for female mate choice.Assuming that olfactory cues remain unaffected by levels of turbidity, our finding of a significant effect of turbidity on mate choice suggests that olfactory cues might not play a major role in the female's ability to choose more colourful males.
Although the exact role of chemical communication in mate choice is not well understood in guppies, previous experiments have shown that male and female guppies can use chemical cues to locate mates and ascertain their quality.In particular, male guppies are able to discern the reproductive status of females from chemical cues (Saoshiro & Karino, 2023) and female guppies show preferences for males based on olfactory cues, although this contrasts with choices made using visual cues (Shohet & Watt, 2004).If females are using olfactory cues to discriminate male quality in our experiment, our results suggest that either these cues signal different aspects of quality than visual signals or that olfactory cues were also affected by our water turbidity treatment.Future studies that disentangle female reliance on olfactory versus visual cues under turbid conditions (e.g. using water-permeable opaque barriers between males and females; Chung et al., 2019) would be insightful.
Our use of a two-choice experimental paradigm means that our experiment is detecting the relative preference for a male (Dougherty & Shuker, 2015).A substantial number of females in our experiment spent time inspecting only one of the males in the choice trial.Data from these females is difficult to interpret because it means that either they may be exhibiting very strong preferences,  Treatments are considered significantly different if the confidence intervals (CIs) do not overlap.The percentage values refer to the percentage of the confidence intervals, that is, how confidently we can say that the true value falls within that range and L, M, H refer to the low, mid and high turbidity conditions, respectively.The ratio of SEs and the CIs were calculated using methods outlined in Payton et al. (2003).*P < 0.05.
or they may simply been unaware of the male housed at the opposite end of the tank.When excluding trials where females spent time with only one male, the effect of turbidity on mate choice disappeared.Did an unawareness of another male lead to the significant difference in female choice between the turbidity treatments?If the females were unaware of the male at the other end of the tank, we would expect this to occur more often in more turbid water, which was not the case.Closer inspection of the difference in results when comparing all trials to those where both males were inspected reveals a reduced preference in the low turbidity water (73.7 versus 60.9%) with a similar preference in the mid (48.7 versus 53.3%) and high (49.8versus 48.1%) turbidity treatments.This suggests that the lack of effect when excluding trials where only one male was inspected is most likely due to the exclusion of females with strong preferences in low turbidity water and/or a reduction in statistical power.Indeed, of a total of 13 females that spent time with only one of the males in low turbidity water, 11 strongly preferred the colourful males in the first trial and 10 of these 11 females again favoured the same colourful males in their second trial.Our results thus suggest that females in low turbidity water were capable of accurately assessing male quality from a distance and did not require proximity to potential mates.Increased time evaluating mates in more turbid water may lead to fitness consequences for females, either by spending more energy or by being more exposed to predators (St€ ohr, 1998).
In addition to effects of water turbidity, we also found some evidence that female preference for colourful males became stronger in the second trial, following a period of 7e14 days in isolation.Although female preferences did not differ significantly between the first and second trial, when considering all trials where females associated with at least one male, females preferred more colourful males in the second trial but not the first.There may be several explanations for this result.First, females might have been more familiar with the mate choice trial set-up and thus more inclined to exhibit a choice (Godin & Briggs, 1996).Alternatively, females might remember males from the previous trial and were able to assess their quality more easily.If so, we would expect more females to spend time with only the preferred males in the second trial, but we instead found a similar number of females spending time with only one of the males (N ¼ 47 for the first trial, 50 for the second trial).Female guppies have previously been shown to prefer novel males, which suggests some memory of previous encounters with males (Graber et al., 2015;Potter et al., 2023).However, other studies show that in a schooling context, familiarity takes 12 days to develop in guppies (Griffiths & Magurran, 1997).Although we cannot identify the underlying mechanism, the stronger preference in the second trial underscores the potential for females to refine their mate choice when encountering the same pair of males again, and the period of 7e14 days in isolation did not erode this potential.Conducting more trials per female with varying time intervals would help assess habituation to the experimental conditions and female memory for male quality.
We expected repeatability of female mate choice to be high in low turbidity water and to decrease with increasing turbidity.In mosquitofish, Gambusia holbrooki, male mate choice was found to be repeatable when males were allowed to interact freely with females in clear water (Hoysak & Godin, 2007).Further, a previous study on guppies, using a similar experimental design to what we used here (i.e.binary mate choice between a colourful and noncolourful male), also showed that females repeatedly preferred brightly coloured males in clear water when tested across 2 successive days (Godin & Dugatkin, 1995).Our study shows that repeatability of mate choice in low (or clear) turbidity water is maintained across longer time periods but not necessarily across environmental contexts.Specifically, mate choice was not repeatable in the mid turbidity treatment but was in the high turbidity treatment.Since female choice exerts strong selection on male coloration (Kodric-Brown, 1985), the low repeatability of mate choice in the mid turbidity water could lead to reduced selection on male sexual traits.While the preference was significantly repeatable in high turbidity water, note that it did not deviate from random chance.The high repeatability of the absence of mate preference still does not confer advantages in terms of selecting higher quality males to elevate reproductive value of their offspring.Future research should investigate the long-term effects of turbidity on female mate choice, specifically with respect to how it can influence the evolution of male coloration.

Conclusion
In summary, we found (1) a significant effect of water turbidity on mate choice with females preferring more colourful males in low turbidity water, but choosing at random with respect to male colour in the mid and high turbidity treatments; (2) female mate choice showed low repeatability in mid turbidity water, further suggesting weakened selective pressures on male coloration; (3) an increased preference for colourful males during the second trial, possibly indicating a more accurate assessment of male quality upon subsequent encounters.Our results highlight the importance of the visual environment for allowing female guppies to choose high quality mates and suggest that increased turbidity associated with anthropogenic activities is likely to disrupt the mating system of fish that rely on visual cues in sexual selection.

Figure 2 .
Figure 2. (a) The effect of turbidity on female choice when considering all females (low turbidity: N ¼ 53; mid turbidity: N ¼ 60; high turbidity: N ¼ 52).(b) The effect of trial number on female choice when considering all females.The mean and SE for each treatment are shown with individual data points.The data for all treatments are combined.The lines at 0.5 indicate a random choice.The shaded areas represent the distribution of the data points.

Table 1
Schuster et al. (2017)e mate choice for the colourful males.All values presented are original-scale approximations.In cases where the confidence interval (CI) includes zero but the repeatability estimates are high and P < 0.05, evidence for repeatability is outlined inSchuster et al. (2017).

Table 2
Pairwise comparisons of repeatability between treatments.