Phenotypically plastic mating behavior may allow males to modify their reproductive behavior to suit the prevailing social conditions, but we do not know if males only react to immediate social stimuli or change their inherent mate preferences according to their social history. Here we examine the effect of social experiences on the subsequent reproductive behavior of male guppies under standard conditions, allowing us to distinguish the effect of past and immediate social conditions. Males experienced experimental conditioning periods during which they interacted with three females, either of variable size or of similar size. Females arrived either simultaneously or consecutively. In subsequent standard assays, only males that had experienced females of variable size preferentially courted large females. Further, males exposed to sequential female arrival courted subsequent females more vigorously than males that had experienced simultaneous female arrival. In contrast, males did not alter their coercive mating attempts in relation to their recent social history. These results demonstrate that males use past experiences to modify their subsequent reproductive behavior rather than reacting only to immediate stimuli, and reveal the sophisticated ways in which males alter their reproductive tactics to suit the social environment and maximize fitness across changing selective landscapes.

Since the early 1990s there has been a growing acceptance that phenotypic variation in mate choice may influence the mode, direction, and strength of sexual selection on male traits (Partridge 1988; Ritchie 1992; Jennions et al. 1995; Murphy and Gerhardt 2000; Brooks and Endler 2001; for an excellent review, see Jennions and Petrie 1997). For the most part, however, this work has concerned mate choice variation among females; such variation is significant because a male does not earn matings as a consequence of his mean attractiveness to all females, but rather because of his attractiveness to individuals. Thus a male that is unattractive to most but strongly preferred by a few females may have higher mating success than a male that is moderately attractive to all females, even if the latter's mean attractiveness is higher.

Female preference functions (for certain traits in males), or choosiness (as effort toward finding a preferred mate) have been shown to covary with a range of factors including female condition (David et al. 2000; Hingle et al. 2001), parasite load (López 1999), social environment (Cornwallis and Uller 2010), experience of previous mates (Rosenqvist and Houde 1997; Pitcher et al. 2003), and predation risk (Godin and Briggs 1996; Gong and Gibson 1996). This plasticity in mate preferences may be adaptive if it allows females to adjust their mate choice behavior to suit the prevailing social conditions (Qvarnstrom 2001). When the fitness returns of mating with a partner of a certain phenotype vary with changing social conditions (Qvarnstrom et al. 2000), plasticity in female preferences may be subject to positive selection if it allows females to maximize their fitness over separate mating attempts (Chaine and Lyon 2008). This variation can have important consequences for the evolution of male traits, including dampening the net sexual selection that operates over a breeding season (Brooks 2002; Chaine and Lyon 2008; Lehtonen et al. 2010) and maintaining genetic variance within populations.

In recent years, the focus on female choice is being balanced by reports of male mate choice. The nature and targets of male mate choice are often different from the more intensively studied examples of female mate choice, and the theoretic issues are also somewhat divergent. Male choice is often expressed in terms of decisions about which female a male should court, and how intensely to do so. Often these preferences are followed by female choice; accepting or ignoring the courtship attempt. Male preferences for larger or more fecund females have been found in moths (Xu and Wang 2009), newts (Verrell 1985) finches (Jones et al. 2001), Drosophila melanogaster (Byrne and Rice 2006), and many other groups (review; Bonduriansky 2001). Males may also prefer to mate with females based on other traits, such as female coloration (Amundsen and Forsgren 2001), mating status (Guevara-Fiore et al. 2010), pheromonal signature (Andrade and Kasumovic 2005), or likelihood of successful insemination (Preston et al. 2003).

There is also mounting evidence that, just as for female preferences, male preferences are not fixed, but may change according to local conditions. Males that can alter the direction and intensity of courtship behavior according to the level of competition from rival males, as well as their own relative competitive ability, are predicted to outperform males with fixed preferences (Fawcett and Johnstone 2003; Härdling and Kokko 2005; Venner et al. 2010; Jordan et al., unpubl. ms.). Males may use local information to modify their reproductive strategies depending on factors such as competition from rival males (Shine et al. 2003; Bel-Venner et al. 2008; Candolin and Salesto 2009), internal condition (Kotiaho 2002), the abundance of mates (Svensson et al. 2010), or the reactions of courted females (Patricelli et al 2002; Royle and Pike 2010). However, because these studies tested the behavior of males under variable social conditions, they could not determine whether males modified their behavior in relation to immediate social conditions, or if social experience actually changed the mate preferences that males then expressed in subsequent encounters.

Evidence from spiders and crickets (Kasumovic and Andrade 2006; Bailey et al. 2010) suggests males can alter their future reproductive behavior based on past experiences, changing developmental trajectories in response to their anticipated adult social environment. The changes observed in these invertebrates are developmental, affecting males’ body size, condition and gonad size at eclosion, and hence are nonreversible. For species in which male reproductive strategy involves decisions about allocation of courtship effort among a range of potential mates, the manner in which past social experience affects subsequent behavior is far less clear.

There is an important conceptual distinction between reproductive tactics that continually change according to immediate stimuli, and plastic preferences that are altered by social experience but transferable to subsequent mate encounters. In the former case, all individuals would be expected to react similarly to a standard stimulus regardless of recent experience, whereas in the latter there may be true variation in mate preferences even under standard conditions. We do not yet know whether such persistent variation in mating and courtship preferences is possible or if individuals are only able to react to stimuli with which they are immediately presented. This is a crucial question that must be addressed to understand the evolution of plasticity in mating behavior.

Here we examine the effects of different social environments on subsequent mating behavior and mate preferences in male guppies, Poecilia reticulata under standard conditions. Guppies are particularly useful for examining these questions, both because a great deal is known about female preferences (Bischoff et al. 1985; Kodric Brown 1985; Houde 1987; Houde and Endler 1990) and how these preferences may shift (Rosenqvist and Houde 1997), and also because male guppies have a preference for courting larger, and hence more fecund females (Abrahams 1993; Houde 1997), as well as changing their behavior in response to the social environment (Jirotkul 1999a; Jirotkul 1999b). We test whether male P. reticulata preference change in response to the variation in quality of potential mates, the abundance of potential mates, or a combination of these factors. By experimentally manipulating male's social histories and then comparing their behavior under standard conditions, we experimentally decouple the effects of recent and immediate social conditions on male courtship strategies and preferences.


Guppies, P. reticulata, used in experiments were descendents of feral fish collected at Alligator Creek, 30 km southwest of Townsville, Australia. Each fish was fed every day with newly hatched Artemia salina, and maintained at 26°C and 12 h light per day. Males and females used in studies were housed in mixed-sex tanks containing between 20 and 30 adult fish prior to inclusion in experiments, and were therefore nonvirgin. Females that had given birth within 3 days of collection were not used in experiments, because postparturient females are known to be more receptive to male mating attempts than females that have not recently given birth. Females were collected and sorted into three size classes, small (25–30 mm SL), medium (30–35 mm SL), and large (35–40 mm SL). To control for effects of male size, we only used males between 23- and 27-mm standard length. Throughout the experiment, fish were housed at 25°C, with light regime of 2 h simulated dawn (1.5 μmol/sec/m2) followed by 8 h full light (8.7 μmol/sec/m2), followed by 2 h simulated dusk (1.5 μmol/sec/m2), followed by 12 h of darkness.

We sought to determine the effect of social environment during a six-day experimental period on male reproductive behavior in a subsequent standardized assay. Social environments in our experiments differed in two ways, firstly in the variation in potential mate quality and second in the encounter rate with potential mates. We manipulated variation in female quality by housing males with either three medium-sized females (i.e., no variation), or with one female of each size class (i.e., high variation). We manipulated encounter rate by either presenting a male with all three females throughout the six-day conditioning period (i.e., simultaneous encounters), or by presenting only one female at a time (i.e., sequential encounters).

We used 12 males per treatment, without repetition of males, in the following four treatment combinations: “Sequential/variation”: one male was housed with one female in a 4-L tank (density 1 fish/2 L water). The female was replaced after simulated nightfall each day with a female of a different size that had not previously been used in any trial. Females of different size classes were presented in a random order, with the constraint that each male was housed with each female size class twice in the conditioning period. “Sequential/no variation”: as for sequential/variation, but introduced females were always of medium size. “Simultaneous/variation”: one male housed with three females simultaneously (small, medium, and large) in an 8-L tank (density 1 fish/2 L water, identical to “sequential” treatments). All three females were replaced after simulated nightfall each day with three new females (small, medium, and large) that had not previously been used in any trial. “Simultaneous/no variation”: as for simultaneous/variation, except that the three introduced females were always of medium size.

We initially conducted treatments sequential/variation, sequential/no variation, and simultaneous/variation. It became apparent that simultaneous/no variation would also provide useful information, and seven days after the end of observations the first experiment, we reran simultaneous/variation and added simultaneous/no variation. Although the treatments were separated in time, a one-way ANOVA of the treatment combinations simultaneous/variation (experiment I) and simultaneous/variation (experiment II) showed no significant differences for any response variable (proportion of consensual mating attempts F1,22= 0.196, P = 0.662; total consensual mating attempts F1,22= 0.017, P = 0.898; proportion of coercive mating attempts F1,22= 0.031, P = 0.862; total coercive mating attempts F1,22= 0.047, P = 0.831). Because there were no temporal effects, we directly compared the treatment combinations sequential/no variation, sequential/variation, simultaneous/no variation, and simultaneous/variation (experiment II) in a single statistical analysis, described below.


Behavioral assays were conducted on the seventh day (i.e., the day after the last day of exposure to stimulus females) after removing stimulus females the previous night. Trials were conducted within 90 min of the start of simulated dawn, for 10 min per trial. Six behavioral assays were conducted per day. The observed fish were not able to see the observer, who was in a darkened section of the room. One female of each size class was taken from the holding tanks, introduced into one of two 100-L observation tanks, under the same conditions as the holding tanks, and allowed to settle for 10 min prior to a male being introduced. A male was then introduced to the same tank and allowed to settle for 1 min prior to measurements. During the 10-min observation period, we measured the number of consensual mating attempts (sigmoid displays), and the number of coercive mating attempts (gonopodial thrusts and nips; for detailed descriptions of these behaviors, see Baerends et al. 1955) directed to each of the three females. At the end of each assay, males were removed and females allowed to settle for a further 10 min while observations were conducted in the alternate tank. Three males were tested in each testing tank each day, and day, order of testing, and testing tank were included in statistical models.


Behavioral observations were recorded using JWatcher 1.0 (Blumstein et al. 2006), and analyzed using PASW 18.0 (SPSS Inc., Chicago, IL). For each experiment, we fitted a separate two-factor ANOVA for the total number of consensual mating attempts, the total number of coercive mating attempts, the proportion of consensual mating attempts made toward large females, and the proportion of coercive mating attempts made toward large females, with sequential/simultaneous arrival of females and no-variation/variation of female size as fixed effects. We also included observation tank, order of observation (i.e., first, second, or third observation that day), and female block (i.e., the group of three females to which the male was introduced in behavioral assays) as random effects, but none of these factors had significant effects and were dropped from the final models. We also compared the proportion of mating attempts to a null distribution of 0.333 (i.e., random choice—one-third of attempts to each of the three females) using one-sample t-tests for each treatment combination. Data for the total number of consensual and coercive mating attempts were nonnormally distributed and were log10-transformed to achieve normality. In all cases residuals were normally distributed.



Female density in the conditioning treatments had a significant effect on the number of subsequent consensual mating attempts in standardized behavioral assays, with males that had experienced sequential female arrival making more consensual mating attempts than males that had experienced simultaneous females arrival (Table 1 and Fig 1A). There was no main effect of variation in the size of recently encountered females, or an interaction with density, on consensual mating attempts (Table 1 and Fig 1A). There were no effects of any treatment level on the total number of coercive mating attempts (Table 1 and Fig 2A).

Table 1.  Results of two-way ANOVAs on four different measures of male courtship behavior (total number of consensual mating attempts, proportion of consensual mating attempts directed to the largest female, total number of coercive mating attempts and proportion of coercive mating attempts directed to the largest female) in relation to the variation in female size and whether the females were presented sequentially or simultaneously. Bold type indicates significant effects (P < 0.05).
 Total consensual mating attemptsProportion of consensual mating attempts toward large femaleTotal coercive mating attemptsProportion of coercive mating attempts toward large female
F,dfPF, dfPF, dfPF, dfP
Sequential versus simultaneous8.048 1,440.007 0.097 1,44 0.7570.461 1,440.5011.690 1,440.2
No variation versus variation0.100 1,440.75467.18 1,44< 0.0010.001 1,440.9733.850 1,440.056
Interaction term0.334 1,440.566 8.375 1,44 0.0060.635 1,440.430.007 1,440.933
Figure 1.

(A) Mean (± SE) number of consensual mating attempts (sigmoid displays) per minute by males that recently experienced sequential or simultaneous arrival of females that did not vary in size (shaded bars) or that varied in size (open bars) (B) The proportion of total consensual mating attempts directed toward the largest of three stimulus females (mean ± 95% confidence interval). Dashed line indicates the predicted proportion of attempts if males divided consensual mating effort equally among all stimulus females, asterisks indicate significant difference from a null distribution of 0.333 using one sample t-tests.

Figure 2.

(A) Mean (± SE) number of coercive mating attempts (sneak copulations) per minute by males that recently experienced sequential or simultaneous arrival of females that did not vary in size (shaded bars) or that varied in size (open bars). (B) The proportion of total coercive mating attempts directed toward the largest of three stimulus females (mean ± 95% confidence interval). Dashed line indicates the predicted proportion of attempts if males divided coercive mating effort equally among all stimulus females, asterisks indicate significant difference from a null distribution of 0.333 using one sample t-tests.


Of the total consensual mating attempts made by males, we measured the proportion that was directed toward large females, as a measure of effect of recent social history on male mating preferences. Recent experience of sequential versus simultaneous arrival of females did not affect the proportion of consensual attempts toward large females (Table 1), but the variation in female size recently experienced by males had a significant effect on subsequent consensual mating effort (Table 1). There was a significant interaction between treatments for the proportion of consensual mating attempts toward large females, indicating that the effect of experiencing variable female size was stronger when females were presented simultaneously than when presented sequentially. Males that had recently experienced variation in female size directed significantly more of their consensual mating attempts toward large females than males that had not experienced variation in female size throughout the conditioning period. In contrast, recent social history did not significantly affect the distribution of coercive mating attempts (Table 1). We also compared the distribution of male mating attempts to a random distribution expected if males were attempting to mate with all three females equally. Males that had experienced variation in female size (whether sequentially or simultaneously) attempted consensual matings with large females significantly more often than expected under a random distribution, whereas males that did not experience variation in female size did not perform consensual mating attempts toward large females any more often than expected under a random distribution (Fig 1B). Coercive mating attempts were directed toward large females significantly more often only by males in the sequential/variation treatment combination (Fig 2B)


Male guppies are famous for their persistent courtship and sneak copulation attempts. With this reputation, we might expect them to court indiscriminately, or at the very least to follow simple facultative rules for courtship. Instead we find that males are sensitive to their social environment, displaying complex modifications to their reproductive behavior in response to their recent social experiences. Both the abundance and the variation in size of females in a six-day conditioning period influenced subsequent male reproductive behavior under standard conditions. When males had recently experienced sequential female arrival, their subsequent courtship effort in standardized assays was greater than that of males that had recently encountered females simultaneously. This suggests that males are assessing the likelihood of encountering mates in the future based on the recent rate of encounter rather than reacting to immediate stimuli. Further, the variation in female quality males had recently experienced strongly modified their preferences for high-fecundity females—only males that had recently seen a range of female sizes showed a significant preference for large females in subsequent standard assays.

In contrast, alternative (coercive) reproductive strategies appeared to follow an opportunistic pattern expected if male guppies indiscriminately attempt sneak copulations. Social history did not significantly change the total coercive mating attempts or the distribution of these attempts, and although males from the sequential/variation treatment directed more coercive attempts at large females than expected at random, there was no overall effect on the distribution of coercive mating attempts of any treatment. Coercive mating attempts in guppies consist of males approaching females from behind and rapidly inserting their gonopodium into the female's gonopore, thereby attempting to circumvent female choice at the precopulatory stage. This strategy avoids the costs of courtship display, and so may not be subject to the pressures leading to prudent allocation of courtship effort. Although coercive mating only rarely results in fertilization, the fitness returns of a coercive strategy would still be higher when directed toward more fecund females.

Female guppies prefer males with higher courtship rates (Houde 1997), but for males, the fitness returns of courtship effort are expected to vary with the social environment. When females are common, missing a reproductive opportunity will not carry a great opportunity cost, but where encounters with females are rare, there is no benefit to withholding effort for subsequent courtship bouts that might never occur (Parker 1974; see also Jordan et al. unpubl. ms.). We observed that males that were continually exposed to multiple females in the conditioning period courted far less than males that experienced only one female at a time. One interpretation is that males are prudently allocating courtship effort according to their perceived chances of encountering females in the near future (Härdling and Kokko 2005; Härdling et al. 2008). Previous work has shown significant lifetime costs of increased reproductive effort in guppies (Jordan and Brooks 2010), and males may seek to court at lower than maximum rate to reduce these costs. Alternatively, males from simultaneous encounter treatments may have had lower energy reserves, and, rather than being prudent, were simply not able to sustain courtship effort at the same level as males from sequential treatments. A similar interpretation has been suggested for the fact that male guppies with high parasite loads court less than males carrying fewer parasites (Kolluru et al. 2009). We attempted to minimize the effect of male exhaustion during the conditioning treatments by keeping total density identical between simultaneous and sequential treatments, but we cannot rule out effects of male exhaustion on subsequent mating behavior. It is also possible that females in the sequential treatment were less able to resist male mating attempts during conditioning, and that it was the reactions of females that altered subsequent male mating behavior (Patricelli et al. 2002; Royle and Pike 2010).

Recent exposure to variation in female quality modified subsequent male preference for larger (high-fecundity) females, with males from the treatment “simultaneous/variation” showing the strongest preference for large females. This treatment combination was identical to the testing conditions, raising the possibility that familiarity with the treatment conditions rather than actual previous experience led to observed effects. However, males from the “sequential/variation” treatment showed similar strong preferences for large females, despite that treatment being different to the testing conditions, and the treatment “simultaneous / no variation,” which was similar to testing conditions, did not elicit similar levels of response. This demonstrates that previous social experience and not familiarity with testing conditions affected subsequent reproductive effort. It is possible that the effects we observed are a result of males needing to see variation in female size to be able to discriminate between females in subsequent encounters, consistent with evidence that female guppies raised in the presence of males with varying amounts of orange coloration develop stronger preferences for orange than females raised with males that lacked variation in color (Rosenqvist and Houde 1997). We found that the effect on consensual mating tactics of seeing variation in female size was stronger in simultaneous than sequential treatments, suggesting males are better at discerning large females in subsequent encounters when they have previously had the opportunity to compare females of differing quality simultaneously.

Male preferences for large females were modified by recent social experience despite the fact that the immediate benefit of courting large females remains unchanged regardless of their previous experiences; a male will always receive higher fitness returns from mating with high-fecundity females, especially when competition is absent. In the wild guppy shoals are open fission–fusion systems in which group membership continually changes (Croft et al. 2004), and attempting to continually assess the immediate social conditions may be prohibitively difficult in such a rapidly changing social environment. Although it would likely be best for a male to tailor his courtship strategy according to the immediate social conditions, this is likely not possible because of constraints on gathering perfect social information. It is more likely that a male's perception of his social environment is built on numerous interactions, and relying on established preference functions may be quicker and more effective than attempting to continually adjust reproductive strategy. This suggests that the differences in male behavior among treatments would likely disappear the longer males were left in the standard conditions of the behavioral assays, although it is not known how long it takes males to match their mating preferences to the current conditions.

Social conditions in nature are subject to rapid changes that can modify the interactions between potential mates and competitors (Oh and Badyaev 2010). By modifying their underlying mate preferences according to recent social history, males may maximize their fitness across changing social environments. The results presented here suggest male mate preferences are plastic in response to changing social environments, as has been shown for female preferences (Chaine and Lyon 2008; Lehtonen et al. 2010). This has consequences for sexual selection acting on both males and females. Just as variation in female preferences can maintain genetic variance for male sexual traits, changing male preferences may alter the selection acting on female phenotypes. Similarly, shifts in male mate preferences may alter the selection acting on males by allowing males to preferentially invest in reproductive encounters that are likely to yield a higher fitness return under the prevailing social conditions. We have shown that male guppies are able to assess the social environment and modify their subsequent mating preferences as a function of past experience. Such a sophisticated response to social conditions suggests a previously undescribed level of nuance and complexity in the process of male mate choice.

Associate Editor: U. Candolin


We wish to thank two anonymous reviewers for comments that vastly improved the manuscript. Funding for the research was provided by an Australian Research Council grant to RB.