- Email: [email protected]
Male interactions and female mate choice in the sand goby, Pomatoschistus minutus
ANIMAL BEHAVIOUR, 2001, 61, 425–430 doi:10.1006/anbe.2000.1596, available online at http://www.idealibrary.com on
Male interactions and female mate choice in the sand goby, Pomatoschistus minutus NUUTTI KANGAS & KAI LINDSTRO }M
Department of Ecology and Systematics, University of Helsinki (Received 16 February 2000; initial acceptance 11 May 2000; final acceptance 24 July 2000; MS. number: A8715)
To examine whether male–male interactions constrain female choice in the sand goby, we compared female preferences expressed under three conditions that varied in the degree of male–male interaction with matings that resulted from free interactions. The experiment consisted of two phases. In the initial phase a transparent divider separated the female from two males. During this stage the males were either isolated by an opaque divider, had visual contact through a transparent divider, or were allowed full interaction by having access to each other’s compartments. The second phase of the experiment was initiated by the removal of all dividers and terminated when the female spawned with one of the males. When there was no initial contact between the males, females spawned with the male they showed a preference for in the first phase of the experiment. In the other two male treatments, females were inconsistent in their spawning choice. Females did not seem to choose dominant males and neither did male aggressive interactions seem to limit a female’s possibilities to spawn with a particular male. The results suggest that male–male interaction may affect the ability of females to evaluate males, in part because of courtship interference between males. Male–male interaction therefore seems to impede female choice and consequently may hinder sexual selection due to mate choice.
aculeatus, from an eastern Canadian population, which show no preference for more aggressive or dominant males (Ward & FitzGerald 1987) and female three-spines from western Canada, whose courtship of ‘more attractive’ males is disrupted by less preferred suitors (McLennan & McPhail 1990). Because of these dynamics, authors have suggested that male–male interactions should be controlled in studies examining female mate choice (e.g. see Bateson 1987; Andersson 1994; Houde 1997). The argument behind this concern is that when males are allowed to interact in a mate choice set-up they may influence each other’s abilities to attract the female. For example, a male may so completely dominate a competitor that it is impossible for the female to assess both males equally. Female choice may therefore be constrained by male–male interactions. This is especially problematic when females prefer male traits that are not associated with increased male dominance (e.g. Hoelzer 1990; Forsgren 1997a). Whether these situations also apply under natural conditions is largely unknown. The importance of male–male interaction in modifying female choice will certainly depend on the spatial distribution of males in the field. In areas where territories are dense one may expect more influence from male–male interactions than in sparse populations.
Sexual selection arises due to an unequal distribution of mating success among individuals of the same sex (Darwin 1871). Among males, mating success may differ because some males are more successful in monopolizing resources needed by females for mating or because some males are more attractive to females as mating partners. These two components of sexual selection are termed intrasexual and intersexual selection, respectively (Andersson 1994). Intrasexual selection is typically prevalent among males, while mate choice (intersexual selection) is usually performed by females. There are two ways in which inter- and intrasexual selection can interact over evolutionary time. In some cases, the two forces will act in the same direction on the same trait, providing mutual reinforcement. For example, females have been shown to prefer dominant males to subordinate ones in pupfish (Kodric-Brown 1995) and at least one population of stickleback (Candolin 1999). In other cases, the two forces might operate on the same trait in opposite directions, thus interfering with, or even cancelling, each other. Examples of this kind of dynamic include female three-spined sticklebacks, Gasterosteus Correspondence: N. Kangas, Department of Ecology and Systematics, Division of Population Biology, P.O. Box 17, 00014 University of Helsinki, Helsinki, Finland (email: [email protected]). 0003–3472/01/020425+06 $35.00/0
2001 The Association for the Study of Animal Behaviour
425
2001 The Association for the Study of Animal Behaviour
426
ANIMAL BEHAVIOUR, 61, 2
Few studies have systematically examined the influence of male–male interactions on female mate choice. The purpose of this study was to examine whether male interactions affect female choice in the sand goby, a small marine fish. In this species males construct nests by excavating the sand underneath rocks or mussel shells. They then attempt to attract females to the nest by courting them. Once a female has been attracted into the nest the male emits courtship sounds (Lindstro ¨ m & Lugli 2000). If a female is interested in a particular male she attaches her eggs to the ceiling of the nest. Nest density varies extensively both between and within populations (Forsgren et al. 1996) and consequently the opportunity for male interference during courtship and female choice can be expected to vary. In our study population nest availability is on average low and there is intense male–male competition for available nest sites (Lindstro ¨ m 1988, 1992). In this study we varied the level of male–male interaction, then compared the result of female choice trials in those different situations. METHODS The experiments were done during May–June 1996 at the Tva¨ rminne Zoological Station, situated on the shoreline of the northern Baltic Sea, the south coast of Finland (60N, 23E). Fish for the experiments were caught from a nearby breeding site using a beach seine. The sexes were kept in separate large storage tanks (150-litre) with a constant through-flow of sea water. During this time the fish were fed ad libitum live Neomysis integer shrimp and frozen Chironomidae larvae. Experimental aquaria measured 704040 cm. Each experimental tank was provided with a 4-cm layer of fine sand on the bottom of the tanks and two artificial nest sites (ceramic tiles measuring 7.57.5 cm) placed at opposite ends of the tank. The tank was divided lengthwise using a transparent Plexiglas divider so that the nests were in the section representing two-thirds of the bottom surface (Fig. 1). Due to the set-up no through-flow of water was provided during the experiment but the water was exchanged between replicates. We simultaneously added two males to the nest section and allowed them 24 h to build nests. If one or none of the males had finished his nest by this time we replaced both males. To minimize any asymmetries in dominance relations between the two males, we selected males of matched body size. After the nest-building phase we added a female to the smaller of the two lengthwise sections (see Fig. 1). Depending on the treatment, males were freely interacting, separated by a transparent barrier, or separated by an opaque barrier (see below). The female had complete visual access to both males because she could freely move over the total length of the tank and the males were free to observe and visually interact with the female. No individual was used in more than one replicate. The fish were returned to their site of origin after the experiment. We conducted experiments in two phases (phases 1 and 2).
B
A
Figure 1. Tank design. The experimental tanks were initially divided into three sections for all treatments except the interaction treatment. The female was placed in the long section and she had visual contact with the males through the transparent divider A. The two males were placed in the two rear compartments separated by the divider B. Divider B was either an opaque or transparent Plexiglas wall except during the interaction treatment, in which divider B was not present. In the second phase of the experiment all dividers were removed and all three fish had free access to all parts of the tank. The grey squares represent the ceramic tiles used as nest sites.
Phase 1 In phase 1 a transparent divider separated the female from the two males, which were either freely interacting, or separated by a transparent or opaque divider (see below). During this phase, which lasted 4 h, the female and males were observed for a total of 50 min. The observation time consisted of five 10-min sessions: immediately, 1, 2, 3 and 4 h after the introduction of the female (all observations were done by N.K.). We recorded the time the female spent on each male’s side. The dividing line was the centre of the tank. The male on whose side the female spent more time was considered to be the preferred male. In addition we recorded the time both males spent courting the female and interacting with each other. Courtship in the sand goby involves fin displays, jerky swimming towards and around the female, lateral display and finally lead swims towards the nest (also see Forsgren 1997b). Male–male interactions in the sand goby are usually aggressive, however, in the present study, we observed no mortality or injuries associated with these interactions. We recorded fin displays, chases, darts against the opponent, lateral displays and physical fighting. Behaviours presented as percentages were obtained by dividing the total time in certain behaviour by the total observation time. Frequencies were obtained by dividing the number of times a behaviour was performed with the observation time. The mean length of courtship bouts was obtained as the total time spent courting divided by the number of times the male courted.
Phase 2 In phase 2 we removed the lengthwise divider separating the males and the female and any other dividers between the males (see below), at which stage all three individuals had physical contact with each other. We observed the behaviour of the female and the two males
KANGAS & LINDSTRO } M: FEMALE MATE CHOICE IN GOBIES
for 20 min immediately after the removal of the dividers and then 10 min each full hour for the first 4 h or until the female had spawned, whichever came first. The experiment was ended when the female spawned.
Phase 1 treatments The two males in phase 1 were presented to the female in three different ways: (1) freely interacting (male–male interaction treatment), (2) separated by a transparent barrier (male–male visual treatment), or (3) separated by an opaque barrier (male–male noninteraction treatment).
Interaction treatment. In this treatment there was no divider between the two males so nothing prevented them from interfering with each other’s courtship or other activities. It should thus be as difficult for the female to independently assess the males in phase 1 as in phase 2. If male interference modifies female choice, then we would not expect the choice performed in phase 1 to differ from the eventual spawning decision made in phase 2.
(1) If males can physically suppress each other’s courtship performance, then the spawning decision in phase 2 should differ most from the choice in phase 1 for the noninteraction and the visual interaction treatments. This is because males in these treatments could not physically interact with each other during phase 1 and therefore female choice would not be affected until phase 2. This also requires that females show consistent preferences in phase 1 and 2 in the interaction treatment. (2) If the outcome differs only for the noninteraction treatment, where no visual or physical interaction was possible between males, then courtship competition is at least equally important in interfering with female choice as is direct male–male interference. This is because in both the interaction and visual treatment, female choice in phase 1 could be affected by male–male interactions and her choice should not be independent of these interactions. RESULTS
Female Preference Visual treatment. The two males were separated by a transparent divider. They therefore had visual contact with each other and could engage in courtship competition (i.e. a situation where the two males’ courtship depended upon one another), as well as threat displays, but one male could not physically dominate or interfere with the other one. A visual interaction between the males may lead to an escalation of courtship. Such courtship competition may cause males to invest more in courtship than they would in the absence of competition, which may interfere with a female’s ability to independently assess male quality. If this is the case then females should show consistent choice in this treatment because courtship competition could be present both before and after the removal of the transparent divider.
Noninteraction treatment. An opaque Plexiglas divider separated the two males, so each male could only see the female, while the female was free to observe each male. In this treatment there was no interaction between the two males and the female should have been able to independently assess them. If the female were to change her choice in the second phase from her initial choice, this would constitute strong evidence that male–male interactions indeed affect a female’s ability to independently express choice. If females show consistent choice in this treatment, this may suggest that female choice is unlikely to be affected by male–male interactions and that females are able to independently assess male quality even in the presence of male–male interactions. It may, however, also be an indication that females use the same traits to assess male quality that make males successful in male–male interactions. A consistent outcome in this treatment would therefore be less conclusive as it would be expected under several different scenarios, depending on the importance and type of male–male interference during courtship. Therefore, to distinguish between these various outcomes, we evaluated the outcomes from the different treatments based on the following predictions.
In 64% of the trials females chose the same male in phase 1 as in phase 2 of the experiment. There were, however, differences between treatments (chi-square test: 22 =5.88, P=0.053). In the noninteraction treatment, where males lacked visual and physical contact, females were most consistent in their choice. In 13 of the 15 replicates, the female spawned with the male she spent the most time with in phase 1 (binomial test: P=0.004). When the males had visual contact with each other, the female showed consistent choice in nine of 15 trials (binomial test: P=0.3). Similarly, in the treatment with physical contact between the two males, the female mated with the same male she spent the most time with in seven out of the 15 replicates (binomial test: P=0.5). Thus, females did not change their initial choice when an opaque divider separated the males and male–male interactions were prevented. In the other treatments, however, females showed extensive inconsistency in their eventual spawning decision. It is possible that females in the noninteraction treatment became fixated on their preferred males during phase 1, and therefore showed a more consistent choice in phase 2 than females in the other treatments. If this were the case, one might predict that females in the noninteraction treatment would also spend more time with their preferred male than females in the other treatments. There was, however, no difference between treatments in the proportion of time that females spent with their preferred males (XSD percentage of time on winner’s side: noninteraction treatment: 74.933.2, N=15; visual treatment: 61.448.0, N=15; interaction treatment: 58.046.0, N=15; Kruskal–Wallis test: H2 =1.773, P=0.412), suggesting that females in the noninteraction treatment were no more fixated on their preferred males than females in other treatments.
Male–Male Interaction In the two treatments in which males could interact (i.e. visual and interaction), interactions between the
427
ANIMAL BEHAVIOUR, 61, 2
200
Phase 1 Phase 2 10
5
* 0
Noninteraction
Visual Treatment
Interaction
Figure 2. The frequency of aggressive interactions (attacks/h+SE) during phase 1 and 2 of the experiment. Because males had no contact during phase 1 in the noninteraction treatment (star) no interactions could be measured. Attack frequency was highest during phase 2 in the noninteraction treatment. : Phase 1, before the removal of the dividers; : phase 2, when all three individuals were in physical contact.
males occurred frequently (XSD: frequency of attacks: 2.5/h; percentage of time spent in interaction: 6.6 10.8%), but there was no difference in the frequency of aggression between the two treatments (Mann–Whitney U test: U=117, N1 =N2 =15, P=0.87; Fig. 2). In phase 2, when males were in physical contact in all treatments, the frequency of aggressive interactions differed significantly between treatments (Kruskal–Wallis test: H2 =6.842, N=45, P=0.033; Fig. 2). The frequency of aggression was highest in the noninteraction treatment, when males were isolated during phase 1, and lowest in the treatment that allowed physical contact during both phases (Fig. 2). There was a tendency for the frequency of aggression to be higher during phase 1 than phase 2 in the interaction treatment (Wilcoxon signed-ranks test: Z=1.886, N=15, P=0.059; Fig. 2). This high level of aggression could have contributed to the high degree of inconsistency in female preference patterns. No such difference was found for the visual treatment (Wilcoxon signed-ranks test: Z=0.560, N=15, P=0.575; Fig. 2). To establish whether the female mated with the dominant male we compared the frequency of aggressive initiatives made by the two males during phase 1 for the visual and interaction treatments. In the visual treatment the male that eventually mated with the female in phase 2 made the most aggressive initiatives during phase 1 (XSD attacks/h: winner: 1.92.6; loser: 0.10.5; Wilcoxon signed-ranks test: Z=2.384, N=15, P=0.017). This was not the case for the interaction treatment (winner: 1.22.0 attacks/h; loser: 1.84.4 attacks/h; Wilcoxon signed-ranks test: Z=0.213, N=15, P=0.831). During phase 2 of the experiment we measured male dominance as the percentage of time a male spent on his opponent’s side. There were no differences between winners and losers in any of the treatments (XSD percentage of time: noninteraction treatment: winner: 7.69.5%; loser: 5.212.8%; visual treatment: winner:
Courtship bout length (s)
15 Frequency of aggression
428
Phase 1 Phase 2 150
100
50
0
Noninteraction
Visual Treatment
Interaction
Figure 3. The average+SE length of uninterrupted courtship bouts (in seconds). All courtship behaviours have been combined. : Phase 1, before the removal of the dividers; : phase 2, when all three individuals were in physical contact.
4.313.2%; loser: 1.55.2%; interaction treatment: winner: 2.05.0%; loser: 0.31.0%; Wilcoxon signedranks test: noninteraction treatment: Z=1.050, N=15, P=0.294; visual treatment: Z=0.267, N=15, P=0.789; interaction treatment: Z=1.336, N=15, P=0.181). As an additional measure of dominance we used the number of times a male aggressively chased his rival away from his side of the tank. The sample size was too small to be analysed separately for the three treatments, but in the 11 replicates where this occurred it was always the male that eventually mated that chased his opponent (binomial test: P=0.0005).
Courtship When males were able to see each other, but not interact physically, their courtship times (the percentage of time spent courting) were positively correlated (Pearson correlation: r13 =0.69, P=0.004). Indeed, we observed that as one male would start to court, the other male would quickly initiate his courtship displays, although these bouts were soon interrupted by aggressive displays. This observation suggests that courtship competition was taking place. Such a correlation was not observed in the interaction treatment, where males had physical contact with each others (r13 = 0.13, P=0.66) nor in the noninteraction treatment with the opaque divider (r13 = 0.33, P=0.27). In the interaction treatment males often interrupted each other’s courtship through interference, and in the noninteraction treatment, males were probably unaware of each other’s behaviour during the first phase of the experiment. Courtship during the first phase of the experiment differed between different treatments (Kruskall-Wallis test: H2 =7.322, N=45, P=0.03; Fig. 3). Courtship bouts were longest in the noninteraction treatment, when the males were undisturbed, while in the other two treatments the possibility for physical and/or visual interaction made long courtship bouts unlikely (Fig. 3). In the noninteraction treatment the male that courted most
KANGAS & LINDSTRO } M: FEMALE MATE CHOICE IN GOBIES
eventually also mated in most of the cases (XSD proportion of time spent courting: winner: 27.724.6 s; loser: 13.215.8 s; Wilcoxon signed-ranks test: Z=1.989, N=15, P=0.05). We found no differences in the amount of courtship between mated and nonmated males during phase 1 in the visual and interaction treatments. In phase 2 of the experiment the duration of courtship bouts also did not differ between treatments (Kruskal–Wallis test: H2 =1.422, N=45, P=0.49; Fig. 3). DISCUSSION A general concern in studies on mate choice is that interactions between members of the competing sex may prevent the choosing sex from expressing its preferences. Thus dominant individuals may suppress the courtship behaviour of subordinates or may limit the access of the choosing sex to other mates (Bisazza et al. 1989; Rosenqvist 1990; Morris et al. 1992). Consequently studies on female choice have attempted to eliminate the possibility of male–male interactions by various means, for example by using designs that separate males (for reviews see Bateson 1987; Andersson 1994; Houde 1997). The aim of this study was to test to what extent interactions between males may affect a female’s eventual spawning decision. We did this by presenting females with a two-stage choice situation: phase 1, during which the degree of male interaction was controlled and phase 2, during which males were allowed to interact fully. The logic behind this experiment was that if a female made the same choice in phase 1 and 2, then her choice was not additionally affected by the physical interaction between the males during phase 2. We found that when an opaque divider prevented male interactions during phase 1, females showed the most consistent choice behaviour (i.e. they spawned with the same male that they showed a preference for during phase 1; also see Forsgren 1992). That the males in the noninteraction treatment could not interact in phase 1 is supported by the fact that male courtship showed no correlation between the two males and that male–male interactions in phase 2 were more frequent compared with the other treatments. Because no interactions between the males were possible during the first phase, this result either suggests that a female’s decision is not altered to a large extent by dominance relationships between males, or that males somehow prevent females from mating with a particular male. It could also imply that females prefer males that are dominant, but they use traits other than those of male–male interactions to assess this. A consistent outcome in the noninteraction treatment is the least conclusive concerning the role of male–male interactions in female choice as it can be predicted by several different hypotheses. However, in one case where the female switched her ‘preference’, the male that she spawned with clearly hampered the male she had initially preferred (determined during phase 1 as the male on whose side the female spent the most time) by taking over his nest at the moment the female was about to spawn. In the other two treatments males were able to interact to various extents. In these treatments females showed
much more ambiguous choices; about half the females mated with a different male from the one they had initially showed a preference for in phase 1. Our original expectations were that, if male–male interactions are important in female choice, then females in the noninteraction treatment would show the greatest variation in their choice of males between phases 1 and 2 and those in the interaction treatment would show the least variation. However, our results showed just the opposite, and in fact, suggest that the female had a more difficult decision task when she had previously observed some sort of interaction between the males. One reason for this could have been that in treatments where males were allowed to interact the males had to interrupt their courtship bouts more often in order to respond to the rival male. Uninterrupted courtship bouts were longest in phase 1 of the noninteraction treatment where no male–male interactions were possible. For the other treatments, courtship bouts were shorter while aggression was more frequent (Figs 2, 3) suggesting that there might be a trade-off between aggression and courtship. In sand gobies courtship activity seems to be an important trait used by females in making decisions about mating (Forsgren 1997b). Thus, interrupted courtships could have made it more difficult for the females to assess the two males during phase 1, and therefore their choices were less consistent between phases 1 and 2 of the experiment. The addition of male aggression in the visual and interaction treatments also could have contributed to the ambiguous female choice, especially in the interaction treatment, where aggression levels tended to be higher during phase 1. There was also some indication from the visual treatment that males that displayed more aggressive interactions were more successful in spawning. Successful males also chased unsuccessful males more often in phase 2, but this could be a consequence of success in attracting females as well as the reason for mating success. In general, however, we found no strong support for dominant males being more successful. In sticklebacks male–male interactions enhance differences in courtship coloration thus enhancing female choice (Candolin 1999) and females have been shown to choose males that experience the least interference from other males (Sargent 1983). In other cases aggression seems to have a negative impact on female preferences or at least it seems females do not prefer more aggressive or dominant males (Ward & FitzGerald 1987). In the sand goby, females do not prefer dominant males but rather males that provide good-quality parental care in terms of a high egghatching success (Forsgren 1997a). The study by Forsgren (1997a) also shows that dominant males are not necessarily able to prevent a female from mating with a subdominant male, which was also supported by our results for the noninteraction treatment. We conclude from our study that females do not prefer dominant males in a systematic way. A problem with our interaction treatment was that males were free to move around the whole male compartment. Thus a female could, in principle, be staying close to her preferred male, yet we would record her on the side of the other male if the preferred male was also on that
429
430
ANIMAL BEHAVIOUR, 61, 2
side. We do not think that this occurred to a large extent because the males would generally stay in their own half of the tank for most of the experiment time and there were no major differences between the visual and interaction treatments in the proportion of time spent by the female on either side before or after the partitions were removed. Also, if the female had been following one male we would expect a correlation between the time the female spent on a particular side and the time the mated male spent on that side. This was, however, not the case (Spearman rank correlation: rS =0.007, N=15, NS). Therefore, even though we cannot rule out this possibility completely, we do not think that it had a qualitative impact on our results. In conclusion our results seem to suggest that male– male interactions do affect the ability of females to make an unequivocal choice, as evidenced by the disparity of outcomes in the visual and interaction treatments. However, it also suggests that male–male interactions do not alter a female’s spawning decision to a large extent, as shown by the outcomes of the noninteraction treatments, where female behaviour did not change when males were allowed to interact physically in phase 2. Instead male–male interaction in terms of courtship competition and aggressive interactions may make it more difficult for females to make a definitive decision, thus producing the disparity in the outcomes. Nest densities in the field show a great deal of variation. On sand bottoms densities typically range between 0.05 and 3.0 nests/m2 (unpublished data) but can be considerably higher in areas with an accumulation of nest sites. In a natural setting a female may therefore be forced to swim a considerable distance before she encounters a nest and these situations would probably be equivalent to our noninteraction treatment. On the other hand a female will often encounter situations where two or more males would simultaneously try to court her (Forsgren 1997b; personal observation). These situations would most resemble our interaction treatment. Our results suggest that in these cases females may not always be able to manifest their preferences. This could allow males to specialize in different strategies to achieve matings: they could be attractive to females, for example by being able to provide good-quality care or they could specialize in ‘stealing’ mating opportunities from other males. In the sand goby, nest densities can easily be manipulated in the field to create situations in which the potential for male– male interference is comparable to the experiments presented here. Research is currently under way in order to test the findings of this study in the field. Acknowledgments We thank the Tva¨ rminne Zoological Station for offering excellent facilities and a nice surrounding to pursue the work. Colette St Mary and Hannu Pietia¨ inen provided valuable comments on the manuscript. This work has been financed by grants from the Finnish Cultural Foundation and Emil Aaltonen foundation (to N.K) and the Academy of Finland (to K.L.). All experimental
procedures adhered to the Guidelines for Use of Animals in Research and were carried out under Finnish permit 124/97 184. The research methods presented here were subsequently evaluated and approved by the Animal Behavior Society’s Animal Care Committee on 25 July 2000. References Andersson, M. 1994. Sexual Selection. Princeton, New Jersey: Princeton University Press. Bateson, P. 1987. Mate Choice. Cambridge: Cambridge University Press. Bisazza, A., Marconato, A. & Marin, G. 1989. Male competition and female choice in Padogobius martensi (Pisces, Gobiidae). Animal Behaviour, 38, 406–413. Candolin, U. 1999. Male–male competition facilitates female choice in sticklebacks. Proceeding of the Royal Society of London, Series B, 266, 785–789. Darwin, C. 1871. The Descent of Man, and Selection in Relation to Sex. London: Murray. Forsgren, E. 1992. Predation risk affects mate choice in a gobiid fish. American Naturalist, 140, 1041–1049. Forsgren, E. 1997a. Female sand gobies prefer good fathers over dominant males. Proceeding of the Royal Society of London, Series B, 264, 1283–1286. Forsgren, E. 1997b. Mate sampling in a population of sand gobies. Animal Behaviour, 53, 267–276. Forsgren, E., Kvarnemo, C. & Lindstro ¨ m, K. 1996. Mode of sexual selection determined by resource abundance in two sand goby populations. Evolution, 50, 646–654. Hoelzer, G. A. 1990. Male–male competition and female choice in the Cortez damselfish, Stegastes rectifraenum. Animal Behaviour, 40, 339–349. Houde, A. E. 1997. Sex, Color and Mate Choice in Guppies. Princeton, New Jersey: Princeton University Press. Kodric-Brown, A. 1995. Does past reproductive history predict competitive interactions and male mating success in pupfish? Animal Behaviour, 50, 1433–1440. Lindstro ¨ m, K. 1988. Male–male competition for nest sites in the sand goby, Pomatoschistus minutus. Oikos, 53, 67–73. Lindstro ¨ m, K. 1992. The effect of resource holding potential, nest size and information about resource quality on the outcome of intruder–owner conflicts in the sand goby. Behavioral Ecology and Sociobiology, 30, 53–58. Lindstro ¨ m, K. & Lugli, M. 2000. A quantitative analysis of the courtship acoustic behaviour and sound patterning in male sand goby, Pomatoschistus minutus. Environmental Biology of Fishes, 58, 411–424. McLennan, D. A. & McPhail, J. D. 1990. Experimental investigations of the evolutionary significance of sexually dimorphic nuptial coloration in Gasterosteus aculeatus (L.): the relationship between male colour and female behaviour. Canadian Journal of Zoology, 68, 482–492. Morris, M. R., Batra, P. & Ryan, M. J. 1992. Male–male competition and access to females in the swordtail Xiphophorus nigrensis. Copeia, 1992, 980–986. Rosenqvist, G. 1990. Male mate choice and female–female competition for mates in the pipefish Nerophis ophidion. Animal Behaviour, 39, 1110–1115. Sargent, R. C. 1983. Territory quality, male quality, courtship intrusions and female nest-choice in the threespine stickleback, Gasterosteus aculeatus. Animal Behaviour, 30, 164–174. Ward, G. & FitzGerald, G. J. 1987. Male aggression and female mate choice in the threespine stickleback, Gasterosteus aculeatus L. Journal of Fish Biology, 30, 679–690.
Male interactions and female mate choice in the sand goby, Pomatoschistus minutus NUUTTI KANGAS & KAI LINDSTRO }M
Department of Ecology and Systematics, University of Helsinki (Received 16 February 2000; initial acceptance 11 May 2000; final acceptance 24 July 2000; MS. number: A8715)
To examine whether male–male interactions constrain female choice in the sand goby, we compared female preferences expressed under three conditions that varied in the degree of male–male interaction with matings that resulted from free interactions. The experiment consisted of two phases. In the initial phase a transparent divider separated the female from two males. During this stage the males were either isolated by an opaque divider, had visual contact through a transparent divider, or were allowed full interaction by having access to each other’s compartments. The second phase of the experiment was initiated by the removal of all dividers and terminated when the female spawned with one of the males. When there was no initial contact between the males, females spawned with the male they showed a preference for in the first phase of the experiment. In the other two male treatments, females were inconsistent in their spawning choice. Females did not seem to choose dominant males and neither did male aggressive interactions seem to limit a female’s possibilities to spawn with a particular male. The results suggest that male–male interaction may affect the ability of females to evaluate males, in part because of courtship interference between males. Male–male interaction therefore seems to impede female choice and consequently may hinder sexual selection due to mate choice.
aculeatus, from an eastern Canadian population, which show no preference for more aggressive or dominant males (Ward & FitzGerald 1987) and female three-spines from western Canada, whose courtship of ‘more attractive’ males is disrupted by less preferred suitors (McLennan & McPhail 1990). Because of these dynamics, authors have suggested that male–male interactions should be controlled in studies examining female mate choice (e.g. see Bateson 1987; Andersson 1994; Houde 1997). The argument behind this concern is that when males are allowed to interact in a mate choice set-up they may influence each other’s abilities to attract the female. For example, a male may so completely dominate a competitor that it is impossible for the female to assess both males equally. Female choice may therefore be constrained by male–male interactions. This is especially problematic when females prefer male traits that are not associated with increased male dominance (e.g. Hoelzer 1990; Forsgren 1997a). Whether these situations also apply under natural conditions is largely unknown. The importance of male–male interaction in modifying female choice will certainly depend on the spatial distribution of males in the field. In areas where territories are dense one may expect more influence from male–male interactions than in sparse populations.
Sexual selection arises due to an unequal distribution of mating success among individuals of the same sex (Darwin 1871). Among males, mating success may differ because some males are more successful in monopolizing resources needed by females for mating or because some males are more attractive to females as mating partners. These two components of sexual selection are termed intrasexual and intersexual selection, respectively (Andersson 1994). Intrasexual selection is typically prevalent among males, while mate choice (intersexual selection) is usually performed by females. There are two ways in which inter- and intrasexual selection can interact over evolutionary time. In some cases, the two forces will act in the same direction on the same trait, providing mutual reinforcement. For example, females have been shown to prefer dominant males to subordinate ones in pupfish (Kodric-Brown 1995) and at least one population of stickleback (Candolin 1999). In other cases, the two forces might operate on the same trait in opposite directions, thus interfering with, or even cancelling, each other. Examples of this kind of dynamic include female three-spined sticklebacks, Gasterosteus Correspondence: N. Kangas, Department of Ecology and Systematics, Division of Population Biology, P.O. Box 17, 00014 University of Helsinki, Helsinki, Finland (email: [email protected]). 0003–3472/01/020425+06 $35.00/0
2001 The Association for the Study of Animal Behaviour
425
2001 The Association for the Study of Animal Behaviour
426
ANIMAL BEHAVIOUR, 61, 2
Few studies have systematically examined the influence of male–male interactions on female mate choice. The purpose of this study was to examine whether male interactions affect female choice in the sand goby, a small marine fish. In this species males construct nests by excavating the sand underneath rocks or mussel shells. They then attempt to attract females to the nest by courting them. Once a female has been attracted into the nest the male emits courtship sounds (Lindstro ¨ m & Lugli 2000). If a female is interested in a particular male she attaches her eggs to the ceiling of the nest. Nest density varies extensively both between and within populations (Forsgren et al. 1996) and consequently the opportunity for male interference during courtship and female choice can be expected to vary. In our study population nest availability is on average low and there is intense male–male competition for available nest sites (Lindstro ¨ m 1988, 1992). In this study we varied the level of male–male interaction, then compared the result of female choice trials in those different situations. METHODS The experiments were done during May–June 1996 at the Tva¨ rminne Zoological Station, situated on the shoreline of the northern Baltic Sea, the south coast of Finland (60N, 23E). Fish for the experiments were caught from a nearby breeding site using a beach seine. The sexes were kept in separate large storage tanks (150-litre) with a constant through-flow of sea water. During this time the fish were fed ad libitum live Neomysis integer shrimp and frozen Chironomidae larvae. Experimental aquaria measured 704040 cm. Each experimental tank was provided with a 4-cm layer of fine sand on the bottom of the tanks and two artificial nest sites (ceramic tiles measuring 7.57.5 cm) placed at opposite ends of the tank. The tank was divided lengthwise using a transparent Plexiglas divider so that the nests were in the section representing two-thirds of the bottom surface (Fig. 1). Due to the set-up no through-flow of water was provided during the experiment but the water was exchanged between replicates. We simultaneously added two males to the nest section and allowed them 24 h to build nests. If one or none of the males had finished his nest by this time we replaced both males. To minimize any asymmetries in dominance relations between the two males, we selected males of matched body size. After the nest-building phase we added a female to the smaller of the two lengthwise sections (see Fig. 1). Depending on the treatment, males were freely interacting, separated by a transparent barrier, or separated by an opaque barrier (see below). The female had complete visual access to both males because she could freely move over the total length of the tank and the males were free to observe and visually interact with the female. No individual was used in more than one replicate. The fish were returned to their site of origin after the experiment. We conducted experiments in two phases (phases 1 and 2).
B
A
Figure 1. Tank design. The experimental tanks were initially divided into three sections for all treatments except the interaction treatment. The female was placed in the long section and she had visual contact with the males through the transparent divider A. The two males were placed in the two rear compartments separated by the divider B. Divider B was either an opaque or transparent Plexiglas wall except during the interaction treatment, in which divider B was not present. In the second phase of the experiment all dividers were removed and all three fish had free access to all parts of the tank. The grey squares represent the ceramic tiles used as nest sites.
Phase 1 In phase 1 a transparent divider separated the female from the two males, which were either freely interacting, or separated by a transparent or opaque divider (see below). During this phase, which lasted 4 h, the female and males were observed for a total of 50 min. The observation time consisted of five 10-min sessions: immediately, 1, 2, 3 and 4 h after the introduction of the female (all observations were done by N.K.). We recorded the time the female spent on each male’s side. The dividing line was the centre of the tank. The male on whose side the female spent more time was considered to be the preferred male. In addition we recorded the time both males spent courting the female and interacting with each other. Courtship in the sand goby involves fin displays, jerky swimming towards and around the female, lateral display and finally lead swims towards the nest (also see Forsgren 1997b). Male–male interactions in the sand goby are usually aggressive, however, in the present study, we observed no mortality or injuries associated with these interactions. We recorded fin displays, chases, darts against the opponent, lateral displays and physical fighting. Behaviours presented as percentages were obtained by dividing the total time in certain behaviour by the total observation time. Frequencies were obtained by dividing the number of times a behaviour was performed with the observation time. The mean length of courtship bouts was obtained as the total time spent courting divided by the number of times the male courted.
Phase 2 In phase 2 we removed the lengthwise divider separating the males and the female and any other dividers between the males (see below), at which stage all three individuals had physical contact with each other. We observed the behaviour of the female and the two males
KANGAS & LINDSTRO } M: FEMALE MATE CHOICE IN GOBIES
for 20 min immediately after the removal of the dividers and then 10 min each full hour for the first 4 h or until the female had spawned, whichever came first. The experiment was ended when the female spawned.
Phase 1 treatments The two males in phase 1 were presented to the female in three different ways: (1) freely interacting (male–male interaction treatment), (2) separated by a transparent barrier (male–male visual treatment), or (3) separated by an opaque barrier (male–male noninteraction treatment).
Interaction treatment. In this treatment there was no divider between the two males so nothing prevented them from interfering with each other’s courtship or other activities. It should thus be as difficult for the female to independently assess the males in phase 1 as in phase 2. If male interference modifies female choice, then we would not expect the choice performed in phase 1 to differ from the eventual spawning decision made in phase 2.
(1) If males can physically suppress each other’s courtship performance, then the spawning decision in phase 2 should differ most from the choice in phase 1 for the noninteraction and the visual interaction treatments. This is because males in these treatments could not physically interact with each other during phase 1 and therefore female choice would not be affected until phase 2. This also requires that females show consistent preferences in phase 1 and 2 in the interaction treatment. (2) If the outcome differs only for the noninteraction treatment, where no visual or physical interaction was possible between males, then courtship competition is at least equally important in interfering with female choice as is direct male–male interference. This is because in both the interaction and visual treatment, female choice in phase 1 could be affected by male–male interactions and her choice should not be independent of these interactions. RESULTS
Female Preference Visual treatment. The two males were separated by a transparent divider. They therefore had visual contact with each other and could engage in courtship competition (i.e. a situation where the two males’ courtship depended upon one another), as well as threat displays, but one male could not physically dominate or interfere with the other one. A visual interaction between the males may lead to an escalation of courtship. Such courtship competition may cause males to invest more in courtship than they would in the absence of competition, which may interfere with a female’s ability to independently assess male quality. If this is the case then females should show consistent choice in this treatment because courtship competition could be present both before and after the removal of the transparent divider.
Noninteraction treatment. An opaque Plexiglas divider separated the two males, so each male could only see the female, while the female was free to observe each male. In this treatment there was no interaction between the two males and the female should have been able to independently assess them. If the female were to change her choice in the second phase from her initial choice, this would constitute strong evidence that male–male interactions indeed affect a female’s ability to independently express choice. If females show consistent choice in this treatment, this may suggest that female choice is unlikely to be affected by male–male interactions and that females are able to independently assess male quality even in the presence of male–male interactions. It may, however, also be an indication that females use the same traits to assess male quality that make males successful in male–male interactions. A consistent outcome in this treatment would therefore be less conclusive as it would be expected under several different scenarios, depending on the importance and type of male–male interference during courtship. Therefore, to distinguish between these various outcomes, we evaluated the outcomes from the different treatments based on the following predictions.
In 64% of the trials females chose the same male in phase 1 as in phase 2 of the experiment. There were, however, differences between treatments (chi-square test: 22 =5.88, P=0.053). In the noninteraction treatment, where males lacked visual and physical contact, females were most consistent in their choice. In 13 of the 15 replicates, the female spawned with the male she spent the most time with in phase 1 (binomial test: P=0.004). When the males had visual contact with each other, the female showed consistent choice in nine of 15 trials (binomial test: P=0.3). Similarly, in the treatment with physical contact between the two males, the female mated with the same male she spent the most time with in seven out of the 15 replicates (binomial test: P=0.5). Thus, females did not change their initial choice when an opaque divider separated the males and male–male interactions were prevented. In the other treatments, however, females showed extensive inconsistency in their eventual spawning decision. It is possible that females in the noninteraction treatment became fixated on their preferred males during phase 1, and therefore showed a more consistent choice in phase 2 than females in the other treatments. If this were the case, one might predict that females in the noninteraction treatment would also spend more time with their preferred male than females in the other treatments. There was, however, no difference between treatments in the proportion of time that females spent with their preferred males (XSD percentage of time on winner’s side: noninteraction treatment: 74.933.2, N=15; visual treatment: 61.448.0, N=15; interaction treatment: 58.046.0, N=15; Kruskal–Wallis test: H2 =1.773, P=0.412), suggesting that females in the noninteraction treatment were no more fixated on their preferred males than females in other treatments.
Male–Male Interaction In the two treatments in which males could interact (i.e. visual and interaction), interactions between the
427
ANIMAL BEHAVIOUR, 61, 2
200
Phase 1 Phase 2 10
5
* 0
Noninteraction
Visual Treatment
Interaction
Figure 2. The frequency of aggressive interactions (attacks/h+SE) during phase 1 and 2 of the experiment. Because males had no contact during phase 1 in the noninteraction treatment (star) no interactions could be measured. Attack frequency was highest during phase 2 in the noninteraction treatment. : Phase 1, before the removal of the dividers; : phase 2, when all three individuals were in physical contact.
males occurred frequently (XSD: frequency of attacks: 2.5/h; percentage of time spent in interaction: 6.6 10.8%), but there was no difference in the frequency of aggression between the two treatments (Mann–Whitney U test: U=117, N1 =N2 =15, P=0.87; Fig. 2). In phase 2, when males were in physical contact in all treatments, the frequency of aggressive interactions differed significantly between treatments (Kruskal–Wallis test: H2 =6.842, N=45, P=0.033; Fig. 2). The frequency of aggression was highest in the noninteraction treatment, when males were isolated during phase 1, and lowest in the treatment that allowed physical contact during both phases (Fig. 2). There was a tendency for the frequency of aggression to be higher during phase 1 than phase 2 in the interaction treatment (Wilcoxon signed-ranks test: Z=1.886, N=15, P=0.059; Fig. 2). This high level of aggression could have contributed to the high degree of inconsistency in female preference patterns. No such difference was found for the visual treatment (Wilcoxon signed-ranks test: Z=0.560, N=15, P=0.575; Fig. 2). To establish whether the female mated with the dominant male we compared the frequency of aggressive initiatives made by the two males during phase 1 for the visual and interaction treatments. In the visual treatment the male that eventually mated with the female in phase 2 made the most aggressive initiatives during phase 1 (XSD attacks/h: winner: 1.92.6; loser: 0.10.5; Wilcoxon signed-ranks test: Z=2.384, N=15, P=0.017). This was not the case for the interaction treatment (winner: 1.22.0 attacks/h; loser: 1.84.4 attacks/h; Wilcoxon signed-ranks test: Z=0.213, N=15, P=0.831). During phase 2 of the experiment we measured male dominance as the percentage of time a male spent on his opponent’s side. There were no differences between winners and losers in any of the treatments (XSD percentage of time: noninteraction treatment: winner: 7.69.5%; loser: 5.212.8%; visual treatment: winner:
Courtship bout length (s)
15 Frequency of aggression
428
Phase 1 Phase 2 150
100
50
0
Noninteraction
Visual Treatment
Interaction
Figure 3. The average+SE length of uninterrupted courtship bouts (in seconds). All courtship behaviours have been combined. : Phase 1, before the removal of the dividers; : phase 2, when all three individuals were in physical contact.
4.313.2%; loser: 1.55.2%; interaction treatment: winner: 2.05.0%; loser: 0.31.0%; Wilcoxon signedranks test: noninteraction treatment: Z=1.050, N=15, P=0.294; visual treatment: Z=0.267, N=15, P=0.789; interaction treatment: Z=1.336, N=15, P=0.181). As an additional measure of dominance we used the number of times a male aggressively chased his rival away from his side of the tank. The sample size was too small to be analysed separately for the three treatments, but in the 11 replicates where this occurred it was always the male that eventually mated that chased his opponent (binomial test: P=0.0005).
Courtship When males were able to see each other, but not interact physically, their courtship times (the percentage of time spent courting) were positively correlated (Pearson correlation: r13 =0.69, P=0.004). Indeed, we observed that as one male would start to court, the other male would quickly initiate his courtship displays, although these bouts were soon interrupted by aggressive displays. This observation suggests that courtship competition was taking place. Such a correlation was not observed in the interaction treatment, where males had physical contact with each others (r13 = 0.13, P=0.66) nor in the noninteraction treatment with the opaque divider (r13 = 0.33, P=0.27). In the interaction treatment males often interrupted each other’s courtship through interference, and in the noninteraction treatment, males were probably unaware of each other’s behaviour during the first phase of the experiment. Courtship during the first phase of the experiment differed between different treatments (Kruskall-Wallis test: H2 =7.322, N=45, P=0.03; Fig. 3). Courtship bouts were longest in the noninteraction treatment, when the males were undisturbed, while in the other two treatments the possibility for physical and/or visual interaction made long courtship bouts unlikely (Fig. 3). In the noninteraction treatment the male that courted most
KANGAS & LINDSTRO } M: FEMALE MATE CHOICE IN GOBIES
eventually also mated in most of the cases (XSD proportion of time spent courting: winner: 27.724.6 s; loser: 13.215.8 s; Wilcoxon signed-ranks test: Z=1.989, N=15, P=0.05). We found no differences in the amount of courtship between mated and nonmated males during phase 1 in the visual and interaction treatments. In phase 2 of the experiment the duration of courtship bouts also did not differ between treatments (Kruskal–Wallis test: H2 =1.422, N=45, P=0.49; Fig. 3). DISCUSSION A general concern in studies on mate choice is that interactions between members of the competing sex may prevent the choosing sex from expressing its preferences. Thus dominant individuals may suppress the courtship behaviour of subordinates or may limit the access of the choosing sex to other mates (Bisazza et al. 1989; Rosenqvist 1990; Morris et al. 1992). Consequently studies on female choice have attempted to eliminate the possibility of male–male interactions by various means, for example by using designs that separate males (for reviews see Bateson 1987; Andersson 1994; Houde 1997). The aim of this study was to test to what extent interactions between males may affect a female’s eventual spawning decision. We did this by presenting females with a two-stage choice situation: phase 1, during which the degree of male interaction was controlled and phase 2, during which males were allowed to interact fully. The logic behind this experiment was that if a female made the same choice in phase 1 and 2, then her choice was not additionally affected by the physical interaction between the males during phase 2. We found that when an opaque divider prevented male interactions during phase 1, females showed the most consistent choice behaviour (i.e. they spawned with the same male that they showed a preference for during phase 1; also see Forsgren 1992). That the males in the noninteraction treatment could not interact in phase 1 is supported by the fact that male courtship showed no correlation between the two males and that male–male interactions in phase 2 were more frequent compared with the other treatments. Because no interactions between the males were possible during the first phase, this result either suggests that a female’s decision is not altered to a large extent by dominance relationships between males, or that males somehow prevent females from mating with a particular male. It could also imply that females prefer males that are dominant, but they use traits other than those of male–male interactions to assess this. A consistent outcome in the noninteraction treatment is the least conclusive concerning the role of male–male interactions in female choice as it can be predicted by several different hypotheses. However, in one case where the female switched her ‘preference’, the male that she spawned with clearly hampered the male she had initially preferred (determined during phase 1 as the male on whose side the female spent the most time) by taking over his nest at the moment the female was about to spawn. In the other two treatments males were able to interact to various extents. In these treatments females showed
much more ambiguous choices; about half the females mated with a different male from the one they had initially showed a preference for in phase 1. Our original expectations were that, if male–male interactions are important in female choice, then females in the noninteraction treatment would show the greatest variation in their choice of males between phases 1 and 2 and those in the interaction treatment would show the least variation. However, our results showed just the opposite, and in fact, suggest that the female had a more difficult decision task when she had previously observed some sort of interaction between the males. One reason for this could have been that in treatments where males were allowed to interact the males had to interrupt their courtship bouts more often in order to respond to the rival male. Uninterrupted courtship bouts were longest in phase 1 of the noninteraction treatment where no male–male interactions were possible. For the other treatments, courtship bouts were shorter while aggression was more frequent (Figs 2, 3) suggesting that there might be a trade-off between aggression and courtship. In sand gobies courtship activity seems to be an important trait used by females in making decisions about mating (Forsgren 1997b). Thus, interrupted courtships could have made it more difficult for the females to assess the two males during phase 1, and therefore their choices were less consistent between phases 1 and 2 of the experiment. The addition of male aggression in the visual and interaction treatments also could have contributed to the ambiguous female choice, especially in the interaction treatment, where aggression levels tended to be higher during phase 1. There was also some indication from the visual treatment that males that displayed more aggressive interactions were more successful in spawning. Successful males also chased unsuccessful males more often in phase 2, but this could be a consequence of success in attracting females as well as the reason for mating success. In general, however, we found no strong support for dominant males being more successful. In sticklebacks male–male interactions enhance differences in courtship coloration thus enhancing female choice (Candolin 1999) and females have been shown to choose males that experience the least interference from other males (Sargent 1983). In other cases aggression seems to have a negative impact on female preferences or at least it seems females do not prefer more aggressive or dominant males (Ward & FitzGerald 1987). In the sand goby, females do not prefer dominant males but rather males that provide good-quality parental care in terms of a high egghatching success (Forsgren 1997a). The study by Forsgren (1997a) also shows that dominant males are not necessarily able to prevent a female from mating with a subdominant male, which was also supported by our results for the noninteraction treatment. We conclude from our study that females do not prefer dominant males in a systematic way. A problem with our interaction treatment was that males were free to move around the whole male compartment. Thus a female could, in principle, be staying close to her preferred male, yet we would record her on the side of the other male if the preferred male was also on that
429
430
ANIMAL BEHAVIOUR, 61, 2
side. We do not think that this occurred to a large extent because the males would generally stay in their own half of the tank for most of the experiment time and there were no major differences between the visual and interaction treatments in the proportion of time spent by the female on either side before or after the partitions were removed. Also, if the female had been following one male we would expect a correlation between the time the female spent on a particular side and the time the mated male spent on that side. This was, however, not the case (Spearman rank correlation: rS =0.007, N=15, NS). Therefore, even though we cannot rule out this possibility completely, we do not think that it had a qualitative impact on our results. In conclusion our results seem to suggest that male– male interactions do affect the ability of females to make an unequivocal choice, as evidenced by the disparity of outcomes in the visual and interaction treatments. However, it also suggests that male–male interactions do not alter a female’s spawning decision to a large extent, as shown by the outcomes of the noninteraction treatments, where female behaviour did not change when males were allowed to interact physically in phase 2. Instead male–male interaction in terms of courtship competition and aggressive interactions may make it more difficult for females to make a definitive decision, thus producing the disparity in the outcomes. Nest densities in the field show a great deal of variation. On sand bottoms densities typically range between 0.05 and 3.0 nests/m2 (unpublished data) but can be considerably higher in areas with an accumulation of nest sites. In a natural setting a female may therefore be forced to swim a considerable distance before she encounters a nest and these situations would probably be equivalent to our noninteraction treatment. On the other hand a female will often encounter situations where two or more males would simultaneously try to court her (Forsgren 1997b; personal observation). These situations would most resemble our interaction treatment. Our results suggest that in these cases females may not always be able to manifest their preferences. This could allow males to specialize in different strategies to achieve matings: they could be attractive to females, for example by being able to provide good-quality care or they could specialize in ‘stealing’ mating opportunities from other males. In the sand goby, nest densities can easily be manipulated in the field to create situations in which the potential for male– male interference is comparable to the experiments presented here. Research is currently under way in order to test the findings of this study in the field. Acknowledgments We thank the Tva¨ rminne Zoological Station for offering excellent facilities and a nice surrounding to pursue the work. Colette St Mary and Hannu Pietia¨ inen provided valuable comments on the manuscript. This work has been financed by grants from the Finnish Cultural Foundation and Emil Aaltonen foundation (to N.K) and the Academy of Finland (to K.L.). All experimental
procedures adhered to the Guidelines for Use of Animals in Research and were carried out under Finnish permit 124/97 184. The research methods presented here were subsequently evaluated and approved by the Animal Behavior Society’s Animal Care Committee on 25 July 2000. References Andersson, M. 1994. Sexual Selection. Princeton, New Jersey: Princeton University Press. Bateson, P. 1987. Mate Choice. Cambridge: Cambridge University Press. Bisazza, A., Marconato, A. & Marin, G. 1989. Male competition and female choice in Padogobius martensi (Pisces, Gobiidae). Animal Behaviour, 38, 406–413. Candolin, U. 1999. Male–male competition facilitates female choice in sticklebacks. Proceeding of the Royal Society of London, Series B, 266, 785–789. Darwin, C. 1871. The Descent of Man, and Selection in Relation to Sex. London: Murray. Forsgren, E. 1992. Predation risk affects mate choice in a gobiid fish. American Naturalist, 140, 1041–1049. Forsgren, E. 1997a. Female sand gobies prefer good fathers over dominant males. Proceeding of the Royal Society of London, Series B, 264, 1283–1286. Forsgren, E. 1997b. Mate sampling in a population of sand gobies. Animal Behaviour, 53, 267–276. Forsgren, E., Kvarnemo, C. & Lindstro ¨ m, K. 1996. Mode of sexual selection determined by resource abundance in two sand goby populations. Evolution, 50, 646–654. Hoelzer, G. A. 1990. Male–male competition and female choice in the Cortez damselfish, Stegastes rectifraenum. Animal Behaviour, 40, 339–349. Houde, A. E. 1997. Sex, Color and Mate Choice in Guppies. Princeton, New Jersey: Princeton University Press. Kodric-Brown, A. 1995. Does past reproductive history predict competitive interactions and male mating success in pupfish? Animal Behaviour, 50, 1433–1440. Lindstro ¨ m, K. 1988. Male–male competition for nest sites in the sand goby, Pomatoschistus minutus. Oikos, 53, 67–73. Lindstro ¨ m, K. 1992. The effect of resource holding potential, nest size and information about resource quality on the outcome of intruder–owner conflicts in the sand goby. Behavioral Ecology and Sociobiology, 30, 53–58. Lindstro ¨ m, K. & Lugli, M. 2000. A quantitative analysis of the courtship acoustic behaviour and sound patterning in male sand goby, Pomatoschistus minutus. Environmental Biology of Fishes, 58, 411–424. McLennan, D. A. & McPhail, J. D. 1990. Experimental investigations of the evolutionary significance of sexually dimorphic nuptial coloration in Gasterosteus aculeatus (L.): the relationship between male colour and female behaviour. Canadian Journal of Zoology, 68, 482–492. Morris, M. R., Batra, P. & Ryan, M. J. 1992. Male–male competition and access to females in the swordtail Xiphophorus nigrensis. Copeia, 1992, 980–986. Rosenqvist, G. 1990. Male mate choice and female–female competition for mates in the pipefish Nerophis ophidion. Animal Behaviour, 39, 1110–1115. Sargent, R. C. 1983. Territory quality, male quality, courtship intrusions and female nest-choice in the threespine stickleback, Gasterosteus aculeatus. Animal Behaviour, 30, 164–174. Ward, G. & FitzGerald, G. J. 1987. Male aggression and female mate choice in the threespine stickleback, Gasterosteus aculeatus L. Journal of Fish Biology, 30, 679–690.