Behavioral Ecology and Sociobiology

, Volume 56, Issue 1, pp 59–64

Female courtship in the Banggai cardinalfish: honest signals of egg maturity and reproductive output?


    • Department of Animal Ecology, Evolutionary Biology CentreUppsala University
Original Article

DOI: 10.1007/s00265-003-0754-5

Cite this article as:
Kolm, N. Behav Ecol Sociobiol (2004) 56: 59. doi:10.1007/s00265-003-0754-5


Despite the vast literature on male courtship behaviour, little is known about the function and information content of female courtship behaviour. Female courtship behaviour may be important in many species, particularly where both sexes invest heavily in the offspring, and if such behaviours contain honest information regarding a female’s potential reproductive investment, they may be particularly important in male mate choice. Using observations of two female courtship behaviours (the “rush” and the “twitch”) from experimental pairings in the Banggai cardinalfish (Pterapogon kauderni), I addressed the question of whether these courtship behaviours contained information on female reproductive output (clutch weight) and egg maturity (proximity to spawning), traits commonly associated with male mate choice. I especially focused on the importance of these courtship behaviours in relation to other female characters, such as size and condition, using multiple regression. I found that one of these behaviours, the rush, was strongly associated with fecundity, whereas size, condition and the twitch were not. Further, I found that the “twitch” behaviour was associated with how close to actual spawning a female was. The results suggest that female courtship behaviour may convey highly important information in a mate choice context. I discuss the adaptive value of honest information in female courtship behaviour in light of these results.


Female courtship displayHonest informationDifferential allocationMale mate choice


Male courtship behaviour is a common trait in all animal taxa and may convey important information about a male’s value as a partner in terms of, for instance, parental abilities or genetic quality (e.g. Bradbury and Vehrencamp 1998; Andersson 1994). The significance of female courtship, however, is a rather neglected area in behavioural ecology. As females always carry a substantial reproductive burden in the production of eggs (Clutton-Brock 1991), we expect, in most animal species, that courtship behaviour is limited to males. This is the case because this reproductive burden most often translates into females being the more limited sex with regards to sexual selection of courtship displays (e.g. Clutton-Brock 1991; Andersson 1994; Kvarnemo and Ahnesjö 1996). However, in species where both sexes invest heavily in the offspring, selection may also favour choosy males (Crowley et al. 1991; Johnstone et al. 1996). This may result in mutual mate choice or sex role reversal, and hence the use of female courtship behaviour to attract potential partners (e.g. Andersson 1994; Kolm 2002; Berglund and Rosenqvist 2003). Several female phenotypic traits, for instance female size and female condition, are positively correlated to reproductive output (i.e. fecundity and egg size) in virtually all taxa (e.g. Mosseau and Fox 1998). Such traits are therefore commonly preferred by choosy males in species with mutual mate choice or sex role reversal (Côté and Hunte 1989; Jones and Hunter 1993; Andersson 1994; Kraak and Bakker 1998; Sandvik-Widemo 2003). There are, however, very few empirical studies that have investigated the importance of female courtship behaviours, particularly for male mate choice.

Of special interest is the information content of such female courtship behaviour. Honest signalling of potential reproductive output could be a way for females to display their reproductive value to potential mates, especially if there is variation in reproductive output between females of the same size classes (e.g. Kolm 2001). This is expected when there is a particularly steep trade-off between current and future reproductive events. One such example is species where females exhibit differential allocation (Burley 1986). Differential allocation, which occurs when females increase the reproductive investment when paired to attractive mates, may be a way for females to obtain and/or retain an attractive mate (Burley 1986). As mentioned earlier, few studies have demonstrated the use of female courtship behaviour to attract a mate, and no study has, to my knowledge, demonstrated any plastic female character correlated to the female reproductive investment in species exhibiting differential investment. Before such a correlation is demonstrated, we are unable to conclude whether differential investment may be a way for females to gain access to mates of greater attractiveness, and hence increase their reproductive success (Burley 1986, 1988). The demonstration of such plasticity in a courtship behaviour correlated to a matching plasticity in female reproductive investment, however, could indicate a much greater influence of females on sexual selection than previously thought. Furthermore, information on egg maturity (i.e. how close to spawning a female is) can be highly important for choosy males, and may be transferred via plastic female traits such as courtship behaviour (Sargent et al. 1998).

The Banggai cardinalfish (Pterapogon kauderni) is a species exhibiting mutual mate choice, where females produce a small clutch (maximum 90) of large eggs (2–3 mm in diameter), which the male mouthbroods for 30 days without feeding. Pairs are formed up to 2 weeks prior to spawning and the female courts the male prior to, and after, pair formation until spawning (personal observation). The female courtship mainly consists of two distinct behaviours: the “twitch” and the “rush”. The twitch consists of the female folding her pelvic, anal and dorsal fins together while twitching her body close to the male. Females perform the rush by folding the pelvic, dorsal and anal fins and quickly swimming past the male for a distance of 10–40 cm (Kolm 2001). Total clutch weight is likely to be the most important trait for males to maximise their fitness in this species as it reflects both egg size and fecundity (Kolm 2001, 2002). Furthermore, although both fecundity and egg size are positively correlated to female size (Kolm 2002) in this species, females perform the rush behaviour more intensively towards larger males, for which they also produce heavier clutches consisting of larger eggs (Kolm 2001). In this species, it is therefore highly plausible that female courtship behaviour may convey information that is important for choosy males.

To investigate this, I have analysed behavioural data from two experiments. I used multiple regression analysis to compare the information content of female courtship behaviour to size and condition in relation to reproductive output. This allowed me to first assess which of these traits contain most information about the reproductive output (i.e. clutch weight). Second, I determined whether female courtship behaviour may be a signal of egg maturity, another important cue for potential mates. Finally, to determine whether or not the intensity of female behaviour might be related to male mate choice, I asked whether the intensity of the female behaviours was correlated with spawning latency, a trait apparently indicative of male mate choice in the Banggai cardinalfish (Kolm 2002).


Stocking and experimental system

All fish were wild caught and obtained from a fish dealer, and held in eight 400-l aquaria prior to experiments. The aquaria were connected in parallel, with water being filtered through a central filtering unit consisting of a 400-l plastic tank, a mechanic filter, a protein skimmer and a 100-l biofilter. A pump provided circulation through the aquaria. Each individual aquarium was supplied with 4 separate in- and outlets, forming in total 32 separate 100-l compartments (divided by opaque PVC sheets). Temperature was held constant at 27°C and salinity ranged between 32 and 34‰. The photoperiod was 10 h light 14 h dark. Animals were fed frozen brine shrimp and mysids ad libitum once per day.

Behavioural data

This study includes behavioural data from two separate experiments conducted in 2000 and 2001. Data originate from experiments where sexually mature females of different sizes were paired to sexually mature males of different sizes in a manner ensuring that male size and female size were not correlated. The female courtship behaviours in this species are the “twitch”, and the “rush” (Kolm 2001). Females perform both these behaviours before, as well as after, pair formation in aquaria and in the field, until spawning (personal observations). As all fish were individually isolated for 3 weeks prior to pairings, the observations during experimental pairings will thus include both the phase prior to pairing and after actual pairing. From the day of experimental pairing, twitch and rush behaviours were counted for a period of 5–10 min daily between 1300 hours and 1700 hours (the peak in courtship and spawning behaviour is usually in the afternoon in the Banggai cardinalfish) until spawning. As data on egg weights were not always collected where observations of courtship behaviour were collected (note that both rush and twitch behaviours have been displayed by females in all experimental pairings to date), the sample sizes differ between analyses including egg investment and female courtship behaviour, and correlations between spawning latency and female courtship behaviour. For the analysis on female traits and clutch weight, I only included replicates where all these parameters were available. This gave a total sample size of N=19 for this analysis. For the analysis on courtship rates in relation to spawning latency, however, I had access to more replicates with data on display rates (N=29), and hence the sample size differs between the analyses. For the analyses on the relation between female traits and clutch weight, I only included behavioural data from the 13 days prior to spawning. By applying this procedure, I avoided problems with different pairs having been observed for potentially different phases in their courtship displays. As pairs normally are formed not longer than 14 days, and usually shorter, prior to spawning (personal observations), this method still allowed me to include most, if not all, of the courtship-display information that would occur in naturally formed pairs both prior to, and after, pair formation. To standardise each observation, the ratio of number of behaviours per minute of observation was used after verifying that there was no difference in ratios between 5- and 10-min observations (t-test: t27=1.5, P=0.13). Details concerning the methods of such pairing experiments have been described elsewhere (Kolm 2001, 2002). The number of days from pair formation until spawning (spawning latency) was also recorded for each pair. Directly after spawning, the egg clutch was removed from the mouth of the male using a pair of forceps. The wet weight of the entire egg clutch was measured, as described by Kolm (2001), and used as the measure of female investment. Female wet weight was also measured and female condition was quantified as the residuals from the regression of female weight on size.

Statistical analysis

Analyses on display information were performed using parametric procedures. When required, the data were log-transformed to achieve normality. In the analysis of courtship behaviour as indicators of egg maturity, the number of observations differed between pairs due to different spawning latencies between pairs. To control for this, and to avoid pseudoreplication which would result from a simple regression analysis including all pairs for all days, the average of all pairs’ correlation coefficients from the correlation between day prior to spawning and number of behaviours was calculated after log-transformation. The average correlation coefficient was then tested against the student’s t-distribution with H0: r=0, df=n−1 (N=29). All statistical analyses were carried out using the STATISTICA (StatSoft 2001) statistical package.


Females displayed the twitch more often than the rush (twitch/minute of observation time (mean±SD)=1.41±0.15; rush/minute of observation time: 0.24±0.66; dependent t-test on log-transformed data: t18=11.4, P<0.0001, Fig. 1a). Further, the twitch behaviour was performed earlier after pair formation as compared to the rush (days until first display after pairing (mean±S. D.): twitch 2.9±1.4; rush 4.7±3.5; dependent t-test on log-transformed data: t18=2.7, P=0.02, Fig. 1b). The display rates of the two behaviours were not significantly correlated to each other (Pearson correlation: r19=0.23, P=0.34). The spawning latency of formed pairs was (mean±SD) 16.6±8.4 days.
Fig. 1

a Difference in display intensity between the twitch (mean±SD; unfilled bar) courtship behaviour and the rush (mean±SD; filled bar). b Difference in time until first display after pairing between the twitch (mean±SD; unfilled bar) and the rush (mean±SD; filled bar) (*P<0.05, ***P<0.001). See text for details

The rush intensity held the strongest information content regarding clutch weight in the multiple regression (Table 1). Female size, female condition and the twitch behaviour did not explain the variation in clutch weight (Table 1). There was also a negative association between the total intensity of rushing and spawning latency (Pearson correlation: r29=−0.43, P<0.02). Together, these results suggest that the rush behaviour is most important for a male to predict female reproductive investment in comparison to the female traits size, condition and twitch behaviour.
Table 1

Results from multiple regression with clutch weight as the dependent variable. Included as independent variables are female size, female condition, rush behaviour and twitch behaviour. All independent variables were log-transformed prior to analysis. β is the standardised regression coefficient (multiple r2=0.49, F(4, 14)=3.3, P=0.04)

Independent variables





Female size





Female condition















The number of twitches performed on a given day was strongly associated with how close to spawning a female was (Fig. 2a). The number of twitches increased closer to spawning and showed an exponential relationship with day prior to spawning [average r for all correlations between number of twitches (log-transformed) and day prior to spawning±SD=−0.61±0.24, t28=13.4, P<0.0001, Fig. 2a]. Also, the number of rush behaviours increased closer to the day of spawning, but linearly so (average r for all correlations between number of rushes and day prior to spawning±SD=−0.31±0.24, t28=6.8, P<0.001, Fig. 2b). The slopes of the curves differed significantly (dependent t-test: t28=4.6, P<0.0001), indicating that the twitch is a stronger predictor of how close to spawning a female is as compared to the rush.
Fig. 2

a Correlation between twitch intensity (mean±SE for each day) and proximity to actual spawning. b Correlation between rush intensity (mean±SE for each day) and proximity to actual spawning. The slope of the curves differed significantly (dependent t-test: t28=4.6, P<0.0001) indicating that the twitch is a stronger predictor of how close to spawning a female is as compared to the rush. See text for details


This study suggests that female courtship behaviour in the Banggai cardinalfish contains information that may be highly important for choosy males. The twitch behaviour was displayed earlier and more commonly than the rush. Out of the four measured female traits (size, condition, rush and twitch), only rush was significantly and positively associated with clutch weight in the multiple regression (Table 1). Further, spawning latency was negatively related to rush behaviour, suggesting that males may spawn quicker with females that display intensively. Both rush and twitch behaviours increased closer to spawning but twitch was a stronger predictor of how close a female was to spawning (Fig. 2).

Interestingly, the female rush behaviour was associated more strongly with clutch weight as compared to female size and condition. Many examples exist where larger females and/or females in better condition lay larger clutches and/or larger eggs (e.g. Shuster 1981; Wiewandt 1982; Berglund et al. 1986; Bryant 1988; Howard 1988; Honek 1993; Mosseau and Fox 1998). Although these traits are also important in the Banggai cardinalfish, where larger females lay larger clutches and larger eggs (Kolm 2002), there is thus ample variation in reproductive output within females of similar size and condition. Hence, the rush behaviour may convey more important information to potential mates than a female’s size and condition.

Further, the link between rush intensity and clutch weight suggests that female courtship behaviour contains information about a female’s potential reproductive effort in a differential allocation context. Previous experiments have shown that female Banggai cardinalfish display the rush behaviour more often towards larger than towards smaller males, and that females also invest more in reproduction by producing larger eggs (resulting in heavier egg clutches) for larger than for smaller males (Kolm 2001). Here I show that the rush behaviour contains information on female reproductive output (Table 1). This is congruent with Burley’s (1986) differential allocation hypothesis, where differential allocation may be a way for females to obtain or retain high-quality mates. As the Banggai cardinalfish does not maintain long pair bonds (Vagelli 1999; personal observations), such behavioural display may provide males with an opportunity to assess the value of a mating with a certain female, again in addition to her size and condition. However, to fully disentangle the effect of male size from that of female reproductive investment on the rush behaviour, one needs to experimentally manipulate egg investment and see how the rush intensity is affected.

Female advertisement of fecundity has been suggested to be a requisite for mutual mate choice (Kokko and Johnstone 2002). Therefore it also makes sense in this species which exhibits mutual mate choice (Kolm 2002). Both sexes invest heavily in reproduction, with females investing in an egg clutch consisting of remarkably large eggs and males investing in a long brooding period (Vagelli 1999). To my knowledge, the only other example of a plastic signal that is related to fecundity was recently found in the lagoon goby (Knipowitschia panizzae) (Massironi, unpublished data). Here, female belly colour was positively correlated to fecundity, also suggesting honest signalling of an important trait for male mate choice in this species with more traditional sex roles.

But is this courtship behaviour a costly, and hence, honest signal? Honest signalling is expected when the balance of benefit and costs of a display yields increasing display effort with increasing quality (e.g. Grafen 1990; Kokko 1998). This is likely in the present scenario for several reasons. The rush behaviour was displayed rarely, especially in comparison to the twitch behaviour (Fig 1a). Since a rushing female swims past the male in a highly conspicuous manner, it is easy to imagine that this behaviour may draw the attention of potential predators. Moreover, a male may call a bluff as he broods the eggs in his mouth. A male receiving a clutch smaller than the female has “promised” through her courtship display can simply eat the eggs and spawn with another female. Brood cannibalism is frequent in other cardinalfish (e.g. Okuda et al. 1997) and is found also in the Banggai cardinalfish (personal observation).

Males may make use of the rush information, as spawning latency was negatively associated with increasing rush intensity, which may indicate that males actually choose females based on their display. A previous study (Kolm 2002) suggested that males, rather than females, may determine when to spawn, as in other sex-role reversed species, e.g. the deep-snouted pipefish (Syngnatus typhle) (Berglund et al. 1986). However, although males are likely, at least to some extent, to control the timing of spawning in the Banggai cardinalfish, I cannot rule out that this result is simply an effect of females closer to spawning rushing more often.

The twitch behaviour, however, showed no significant association with reproductive output. However, the relatively small sample size in this analysis may mask an actual association between twitch and reproductive output. Interestingly though, twitch was not correlated to the rush behaviour, suggesting different information contents between the two behaviours. Twitch was, however, displayed earlier after pairing than the rush (Fig. 1b), suggesting that it is important early with regards to pair formation. Further, both the rush behaviour and the twitch frequencies increased closer to actual spawning (Fig. 2); however, the twitch more strongly so (steeper slope on the correlation between the intensity of the behaviour and time until spawning) and, furthermore, the frequency of the twitch behaviour showed an exponential relation to the proximity to the actual spawning. This suggests that the twitch behaviour holds accurate information regarding how mature the eggs of a certain female are, and particularly so in the later part prior to spawning. By forming a pair with a female with more mature eggs, a male can thus increase his reproductive rate and also match the readiness of a certain female to his own readiness to spawn. This may be important as males in this species need at least a week to regain enough energy reserves to be able to brood another clutch after completion of a brooding cycle (personal observations). Such female displays indicating the stage of egg maturation have been reported earlier, in the sailfin molly (Poecilia latipinna) (Sumner et al. 1994) and the threespine stickleback (Gasterosteus aculeatus) (Wootton 1974). Also, female coloration is correlated to egg maturity in sticklebacks (Rowland et al. 1991; McLennan 1995) and, in mammals, olfactory cues may signal a female’s ovulatory cycle, as for instance in the rhesus monkey (Macaca mulatta) (e.g. Bonsall and Michael 1980). That the intensity of both female courtship behaviours increased closer to spawning may be due to females investing more into the eggs as they get riper. In the Banggai cardinalfish, females release the unfertilised egg clutch on the bottom if one removes the male prior to actual spawning (personal observations). This suggests that females cannot reabsorb the eggs and, in the case of male desertion, the female wastes a costly reproductive investment and hence the increased courtship closer to spawning.

To conclude, different female courtship displays apparently held fine-scale information regarding both the potential reproductive effort measured as clutch weight and how close to spawning a certain female was. Further, that both behaviours were displayed right up to actual spawning, suggests that they are likely to be important both to obtain and to keep a mate, and that quite substantial negotiation between the sexes may be necessary prior to spawning. I suggest that this is an example of how the mating game consisting of both conflict and common fitness interests between the sexes may select for complex sexual communication through displays between mating partners (Badyaev and Qvarnström 2002). However, experimental studies are needed to verify the exact causality behind these behaviours, and how males may use these female courtship behaviours when assessing potential partners.


I thank Jens Olsson for help with collection of behavioural data. Comments by Ben Sheldon, Anders Berglund, Anna Qvarnström, Claudia Fricke, Urban Friberg, Sarah Robinson-Wolrath, Göran Arnqvist, Anssi Laurila and three anonymous referees helped to improve earlier drafts. This study was financed partly by the Zoological Foundation and Inez Johanssons Foundation.

Copyright information

© Springer-Verlag 2004