Elucidating current mechanisms of sexual selection within a species may help us understand past selective forces driving the processes of assortative mating, reproductive isolation, divergence and speciation (Darwin 1981; West-Eberhard 1983; Coyne & Orr 2004). Much of the empirical work investigating sexual selection has focused on visual and acoustic signals such as plumage or calls in organisms such as birds, fishes and invertebrates (Wing 1946; Armstrong 1965; Gilliard 1969; Milinski & Bakker 1990; Snedden & Sakaluk 1992; Kotiaho 2000), whereas comparatively few studies have looked at the importance of chemical cues such as pheromones in mate choice (Pivnick et al. 1992; Jones & Hamilton 1998; Marco et al. 1998; Martin & Lopez 2000).

Pheromones are species-specific chemical compounds predominantly used for intraspecific communication, yet they may also communicate information among heterospecific individuals of closely related species (Phelan & Baker 1987; Löfstedt et al. 1991). Pheromones often function as mate attractors among conspecifics (Eisner & Meinwald 1995) and may convey information about the prospective mates, such as quality and quantity of nuptial gifts (Dussord et al. 1991), developmental stability (Thornhill 1992), dominance status (Moore et al. 1997), body size (Shine et al. 2003) and degree of relatedness (Smith 1983).

Studies investigating the role of sex pheromones in the Lepidoptera have mostly focused on moths, where a total of 1500 species have known sex attractants (Eisner 1980; Arn et al. 1986). Sex pheromones in lepidopterans are divided into long-range and close-range signals. The former are involved in the attraction of mates over long distances, whereas the latter play a role in courtship behaviour (Hartlieb & Anderson 1999). When visual cues are not the primary means of mate location, females may release the long-range sex pheromones from abdominal glands that attract males to fly towards the source from long distances (Hartlieb & Anderson 1999). Males, on the other hand, usually release the close-range sex pheromones from glands on the wings accompanied by visual and mechanosensory cues when courting females (Grant 1987). Aside from their role in courtship behaviour and as anti-aphrodisiacs (Eisner 1980; Pivnick et al. 1992; Schulz et al. 1993; Anderson et al. 2000, 2004), these close-range sex pheromones can also facilitate reproductive isolation between closely related species (Phelan & Baker 1987; Löfstedt et al. 1991). Although not directly examined, many studies have inferred the use of chemical cues as an additional signal to visual cues in mate recognition in lepidopterans (Van Dongen et al. 1998; Jiggins et al. 2001; Fordyce et al. 2002).

Male sex pheromones in lepidopterans are usually associated with morphological structures called androconia, many containing modified scales in the form of hair pencils and folds on the wings (Wuest 1998; Hall & Harvey 2002). The scents are emitted from pheromone glands in the androconia during wing fluttering in courtship (Grant & Brady 1975; Baker & Carde 1979; Rutowski 1980; Pivnick et al. 1992; Schulz et al. 1993).

Males from the 80-species-rich African butterfly genus, Bicyclus, are primarily differentiated on the basis of the location and number of the scent glands and modified hair-like scales that cover these glands (Condamin 1973; Larsen 2005), presumably to assist in spreading male scents. As such, it is plausible to hypothesize that diversification of male androconial organs and associated scents may have been important during the speciation and divergence within this genus. To date, however, sexual selection mechanisms operating within the genus, Bicyclus, have been investigated solely for visual cues in a single member, Bicyclus anynana. It was found that females exert stabilizing selection on the dorsal pattern of the forewings in males, preferring to mate with males with intermediate-sized UV-reflective pupils, present in the centre of the eyespots (Robertson & Monteiro 2005). The use of multiple signals in mating systems in Bicyclus and the importance of these visual cues relative to other potential cues such as chemical cues remains unknown.

Here, we investigate whether putative androconial organs present on both the wing pairs of male B. anynana butterflies play an important role in female mate choice. In addition, we determine the relative importance of visual versus olfactory cues in sexual selection in these butterflies. Specifically, we ask whether (i) androconial organs present on the hindwing and/or (ii) androconial organs present on the forewing of males affect sexual preference in the females, (iii) androconial organs from the hind- and forewings are equally important in B. anynana female choice, and (iv) olfactory cues are more or less important relative to visual cues in B. anynana female choice. We address each question in a series of laboratory experiments.

We tested female choice inside a greenhouse by placing single virgin females in cylindrical hanging net cages (30cm diameter×40cm height) with two virgin males. The two males within one experimental unit were randomly chosen with the exception of having similar wing sizes. The butterfly treatments were performed 24h prior to observations, and a female was added to each cage immediately prior to mate-choice experiments.

The time to mating and the female mate choice were recorded for each trial. All three butterflies were then sacrificed and kept in a freezer for later analysis. Trials in which no mating occurred within 3h were excluded from the analysis. This time limit was arbitrarily chosen in a previous study (Robertson & Monteiro 2005) and was kept the same in this study so that the results could be compared across studies. All observations were done in a greenhouse, only on sunny days to control for the effect of weather on butterfly behaviour (Scott 1973; McDonald & Nijhout 2000).

A sign rank test was used to analyse differences in female mate choice in all the four experiments. The number of trials the females chose each of the two male treatments was recorded, and the probability of the difference between these values being significantly different from zero was obtained. A two-sample t-test was used to analyse differences in the length of time it took females to mate in the first experiment.

Figure 1 Methods used to prepare male treatments for (a) blocking androconia (including hair pencils) on the dorsal surface of the hindwing. Inset shows the proximal straw-coloured and the distal dark-coloured hair pencils covering gland-producing structures (not (more ...)

Figure 2 (a) Clear nail solution treatment applied to the dorsal surface of the male hindwing. (b) Untouched dorsal surface of control male hindwing. (c) Ventral hindwing of control males with applied solution. (d) Nail solution applied to the single antenna marked (more ...)

Forty-four virgin females were placed individually in a cage, each with two alternative virgin males. The ‘blocked’ males had both fore- and hindwing androconial organs blocked with identical methods used in experiments 1 and 2 (figure 2a,e). The control males received sham treatments on both fore- and hindwings with methods identical to experiments 1 and 2 (figure 2c,f). In addition to the sign test, a Z-test was performed on the proportion of females that chose the males without their androconia blocked (wild-type) and the proportion of females that chose the wild-type wing pattern from the female choice wing pattern manipulation conducted by Robertson & Monteiro (2005). Both of these experiments were conducted with identical methods other than the treatments administered (see Robertson & Monteiro 2005). Each experiment compared consisted of mate-choice trials with one virgin female and two alternative virgin males.

In all four experiments, observations concluded that it was female choice that dictated which males mated with the respective females rather than male–male competition. In most trials, both males courted the females with behaviours such as wing fluttering and wing vibrations while slowly approaching the females. Females illustrated a preference behaviourally in that at times they would refuse a male's courtship attempt, while being receptive to the other male's courting. Once the female started copulating with a male, the other male would sometimes attempt to disrupt copulation, but was always unsuccessful. In experiment 3, we observed that control males would often attempt to court the other male in the cage that had all of its androconia blocked. It is possible that the males emitting no male scent were perceived by the other males as a female.

Figure 3 Number of females that mated with ‘blocked’ and ‘control’ males across all experiments. All blocked males had androconia blocked for each respective experiment, whereas the control males had respective sham treatments. (more ...)

Behavioural observations throughout the trials suggest that it is female mate choice rather than male–male competition driving sexual selection in B. anynana in our laboratory setting. We found that androconia present in both the fore- and hindwings of male B. anynana butterflies play a role in female choice. These pheromones are perceived, at least for the most part, by the female antennae, because when these are blocked the females no longer discriminate between scented and unscented males. Surprisingly, females with blocked antennae take a similar amount of time to mate with soliciting mates as females with unblocked antennae. This result seems to indicate that when multiple stimuli are available to females to evaluate males, the presence of a correct stimulus in males (in this case the visual stimulus) is sufficient to lead females to copulation. When both stimuli are available, however, the males with the correct scent and visual stimuli are favoured to those only with the correct visual stimulus.

Pheromone glands and pheromone receptors can occur in a variety of locations in the Lepidoptera. In Colias butterflies, the glands are also found on the wing and the pheromones are believed to be secreted during the ‘wing clapping’ male courtship behaviour (Rutowski 1980). In moths, glands were found on the femur and tibia (Percy & Weatherston 1974), and the thorax and genitalia (Atkinson 1982). In addition, other Bicyclus species, but not B. anynana (Condamin 1973), have patches of velvety scales on the dorsal surface of the forewing. It is possible that these patches of scales represent androconia as well, but this has not been investigated for any of these species. Furthermore, female sensory organs are not limited to the antennae: in some lepidopteran groups, females are receptive to olfactory or chemical cues via their mouthparts as well (Bloomquist & Vogt 2003). It is possible that these additional sensory organs are also present in B. anynana, but our experiments (experiment 1) showed that just by blocking antennae, the strong preference towards ‘scented’ males was completely eliminated, suggesting the importance of antennae as primary sensory organs in the reception of male chemical signals in this species. The potential presence of additional pheromone glands in male B. anynana, not investigated in this study, could contribute, however, to us having underestimated the importance of olfactory cues in mate choice in B. anynana.

Assuming that we targeted the most important scent organs in the males, the results of our experiments combined with those of Robertson & Monteiro (2005) suggest that olfactory and visual cues play important roles in female mate choice in B. anynana. In addition, the results of our fourth experiment along with no detection of a difference in female preference between males with no olfactory cues (experiment 3) and males with no visual cues (experiments conducted by Robertson & Monteiro (2005)) suggest that these signals are relatively equal in their importance in our laboratory setting. While the relative importance of olfactory versus visual cues has been quantified in butterflies in other ecological contexts such as flower visitation and in other organisms in the context of mate choice (Kordric-Brown & Strecher 2001; Candolin 2003; Omura & Honda 2005), this is the first study to address the relative importance of multiple signals in sexual selection in butterflies. It is possible, however, that in this laboratory study, we have underestimated the importance of the visual cues relative to natural field conditions and, as a result, may also have underestimated their relative importance to olfactory cues relative to field conditions. We specifically targeted the UV-reflective pupils, present in the centre of the eyespots of the males as the visual cues because these were shown to be the only visual traits that influenced a female's mate choice (Robertson & Monteiro 2005). However, UV reflectance inside the mesh cages might have been reduced due to the reduction of UV light reaching the inside of these cages. In the greenhouse, the total UV content at 08.00 on September 18 was 1.48mWcm⁻², whereas outside the greenhouse it was more than double, 3.8mWcm⁻². Inside the mesh cages, total UV was further reduced to 0.29mWcm⁻². UVB content of light, however, was not so affected by the greenhouse setting (0.02mWcm⁻² both outside and inside the greenhouse and 0.01mWcm⁻² inside the cage). These measurements were similar at 13.00 only differing for UVB content outside the greenhouse which became 0.15mWcm⁻². Until we know more about the visual system of B. anynana, it may be difficult to extrapolate about the relative contribution of these important cues to the field situation.

It has been proposed that the magnitude of female preference may be positively related to the number of concurrent male sexual traits as in the three-spined stickleback fish (Künzler & Bakker 2001). Our study and that of Robertson & Monteiro (2005) also support the idea that females prefer to mate with males with both visual and olfactory cues present relative to males that carry only either a visual or a chemical signal. Despite the presence of two male signals, B. anynana females will mate with males producing only one of these signals, suggesting that the presence of both of these signals simultaneously is not vital for mate discrimination. Similarly, female Cycnia tenera moths will mate with courting males when chemical or acoustic signals were presented either alone or together (Conner 1987). Although in our study female Bicyclus butterflies will mate with males with only one signal present (experiment 4), they prefer to mate with males when they can perceive both signals (experiment 1).

Although we have identified the importance of chemical and visual cues in female mate choice in Bicyclus butterflies, the function of these cues has yet to be identified. Theoretical models presented by Van Doorn & Weissing (2004) suggest that female preference for multiple traits may evolve when the cost of choice is low and when different signals indicate different components of quality. It is possible that these two signals communicate different components of male quality, although a more thorough investigation of the function of these cues is needed. However, because females will mate with males with only one signal (experiment 4), the information conveyed by each signal, whether it may be redundant or not, is not necessary for a successful mating. Furthermore, it would be interesting to determine whether female Bicyclus would mate with males with both of the visual and olfactory cues blocked in further studies. This would not only confirm that Bicyclus can exhibit strong sexual selection in the experimental cages, but may also detect the existence of other cues and possibly a multimodal system.

The function of multiple signals may also differ on a spatial and temporal scale. Certain signals may also function over long distances or in close range (Candolin 2003; Lopez & Martín 2001), or function during different times in the courtship ritual (Shine & Mason 2001). In butterflies, it is suggested that visual cues function in mate location and that tactile and chemical cues follow during courtship (Siberglied 1984; Rutowski et al. 1991; Fordyce et al. 2002). Our study did not specifically look at the relative importance of these cues over long- versus close-range mate location and discrimination but does illustrate that they both are important in close-range interactions.

One last aspect that deserves consideration refers to the observation that in butterfly species in mimicry complexes (Schulz et al. 1993; Vane-Wright & Boppré 1993), there is often a shift from dependence on visual cues during intraspecific mate choice towards chemical cues such as pheromones. In the milkweed butterfly genus Amauris, for instance, there is a shift towards chemical communication among closely related Müllerian mimics. Although these chemical signals may not solely mediate species isolation, they can provide information on the identity of the male to the female and aid in species recognition in mimicry complexes where visual communication can be problematic (Schulz et al. 1993; Vane-Wright & Boppré 1993).

Members of the African endemic genus, Bicyclus, are not currently understood as belonging to mimicry complexes either with each other or with closely related genera, but perhaps they should. In support of this idea, we observe that all 80 Bicyclus species share fairly similar ventral wing patterns many occupy the same forests and some of the males in the genus are quickly identifiable in the field by their strong musky odour. For instance, it is possible to collect over 20 different species within a small patch (1mile²) of rainforest in at least two different localities (Kibale Forest in Uganda, and the Dja Reserve in Cameroon; A. Monteiro 1997, 1998, personal observation). Bicyclus anynana, in particular, can be trapped with at least four other members of the genus in Malawi displaying similar wing patterns (Condamin 1973; Windig et al. 1994; Roskam & Brakefield 1999). In Lepidoptera, wing patterns have been suggested to aid in reinforcement of recently diverged species (Howard 1993; Fordyce et al. 2002; Lukhtanov et al. 2005; Mavárez et al. 2006). The ability of members of the genus Bicyclus not to rely exclusively on visual cues in sexual selection would decrease the chance of maladaptive matings in forests populated with heterospecifics. Therefore, the use of multiple signals in Bicyclus may benefit the female's intra- and interspecific mate evaluation.

In conclusion, this study found that in addition to visual cues such as wing patterns, chemical cues also play a role in intraspecific communication for mate choice in B. anynana. In addition, we found that both signals are equally important in the process of sexual selection by female choice in the laboratory and that females will mate with males exhibiting only one of these signals. The function of these cues and the adaptive significance of multiple signals in this mating system need to be further evaluated.

We thank Kendra Robertson for advice on the practical aspects of these experiments and William Piel, Diane Ramos, Daniel Poland, Derrick Dupois and three anonymous reviewers for comments on the manuscript. We also thank Amanda Diaz, Bin Chen, Gary Glaser, Kyle Golden, Andrew Goldman and Diane Ramos for the maintenance of laboratory butterflies used in this study, and Dennis Pietras for the plant husbandry. This study was funded by NSF award IOB-0316283 to A.M.