The variable colours of the fiddler crab Uca vomeris and their relation to background and predation

A lively scientific debate has recently developed on the evolution of animal signals in the context of sexual selection (e.g. Endler and Basolo, 1998; Price, 2006), on the information content of signals (e.g. Grether, 2000; Griffith et al., 2006), and on the cost of signalling (e.g. Olson and Owens, 1998; Grether et al.,2001; Møller and Nielsen, 2006) (for a review, see Maynard-Smith and Harper,2003). The two important costs animals are considered to incur are the energy needed to produce signals and the fact that signals make animals more conspicuous to predators. The question of how costs and benefits have driven the evolution of signalling systems is still largely unresolved. A full understanding requires insight into the natural operating conditions of communication through an accurate analysis of the signal itself, of how it is produced and perceived, and of the environmental conditions in which communication takes place. These crucial relationships between sensory systems, signals and the environment are largely unexplored (e.g. Endler and Basolo, 1998).

We report here on an unusually flexible system of colour signalling in fiddler crabs, which may eventually allow us to investigate and manipulate the costs of signalling in minute detail. Both male and female Uca vomeris have individually distinct colour patterns on their posterior carapace, which can change in a matter of minutes(Zeil and Hofmann, 2001). However, over years, neighbouring colonies of crabs can be persistently colourful or dull mottled.

Colour changes in fiddler crabs have been known to occur for a long time. Crane (Crane, 1944; Crane, 1975) and Altevogt(Altevogt, 1957; Altevogt, 1959), for instance,described how the `display colours' of fiddler crabs increase in intensity during the period of activity and how crabs may lose them rapidly during a fight. The changes in carapace colouration are also driven by endogenous rhythmic signals, which are synchronised with the local tidal regime (reviewed by Palmer, 1995). The physiology of these rhythmic colour changes has been extensively studied(Rao, 1985; Thurman, 1988). The colours are generated by a system of cuticular chromatophores having different spectral characteristics (black, red, white, yellow, blue)(Green and Neff, 1972; Robison and Charlton, 1973; McNamara and Moreira, 1983; Rao, 1985) in which the distribution of pigment is under the control of at least two hormones, a pigment-dispersing and a pigment-concentrating hormone (e.g. Quackenbush, 1981; McNamara and Ribeiro, 2000; Thurman, 1988; Nery and de Lauro Castrucci,1997) (reviewed by Rao,1985; Rao, 2001). The input to this hormonal control system appears to be related to endogenous rhythms (Granato et al.,2004), mating and social status, stress and possibly also fitness. Although rhythmic colour changes and colour variations related to stress have been known to exist in fiddler crabs for over 100 years (see Crane, 1975; Thurman, 1988), this phenomenon of facultative colour change has to our knowledge never been considered in the behavioural and ecological context in which the crabs live.

Different fiddler crab species – at least in Australia – show a high degree of colour variation, with their red, yellow, pink and white claws and legs and their blue, green, turquoise, yellow and white carapace colours. Not surprisingly, colour as perceived by humans is an accepted taxonomic feature (e.g. George and Jones,1982), and the colour patterns on the males' enlarged claw are used by the crabs themselves for species recognition and those on the posterior carapace for recognizing neighbours(Detto et al., 2006). The crabs may be able to employ colour vision in these recognition tasks, because behavioural tests (Hyatt,1975) and recent electroretinogram recordings(Horch et al., 2002) have provided evidence that fiddler crabs may be dichromats, with peak spectral sensitivities at 430 nm and 530 nm.

There are at least three, possibly interrelated, functions of carapace colours in fiddler crabs, in addition to their social signalling role: they could serve as a means of camouflage directed against predators and/or fellow crabs, they could signal distastefulness to bird predators, and they could help to control temperature (Thurman,1990). These functions may be modulated by motivation and/or stress, by the properties of the diet and the substratum, by age and by the population genetics. We will focus here on the first two functions, namely signalling and camouflage.

Measurements and experiments were performed with colonies of Uca vomeris (McNeill) (Ocypodidae: Brachyura: Decapoda) off Cungalla, at Bowling Green Bay, Townsville, Queensland, Australia (19°24.3′S,147°6.9′E) in October 2000, 2003 and 2004.

In order to measure spectral reflectance, we selected three distinct crab colonies that we had classified as `colourful', `dull' and `intermediate'during yearly visits starting in 1997. We caught male and female crabs randomly at these three sites, marked their burrows, and within 1 min of capture measured their reflectance relative to a white `spectralon' standard on three points on the posterior carapace using an Ocean Optics USB2000(Dunedin, FL, USA) fibre optics spectrograph. Illumination was provided by a Tungsten/Deuterium lamp combination through a bifurcated light guide, which ensured that measurements were restricted to 1 mm² of illuminated area on the carapace. Crabs were subsequently photographed and their carapace width determined with callipers to the closest 0.5 mm. Some of the most colourful crabs were kept in transparent plastic vials, filled to a depth of 5 mm with seawater, and were remeasured at about 15 min intervals for up to 1 h after capture to document colour changes under the influence of stress (see also Zeil and Hofmann, 2001). All crabs were subsequently returned to their marked burrows.

In order to assess how conspicuous crabs are against the mudflat background, from the perspective of predatory birds, we scanned crabs from a height of about 1 m and a viewing angle of 45° using a spectrographic imager (CASI, Itres, Canada) and a UV-sensitive camera (Hamamatsu Beamfinder II, with long-wavelength cut-off filters Oriel 5172+51124). The CASI is a fully calibrated spectrograph that allowed us to determine radiance throughout a scene in units of photons sr^–1 μm^–2s^–1 nm^–1. Details are given elsewhere(Zeil and Hofmann, 2001).

To relate carapace colours to vision in crabs and birds, we estimated the signals that these colours yield in the photoreceptors of crabs and birds and modeled chromatic and achromatic contrasts. There is electroretinogram evidence that fiddler crabs Uca tayerii possess two spectral types of photoreceptors: a short-wavelength receptor (S, peak at 430 nm) and the long-wavelength receptor (L, peak at 530 nm)(Horch et al., 2002). Birds typically possess four spectral types of single cones (VS, S, M, L, for very short-, short-, middle- and long-wavelength, respectively) and one type of double cone (D) (for a review, see Hart,2001). Visual systems of birds belong to two distinct groups:those that have VS cones with peak sensitivity in the UV range and those having VS cones with peak sensitivity in the violet range, with little variation within these groups (Hart,2001; Ödeen and Håstad, 2003; Hart and Vorobyev, 2005). To model a visual system of a bird having a UV receptor, such as terns [Laridae(Ödeen and Håstad,2003)], which are the main predators on fiddler crabs at our study site (Land, 1999), we used microspectrophotometry data obtained for the starling Sturnus vulgaris (Hart et al.,1998). A visual system of a bird having a violet receptor, like plovers and oystercatchers (Ödeen and Håstad, 2003), which are also known to prey on fiddler crabs(Ribeiro et al., 2003; Ens et al., 1993), was modeled using data obtained for the peacock Pavo cristalisurnus vulgaris(Hart, 2002). The spectral sensitivities were modeled using the analytical approximation of visual pigments (Govardovskii et al., 2000) and oil droplet spectra(Hart and Vorobyev, 2005).

To determine differential bird predation in different crab colonies, the bird presence and behaviour around a 5 m long marked transect at three sites was monitored by three observers, for 3 h before low tide on three consecutive days in October 2000. Observers were randomly assigned to each of the sites,starting 3 h before low tide, and rotated to a new site every hour. Observers judged the following categories of bird behaviour: whether birds approached the observation area to within 20 m of the transect, either walking, or flying high or low above ground. Further details are given in the Results.

Our aim was to quantify possible long-term colour changes under dummy predation pressure, by continuously monitoring and comparing the colour patterns between treated and untreated animals, which were otherwise left as undisturbed as possible. We exposed colourful individuals to artificial bird predation by repeatedly moving a 3 cm black Styrofoam sphere at a height of 20 cm directly over their burrows. Such a bird dummy is designed to mimic the trawling flights of terns (Land,1999) and is highly effective in eliciting escape responses in fiddler crabs (Hemmi, 2005a; Hemmi, 2005b). The dummy could be moved from a distance by a monofilament line attached to it, which ran through a system of stainless steel rod supports(Fig. 1). We selected pairs of colourful females with burrows about 80 cm apart and erected a plywood board screen 20 cm high and 1.2–2.4 m long between the two burrows, so that the bird dummy could only be seen by the experimental animal, and not by the control crab. We chose to work with females because they were more likely to be burrow-constant for several days. The longest time we managed to work with a pair of females was 7 days. We filmed an area of approximately 15 cm×20 cm around the burrow of each crab from above using two Hi8-Video Camcorders (Sony DCR-TRV250E; Shinagawa-ku, Tokyo, Japan) each day continuously for the whole activity period of the crabs (up to 4.5 h per day). The test crab was exposed to the bird dummy moving overhead every 3 min, a frequency that is quite close to what we determined the natural frequency of terns to be over several more exposed colonies (0.5–1 h counts on three occasions showed average tern return times of 2.5, 2.2 and 1.7 min).

We sampled video films at a rate of 1 frame s^–1 to dv-AVI format using a firewire IEEE 1394 interface and custom-made software (see Hemmi and Zeil, 2003). We then extracted individual frames whenever they showed a crab close to its burrow,together with their time stamp, and saved them as ppm files. Since illumination conditions changed throughout a day and especially between days,we matched the RGB histograms of all images, which we recovered from one experiment, to the RGB histogram of a background image section of one reference image, using ENVI software (ITT, Boulder, CO, USA) (see inset Fig. 1). The normalized images were cropped to an area of 150×150 pixels containing the crab. In the blue channel of these cropped images, we determined the location and number of pixel values larger than 200 (out of 256) or within other ranges of pixel values as indicated in the figures. This thresholding operation identified most of the area of bright colour pattern on the posterior carapace of the crabs, when mapped back onto the colour image (see Fig. 1). The numbers of pixels above threshold thus provided us with the simplest way of quantifying colour change over time.

The statistical analysis of the data presented in Figs 6 and 9 was done in R(R Development Core Team,2005) using a generalised linear model for count data(family=quasipoisson). The analysis for Fig. 6 was designed to ask whether there is a significant difference in the number of birds between the three sites and whether there is a difference in the number of birds across days. For the bird counts at three different crab colonies(Fig. 6B), the statistical model therefore was: bird_count∼site+date, with a residual deviance of 43.1.

To assess whether the colouration of crabs depends on dummy predation(Fig. 9B) we compared the colour changes during the activity period of test and control crabs. A significant interaction between the time of day and whether or not the crabs were exposed to dummy predation would indicate that the crabs' colouration depends on predation. We therefore correlated our measure of colouration(pixel_count) with crab, time from start of activity, state of predation and experimental day. Only significant terms were included in the final model, which had the following form: pixel_count∼crab+time*predation, with a residual deviance of 1541.

Both male and female Uca vomeris have individually distinct colour patterns on their posterior carapace, which can range from white, through brilliant blue, to a mottled appearance(Fig. 2). The patterns appear to develop first as two symmetrical spots on the ventral carapace (visible only to other crabs), which – as a crab becomes more colourful –fuse into a band before the more dorsal and lateral parts of the carapace become bright (visible also to birds; e.g. Fig. 2A, left row and Fig. 3A, row 4). This observation indicates that these colours have an intra-specific signalling function and are not primarily directed at bird predators in order to, for instance, signal distastefulness.

Reflectance measurements determined using an Ocean Optics Spectrograph from three points on the posterior carapace are shown in Fig. 3B. Female reflectance spectra are shown in light blue, male spectra in purple. Spectra in both males and females rise steeply towards a maximum between 450 and 500 nm, or between 400 and 450 nm. Because the mudflat reflects very little short wavelength light, the white and blue colours of Uca vomeris contrast strongly against this background and should therefore be visible to predatory birds,which are known to be sensitive to wavelengths in the UV-Blue part of the spectrum (Hart, 2001). The significance of cuticular colours in Uca vomeris in the context of predation is illustrated by the spectrographic imager scans in Fig. 4A and the UV images in Fig. 4B, which were taken from above looking down onto the mudflat. Note that the spectrographic imager is insensitive below 400 nm and measures radiance, the combined effects of the colour of illumination and the reflectance of the cuticle. Imager spectra therefore differ from the reflectance spectra shown in Fig. 3. They do show, however,that from the perspective of predatory birds, the blue-white carapace of colourful crabs are likely to be highly conspicuous both at human-visible wavelengths (Fig. 4A) and in the UV between 300 and 400 nm (Fig. 4B) against the mudflat background, which is brown but also riddled with specular reflections and shadows. We will justify this statement by quantitative modelling of chromatic and achromatic contrast below.

Uca vomeris crabs lose their brilliant colours within 15–20 min of capture, their individual patterns becoming visually less distinct(Fig. 5) (see also Zeil and Hofmann, 2001). The crabs are thus able to control the colour patterns they display on a short time scale and their darkening upon being handled indicates that stress may be one of the factors that elicit colour changes. Yet, despite this short-term control over their body colouration, we found over the years crabs to be reliably very colourful in some places at our study site at Bowling Green Bay and dull in closely neighbouring colonies. The colours of crabs ranged from predominantly mottled at the site we call `Central', intermediate at site`North', and very colourful at site `South' (see map inset Fig. 6). One possible explanation for such long-term differences is that crabs modify their colours depending on their perception of bird predation. We tested this hypothesis by recording the presence and the behaviour of birds (terns, herons, curlews,seagulls, kites) simultaneously at three U. vomeris colonies. The bird activity records in Fig. 6B and C show that the total amount of bird exposure is highest for the`Central' colony and lowest for the colourful colony at site `South'. Statistical analysis (see Materials and methods) shows that the only significant effect on the number of birds was site: d.f. 2/4; deviance 187.1; P(χ²)≤0.001. There was no significant change in the number of birds over the 3 days of the experiment (d.f. 2/4; deviance 14.5; P=0.51). Our own and other observations(Land, 1999) show that gull-billed terns Gelochelidon nilotica, in particular, are systematic and successful predators of U. vomeris at Bowling Green Bay.

We subsequently caught crabs as randomly as possible at each of the three sites to measure the reflectance spectra of their posterior carapace(N=21, 11, 10 for `South', `North' and `Central' sites,respectively). We found that the average reflectance of the crabs' cuticle was significantly different at the `South' colony, compared with the other two sites (Fig. 6A). The wavelength range over which there was no overlap between the average spectra at the mean± 2 s.e.m. level, is shown as black bars below the right panel(Fig. 6A), indicating the range in which the three mean spectra differ significantly (P<0.01).

Can crabs and predatory birds actually see these differences in carapace colouration? To answer this question, we modelled chromatic and luminance contrast as seen through crab and bird spectral sensitivities(Fig. 7, see Materials and methods for details). The colourful crabs in the `South' colony provide significantly higher chromatic distance to the background than the crabs from`Central' and `North' colonies, both for bird and crab visual systems(P≤0.001, t-test), and thus are clearly more conspicuous than the duller crabs at the other two sites. The differences in chromatic contrasts between `Central' and `North' colonies were not significant(P=0.16, 0.43 and 0.09 for a bird with UV receptor, with Violet receptor and for a crab, respectively, t-test). Chromatic distance to the white point and luminance signals between crabs from the `South' colony and those from the `Central' and `North' colonies was not significantly different for bird visual systems (P=0.15, 0.17 for a bird with UV and Violet receptor, respectively, t-test), but was significantly different for a crab visual system (P=0.01, t-test). The difference between the luminance contrast calculated for the visual systems of birds and crabs reflects the shorter wavelength tuning of the crab luminance system compared with that of birds (Fig. 7A) and the shape of reflectance spectra of colourful crabs, which differ from those of the dull crabs predominantly in the blue-green part of the spectrum (see Fig. 6).

Colonies with predominantly mottled crabs experience more passing and hunting birds than the colony with predominantly brightly coloured crabs(Fig. 6). However, there are many other factors, including differences in substratum and food availability that may underlie this correlation. We therefore tested directly whether perceived risk of bird predation can elicit colour change, by exposing individual colourful crabs living in an area close to the South colony over several days to dummy birds (for details see Materials and methods and Fig. 1).

Crabs respond to persistent dummy predation in a variety of ways, including ceasing activity on the surface early by closing their burrows, or abandoning the burrow altogether and wandering off in search for a new one. However, we were able to follow two pairs of female crabs (a control crab and a crab treated to dummy bird predation) over several days to determine the effect of dummy predation on their carapace colouration. Measuring this effect is complicated by the fact that the colours of individual crabs change throughout an activity cycle: crabs tend to emerge rather dull and then brighten over the course of the day, but the extent and speed of colour change differ on successive days. In addition, dummy predation does not affect carapace colouration as rapidly and clearly as catching and handling crabs does (see Fig. 5).

Over the course of days, however, crabs do indeed modify their bright colour in response to dummy bird predation, as documented by the images of a female crab in Fig. 8. The crab had been treated to dummy predation for 3 days (days 1–3) and served as a control on day 4 (see also Fig. 9A). On the first day, the crab brightened significantly in the course of 1 h but closed her burrow early, apparently in response to dummy predation. On days 2 and 3 of persistent dummy predation, the crab remained dull throughout her activity period, which she again ended early by closing her burrow. On the following day (day 4), dummy predation treatment was switched to the second (control) crab and the former test crab now stayed active much longer and regained, compared to day 1, most of her initial body colouration. To account for these long-term and dynamic aspects of body colouration, we present in Fig. 9 the full data set for two successful experiments involving both test and control crabs and in Fig. 10 the colour changes of an individual female subjected to three days of dummy predation. The images in Fig. 8 are those of Crab 1, Experiment 1, Fig. 9A. For each experimental day, we determined the number of pixel values above 200 on the carapace of crabs in consecutive video images and averaged these numbers over 30 min intervals. Regression lines were then calculated through these mean values over time.

As we have already seen in Fig. 8, the test crab 1 in Experiment 1(Fig. 9A) brightened significantly during the first 1.5 h of dummy bird predation, but closed her burrow early. On day 2 of dummy bird treatment test crab 1 emerged dark,remained dull for the remainder of the activity cycle and closed her burrow much earlier than the control crab 2. On day 3, test crab 1 had lost her colour to the extent that the previously colourful carapace patches did not reach threshold values in the normalized images, but could only be detected in the pixel value range of 100–155 out of 256 grey level values. On day 4,the treatment was switched to crab 2, which had served as the control on day 1–3. This crab largely ignored the bird dummy. Compared to day 2 and 3,crab 1 became significantly brighter again on day 4 and also stayed much longer on the surface. The reversal of treatment in this particular experiment could not be continued, because the bird dummy did not elicit escape responses in crab 2 and both females eventually abandoned their burrows on day 5. Although dummy predation clearly had a strong effect on crab colouration in this experiment, the structure of the data makes a statistical treatment too complicated.

Preceding the second complete experiment(Fig. 9B), crab 1 had been paired for 3 days with a second crab whose colour was restricted to the ventral parts of the carapace and could not be reliably detected and quantified from video images. Crab 1 had served as a test animal subjected to dummy predation on preceding days 1–3(Fig. 9B, top left panel: red regression lines). On day 1 of the present experiment, crab 1 served as a control for a new test female (crab 2) who continuously lost colour in response to dummy predation on day 1 (Fig. 9B, second panel from left). Treatment was then switched back to crab 1 on days 2 and 3, with crab 2 serving as the control. The two crabs responded very differently to the dummy predation treatment with the main effect being the rate of brightening or darkening throughout the activity period. Whereas crab 1 continued to brighten over the 3 days of dummy predation preceding the pairing with crab 2, losing colour only on day 3(Fig. 9B, top left panel), crab 2 rapidly lost colour on her first day of dummy predation(Fig. 9B, day 1, Fig. 9B, second panel row). During the following 2 days of switched treatment, crab 2 as a control became brighter throughout the activity period, particularly on day 3, while crab 1 became darker on both day 2 and day 3 of dummy predation. Both animals darkened during the day when they were subjected to dummy predation but brightened when used as control. This change in the slope of the colouration due to dummy predation (see Materials and methods) is statistically significant: the number of bright pixels is a function of the crab [d.f.:1/33; deviance 1108; P(χ²)≤0.001], the time since the start of activity and the predation state. There was a significant interaction between time and predation [d.f.:1/33; deviance 124.08; P(χ²)=0.01]. Both crabs respond in the same way to dummy bird predation: this interaction is not different for the two crabs[d.f. 1/28; deviance 26.84; P(χ²)>0.2].

Up to this point, our analysis of predation-induced colour change does not capture many of the subtleties that appear to be involved in the responses of crabs to predation stress. We have seen that the animals have a number of options when faced with increased predation pressure, including sealing their burrows early in the activity cycle, and that there are individual differences in how crabs respond. On a finer temporal and spatial scale, however, it becomes clear that colour changes are quite subtle indicators of predation stress. We illustrate this conjecture in Fig. 10, which shows the detailed temporal change of body colour of a female that has been exposed to dummy bird predation for 3 days. The figure shows the development of raw data over time, that is to say each consecutive measurement that we took, not only of the number of pixel values between 201 and 255(Fig. 10, filled circles), but also of the number of pixel values between 150 and 200(Fig. 10, open circles),together with 30 min mean values (Fig. 10, large red symbols, thick red lines). The record shows that the animal did not maintain high-intensity colouration under the stress of dummy bird predation. On the third day, the crab closed her burrow after rapidly losing colour during her first hour of activity on the surface and emerged an 1 h later with much of her colour regained. She then continued to be active on the surface for over 1 h, despite continuing dummy bird predation, but evidently could not sustain her bright body colouration. The body colours of fiddler crabs thus reflect, in quite subtle ways, either the stress the animals are exposed to or the effect of lost feeding opportunities.

We have shown that both males and females of Uca vomeris in Australia can be brightly coloured on their carapace, that these colours contrast strongly against the mudflat for both bird and crab visual systems,and that the crabs have short-term control over their colouration. Crab colonies differ in their colouration on a local scale and this difference correlates with local differences in bird predation. Colourful crabs become dull over the course of days, when exposed to regular dummy bird predation,demonstrating that fiddler crabs modify their social signalling system under the pressure or the perceived risk of predation. The strength of that response varies greatly between individuals, indicating that there are many factors that determine the maintenance of carapace colouration, including metabolic efficiency, time in the tidal cycle and reproductive and/or territorial status.

Examples of other animals modifying their body colours on a short time-scale in response to predation are rare. Salamander tadpoles, for instance, respond to the presence of fish chemical cues by changing colour to better match the available background(Garcia and Sih, 2003). In cephalopods, which can change their body colour patterns on the time scale of seconds, the high variability may serve functions in both camouflage and communication (Hanlon et al.,1999). However, the structure of these complex visual signals and their biological significance remain to be fully understood (e.g. Crook et al., 2002; Anderson et al., 2003). In most other cases, as far as we are aware, predation-related colour change is a long-term process. Local populations of guppies evolve different body colours,which are used in mating interactions, depending on pressure from visually hunting predators (Endler,1991). Guppies, however, do not appear to be able to change their colours on a short time scale. Equally, variable background colour matching in crab spiders has been described in relation to bird predators and insect prey(Théry and Casas, 2002; Heiling et al., 2005), but again, it does not appear to be the case that the spiders can adjust their colouration on short time scales.

The facultative change of body colouration in fiddler crabs thus raises a number of interesting issues: what is the meaning of colour patterns in fiddler crabs, and how do colour changes affect the roles these colours play?How are these colours produced and changed? How do crabs assess predation risk?

The function of body colours in crabs can range from providing crypsis,mimicry and temperature control to social signals with a variety of meanings(for a review, see Detto et al.,2004). In general, however, very little is known about the roles of colours in fiddler crabs. For the present case of Uca vomeris, we can rule out that these bright colour patterns make the crabs cryptic against the mudflat background, because as we have seen, they in fact contrast strongly, since the mudflat reflects little light at short wavelengths (Figs 4, 6 and 7). These colour patterns,especially when they extend to cover the dorsal carapace, are therefore very conspicuous to birds.

Equally, it seems unlikely that the blue-white carapace colours protect U. vomeris from predation by mimicking the colours of soldier crabs(Mictyridae), which can occur in huge numbers close to U. vomerishabitats and are considered to be at least slightly unpalatable to birds. First, terns and waders (e.g. Land,1999) feed on both soldier and fiddler crabs, including U. vomeris, and we have observed soldier crab activity close to both dull and colourful U. vomeris colonies (J.M.H., W.P. and J.Z., personal observations). Second, as we have shown, harassed crabs become duller, not brighter, as one would expect from a warning colouration. Third, populations of U. vomeris that are exposed to intense natural predation (e.g. Land, 1999) are dull,indicating either that the crabs in these populations are unable to maintain bright colours, or that the colours do not offer protection. Finally, if carapace colouration does provide a significant protection against bird predators, one would expect the colouration to be mainly associated with the dorsal part of the carapace, where it is clearly visible to birds, and not with the ventral carapace, where it is mainly visible to other crabs. However,it remains to be shown that colourful crabs are not avoided because of their colouration.

One consequence of bright reflecting body colours is that they do lower the crabs' core temperature, compared to darker colours(Thurman, 1990). At least in U. vomeris, if a crab becomes colourful, it always appears to do so starting with two blue spots on the lower back, a position that is largely irrelevant for temperature control.

The function that has clearly been identified for the type of colour patterns we observe in Uca vomeris is the social role they play in fiddler crab societies. In behavioural experiments, it has recently been shown that U. mjoebergi females identify conspecific males by the colour pattern of their enlarged claws and that U. capricornis males recognize their female neighbours by the individually distinct colour patterns on the female's carapace (Detto et al.,2006). The two Australian fiddler crab species U. vomerisand U. capricornis, in which the carapace colour patterns of both male and female crabs are known to be individually distinct (see also Shih et al., 1999), appear to form stable neighbourhoods, and individual recognition could help reduce agonistic interactions by maintaining this stability(Backwell and Jennions, 2004; Detto et al., 2006). Uca vomeris males, for instance, keep track of significant neighbours by remembering the location of their burrows, an ability based on path integration (Zeil and Layne,2002). Individually distinct colour patterns would aid in this recognition by increasing the detectability of neighbours and making them identifiable even at a distance. In the case of U. vomeris, the spectral characteristics of bright carapace colours do not vary much between individuals. Their dominant visual effect appears to be to provide a high-contrast, individually distinct pattern. An example in which colour itself carries intraspecific social information is that of the semaphore crab, Heloecius cordiformis, a sister genus of Uca amongst the Ocypodidae, in which claw colouration correlates with the size of crabs and differently so in males and females (Detto et al., 2004).

There are thus good reasons to believe that the body colours of Uca vomeris are part of a social signalling system and that the social system must be affected by the crabs' ability to modify their signals on both short and long time scales. Given the findings by Detto et al. that claw colouration is used for species recognition and carapace colour patterns for individual recognition (Detto et al.,2006), it is interesting to note that the claw colours of male fiddler crabs do not appear to show similar short-term variations as we have documented them here for the colour patterns on the posterior carapace of both males and females. An important question now is what social costs the crabs incur when they cannot or choose not to maintain their body colouration under the risk of predation. One testable prediction we would make, for instance, is that crabs in dull colonies, compared to colourful ones, should engage in more and repeated agonistic interactions, because it should be harder for them to recognize individual neighbours.

In contrast to most visual communication systems where social signalling is being studied, both male and female U. vomeris can be very colourful,suggesting that body colouration is important in a wider social context, for instance in species and neighbour recognition and territorial interactions, in addition to courtship and mate choice (see also Cheroske and Cronin, 2005). Furthermore, the crabs have the ability to modify their colours on a short time scale, potentially enabling them to tune their signals depending on prevailing conditions. This ability is interesting in itself, but also opens unique opportunities for the experimental manipulation of this signalling system. Facultative adjustments of visual signals are little understood (e.g. Bradbury and Vehrencamp, 1998),but may allow us to study in detail the information feeding into the crabs'assessment of the costs and the benefits of signalling.

We are grateful for the generous support we received from Itres, Canada,the manufacturers of the CASI. We acknowledge financial support from the ARC Centre of Excellence in Vision Science to J.Z. and M.V., from the Centre of Visual Sciences (CVS) to J.M.H., M.V. and J.Z., from the Swiss National Science Foundation to J.M.H., from the Human Frontiers Science Program (HFSP)to J.Z. and from an ARC Linkage Project Grant to J.M., M.V. and J.Z. We are grateful to Liz Howlett, Australian Institute of Marine Science (AIMS) for facilitating our work there. We thank Martin Hofmann and Katharina Siebke for technical support and Pat Backwell, Tanya Detto and Mandyam Srinivasan for their comments on the manuscript. We are grateful to referees for constructive comments that helped us improve our argument.