Is honesty the best policy? Testing signal reliability in fiddler crabs when receiver-dependent costs are high

Authors


Correspondence author. E-mail: r.wilson@uq.edu.au

Summary

1.Contests between conspecific males are an important method of establishing mating rights or territories, yet the potential costs of injuries are high. To reduce potential risks, males can use signals to convey information about their underlying strength to competitors, and although individuals could signal deceptively to gain an advantage, most signals seem to be reliable.

2.Theory suggests that signal reliability is maintained because individuals that signal unreliably may be punished (i.e. receiver-imposed costs). Manipulative studies support this idea, showing that low-quality individuals that are given high-quality signals bear substantial costs.

3.Here, we explore the importance of receiver-imposed costs in natural, un-manipulated populations. Specifically, we show how the likelihood of being exposed (i.e. potential receiver density) and the potential severity of punishment (i.e. average receiver size) affect the predominance of unreliable signalling among populations.

4.Male fiddler crabs, Uca vomeris, use their enlarged claws as signals and weapons in combat, and the relationship between claw size (signal) and strength (quality) determines whether or not a male is a reliable signaller. We predicted that the prevalence of unreliable signalling among individual males would increase in populations where the receiver-imposed costs of deception were low. That is, males should produce unreliable (large but weak) claws.

5.We show that individual crabs produce more reliable signals of strength in populations with a high biomass, where there are both higher densities and larger average body sizes of potential receivers. Our study provides evidence that receiver-imposed costs can maintain signal reliability in natural un-manipulated populations and supports contemporary models of aggressive signalling.

Introduction

Studies of animal communication have sought to identify the mechanisms that maintain signal reliability (Maynard Smith & Harper 1995; Johnstone 1997). Reliable signals are those in which the magnitude of the signal accurately represents the intrinsic quality being advertised by the signaller (Maynard Smith & Harper 2003). Although low-quality signallers might benefit from producing high-quality (i.e. unreliable) signals, most signals appear to be reliable. When the interests of signallers and receivers oppose (e.g. aggressive contests), reliability is thought to be maintained in one of two ways (Searcy & Nowicki 2005). Firstly, the magnitude of the signal may be mechanistically linked to the quality being advertised, such that poor-quality individuals are physically incapable of replicating a larger signal (‘indices’) (Rohwer & Ewald 1981; Fitzgibbon & Fanshawe 1988). For example, the pitch of roars produced by male red deer is constrained by the size of the larynx, such that larger deer inherently produce calls that are lower in frequency (Reby & McComb 2003). Secondly, signal reliability can be maintained via costs. These can be the costs of producing or bearing the signals (receiver-independent costs) or those imposed by receivers on unreliable signallers (receiver-dependent costs) (Yachi 1995; Kotiaho 2000; Maynard Smith & Harper 2003; Hurd & Enquist 2005).

During aggressive interactions, signals can be used by receivers to assess the resource-holding potential of an opponent and estimate the likelihood of success should physical conflict ensue (Bradbury & Vehrencamp 1998; Hughes 2000). This provides weaker individuals an opportunity to retreat without injury and allows stronger individuals to maintain dominance unopposed (Berglund, Bisazza & Pilastro 1996; Briffa & Sneddon 2007; Arnott & Elwood 2010). Thus, signalling can reduce costs for both of the prospective combatants. Although some aggressive signals are indices (e.g. Bee, Perrill & Owen 2000; Reby & McComb 2003), most are expected to be reliable as a result of receiver-dependent costs (Szamado 2011a) such as proximity to aggressive opponents (proximity costs) or punishment of unreliable signalling (social costs). Proximity costs can maintain signal reliability as individuals living closer together are more likely to encounter potential rivals, increasing the chances of physical combat and reducing the likelihood of an unreliable signal remaining undetected (Molles & Vehrencamp 2001; Szamado 2008). The evidence for proximity costs centres more around one-to-one threat displays; however, reliability of dominance displays (i.e. one-to-many) can be maintained via the commitment of individuals to the defence of long-term resources, and again unreliability is a less successful strategy (Szamado 2011b). Social costs can maintain signal reliability via punishment to weak individuals that falsely display high-quality signals and that are exposed as frauds by competitors. For example, female paper wasps (Polistes dominulus) use facial badges to signal their potential fighting ability to conspecifics (Tibbetts & Dale 2004). Researchers experimentally manipulated these badges so that some low-quality females displayed high-quality signals and found that these unreliable females endured more directed aggression from conspecifics than females with similar-sized but reliable badges (Tibbetts & Dale 2004; Searcy & Nowicki 2005; Tibbetts & Izzo 2010). These studies, which decouple signal magnitude from underlying quality, provide compelling evidence for the importance of receiver-dependent costs in maintaining signal reliability (Rohwer 1977; Lailvaux, Reaney & Backwell 2009; Tibbetts & Izzo 2010). Surprisingly, the importance of receiver-dependent mechanisms has yet to be explored in natural and un-manipulated populations (Searcy & Nowicki 2005).

In nature, signallers in populations with higher levels of intraspecific aggression should incur greater proximity and social costs as deception is more likely to be exposed and punished (Dawkins & Guilford 1991). Increased frequency or intensity of male–male contests (Bellows 1981; Arcese & Smith 1988) could occur via increases in (i) population density and/or (ii) the average body size of potential receivers (Szamado 2011b). Greater population density should increase the likelihood that an individual's signal will be tested and larger body sizes among receivers should elevate the physical cost of fighting or punishment. While fighting intensity increases when individuals in contests are evenly size matched, larger individuals have the capacity to inflict more severe injuries (Arnott & Elwood 2009). As a result, we expect reliable signalling to be ubiquitous in populations with greater and more costly intraspecific aggression. Detecting a relationship between the level of male–male competition and signal reliability would provide strong evidence for the importance of receiver-dependent costs in the maintenance of signal reliability in natural populations at an evolutionary equilibrium.

Fiddler crabs of the genus Uca offer an ideal system for studying the reliability of signals used in aggressive communication (Backwell et al. 2000; Lailvaux, Reaney & Backwell 2009). They live socially in mixed age and sex populations that can markedly vary in density and resource availability (Zeil & Hemmi 2006). Male fiddler crabs possess a greatly enlarged (major) claw that is used to signal to females during courtship and to other males during prefight assessment (Jordao & Oliveira 2001; Lailvaux, Reaney & Backwell 2009). Assessment of a rival's claw is the first stage in aggressive interactions between males and determines the initiation, escalation and duration of fights (Lailvaux, Reaney & Backwell 2009). As a consequence, it is the presentation of the claw to rival males that is the primary signal used to demonstrate the overall fighting ability for male crabs (Morrell, Backwell & Metcalfe 2005). Both threat and dominance displays are employed during prefight assessment. During physical fights, the ability of male fiddler crabs to inflict costly injuries upon opponents is mediated through their underlying claw strength. It is the relationship between claw size (signal) and strength (quality) that determines whether an individual is a reliable or unreliable signaller. Reliable males are those where claw strength matches claw size, while unreliable males are those where claw size is a poor indicator of strength. Such a mismatch between signal size and strength can result from the expression of different claw morphs (weak regenerated versus strong original claws) (Backwell et al. 2000; Lailvaux, Reaney & Backwell 2009; McLain et al. 2010) or from variation in the amount and quality of muscle present within the claw [e.g. freshwater crayfish; (Wilson et al. 2007)].

In this study, we examined how population density, average male body size and biomass could mediate the reliability of signals of strength in male fiddler crabs, Uca vomeris, with original claws. Based upon theoretical models (Szamado 2008, 2011a,,b,), we predicted that unreliable signalling would be more prevalent in populations when the receiver-dependent costs of deception remain low (i.e. low male–male competition). In particular, we predict that males should have lower relative claw strength when competition is low and underlying strength is less likely to be tested.

Materials and methods

We collected 575 male U. vomeris across ten populations situated along the south-east Queensland coast between September 2008 and April 2009. Crabs were collected by hand from mudflats of relatively exposed bays and local creeks. Individuals were communally housed within their population groups in 40-L containers maintained at 20 °C that contained a gravel substrate and shelter. Each container housed fewer than 10 individuals. All individuals were tested within 3 days of collection and then returned to their site of capture. For each individual, we recorded body mass, carapace width (anterolateral angle to anterolateral angle), length and depth and claw size. Body dimensions were measured using electronic digital callipers (±0·01 mm) (Model Q-1382; Dick Smith Electronics, Brisbane, Queensland, Australia) and mass with an electronic balance (±0·01 g) (Sartorius Mechatronics Australia Pty Ltd, Dandenong South, Victoria, Australia). All statistical analyses were performed using R (Version 2.11.1, R Foundation for Statistical Computing, URL http://www.R-project.org/.

Following the injury or loss of the major claw, the males of several species of Uca are able to regenerate a replacement (Crane 1975). Previous studies have identified morphological and physiological differences between the original and regenerated claw types (Backwell et al. 2000; Reaney et al. 2008; Lailvaux, Reaney & Backwell 2009; McLain et al. 2010). Of all the individuals collected, only 10% possessed regenerated claws (population proportion mean: 0·124 ± 0·034), and as males possessing regenerated claws are noticeably weaker compared with those with original claws of the same size (Lailvaux, Reaney & Backwell 2009), they were removed to avoid any effect on measurements of signal reliability. Analyses were performed with and without regenerated individuals; however, none of the major results were influenced by their removal.

To measure claw size, photographs were taken of each major claw using a digital camera (model#DSC-W5; Sony, Kensington, Victoria, Australia), against a background of graph paper to calibrate. Digital images were analysed using morphometric software (SigmaScan Pro5, Systat Software Inc., San Jose, California, USA), and seven measurements were recorded for each claw (Fig. 1). These measurements included (i) width at heel, (ii) width at dactyl/manus joint, (iii) length of manus from heel to joint, (iv) width of pollex at dactyl joint, (v) width of dactyl (at joint), (vi) length of pollex (tip to joint) and (vii) length of dactyl (tip to joint). We ran a principal components analysis (PCA) using the seven measurements recorded for all individuals combined to provide an overall measure of claw size and claw shape. For initial reliability comparisons, claw size was calculated based on raw, uncorrected values. A second PCA (PCAC) was performed on claw measurements corrected for body size (carapace width) to allow for other principal components to be identified. This was also calculated to ensure that for interpopulation comparisons, results were independent of any body size variability between sites.

Figure 1.

Diagram of the major claw of male Uca vomeris identifying, (a) The main components of the major claw. The dactyl is the moveable section. (b) The seven measurements recorded to describe claw size and shape using a principal components analysis.

Maximum strength was measured for the major claw of each male using a custom-built force transducer. The transducer consisted of a strain gauge (RS Components Pty Ltd, Smithfield, New South Wales, Australia) attached to the outer side of one of two parallel pieces of metal (0·7 mm thick) separated by a pivot point (1·7 mm thick). The strain gauge recorded the degree of flexion of the metal that was proportional to the force applied by the crab clamping down. Data were recorded in millivolts using Chart 5.0 for Windows via a powerlab (ADInstruments Pty Ltd, Bella Vista, New South Wales, Australia), which was connected to the transducer by BridgePod amplifiers (ADInstruments Pty Ltd, Bella Vista, New South Wales, Australia) and a custom-made Wheatstone bridge. Each crab was encouraged to grab the transducer at least five times, then rested for 5 min before being encouraged to grab the transducer another five times. The greatest grabbing force was taken as an individual's maximum claw strength. The transducer was calibrated daily using known weights, and force measurements were converted into Newtons (N) for analyses. Claw strength was square root transformed to increase normality (Quinn & Keough 2002), and only transformed values were used in all analyses. Maximum strength for each individual was compared against corrected and uncorrected size and shape measures using multiple linear regressions and model simplification to establish which morphological components of the claw best predict the variation seen in strength between male U. vomeris.

We examined three ecological factors that may influence major claw strength among populations of U. vomeris. For all ten populations, we estimated the average individual body mass, density and total biomass of males for a population. Density was recorded by counting the number of males present on the mud surface within 20-m transects. A transect consisted of multiple adjoining 2 × 5 m quadrats, and two transects were placed concurrently at each site (maximum total area of 200 m2). Two counts for each transect were recorded approximately 30 min apart. Given that the total area of each study site varied considerably, total transect area varied from 70 to 200 m2 depending on the size of each site (i.e. min: 7 quadrats, max: 20 quadrats). However, where possible, transects remained the same size and at all sites covered at least 50% of the total crab collection area. Also where possible, counts were conducted under approximately the same weather conditions, only when low tide fell between 10 am and 2 pm and most occurred during the week preceding the new moon to minimize variation in density because of the lunar cycle. Average mass of individuals within each population was calculated based upon the individuals collected at each site that were used in strength analyses. The number of male crabs collected was dependent on the density at each site and the ease of capture (min: n = 13, max: n = 98). Typically, fewer individuals were captured at sites with more compact and rocky substrates. Biomass was calculated by multiplying the population density and average population body mass to produce a measure in g m−2 of crab.

We quantified the level of competition in each population by measuring the number of male–male surface interactions using video recordings. At each site, two video cameras were mounted on tripods and placed on the mudflats, so they could record the behaviour of the crabs within an approximate 5 m by 5 m area. Cameras began recording 15 min after they were placed on the mudflats to ensure crabs were active and interacting again at the surface. The direction of the cameras was then subsequently panned every 5 min for one hour to ensure a different 5 m by 5 m area was being filmed. Shifting the direction of the camera did not disturb the crabs on the mudflats. From the video footage, we quantified the proportion of focal males from each population that aggressively interacted with other males during these 5-min segments. Focal males were chosen from any individuals observed within each segment of footage and as many males as possible were measured. The most discernible male was selected as the first focal individual and any males that interacted with the focal were then excluded to maintain independence of observations. Subsequent males were selected based on independence from previous focal males and ease of observation. A male interaction was defined as any action that occurred between conspecific males within three body lengths of each other and included waving, when the recipient was known, as well as direct physical contact. Focal males were considered as the units of replication and sample size varied for each population (min: n = 10, max: n = 85). The proportion of interacting focal males from each population was compared with density using generalized linear models with a binomial distribution.

In addition to claw strength, we also compared biomass, density and average receiver size to relative claw strength and population signal reliability. Measures of relative claw strength for each population were obtained by calculating the least square means of claw strength on claw size (uncorrected PC1). Means and standard errors were used for comparisons to ecological measures.

Population signal reliability explains the variability of signal quality within a given population. This represents the potential for receivers to accurately predict claw strength based on an assessment of claw size alone. Our measure of reliability differs from previous studies that have calculated residuals of behavioural signals (Hughes 2000). Because the enlarged claws of male fiddler crabs serve as both a signal and a weapon, one can use the relationship between the structure (size) and function (strength) of this weapon to assess the reliability of the signal directly. Thus, the possession of a weapon that is weak for its size is direct evidence of an unreliable signal of strength. Theoretically, a weapon that is stronger for its size could also be considered unreliable as size is not indicative of strength. However, for the purpose of this study, we consider only the former. We assume that a claw that is strong for its size elicits the same response from a receiver as a claw for which size and strength are evenly matched. In terms of physical contests, receivers would only be negatively impacted if they assume large claws to be weak and not if they assume large claws to be strong (in any capacity). For this analysis, the relationship between claw strength and claw size was examined for each population, independently. This provided a measure of relative reliability for individuals within their own populations. Residual values of strength were calculated by correcting for claw size (uncorrected PC1). Values were then squared to remove negatives for each population, and means and standard errors were derived for each. A population mean closer to zero was representative of an increasingly reliable population whereby claw strength would be highly predictable for any given claw size. Populations with low signal reliability (a higher mean value) would have greater signal variability between individuals causing strength to be highly unpredictable. Population reliability was compared with the ecological predictors (as above) using linear regressions.

Results

Claw size and shape

The first component of the PCA based on uncorrected claw measurements explained 95% of the variation observed in the data, with all vectors loading in the same direction. PC1 was considered as an overall measure of claw size. PC2 and PC3 accounted for <1% of variation in the data combined. In contrast, PCC1 was less powerful and PCCs 1–3 combined explained only 85% of the variation in the data. PCC2 was deemed a measure of claw shape based on the loadings generated from the analysis (Table 1). Manus size negatively covaried with dactyl and pollex length (as the relationships of the dactyl and pollex are similar, we will henceforth only use dactyl as our example). PCC3 represents a second descriptor of claw shape that indicates individuals with a larger heel width have a narrower dactyl (Table 1).

Table 1. The principal component loadings of size-corrected claw measurements (PCAC). Values represent the relative contribution of each of the seven claw measurements (see Fig. 1b) towards the data variation explained by each principle component. Missing values indicate loadings of <0·1. PCC1 represents claw size as all values have similar loadings in the same direction. PCC2 and PCC3 represent claw shape and describe variation in claw proportions and claw widths, respectively
 Claw measurementPCC1PCC2PCC3
1Width at heel−0·275−0·5200·573
2Width at dactyl/manus joint−0·466−0·185 
3Length of manus from heel to joint−0·403−0·256 
4Width of pollex at dactyl joint−0·3750·230−0·469
5Width of dactyl (at joint)−0·332−0·356−0·568
6Length of pollex (tip to joint)−0·3900·4750·165
7Length of dactyl (tip to joint)−0·3770·4740·305

Relationship between biomass and male–male competition

The proportion of focal males within a population that aggressively interacted (waving or fighting) with other males during the 5-min observation period was positively associated with population density (z = 2·301, P = 0·0214) (min: 0·18, max: 0·63). Thus, males from those populations with higher densities were more likely to be involved in aggressive interactions with other males, suggesting that male competition is density dependent. In contrast, the average mass of males within a population was not significantly associated with the proportion of males that were observed to aggressively interact (z = −0·705, P = 0·481). Although there was a general positive trend between a population's biomass and the proportion of males that interacted aggressively, this association was not significant (z = 1·899, P = 0·058).

Variation in strength and reliability among populations

Claw size was positively associated with body size (carapace width) (t = −87·37, d.f. = 550, P < 0·0001, r2 = 0·93) and claw strength (t = −15·81, d.f. = 550, P < 0·0001). Larger individuals possessed larger major claws that were stronger than claws of smaller individuals (Fig. 2). Maximum strength appears to be constrained by the size of the claw; however, below this upper limit strength was highly variable. This is highlighted by the relatively poor model fit (r2 = 0·31).

Figure 2.

The relationship between claw size and claw strength for 552 male two-toned fiddler crabs (Uca vomeris) across ten populations. The black dotted line represents the apparent upper limit of strength based on size, and the full black line is the linear relationship (t = −15·81, d.f. = 550, P < 0·0001, r2 = 0·313).

Once claw measurements were corrected for body size, claw size (PCC1) was not a predictor of claw strength (t = −0·63, P = 0·53). Both claw shape (PCC2) and claw width (PCC3) significantly influenced claw strength (t = −5·88, d.f. = 549, P < 0·0001 and t = 5·93, d.f. = 549, P < 0·0001 respectively); however, there was no significant interaction between these two predictors (t = −0·24, P = 0·82). Individuals with shorter (high PCC2) and wider dactyls (low PCC3) and a larger manus (high PCC2) were more likely to produce stronger claw forces.

Average claw strength significantly differed with population biomass (t = 2·48, d.f. = 8, P = 0·038). Individuals from populations with a low biomass possessed weaker claws than those living in populations with a high biomass (Fig. 3). Relative claw strength was also significantly positively associated with population biomass (t = 4·35, d.f. = 8, P = 0·002). Individuals were weaker for a given claw size in populations with a low biomass (Fig. 4). Neither average body mass nor the density of a population was independently associated with relative claw strength (density: t = −2·01, P = 0·09 and average mass: t = −1·11, P = 0·31). Males had relatively stronger claws in populations that had both large average body masses and high density. Claw size, shape and width (size corrected) were not significantly different among populations (PCC1: t = −0·88, P = 0·41, PCC2: t = −0·92, P = 0·39 and PCC3: t = 0·57, P = 0·58). In addition, there was no difference in average body size (carapace width) among populations (t = 0·012, P = 0·99). Average male body size and average body mass were significantly influenced by site area by proxy of transect area (width: t = 4·475, P = 0·002, average mass: t = 4·137, P = 0·003).

Figure 3.

The relationship between average claw strength and population biomass for 10 populations of Uca vomeris. There was a significant linear relationship present (t = 2·480, d.f. = 8, P = 0·038). Populations with a greater biomass were stronger.

Figure 4.

The relationship between relative claw strength and population biomass for 10 populations of Uca vomeris. Relative claw strength was calculated using least square means of strength (N) on claw size (PC1). A simple linear regression showed that population biomass has a significant effect on signal reliability (t = 4·35, d.f. = 8, P = 0·002).

Population signal reliability did not significantly differ across different biomasses (t = 1·10, d.f. = 8, P = 0·30) (Fig. 5), densities (t = 1·35, P = 0·22) or the average body masses of individuals (t = −0·39, P = 0·71), indicating that signal quality varied equally across all population.

Figure 5.

The relationship between population reliability and biomass using means and standard error for 10 populations of Uca vomeris. Mean reliability closer to 0 represents populations that were more reliable, that is, claw size is an accurate predictor of claw strength, and signals were less variable between individuals. Biomass does not influence the reliability of populations (t = 1·101, d.f. = 8, P = 0·303).

Discussion

Our study tested theoretical models of animal signalling that suggest receiver-dependent costs should be one of the primary mechanisms maintaining signal reliability at equilibrium (Szamado 2008, 2011a,,b). We predicted that among natural populations of the fiddler crab, U. vomeris, the level of unreliable signalling would be greater when the receiver-dependent costs remain low, that is low intraspecific competition. Thus, we expected signal variability would be greater in populations of U. vomeris with high biomass. However, we found no association between biomass and a population's overall signal reliability. In fact, we found substantial variability in signal quality among the ten populations of U. vomeris, which indicates that no population was made up entirely of reliable or unreliable signals. This finding supports theoretical assumptions that the success or failure of unreliable signalling depends on the persistence of at least some reliable signallers in the population (Krakauer & Pagel 1995). The fact that there were no differences in signal variability between biomasses may also indicate that it is equally costly (or beneficial) under any environmental conditions to signal unreliably (Szamado 2011b).

In contrast, and in support of our prediction that receiver-imposed costs should affect relative claw strength, we found that individuals from populations with lower biomasses were relatively weak for their claw size compared with than those from populations with higher biomasses. Moreover, our work is the first to examine the importance of receiver-imposed costs and signal reliability in natural, un-manipulated populations. Previously, only studies that have manipulated signal size (to decouple signal from quality) have found a relationship between receiver-imposed costs and signal reliability (e.g. Rohwer 1977; Rohwer & Rohwer 1978; Fugle & Rothstein 1987; Gonzalez et al. 2002; Tibbetts & Dale 2004; Tibbetts & Izzo 2010). For example, Rohwer (1977) found that male Harris's sparrows with badges that were manipulated to mimic a dominant individual received a higher proportion of attacks than the un-manipulated subordinates with similar badge sizes.

Some models of aggressive contests predict that when potential combatants are in close proximity and are matched in size and intent, prefight displays will not be adequate to resolve contests (Maynard Smith 1982; Hurd 2004; Szamado 2008). In these cases, it is more likely that (i) contests would escalate to physical contact (high proximity costs), (ii) deceptive displays would be exposed and (iii) weaker individuals would be forced to pay the increased combat costs (Dawkins & Guilford 1991; Lailvaux, Reaney & Backwell 2009; Wilson et al. 2009). Specifically, greater population densities should elevate the frequency of fights, and larger average body sizes should increase the magnitude of fighting costs. In our study, we found that male U. vomeris from populations with higher densities were more likely to be involved in aggressive interactions with other males. Thus, our data support the idea that density can alter the frequency of aggressive interactions and the likelihood of unreliable signals being discovered by receivers. Similar patterns have been observed in penguins (Waas 1991) and fulmars (Enquist, Plane & Roed 1985), where distance to rivals affected the intensity of displays, and fighting occurred more frequently when individuals were closer to each other. Waas (1991) also found that certain displays by little blue cave penguins were more likely to provoke attack by an opponent, and these high-risk displays were only performed when individuals were in close proximity. Further research is needed to examine and directly quantify the costs of fighting for individuals in populations where individuals have larger body sizes.

Variation found in relative claw strength among populations of male U. vomeris was likely due to differences in the amount of muscle present within the claw and its potential to generate force. We found that males produced weaker claws in populations with lower biomasses; however, why this occurs still remains unclear. The long-term effects of high competition or access to resources could explain why some populations have higher relative strength, even though signal variability was not seen within populations, as selection would likely favour strong claws when competition is high and weaker claws when low. When proximity to and size of opponents was low (i.e. low biomass), males were less likely to engage in costly combat; therefore, the importance of strength is likely to diminish under such conditions. As claw muscle is energetically expensive to maintain, males would therefore benefit from producing relatively weak claws by decreasing the quantity of muscle and thus minimize these costs. Nonetheless, the benefits of large claws are evident, and for male fiddler crabs, possessing a large claw often determines courtship success and eliminates the need for direct combat via successful prefight displays (Backwell et al. 2000; Reaney et al. 2008). Previous studies have found that males with smaller claws received fewer visits from females and retreated prior to physical contact more often than males with larger claws (Reaney et al. 2008; Lailvaux, Reaney & Backwell 2009; McLain et al. 2010). Alternatively, the availability of resources within a specific site may influence the amount of energy able to be allocated to claw and muscle growth (McLain & Pratt 2010). While we did not directly quantify resource availability, we did find that crab size (mass or carapace width) was correlated with the total area of a site; thus, potential resources available may be affecting both density and average crab size. However, even if the overall condition of the crab is influenced by resource availability, we suspect that this would not be driving the variability seen in signal quality, as claw size ought to be constrained under these circumstances as well as strength.

In U. vomeris, signal reliability is determined by the relationship between the claw size (signal) and claw strength (quality). We found that male strength differed greatly for any given claw size, but below an apparent upper limit that varied with claw size (see Fig. 2). Individual investment in claw muscle quality (e.g. Wilson et al. 2007; Mowles, Cotton & Briffa 2011) represents one of the possible explanations for this variation. Other factors likely include: (i) variation in the shape of the claw (e.g. Levinton & Allen 2005) and/or (ii) an individual's stage in the moult cycle (e.g. Skinner 1966). In our study, claw shape substantially influenced the maximum force that could be generated by the claw; stronger claws were more likely to have a shorter dactyl and pollex, and broader claw heel width (see Table 1). These characteristics impose biomechanical constraints on the generation of force by influencing the leverage achievable by a claw. Changes in leverage are also believed to affect force production in the regenerated claws of several fiddler crab species (Rosenberg 2002; Lailvaux, Reaney & Backwell 2009; McLain et al. 2010). Importantly, we did not find any differences in claw shape among populations. Claw strength is also dependent on the cross-sectional area of the muscle within a claw and the contractile ability of the muscle fibres. Crustacean claw muscle undergoes substantial atrophy to moult and grow (Mykles & Skinner 1982). Tissue atrophy alters the ability of muscle fibres to contract efficiently, and as a consequence, reduces muscle strength (Skinner 1966; Ismail & Mykles 1992). Although untested in crustaceans, it is probable that some of the variation in the signal quality of U. vomeris that we observed is because of the level of muscle atrophy at the time of testing.

Differences in signal reliability among populations of U. vomeris are not likely to result from genetic divergence. Fiddler crabs have complex life cycles that include a benthic adult phase and a planktonic larval phase, which serves as the primary means of dispersal for fiddler crabs (Brodie et al. 2007; Lopez-Duarte & Tankersley 2009). Larvae are released by females at high tide and migrate from estuaries to coastal waters to develop before returning to estuarine waters as megalopae to metamorphose (Godley & Brodie 2007). As a result of larval mixing within coastal shelf waters, it is expected that fiddler crabs have high gene flow and little differentiation between close populations. For example, a study by Sanford et al. (2006) found no geographical genetic variation among several populations of Uca pugnax found along Cape Cod, Massachusetts. Similarly, our populations were all situated in estuaries that flow tidally into Moreton Bay. Thus, we expect there to be little genetic differentiation between populations of U. vomeris sampled in this study as a result of larval mixing within the bay. We believe that differences in signal reliability and relative claw strength among the populations may be driven by variation in the social environment and resource availability during postmetamorphic growth (Kasumovic et al. 2011).

Conclusions

Our results show that the major claws of male U. vomeris were weaker for a given size when living in populations of low biomass. We suggest that this could be the result of receiver-imposed costs, even though level of signal variability did not differ between populations. This supports recent theoretical models (e.g. Szamado 2011a,,b) that suggest receiver-imposed and commitment costs may drive the reliability of a signalling system. Importantly, we demonstrate how an organism's social environment can affect both its physiology and behaviour in un-manipulated natural ecosystems.

Acknowledgements

We would like to thank G. David, S. Cameron, B. Barth, V. van Uitregt, D. Bywater and A. Lamont for assistance in the field and data collection. We also thank C. Condon for assistance with statistical analyses and A. Niehaus for manuscript revisions. This manuscript was substantially improved by comments from R. Elwood, I. Booksmythe and two anonymous reviewers. All research was completed in accordance with the University of Queensland Native/Exotic Wildlife and Marine Animals (NEWMA) guidelines.

Ancillary