Extravagant ornaments of male threadfin rainbowfish (Iriatherina werneri) are not costly for swimming


Correspondence author. E-mail: r.wilson@uq.edu.au; www.wilsonperformancelab.com (Lab)


  1. Exaggeration of male sexual ornaments should be costly, in terms of metabolic expenditure, resource allocation or even locomotor function. For example, many male ornaments are predicted to affect the aerodynamics, drag or biomechanics of movement and thus inhibit the speed or manoeuvrability of individuals; but empirical support for this is equivocal.
  2. We tested the locomotor and metabolic costs of exaggerated male ornaments in the threadfin rainbowfish (Iriatherina werneri), an Australasian native fish characterized by excessively long fin streamers. We predicted that males with greater relative ornamentation would have reduced escape abilities (i.e. burst swim speeds) as well as higher metabolic costs when resting or swimming. Furthermore, we evaluated the benefits of the signal by comparing the preference of females for males with differing amounts of ornamentation.
  3. As expected, we found that females spent more time observing (i.e. preferred) males with longer relative fins. We also experimentally reduced threadfin length and found that females continued to show preference for males with longer fins, rather than a preference for particular males.
  4. Male I. werneri with longer ornaments had higher resting metabolic rates, but we found no effect of ornament size on metabolic rates during swimming. Males with longer threadfins tended to swim faster, but our manipulation of fin length had no effect on burst swimming speed, indicating swimming abilities are not causally related to threadfin length.
  5. Overall, we found no evidence that the extravagant ornaments of male threadfin rainbowfish increase the metabolic or functional costs associated with swimming. Our results are surprising, given the high viscosity of water and the extreme length of I. werneri's ornaments. We suggest that future work should focus on the fitness costs of threadfin length, relative to reproductive output or survival under more natural conditions.


One of the initial challenges to Darwin's theory of natural selection was based on the observation that the males of many species possessed extravagant ornaments with such an obvious mechanism of reducing survival that their evolution seemed to represent a paradox. Darwin's answer to this conundrum was his theory of sexual selection, whereby, in contrast with natural selection, the success of an individual ‘depends, not on a struggle for existence, but on a struggle between the males for possession of the females; the result is not death to the unsuccessful competitor, but few or no offspring’ (Darwin 1859). In modern evolutionary biology, sexual selection is simply viewed as differences in reproductive success caused by competition over mates related to the expression of a trait (Andersson 1994). Male traits associated with reproduction can take many forms, including the extravagant ornaments used to attract females and the weapons used in battle or to advertise dominance (Endler 1992; Václav & Hoi 2002; Allen & Levinton 2007; Fitze et al. 2008). Exaggeration of male traits is known to increase reproductive potential in many species (Andersson 1982; Møller 1988, 1990), but traits cannot become increasingly exaggerated without a cost (Kotiaho 2001). Exaggeration is limited by counterbalancing selection, which can decrease the viability of an individual (Andersson 1994) via reduced growth (Royle, Metcalfe & Lindstrom 2006), increased predation (Forsgren 1992; Rosenthal et al. 2001; Walker et al. 2005), compromised immune response (Clotfelter, Ardia & McGraw 2007; Garvin et al. 2008), lowered physical fitness (Emlen 2001) or decreased fecundity from the preferential allocation of metabolic energy to the production of ornaments (Ghalambor, Reznick & Walker 2004). The most obvious and well studied of these costs is increased predation due to an increased conspicuousness to predators (Andersson 1994). For example, male three-spine sticklebacks with sexy red throats are twice more likely to receive predatory attacks than those individuals with dull throats (Moodie 1972).

Extravagant male ornaments may also hinder a male's locomotor abilities (Møller 1996; Oufiero & Garland 2007; Husak et al. 2011), making him more susceptible to capture by predators. One only has to imagine the difficulties experienced by male peacocks or birds of paradise, which carry around such ostentatious ornaments, to appreciate the logic behind this idea. Although it seems logical, the link between exaggerated ornaments and locomotor costs is poorly supported. Studies of the tail streamers of male barn swallows (Hirundo rustica) suggest that long tails are cumbersome and affect manoeuvrability (Møller 1994; Barbosa & Moller 1999). Manipulative experiments of the tail streamers of barn swallows indicate natural tail lengths of males are c. 10% longer than the optimum for flight manoeuvrability, and this reduces flight times through mazes by around 20% (Rowe, Evans & Buchanan 2001). However, recent manipulative experiments suggest the locomotor costs of many avian tail ornaments may be minimal at best. Clark & Dudley (2009) lengthened tail streamers of male Anna's hummingbirds (Calypte anna) by more than sixfold (3–19 cm) and found the metabolic costs of flight increased by only 11%, while the maximum flight speed decreased by only 3·4%. Similarly, substantial elongation of the tail ornaments of male red-billed streamertails (Trochilus polytrus) posed no increase in the costs of low-speed flight manoeuvrability (Clark 2011). Studies of male ornaments in fish are equally equivocal. Male swordtail fish (Xiphophorus montezumae) with intact swords have higher rates of oxygen consumption during both routine and courtship swimming than those males with swords experimentally removed (Basolo & Alcaraz 2003). However, male X. helleri with longer swords had higher fast-start escape performance than those with shorter swords (Royle, Metcalfe & Lindstrom 2006), but no influence of sword removal on their burst swimming performance (Baumgartner, Coleman & Swanson 2011), and no detectable relationship was found between sword length and swimming endurance for X. nigrensis (Ryan 1988).

Several alternate mechanisms could help explain the limited empirical support for the existence of locomotor costs for male ornamentation (Wilson et al. 2010). One idea is that organisms may possess one or more other traits that can effectively compensate for the negative effects of an exaggerated sexual trait on locomotor function (Balmford et al. 1994; Møller 1996; Oufiero & Garland 2007; Husak et al. 2011), thereby making the detection of any associated costs empirically difficult. Husak et al. (2011) reported that for sexually dimorphic species of stalk-eyed flies, males with larger relative eyespans (signals) were more likely to also possess greater relative wing areas (linked to flying ability). The locomotor costs for possessing larger eyespans were masked by the compensatory mechanism of growing larger wings. Another valid hypothesis is that many of the elaborate ornaments exhibited by males may simply have only subtle or negligible effects on locomotor function – the extravagant ornaments of many birds and fish may simply not bear a large enough additional weight or drag to affect locomotor function. One way of empirically disentangling the effects of compensatory mechanisms from the actual costs of possessing ornaments is by using manipulative experiments of ornament size or function. Only via the manipulative experiments of avian tail ornaments has it become increasingly apparent that their tail streamers can have minimal impacts on locomotor efficiency (Clark & Dudley 2009; Clark 2011). Although the males of many fish also display extreme ornaments, there is a paucity of manipulative experiments examining the costs and benefits of these structures (but see Basolo & Alcaraz 2003 and Kruesi & Alcaraz 2007).

We used males of the small Australasian native fish, the threadfin rainbowfish (Iriatherina werneri), to quantify the locomotor and metabolic costs associated with extreme ornamentation using both a correlative and manipulative approach. Threadfin rainbowfish are native to tropical Australia and Papua New Guinea and frequent thick-vegetated freshwater swamps and slow-flowing creeks where they grow to c. 5 cm in length, and the males possess extravagant fins that they use to attract females and intimidate other males. Male threadfin rainbowfish frequently flare their extravagant fins during both courtship and intrasexual displays, and the extensions of the filamentous rays of the second dorsal and anal fins are commonly more than twice their total body length (Allen, Midgley & Allen 2002) (Fig. 1a). It is likely these extended fin filaments increase hydrodynamic drag because they extend beyond the fish's body and regularly flutter during burst and routine swimming. The fins of this species are also amenable to manipulative studies because they are soft cartilaginous rays that can be altered surgically. First, we investigated the reproductive benefits of exaggerated male ornamentation using staged female preference trials. We predicted males with longer relative threadfins would gain more attention from females than males with shorter threadfins. Second, we examined the possible costs of increased ornamentation for (i) the metabolic demands of routine swimming, (ii) burst escape performance and (iii) resting metabolic rate. Based on the idea these ornaments will be costly, we predicted that fish with greater relative ornamentation would incur greater metabolic costs of swimming, reduced escape speed and greater maintenance metabolic costs. Finally, we manipulated fin length to examine its effects on attractiveness to females and burst swimming performance.

Figure 1.

(a) Individual male threadfin rainbowfish (Iriatherina werneri) displaying his long filamentous fins to a rival male (photograph: Dave Wilson). (b) The fin length measurements taken from each male were (1) length of the 1st dorsal fin, (2) length of the 2nd dorsal fin to the beginning of the fin extension, (3) the first threadfin extension on the 2nd dorsal fin, (4) the length of the second threadfin extension on the 2nd dorsal fin, (5) anal fin to the beginning of the fin extension, (6) length of the first threadfin extension on the anal fin and (7) the second threadfin extension on the anal fin. (c) The lines indicate the locations where manipulations of the threadfins occurred for the procedural control (dotted line) and the threadfin removal fish (dashed line).

Materials Methods

Eighty male (standard length: 2·6–4·0 cm) and 50 female (standard length: 2·5–3·1 cm) I. werneri that were collected from northern Australia were purchased from Eureka Aquarium Supplies (Wamuran, QLD, Australia) and transported to the University of Queensland where they were housed in 60-L single-sex aquaria at a density of 0·25 L−1. Water temperature was maintained by placing aquaria within a constant temperature room set to 28 ± 0·5 °C, and photoperiod was controlled by overhead fluorescent lights timed to 14:10 L:D. These conditions are similar to those encountered in their natural environments in tropical Australia. All fish were fed daily live brine shrimp nauplii. Fish were maintained in these conditions for 3 weeks prior to the beginning of the experiment where individual males and females were transferred into separate 4-L glass aquaria (15 × 15 × 20 cm) with a 2-cm gravel bed, a small amount of vegetation for shelter and constant aeration. All experiments were approved by the University of Queensland Animal Welfare and Ethics Committee prior to the beginning of the study.

Male ornamentation as a sexually selected trait

We assessed male attractiveness by quantifying the amount of female interest directed towards each male (N = 51) during a 10-min observation trial. At 1800 h on the day prior to observation, a tank containing one male was placed next to a randomly selected tank containing one female (N = 30 females). Black dividers were positioned around four sides of each male tank to prevent unobserved interactions between individuals. At 09.00 h the following day, the dividers between the focal male and female tank were removed, and the length of time that a female directed interest towards the male was recorded. Females were deemed ‘interested’ when they oriented towards the male and remained close (within 3 cm) to the side of the aquarium adjacent to the male tank. Four separate observations were conducted for each male with a different randomly chosen female over 12 days, with each male rested for at least 3 days and females 1 day between exposures. Male attractiveness was scored as the proportion of time that females directed interest toward a given male. As a males’ relative attractiveness is completely dependent on the female, we attempted to control for this variation by calculating a male's attractiveness as an average across the four trials.

Metabolic cost

Resting metabolic rate

To establish resting metabolic rate, fish were fasted for 48 h and then placed into a 50-mL syringe (i.e. respirometry chamber), covered with a dark cloth and allowed to settle for 6 min. A fibre-optic oxygen sensor (PreSens NTH oxygen micro-sensor; AS1 Ltd, Palmerston North, New Zealand) and meter (PreSens Microx TX3, Precision Sensing GmbH, Regensburg, Germany) connected to a PC running data logging software (PreSens Oxyview 5.31, Precision Sensing GmbH, Regensburg, Germany) were used to monitor and record the rate of change in oxygen content of the water as the fish respired over a 45-min period. Rate of oxygen consumption (inline image), a proxy for metabolic rate, was then calculated according to the equation:

display math

where mf is the slope of the relationship between oxygen saturation and time with a fish in the syringe (percentage air saturation h−1), mc is the slope for a control trial (i.e. a syringe without a fish, percentage air saturation h−1), V is the water volume of the respirometry chamber (0·05 L), and βo2 is the oxygen capacitance of air-saturated fresh water at 28 °C (5·77 mL L−1) (Riley & Chester 1971). The (mfmc)/100 is calculated to convert the saturation of water to fractional saturation.

Active metabolic rate

The metabolic rate of each individual rainbowfish was also analysed during active swimming. A sealed 250-mL respirometry chamber containing a magnetic stir bar was placed on top of a magnetic stirring device. The respirometry chamber was submerged in an aquarium maintained at 28 ± 0·1 °C. The respirometer was calibrated via particle image velocimetry to a swimming speed of 3 cm s−1 (c. 1·5 body lengths s−1).

For each measurement, fish were placed into the chamber and allowed to settle for 5 min. The water flow was then gradually increased to a stable current by mixing the magnetic stir bar at a low speed at the bottom of the vessel. The fish then swam against the water flow for 25 min, and the change in the oxygen content of water was assessed as described for the resting metabolic rate (RMR) measurements. To determine the metabolic costs of activity, we subtracted RMR from active metabolic rate (AMR) (Ohlberger, Staaks & Holker 2006).

Swimming performance

To determine the maximum escape velocity of each male, we recorded startle responses via a high-speed digital video camera. Startle or burst responses are used during predator evasion (Beddow, Van Leeuwen & Johnston 1995) and are frequently used as a measure of whole-animal performance capacity in fish (Condon & Wilson 2006), amphibians (Niehaus et al. 2012) and reptiles (Husak 2006). Threadfin rainbowfish live in environments that have abundant predatory fish, and burst swimming performance is likely to be an important means of escaping predation attempts. Startle responses were elicited within a glass swimming arena (30 × 20 × 5 cm) by tapping the glass with a fine metal rod 1 cm behind the caudal fin. A high-speed digital video camera (Redlake Motionscope model PCI 500, 8–48 mm Canon 1:1·0 lens, filming at 200 Hz; Tallahassee, FL, USA) filmed a dorsal view of swimming responses from a mirror suspended at 45° directly above the arena. For each male, at least four C-start responses were recorded. A C-start is the standard burst-startle response exhibited by most fish, when the initial contraction of all muscle on one side of the body curves fish into a ‘C’-shape directly followed by at least one vigorous tail stroke (Temple & Johnston 1998). All recordings were analysed with the accompanying Redlake software by digitizing the snout and tracking the displacement frame by frame across the first 20 ms of each response. Instantaneous measures of velocity (Umax) were subject to a three-point moving average filter (Wilson & Franklin 1999), and the greatest Umax from each fish was used as a measure of maximum escape velocity.


Fin size was determined from digital images of each male and analysed with ImageJ v1.4 G (National Institutes of Health, UK) morphometric software. Body length and 5 fin length measurements were recorded for each male (Fig. 1b), including (i) length of the 1st dorsal fin, (ii) length of the 2nd dorsal fin to the beginning of the fin extension, (iii) the sum of the first and second threadfin extension lengths on the 2nd dorsal fin, (iv) anal fin to the beginning of the fin extension and (v) the sum of the first and second threadfin extension lengths on the anal fin (Fig. 1b). We treated each fin separately, but if there were multiple ornaments coming from the one fin, then the lengths of these ornaments were summed to create an overall measure of ornament length for each fin. Principal component analysis of the five fin measurements was used to create new orthogonal predictor variables of overall fin size.

Manipulation of fin length

We then examined the effect of experimentally altering threadfin length on male attractiveness and swimming performance. Forty-four male rainbowfish were randomly allocated into either a control or trimmed fins treatment group. The soft ray tissue of both threadfin extensions for the anal (6 and 7 from Fig. 1b) and 2nd dorsal (2 and 3 from Fig. 1b) fins were excised under hypothermically induced anaesthesia, constituting a 40% average reduction in fin lengths (range was 20–75%). Male threadfin rainbowfish can vary substantially in their overall fin size, and the magnitude of fin manipulation used in our study was within this natural variation. This technique has been performed successfully in several other studies (Basolo & Alcaraz 2003; Kruesi & Alcaraz 2007), and soft fin damage is frequently regenerated within weeks in most teleost fishes. The control group underwent a procedural treatment whereby only 1 mm of the tips of the threadfins was removed, but otherwise experienced the same anaesthesia and handling as the trimmed treatment. Fish from both groups were then re-tested for both attractiveness to females and burst swimming performance using the same methods as previously described, and the results of these second trials were compared with original data.

Statistical analysis

Principal components (PCs) were calculated from fin size data to collapse correlated measures of ornament size into new orthogonal variables. For behavioural analysis of female interest, PCs were calculated on raw fin size data uncorrected for male body size. We used uncorrected fin measurements because we found only a weak and non-significant correlation between male body length and female interest (Pearson's product–moment: r = 0·26, 95% CI = −0·016, 0·5, t = 1·892, df = 49, P = 0·064). Arcsine transformation was used to normalize the proportion of observation time that females were interested in males prior to linear regression. As male standard body length was found to have a significant correlation with RMR (r = 0·62, 95% CI = 0·43, 0·76, df = 55, t = 5·91, P < 0·001), we recalculated PC on the residual scores of each of the 5 fin measurement regressed against standard length to control for body size (log–log scale). PCs on the size-corrected fin size data (PCc) were used for all analyses of metabolic rate and swimming performance. For swimming performance data, burst velocities were not corrected for male body size as no correlation between standard length and swimming performance was detected (= 0·18, t = 1·3531, = 0·18).

To examine the effect of the removal of male threadfins, we used a mixed-effect linear model to test the interaction between treatment group and time (initial and post-fin excision). Each individual fish was also included as a random effect to control for individual variation in behavioural and burst swimming responses between initial trials and after fins were experimentally cut. All data analyses were performed in R (CERN). Results are presented as means ± standard errors. Significance was taken at the level of P < 0·05.


We found the first dimension of the principal component (PC1) analysis for all five measures of fin length explained 80·5% of the total variation. Thus, PC1 was a good indicator of overall fin size, as eigenvector loadings across all 5 fins were approximately equal (Table 1). In contrast, PC2 indicates a negative correlation between the dorsal fins and anal fins.

Table 1. Principal components calculated on five measures of fin length: 1st dorsal fin, 2nd dorsal fin, 2nd dorsal threadfins combined, anal fin, and anal fin and threadfins combined. Principal components were initially calculated on raw measures of fin length. To remove variation due to body size, we recalculated principal components on the residuals of standard length and each fin measurement. Only the first two principal components for each analysis are presented
 UncorrectedBody size corrected
1st Dorsal0·43−0·460·430·87
2nd Dorsal0·45−0·420·450·03
2nd Dorsal threadfins0·45−0·200·45−0·25
Anal fin0·460·500·45−0·32
Anal fin–threadfins0·450·570·45−0·30
% Variance explained80·68·094·92·8

Female rainbowfish maintained a high level of interest in viewing males across all experimental trials and spent c. 75% of the total observation time observing the males (448·34 ± 13·97 s). Male rainbowfish with larger fins (PC1) gained more attention from females during the observation period than males with smaller ornaments (r2 = 0·10, t = −2·15, P = 0·03) (Fig. 2). Although PC2 explained an additional 7·8% of variation in fin size, this trait had no significant effect on male attractiveness (r2 = 0·03, t = −1·147, = 0·25).

Figure 2.

Relationship between overall threadfin sizes (PC1) and the proportion of time females spent directing attention towards individual males during a 10-min observation period (r2 = 0·10, t = −2·15, P = 0·03).

Our body size-corrected measure of fin length (PCC1) explained nearly 95% of the total variation in fin size (Table 1). Eigenvector loadings across all fins were approximately equal indicating that the correlation among the traits was largely due to size. PCC2 explained only 2·8% of fin variation; however, the fan-shaped 1st dorsal fin had a large loading along this vector (0·86) and a weak negative association with the morphologically distinct 2nd dorsal threadfin extension, anal fin and the elongated anal threadfin (Table 1).

We found that overall, males' O2 consumption was 4·5 times greater while actively swimming (0·376 ± 0·012 mLO2 h−1) than at rest (0·085 ± 0·002 mL O2 h−1). Male rainbowfish with larger fins had higher RMR (r2 = 0·29, t = −4·007, = 0·009, Fig. 3a), but there was no significant effect of threadfin size (PCC1) on AMR (r2 = 0·01, t = −0·007, = 0·9). PCC2, a measure of the size of the 1st dorsal fin, was significantly related to metabolic scope (i.e. the difference between AMR and RMR) (r2 = 0·09, t = 2·34, = 0·023). However, due to the relatively small (2·8%) amount of variation explained by PCC2, the biological significance of the metabolic cost of activity induced by the 1st dorsal fin may be negligible.

Figure 3.

Relationship between relative threadfin sizes (PCC1) with (a) the resting metabolic rate (r2 = 0·29, t = −4·007, P = 0·009) and (b) maximum burst swimming speed (r2 = 0·15, t = −1·694, = 0·096) of male threadfin rainbowfish (Iriatherina werneri).

There was a suggested trend between relative fin size (PCC1) and mean escape performance for male threadfin rainbowfish (r2 = 0·15, t = −1·694, = 0·096, Fig. 3b). Contrary to our predictions, males with larger relative fin sizes achieved greater average burst velocities across swimming trials.

We found a significant interaction between treatment groups when we compared the behaviour of females before and after manipulation (t = −2·06, = 0·04, Fig. 4a). Females spent significantly less time observing those males with experimentally shortened threadfins (29·0 ± 2·0% of total time) than control males that had only 1 mm of their fin removed (40·0 ± 2·0% of total time). We found no interaction between treatment group and experimental round on mean burst escape velocity (t = 0·63, = 0·53, Fig. 4b). There was also no significant difference in burst swimming velocity between trimmed and control groups (t = −0·61, = 0·54). However, burst swimming velocity was affected by experimental round (t = −3·73, < 0·001). Burst swimming performance across both the trimmed and control treatments post-fin removal was 13·5% lower than in the initial experiment possibly due to unintended physiological stress of the removal procedure (initial: 118·8 ± 2·7 cm s−1, 2nd round: 103·0 ± 2·1 cm s−1).

Figure 4.

Influence of manipulation of male threadfin length on (a) the proportion of time females directed interest towards the individual males and (b) the burst swimming performance of the individual males. Control (Cont) males received a procedural control of 1 mm removal of the four fin ornaments, while the full treatment males (Trim) had their ornaments removed back to the beginning of the dorsal and anal fins (see Fig. 1c). The first two columns refer to the data collected before treatments (pre-), and the last two columns are for the data after fin cutting treatments (post-). There was an effect of time and treatment on the proportion of time females directed attention towards the males, with the Trim treatment decreasing female attention (denoted by asterisk). Treatment did not affect burst swimming performance, but there was an overall influence of time.


We found no evidence that the extravagant ornaments of male threadfin rainbowfish (Iriatherina werneri) increase the metabolic costs of swimming or decrease maximum burst swim speed. In fact, male rainbowfish with longer fins tended to possess faster burst swimming speeds than those with shorter fins, although this specific correlation was non-significant. This result is somewhat surprising when one considers that water has a much greater viscosity than air and the effects of drag should pose greater demands on exaggerated ornaments in this medium (Ohlberger, Staaks & Holker 2006). Royle, Metcalfe & Lindstrom (2006) also found that male green swordtails with the longest sexual ornaments (swords) were faster, while the relationship between ornamentation and burst swimming performance was even more complex for males of the Pacific blue-eye fish (Pseudomugil signifer) (Wilson et al. 2010). In that experiment, the length of their first dorsal fin was positively correlated, while the anal fin was negatively correlated, with burst swimming performance (Wilson et al. 2010). In our study, the experimental reduction in male ornament length did not affect their burst swim speeds, showing that there seems to be no causal relationship between the longer fins of male I. werneri and the tendency for faster swimming speeds. This is unsurprising as the extended fin filaments trail behind the males when they swim quickly and are unlikely to provide any propulsive benefits. Given that the threadfins have no propulsive value and only trail behind when swimming, it is also highly unlikely that there is an optimum threadfin length for swimming that was greater than the length of those experimentally manipulated, but less than those un-manipulated. We were also able to discount the possibility that compensatory mechanisms were obscuring the detection of locomotor costs in threadfin rainbowfish because we could experimentally manipulate ornament size to make it independent of individual quality.

Male threadfin rainbowfish with longer relative fins received greater attention from females than those with shorter fins. This greater attention is indicative of a strong female mating preference for males with greater ornamentation. Female fish are known to prefer males with larger and brighter ornaments across a number of species (Basolo 1990; Brooks & Caithness 1995; Endler & Houde 1995; Karino 1997; Brooks & Endler 2001; Suk & Choe 2002). However, as we did not control (or quantify) the activity of the males during our female preference trials, it is possible that the females were responding not only to the male's ornamentation but also to his behaviour as well. As we found the males with longer fins possessed higher resting metabolic rates, it is possible that the fish with longer fins were indeed more active; however, we have no direct evidence that this is the case. The higher resting metabolic rates of males with longer fins are unlikely to be related to any metabolic costs of extra fin tissue, because the mass of the additional tissue is negligible. Instead, it is more likely that the greater metabolic rate of the longer-finned males is related to the hormonal demands of expressing a high-quality signal. For example, males of the house sparrow (Passer domesticus) that were experimentally administered with testosterone developed both larger signals of dominance (badge of status) and basal metabolic rates (Buchanan et al. 2001). A similar mechanism may operate for male threadfin rainbowfish, and further studies may reveal a hormonal basis for fin development and/or metabolic performance in this species.

Preference by females for males with longer threadfins provides evidence that male ornaments are sexually favoured, but it seems these structures pose no obvious locomotor costs. Costs are critical for the maintenance of signal honesty, and such theoretical predictions rely on two key requirements: (i) signalling is costly, and (ii) higher-quality signallers have higher absolute costs of signalling than lower-quality signallers, meaning that bigger signals cost more (Bywater & Wilson 2012). The case that locomotor costs are an important mechanism maintaining the honesty of animal signals remains somewhat limited. There are now several studies that have shown that ornaments or armaments have substantial locomotor costs (Rowe, Evans & Buchanan 2001; Basolo & Alcaraz 2003; Allen & Levinton 2007; Wilson et al. 2009), but there are just as many studies showing these costs are negligible or absent (Nicoletto 1991; Royle, Metcalfe & Lindstrom 2006; Clark & Dudley 2009; Wilson et al. 2010; Clark 2011), although this relationship may be somewhat obscured by compensatory mechanisms (Husak et al. 2011). Evidence that there is variation in the viability–fecundity trade-off between individuals of differing quality is almost entirely absent for those traits expected to affect locomotor function (but see Murai, Backwell & Jennions 2009). Revealing the importance of locomotor costs for the evolution of ornaments or armaments relies on such analyses. Furthermore, we suspect that the type of ornament or armament examined will have a substantial influence on the expected locomotor costs. The locomotor costs of armaments, such as fiddler crab and crayfish claws, seem to have more predictable and obvious costs (Allen & Levinton 2007; Murai, Backwell & Jennions 2009 Wilson et al. 2009) than many of the more ornamental structures of other taxa (Rowe, Evans & Buchanan 2001; Basolo & Alcaraz 2003). It is possible that sexually selected male traits only have associated locomotor costs when these ornaments require structural modification of the actual functional machinery involved in locomotion and directly hinder the functional capacity of the organism to move. At this stage, it is unclear whether the type or magnitude of costs for exaggerated sexually selected traits should differ between those structures used to attract females versus those used to fight other males, but we suggest this is an idea worth exploring.

The reliability of signals is maintained via the following: (i) morphological or physiological constraints on signal production, (ii) costs associated with developing or maintaining signals or (iii) costs imposed on dishonest signallers by conspecific receivers (reviewed by Searcy & Nowicki 2005). We found that the elongated fins of male threadfin rainbowfish appear unconstrained by physical traits, are not costly to metabolism or locomotion, and probably require minimal energetic expenditure to grow. Therefore, the reliability of this sexual signal is probably maintained via the costs associated with detection by predators or those enforced by competitors. Larger, showier fins are likely to increase predation risk by attracting the attention of predators and may increase the risk of combative injury by attracting the attention of other large-finned threadfin males. However, we cannot discount the possibility that fin size in these fish is related to testosterone production, which could decrease immunological function and shorten life span (Salvadora et al. 1995; Verhulst, Dieleman & Parmentier 1999); further studies should examine the role of hormones in this species.

An important theoretical concern for empiricists studying the costs of male ornamentation is that there are no quantitative predictions as to the actual magnitude of costs that should be expected. Put simply, when are the costs of great enough magnitude to ensure the reliability of signals? For example, a sixfold lengthening of the tail streamers of male Anna's hummingbirds only increased the metabolic costs of flight by 11% and decreased maximum flight speed by only 3·4% (Clark & Dudley 2009). Are these costs great enough to act as a mechanism that constrains only subtle increases in tail streamers? Male barn swallows possess tails that are around 10% longer than the optimum for flight (Bro-Jørgensen, Johnstone & Evans 2007), but why is this elongation similar for birds of differing quality? How does this flight cost actually translate into fitness? Theoretical models of signal honesty assume the currency of cost is fitness. Costs that are not measured in this form can then be difficult to interpret. This is not merely a semantic issue, and functional ecologists wishing to make significant advances in our understanding of the maintenance of reliable signals should consider measuring the associated fitness costs of ornaments or armaments that have a clear influence on locomotor function. Better still, studies could measure the fitness consequences of experimental manipulations in ornament size. This would undoubtedly allow us to better appreciate the importance of constraints on locomotor function, however minor, for an animal's reproductive success and survival.


Thanks to Billy van Uitregt, Ben Barth and Candice Bywater for assistance with the experiments. We also thank Amanda Niehaus, Michael Kasumovic and two anonymous reviewers for comments and suggestions that substantially improved the manuscript.