Multiple selective pressures apply to a coral reef fish mimic: a case of Batesian–aggressive mimicry

Abstract

Mimics closely resemble unrelated species to avoid predation, capture prey or gain access to hosts or reproductive opportunities. However, the classification of mimicry systems into three established evolutionary mechanisms (protection, reproduction and foraging) can be contentious because multiple benefits may be gained by mimics, causing the evolution of such systems to be driven by more than one selective agent. However, data on such systems are generally speculative or anecdotal. This study provides empirical evidence that dual benefits apply to a coral reef fish mimic in terms of increased access to food (aggressive mimicry) and reduced predation risk (Batesian mimicry). Bicolour fangblennies Plagiotremus laudandus gained access to more reef fish victims, which they attack to feed on fins and scales, when they spent more time associated with their model Meiacanthus atrodorsalis. Furthermore, exact replicas of P. laudandus incurred fewer approaches from potential predators compared with control replicas that varied in resemblance to P. laudandus. Predators with trichromatic visual systems (three distinct spectral sensitivities) could potentially distinguish between replicas based on colour based on theoretical vision models. Therefore, this mimicry system could be best described as Batesian–aggressive mimicry in which mimicry evolution is driven by multiple simultaneous selective pressures.

1. Introduction

Mimicry is a classical example of Darwinian adaptation by natural selection in which mimics evolve to resemble other species to gain a fitness advantage. Traditionally, the classification of mimicry systems is based on the functional and ecological relationships between participants (Pasteur 1982), and in recent studies have been condensed into three mutually exclusive categories: protective, aggressive (or foraging) and reproductive mimicry (Zabka & Tembrock 1986; Starrett 1993) based on benefits accrued by the mimic. However, systems in which mimics gain multiple benefits spanning functional categories may exist (Malcolm 1990). Such mimicry systems may challenge the assumption that mimicry types are driven by three discrete evolutionary mechanisms (protection, reproduction and foraging), and simultaneous selective pressures may act on these systems.

For example, flies from the family Asilidae and the syrphid fly Volucella are considered to be aggressive mimics—defined as a predator or parasite that resembles another non-threatening (or even inviting) species to gain access to prey or hosts (Cott 1940; Wickler 1968)—of hymenoptera (bees and wasps), which they resemble closely. Asilids prey on smaller flying insects and it is thought that their similarity to hymenoptera allows them to approach their victims more closely before carrying out an attack. However, asilids do not approach their prey in a manner that would suggest that they are under disguise, and classical Batesian mimicry—defined as palatable species resembling toxic ones to avoid predation (Bates 1862)—may also be used to describe this form of mimicry (Poulton 1904; Evans & Eberhard 1970; Rettenmeyer 1970). Interestingly, this is a long-standing debate, as a rather heated discussion occurred between Poulton and Bateson in letters to Nature in 1892 on whether Volucella was an aggressive or a Batesian mimic (starting with Bateson 1892). However, a combination of mimicry types, termed Batesian–aggressive mimicry, may explain this type of mimicry system more accurately.

Another example involves snake mimics, such as king snakes from the Lampropeltis genus, which resemble venomous coral snakes. They are apparent Batesian mimics as they are protected from predators only in areas where the model is present and not in areas where the model is absent (Pfennig et al. 2001). However, it has been suggested that their conspicuous appearance may also assist in foraging, either by accessing bird nestlings as their parents react strongly to their bright coloration, or for other snake species as their presence may inhibit other predators attacking potential prey (Pough 1988; Malcolm 1990). They could therefore benefit both in terms of protection from potential predators and increased foraging access to prey (Malcolm 1990). However, to date, evidence to support mimicry types in which dual benefits may apply has been putative or circumstantial (robber flies, Poulton 1904; Evans & Eberhard 1970; Rettenmeyer 1970; coral snake mimicry, Malcolm 1990; reef fishes, Moland et al. 2005).

In this study, the putative example of Batesian–aggressive mimicry was investigated on coral reefs in the Indopacific (Moland et al. 2005). The bicolour fangblenny, Plagiotremus laudandus, closely resembles the forktail blenny, Meiacanthus atrodorsalis (Losey 1972; Springer & Smith-Vaniz 1972; Moland et al. 2005). The colour patterns of P. laudandus and M. atrodorsalis are nearly identical; both are blue anteriorly, and yellow posteriorly (figure 1a). When threatened, M. atrodorsalis delivers toxic bites to predators using enlarged canine teeth with venom glands at the base of each tooth (Randall et al. 1997). During experiments in which living Meiacanthus spp. were fed to predators, the majority of individuals were taken into the predators' mouth and promptly rejected (Losey 1972; Springer & Smith-Vaniz 1972). Plagiotremus laudandus is therefore thought to act as a Batesian mimic benefiting from its resemblance to M. atrodorsalis by gaining protection from predation (Losey 1972). However, P. laudandus is also considered to be an aggressive mimic (Losey 1972; Moland et al. 2005); it attacks passing coral reef fishes to feed on scales, fins and tissue (Randall et al. 1997) and may benefit from associating with M. atrodorsalis by being undetected by potential victims.

Figure 1.

(a) Photographs of (i) M. atrodorsalis and (ii) P. laudandus in the field. (b) A photograph of P. laudandus (replica 1) was manipulated to create replicas that varied in colour and pattern (replicas 2–3). A replica of Plagiotremus rhinorhynchos (replica 4) was used as a control.

While Batesian-type and aggressive-type benefits have been independently evaluated experimentally in coral reef fishes (Caley & Schluter 2003; Moland & Jones 2004; Cheney & Côté 2007), this study aims to provide empirical evidence that dual benefits can apply to a single mimicry system. Plagiotremus laudandus was assessed to determine whether it benefited in terms of (i) increased foraging success and (ii) reduced predation risk from its association with, and resemblance to, the model M. atrodorsalis in areas which varied in the abundance of models and mimics.

2. Material and methods

(a). Behavioural observations

Observations were conducted for 30 min on 40 P. laudandus that were haphazardly located on 10 coral reefs around Pulau Hoga 05°28′ S, 123°45′ E, southeast Sulawesi, Indonesia, in July–August 2006. Individuals were located using self-contained underwater breathing apparatus (SCUBA) at depths of 2–18 m and were found at least 10 m apart, therefore the assumption was made that individuals were not observed on more than one occasion. During each observation, the following was recorded: the total number of attacks made by a mimic, defined as a mimic darting towards a potential victim; the number of successful attacks, defined as clear contact made with the victim; and the time that each mimic spent within 30 cm of the nearest M. atrodorsalis model. After each observation was conducted, the abundance of all P. laudandus and M. atrodorsalis was recorded within a 5 × 5 m quadrat centred on each focal individual's location at first sighting. The numbers of potential victims/predators were also estimated at each site with five 5 min point counts, noting the number and species within or passing through a 3 × 3 m quadrat.

(b). Replicas of P. laudandus

The response of predatory wild fish to four fish replicas (figure 1b) during a 10 min trial was tested on reefs around Pulau Hoga and on fringing reefs off Lizard Island (14°39′ S, 145°26′ E), Great Barrier Reef, Australia. Replicas were constructed by gluing a photograph of a fish on either side of a 2 mm thick Perspex outline and then waterproofed with boat resin. The replicas were tested at eight sites on reefs off Pulau Hoga, all of which had both P. laudandus and M. atrodorsalis present in relatively high numbers (mean ± s.d. 100 m⁻²: P. laudandus 1.3 ± 0.2; M. atrodorsalis 3.2 ± 1.3), 10 sites on reefs at Lizard Island that had P. laudandus and M. atrodorsalis present in lower numbers (P. laudandus 0.5 ± 0.1; M. atrodorsalis 1.2 ± 1.3) and eight sites on reefs that did not have any P. laudandus and M. atrodorsalis present. Depths varied from 2 to 5 m.

The four replicas were constructed using the following photographs: (i) a true photograph of P. laudandus, (ii) and (iii) manipulated photographs of P. laudandus that varied in colour and patterns, and (iv) a photograph of P. rhinorhynchos (figure 1b). Photographs were manipulated using Adobe Photoshop CS by altering their colours to red and green (replica 2) or altering the colour information to shades of grey (replica 3). A replica of P. rhinorhynchos was used to control for the fact that predators could approach the manipulated replicas (2 and 3) more frequently owing to a novelty effect.

Methods for replica trials were modified from Caley & Schluter (2003). A trial involved observing a single replica placed at a fixed location for 10 min on snorkel or SCUBA. Each replica was given buoyancy by attaching one end of a section of monofilament line to a point on its upper back, and attaching a small piece of foam to the line approximately 2 m from the replica. The monofilament line was attached to a small piece of PVC pipe, around which the line could be wrapped and held by the observer. This line was used to create motion to the replica by the observer in a sequence of regular upward and downward movements. A second monofilament line was attached to the bottom of the replica and a small fishing weight was added approximately 30 cm from the replica. Separate trial sites were spaced at least 20 m apart to ensure that the same potential predators were not repeatedly exposed to the same replica. Replica type was randomized between days, sites and time. Individual sites were not visited more than once per day. All trials were conducted between 08.00 and 14.00.

The number and species of fishes approaching a replica was recorded during a 10 min trial. An approach was recorded when an individual fish swam towards the replica and either halted and visually inspected it within a distance of 2 m or actually bit the replica. Repeated approaches by the same fish were counted as a single approach. Fish species were classified as piscivores or non-piscivores (from fishbase.org). Non-piscivores were recorded approaching the replicas to examine whether another undefined aspect of a particular replica was attracting or deterring fishes.

The Vorobyev and Osorio colour discrimination opponent theoretical vision model was used (Vorobyev & Osorio 1998; Vorobyev et al. 2001) to determine whether a potential fish predator that had been observed approaching the replicas (Lutjanus bohar) and a potential victim that had been observed being attacked by P. laudandus (Acanthurus nigroris) had the visual capabilities to distinguish between the replicas based on colour alone. The vision model estimates a ‘colour distance’ (ΔS) between coloured spectra within a visual space, determined by the visual system of a signal receiver. Essentially, colours that appear similar to a signal receiver either because of the nature of the visual system or because of a small difference in the reflectance spectra of colours result in small ΔS values, while those that have high chromatic contrast have large ΔS values.

The spectral reflectance of each colour patch on each replica was measured with an Ocean Optics (Dunedin, FL) USB2000 spectrophotometer with a PX-2 pulsed xenon light source, a 200 µm bifurcated fibre optic cable and a Spectralon 99 per cent white reflectance standard (LabSphere, Inc., North Sutton, NH). The spectral sensitivity of L. bohar and A. nigroris is known from microspectrophotometry work performed previously (L. bohar: λ_max = 424, 494 and 518 nm; Lythgoe et al. 1994; A. nigroris: λ_max = 446, 514 and 526 nm; Losey et al. 2003). The proportions of spectrally distinct cone types and the Weber fraction were used as per a previous study (Cheney et al. 2009). Colour distances were calculated using illumination measurements at a water depth of 5 m.

(c). Statistical analyses

A multiple linear regression model was used with the success of attacks considered as the dependent variable; ratio of model : mimic, time spent with model and number of predators/victims in the local area as predictor variables. However, there was strong multicollinearity between the ratio of model : mimic and time spent with model, therefore the effects of time spent with model and ratio of model and mimic could not be disentangled without further experimentation. Correlation analyses were therefore conducted for individual variables. A repeated measures analysis of variance (ANOVA) was used to assess differences in predator and non-predator approach rates, replicas 1–4 were treated as within-subject factors, and sites (Lizard Island—no models/mimics present, Lizard Island—models/mimics present, Pulau Hoga) were between-subject factors. Statistical analyses were conducted using SPSS v. 11.0.

3. Results

(a). Behavioural observations

The total number of attacks made by a mimic ranged from 1 to 12 per 30 min (mean ± s.d.: 5.68 ± 3.06) and the number of successful attacks ranged from 0 to 9 per 30 min (mean ± s.d.: 2.98 ± 2.25). The time that each mimic spent within 30 cm of the nearest M. atrodorsalis model ranged from 0 to 30 min (mean ± s.d.: 13.78 ± 9.98). Both the total number of attacks and the percentage of successful attacks were positively correlated with time spent by the mimic in proximity to its model (total number of attacks: Pearson's correlation r = 0.66, n = 40, p < 0.001; successful attacks (%): Pearson's correlation r = 0.64, n = 40, p < 0.001; figure 2).

Figure 2.

The percentage (%) of successful attacks by mimic bicolour fangblennies (P. laudandus) in relation to (i) time spent within 30 cm of models (M. atrodorsalis) (white diamonds) and (ii) ratio of models to mimics (black circles); n = 40.

The maximum number of models and mimics found in one location was four models and two mimics, and a maximum of three mimics were found in the same location. Nine individual mimics were found without a model in close proximity. The ratio of models to mimics at each location was correlated with the percentage of successful attacks (Spearman rank r_s = 0.43, n = 40, p = 0.005; figure 2).

The total number of potential victims/predators ranged from 5 to 81 species (mean ± s.d.: 36.0 ± 20.1). The number of potential victims/predators did not significantly account for variation in the total number of attacks (t = 0.20, p = 0.84) or the proportion of successful attacks (%) (t = 0.34, p = 0.72).

(b). Replicas of P. laudandus

A total of 1230 fish approached the replicas during 104 trials. Piscivores included wrasses (Labridae; mainly Epibulus insidiator and Cheilinus diagrammus), groupers (Serranidae; mainly Cephalopholis cyanostigma) and snappers (Lutjanidae; mainly Lutjanus fulviflamma). On reefs where there were no mimics and models present (Lizard Island), there were no differences in the number of predators that approached the true replica compared with the other replicas (F_3,21 = 1.19, p = 0.34; figure 3a). However, at locations where both P. laudandus and M. atrodorsalis were present, significantly fewer predators approached the true replica of P. laudandus in comparison with manipulated replicas or the control replica of P. rhinorhynchos (range: 0–7 per 10 min; mean ± s.d.: 2.6 ± 1.4; repeated measures ANOVA: Lizard Island (with models and mimics): F_3,27 = 3.12, p = 0.04, figure 3b; Pulau Hoga: F_3,21 = 9.95, p < 0.001, figure 3c). Non-predators included damselfish (Pomacentridae), angelfish (Pomacanthidae), wrasses (Labridae) and surgeonfish (Acanthuridae) (range: 4–23 per 10 min; mean ± s.d.: 19.2 ± 10.4).

Figure 3.

The mean number of approaches by predators and non-predators towards replicas at (a) Lizard Island (no models/mimics present), (b) Lizard Island (models/mimics present), and (c) Pulau Hoga (models/mimics present). Error bars represent 1 s.e.m. Asterisk indicates significant difference from other bars (least significant difference post hoc, p < 0.05).

(c). Spectral capabilities of predators/victims viewing models and mimics

There was no overlap between the colours of replicas in the chromaticity diagram, and colours were distributed across different areas of the plot (figure 4); therefore, both fish have the theoretical capability of distinguishing between the various colours of the replicas.

Figure 4.

The trichromatic visual system of L. bohar (predator), dark circles, and A. nigroris (victim) open circles, represented by a two-dimensional chromaticity diagram corresponding to the photoreceptor noise-limited colour opponent model. The colours from the replica models were plotted according to eqn (B5) in Kelber et al. (2003). Chromaticity values that plot at a distance of less than one unit are unlikely to be discriminable along that axis. As plots increase in distance from one another, colours are more distinguishable. The origin corresponds to all achromatic colours including white, black and grey.

4. Discussion

This study provides empirical evidence that the P. laudandus–M. atrodorsalis mimicry system may be best described as Batesian–aggressive mimicry, as dual benefits are received by the mimic in terms of increased success in attacking reef fishes when in the presence of a model, independent of the number of potential victims in the local area, and a reduction in predation risk owing to mimetic resemblance. The avoidance of replicas by predators could have been a function of aposematic coloration, defined as the use of conspicuous coloration to warn potential predators they are chemically or otherwise defended, or an effect of increased detectability; however, we did not find an avoidance of P. laudandus-type mimics on reefs where M. atrodorsalis was not present. The relative benefits of increased foraging versus protection—both of which provide a significant fitness advantage—remains to be determined, and may underpin our understanding of the evolution of such mimicry systems.

Aggressive and Batesian mimics are typically understood to be in an evolutionary arms race (Dawkins & Krebs 1979) with signal receivers: mimics are under selection pressure to appear more similar to the models to prevent detection, while signal receivers are under selection to improve discrimination between models and mimics. In Batesian–aggressive mimicry, the selective agents driving the mimicry system are both predators and victims of attack, causing additional evolutionary pressures on the mimic to exist compared with either types of mimicry alone. Mimics may therefore evolve a more accurate resemblance to their models to ensure that they are not detected by a wide range of signal receivers with varying visual and perceptual capabilities. Reef fishes certainly vary in their visual capabilities (Losey et al. 2003), and therefore mimics may have to appear visually similar to each type of system to prevent discrimination. Indeed, the seemingly exact mimicry of P. laudandus–M. atrodorsalis could be a result of simultaneous defence from predators and foraging selection by prey.

In both Batesian and aggressive mimicry systems, mimics benefit from their resemblance to the model, while models incur costs (Huheey 1988; Côté & Cheney 2004, but see Rowland et al. 2007). In Batesian mimicry, models typically display a warning signal which predators learn to avoid. If predators come into contact with palatable Batesian mimics displaying similar warning signals, then predator learning is disrupted and models incur costs in the form of increased attack rates from predators. Similarly, in aggressive mimicry systems, mimics impose costs on their models including reduction in foraging rates (Côté & Cheney 2004), reduced reproductive opportunities and/or death (Lloyd 1965). However, in the P. laudandus–M. atrodorsalis system, predators will receive an adverse stimulus from aggressive mimics (in terms of being attacked), and therefore models could potentially benefit from the presence of mimics, as predators will learn to avoid both models and mimic. Thus, this mimicry system could also include aspects of Müllerian mimicry, defined as two or more defended species that evolve to resemble one another to enhance predator learning (Müller 1879). Instead of M. atrodorsalis models evolving new colour patterns to escape its mimic, the model and mimic may evolve to resemble one another. However, whether P. laudandus mimics do benefit M. atrodorsalis in this mimicry system remains to be determined.

Plagiotremus laudandus may enhance their foraging success as a Batesian mimic if they alter their behaviour in the presence of a model, for example, by spending more time in the open or less time engaging in anti-predator behaviour. However, behavioural observations of P. laudandus suggest that there is no difference in such activities between individuals that spend a lot of time with models compared to those that do not (K. L. Cheney 2006–2007, unpublished data). Further observations should be conducted to confirm this.

Empirical data should be collected on other types of putative Batesian–aggressive mimicry systems. For example, the false cleanerfish (Aspidontus taeniatus) closely resembles cleaner wrasse (Labroides dimidiatus) (Wickler 1968). Instead of removing ectoparasites from client reef fishes, these mimics have also been observed to attack fishes to remove scales, fins and body tissue. The mimic is reported to use its resemblance to cleaner fish in order to attract clients. However, recent studies have shown that the fangblenny mimic is rarely observed attacking fishes (K. L. Cheney 2006–2008, unpublished data; Moland et al. 2005). Indeed, fish eggs and demersal material are found more frequently in their diet (Kuwamura 1981). Therefore, fangblennies may also benefit from protection from predation by resembling the cleaner fish, which is thought to be generally immune from predation (Côté 2000).

This is the first study, to my knowledge, to provide empirical evidence that dual benefits can be bestowed on a mimic, and may demonstrate a system in which multiple selective agents can drive the evolution of a mimetic species. Despite the intuitive appeal of classifying mimicry systems under discrete, mutually exclusive headings such as Batesian, Müllerian, aggressive, or reproductive mimicry, such a classification system appears to be too simplistic. Indeed, a number of recent papers have questioned the classification of mimicry systems. For example, traditional Batesian versus Müllerian mimicry classification has been challenged by the quasi-Batesian mimicry theory (Speed 1999; Speed & Turner 1999). Mimicry systems are perhaps not such a clear dichotomy as has been previously supposed, and many examples may lie on a mimicry spectrum in which there is overlap between mimicry categories.