5 3 4

Th e most parsimonious scenario to explain the distribution of cleaning in gobies is that this behaviour has evolved independently twice in this group (see also Rüber et al., 2003). One evolutionary transition probably occurred following the divergence of the Elacatinus and Tigrigobius clades (Fig. 4.3.1), explaining the occurrence of cleaning in the Pacifi c species E. puncticulatus and the Atlantic species E. evelynae, E. oceanops, E. genie, E. illecebrosus, E. prochilos, E. fi garo and E. randalli. Cleaning behaviour was then lost in E. atronasus, perhaps with the adoption of a schooling, non-benthic habit (Taylor and Akins, 2007), and in the horsti group (Rüber et al., 2003). It is likely that cleaning in the two newly described Brazilian species E. phthirophagus (Sazima et al., 2008) and E. pridisi (Guimaraes et al., 2004) also arose in the same event. A second origin of cleaning must have occurred some time aft er the separation of the Atlantic and Pacifi c species of the Risor clade (Fig. 4.3.1), which explains the presence of cleaning in four of the fi ve species in this group. Th e fi ft h, Elacatinus janssi, does not appear to clean (although little is known about this species). In the Elacatinus clade, the absence of cleaning appears to be associated with a sponge-dwelling habit or, conversely, the presence of cleaning is associated with living on substrata other than sponge (Rüber et al., 2003; Taylor and Hellberg, 2005). Th is alternative substratum is oft en, but not always, live coral (e.g., Sazima et al., 2000; Whiteman and Côté, 2002; Sazima et al., 2008). Interestingly, this habitat-feeding mode link also occurs intra-specifi cally. For example, in Barbados, the broadstripe cleaning goby E. prochilos occurs on both sponge and live coral (Whiteman and Côté, 2004a), and its foraging mode and social systems diff er between the two substrata. Coral-dwelling E. prochilos are active cleaners (Arnal and Côté, 2000; Whiteman and Côté, 2002) that live in singly or in small groups (Whiteman and Côté, 2002) while sponge-dwelling E. prochilos occur in large, dominance-structured groups and feed predominantly on polycheate worms that burrow within sponge tissues (Whiteman and Côté 2004b). Variable habitat use has been noted in other species, such as E. evelynae (White et al., 2007) and E. fi garo (Rocha et al., 2000), and in both cases, sponge-dwelling individuals did not clean. Note that there appears to be geographic variation in habitat associations: E. evelynae in Barbados occur virtually exclusively on live coral (Whiteman and Côté, 2002) while in the US Virgin Islands, they are found on coral as well as in basket sponges (White et al., 2007). Habitat associations are also apparent in the Risor clade (Fig. 4.3.1). Groups of non-cleaning species in this clade are specifically associated with chiton burrows, sea urchin spines, live coral or sponges (Taylor and

Hellberg, 2005). However, the group in which cleaning has arisen lacks strong habitat specialization. In both the Elacatinus and the Risor clades, it therefore appears that specialization in feeding mode (i.e., cleaning) has emancipated gobies from habitat specialization, although both types of specialization effectively lead to ecological separation that may have promoted the early stages of diversifi cation in this group (Streelman and Danley, 2003). Foraging specialization is often accompanied by morphological adaptations. It has been long thought that a ‘picker-type’, terminally located mouth may pre-adapt fish for cleaning (Hobson, 1971, 1976; McCourt and Th omson, 1984). Most Elacatinus (sensu lato) gobies do indeed have a terminal mouth; however, this ancestral morphology has been replaced in four species in the Elacatinus clade-E. evelynae, E. oceanops, E. genie and E. illecebrosus-by a subterminal, inferior mouth. Molecular evidence suggests a polyphyletic origin for this new mouth position (Taylor and Hellberg, 2005). In hindsight, the initial notion that terminal mouths would be well-suited to removing ectoparasites was a sensible one, but only for species that clean in mid-water, such as the damselfishes and wrasses observed by Hobson (1971, 1976) and McCourt and Thomson (1984). For species with a benthic habit, such as gobies which lie on their clients as they would on the substratum, a ventrally located mouth would facilitate cleaning. If so, the question becomes why has an inferior mouth evolved in some but not all cleaning gobies? Th e retention of a terminal mouth could be linked to fl exibility in foraging mode. Th is could explain mouth morphology in cleaning gobies of the Risor clade, which have been described as facultative cleaners (Humann, 1993; Allen and Robertson, 1994), and in E. prochilos and E. fi garo, which are both known to exhibit diff erent foraging modes when occupying diff erent substrata (Rocha et al., 2000; Whiteman and Côté, 2002). However, E. evelynae also shows habitatdependent foraging modes (White et al., 2007) and yet has a subterminal mouth. Some insight into the evolutionary link between feeding morphology and cleaning behaviour could be gained by examining other cleaning taxa (e.g., the family Labridae). The final potential correlate of cleaning behaviour is colour pattern. The presence of a cleaner signaling coloration was posited long ago by researchers who were struck by the similarity in colour patterns of cleaning species from various parts of the world (Eibl-Eibesfeldt, 1955; Wickler, 1963; Potts, 1973a). Côté (2000) provided the fi rst test of this hypothesis and showed that cleaning gobies have a longer lateral stripe, relative to body length, than their non-cleaning relatives. By contrast, among more

varied taxa of Caribbean facultative cleanerfi sh, there was no evidence that cleaners were more likely to be striped but they tended to be more brightly coloured than non-cleaning congeners. Although this analysis was based on taxonomy, the conclusions were later supported by Arnal et al. (2006) who, using a well-resolved molecular phylogeny, found that cleaning behaviour in wrasses is associated with the presence of a dark lateral stripe. In addition, in fi eld experiments using wooden models of cleaner wrasses with and without stripes, the former attracted many more fi sh clients (Stummer et al., 2004). In the Elacatinus clade, there appears to be an association between a subterminal mouth position and blue lateral stripe (Taylor and Hellberg, 2005). Incidentally, the spectral refl ectance of the blue stripe in gobies appears to be similar to that of the ubiquitous Indo-Pacifi c cleaner wrasse Labroides dimidiatus (Taylor and Hellberg, 2005), which is distinct from the blue hue of virtually all other coral reef fi shes (except for the cleanerfi sh mimic Plagiotremus rhynorhinchos, Cheney et al., 2008). In the Risor clade, however, the four cleaning species, as well as their non-cleaning clade-mates are variously banded or spotted rather than stripped. Taken together, these studies suggest that a cleaner ‘guild coloration’ exists but it may not be universal. Much work remains to understand the exceptions to the stripe rule.