PAPERS
The Ethology
RESEARCH Meaning of Jolts by Fish Clients of Cleaning Gobies
Marta C. Soares*, Redouan Bshary, Sonia C. Cardoso* & Isabelle M. Cote
* School of Biological Sciences, University of East Anglia, Norwich, UK Institut de Zoologie, Eco-Ethologie, Universite de Neuchatel, Neuchatel, Switzerland Department of Biological Sciences, Simon Fraser University, Burnaby, BC, Canada
Correspondence Isabelle M. Cote, Department of Biological Sciences, Simon Fraser University, Burnaby, BC V5A 156, Canada. E-mail: [e-mail]
Abstract Cooperative interactions offer the inherent possibility of cheating by each of the interacting partners. A key challenge to behavioural observers is to recognize these conicts, and nd means to measure reliably cheating in natural interactions. Cleanersh Labroides dimidiatus cheat by taking scales and mucus from their sh clients and such dishonest cleaning has been previously recognized in the form of whole-body jolts by clients in response to cleaner mouth contact. In this study, we test whether jolts may be a general client response to cheating by cleaners. We experimentally varied the ectoparasite loads of yellowtail damselsh (Microspathodon chrysurus), a common client of the cleaning goby Elacantinus evelynae, and compared the rates of jolts on parasitized and deparasitized clients. As predicted if jolts represent cleaner cheating, deparasitized clients jolted more often than parasitized clients, and overall jolt rates increased over time as client parasite load was presumably reduced by cleaning activity. Yellowtail damselsh in the wild jolted signicantly less frequently than those in captivity, which is consistent with a loss of ectoparasites during capture. Our results suggest that jolts by clients of cleaning gobies are not related to the removal of ectoparasites. Client jolts may therefore be a generally accurate measure of cheating by cleanersh.
Received: April 28, 2007 Initial acceptance: August 14, 2007 Final acceptance: September 17, 2007 (K. Reinhold)
doi: 10.1111/j.1439-0310.2007.01471.x
Introduction The evolution of cooperation has long appeared paradoxical because of the so-called cheating problem (Noe 2006). In any cooperative interaction, there is strong selection for partners derive benets from the interaction without incurring costs. Cheating therefore occurs when either of the cooperative partners deceives the other by providing a dishonest service, which can range from a subtle reduction in service value to not delivering the expected commodity at all (Dugatkin 1997, 2002; Noe 2001, 2006; Bshary & Noe 2003; Sachs et al. 2004). One biological system in which conicts over cheating have been demonstrated is the interactions between cleanersh and their sh clients. Clients visit cleaning stations, i.e. small territories held by
Ethology 114 (2008) 209214 2008 The Authors Journal compilation 2008 Blackwell Verlag, Berlin
cleanersh, to have their parasites and dead or infected tissues removed (reviewed by Losey et al. 1999; Cote 2000) but cleaners potentially have access to other commodities, such as mucus and scales, on their clients bodies. Conicts between cleaners and clients over what cleaners should feed on are evident (Bshary & Wurth 2001). In an exper iment in which anaesthetized parasite-free surgeonsh clients were presented to bluestreak cleaner wrasses Labroides dimidiatus, the commonest IndoPacic cleanersh, most cleaners scraped the body surface of their clients rather than feed on prawns which were provided as an alternative food source (Bshary & Grutter 2002). In addition, in food preference trials in the laboratory, L. dimidiatus fed on mucus rather than on gnathiid ectoparasites (Grutter & Bshary 2003), and preferred energy-rich mucus
209
Meaning of Jolts by Client Fish
M. C. Soares et al.
(e.g. of parrotsh) over energy-poor mucus (e.g. snapper) (Grutter & Bshary 2004). Cleaners therefore appear to prefer mucus over ectoparasites, whereas clients would prefer them to forage only on the latter. A key challenge for empirical biologists is to nd means to reliably measure cheating in natural interactions. For marine cleaning mutualism involving the cleaner wrasse L. dimidiatus, client jolts wholebody shudders that appear to be painful in response to cleaner sh mouth contact appear to be a good correlate of cheating by cleaners, i.e. the eating of mucus or scales (Bshary 2001). The fact that experimentally deparasitized clients of L. dimidiatus jolt more often than highly parasitized clients supports this interpretation (Bshary & Grutter 2002). Stomach content analyses of cleanersh species other than L. dimidiatus have shown that all cleaner species feed not only on ectoparasites but also mucus and scales to varying extents (Randall 1958; Gorlick 1984; Cheney & Cote 2005; M. C. Soares, pers. obs.). It therefore becomes of interest to determine whether incidences of cheating by cleanersh other than L. dimidiatus can also be detected through client behaviour. Cleaning gobies (Elacatinus spp.) are the most ubiquitous cleaners in the Caribbean region. The clients of cleaning gobies also jolt during cleaning interactions (e.g. Cote & Molloy 2003; Soares et al. 2007), but the signicance of this behaviour has not yet been determined. In a eld study, Soares et al. (2007) failed to nd a relationship between client jolting rate and client ectoparasite loads. Moreover, predatory clients, which should be able to enforce honesty in cleaners through risk of predation, did not jolt less frequently during cleaning interactions than harmless clients. Both results suggest that jolts by clients of cleaning gobies could be unrelated to the delivery of cheating bites. To investigate the meaning of jolts by clients of cleaning gobies, we examined the interactions between cleaning gobies and both parasitized and experimentally deparasitized yellowtail damselsh (Microspathodon chrysurus) under captive conditions. We tested three main predictions. First, if client jolts are indeed linked to cleaner cheating, then deparasitized clients should jolt more than parasitized clients. Second, differences in jolt rates between parasitized and unparasitized clients should diminish as cleaners clean, thus reducing the ectoparasite load differences between both groups. Third, client jolt rate should increase over time as client parasite load is reduced by cleaning activity. Finally, to verify the extent to
210
which behaviours observed in captivity reect occurrences in the wild, we compared our experimentally controlled data to eld observations of yellowtail damselsh behaviour during interactions with cleaning gobies. Methods
Study Species and Fish Collection
The study was carried out between Apr. and Nov. 2005 at the Bellairs Research Institute, in Barbados, West Indies. We focussed on the sharknose goby (Elacatinus evelynae), one of two species of cleaning gobies present on Barbadian fringing reefs. These cleaning gobies are small (1.23.5 cm total length) and found solely on coral (usually Siderastrea spp. and Montastrea spp.), where individuals or pairs of sh establish small territories that act as cleaning stations. The bulk of the sharknose goby diet consists of ectoparasites gleaned from visiting client species (Whiteman & Cote 2002). As a focal client, we selected yellowtail damselsh because they are among the most frequent visitors to cleaning stations and they usually have a relatively higher ectoparasite loads than other client species at this location (Sikkel et al. 2000, 2005; Soares et al. 2007).
Behavioural Observations in the Wild
To provide a comparison for experimental trials with captive sh, in situ observations of interactions between yellowtail damselsh and sharknose cleaning gobies were carried out while diving or snorkelling on eight fringing reefs near the Bellairs Research Institute. Forty-three cleaning stations were selected haphazardly across the reefs, with four to eight stations per reef. Each cleaning station was observed once for 30 min, between 10.00 and 17.00 hours. In fact, most of our eld observations were carried out in the early afternoon, similarly to our laboratory experimentation trials, to make comparisons more meaningful. Observations were made from a distance of 23 m and began after a 2- to 5min delay to allow the sh to become accustomed to the presence of the observer. During each observation period, we recorded on plastic slates client total length (estimated visually to the nearest cm), the duration (in s) of inspection and the number of jolts, i.e. apparently painful reactions to a cleanersh bite, for each visiting yellowtail damselsh. Because yellowtail damselsh were unmarked, we assumed that all visits during an observation period were made by
Ethology 114 (2008) 209214 2008 The Authors Journal compilation 2008 Blackwell Verlag, Berlin
M. C. Soares et al.
Meaning of Jolts by Client Fish
different individuals. Focal observations carried out on 23 individual yellowtail damselsh on three of the study reefs revealed that fewer than half (10 of 23 individuals) repeatedly visited the same cleaner in a 30-min period and successive visits to cleaners were made to different cleaning stations (M. C. Soares, pers. obs.), suggesting a limited extent of double counting in our main observations.
Experimental Design
We caught cleaning gobies and yellowtail damselsh from three of the reefs on which the in situ behavioural observations were carried out. Twelve cleaning gobies were collected 34 wk prior to the beginning of experiments to acclimatize to laboratory conditions. A mixture of clove oil (a natural anaesthetic), ethanol and water was sprayed over each individual cleaning goby in order to induce a temporary reduction in activity. Gobies were then easily caught with hand nets and placed individually in sealed plastic bags lled with seawater. Once in the laboratory, gobies were measured (total length to the nearest mm) and placed in individual glass aquaria (61 cm long 38 cm wide 46 cm high) with running seawater. Each aquarium had 12 cm of sand and gravel at the bottom. Several pieces of dead coral (1040 cm diameter) were placed in a mound at one end of each aquarium to provide shelter and a vantage point. Aquaria were separated by opaque partitions. During the acclimation period, cleaning gobies, which ranged in total length from 2.2 to 3.3 cm, were fed daily with brine shrimp Artemia spp. We collected 24 yellowtail damselsh from three of the study reefs between 07.00 and 12.00 hours. Individuals were targeted haphazardly and herded into a barrier net placed near the edge of their territories. The sh were then caught with a hand net, rapidly placed individually into hermetically sealed plastic bags lled with seawater, and immediately brought to the laboratory. Damselsh were randomly assigned to one of two groups: ectoparasite removal or control procedure. The ectoparasite removal method was similar to that of Sikkel et al. (2004), and has been shown to remove 98100% of all ectoparasites present on coral reef sh (Grutter 1995a,b). Damselsh were placed into individual containers with variable amounts of seawater and two to three drops of clove oil. The damselsh were then transferred to a freshwater bath for 10 min, during which their body surface was gently brushed with a soft-bristle paintbrush. Finally, sh were
Ethology 114 (2008) 209214 2008 The Authors Journal compilation 2008 Blackwell Verlag, Berlin
placed in seawater-lled containers to recover for a minimum of 10 min. Full recovery was deemed to have occurred when the sh were swimming actively. The control procedure was identical, except that damselsh were placed in a seawater bath rather than a freshwater bath, which is less effective at removing ectoparasites (see Results). Capture, parasite removal (or control procedure) and behavioural testing occurred on the same day. After each behavioural trial, damselsh were measured to the nearest mm (total length) and released at their capture location. All uids from the freshwater and saltwater baths were ltered separately for each sh. These samples were later examined under a binocular microscope. Ectoparasites were counted and identied to family, focussing on the families Bomolochidae, Caligidae, Ergasilidae, Gnathiidae and Hatschekiidae.
Experimental Behavioural Observations
The yellowtail damselsh used for experimentation ranged in standard length from 9.2 to 12.3 cm. Of these, 12 had been deparasitized (i.e. subjected to a freshwater bath) and 12 were parasitized (i.e. subjected to a saltwater bath). There was no size difference between the two groups (independent t-test: t20 = )0.69, p = 0.50). Each damselsh was used in a single trial. Each of the 12 cleaning gobies was used twice: once with a parasitized client and once with a deparasitized client, with testing occurring on different days. All trials were carried out between 12.00 and 15.00 hours. Each damselsh client was placed in a test aquarium with one cleaning goby for 30 min. Cleaning interactions were videotaped with a Sony Handycam digital videocamera (model DCR-TRV10E), which was placed 60 cm from the front wall of the aquarium, allowing all interactions to be clearly observed. The videotapes were played back on a 35-cm television. We recorded the duration of each inspection bout and the number of jolts by yellowtail damselsh.
Statistical Analysis
Because the ectoparasite loads of experimental yellowtail damselsh were low (see Results), we surmised that any difference in parasite numbers between parasitized and deparasitized clients would be rapidly reduced by cleaning activity. We therefore initially considered only the rst interaction between cleaners and damselsh under experimental
211
Meaning of Jolts by Client Fish
M. C. Soares et al.
conditions. Client jolt rates (always expressed as the number of jolts 100 per s of inspection time) were compared between parasitized and deparasitized clients using paired tests centred on individual cleaning gobies. We also examined how jolt rates changed over time by considering the rst three interactions between cleaners and both parasitized and deparasitized clients. These were analysed using two-way repeated measures analyses of variance (anova) with time as a within-subject factor and parasite load as a between-subjects factor. Finally, to compare the jolt rates of captive and wild sh, we obtained jolt rates for individuals across all observed cleanerclient interactions and carried out pairwise comparisons between wild sh and each of the two categories of captive sh (parasitized and unparasitized) using unpaired t-tests. Results
Effect of Ectoparasite Removal and Control Procedures
Effect of Client Ectoparasite Load on Client Jolt Rate
Parasitized damselsh jolted signicantly less often during the rst cleaning interaction than deparasitized damselsh (paired t-test: t11 = )2.37, p = 0.04; Fig. 2). When considering the rst three interactions, jolt rate increased signicantly over time (two-way rm-anova: F2,44 = 3.98, p = 0.03), but did not vary between deparasitized and parasitized clients (twoway rm-anova: F1,22 = 1.24, p = 0.28). There was no interaction between time and client parasite load (two-way rm-anova: F2,44 = 0.42, p = 0.66). Yellowtail damselsh in the wild jolted signicantly less frequently than captive damselsh. Jolt rates in the wild were 79% lower than in captive, deparasitized damselsh (independent-samples t-test: t44.53 = 4.34, p < 0.0001), and 66% lower than in captive, parasitized sh (independent-samples t-test: t45.36 = 2.36, p = 0.02; Fig. 2). Discussion Our results provide the rst experimental evidence that jolts by clients of cleaning gobies are unlikely to be related to ectoparasite removal. If jolts by clients were responses to the physical removal of ectoparasites by cleaning gobies, then clients with higher parasite loads should have exhibited more frequent jolt reactions during cleaning inspection. Instead, we found that deparasitized clients jolted signicantly more than their parasitized counterparts. This result supports our rst prediction and corroborates previous work carried out with the cleaner wrasse L. dimidiatus (Bshary & Grutter 2002). Neither we nor previous authors have unambiguously shown that jolt-inducing bites are dishonest, i.e. that they result in the removal of items such as scales and mucus, which may be costly for clients to replace and or make clients more vulnerable to disease. However, given the frequency with which such items are found in the gut contents of cleanersh (Randall 1958; Gorlick 1984; Grutter 1997; Cheney & Cote 2005, M. C. Soares, pers. obs.), it seems parsimonious to conclude that cheating by cleanersh is prevalent and that client jolts are a reection of this dishonest behaviour. We had also predicted that as cleaning gobies clean, and thereby reduce the ectoparasite load differences between parasitized and unparasitized clients, differences in jolt rates between both groups should diminish, and overall jolt rate should increase. Only the latter prediction was supported. The signicant anova interaction between time and
Ethology 114 (2008) 209214 2008 The Authors Journal compilation 2008 Blackwell Verlag, Berlin
Signicantly more ectoparasites were removed from yellowtail damselsh in the deparasitized group (i.e. subjected to a freshwater bath) than from those in the parasitized group (i.e. subjected to a seawater bath) (independent t-test: t22 = 2.61, p = 0.02; Fig. 1). Assuming that sh in both groups had similar numbers of ectoparasites upon capture, damselsh in the parasitized group entered the behavioural trials with more ectoparasites than those in the deparasitized group.
Number of ectoparasites removed
*
5 4 (12) 3 2 (12) 1 0 Deparasitised Parasitised
Fig. 1: Number of ectoparasites removed from yellowtail damselsh subjected to a freshwater bath (= deparasitized) or to a saltwater control bath (= parasitized) before entering behavioural trials. Means are shown 1 SE. *p < 0.05 with paired t-test. Sample sizes (= number of individuals) are given in parentheses.
212
M. C. Soares et al.
Meaning of Jolts by Client Fish
50
Client jolts/100 s
40 30 20 10
Deparasited (lab.) Parasited (lab.) Wild (12) (12) (12) (12) (12) (12) (85) Second Third Overall
Fig. 2: Number of jolts per 100 s of inspection by captive deparasitized (white bars) and parasitized (black bars) yellowtail damselsh, in their rst three interactions with cleaning gobies, and by wild yellowtail damselsh (grey bar). Means are shown 1 SE. Sample sizes (= number of individuals) are given in parentheses.
0 First
parasite load that would have supported the former was not observed. Experiments similar to ours, but carried out using clients that were heavily parasitized (x 1 SD: 42 33 parasites) on one side of their body and deparasitized on the other, showed that client jolting rate was signicantly higher after 25 h than in the rst 15 min of exposure to L. dimidiatus cleaners (Bshary & Grutter 2002). Moreover, clients jolted at similar rates regardless of which side of the body the cleaner wrasses were foraging, suggesting that over the course of the day, the cleaners had reduced parasite numbers to equal levels on both sides of their clients (Bshary & Grutter 2002). Given that our parasitized clients entered the behavioural trials with relatively few parasites, as expected with the low ectoparasite intensities found in the Caribbean (Arnal et al. 2001; Cheney & Cote 2005; Soares et al. 2007), they probably became deparasitized quickly. This rapid levelling of parasites loads across sh groups may explain the absence of statistical difference in jolt rates between parasitized and deparasitized clients over time. Despite the fact that jolt behaviour has often been used to infer cheating by cleanersh under experimental conditions and in the wild (Bshary & Wurth 2001; Bshary & Grutter 2002; Bshary & Schaffer 2002; Soares et al. 2007), there have been to date no direct comparisons of jolt rates by captive and wild clients of the same species in the same location. Wild yellowtail damselsh jolted signicantly less frequently than both captive deparasitized and parasitized sh, suggesting that wild sh may have more parasites than those held in captivity. This is possible as gnathiids, in particular, are highly mobile ectoparasites (Davies & Johnston 1976; Grutter 1995a), which will readily leave disturbed hosts (Grutter 1994, 1995a). The capture procedure may therefore have reduced ectoparasite loads on the experimental damselsh, prior to further reductions during deparasitization. Given the link established here between clients jolting and cleaning goby cheating in captivity, and
Ethology 114 (2008) 209214 2008 The Authors Journal compilation 2008 Blackwell Verlag, Berlin
given that jolting is frequently observed in the wild, one may predict that the clients of cleaning gobies will have strategies to control cleaner cheating. Such strategies have been documented in L. dimidiatus. For example, clients of L. dimidiatus that have the possibility to choose among several cleaners, end cleaning interactions after a jolt-inducing bite and withhold revisiting the cheating cleaner in favour of alternative cleaners (Bshary 2001; Bshary & Schaffer 2002). By contrast, clients that have access to a single cleaning station, owing to a small territory size, respond to jolt-inducing bites by chasing aggressively the cheating cleaners, which appears to reduce the likelihood of cleaner dishonesty in subsequent encounters (Bshary & Grutter 2002; Bshary & Noe 2003). Partner control, through punishment strategies that enforce honesty, is an essential feature in the maintenance of interspecic mutualistic interactions (Bshary & Grutter 2005). Further studies are needed to determine if and how the clients of cleaning gobies control cheating by their cleaners. Acknowledgements We are grateful to the staff of Bellairs Research Institute for providing facilities and nancing the building of new aquaria specically for our experimentation procedures. We thank Stanton Thomas for support and friendship during the collection...