Cleaner Fish Use Tactile Dancing Behavior As a Preconflict Management Strategy

View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by Elsevier - Publisher Connector Current Biology, Vol. 14, 1080–1083, June 22, 2004, 2004 Elsevier Ltd. All rights reserved. DOI 10.1016/j.cub.2004.05.048 Cleaner Fish Use Tactile Dancing Behavior as a Preconflict Management Strategy

Alexandra S. Grutter load. Field observations suggest that cleaner fish use School of Life Sciences preconflict management behavior, which involves the University of Queensland use of tactile stimulation, where they rub clients with Brisbane, Queensland 4072 their body or fins, as a strategy to reduce conflict that Australia might arise from a client’s decision [11]. Because of the severe consequences (death) of conflicts with predators, it was also proposed that cleaner fish should pro- Summary vide tactile stimulation unconditionally to predators as compared to nonpredators, who receive tactile stimula- The most commonly asked question about cooperation only in certain situations [11]. This has never been tive interactions is how they are maintained when tested experimentally. Most other antipredator behav- cheating is theoretically more profitable [1]. In clean- iors, such as predator inspection, harassment, and mob- ing interactions, where cleaners remove parasites bing [12], however, are costly. Therefore, identifying po- from apparently cooperating clients, the classical tentially dangerous individuals would be advantageous question asked is why cleaner fish can clean piscivo- to cleaners. Another controversial form of interspecific rous client fish without being eaten, a problem Trivers communication in cleaning interactions involves cleaner [2] used to explain reciprocal altruism. Trivers [2] sug- fish “dancing,” where cleaner fish swim in an oscillating gested that predators refrain from eating cleaners only fashion [13]. Although dancing is assumed to play a when the repeated removal of parasites by a particular role in advertising cleaning services [6, 14, 15], others cleaner results in a greater benefit than eating the propose that it may reduce aggression [13–16], yet this cleaner. Although several theoretical models have ex- has not been tested. One of the most striking behaviors amined cheating behavior in clients [3, 4], no empirical of clients during cleaning interactions is the highly ste- tests have been done (but see Darcy [5]). It has been reotyped posturing [17]. Posing appears to signal a cli- observed that cleaners are susceptible to predation ent’s desire to be cleaned because it increases the cli- [6, 7]. Thus, cleaners should have evolved strategies ent’s likelihood of being cleaned [17]. However, it has to avoid conflict or being eaten. In primates, conflicts been proposed that posing might also serve as an ap- are often resolved with conflict or preconflict manage- peasement behavior by potentially dangerous clients ment behavior [8]. Here, I show that cleaner fish tacti- toward cleaners [17]. Ectoparasite load is known to af- cally stimulate clients while swimming in an oscillating fect a client’s desire to seek a cleaner [18], but whether it “dancing” manner (tactile dancing) more when ex- influences its posing behavior is unknown. To determine posed to hungry piscivorous clients than satiated whether cleaner and client behaviors differed according ones, regardless of the client’s parasite load. Tactile to client hunger (satiated or hungry) and ectoparasite dancing thus may function as a preconflict manage- loads (few or many parasites), I tested the effect of these ment strategy that enables cleaner fish to avoid con- factors, and the interaction between the two, on the flict with potentially “dangerous” clients. behaviors of the cleaner fish Labroides dimidiatus and the large piscivorous client, the coral trout Plectropo- mus leopardus. Coral trout eat a wide range of fishes, Results and Discussion including fishes in the same family as L. dimidiatus [19]. Including a non-cleaner fish, Halichoeres melanurus, The iterated prisoner’s dilemma has long been used to helped to determine whether another fish perceived the explain the evolution of cooperation between unrelated client as a potential threat. Tactile stimulation by clean- individuals, although some of its limitations have been ers, with and without cleaner dancing, and client posing illustrated with the cleaner fish mutualism [9]. Recently, rates were measured. biological market theory, in which traders exchange I found that the frequency of tactile dancing was goods, services, or both, was proposed as an alternative higher in cleaner fish exposed to hungry piscivorous for understanding cooperation in many systems, includ- client fish than in cleaner fish exposed to satiated cli- ing cleaning symbioses [10]. Client ectoparasites and ents, regardless of client parasite level, and this was cleaning services are the main goods traded in the consistent in both experiments (Figure 1). Because con- cleaner fish market [10]. Thus, the probability of aggres- flicts with hungry clients are likely to be riskier, tactile sion directed at a cleaner by a client is likely to increase dancing may function as a preconflict management as a client’s ectoparasite load decreases and its need strategy [8, 11], conditional on client characteristics. The for the cleaner’s services subsequently declines. Also, results also support the hypothesis that cleaners exploit as client hunger levels increase, the benefits of eating the sensory system of clients with tactile stimulation a cleaner for a client should also increase. Thus, I hy- [20]. Parasite load, however, did not affect the rate of pothesized that cleaners should respond differently to tactile dancing toward hungry clients, as was predicted. hungry piscivorous clients compared with satiated ones This suggests that, for hungry client fish, the potential and that this should vary according to client parasite benefits of eating a cleaner fish relative to the benefits of parasite removal are higher. Tactile dancing ap- Correspondence: [email protected] pears reminiscent of predator inspection behavior Tactile Dancing in a Cleaner Fish Reduces Conflict 1081

Figure 2. The Rate of Tactile Stimulation by Cleaner Fish without Figure 1. The Rate of Tactile Dancing in Cleaner Fish Cleaner Fish Dancing The number of times a cleaner fish engaged in tactile dancing while The number of times a cleaner fish engaged in tactile stimulation exposed to a piscivorous client fish for 90 min of observation. (A) of a piscivorous client fish without cleaner fish dancing per 90 min Experiment 1. (B) Experiment 2. Clients were satiated or hungry of observation. (A) Experiment 1. (B) Experiment 2. Clients were and had many or few parasites. There was a significant difference satiated or hungry and had many or few parasites. There was a ϭ ϭ between satiated and hungry fish (F1,91 8.26, p 0.005) and significant difference between fish with many and fish with few between the two experiments (F ϭ 5.77, p ϭ 0.019); all other 1,91 parasites (F1,91 ϭ 4.64, p ϭ 0.034) and between the two experiments factors and interactions were not significant (p Ͼ 0.05). Data are (F1,91 ϭ 22.98, p ϭ 0.0001); all other factors and interactions were least square means, and bars show standard errors. Black circles not significant (p Ͼ 0.05). Data are least square means, and bars indicate many parasites; white circles indicate few parasites. show standard errors. Black circles indicate many parasites; white circles indicate few parasites. where the most vulnerable prey approach predators more closely [12]. 0.685; satiated: 12, 1.6 [0/16.4]; hungry: 11, 2.1 [0/13.6]). How Labroides dimidiatus identified the hungry client In experiment 2, the frequency of cleaner fish dancing fish is unclear; no effect of hunger on client behavior without tactile stimulation also did not vary with hunger was detected. It has been proposed that cleaner fish level when cleaners were exposed to client fish with few may recognize predators that are not intent on feeding parasites (S ϭ 169, Z ϭ 1.19, P ϭ 0.233, satiated: 12, 0 or that predators may posture and advertise this state [0/5.2], hungry: 12, 0.5 [0/24.6]) or many parasites (S ϭ [21], but the latter was not apparent. 118.5, Z ϭ 0.25, P ϭ 0.801; satiated: 12, 0 [0/2.0]; hungry: The fishes’ use of tactile stimulation as a preconflict 10, 0 [0/3.9]). In experiment 1, the frequency of cleaner management strategy while dancing differed in appear- fish dancing without tactile stimulation did not vary with ance, but apparently not function, from a previously parasite load when cleaners were exposed to satiated proposed strategy in which tactile stimulation is directed client fish (S ϭ 144, Z ϭ 0.72, P ϭ 0.473; few parasites: at the dorsal fin area only (not observed in this study); 11, 3.0 [0/31.2]; many parasites: 12, 1.6 [0/16.4]) or hun- this previous study did not take possible dancing into gry client fish (S ϭ 153.5, Z ϭ 1.32, P ϭ 0.187; few account [11]. Potentially, cleaner fish need to use differ- parasites: 12, 1.0 [0/11.1]; many parasites: 11, 2.1 [0/13.6]). ent conflict management behaviors with different client In experiment 2, the frequency of cleaner fish dancing fish. The fact that most (88%) tactile dances were di- without tactile stimulation also did not vary with hunger rected at the body of the client with only 8% and 5% level when cleaners were exposed to satiated client fish directed at the tail and head, respectively, with none (S ϭ 143, Z ϭϪ0.47, P ϭ 0.640; few parasites: 12, 0 directed at the mouth and gills, suggests that cleaner [0/5.2]; many parasites: 12, 0 [0/2.0]) or hungry client fish behaved cautiously during tactile dancing by fish (S ϭ 97, Z ϭϪ1.29, P ϭ 0.196; few parasites: 12, 0.5 avoiding potentially dangerous sites on the client body. [0/24.6]; many parasites: 10, 0 [0/3.9]). Clearly, dancing The idea that cleaner fish were indeed vulnerable to alone is not used as a potential form of communication predation from the client fish was supported by the fact but rather is combined with tactile stimulation (tactile that three cleaner fish were eaten within 8 min of the dancing). Furthermore, tactile stimulation without danc- start of the trials. Client fish also ate dead cleaner fish ing was directed at all sites, with 32%, 16%, 23%, 32%, offered to them at the end of the study. Control fish and 19% directed at the body, tail, head, mouth, and almost always hid from client fish, suggesting that they gills, respectively. This agrees with observations that usually felt threatened regardless of client fish parasite client parasites were found on all sites of the fish (A.S.G. load or hunger level. Thus, their behavior does not repre- and C. Fury, unpublished data). sent predation risk to cleaner fish. The frequency of tactile stimulation without dancing The frequency of cleaner fish dancing without tactile was higher on fish with many parasites than those with stimulation was low compared to the tactile dancing few, regardless of hunger level, and this was also consis- rate and was not affected by hunger level and parasite tent in both experiments (Figure 2). The fact that foraging load. In experiment 1, the frequency of cleaner fish danc- tactile stimulation was higher on fish with many para- ing without tactile stimulation did not vary with hunger sites suggests it is related to feeding. level when cleaners were exposed to client fish with few Client posing frequency, a measure of a client’s desire parasites (Wilcoxon Rank Sum test S ϭ 152.5, Z ϭ 1.26, to be cleaned [17], was higher for fish with many para- P ϭ 0.209, n, median [10%/90% quartile per 90 min sites, regardless of hunger level, and this also was con- observation]; satiated: 11, 3.0 [0/31.2]; hungry: 12, 1.0 sistent in both experiments (Figure 3). This is the first [0/11.1]) or many parasites (S ϭ 139, Z ϭ 0.41, P ϭ study to show, experimentally, that client posing rates Current Biology 1082

144 hr). One fish from each of the pools was tested each day. Clients were placed in test aquaria (37 cm high ϫ 90 cm long ϫ 36 cm wide) the evening before trials. Two upright, covered, clear Plexiglas tubes (10 cm ϫ 50 cm) were each placed in two aquaria on the night before the trials, then 90 min prior to trials for the remaining trials, so that clients could acclimate to the tubes. Test aquaria had shelters (2 cm ϫ 15 cm opaque pipes) for the cleaner and non- cleaner fish. Recording began when a cleaner and a non-cleaner fish were placed in the tube to acclimate to the client and aquarium. After 5 min, fish were released from the tube. Two 90 min trials were recorded at 0700, 0900, 1100, and 1300 hr with two video cameras (Sony Hi8 TRV89E and TRV87E). Both cleaner fish and client fish were diurnally active [26, 27]. The time of day was random but Figure 3. The Rate of Client Posing for Cleaner Fish balanced across treatments with each treatment-by-time-of-day combination tested three times. Thus, all treatments were tested The number of times piscivorous client fish posed for cleaner fish each day and at all four times of the day. At 1700 hr, tested clients per 90 min of observation. (A) Experiment 1. (B) Experiment 2. Clients were offered a chopped pilchard fish so that their hunger levels were satiated or hungry and had many or few parasites. There was could be estimated. The proportion of clients that fed on pilchards a significant difference between fish with many and fish with few was higher for hungry (88, 55%) than for satiated fish (50, 11%,) in parasites (F ϭ 4.33, p ϭ 0.040) and between the two experiments 1,91 experiments 1 and 2, respectively. Tested clients were returned to (F ϭ 4.76, p ϭ 0.032); all other factors and interactions were not 1,91 the same pool. For identification, they were anaesthetized with 2 significant (p Ͼ 0.05). Data are least square means, and bars show g /25 L of MS222 (Sigma) in seawater for 7 min and measured, and standard errors. Black circles indicate many parasites; white circles individual markings were recorded. Fish were also tagged by a piece indicate few parasites. of flexible plastic (1.5 x 1.5 cm) being threaded with surgical thread into the dorsal fin, but some of the tags were lost during the course are related to client ectoparasite load. Whether posing of the study. Separate anaesthetic baths were used for fish with was initiated by the cleaner fish or the client fish [17], few and many parasites. The day after all trials were completed, however, could not be determined. Thus, it is unclear fish were fed with pilchards. Afterward, five dead cleaner fish were offered, one at a time, to the client fish in the pools (three pools whether posing was a response to ectoparasite irritation with previously hungry fish and two pools with satiated fish). All or to the tactile stimulation without dancing, which was cleaner fish were eaten. Mean (s.e.) client total length was 44.0 (0.8) higher on fish with many parasites. The finding that client cm and did not vary among treatments (ANOVA p Ͼ 0.05). Emptying posing did not vary with hunger level, however, suggests all test aquaria after trials and soaking them in freshwater for 30 that it does not serve as a form of appeasement behavior min killed any detached parasites. The experiment was repeated by potentially dangerous fish toward cleaner fish [17]. with randomly selected fish. Behaviors were analyzed with Observer Video Pro 4 (Noldus). Tactile dancing as interspecific signaling to manage conflicts, in addition to manipulating, reconciling [11], Behaviors Recorded using altruism [22], and recognizing clients [23], adds Behaviors were defined as follows: tactile dancing involved the to the increasing list of the cleaners’ abilities to deal cleaner oscillating its posterior body in a dorsal-ventral manner [6, with complex social environments, abilities usually the 14] within 15 cm of the client and mainly in one place while often focus of cognitive studies of primates [24]. contacting the client with its body. Tactile stimulation without dancing involved the cleaner contacting the client’s body with its pelvic Experimental Procedures fins and often also nibbling on the client’s body. Posing involved the client opening the gills and/or mouth [17]. Non-cleaner-fish hiding Fish Collection involved sheltering in a pipe or in a corner. Clients (n ϭ 56) were collected with hook and line at Lizard Island, Great Barrier Reef, 6–26 days before trials. Fish were initially held Statistical Analyses in four pools (3 ϫ 1 m) and fed pilchards every other day. Cleaner Data were analyzed with a repeated-measures ANOVA with hunger fish (n ϭ 48), a mean (s.e.) 7.4 (0.1) cm in total length (TL), and non- level and parasite load as the between-subjects factors, experiment cleaner fish (n ϭ 48), 5.85 (0.1) cm TL, were collected [25] at the as the within-subjects/repeated factor, and client pool as the factor same time and held in aquaria. nested within hunger level and parasite load. Because the effects of pools were all highly not significant (p Ͼ 0.25), they were omitted Fish Handling and Experimental Treatments from the final analyses to increase power according to Underwood’s Up to 10 days before trials, some fish were fed daily, whereas others [28] rules. One trial, in which the client appeared to be agitated, were not; fish that received meals were considered to be “satiated,” was omitted. Three clients ate a cleaner and so trials were omitted and the others were “hungry.” Two to 8 days before trials, fish from the analyses. Three non-cleaner fish spent no time hiding in a randomly allocated to the treatment “few parasites” were treated shelter and were omitted from the analysis to satisfy the assumption for parasites with two baths (in 0.5 g Praziquantal in 3 ml ethanol/ of homogeneity of variance of the analysis of variance. Non-cleaner 50 L seawater for 2.5 hr, ICN Biomedicals Inc., Aurora, then 3 min fish spent 98% of their time in a shelter; this was not affected by in freshwater 24 hr later) and those allocated to “many parasites” experiment, hunger, or parasites (all p Ͼ 0.230). Client fish had 15 were given two control seawater baths. Six clients, plus one for different types of ectoparasites, mostly monogeneans and cope- ectoparasite estimates (not tested), were transferred to each of eight pods (A.S.G. and C. Fury, unpublished data). The mean (s.e.) parasite pools with two pools per treatment combination. The eight untested load for the subsample of fish in experiments 1 and 2 was 2041 fish and the last eight fish tested were examined for parasites after (715) and 1456 (785) for fish with many parasites and 169 (934) and experiments 1 and 2, respectively. Parasites were removed with a 26 (8) for fish with few parasites, respectively. Separate log-linear 5 min freshwater bath and filtered at 62 ␮m (the former were first analyses, for which R 1.6.2 software and the generalized linear killed with an overdose of clove oil, and their gills and fins were model procedure were used, tested whether the probability (pres- examined separately with a microscope). ence/absence of the behavior per observation) of tactile dancing and tactile stimulation without dancing differed among client body Behavioral Observations sites (body, mouth, head, gills, and tail) and with hunger, parasites, Two identical, 6-day-long experiments were conducted 5 days apart and experiment. A full model was initially fitted with interaction terms (n ϭ 12 trials per treatment combination for a total of 96 trials and and no intercept. All interaction terms were eliminated because they Tactile Dancing in a Cleaner Fish Reduces Conflict 1083

were nonsignificant. The probability of these behaviors occurring 20. Losey, G.S. (1987). Cleaning symbiosis. Symbiosis 4, 229–258. at different sites of the client’s body varied with site but did not 21. Hobson, E.S. (1971). Cleaning symbiosis among California in- vary with hunger level, parasite load, or experiment (log-linear analy- shore fishes. Fishery Bulletin 69, 491–523. ses, all p Ͼ 0.05). 22. Bshary, R. (2002). Biting cleaner fish use altruism to deceive image-scoring client reef fish. Proc. R. Soc. B, Biol. Sci. 269. Acknowledgments 23. Tebbich, S., Bshary, R., and Grutter, A.S. (2002). Cleaner fish Labroides dimidiatus recognise familiar clients. Anim. Cogn. 5, I thank C. Fury, F. Grutter, M. Johnson, E. Hunley, and the Lizard 139–145. Island Research Station staff for field and laboratory assistance, N. 24. Bshary, R., Wickler, W., and Fricke, H. (2002). Fish cognition: a Khan for statistical advice, S. Baldwin, J. Becker, M. Blows, R. primate’s eye view. Anim. Cogn. 5, 1–13. Bshary, C. Jones, H. McCallum, and K. Warburton for comments 25. Grutter, A.S. (1997). Spatio-temporal variation and feeding se- on previous versions of the manuscript, and the Australian Research lectivity in the diet of the cleaner fish Labroides dimidiatus. Council for funding. Copeia 1997, 346-355. 26. Grutter, A.S. (1996). Parasite removal rates by the cleaner Received: March 8, 2004 wrasse Labroides dimidiatus. Mar. Ecol. Prog. Ser. 130, 61–70. Revised: April 20, 2004 27. Zeller, D. Home range and activity patterns of the coral trout Accepted: April 20, 2004 Plectropomus leopardus (Serranidae). Marine Ecology Progress Published: June 22, 2004 Series 154, 65–77. 28. Underwood, A.J. (1997). Experiments in ecology: their logical References design and interpretation using analysis of variance (New York: Cambridge University Press). 1. Dugatkin, L.A. (1997). Cooperation Among Animals: An Evolu- tionary Perspective (New York: Oxford University Press). 2. Trivers, R.L. (1971). The evolution of reciprocal altruism. Q. Rev. Biol. 46, 35–57. 3. Poulin, R., and Vickery, W.L. (1995). Cleaning symbiosis as an evolutionary game: to cheat or not to cheat? J. Theor. Biol. 175, 63–70. 4. Freckleton, R.P., and Coˆ te´ , I.M. (2003). Honesty and cheating in cleaning symbioses: evolutionarily stable stragies defined by variable pay-offs. Proc. R. Soc. Lond. B, Biol. Sci. 270, 299–305. 5. Darcy, G.H., Maisel, E., and Ogden, J.C. (1974). Cleaning prefer- ences of the gobies Gobiosoma evelynae and G. prochilos and the juvenile wrasse Thalassoma bifasciatum. Copeia 2, 375–379. 6. Wickler, W. (1968). The origin of the cleaner mimic. In Mimicry in plants and animals, W. Wickler, ed. (London: Weidenfeld and Nicolson), pp. 157–176. 7. Lobel, P.S. (1976). Predation on a cleanerfish (Labroides)bya hawkfish (Cirrhites). Copeia 1976, 384–385. 8. Aureli, F., and Waal, F.B.M. (2000). Natural Conflict Resolution (Berkeley: University of California Press). 9. Hammerstein, P., and Hoekstra, R.F. (1995). Mutualism on the move. Nature 376, 121–122. 10. Bshary, R. (2001). The cleaner fish market. In Economics in Nature: Social Dilemmas, Mate Choice and Biological Markets, R. Noe, J.A.R.A.M. van Hooff and P. Hammerstein, eds. (Cam- bridge: Cambridge University Press), pp. 146–172. 11. Bshary, R., and Wu¨ rth, M. (2001). Cleaner fish Labroides dimidiatus manipulate client reef fish by providing tactile stimulation. Proc. R. Soc. Lond. B, Biol. Sci. 268, 1495–1501. 12. Dugatkin, L.A., and Godin, J.-G. (1992). Prey approaching predators: a cost-benefit perspective. Annales Zoologici Fennici 29, 233–252. 13. Potts, G.W. (1973). The ethology of Labroides dimidiatus (Cuv. & Val.) (Labridae, Pisces) on aldabra. Anim. Behav. 21, 250–291. 14. Youngbluth, M.J. (1968). Aspects of the ecology and ethology of the cleaning fish, Labroides phthirophagus Randall. Z. Tierp- sychol. 25, 915–932. 15. Sazima, I., Moura, R.L., and Gasparini, J.L. (1998). The wrasse Halichoeres cyanocephalus (Labridae) as a specialized cleaner fish. Bull. Mar. Sci. 63, 605–610. 16. Losey, G.S. (1971). Communication between fishes in cleaning symbiosis (Baltimore: University Park Press). 17. Coˆ te´ , I.M., Arnal, C., and Reynolds, J.D. (1998). Variation in posing behaviour among fish species visiting cleaning stations. J. Fish Biol. 53, 256–266. 18. Grutter, A.S. (2001). Parasite infection rather than tactile stimulation is the proximate cause of cleaning behaviour in reef fish. Proc. R. Soc. Lond. B, Biol. Sci. 268, 1361–1365. 19. St John, J., Russ, G.R., Brown, I.W., and Squire, L.C. (2001). The diet of the large coral reef serranid Plectropomus leopardus in two fishing zones on the Great Barrier Reef. Australia. Fishery Bulletin Seattle 99, 180–192.