Pacific Science (1990), vol. 44, no. 4: 401-406 © 1990 by University of Hawaii Press. All rightsreserved Shell Exchange in Hawaiian Hermit Crabs1 BRIAN A. HAZLETT 2 ABSTRACT: Shell exchange behavior of intertidal Hawaiian hermit crabs was observed in the laboratory. Outcomes of 255 shell-related interactions were analyzed to test the predictive powers of two models of resource exchange . In the case ofintraspecific interactions, the negotiations model (which predicts that exchanges should occur only when both crabs will gain in shell value) was superior to an aggression model ofresource exchange . In the case of interspecific interactions, the negotiations model accurately predicted outcome of Calcinus­ Calcinus interactions, but poorly predicted outcome of Calcinus-Clibanarius interactions. HERMIT CRABSNEED GASTROPOD shells for pro­ model. Before the development of the negoti­ tection, and one means of acquiring them is ations idea as an alternative model of resource by exchange of shells with other crabs (Hazlett exchange, an earlier study ofHawaiian species 1966). Shells can be a limiting resource for (Hazlett 1970) emphasized the aggressive/ crabs (Hazlett 1981), and the behavior pat­ competitive nature ofhennit crab interactions. terns associated with the exchange of this re­ source have often been viewed as fights (Hazlett 1967, 1972, Bach et al. 1976, Dowds MATERIALS AND METHODS and Elwood 1983, 1985). Indeed, an earlier paper on Hawaiian hermit crabs (Hazlett Crabs used in the study were collected at a 1970) viewed shell fights as a mechanism of variety of locations around the island of ecological competition. Oahu, Hawaii. The crabs were placed in water The idea that exchanges ofshells occur pri­ tables at the Hawaii Institute of Marine Bio­ marily when both crabs gain in resource value logy, Coconut Island, Oahu. The crabs were has been termed negotiations (Hazlett 1978), not experimentally treated in any way and and this type of behavioral interaction can were interacting with each other in an envi­ result in ecological processes that differ quali­ ronment that was similar to a 0.5 x 1.5 m tide tatively from competition (Hazlett 1987, pool. Observations were made during day­ Vandermeer et al. 1985). These ecological pro­ light hours, using natural lighting, during the cesses can occur both intraspecifically months of January through May of 1988. (Abrams 1982a) and interspecifically (Hazlett Crabs were periodically fed with algae and 1983). detritus on rocks placed in the water table. Because most ofthe results that support the Species composition and the relative propor­ negotiations model of resource exchange were tions of different species changed over the 5 obtained from observations of Caribbean months ofobservation as new specimens were species of hermit crabs, it is of interest to continually added to the water table . Thus no examine the resource exchange patterns of quantitative statements about the relative Pacific species to test the generality of the amounts of inter- versus intraspecific inter­ actions can be made. The observer sat near the water table and continually scanned the water table for social 1 Contribution no. 807 from the Hawai i Institute of Marin e Biology. Manuscript accepted 31 January 1990. behaviors (the results of reproductive inter­ 2 Department of Biology, University of Michigan , Ann actions are reported elsewhere [Hazlett 1989]). Arbor, Michigan 48109. A shell-related interaction was recorded when- 401 402 PACIFIC SCIENCE, Volume 44, October 1990 ever shell rapping was seen. This rapid bring­ gastropod for at least 48 hr. Shell sizes were ing together of shells by an initiating crab chosen to cover the size range possibly used occurs only in the context of shell exchange by the crabs being tested. Each crab had about attempts, and shell exchanges do not occur 201 shells to choose from (its own and 200 without shell rapping. Whenever shell rapping empty ones) since crabs can only occupy one occurred (see Hazlett 1966 for description), shell at a time. After this period, the crabs special attention was paid to that interaction. were removed and the crab and shell para­ Although attention was focused on the shell­ meters were measured as outlined above. related interaction, the rest of the water table From these data, a regression line between was also scanned for additional interactions. crab size and shell size for each particular Up to five interactions at a time could be crab-gastropod species combination was cal­ followed; however, usually only one inter­ culated. These regressions were obtained only action was in progress at a time. It is import­ for the more common crab-shell combina­ ant to remember that these observations were tions. ofcrabs that were not manipulated in any way The data were then analyzed, case by case, other than being collected and placed in a to determine if the outcome of each inter­ water table.:Crabs that have been experimen­ action fit one, both, or neither of two models tally manipulated behave differently in shell­ of resource exchange (Hazlett 1983, 1987). related interactions (Hazlett, unpublished Although the initiating individual was usually data). larger than the noninitiator, either individual Shell-related interactions terminate in one could be the larger crab in a given interaction. of two ways. Either the noninitiator comes In the aggression model, it is predicted that out of its shell and an exchange of shells oc­ a shell exchange will occur only when the ex­ curs (exchange) or the initiator stops rapping change will lead to a better shell for the initi­ behavior and walks away from the shell it was ating crab and the initiator is larger. A parti­ attempting to obtain (no exchange). Once the cular shell could be better in terms of size of nature of the outcome of the interaction was the shell (closer to the desired shell size as determined by the behavior ofthe crabs, both determined by the free-access experiments) or crabs and their shells were removed from the the species of shell, if there are clear inter­ water table and set aside for later measure­ specific preferences. The negotiations model ments. After a period of observations was (Hazlett 1978) predicts that shell exchanges finished, data regarding the interactants were will occur only ifboth participants will obtain recorded. The species, sex, reproductive con ­ a better shell as a result of the exchange. If dition of females (berried or not), and size in only one crab (the initiator) will obtain a bet­ terms of millimeters of cephalothorax length ter resource, no exchange will occur, accord­ were recorded. The species ofgastropod shells ing to the negotiations model. Each shell­ ofthe initiator and noninitiator was recorded, related interaction was scored as to the accu­ and the shells were placed in a 60° C drying racy of each model in predicting the outcome oven for at least 48 hr. The shells were then of the interaction. The outcome of an inter­ weighed , first each shell by itself and then action could be correctly predicted by one, when filled with fine sand of known specific both, or neither model. gravity. The latter measurement allowed sub­ After measurements of its cephalothorax, sequent calculation of the internal volume of each crab was given a new shell and returned the shells. to the intertidal zone , thus limiting its data The desired shell size of several species of input to one interaction. This process was not crab with regard to several species of gastro­ followed for two species, each represented by pods was determined by free-access experi­ just one individual. One of these was a single ments (Hazlett 1970). Thirty to 60 crabs of a specimen of Calcinus haigae and the second a particular species were placed with an excess new species of Calcinus currently being des­ (100-200) ofempty shells ofa given species of cribed by P. A. McLaughlin. Hermit Crab Shell Exchange-c-Hxztrrrr 403 RESULTS either initiator or noninitiator were observed (Table 2). There were two exceptions to this For five crab species (Clibanarius zebra , generality. Individuals of Clibanarius zebra Calcinus laevimanus, Calcinus latens, Calcinus rarely initiated interactions with an individual seuratic, and Calcinus elegans) the relation­ of any Calcinus species. A few such inter­ ship between crab size and desired shell vol­ actions were observed but they were indeed ume was determined for Trochus intextus. not common and none involved individuals of Interestingly, the slope ofthe crab size-desired C. laevimanus. All four Clibanarius-initiated shell size relationship was essentially the same interspecific interactions resulted in no shell for all five species ofcrab (the crab species by exchange. Second, individuals of Calcinus crab size interaction term from an analysis of seurati were rarely the noninitiator in inter­ variance analyzing variation in shell volume specific interactions. This may be the result chosen with crab size as a covariate; F = 0.667, of the rapid locomotory capabilities of C. P = 0.616). However, both crab size (F = seurati, which allow it to run away from a 307; df = 4, 1; P < 0.001) and crab species potential shell-related interaction initiated (F = 3.2, P = 0.003) significantly affected by slower crabs (i.e., individuals of other shell volume chosen. This indicates that al­ species). though the slopes are not different for the The outcomes of all of the interspecific in­ different crab species, the intercepts are dif­ teractions between individuals in the genus ferent. That is, crabs the same size but of Calcinus were predicted well by the negoti­ different species selected different-sized shells. ations model. The precentage of cases cor­ A total of255 shell-related interactions was rectly predicted ranged from 100% (c. observed. Ofthese, 144 were intraspecific and seurati, C. haigae, C. latens, and C. elegans as III were between crabs of different species.