Habitat and mutualism affect the distribution and abundance of a shrimp-associated goby

CSIRO PUBLISHING www.publish.csiro.au/journals/mfr Marine and Freshwater Research,,, Habitat and mutualism affect the distribution and abundance of a shrimp-associated goby A. R. Thompson Department of Ecology, Evolution, and Marine Biology, University of California, Santa Barbara, California, United States of America. Current address: Vertebrates Ichthyology, Natural History Museum of Los Angeles County, 9 Exposition Blvd., Los Angeles, CA 9, USA. Email: andrew@yahoo.com Abstract. To elucidate factors that affect patterns of habitat use by a goby, Ctenogobiops feroculus, which interacts mutualistically with an alpheid shrimp, Alpheus djeddensis, goby density and habitat type were surveyed along the north shore of Moorea, French Polynesia. Although linear and quadratic multiple regression models both described the relationship between goby density and habitat as being statistically significant, more than twice the variation was accounted for by the quadratic model compared with the linear model. Further evaluation using univariate scatter plots confirmed that goby density correlated non-linearly with habitat as maximal goby density occurred in locations with approximately % sand and 9% rubble. Next, an aquarium-based, habitat choice experiment demonstrated that shrimp prefer to burrow in sand rubble compared with pure sand, and that shrimp are unable to build tunnels in pure sand because burrows collapse in this substrate. This research shows that habitat choice by one organism (e.g. a shrimp) can, because of positive interactions, affect the distribution of another species (e.g. a goby). In addition, it highlights the importance of assessing non-linear relationships in studies of habitat use and, when appropriate, using statistical methods that do not rely on linear assumptions. MF9 Gobyand A. R. Thompson shr imp habi tat use Extra keywords: coral reef fish, Gobioidei, snapping shrimp, symbiosis. Introduction Organisms are dispersed non-randomly (at least at some spatial scale) in every environment, and a principal goal of ecological research is to elucidate factors that cause these patterns. On coral reefs, for example, fish are distributed heterogeneously at scales spanning several orders of magnitude (Sale 998). These patterns are influenced by a combination of both physical (habitat characteristics, environmental disturbance and oceanographic processes) and biological (competition, predation, commensalisms and mutualism) factors. Studies examining the role of physical factors on reef fish distribution and abundance have produced equivocal results. On the one hand, many fishes utilise specific microhabitats and their distribution patterns are restricted by the occurrence of this microhabitat (Munday ; Holbrook et al. ). Accordingly, fish distribution (Letourneur et al. ) and abundance (Munday et al.99; Munday ; Schmitt and Holbrook ; Clarke and Tyler ) have been shown to be affected by microhabitat availability. Conversely, because habitat preferences often overlap greatly among heterospecific fishes (Sale et al. 98), habitat may be a poor predictor of fish distribution if incoming settlers cannot displace current residents and the establishment of fish on reefs is determined largely by chance settlement events (Munday et al. ). Indeed, several studies have failed to detect patterns between habitat characteristics and the distribution of at least some fish species (Sale and Douglas 98; Roberts and Ormond 989; Green 99; Ault and Johnson 998; Bean et al. ) or have found scale-dependent trends (Tolimieri 99; Munday ). In addition to physical parameters, biological interactions (e.g. predation, competition, commensalism, and mutualism) can affect the distribution and abundance of reef fish. Although competition (Munday et al. ; Scholfield ) and predation (Caley and St John 99) have frequently been found to impact reef fish populations (see review by Jones 99), studies on the role of positive interactions (commensalism and mutualism) are less CSIRO./MF99 -//

Marine and Freshwater Research A. R. Thompson Fig.. Ctenogobiops feroculus and Alpheus djeddensis. Note that the burrow is buttressed by coral rubble and that the shrimp is maintaining antennal contact with the goby. Photograph by Daniel Geiger, reproduced with permission. common. Nevertheless, it is becoming clear that commensal and mutualistic interactions are important on reefs, as many fishes associate exclusively with living organisms such as coral (Munday et al.99; Schmitt and Holbrook ), algae (Carr 99) and anemones (Schmitt and Holbrook ). In situations in which the distribution and abundance of fish is impacted directly by the presence of a habitatproviding organism, factors influencing the dynamics of that organism ultimately control the spatial arrangement of fish. Fluctuations in the abundance of coral-dwelling gobies, for example, have been shown to mirror changes in coral density (Munday et al. 99). Moreover, associations with fish can enhance the growth rates of coral (Meyer et al. 98) and anemones (Schmitt and Holbrook ). Hence, some fishes have the ability to augment habitat availability, thereby reducing competition for shelter. Consequently, heterospecific positive interactions are likely to play an important, but underappreciated, role in the distribution and abundance of coral reef fish. Given the lack of attention currently paid to populationlevel impacts of positive interactions in coral reefs (and in ecology in general cf. Bruno et al. ), I explored the roles of abiotic (physical habitat) and biotic (interaction with a mutualist) factors on the patterns of distribution and abundance of a common reef fish, the fierce shrimp goby, Ctenogobiops feroculus. I selected this goby as a model species because it is widespread on coral reefs in the Indo-West Pacific and interacts mutualistically with the snapping shrimp, Alpheus djeddensis (Polunin and Lubbock 9) (Fig. ). Associations between gobies and shrimp are found on many tropical reefs, and more than species of goby and species of shrimp share this mutual interaction (Karplus 98; A. Anker, personal communication). The relationship between gobies and shrimp is mutually beneficial (review by Karplus 98) because shrimp construct burrows in which gobies reside, and gobies warn shrimp of potential predators through rapid tail flicks that are felt by the shrimp s antennae (Karplus 99). The relationship is obligate for many species of shrimpassociated gobies (including C. feroculus) because goby mortality is very high when gobies are excluded from burrows (A. Thompson, unpublished data; Karplus 99). Gobies also aid shrimp, as shrimp do not emerge from burrows and burrow less when alone (Karplus et al. 9). Because shrimp obtain nutrition by uncovering food items, such as microcrustaceans, while digging (Magnus 9; Harada 99), reduced burrowing activity results in slower rates of shrimp growth (A. Thompson, unpublished data). To understand better the factors that influence the distribution and abundance of fierce shrimp gobies, I quantified habitat parameters and goby abundance at scales of m and m along the north shore of Moorea, French Polynesia. I then assessed whether shrimp gobies were associated with a particular type of habitat. Next, I determined whether spatial location had an influence on goby abundance by comparing goby abundance among the four lagoons on Moorea s north shore at both scales of sampling. Finally, I ascertained whether patterns of goby

Cooks Bay Opunahu Bay Goby and shrimp habitat use Marine and Freshwater Research (a) Moorea Open ocean Reef crest Scale: :, Kilometres (b) North Shore Lagoon CSP- CSP- CSP- CSP- Opunahu Bay FSP-FSP- - FSP- FSP- Land Fig.. (a) Island of Moorea, French Polynesia. (b) Location of survey sites along the north shore of Moorea. Surveys were taken at intervals of m at the lagoon scale; that is, coarse-scale plot (CSP) and at -m intervals within fine-scale plots (FSP). distribution and abundance are influenced by the ability of shrimp to construct burrows in various habitats by conducting habitat choice experiments within aquaria. Materials and methods Study location The study took place on Moorea, French Polynesia ( S, 9 W) between June and September, 999. Moorea is a volcanic island surrounded by a barrier reef that encloses relatively shallow lagoons (depth,. m; width, m). Live and dead coral (Porites spp., Pocillopora spp., Millepora spp. and Acropora spp.) are both abundant within the lagoons. Habitat types are distributed patchily within the lagoons (Thompson ) but, in general, coral cover is greatest near the reef crest, whereas sand and rubble predominate near the shore. Field study: substrate and shrimp goby surveys I quantified the spatial arrangement of gobies and habitat parameters by conducting surveys at two spatial scales within lagoons on the north shore of Moorea. At a coarse scale, I counted the number of gobies and quantified habitat structure within m quadrats that were placed at -m intervals along north south and east west axes within the four north-shore lagoons. (Henceforth, I will refer to these lagoons as coarse-scale plots (CSPs) with CSP being the furthest east and CSP being the furthest west; Fig. ). I determined the location of sampling sites by orienting myself with landmarks taken from a Cooks Bay N detailed nautical map and, when necessary, verified the distance between sample locations with a hand-held global positioning system (GPS) unit (GPS II plus, Garman Corporation, Olathe, KS, USA). Once a survey site was located, I placed a m quadrat on the substratum and quantified the type of habitat touching evenly spaced points on the quadrat. Specifically, I recorded the following habitat types: sand; rubble (dead, highly eroded coral < cm diameter); live branching coral (mostly Acropora spp.); live non-branching coral (mostly Porites spp., Pocillopora spp.); dead coral; and pavement (flat and eroded coral rock). After assessing the habitat composition of a sample site, I counted the number of gobies within the quadrat. Although almost every goby was clearly associated with a shrimp, I did not directly count shrimp because frequently they did not emerge from a burrow. Because coarse-scale sampling may have failed to capture variation between sample points, I conducted additional surveys at a finer scale. Here, I used the techniques described above to quantify habitat parameters and goby abundance, but conducted surveys at -m intervals within m plots. (As with the CSP, I will refer to fine-scale plots (FSP) as FSP, with FSP being furthest east). Fine-scale plots and were located within lagoon, FSP in lagoon and FSP in lagoon (Fig. ). The FSPs were separated by m and were located m from shore. I chose the position of FSPs to include at least some locations in which shrimp gobies were present. All surveys were carried out using a mask and snorkel. To elucidate relationships between goby density and habitat parameters, I first conducted a multiple linear regression using proportions of sand, rubble, coral and pavement as independent variables. Although I measured proportions of branching live, non-branching live and dead coral, I combined all three as coral because initial inspection of the data indicated that gobies responded similarly to each coral type. Because several of the relationships between goby density and habitat appeared to be non-linear, next I performed a multiple regression analysis using both linear and quadratic terms of each habitat parameter. I then created a scatter plot of each habitat type against goby density and used least-squares analyses to fit second- or third-order polynomial trend lines to the data. Finally, to better visualise relationships between habitat and density, I correlated mean goby density, which was calculated within. bins of each habitat parameter (e.g. I calculated mean goby density at locations with.. proportion sand,.. sand, etc.), against each habitat type. I determined whether goby density was spatially variable at a lagoon scale by comparing goby density among CSPs, using one-way analysis of variance (ANOVA) and post-hoc Tukey s test. Next, I restricted the ANOVA to include only those sample points with usable habitat (i.e. 9% sand and % rubble; discussed later in Results) to assess whether differences in goby density among lagoons were influenced by variability in appropriate habitat. I then used ANOVA and Tukey s test to determine whether habitat was the primary factor affecting distribution at the fine scale by comparing goby density among FSPs, using only those samples with appropriate habitat. Laboratory study: habitat choice experiments Because field surveys showed that goby density was relatively low in locations with close to % sand (see Results), using an aquarium experiment I determined whether this pattern was caused by active habitat selection. I placed one goby and one shrimp (average goby total length (TL) ± s.e. = 8. ±. mm, average shrimp TL ± s.e. =.8 ±. mm) in an aquarium (. m ) in which the substrate was pure sand on one side and a mixture of approximately % coral rubble and % sand (by volume) on the other. I added gobies and shrimp to the aquarium in the morning and after h assessed whether a burrow was constructed and where it was located. I conducted replicates using different gobies and shrimp each time. I followed the habitat choice

8 Marine and Freshwater Research A. R. Thompson Table. Multiple regression of proportion of habitat per quadrat and goby density (no. gobies per 9 m ) using only linear terms Term Parameter estimate F-ratio P Intercept.. Sand.. <. Rubble 9.8 8. <. Coral...8 Pavement.8.. Overall model results: F =.; P <.. experiment by placing one goby and one shrimp into an aquarium with a pure sand substrate and after h determined whether a burrow had been constructed and whether the goby associated with the shrimp. I again conducted trials using the gobies and shrimp from the previous experiment. I used a χ goodness of fit test to determine whether burrow location deviated from the null expectation of an even number of burrows in sand and sand rubble habitats. Finally, I visually inspected pure sand tanks to determine whether this environment precluded shrimp from constructing burrows. All statistical analyses were performed using JMP.. (SAS Institute Inc., Cary, NC, USA). Results Field study Although linear (Table ) and quadratic (Table ) multiple regression models both significantly explained the variability in goby density, the quadratic model (r =., F = 9., d.f. = 8, 8; P <.) accounted for more than twice the variation than the linear model (r =., F =., d.f. =, 8; P <.). Furthermore, in comparison with the linear model, the quadratic version reduced collinearity effects among the independent variables as all four habitat parameters contributed significantly to the quadratic model (Table ), but only two (sand and rubble) were significant in the linear model (Table ). Univariate scatter plots further emphasised the non-linear relationship between goby density and habitat, as a third-order polynomial equation best described the correlation between goby density and sand, and second-order polynomials best described the relationship between goby density and rubble, coral and pavement (P <. for all) (Fig. ). Finally, that goby density correlated non-linearly with habitat was illustrated most clearly through examination of scatter plots of average goby density against binned habitat parameters (Fig. ). Here, a third-order polynomial equation best described the relationship between mean goby density v. sand and rubble, whereas a second-order polynomial equation characterised the relationship between average goby density v. coral and pavement (P <. for all) (Fig. ). By evaluating the first derivative of equations for the best-fit lines and solving for, I determined that, on average, maximal goby density occurred when the substrate was composed of % sand and 9% rubble. Table. Multiple regression of proportion of habitat per quadrat and goby density (no. gobies per 9 m ) using quadratic terms Term Parameter estimate F-ratio P Intercept. <. Sand. 9.8 <. Sand 9..9 <. Rubble.. <. Rubble.. <. Coral... Coral..8. Pavement 8... Pavement 9. 9.. Overall model results: F = 8,99; P <.. Comparisons of goby density among lagoons (i.e. CSPs) indicated that density was significantly greater in CSPs and rather than, but that there was no significant difference between CSP and any of the other CSPs (Fig. a; F,9 =., P =.). The difference in goby density among lagoons, however, was explained by inter-lagoon habitat variation, as there was no difference in density when analysis was restricted to points with suitable habitat (i.e. 9% sand and % rubble) (Fig. b; F, =., P =.). Analysis of goby density among FSPs indicated that goby densities differed among plots even when points only with appropriate habitat were considered (Fig. ; F, = 9., P <.). Specifically, goby density was significantly lower in FSP compared with the other plots. Laboratory experiment Habitat choice experiments indicated that shrimp avoided pure sand habitats, as all burrows were constructed in the rubble sand habitat (Table ). Furthermore, in each trial, the goby associated with the shrimp and utilised its burrow for shelter. Examination of gobies and shrimp placed in pure sand habitats demonstrated that shrimp were incapable of constructing burrows in this environment as burrows consistently collapsed and left gobies and shrimp in a shallow sand depression that provided little shelter (Table ). Discussion Although competition and predation (review by Jones 99) have been shown to influence the distribution and abundance of fishes on coral reefs, the role of positive, interspecific interactions has received less explicit attention. My results suggest that the distribution of the fierce shrimp goby, C. feroculus, is affected significantly by the habitat preference of a mutualistic shrimp, A. djeddensis, and that this pattern was prevalent at multiple spatial scales. Field surveys indicated that gobies are restricted largely to habitats with a mixture of sand and rubble because shrimp are best able to construct burrows in such environments. Even though I expected that shrimp would not occupy habitats dominated

Goby and shrimp habitat use Marine and Freshwater Research 9 Goby density (no. gobies per 9 m ) Goby density (no. gobies per 9 m ) (a) r =.9....8 (c) r =.....8 Proportion coral (b) r =.9....8 Proportion rubble (d) r =.....8 Proportion pavement Fig.. Scatter plots and best-fit trend lines depicting relationships between habitat and goby density using all sample points from coarse and fine scales. Trend lines were characterised by the following equations: (a) y =.8x + 88.x.8x +. (sand); (b) y =.x +.x +. (rubble); (c) y =.9x.x +. (coral); (d) and y =.x.x +. (pavement), whereby y represents goby density and x is the proportion habitat at a sample location. by hard substrates, such as coral or pavement, because shrimp cannot tunnel through solid substrate, I unexpectedly found that pure sand substrate also precluded burrow construction because at least some hard substrate is necessary to keep tunnels from collapsing. Hence, the distribution of A. djeddensis is constrained by appropriate habitat, which, in turn, apparently dictates the distribution and abundance of C. feroculus. In addition to C. feroculus and A. djeddensis in Moorea, the distribution patterns of several species of gobyassociated shrimp are likely to be affected by habitat parameters. In Japan, Yangisawa (98) found that the goby Amblyeleotris japonica and its shrimp partner Alpheus bellulus are found primarily in locations with a mixture of sand, pebbles, coral debris and shell fragments, but are scarce in pure sand habitats. In addition, four species of goby-associated shrimp that reside in lagoon habitats in the northern Red Sea each inhabit specific microhabitats that are characterised by sediment grain size and depth (Karplus et al. 98). In contrast to the shrimp, gobies are associated more weakly with particular habitats in the Red Sea and are frequently associated with several species of shrimp (Karplus et al. 98). As opposed to the Red Sea, where nine species of gobies and eight species of shrimp are found in close proximity, there are only three goby and three shrimp species in Moorea and their distributions are allopatric (A. Thompson, personal observation). Qualitative observations showed that the goby Amblyleotris wheeleri and the shrimp Alpheus sp. are found outside of lagoons, in relatively deep (> m) water with coarse sediment. In addition, the goby Vanderhorstia ornatissima and its shrimp partner Alpheus sp. reside solely in silty habitats within lagoons (A. Thompson, personal observation). Hence, habitat may influence the distribution patterns of multiple species of gobies and shrimp in Moorea. An important finding of the present study is that the carrying capacity of one organism (a goby) is influenced directly by a mutualistic interaction with another species (shrimp). Although the goby shrimp interaction is a highly co-evolved mutualism (Karplus 98), the general processes that influence the dynamics of this system are likely to apply to many reef organisms. Trapeziid crabs (Trapezia sp.) and alpheid shrimp (Alpheus lottini), for example, interact mutualistically with branching corals (Glynn 9). In such

Marine and Freshwater Research A. R. Thompson Goby density (no. gobies per 9 m ) Goby density (no. gobies per 9 m ) 9 8 r =.99........8.9 (c) (a) r =.99........8.9 (b) 9 r =.9 8........8.9 (d) r =.998........8.9 Fig.. Scatter plots and best-fit trend lines depicting relationships between habitat and goby density using data calculated at. intervals of each habitat type. Error bars represent ± s.e. Trend lines were used to fit values taken from each bin and were characterised by the following equations: (a) y =.9x +.9x.x +.8 (sand); (b) y = 9.x x +.x. (rubble); (c) y =.x.9x +. (coral); and (d) y =.8x.9x +. (pavement), whereby y is the mean goby density and x is the proportion of each type of habitat. systems, corals provide shelter and food (in the form of mucus secretion) for the crustaceans, whereas crustaceans actively defend corals against predators such as the crown of thorns starfish (Acanthaster planci) (Glynn 9). Similarly, majid crabs (Mithrax sp.) utilise coralline algae (Stachowicz and Hay 99) or coral (Stachowicz and Hay 999) for shelter and benefit their hosts by consuming epiphytic algae, which would otherwise overgrow them. A comparable system involves damselfishes and anemones; anemones provide damselfishes with protection and the damselfishes defend anemones from predators (Schmitt and Holbrook ). The common thread between these interactions is that one species directly chooses a habitat whereas the other utilises its host as habitat. In such systems it is critical to assess simultaneously the dynamics of both species to elucidate causes of their distribution and abundance. In addition to direct mutualistic interactions, numerous reef fishes reside commensally on or within larger sessile or slowly moving invertebrates (e.g. sponges, corals or echinoderms). As with the goby examined in the present study, the population ecology of such fishes is ultimately influenced by the distribution patterns of another organism. The dependence of fish on coral dynamics was demonstrated in a study that examined habitat use by obligate coral-dwelling gobies, Gobiodon sp. (Munday et al. 99). These gobies reside within specific coral species and their populations vary in response to coral density. Similarly, the abundance of another coral-dwelling goby, Paragobion echinocephalus, was found to fluctuate in concert with a branching coral, Stylophora pistillata (Kuwamura et al. 99). Consequently, to understand the factors influencing coral-dwelling fish populations, it is often necessary to examine the dynamics of organisms that provide shelter rather than merely treating those organisms as habitat. The present study accentuates two points that should be considered in studies of the effect of habitat on fish density. First, it is necessary to determine whether a linear relationship exists between habitat parameters and fish density. On the north shore of Moorea, I found that shrimp and goby density correlated non-linearly with both sand and rubble. Although linear and quadratic multiple regression models both indicated that a significant relationship existed between habitat and goby density, the quadratic model accounted for more than twice the variability in goby density than the linear

Goby and shrimp habitat use Marine and Freshwater Research Mean goby density (no. 9 gobies per m CSP ) Mean goby density (no. gobies per 9 m CSP ) 9 8 (a) (b) Course-scale plot Course-scale plot Fig.. Mean goby density per coarse-scale plot (i.e. lagoon). (a) Densities are calculated using data from all coarse-scale plots. (b) Densities are calculated using sample points only with appropriate habitat ( 9% sand and % rubble). Error bars represent ± s.e. model (Tables, ). This result highlights the importance of testing for non-linear relationships in studies of habitat use. If non-linear patterns exist, then problematic conclusions might be reached if common examination techniques that are based on linear correlation (e.g. principle component analysis, multiple regression) are used to analyse data. Hence, non-linear relationships between habitat and fish density might contribute to the relatively common finding that fish abundance is not affected by habitat (e.g. Roberts and Ormond 98; Green 99; Ault and Johnson 998). Second, my results stress the need to quantify accurately the precise microhabitat requirements of a particular fish before searching for patterns of broad-scale habitat use. If appropriate microhabitat is not measured when researching fish habitat relationships, there will be little ability to find meaningful relationships between habitat parameters and fish density. For example, although Sale (9) did not consistently detect strong relationships between the density of the damselfish, Dascyllus aruanus and habitat, Holbrook et al. () found that habitat accurately explained its density over several spatial scales. Holbrook et al. () attributed this discrepancy to their ability to identify the precise microhabitat requirements of D. aruanus rather than relying on coarse categorisations of habitat type. In addition, although Jenkins and Hamer () found no relationship between seagrass cover and King George whiting (Sillaginoides punctata) density, they detected a significant correlation between habitat and S. punctata after accounting for both interstitial prey (microcrustaceans) and seagrass among sample sites. Similarly, my ability to detect the effect of habitat on shrimp gobies would have been compromised had I not considered both sand and rubble. The goal of the present study was not to downplay the role of competition and predation on reef fish distribution, but to heighten the attention paid towards the important role of mutualistic and commensal interactions among many reef organisms. In addition to habitat type and mutualistic interactions, the distribution of shrimp gobies in Moorea, like most reef fishes, are probably affected by a variety of abiotic and biotic forces (see Jones 99; Schmitt and Holbrook ; Shima ) such as recruitment variability (Doherty and Fowler 99), predation (Hixon and Beets 99) and intraspecific competition for mutualists (Karplus et al. 9; Cushman and Whitham 99). For example, although habitat heterogeneity explained the abundance of gobies among lagoons (Fig. b), goby abundance within appropriate habitats was significantly less in FSP than in the other FSPs (Fig. ). Hence, a factor other than habitat type affected abundance patterns in FSP. A potential (but untested) explanation for the relatively low abundance of gobies in FSP is that shrimp and/or goby recruitment was low at this location. Rates of recruitment, which are dictated by larval input and post-settlement processes such as competition and predation, have been shown to strongly influence the size of marine invertebrate (Roughgarden et al.988) and fish (Doherty and Fowler 99) populations in many systems, including Moorea (Schmitt and Holbrook ; Shima ). In Moorea, water comes into the lagoons over the reef crest and flows out through channels in a Table. Influence of substrate on shrimp habitat use Given are: (i) the number of burrows constructed by shrimp when given a choice between a substrate comprising either pure sand or a mix of sand and rubble; and (ii) the number of burrows constructed when offered only a pure sand habitat Habitat choice experiment Pure sand experiment No. burrows constructed in: Sand Sand + rubble χ P No. trials No. burrows <.

Marine and Freshwater Research A. R. Thompson Mean goby density (no. gobies per 9 m FSP ) 8 Fine scale plot Fig.. Mean goby density per fine-scale plot. Samples only with appropriate habitat ( 9% sand and % rubble) were used for this analysis. Error bars represent ± s.e. unidirectional manner (Shima 999). Shima () demonstrated that rates of larval settlement of six bar wrasse, Thalassoma hardwicke, were greatest near the reef crest and declined towards shore. Hence, T. hardwicke recruitment on patch reefs was a direct function of settlement rates at near-shore sites, but was influenced more strongly by density-dependent morality near the reef crest where suitable habitat was consistently saturated. Because FSP was further from the reef crest than the other FSPs, and gobies and shrimp both have pelagic larvae (A. Thompson, personal observation), perhaps a reduced supply of larva limited the abundance of shrimp gobies in this location. Consequently, multiple factors must be considered in order to gain a complete picture of which factor(s) influence the distribution and abundance of reef fish (Jones 99; Schmitt and Holbrook ; Shima ). In summary, this is the first study to quantify the effects of environmental heterogeneity on the distribution and abundance of a shrimp-associated goby. My findings emphasise that habitat use studies need to: (i) deduce the precise habitat requirements of organisms; (ii) assess whether or not relationships between habitat and organism density are non-linear and, if necessary, use appropriate statistical analyses that account for non-linearity; and (iii) consider the dynamics of multiple organisms simultaneously because many organisms interact positively with and depend on other species for habitat. Incorporating this approach when studying fish habitat use should provide insight towards which factors ultimately control the population ecology of many species of reef fish. Acknowledgments I thank A. Anker, S. Cooper, S. Holbrook, I. Karplus, B. Larson, A. Rassweiler, W. Rice, D. Sax and R. Schmitt for their critical discussions; N. Davies, B. Williamson and J. White for their technical assistance; D. Geiger for taking excellent photographs of gobies and shrimp; and J. Blythe, D. Craig, J. Lewis and M. Schmitt for their assistance in the field. Funding for this work was provided by RTG and GRT Programs in Spatial Ecology (NSF BIR9- and NSF GER9-8, both to W. Murdoch). This paper is contribution # of UC Berkeley s Richard B. Gump South Pacific Research Station, Moorea, French Polynesia. References Ault, T. R., and Johnson, C. R. (998). Spatially and temporally predictable fish communities on coral reefs. Ecological Monographs 8,. Bean, K., Jones, G. P., and Caley, M. J. (). Relationships among distribution, abundance and microhabitat specialization in a guild of coral reef triggerfish (family Balistidae). Marine Ecology Progress Series,. Bruno, J. F., Stachowicz, J. J., and Bertness, M. D. (). Inclusion of facilitation into ecological theory. Trends in Ecology & Evolution 8, 9. doi:./s9-()-9 Caley, J. M., and St John, J. (99). Refuge availability structures assemblages of tropical reef fishes. Journal of Animal Ecology, 8. Carr, M. H. (99). Effects of macroalgal dynamics on recruitment of a temperate reef fish. Ecology,. Clarke, R. D., and Tyler, J. C. (). Differential space utilization by male and female spinyhead blennies, Acanthemblemaria spinosa (Teleostei: Chaenopsidae). Copeia,. Cushman, J. H., and Whitham, T. G. (99). Competition mediating the outcome of a mutualism: protective services of ants as a limiting resource for membracids. American Naturalist 8, 8 8. doi:.8/8 Doherty, P. J., and Fowler, A. J. (99). An empirical test of recruitment limitation in a coral reef fish. Science, 9 99. Glynn, P. W. (9). Some physical and biological determinants of coral community structure in the eastern Pacific. Ecological Monographs,. Green, A. L. (99). Spatial, temporal and ontogenetic patterns of habitat use by coral reef fishes (Family Labridae). Marine Ecology Progress Series,. Harada, E. (99). On the interspecific association of a snapping shrimp and gobiid fishes. Publications of the Seto Marine Biological Laboratory,. Hixon, M. A., and Beets, J. P. (99). Predation, prey refuges, and the structure of coral-reef fish assemblages. Ecological Monographs,. Holbrook, S. J., Brooks, A. J., and Schmitt, R. J. (). Variation in structural attributes of patch-forming corals and in patterns of abundance of associated fishes. Marine and Freshwater Research,. doi:./mf Holbrook, S. J., Forrester, G. E., and Schmitt, R. J. (). Spatial patterns in abundance of a damselfish reflect availability of suitable habitat. Oecologia, 9. Jenkins, G. P., and Hamer, P. A. (). Spatial variation in the use of seagrass and unvegetated habitats by post-settlement King George whiting (Percoidei: Sillaginidae) in relation to meiofaunal distribution and macrophyte structure. Marine Ecology Progress Series, 9 9. Jones, G. P. (99). Postrecruitment processes in the ecology of coral reef fish populations: A multifactorial perspective. In The Ecology of Fishes on Coral Reefs. (Ed. P. F. Sale.) pp. 9 8. (Academic Press: San Diego, CA.)

Goby and shrimp habitat use Marine and Freshwater Research Karplus, I. (99). The tactile communication between Cryptocentrus steinitzi (Pisces, Gobiidae) and Alpheus purpurilenticularis (Crustacea, Alpheidae). Zeitschrift fur Tierpsychologie 9, 9. Karplus, I. (98). The association between gobiid fishes and burrowing alpheid shrimps. Oceanography and Marine Biology,. Karplus, I., Szlep, R., and Tsurnamal, M. (9). Associative behavior of the fish Cryptocentrus cryptocentrus (Gobiidae) and the pistol shrimp Alpheus djiboutensis (Alpheidae) in artificial burrows. Marine Biology, 9. Karplus, I., Szlep, R., and Tsurnamal, M. (9). The burrows of Alpheid shrimp associated with Gobiid fish in the northern Red Sea. Marine Biology, 9 8. Karplus, I., Szlep, R., and Tsurnamal, M. (98). Goby-shrimp partner specificity. I. Distribution in the northern Red Sea and partner specificity. Journal of Experimental Marine Biology and Ecology, 9. doi:./-98(8)9-9 Kuwamura, T., Yogo, Y., and Nakashima, Y. (99). Population dynamics of goby Paragobiodon echinocephalus and host coral Stylophora pistillata. Marine Ecology Progress Series,. Letourneur, Y., Ruitton, S., and Sartoretto, S. (). Environmental and benthic habitat factors structuring the spatial distribution of a summer infralittoral fish assemblage in the north-western Mediterranean Sea. Journal of the Marine Biological Association of the UK 8, 9. Magnus, D. B. E. (9). Zur okologie sedimentbewohnender Alpheus garnelen (Decapoda Natantia) des Roten Meeres. Helgolander Wissenschaftliche Meeruntersuchungen,. Meyer, J. L., Schultz, E. T., and Helfman, G. S. (98). Fish Schools an Asset to Corals. Science, 9. Munday, P. L. (). Interactions between habitat use and patterns of abundance in coral-dwelling fishes of the genus Gobiodon. Environmental Biology of Fishes 8, 9. doi:./a: 899 Munday, P. L. (). Does habitat availability determine geographical-scale abundances of coral-dwelling fishes? Coral Reefs,. Munday, P. L., Jones, G. P., and Caley, M. J. (99). Habitat specialization and the distribution and abundance of coral-dwelling gobies. Marine Ecology Progress Series, 9. Munday, P. L., Jones, G. P., and Caley, M. J. (). Interspecific competition and coexistence in a guild of coral-dwelling fishes. Ecology 8, 89. Polunin, N. V. C., and Lubbock, R. (9). Prawn-associated gobies (Teleostei-Gobiidae) from Seychelles, Western Indian Ocean - systematics and ecology. Journal of Zoology 8,. Roberts, C. M., and Ormond, R. F. G. (98). Habitat complexity and coral reef fish diversity and abundance on Red Sea fringing reefs. Marine Ecology Progress Series, 8. Roughgarden, J., Gaines, S., and Possingham, H. (988). Recruitment dynamics in complex life-cycles. Science,. Sale, P. F. (9). Influence of corals in the dispersion of the pomacentrid fish, Dascyllus aruanus. Ecology,. Sale, P. F. (998). Appropriate spatial scales for studies of reef-fish ecology. Australian Journal of Ecology, 8. Sale, P. F., and Douglas, W. A. (98). Temporal variability in the community structure of fish on coral patch reefs and the relation of community structure to reef structure. Ecology, 9. Sale, P. F., Doherty, P. J., Eckert, G. J., Douglas, W. A., and Ferrell, D. J. (98). Large-scale spatial and temporal variation in recruitment to fish populations on coral reefs. Oecologia, 9 98. Schmitt, R. J., and Holbrook, S. J. (). Habitat-limited recruitment of coral reef damselfish. Ecology 8, 9 9. Schmitt, R. J., and Holbrook, S. J. (). Mutualism can mediate competition and promote coexistence. Ecology Letters, 898 9. doi:./j.-8...x Shima, J. S. (999). Variability in relative importance of determinants of reef fish recruitment. Ecology Letters,. doi:./j.-8.999.89.x Shima, J. S. (). Recruitment of a coral reef fish: Roles of settlement, habitat, and postsettlement losses. Ecology 8, 9 99. Scholfield, P. J. (). Habitat selection of two gobies (Microgobius gulosus, Gobiosoma robustum): influence of structural complexity, competitive interactions, and presence of a predator. Journal of Experimental Marine Biology and Ecology 88,. doi:./s-98()- Stachowicz, J. J., and Hay, M. E. (99). Facultative mutualism between an herbivorous crab and a coralline alga: advantages of eating noxious seaweeds. Oecologia, 8. Stachowicz, J. J., and Hay, M. E. (999). Mutualism and coral persistence: the role of herbivore resistance to algal chemical defense. Ecology 8, 8. Thompson, A. R. (). Habitat-use patterns of the fierce shrimpgoby (Ctenogobiops feroculus) at multiple scales: a geospatial approach. In Proceedings of the 9th International Coral Reef Symposium. (Eds M. K. Kasim Moosa, S. Soemodihardjo, A. Nontji, S. A., K. Romimohtarto, Sukarno, and Suharsono) pp. 9. (Ministry of Environment, the Indonesian Institute of Sciences and the International Society for Reef Studies: Bali, Indonesia.) Tolimieri, N. (99). Effects of microhabitat characteristics on the settlement and recruitment of a coral reef fish at two spatial scales. Oecologia,. Yangisawa, Y. (98). Studies on the interspecific relationship between gobiid fish and snapping shrimp. II. Life history and pair formation of snapping shrimp (Alpheus bellulus). Publications of Seto Marine Biological Laboratory 9, 9. Manuscript received 9 July ; revised September, November ; accepted November. http://www.publish.csiro.au/journals/mfr