Individual differences in microhabitat use in a Caribbean cleaning goby: a buffer effect in a marine species?

Authors

  • ELIZABETH A. WHITEMAN,

    1. Centre for Ecology, Evolution and Conservation, School of Biological Sciences, University of East Anglia, Norwich, NR4 7TJ, UK
    Search for more papers by this author
  • ISABELLE M. CÔTÉ

    Corresponding author
    1. Centre for Ecology, Evolution and Conservation, School of Biological Sciences, University of East Anglia, Norwich, NR4 7TJ, UK
    Search for more papers by this author

I.M. Côté, Centre for Ecology, Evolution and Conservation, School of Biological Sciences, University of East Anglia, Norwich NR4 7TJ, UK. E-mail: i.cote@uea.ac.uk

Summary

  • 1Buffer effects occur when changes in population size result in the disproportionate use of poor-quality habitats. Thus, at low population sizes high-quality habitats are used preferentially. As population size increases an increasing proportion of the population uses poorer-quality habitats. Assessment of the temporal and spatial variation in patterns of habitat occupancy can therefore shed light on the differences in quality between habitats and the individual fitness consequences of habitat choice.
  • 2We provide the first evidence of the potential operation of a buffer effect for a site-attached marine species. Caribbean cleaning gobies Elacatinus prochilos (Böhlke & Robins) occupy coral and sponge on fringing reefs in Barbados. For adult gobies, resource selection indices suggested a preference for sponge. However, as cleaning goby population size increased, the number of adult cleaning gobies occupying sponge increased more rapidly than the number occupying coral. In contrast, adults preferentially re-colonized coral following experimental removals and at these low population densities the rate of population increase was greater on coral. Our results suggest that coral may be the preferred habitat, but in Barbados this habitat becomes saturated at very low population densities as a consequence of low client densities and ectoparasite loads. Thus, a larger proportion of the population occupies sponge at most observed population densities.
  • 3Patterns of habitat occupancy with population size for recruits and juveniles suggest only a small difference in habitat quality between sponge and coral. Indeed, recruits and juveniles do not discriminate between sponge and coral. The population shift towards sponge rather than coral occupancy between recruitment and maturity may arise as a combination of differing survival of recruits and juveniles on coral and sponge and active movement of individuals towards sponge.
  • 4Our results demonstrate that interactions among individuals are an integral part of population distribution and dynamics and are therefore important in future studies of habitat choice and its associated fitness consequences.

Introduction

Local populations of terrestrial and marine species often show spatial variation in abundance. Such differences are particularly apparent in coral reef fishes. Differences in habitat use have been demonstrated to arise either through habitat selection at settlement (e.g. Stegastes spp.; Gutiérrez 1998) or after settlement as a result of interspecific (e.g. Stegastes spp.; Robertson 1996; Gobiodon spp.; Munday, Jones & Caley 1997) or intraspecific competition (e.g. Stegastes planifrons threespot damselfish; Levin et al. 2000). These differences in habitat occupancy can often be linked to individual differences in size or developmental stage (e.g. Epinephelus striatus Nassau grouper; Eggleston 1995; Parupeneus multifasciatus goatfish; McCormick & Makey 1997; Plectropomus leopardus coral trout; St John 1999). By contrast, occupancy by reef fishes of two or more qualitatively different habitats by a single age class is observed more rarely. Adults may differ in density across reef zones (e.g. Acanthurus bahianus ocean surgeonfish; Lawson, Kramer & Hunte 1999) but we are currently aware of only one marine example in which adults display a dichotomous distribution between two different substrata. The Caribbean cleaning goby Elacatinus prochilos occupies both barrel sponges Xestospongia muta and coral heads. We have previously shown for a Barbadian population that these two distinct substrata were linked to differences in social behaviour, diet and cleaning activity (Whiteman & Côté 2002a).

Habitat or dietary specialization within a single population may arise through a variety of mechanisms. If individuals differ in their ability to exploit differing habitats through, for example, differing morphology or foraging behaviours, then specialization on a particular habitat may maximize the rate of return for each individual. Such resource polymorphisms are frequent among freshwater fish (e.g. Salvelinus fontinalis brook charr; McLaughlin 2001; Lepomis macrochirus bluegill sunfish; Ehlinger 1990) and may have arisen in response to competition for resources. Proximately, intraspecific competition can generate varying patterns of habitat use and distribution within a population.

If similar individuals experience exploitative intraspecific competition, the ideal free distribution predicts that individuals will choose the habitat with the highest net profitability such that the average fitness of individuals at high density in a higher-quality habitat is the same as the average fitness of individuals at lower densities in a poorer-quality habitat (e.g. Rhinichthys atralulus minnows; Tyler & Gilliam 1995). In contrast, if individuals differ in competitive ability, interference competition will result in competitively superior individuals occupying the higher-quality habitat. Competitively inferior individuals are forced into lower-quality habitats. Although fitness is maximized for each individual, individuals will differ in fitness between habitats. For example, Limosa limosa (black-tailed godwits) on low-quality estuaries have lower prey-intake rates, lower survival rates and arrive later in Icelandic breeding grounds than godwits on high-quality estuaries (Gill et al. 2001). Similarly, juvenile Perca flavescens (yellow perch) at high densities split into a fast-growing littoral group and a slower growing pelagic group as a result of density-dependent interference competition for the higher-quality habitats (Post, Johannes & McQueen 1997).

The fitness consequences of habitat choice may therefore vary significantly according to the mechanisms creating patterns of habitat occupancy. Among many reef fishes, however, data necessary to establish fitness differences between individuals are difficult to collect. Mobile fish species are hard to track, observations of spawning and larval development necessary for measuring reproductive success are often impossible, and small increments of growth in small-bodied species hinder accurate measurement of growth rates. Alternative methods of addressing the individual consequences of habitat choice, as well as concomitant relative habitat quality, are therefore required.

Assessment of temporal and spatial variation in patterns of habitat occupancy can shed light on the differences in quality between habitats, the individual fitness consequences of habitat choice and ultimately the processes generating disparate habitat use. Shifts in habitat occupancy through time have been demonstrated most frequently in terrestrial vertebrates and freshwater habitats and may occur through a buffer effect (e.g. L. limosa; Gill et al. 2001). Buffer effects occur when changes in population size result in the disproportionate use of poor-quality sites or habitats. Thus, at low population sizes high-quality habitats are preferentially used. As population size increases an increasing proportion of the population uses poorer-quality habitats (Brown 1969). Habitat quality can therefore be inferred by the rate at which habitat-specific sections of the population grow in relation to total population size, assuming that habitat quality remains constant. Although buffer effects have been reported widely in terrestrial animals, they have rarely been demonstrated in marine species (e.g. Gadus morhua Atlantic cod; Hutchings 1996).

We use these methods of analysing the spatial and temporal patterns of habitat occupancy to shed light on the processes creating the disparate habitat use in the Caribbean cleaning goby E. prochilos. In Barbados, at coral cleaning stations E. prochilos are found solitarily or in small groups, while groups of up to 40 individuals are observed on sponge cleaning stations (Whiteman & Côté 2002a). Coral-dwelling E. prochilos spend 25 times longer cleaning and take 16 times more bites on fish clients than those on sponge, while sponge-dwelling individuals forage largely on the parasitic polychaetes Haplosyllis spp., which live within the aquiferous canals of the sponges (Whiteman & Côté 2002a). Furthermore, sponge-dwelling social groups exhibit highly stable, size-structured dominance hierarchies in which larger, competitively dominant individuals monopolize areas with the highest food density (Whiteman & Côté 2004). The relative quality of sponge and coral habitats is not directly known. However, gobies display spatial awareness. Individuals have been observed moving between a home sponge and a nearby coral head and can locate a home sponge from more than 40 m away (E. Whiteman, unpublished data), suggesting an ability to locate and select between different habitat types. We therefore predict that patterns of habitat occupancy will reveal a buffer effect in which saturation of either sponge or coral leads to an increase in the proportion of gobies occupying the alternative habitat with population increases. We also investigate whether the pattern of habitat occupation of E. prochilos stems from processes occurring soon after settlement or later in life by comparing the rates of occupancy on both substrata in relation to population size between recruits, juveniles and adults. Finally, we use experimental removals of adults from patch reefs to determine whether individuals preferentially select coral or sponge.

Methods

field site and study species

The study was carried out in Barbados (13°10′N, 59°30′W), West Indies, between February 2000 and June 2002. All observations were made on the North and South sections of the Bellairs Reef in the Barbados Marine Reserve, a 2·2-km stretch of coast containing fringing reefs, on the west coast of the island.

We focused on the broadstripe cleaning goby E. prochilos. This small (maximum total length, 4 cm) species is conspicuously coloured, with a white stripe running laterally from snout to tail on a black body. E. prochilos are sexually monomorphic in colouration and gross morphology. However, sexes can be distinguished by examining the shape of the urogenital papilla, which is long and conical in males and short, truncate and often lobed in females (Thresher 1984; E. Whiteman, personal observations).

fine-scale patterns of distribution and habitat choice

In February 2001, two census areas, each 40 m × 40 m, were identified, one in the patch reef near the seaward edge of the North Bellairs Reef and one in the spurs and grooves zone on South Bellairs Reef. In each census area the position of all cleaning stations occupied by E. prochilos were recorded. We refer to both corals and sponges occupied by E. prochilos as cleaning stations. For cleaning stations on coral and the sponge Ircinia strobilina the species was noted and the surface area of each was estimated by counting the number of 5 cm squares covered in a grid placed over it. For cleaning stations on sponges Xestospongia muta the maximum length, height and depth (cm) of each sponge were recorded and surface area estimated as [(2 × length × height) + (2 × length × depth)]. The height above the reef, height above sand and the distance from the edge of the patch (or spur) of each coral head and sponge were also recorded. Potential cleaning stations were identified as every sponge and coral head of each species not occupied by cleaning gobies that was equal in size or larger than the smallest cleaning station recorded. These potential cleaning stations were then mapped and physical characteristics recorded in the same way as occupied cleaning stations. Both plots were then re-censused in April, June and August 2001. Gobies were classified by size into recruits (< 12 mm), juveniles (12–25 mm) and adults (> 25 mm). Total length was estimated by noting the position of the head and tail of an individual against the substratum and measuring the distance between these points.

Resource selection ratios were used to determine the specific substratum preferred by adults and recruits. Data were analysed from all censuses for patterns of adult and juvenile selection but only from the February and April censuses for patterns of recruit selection as the total numbers of recruits were low in the June and August censuses. Selection ratios were estimated using the formula:

Si = xi/yi

where xi is the proportion of all individuals within an age class occupying substratum species i and yi is the proportion of total available area of all substrata that is species i.

Substrata which were available on each patch but which were not inhabited by gobies were assigned a value of zero. Coralline rock, Agaricia agaricites and Porites astreoides were excluded, as the area of suitable habitat could not be calculated accurately. The use of a substratum species was considered to be disproportional to its availability where the 95% CI did not encompass 1 (Manly, McDonald & Thomas 1993). Within the two census plots, there were no significant differences between census times in the resource selection ratios of adults, juveniles and recruits for each substratum species when North and South Bellairs reefs are considered together and separately (repeated-measures anova: P > 0·05 in all cases). There were also no differences between reefs in the selection ratio for each substratum when censuses were considered together and separately (paired-t: P > 0·05 in all cases). Data from all four censuses were therefore combined.

To sex individuals and accurately measure adult sizes for comparison of individuals between sponge and coral, all adults within 1200 m2 of each census area were captured in hand nets using a clove oil and ethanol solution (1 : 4) during the February and June censuses. Individuals were then measured, sexed and returned to their cleaning station. The maximum time elapsed between capture and release was 1 min and the whole procedure was carried out near the reef surface to minimize stress for the fish. Fish recovered from the effects of the clove oil within 2 min and there were no obvious effects of capture (Whiteman & Côté 2002b).

population monitoring: potential buffer effect

In February 2000, five areas of reef were identified, three patch reefs on North Bellairs Reef and two rectangular plots in the spurs and grooves zone of South Bellairs Reef. The maximum and minimum depth of each patch ranged from 5·25 to 7 m and 4–6·25 m, respectively. Patches ranged in size from 13·8 to 28·0 m2.

In each patch the position and coral or sponge species of all cleaning stations occupied by E. prochilos were recorded. Sponges X. muta and coral heads co-occur across the patch reefs and spurs and grooves zones of these reefs, with no obvious aggregation of either substratum. All patches were censused fortnightly between February and August 2000 and weekly between February to September 2001 and February to June 2002. During each census the number of gobies of each species on each coral head or sponge was noted. Gobies were classified as recruits, juveniles and adults, as described above.

To investigate short-term variability in the number of gobies recorded, each site was censused three times on three consecutive days during seven censuses from February to July 2000. There were no significant differences between each of the three censuses during any census period in the number of adults, juveniles and recruits of E. prochilos (repeated-measures anova: P > 0·05 in all cases). Therefore, only the first census from each set was used in the subsequent analysis. A total of 61 census periods were analysed from 2000 to 2002. Each age class was classified into two categories according to the substratum on which they were observed: sponge and coral (including coralline rock). The number of gobies of each age class on each habitat type was then related to total population size and the relative rates of habitat-specific population increase compared.

experimental test of habitat choice

To test experimentally adult preference for coral or sponge, in June 2001, 10 patch reefs/truncated spurs were identified on the seaward edge of the Bellairs reefs (North Bellairs: five patches, South Bellairs: five spurs). Distance from shore ranged from 120 to 200 m and patches ranged in size from 20 to 85 m2. On each patch reef, the location and coral or sponge species of all cleaning stations were mapped. Gobies were classified as recruit, juveniles and adults, and sizes estimated as described above.

All cleaning gobies were then removed from five patches and gobies classified as adults were removed from the remaining five patches. Fish were caught in hand nets using a clove oil and ethanol solution (1 : 4). Anaesthetized gobies were transferred to plastic bags and allowed to recover for 5 min from the effects of the clove oil before being released more than 50 m from their home territory. These distances were necessary to preclude a return to the home sponge (E. A. Whiteman, personal observation). Previous experiments have indicated no adverse effects of capture and release (Whiteman & Côté 2002b). Each patch was then re-censused fortnightly from June to September 2001 and from February to April 2002 and the location and species of each coral head or sponge occupied by adults was recorded. A total of 14 census periods were therefore recorded.

Five unmanipulated patches identified for population monitoring (above) served as control sites for the experiment. The number of adults, juveniles and recruits on each patch from which gobies had been removed declined from June to September 2001 but then increased from February to April 2002. This was consistent with a recruitment peak observed during February 2002 on the five unmanipulated patches (E. A. Whiteman, unpublished data). The rates of recolonization of sponge and coral were therefore compared both in 2001, immediately following goby removal, and in 2002, from the first census period in that year.

In addition, potentially available coral heads and sponges were estimated as all the coral heads and sponges on which gobies were observed during the monitoring periods. This represented more than 95% of the coral heads which were equal in size or larger than the smallest observed occupied coral head. The observed proportion of adults on coral was then compared to expected values taking into account variation between patches in the relative availability of each substratum type.

Finally, adults were classified into two categories according to the substratum on which they were observed: sponge and coral (including coralline rock). The number of adults on each habitat type was then related to total population size and the relative rates of habitat-specific population increase compared.

Results

Cleaning stations occupied by adult E. prochilos were observed on the sponge X. muta and on seven coral species: Siderastrea siderea, Montastrea cavernosa, M. annularis, Diploria strigosa, Porites astreoides, Agaricia agaricites, Millepora squarrosa and coralline rock. Recruits and juveniles were also observed on the sponge Ircinia strobilina and the corals Diploria labyrinthiformis and Colpophyllia natans.

patterns of adult distribution

Fine-scale patterns of adult distribution

The number of adult gobies in both the North and South Bellairs census areas increased from the February census to a maximum in the June census (Fig. 1a). However, throughout the study densities of adults on South Bellairs Reef were consistently more than three times higher than on North Bellairs Reef.

Figure 1.

Total number of adult gobies (a) and the percentage of sponges (b) and corals (c) occupied by adults during the February, April, June and August censuses on North (filled points and bars) and South Bellairs reefs (open points and bars).

The relative availability of sponge and coral differed between the North and South Bellairs census plots. X. muta comprised 23% of the mapped potential and occupied cleaning stations on South Bellairs Reef but only 6·5% on North Bellairs Reef. Similarly, the surface area of sponges available on South Bellairs Reef was 33% of the total mapped area of available substratum on South Bellairs but only 11·1% on North Bellairs. Coral heads represented the remaining mapped potential and occupied cleaning stations. Adult gobies therefore occupied only the sponge X. muta significantly more than expected given the availability of each substratum species (Fig. 2a). All coral substrata were occupied significantly less than expected except for D. strigosa, which was occupied in proportion to its availability, but only on North Bellairs Reef (Fig. 2a).

Figure 2.

Selection ratios (+ 95% CI) for each sponge and coral substratum by adult (a), juvenile (b) and recruit E. prochilos (c) on North (filled bars) and South (open bars) Bellairs reefs. D. labyrinth. refers to Diploria labyrinthiformis. The horizontal line represents resource selection in exact proportion to its availability.

Maximum adult occupancy was 78% and 62% of the available sponges on North and South Bellairs, respectively (Fig. 1b). The proportion of adult-occupied sponges increased by 11% with an increase in total population size from February to a maximum in April on North Bellairs and by 18% from February to a maximum in June on South Bellairs. Both reefs experienced a decline in the number of adult-occupied sponges with a decline in the total population size between June and August (North Bellairs: 12%, South Bellairs: 34%, Fig. 1b). In contrast, the maximum number of adult-occupied coral heads was only 18 of 129 (14%) on North Bellairs and nine of 115 (8%) on South Bellairs. However, on both reefs the number of occupied coral heads almost tripled between the minimum and maximum population sizes in February and June (North Bellairs: 4·6–14·0%; South Bellairs: 2·6–8·0%, Fig. 1c).

In June, during the period of maximum observed occupancy of coral and sponge cleaning stations, there were few physical differences between adult-occupied and unoccupied cleaning stations. On South Bellairs adult-occupied cleaning stations were further from the edge of a patch/spur than unoccupied sponges (occupied: 115·95 ± 137·15 cm, unoccupied: 15·77 ± 60·75 cm, t29 = 2·92, P = 0·007) while coral cleaning stations were located on larger coral heads than unoccupied corals (occupied: 4370·75 ± 2242·25 cm2, unoccupied: 1841·25 ± 2286 cm2, t98 = 2·63, P = 0·01). However, there were no differences in the height of occupied or unoccupied coral heads from the reef or sand (t-tests: P > 0·05). No differences in any measured physical characteristic were found on North Bellairs (t-tests: P > 0·05).

Across both reefs the total number of adults was highly variable between sponges (ranges: February: one to nine adults, April: one to 15 adults, June: one to 22 adults, August: one to 13 adults) and also increased on each sponge with the population increase from February to June (Fig. 3a). In contrast, only one to three adults were observed per coral head in each of the censuses (Fig. 3a). Thus, on South Bellairs, 85–95% of the adult population occupied sponge rather than coral cleaning stations while on North Bellairs, 26–56% occupied sponges.

Figure 3.

Mean group sizes (+ SE) for adults (a) and recruits (b) on sponge (filled bars) and coral (open bars) cleaning stations across both North and South Bellairs reefs during each of the four census months.

Is there evidence of a buffer effect in adult distribution?

The number of adult gobies occupying sponge increased significantly with total adult population size (r2 = 0·96, N = 61, P < 0·001, Fig. 4a). By contrast, total population size was not related to the number of adults occupying coral (r2 = 0·08, N = 61, P = 0·28, Fig. 4a). The number occupying sponge therefore increased at a higher rate than the number occupying coral (comparison of regression coefficients: t118 = 15·89, P < 0·001).

Figure 4.

The relationships between total population size and the numbers of E. prochilos adults (a), juveniles (b) and recruits (c) occurring on sponge (closed circles) and coral (open circles) substrata (regression equations; adults: sponge: y = 0·95x − 4·88, coral: y = 0·05x + 4·88, juveniles: sponge: y = 0·85x − 1·33, coral: y = 0·15x + 1·33, recruits: sponge: y = 0·62x + 0·49, coral: y = 0·38x + 0·04).

patterns of juvenile distribution

Fine-scale patterns of juvenile distribution

The number of juveniles on North Bellairs Reef increased to a maximum in June (13 juveniles). By contrast, numbers of juveniles on South Bellairs Reef decreased from February (97 juveniles) to August (12 juveniles).

Juveniles occupied a broad range of substrata. No substratum could be identified as being occupied disproportionately more or less than predicted by its availability (Fig. 2b). During the peak numbers of observed juveniles in June on North Bellairs, only two of 13 juveniles (15·4%) were observed on X. muta. In contrast, on South Bellairs 77 of 97 juveniles (79·4%) were observed on X. muta during the peak in February. Juveniles on coral were most frequently alone; however, groups of up to five juveniles were observed. Juveniles occupied the same coral head as an adult in only two instances in both February, when juvenile numbers were highest, and June, when adult abundances were at their maximum. By contrast, juveniles occupied sponges in groups of one to 20 individuals alongside recruits and adults.

Is there evidence of a buffer effect in juvenile distribution?

The number of juvenile gobies occupying sponge increased significantly with total adult population size (r2 = 0·96, N = 61, P < 0·001, Fig. 4b). Similarly, the number of juveniles on coral increased with increases in total population size (r2 = 0·46, N = 61, P < 0·001, Fig. 4b). However, the rate of increase in juvenile numbers were significantly higher on sponge (comparison of regression coefficients: t118 = 15·17, P < 0·001).

patterns of recruit distribution

Fine-scale patterns of recruit distribution

A peak of recruitment was observed in February on North Bellairs Reef and April on South Bellairs Reef. The peak density of recruits on North Bellairs Reef was less than 50% of the peak density of recruits on South Bellairs Reef (North: 52 recruits, South: 129 recruits).

Recruits occupied a broad range of substrata. No substratum could be identified as being occupied disproportionately more or less than predicted by its availability as recruit preferences were variable between the two censuses (Fig. 2c). During the peak numbers of observed recruits in February on North Bellairs, only nine of 52 recruits (17·3%) were observed on X. muta. In contrast, on South Bellairs 96 of 129 recruits (74·4%) were observed on X. muta during the recruitment peak in April.

In February, during the period of maximum observed number of recruits, there were few physical differences between adult-occupied and recruit-occupied cleaning stations on North Bellairs. Adult-occupied cleaning stations were further from the edge of a patch/spur than coral heads occupied by recruits (adult-occupied: 50·42 ± 56·22 cm, recruit-occupied: 9·69 ± 25·26 cm, t26 = 2·58, P = 0·016). No differences were found on South Bellairs in April, when recruits were most abundant on this reef.

The number of recruits per sponge was highest at low total population sizes in February and decreased slightly through the year (Fig. 3b). By contrast, the number of recruits observed per coral reached a maximum at the highest recruit densities in April but remained low at all other times (ranges: February: 1–4; April: 1–23; June: 1–4; August: 1–2; Fig. 3b). Total group sizes on sponges, including juveniles and recruits, also increased with the population increase from February to April (February: 1–32; April: 1–52; June: 1–27; August: 1–17), while total group sizes on coral cleaning stations remained relatively small (February: 1–7; April: 1–6; June: 1–3; August: 1–2).

Is there evidence of a buffer effect in recruit distribution?

Both the number of recruits on sponge and the number of recruits on coral were significantly related to the total number of recruits (sponge: r2 = 0·83, N = 61, P < 0·001, coral: r2 = 0·62, N = 61, P < 0·001). Both substrata experienced an increase in the number of occupying recruits with increases in population size; however, the rate of increase was greater on sponges (comparison of regression coefficients: t118 = 3·69, P < 0·001, Fig. 4c).

experimental test of adult habitat choice

Following goby removal in June 2001, coral heads were re-colonized by adult E. prochilos before sponge substrata on five of 10 reef patches. There were no differences between patches from which all gobies had been removed and patches from which only adults were removed in the frequency with which coral was occupied first (χ2 = 1·67, d.f. = 1, P = 0·2). By comparison, on the first census in 2002, adult gobies were present on only five patches and occupied cleaning stations on coral heads (five fish) and sponges X. muta (two fish). In 2002, adults colonized coral heads before sponges on the remaining five patches. Similarly, there was no difference between patches from which all gobies and adults only were removed (χ2 = 0·48, d.f. = 1, P = 0·5).

Taking into account differing availability of coral and sponge between patches, the observed proportion of adults on coral did not differ significantly from expected values on any date (t-tests: P > 0·05 in all cases). There were also no significant differences between the proportion of adults on coral before experimental removals and on the first and last monitoring date in 2001 and 2002 (repeated measures anova: P > 0·05). However, on only three patches were the first corals to be colonized the same coral heads from which E. prochilos adults had previously been removed. On the remaining seven patches new coral heads were colonized first.

Across the total monitoring period, both the number of adults on sponge and the number of adults on coral were significantly related to the total number of adults (sponge: r2 = 0·43, N = 14, P = 0·01, coral: r2 = 0·66, N = 14, P < 0·001). Both substrata experienced an increase in the number of occupying adults with increases in population size; however, the rate of increase was greater on corals (comparison of regression coefficients: t24 = 4·70, P < 0·001, Fig. 5).

Figure 5.

The relationship between total population size on experimental patches and the numbers of E. prochilos adults occurring on sponge (closed circles) and coral (open circles) substrates (regression equations: coral, y= 0·62x+ 0·81; sponge, y= 0·39x+ 0·66).

adult characteristics in relation to substratum

In February, eight of 50 females (16%) and two of 11 males (18%) found across both reefs occupied coral heads and there were no significant differences in male and female sizes between sponge- and coral-occupying gobies (t-tests: P > 0·05 in both cases). Similarly, 10 of 76 females (13%) and eight of 42 males (19%) recorded in the June survey were observed on coral heads. In this month, there was also no size difference between males and females occupying sponge or coral (t-tests: P > 0·05 in both cases). Coral cleaning stations were most frequently occupied by a single cleaning goby (seven females and three males), although male–female pairs (three cleaning stations) and male–male pairs (one cleaning station) were observed. There were no size differences between mated coral-dwelling males and females and sponge-dwelling males and females or between unmated coral-dwelling males and females and sponge-dwelling males and females (t-tests: P > 0·05 in all cases). However, lone males on coral heads tended to be smaller than males paired with a female on a coral head (single males: 25·4 ± 1·9 mm, paired males: 28·7 ± 2·3 mm, t6 = 2·15, P = 0·07).

Discussion

For adult gobies, resource selection indices suggested a preference for sponge occupancy. However, as cleaning goby population size increased, the number of adult cleaning gobies occupying sponge increased more rapidly than the number occupying coral. By contrast, at low population densities on experimental patches, the rate of adult population increase was greater on coral than sponge. These patterns are not consistent with that predicted by a buffer effect if sponge habitat is of higher quality than coral. Instead, our results suggest that coral may be the preferred habitat, but in Barbados this habitat becomes saturated at very low population densities, thus a larger proportion of the population occupies sponge at most observed population densities.

A buffer effect should be expected only when all good-quality habitat is occupied. This was not the case for E. prochilos on sponge as even at peak population size, 20% of sponges remained unoccupied and these sponges did not harbour fewer numbers of the preferred prey, Haplosyllis spp. (Whiteman & Côté 2004). Our observational results could be interpreted as showing the pattern of occupancy expected prior to spillover into poorer habitat. However, even at very low population densities, a significant proportion of the population was observed on coral.

The observation of a remarkably consistent number of adults occupying coral instead suggests that the coral habitat becomes saturated at very low population densities and that, as population size increases, individuals are forced to occupy a lower-quality habitat, i.e. sponge. Indeed, at the very low densities resulting from goby removals the rate of population increase is initially higher on coral. Our study therefore provides the first evidence of a buffer effect for a site-attached and habitat-specialized marine species.

There were few physical differences between occupied and unoccupied coral heads, and following goby removals recolonizing adults rarely occupied previously occupied corals. Thus the number of occupied coral heads remained very low relative to the total number of coral heads apparently available. Cleaning gobies may not be limited by habitat directly but by the availability of food resources associated with coral occupancy. Client-gleaned items represent a significant proportion of the diet of coral-occupying gobies (Whiteman & Côté 2002a). However, ectoparasite loads in Barbados are low in comparison to other Caribbean locations (Losey 1974; Sikkel, Fuller & Hunte 2000; Cheney & Côté 2001), suggesting that the local capacity to support coral-dwelling gobies is limited. The spatial distribution of cleaning stations may therefore play a more important role in the choice of coral heads for cleaning station establishment than the physical characteristics of the coral. By contrast, the abundance of the polychaetes Haplosyllis spp. reported in Barbados (Whiteman & Côté 2004) is higher than elsewhere (Magnino & Gaino 1998; Magnino et al. 1999). This, in addition to the presence of unoccupied sponges with high polychaete densities, may also explain why there was no apparent saturation of sponge habitat even at the highest population densities.

Saturation of the coral habitat at low densities in this population may also partly be a consequence of competitive interactions with the sharknose cleaning goby E. evelyna, which occurs sympatrically with E. prochilos in Barbados. If, for example, E. evelynae is competitively dominant, the coral habitat available for E. prochilos may be restricted even further. These species differ in morphological characteristics which may affect the ability of each to switch to non-client-gleaned food sources when available. Colin (1975), for example, describes the inferior position of the mouth in E. evelynae as an advanced specialization for cleaning. E. prochilos may therefore be more likely or able to switch to alternative foods but the competitive relationships between these species have never been studied.

At large geographical scales there appears to be a general link between the density of coral-occupying gobies and both the density of client species and the density of ectoparasites that they harbour. For example, sharknose cleaning gobies which are observed only as lone gobies or as male–female pairs in Barbados, occur in large groups on coral heads in Curaçao, where ectoparasite loads and client densities are higher (K. Cheney, unpublished data). The local availability of client ectoparasites has also been linked to geographical variation in the diet of cleaner wrasses (Labroides spp.; Grutter 1997) but has rarely been extended to include consequences for social interactions or population-level effects. This large-scale mechanism may be operating on a much smaller scale on the Bellairs fringing reefs with significant consequences for the social and population structure of E. prochilos. Moreover, analyses of the distribution patterns of E. prochilos gobies in other locations may well reveal similar buffer effects and increase our understanding of the role of resources in generating disparate habitat use and variable social systems.

Precise habitat selection by settling larvae can also influence the distribution of juveniles and adults within habitats (e.g. Pomacentridae; Ohman et al. 1998; Gutiérrez 1998). Interestingly, the patterns of habitat occupancy with population size for recruits suggest only a small difference in habitat quality between sponge and coral. Selection ratios also showed that recruits and juveniles do not discriminate between sponge and coral. Wilson & Osenburg (2002) provided evidence that settlement of E. evelynae and E. prochilos to patch reefs in St Croix is density-dependent and suggested that observed discrepancies in the strength of density dependence may have resulted from a correlation between survival and habitat quality. Our results provide some support for this contention but also suggest that spatial dynamics and individual movements may be important. The population shift towards sponge rather than coral occupancy we observed between recruitment and maturity may arise either through differing survival of recruits and juveniles on coral and sponge or as a result of active movement of individuals towards sponge. Recruits occupy coral heads even when adult coral occupancy appears to be saturated suggesting that recruits were observed on poor-quality corals, although there were no clear differences in physical characteristics between adult-occupied and recruit-occupied coral heads. However, recruits could exploit different food resources than adults. For example, they may consume relatively fewer ectoparasites or more client mucus and thus client availability may not limit coral occupancy. In addition, in our study population the maximum density of recruits never exceeded greatly the maximum observed adult density on any site (E. Whiteman, unpublished data) suggesting that, although post-recruitment mortality undoubtedly occurs and may be linked to substratum choice, its role in shaping adult distribution may be limited. Active post-settlement movement of recruits toward sponges may be relatively more important in generating the dichotomous distribution of juveniles and adults.

Habitat choice in response to resource availability as predicted by a buffer effect appears to govern the distribution patterns of cleaning gobies. However, it remains unclear whether adults occupying sponges suffer fitness costs relative to those on coral. There were no size differences of males and females between sponge and coral, although the number of coral-dwelling gobies was low, perhaps obscuring any size differences between habitats. It is also possible that, although the specific costs and benefits of occupancy of each habitat may differ, the fitness of equally sized individuals in both habitats may be similar, on average. At low population densities, and therefore small group sizes, the fitness consequences of habitat choice may arise mainly from differences in foraging opportunities. Unfortunately, the energetic benefit of feeding on sponge-dwelling polychaetes Haplosyllis spp. vs. client-gleaned ectoparasites and mucus is not known. With increasing population size, intraspecific competition within sponge-dwelling groups should increase. Indeed, the presence of stable dominance hierarchies within these groups highlights the importance of such competition in creating within-group differences in foraging rates (Whiteman & Côté 2004). However, gobies occupying sponge may also experience a group-size fitness benefit that may compensate for the costs of increased competition. Such a mechanism could include reduced costs of finding a mate, through reduced travel or sampling costs (e.g. Halichoeres melanurus tailspot wrasse; Karino et al. 2000; Pomatoschistus minutus sand goby; Forsgren 1997). The interaction between costs and benefits involving both mating and foraging considerations may therefore make the calculation of the relative fitness of sponge- and coral-dwelling very complex.

In conclusion, our results show how dynamic patterns of distribution, particularly with population changes over time, can be informative as to the processes generating disparate use and the differences in quality between habitats. We provide the first evidence of the potential operation of a buffer effect for a site-attached marine species. Furthermore, our results demonstrate that interactions among individuals are an integral part of population distribution and dynamics and are therefore important in future studies of habitat choice and its associated fitness consequences.

Acknowledgements

We are grateful to the staff of the Bellairs Research Institute for their help during this study. We also thank Jenny Gill, Bill Sutherland and Karen Cheney for discussions and comments on the manuscript. Funding for fieldwork was provided by the John & Pamela Salter Charitable Trust and by the University of East Anglia. E.A. Whiteman was supported by a UK Biotechnology and Biological Sciences Research Council PhD studentship.

Ancillary