MARINE E C O L O G Y - PROGRESS SERIES

Mar. Ecol. h o g . Ser.

i Published August 30

Community patterns of coral-associated decapods

Nicholas J. Gotelli and Lawrence G. Abele
Department of Biological Science, Florida State University, Tallahassee, Florida 32306, USA

ABSTRACT: Community patterns of abundance and CO-occurrenceare described for decapod cmsta-
ceans associated with the coral Pocillopora damicornis in the Eastern Pacific Ocean. Most reef flank
(deep) corals have lower densities of decapods than reef flat (shallow) corals. This difference affects the
species-area relation for the 2 habitats. Seasonally, abundance is highest in June and April, lowest in
August and January. However, the species-area relation does not differ greatly through time. Most
coral heads support fewer species than expected if the individuals from a sample were distributed
randomly among heads. Principal components analysis did not reveal any distinct patterns of covaria-
tion among species' densities. Likewise, pairwise comparisons between common species revealed few,
if any, significant correlations of abundance or CO-occunence.These results suggest a considerable
degree of independence in the density and occurrence of decapod species associated with Pocillopora.

INTRODUCTION interspecific association. In this analysis, we compare

species-area relationships for different locations (flat
Decapod crustaceans are the most abundant macro- vs. flank following the terminology of Glynn, 1976) and
invertebrates inhabiting the branching coral Pocillo- seasons on the same reef. The effect of species' abun-
pora (Patton, 1974; Abele and Patton, 1976; see also dances on the species-area relationship is also as-
Reed et al., 1982 for a different coral community). sessed. This important factor is usually ignored in
Many of these decapod species are obligate symbionts species-area comparisons (Haila and Jarvinen, 1981).
of live pocilloporid corals and feed on coral mucus In addition to the area effect, interspecific interac-
(Knudsen, 1967; Patton, 1974). The fauna is relatively tions may influence patterns of abundance and co-
well known. Modern collections have been taken in occurrence. If competition were to influence cornrnu-
the Eastern Pacific Ocean (Abele, 1976a, b; Abele and nity structure, pairs of interacting species might be
Patton, 1976; Glynn, 1976; Abele, 1979; Castro, in negatively correlated or associated with each other.
press), Hawaii (Barry, 1965; Preston, 1971, 1973; Coles, Likely competitors in the Pocillopora community are
1980), Central Pacific Ocean (Garth, 1964; Knudsen, xanthid crabs of the genus Trapezia. Trapezia spp. are
1967), Indian Ocean (Garth, 1974), and Australian obligate associates of live pocillopond corals (Castro,
waters (Patton, 1966, 1974; Austin et al., 1980; Abele, 1976). Adult crabs often associate as male-female
in press). pairs (Preston, 1971, 1973), and individuals frequently
Glynn (1973, 1976) has made detailed studies of the migrate between coral heads (Castro, 1978).Both intra-
branching coral Pocillopora damicornis (Linnaeus) in and interspecific agonistic interactions have been
the Eastern Pacific. In one study (Glynn, 1976), 119 reported between adult pairs (Preston, 1971, 1973),and
heads of Pocillopora were collected from Uva Island, a in Hawaii, some species of Trapezia are negatively
patch reef in the Gulf of Chiriqui, Panama. Glynn associated in occurrence (Preston, 1973). Here, we
(1976) reported on the distribution and abundance of report indices of association and correlation (Hurlbert,
the most conspicuous symbionts, the xanthid crab 1969), which may reflect interactions between pairs of
Trapezia spp. and the snapping shrimp Alpheus lottini species (or indirect interactions that occur through the
Guerin. These species actively defend their coral host influence of a third species; Whittam and Siegel-
from attacks by the predatory crown-of-thorns seastar Causey, 1981).
Acanthasterplanci (Linnaeus) (cf. Glynn, 1976, 1980). Community patterns may be detected on a broader
Here, we examine Glynn's (1976) Uva Island collec- scale if groups of interacting species covary in their
tions with respect to 2 major community patterns: abundance or occurrence. These patterns would not
species-area relationships (Abele, 1976a, 1979) and necessarily emerge from pairwise comparisons of a few

O Inter-Research/Printed in F. R. Germany
 132 Mar. Ecol. Prog. Ser. 13: 131-139, 1983

species. A principal components analysis is one way of species, not to 1. Two of the species, T. ferruginea
detecting patterns of covariation among groups of Latreille and T, corallina Gestaecker, are very com-
species (Pielou, 1977). A principal components solu- mon, while T, Formosa Smith is very rare. Therefore, all
tion defines new orthogonal variables, which account of the January samples were increased by 1 species.
for variation in the densities of all species considered. The species-area analysis was repeated without the
The descriptive approach has a strong tradition in January data, and the results did not change. For the
community ecology (MacArthur, 1972). However, the partial correlation coefficients and the principal com-
dangers of overinterpreting non-experimental data are ponents analysis, only complete samples were used
becoming more apparent in community studies (Hair- (n = 58). The influence function, which identifies in-
ston, 1981; Lawton and Strong, 1981). Because this fluential cases in least-squares regressions, was com-
analysis is only descriptive, mechanistic explanations puted from the procedure in Weisberg (1980). Regres-
cannot be firmly attached to any patterns that emerge. sions, partial correlation coefficients, and the principal
Instead, the results should b e used to suggest hypo- components analysis were calculated by SPSS, version
theses that warrant closer study. 9.0. All computations were performed on the Cyber 760
computer, at Florida State University.

MATERIALS AND METHODS

RESULTS
The study site at Uva Island is a patch reef 100 m off
shore in the Gulf of Chiriqui (Glynn et al., 1972; Glynn, Patterns of species richness
1976). Sea surface temperatures are relatively constant
year-round. Seasonal upwelling has not been detected Species number is significantly correlated with both
at this site. Annual sea surface salinity ranges from 31 coral head size and abundance of individuals (Fig. l a ,
to 33 ppt, and sea surface temperature ranges from 28 b). Typically, these relationships are plotted on a log-
to 29 "C (Abele, 1976b). The Gulf of Chiriqui experi- log scale. However, for these data, a square root y vs.
ences distinct wet and dry seasons, with high rainfall, log,,x transformation provided a slightly better linear
increased sedimentation, and reduced salinity and vis- fit. These transformations were selected by examina-
ibility during the wet season, from May to November tion of residual plots. The log-log model gave similar
(Dana, 1975). The maximum tidal range is 5 m at Uva results, but the residuals did not appear as normally
Island, and reef flat corals occasionally suffer extensive distributed, so the square root transformation was
mortality from low-tide exposures (Glynn, 1976). favored. In this case, we wished to test more detailed
Pocillopora cover ranges from 40 to 80 % on the reef regression models, so a good linear fit was important.
flat, with near continuous cover on the reef crest. Pocil- Actually, both transformations describe a biologically
lopora cover on the reef flank drops off quickly with realistic curve that increases asymptotically as area is
depth. Almost no pocilloporid corals are found below increased.
6 m (cf. Glynn, 1976: Fig. 2). Although area is well correlated with species
Coral heads were collected by P. W. Glynn in 1973. number over several orders of magnitude, there is
Upon collection, corals were sealed in plastic bags and considerable scatter in the data; coral head size ex-
carried to the surface (Abele and Patton, 1976; Glynn, plains only 30 % of the variation in species number.
1976). L. G. Abele, in 1973, sorted the samples, mea- Some of the scatter may result because data from 4
sured the corals and identified the associated decapod different times and 2 different locations are plotted
crustaceans. All epifaunal organisms were removed together. In an attempt to explain more residual var-
from the live portions of the coral; animals associated iance, analysis of covariance models were tested that
with the deal coral base were excluded. A single include parameters for season, location, and density
endolithic gall crab, Haplocarcjnus marsupialis Stimp- (Table 1).
son, was collected, but not included in these analyses. There are no statistical differences in either the slope
Coral volume was estimated by l x wx h, measured in or the intercept of the species-area relationship at
centimeters. The collection of 119 coral heads yielded different times. However, a significant pattern was
52 species and 4724 individuals of decapods. Coral detected for differences in location. The slope of the
heads were collected from the reef flank and reef flat in reef flank samples is significantly steeper than that of
January (flank n = 28, flat n = 28), April (flank n = 12, the reef flat samples (F1,,,,= 5.07, p < .05). The inter-
flat n = l ? ) , June (flank n = 12, flat n = 12) and August cepts do not differ significantly (F,,,,, = 1.89, p > .05).
(flank n = 5, flat n = 5). Inclusion of the slope term adds about 10 % more
Unfortunately, the January samples were pooled variance explained in species number.
before it was realized that the Trapezia belonged to 3 For the species-area relation, most observations had
 Gotelli and Abele: Community patterns of decapods

Fig. 1. (A) Species-area plot for

Pocillopora decapods; regression
equation: square root (species
+
number) = 0.121 0.907 (log,,,
coral head size); R2 adjusted
= 0.30. (B) Species-individuals
plot for Pocillopora decapods; re-
gression equation: square root LOG,, ( Area ) LOG,, ( Abundance 1
(species number) = 1.08 1.33 +
(loglo abundance); R2 adjusted
= 0.61. (C) Individuals-area plot
for Pocillopora decapods; regres-
sion equation: square root (abun-
+
dance) = - 8.01 4.28 (log,,, cor-
al head size); R2 adjusted = 0.44.
(D) Species-area regressions for
reef flank and reef flat samples.
Reef flank: Y = - 1.14 + 1.27 X;
R2 adjusted = 0.38. Reef flat:
+
Y = 0.868 0.691 X ; adjusted
= 0.25. Open circles = reef flat
samples; closed circles = reef
flank samples. (E) . . Individuals-
area regressions for reef flat and LOG,, ( Area 1
LOG,, ( Area 1
reef flank samples. Reef flank:
+
Y = - 13.80 5 . 8 7 ~ R2; adjusted
= 0.56. Reef flat: Y = - 5.29 + -
E
3 . 5 2 ~ R2
; adjusted = 0.41. Open
circles = reef flat samples; closed 11.0-
circles = reef flank samples. (F)
Factor loadings for a principal
5
U
C
components analysis of species 2 8.0-
densities; factor loadings on the 4
first (PCI) and second (PCII) axes W

are shown. Circles = palae- l-

5.0-
monids; squares = majids; trian-
0
gles = alpheids; open hexagons V)
= xanthids; closed hexagons = o
porcellanids. 1 = Trapezia fer- 2.0- 1 I I I I I
ruginea; 2 = Trapezia corallina; 2.00 3.00 4.00
3 = Trapezia formosa LOG,, ( Area 1

relatively little influence on the observed slope and Heterogeneous slopes are troublesome in analyses of
intercept. However, 2 cases were especially influen- covariance, because the treatment effect depends on
tial. One case is a large coral head (11,340 cm3) from the value of the covariate. The Johnson-Neyman
the April flank samples. The predicted number of method (Johnson and Neyman, 1936; Huitema, 1980)
species from the regression is 14, but the observed was used to evaluate the heterogeneous slopes and
number is only 8. The other case is also a large coral estimate a non-significance xinterval (Table 2). Within
head (10,875cm3) from the June flank samples. The this interval, species richness of the flank and flat
predicted number of species from the regression is 14, samples does not differ significantly. Above the inter-
but the observed number is 20. After these 2 cases were val, reef flank samples (with a steeper slope) have
deleted, the analysis of covariance was recomputed. more species; below the interval, reef flat samples
Again, there were no significant seasonal effects, but (with a flatter slope) are more species rich. The non-
the F ratio for locational differences in slope increased significance X interval ranges from 2.34 to 6.82 (log
from 5.07 (p < .05) to 7.72 (p < .01).This result strength- scale), which encompasses the range of collected data.
ens the assertion that the species-area regression is Thus, although the flank and flat regression slopes are
significantly steeper for reef flank samples. unequal, average species richness does not differ over
 Mar. Ecol. Prog. Ser. 13: 131-139, 1983

Table 1. Analyses of covariance results for seasonal and locational effects on abundance and species number. (Influential cases
deleted)

Y-variable Covariate (S) Factor n Test for dif- Test for dif-
ferences among ferences among
slopes intercepts
(F values) (F values)

Species Area Season 117 .43[3.109) 2.29(3,1~2]

Number Location 7.72(1,1131.. 1.91(1,114,
Species Abundance Season (3,110) 1.88y,~~3)
Number Location 2.24(~,~~4~ 1.73(~,~~s]
Abundance Area Season .32(3,~io) 4.51(3,113)"
Location 5.03(1,11,). 9.66[1,115)..
Species Area, Season .81(6,105) 1.98(3,~~~)
Number Abundance Location 1.72(z,~~~l 1.51(~,~~3)

Significance of F values: ' p < .05

" pe.01

Table 2. Contrasts and non-significance intervals for the range of coral head sizes collected. The flank and
ANCOVA results flat samples differ statistically, but probably not
I I biologically; the two habitats have similar numbers of
Adjusted means for abundance/coral head (values in
parentheses estimate untransformed densities) species per coral head.
- The slope and intercept of the species-individual
Season y j
regression differ neither seasonally nor by location
Jan 5.64 (31.78) (Fig. lc). However, there are strong differences in the
A P ~ 6.63 (43.94) individuals-area regressions for the different samples.
Jun 6.49 (42.06) Seasonally, the slopes are equal, but the intercepts
Aug 5.71 (32.58) differ (Table 1). Density was highest in April (y = 43.9
Pairwise comparisons of adjusted mean abundances. A individuals/coral head) and June (Y= 42.1) and lowest
significant value indicates that the 2 samples differ in
mean abundance/coral head
in January (Y= 31.8) and August (v = 32.6). Average
density of dry season samples (January and April) did
l Contrast Fisher's protected least not differ significantly from that of wet season samples
significant difference
(June and August). The contrast of April-June samples,
however, was significantly different from that of Janu-
Jan-Apr
Jan-Jun ary-August samples (Table 2).
Jan-Aug For the flank-flat comparison, both slopes and inter-
Apr-Jun cepts differ significantly (Table 1). The Johnson-Ney-
Apr-Aug man non-significance interval for the individuals-area
Jun-Aug
regression is 2,570 cm3 to 35,480 cm3. Thus, for coral
( Scheffe's contrasts of adjusted mean abundances heads smaller than 2570 cm3, reef flat samples had

I January-April/June-August
(dry season) (wet season)
significantly more individuals than reef flank samples.
Of the coral heads 6 7 % were below this size. The
upper end of the non-significance interval (35,480 cm3)
exceeds the largest coral head collected. Therefore,
Johnson-Neyman non-significance intervals for hetero-
geneous slopes of flank-flat comparison. Within this X large reef flank heads do not contain significantly
interval, the flank and flat data do not differ significantly more individuals than large reef flat heads.
I Model Interval I It is noteworthy that the individual-area regressions
cross in the same fashion as the species-area regres-
sions (Fig. I d , e). Thus, the heterogeneous slopes of the
species-area relation may be caused by differences in
abundance.
To test this idea, we repeated the analysis, using
both area and the number of individuals as covariates.
 Gotelli and Abele: Community patterns of decapods 135

Although the covariate term is significant, slope and Table 3. Observed and expected species number per coral
intercept differences are not (Table 1). ln particular, head. Expected values and a 95 % confidence interval are
generated by rarefaction for each sample. Numbers of coral
note that the significant difference in slopes for the
heads with significantly fewer species are given in paren-
species-area relations (F,,,,, = 7.72, p < is elimi- theses (p < .05, 2-tailed); no significant positive deviations
nated when abundance of individuals is included in were found
the model (F,,,,, = 1.72, p > .05). A locational differ-
ence in the individuals-area relation produces a cor- Samples Observed species Observed species
responding difference i n the species-area relation. The number < expected number > expected
species-area regression is steeper for reef flank corals,
January Flat 25 (10) 3
in part, because the individuals-area regression is also January Flank 25 ( 6 ) 3
steeper.
April Flat 16 ( 6) 1
Thus, the simplest model accounting for species
April Flank 12 ( 4) 0
richness includes only coral head size and abundance
of individuals. This multiple regression explains 61 % June Flat 9 ( 3) 3
June Flank 1 1 ( 6) 1
of the variation in species number. However, this is no
greater than a simple regression of species on indi- August Flat 3 ( 0) 2
August Flank 4 ( 1) 1
viduals. Furthermore, the coefficient for area in the
multiple regression does not differ significantly from Total l05 (36) 14
0 (p = .449). These results suggest that the most impor-
tant proximate variable affecting species number is the
abundance of individuals. Area may be of importance axis accounted for only 21 % of the variation in density.
through its effect on abundance. All of the factor loadings were positive, which might
be expected because species abundances are usually
correlated positively with coral head size. In fact, log,,
Rarefaction
coral head size accounts for 38 % of the variance
(p < .001) in the scores on the first principal compo-
To evaluate further the species-individual relation,
nent. Subsequent axes account for little residual var-
we generated rarefaction curves. Rarefaction estimates
iance. Four axes are required to explain 5 0 % of the
the expected number of species in a small sample of
variance, and nine axes to explain 75 %.
individuals drawn randomly from a larger collection
The factor loadings for the different species reveal
with a given species-individual distribution (Hurlbert,
no obvious patterns on the first 2 axes (Fig. If). This
1971; Simberloff, 1978). The data from all coral heads
result is expected, because of the small amount of
collected in one region and at one time were pooled to
variance explained by the first 2 principal component
generate rarefaction curves. These curves give the
axes. Note that Trapezia ferruginea and T. corallina,
expected number of species (and a 95 % confidence
which are significantly correlated in abundance
interval) in each coral head if all individuals in the
(r = .42, p = .002), have similar loadings on the first 2
sample were distributed randomly among heads.
axes. Plotting the data by samples did not reveal any
These curves were compared to the actual number of
obvious seasonal or locational patterns.
species observed in the heads.
A principal components analysis was also performed
In general, the data are poorly fit by the rarefaction
on presence-absence data for all species. Because
curves (Table 3). For all samples, observations consist-
many species are rare, this technique gives more equal
ently fall below the expected values. Thirty-six of the
weight to all species, whereas the PCA on abundances
119 coral heads had significantly fewer species than
reflects the fluctuations of a few common species. Pre-
expected (p < ,025). Only 14 coral heads had more
sence-absence data violate the assumption of mul-
species than expected, and none of these observations
tivariate normality, but the violation may not b e impor-
was significant. Thus, the distribution of species is
tant unless statistical tests are performed (Green,
'clumped'. Coral heads have too few species if all
1979).
individuals from a collection were distributed ran-
For the presence-absence data, the first axis extracts
domly among heads.
only 11 % of the variance. Eight axes are required to
explain 50 % of the variance, and 15 axes to explain
Covariation among groups of species 75 %. Unlike the abundance analysis, the first princi-
pal component for the presence-absence data has both
To examine covariation among species, a principal positive and negative loadings. Both the first and the
components analysis was performed on the loglo second principal components are weakly correlated
abundances of the 23 most common species. The first with coral head size (PC1 r = .25, p < .05; PC11 r = .35,
136 Mar. Ecol. Prog. Ser. 13: 131-139. 1983

p < .01). Again, no obvious patterns emerge from the all the data, only 2 comparisons were significant
factor loadings on the first 2 axes. (p < .05, 2-tailed) and both were positive correlations
(Table 4). The congeneric xanthids Trapezia fer-
ruginea and T. corallina were positively correlated.
Covariation among species pairs T. corallina was also positively correlated with the
snapping shrimp Alpheus lottini. None of the other 19
To study patterns of interspecific association, corre- pairs was significant in either direction.
lation coefficients of abundance were calculated for However, the results are different if only those sam-
selected species pairs. This statistic reveals whether ples with both species present are used. In this case, 2
the abundances of pairs of species covary positively, significant negative correlations also appear (Table 4).
negatively, or independently of each other. Hurlbert The congeners Trapezia corallina and T. formosa are
(1969) warns that correlation coefficients may be mis- negatively correlated. T. forrnosa is also negatively
leading because they reflect both covariation among correlated with the palaeomonid shrimp Fennera
species densities (correlation) and the degree of associ- chacei Holthuis. T. corallina is positively correlated
ation between species (as measured by presence-ab- with both T. ferruginea and the shrimp Harpiliopsis
sence data). In addition, abundance data usually vio- depressus (Stimpson). Although some of the signifi-
late the assumption of normality required for correla- cance levels change, the patterns are similar if the data
tion. are divided by season or location.
However, the correlation coefficient is still a useful Statistical tests are problematic for these sorts of
measure of covariation, because confounding variables data. With 21 comparisons, at least 1 or 2 significant
can be factored out. For example, most species abun- results are expected by chance alone. If the signifi-
dances increase with coral head size, so positive corre- cance level is appropriately reduced to maintain an
lations between densities may be spurious. A partial overall error rate of .05, then only the positive correla-
correlation coefficient, factoring out the effect of coral tion between Trapezia ferruginea and T. corallina is
head size, would more accurately reflect the degree of significant.
covariation between the species. In addition, the effect Overall, species abundances do not appear strongly
of other species can be partialled out of any pairwise correlated. In particular, note that negative correla-
comparison. tions only emerge on coral heads where both species
Correlations among the seven most common species are present. Total abundances covary positively in 2
were examined, using log densities to meet better the cases, independently in all the rest. Positive correla-
assumption of normality. For each pairwise compari- tions are just as frequent as negative ones, and both
son, the effects of coral head size and the abundance of congeneric and non-congeneric pairs may be signifi-
the other 5 species were statistically partialled out. For cantly correlated. Patterns of covariation do not appear

Table 4. Partial correlation coefficients of abundances of pairs of common species. Coral head size and the abundance of other
species factored out; all samples used. (Only samples with both species present used for values in parentheses)

Trapezia Trapezia Alpheus Harpiliopsis Fennera Teleophrys

corallina formosa lottini depressus chacei cristulipes

Trapezia .42" 14 .05 .02 .l8 .21

ferruginea (.40)' ' (-.l l ) (.05) (c151
Trapezia -.04 .31' .24 -.l8 .l6
corallina (-.42) ' (.23) (.31) (--06) (.06)
Trapezia .21 14 11 -.05
formosa (.lO) b31) (- 54) ' ' ( 24)

Alpheus .05 .l3

lottini -.03
(-.08) (-.l41
(-.05)
Halpiliops~s .25 -.OS
depressus (.24) (-.09)
Fenn era .05
chacei (.01)

' <.05
" < .01
 Gotelli and Abele: Community patterns of decapods

to change much in different seasons or different loca- areas of reef flat corals may be killed by lengthy low-
tions. tide exposures (Glynn, 1976). This mortality may affect
Hurlbert's (1969)C, index was calculated to measure the community structure of reef flat decapods. The
species associations. This statistic measures associa- fauna associated with dead Pocillopora differs in both
tion based upon presence-absence data, whereas the associated species richness and identity from that of
correlation coefficient measures the degree of covaria- live coral (Abele, 1976b; Coles, 1980).
tion between the densities of species. As measured by Coral eating fishes are very common on the reef
C,, most associations of species pairs are very close to flank at Uva (Glynn et al., 1972) and may affect the
0. Trapezia ferruginea and T. corallina are not associ- densities of decapod associates. In laboratory experi-
ated with any other species (C8 = 0) because they are ments, Castro (1978) showed that Trapezia migration
present on every coral head. T. formosa and Fennera was considerably reduced in the presence of carni-
chacei show weak positive association (C, = 0.34, vorous fish. Thus, predation could account for lower
p > .05) but covary negatively in their abundances abundances on small reef flank corals. Decapod associ-
(r = - 0.54, p = ,009). Association indices were also ates of Stylophora in the Red Sea are also less common
calculated for some congeneric pairs of synalpheid on reef flank corals (Edwards and Emberton, 1980).
shrimps and porcellanid crabs (Table 5). The con- At Uva, seasonal differences in density are evident,
but did not translate into differences in species rich-
Table 5. Association indices (C,) for some congeneric pairs of ness; the species-area relation remains constant
species throughout the year. Moreover, these density changes
are not coincident with the wet-dry season cycle in the
Harpiliopsis spinigems X Harpiliopsis depressus -.06 Eastern Pacific. This may not be so surprising, because
Petrolisthes edwardsi X Petrolisthes haigae .06 physical conditions in the Gulf of Chiriqui are rela-
Petrolisthes edwardsi X Petrolisthes glasselli .OO tively constant (Abele, 1976a); seasonal upwelling
Petrolisthes haigae X Petrolisthes glasselli .OO does not occur, and annual temperature changes are
Synalpheus charon X Synalpheus mexicanus .26' slight, though significant (Dana, 1975).
Synalpheus charon X Synalpheus diguieti -09 It is important to note the effects of including abun-
Synalpheus mexicanus X Synalpheus diguieti .OO dance in the species-area model. Because abundance
' pc.05 directly influences the shape of the species-area
regression, analyses based entirely on presence-ab-
sence data may be incomplete. Geographic compari-
geners Synalpheus charon (Heller) and S. cf. mex- sons of species-area regressions may be confounded by
icanus are positively associated (C, = 0.26, p < .05), density effects.
but all other comparisons are close to 0. On the spatial scale of a single coral head, species
interactions may also affect patterns of CO-occurrence
and abundance. Several 'neutral models' (sensu Cas-
DISCUSSION well, 1976) were examined with the hope of detecting
such interactions. Rarefaction curves suggest that most
For decapods inhabiting Pocillopora, the species- coral heads have fewer species than expected if indi-
area relation is a function of both reef location and viduals from a sample were distributed randomly
density of individuals. Small coral heads support more among heads. Interspecific competition could produce
individuals on the reef flat than on the reef flank, and this result, although alternative mechanisms are
this pattern influences the shape of the species-area equally plausible (Dillon, 1981). For the Pocillopora
regression. One explanation for this finding is pro- community, a partial explanation is that coral heads
vided by Preston (1962). Reef flank corals are patchily are colonized by male-female pairs, rather than single
distributed and might behave like true 'isolates' with individuals of a species. All species of Trapezia, the
low migration of decapods between coral heads. Coral snapping shrimp Alpheus lottini, and some species of
cover is more continuous on the reef flat, where the genus Synalpheus associate as male-female pairs
increased decapod migration could produce a shal- on coral heads (Abele, 1979).
lower species-area regression. However, turnover rates A principal components analysis was used to detect
have never been determined for decapods of Pocillo- any major patterns of covariation among species'
pora. It would be premature to suggest that equilib- densities. The low percent of variation explained by
rium is maintained as in the MacArthur-Wilson (1967) any single axis suggests that most species covary inde-
model. pendently in abundance. Our findings are similar to
Furthermore, the reef flat and reef flank at Uva are those of Shulenberger (1980) for hyperiid amphipod
subject to very different abiotic regimes. Extensive assemblages and Wiens and Rotenberry (1981) for
 Mar. Ecol. Prog. Ser. 13: 131-139, 1983

shrubsteppe bird communities. Neither study detected Abele. L. (1979). The community structure of coral-associated
any major patterns of covariation in species' abun- decapod crustaceans in variable environments. In: Living-
ston, R . J. (ed.)Ecological processes in coastal and marine
dances. In this case, the first component of variation systems. Plenum Press, New York, p. 265-287
seems to be a function of coral head size. Abele, L. G . (in press). Biogeography, colonizat~on and
Association and correlation indices also suggest that experimental structure of coral-associated crustaceans. In:
most species covary independently of each other. Few Strong, D. R., Simberloff, D. S., Abele, L. G., Thistle, A. B.
significant correlations emerged from 21 pairwise (ed.) Ecological communities: conceptual issues and the
evidence. Princeton University Press, Princeton, N. J.
comparisons. For the 3 Trapezia species, 1 pair was
Abele, L. G., Patton, W. K. (1976). The size of coral heads and
positively correlated in abundance, and 1 pair nega- the community biology of associated decapod crustaceans.
tively correlated. Association was near 0 because J . Biogeogr. 3: 35-47
T. fermginea and T. corallina are found on every coral Austin, A. D., Austin, S. A., Sale, P. F. (1980). Community
head. structure of the fauna associated with the coral Pocillopora
damicomis (L.) on the Great Barrier Reef. Aust. J. mar.
These findings contrast sharply with Preston's (1973) Freshwat. Res. 31: 163-174
analysis of the 5 species of Trapezia in Hawaii. Preston Barry, C. K. (1965). Ecological study of the decapod crusta-
(1973) found 7 out of 10 pairwise comparisons had ceans commensal with the branching coral Pocillopora
negative values for C,. This result was obtained meandrina var. nobilis Verrill. M.S. thesis, University of
Hawaii, Honolulu, Hawaii
because Trapezia species in Hawaii are preferentially
Castro. P. (1976). Brachyuran crabs symbiotic with scleracti-
associated with different sized coral heads (Preston,
nian corals; a review of their biology. Micronesia 12:
1973). Thus, C, values are confounded by size effects 99-110
and may not be indicative of species interactions. In Castro, P. (1978). Movements between corals colonies in
contrast, the Trapezia species in the Eastern Pacific Trapezia ferruginia (Crustacea: Brachyura), a n obligate
Ocean can be found on all sizes of coral heads and symbiont of scleractinian corals. Mar. Biol. 46: 237-245
Castro, P. (in press). Notes on symbiotic decapod crustaceans
have association indices (C,) of 0. These differences from Gorgona Islands. Columbia, with a preliminary revi-
between Hawaiian and Eastern Pacific Trapezia are sion of the eastern Pacific species of Trapezia (Brachyura,
striking. The Pocillopora habitat is similar, and 2 of the Xanthidae), symbionts of the scleractionian corals. An.
species ( T . ferruginea and T.digitalis) are present in Inst. Inv. Mar. Punta Betin 12
Caswell, H. L. (1976). Community structure: a neutral model
both areas. However, Pocillopora heads are widely
analysis. Ecol. Monogr. 46: 327-354
scattered in Hawaii (Preston, 1973) and form dense Coles, S. L. (1980). Species diversity of decapods associated
reefs in the Eastern Pacific (Glynn, 1976). with living and dead reef coral Pocillopora meandrina.
Both inter- and intraspecific interactions have been Mar. Ecol. Prog. Ser. 2: 281-291
reported for Pocillopora decapods (Preston, 1971; Las- Dana, T. F. (1975). Development of contemporary Eastern
Pacific coral reefs. Mar. Biol. 33: 355-374
sig, 1977; Castro, 1978).Yet, most species show consid-
Dillon, R. T. (1981). Patterns in the morphology and distribu-
erable independence in their patterns of abundance tion of gastropods in Oneida Lake, New York, detected
and CO-occurrence.Either species interactions are not using computer-generated null hypotheses. Am. Nat. 118:
strong enough to influence these patterns, or the 83-101
interactions are complex and are not manifest in these Edwards, A., Emberton, H. (1980). Crustacea associated with
the scleractinian coral. Stylophora pistillata (Esper),in the
statistics. An experimental study is needed to distin- Sudanese Red Sea. J. exp. mar. Biol. Ecol. 42: 225-240
guish between these alternatives. An obvious ap- Garth, J. S. (1964). The Crustacea Decapods (Brachyura and
proach would be to manipulate species pairs with Anomura) of Eniwetok Atoll, Marshall Islands, with spe-
negative association or correlation indices. These cial reference to the obligate commensals of branching
corals. Micronesia 1: 137-144
manipulations would be most likely to reveal any
Garth, J. S. (1974). Decapod crustaceans inhabiting reef-
interspecific effects. building corals of Ceylon and the Maldive Islands. J. mar.
biol. Ass. India 15: 195-212
Acknowledgements. We thank Peter Glynn for access to the Glynn, P. W. (1973). Acanthaster effect on coral reef growth
Uva Island collections. Don Strong, Joe Travis, Craig Young in Panama. Science, N. Y. 180: 504-506
and 3 anonymous reviewers provided useful comments on the Glynn, P. W. (1976). Some physical and biological determin-
manuscript. ants of coral community structure in the eastern Pacific.
Ecol. Monogr. 46: 431-456
Glynn, P. W. (1980). Defense by symbiotic crustacea of host
LITERATURE CITED corals elicited by chemical cues from predators. Oecologia
47: 287-290
Abele, L. G. (1976a). Comparative species richness in fluc- Glynn, P. W., Stewart, R. H., McCosker, J. E. (1972). Pacific
tuating and constant environments: coral-associated coral reefs of Panama: structure, distribution, and pre-
decapod crustaceans. Science, N. Y 192: 461-463 dators. Geol. Rdsch. 61: 483-519
Abele, L. G. (1976b). Comparative species composition and Green, R. H. (1979). Sampling design and statistical methods
relative abundance of decapod crustaceans in marine for environmental biologists. John Wiley and Sons, New
habitats of Panama. Mar. Biol. 38: 263-278 York
 Gotelli and Abele: Community patterns of decapods

Haila, Y., Jarvinen, 0. (1981).The underexploited potential of Pielou, E. C. (1977). Mathematical ecology. John Wiley and
quantitative bird censuses in insular ecology. Studies in Sons, Inc., New York
Avian Biology 6: 559-565 Preston, F. W. (1962). The canonical distribution of common-
Hairston, N. G. (1982). An experimental test of a guild: sala- ness and rarity. Ecology 43: 185-215; 410-432
mander competition. Ecology 62: 65-72 Preston, E. M. (1971).Niche overlap and competition among
Huitema. B. E. (1980). The analysis of convariance and alter- five syrnpatric congeneric species of xanthid crabs. Ph. D.
natives. John Wiley and Sons, Inc., New York thesis, University of Hawaii. Honolulu, Hawaii
Hurlbert, S. H. (1969). A coefficient of interspeciflc associa- Preston, E. M. (1973). A computer simulation of competition
tion. Ecology 50: 1-9 among five sympatric congeneric species of xanthid crabs.
Hurlbert. S. H. (1971). The nonconcept of species diversity: a Ecology 54: 4 6 9 4 8 3
critique and alternative parameters. Ecology 52: 577-586 Reed, J. K., Gore, R. H., Scotto, L. E., Wilson, K. A. (1982).
Johnson, P. O., Neyman, J. (1936). Tests of certain linear Community composition, structure, area1 and trophic rela-
hypotheses and their application to some educational tionships of decapods assoc~atedwith shallow- and deep-
problems. Statist. Res. Mem. Univ. Coll. Lond. 1: 57-93 water Omlina varicosa coral reefs: studies on the decapod
Knudsen. J. W. (1967). Trapezia and Tetralia (Decapoda, crustacea from the Indian River region of Florida, XXIV.
Brachyura, Xanthidae) as obligate ectoparasites of pocil- Bull. mar. Sci. 32: 761-786
loporid and acroporid corals. Pacif. Sci. 21: 51-57 Shulenberger, E. (1980). Factor analyses of a hyperiid
Lassig, B. R. (1977). Communication and coexistence in a amphipod assemblage from the North Pacific Central
coral community. Mar. Biol. 42: 85-92 Gyre. Mar. Ecol. Prog. Ser. 2: 109-120
Lawton, J. H., Strong, D. R., Jr. (1981). Community patterns Simberloff, D. S. (1978). Use of rarefaction and related
and competition in folivorous insects. Am. Nat. 118: methods in ecology. In: John Cairns, Jr., Dickson, K. L.,
317-338 Livingston, R. J . (ed.) Biological data in water pollution
MacArthur, R. H. (1972). Geographical ecology: patterns in assessment: quantitative and statistical analyses, ASTM
the distribution of species. Harper and Row, New York STP 652 (Am. Soc. for Testing and Materials), p. 150-165
MacArthur, R. H., Wilson, E. 0. (1967). The theory of island Weisberg, S. (1980). Applied linear regression. New York.
biogeography. Princton University Press, Princeton, N. J . Wiley
Patton, W. K. (1966). Decapod crustacea commensal with Whittam, T. S., Siegel-Causey, D. (1981). Species interactions
Queensland branching corals. Crustaceana 10: 271-295 and community structure in Alaskan seabird colonies.
Patton, W. K. (1974). Community structure among the animals Ecology 62: 1515-1524
inhabiting the coral Pocillopora damicomis at Heron Wiens, J . A., Rotenbeny. J. T. (1981). Habitat associations and
Island, Australia. In: Vernberg, W. B. (ed.) Symbiosis in community structure of birds in shrubsteppe environ-
the sea. University of South Carolina Press, Columbia, p. ments. Ecol. Monogr. 51: 2 1 4 1
219-243

This paper was submitted to the editor; it was accepted for printing on June 10, 1983