232 
H.S. Gill, P. Humphries 
individuals of each species. Although the densities 
recorded in the field were generally at the lower 
end of the range used for these trials, i.e., 10 
individuals nr 2 (Humphries and Potter 1993), it 
was considered important to use at least this 
density to provide a sufficiently large expected 
frequency for statistical purposes (Sokal and Rohlf 
1981). Trials with all combinations of species were 
conducted: i.e., P. olorum/A. suppositus, P. olorum/F. 
lateralis, A. suppositus/F. lateralis, P. olorum/A. 
suppositus/F. lateralis. Each fish was only used once 
in the experiments. 
The G-statistic was used to test results for 
significant deviation from expected frequencies. 
This test being theoretically superior and 
computationally simpler than the y 2 test (Sokal and 
Rohlf 1981). Readings and replicates were tested 
for homogeneity using an 'interaction' or 
'heterogeneity' G-test. If all readings were 
homogeneous, the results were pooled. In only five 
of the 78 (6.4%) trials carried out were the 
replicates not homogeneous. In single species trials, 
observed frequencies were tested against expected 
frequencies of 50:50 using a goodness of fit G-test 
and a pooled G-statistic was calculated. For mixed 
species trials, observed frequencies were tested 
against the frequencies obtained using the same 
number of fish as in single species trials, using a G- 
test for independence. This type of comparison 
means that the only difference between the single 
species experiments and those involving mixed 
species was the addition of another species. 
When on their own A. suppositus showed a very 
strong preference for the seagrass, irrespective of 
the densities of fish or the time of day. A similar, 
but less pronounced, choice was generally exhibited 
by F. lateralis. Pseudogobius olorum showed no 
consistent preference for either habitat. In mixed 
species experiments, A. suppositus maintained its 
strong preference for seagrass, whereas F. lateralis 
reversed its choice of habitat to that of sand in the 
presence of A. suppositus. Pseudogobius olorum 
showed a clear preference for a particular habitat, 
namely seagrass, only when all three species were 
together. In all combinations all species sought 
refuge in the seagrass when disturbed. 
In these laboratory experiments, the habitat 
choice of A. suppositus and P. olorum when on their 
own paralleled their distribution in the field, i.e., in 
both the laboratory and the field A. suppositus 
showed a very strong preference for the artificial 
seagrass, while in the case of P. olorum the lack of 
any consistent preference for either seagrass or 
sand in the laboratory is consistent with the 
greatest densities for this species being recorded 
from patchy Ruppia habitat in the field (Humphries 
and Potter 1993). In contrast, although F. lateralis 
chose artificial seagrass in the laboratory it was 
only ever recorded at low densities in Ruppia in the 
field (Humphries and Potter 1993), while in other 
systems both Shaw (1986) and Bell and Westoby 
(1986) reported that F. lateralis was typically 
associated with a bare sand habitat. 
In the mixed species experiments, the choice of 
habitat by each of these species was significantly 
influenced by the presence of other species. The 
most dramatic effect was seen with F. lateralis, 
which showed a significant increase in the 
utilisation of sand in the presence of A. suppositus. 
This parallels the distribution of these species in 
the field and also the situation reported by 
Wiederholm (1987) for another goby species 
(Pomatoschistus microps), which when alone utilised 
artificial vegetation, but in mixed species 
experiments and in the field was found in an open 
habitat. However, although Wiederholm (1987) 
reported a shift in habitat utilisation and 
considered that the small size of P. microps may 
make it vulnerable to displacement by larger, more 
aggressive species, other workers have found that 
differences in habitat use are maintained both in 
the absence and presence of species at the same 
trophic level and are probably more a function of 
differences in morphology and species-specific 
responses to environmental factors (Schlosser and 
Toth 1984). A comparison of the body lengths in 
the present study suggests that size alone cannot 
account for changes in habitat choice. While A. 
suppositus is on average the largest of the three 
goby species and therefore could conceivably win 
agonistic interactions with F. lateralis if size was the 
sole criterion, it did not affect the habitat choice of 
P. olorum, the smallest of the three species. 
Furthermore, the presence of F. lateralis appeared 
to lead to a greater number of A. suppositus 
entering the artificial seagrass. Thus, there would 
appear to be a mutual interaction between F. 
lateralis and A. suppositus. The fact that A. suppositus 
shows a very strong preference for seagrass would 
probably make it difficult to displace from this 
habitat. Humphries and Potter (1993) reported that 
not only did Afurcagobius suppositus and F. lateralis 
share several dietary taxa but that they were one of 
only two pairs amongst three gobiid and three 
atherinid species whose diets were often 
correlated. Moreover, the similar morphology of 
the terminal/superior mouths, along with short 
guts and ambush-type feeding strategy of A. 
suppositus and F. lateralis contrast with the sub¬ 
terminal mouth, long gut and active omnivorous 
foraging mode exhibited by P. olorum (Gill and 
Miller, 1990; Gill, 1993; Gill and Potter 1993). These 
similarities may make competitive interactions 
more likely between A. suppositus and F. lateralis 
than between these species and P. olorum. 
ACKNOWLEDGEMENTS 
The authors are indebted to Stuart Bradley, Mike 
