1 The species occurring on 128 small cays in the central Exumas and 54
small cays near Andros, Bahamas, were surveyed. 2 Step-wise multiple
linear regressions selected the log transformations of vegetated area
or total area as the variables explaining the most variation in specie
s number at Andros. In the Exumas, the ratio of vegetated area to tota
l area, which is an expression of both insular area and exposure, was
the most frequent first-selected variable. 3 In both archipelagoes, ca
ys with vegetation were significantly larger, higher and closer to a m
ainland island than cays naturally devoid of vegetation. 4 High degree
s of nestedness were observed in both archipelagoes, both at the level
of the entire species assemblage and at the level of individual speci
es. Ninety per cent of the species that did not display significant ne
sted patterns were rare species, occurring on less than or equal to 3
cays each. 5 Step-wise multiple logistic regressions revealed that the
incidence of most species was positively related to some measure of a
rea (14 of 14 species in the Exumas, 8 of 12 species on Andros) and ne
gatively related to distance (12 of 14 species in the Exumas, 3 of 12
species on Andros). Correct classification was > 75% for all but one s
pecies. 6 Species turnover was measured for 77 cays in the Exumas, whi
ch were surveyed annually over a 4-year period. Turnover was very low,
and most of the observed turnover was due to immigrants that never be
came established. Contributing factors to the observed low turnover in
cluded the absence of annuals and the physical stability of these pred
ominantly rocky cays. 7 Propagules of Sesuvium portulacastrum were int
roduced to 10 cays naturally lacking vegetation. After 40 months, S. p
ortulacastrum was still present on nine of 10 cays and had increased i
n vegetated area from approximately 1 m(2) up to 650 m(2), indicating
that non-vegetated cays were physically able to support plant life, at
least in the short term. 8 A 'small island effect' (i.e. a weak speci
es-area relationship on very small islands) was apparent at Andros and
probably resulted from smaller cays being washed over by waves.