Coral reef fishes occupy habitats that are patchy and subject to frequent n
atural disturbances. Although different types of disturbance are likely to
generate different community responses, the relationship between different
disturbance agents and their effects on reef fish communities has not been
examined experimentally. We studied a set of natural patch reefs, dominated
by a diverse array of soft and hard coral cover, at Lizard Island on the G
reat Barrier Reef (northeastern Australia). The fish assemblages on the ree
fs were sampled over 4 mo to establish baseline values and then experimenta
lly disturbed. Two types of disturbance were carried out in a factorial com
bination: pulsed mortality by removing all fish from reefs and pulsed habit
at disturbance, Habitat disturbance was applied at two levels: Level 1 cons
isted only of damaging all live hard corals with a hammer; Level 2 consiste
d of damaging all live hard corals, and in addition, using a hammer to redu
ce the height and complexity of the reef matrix. We then monitored the expe
riment for a further 19 mo, including two recruitment seasons.
Unmanipulated control assemblages persisted through time, and despite large
changes in total abundance, species composition remained consistent relati
ve to disturbed treatments. Assemblages disturbed by fish removal were resi
lient, with recolonization from both immigration and larval settlement effe
ctively removing differences between removal treatments and controls 3 mo a
fter manipulation. Habitat disturbance alone generated differences between
experimental and control assemblages, which persisted for the duration of t
he experiment. The more extreme level of habitat disturbance generated more
extreme changes in fish assemblages when no pulsed mortality occurred. Hab
itat disturbance in combination with pulsed mortality generated similar com
munity responses as the habitat disturbance treatment alone. However, fish
removal had the effect of eliminating the difference between fish assemblag
es on reefs subjected to different levels of habitat disturbance. Community
response to habitat disturbance was driven by species-specific patterns of
reduced abundance of species associated with live coral in combination wit
h increased numbers of those associated with rubble. Declines in the abunda
nce of coral associates on damaged reefs were abrupt, with no recovery obse
rved for the duration of the experiment. In contrast, increases in the abun
dance of rubble associates were more ephemeral, in that initial high levels
of recruitment and immigration were followed by a high rate of loss. Habit
at disturbance also generated reefs that typically supported lower fish abu
ndance, fewer species, and increased evenness relative to controls.
Our results support a model of patch-reef fish assemblages organized by a c
ombination of deterministic factors (such as habitat structure) and stochas
tic processes (such as recruitment). These disparate mechanisms operate in
concert to generate reasonably consistent patterns of community structure.
Habitat structure appears to mediate much of the apparent determinism and i
s likely to operate both as a reflection of species-specific habitat prefer
ences and by modifying interactions among fish species. Consequently, distu
rbance plays a substantial role in structuring communities of coral-reef fi
shes by modifying both spatial and temporal heterogeneity.