We refined established methods for examining the temporal organization of b
ehavioural events and applied them to the study of bats emerging from roost
sites. Previous studies have shown that as roosting numbers of bats increa
se, temporal patterning (clustering) in their emergence behaviour, as measu
red by chi(2) or G statistics, also increases. Expressing the extent of tem
poral clustering using chi(2) or G, however, confounds two effects. The sta
tistic may increase because clustering increases. However, an increase will
occur, even if the amount of clustering stays constant, if the sample size
of observations increases. A positive relation between emergence size and
chi(2) or G may not necessarily reflect increased clustering. We developed
a method to separate these effects using computer-modelled event streams. U
sing this method, we examined intra- and inter-roost variation in the tempo
ral patterning of emergences of the pipistrelle bat, Pipistrellus pipistrel
lus, from four roosts in northeast Scotland. The 'corrected' expression of
clustering decreased as roost size increased, which was opposite to the eff
ect found with the uncorrected expression and also opposite that predicted
a priori from the bottleneck hypothesis. This novel result suggests bottlen
ecks may disrupt clustering rather than promote it. This latter interpretat
ion was supported by observations that clustering was significantly reduced
in the middle of large emergences. Variation in clustering between roosts
was significant, suggesting exit topology may have an important influence o
n clustering behaviour. (C) 1999 The Association for the Study of Animal Be
haviour.