1. Spectral analysis of 66 years of locust swarm abundance data failed
to reveal any significant cycles although the dominant cycle detected
in a de-trended series, adjusted to take,account of a significant par
tial autocorrelation for a lag of 1 year, had a periodicity of 16 year
s. 2. Although some estimates of the intrinsic rate of increase (r) of
desert locusts are indicative of chaos, reconstructions of locust dyn
amics using response surface methodology (RSM) suggested exponential s
tability. This was also true for data for the West African Region alon
e and inclusion of rainfall data from the Sahel improved the significa
nce of an RSM model for West Africa. 3. The observed positive relation
ship between locust abundance and rainfall in West Africa confirmed th
e importance of rain; but the variance of the locust abundance also in
creased with rainfall, making rainfall alone a poor predictor. However
, this heteroscedastic pattern was reproducible by a simple logistic m
odel with a chaotic r and a variable K. This was not the case when a s
table value for r was used. 4. The available data and current methodol
ogies are insufficient to provide unequivocal conclusions on locust dy
namics, which are complicated by phase changes and associated switches
in r values.