The population structure of the malaria vector Anopheles arabiensis was inv
estigated using data from six microsatellite loci in samples from localitie
s in Mozambique and Tanzania. Genotype frequencies were neither significant
ly different between houses in a village in Tanzania nor between villages w
ithin a 20-km radius in Mozambique. Thus a deme has an area greater than 20
km in radius. At five of the six loci the heterozygosity of the population
from Mozambique was lower than that from Tanzania, implying a lower effect
ive population size (N-e) at this southern edge of the species range. There
were significant differences in genotype frequencies between the Tanzanian
and Mozambique populations at five of the six loci (P < 0.05). Values for
both F-ST (mean = 0.069) and R-ST (mean = 0.025) were significantly differe
nt from zero (P < 0.05) at four and three out of five loci, respectively, b
ut there was no significant correlation between the two statistics. The wid
e variation in values of F-ST and R-ST across loci suggests that care shoul
d be taken in interpreting values derived from averaging across loci. Wheth
er the variation results from sampling effects or selectional constraints o
n some loci is unclear. Although there is evidence for significant differen
tiation between these populations, estimates of gene flow (Nm) calculated f
rom mean F-ST and R-ST statistics were relatively high, 3.4 and 4.9, respec
tively. We argue that this is more likely to reflect recent separation of t
hese populations and/or large effective population size rather than large-s
cale present day migrations.