A stochastic cellular automata model for the population dynamics of th
e army ant Eciton burchelli on Barro Colorado Island in Panama is set
up. It is simulated on the computer and shown to give good agreement w
ith biological data. It is analysed using two approximations akin to t
he mean field approximation in statistical mechanics, and good agreeme
nt with the simulations is obtained. Finally, the role of distance bet
ween successive statary phase bivouacs is discussed with regard to the
rate of colony growth. There are two aspects of the biological system
studied here that make it of general importance. First, the populatio
n is structured, since the size of each colony of army ants is crucial
. Second, the spatial behaviour of the population, as in many others,
is not diffusion-like, although it is random. This has implications fo
r the kind of model that is chosen.