BACTERIAL colonies must often cope with unfavourable environmental con
ditions(1,2). To do so, they have developed sophisticated modes of coo
perative behaviour(3-10). It has been found that such behaviour can ca
use bacterial colonies to exhibit complex growth patterns similar to t
hose observed during non-equilibrium growth processes in non-living sy
stems(11) some of the qualitative features of the latter may be invoke
d to account for the complex patterns of bacterial growth(12-18). Here
we show that a simple model of bacterial growth can reproduce the sal
ient features of the observed growth patterns. The model incorporates
random walkers, representing aggregates of bacteria, which move in res
ponse to gradients in nutrient concentration and communicate with each
other by means of chemotactic 'feedback'. These simple features allow
the colony to respond efficiently to adverse growth conditions, and g
enerate self-organization over a wide range of length scales.