Dispersive flight by Triatoma infestans was modelled by assuming either unidirectional flight in a straight line, or that flight followed a pattern of ‘random walk’ analogous to two-dimensional Brownian motion. Parameter estimates for the models were derived from previous fieldwork in Argentina and Brazil. Calculated probabilities of houses becoming infested over a 1 year period were compared with actual changes of house infestation in a small farming community in central Brazil. It was found that infestation could be explained partly by the dispersive flight of the bugs according to the distances between houses, and partly by the predisposition of houses to infestation by virtue of their construction, since houses with cracked mud walls were particularly susceptible to infestation. Infested houses lying more than 200 m from uninfested houses appeared to make an almost negligible contribution to infestation of the uninfested houses. This suggests that buffer zones of this size around treated areas could help to minimize the risk of reinfestation after treatment.