Models for fire interval distributions in ecological communities are propos
ed, based on an understanding of the processes that influence the probabili
ty of fire, especially changes to the amount and condition of the fuel. The
models represent changes in the probability of fire as a function of time
since last fire. Despite considerable differences in the probability distri
butions of fire intervals, the models generate very similar age distributio
ns when the mean fire interval is the same. Therefore, fitting the theoreti
cal distributions to observed landscape age structure is unlikely to allow
discrimination between different models. Previously, the most commonly used
models of fire intervals have been based on the Weibull probability distri
bution. We believe that this is unnecessarily restrictive, and a broader ra
nge of models should be considered. The models may be based on an a priori
understanding of the ecosystem being studied. They should assist interpreta
tion of observed or inferred fire interval distributions.