Considerable work has been done in recent years by geographers, spatial economists, regional scientists, and allied disciplines, on dynamic modelling. A main emphasis of this work is description of spatio-temporal paths of geographical systems. Good examples of dynamic models for spatial problems are provided in Griffith and MacKinnon (1981). Unfortunately these sorts of dynamic models tend to be linear in form, and tend to treat locations in space in an independent fashion. This latter feature is analogous to constructing a set of n time series models that are independent of one another. But spatial autocorrelation mechansims, space-time processes and the relative nature of space highlight the inappropriateness of most dynamic spatial models for most real world problems. As soon as non-linear formulations embracing interaction effects are introduced, the modelling game is dramatically changed. Movements through time become irreversible, and bifurcations of trajectories become possible, perhaps even likely.