Recently several authors have claimed that long term vegetation change, at least in the case of forest ecosystems, can usefully and quite accurately be described by stationary Markov chain models (Anderson, 1966; Waggoner & Stephens, 1970; Horn, 1974, 1975a, 1975b, 1976) or their deterministic counterpart, coupled differential equation models (Shugart et al., 1973). Markov chain models do indeed form a class of unusually versatile and well-studied models and their exploratory use as models of succession would seem quite appropriate (for an introduction see van Hulst 1979 and references therein). Markov models, for example, can mimic not only simple linear succession (with species or community B replacing species or community A C replacing B, and so on), but also successions involving such phenomena as reversals or ‘sticky’ states, cyclical successions, indeterminate situations, a gradual approach to a steady state, and several other complications (see e.g. Horn 1976).