Differences in vegetation dynamics over a period of two years along an environmental gradien t of shading on a lawn are examined. 'Communities' (groups) recognized by numerical classification are correlated with degree of shading, season and differences between years by means of generalized linear model analysis. More of the variance is explained if environmental position is used instead of degree of shading as spatial distribution of a strongly competitive species (Trifolium repens) is confounded with shading. Transition probabilities are related to environmental position and season. Simulation with Markov matrices for each season and position demonstrate markedly different successions for different positions. These simulations have no predictive value however as accurate estimate of transition probabilities requires knowledge of the state of adjacent quadrats, i.e. individual observations of transition probabilities are not independent.Transition matrices are unlikely to be useful for predictive analysis of succession when spatial pattern is of significance in the community.