Summary 1. Natural aquatic communities or habitats cannot be fully replicated in the wild, so little is known about how initially identical communities might change over time, or the extent to which observed changes in community structure are caused by internal factors (such as interspecific interactions or traits of individual species) versus factors external to the local community (such as abiotic disturbances or invasions of new species).
2. We quantified changes in seven initially identical fish assemblages, in habitats that were as similar as possible, in seminatural artificial streams in a 388-day trial (May 1998 to May 1999), and compared the change to that in fish assemblages in small pools of a natural stream during a year. The experimental design excluded floods, droughts, immigration or emigration. The experimental fish communities diverged significantly in composition and exhibited dissimilar trajectories in multivariate species space. Divergence among the assemblages increased from May through August, but not thereafter.
3. Differences among the experimental assemblages were influenced by differences that developed during the year in algae cover and in potential predation (due to differential survival of sunfish among units).
4. In the natural stream, fish assemblages in small pools changed more than those in the experimental units, suggesting that in natural assemblages external factors exacerbated temporal variation.
5. Our finding that initially identical assemblages, isolated from most external factors, would diverge in the structure of fish assemblages over time suggests a lack of strong internal, deterministic controls in the assemblages, and that idiosyncratic or stochastic components (chance encounters among species; vagaries in changes in the local habitat) even within habitat patches can play an important role in assemblage structure in natural systems.