Species assemblages of systems of islands and other fragmented habitats frequently show a "nested-subset" structure in which the biotas of sites with low species richness are non-random subsets of the biotas of richer sites. Much literature suggests that extinction is more likely to produce strongly nested patterns than colonization, although very few experiments have been conducted on the generation of nested-subset patterns. Here, we describe an experiment on nestedness of benthic invertebrates occupying rocks in the littoral zone of a lake in western Victoria, Australia. Data collected from previous work indicated that the invertebrate biotas of rocks were nested. We used initially defaunated, different-sized habitat patches (half house-bricks, full house-bricks, double house-bricks) and followed the time-course of occupation to assess whether nested-subset patterns would emerge, and whether extinction or colonization best accounted for the patterns. We predicted that nestedness would increase through time up to the end of the experiment. This was found to be the case with colonization dominating the establishment of a strongly nested system; extinction appeared to be of relatively little importance over the duration of the experiment. These results suggest that at least in some circumstances, differential colonization may be influential in producing nested-subset patterns but experiments conducted over longer times may be needed to more completely understand the respective roles of extinction and colonization in generating nested-subset patterns in this system.