The chance of encounter—and hence mating—between two animals may be limited by social as well as spatial factors, thus dividing a population into a series of discrete territories with little movement between them. Only within the sub-units will random mating take place, and drift may cause a considerable amount of non-adaptive genetical change if the units are very small. Work on House mouse (Mus musculus) populations in the laboratory and particular ecological situations have suggested that the effective breeding size of these units may be as small as four. This would mean that a considerable amount of random change would be expected in mouse populations. However, a six year study (1964–69) of movement and territoriality on the 244 acre (100 ha) Welsh island of Skokholm during which over 3000 animals were marked and released, showed that more than 20% of individuals breed in an area other than the one in which they were born, i.e. a considerable amount of population churning takes place. This conclusion is supported by evidence of the spread of three rare biochemical variants in the population. Chance seems to play little part in the determination of the genetical constitution of the Skokholm mice. In general, estimates of the size of effective breeding units must always be qualified by an understanding of the ecology of the population in question.