The Office for National Statistics (ONS) Longitudinal Study (LS) is an exceptional resource for exploring dynamic processes in individuals' lives for a representative sample of the population of England and Wales and across a thirty year period, including how those processes vary by ethnic group. However, analyses tend to assume a certain stability in the meaning of the ethnic group being studied: the insights into ethnic group differentiation are premised on the fact that the group has the same meaning over time. Here we show how the LS allows us to challenge such notions of group stability. This has practical implications for the ways we measure and conceive of Britain's minority ethnic groups. We illustrate this point with two examples: one exploring the change in ethnic group identification by the same individuals between 1991 and 2001, and the second exploring how loss to follow up is differentially experienced according to ethnic group. We provide some suggestions on the implications of this ethnic group instability for other research.