Abstract
The article builds on earlier research suggesting that it is fairly difficult to extrapolate current trends in school performance into the future (see Gray, Goldstein & Thomas, British Educational Research Journal, 27, 2001). It explores two areas of research on school improvement which have, to date, received relatively little attention--trends in performance over time and the incidence of time-lagged phenomena. In responding to a critique by Pugh and Mangan (2002), it suggests that, whilst of interest, time-trend analyses favoured by economists may be less useful than other forms of analysis which search for consistent patterns over shorter periods of time. Some strategies for thinking about their identification are put forward. The implications of these approaches for target-setting and related exercises, which are premised on linear assumptions about school improvement, are also discussed.