There is a trend toward increasing intensification in dairy farming in the United Kingdom. In particular, there is concern over systems in which cows are housed throughout the year, as the behavioural restriction implicit in these systems is associated with poor welfare in other species. The aim of this on-going project is to determine how this affects the behaviour and welfare of dairy cattle. A range of behavioural, physical and health measures are being used to assess cow welfare on about 40 commercial British dairy farms. Initially, five farm types were identified from analysis of returns from a farmer questionnaire. Milk production level and housing type were the principal factors explaining variation in farm type. The sample groups are: high-, mid- and low-production cubicle-housing units, mid-production straw-court units, and cubicle-housing high-production zero-grazing units. Observations will take place over three winter housing periods (2000/01 to 2002/03), with recording on each unit taking five days. Our main hypothesis is that the behavioural and physical responses of ‘at-risk’ younger cows provide a sensitive indication of farm-level ‘stress’. Cows are marked according to age, and the feed-face videotaped continuously to record feeding time and social interactions. Temporal organisation of behaviour will be analysed using fractal mathematics and qualitative assessment approaches used in a human interaction test. Cow cleanliness, condition score, response to a novel object and the incidences of lameness and leg injury are recorded, and building quality is assessed. Ultimately, multivariate methods will be used to test our underlying hypothesis and to assess effects of housing and production type on behaviour and welfare. This analysis may identify key objective measures of welfare for use in farm assurance schemes.