The study aimed to examine dietary changes and their associations with demographic, lifestyle and health variables in a random sample of British adults. The Health and Lifestyle Survey of 1984–5 (HALS1) identified four main dietary components by principal component analysis from food frequency data. Comparison of the HALS1 dietary component scores with those of the follow-up survey of 1991–2 (HALS2) for the same individuals revealed increases on component 1 (high weightings for fresh fruit, salads, ‘brown’ bread, fruit juice and green vegetables but low weightings for chips, fried food and processed meat). There were substantial decreases on component 2 (high weightings for puddings/pies, cake, potatoes, biscuits, preserves, pulses and meat), small increases on component 3 (high weightings for crisps, soft drinks and chips) and increases on component 4 (high weightings for confectionery, biscuits and cake and low weightings for vegetables of all kinds). Except for women on component 3 the changes were all significant, P < 0·001. Unadjusted score changes were smallest in elderly respondents for all components. Differences in score changes between groups were based on an analysis of covariance adjusting for the HALS1 score. On component 1 the largest score increases were associated with non-manual groups, improvements in lifestyle and good health. For component 2 the greatest changes in score were associated with changes in household size, smoking habit and heavy drinking. Score increases on component 3 were also associated with heavy drinking, whilst the largest rises on component 4 were amongst the non-manual, the non-drinkers and the non-smokers and, for women only, those who had few malaise symptoms or who lived in Scotland. The results show that there have been overall dietary changes and that changes have been associated with longitudinal alterations in socio-demographic, lifestyle and health circumstances.