Diet is important in the aetiology and management of many conditions in primary care. Although valid dietary assessment is required for both clinical work and research, no dietary assessment instruments have been validated among patients seen in primary care. A range of simple self-completion dietary assessment questionnaires and established research instruments were compared with an accepted reference standard, a seven-day weighed record, in 111 subjects assessed in a practice nurse-run treatment room. Simple self-completion tools based on food groups and portion sizes perform as well (likelihood ratios for a positive test = 2 to 3) as much more time-consuming instruments. The error in using such instruments is comparable with the error of the standard itself. There is little justification for using time-consuming dietary assessment questionnaires, since simple tools are accurate enough to be clinically useful--to allow practice nurses to target patients for counselling and waste less time on inappropriate counselling--and also useful for research.