The emergency and outpatient referral rates of practices taking part in the 2nd and 3rd National Morbidity Studies (NMS-2, NMS-3) were studied. Variation in referral rate between individual practices was contrasted with variation between different practice and patient sub-groups. Variability in referral rates between practices to outpatient departments was shown greatly to exceed variability of these rates associated with different practice locations (urban/rural), and between sexes, age-groups or social groups of patients. A similar contrast was seen when exploring rates for inpatient referrals (with exception made for rates in patients aged 75+). There were strong correlations present between individual practice referral activity in the first two years of NMS 2 (R = 0.72, n = 43, P = less than .001), and between NMS-2 and NMS-3 for those practices which participated in both studies (R = 0.80, n = 18, P = less than .001), indicating a high degree of consistency of referral behaviour. In the 3rd study total referrals from general practice were 10.6 per 100 population compared with 20.7 per 100 reported in hospital statistics. Direct access to Accident and Emergency (A and E) Departments and internal hospital referrals between specialist departments account for the bulk of the difference. The cost implications of variability in referral rates among practitioners are considerable. More data about referral are required.