A 20-fold variation in referral rates from general practitioners to hospital outpatient departments has been shown in studies published to date. Most of the hypotheses proposed to account for this variation have not been supported by these studies. A simple model was constructed, which showed that a significant part of the variation may be due to the fairly small numbers of referrals in most studies. Real differences may have been swamped by random variations in the small numbers of referrals. The statistical power of the studies may not have been high enough to determine which factors were significant in describing the variation and how much of the variation was due to differing clinical practice. There remains a substantial part of the variation that cannot be accounted for.