A review of 15 population-based glaucoma prevalence surveys in Western Europe, the US, the West Indies and Japan shows that the proportion of patients with the condition who had previously gone undetected was generally at least 50%. Possible reasons for underdetection of glaucoma have been considered in relation to England and Wales, where most patients with glaucoma are initially detected during the course of sight tests in connection with providing spectacle lenses. It was found that: (i) a high proportion of the population over 40 years of age attends fairly regularly for a sight test, (ii) the standard of primary testing for glaucoma is very uneven--those examiners who test comprehensively detect about 50% more cases than average; and (iii) referral criteria, which reflect the need not to overload hospital eye clinics, inevitably exclude many patients who are in apparently low risk categories. Both the population survey data and the subsequent analysis suggest that underdetection is most pronounced in patients with glaucoma of the normal pressure type.