This paper reports on the accuracy of tooth counts conducted in 22 subjects by 10 trained nurses as part of a large longitudinal study of a pharmacological agent. These nurses participated in a training course consisting of seminars, discussion, demonstrations, and practice examinations. Each of the nurses then counted the teeth of 22 subjects and recorded their findings independently. The counts of the nurses were compared with those of the dentists to assess the accuracy of the nurses' counts. We found that nurses and dentists were in perfect agreement for 86% of the patient counts conducted. Individual nurses' levels of agreement with dentists ranged from 73% to 100%, with pairwise kappa statistic values ranging from 0.70 to 1.00. In addition, both Pearson correlation and interclass correlation measures exceeded 0.98 for every comparison of dentist and nurse counts. The results of this study suggest that training nondental health care workers may be an accurate and low-cost way of obtaining tooth loss data and other oral health measures, particularly when oral health data are collected as part of larger, multi-disciplinary studies.