The utility of fine-needle aspiration biopsy to detect carcinoma in thyroid nodules was evaluated by a decision-analysis approach in 102 patients. The procedure caused no morbidity. Cytologic diagnoses were categorized as unsatisfactory (4), no abnormality detected (61), atypical (13), suspicious for malignancy (14), malignancy (2), and inflammation (8). The duration of follow-up averaged 13 months. Of 21 thyroidectomy patients, 10 (48%) had carcinoma. Half of the ten patients operated upon for suspicious cytologic findings were found to have malignancy. Assuming criterion I, that atypical, suspicious, or malignancy results indicated cancer, sensitivity was 90%, specificity 77%, false positive fraction 23%, positive predictive value 31%, negative predictive value 99%, and accuracy 79%. Assuming criterion II, that only suspicious or malignancy cytologic findings represented carcinoma, sensitivity was 70%, specificity 90%, false positive fraction 10%, positive predictive value 44%, negative predictive value 96%, and accuracy 88%. We conclude that sensitivity and specificity of fine-needle aspiration biopsy vary depending upon the use of criterion I or II. Accuracy is highest if atypical results are not considered to represent carcinoma. Positive predictive values remain low and negative predictive values are high in either case. The utility of fine-needle aspiration biopsy when interpreted in relation to clinical criteria is supported by these results.