This article emphasizes some controversies concerning the preoperative selection of nonpalpable thyroid tumors. The prevalence of occult thyroid carcinoma in surgical series (1.8-10%) is not higher than in autopsy thyroid series (2.7-24%). The prevalence of occult thyroid carcinoma in thyroid glands examined in the same institution by ultrasound, for a clinical thyroid abnormality or for investigation of other neck structures without clinically evident or suspected thyroid disease, varies from 3% to 8% and is very similar independent of the fact that a thyroid abnormality is or is not the indication for ultrasonography. These data suggest that the presence of a thyroid disease is not a risk factor for harboring an occult thyroid carcinoma (except for C-cell hyperplasia in the rare case of MEN 2 syndromes). As it is not cost effective to examine all the nonpalpable lesions with fine-needle aspiration (FNA) guided by ultrasounds, it is necessary to define to which extent ultrasound is useful in selecting those lesions to be examined cytologically by FNA. The use of ultrasound to select these lesions is very controversial. Ultrasound-guided cytologic diagnosis of nonpalpable nodules is not as accurate as in the case of palpable nodules. Sampling of material adequate for cytologic analysis depends on the lesion size; it is 64% for a 0.7-cm lesion and it increases to 86.7% for a mean size of 1.1 cm. For the diagnosis of occult thyroid carcinomas (< or =1 cm), sensitivity is 35.8% and false-negative results are 49.3%. Nonpalpable nodules with a size of 1.5 cm represent an absolute indication to perform an ultrasound-guided FNA because this is the size limit for dividing thyroid nodules in probably innocuous or potentially dangerous categories and because the cytologic diagnosis of nodules of this size is sufficiently reliable. For the smaller incidentally discovered thyroid nodules following ultrasound, physicians should discuss with the patient whether and when to perform an ultrasound-guided FNA considering the patient's data (risk factors, age, health state, etc.), the natural history of a small thyroid carcinoma, as well as the accuracy of ultrasound and ultrasound-guided FNA in the specific institution.