Despite several recent studies, the biological status and clinical relevance of telomerase expression in tumours derived from the thyroid follicular cell remain controversial. This study has analysed a series of normal, benign, and malignant thyroid samples using two novel approaches: the use of purified epithelial cell fractions to eliminate false-positives due to telomerase-positive infiltrating lymphocytes; and the simultaneous measurement of telomere length to provide a clearer interpretation of telomere dynamics in thyroid neoplasia. The data obtained support the prediction that the epithelial component of non-neoplastic thyroid and of follicular adenomas is telomerase-negative, any positive results being explicable by lymphocyte infiltration. In contrast, many malignant tumours, both follicular and papillary, were telomerase-positive. However, serial dilution of extracts indicated a wide spectrum of activity in these cancers, possibly related to variation in the proportion of telomerase-positive cells. Furthermore, an unexpectedly high proportion were telomerase-negative, a finding which was not explicable by technical problems such as TRAP (telomeric repeat amplification protocol) assay sensitivity. Many of these apparently telomerase-negative tumours had abnormally long telomeres. Correlation of telomerase and telomere length data suggests that thyroid cancers fall into three biological groups: telomerase-positive lesions, consistent with the conventional model of telomere erosion followed by telomerase reactivation; telomerase-negative tumours, which maintain telomere length by a mechanism independent of telomerase; and telomerase-negative tumours which are still undergoing telomere erosion and may therefore be composed of mortal cancer cells. From a clinical standpoint, it is concluded that telomerase detection on unfractionated tissue, such as fine needle aspirates, is of no value as a marker of malignancy in follicular lesions, due to both low sensitivity and specificity.