Enamel thickness was investigated in the mesial cusp section of 167 unworn human molars by means of non-destructive micro-CT based methodology. Serial sections of the entire crown were taken at a voxel resolution of 28 microns, and the initial volume data set of each molar was standardized in orientation to obtain a vertical section that accurately contains the dentine tips of the two mesial cusps. Enamel thickness at the cusp tips, occlusal basin, and lateral crown face was measured in the mesial cusp section and in sections offset from that section by 0.6 mm. We found that thickness at the cusp tips may be overestimated in offset sections by up to about 1 mm, and those of the occlusal basin overestimated or underestimated by up to about 0.5 mm. We also found that maximum 'radial' thickness of the lateral crown face was least affected by section position, usually with discrepancies of less than about 0.1-0.2 mm. In all serial positions in both upper and lower molars, a 'functional' (lingual in uppers and buccal in lowers) to 'non-functional' side gradient in enamel thickness was observed in cusp tip, occlusal basin, and lateral crown face enamel, with the exception of the characteristically thin enamel at the protoconid and paracone cusp tips. Serial differences in thickness were seen between the thinner M1 and the two posterior molars in many but not all measures of thickness, the pattern of which appears to be influenced by the thin M1 mesiobuccal cusp enamel. Individual variation of maximum lateral thickness, the least variable measure of thickness, was found to be substantial (a 30-60% range) even with serial and buccolingual positions controlled. Correlation between whole crown average enamel thickness and maximum lateral thickness was high, indicating that the latter is a potentially useful predictor of overall enamel thickness of the molar crown. The present results indicate that interspecific comparisons of enamel thickness must be made with careful attention to positional placement of thickness measures, potential serial differences, and intraspecific variation.