The determination of the precise position of permanent first and second modern human molars, following standard tooth identification criteria, is often difficult because of their morphological similarities. Here, we proposed to evaluate the suitability of two-dimensional crown contour shape analysis in achieving this objective. The method was tested separately on 180 first and second maxillary molars (UM) and 180 first and second mandibular molars (LM) securely identified (in anatomical position in their sockets). Generalized Procrustes superimposition is used to normalize the outlines prior to applying elliptic Fourier analyses ("EFAproc" method). Reliability and effectiveness of this morphometric procedure was evaluated by comparing the results obtained for the same dataset with four other morphometric methods of contour analysis. Cross-validated ("leave one individual out") percentages of misclassification yielded by linear discriminant analyses were used for determining the anatomic position of modern human molars. The percentages of misclassifications obtained from every method of contour analysis were low (1.67% to 3.33% for the UM, 5.56% to 6.67% for the LM) indicating the high suitability of crown contour analyses in correctly identifying molars. A reliable protocol, based on predictive linear discriminant analyses, was then proposed for identification of isolated molars. In addition, our results confirmed that the EFAproc method is suitable for normalizing outlines prior to undertaking elliptic Fourier analyses, especially in the case of nearly circular outlines: it obtained better classification than the classic method of normalization of Fourier descriptors for UM and provided also some advantages over the three landmarks-based methods tested here. Am J Phys Anthropol, 2013. © 2013 Wiley Periodicals, Inc.