Aims: To conduct an internet-based study using virtual slides (VS) of sterotactic core biopsy specimens of non-palpable breast lesions in order to evaluate interobserver reproducibility between pathologists. Methods and results: A total of 18 breast lesions, determined to be histologically complex by two pathologists, were selected. Digitized VSs were then created using QuickTime Virtual Reality technology (Apple, Cupertino, CA, USA) and posted on the world-wide web. In all, 10 pathologists completed the evaluations of 18 VSs using the five diagnostic categories (B1-B5) from the European guidelines for quality assurance in breast cancer screening and diagnosis. Their results were compared with those of every other participating pathologist, and were then individually compared with the results of a highly experienced breast pathologist (referee). Of the 18 cases, 10 (56%) were classified by the referee as borderline (B3 and B4). Comparisons with reference values showed a less than satisfactory level of reproducibility (median κ w = 0.60). As regards interobserver reproducibility, results showed that, in general, the level of agreement was not satisfactory (median κ w = 0.53). Conclusions: Overall, the findings are comparable to those quality control studies using circulating slides when analysis is done on borderline cases. © 2010 Blackwell Publishing Limited.