On the basis of previous research, the stone size effect is not always ruled by well-known statistical laws such as the Weibull or Bazant law because of the intrinsic complex structure and features of the materials analysed. According to the EN 12372 test method for flexural strength under concentrated load on natural stones, the suggested dimensions are 50 × 50 × 300 mm. Other dimensions are possible, but some dimensional requirements (on thickness, width and distance between supporting rollers) need to be fulfilled. In this research, to investigate the thickness effect further, in relation to stone structure and heterogeneity, eight different stones were tested: four with anisotropic features and four homogeneous. The distance between the supporting rollers was proportional to the specimens' length. The thickness effect on the flexural strength of most of the materials tested is negligible considering the uncertainty (except for gneiss). From the data obtained some remarks have been made on the test method and on the specimens' thickness. Moreover, with the aim of obtaining reliable results with low uncertainty by means of a flexural strength test, a minimum thickness specimen both for fine- and coarse-grained stones has been determined (respectively 15 and 20 mm).