The results of three different upsetting tests on steel C1530 are presented. The die design is shown to have a marked effect on free surface fracture. A popular empirical ductile fracture criterion (Oyane's criterion) is assessed by means of the experimental data obtained and by a theoretical approach for determining the workability diagram from free surface fracture data. It is shown that the criterion is not appropriate for the material investigated. The range of free surface conditions where experimental fracture data are required for formulating an accurate fracture criterion is specified and a number of tests to create such conditions is proposed.