Recent disintegration of ice shelves on the Antarctic Peninsulahas highlighted the need for a better understanding of ice shelf fracture processes generally. In this paper we present a fracture criterion, incorporating new experimental fracture data, coupled with an ice shelf flow model to predict the spatial distribution of surface crevassing on the Filchner-Ronne Ice Shelf. We have developed experiments that have enabled us to quantify, for the first time, quasi-stable crack growth in Antarctic ice core specimens using a fractureinitiation toughness, Kinit, for which crack growth commences.The tests cover a full range of nearsurface densities, r = 560-871 kg m-3 (10.9-75.7 m depth). Results indicate anapparently linear dependence of fracture toughness on porositysuch that Kinit = 0.257 r-80.7, predicting a zero-porosity toughness of Ko = 155 kPa m(1/2). We have used this data to test the applicability to crevassing of a two-dimensional fracture mechanics criterion for the propagation of a small sharp crack in a biaxial stress field. The growth of an initial flaw into a larger crevasse, which involves a purely tensile crack opening, depends on the size of the flaw, the magnitude of Kinit, and the nature of the applied stress field. By incorporating the criterion into a stress map of the Filchner-Ronne Ice Shelf derived from a depth-integrated finite element model of the strain-rate field, we have been able to predict regions of potential crevassing. These agree well with satellite imagery provided an initial flaw size is assumed in the range 5-50 cm.