During this PhD thesis, a new fracture test geometry is designed for the accurate measurement of the fracture properties of brittle solids, subsequently applied to study failure in PMMA. At high crack speeds, their fracture surfaces are optically smooth. But below vc = 15 mms-1, a transition to rough surfaces occurs through the formation of puzzling triangular patterns. These patterns lead to significant toughening of the material that reflects through the pinned shape of the crack front as it crosses triangles. In addition, these triangles are found to be decorated by faceted features reminiscent of the crack front fragmentation instability in mode I+III. Assuming a shear-dependent fracture energy Gc(KIII/KI) = GcI[1+ (KIII/KI)2] we theoretically predict a fragmentation threshold (KIII/KI)thc that can be as low as a few percent while earlier models (that assumes = 0) predict a much larger value, inconsistent with various experimental observations. Applied to our experiments, this model allows us to measure exp from the deformation amplitude of the pinned front and the amount of applied shear (KIII/KI)exp from the facet inclination which is found to be compatible with the theoretically predicted threshold (KIII/KI)thc . Using the values (KIII/KI)exp and exp thus determined, one finally predict a drift of the facets from the propagation direction accounting for the triangle angle observed experimentally. To conclude, our study shows that the roughening transition in PMMA is a signature of front fragmentation under mode I+III. As a result, deciphering the triangular patterns at the transition led to significant improvements in the understanding of this instability.