Several analytical and numerical models proposed to predict the effective fracture toughness (KIC) of dispersion reinforced brittle matrix composites require the knowledge of the in situ fracture toughness of the included phase [1– 4]. These values are usually unavailable and they may differ significantly from data obtained on bulk materials. Ceramic platelets (e.g. Al2O3 and SiC platelets) have a high potential to be used as reinforcement in ceramic and glass matrices [5–13]. In order to be able to predict the effective toughness of platelet-reinforced composites it is necessary to know the value of the platelets’ intrinsic toughness. However, the data are not easily available. For example, in the case of single crystal Al2O3 platelets, many uncertainties exist in the choice of the appropriate toughness value. A very large range of data is reported in the literature, i.e. KIC = 1.7–5 MPa m1/2 [14, 15]. These two extreme values would result in a very different prediction of the fracture behavior and effective fracture toughness of the composite. Moreover, these values are often reported for bulk alumina ceramics and not for the single crystal. In fact, experimental determination of fracture toughness on loose platelets is difficult to achieve. In the present work, a parametric study using a numerical model has been carried out in order to determine a realistic value of fracture toughness for alumina platelets embedded in a borosilicate glass matrix. The method is based on numerical simulations and on the comparison between the numerical results and available experimental observations. A previous experimental study [11] has demonstrated that in borosilicate glass reinforced with alumina platelets crack deflection was more likely to occur when the platelets were inclined at certain angles with respect to the crack propagating direction. There was a critical angle between platelet and crack direction for which the crack began to deviate. From the experimental observation of several crackplatelet interactions, this angle has been found to be approximately 50◦ [11]. At greater angles of intersection, the platelet at the front of the propagating crack would be fractured, with little or no previous deflection of the crack. Typical examples of crack/platelet interactions in these composites, documenting their behavior