It is reasonably well accepted that the standard procedures developed for isotropic homogeneous metals using linear elastic fracture mechanics models are not appropriate for either continuously or discontinuously reinforced metal matrix composites. For example, the ASTM plane strain fracture toughness test methods typically give widely different values of fracture toughness depending on the particular test specimen geometry as well as the fiber orientation. For unidirectional boron/aluminum composites one finds approximately a factor of two difference between the measured values of fracture toughness obtained from a center-notched test coupon and that given by a compact tension specimen. In particular for unidirectional composites, and to a slightly lesser degree for angle ply laminates, the dominant controlling mechanism for this behavior is matrix plasticity. A secondary toughening mechanism, resulting from the matrix plasticity, is stable transverse fiber failure. The present paper will focus on both the influence of the large plastic zone at the end of the notch and on the constraint that the fibers impose on the shape of this zone, as well as the transverse crack growth. First, a review of some particular experimental studies and methods of analysis for predicting crack growth and fracture of notched unidirectional metal matrix composites is given. Next, two mechanistic models for unidirectional composites with damage are presented. The first is an improved shear lag model that accounts for both of the above damage modes, and the second describes a recent extension of the shear lag concept in an attempt to include transverse stresses. A related finite width laminate model is then discussed, and it is indicated that an isotropic finite width correction factor is reasonably accurate for most center-notched test coupons.