Models for rubber particle cavitation, shear yielding, and crazing are reviewed, and their ability to predict the large-strain deformation behavior of toughened polymers is discussed. An existing model for void initiation and expansion in rubber particles correctly predicts the observed trends: cavitation resistance increases when either the shear modulus or the surface energy of the rubber is increased, or the particle size is reduced. However, further work is needed to improve quantitative modeling of the thermally- and stress-activated void nucleation step. Shear yielding, which is also a rate process, is much better understood; here, the main problems in modeling relate to the formation and evolution of porous shear bands. Craze growth and failure are also reasonably well understood, but previous attempts at modeling have been hampered by uncertainties about craze initiation. To overcome these difficulties, a new theory of crazing is proposed, which treats initiation as a fracture process, and defines a new materials property, G(nasc), the energy required to form unit area of nascent craze. Because nascent crazes are similar to 20 nm thick, G(nasc) is low: calculations give values <0.5 J m(-2) for polystyrene. A new criterion incorporating a plasticity factor fits the data of Sternstein and coworkers on crazing under biaxial loading. In combination with theories of particle cavitation and shear yielding, the fracture mechanics model explains why the balance between crazing and shear yielding is governed by particle size, for example in ABS.