A model has been proposed recently, which describes the experimentally observed mechanical behavior of some shape memory polymers. It considers a purely thermoelastic behavior, without strain rate effects, and assumes essentially that the polymer can be considered as a two-phase composite, with glassy and rubbery phases having volume fractions that depend on temperature only. Since a uniform stress hypothesis was used in the original formulation, with an inconsistency when thermal expansion was considered, this model is revisited here by taking advantage of many results that have been established in the theory of composite materials. It is shown, especially, that a uniform strain hypothesis is more appropriate than assuming a uniform stress.