The major results of this paper are the pointing out of deficiencies in existing thermal remanent magnetization theories for nonuniformly magnetized grains and the suggestion of ways in which some of these deficiencies might be removed. Most models of pseudo single domain behavior fail to predict certain critical experimental observations of TRM, particularly those dealing with magnetic stability. At least part of the problem with understanding TRM in small nonuniformly magnetized grains is that the size and role of the demagnetization field has been grossly overestimated. TRM in small multidomain grains is probably acquired by domain wall movement, except for the smallest (submicron) grains in which pseudo single domain behavior associated with wall moments probably dominates. A collective response of domain walls seems to be required to explain magnetic stability observations, although the degree and the precise nature of this collective response has not yet been determined.