Three very different mechanisms of martensitic nucleation have been proposed over the past 3 decades. The most classical approach is that of Suezawa and Cook wherein heterophase fluctuations are visualized to generate fully-formed martensitic particles when aided by elastic interaction with very potent, pre-existing superdislocations. In contrast, Clapp suggested that a local mechanical instability of the parent phase, such as a localized soft phonon mode, may set in near an existing defect under the right thermodynamic conditions. As with higher-order displacive transformations, one can describe fluctuations that might result in a "strain-spinodal" via specific phonon modes. Suzuki and Wuttig treated this idea and concluded that the gradient energy between the local instability and the coherently-coupled surrounding matrix would have to be negative in order to allow spontaneous nucleation. The third mechanism proposed by Olson and Cohen invokes arrays of pre-existing lattice dislocations as the fundamental nucleating defect. They provide a detailed description of how such defects can dissociate to achieve the nucleation process: Using FCC as an exemplary structure, consider an array of closely-spaced lattice dislocations such as may occur at a grain boundary, inclusion interface or other defect, dissociating into two arrays of partial dislocations which repel each other but are severely restricted by the high energy of the faulted structure lying between the two arrays. It turns out that it is possible for these arrays to further dissociate if the fault energy becomes negative. Such a condition exists at the nucleation stage of a martensitic transformation. The underlying thermodynamic conditions necessary for this to occur are described in relation to the observed onset of transformation viz., Ms. As well, the structure changes resulting from defect dissociation mechanisms are described in terms of the associated burgers vectors of the partial dislocations.