Using the epoxy-aluminum wedge specimens defined and analyzed in Part 1, we measure the number of cycles required to initiate an interfacial fatigue crack near the apex, where the stress field is predicted to be singular. The eigenvalue, lambda, and the generalized stress intensity factor, Q, are varied via the wedge angle, and via the beam deflection, respectively. Crack initiation is detected using a strain gage bonded near the tip of the wedge. Following the methodology developed in Part 1, the fatigue data are then used to construct a fatigue initiation criterion characteristic of the bimaterial interface. This criterion is a 3-D surface, with the ordinate representing the generalized stress intensity factor and the two horizontal axes representing the number of cycles to initiation and the eigenvalue, respectively. Three key assumptions of the model are found to be satisfied in the specimens tested herein: (1) geometric imperfections at the apex are smaller than the singular region, (2) the plastic zones near the apex are also smaller than the singular region, and (3) the locus of initiation is near interfacial. Finally, a thermomechanical analysis indicates that the residual thermal stresses generated during the fabrication process make a significant contribution to the critical stress intensity factor. With high T-g adhesives and under unfavorable conditions thigh modulus, high CTE, poor adhesion), we predict that the residual stresses alone could be sufficient to cause debond initiation.