In the absence of a mechanical defect at the surface, failure is almost always initiated in the bulk of a sample at a material flaw. A variety of flaw morphologies are observed; however, the extent of cyclic damage could not be related to either flaw size or flaw type. Energy dispersive spectographic analysis did not reveal any differences in chemistry between a flaw site and regions remote from a flaw. The size and definition of a flaw increase with an increase in carbon-black loading. Initiation flaw sites are enveloped by fan-shaped or penny-shaped regions which develop during cycling. With the highly loaded carbon-black samples, these regions consist of concentric bands which increase in number and become more prominent with an increase in test temperature. This is attributed to the presence of knotty tearing. With low carbon-black loadings, the absence of any significant disorientations in the tearing paths during either a cycle or a series of cyclic events results in the absence of individual resolvable bands within the fatigue-tear regions. The size and morphology of a fatigue-tear region appears to be independent of the fatigue load or the extent of the damage (strength loss). By contrast, either an increase in cycling load or an increase in damage at constant load increases the definition of the fatigue-region morphology for all formulations of carbon black. The morphologies in regions leading to failure during the final cycle are identical to the morphologies of fractured surfaces obtained in conventional tensile tests—tear steps or ridges at low magnifications and slow crack-growth with coarse tear paths and fast crack-growth with fine tear paths at high magnifications. Morphologies typical of slow crack-growth are also seen within the banded (fatigue) regions. On the finest scale, the morphology can be described in terms of tearing of individual groups of rubber strands, collapsing to form a cell-like structure.