The fatigue-life variability in an ␣ ϩ  Ti alloy (Ti-6Al-4V) has been examined through a probabilistic micromechanical model that treats the crack-initiation and growth processes at the grain-size level. First, a physics-based crack-initiation model is described. This is followed by a summary of a physicsbased fatigue-crack-growth model. The combined model is applied to predict the variability of crack initiation and growth lives due to microstructural variations in Ti-6Al-4V. Finally, possible fatigue mechanisms or scenarios that can lead to the worst-case fatigue life are elucidated via probabilistic modeling of the fatigue-crack-initiation process, the driving force of the grain-sized cracks, as well as the intrinsic (closure-free) threshold and the closure-affected threshold of the small cracks. In the absence of preexisting cracks, the worst-case total fatigue life is obtained when two conditions are met:(1) the crack size at initiation is on the order of 1 to 2 times the grain size, and (2) the driving force (applied ⌬K ) exceeds the intrinsic threshold of the small cracks. The probabilistic results are also used to elucidate the conditions for the occurrence of dual fatigue limits in high-cycle fatigue (HCF) or giga-cycle fatigue.