A combined experimental and analytical approach is used to study damage initiation and evolution in three-dimensional second phase particle fields. A three-dimensional formulation of a damage percolation model is developed to predict damage nucleation and propagation through random-clustered second phase particle fields. The proposed approach is capable of capturing the three-dimensional character of damage phenomena and the three stages of ductile fracture, namely, void nucleation, growth, and coalescence, at the level of discrete particles. An in situ tensile test with X-ray tomography is utilized to quantify material damage during deformation in terms of the number of nucleated voids and porosity. The results of this experiment are used for both the development of a clustering-sensitive nucleation criterion and the validation of the damage percolation predictions. The evolution of damage in aluminum alloy AA5182 has been successfully predicted to match that in the in situ tensile specimen. Two forms of second phase particle field input data were considered: (1) that measured directly with X-ray tomography and (2) fields reconstructed statistically from two-dimensional orthogonal sections. It is demonstrated that the adoption of a cluster-sensitive void nucleation criterion, as opposed to a cluster-insensitive nucleation criterion, has a significant effect in promoting predicted void nucleation to occur within particle clusters. This behavior leads to confinement of void coalescence to within clusters for most of the duration of deformation followed by later development of a macrocrack through intracluster coalescence. The measured and reconstructed second phase particle fields lead to similar rates of predicted damage accumulation and can be used interchangeably in damage percolation simulations.