Cognitive changes in normal aging have been explained by the frontal-executive hypothesis, but the assumptions made by this hypothesis concerning the neurobiological causes are still a matter of debate. Executive functions (EF) may activate neural networks that include disparate grey matter regions, and rely on the integrity of white matter connections. In 118 adults (50-90 years old) from the GENIE study, white matter integrity was measured using diffusion tensor imaging, and information processing speed, fluid intelligence and EF were assessed. A theory-driven structural equation model was developed to test associations between variables. The model was revised, removing non-significant paths. The adjusted model explained well the covariance in our data; and suggested that the reduction in white matter integrity associated with age directly affected only working memory. Fluid intelligence was mediated by all measured cognitive variables. The results suggest that white matter integrity may be particularly important for abilities activating complex neural networks, as occurs in working memory. Integration of the information processing speed and frontal-executive hypotheses may provide important information regarding common, unique, and mediating factors in cognitive aging.