Vascular dementia (VaD) relates to different vascular mechanisms and changes in the brain and has different causes and clinical manifestations, reflecting complex interactions between vascular etiologies, changes in the brain, host factors, and cognition. Critical elements to the concept and diagnosis of VaD are defining the vascular causes, the vascular etiologies, and changes in the brain. Verifying the relation between brain lesions and cognition (i.e., the extent to which brain changes cause, compound, or coexist with cognitive impairment) and establishing the types, extent, side, site, and tempo of brain lesions that relate to incident cognitive impairment are major diagnostic challenges. Previous work on interactions between brain lesion and cognition in to cerebrovascular disease (CVD) have shown variation in the definitions and measures of cognitive impairment, in the techniques and methods used to reveal different brain changes, and in the selection of patient populations. Furthermore, small sample sizes and the absence of multivariate statistics have been design limitations. Accordingly, the different sets of criteria used and methods applied identify different numbers and clusters of subjects and different distribution of brain changes. Furthermore, this heterogeneity is reflected in variation in natural history such as the rate of progression of decline in different cognitive domains over time. All these factors have hampered optimal designs of clinical drug trials. A summary of generalizations regarding lesion and cognition interaction in VaD can be made. (1) Not a single feature, but a combination of infarct features--extent and type of white matter lesions (WMLs), degree and site of atrophy, and host factor characteristics--constitues correlates of VaD. (2) Infarct features favoring VaD include bilaterality, multiplicity (>1), location in the dominant hemisphere, and location in the limbic structures (fronto- and mediolimbic). (3) WML features favoring VaD are extensive WMLs (extensive periventricular WMLs and confluent to extensive WMLs in the deep WM). (4) It is doubtful that only a single small lesion could provide imaging evidence for a diagnosis of VaD. (5) Absence of CVD lesions on computed tomography or magnetic resonance imaging is strong evidence against a diagnosis of VaD. In forthcoming protocols on CVD-associated cognitive impairment, the following brain imaging features should be specified: detailed characterization of brain changes; use of possible predefined subtypes based on brain imaging; use of rating of vascular burden; defining the type and extent of WMLs favoring a diagnosis of VaD; defining the extent of medial temporal lobe atrophy disfavoring a diagnosis of VaD; and technical harmonization of methods of scanning and analysis.