Abstract Cerebral white matter hyperintensities (WMH), detected in vivo with magnetic resonance imaging (MRI), are commonly used to assess cerebrovascular burden in cognitive impairment. However, the association between WMH and cognition is not consistent across the literature. The present review examines evidence from published longitudinal studies. We reviewed the PubMed data base from January 1990 to March 2013 and included studies investigating the association of WMH with (1) the risk of dementia in the general population, (2) the risk of conversion to dementia in the mild cognitive impairment (MCI) population, and (3) cognitive decline in the general population. WMH were associated with all types of dementia in the general population, but not in MCI patients. Results are discrepant for global decline. WMH appear to be early predictors of the risk of dementia, but this association appears to be modulated by cognitive reserve, age and the spatial distribution of lesions. There are, however, some limits in the use of WMH as a marker of vascular burden. In addition to their ischaemic origin, WMH may be the result of co-occurring morbidity. Further research is needed to elucidate to what extent WMH actually reflect vascular risk to evaluate the likely efficacy of interventions specifically targeting WMH reduction.