<i>Background:</i> There are no data concerning the relative representation of clinical vascular risk factors and radiological lesions in cases that have been ruled in and ruled out for probable vascular dementia (VaD) according to NINDS-AIREN criteria. <i>Methods:</i> Three months after their index stroke, a psychiatrist interviewed patients and made a diagnosis of VaD according to both DSM-IV and NINDS-AIREN criteria for probable VaD. Patients who fulfilled the DSM-IV criteria for VaD were divided into two groups: those who were ruled in and ruled out according to NINDS-AIREN criteria as probable VaD. Demographic characteristics, vascular risk factors, clinical features of the index stroke and radiological findings were then compared between the two groups. <i>Results:</i> Of the 297 patients screened, 56 (18.8%) had a DSM-IV diagnosis of dementia. Among these demented patients, 55 (98.2%) and 22 (39.3%) fulfilled DSM-IV and NINDS-AIREN diagnosis of VaD, respectively. The concordance and level of agreement (kappa statistic) between DSM-IV and NINDS-AIREN diagnoses were 40% and 0.02, respectively. Reasons of failure to meet NINDS-AIREN criteria included the lack of temporal relationship between dementia and stroke (n = 20), the absence of focal neurological signs and/or radiological evidence of stroke (n = 6) and both of the above (n = 7). There was no significant difference between the above two groups in terms of demographic data, features of index stroke, vascular risk factors and CT scan findings, except that leukoaraiosis (p = 0.021) and bilateral lesions (p = 0.015) were more frequent in subjects diagnosed according to NINDS-AIREN criteria of probable VaD. The difference between these two groups with respect to the number of lesions was borderline for significance (p = 0.052). <i>Conclusions:</i> The use of NINDS-AIREN criteria for VaD for case selection in poststroke dementia research may exclude a number of subjects with VaD.