Leakage of serum proteins into the brain parenchyma has been repeatedly used as evidence of blood-brain barrier (BBB) damage in experimental and human studies. However, there is no consensus in the literature concerning this phenomenon in Alzheimer's disease (AD). We have examined this question by comparing frontal lobe sections in seven groups of patients: Multi-infarct dementia (n = 6), AD with (n = 10) and without (n = 10) infarcts, age-matched controls with (n = 10) and without (n = 10) infarcts, controls with neurodegenerative diseases other than AD, and young controls (n = 10). An additional series compared prospectively followed patients with a diagnosis of either multi-infarct dementia (n = 5) or AD (n = 4). Albumin was detected in white-matter astrocytes in all cases, without significant variation in intensity. In addition, diverse combinations of neurons, astrocytes, and (in AD patients) senile plaques were present in the cerebral cortex in an inconsistent manner. Semiquantitative analysis showed no statistically significant differences among groups. Anti-IgG labeled astrocytes in infarcts only. Complement C3c component was detected in rare amyloid plaques in a minority (15%) of AD cases. Selective labeling of AD-specific lesions in a patchy manner was observed for serum amyloid P. We conclude that there is no immunohistochemical evidence of alteration of the BBB in Alzheimer's disease with or without vascular factors or in old age. Serum amyloid P binds avidly to AD lesions, but our findings are consistent with leakage through the BBB during the agonal or immediate postmortem period. Finally, no specific pattern of abnormality in the BBB was detected in multi-infarct dementia.