Antibodies to multiple epitopes spanning the length of the tau molecule were used to study Alzheimer neurofibrillary tangles (NFT) using immunocytochemical methods and several different methods of fixation and tissue processing, including staining of vibratome sections, hydrated autoclaving of paraffin sections and immunofluorescence of NFT isolated from fresh brain tissue. Smears and sections were pretreated with trypsin and/or phosphatase to further characterize antibody binding. In tissue fixed briefly in periodate-lysine-paraformaldehyde, tau immunoreactivity was detected in astrocytes, but only a few tau epitopes were detected in NFT with this fixation method. In contrast, all tau epitopes were detected in NFT in tissue fixed in formaldehyde for prolonged periods of time. In the hippocampus, the number of NFT detected in the dentate fascia was in proportion to the duration of dementia, as we previously noted. Dentate fascia NFT were intracellular (i-NFT) and were reactive with antibodies recognizing epitopes in both the carboxy- and amino-terminal regions of tau, but not the microtubule-binding domain of tau, suggesting that microtubule-binding domain epitopes are hidden in i-NFT. In contrast, NFT in the subiculum and layer II of the parahippocampal cortex were mostly extracellular (e-NFT), especially in severe cases of long duration. e-NFT were immunoreactive with antibodies to the microtubule-binding domain, but only weakly reactive with antibodies to carboxy- or amino-terminal epitopes, suggesting that e-NFT may contain fragments of tau. In both isolated NFT and NFT in sections, amino-terminal epitopes, including the Alz-50 epitope, were sensitive to trypsin proteolysis, which suggests that the lack of staining of e-NFT by antibodies to the amino-terminal regions of tau is due to proteolysis. Antibodies reactive with amino-terminal epitopes also stained fewer NFT following hydrated autoclaving, while those reacting with the carboxy half of tau stained more NFT after hydrated autoclaving. Thus, although carboxy-terminal regions are not detected in e-NFT, they are probably masked, rather than proteolytically cleaved, since they can be revealed by hydrated autoclaving. Finally, phosphatase treatment of isolated NFT revealed enhanced immunostaining not only with Tau-1, as in previous studies demonstrating abnormal phosphorylation of tau proteins in NFT, but also with an antibody to exon 2, which reveals yet another phosphorylation site in tau of NFT.