There is a patch of tissue sitting low in the brain, behind the eyes and above the roof of the mouth, that no one can point to on a standard scan without help. It is called the basal forebrain, and it is less a structure than a scattered neighbourhood: the nucleus basalis of Meynert, the diagonal band of Broca, the septal nuclei. Its cells manufacture acetylcholine and pipe it out across the cortex, which is why the oldest drugs still prescribed in Alzheimer's care, the cholinesterase inhibitors, exist at all. It is also, by long-standing neuropathology, one of the first places where tau tangles appear, and the cholinergic neurons there degenerate selectively in Alzheimer's disease. For decades, the price of admission to this region was a post-mortem slide. In living people, tau PET imaging has mostly stayed where the measurement is easy: the entorhinal cortex and the wider temporal lobe.
The reason is not indifference. It is physics and chemistry. The basal forebrain is small, PET resolution is coarse, and the region sits next to the thalamus and pallidum, where some tau tracers produce off-target signal, a bright reading that is not tau at all. Any number you pull out of that neighbourhood could be spillover from a neighbour. A methods paper published open access in Neuropsychopharmacology on 22 September 2026, from Theresa Harrison, Trevor Chadwick, William Jagust and colleagues at the University of California, Berkeley and Lawrence Berkeley National Laboratory, set out to test whether the number can be trusted.
The design is a cross-sectional feasibility study, not a treatment trial and not a prediction study. The team pooled scans from 1397 cognitively unimpaired older adults in two separate cohorts imaged with two different tracers: 509 people scanned with flortaucipir, drawn from the Alzheimer's Disease Neuroimaging Initiative (416) and the Berkeley Aging Cohort Study (93), and 888 people scanned with MK6240 from the POINTER Imaging study. The two tracers matter because they fail differently; their off-target signal patterns are distinct, so a result that survives both is less likely to be an artefact of either. Signal was quantified in three categories of region: the basal forebrain itself, the usual temporal tau regions (entorhinal cortex and the temporal MetaROI), and off-target regions adjacent to the basal forebrain.
The logic of the test is elegant in its plainness. If basal forebrain readings were just noise bleeding in from next door, they should behave like the off-target regions. They did not. Basal forebrain tau and temporal tau were both associated with global amyloid burden; the off-target regions were not. Basal forebrain tau and medial temporal lobe tau showed similar associations with age, with basal forebrain volume, with hippocampal volume and with multi-domain cognition, and those associations held in both the flortaucipir cohort and the MK6240 cohort. In whole-brain morphometry, medial temporal tau was negatively correlated with basal forebrain volume: more tau in one place, less tissue in the other. The authors' conclusion is deliberately narrow, that it is feasible to use PET to measure tau signal in the basal forebrain in unimpaired older adults.
For a family living with a diagnosis, nothing changes this week. No scan protocol has been approved on the strength of this, no clinic will report a basal forebrain tau number tomorrow, and the people studied here had no cognitive impairment at all. What changes is the research vocabulary. The cholinergic system has been the target of Alzheimer's drugs since the 1990s while remaining largely invisible to the imaging that now stages the disease. If a tau reading can be taken from that hub in living people, then questions that were previously answerable only at autopsy become askable in a scanner: whether tau arrives there before or alongside the temporal lobe, whether its arrival tracks the thinning of the region, and eventually whether any of it predicts who declines. Those are the next studies, and this paper is the ruler, not the result.