Eye Movements May Reveal Which Form of Tau Pathology Is Affecting the Aging Brain

A six-minute eye-tracking test may offer a new way to distinguish biological forms of Alzheimer’s-related pathology in older adults with mild cognitive impairment, according to a study published in GeroScience. The research found that people with two different cerebrospinal-fluid biomarker profiles did not primarily differ in how large their eye responses were. Instead, they differed in when their eyes and pupils reached their maximum response while they performed an attention task. The result suggests that the timing of subtle eye movements could provide a non-invasive functional complement to lumbar puncture, brain imaging and emerging blood tests used to characterize neurodegenerative disease.

The study focused on the AT(N) framework, which classifies Alzheimer’s biology according to amyloid-β accumulation, phosphorylated tau and neurodegeneration. The researchers compared 38 people with mild cognitive impairment: 26 had an A+T+ profile, meaning that both amyloid and tau biomarkers were abnormal, while 12 had an A−T+ profile, indicating tau abnormalities without detectable amyloid pathology. A+T+ is considered the biological profile of Alzheimer’s disease under current research criteria. A−T+, by contrast, may reflect primary age-related tauopathy or another non-Alzheimer’s tauopathy, although cerebrospinal-fluid testing alone cannot determine the precise underlying cause.

Participants were recruited from two hospitals in Barcelona and had already undergone lumbar puncture as part of their clinical evaluation. The researchers classified them using validated, hospital-specific cerebrospinal-fluid thresholds for amyloid-β42, phosphorylated tau and total tau. Two participants with isolated total-tau elevation but no amyloid or phosphorylated-tau abnormality were grouped with the A−T+ participants because total tau can indicate neurofibrillary degeneration across several tauopathies. A sensitivity analysis excluding those two people produced broadly similar findings, although one borderline vergence measure no longer reached statistical significance.

During the experiment, participants viewed strings of meaningless letters on a laptop screen while a remote binocular eye tracker recorded their gaze and pupil diameter. Most strings were blue distractors, appearing on 80 percent of trials, while 20 percent were red targets. Participants were instructed to press a button whenever they detected a red string. The visual oddball task is widely used to study attention because rare, salient stimuli recruit systems involved in arousal, target detection and decision-making. The test lasted about six minutes, and the tracker sampled eye position 33 times per second—sufficient for the relatively slow vergence and pupil responses that unfold over roughly half a second to two seconds after a stimulus appears.

The investigators calculated cognitive vergence, the small coordinated change in the angle between the two eyes that accompanies attention and visual processing, as well as changes in pupil diameter. They extracted several characteristics from each response, including initial, global and late slopes; cumulative response area; peak amplitude; and time to peak. These measurements allowed the team to distinguish the strength of a response from its temporal organization. Statistical models accounted for repeated observations within individuals, while penalized logistic regression was used to examine whether participant-level response patterns were associated with biomarker profile. The analysis was exploratory rather than a diagnostic-classifier study, and the sample was too small to establish sensitivity, specificity or clinical accuracy.

Across the full group, red target stimuli produced larger vergence and pupil responses than blue distractors. Yet average response magnitude did not distinguish the A+T+ and A−T+ groups. The important differences emerged in the interaction between biological profile and stimulus condition. During distractor trials, people in the A−T+ group generally reached their vergence and pupil peaks later than those in the A+T+ group. During target trials, the pattern reversed: A+T+ participants showed the most delayed peak responses, whereas A−T+ participants responded relatively earlier. Vergence global slope also differed between profiles during target trials, with the A−T+ group showing steeper dynamics. These effects indicate that the groups did not simply differ in overall slowing. Rather, their timing changed differently depending on whether attention was required.

Behavioral performance followed the same condition-dependent pattern. The two groups performed almost identically when they had to withhold responses to distractors, with accuracy close to 100 percent. On target trials, however, A−T+ participants detected 89.7 percent of the red strings, compared with 82.5 percent among A+T+ participants. That difference was statistically significant. At the individual level, the difference between target and distractor timing was associated with the likelihood of belonging to the A+T+ group for both vergence time to peak and pupillary time to peak. The pupil association was particularly stable: its direction remained unchanged when each participant was removed from the analysis, and statistical significance persisted in 37 of 38 leave-one-participant-out tests. The vergence association was less robust, remaining significant in only eight of those 38 refits.

The researchers interpret the findings through the biology of the locus coeruleus, a small noradrenaline-producing nucleus in the brainstem that helps regulate alertness, attention and responses to salient events. Post-mortem studies suggest that the locus coeruleus is among the earliest sites of tau accumulation in Alzheimer’s disease. Its connections influence pupil control through pathways linked to the Edinger–Westphal nucleus and may affect vergence indirectly through the superior colliculus, which participates in three-dimensional eye-movement control. Pupil diameter is therefore often used as an indirect index of locus-coeruleus activity, although it is also affected by light, medication, autonomic function and other physiological factors.

One possible explanation is that isolated tau pathology and combined amyloid-tau pathology interfere with attention through partly different routes. Under low-demand distractor conditions, the A−T+ pattern of delayed responses could reflect altered tonic regulation of arousal by the locus coeruleus. Under target conditions, successful detection requires a rapid, coordinated response involving the dorsal and ventral attention networks, frontoparietal control systems, hippocampus and phasic locus-coeruleus signaling. Amyloid pathology is known to affect hubs of the default mode network, including the posterior cingulate cortex and precuneus, and may impair the suppression of that internally oriented network when goal-directed attention is needed. The additional cortical network disruption in A+T+ participants could help explain their delayed target-related eye and pupil responses and lower detection accuracy. The authors emphasize, however, that the study did not directly measure locus-coeruleus integrity, default-mode connectivity or compensatory brain activity, so this mechanism remains a hypothesis rather than a demonstrated cause.

The potential clinical appeal lies in the simplicity of the measurement. The protocol required a calibrated remote eye tracker and an ordinary computer, with no consumables, radiation or invasive procedure. A portable test of this kind could eventually help identify people who need more definitive biomarker assessment, particularly in settings where cerebrospinal-fluid analysis or positron-emission tomography is expensive or difficult to access. It could also add a functional dimension to blood biomarkers such as plasma phosphorylated tau, which estimate molecular pathology but do not show how the brain responds to cognitive demand. The eye signal might therefore be useful not because it replaces molecular tests, but because it captures the performance of attention and arousal networks in real time.

The evidence is not yet ready for clinical deployment. The study was cross-sectional, involved only 38 participants, lacked a cognitively unimpaired control group and included an imbalanced number of people in the two biomarker categories. The A−T+ group itself is biologically heterogeneous, potentially containing people with primary age-related tauopathy and other tauopathies. Medication use and autonomic dysfunction—both of which can influence pupil responses—were not fully assessed. Larger, balanced and independently replicated studies will need to test whether the timing signatures generalize across devices, languages, clinical populations and stages of disease. Longitudinal research will also be necessary to determine whether these eye-movement patterns track progression or treatment response, or appear before measurable cognitive decline. For now, the study offers a striking possibility: in the earliest stages of cognitive impairment, the brain’s molecular history may be reflected not in how dramatically the eyes react, but in the precise moment at which they do so.