Ask most people what post-traumatic stress disorder is and the answer will be about memory — flashbacks, intrusive images, the past arriving uninvited. That description is accurate as far as it goes, and it is where the popular account stops. What the measurement side of the field has been doing for several years is different, and the direction of travel is worth noticing.

The Assumption

If the condition is essentially about memory, then assessment should be about memory too: what someone recalls, how intrusively, how often. That is broadly how clinical assessment works — a structured conversation about symptoms, conducted retrospectively.

It is not a foolish model. Intrusive re-experiencing is a defining feature, and asking is the only way to reach much of what matters.

But it has two limitations that are structural rather than incidental. It depends on recall of symptoms, which is itself affected by the condition. And it samples a person at the moments they happen to be in a clinic.

Both limitations point the same way. A person is asked, at an appointment, to summarise weeks they may have experienced as a blur, using a faculty the condition itself disturbs — and whatever happened on the nights and mornings between appointments is reconstructed rather than observed.

What Else Is Being Measured

A 2026 paper in npj Digital Medicine proposes a framework built on the observation that a great deal of what PTSD affects is neither verbal nor retrospective.

The authors propose a novel framework and measurement model that frames PTSD in the context of brain health capital, using advances in smart devices — wearables and phones — together with AI, to better capture the multimodal physiological, behavioural and cognitive features of the condition.

Their stated intention includes giving patients complex information in a meaningful way, with the aim of identifying personal change mechanisms and supporting empowered, patient-led decision-making.

The label matters and is easy to lose in summary: this is a proposal. The authors say so — they propose a framework and a measurement model. No new data are reported, no validation is presented, and nothing in it has been shown to improve assessment yet.

What Continuous Measurement Actually Found

A separate 2026 study did collect data of exactly the kind the framework describes, and its result is the strongest argument for the wider view.

It monitored 211 survivors of the Supernova music festival mass-trauma event and 113 matched comparison participants with wearable sensors tracking heart rate, activity and sleep–wake cycles, in real-world settings, for a month at two to six months post-event. PTSD was assessed again at eight to eleven months (Translational Psychiatry, 2026).

Survivors showed reduced circadian stability — greater interday variability in sleep timing.

Notice what that signal is not. It is not a memory, not a symptom report, not something anyone could have described in an interview. It is a pattern only visible by measuring the same person continuously across weeks, and it is covered at length in what the wearable data showed.

Why the Wider Frame Fits the Condition Better

Sleep, autonomic tone, attention and daily activity are not incidental accompaniments to PTSD; they are among the things people most consistently report as difficult, and they are continuously measurable in a way that recall is not.

A condition that disturbs sleep architecture, resting heart rate, attentional allocation and daily patterning is not well summarised as a disorder of memory, even if intrusive memory is its most recognisable feature.

There is also a difference in what the two approaches detect. An interview is unmatched for meaning — what an experience was like, what it changed, what someone is afraid of — but it samples a person occasionally and relies on recall. Continuous physiological measurement is unmatched for pattern: regularities and irregularities across weeks that nobody is positioned to notice about themselves, gathered without asking anything. Neither substitutes for the other, and the interesting proposals use both.

What Should Not Be Read Into This

A framework is not a finding. The 2026 proposal offers a way to organise measurement. It does not demonstrate that digital biomarkers improve diagnosis, prediction or treatment, and it should not be cited as though it did.

Wearable data is not diagnostic. The circadian study reports a group-level association in a research cohort after one specific event. No threshold exists, no device reports anything clinically meaningful about trauma, and reading a personal sleep-variability figure as a risk indicator would be a serious over-extension.

Continuous measurement raises its own problems. Data collected from someone’s phone and watch across months is intimate, and ‘AI to interpret it’ is a claim requiring validation that this proposal does not contain.

And a wider footprint is not a wider diagnosis. Measuring sleep, heart rate and activity in people with PTSD does not make every disturbance of those things a sign of trauma. They are affected by a great many ordinary conditions, which is exactly why a signal found in a research cohort does not transfer to an individual reading.

And none of it displaces the memory account. Intrusive re-experiencing remains central to the diagnosis. The argument is that it is not the whole of what the condition does.

What is genuinely supported: the measurable footprint of PTSD extends well beyond what a person can report, and the field is building the tools to look at the rest of it.