The number 10,000 came from a 1960s Japanese pedometer ad. The number 9,826 came from a single observational study that can't prove what the headlines claim. Here's what the evidence actually supports
Jose-Alberto, the question your piece raises by accident is why the number keeps outliving its own evidence. I think it is because the number was never really doing evidential work. It was doing emotional work.
Dementia is the worst kind of fear to sit with, slow and impossible to picture in advance. A step target converts that into something you can finish by bedtime, and the watch hands you a verdict every night. Swapping 10,000 for 7,000 changes nothing that matters, because the problem was never the figure but the promise it is asked to carry: protection from a future no trial has shown walking can buy. The count becomes a receipt for something the evidence has not sold.
Exactly right. The number was never built to carry evidential weight, it was recruited to carry it after the fact, because something had to fill the gap. Surveys in the UK and elsewhere consistently put dementia first or second among the most feared health conditions, usually trading places with cancer, with stroke or vision loss competing for third depending on the study. Part of why is structural: you can lower your LDL and watch the number move, and feel that the risk vanishes, or at least goes down. With dementia we don't have something that easy, at least not yet (we now have blood-based tests for AD, but whether they move with changing risk factors is still unknown). So most modifiable risks at least offer something to check. Dementia, so far, offers nothing.
A step count fills that gap with something checkable. There's a name for the mechanism in psychology: the "illusion of control," the tendency to treat a controllable, proximal action as if it purchases protection against a distal, uncontrollable outcome, even with no evidence the two are connected. The watch isn't measuring dementia risk. It's measuring whether the ritual got performed.
I'd push your point one step further into my own piece's territory: even the receipt is counterfeit at the measurement level, not only the evidential one. The number on someone's wrist carries 23-28% error against research-grade devices in free-living conditions. So it isn't just an unvalidated promise, it's an unvalidated promise measured wrong. Whether wearables are useful for anything is worth its own post, and I'll take it on later this year.
However, I don't think that makes the number completely useless. Behavior-change research shows that an arbitrary but trackable target beats a vague instruction like "move more" for adherence, regardless of whether the target has any biological grounding. So the count may be epidemiologically empty and behaviorally effective at the same time, which is an uncomfortable place for my own piece (built on "the evidence doesn't support this number") to land, since the piece evaluates the evidence and not the coping mechanism it's tangled up with.
That last paragraph reads to me as the honest place to land, and it does not cost you the piece. The two findings only seem to collide because they sit on different axes: one asks whether the number is true, the other whether it works. Your piece goes after the first, where the harm actually lives, and the behavioral usefulness is what gives the false promise its cover. The target keeps people moving well enough that nobody notices it is also claiming, silently, to buy down a risk no trial has priced. Debunking that claim does not abolish the tool. It lets the count be prescribed honestly, as a way to get someone moving, without the watch pretending it also buys protection. So the discomfort may be the next post rather than a flaw in this one: the evidence clears the ground the coping mechanism stands on.
Jose-Alberto, the question your piece raises by accident is why the number keeps outliving its own evidence. I think it is because the number was never really doing evidential work. It was doing emotional work.
Dementia is the worst kind of fear to sit with, slow and impossible to picture in advance. A step target converts that into something you can finish by bedtime, and the watch hands you a verdict every night. Swapping 10,000 for 7,000 changes nothing that matters, because the problem was never the figure but the promise it is asked to carry: protection from a future no trial has shown walking can buy. The count becomes a receipt for something the evidence has not sold.
Exactly right. The number was never built to carry evidential weight, it was recruited to carry it after the fact, because something had to fill the gap. Surveys in the UK and elsewhere consistently put dementia first or second among the most feared health conditions, usually trading places with cancer, with stroke or vision loss competing for third depending on the study. Part of why is structural: you can lower your LDL and watch the number move, and feel that the risk vanishes, or at least goes down. With dementia we don't have something that easy, at least not yet (we now have blood-based tests for AD, but whether they move with changing risk factors is still unknown). So most modifiable risks at least offer something to check. Dementia, so far, offers nothing.
A step count fills that gap with something checkable. There's a name for the mechanism in psychology: the "illusion of control," the tendency to treat a controllable, proximal action as if it purchases protection against a distal, uncontrollable outcome, even with no evidence the two are connected. The watch isn't measuring dementia risk. It's measuring whether the ritual got performed.
I'd push your point one step further into my own piece's territory: even the receipt is counterfeit at the measurement level, not only the evidential one. The number on someone's wrist carries 23-28% error against research-grade devices in free-living conditions. So it isn't just an unvalidated promise, it's an unvalidated promise measured wrong. Whether wearables are useful for anything is worth its own post, and I'll take it on later this year.
However, I don't think that makes the number completely useless. Behavior-change research shows that an arbitrary but trackable target beats a vague instruction like "move more" for adherence, regardless of whether the target has any biological grounding. So the count may be epidemiologically empty and behaviorally effective at the same time, which is an uncomfortable place for my own piece (built on "the evidence doesn't support this number") to land, since the piece evaluates the evidence and not the coping mechanism it's tangled up with.
That last paragraph reads to me as the honest place to land, and it does not cost you the piece. The two findings only seem to collide because they sit on different axes: one asks whether the number is true, the other whether it works. Your piece goes after the first, where the harm actually lives, and the behavioral usefulness is what gives the false promise its cover. The target keeps people moving well enough that nobody notices it is also claiming, silently, to buy down a risk no trial has priced. Debunking that claim does not abolish the tool. It lets the count be prescribed honestly, as a way to get someone moving, without the watch pretending it also buys protection. So the discomfort may be the next post rather than a flaw in this one: the evidence clears the ground the coping mechanism stands on.