You have almost certainly heard that human attention span has fallen to eight seconds, shorter than a goldfish. It is one of the most repeated statistics of the last decade, it appears in conference talks and newspaper columns, and it has no research behind it at all. Attempts to trace it lead to a marketing summary citing a source that never made the claim, and nobody has produced the goldfish study. The BBC and others have taken it apart in detail.
It is worth starting there, because the myth crowds out a real finding that is more interesting and considerably more actionable.
The number that is real: 47 seconds
Gloria Mark, an informatics researcher at UC Irvine, has spent two decades measuring what people actually do at computers — not self-reports, but logged screen activity and, in some studies, heart-rate and stress data alongside it.
The headline result: the average time people spend on a single screen before switching has fallen from roughly two and a half minutes in the mid-2000s to about 47 seconds in more recent work. That is not the capacity of human attention. It is what attention does in an environment engineered for interruption.
The distinction matters. "Attention span has shrunk" implies a biological decline. What the data shows is behavioural: the same brains, in an environment where switching is one click away and something is always waiting, switch constantly. People still read long novels and watch three-hour films. The capacity is intact; the conditions are hostile.
Why switching is expensive
Mark's earlier work with Gudith and Klocke put numbers on the cost of interruption, and the result was counter-intuitive. Interrupted work was often completed faster — people compensated by working at higher intensity. But they paid for it in stress, frustration and mental effort. You do not lose the time; you lose the ease.
The lingering half of the cost is attention residue: after switching, part of your mind stays on the previous task, and the new task gets a degraded version of you. Stack enough switches and you spend the day at partial capacity while feeling extremely busy. That is why a fragmented day can be exhausting and unproductive at the same time.
The related trap is that switching feels efficient. It is not — see multitasking and context switching.
The rhythmic view: attention is not a battery
A second finding from Mark's work is more encouraging. Attention is not a single resource that drains. It varies in kind through the day, and different kinds suit different work.
She describes states along two axes, engagement and challenge, producing recognisable modes: focused work, rote work, bored, and frustrated. Most people have a rough daily shape — for many, focus peaks late morning and again mid-afternoon, with a trough between. Trying to do demanding work in the trough tends to fail; doing rote work there tends to go fine.
The practical implication is not "focus harder all day". It is to match the work to the state, and to accept that a proportion of the day will be rote. Attempting eight hours of deep work is a plan that fails and then feels like a personal defect.
What actually extends attention
Ranked roughly by how much they matter:
- Remove the trigger, not the temptation. Notifications are the dominant cause of switching, and willpower is a poor defence against a device designed to interrupt. Turning them off is unglamorous and does more than any technique.
- Close the open loops before starting. An unfinished task keeps pulling — the Zeigarnik effect. Writing pending items down releases most of that pull, which is the whole argument for a trusted note inbox.
- Match the block to the state. Demanding work in your peak window, admin in the trough. See time blocking.
- Take real breaks. Mark's data suggests short, genuinely restorative breaks sustain attention better than pushing through, and that scrolling is not a break — it is another screen. See attention restoration theory.
- Reduce the load being carried. Attention spent holding things in mind is attention unavailable for the task; see cognitive load and external memory.
Two honest caveats
Averages hide enormous variation. A 47-second mean covers a distribution with long sessions in it. Your own pattern, your work, and your environment matter more than the population figure.
Not all switching is failure. Some jobs are genuinely interrupt-driven, and some creative work benefits from wandering — see mind wandering. The problem is involuntary switching during work that needs continuity, not switching as such.
What this means for note-taking
If attention now runs in short segments by default, the useful design response is to stop requiring long ones for capture. A system that only works when you have ten uninterrupted minutes will be empty, because those ten minutes are increasingly rare. A system that works in eight seconds will fill up.
That is the bet behind Clair Mind: one gesture to offload a thought, connections drawn afterwards rather than demanded up front. The environment is not going to get quieter, so the tools have to fit the attention people actually have.
More in attention and focus, deep work, and memory and focus.