Working memory, and why complex tasks feel harder some days

Performance · updated 2026-09-05

Some days a task that is objectively routine feels impossible to hold together. Other days the same task is trivial. The most useful single concept for understanding that difference is working memory: the small, temporary workspace where you keep the parts of a problem while you work on it.

The workspace is small

The defining feature of working memory is how little fits. Estimates vary with how you measure it, but every estimate is small: a handful of independent items at once, not dozens. Everything else has to be written down, offloaded, or reconstructed from scratch each time you need it.

This is why interruptions are so expensive. The cost is not the thirty seconds the interruption takes. It is that the workspace is cleared and has to be rebuilt, and rebuilding it can take minutes.

Why difficulty is about structure, not size

A task with twenty steps that can be done one at a time is easy. A task with four steps that all depend on each other is hard. The difficulty is not the number of steps; it is how many things must be held simultaneously to make progress.

This explains a lot of otherwise puzzling experiences. Reading a dense paragraph is hard because each clause modifies the last and you cannot let any of them go. Mental arithmetic with carrying is hard because you hold a running total and a partial result at once. Debugging is hard because the cause and the symptom are far apart and you must keep both in view.

If a task feels harder than its description suggests, count how many things you have to hold at once. That number, not the step count, is what makes it heavy.

What actually helps

Because the constraint is capacity, the interventions that work are the ones that reduce what you have to hold, rather than the ones that try to expand the workspace.

• Externalise aggressively. Anything written down is something you are not holding. A scrappy diagram beats a tidy mental model.
• Close the loops you can close. An unfinished sub-task you are keeping in mind costs capacity until it is either finished or written down.
• Batch by shape, not by topic. Switching between two tasks that need different mental setups costs more than the clock says.
• Do the dependent parts first. When a task has both a tangled core and routine edges, spend your best attention on the core; the edges survive interruption.
• Reduce ambient load. A visible notification you are choosing to ignore is still occupying part of the workspace.

On brain training

The obvious question is whether you can simply train working memory upward. This has been studied a great deal, and the honest summary is that people reliably get better at the trained task, while transfer to different, untrained tasks has been much harder to demonstrate. Claims that a training app will make you generally smarter run well ahead of the evidence.

That is not a reason to avoid the tasks, which are good at showing you your own limits. It is a reason to be sceptical of anything sold on the promise of raising a general capacity.

Why it varies day to day

Available capacity is not fixed. Short sleep, competing worries, and cognitive fatigue from earlier in the day all reduce what you can hold, which is why the same task can be easy on Tuesday and impossible on Thursday. Recognising that a bad day is a capacity problem rather than a character problem is genuinely useful: the right response is to switch to work that needs less holding, not to try harder at the work that needs more.