Crossword-doers decline more slowly. That is not the same as crosswords working.

People who do crosswords, read, and keep mentally busy tend to reach old age with sharper minds, and cohort studies find it repeatedly. The tempting conclusion — that the puzzles are doing the protecting — is one the evidence does not support, because the most likely explanation runs in the opposite direction. The earliest, silent phase of dementia can itself reduce the appetite for mentally demanding activity years before a diagnosis, which makes the activity look protective when it may partly be a symptom of who was still well.

This is reverse causation, and in dementia research it is not a hypothetical worry. It is visible inside the data usually cited as evidence that mental activity helps.

The finding that contains its own warning

A widely cited analysis in Neurology followed 1,157 older adults who were free of dementia at enrolment and had rated how often they took part in mentally stimulating activities [s1]. Over follow-up, 614 remained cognitively normal, 395 developed mild cognitive impairment and 148 developed Alzheimer's disease [s1].

Among those who stayed cognitively healthy, more cognitive activity tracked with markedly slower decline: each additional point on the activity scale was associated with a 52% lower annual rate of decline [s1]. Read alone, that is the headline the puzzle industry wants. But the same study found something that complicates it. Once people had developed Alzheimer's, higher prior cognitive activity was associated with faster decline — the annual rate rose by 42% for each point on the same scale [s1]. In the mild-cognitive-impairment group, activity was unrelated to the rate of decline at all [s1].

The study's authors read this pattern as a form of cognitive reserve compressing the illness — mentally active people staving off symptoms until later, then declining more steeply once the disease breaks through [s1]. That is a plausible and partly hopeful interpretation. But the same U-shaped pattern is exactly what you would expect if cognitive activity is, in part, a readout of underlying brain health rather than a cause of it. A brain already accumulating pathology may quietly withdraw from demanding activity first; the steep later decline is then the disease revealing itself, not the crosswords backfiring. The honest position is that these two readings are hard to separate, and the simple observational story — active people decline less, therefore stay active to decline less — assumes a causal direction the design cannot confirm.

What a fairer test would need, and what the best one showed

To get past reverse causation you have to randomise: assign some people to puzzles and some not, and follow them. Prevention trials in cognitively healthy people are long, expensive and rare. The closest rigorous evidence comes from a trial in people who already had mild cognitive impairment — and it was set up to answer a narrower question.

That trial randomised 107 participants with mild cognitive impairment to 78 weeks of either home-based computerised cognitive games or web-based crossword puzzles [s2]. The result surprised its own designers: crosswords did better. On the ADAS-Cog scale, where higher scores mean worse cognition, the games group worsened slightly while the crosswords group improved, a difference of 1.44 points in favour of crosswords (95% confidence interval, 2.83 to 0.06; P=0.04) [s2]. Mean ADAS-Cog scores moved from 9.53 to 9.93 for games and from 9.59 to 8.61 for crosswords, and MRI showed less shrinkage of the hippocampus and cortex in the crossword group [s2].

That is a genuine point in favour of old-fashioned puzzles over branded games — but note what it is not. It compared two active interventions in people already impaired; there was no do-nothing arm, and no cognitively healthy participants. It cannot tell you whether crosswords prevent dementia in a healthy 55-year-old, because it did not study prevention or healthy people [s2].

The honest answer

There is no strong evidence that doing crosswords, sudoku or brain puzzles prevents dementia in a healthy person, and the large observational associations that seem to say otherwise are heavily exposed to reverse causation and confounding — the puzzle-doers differ from non-puzzle-doers in education, health and habits, and incipient disease itself changes behaviour before it is diagnosed. What the evidence does support is smaller and still worth having: staying mentally engaged is associated with better cognitive ageing, mentally demanding puzzles are a reasonable way to do that, and in people with mild impairment, crosswords held up at least as well as a purpose-built game [s2].

The reason to do the crossword, in other words, is that you enjoy it and it keeps you engaged — not because the puzzle is buying down a dementia risk the data cannot show it touches.

This article is informational and is not medical advice.