Ask which puzzle is better for your brain and you will get a confident answer from almost every source that has an interest in one of them. The accurate answer is duller and more useful: nobody has run the comparison. There is no study in which people were randomised to Sudoku or to crosswords and followed to see which group fared better. Every claim that one beats the other is an inference across studies that used different people, different measures, and different decades.
What is true is that crosswords have accumulated considerably more supporting evidence. That asymmetry is real, and it is worth understanding precisely, because the reason for it turns out to have very little to do with crosswords being better. As with the broader question of puzzles and cognition, what follows is a summary of published research for general interest, not medical advice.
Why crosswords have the head start
The landmark studies in this field are long-running cohorts that recruited older adults and followed them for years or decades. Their questionnaires asked about the leisure activities that existed when the cohort was enrolled — reading, cards, board games, musical instruments, crosswords.
Sudoku was not on those lists, and could not have been. The Times of London printed its first Sudoku on 12 November 2004, and the global boom followed in 2005 — a chronology covered in the history of the puzzle. The cohorts that produced the famous crossword findings had been collecting data for years by then. Crosswords are better evidenced because researchers had the opportunity to measure them, not because anyone weighed the two and found crosswords stronger. In a field where the evidence takes decades to accumulate, a thirty-year head start is an enormous advantage that says nothing about the puzzles themselves.
What the crossword findings actually say
Two results from the Bronx Aging Study do most of the work in popular coverage, and both are more interesting than their headlines.
Joe Verghese and colleagues, writing in the New England Journal of Medicine in 2003, followed older adults and found that those doing crossword puzzles four days a week had a dementia risk roughly 47% lower than those doing them once a week. Reading, board games and playing a musical instrument showed similar associations; among physical activities, only dancing did. That last detail is a useful corrective to the way the study gets cited — the finding was about cognitively engaging activity broadly, and crosswords were one instance of it rather than a uniquely protective one.
The follow-up is subtler. Pillai and colleagues, in 2011, looked specifically at the people in that cohort who went on to develop dementia, and asked when their memory decline began to accelerate. Crossword participation at baseline was associated with that acceleration starting about 2.54 years later. Read carefully, this is not a story about preventing disease. It is the signature of cognitive reserve: the same underlying pathology, tolerated longer before it becomes visible. Worth having, and a different claim from the one the headlines make.
Both are observational, and inherit the interpretive problems that come with that — reverse causation, confounding by education and health, and the possibility that early undiagnosed decline is what stops people doing puzzles rather than the reverse.
The one trial that actually randomised anyone
In 2022, Devanand, Doraiswamy and colleagues published a genuine randomised controlled trial in NEJM Evidence. Adults with mild cognitive impairment, average age around 71, were assigned to twelve weeks of intensive home-based training on either web-based crossword puzzles or computerised cognitive-training games, with booster sessions, and were followed for 78 weeks.
Crosswords won. On the primary cognitive measure, the crossword group improved by week 78 while the computer-games group worsened, with benefits also appearing on functional measures and less brain shrinkage on imaging. In a literature where controlled trials routinely find nothing, that is a genuinely notable result.
Note carefully what it does and does not establish. The comparison was crosswords versus commercial-style brain-training games — the very products the brain-training reviews have repeatedly found wanting. Sudoku was not an arm of this trial. The result is strong evidence that a rich, self-directed, effortful puzzle beat a gamified training app in one clinical population. It is not evidence that crosswords beat Sudoku, because that experiment has never been done.
Where Sudoku does have evidence
The best data on Sudoku-like puzzles comes from the PROTECT cohort in the UK, which in 2019 published two companion papers on roughly 19,000 adults aged 50 to 93. One examined self-reported frequency of word puzzle use; the other examined numberpuzzle use. Both found that more frequent puzzling was associated with better performance across measures of attention, reasoning and memory, and the number-puzzle associations were statistically strong across all fourteen cognitive measures tested.
Two limitations matter, and they are the reason this cannot settle the argument. The analyses were cross-sectional — a snapshot relating puzzle habits to cognitive scores at the same moment, which cannot tell you which came first. And although the two papers used the same cohort, word and number puzzles were analysed separately rather than pitted against each other. Even here, where the same researchers had both kinds of data in hand, no head-to-head comparison was reported.
What the two puzzles actually ask of you
This is the part that is genuinely knowable, because it does not depend on outcome studies at all. The two puzzles load different machinery.
- A crossword draws on stored knowledge. Vocabulary, general knowledge, wordplay conventions, and the retrieval of a word you know but cannot immediately reach. It rewards what psychologists call crystallised ability, which is why crossword skill tends to hold up well with age and why a lifetime of reading pays off in it directly.
- A Sudoku draws on almost no stored knowledge. The digits are arbitrary labels; you could solve the identical puzzle with nine colours. What it demands is constraint reasoning held in working memory — tracking what a cell cannot be, keeping several partial deductions alive at once, and noticing a structure across a region of the grid.
That difference has a practical consequence. A crossword can be hard because you do not know the answer, and no amount of reasoning will retrieve a fact you never learned. A Sudoku is never hard in that way: everything needed is present in the grid, which is why being stuck is always a failure of seeing rather than of knowing. Whether that makes it better for you is unknown; it does make it a different exercise, and it makes the two poor substitutes rather than interchangeable options.
The practical answer
Pulling it together honestly:
- No study has compared them, so anyone who tells you which is better is extrapolating.
- Crosswords have more evidence, largely because they were available to be studied decades earlier.
- The strongest trial in the area supports crosswords over brain-training games, not over Sudoku.
- What evidence exists for both points the same direction: frequency and engagement matter more than which puzzle you chose.
Which suggests the honest recommendation is not to pick a winner. If the mechanism that matters is sustained, effortful engagement — the case laid out in why easy Sudoku stops working your brain — then the puzzle you will still be doing in five years beats the one that is theoretically superior and quietly abandoned in March. Doing both is entirely reasonable, and on the evidence available it is at least as defensible as agonising over the choice.
If Sudoku is the one that holds you, the way to keep it demanding is to keep meeting patterns you have not automated. The learning-order guide sequences them, and the practice pages let you drill a specific technique rather than waiting for one to appear.
Sources referred to
- Verghese, J., et al. (2003). Leisure activities and the risk of dementia in the elderly. New England Journal of Medicine, 348, 2508–2516.
- Pillai, J. A., et al. (2011). Association of crossword puzzle participation with memory decline in persons who develop dementia. Journal of the International Neuropsychological Society, 17(6), 1006–1013.
- Devanand, D. P., et al. (2022). Computerized games versus crosswords training in mild cognitive impairment. NEJM Evidence, 1(12).
- Brooker, H., et al. (2019). An online investigation of the relationship between the frequency of word puzzle use and cognitive function in a large sample of older adults. International Journal of Geriatric Psychiatry, 34(7), 921–931.
- Brooker, H., et al. (2019). The relationship between the frequency of number-puzzle use and baseline cognitive function in a large online sample of adults aged 50 and over. International Journal of Geriatric Psychiatry, 34(7), 932–940.
Summaries above are the author's reading of these papers and are simplified for a general audience; the originals carry caveats worth reading. Corrections are welcome via the about page.