Sudoku improves the skills it uses, and may modestly improve a few closely related ones. It does not appear to broadly upgrade memory. The honest reading of the research is more interesting than the wellness-blog version: regular Sudoku is consistent with the cognitive-reserve hypothesis and tracks with better cognitive scores in older adults, but it does not transfer to abilities outside its own narrow exercise.
This post separates what controlled and large observational studies support from what gets implied.
The short answer. Sudoku trains working memory under interference, logical reasoning, and visual scanning. In a 2019 study of more than 19,000 adults aged 50 to 93, frequent number-puzzle use was associated with better short-term memory, working memory, and attention. The link is real and the effect sizes are modest. Sudoku is a genuinely good habit; it is not a memory cure.
What Sudoku actually trains
A Sudoku grid is a constraint-satisfaction puzzle. You hold a candidate digit in working memory, scan rows, columns, and boxes for conflicts, and revise. Done repeatedly, that drills three cognitive abilities well:
- Working memory under interference. You are tracking multiple candidate digits while the visual scene presents distractors. This is exactly the load that working-memory tests measure.
- Logical reasoning. Sudoku rewards if-then chains, eliminating options based on combinations of clues. This is a different skill from word retrieval, the one crosswords mostly train.
- Sustained visual attention. Solving a hard puzzle requires holding focus for tens of minutes on a small region of a small grid.
What Sudoku does not heavily train is vocabulary, name and face recall, prospective memory, or attention under emotionally loaded interference. Those are different abilities, exercised by different tasks.
What the PROTECT data shows
The largest single dataset on Sudoku and cognition comes from PROTECT, an online cohort of UK adults aged 50 to 93 maintained by researchers at the University of Exeter and King's College London. In 2019, Brooker and colleagues analyzed 19,078 PROTECT participants, asking how often they did number puzzles and how well they performed on a cognitive battery covering memory, attention, and executive function.
The finding: adults who reported doing number puzzles more frequently showed higher scores on tests of grammatical reasoning, short-term memory, working memory, attention, and executive function. The effect was dose-dependent. Daily puzzlers scored higher than weekly puzzlers, who scored higher than less-than-monthly puzzlers. The authors reported the equivalent of being eight to ten years younger on some tests for the most frequent puzzlers compared with the least frequent.
"The more regularly participants engaged with the puzzles, the better their performance," wrote the team. The pattern held across age bands and after adjusting for education.
Two important caveats stay with this finding. First, this is associational. The study cannot tell whether Sudoku makes older adults sharper, or whether sharper older adults are more likely to do Sudoku, or both. Second, the effect is on the kinds of tests Sudoku rehearses (working memory, attention, executive function). The link is weaker for episodic memory, the system most relevant to remembering people, conversations, and events.
What the near-versus-far-transfer problem means here
The central debate in cognitive training, covered in our cognitive training guide, is whether training one skill spills over to others.
- Near transfer is improvement on tasks similar to the trained one. Sudoku reliably produces near transfer to other working-memory and reasoning tasks.
- Far transfer is improvement on a meaningfully different ability. Sudoku transferring to remembering names or appointments, for example. The evidence for far transfer in puzzle-style training is weak.
Daniel Simons and colleagues stated this directly in their 2016 consensus review: "Brain-training interventions improve performance on the trained tasks." The honest interpretation of the Sudoku research follows the same shape. Frequent Sudoku makes you better at Sudoku and at closely related lab tasks. It does not appear to make you generally smarter or better at memory in everyday life.
Sudoku versus crosswords versus targeted training
If the question is what to do with a daily 15 minutes of mental engagement, the meaningful differences are below:
| Activity | Main skills trained | Best evidence | Common limit |
|---|---|---|---|
| Crosswords | Vocabulary, word retrieval | NEJM Evidence 2022 trial in MCI; large observational links | Trains word knowledge, not working memory or processing speed |
| Sudoku | Working memory under interference, logic, visual scanning | PROTECT 2019 (n=19,078); associational | Trains the puzzle; weak far transfer |
| Targeted cognitive training (e.g., ACTIVE-style) | Processing speed, working memory, attention | ACTIVE trial 10-year follow-up; randomized | Requires structured, adaptive sessions, not casual play |
The honest hierarchy from the literature: if you are mostly worried about word-finding, crosswords are well matched. If you enjoy logic and want to hold working memory in shape, Sudoku is well matched. If you want the kind of training with the strongest randomized evidence for dementia-risk reduction, that is structured speed-of-processing training, not a puzzle book.
What the cognitive-reserve angle adds
The observational link between mentally demanding activity and slower cognitive decline is a real one, and Sudoku fits inside that pattern. John Verghese and colleagues' landmark 2003 New England Journal of Medicine study followed 469 older adults for over five years and found that frequent participation in cognitive leisure activities, including board games and puzzles, was associated with a significantly lower risk of dementia.
The mechanism the field has converged on is cognitive reserve: the brain's ability to compensate for age-related changes through richer neural networks built up over a lifetime of demanding mental work. Our cognitive reserve article goes into the framework. The short version: people with more education, more cognitively demanding work and hobbies, richer social lives, and lifelong learning show better-preserved function later in life, even when underlying disease is present.
Sudoku contributes to that reserve. So do crosswords, language learning, music practice, complex hobbies, and demanding social conversation. There is no evidence that Sudoku specifically is uniquely protective. The pattern is about consistent mental engagement, not about any single puzzle format.
How much, and what kind?
If you want to do Sudoku for its cognitive benefits, the research points to a few practical rules:
- Daily beats weekend marathon. The PROTECT 2019 dose-response was monotonic in frequency. Daily puzzlers outperformed weekly puzzlers on every test in the battery.
- Push your difficulty. A Sudoku you solve on autopilot trains less than one that sits at the edge of your competence. This mirrors the broader cognitive-training principle that adaptive difficulty matters.
- Pair it with other domains. Sudoku trains logic and working memory. It will not improve word retrieval or face recall. If those are the lapses that worry you, add training that targets them specifically.
The honest takeaway
Does Sudoku improve memory? In a specific way: it strengthens working memory under interference, logical reasoning, and visual scanning, and the observational link to better cognitive scores in older adults is real. It does not broadly upgrade everyday memory. Like crosswords and other mentally demanding hobbies, it earns its place as one good habit among several.
If the lapses you are trying to fix are name recall, face recognition, lists, and focus under interference, those need their own targeted training. That is the design space BrightYears was built around, brief daily memory practice for adults 45+, on the everyday skills that matter, in about five minutes a day.