The short answer
Practicing a mental skill usually makes you better at that skill, and some of that gain can last for years. People who do puzzles regularly also tend to score better on thinking tests. But the research does not show that crosswords, sudoku or brain-training apps keep dementia away, and practice mostly makes you better at the task you practiced rather than sharper in general. The best reason to play is still the oldest one: you enjoy it.
The biggest trial: ten sessions, ten years
The best-known study is ACTIVE, a trial in six US cities that put people into groups at random. It gave 2,832 people living independently, with an average age of about 74, ten training sessions in memory, reasoning or speed of processing, or no training at all.[1][2]
Ten years later, the reasoning and speed groups still did better than the control group on tests of the skill they had trained. The memory group's advantage on memory tests had faded. All three trained groups also reported less difficulty with everyday tasks than the untrained group did.[1]
The limits matter. Everyone's scores declined over the ten years; the reasoning and speed groups simply declined less.[2] Only 44% of the original volunteers took part in the ten-year follow-up,[2] and the benefit for everyday tasks showed up in what people reported, not in the researchers' own tests of everyday tasks.[1] The training was a structured course, not a puzzle book, so ACTIVE tells us little about crosswords.
Brain-training apps: better at the app
In 2010, a study run through the BBC's Lab UK website asked 11,430 adults aged 18 to 60 to train online for six weeks. They got better at the training tasks, but there was no sign that the gains carried over to other mental skills, even ones that use similar parts of the brain.[3]
In 2016, a team of psychologists reviewed the published evidence, including the studies that brain-training companies themselves cite. They concluded that the training improves performance on the trained tasks, with less evidence for closely related tasks and "very little evidence" that it improves everyday thinking.[4] A 2020 Cochrane review of eight trials, with 1,183 healthy older people, found a slight improvement in overall thinking straight after 12 or more weeks of computer training, rated as low-quality evidence, and no evidence that it lasted a year.[5]
Regulators have acted on the marketing, too. In 2016 the maker of a well-known brain-training app agreed to pay $2 million to settle charges that it had advertised, without good evidence, that its games could delay age-related decline and protect against dementia.[9]
Crosswords: one interesting trial
Crosswords got a rare head-to-head test in 2022. Researchers at Columbia and Duke randomly gave 107 people with mild cognitive impairment, average age 71, either crosswords or brain-training games, both played on a computer at home, for 12 weeks with top-up sessions over 78 weeks. The crossword group did better on a standard test of thinking and memory (ADAS-Cog) at 12 and 78 weeks, and on everyday function at 78 weeks.[6]
It is one small study, in people who already had memory problems, and both groups were doing a mental activity: there was no group that did nothing. The authors themselves call for a larger trial with an inactive comparison group before anyone draws firm conclusions.[6] It's encouraging for crossword fans, and it is not proof.
Puzzlers score better. That's not the same as cause
Two large studies show that people who play more puzzles and games tend to do better on thinking tests. In the UK PROTECT study, more than 19,000 adults aged 50 and over who did word and number puzzles more often scored better on tests of attention, reasoning and memory. The lead researcher was careful: "We can't say that playing these puzzles necessarily reduces the risk of dementia in later life."[7]
In Scotland, more than 1,000 people born in 1936 were tested from the age of 70 to 79. Those who played more games such as cards, chess, bingo and crosswords scored better, and those who increased their game playing in later life declined less in memory and thinking speed, even after allowing for their childhood test scores, education and background. The researchers describe the link as "small but detectable".[8]
Studies like these can't separate cause from effect. People who stay sharp may simply keep playing, rather than playing keeping them sharp.
What the dementia experts focus on
The 2024 report of the Lancet Commission on dementia lists 14 risk factors that can be changed, which together are linked to nearly half of dementia cases: less education in early life, hearing loss, high blood pressure, smoking, obesity, depression, physical inactivity, diabetes, heavy drinking, head injury, air pollution, social isolation, high LDL cholesterol and untreated vision loss.[10]
Puzzles are not on that list by name. The Commission does recommend staying cognitively active in midlife, and it links social isolation in later life to an estimated 5% of cases, as much as less education in early life; the largest shares are hearing loss and high LDL cholesterol, at 7% each.[10] That is one reason we like games you play with someone: a weekly game of chess with an old friend is time spent together, whatever it does or doesn't do for the brain.