What different word puzzles actually train, according to research

The rule: you get better at what the puzzle makes you do

The clearest evidence comes from people who have practised one word game for years. Competitive Scrabble players, tested at the University of Calgary, were faster than matched non-players at deciding whether a string of letters is a real word, and less slowed down when the word was printed vertically, the way it often sits on a board (Hargreaves and colleagues, 2012). They were also better at anagramming and at producing words from a given letter. All of that is Scrabble, practised.

A follow-up by the same group then asked whether the expertise spreads further. On a task that looked similar but used symbols instead of letters, the Scrabble experts performed no better than anyone else, even though brain imaging showed them approaching it differently (van Hees and colleagues, 2016). The skill was real, and it stayed close to home.

That is the pattern across this whole area of research, and it is good news rather than bad. It means puzzles are not interchangeable. If you want a particular skill, the question to ask of a puzzle is simply what it makes the solver do, and the rest of this article goes through the common ones.

Crosswords: recalling words from their meaning

A crossword clue gives you a meaning and the grid demands the word. That makes a crossword a form of what psychologists call retrieval practice: pulling something out of memory rather than looking at it again. It is one of the most thoroughly replicated findings in the science of learning, and the most famous demonstration of it happens to use vocabulary.

In a 2008 study in Science, Jeffrey Karpicke and Henry Roediger had university students learn 40 Swahili-English word pairs, such as MASHUA - boat. Once a pair had been recalled correctly, it was either tested again, studied again, both, or neither. A week later the results split along one line.

A 2014 meta-analysis by Christopher Rowland, pooling 61 studies, found the same advantage for testing over restudying (g = 0.50), and two details from it map straight onto crosswords. Producing an answer from a cue helped about twice as much as recognising it among options (0.61 against 0.29), and seeing the correct answer afterwards nearly doubled the benefit again (0.73 against 0.39).

There is also direct evidence from crossword-style practice. Shane Mueller and Elizabeth Veinott (2018) ran four controlled experiments with 265 psychology students, who practised 90 course terms in a game that showed a clue and a partly filled answer, like one crossword entry midway through a solve. Practised terms were remembered better than unpractised ones, and more practice helped more. But difficulty mattered: when only three letters were shown, repeated practice still paid off a day later; when all but three were shown, it did not. With most of the letters on the page, the answer can be read rather than recalled.

What it trains: recalling a word from its meaning, which is the half of vocabulary knowledge that lets you actually use a word rather than just recognise it. A crossword built from a word list is one of the few puzzles that practises it directly.

Cryptic crosswords: seeing past the obvious reading

In a cryptic clue, the sentence as a whole means one thing and its parts mean something else entirely. A large study of cryptic solvers, by Kathryn Friedlander and Philip Fine (2016), surveyed 805 of them. The researchers describe cryptic solving as depending on reasoning and on "breaking frame": escaping the misleading surface reading a setter plants on purpose. The strongest solvers were disproportionately from mathematical and scientific backgrounds; 57.6% of the experts had studied a STEM subject.

What it trains: the evidence here describes who is good at cryptics, not what solving them does to you, so the honest claim is narrower. Cryptics exercise flexible, frame-breaking thinking about language, and they reward the kind of analytical habits that the best solvers bring with them. For how clues are put together, see our crossword guides.

Anagrams: working with letter patterns

An anagram asks you to rearrange letters until a word appears, so it works at the level of spelling rather than meaning. The Scrabble experts above were better at it, which fits: both games reward a fluent sense of which letter combinations English allows.

Anagram research also turned up a useful quirk. When part of the scrambled letters already forms a familiar word, solvers take longer, because the familiar chunk is hard to see past; their eye movements suggest they take the chunk in at once but struggle to fit the remaining letters around it (Ellis and Reingold, 2014). And an older study found that easy anagrams led to better later memory of the solution word than hard ones, even though hard ones took twice as long, perhaps because a hard anagram pulls attention onto the letters rather than the word (Foley and colleagues, 1989).

What it trains: manipulating letters and noticing spelling patterns, plus the habit of breaking up a chunk that looks right but is not. If the goal is remembering the words themselves, keep the anagrams easy. The Anagram Generator shows every arrangement of a set of letters, which makes it easy to check an answer or build a puzzle.

Scrabble: fast, spelling-first word recognition

The Calgary studies found one more thing about expert players: meaning mattered less to them. Most people recognise concrete words such as TRUCK faster than abstract ones such as TRUTH; for Scrabble experts that gap was smaller, as though they were judging words by their letters first. That makes sense for a game in which knowing that a word is valid matters more than knowing what it means.

What it trains: quick, accurate recognition of valid spellings, in any orientation, along with anagramming and letter-prompted word finding. It does not specially train meaning, which is the crossword's territory. The Scrabble Word Finder and our Scrabble word lists are built for exactly this kind of practice.

Word searches: scanning and the shape of words

A word search gives you the word and asks you to find it among a grid of distracting letters. That is a visual search task, and it keeps the exact spelling of each word in front of the solver the whole time: to find PHOTOSYNTHESIS running diagonally backwards, you have to hold its letters in mind in order. It is also low-pressure and open to almost anyone, which is a real strength.

Here we have to be straightforward about the evidence. We found no controlled study showing what word searches improve; what exists is mostly small classroom studies comparing a group only with itself before and after, which cannot separate the puzzle from everything else that happened in between. And because the word is printed in the list, a word search does not practise recalling it, the thing that makes crosswords useful for vocabulary.

What it trains: scanning, sustained attention, and familiarity with how words are spelled. For a stronger vocabulary workout, print the puzzle without its word bank and give definitions instead, so each word has to be recalled before it can be found. The Word Search Maker can switch the word bank off for exactly that.

Choosing and making the right puzzle

Put together, the research suggests a few simple choices:

• For learning what words mean, use a crossword with clues that point at the meaning: a definition, a synonym, or a sentence with a gap. Leave the word bank off, since a list of every answer turns recall back into matching.
• For spelling and familiarity, use a word search or anagrams. They keep the letters of each word in front of the solver.
• Use both on the same list. A word search to meet the words, then a crossword a few days later to recall them, covers form and meaning in the right order.
• Aim for hard but solvable. Mueller and Veinott's easy hints taught nothing lasting, while in Rowland's data retrieval that mostly failed without feedback taught nothing either. Crossing letters and a way to check or reveal an answer keep a crossword in that zone.
• Come back to it. Repeated recall paid off in both the crossword study and the Swahili experiment, and the benefit of recall grows over longer delays. A second puzzle from the same list a few days later is worth more than a longer first one.

The Crossword Maker and Word Search Maker both turn the same word list into a printable puzzle or a link to solve online. In the Crossword Maker, the word bank is off by default, suggested clues are definitions you can edit, and Regenerate builds a fresh layout from the same words for the return visit, keeping every clue you wrote. Its solving link lets you choose whether solvers can check their answers and reveal a word they are stuck on.

---