Working Memory, Explained: What Puzzles Actually Exercise

“Memory” is a vague word that covers several genuinely different mental systems, and puzzle games mostly exercise just one of them: working memory. Understanding what that term actually refers to makes it much easier to see what a game like Memory Match or Sudoku is doing to your brain in the moment — and, per Do Brain Games Actually Improve Memory?, what it probably isn’t doing beyond that moment.

Working memory vs. long-term memory

Long-term memory is what most people mean by “memory” in everyday conversation — the store of facts, experiences, and skills you can retrieve minutes, days, or years later. Working memory is different: it’s the small amount of information you can actively hold and manipulate right now, while you’re using it. Classic estimates put working memory capacity at around four to seven discrete items for most people, though the exact number depends heavily on how “item” is defined and what kind of information it is.

A useful analogy is a desk versus a filing cabinet. Long-term memory is the filing cabinet — large capacity, but you have to go retrieve something before you can use it. Working memory is the desk — whatever’s currently spread out in front of you, immediately available, but with very limited surface area. When you’re solving a Sudoku cell, the candidate digits you’re currently considering are on the desk; the general rules of Sudoku you learned once are in the filing cabinet.

Why working memory feels effortful

Because working memory has such a small capacity, it’s easy to overload. Try to hold six unrelated numbers in mind while also reasoning about a seventh piece of information, and something usually falls off the desk — a phenomenon researchers call interference. This is why Memory Match gets harder as difficulty increases: each additional pair of cards adds another position-symbol association competing for the same limited desk space, and past a certain number of pairs, most people can’t hold the whole board in working memory at once and have to rely on partial, chunked recall instead (see Memory Match Strategy for how chunking helps manage this directly).

Working memory has sub-types

Researchers generally distinguish at least two components: a visuospatial sketchpad, which holds spatial and visual information (where things are, what they look like), and a phonological loop, which holds verbal or sound-based information (words, numbers you’re rehearsing by “saying” them internally). These systems draw on somewhat separate cognitive resources, which is why you can often hold a phone number in mind while also visually tracking where you left your keys — you’re using two different working-memory subsystems rather than one shared pool.

Different puzzles lean on different subsystems. Memory Match is heavily visuospatial — you’re remembering where symbols are, not naming them aloud in sequence. Sudoku draws more on a combination of visuospatial tracking (which cells you’ve checked) and a more abstract, rule-based reasoning process layered on top. Word Search is visuospatial in a different way — pattern-matching a shape against a grid rather than remembering discrete positions.

What “exercising” working memory actually means moment to moment

When a game pushes you right up against your working-memory limit, you’re forced to develop strategies to cope with the overload — chunking information into fewer, larger units; externalizing part of the load (writing pencil-mark notes in Sudoku instead of holding candidates in your head); or systematically structuring your search so you don’t need to remember as much at once (scanning a word-search grid in a fixed row-by-row order, discussed in Word Search Tips). These coping strategies are themselves a skill, and they do transfer somewhat across similar tasks — this is part of why near transfer (see the companion article on brain-training research) is more reliably observed than far transfer.

Working memory training programs, specifically

Some research programs go further than casual puzzle games, using repetitive tasks (like the n-back test, where you track whether a current stimulus matches one from several steps earlier) specifically designed to push working-memory capacity under controlled conditions. These programs have shown some of the more credible transfer effects in the brain-training literature — modest gains on related working-memory measures, though typically fading over weeks without continued practice. Casual puzzle games aren’t designed with this rigor and shouldn’t be expected to match those specific, narrow results, even though they engage some of the same underlying processes.

The central executive: managing the desk itself

Beyond the visuospatial sketchpad and phonological loop, most models of working memory include a third component often called the central executive — the process that decides what gets attention, switches between subtasks, and coordinates the other two systems rather than storing content itself. In Sudoku, the central executive is what decides whether to keep scanning for naked singles or switch to checking for pointing pairs; in Minesweeper, it’s what decides whether to keep applying certain deductions or shift into probability estimation for an ambiguous region.

This coordinating function is arguably the most effortful part of working memory to use well, because it requires monitoring your own reasoning process, not just holding content. Puzzles that require switching between different types of reasoning mid-solve — as harder Sudoku and Minesweeper boards often do — exercise this executive function more than puzzles with a single, repetitive reasoning pattern throughout.

Why difficulty scaling matters for working memory specifically

Because working memory has a roughly fixed capacity per person, a puzzle that’s too easy doesn’t meaningfully load it at all — you can solve it using long-term memory and pattern recognition without ever approaching your working-memory limit. This is worth knowing if your specific goal is to engage working memory rather than just enjoy a puzzle: an easy Memory Match board with only a few pairs may be relaxing, but it’s not pushing the same cognitive system as a hard board with many more pairs to track simultaneously. If working memory engagement specifically is what you’re after, difficulty selection matters more than game selection.

The practical upshot

Puzzles that push you close to your working-memory limit — a hard Memory Match board, a Sudoku puzzle requiring you to track several candidate digits at once — are genuinely exercising a real, well-defined cognitive system in the moment you’re playing. That’s a legitimate and satisfying thing to do with your attention, independent of whether it produces durable, general cognitive benefits afterward. If you’re building this into a regular habit rather than an occasional session, Building a Daily Brain-Training Habit covers how to make that stick.