Working Memory and Reasoning Are Closely Related. They Are Not the Same Thing

Working memory is one of the few technical terms in cognitive psychology that has escaped into ordinary speech, usually in a slightly wrong form. It does not mean short-term storage, and it is not a measure of how many things you can remember. It means holding information available while operating on it — keeping the first half of a sentence live while parsing the second, tracking two constraints while checking whether a third is consistent with them.

How it is measured

The standard laboratory tools are complex span tasks. You are given something to remember, then interrupted with an unrelated judgement, then given the next thing to remember, and so on, and finally asked to recall the sequence. The interruption is the point: simple storage would be easy, and the task is built so that storage and processing compete.

Clinical batteries measure something adjacent. The Wechsler scales, for instance, carry a working memory index built from subtests such as repeating digit sequences forwards and backwards. It is not identical to a laboratory complex span task, which is one reason effect sizes differ across studies that all say "working memory".

The relationship with reasoning

Reasoning, in this literature, usually means fluid reasoning: solving novel problems that do not depend on stored knowledge — matrix items, series completion, logical inference from unfamiliar premises. Across a large number of studies, working memory measures and fluid reasoning measures correlate substantially and positively. That finding is robust and has been replicated many times.

What the correlation means is where the field divides. One camp reads it as evidence that the capacity to maintain and control information is close to the core of what reasoning ability is. Another argues the overlap is real but has been inflated by shared task demands — both kinds of test require you to sit still, follow instructions, resist distraction and work under time pressure — and that stripping those shared demands out leaves two related but genuinely distinct abilities.

The training question

This is not an academic dispute, because a whole product category rests on the answer. If working memory capacity is close to the core of reasoning, and if it can be trained, then training it should raise reasoning ability. That syllogism is the entire premise of the brain-training industry.

The evidence has been unkind to it, and specifically to the second and third steps. Training on a working memory task reliably improves performance on that task, sometimes dramatically. Improvement on closely similar tasks — what researchers call near transfer — appears sometimes. Improvement on genuinely different measures, including reasoning tests, is what is called far transfer, and it is the claim that has repeatedly failed to hold up in larger, better-controlled studies and in meta-analyses that account for the quality of the control condition.

What is safe to say

• Working memory and fluid reasoning are substantially correlated. This is well replicated and not seriously contested.
• The size of the correlation depends heavily on how each construct was measured, so single reported figures should be treated as one estimate among many.
• Whether the overlap reflects a shared underlying capacity or shared task demands is genuinely open.
• Training a working memory task improves that task. That much is not in doubt.
• Far transfer from that training to general reasoning is not established, and the better-controlled the study, the smaller the effect tends to be.

The useful takeaway is not a number. It is that "related" and "the same" are different claims, that most popular writing collapses them, and that a product built on the collapsed version is selling a step the research has not delivered.

Common questions