How Is Working Memory Related to Reasoning?

Working memory and fluid reasoning are so strongly related that
psychologists have spent decades trying to work out whether one explains
the other. The intuitive story is compelling. A difficult reasoning problem asks
you to keep several things in mind at once. If some people can hold more
information there, perhaps that larger mental workspace is what allows
them to reason better. It is a beautiful explanation. It may also be too simple.

The capacity
hypothesis meets an awkward test

Consider a matrix problem that requires several rules to be
coordinated. A straightforward capacity account predicts that as the
number of relevant relations rises, people with greater working-memory
capacity should gain an increasing advantage. The correlation between
working memory and reasoning should become stronger as the presumed
memory demand of the reasoning item increases. Researchers have tested versions of that prediction for years. A 2025 systematic review by Đokić, Koso-Drljević and Bilalić found
mixed evidence across the literature. Their own analyses, using a
substantially larger sample and several ways of representing reasoning
difficulty, did not support the idea that the close
working-memory–fluid-intelligence relationship is simply a function of
increasing storage demands. That does not weaken the association itself. It changes what the
association can be taken to mean.

Perhaps the important
word is not “memory”

Working memory is not merely a temporary box. While solving a problem, information has to remain accessible, but
goals also have to be protected from distraction. No-longer-useful
representations must stop controlling behavior. Information may need to
be retrieved, updated or rebound to the right context. Imagine discovering that a candidate rule works for the first two
examples but fails for the third. You now have to preserve the
contradiction, abandon a tempting hypothesis and continue searching
without losing the original task. That looks less like filling a storage container and more like
controlling an unfolding cognitive episode. This is why theories emphasizing executive attention and interference
management have been influential in explaining the relationship between
working memory and reasoning.

A very high
correlation is still not an identity

At the latent level, working-memory capacity and fluid reasoning can
correlate strongly enough to raise an obvious question: are
psychologists measuring the same underlying thing twice? Not necessarily. The answer depends partly on which tasks are used and how the models
are specified. Working-memory tasks can require temporary maintenance
and control without demanding novel relational inference. Reasoning
tasks require discovering or applying relations even when their
immediate storage demands are modest. The constructs overlap substantially without becoming universally
interchangeable. This distinction matters because a correlation—even an impressive
one—is a pattern to be explained, not a mechanism.

Training provides
another reality check

If working-memory capacity were a straightforward bottleneck causing
general reasoning ability, training that expands the bottleneck should
improve reasoning broadly. Working-memory training reliably improves performance on practiced
tasks and sometimes on closely related ones. Evidence for durable far
transfer to intelligence and substantially different reasoning measures
is much weaker, particularly in studies with strong control
conditions. That result does not prove that working memory plays no causal role
in reasoning. Training may fail to change the relevant underlying
processes. But it makes the simplest causal story harder to sustain. The puzzle therefore survives. Reasoning clearly needs a way to keep goals and intermediate
information usable while a problem unfolds. Working memory is deeply
involved in providing that infrastructure. What has become less
convincing is the idea that the relationship can be explained by a
single quantity—the number of things that fit in mind at once. The mind may need room to reason. But it also needs control over what
stays in that room, what leaves, and what gets brought back when the
problem changes.