Does working memory capacity correlate with fluid intelligence and science learning achievement?

What it means to say working memory and fluid intelligence are correlated

Working memory is the mental workspace where you hold information briefly while using it—remembering the first half of a sentence while reading the second half, or keeping track of numbers while doing a calculation. Fluid intelligence is the ability to reason and solve novel problems without relying on previously learned knowledge. A review of working memory research found that correlational studies consistently support a close relationship between working memory and measures of fluid intelligence [1].

That correlation also shows up in laboratory tasks. In one study, 103 young adults completed visual working memory tasks using either simple colored circles or images of real-world objects, along with fluid and crystallized intelligence tests. All the working memory tasks correlated reliably with fluid intelligence scores, and the strength of the correlation did not differ between simple and meaningful stimuli [5]. Working memory performance was not reliably correlated with crystallized intelligence—the kind of intelligence based on accumulated knowledge—in any task [5]. That distinction matters: the link appears specific to reasoning ability rather than to knowledge in general.

Why remembering real objects may say something extra about reasoning

The same study found something more specific: fluid intelligence scores correlated with the difference in working memory performance between real-world objects and colored circles [5]. In other words, people who gained more from meaningful stimuli—remembering objects better than simple features—tended to have higher fluid intelligence. The authors suggest that maintaining real-world objects and colored circles largely rely on shared cognitive processes that may partly underlie individual differences in fluid intelligence [5]. This is a proposed interpretation, not a proven mechanism, but it hints that the working memory–intelligence link is not just about raw storage space.

Does the same link extend to science learning?

The review that examined working memory, fluid intelligence, and science learning reported that correlational studies supported a close relationship between working memory and measures of science achievement as well [1]. That fits the intuition that science learning places heavy demands on holding and manipulating information—keeping a hypothesis in mind while evaluating evidence, for example. But the same review found no experimental studies on the impact of working memory training on science achievement [1]. Without experiments, we cannot tell whether working memory is a cause of science achievement, a consequence of it, or simply a marker of some other underlying ability.

A separate study on computerized cognitive training set out to examine effects on working memory, fluid intelligence, and science achievement, but its available abstract is too fragmentary to report what it found [2]. It is therefore not evidence that training works; it is at most a sign that the question has been asked.

When higher ability does not mean a bigger benefit

Two recent studies looked at how working memory capacity and fluid intelligence moderate the pretesting effect—the finding that guessing answers before learning improves memory. Among undergraduate students, lower fluid intelligence was associated with a larger pretesting effect, stemming from lower performance on read items [3][4]. In a second study with adult online participants, patterns were less consistent, but lower working memory capacity was associated with larger pretesting effects, again due to lower performance on read items [3][4].

These results do not contradict the positive correlation between working memory and fluid intelligence. They show that a higher-capacity learner does not always benefit more from a particular learning technique; pretesting appeared to help lower-ability learners catch up somewhat, which the authors describe as homologizing memory ability across individuals [3][4]. The pattern varied across learner populations, so it should be treated as a suggestion rather than a fixed rule.

Sources used in this answer

Related research questions