Individual Variation in Working Memory Capacity (WMC): a First Step Down the Research Rabbit Hole

There are many, many studies that measure variation in WMC vs variation in other metrics.

Want to get notified about new posts? Join the mailing list and follow on X/Twitter.

Variation in WMC is often quantified by measuring one’s top performance on WM-loading tasks.

The most well known of these tasks is “digit span,” i.e., I read off a list of N digits and you repeat them back to me in the correct order. People typically max out around N=7, but some people can go to a higher N while others max out at a lower N. There’s a pretty good Wikipedia article on this.

Digit span was one of the earliest studies on working memory and since then, a number of other tasks have been created. Barbara Dosher’s “Working Memory” article in the 2003 Encyclopedia of Cognitive Science elaborates on plenty of them.

I should note that it is sometimes possible to train up one’s performance on a specific WM-loading task without the performance actually representing an increase in WMC (you can tell because the increased performance doesn’t actually carry over to other WM-loading tasks). So if you’re measuring WMC you typically want to base your measurement on a variety of loading tasks instead of relying on just one.

There are many, many studies that measure variation in WMC vs variation in other metrics. You can find a bunch of these on Google Scholar if you just type in “working memory capacity vs ___” where you fill in the blank with whatever you’re curious about. For example: “working memory capacity vs academic achievement”

Many of these studies are super interesting. For instance, did you know that WMC is a better predictor than IQ when predicting a young student’s future academic success? Check out Alloway & Alloway, 2010.

Turns out that WMC impacts a lot of things: perceived effort, abstraction/generalization ability, learning speed… I’ve written plenty about this here if you’re interested.

Point of Nuance: Fluid / Crystallized Intelligence vs WMC / LTM

If you dive into the literature, you might also see discussion of fluid & crystallized intelligence (abbreviated Gf and Gc, respectively). It’s a common point of confusion how these are different from working memory capacity (WMC) and long-term memory (LTM).

Basically: these pairs are related, but WMC and LTM are at lower levels of scale.

There’s less baked into the definitions of WMC and LTM, and more baked into the definitions of Gf and Gc.

For instance, something like “generalization ability” is baked into the definition of Gf, whereas it’s not baked into the definition of WMC.

To really understand this, it’s necessary to think all the way back to how these things are measured. Gf is measured by something like performance on Raven’s Progressive Matrices which directly tests generalization ability. WM is measured by something like digit span which does not directly test generalization ability.

Want to get notified about new posts? Join the mailing list and follow on X/Twitter.