Cognitive Psychology

Working-Memory Load Laboratory

Hold a few things in mind. Now do something else for five seconds. Now report what you were holding. How much you lose depends on how many items you were holding — and, just as much, on whether the something else was made of the same stuff.

Illustrative browser timing — a demonstration of a demand, not a measurement of your capacity

Learning objective

By the end you should be able to keep storage load, attentional control and similarity apart when reading a load function.

About 30 minutes. Nothing you do here is saved or sent anywhere.

Before you start — what is being measured

This operationalises a demand; it does not measure your capacity. "Working-memory capacity" as used in the literature is estimated from long, carefully standardised procedures and reported with error. What happens here is that a demand is varied and performance moves. Those are different claims, and only the second one is on offer.

The concurrent judgements are paced. Each one waits 1.8 seconds and then moves on, so the retention interval is the same length in every condition. Nothing else on the page is timed, nothing flashes, and the worked example reaches every conclusion without running a single trial.

  1. Answer the prediction below — it unlocks the laboratory.
  2. Choose what to hold in mind, and what to do while you hold it.
  3. Run the practice trials, then a run of twelve — four at each load.
  4. Read the load function, then change the concurrent task and run again.
  5. Compare the runs: same-domain against different-domain.
  6. Prefer not to perform it? Load the worked example: four documented simulated runs with fixed seeds.

First, a prediction

You will hold either a set of letters or a set of positions in a grid. While holding it you will make three quick judgements — either about how many syllables a word has, or about whether a pattern is symmetrical. Then one item is shown and you say whether it was in the set.

Which combination do you expect to damage memory most?

The laboratory

A trial has three parts: hold the set, judge three items, then answer one question about the set. Only the middle part is paced.

Key terms
Storage load
How many items you are holding in mind. This activity uses two, four and six.
Concurrent task
What you do while you are holding the items, before you report them.
Same-domain and different-domain
Whether the concurrent task uses the same kind of material as the thing you are holding — verbal against verbal — or a different kind, such as spatial against verbal.
Load function
Accuracy plotted against the number of items held. It is the result each run produces.

Hold it, do something else, report it

Twelve trials — four each at two, four and six items. Results are withheld until the run ends.

Phase Answer the question above to unlock this.

Set up the run

Run "nothing" first as a baseline, then a matching task, then a mismatched one.

Run

Four simulated runs, seeded. Not anybody's data.

Answer the question above to unlock the laboratory.

Challenge — three things that are not the same

Storage load is how much is being held. Attentional control is what the concurrent task demands of the general-purpose machinery — keeping the task set, resisting the wrong response, coordinating two things at once. Task similarity is whether the two draw on the same kind of representation. Every result below is routinely described with the wrong one of the three.

Five findings, each to be assigned to the factor that best explains it
Finding Which factor is doing the work?

What this demonstrates

Load has a shape

Memory for a small set is close to perfect and stays that way for a while; then it falls, and it falls faster the more the concurrent task takes. What that decline is not is a cliff at a fixed number of items. Talking about a capacity of a particular number of items is a summary of a curve, not a description of a container.

Not all interference is equal

A concurrent task of the same objective difficulty costs far more when it is made of the same kind of material as the thing being held. Judging syllables while holding letters is worse than judging syllables while holding positions, and the reverse holds for symmetry judgements. That selective pattern is the original evidence for separable stores rather than one undifferentiated pool. It is also why "how hard is the second task?" is the wrong first question to ask.

Three explanations that get run together

A drop in performance under load can come from having more to hold, from the concurrent task demanding more control, or from the two competing for the same representational format. These make different predictions and are separated by different manipulations. This laboratory separates the first and the third cleanly. It does not manipulate attentional control at all, which is why the challenge includes findings this page cannot speak to.

A demand, not a capacity

The number that comes out of a run here is your accuracy on twelve trials of a particular task on a particular afternoon. Estimates of working-memory capacity in the literature come from long standardised procedures, are reported with confidence intervals, are checked for reliability, and still predict individual outcomes only modestly. The distance between those two things is worth keeping in view.

Download activity HTML

For teaching elsewhere: take this activity as one self-contained block of HTML, on the clipboard or as a file. Either way it is styled so that it will not disturb the page you put it into.