Executive Function: What It Is, How It Is Measured, and Why the Measures Disagree

What executive function is

Nearly everything you do that is not automatic requires executive function. Resisting the biscuit. Holding a phone number in mind while someone talks to you. Abandoning a plan that has stopped working. The common thread is that a goal you are holding internally has to override a response the environment is supplying for free.

Adele Diamond's synthesis, the most-cited framing in the field, identifies three core executive functions from which the higher-order ones are built.

From these, Diamond argues, the higher-order capacities follow: reasoning, problem-solving and planning. They are not separate ingredients so much as what the core three do when combined and sustained.

Unity and diversity

The structural question — is executive function one thing or several? — was addressed definitively by Miyake and colleagues in 2000, using latent variable analysis to strip out the noise from individual tasks.

Their answer was both. The three components are clearly separable: they load on distinct latent factors and contribute differently to complex tasks. But the factors are moderately correlated with one another, which means something is shared across all of them. Neither the “executive function is a single resource” view nor the “executive function is a bag of unrelated skills” view survives.

The task impurity problem

This is the methodological difficulty that shapes everything else, and it is rarely explained outside the literature.

An executive function, by definition, operates on some other cognitive process. There is no such thing as inhibiting in the abstract — you inhibit reading a word, or reaching for a button. Any task that measures inhibition therefore also measures reading, or motor control, or visual processing. The executive variance is entangled with non-executive variance in every single task.

This is why Miyake's group used latent variable analysis rather than single tasks. If you administer several inhibition tasks that share the executive demand but differ in their non-executive demands, the shared variance across them is a cleaner estimate of inhibition than any one task provides. The individual task scores, taken alone, are often unreliable.

The measures disagree with each other

There are two established ways to assess executive function. Performance-based tasks put a person in front of a computer and time their responses. Rating scales ask the person, or a parent or teacher, how well they manage in daily life — do they lose things, do they interrupt, do they finish what they start.

Both are used routinely, often in the same clinical assessment, and they are commonly assumed to be two routes to the same construct. Toplak, West and Stanovich tested that assumption directly.

Three-quarters of the comparisons were not statistically significant, and the authors note that even the median of .19 is probably an overestimate, because some studies declined to report their non-significant correlations.

That reframing is the useful part. A child can have an intact ability to inhibit a response under laboratory conditions and still be, in the ordinary sense, disorganised — because succeeding at goals in the world requires more than the capacity to inhibit when instructed to. Efficiency is not the same as deployment.

How executive function relates to intelligence

Closely, through one component in particular.

Working memory — the updating component — correlates strongly with fluid intelligence, at a median of roughly r = .72 across large reanalyses. Shifting and inhibition correlate with fluid reasoning much more weakly. The executive-intelligence relationship is therefore not evenly distributed across the construct; it runs mostly through updating.

This matters when interpreting a claim like “executive function predicts academic success”. Much of that prediction is carried by working memory capacity, which is itself close to fluid intelligence. Whether executive function adds anything beyond intelligence, once measurement error is handled properly, remains a live question.

Can executive function be trained?

The narrow answer is yes. The interesting answer is that the training does not go anywhere.

People who practise n-back get better at n-back. People who practise task-switching get faster at task-switching. What has repeatedly failed to appear is far transfer — improvement on untrained measures of reasoning or intelligence. Melby-Lervåg, Redick and Hulme's meta-analytic review of working-memory training found no convincing evidence of transfer to intelligence or other far-transfer measures.

Diamond's own reading of the intervention literature is instructive here: the programmes that show the broadest benefits are not computerised drills but activities that demand sustained executive effort while also being interesting, socially embedded and progressively harder. Whether that is training executive function or training something else is exactly the kind of question the task impurity problem makes hard to answer.

Myths and facts about executive function