Can attention be trained, and what would that even mean?

Three lessons have taken three famous claims apart. This one asks the question all three were really about: can you get better at this?

It is also the only lesson in the course with something constructive in it, and the constructive part comes last and is smaller than you want it to be.

The word that decides the question

Transfer is what a psychologist calls the thing you actually want when you train. You practise one task, and the question is what else improves.

The literature separates three levels and the separation isn't a technicality. It is the whole argument.

Near transfer. You get better at the thing you practised, or at something almost identical. You do a number-sequence game every day and your score on that game rises.

Intermediate transfer. You get better at a different task that taxes the same system. Your score rises on a working-memory test you were not practising.

Far transfer. You get better at something out in your life that the system is supposed to support. Reasoning. Reading comprehension. Arithmetic. This is the only level anybody is buying, and it is the level the advertising implies.

What the evidence says

Melby-Lervåg, Redick and Hulme pooled the working-memory training literature in 2016.¹ This course has read its abstract verbatim; the body of the paper was not opened.

The scale, in their words: "87 publications with 145 experimental comparisons".¹

The result, in their words, and it is worth having in full: "Immediately following training there were reliable improvements on measures of intermediate transfer (verbal and visuospatial working memory). For measures of far transfer (nonverbal ability, verbal ability, word decoding, reading comprehension, arithmetic) there was no convincing evidence of any reliable improvements when working memory training was compared with a treated control condition."¹

Read "treated control condition" carefully, because it carries the argument. A treated control is a group that does something else: another activity, another training programme, anything that takes similar time and attention. Comparing training against a group that did nothing measures the expectation of improvement, the attention from the researchers, and the practice at being tested, all of which move scores on their own. A do-nothing comparison cannot separate the training from any of that, which is why the meta-analysis restricts its far-transfer verdict to studies with a treated control.

Then the mediation check, which is the finding this lesson would keep if it could keep only one: "mediation analyses indicated that across studies, the degree of improvement on working memory measures was not related to the magnitude of far-transfer effects found."¹

That is awkward for the whole idea. If training works by strengthening a system, and a stronger system is what carries improvement outwards, then people who improved most on the trained measure should show the most far transfer. They did not.

And then the sentence that is unusually blunt for a meta-analysis: "analysis of publication bias shows that there is no evidential value from the studies of working memory training using treated controls."¹

And their statement of what training does produce, which is the honest replacement for any number, is the first half of the finding quoted above: reliable improvement on measures of intermediate transfer, and no convincing evidence of improvement on far transfer against a treated control.¹ That is the course's own paraphrase of the abstract sentence, not a second quotation, because the abstract is the whole of what this course read.

Their conclusion, in their words: "These results seriously question the practical and theoretical importance of current computerized working memory programs as methods of training working memory skills."¹

What about meditation?

This is the question everybody asks next and this course has to answer it carefully, because it hasn't read the evidence.

At search-summary level, two syntheses from 2021 are worth naming: one by Yakobi and colleagues and one by Verhaeghen.² Both report small effects, with significant results on some attention measures and not others. This lesson is not going to give you their figures, because the research file says in terms that a course at this read level must not put one in a learner's hands, and a number remembered is a number quoted.

Three things follow and the third is the one to keep.

First, the syntheses themselves call those effects small.

Second, the active-control question applies here too, and this course doesn't know how those syntheses handled it. That is not a criticism of them; it is a statement about what twenty minutes of searching established.

Third, and this is the point: this course hasn't read those papers, so it isn't going to put a figure in your hands as though it had.² What it can tell you is the direction and the read level, which is: reported effects are small, the syntheses exist, and somebody who wants an answer should read them rather than take this course's word for it.

The hour, and what is holding it up

This is the only constructive section in the course, and every element below is labelled.

One: a context you meet reliably. Same place, same time, something you encounter anyway. Evidence, plus a step this course is taking from it. The evidence is Habits and Self-Discipline lesson 3: two field experiments showing that an established behaviour runs off its context, and that removing the context stops it. The step is that choosing a reliable context will help institute a new hour, which those experiments do not test, because they measured behaviours people already had. This course rates it the strongest borrowed element in the list.³

Two: remove the interruptions rather than compensating for them. Evidence, plus a step. The evidence is lesson 2: people compensate for interruptions by working faster and pay for it in stress, frustration, time pressure and effort, in 48 people in a 90-minute session. What that establishes is that a cost is hidden, not that removing interruptions improves the hour, and the step from one to the other is this course's.³

Three: an hour. The course's own judgement.³ Nothing this course read tests the length of a work block. An hour is long enough to get into something and short enough to protect, and if somebody tells you ninety minutes is optimal, ask them what measured it.

Four: the same hour each day. The course's own judgement,³ extending element one by analogy: a fixed time is easier to make a reliable context out of. Habits lesson 3's own misconception list says a clock time is a weak context on its own, so this is weaker than element one and depends on the hour having something else attached to it.

Five: one measure, written down. The course's own judgement,³ and the reason is lesson 1: a self-report and a count answer different questions, so keep one of each if you can. Whatever you got through, and an effort rating.

Six: the phone. Not evidence. Lesson 5 found the pooled effect small, in memory rather than attention, and absent in one region's samples. Do what you like with your phone and don't tell yourself the research decided it.

Notice the shape of that list. Two elements have evidence behind them and a step away from it in each case, and one of the two is borrowed from the previous course. Three are judgement outright. One is explicitly not supported. That is what an honest constructive section looks like in this subject, and a course offering you seven techniques has either read something this one hasn't or is not telling you which is which.

Worked: one claim, taken apart by transfer level

The claim is invented and the reasoning is the course's own.³ It is built to be ordinary rather than egregious.

"Ten minutes a day of our memory game. Users improve their scores by 62 percent in six weeks, and report feeling sharper at work." Three sentences, three different levels, and they are stacked so that the first supports the third.

1. "Improve their scores by 62 percent." That is near transfer, measured on the trained task, and it is the one thing the meta-analysis found large and reliable. It is also the level that tells you least, because getting better at a thing you practise daily is what practice is.
2. "Feeling sharper at work." That is far transfer, and it is what the buyer is buying. It is also a self-report, which lesson 1 says is a different instrument from the score, so the sentence contains two kinds of evidence presented as one.
3. What is missing is the middle. No intermediate transfer is claimed, which is the level where a real effect would show up first, and no comparison group is mentioned at all.

So the sort is: the first claim is measurable and probably true; the third is measurable and untested here; and the sentence's work is done by putting them next to each other. Nothing in it is a lie, which is what makes it worth working.

Three things people get wrong about this

"Focus is a muscle." The training literature's own meta-analysts say there is no evidential value in the treated-control studies of working-memory training, and that improvement on the trained task doesn't predict improvement anywhere else.

"Brain training was debunked, so nothing helps." Two different claims. What failed is that practising a component improves the system. Arranging your situation is a separate proposition and lesson 2 and Habits lesson 3 both bear on it.

"Meditation is proven to improve attention." The syntheses report small effects and this course hasn't read them, which is a statement about this course rather than about meditation.

Practice

Connections

Back. Lesson 1's instruments, lesson 2's finding, lesson 3's measurement and lessons 4 and 5's pooled evidence all feed the hour. Habits and Self-Discipline lesson 3 supplies element one and is the strongest thing in the design. How to Learn Anything taught deliberate practice, which is the nearest relative of the transfer question and is not re-taught here.

Forward. Lesson 7 is what stays useful when everything in this lesson has dated, and the sorting exercise above is its first rehearsal. And the course's project is this lesson's hour run for four weeks: its first section is the six elements above, each labelled the way you labelled them here.

Go deeper

• Working Memory Training Does Not Improve Performance on Measures of Intelligence or Other Measures of "Far Transfer" (Perspectives on Psychological Science, 2016). This course has read the abstract verbatim and not the body. The abstract alone is worth it: it states the scale, the result, the mediation check and the publication-bias verdict in about two hundred words, which is unusually direct for a paper of this size.
• The pooled evidence on phones and cognitive performance (Behavioral Sciences, 2023), open access, read in substantial part. Lesson 5's subject. Read it beside this one for the contrast in verdicts: there, an effect smaller and narrower than its name; here, a mechanism that does not transfer at all.

Sources

1. Monica Melby-Lervåg, Thomas S. Redick and Charles Hulme, "Working Memory Training Does Not Improve Performance on Measures of Intelligence or Other Measures of 'Far Transfer': Evidence From a Meta-Analytic Review", Perspectives on Psychological Science 11(4), 2016, pages 512 to 534. Abstract read verbatim; the body of the paper was not opened, and the body of this lesson says so. Supports: the quoted count of 87 publications and 145 experimental comparisons; the quoted statement of intermediate and far transfer results; the quoted mediation finding; the quoted publication-bias sentence; and the quoted conclusion. No effect size from this paper appears in the lesson, because the abstract this course read reports none, and the body was not opened.
2. Two mindfulness syntheses, at search-summary level only; neither was opened, which the body states where they appear: Yakobi and colleagues, 2021, and Verhaeghen, 2021, "Mindfulness as Attention Training". Both report small effects, with significant results on some attention measures and not others. research/SOURCES.md says in terms that a course at this read level must not put a figure in a learner's hands, so this lesson gives none and neither does this entry. No claim in this lesson rests on either beyond the direction and the read level.
3. Elements three, four and five of the hour are this course's own judgement, each labelled inline at the point of use, and nothing this course read tests the length of a work block, the value of a fixed time, or which measures to keep. Element four's analogy to element one is explicitly weaker than element one, and the lesson says so by pointing at Habits lesson 3's own misconception about clock times.