You Can't Cross-Train Your Brain

In 1901, Edward Thorndike and Robert Woodworth sat students down and had them estimate the area of rectangles. Over many sessions, the students got very good at it. Then Thorndike and Woodworth asked them to estimate the area of triangles and circles.

Nothing transferred. Practice on rectangles made them better at rectangles. That's it.

This was a direct challenge to what educators called "formal discipline," the 19th-century belief that studying rigorous subjects like Latin and geometry trained general mental faculties. The idea was that hard subjects exercised the mind the way physical exercise strengthens a muscle, and that a mind strengthened by Latin would think more clearly about everything.

Thorndike and Woodworth said: let's actually test that. They did. It wasn't true.

That was 125 years ago. The formal discipline doctrine is still alive, dressed up in new clothes, embedded in arguments made every day about chess programs, music lessons, coding education, and brain training apps. The research hasn't changed. The belief has.

What Transfer Actually Means

Transfer of learning is when skills or knowledge from one domain improve performance in a different domain.

Near transfer is when the domains are closely related. Learning to add fractions helps with adding mixed numbers. That happens reliably. Nobody debates it.

Far transfer is the interesting one. The ambitious one. The one most people actually care about. Chess sharpening your math reasoning. Music lessons improving verbal memory. Coding teaching you to think logically about any problem. A general mental toughness built through difficult experience that makes you better at everything.

Far transfer is what self-improvement is mostly built on. The implicit promise behind "challenge yourself" culture, behind learning hard things for the sake of the discipline rather than the skill.

Barnett and Ceci spent their 2002 paper in Psychological Bulletin trying to build a useful framework for when transfer occurs. They identified nine dimensions along which transfer varies: how similar the domains are, how much time has passed, how different the physical context is, and so on. Their core finding was that transfer becomes less likely the more different the original learning and the new task are. The farther you get from the original context, the more the transfer evaporates.

That's the problem. The transfers people care about are the far ones.

Chess

There are over a thousand schools running chess programs. Most of them cite cognitive benefits. Better math performance. Stronger reading comprehension. Higher IQ scores. The programs are popular, well-funded, and built on a story that sounds reasonable: chess is a game of pure thought, and practicing it trains the thinking skills underneath.

Giovanni Sala and Fernand Gobet meta-analyzed 40 studies on chess instruction in 2016. They were published in Educational Research Review. What they found: no significant effect of chess training on academic outcomes. Math scores didn't improve. Reading scores didn't improve. General cognitive ability didn't move.

The studies that showed positive results were almost uniformly ones without active control groups. Students who played chess did better than students who did nothing. But students who played anything did better than students who did nothing. When Sala and Gobet looked only at studies with active controls, where chess was compared to another engaging activity, the advantage disappeared.

The chess-trained students got better at chess. That's the whole story.

Music

The music argument runs similar logic. Music involves pattern recognition, abstract reasoning, and sustained practice. Of course it trains your brain. Countless parents cite it when explaining why their kids are in piano lessons.

Sala and Gobet ran a second meta-analysis in 2017, this time on music training and cognitive or academic skills. Published in Educational Research Review. 38 studies, 118 effect sizes, 3,085 participants.

Overall effect size: d = 0.16. Small. And when they filtered to studies with active control groups and random assignment, the effect dropped to near zero.

The same story. Music training improved music performance. It did not reliably improve intelligence, memory, reading, or math in ways that a well-designed study could detect.

(The researchers weren't anti-music. They were clear that music has intrinsic value and that kids should learn it. But the cognitive transfer argument doesn't hold up, and they said so.)

Brain Training Apps

Lumosity, Cogmed, BrainHQ, and dozens of others have been marketed explicitly as cognitive enhancement tools. Do these puzzles daily and you'll think faster, remember more, focus better. The logic sounds solid: cognition is a skill, skills improve with practice, therefore practice this cognitive task to improve cognition.

Daniel Simons and a team of six researchers published a 132-page review in Psychological Science in the Public Interest in 2016. They read every serious study on commercial brain training. Their conclusion: these programs improve performance on the trained tasks. That's it. There was no compelling evidence they improved general cognitive function beyond the specific tasks being practiced.

Lumosity eventually settled with the FTC for $2 million for deceptive advertising. The science was what it was.

Why People Believe It Anyway

Thorndike identified the mechanism 125 years ago and it's still the right answer. Skills transfer to the degree that the new task shares specific elements with the trained task. This is called the identical elements theory, and it's held up well.

Learning to touch-type improves your typing speed on all keyboards because the elements are identical. Learning Spanish speeds up learning Portuguese because the elements overlap substantially. Learning chess does not improve algebra because there are essentially no identical elements.

But humans are pattern-seeking in a way that overestimates connection. Chess feels like math because both involve formal rules and calculation. Music feels like memory training because it requires memorizing. Coding feels like general logical thinking because it involves if-then structures.

These similarities are real but surface-level. The underlying cognitive operations are domain-specific, not domain-general.

There's also confirmation bias at scale. Schools run chess programs and track outcomes. When a student improves in math the same year they're in chess club, chess gets the credit. The thousands of students in chess programs who didn't improve in math don't generate case studies.

The Near Transfer That Actually Works

None of this means transfer is impossible. Near transfer is consistent and reliable.

If you want to get better at writing, write more. The transfer to related writing tasks is real and large. Writing essays makes you better at writing emails and reports and arguments. The elements overlap enough that the skill moves.

If you want better working memory for math, practice math. Retrieval practice with math problems improves performance on math tests in ways that retrieval practice with unrelated material does not.

Sport science researchers have documented near transfer extensively. A basketball player practicing ball-handling drills does transfer improvement to game ball-handling. The farther the drills get from game conditions, the less transfer occurs. Schmidt and Bjork's work on contextual interference (which I covered in article four of this series) shows that variable practice within a domain transfers better within that domain. But the boundaries of "within that domain" matter.

The principle is clear: train the thing. Near transfer from closely related practice is real. Far transfer from structurally unrelated tasks is largely fantasy.

What This Breaks

A lot of self-improvement content runs on far transfer logic.

Cold showers build discipline that transfers to hard work. (Maybe. Maybe what actually transfers is the story you tell yourself about being a cold shower person.)

Fasting builds willpower that transfers to resisting other temptations. (The research on willpower depletion is actually contested. Ego depletion effects haven't replicated well.)

Learning a language trains general cognitive flexibility. (The bilingualism advantage claim has had serious replication problems since about 2015.)

None of this means cold showers, fasting, or language learning are bad. They might have real benefits. But the far transfer argument, that hard thing X trains some general cognitive muscle that makes you better at hard thing Y, is almost always unsupported.

If you want to be more disciplined about work, practice being disciplined about work. If you want better focus for deep reading, practice deep reading. The brain doesn't have a general discipline or focus reservoir that fills up from any difficult activity and depletes for any challenging task. It has specific circuits that strengthen with specific practice.

What Expertise Literature Actually Says

The chunking research in article seven made this point from another angle. Chess grandmasters don't have better general memory. They have better chess memory, built from tens of thousands of chess-specific patterns accumulated over years.

When Chase and Simon ran the experiment in 1973, they were explicit: the memory advantage disappeared completely with random board positions. The expertise was specific. The chunks were chess chunks. They didn't make grandmasters better at remembering phone numbers or solving physics problems or navigating cities.

Ericsson spent his career emphasizing the same thing. Deliberate practice builds mental representations in specific domains. Those representations don't transfer. A surgeon with 10,000 hours of deliberate practice in abdominal surgery has elite mental representations of abdominal anatomy, tissue response, and instrument control. That expertise doesn't make them a better orthopedic surgeon. It doesn't make them a better diagnostician. The specificity is the point.

The Uncomfortable Math

This creates a real problem for how most people approach skill-building.

We want leverage. We want to learn things that compound across other things. We want the efficient portfolio of skills that unlocks a hundred others. The formal discipline doctrine survives because people want it to be true, not because the evidence supports it.

The evidence says: want to be a better writer? Write. Want to be a better programmer? Program. Want to understand statistics? Do statistics problems.

There's no cognitive cross-training hack. Playing Go won't make you a better investor. Learning an instrument won't make you a better engineer. Doing daily logic puzzles won't sharpen your emotional intelligence.

What you practice, you get better at. What you don't practice, you don't. The portability problem is that skills mostly aren't portable.

That's not depressing. It's clarifying. Stop looking for the meta-skill that trains everything else. Put in the hours on the actual thing you want to be good at. Thorndike figured this out in 1901. It still takes most people by surprise.

Sources

• Thorndike & Woodworth, "The Influence of Improvement in One Mental Function upon the Efficiency of Other Functions" (1901, Psychological Review)
• Barnett & Ceci, "When and Where Do We Apply What We Learn? A Taxonomy for Far Transfer" (2002, Psychological Bulletin)
• Sala & Gobet, "Do the Benefits of Chess Instruction Transfer to Academic and Cognitive Skills? A Meta-Analysis" (2016, Educational Research Review)
• Sala & Gobet, "When the Music's Over: Does Music Skill Transfer to Children's and Young Adolescents' Cognitive and Academic Skills? A Meta-Analysis" (2017, Educational Research Review)
• Simons et al., "Do 'Brain-Training' Programs Work?" (2016, Psychological Science in the Public Interest)
• Chase & Simon, "Perception in Chess" (1973, Cognitive Psychology)
• Classics in the History of Psychology: Thorndike & Woodworth (1901)

Part of the Practice Paradox series. Read the previous article: Productive Failure and the Generation Effect. Continue with the final article: Error, Prediction, and the Learning Brain.