Expertise transfers between domains far less than the people holding it expect

The appealing idea about general skill

A comfortable belief about learning is that practising hard thinking in one area produces better thinking generally. Study a difficult subject and you become more rigorous; solve complex problems at work and you acquire a transferable problem-solving faculty. It is the premise behind a great deal of education, a fair amount of hiring, and most claims that a particular activity sharpens the mind.

The research picture is considerably less encouraging. What repeatedly turns out to be true is that people get better at what they practise and at things closely resembling it, and that the improvement falls off sharply as the new task moves away from the trained one. This is usually described as the difference between near and far transfer, and far transfer is the rare one.

Why skill is tied to the material it was built on

The classic demonstrations come from board games. Strong players can reconstruct a position from a brief glance far better than weaker players, which looks like superior memory. Show them a board with the pieces placed at random and the advantage largely disappears. What they possess is not a general memory capacity but a large store of meaningful configurations, which is useless when the configurations are meaningless.

This pattern has been found again and again in other domains. The expert advantage lives in recognising structure, and the structure is domain-specific. That is not a limitation of the people; it is what expertise consists of. A faculty of general reasoning, detachable from content, is not what the evidence suggests is being built.

The training claims and how they have held up

The most heavily tested version is training designed to improve general cognitive ability through repeated exercises. The consistent finding is improvement on the trained task, some improvement on very similar tasks, and little reliable evidence of improvement on distant ones or on everyday functioning. Sizeable reviews and large registered studies have converged on that reading, though the literature is not unanimous and proponents contest both the analyses and the choice of outcome measures.

The honest summary is that near transfer is well supported, far transfer is not, and the burden of evidence sits with anyone claiming the latter. That burden is rarely met and almost never acknowledged in the marketing of such programmes.

What does seem to travel

Some things transfer better than skills, and they are mostly abstract rules with wide application. There is evidence that explicit instruction in statistical and methodological principles changes how people reason about everyday problems outside the training context, which is a genuine instance of transfer, though the effects reported are modest rather than transformative.

What such training has in common is that the principle is taught as a principle, with varied examples, and with explicit attention to recognising when it applies. That last part appears to be the bottleneck: people often possess a relevant rule and fail to notice that the situation in front of them is one where it belongs.

The dangerous case is confidence, not competence

The practical hazard is not that an expert in one field performs poorly in another. It is that the feeling of competence transfers even where the competence does not. Somebody who is genuinely excellent at reading one kind of system arrives in a different one carrying the same certainty, and that certainty was calibrated somewhere else entirely.

This has a compounding effect, because expertise also buys standing. The confident outsider is listened to, their errors are less likely to be challenged, and the feedback that might have recalibrated them is softened by their reputation. It is a reliable way to be wrong in public with support.

Making transfer more likely

Three things help, none of them dramatic. Learn a principle from varied examples rather than repeated similar ones, since variation is what forces the abstraction. Practise identifying which principle a novel situation calls for, rather than only applying a principle you were told to use. And state the structure explicitly, because implicit pattern-learning stays tied to the surface it was learned on.

And when entering a new domain, treat the transfer as unproven until something confirms it. The parts of your judgement that came from general principles may hold. The parts that came from recognising familiar patterns will not, and from the inside those two feel exactly the same.