Transfer of learning is the carrying of knowledge, skill, or strategy acquired in one situation into a new and different one. It is the implicit promise behind almost all teaching: that a multiplication drill, a grammar rule, or a problem-solving habit learned today will pay off on tasks the learner has never seen. When it helps, the transfer is positive; when an old habit interferes with the new task, it is negative. The hard and humbling finding of a century of research is that useful transfer, especially across distant situations, is far rarer and harder to engineer than teachers hope.
The distance between the learning context and the application context defines the two poles. Near transfer moves a skill into a situation that strongly resembles the one in which it was learned, as when a learner who has practised one division algorithm applies it to similar problems on a test. Far transfer reaches a situation that looks quite different on the surface, as when training in chess or music is expected to sharpen general reasoning. Near transfer happens readily; far transfer is the prize, and the problem.
The first systematic account came from Edward Thorndike and Robert Woodworth, whose 1901 experiments tested whether training one mental task improved a related but distinct one. Their identical-elements theory held that transfer occurs only to the degree that the two tasks share actual components, the same stimuli and responses. The claim demolished the older "mental discipline" idea that studying Latin or geometry strengthened the mind in general, like a muscle. On this view there is no all-purpose reasoning faculty to exercise; what transfers is specific, and only where genuine overlap exists.
Gavriel Salomon and David Perkins (1988, 1989) added a mechanism. Low-road transfer is the near-automatic application of a well-practised skill when a new situation closely resembles the old, requiring no conscious thought, like a driver moving from car to truck. High-road transfer depends instead on mindful abstraction: the learner deliberately pulls a principle out of its original setting and searches for where it might apply, which is how a strategy can reach a genuinely dissimilar problem. High-road transfer is effortful and must usually be taught, by prompting learners to articulate the principle and look actively for its reach.
Across analogy research, intelligence, expertise, industrial training, and education, far transfer has stubbornly resisted demonstration. The brain-training and cognitive-training industry has tested the strongest version of the claim, that exercising working memory or playing chess yields general cognitive gains, and the evidence is discouraging: Sala and Gobet (2017) reviewed chess, music, and working-memory training and found that gains stay near the trained skill and do not spread to broad ability. The honest summary is that learners transfer what overlaps and little else, that "learning to learn" and brain-training promises routinely overreach, and that transfer is best cultivated by teaching for it directly, varying contexts, surfacing underlying principles, and practising the act of application, rather than assuming it will happen by itself.
The construct itself is contested. Some researchers argue that "transfer" as classically defined, the clean application of an inert skill to a new task, may be the wrong unit of analysis, since real performance is so bound to its context that asking whether a decontextualised skill "transfers" can be misframed. Even where transfer clearly occurs it is hard to measure cleanly, because near and far sit on a continuum with no sharp line, and a study's verdict can hinge on how similar the experimenter judged two tasks to be. The field's pessimism is robust but not total: positive results exist where instruction explicitly bridges contexts, so the lesson is less that far transfer is impossible than that it almost never comes free.
In language teaching the word transfer names a different and narrower idea, easy to confuse with this one. Language Transfer and the broader notion of crosslinguistic influence concern how a learner's first language shapes their second, as when a Vietnamese speaker's article system surfaces in their English. That is influence between two language systems in one mind, not the application of a learned skill to a new domain. The two share a label and almost nothing else; keeping them apart avoids real confusion when both come up in a methodology discussion.
If transfer rarely happens on its own, it has to be designed in. Practising a skill across varied surface contexts loosens it from the situation where it was first learned. Making the underlying principle explicit, and asking learners to name where else it might apply, supports the deliberate high-road path. Worked examples paired with later varied problems, and prompts that ask learners to compare a new task with familiar ones, all push toward generalisation. The realistic goal is reliable near transfer plus carefully scaffolded reaches toward farther application, not a vague hope that broad mental powers have been strengthened.