The transfer problem
The finding that a skill learned in one task rarely carries over to a different task on its own. Schooling assumes the opposite — that learning to write essays makes someone a better writer in general, that learning to reason makes them reason better everywhere — but a century of experiments shows the carry-over is far narrower and less reliable than that hope. Transfer to closely similar tasks is fairly robust; transfer to tasks that share a deep structure but look different is rare, and improves mainly when that shared structure is taught and pointed out directly.
The empirical puzzle that skill acquired in one task does not, in general, generalise to neighbouring tasks without further deliberate work in the new task. The construct names the gap between two intuitions that schooling depends on: the intuition that learning addition makes a student better at arithmetic in general, learning to write essays makes them better at writing in general, and learning to think makes them better at thinking generally — and the experimental finding that, across a hundred years of careful work, transfer is consistently narrower and less reliable than the intuition would predict.
Etymology§
The transfer literature begins with Edward Thorndike and Robert Woodworth's 1901 Psychological Review paper The Influence of Improvement in One Mental Function upon the Efficiency of Other Functions, which set out the first careful experimental programme on the question and reported what is now called the theory of identical elements — transfer occurs to the extent that the new task shares specific elements with the trained task, and not otherwise. The construct's twentieth-century course runs through Charles Judd's 1908 generalisation theory, the Gestalt psychologists' work on productive thinking in the 1930s and 1940s, and David Perkins and Gavriel Salomon's near versus far transfer distinction in the late 1980s. The consolidating volume for the topic's purposes is Douglas Detterman and Robert Sternberg's edited Transfer on Trial (Ablex, 1993), which assembled the empirical case against expansive transfer claims into a single reference.
The puzzle's force comes from the failure of an intuition the educational system depends on. Most curricula are built around the assumption that competence transfers — that the student who learns to write a five-paragraph essay will, by the end, be better at writing, and that the writing skill thus acquired will help with the next essay, with letters, with reports, with fiction. Hundreds of carefully designed experiments through the twentieth century say otherwise. The student who has learned to solve a particular kind of physics problem does not, in general, solve a different kind without re-learning. The student who has been trained on syllogistic reasoning does not, in general, reason more logically about everyday problems. Near transfer — to tasks that closely resemble the trained task — is robust; far transfer — to structurally similar tasks in a different surface domain — is rare and unreliable.
Transfer on Trial is the consolidating reference. Detterman's opening chapter, The case for the prosecution: Transfer as an epiphenomenon, argues that transfer effects appearing in the literature are typically small, often non-replicable, and explicable as artefacts of method when they do appear. Sternberg's counter-chapters argue for a more graded view — transfer is limited but not non-existent, and the question is what conditions expand it. The volume is the standard reference cited when transfer claims appear in subsequent education-research papers.
The empirical pattern across studies is consistent. Transfer is greater when the trained task and the transfer task share surface features (the same problem context, similar wording, similar visual elements). Transfer is greater when the learner has been explicitly taught the deep structure of the trained task — what abstract problem the surface features instantiate — and when the transfer task is presented with cues to that deep structure. Transfer is greater under instruction that asks the learner to articulate the principle, generate examples, and identify the abstract pattern. Transfer is smaller under most ordinary classroom conditions.
The construct's implications for the topic's other concepts are substantial. The deliberate-practice framework predicts narrow transfer as a matter of theory: skill is built through schemas specific to the practised task, and a new task requires new deliberate work to build new schemas. The cognitive-apprenticeship framework treats the exploration phase — transferring the technique to new domains — as the place where transfer either happens or doesn't, and treats teaching for transfer as an explicit pedagogical move rather than an emergent property. The cognitive-load framework predicts transfer to depend on whether the new task's elements can be processed using existing schemas; novel elements pull working memory back to the bottleneck and the trained skill does not travel.
Skill at writing one kind of thing does not straightforwardly produce skill at writing another; skill at one research method does not carry over to the next. The deliberate-practice and cognitive-apprenticeship work that builds the first skill has to be partially redone for each transfer target, and the framework makes the difficulty visible rather than fixing it. The literature's residual disagreement is about how much help the right instruction can supply at the transfer step — meaningful, on Sternberg's reading; modest, on Detterman's.