Evidence for Transfer: making learning stick beyond the test #

Every substantive claim on the Transfer: making learning stick beyond the test page is checked against current research. Here is each claim, how well today’s evidence supports it, and the sources. The full, de-duplicated source list lives on the references page.

Supported · strong evidence — Transfer of learning to a new situation is not automatic, and it becomes rarer the larger the gap between the learning context and the application context.

Barnett & Ceci’s taxonomy and the broader literature establish that transfer falls off as the distance between training and transfer context grows; that far transfer is difficult and not guaranteed remains the consensus view in 2026.

Sources: Barnett, S. M., & Ceci, S. J. (2002), When and where do we apply what we learn? A taxonomy for far transfer. Psychological Bulletin — https://doi.org/10.1037/0033-2909.128.4.612 · full reference ›

Supported · strong evidence — Transfer can be classified along multiple dimensions (e.g. knowledge domain, physical and social context, modality, time elapsed), and ’near’ versus ‘far’ transfer is a matter of how many of these dimensions differ between learning and application.

The multi-dimensional near/far taxonomy is exactly what Barnett & Ceci (2002) proposed and it is widely adopted as the standard framework for describing transfer; it remains uncontested as a descriptive scheme in 2026.

Sources: Barnett, S. M., & Ceci, S. J. (2002), When and where do we apply what we learn? A taxonomy for far transfer — https://doi.org/10.1037/0033-2909.128.4.612 · full reference ›

Supported · moderate evidence — Transfer tends to fail because learners bind knowledge to the surface features of the problem they trained on, so it is not retrieved when a new problem shares the deep structure but differs on the surface.

The surface-versus-structure account of transfer failure is well established from analogical-transfer research (Gick & Holyoak’s classic studies show people often fail to apply a relevant prior solution when surface features differ); it is a standard explanation in the 2026 literature, though magnitudes depend on task and cueing.

Sources: Gick, M. L., & Holyoak, K. J. (1983), Schema induction and analogical transfer. Cognitive Psychology — https://doi.org/10.1016/0010-0285(83)90002-6 · full reference ›

Supported · moderate evidence — Studying a principle through multiple varied examples (rather than a single example) promotes transfer, because varying the surface forces abstraction of the shared deep structure.

Gick & Holyoak found that giving two analogues, and prompting comparison, sharply increased schema induction and spontaneous transfer over a single example; the benefit of multiple/varied examples and explicit comparison is a robust and current finding, with effect sizes varying by guidance.

Sources: Gick, M. L., & Holyoak, K. J. (1983), Schema induction and analogical transfer — https://doi.org/10.1016/0010-0285(83)90002-6 · full reference ›

Supported · moderate evidence — Explicitly abstracting and stating the underlying principle, rather than relying on it to emerge implicitly, improves the chance that knowledge transfers to a structurally similar new problem.

Inducing or supplying an abstract schema (and prompting learners to articulate the common principle across cases) reliably raised transfer in Gick & Holyoak’s work and in later comparison-based studies; the direction of the effect is well supported, though spontaneous (uncued) far transfer remains hard.

Sources: Gick, M. L., & Holyoak, K. J. (1983), Schema induction and analogical transfer — https://doi.org/10.1016/0010-0285(83)90002-6 · full reference ›

Supported · moderate evidence — Practising under conditions that resemble the target situation reduces the transfer distance and makes the learning more likely to be available when it is needed.

Because transfer declines with contextual distance (Barnett & Ceci 2002), narrowing the gap between practice and application is a sound implication; the practical benefit of context-similar and realistic practice is supported, though the magnitude varies by domain and task.

Sources: Barnett, S. M., & Ceci, S. J. (2002), When and where do we apply what we learn? A taxonomy for far transfer — https://doi.org/10.1037/0033-2909.128.4.612 · full reference ›

Supported · strong evidence — Cognitive and working-memory ‘brain training’ programs do not produce broad, far transfer to general cognitive ability or real-world performance; gains are largely confined to the trained tasks (near transfer).

This row debunks the popular ‘brain-training makes you generally smarter’ claim. Sala & Gobet’s syntheses across working-memory, chess, music and video-game training find reliable near transfer but little to no far transfer to general cognition; meta-analyses and the 2014/2016 consensus statements concur, so the debunk is the well-supported position in 2026.

Sources: Sala, G., & Gobet, F. (2017), Does far transfer exist? Negative evidence from chess, music, and working memory training. Current Directions in Psychological Science — https://doi.org/10.1177/0963721417712760 · Simons, D. J., et al. (2016), Do ‘brain-training’ programs work? Psychological Science in the Public Interest — https://doi.org/10.1177/1529100616661983 · full reference ›

Supported · strong evidence — Training a specific cognitive task reliably improves performance on that task and very similar ones (near transfer), even though it does not generalise broadly.

That practice yields robust task-specific (near) gains while failing to generalise is the consistent pattern in Sala & Gobet’s reviews and the wider cognitive-training literature; the asymmetry between strong near transfer and weak far transfer is well established in 2026.

Sources: Sala, G., & Gobet, F. (2017), Does far transfer exist? Negative evidence from chess, music, and working memory training — https://doi.org/10.1177/0963721417712760 · full reference ›