Transfer of learning: the hardest problem in education, stated honestly

Evidence grade: STRONG on the problem, PROMISING on the helps. That transfer is harder and rarer than intuition expects is one of educational psychology's most consistent (and oldest) findings. The interventions that improve it — varied examples, explicit abstraction, bridging — have decent support; anything promising cheap far transfer deserves your suspicion by default.

Everything taught in school is taught somewhere it won't be used: the classroom. Transfer — knowledge showing up where it's needed rather than where it was learned — is therefore not a nice-to-have; it's the entire wager. Which makes the field's foundational finding uncomfortable: transfer is hard, and the further the distance, the rarer it gets.

The century of sobering results

The modern study of transfer began with a debunking: Thorndike and Woodworth (1901) tested the then-universal "formal discipline" doctrine (Latin and geometry as mind-training that strengthens general faculties) and found improvement stubbornly specific — training on one task transferred roughly in proportion to shared identical elements with the target. The pattern has recurred for a century: chess masters' prodigious memory evaporates for random-piece boards (Chase & Simon — the expertise is chess-shaped, not memory-shaped); problem-solving strategies mastered in one cover story fail to surface under another (Gick & Holyoak's radiation-problem studies: spontaneous analogical transfer around 20% — until a hint to use the prior story triples it, which is the finding's hopeful half); and the "brain-training" industry re-ran formal discipline with software and meta-analyzed to the same verdict (Simons et al., 2016: practiced tasks improve; general cognition doesn't). Computational thinking's transfer claims inherit exactly this history, which is why we grade them cautiously.

The useful vocabulary (Barnett & Ceci's taxonomy): near transfer — target resembles training in surface and structure — is common and reliable; far transfer — new domain, new surface, shared deep structure only — is rare, effortful, and never free.

What actually helps

The Gick & Holyoak hint result names the core problem: knowledge exists but isn't recognized as relevant — filed under its surface features (the story about the general) rather than its structure (the convergence principle). The helps all attack that filing problem:

Varied examples and contexts — the same principle met in multiple costumes forces the abstraction out of any one costume (why we teach loops on a golem and a turtle, conditionals in charms and spreadsheets).
Explicit abstraction and comparison — having learners state the shared principle across two cases dramatically beats either case alone (the analogical-encoding literature: comparing is the abstraction event).
Explicit bridging — naming the destination ("you'll use this exact move in your email, your documents, everywhere") isn't motivational garnish; it's pre-filing the knowledge under its future retrieval cues.
Strong initial learning — nothing transfers that wasn't learned well; mastery is transfer's precondition, not its rival.

What the evidence doesn't say

It doesn't make teaching pointless — near transfer is dependable, and most of what schools need (skills used on real documents, real spreadsheets, real browsers) is near transfer, buildable by training on the genre's shared elements.
It doesn't rule out far transfer — it prices it: achieved deliberately, via abstraction and bridging, never as a side effect.
It doesn't justify surface-mimicry-only training either — identical elements include deep structure; teaching the concept ("reply-all goes to everyone") transfers where teaching the button position doesn't.

In the classroom

Teach for the identical elements that matter: concepts and conventions over vendor pixels.
Two costumes minimum for anything abstract — and make the comparison explicit: "where have you seen this before?"
Bridge out loud, constantly — one sentence connecting today's skill to its real destinations, every lesson.
Distrust cheap far transfer in product claims — "coding teaches general problem-solving," "chess raises IQ": ask for the controlled studies; the base rate says they'll disappoint.

How Wiz Kids applies this

Transfer strategy is threaded through the design: simulations train genre conventions (the identical elements real apps share), core constructs appear twice in different clothes with the echo made explicit, the fiction bridges outward relentlessly ("this is how you'll find any file you ever lose"), and the capstone projects force skills out of their teaching contexts into one integrated job — near transfer, engineered on purpose, with far-transfer claims deliberately absent from our marketing (claims discipline applies to ourselves first).

References

Thorndike, E. L., & Woodworth, R. S. (1901). The influence of improvement in one mental function upon the efficiency of other functions. Psychological Review, 8.
Gick, M. L., & Holyoak, K. J. (1980; 1983). Analogical problem solving / Schema induction and analogical transfer. Cognitive Psychology, 12; 15.
Chase, W. G., & Simon, H. A. (1973). Perception in chess. Cognitive Psychology, 4.
Barnett, S. M., & Ceci, S. J. (2002). When and where do we apply what we learn? A taxonomy for far transfer. Psychological Bulletin, 128(4).
Simons, D. J., et al. (2016). Do "brain-training" programs work? Psychological Science in the Public Interest, 17(3).

© Glu IO Pty. Ltd. — Wiz Kids (wiz.kids). Link freely; republication requires permission — see terms. Found an error in our reading of the research? We correct fast: tell any teacher piloting Wiz Kids.