When Learning Travels: Transfer
Transfer — applying learning outside the setting where it was acquired — is rarer than intuition expects: near transfer to similar problems is common, far transfer to distant ones is scarce, and it is earned by learning underlying structure through multiple varied examples. · 12 min
You study a skill in one place and quietly assume it will follow you everywhere it applies. Sometimes it does. Often it does not. A student who can solve a problem in the textbook stalls on the same problem dressed in new words; a fact learned for one class never shows up in another where it would help. This is the problem of transfer — carrying learning to a new situation — and a century of research says it happens far less freely than it feels like it should. This folio explains why, and what actually makes learning travel.
Guess before you learn
You solve a tricky puzzle using a clever move. Minutes later you meet a second puzzle — different cover story, but the exact same underlying move solves it. What fraction of people spontaneously use the move they just learned?
In Gick and Holyoak's classic experiments, only about a third of people spontaneously applied a solution they had just used to a structurally identical problem. If you guessed 'almost everyone', you are in good company — transfer feels automatic from the inside. It usually is not. The rest of this folio is about the conditions that raise that fraction.
9–12
3–5
When you learn something in one place, it does not always come with you to another. Math you practiced with apples should work with coins too — same idea, new objects. That is called transfer.
Transfer is easy when the new problem looks like the old one. It is hard when the new problem looks different, even if the hidden rule is the same. Seeing many examples helps you find the rule.
6–8
Transfer is applying something you learned in one setting to a new one. It comes in degrees. Near transfer reaches a problem much like the one you practiced — usually easy. Far transfer reaches a distant situation that shares only the underlying structure — and it is surprisingly rare. Most studying produces knowledge that stays close to where it was learned.
The reason is that we tend to bind a fact to its surface details — the cover story of the example, not its deep structure. To make learning travel, you have to learn the structure itself, and the reliable way to do that is to meet the same structure in several different examples.
9–12
Psychologists distinguish near transfer (to similar tasks and contexts) from far transfer (to distant ones). Edward Thorndike's early experiments (Thorndike and Woodworth, 1901) demolished the old 'formal discipline' idea that studying Latin or geometry trains the mind generally: practice on one task improved another only to the extent they shared identical elements. Transfer is not a mental muscle; it rides on shared structure.
Gick and Holyoak (1983) showed how fragile spontaneous far transfer is: given a story whose solution fit a later puzzle, only about a third of solvers used it on their own — yet showing two analogous stories and asking students to compare them roughly doubled spontaneous transfer, because comparison surfaces the shared schema that a single example leaves buried in its cover story.
K–2
You learned to tie your shoes. Now you can tie a ribbon on a present too. The knot is the same, even though the ribbon is new. Your learning traveled to a new place.
But learning does not always travel. Tying shoes will not help you ride a bike. When two things share a hidden pattern, learning can jump across. When they do not, it stays put.
Undergrad
Barnett and Ceci's (2002) taxonomy replaces a single near/far axis with several dimensions — knowledge domain, physical and temporal context, functional and social context, modality — so that 'far' names a distance along many axes at once. This dissolves fruitless debates: transfer is always transfer of something across some distance, and empirical claims must specify both.
The design lever is variability of practice. A single context invites students to encode incidental surface features as if essential; varied examples that hold structure constant while changing the surface let students abstract the invariant. This trades short-term performance — varied practice feels harder and slower — for the flexible, transferable representation a single fixed context cannot build: a desirable difficulty (folio 11).
Postgrad
The classic pessimism — Detterman's claim that significant far transfer is rare to nonexistent — is best read as a boundary condition, not a dead end. Transfer failures are typically failures of retrieval and representation: the target situation does not reinstate the cues that would access the relevant knowledge, and that knowledge was encoded too context-bound to be recognized as relevant in the first place.
Interventions that work attack exactly those failures. Contrasting cases and analogical comparison (Gentner, Loewenstein, and Thompson) build schema-level representations; 'preparation for future learning' assessments (Bransford and Schwartz, 1999) reveal transfer that static tests miss. The practical upshot for a studying reader is unromantic: expect little transfer for free, and buy what you can with multiple varied examples and explicit comparison.
transfer
Applying learning in a setting other than the one where it was acquired. Near transfer reaches similar problems; far transfer reaches distant ones and is much rarer.
So transfer is not free — but it can be bought. The purchase is always the same: pull the structure out from under the surface, and the way to do that is to see the structure wear more than one costume. One example teaches the costume; several examples that differ on the surface but agree underneath teach the thing itself. Here is the move, step by step.
Build a practice set that earns transfer — the steps fade as you master them
Structure: many small contributions combine on one target
a budget, a recipe, a dosage — all 'combine parts to a total'
"Each splits a fixed total among parts."
new story -> does the same rule apply?
Hold on to the honest version: expect little transfer for free, and buy the rest deliberately. One example teaches its cover story; several varied examples, compared, teach the structure that can travel. Next folio leaves the desk entirely for the pillow — because some of the filing that makes learning durable happens only while you sleep.
Note
Each gate you clear here returns in the Fading Ink — review what's fading — on the schedule folio 8 explains, so the structure has more than one chance to settle.
Practice — new ink and old, interleaved
1.From folio 9: why does interleaving related problem types beat blocking them by type?
2.In one sentence, explain why comparing two examples helps you transfer more than studying one twice.
3.You are memorizing the water cycle. Which addition uses dual coding?
4.From folio 7: the best review gap is roughly 10–20% of how long you need to remember. For a test 30 days away, the middle of that range is a gap of about how many days?
5.A meta-analysis reports retrieval practice at g ≈ 0.61. What does that number mean?
6.In the 2020 randomized classroom trial, blocked classes averaged 38% on the month-later test. What did the interleaved classes average?
7.From folio 13, without looking back: what is dual coding, and why does a matching picture beat a decorative one?
Dual coding is pairing words with a matching visual so one idea is stored two ways, giving two retrieval routes. A matching picture carries the idea's structure and adds a real second trace; a decorative picture carries no meaning and can split attention.
How close were you? Grade yourself honestly — it sets your review date.
8.From folio 5: on the week-delayed test, the repeated-recall group kept about what percentage of the passage, against about 40% for the rereaders?
9.A tutoring program claims that playing memory games will make children better at school in general. On the evidence about transfer, how skeptical should you be?