What is the difference between near and far transfer?

Near transfer is using what you learned on a task that closely resembles the one you trained on, such as applying a division method to slightly larger numbers. Far transfer is using it on something that shares underlying structure but almost no surface features, such as recognising exponential growth in a rumour after studying compound interest. Near transfer is common. Far transfer is rare and needs deliberate support.

Does brain training transfer to everyday thinking?

Mostly not, according to large reviews and pre-registered trials. People reliably improve at the trained games and sometimes at very similar tasks, but improvements in general attention, reasoning or daily functioning have been small or absent. That pattern matches the wider transfer literature going back to Thorndike, so it is not a surprise.

How can I make my own learning transfer further?

Meet each principle in several different disguises rather than one, mix problem types instead of practising them in blocks, state the underlying principle in your own words, and test yourself with varied questions instead of rereading. All of these stop the knowledge from being tied to a single surface, which is what blocks transfer in the first place.

Why would story-based learning transfer better than drills?

Because a story stores causes, motives and constraints rather than one screen layout, and those structures recur across situations. A speculative bubble, a chain of small compliances, a city choosing where to build a wall: each is a pattern you can recognise elsewhere. MindSnap, our app, is built around that idea, though the principle holds regardless of what you use to learn.

Learning Science

Transfer of Learning Explained: Why Skills Do Not Always Travel

Ingrid Larsen

Learning Science Writer

Sep 9, 2026

Last updated: August 2026

8 min read

TL;DR

Transfer is the use of something learned in one setting in a different one. Near transfer, to closely similar tasks, happens readily. Far transfer, to genuinely different domains, is rare and has disappointed researchers since Thorndike tested the idea over a century ago. It is also the promise brain-training marketing rests on. What does improve transfer is varied practice, multiple examples across different surfaces, and learning the underlying principle rather than the wrapping, which is why knowledge learned as a story with causes tends to travel better than a drill.

The whole point of learning something is to use it somewhere other than where you learned it. Nobody studies percentages in order to be good at that worksheet. So the question of whether learning travels, and how far, is not an academic curiosity. It is the question that decides whether your study time was an investment or a hobby.

The research answer is narrower than most of us would like, and the disappointment is old.

Near and far

Transfer is usually described along a distance axis.

Near transfer means applying learning to a task that closely resembles the one you trained on. Learning to drive one hatchback and then driving a different hatchback. Practising long division with three digit numbers and then handling four. Near transfer happens easily and reliably, which is why practice works at all.

Far transfer means applying learning to a task that shares deep structure but almost no surface features. Learning about compound interest and recognising the same exponential shape in a spreading rumour. Learning chess and becoming a better strategist at work. Far transfer is what everyone actually wants, and it is where the evidence gets thin.

A century of disappointment

The formal story starts around the turn of the twentieth century, when schools taught Latin and geometry partly on the theory of mental discipline: hard abstract subjects were thought to strengthen the mind in general, the way lifting strengthens the arm regardless of what you later lift.

Edward Thorndike, with Robert Woodworth, tested pieces of that assumption and found the strengthening did not generalise the way the theory required. Training people to judge one kind of magnitude improved that judgement and left closely related ones largely untouched. Later work in the same tradition kept finding the same shape of result: improvement clung tightly to what was actually practised. Thorndike's identical elements account said transfer happens to the extent that two tasks share components, which is a deflating but durable summary.

The finding has been rediscovered many times since. Expertise studies in the 1970s and 1980s found that chess masters had extraordinary memory for realistic board positions and unremarkable memory for randomly scattered pieces. Their skill was not a general memory upgrade, it was deep familiarity with meaningful patterns in one domain.

The modern rerun of this argument is commercial. Cognitive training apps promise that practising working memory games will improve attention, reasoning and everyday functioning. Large reviews and pre-registered trials have generally found that people get better at the trained games, sometimes better at very similar tasks, and rarely better at the everyday outcomes that were the whole selling point. We went through that literature in brain training versus real learning, and the shape is exactly Thorndike's: improvement stays close to home.

Why far transfer is so hard

The core obstacle is that people encode surfaces, not structures. Give someone a problem about army divisions converging on a fortress and then a problem about radiation beams converging on a tumour, and most will not spot that the second is the first in different clothing. The classic analogical reasoning experiments found exactly that: without a hint, the majority miss the connection, and with a hint most can use it. The knowledge was there. The retrieval cue was not, because the cue they had stored was 'fortress'.

This is also why context matters more than we assume. Learning bound to one room, one format, one type of example gets tagged with that context and does not come to mind elsewhere, a point that overlaps with context-dependent memory.

What actually improves transfer

The good news is that transfer is trainable to a degree, and the interventions are all about how you practise rather than how much.

1Vary the surface. Meet the same principle in several different guises: a physics version, a money version, a history version. Multiple examples make the shared structure available for extraction, while one example teaches the example.
2Mix problem types rather than blocking them. Practising several kinds together forces you to identify what kind of thing you are facing before solving it, which is the part real life demands and blocked practice quietly skips. That mechanism is the subject of our piece on interleaving.
3Name the principle out loud. Comparing two worked examples and stating what they share produces better transfer than studying either alone. If you cannot say the principle in a sentence, you have not extracted it.
4Retrieve rather than review. Pulling knowledge out under slightly different questions each time creates more routes back to it, which is the practical value of retrieval practice.
5Accept some friction. Conditions that feel harder in the moment and yield better later flexibility are the whole point of desirable difficulty.

Notice the through-line. Every one of these makes the learning less tightly bound to a single situation. Transfer is not a separate skill you bolt on. It is what happens when learning was never stored in only one shape.

Why stories travel better than drills

This gives a defensible reason to prefer knowledge learned as narrative over knowledge learned as isolated items, and it is not a marketing claim, it is a claim about cues.

A drill gives you one surface: this format, this instruction, this button. A story gives you people, motives, constraints, a decision and a consequence. That is a rich structure with many hooks, and it maps onto other situations because human motives and constraints recur. Learn how a speculative bubble inflated among Dutch tulip traders and you have a template with a shape: a novel asset, a story about scarcity, credit that lets buyers commit money they do not have, and a settlement date nobody thought hard about. That shape shows up again in every century, which is why our collection on Tulip Fever does more for your judgement than a hundred memorised dates would.

Similarly, learning how ordinary volunteers were talked step by step into doing something they found repellent, the substance of The Obedience Files, gives you a structure you will recognise in an office, a committee or a family. This is transfer through causal structure rather than through surface similarity, and it is the kind that actually shows up in conversation and decisions.

MindSnap, which is our app, is built on that bet: two minute story-driven lessons with five flagship collections plus unlimited topics, and interactive Turning Points scenarios that ask you to make the decision yourself before revealing what really happened. Choosing before you know is a small act of retrieval and prediction, and prediction errors are one of the better ways to make a principle stick.

What to expect, honestly

Do not expect any single subject to make you globally smarter. A century of evidence says that is not how it works, and anyone selling it is either overreaching or not reading their own field. Expect instead that broad, well-connected knowledge makes new learning faster, because the hooks are already there, and that principles met in several guises will be recognised again. That is a smaller promise than the brain-training pitch and it has the advantage of being true.