Skip to content

Transfer of Learning

Principle · All principles
Evidence · 30 claims (19 for, 11 mixed) · 32 studies (12 causal, 8 review, 5 theoretical, 4 quant-synthesis, 2 qualitative, 1 associational), q1–q4 · 6 of 32 report an effect size · 26 claims rest on one study

Conditional relationship

This page owns the most general relationship in the transfer family: whether what a learner has learned in one setting is used in a new task, situation or problem. Consider a learner who can do a task, or state a principle, in the form it was taught. Their starting response on a new case might be to treat it as unfamiliar, answer from its surface features, apply the taught method where it does not fit, or succeed only when told "this is like last week". The expectation is that instruction which shows the same underlying structure in several surface forms, asks the learner to compare those forms and name what they share, and then has them retrieve and use the structure on changed items produces more near-transfer to new cases of the same kind than practising one version of the task in one context. The target is a principle, concept, procedure or complex-skill whose structure can appear in more than one surface form. Evidence for far-transfer to a different domain, and for gains at delayed-retention, is thin, and the strongest results are near and short-term.

No claim in the wiki tests this general relationship as stated, that is, instruction across varied contexts against instruction in one context, with a transfer test. The idea that varied encoding produces decontextualised knowledge reaches the wiki only second-hand, through a review, and the whole-task argument is a training-design proposal that no recorded study tests. What the wiki does test are four configurations inside the relationship, each with its own comparison: comparing two provided cases against studying them separately, mixing problem types against blocking them, retrieval practice against restudy on new questions, and faded worked examples with principle prompts. All four measure transfer on study-defined tasks close to the training domain. Two experiments show far transfer staying at the floor in every condition. Whether the learner's original learning is secure, whether the varied cases really share the structure, and what counts as "new" for the objective must be established locally.

How the converted siblings sit inside this. Analogical Reasoning owns mapping one provided or familiar case onto a new one, and holds the case-comparison evidence in detail. Cognitive Flexibility owns holding the same knowledge in several cases, perspectives or representations so it can be reassembled for an ill-structured situation. Retrieval Practice owns retrieval as such. Guided Practice and Worked Examples own how the original learning is secured. This page does not repeat their models. It says what a designer adds when transfer itself is the objective, and how to tell transfer from recall. The 4C/ID pattern and the case-based learning pattern hold converted policies for whole-task and case-based designs. The Hugging and Bridging for Transfer strategy holds two teacher techniques taken from the earlier version of this page.

Default design, while the relationship is untested

The earlier page's guidance (Schunk; Souza; Perkins and Salomon), kept as a concrete default a designer can act on and revise. Evidence status is stated beside each step. Every quantity is an untested proposal unless a claim is named.

  1. Secure the original learning before asking for transfer. Souza's first factor is "context and degree of original learning". Before the first transfer item, the learner should do the taught task unaided and explain one step. Untested here as a precondition. How to get there is modelled in Guided Practice and Worked Examples.
  2. Teach each principle through at least two, preferably three, cases that differ in surface and share the structure. Examples from the earlier page: a proportional-reasoning strategy in recipes, map scales and budgets, or one clinical reasoning routine across different patients. Evidence: comparing analogous cases beats studying the same cases separately [+S]. A second-hand review reports that examples from varied contexts produce decontextualised knowledge [+W]. Two position papers argue that varied tasks of one type [+W] and learning in multiple contexts [+W] lead to transfer. A review reports that repeating one task improved that task and not a new task of the same type [~M]. Two cases rather than one is supported. The number of cases is not settled.
  3. Put the cases side by side and ask what corresponds and what is the same, then state the principle. Evidence: in the case-comparison meta-analysis the benefit was larger when learners looked for similarities and when the principle came after the comparison (claim [+S]). In worked-example learning, prompts to identify the principle, with backward fading, improved near and far transfer [+M] on study-defined probability tasks.
  4. Name the critical attributes and where the structure does not apply. Include at least one case that looks similar but has a different structure, and ask the learner to say why the principle does not fit. This is from the earlier page (Souza's "critical attributes"; negative transfer) and is an untested proposal. Mixing confusable problem types once each has been introduced is the nearest tested form: interleaving improves classification of new examples [~M], strongly for visual categories, less for mathematics, and reversed for word categories.
  5. Practise retrieval on changed items, not the same ones. Ask application and inference questions in a changed format rather than repeating the taught items. Evidence: retrieval practice beats restudy on transfer questions [~S], most clearly for application and inference questions and changed formats. Bias-corrected estimates show little transfer without those features.
  6. Prompt bridging in both directions. While learning, ask "where else could you use this?" (forward-reaching). On a new problem, ask "what earlier problem is this like?" (backward-reaching). Fade the prompt once the learner names a correct source without it. From the earlier page (Perkins and Salomon's high-road transfer; Hugging and Bridging). Untested proposal.
  7. Keep the whole task visible when the target is a complex skill. Practise integrated tasks that keep the structure of later performance, not only isolated parts. Evidence: only a training-design argument that no recorded study tests [+W]. For learners new to a high-interactivity task, practise isolated parts first and integrate them soon after: part-task practice reduced load for absolute novices [~M].
  8. Judge transfer on its own items. Score near-transfer items (same structure, new surface) and far-transfer items (different domain or form) separately from recall and from the taught items. Give at least one at a delay of a week or more. Expect far-transfer scores to be low: all conditions sat at a floor on far transfer in a computer-based engineering study [~M], and no condition raised far-transfer scores in a game-learning experiment [~W]. Do not read a far-transfer null as a failure of the near-transfer design.
  9. Spread the arc across sessions, with the source of cases moving outward. The following arc is an untested proposal for a unit of about four to six sessions. Early (sessions 1–2): about 20 minutes per session on the taught case and one near variant, with the comparison prompt given and the principle stated by the teacher afterwards. Middle (sessions 3–4): about 15 minutes per session on two or three new surface forms, problem types mixed, and retrieval questions in a changed format. The learner states the principle, and the teacher checks it. Late (sessions 5–6): about 10–15 minutes per session on unprompted new cases, including one the learner brings from their own work or life, and one far case. Open each session after the first with two retrieval items on earlier cases in a new surface form. Move a learner on from one stage to the next when they solve two of three unprompted near cases in a new surface form and can name the shared structure. Move a learner who already does this at the start straight to the late stage (expertise reversal [~M]). Repeat a near and a far item at least a week after the last session.

When a condition is not met: if only one context can be taught (one workplace procedure, one textbook case), vary the problems within it and add a bridging prompt about where else it applies. Test transfer to the context that matters and do not claim a general gain. If learners cannot yet do the taught task, transfer items will only measure that.

Fitting the design to a situation

The three facts that most change the decision:

  • How far the target situation is from the taught one. A new surface in the same domain (near-transfer) is supported here. A different domain or form (far-transfer) is barely moved in the evidence the wiki holds. If the brief does not say, ask: what new situations must the learner handle, and how different are they from the ones in the course?
  • Whether the original learning is secure. A learner who cannot do the taught case unaided has nothing to transfer, and extra varied cases add load. If the brief does not say, ask: can the learner do one taught item now, unaided, and explain why it works?
  • Whether the target setting supports the use. In workplaces, transfer depends on cases, equipment and colleagues in the setting as well as on the course. If the brief does not say, ask: will the learner meet the situations, and have the resources, to use this within weeks of the course?
If the brief says… Then change… Basis
Learners are novice on this task, or the task is high in element interactivity Secure the taught case first with worked examples and guided practice; start with two cases, not four; practise parts briefly before integrating; delay mixing until each type has been introduced Part-task practice reduces cognitive load for absolute novices during initial skill acquisition. [~M] (load, not transfer); interleaving [~M] (claim page notes novel material can reverse the benefit)
Learners are advanced, or already solve the taught case unaided Skip the early stage; give unprompted varied and far cases with selective feedback, and ask learners to generate their own cases Instructional guidance that helps novices can become redundant or counterproductive as expertise grows. [~M] (review of cognitive-load studies; not about transfer designs)
Learners are child (primary age) Use concrete, familiar surface forms; give the comparison prompt every time and state the principle yourself; keep to two cases per session untested; the case-comparison and retrieval-transfer evidence here is mostly with undergraduates
Learners are adult or professionals Take the late-stage cases from their own work; ask them to bring one; check that their workplace offers the cases and equipment to use the skill The closer training content is to authentic clinical scenarios, the smoother its translation into practice [+W] (interviews with neurosurgical nurses; perceived, not measured)
Goal is a principle or concept to be used in new cases Pair cases and prompt comparison before stating the principle; test on new cases rather than definitions Analogical Reasoning Improves Transfer [+S] (undergraduates, negotiation; meta-analysis across domains)
Goal is choosing among several procedures (which method fits which problem) Mix problem types in practice once each is introduced, so the learner must decide which applies; expect slower, error-prone practice Interleaving category examples improves inductive category learning for visual and mathematical materials, but not for expository texts or word categories [~M] (classification of new examples; smaller for mathematics, reversed for word categories)
Goal is spoken or performed fluency on a type of task (e.g. a language task) Vary the task within the type rather than repeating one task; repeat a task only to build fluency on that task Task rehearsal improves fluency and complexity on the repeated task but does not transfer to a new task of the same type [~M] (second-hand report of one language-learning study)
Setting is online-self-paced, no teacher present Build the comparison prompt and the principle statement into the materials; use automatically checked retrieval questions in changed formats; add an unprompted transfer item before the next unit Retrieval Practice Improves Transfer [~S] (lab and prose-passage studies; carried to self-paced courses untested)
Time is one session only Teach two cases side by side, prompt the comparison, state the principle, and end with one near-transfer item; do not claim far or delayed transfer Analogical Reasoning Improves Transfer [+S] (immediate tests favoured in the meta-analysis)
classroom with one teacher and a large group Pairs compare the two cases on one sheet and write what is the same; the teacher collects answers and states the principle from them; every learner does one unaided transfer item on paper untested; the comparison evidence includes classroom experiments in the meta-analysis but no class-size result
Stakes are high, or the objective is far transfer to another domain Add cases from the target domain itself rather than relying on transfer from another; assess on representative target tasks after a delay All learners … showed a floor effect on far-transfer problems [~M]; no increases for far-transfer problems [~W] (one study each, engineering and a game task)
workplace-clinical, and the course teaches advanced competencies the setting rarely uses Agree with the workplace when the learner will meet the cases; plan refresher cases or simulation if they will not; report use separately from knowledge Transfer of advanced competencies is limited by institutional case exposure, technical platforms, and multidisciplinary maturity rather than training deficiencies [~W] (interviews, one programme in China)

Ten of the twelve rows cite a claim. Five of them (concept, procedure choice, self-paced, one session, far transfer) rest on comparison studies with a transfer outcome, and even these carry undergraduate laboratory or classroom results to other learners and settings. The other five carry a review, interview studies or a load result to this setting, and they say so. The child and classroom rows are untested proposals.

Observation, state and explanation

What can be observed is a response to a new case under stated conditions: the case, how it differs from the taught ones (surface, structure, domain), whether a hint or reminder was given, the answer, and any explanation. "Has abstracted the principle", "knowledge is context-bound" and "negative transfer" are inferred states. Record whether the learner was told that an earlier case was relevant. Success with that hint shows that the learner can apply the principle, not that they can recognise when it applies.

Response Plausible explanations Observation that could change the interpretation
Solves the taught items, fails a new case with a different surface Learning bound to the taught surface; original learning shallower than it looked; the new case also differs in structure Give a new case with the taught surface and new numbers, then one with a new surface and the same structure; ask what the case is "like".
Solves the new case only when told "this is like the earlier one" Principle held but not retrieved spontaneously; the hint gave away the method Give matched new cases with and without the hint, and one that looks like the earlier case but needs a different principle.
Applies the taught method to a case it does not fit Surface similarity triggered it (negative transfer); critical attributes not learned; only one method available Ask what would have to be true for the method to work here; give a pair of look-alike cases with different structures to sort.
States the principle correctly, cannot use it on a new case Verbal knowledge without a procedure; the new case needs other knowledge; the item is a far case Give a near case in the taught domain, then the far one; check whether the far case's own domain knowledge is held.
Transfers at the end of the unit, not a week later Little consolidation; the transfer item was close to practice items; interference from later work Re-test with a fresh case at the delay and record all practice in between.
Uses the skill in the course, not on the job No suitable cases at work; missing equipment or support; the skill was adapted locally and looks different Ask the learner and a supervisor which cases occurred and what was done; observe one case if possible.

These rows are proposals, untested with learners as diagnostics. A transfer probe is itself a varied case and teaches as it measures, so record each one. Keep the explanations that remain open.

Evidence and qualifications

No claim compares instruction across varied contexts with instruction in one context on a transfer outcome at first hand. The claims below test configurations inside the relationship or limit it. Learners, comparisons and outcomes differ, so do not rank them by effect label.

Taken together: comparing provided cases, mixing confusable types, retrieval on application questions, and principle prompts during example study each have experimental or meta-analytic support for near transfer at short horizons, mostly with undergraduates. Varying contexts as such is supported only second-hand and by argument. Far transfer is not moved in the two studies that measured it separately.

The evidence by question

Further evidence, not yet read against this model

Claims bearing on transfer that the wiki holds and this model does not use. None was cited by the earlier page; each label says what it is.

Objective and learner-valued goal

The designer's objective is usually unaided performance on a new case of a stated kind (near or far), with stated aids, at a stated delay. Elicit separately what the learner wants to be able to do and where: pass the course test, handle a situation at work next month, or use the idea in a project of their own. Agreement and divergence are both observations. A learner who values the test may resist varied cases that slow them down on the taught items. A learner with a specific target situation is the best source of the late-stage cases.

For example, a nurse on a specialty course may want to manage one kind of patient on their own ward, while the designer's objective is reasoning across a family of neurosurgical cases. Use the learner's ward cases as the late transfer items, and record use on the ward separately from scores on course cases, so that success in the course is not read as the capability the learner came for.

What would revise this model?

The expectation should weaken if first-hand comparisons of varied-context with single-context instruction, with comparable learners and transfer items scored separately from taught items, show no near-transfer advantage, or one that disappears at a week. If comparison and principle prompts help only when the cases are given side by side, and not when learners must retrieve an earlier case, the bridging step should be dropped or redesigned. If studies that measure far transfer go on finding floors in every condition, the page should stop presenting far transfer as an objective that instructional variation alone can reach, and point to teaching in the target domain. If whole-task and part-task practice are compared on transfer and part-task does as well for novices, step 7 changes. Do not protect the model by calling a failed transfer test "too far" after the fact. Say how far the item was before it is given.

Recall of the taught items, near transfer, far transfer, delayed transfer and use in the learner's own setting are separate claims. The present evidence establishes no number of cases, no session dose and no criterion for moving on.

Claims

Examples

Key Sources

  • Perkins, D. N., & Salomon, G. (1992). Transfer of learning. International Encyclopedia of Education.
  • Schunk, D. H. (2004). Learning theories: An educational perspective (4th ed.). Pearson.
  • Souza, D. A. (2016). How the brain learns (5th ed.). Corwin Press.
  • Helfenstein, S. (2005). Transfer: Review, reconstruction, and resolution. University of Jyväskylä.
  • Arduini-Van Hoose, N. (2020). Teacher's perspectives on learning. In Educational psychology. Retrieved from https://edpsych.pressbooks.sunycreate.cloud. CC BY-NC-SA 4.0.
  • Nguyen, T., Duncan, R.J., & Bailey, D.H. (2019). Theoretical and methodological implications of associations between executive function and mathematics in early childhood. Contemporary Educational Psychology. https://doi.org/10.1016/j.cedpsych.2019.04.002
  • Merrill, M. D. (2002). First principles of instruction. Educational Technology Research and Development, 50(3), 43-59. https://doi.org/10.1007/BF02505024
Page metadata
Field Value
type principle
id transfer-of-learning
aliases ef-interventions-target-general-factors-for-math-transfer
title Transfer of Learning
description Showing the same structure in several surface forms, prompting learners to compare them and name what they share, and then having them retrieve and apply it on changed items is expected to raise near transfer to new cases once the original learning is secure; no claim tests varied against single-context instruction directly, and far transfer stayed at floor where it was measured.
status review
moved_from pattern/merrills-first-principles-integration
canonical True
generated by claude/unspecified on 2026-10-05
trust unverified — no one has confirmed this page's content
sources 2 — nguyen-2019, merrill-2002

Frontmatter is the page's machine-readable record — see the Schema & Guide.