Skip to content

Holistic Learning

Principle · All principles
Evidence · 18 claims (9 for, 9 mixed) · 24 studies (7 causal, 7 review, 6 theoretical, 4 quant-synthesis), q1–q4 · 4 of 24 report an effect size · 13 claims rest on one study

Conditional relationship

For a learner who does not yet hold an organisation of their own for a body of interconnected content (a novice for that content's structure, not necessarily for the wider field), giving them a usable whole early (a short map of the few most general ideas and how they relate, a working model, or one complete simple version of the target task) and keeping each later detail attached to that whole is expected to improve performance on tasks that draw on the whole (principle and concept goals: explaining, deriving, solving problems; near-transfer), compared with learning the same details as isolated pieces. The relationship is conditional on four things: how much the ideas interact against what the learner already knows (a whole that is too complex to follow may need its parts first), whether the tasks the learner will face match the ideas placed at the top of the whole, the learner's prior attainment, and whether the learner works on the whole (arranges, places, redraws, applies) rather than only being shown it.

No claim in this wiki tests "whole first" against "parts first" as an order, or tests holistic design as a whole. The nearest test is one report of small physics experiments: a hierarchical organisation of the same knowledge, general ideas above detail, beat a single-level one on complex recall and problem-solving tasks, helped only the tasks that matched its top level, did not detectably help tasks resting on single details, and did not detectably help low-ability students. The claim that practising whole tasks improves transfer is a training-design argument with no study recorded behind it. One experiment points the other way for absolute novices on highly interacting material. So the default design below is a labelled proposal built from the earlier page's guidance.

How the converted neighbours sit beside this: Elaboration Theory is the reusable sequence for content (an epitome elaborated level by level and tied back by synthesis) and carries the detailed evidence on hierarchical organisation; 4C/ID is the reusable sequence for whole tasks in a complex-skill; Clear Structure owns signals, organizers and outlines that make a structure visible; Chunking owns how large one unit is; Cognitive Load Theory owns isolating interacting elements before the whole for task-specific novices. This page owns the general stance that joins them: decide what the whole is, give it early in a form the learner can use, and keep the parts attached to it. Use the pattern pages for the step-by-step sequence; do not treat this page as a third sequence.

Default design, while the relationship is untested

The earlier page's guidance, made concrete. Each step is a design proposal; its evidence status is beside it.

  1. Decide what the whole is before planning parts. For organised content (a theory, a system, a body of concepts) the whole is a structure: the few most general ideas and how they relate. For a skill (writing, programming, a clinical or design workflow) the whole is a complete task: the full workflow or product, in its simplest real version. Choose the ideas at the top from the tasks learners will later face, since the organisation helped the tasks that matched it (organisation–task match [~M]). If two kinds of task matter, prepare two views rather than one whole that favours one. Choosing content versus task is an untested proposal.
  2. Ask before you show (5 minutes). Before any instruction, ask learners to sketch or list how they think the topic fits together and to place two or three specific details in it. A learner who does this correctly skips the overview and starts where their explanation fails (expertise reversal [~M], a review of cognitive-load studies; gives no crossover point). The elicitation itself is an untested proposal.
  3. Give a usable whole, not a slogan (10–15 minutes in the first session). Present three to five top-level ideas as a labelled diagram or model, or one complete simple version of the task worked through end to end, with the purpose said in one sentence. "Here is the big picture" without a structure learners can place things in does not count. Basis: a hierarchical organisation beat a single-level one on complex tasks in one experiment (hierarchical organisation [+M], 36 college physics volunteers). The three-to-five figure is carried from memory research on how many chunks novel material can hold (chunking [+S], not a tested overview size), so treat it as a starting size.
  4. If learners cannot follow the whole, pre-teach its parts briefly, then return. When the top-level ideas only make sense together and learners know none of them, teach the one or two elements the whole depends on in isolation (one short segment), then present the whole again in the same session. Basis: novices on high-interactivity tasks did better with isolated parts first (part-task practice for novices [~M], one experiment), and naming a system's parts before a complex lesson helped later transfer (pretraining [+M], 93 participants, one VR lesson). The "then return" is an untested proposal.
  5. Every later session: open on the whole, teach one part, close by placing it (about 5 minutes at each end). Show the map or the whole task for two or three minutes and say which part today elaborates; teach the part; end with learners placing the new detail on the map, arranging items into it, or saying how the part changes the whole. Basis: learners who marked charts and arranged items into them outperformed learners who wrote a structured summary of the same materials for the same time (learner-built organisation [+M], 20 high-school physics students). Learners building concept maps beat most comparison conditions, but not time-matched retrieval practice (concept mapping [~S]), so from the middle sessions on have learners redraw the map from memory before checking it.
  6. Apply to a whole task every one or two sessions. Give one integrative task (explain a case, solve a problem that uses several parts, produce a complete small version of the product) so the parts are used together, not only placed. Basis: an argument, not a study (whole-task practice and transfer [+W]); the cadence is an untested proposal.
  7. Check low-level detail separately. Detail placed low in the structure was recalled least two weeks later (in one group 50% at the top level falling to about 5% at the lowest; recall by hierarchy level [~M]). Where specific details matter, give them their own retrieval practice cued from their parent idea; do not assume the whole carries them.
  8. Fade the provided whole across sessions. Early sessions (about the first quarter of a unit): the teacher's map or model stays visible throughout and the teacher leads the closing placement aloud. Middle sessions: learners redraw the map from memory at the start, then compare with the reference, and do the closing placement themselves. Late sessions: learners build or extend their own map for new material and do the integrative tasks with the map hidden. Move a learner on from the overview once they can, unaided, place two new details correctly and explain one relation two levels down, in two consecutive sessions. The arc and the threshold are untested proposals; keep a learner who already meets the threshold out of overview work (expertise reversal [~M]).
  9. Do not force a whole onto list-like goals. For facts, names or vocabulary (verbal-association), use plain grouping and retrieval practice instead; tasks resting only on single details showed no detected benefit from a hierarchical organisation (no effect on local tasks [~M], a non-significant difference, not equivalence).
  10. Do not rely on the whole alone for learners with weaker prior attainment. Add the closing placement at every session, a check on the top-level ideas before moving on, and more practice per part. Basis: low-ability students showed no significant benefit from the hierarchical organisation (ability moderates the benefit [~M]), and overview organizers helped high-ability learners more than low-ability ones (advance organizers [~M]).

When a condition is not met: if no one can see learners' maps or syntheses (a large lecture, self-paced material), make the closing placement an arrange-or-label task that software or an answer key checks; if learners already hold the structure, skip steps 2–3 and start with step 6.

Fitting the design to a situation

The three facts that most change the decision:

  • Whether learners already hold an organisation for this content or task. Without one they need the whole and the closing placements; with one they need a quick check and the part they lack, and an overview costs time. If unstated, ask: before instruction, can they sketch how the main ideas relate, or carry out a simple complete version of the task, and place one detail?
  • What they will be asked to do. A whole organised around general ideas helped tasks that drew on those ideas and showed no detected benefit on tasks resting on single details. If unstated, ask: will they explain, solve and transfer, or recall specific facts, and in what form will they be assessed?
  • How much the ideas interact, against the learner's prior knowledge. If the whole only makes sense once several new ideas are held together, a novice may not be able to follow it first. If unstated, ask: can one of the top-level ideas be understood without the others?
If the brief says… Then change… Basis
Learners are novice and the content is high in element interactivity Keep the first whole to two or three ideas; pre-teach the one or two elements it depends on in a short isolated segment, then present the whole in the same session; make each later part small Part-task practice for novices [~M] (one experiment; sample and horizon not reported); pretraining [+M] (one VR lesson, carried to other content)
Learners are intermediate or advanced for this content Replace the overview with the opening sketch; start at the first part they cannot explain; spend the time on integrative tasks Expertise reversal [~M] (review of cognitive-load studies, carried to overviews; no crossover point)
Learners have weaker prior attainment, or low confidence Add a closing placement every session, a check on the top-level ideas before moving down, extra practice per part; praise correct placements privately Ability moderates the benefit [~M] (college physics volunteers); advance organizers [~M] (older meta-analyses); the confidence part is untested
Learners are child or adolescent Make the whole concrete (a picture, story or physical model of the whole), keep each part to one lesson, and have the teacher lead the placement aloud before learners do it untested proposal; the hierarchy studies used college and advanced high-school students
Learners are adults in a work setting (adult-workplace) Take the whole from their own work: the complete workflow or a real case they handle; elicit what they already do and start at the part where their account breaks down untested proposal; whole-task practice and transfer [+W] is an argument, not a study
The goal is verbal-association (facts, names, vocabulary) Do not build a whole-first sequence; group the items plainly and use retrieval practice No effect on local tasks [~M] (non-significant, not equivalence)
The goal is a complex-skill (whole-task performance) Make the whole a complete simple task, sequence tasks from simple to complex with the 4C/ID pattern, and keep the conceptual map as supporting information around the tasks untested proposal; whole-task practice and transfer [+W]
The assessment is high-stakes and uses one kind of task Put the ideas that task draws on at the top of the map and of every closing placement; assess a matched task, not a recall of the map Organisation–task match [~M] (20 high-school physics students, two-week test)
online-self-paced, no teacher present Keep the map on every screen with today's part highlighted; make the closing placement an arrange-or-label activity the system checks; add a redraw-from-memory item from the middle of the course Learner-built organisation [+M] (classroom study, untested online); concept mapping [~S] for the redraw; the persistent map is untested
Mixed mode: some learners remote, some in the room, or phone-only Share the map as one image everyone can see on a small screen (few labels, large type); do the closing placement as a numbered poll or chat answer ("which of the five ideas does this detail belong to?") so remote and in-room answers are seen together untested proposal
One teacher with a large group Have every learner show their placement at once (cards, mini-whiteboards, a poll), address the commonest misplacement to the whole group, then go to those still misplacing untested proposal
Time is one session only Give the whole and one part only; end with one placement and one integrative task; do not compress later parts into the same session or claim retention untested proposal
Learners study in a second language, or have low literacy State the whole as a labelled diagram with short phrases; keep the same term for the same idea in every session; let learners place items by arranging cards rather than writing a summary untested proposal
Learners will produce something for a real audience (a report, performance, product, or a talk to neighbours) Make that product the whole: show a complete simple example of it first, then teach parts as the pieces of the product, and end with learners making a short complete version untested proposal; whole-task practice and transfer [+W]
The syllabus or textbook is fixed in topic-by-topic order Keep the order but overlay a one-page map shown at the start and end of each topic, with the closing placement; add one integrative task per two or three topics untested proposal

Eight of the fifteen rows cite a claim. Each rests on one small study, an older synthesis, a review or an argument, and most carry a finding to learners or settings it did not test, as each row says; the other seven rows are untested proposals.

Observation, state and explanation

What can be observed is a response under stated conditions: the learner's sketch of the whole, where they place a new detail, what they say when asked how a part changes the whole, and their score on an integrative task and on a detail task, at a stated delay. "Has a coherent model", "is fragmenting knowledge" and "is lost in detail" are inferred states.

Response Plausible explanations Observation that could change the interpretation
Restates the overview but cannot place a new detail in it The overview is held as words, not as a structure; the detail's link to the whole was never stated; the overview's top level does not fit this kind of detail Give the detail and ask which top-level idea it belongs to and why; then give a detail from an earlier part. If both fail, add arrange-the-items placements; if only the new one fails, state its link.
Places details correctly but cannot solve an integrative problem Knows where things go but has not used parts together; the problem draws on ideas not at the top of the whole Give a matched problem with the relevant two ideas named; if that succeeds, add integrative tasks; if not, re-teach the relation.
Recalls the top-level ideas a week later but not the specifics Expected loss of low-level detail; detail never practised on its own Cue the detail with its parent idea; if cueing restores it, add cued retrieval; if not, re-teach.
Cannot follow the overview at all The ideas interact too much for the learner's prior knowledge; vocabulary is missing; the overview is too abstract Present one top-level idea on its own and ask what it means; if that works, simplify or make the whole concrete; if not, pre-teach the elements.
Already draws a full, correct map before instruction Holds the organisation; may hold a different but valid one Ask for a relation two levels down; skip the overview if it is explained, and compare their organisation with yours if it differs.

These rows are untested as diagnostics. A probe is also practice and changes what it measures, so record each one and keep explanations open until an observation separates them.

Evidence and qualifications

No claim compares a whole-first with a parts-first order or tests a holistic course design. The claims below test neighbouring parts of the relationship. Six of them come from one report (Eylon & Reif, 1979, three small experiments with physics students), so they are one research programme, not replications; the Elaboration Theory pattern reads them in more detail.

Learners, comparisons and outcomes differ across these claims (college and high-school physics students, novices on complex tasks, older syntheses, a VR lesson), so do not combine them into a predicted effect for holistic design.

The evidence by question

Further evidence, not yet read against this model

Objective and learner-valued goal

The designer's objective is usually integrative: explain how parts relate, solve problems that draw on several of them, or carry out the whole task, at a stated horizon. Elicit separately what the learner wants from the topic, because that decides which whole is worth giving first. A learner who wants to pass a fact-heavy test may see an overview as a delay; one who wants to do a real job may want the complete workflow before any theory. Agreement and divergence are both observations.

For example, a volunteer learning about a river catchment may want to explain to neighbours why it floods, while the designer's objective is to run a set of monitoring procedures. A causal whole (rain, land, channel, people) serves the volunteer's goal; a whole built from the procedures serves the designer's. Showing both, and saying which part of each the session elaborates, keeps one from being read as success on the other.

What would revise this model?

The expectation should weaken if comparisons of a whole-first against a parts-first order, with the same content, comparable learners and an aligned integrative outcome at a delay, show no advantage; if a learner-built whole does no better than retrieval practice of the parts for the same time; or if the advantage of a hierarchical organisation does not replicate beyond small physics samples. It should narrow further if weaker learners consistently gain nothing from the whole, in which case the default for them becomes parts first with an integrative task at the end. Do not protect the model by calling every failure "a whole too abstract" after the fact.

Performance on tasks matched to the whole, recall of details, delayed retention and transfer are separate outcomes. The present evidence establishes no overview size, no number of parts per session and no criterion for leaving the overview behind.

Claims

Examples

Illustrative

Reigeluth's Elaboration Theory — Learners begin with an epitome or simple whole, then elaborate progressively while keeping the larger structure in view.

Whole-task previews before subskill focus — In writing, programming, or design, learners first see the complete workflow or product before analyzing individual components.

Concept maps that return throughout a unit — The same framework is revisited as new detail is introduced, helping learners keep parts integrated into the whole.

Four-Component Instructional Design — Whole tasks of increasing complexity for a complex skill, with part-task practice only for routines that must become automatic.

Key Sources

  • Reigeluth, C. M. (1979). In search of a better way to organize instruction: The elaboration theory. Journal of Instructional Development, 2(3), 8-15. doi:10.1007/bf02984374
  • van Merriënboer, J. J. G., & Kirschner, P. A. (2018). Ten steps to complex learning (3rd ed.). Routledge. https://doi.org/10.4324/9781315113210
Page metadata
Field Value
type principle
id holistic-learning
title Holistic Learning
description For a learner without an organisation of their own for interconnected content, giving a usable whole early (a map of the main ideas, a model or a complete simple task) and keeping later details attached to it is expected to help integrative tasks more than learning details in isolation, unless the whole is too complex for a novice to follow; no claim tests whole-first against parts-first, and one small set of physics experiments on hierarchical organisation bears on it.
status review
canonical True
generated by claude/unspecified on 2026-10-05
trust unverified — no one has confirmed this page's content
sources 1 — van-merriënboer-2018

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