Skip to content

Redundancy Principle

Principle · All principles
Evidence · 22 claims (6 for, 16 mixed) · 19 studies (6 causal, 6 review, 5 quant-synthesis, 2 theoretical), q1–q4 · 5 of 19 report an effect size · 18 claims rest on one study

Conditional relationship

This page owns one decision: whether to present the same information twice, most often the same words both spoken and written, and more broadly a second source (a caption, an explanatory note, a transcript) that repeats what another source already carries. The answer depends on what else is on the screen and who is learning.

When a picture, diagram or animation is explained by narration, adding on-screen text that repeats the narration word for word is expected to lower transfer for learners new to the content. In Mayer's programme, learners given animation and narration did better on a transfer test than learners given the same animation, narration and on-screen text in 5 of 5 experiments, median d = 0.72. A 63-study review places the harm in the same place: written text competing with a visualization. When no picture competes for the eyes, the same words spoken and written do not hurt: across a meta-analysis of verbal redundancy they did not differ from written-only presentations and beat spoken-only ones, with the advantage largest for second-language and less fluent readers. And as expertise grows, explanatory text that novices need becomes redundant: experienced trainees did better with a diagram alone than with the text integrated into it.

The relationship is conditional on six things, each from the evidence below:

  • Whether a visualization is present. Spoken plus written words beat spoken-only words with no image (g = 0.45) but not when images or animation were included (g = 0.06 in both cases) (Adesope & Nesbit 2012 [~M]). The Mayer harm was measured with animations.
  • Who the readers are. The spoken–written advantage was g = 0.82 for nonfluent first-language readers, 0.58 for second-language readers and 0.24 for fluent readers; for low prior knowledge 0.29, for high prior knowledge −0.05 (a marginal difference).
  • Pacing. The spoken–written advantage appeared for system-paced material (g = 0.34) and not for learner-paced material (g = 0.12); multimedia design principles in general mattered more in system-paced media (g = 0.41 against 0.27 learner-paced; Noetel et al. 2022).
  • Verbatim or abridged. Key terms extracted from the narration went with better outcomes than verbatim duplicated text; the review reports abridged text as less harmful than verbatim.
  • The outcome measured. The spoken–written advantage was g = 0.53 on retention and 0.01 on transfer; 0.63 on multiple-choice tests against 0.11 on mixed formats. Mayer's harm from duplicated text was measured on transfer.
  • Age. In a middle-school quasi-experiment on electricity, adding written text to narrated animation made no detectable difference; the review lists learner age among the moderators that reverse the effect.

How this sits inside its umbrella: Multimedia Learning states the general relationship (a relevant picture with the words that explain it) and names redundancy as one of its qualifications; this page is that qualification in depth and does not repeat the umbrella's model. It borders three neighbours. The modality question (speak the words or print them, when a picture is present) is the umbrella's; this page starts after it is answered, and asks whether to add the printed copy back. Split attention concerns two sources that each need the other (Spatially integrating text and diagrams [+M]); redundancy concerns a source that is intelligible alone, so the remedy is removal, not integration. Coherence concerns material that is irrelevant; redundant material is relevant but repeated. Behind all three sits Cognitive Load Management: redundancy is one source of avoidable load, and as with the other supports, what helps a novice can cost an expert.

Default design, while the dose is untested

The claims establish when duplication hurts and when it helps; none sets a dose (how many on-screen words, how long a segment, when to withdraw text). The rules below follow the evidence; the numbers and the progression are untested proposals.

  1. Inventory each segment's channels. For every screen, slide or page, list what is spoken, what is written and what is pictured, and mark each written item as verbatim (repeats the narration), abridged (key terms of it) or complementary (says something the narration does not). (Untested proposal: the audit itself.)
  2. Narrated picture or animation: remove the verbatim on-screen text. Keep narration plus picture, and at most a few key terms (untested proposal: about one to five words per screen) placed on the parts of the picture they name. Basis: Redundancy principle: animation plus narration beats animation, narration, and on-screen text for transfer [+M]; Redundant on-screen text impairs learning when it competes with a visualization [~M]; key terms over verbatim text from On-screen text that duplicates the narration of a diagram can hurt learning, but spoken–written redundancy is not harmful on average and helps over spoken-only presentation in some conditions [~S].
  3. Words with no picture (an audio explanation, a read-aloud, a lecture recording without slides): give the words both spoken and written for second-language, less fluent or low-prior-knowledge readers, and especially for system-paced, unsegmented audio. Basis: Spoken–written advantages are larger for second-language and nonfluent first-language readers and for low prior knowledge learners [+M]; Spoken–written presentations do not outperform written-only presentations but do outperform spoken-only presentations [~M].
  4. Self-paced reading pages: written text alone is enough for fluent readers; offer the audio on demand rather than by default. Basis: spoken–written did not beat written-only (g = −0.04), and learner-paced material showed no detectable advantage (system-paced only [~M]).
  5. Video captions: on by default for second-language learners and for anyone who needs them for access; a toggle for everyone else. Basis: Captioning second-language videos produces large benefits for comprehension and vocabulary acquisition [+M]; the toggle for fluent viewers carries step 2 to video, an untested proposal.
  6. Across sessions, withdraw explanatory text as expertise grows. Early sessions (learners new to the diagrams or notation): explanatory text integrated into the diagram. Once a learner can explain a diagram from the diagram alone (untested proposal: on two consecutive sessions, on a diagram not seen before), switch that learner to diagram-only material and keep the text available on request. If the unaided explanation fails, return the text. Basis: The split-attention benefit for novices reverses into a redundancy effect for more experienced learners [~M]; Instructional guidance that helps novices can become redundant or counterproductive as expertise grows [~M]; the switching rule is an untested proposal.
  7. Segment length is not set by these claims. Keep each narrated segment to one step of the explanation (untested proposal: two to four minutes), and note that the spoken–written advantage was smaller for segmented presentations (g = 0.15) than unsegmented ones (g = 0.53) (segmentation moderates verbal redundancy [~M]): once audio is segmented, written duplicates matter less.
  8. Judge the configuration on a transfer item, not on a retention or multiple-choice test alone, where duplication looks better than it is. Basis: Spoken–written advantages are larger for retention outcomes and multiple-choice test formats [~M]. Use the brief's own assessment when it gives one; add no test it did not ask for, and instead read its transfer items, if it has any, separately.

Fitting the design to a situation

Three facts change the decision most. Ask for them when the brief omits them: is a picture, diagram or animation on screen while the words are spoken? (if yes, remove verbatim text; if no, spoken plus written can help); how fluent are the learners as readers of this language, and how much do they already know of this content?; and who controls the pace (a video or lecture that runs on, or pages and pause controls the learner uses)?

If the brief says… Then change… Basis
A narrated animation, screencast or explainer video carries the full narration as on-screen text, for novice learners Remove the on-screen transcript; keep narration, picture and a few key terms on the picture Redundancy principle (Mayer 2008) [+M] (5 of 5 experiments, median d = 0.72, transfer); Redundant on-screen text competing with a visualization [~M]
The words have no picture with them (audio lesson, podcast, read-aloud), learners are low on prior knowledge Give the words spoken and written together Spoken–written beats spoken-only, not written-only [~M] (g = 0.29 against spoken-only); Verbal redundancy and images [~M] (g = 0.45 without images)
Learners read in a second language, or are less fluent first-language readers Keep spoken and written words together; for video, captions on by default Reader fluency and prior knowledge moderate redundancy [+M] (L2 g = 0.58, nonfluent L1 g = 0.82); Captioning second-language videos [+M] (comprehension g = 0.99)
Learners are advanced or know this content well Drop explanatory text that repeats what a diagram already shows; give diagram-only material and text on request Split-attention benefit reverses into redundancy for experienced learners [~M]; high prior knowledge g = −0.05 on reader fluency and prior knowledge [~M]
Learners are children in the middle-school years Do not expect a cost from adding written text to narrated animation; remove it if it crowds the picture, keep it if learners rely on it The redundancy effect did not hold for middle school students [~M] (no detectable difference; equivalence not tested)
Learners are young children still learning to read Keep the narration; add short written labels only for words the lesson is teaching untested; no claim here tests redundancy with beginning readers
Material is self-paced (pages, pausable video) Treat duplication as a low-stakes choice: fluent readers can ignore one channel; give audio on demand Spoken–written advantage only for system-paced material [~M] (g = 0.12 learner-paced, not detectable)
Material is system-paced (a live lecture, a video that runs on, a broadcast) Apply the picture rule strictly: this is where design choices matter most Design-principle effects larger in system-paced media [~M] (g = 0.41 against 0.27)
Slides with full sentences that the presenter reads aloud Cut slides to key terms and pictures; say the sentences untested in live lectures; carries Redundancy principle (Mayer 2008) [+M] from narrated animation to slides
Print only, with a diagram and text that restates what the diagram shows For novices keep the text, placed next to what it describes; for experienced readers a diagram alone may do Redundant text in a diagram made no difference [~M]; expertise reversal [~M]
VR, 3D simulation or a game with narration Do not float transcript panels in the scene; anchor short labels to objects, captions as a toggle untested; carries the picture rule to a setting no claim here tested
Phone-only, small screen Captions as a toggle, not burned in; no duplicated text panels over the picture untested proposal
A single session Apply steps 2–5 of the default; skip the across-session withdrawal of text Redundancy principle (Mayer 2008) [+M]; the rest is the default above
A fixed short time budget (for example 10 minutes a day) Removing a duplicate costs no time; do not spend minutes building parallel spoken and written versions unless learners are L2 or less fluent readers untested proposal
One teacher for many learners, mixed reading fluency Produce one version with a caption or transcript toggle and let learners choose untested proposal; fluency evidence in the rows above
Learners are deaf or hard of hearing Captions and transcripts are required access, not redundancy; keep them, and keep the picture uncluttered untested; the redundancy evidence assumes learners hear the narration
Learners are blind or have low vision Give one audio track: narration plus description of the visual; do not let a screen reader read on-screen text that repeats the narration untested proposal
The brief supplies its own materials, routine, rubric or assessment Keep them as given; change only verbatim duplicates over pictures; read a retention or multiple-choice result as favouring duplication more than a transfer result would Spoken–written advantages larger for retention and multiple-choice [~M] (retention g = 0.53, transfer g = 0.01)
The brief wants a whole demonstration kept whole (no segmenting) Keep it whole; with no picture, add written key text, which helped more in unsegmented presentations Segmentation moderates verbal redundancy [~M] (g = 0.53 unsegmented, 0.15 segmented)
Private material (recorded patients, clients, learners) Generate any captions or transcripts in the same protected setting as the recording and do not publish them separately untested proposal
High stakes (certification, safety training) Judge the design on transfer items; do not remove captions or transcripts that access rules require Retention and multiple-choice overstate the spoken–written advantage [~M]

Observation, state and explanation

These rows are proposals, untested with learners as diagnostics.

Response Plausible explanations Observation that could change the interpretation
Recalls the narration's words but cannot explain the animation Reading the duplicated text took the eyes off the picture; the test asked for recall Re-run one segment without the on-screen text and ask a "what if this part changed" question.
Asks for the transcript, or turns captions on Hard to follow the speech (language, accent, pace, noise, hearing); a habit Ask why; check comprehension of the spoken words alone. If access is the reason, the text is not redundant for this learner.
An experienced learner rates the integrated diagram as tedious and does no better with it The explanatory text duplicates their schema (expertise reversal) Give a diagram-only version and an unaided explanation item; if they succeed, keep diagram-only.
Retention quiz improves after adding written text, transfer does not Duplication helps recognition and recall, not use Score transfer items separately before keeping the change.

Evidence and qualifications

Taken together: with a picture on screen, verbatim narration-plus-text costs transfer for novices in short system-paced lessons (one programme's experiments, supported by the review); without a picture, spoken plus written words help less fluent and low-knowledge readers and do no harm, with the gains concentrated on retention and multiple-choice tests; for experienced learners, explanatory text becomes the redundant source. No claim in the wiki sets a dose, tests the withdrawal schedule, or tests redundancy in live lectures, VR, or with beginning readers.

The evidence by question

Each question links a general claim that gathers the studies behind it (this page's synthesis spec, built by scripts/synthesize_claim.py; one row per distinct study, no averaged verdict).

Further evidence, not yet read against this model

Objective and learner-valued goal

The designer's objective is usually understanding that transfers: explaining how the system in the animation works, using the idea on a new problem. Ask separately what the learner wants. A second-language learner may value the written words for learning the language itself, which the captioning evidence supports; a learner who will take a multiple-choice recognition test next week may prefer the transcript, and on that test it may well help. Where the learner's goal and the designer's diverge, say so and offer the text on request rather than deleting it by rule or imposing it.

What would revise this model?

The picture rule should weaken if comparisons outside Mayer's programme, with comparable novices and transfer tests, find no cost of verbatim text over narrated animation; the review's reversals for older learners and learner-paced delivery point to where that might happen. The no-picture rule should weaken if spoken plus written words hurt fluent readers on transfer. The expertise rule should be extended, not reversed, by a test of when to withdraw text (the switching rule in the default is a proposal). Studies of live lectures, VR and phone-only video would extend the model to settings it does not yet cover.

Examples

  • Multimedia Learning — the pattern that carries this principle into a lesson ("do not duplicate words over a visualization; do not delete text by rule").
  • Accommodations — captions and transcripts kept for access.

Key Sources

  • Adesope, O. O., & Nesbit, J. C. (2012). Verbal redundancy in multimedia learning environments: A meta-analysis. Journal of Educational Psychology, 104(1), 250–263. doi:10.1037/a0026147
  • Mayer, R. E. (2008). Applying the science of learning: Evidence-based principles for the design of multimedia instruction. American Psychologist. doi:10.1037/0003-066x.63.8.760
  • Trypke, M., Stebner, F., & Wirth, J. (2023). Two types of redundancy in multimedia learning: a literature review. Frontiers in Psychology, 14, Article 1148035. doi:10.3389/fpsyg.2023.1148035
  • Noetel, M., Griffith, S., Delaney, O., Harris, N. R., Sanders, T., Parker, P., del Pozo Cruz, B., & Lonsdale, C. (2022). Multimedia design for learning: An overview of reviews with meta-meta-analysis. Review of Educational Research. doi:10.3102/00346543211052329
  • Kalyuga, S., Chandler, P., & Sweller, J. (1999). Managing split-attention and redundancy in multimedia instruction. Applied Cognitive Psychology, 13(4), 351–371. doi:10.1002/(SICI)1099-0720(199908)13:4<351::AID-ACP589>3.0.CO;2-6
  • Kalyuga, S., Chandler, P., & Sweller, J. (1998). Levels of expertise and instructional design. Human Factors, 40(1), 1–17. doi:10.1518/001872098779480587
  • Kalyuga, S., Ayres, P., Chandler, P., & Sweller, J. (2003). The expertise reversal effect. Educational Psychologist, 38(1), 23–31. doi:10.1207/S15326985EP3801_4
  • Kutbay, E., & Akpınar, Y. (2020). Investigating modality, redundancy and signaling principles with abstract and concrete representation. International Journal of Education in Mathematics, Science and Technology, 8(2), 131–145. ERIC EJ1255527
  • Clinton, V., Alibali, M. W., & Nathan, M. J. (2016). Learning about posterior probability: Do diagrams and elaborative interrogation help? The Journal of Experimental Education, 84(3), 579–599. doi:10.1080/00220973.2015.1048847
Page metadata
Field Value
type principle
id redundancy
title Redundancy Principle
description When a picture, diagram or animation is explained by narration, adding on-screen text that repeats the narration word for word is expected to lower transfer for novices, most in system-paced lessons; when no picture competes for the eyes, the same words spoken and written do no harm and help second-language, less fluent and low-prior-knowledge readers, and text that helps novices becomes redundant as expertise grows. Claims test this: Mayer's experiments, a meta-analysis of verbal redundancy, a 63-study review and a meta-meta-analysis, with boundary conditions for pictures, pacing, reader fluency, prior knowledge, outcome and age.
status review
canonical True
generated by claude/unspecified on 2026-10-14
trust unverified — no one has confirmed this page's content
sources 9 — adesope-nesbit-2012, mayer-2008, trypke-et-al-2023, noetel-et-al-2022, kalyuga-et-al-1999, kalyuga-et-al-1998, kalyuga-et-al-2003, kutbay-akpinar-2020, clinton-et-al-2016

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