Skip to content

Coherence Principle

Principle · All principles
Evidence · 37 claims (15 for, 20 mixed, 2 against) · 39 studies (17 causal, 8 review, 5 quant-synthesis, 5 theoretical, 2 associational, 1 qualitative, 1 design), q1–q4 · 10 of 39 report an effect size · 30 claims rest on one study

Conditional relationship

This page owns one decision inside Multimedia Learning: what to leave out. When learners are studying an explanation of a process, system or idea they do not yet understand, and the lesson carries material that is interesting but not needed for that understanding (a vivid anecdote, a fun fact, a decorative photograph, a tangential video clip, background music), leaving that material out is expected to produce better comprehension and transfer than the same lesson with it in. The added material is called a seductive detail because learners usually like it: it can raise mood or interest while drawing attention, and working memory, away from the explanation.

The relationship is tested in the wiki's claims, and the claims bound it in five ways:

  • How large the effect is depends on who studied it. In Mayer's own programme of short lab lessons, a concise lesson beat an expanded one on a transfer test in 13 of 14 experiments, median d = 0.97 (Mayer 2008). Rey's (2012) meta-analysis confirmed the direction on retention and transfer, with a smaller magnitude than the headline experiments, and Sundararajan and Adesope's (2020) meta-analysis found the effect depends on design features. An independent e-learning study measured a cost of η² = .04, which its authors call very small (González et al. 2019). Pooled across all multimedia principles the average was g = 0.38, and which principle was applied explained most of the variance, so that figure is not coherence's own size either (Multimedia design principles improve learning overall [+M]). Plan for a reliable but modest effect, not d = 0.97.
  • What is added. Interesting but irrelevant text and pictures cost transfer in science lessons (Harp & Mayer 1998); readers recalled the seductive details themselves better and the main ideas worse (Garner et al. 1992). Purely decorative pictures, in three experiments with 7th and 8th graders, neither helped nor harmed learning overall, but weakened the benefit of the instructional pictures beside them (Lenzner et al. 2013). The type of image, the delivery format, language, subject, pacing and the kind of recall question all moderated the effect in the meta-analysis.
  • Who the learner is. The cost was larger for readers with lower working-memory capacity (Sanchez & Wiley 2006), appeared on harder texts only in students with low attentional inhibition (González et al. 2019), and the loss of the instructional-picture benefit was largest for lower prior knowledge (Lenzner et al. 2013). A combined revision of a middle-school mathematics lesson for signalling, contiguity and coherence lowered processing load for low prior knowledge students and raised it for middle prior knowledge students, with no difference in test scores (Clinton et al. 2017). Pooled over all multimedia principles, prior knowledge did not significantly moderate effects (Noetel et al. 2022).
  • Pacing and complexity. Across multimedia design principles, not coherence alone, design mattered more in system-paced media (g = 0.41) than in learner-paced media (g = 0.27), and more for complex material (g = 0.70) than simple material (g = 0.20) (Noetel et al. 2022).
  • The outcome. The evidence is on comprehension, recall of main ideas and transfer, mostly soon after a short lesson. Liking, mood and situational interest are separate outcomes and can move the other way.

How this sits beside its neighbours: Multimedia Learning holds the umbrella model (words with a relevant picture) and its other principles; this page is its "cut what the explanation does not need" principle in depth, and does not repeat that model. Redundancy (the same words twice, spoken and printed) is a separate decision on the umbrella page: redundant words are needed content delivered twice, while a seductive detail is unneeded content delivered once. Cognitive Load Management gives the general account of why extraneous processing costs learning; the overview that pooled load-reducing principles, removing seductive details among them, found the benefit largest for complex material, as the overload account predicts (Cognitive overload degrades learning [+M]). Signalling is the constructive counterpart: cue what matters rather than add what does not.

Default design, while the dose is untested

The claims establish the direction and its moderators; none sets a lesson length or a schedule. The steps follow the evidence where a claim is named; the timings and the progression are labelled proposals, untested.

  1. Name what the explanation is for. Write one sentence saying what the learner should be able to explain or do afterwards, and list the parts of the content that sentence needs. (untested proposal)
  2. Audit every element against that list. For each block of text, picture, sound, clip and animation, ask whether the goal needs it. Cut background music, decorative photographs, tangential anecdotes and fun facts, and extra clips that do not carry the explanation. Check emphasis too: bold, underlining or highlighting on incidental material is incoherence of its own, since in one experiment underlining low-relevance text raised scores on that material (Coherence principle [+S]; Relevancy of emphasis directs attention [+M])
  3. Replace rather than only delete where a picture is wanted. Swap a decorative image for one that shows the relation being explained, such as a sketch of the problem's structure in place of a photograph of its setting. (Decorative photographs distract from the mathematics [+W]; Seductive details effect [~S], where decorative pictures weakened the benefit of instructional ones)
  4. Put interest material outside the explanation. If a story or hook is wanted, place it before or after the core explanation, not interleaved with it. (untested proposal)
  5. Keep each core explanation short. One idea per segment, about 3–6 minutes of narration or one screen sequence, with segmenting handled as on Multimedia Learning. (untested proposal for the length)
  6. Progression across sessions. (untested proposal, except where named)
  7. Early sessions on a new topic: the strict version above, for every learner. Lower prior knowledge, lower working memory and low attentional control are where the claims record the cost (Seductive details effect [~S]).
  8. Moving on: when a learner can explain the core mechanism on a changed case without the materials, add elaborations that carry content (applications, worked contexts, a story that is itself an instance of the idea) after the core explanation.
  9. Later sessions, as learners become experienced: support they no longer need, including explanatory text added to a diagram, becomes extraneous in its turn; cut it (Multimedia principles benefit novices [~M], where experienced trainees did better with the explanatory text removed; Split attention reverses into redundancy for experienced learners [~M]; Expertise reversal [~M], where effects that help novices shrink, vanish and can reverse with expertise).
  10. Judge it on the brief's own assessment of understanding, comparing explanation quality, not on how much learners liked the version. Add no test the brief did not ask for.

Fitting the design to a situation

Three facts change the decision most. Ask for them when the brief omits them: does the added material carry anything the learning goal needs, or is it there to engage (ask: "if this were removed, what could the learner no longer explain?"); how much do the learners already know of this content, and how complex is it (ask what they have studied before and how many interacting parts the explanation has); and who controls the pace (video, live lecture or narration, or pages the learner reads and replays at will).

If the brief says… Then change… Basis
The brief asks for engaging extras in an explanation: stock photos, background music, anecdotes, fun facts Leave them out of the explanation; if interest is needed, use material that carries the content (a relevant example, an informative picture) Coherence principle [+S] (13 of 14 experiments in one programme, median d = 0.97); Seductive details effect [~S] (direction holds, size varies with material and learner)
Learners are novice on the content, and the explanation has many interacting parts Apply the audit at full strength, and expect it to matter most here Design principle effects moderated by pacing and complexity [~M] (g = 0.70 complex, 0.20 simple, across principles); Multimedia principles benefit novices [~M]
Learners have middle or high prior knowledge of the content Do not expect a gain from stripping further; also cut support they no longer need, and check transfer before and after Revised visuals raise processing for middle prior knowledge students [~M]; Revisions had no effect on test scores [~M]; Multimedia principles benefit novices [~M]; Split attention reverses into redundancy for experienced learners [~M]
The material is simple: few parts, a short text Give the audit lower priority; a decorative picture on a simple page costs little Design principle effects moderated by pacing and complexity [~M]; Seductive details effect [~S] (η² = .04 on easy texts)
Learners are children or young adolescents (primary, secondary) Keep decorative pictures out of and away from the instructional pictures; the evidence with 7th and 8th graders found no overall harm but a weaker benefit from the informative picture. For children below that age, treat it as a proposal Seductive details effect [~S] (Lenzner et al., grades 7–8; Garner et al., adults and children); younger children untested
Video, a narrated animation or a live lecture (system-paced) Cut hardest here: learners cannot slow down or skip. No background music, no tangential clips, stories before or after Design principle effects moderated by pacing and complexity [~M] (g = 0.41 system-paced, 0.27 learner-paced, across principles)
online-self-paced pages, learner-paced Still cut, but expect a smaller effect; give learners a clear path through essential content so optional extras can be skipped Design principle effects moderated by pacing and complexity [~M]; Hypermedia usability problems [~M] (distraction and overload in learner-controlled hypermedia)
Slides or print Each slide or page: the essential picture and its words. Remove decorative clip-art and images that cover text Seductive details effect [~S]; eBook images covering the text [~W] (teacher reports)
Phone-only, small screen One idea per screen, no decorative headers or background images, so the essential picture can be large enough to read untested
VR or an immersive simulation Keep the scene to the objects the task needs, and remove ambient sound and decoration that carry nothing for the task untested; the claims in the wiki concern text, pictures, video and narration
A single session, or a fixed short time budget Spend the minutes on the essential explanation and the brief's own check. Do not plan on cutting extras to save time: a combined revision made no reliable difference to time with the lesson Coherence principle [+S]; Revisions made no reliable difference to time [~M]
One teacher for many learners, a live lecture Keep the explanation itself free of digressions; tell the anecdote after the point it illustrates, and only if it is an instance of it Design principle effects moderated by pacing and complexity [~M]; placement untested
Learners with low literacy, or studying in a second language Every extra sentence costs reading effort: cut text the goal does not need first, and keep an informative picture Seductive details effect [~S] (larger cost with lower working memory; language moderated the meta-analysis); Split attention reverses into redundancy for experienced learners [~M] (Yeung et al., as reported in a review: explanatory notes built into a text helped learners with low language competence and hindered more knowledgeable ones, so what to cut depends on the reader's language level); second-language learners as such carried over, untested
Learners with attention difficulties, or low attentional control Treat removal as essential, not optional; this is where the cost was measured on harder texts Seductive details effect [~S] (González et al., Study 2: harm only with low attentional inhibition)
Learners who are blind or partially sighted, or deaf or hard of hearing Mark decorative images as decorative (empty alt text) so a screen reader skips them; give no background music under narration; caption only the essential audio untested
The brief supplies its own materials, routine or rubric Keep them as given. Audit the supplied materials only for additions that carry no content and remove those; run the brief's own assessment as given untested
The brief wants culturally relevant imagery, or learners to see people like themselves Keep it where learners value it, and make it carry content where possible (people from the community doing the task being taught); keep it out of the explanatory diagram Hispanic adults valued culturally relevant visual design [~W] (preference, not a learning outcome)
The brief wants problems set in learners' own interests or contexts A context that carries the problem's structure is not a seductive detail; keep it, and remove only the setting details the mathematics does not use Personalized problem contexts [~W] (practice-guide assertion); Decorative photographs distract from the mathematics [+W]
A whole task is kept whole: a realistic case or authentic problem Do not strip details the real performance has to deal with; deciding what is irrelevant may be the skill. Cut only presentation extras around the task Causal knowledge helps novices with cases containing irrelevant findings [~W] (carried over: a test case with irrelevant findings, not a lesson with seductive details)
The brief involves private material: learners' own cases, records or stories Use only the details the explanation needs; personal detail that carries nothing for the goal is both extraneous and a privacy cost untested
The brief is about the physical classroom: walls and displays Aim for moderate display, not bare walls and not clutter Classroom design affects learning progress [~M] (associational; moderately decorated rooms outperformed bare and heavily decorated ones)
High stakes: certification, a safety-critical procedure Apply the audit, and check understanding on the brief's own assessment at the delay that matters, since the recorded effects are mostly on immediate tests untested for delayed outcomes

Observation, state and explanation

These rows are proposals, untested with learners as diagnostics.

Response Plausible explanations Observation that could change the interpretation
Learners like the embellished version and explain the mechanism no better Interest and mood rose without informing; the added material drew attention from the explanation Remove the addition and compare explanation quality on a changed case, not liking.
Learners recall the anecdote or fun fact and not the main idea The seductive detail captured processing that the main idea needed (as in Garner et al.) Ask for the main idea in their own words with and without the addition; move the anecdote after the explanation.
One learner struggles with the illustrated version while others do not Low attentional control or lower working memory; low prior knowledge Re-present the essential picture alone to that learner; ask about prior study of the topic.
A stripped-down version is read more slowly, or with more effort, by learners who knew some of the content Middle prior knowledge learners may be processing an unfamiliar layout, or the cut removed context they used Compare test performance, not only effort or time; ask what they missed.
No difference between versions The material was simple; the learners were already experienced; the addition was ignored (as decorative pictures were in eye-tracking); the test measured recall the addition did not touch Check complexity and prior knowledge; use a transfer question before concluding.

Evidence and qualifications

  • Coherence principle: irrelevant material hurts learning [+S]: Mayer's (2008) review of his programme, concise lessons beat expanded ones on transfer in 13 of 14 experiments, median d = 0.97; Harp and Mayer (1998) and Mayer, Heiser and Lonn (2001), experiments in which added interesting material reduced transfer; and Rey's (2012) meta-analysis, which confirmed the direction with a smaller average effect. The d = 0.97 is one research group's short lab lessons with immediate transfer tests; read it as that programme's result.
  • Seductive details effect [~S]: Sundararajan and Adesope's (2020) meta-analysis, read as an abstract, found seductive details can hinder learning with the effect depending on image type, delivery format, language, subject, pacing and question type; Garner et al. (1992), poorer recall of main ideas; Sanchez and Wiley (2006), a larger cost for lower working-memory readers; Lenzner et al. (2013), no overall harm from decorative pictures for 7th and 8th graders but a weaker benefit from instructional pictures, most for lower prior knowledge; González et al. (2019), a cost of η² = .04 in Study 1 and, on harder texts, harm only for low attentional inhibition (interaction η² = .15).
  • Design principle effects moderated by pacing and complexity [~M]: Noetel et al. (2022), across multimedia principles pooled, not coherence alone. System-paced g = 0.41 against learner-paced 0.27; complex 0.70 against simple 0.20; prior knowledge not a significant moderator (p = 0.14, which does not establish equivalence). The same overview's pooled average across principles, g = 0.38, with the principle itself explaining most of the variance (Multimedia design principles improve learning overall [+M]), and its reading that complex material is where poor design overloads learners (Cognitive overload degrades learning [+M], which also cites Sweller 1988, read as an abstract only).
  • Multimedia principles benefit novices [~M]: Kalyuga et al. (2003, review) and (1998, three experiments): support that helps novices loses its advantage, and can become a cost, as learners gain experience; the most experienced trainees did better with explanatory text removed from a diagram. It is evidence about redundancy and split attention, carried here to the claim that what counts as extraneous changes with expertise. The same reversal, as reviews report it: integrated text that novice electrical trainees needed became redundant for experienced ones, and integrated explanatory notes helped low language-competence readers and hindered more knowledgeable ones (Split attention reverses into redundancy for experienced learners [~M], Kalyuga et al. 2003 and Ginns 2006 citing Kalyuga et al. 1998 and Yeung et al. 1998); and a 2019 narrative review's statement that such effects first decrease, then disappear and can reverse (Expertise reversal [~M]). None of these tests seductive details.
  • Revised visuals raise processing for middle prior knowledge students [~M] and Revisions had no effect on test scores [~M]: Clinton et al. (2017), 62 middle-school students, one lesson revised for signalling, contiguity and coherence together. Eye-tracking load fell for low prior knowledge students and rose for middle prior knowledge students, and test scores did not differ by condition. The same pattern held for total fixation time on the lesson text (Revised visuals reduce processing of lesson text for low prior knowledge students [~M]) and on the mathematics visuals (Revised visuals reduce processing of math visuals for low prior knowledge [~M]); average fixation length on the visuals was also shorter in the revised version for high prior knowledge students (Revisions to math-relevant visuals affect load by prior knowledge [~M]); and condition made no reliable difference to time with the lesson (Revisions made no reliable difference to time [~M]; a nonsignificant difference does not establish equal time). Because three principles were changed at once, it cannot say what coherence alone did.
  • Relevancy of emphasis directs attention [+M]: Nist and Hogrebe (1985), 67 college freshmen; underlining high-relevance or low-relevance text raised scores on whichever material was underlined. One experiment, F values only, immediate test; carried here to emphasis on incidental material as a form of incoherence.
  • Hypermedia usability problems [~M]: Scheiter and Gerjets (2007, review) name distraction, disorientation and cognitive overload as reasons learner-controlled hypermedia is hard to show effective; it bears on the learner-paced row, not on seductive details as such.
  • Classroom design affects learning progress [~M]: associational; moderately decorated primary classrooms outperformed bare and heavily decorated ones. The physical room, carried over to the wall-display row.
  • Decorative photographs distract from the mathematics [+W]: a curriculum developer's application of the principle, with no new data; it shows the replace-not-only-delete move.

Taken together: leaving out interesting but irrelevant material reliably does not hurt and usually helps comprehension and transfer, most for novices, for complex system-paced material and for learners with lower working memory or attentional control. The size is large in one programme's lab lessons and small in independent work. No claim in the wiki tests coherence alone with delayed outcomes, in VR, with young children or with second-language learners specifically.

The evidence by question

Each question links a general claim that gathers the studies behind it and the narrower claims that bear on it into one table, one row per distinct study, with no averaged verdict (scripts/synthesize_claim.py).

Further evidence, not yet read against this model

Objective and learner-valued goal

The designer's objective is understanding that transfers: explaining a mechanism, solving a changed problem. Ask separately what the learner values. Learners often prefer the decorated version, and decorative pictures improved mood, alertness and perceived difficulty in one set of experiments (Seductive details effect [~S]); a learner who will not open a bare lesson learns nothing from it. Learners may also value seeing people like themselves in materials (culturally relevant visual design [~W]). Where liking and learning diverge, keep what learners value outside the core explanation or make it carry content, and say which outcome each choice serves.

What would revise this model?

The expectation should weaken if independent comparisons with comparable learners, holding the essential content equal, find no comprehension or transfer cost from added interesting material, or if a cost on immediate tests disappears at a delay. It should be qualified, not reversed, if added material that raises interest is shown to improve persistence or later learning in self-paced settings enough to offset its cost. A finding that middle or high prior knowledge learners do worse with stripped materials on learning outcomes, not only on processing measures, would move the boundary in the table. Do not protect the model by relabelling any addition that helped as "relevant" after the fact.

Claims

Examples

Key Sources

  • 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
  • Rey, G. D. (2012). A review of research and a meta-analysis of the seductive detail effect. Educational Research Review, 7(3), 216–237. doi:10.1016/j.edurev.2012.05.003
  • Sundararajan, N., & Adesope, O. (2020). Keep it coherent: A meta-analysis of the seductive details effect. Educational Psychology Review, 32(3), 707–734. doi:10.1007/s10648-020-09522-4
  • Harp, S. F., & Mayer, R. E. (1998). How seductive details do their damage: A theory of cognitive interest in science learning. Journal of Educational Psychology, 90(3), 414–434. doi:10.1037/0022-0663.90.3.414
  • Mayer, R. E., Heiser, J., & Lonn, S. (2001). Cognitive constraints on multimedia learning: When presenting more material results in less understanding. Journal of Educational Psychology, 93(1), 187–198. doi:10.1037/0022-0663.93.1.187
  • Lenzner, A., Schnotz, W., & Müller, A. (2013). The role of decorative pictures in learning. Instructional Science, 41(5), 811–831. doi:10.1007/s11251-012-9256-z
  • González, F. M., Saux, G., & Burin, D. (2019). The decorative images' seductive effect in e-learning depends on attentional inhibition. Australasian Journal of Educational Technology, 35(3). doi:10.14742/ajet.4577
  • Sanchez, C. A., & Wiley, J. (2006). An examination of the seductive details effect in terms of working memory capacity. Memory & Cognition, 34(2), 344–355.
  • Garner, R., Gillingham, M. G., & White, C. S. (1992). Effects of "seductive details" on macroprocessing and microprocessing in adults and children. Cognition and Instruction, 9(4), 359–373.
  • 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, 92(3), 413–454. doi:10.3102/00346543211052329
  • Kalyuga, S., Ayres, P., Chandler, P., & Sweller, J. (2003). The expertise reversal effect. Educational Psychologist, 38(1), 23–31. doi:10.1207/s15326985ep3801_4
  • Kalyuga, S., Chandler, P., & Sweller, J. (1998). Levels of expertise and instructional design. Human Factors, 40(1), 1–17. doi:10.1518/001872098779480587
  • Clinton, V., Cooper, J. L., Michaelis, J., Alibali, M. W., & Nathan, M. J. (2017). How revisions to mathematical visuals affect cognition: Evidence from eye tracking. In Eye-tracking technology applications in educational research (pp. 195–218). IGI Global. ERIC ED574984. doi:10.4018/978-1-5225-1005-5.ch010
Page metadata
Field Value
type principle
id coherence
title Coherence Principle
description When learners study an explanation of something they do not yet understand, leaving out interesting but irrelevant words, pictures, sounds and music (seductive details) is expected to improve comprehension and transfer over the same explanation with them added; the wiki's claims test this, in a meta-analysis and many experiments, and record effects that range from large in one research programme to small or absent elsewhere, depending on the material, its pacing and complexity, and the learner's prior knowledge, working memory and attentional control.
status review
canonical True
generated by claude/unspecified on 2026-10-15
trust unverified — no one has confirmed this page's content
sources 13 — mayer-2008, rey-2012, sundararajan-adesope-2020, harp-mayer-1998, mayer-heiser-lonn-2001, lenzner-et-al-2013, gonzalez-et-al-2019, sanchez-wiley-2006, garner-et-al-1992, noetel-et-al-2022, kalyuga-et-al-2003, kalyuga-et-al-1998 … and 1 more

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