Redundant on-screen text impairs learning when it competes with a visualization, though written text duplicating narration alone can help¶

Claim · All claims
Evidence · 1 study · review r? · q3 · n=63 studies (44 in the narration+on-screen-text "scenario 4")

Adding on-screen text that duplicates narration to a narrated animation or diagram, or adding redundant text to an already comprehensible graphic, is thought to overload working memory and has typically lowered learning outcomes relative to narration-plus-graphics alone. Without a visualization, written text added to narration is the opposite case: the review below reports positive effects there.

Subclaims¶

q3 i? A literature review of 63 studies finds the results of adding written text which duplicates the narration of a narrated visualization highly heterogeneous: among the 33 such studies with redundancy only between narration and written text, some report negative effects (among them classic experiments such as Kalyuga et al. 1999 and Mayer et al. 2001), others positive effects and others no difference. Moderators shift the result: older adults benefited from identical written text where younger adults showed no difference, and abridged or non-identical text was often less harmful than identical text. Adding written text to narration with no visualization is the opposite case: the review reports positive effects there. → Trypke et al. 2023

q3 i? The review classifies the harm as working-memory-channel redundancy between visualizations and written text, and reports positive effects of content redundancy that depend on learners' prior knowledge. → Trypke et al. 2023

Evidence¶

Trypke et al. 2023¶

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

q3 · literature review (63 studies) · i? · no pooled effect size reported · n=63 studies (44 in the narration+on-screen-text "scenario 4") · review · r?

This review classifies redundancy effects across 63 multimedia-learning experiments into "content redundancy" (duplicated information, independent of channel) and "working-memory channel redundancy" (two sources competing for the same processing channel). For the case closest to the classic redundancy principle — narration duplicated by identical on-screen text — the authors report that among the 44 studies adding written text to narrated visualizations, several classic experiments (e.g., Kalyuga et al. 1999, Mayer et al. 2001, Yue et al. 2013) found that "the animation, narration, and identical written text group performed worse on retention and transfer tests," and note the finding "aligns with the findings of the meta-analysis on the verbal redundancy effect, which shows that narrations accompanied by identical written text impaired learning (e.g., Adesope and Nesbit, 2012)." The review also identifies moderators — degree of overlap (abridged/keyword text is less harmful than verbatim duplication), learner age (older adults benefited from identical written text, younger adults showed no difference), prior knowledge, and pacing (in Bernal 2014, a narration-plus-written-text study without a visualization, learner pacing helped overall but redundant and non-redundant conditions did not differ) — that shift the effect. Among the 33 scenario-4 studies with redundancy only between narration and written text, some found negative, some positive and some no effects, which is why the overall pattern is reported as heterogeneous rather than as one pooled effect. The abstract's summary is scenario by scenario: positive effects of content redundancy (depending on prior knowledge), negative effects of working-memory-channel redundancy between visualizations and written text, and positive effects of working-memory-channel redundancy between narration and written text, so the harm is specific to text competing with a visualization.

Studies bearing on this claim¶

When a picture or animation is narrated, does adding on-screen text that repeats the narration hurt learning? This table gathers the studies behind this claim and the 4 narrower claims that bear on it: 5 distinct studies (2 review, 2 causal, 1 quant-synthesis), q2–q3, 1 of 5 with an effect size coded. One row per study, codes as each entry records them; no effect is averaged and no verdict is drawn. Built by scripts/synthesize_claim.py from eval/synthesis/; a row that looks wrong is fixed on the claim it comes from.

Discussion¶

Mechanism. Within cognitive load theory, redundant on-screen text forces learners to split visual attention between the graphic and the text while also trying to reconcile it with the narration — a classic split-attention and redundancy burden [-S]. Learners cannot read and listen to identical text in parallel; reading speed outpaces speech, so the text competes for the same visual channel as the graphic. This parallels the coherence principle: material that adds no new information but adds processing cost hurts learning [-M].

Boundary conditions. The redundancy effect is strongest for novices and for fast-paced, learner-paced-unfriendly multimedia [-M]. It weakens or reverses when: (a) narration is absent (text alone with a graphic is fine — the problem is duplication, not text per se) [~S]; (b) learners are fluent readers with high prior knowledge, who can skim text and ignore narration (an instance of the expertise reversal pattern — see the expertise reversal effect) [~M]; (c) text is short labels or signs rather than full sentences [~M]; or (d) the learner controls pacing and can integrate the channels deliberately [~W].

Design implication. Prefer narration plus a clean graphic; put essential words in the audio channel, and reserve on-screen text for terms, definitions, or content that must be re-read. This complements cognitive load reduction and cognitive load management guidance, and pairs with chunking to keep each channel's load within working-memory limits.

Open questions. Most evidence comes from short, lab-based multimedia lessons; effects on longer, self-paced online courses — where learners routinely expect captions — are less settled [~W], and accessibility needs (deaf and hard-of-hearing learners) can make redundant text a necessary accommodation rather than a redundancy (see accommodations). Designers of captioned video should treat captions as an accessibility layer, not assume they are neutral or beneficial for all learners.

Merged from “Redundancy Effect Impairs Learning” (redundancy-effect-impairs-learning): The redundancy effect is one of the classic effects predicted by Cognitive Load Theory. The proposed mechanism is that when identical information is presented in two formats, working memory resources are consumed holding one representation while the learner maps it onto the other, and the learner may be forced to process both even though one is sufficient [-M]. This is distinct from the coherence principle, where the harm comes from irrelevant rather than duplicated material — redundant material is relevant, just unnecessarily repeated. Both effects operate through the same underlying constraint that cognitive overload degrades learning [-M].

The effect is strongly moderated by learner expertise [~S]. For novices, redundant duplication of essential information can impair learning; for more advanced learners, the same duplication may be harmless or even helpful, consistent with the expertise reversal effect [~S]. Designers should therefore treat redundancy as a novice-focused constraint and fade redundant support as competence develops — the same fading logic that governs when worked examples can become counterproductive for advanced learners.

Boundary conditions worth noting: redundancy in the strict CLT sense refers to same-information duplication. Complementary material that adds new information (e.g., a diagram with genuinely explanatory labels, or narration that elaborates rather than reads the text) is not redundant and does not fall under this claim. Similarly, signposting or structural cues that help learners navigate material may look like redundancy but serve a different function — see clear structure improves learning [+M].

Practical implications for multimedia design: avoid on-screen text that verbatim duplicates narration [-M]; prefer graphics with concise integrated labels over graphics plus redundant audio [-M]; and when accessibility requires captions, consider whether they duplicate or complement the audio track. Where duplication is unavoidable (e.g., regulatory or accessibility requirements), giving learners control over pacing may reduce the coordination cost, though this interaction is not yet well quantified [~W]. Consolidating information so it need not be duplicated at all — for example through chunking — is a complementary mitigation [+M].

Open questions: the one review recorded above reports no pooled effect size, so the magnitude of the effect across domains and media, and how it interacts with learner control (e.g., self-paced versus system-paced multimedia), are not established on this page; primary evidence would be needed before specific design thresholds can be recommended.

Merged from “Redundancy Harms Learning” (redundancy-harms-learning): Mechanism. Under cognitive load theory, redundancy is one of the classic sources of extraneous load: when identical content is duplicated across channels, learners must reconcile the copies, consuming working-memory resources that could otherwise go to schema construction [~M]. The most-studied case is simultaneous narration of identical on-screen text, which forces visual attention and auditory processing onto the same words rather than letting visuals carry the diagram and narration carry the explanation. This is a core multimedia-design principle: duplication across channels degrades learning relative to narration-plus-graphics [~M].

Boundary conditions. The redundancy effect is not universal. It is primarily a novice phenomenon; as expertise grows, redundant formats can become helpful rather than harmful — the same expertise-reversal pattern documented for worked examples and formalized in the expertise reversal effect [~M]. Redundant on-screen text may also help when learners cannot process audio (hearing impairment, noisy environments, non-native listeners), when text serves as a navigational or reference aid rather than a duplicate of narration, or when pacing is fully learner-controlled [~W]. The effect applies to identical duplication; complementary information across channels (e.g., narration plus a non-redundant diagram) is beneficial, not harmful — see the coherence principle for the related harm from irrelevant material.

Design implication. Prefer a single well-chosen presentation of each idea. When using narration with visuals, keep on-screen text minimal and non-duplicative rather than a verbatim transcript. Where accessibility or environment demands captions or transcripts, treat them as an accommodation (see Accommodations) rather than a default for all learners. This pairs naturally with cognitive load management and chunking as part of a broader extraneous-load audit of instructional materials.

Open questions. The one review recorded above reports no pooled effect size, and finds the effect moderated by text overlap, learner age and pacing. The interaction between redundancy and learner-controlled pacing, and the threshold of expertise at which redundancy flips from harmful to helpful, remain under-specified.

Merged from “Redundancy Hurts Multimedia Learning” (redundancy-hurts-multimedia-learning): The redundancy effect is one of the core effects predicted by Cognitive Load Theory. When learners read text and listen to narration that say the same thing, they must coordinate the two streams and reconcile them, consuming working-memory capacity that would otherwise go to schema construction. This interacts with the modality effect: narration plus graphics can outperform text plus graphics, but only when the narration does not duplicate on-screen text.

Key boundary conditions:

• Learner expertise. Redundant material can become beneficial rather than harmful for advanced learners, consistent with the expertise reversal effect — experts may use redundant text as a low-effort verification aid.
• Nature of the redundancy. Harm is strongest for verbatim duplication of narration as on-screen text; a brief caption, heading, or label that complements rather than repeats narration is not redundancy in this sense and may support learning.
• Access conditions. Redundant on-screen text may help learners who cannot hear the narration (hearing impairment, noisy environments, non-native speakers), so accessibility needs can override the effect — a case where cognitive load management trades off against other design goals.
• Pacing. Learner-controlled environments may soften the harm, since learners can pause or ignore one stream; the effect is expected to be strongest under system-paced, continuous presentation.

The page records one review (Trypke et al. 2023); the classic primary experiments and the meta-analytic literature on the redundancy principle are not yet recorded as entries here.

Merged from “Redundancy Principle Hurts Learning” (redundancy-principle-hurts-learning): Mechanism. Within Cognitive Load Theory, redundancy is one of the classic sources of extraneous load: when learners must cross-reference two channels presenting identical content, working memory resources are spent on coordination rather than schema construction. This connects to the broader claim that cognitive overload degrades learning and that irrelevant or incoherent material hurts learning. The effect is distinct from the modality principle: narration plus graphics is generally better than text plus graphics, but adding verbatim on-screen text to narration pushes learners back into redundant processing.

Boundary conditions. The redundancy effect is not universal. Redundant material is less harmful — and sometimes helpful — when learners can control pacing (allowing them to ignore one channel), when the redundant format is not literally identical (e.g., a concise label versus a full restatement), or when learners are novices who might otherwise misinterpret sparse material. As expertise grows, the effect interacts with the expertise reversal pattern described in expertise reversal effect: supports that help novices become redundant for advanced learners. Designers should therefore ask whether each channel adds new information or merely repeats existing information, and remove the repetition when it does not.

Design implications. The most common practical instance is narrated animation or video with on-screen text that verbatim duplicates the narration; the redundancy principle recommends narration plus graphics without the duplicated text, reserving on-screen text for cases where audio is unavailable, the content is technical terminology, or learners are non-native listeners. Concise labels, captions that add information, and signals that direct attention are not redundancy in this sense — they add or restructure information rather than repeating it. This places redundancy alongside coherence and chunking as tools for cognitive load management rather than a blanket prohibition on multiple representations: dual coding of complementary verbal and visual information can still help, provided the two channels are not redundant.

Open questions. The one review recorded above reports no pooled effect size, and the meta-analysis it cites (Adesope & Nesbit 2012) could not be read here, so the size of the effect is not yet established on this page. Key open questions include how effect sizes vary across media (narration vs. text), age groups, and domains, and how much pacing control is needed to neutralize the effect.

Merged from “Redundancy Principle On Screen Text Hurts Learning” (redundancy-principle-on-screen-text-hurts-learning): Mechanism. The redundancy effect is explained within Cognitive Load Theory: when identical verbal information is presented as both narration and on-screen text, learners must coordinate two channels presenting the same content, and written text competes with visual material (diagrams, animations) for the limited visual channel. Narration alone leaves the visual channel free for the graphics. This is closely related to the Coherence Principle — irrelevant material hurts learning and to Cognitive overload degrades learning.

Boundary conditions. The effect is generally strongest when graphics are complex or presented at a fast pace, and when text is verbatim rather than supplementary. Verbatim on-screen text is the harmful case; concise text that adds information not in the narration is not redundancy and may help. The effect also weakens or reverses when learners control pacing (allowing them to read then listen), when there is no pictorial information to compete for visual attention, when learners are hearing-impaired, when the material is in the learner's second language, or when only brief key phrases (labels, headings) are shown rather than full sentences.

Constraints on application. Designers should not treat "never show text" as a rule. Captioning is required for accessibility and benefits second-language learners, so the redundancy cost must be weighed against these needs [~M]. Redundant on-screen text is most harmful in system-paced, graphics-heavy lessons; in self-paced e-learning, learners may simply ignore the text, and the measured harm shrinks or disappears [~W]. Showing only brief labels or key phrases alongside narration avoids the effect and can support reference and note-taking [~M].

Open questions. Most evidence comes from short multimedia lessons in laboratory or controlled settings; the durability of the effect in self-paced, real-world e-learning — where learners can simply ignore redundant text — is less well established. No primary study is recorded as an entry on this page yet.

Merged from “Redundant On Screen Text Hurts Learning” (redundant-on-screen-text-hurts-learning): Mechanism. The redundancy effect is usually explained within Cognitive Load Theory: visual and verbal channels have limited capacity, and when identical text duplicates narration, learners must visually scan the text while listening, splitting attention between two presentations of the same content rather than integrating words with the picture. This connects to the broader claim that cognitive overload degrades learning and that irrelevant or extraneous material hurts learning. Note the contrast with dual coding theory: dual coding predicts benefits when visual and verbal information are complementary (e.g., a diagram plus narration), whereas redundancy involves identical verbal information in two channels, which adds processing without adding information.

Boundary conditions. The effect is expected to be strongest when text and narration are presented simultaneously and verbatim, when the material is fast-paced, and when learners are novices. Redundant text may be less harmful — or even helpful — when learners can control pacing, when the text is not verbatim (e.g., labels or keywords on a diagram), when narration is unavailable or hard to hear, or for learners who benefit from reading support (e.g., second-language learners or those with hearing difficulties). The review recorded above reports text overlap, learner age, prior knowledge and pacing as moderators; the others are plausible from theory but have no entry on this page.

Design implications. Where narration is used, avoid displaying the full transcript on screen at the same time; if text is needed for accessibility, consider making it available on demand rather than simultaneously. Reserve on-screen text for content that is not spoken — labels, keywords, and captions on diagrams — so that the visual channel adds information rather than duplicating it. This pairs naturally with chunking and other cognitive load management strategies.

Open questions. The one review recorded above reports no pooled effect size; the effect size, the role of pacing and learner control, and whether short keyword-level redundancy differs from full-sentence duplication are not yet established on this page.

Related Claims¶

• Irrelevant material hurts learning (coherence principle). — same working-memory logic: extra material that adds no value imposes cost
• Cognitive overload degrades learning. — the overload mechanism redundancy exploits
• Chunking reduces working memory load. — complementary strategy for managing limited working memory
• Worked examples can become redundant or counterproductive for advanced learners. — expertise reversal: redundancy effects depend on learner expertise
• Expertise Reversal Guidance Hurts Experts — related
• Presenting words as spoken narration rather than on-screen text alongside graphics improves learning — related
• 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 — a narrower finding that bears on this claim
• Bimodal captioned input improved L2 listening skills, generalizing to unfamiliar sentences and speakers (attributed to Charles & Trenkic, 2015) — related
• Common language teaching methods such as oral drills, memorization, and fast-paced competitive activities disadvantage older learners — related
• In one middle-school quasi-experiment, adding written text to spoken narration makes no significant difference to achievement with either abstract or concrete animation — a narrower finding that bears on this claim
• A narrative review cites one study in which simultaneous notes, worked example and explanation aided understanding, and another in which temporally contiguous instruction lowered perceived cognitive load — related
• Redundancy principle: animation plus narration beats animation, narration, and on-screen text for transfer — a narrower finding that bears on this claim
• Expertise reversal effect — the moderator that makes redundancy harmful for novices but not experts
• Cognitive Load Theory — the parent theory predicting the effect
• Cognitive load reduction improves learning. — the general principle this claim instantiates
• Clear structure improves learning. — a positive-design counterpart: one well-organized presentation instead of duplicated ones
• Dual coding theory. — explains why complementary (not redundant) visual–verbal pairing helps learning
• Laptop note-takers transcribed more verbatim and did worse on conceptual questions than longhand note-takers in one set of experiments, but a direct replication found no consistent difference in test performance — related
• Multimedia Principle Improves Learning — related
• Redundant text in the diagram did not affect posttest accuracy or difficulty ratings — related
• Interesting but irrelevant details can impair comprehension and transfer, with effects that vary with the material and the learner — related
• Despite ignorance of CLT, surveyed teachers report using some of its principles when designing instructions — related
• Avoiding double load on the visual channel improves visual knowledge (channel-overload effect) — related
• Multiple representations and explanations of math concepts aid student understanding, but faculty lack time to source or develop them — related
• Design principle effects are larger in system-paced and in complex media, and learner prior knowledge was not a significant moderator (p = 0.14) — a broader claim this one bears on
• Working memory includes partially independent subcomponents for auditory/verbal and visual/2- or 3-dimensional information — related
• Reduction of cognitive load can sometimes impair learning rather than enhance it — related
• APA presentations tend to overload the visual channel, especially with high element interactivity content — related
• Redundant speech-gesture cooperation improves recall and social perception more than complementarity — related
• Oral non-interactive teaching can enhance learning more than written teaching for complex materials — related
• 2D pedagogical agents tended to be more effective than 3D agents (g+ = 0.38 vs. g+ = 0.11), as predicted by the redundancy effect and the coherence principle — related
• Synced auditory and visual information in narrated presentations produces better learning than single-channel or out-of-sync presentation — related
• Embodied action can harm learning when it adds motor demands, redundant cues, or irrelevant activity without improving understanding — related