Multimedia learning: how words and pictures combine (and collide)

Evidence grade: STRONG for the core principles. The multimedia-learning research program spans dozens of controlled experiments with replicated effects (multimedia, coherence, redundancy, signaling, modality). Boundary conditions are real and mapped — most effects are strongest for novices and complex material — and the program's lab-heavy origins argue modest humility about exact classroom effect sizes.

Every lesson slide, explainer video, worksheet diagram and app screen makes multimedia decisions — usually by instinct, decoration budget, or template. Richard Mayer's research program spent three decades testing those decisions experimentally, and the results form the closest thing instructional design has to building codes. (Presentation literacy applies them to the slideshow genre; this page is the underlying science.)

The model in one paragraph

Three assumptions, each independently supported: processing runs on two channels (auditory/verbal and visual/pictorial — Paivio's dual coding, Baddeley's working-memory architecture); each channel has sharply limited capacity (the bottleneck); and learning is active integration — selecting, organizing, and connecting words and pictures into one model. Every principle below is a consequence: use both channels, overload neither, spend all capacity on integration rather than on coping with the presentation.

The principles with the strongest support

Multimedia principle: words + relevant pictures beat words alone — the baseline case for illustrating at all.
Coherence (the anti-decoration principle): adding interesting-but-irrelevant material — seductive stories, decorative art, background music — measurably reduces learning. The most violated principle in children's materials, where "engaging" is routinely purchased with exactly the additions the data indicts.
Redundancy: narration + identical on-screen text is worse than narration alone — the channels collide instead of dividing the work.
Modality: complex graphics explained by speech beat the same graphics explained by on-screen text — eyes on the diagram, ears on the words, both channels earning their keep.
Signaling: cues that highlight organization (highlighting the relevant part as it's discussed) help — attention guidance is cheap capacity insurance.
Contiguity: words placed at their referents (labels on parts) beat separated blocks — split attention, spatial edition.
Segmenting & pacing: learner-paced segments beat continuous streams — the pause button is an instructional feature.
Personalization: conversational style ("your golem," "you'll notice") modestly but reliably beats formal style — social framing recruits effort.

The boundary that matters: most effects are strongest for novices with complex material and shrink (occasionally reverse) for experts — expertise reversal again. Multimedia rules are novice-protection rules; that's the primary classroom's exact situation.

What the evidence doesn't say

It doesn't say decoration is forbidden everywhere — it prices decoration inside instructional moments; the celebration screen after the task is a different budget.
It doesn't say text is bad — it says duplicated and displaced text is; labels, key terms, and text-as-the-content follow different rules.
It doesn't fully transfer effect sizes from lab videos to classrooms — the directions are trustworthy; treat exact magnitudes with the usual humility.

In the classroom

Audit materials with the coherence knife: for each decoration, "does this help the point or compete with it?" — the honest answer removes more than expected.
Explain diagrams aloud, not in paragraphs beside them — modality is the most usable principle in daily teaching.
Put words on the thing: label parts on the diagram, instructions beside the field (the same rule our tasks follow).
Give videos a pause culture — segment, stop, retrieve (the loop), continue.

How Wiz Kids applies this

The task surfaces are built to these codes: instructions live at their referents (contiguity), interfaces carry one job's worth of elements (coherence — the fiction's charm lives in language, not in decorative screen clutter competing with tasks), narration via the speak button gives the modality option, everything is learner-paced with clean segments, and the writing is relentlessly conversational (personalization — "your golem," always). The celebration confetti waits until the task is done: decoration on its own budget.

References

Mayer, R. E. (2021). Multimedia Learning (3rd ed.) — the program and principles.
Mayer, R. E. (Ed.) (2014). The Cambridge Handbook of Multimedia Learning — principle-by-principle evidence reviews.
Paivio, A. (1986). Mental Representations: A Dual Coding Approach.
Baddeley, A. (1992). Working memory. Science, 255.
Kalyuga, S., et al. (2003). The expertise reversal effect. Educational Psychologist, 38(1) — the boundary condition.

© Glu IO Pty. Ltd. — Wiz Kids (wiz.kids). Link freely; republication requires permission — see terms. Found an error in our reading of the research? We correct fast: tell any teacher piloting Wiz Kids.