Draw relations

Claim: a relation, a set, a quantity or a process is shown as a figure placed with the words that explain it, not as words alone.

• Grade: measured.
• Evidence: learning science (strand 5), typography and layout (strand 6).
• Check: figures-present (judgment): a long text with no figure or table is asked about. A figure on a page people read is drawn, not typed in characters (typed-figure, a warning). Its SVG is rendered from the source beside it (figure-source, exact).
• Reader styles: strengthened by visual-relational; replaced by an edge list or table for an LLM reader.

What it asks

• A relation between things is a graph; a set or a partition is regions; a quantity is length or position; a process is frames the reader steps through.
• A table when the reader looks up symbols; a graph when the reader compares shapes or trends (Vessey 1991).
• Position and length before angle, area or colour, which are judged less accurately (Cleveland 1984).
• No decorative pictures: interesting but irrelevant material lowers learning, g = -0.33, with strongly varying results (Sundararajan 2020, a meta-analysis); decorative pictures are one kind of it.

Why

A diagram puts related items next to each other, so the reader neither searches nor infers what the page could show (Larkin & Simon 1987). Words with pictures beat words alone across the multimedia literature (Mayer 2020). Building a concept map gave g = 0.82 over activities without one, and studying a map gave 0.37 over them (Nesbit 2006); in the North American studies alone, building gave 0.47.

Limits

Animation beats static pictures only modestly, d = 0.37 overall and more for procedural-motor content (Höffler 2007); it helps when the change itself is what is learned (Ploetzner 2020). Translating between two representations is itself hard and needs support (Ainsworth 2006).