What earns a place on the canvas
A mark belongs on the canvas only if seeing it changes what the reader understands.
#The test
A mark belongs on the canvas only if seeing it changes what the reader understands. Not if it is relevant, not if it is true, not if it was easy to add — if removing it would leave the argument equally clear, it is decoration, and decoration competes with the marks that are doing work.
#What fails it
- Restating the prose. A text mark repeating the sentence beside it adds nothing but clutter.
- Decorative accuracy. Gridlines, ticks and axis labels the argument never refers to.
show-grid,show-ticksandshow-axesexist so they can be turned off. - Everything at once. Four curves drawn together when the argument is about one of them. Reveal them with cue instead, so each arrives when the prose reaches it.
- Marks the reader cannot read. Anything unlabelled that the prose then refers to by position — the upper line — should carry a
label:.
#One pane, one job
A pane should support one claim at a time. Where a scene genuinely needs two things shown side by side — the data and the code that produced it, a function and its residuals — that is a split, not a crowded pane: a scene's detail takes two panes and arranges them with direction and ratio. A pane holding two arguments is a pane the reader has to be told how to read.
continue: #projectile — so the reader keeps whatever they explored, rather than meeting a reset canvas that looks like a new one.#When the canvas should be empty
A scene's detail is optional. A scene that states a definition, poses the question, or draws the conclusion often reads better with nothing beside it — an empty canvas is a legitimate composition, and a stale canvas left over from the previous argument is worse than none.