Tight leading lets neighbouring lines interfere
Aliases: vertical crowding · colliding lines · tight leading
What it is
Once leading is small enough, the failure is not “it looks tight.” Glyphs from the line above invade the identification zone of the line below. In Latin that is g, y, j dropping into b, d, h; in Chinese it is two rows of squares almost face to face, strokes knitting into one grey field. This is interlinear crowding: not an acuity problem of a single character, but neighbouring contours stealing features from the target.
The symptoms are specific: line skips while reading, a finger or caret used as a ruler, a paragraph that becomes a dark rectangle with no countable rows. That is a different fault from “how leading should track measure.” Here the setting is already under the safety line, and the lines have started to eat each other.
Why it happens
Identifying a character needs empty space around it to cut the outline from the ground. When the vertical gap falls below the isolation the face height demands, strokes from both lines enter one crowding zone and the visual system feeds two rows of features into one object. Latin extenders are the spikes of that interference; even when “average” leading still shows a seam, collisions still happen at the deepest descenders. Chinese has no extenders, but the face already fills the em, so the default line box holds almost no spare air, and the same unitless multiplier hits interference sooner.
Crowding has a horizontal cousin: tight letter-spacing, where left and right neighbours steal features. Undersized leading is the vertical version. Together — tight tracking plus tight leading — they hurt more than either alone, because every side of the glyph is occupied.
The damage is not even. Interference is worse mid-paragraph: the first and last lines still have empty margin as a buffer; inside the block the two tracks run closest. Sampling a title therefore does not speak for the three body lines in the middle that have already fused.
Studying it
Classic crowding work flanks a target with neighbours (sides, or above and below) and measures how identification falls as spacing shrinks. A leading study does not need laboratory flankers: compress line spacing and take letter identification or oral line-skips as the observation. Independents: leading multiplier, presence of descenders, CJK versus Latin, weight. Dependents: per-letter accuracy, skip count, errors that treat a stroke from the line above as part of this line.
Tinker listed too-tight leading among the conditions that slow page reading; he was watching whole-page efficiency, not a single-character crowding curve. The two views meet: when page speed drops, collisions of extenders are usually already visible locally. Screen reading adds a wrinkle: subpixel rasterisation and light thin strokes make collisions appear earlier, because the effective ink is fuzzier than the outline on the artboard.
Where it stops holding
A title of one or two lines, even if leading is snug, has no mid-paragraph in which interference can accumulate; weight and size will cover it. Monospaced code that must keep a grid often has leading locked by the editor theme; crowding there is better fought with size and family than by stretching code to prose leading. Thin light type on a dark ground starts to crowd at a larger gap than the same setting on paper-white — bloom thickens the strokes. At the other extreme, leading so loose that lines no longer form a paragraph is a different failure; adding more seam then costs the return sweep its track, it does not “uncrowd.”
Applying it
- Specimen the setting with Latin that contains g, y, j, p and with CJK that carries many horizontal strokes (「量」, 「喜」, 「整」), not with extender-free dummy words. Collisions start on those glyphs.
- Inspect the middle of the paragraph, not the title or the first line. That is where the two tracks run closest.
- If letter-spacing is already tight, open leading before adding other emphasis; a character occupied on four sides cannot be rescued with colour.
- How to check: flatten the paragraph to greyscale and see whether stable horizontal bands can still be counted. When the bands melt into one dark patch, leading is already under the interference line.