Chinese needs more leading than Latin at the same size
Aliases: ideographic line spacing · CJK line-height · Han leading
What it is
The same nominal size and the same line-height: 1.4 can still show rows in a Latin paragraph while a Chinese one has already gone stuffy. Chinese readers are not being fussier; the two scripts fill the em differently. Latin reading lives in a relatively short x-height band, with spare em above and below for extenders. Han face almost fills the em square. Inside the same line box, Chinese therefore inherits less air between rows, and needs more leading before the seam looks equal to the Latin one.
This is a metric difference between scripts, not the general crowding of “leading too tight,” and not the rule that leading should track measure. Even when size and measure are already right, a multiplier tuned on Latin and poured onto Han still undershoots the gap.
Why it happens
Latin metrics slice an em into baseline, x-height, ascender, descender; average ink occupies one of those bands. Han is drawn to fill a square face, with narrow shoulders. Leading computed as “size × multiplier” uses body height, not face height. At 1.4, the space from Latin x-height to the next x-height includes two shoulders plus the extra line box; at 1.4 in Han, what remains is almost only the box minus two faces that already fill it, so the seam is narrower.
Mixed setting makes the gap visible in one line: the x-band of “Design” sits low, the faces of 「设计」 sit full, vertical centres disagree, and the row spacing then looks as if Han is packed while Latin floats. A product with one leading value is usually tuned on Latin, so Chinese screens are systematically tight.
Weight interacts. Bold Han thickens an already full face and the rows gum sooner; bold Latin mainly thickens stems, the x-height band gets heavier, and the shoulders remain. Chinese is therefore more leading-sensitive once weight goes up.
Studying it
No need to invent a “psychology of Chinese.” Measure face-to-face gap, not box-to-box. Overlay a Latin specimen and a Han specimen at the same multiplier and compare: air between x-height bands versus air between Han faces. Independents: script (Chinese / Latin / mixed), nominal multiplier, face-to-em ratio of the font. Dependents: absolute face gap in pixels, rated crowding, reading speed in the Han condition.
Vertical metrics in CJK files (ideographic em-box, winAscent/winDescent, how different engines build the line box) can make “the same 1.4” differ by several pixels across browsers. Report the font file and the engine, or the difference you found is an interpretation of metrics, not a script effect.
Where it stops holding
Chinese titles, plaques, and single-line buttons do not use paragraph leading as a track; a Latin-sized shortfall is barely visible there. Interfaces that are Latin-first with occasional Han proper names can keep Latin leading and add a step only on all-Han paragraphs — often cleaner than raising everything. Vertical Chinese turns “lines” into horizontal column gap; the metric logic is similar but the axis has swapped, so a multiplier tuned for horizontal setting is not a column gap. Very open display setting (posters, covers) may push Han leading past what reading needs; that is style, not the body floor.
Applying it
- Do not let Chinese and Latin share one unitless line-height. Give Chinese body a higher multiplier, or accept the setting on face gap rather than on body multiplier.
- In mixed UI, tune leading on the Han paragraph and let short Latin words live with it — not the other way round, where a whole Chinese passage yields to one English quotation.
- Recheck shoulders when the Chinese face changes. Some designs fill the em harder; a multiplier that was just enough goes stuffy at once.
- How to check: set an English passage and a Chinese passage side by side at the same multiplier. If the Chinese turns into a grey slab while the English still shows countable rows, the multiplier is serving Latin metrics only.