Overshort measure chops reading with constant wraps
Aliases: narrow column · frequent wrapping · too-short line
What it is
When the column is too narrow the failure comes from the other end: the line breaks before a sense-group has finished. Latin parks the, of, and at the start of the next line, or splits a word; Chinese splits four-character compounds, proper names, and number–unit pairs. The eye is forced into constant returns, and each return cuts a phrase that was still being assembled. What the reader feels is stutter, not a “clean column.”
Too short and too long are ends of one ruler, but the mechanisms are not mirrors. Too long is one badly aimed return; too short is an explosion of returns plus chopped linguistic units. Stretching body copy across a wide screen, and slicing a phone column into two rows of seven or eight characters, look like opposite luxuries and both halt the read.
Why it happens
Reading plans chunks — the next few words or characters — not isolated letters. The line end is a hard boundary. When it falls inside a chunk, the structure just built in working memory is taken apart and rebuilt at the next start. Hyphenation makes that wound visible in Latin; even without hyphenation, short lines orphan function words from content words and scatter syntactic cues. Chinese has no spaces; a greedy width wrap will park particles and broken names at line start, so the rhythm fragments harder than in Latin.
The return itself costs time. Short lines mean dense returns; time that should identify characters is spent shuttling. The speed curve is therefore not “shorter is more precise”: past a point, return overhead outruns the aiming advantage of a short line, and net efficiency falls. That is a different ledger from the long line, which is hard to aim but rare.
Justification on a short measure turns the shortage into a disaster: with only a handful of words to flex, the engine must open word spaces into holes or distort letter-spacing. The rag of a short line is already jumpy; justifying it is stretching a cloth that has no fabric left.
Studying it
Line-length experiments have to treat “too short” and “too long” as two regions, with the often-quoted comfort band in between. In Dyson’s screen-reading work, preference tends to sit at moderate lengths, while speed sometimes still rises on longer lines — on the short side, speed and preference usually fall together, because the interruption is noticeable. Dependents besides speed: count of phrases cut by the line end, regressions, unnatural pauses in oral reading.
Independents: words per line or Chinese characters per line, hyphenation on or off, number of columns. For Chinese, note whether wrapping already avoids punctuation and names (line-start/end prohibition is a separate group; here it is only a constraint that short measure trips more often).
Measuring measure with button labels or titles is pointless; they are short by nature. The copy has to be continuous prose, long enough to accumulate many wraps.
Where it stops holding
Navigation, tag clouds, secondary phone columns, and bilingual facing columns sometimes have to be short. Their reading unit was never a paragraph; the cost of shortness shifts from “breaking prose” to “each label still has to show in full.” Verse and lyrics break on metre, so short is structure. Code has its own width tradition; shorter than prose is not a fail. The real leak is reusing a sidebar width on article body. A phone held in portrait has a physical floor on column width, so shortness is almost unavoidable — then drop size or accept denser wraps, and use leading to keep the tracks apart, rather than justifying for neatness.
Applying it
- Do not cut body copy into decorative columns that hold only a few words. Sidebars may be narrow; the body column should let a typical phrase finish on one line.
- Turn justification off on short columns; there are not enough word spaces to adjust.
- In Chinese, inspect line starts for piled particles and split names. If they are common, widening beats editing the wrap dictionary.
- How to check: mark every line end in a paragraph and count how many fall inside a phrase or a name. A high ratio means the column is shorter than the language unit.