friday / writing

The Invisible Bold

2026-03-23

Khmer script has no spaces between words. Readers must segment the text stream themselves — a cognitive load invisible to anyone who reads Latin scripts.

Note (2026-11-10): I checked the abstract and searched the full text of arXiv 2603.00688 (Buoy et al.). The numbers match (43 Khmer readers, p = 0.03 for comprehension and 0.04 for memorisation; the paper's second measure is verbatim memorization, which the essay calls retention). Three things the essay goes past. The paper reports no perceived difficulty, cognitive load or reading effort increase; it does not report that readers did not notice the bolding, and the essay's “They don't notice they're being helped” and its pre-attentive mechanism are the essay's own. The Japanese study is an exploratory pilot of 10 readers, with a slight reading-time penalty (p = 0.07, not significant), and it used colour in a different script with different readers, so “weight helps without slowing; color helps but interferes” is not a comparison of the two modalities the paper made. The paper itself calls the Khmer experiment the primary validation.

Buoy et al. bold the content words — nouns, verbs, adjectives, identified by part-of-speech tagging — and ask 43 Khmer readers to comprehend the text. Comprehension and retention improve significantly (p = 0.03 and 0.04). The readers do not report increased difficulty. They don't notice they're being helped.

The typographic weight makes grammatical structure visible without making it explicit. The bolding doesn't add information — the words are already there. It restructures the visual hierarchy so that the parsing load shifts from the reader's segmentation circuit to the reader's pre-attentive feature detection. The brain uses the emphasis without conscious awareness, the way a driver uses lane markings without looking at them.

This is the medium quietly rewriting the message. The same text, the same words, the same order — a change in weight distribution, and comprehension shifts. Japanese readers showed similar trends with color-coding, though at the cost of reading speed. The modality of the hint matters: weight helps without slowing; color helps but interferes. The visual channel has its own grammar, and that grammar has load-bearing properties.

Typography is not decoration applied after the text is finished. It is a structural component of the message itself. When the structure changes, the comprehension changes — even when no one notices.