A hapax legomenon is a word that appears exactly ONCE in a text — ‘said once’. Astonishingly, they make up a large share of any vocabulary (roughly half in many corpora), and the fraction shrinks as a text lengthens. They’re where an author’s rarest choices live — and a headache for translators of ancient texts. Slide the length and watch the once-words thin out.
Hapax legomena are types with frequency exactly 1. Their count, and their fraction of the vocabulary, is a richness signal: in natural text the hapax fraction is high (often ~40–60% of TYPES) and DECLINES as more tokens arrive and earlier once-words recur. They anchor the tail of the Zipf/frequency spectrum and are diagnostically important in stylometry and in reading damaged or ancient corpora. A fail-loud self-check throws unless a short sample’s hapax fraction is at least as high as a longer sample’s. ◆ real computational stylistics, node-verified.
The stream is illustrative; the once-word count and its decline with length are exact. A high hapax rate flags rare vocabulary, not necessarily ‘good’ writing — noise and typos are hapaxes too.