Hapax legomenon
A hapax legomenon is a word that occurs exactly once within a given context: a single text, an author's complete works, or a language's entire written record. From Greek, 'said once'; plural: hapax legomena.
The numbers
- Share of distinct words in large corpora that are hapax legomena (per Zipf's law): about 40% to 60% (Wikipedia: Hapax legomenon)
- Share of distinct words that are dis legomena (occur twice) in large corpora: about 10% to 15% (Wikipedia: Hapax legomenon)
- Hapax legomena in the Brown Corpus of American English: roughly half of its 50,000 distinct words (Wikipedia: Hapax legomenon)
- Distinct words in Moby-Dick that occur only once: about 44% (Wikipedia: Hapax legomenon)
Definition
A hapax legomenon (Greek for 'said once'; plural hapax legomena, often just 'hapaxes') is a word or expression that occurs exactly once within a defined context. The context matters: a word can be a hapax within one text, within an author's complete works, or within the entire written record of a language, and the claim means something different in each case.
The family extends upwards: a dis legomenon occurs twice, a tris legomenon three times, and a tetrakis legomenon four times.
Hapaxes are the norm, not the exception
Intuition says a word used only once must be unusual. Corpus statistics say the opposite. In line with Zipf's law, roughly 40% to 60% of the distinct words in a large corpus are hapax legomena, and a further 10% to 15% are dis legomena. In the Brown Corpus of American English, about half of the roughly 50,000 distinct words occur exactly once, and in Moby-Dick about 44% of the distinct words appear only once.
What carries information is therefore not the existence of hapaxes but their rate: the proportion of a text's vocabulary used only once, compared against samples of similar length. A high hapax rate signals a writer reaching for fresh words rather than recycling a small working vocabulary.
Use in authorship analysis, and its limits
Hapax counting has a long history in authorship studies; P. N. Harrison popularised its use on Biblical texts. As a standalone attribution method it has faded, because hapax frequency is affected by text length, topic, audience and date of composition, so a difference between two texts can have mundane explanations.
Modern stylometry treats the hapax rate as one lexical-richness signal among many rather than a fingerprint on its own. In computational language processing, hapaxes are often simply discarded from models, since a single occurrence provides little statistical value.
Worked example
Take this 15-word passage: 'I know what I know, and what I know is that nothing here changes quickly.'
The passage has 15 tokens and 10 types. Frequencies: 'I' three times, 'know' three times, 'what' twice, and seven words exactly once each: and, is, that, nothing, here, changes, quickly. Those seven are the hapax legomena, and 'what' is a dis legomenon. Hapax rate: 7 / 10 = 70% of types, or 7 / 15 = 47% of tokens.
Even a deliberately repetitive sentence is mostly hapaxes when counted by type. Always state the basis, per type or per token, when quoting a hapax rate; the two figures differ widely.
Sources
More terms
- Burstiness
- Perplexity
- Sentence Length Variance
- Flesch Reading Ease
- Type-token ratio (TTR)
- Function words
- Lexical density vs lexical diversity
- Stylometric fingerprint
- Idiolect
- Voice vs tone vs style
- Passive voice
ScriptGrain · Why ScriptGrain · Sample voice profiles · The Journal · Support