What is a hapax legomenon
A hapax legomenon is a word that appears only once in a given body of text.
That body of text can consist of one book or one author's complete works. It can also encompass everything written in a language. The term is Greek for "said once." The plural is hapax legomena, and scholars shorten both forms to hapax.
Hapax legomenon definition
A hapax legomenon only counts as one relative to a defined corpus, so the same word can be a hapax in one text and common in another. "Matrimonial" occurs once in Moby Dick, yet any law library prints it every day. That's why careful writers always name the corpus they mean, such as "a hapax in Shakespeare." A word that appears exactly twice has its own name, a dis legomenon. Before you call any word unique, check how the corpus was counted, because spelling variants and inflected forms change the tally.
Famous examples
The best-known hapax legomenon examples come from old texts. In those texts, a word that appears only once can leave translators guessing at what it means, because they have no second passage to compare it with.
Gopher wood in the Bible
Genesis 6:14 tells Noah to build the ark of "gopher wood," and "gopher" appears nowhere else in the Hebrew Bible. Translators have proposed cypress or cedar, but no other passage settles the question. Biblical hapaxes are common. The Cateclesia Institute counts about 3,414 once-used words in the Bible, close to a quarter of its vocabulary.
Honorificabilitudinitatibus in Shakespeare
In Act 5 of Love's Labour's Lost, the clown Costard says this 27-letter word, and it shows up nowhere else in Shakespeare's plays. The word is a mock-Latin form meaning roughly "the state of being able to achieve honours." A writer can also use a hapax to show off.
Hapax ratio in text analysis
The hapax ratio is the share of a text's distinct words that occur only once, and analysts use it as a quick measure of how wide a writer's vocabulary is. To get it, divide the number of hapaxes by the number of distinct words. In Homer's Odyssey, 1,606 of 4,856 words appear once, which gives a ratio of about 33%. In Moby Dick, 44% of distinct words occur only once.
The ratio changes with genre, so compare texts of a similar kind. A technical manual repeats its terms on purpose, and a low ratio there tells you nothing bad about the writing.
Why AI has fewer
AI writing contains fewer one-off words because language models choose likely words, and rare words are unlikely by definition. A model predicts each next word from probabilities it learned in training. Common sampling settings also cut off the long tail where hapaxes sit.
A 2023 study of news text by researchers at the Universidade da Coruña put this to the test. Human articles scored highest on the Measure of Textual Lexical Diversity (MTLD), at 96.51 for human writers, while Falcon 7B scored 57.37. A 2024 Elsevier paper found that ChatGPT-4 matched human lexical diversity, and in some tests it went higher. So check the model and the prompt before you draw conclusions.
How Snoika uses it
Snoika creates and publishes human-like content in more than 135 languages, with human review on top. The hapax idea describes part of what a reviewer looks for. When a draft keeps drawing on the same small set of safe words, it reads as flat, and it leaves out the specific names and terms the subject needs. A reviewer who spots that pattern can fix it by adding the concrete vocabulary a person who knows the field would use.
Related terms
Look up more AI search and SEO terms in our glossary.