Why Does One Word Mean Two Things? Two Words That Collided

Several different situations get described with overlapping terms and separating them is the useful part. Two words may sound alike and be spelled differently, which is one situation. Two may be spelled alike and sound different, which is another. Two may be alike in both, which is the strictest case. And a single word may have several related senses that developed from one another, which is a completely different phenomenon despite looking similar. The first three involve distinct words that happen to coincide, while the last involves one word that has stretched, and the test is whether the meanings are historically connected.

Distinct words come to look or sound the same by several routes:

• •Borrowing from different languages that happen to supply similar forms
• •Sound changes that erase a distinction two words previously had
• •Loss of an ending that formerly distinguished them
• •Spelling reform or standardisation merging previously different forms
• •Shortening of longer words to the same result
• •Pure coincidence, which accounts for a great many

Ambiguity ought to be a problem and mostly is not, because context does almost all the work. The words surrounding an ambiguous form usually eliminate every reading but one before a listener consciously notices, and experiments measuring how quickly people understand such words find no measurable cost in ordinary sentences. Where genuine confusion would arise, languages tend to resolve it, with one of the colliding words falling out of use, acquiring a modifier, or being replaced by a synonym, and historical linguistics documents that happening repeatedly. The classic example is a word for a rooster that was displaced across much of the English-speaking world by a word previously meaning a young male bird, for reasons everybody understands.

Overlapping words are a resource as well as an obstacle, and the uses are worth naming. Puns exploit the overlap directly for humour, and the groan they produce is a recognition that the listener was briefly led along the wrong reading. Riddles depend on presenting a clue that fits two senses. Poetry uses the overlap to make a line mean two things at once without stating either, which is one of the mechanisms by which poetry does more work per word than prose. Advertising slogans use it constantly. Crossword clues in the British tradition are built almost entirely on it, since a cryptic clue reads as one thing and parses as another. And a great many jokes have no other mechanism.

The overlaps that matter are the ones machines encounter, since a computer has no context of the kind a listener uses. Search systems return results for the wrong sense of a query constantly. Automatic translation must choose a sense before producing a word in the target language, and choosing wrongly produces the errors such systems are mocked for. Speech recognition must select between identically sounding forms from context alone. Text to speech must choose a pronunciation for words spelled alike and sounded differently, which is why such systems mispronounce names and technical terms. Resolving which sense is intended is a recognised problem in language technology with a long history and no complete solution.

Words alike in sound, in spelling or in both usually came from separate origins and collided through borrowing, sound change or coincidence, which is different from one word stretching into several related senses. Context resolves the ambiguity so completely that no measurable cost appears in ordinary understanding, and where confusion would arise a language tends to displace one of the words. Machines lacking that context find the overlaps genuinely difficult.