Gavagai
When a native speaker says 'gavagai' as a rabbit runs by, is there any fact of the matter about what they mean?
W. V. O. Quine introduced this example in his 1960 book Word and Object to argue that translation between languages is radically indeterminate. No amount of behavioral evidence can settle what a word refers to, because every possible observation is consistent with multiple incompatible interpretations.
Quine, W. V. O. (1960). Word and Object. MIT Press.
The scenario
A field linguist is learning a previously unknown language by observation. A rabbit runs across a field. The native speaker points and says "gavagai." The linguist writes it down as a candidate translation for "rabbit."
But consider what the speaker might actually mean. "Gavagai" could mean:
- Rabbit
- Undetached rabbit parts
- A rabbit stage (the rabbit as it exists at this moment)
- Rabbithood instantiated here
- A temporal slice of a rabbit
Every behavioral test the linguist can devise is consistent with all of these interpretations. If the linguist points at a rabbit and says "gavagai?" the native nods. If the linguist points at rabbit ears and says "gavagai?" the native still nods, because those are undetached rabbit parts, rabbit stages, and instances of rabbithood just as much as the whole rabbit is.
The indeterminacy of translation
Quine's thesis is that the indeterminacy of translation is not a practical problem but a logical one. There is no fact of the matter that distinguishes the correct translation from the incorrect ones. The totality of behavioral evidence, everything a linguist could ever observe, underdetermines which translation manual is right.
This is not skepticism about whether we can ever know what people mean. It's a stronger claim: there is nothing to know. Two translation manuals that yield systematically different outputs are both equally valid if they both preserve the overall pattern of correct and incorrect assertions.
What this means for language and understanding
If Quine is right, the problem does not disappear when we learn our own first language. English speakers learn their language the same way the linguist does: by observing behavior in context. The rabbit-stage interpretation was never ruled out. We just converged on one manual and use it consistently.
This challenges the commonsense picture of language as a system where words latch onto things in the world. For Quine, there is no deeper fact behind the consistent application of a word. Reference is not a relation between language and reality that exists independently of the practices that organize it. What we call translation is always relative to a chosen scheme, and no scheme is uniquely correct.
Discussion questions
- If you were trying to learn a completely alien language, how would you know when you had the right translation?
- Do you think what words mean is ever fully fixed, or is there always ambiguity underneath?
- Can you think of a miscommunication in your life that came from assuming your interpretation of a word was the only one?
Take it to the dinner table.
Get 3 thought experiments for memorable conversations, designed for dinner, with friends, at events, or anywhere small talk has gone on too long.