"‘Water’ and ‘Water’: On Twin-Earth and the Metaphysics of Words" - James (JTM) Miller (Durham University)

Journal of the American Philosophical Association, Online First

By James (JTM) Miller (Durham University)

Paper Available (open access) here

If you ever took an undergraduate class in the philosophy of language, there is a very good chance that you have heard of Twin-Earth. As Hilary Putnam (1975) describes it, it is a curious place – identical to Earth, but for one difference. Whereas on Earth, the liquid that flows in the rivers, that humans drink and fish swim in has the molecular structure of H2O, on Twin-Earth, the liquid that flows in the rivers, etc., has the molecular structure XYZ. Otherwise, Earth and Twin-Earth are identical. All objects in one have a molecule-for-molecule duplicate including the central figures in our story: Oscar and Twin-Oscar.

Oscar and Twin-Oscar live their lives, in a time before the development of molecular chemistry that might reveal the difference in their worlds. At some point, both utter the word ‘water’. Putnam’s question is: do Oscar and Twin-Oscar mean the same thing as each other when they utter this word? Their intensions will be the same as they are exact duplicates, but, Putnam suggests, this overlooks the important difference in their worlds. Putnam argues that ‘water’ as uttered on Earth by Oscar, and ‘water’, as uttered on Twin-Earth by Twin-Oscar, must mean something different. One ‘water’ refers to H2O, while the other ‘water’ refers to XYZ.

From this, Putnam draws his conclusion of semantic externalism as, roughly, the view that meanings cannot be solely determined by the internal states of individuals and must instead also depend on external factors. Meanings, as Putnam famously quips, therefore ‘just ain’t in the head’ (1975: 144).

It's hard to overstate the impact of this thought experiment on debates in philosophy of language, and related debates in other domains also. Even if some have argued against it, it is clearly a highly influential argument, and, speaking from my own teaching experience, it is an argument that most students find compelling, at least initially. However, my contention is that there is assumption that has gone unnoticed in this argument – an assumption concerning the metaphysics of words.

Before we see what that is, though, we should note that a central part of Putnam’s argument is that Oscar and Twin-Oscar are ‘psychologically’ the same, at least in their narrow psychology. It is only if they are psychologically the same that we are pushed towards the position that the difference in meaning between water’ as uttered on Earth, and ‘water, as uttered on Twin-Earth is to be explained through some external factor. If Oscar and Twin-Oscar are not psychologically the same, then the difference in the meaning of ‘water’ and ‘water’ can be explained by that psychological difference.

Back to words. Why might we think that Oscar and Twin-Oscar are not psychologically the same? I suggest one reason might be that they know different words. The words that we know are part of our psychology. Putnam even gives the example of ‘knowing the alphabet’ as an example of a psychological state (1975: 136). So, now we can ask: are ‘water’ as uttered by Oscar, and ‘water’ as uttered by Twin-Oscar are instances of the same word (or tokens of the same type, if you prefer)? If so, Putnam’s argument holds; if not, then it is based on an unsecured assumption about the sameness of ‘water’ and ‘water’. My argument is that the later is the case. Putnam assumes a controversial claim about ‘water’ and ‘water’ as uttered by Oscar and Twin-Oscar being instances of the same word.

At this point, it seems like we need to know what it is that does make two instances of a word instances of the same word. However, usefully, for my purposes at least, I don’t think we need to fully answer the question of when two words are instances of the same word to make some progress here. That’s a notoriously hard question so it would be good to avoid it if we can. Rather, I think we can instead just argue that there are no good grounds for Putnam to hold that ‘water’ and ‘water’ are instances of the same word. I argue that whatever grounds we use to try to establish that they are the same word are metaphysically implausible and/or would actually result in ‘water’ and ‘water’ being instances of different words contra Putnam’s assumption. Here (briefly) are some of the possible ways we might secure Putnam’s assumption, and why I think they fail.

We cannot use semantic properties of the words as Putnam precisely holds that they are different. Putnam precisely cannot use meanings to easily secure that these are instances of the same word, without begging the question against the semantic internalist. Moving to sets of meanings might seem promising, but will lead to vagueness issues and is generally rejected in the literature on words (hence, I suggest, we have reasons to think it is more broadly metaphysically implausible, even if we could find a version to help Putnam).

We cannot use phonetic properties as even if some particular instances of ‘water’ and ‘water’ uttered by Oscar and Twin-Oscar are exactly the same phonetically. This is because other instances of ‘water’ uttered by them both will not be. This is obviously problematic unless Putnam’s argument is only about those particular utterances. It is also a view that is widely rejected in the metaphysical literature on words.

We cannot appeal to the origins of the words either. This is perhaps initially more promising as it is a view that has been defended in the literature. However, even if it is right, it will not help Putnam because the words we are considering have different origins! One originates on Earth, and the other on Twin-Earth. These worlds might be duplicates, and so the words have exactly resembling origins, but resemblance is different from identity.

The above options assume some version of a type-token ontology of words, but I also argue similar problems arise on other popular ontologies also. Kaplanian stage-continuant ontologies appeals to historical connection to some ‘common ancestor’ to hold that stages are stages of the same continuant. This fails in Putnam’s case as there is no historical connection to the same ancestor across worlds. And, for completeness, adopting nominalism also will not help Putnam – again, even if resemblance is sufficient to account for other phenomena, resemblance is not the identity that Putnam needs. Oscar and Twin-Oscar would not be psychologically the same as they would know resembling, but not identical, words.

Naturally, there are some more details to each of these points, but the moral, I hope, becomes clear. Putnam’s argument for semantic externalism needs to give us a reason why ‘water’ and ‘water’ uttered on Earth and Twin-Earth are instances of the same word, but he cannot. Options are either metaphysically implausible or would not secure what Putnam needs (even if we granted that those theories are, more broadly, viable).

There are similar worries, I think, for related arguments too. Burge, for example, runs a similar but distinct thought experiment in favour of semantic externalism, focused on the meaning of terms like ‘arthritis’. Here too, I think there are some relevant assumptions about the metaphysics of words that might undermine the conclusion. While ‘arthritis’ on Earth and ‘arthritis’ on Twin-Earth might be tokens of the same word, we cannot assume that they are. And, as for Putnam, it is not clear what metaphysics of words would allow the conclusion of Burge’s argument to follow.

All this, it should be stressed, does not disprove semantic externalism itself. Twin-Earth style thought experiments are not the only arguments for semantic externalism, even if they are highly influential ones. I only suggest that these arguments are not as secure as they seem, and that this shows that the metaphysics of words is important for debates beyond pure metaphysical speculation. What we take words to be might impact various arguments elsewhere in the philosophy of language.