In this paper I raise a difficulty for Joseph LaPorte's account of chemical kind terms. LaPorte has argued against Putnam that H2O content is neither necessary nor sufficient to fix the reference of the kind term ‘water’ and that we did not discover that water is H2O. To this purpose, he revisits Putnam's Twin Earth story with the fictional scenario of Deuterium Earth, whose ocean consists of ‘dwater’, to conclude that we did not discover that deuterium oxide is (a kind of) water (usually called ‘heavy water’). Instead, according to LaPorte, by including deuterium oxide in the extension of the term ‘water’, we simply refined our vague use of the term ‘water’. But we could have decided to exclude deuterium oxide from the extension of the term ‘water’. Let us call this the thesis of semantic stipulation.
I raise two problems for LaPorte's Deuterium Earth story. First, I show that ‘dwater’ (i.e. deuterium oxide not as a kind of water) does not have the same scientific credibility of ‘heavy water’ (i.e. deuterium oxide as a kind of water). Second, I argue that for the thesis of semantic stipulation to go through one would need to show that ‘dwater’ is semantically on a par with ‘heavy water’. Namely, one would need to show that ‘dwater’ is a projectible kind term, capable of supporting inductive inferences. But, in fact, it is not, because the term is vulnerable to an unwelcome Goodmanian scenario, unless one surreptitiously reintroduces some Putnamian assumptions about D2O content being necessary and sufficient to fix the reference of the term.