¹ LaPorte, Joseph, Natural Kinds and Conceptual Change (Cambridge: Cambridge University Press, 2004)Google Scholar.

² Op. cit., 97.

³ Op. cit., 47.

⁴ Op. cit., 43.

⁵ Op. cit., 93.

⁶ Op. cit., 104–107.

⁷ Abbott, B. ‘A Note on the Nature of “Water”’, Mind 106 (1997), 311–9CrossRefGoogle Scholar.

⁸ See LaPorte, J., ‘Living water’, Mind 107 (1998), 451–5CrossRefGoogle Scholar.

⁹ Under a Putnamian view, being H₂O is a necessary and sufficient condition for being water insofar as the threshold of impurities is lower than 20% (see Putnam, H., Representation and Reality (Cambridge, Mass.: MIT Press, 1998), 31)Google Scholar. But LaPorte (1998), op. cit. note 8, has attacked this claim by noting how a cutoff at 20% is not sufficiently fine-grained a net to determine the reference of the term ‘water’. By trading on the fact that the content of the Great Salt Lake in Utah contains twenty-eight percent impurities, LaPorte concludes that a jellyfish, a tomato, and an infant would count as ‘water’ in virtue of containing fewer impurities than the lake in Utah.

¹⁰ Op. cit. note 1, 104–107.

¹¹ Op. cit., 105.

¹² Op. cit., 106.

¹³ Op. cit., 107, 110.

¹⁴ Bird, A., ‘A posteriori knowledge of natural kind essences: a defence’, Philosophical Topics 35 (2007), 293–312CrossRefGoogle Scholar.

¹⁵ Op. cit. note 14, 299. See also Bird, A., ‘Discovering the essences of natural kinds’, in Beebee, H. and Sabbarton-Leary, N. (eds) The Semantics and Metaphysics of Natural Kinds (New York: Routledge, 2009), 128Google Scholar.

¹⁶ Bird 2009, op. cit note 15, 135.

¹⁷ Henceforth I use the vernacular kind term for the purpose of comparing it with the Earthlings' vernacular kind term ‘dwater’ under the assumption that both Earthlings and Earth scientists would agree in using the scientific kind term ‘deuterium oxide’ to refer to D₂O, but they would diverge in the use of the vernacular kind term to stress their divergent opinions as to whether D₂O is or is not a kind of water.

¹⁸ Urey, H. C., Brickwedde, F. G., Murphy, G. M. ‘A Hydrogen Isotope of Mass 2’, Physical Review 39 (1932), 164–165CrossRefGoogle Scholar.

¹⁹ See Rutherford, E., et al., ‘Discussion on heavy hydrogen’, Proceedings of the Royal Society A 144 (1934), 1–28CrossRefGoogle Scholar.

²⁰ Op. cit note 19, 14.

²¹ Op. cit, 6.

²² See Dahl, P. F., Heavy water and the wartime race for nuclear energy (Philadelphia: Institute of Physics Publishing, 1999), 29–30Google Scholar.

²³ Op. cit note 22, 41.

²⁴ See op. cit. note 22.

²⁵ In the words of Rae, Howard, Separation of hydrogen isotopes (Washington, DC: American Chemical Society, 1978), 3CrossRefGoogle Scholar, from Chalk River Nuclear Laboratories of Ontario (Canada): ‘This very low value of the deuterium-to-hydrogen ratio in nature of about 150 ppm is the main factor responsible for the high cost of heavy water. It is necessary to process at least 8000 mols of feed per mol of product for all processes…Thus, the overall concentration ratio from feed to product is about 3 × 10⁶. This means that hundreds of separative elements in series are needed to go from natural water to reactor grade heavy water.’

²⁶ Franks, F., Water: a matrix of life (Cambridge: The Royal Society of Chemistry, 2000), 19Google Scholar.

²⁷ Incidentally, this is one of the motivations for Paul Needham's suggestion that microstructural features should not be invoked to classify chemical substances, and instead macroscopic thermodynamic features, such as Gibbs' rule provide more helpful standards for classifying substances (see Needham ‘What is water?’, Analysis 60 (2000), 13–21CrossRefGoogle Scholar; and Needham ‘The discovery that water is H₂O’, International Studies in the Philosophy of Science 16 (2002), 205–226)CrossRefGoogle Scholar. On Needham's account, deuterium oxide would indeed qualify as a separate chemical substance (for a criticism of thermodynamic considerations as a criterion for classifying chemical substances, see Hendry, R. ‘Entropy and chemical substance’, Philosophy of Science 77 (2010), 921–932)CrossRefGoogle Scholar. Although Needham's suggestion is a helpful one, in what follows I stick with scientific orthodoxy in taking the microscopic atomic-number dependent properties as the basis for chemical elements' classifications (as per periodic table), and in taking the macroscopic thermodynamic properties as supervenient on them. In particular, I endorse Robin Hendry's (‘Microstructuralism: problems and prospects’, in Ruthenberg, K. and van Brakel, J. (eds.) Stuff: the nature of chemical substances (Würzburg: Königshausen und von Neumann, 2008), 118Google Scholar) characterization of microstructuralism as ‘committed to two dependence theses: (i) the physical exhaustion of the chemical (take away the microphysical properties and there is nothing chemical left), and (ii) supervenience of chemical kind membership on microphysical properties (there can be no change, or difference, in chemical kind membership without some change, or difference, in microphysical properties’ intended as a non-reductionist thesis. My goal here is to assess LaPorte's criticism of Putnam's view that it is the relation same microstructural kind to paradigm sample of ‘water’ in our actual world that establishes whether deuterium oxide is a kind of water. Whether or not microstructuralism is in turn correct as a view of chemical classification is a separate issue, which falls outside the aim and scope of this paper.

²⁸ Hacking, I., ‘The contingencies of ambiguity’, Analysis 67 (2007), 269–77CrossRefGoogle Scholar.

²⁹ Hendry, Robin (‘Elements, compounds and other chemical kinds’, Philosophy of Science 73 (2006), 864–875CrossRefGoogle Scholar; and ‘The elements and conceptual change’, in Beebee, H. and Sabbarton-Leary (eds) The Semantics and Metaphysics of Natural Kinds (New York: Routledge, 2010), 148)Google Scholar has poignantly stressed this point against LaPorte by noting how atomic mass bears no close relationship to chemical behaviour because isotopes of two different elements may share the same atomic weight.

³⁰ A clarification is in order here. The above observation is not meant to be a sweeping claim to the effect that it would be unconceivable to take isotopic varieties (say, protium and tritium, where the latter is radioactive) as two different substances. Instead, the observation above is meant to be a simple remark that given our current periodic table, and given Bohr's victory over the role of atomic number in the classification of elements, it just happens that in any Putnamian-like baptism of D₂O in our actual world isotopic varieties of water are not normally classified as different substances. Appealing to the intuition that one is radioactive and the other is not, does not begin to show that they should or could be counted as two separate substances.

³¹ I do not want to suggest that low concentration of substances in nature is a ground for not counting them as distinct kinds. If this were the case, then many substances (e.g. rare earths) would not qualify as distinct kinds, which is obviously false. Instead, I am making the more modest claim that given the low percentage of isotopic varieties (such as deuterium oxide) in naturally occurring water, there is no paradigm sample for LaPorte's imagined baptism of D₂O as not a kind of water in our planet.

³² Op. cit. note 1, 107.

³³ Craig, H. ‘Standard for reporting concentrations of deuterium and oxygen-18 in natural waters’, Science vol. 133, No. 3467 (1961), 1833CrossRefGoogle ScholarPubMed.

³⁴ Op. cit note 33, 1833.

³⁵ Op. cit, 1833.

³⁶ See footnote 27.

³⁷ I echo here Fred Dretske's (‘Laws of nature’, Philosophy of science 44 (1977), 248–68CrossRefGoogle Scholar) similar argument for laws of nature intended as necessitation relations between universal properties. In that context, the opacity under co-extensive predicate substitution serves as an argument for the view that laws cannot be understood as true universal generalizations. In my case, the opacity under co-extensive predicate substitution serves as an argument for the view that we cannot interchange co-extensive vernacular kind terms in inductive inferences, salva projectibility.

³⁸ LaPorte's analogous case for biological kind terms (op. cit. note 1, ch. 3) is more successful than his case for chemical kind terms, in my view, precisely because in biology there are competing biological schools (cladistics and evolutionary taxonomy), for classifying biological species and higher taxa. The present article should then be read not as an attempt to criticise LaPorte's view tout court, but instead as raising a difficulty for extending his account to chemical kind terms (at least, in the specific case of deuterium oxide).

³⁹ At the end of Fact, Fiction, and Forecast, Goodman suggested that his analysis of projectibility could be used to distinguish between genuine and artificial kinds, or more genuine from less genuine ones: ‘For surely the entrenchment of classes is some measure of their genuineness as kinds; roughly speaking, two things are the most akin according as there is a more specific and better entrenched predicate that applies to both’, Goodman, N.Fact, Fiction, and Forecast, 3^rd ed. (Indianapolis: Bobbs-Merrill, 1973), 123Google Scholar.

⁴⁰ Boyd, R. ‘Realism, anti-foundationalism, and the enthusiasm for natural kinds’, Philosophical Studies 61 (1991), 127–48CrossRefGoogle Scholar. Kornblith, H.Inductive inference and its natural ground. An essay in naturalistic epistemology (Cambridge, Mass.: MIT Press, 1993)Google Scholar. Hacking, I. ‘Working in a New World: the Taxonomic Solution’, on Horwich, P. (ed.) World Changes. Thomas Kuhn and the Nature of Science (Cambridge, Mass.: MIT Press, 1993), 275–310Google Scholar. Kuhn, T. ‘Afterwards’, in Horwich, P. (ed.) World Changes. Thomas Kuhn and the Nature of Science (Cambridge, Mass.: MIT Press, 1993), 311–39Google Scholar. While Goodman's notion of projectibility is tied to his notion of entrenchment (i.e. how often a predicate or a natural kind term has been successfully projected in the past), some philosophers after Goodman have married projectibility to truth as in Boyd's case, where kind terms are said to be projectible because they latch onto the causal structure of the world (Boyd, op. cit., 139). Other philosophers have retained projectibility without giving an account of it in terms of truth or entrenchment (as in Hacking's case). Others, like Kuhn (1993, op. cit., 315–9), have defended projectibility within the context of a prima facie constructivist position. In what follows, I am going to be noncommittal about the nature of projectibility and I take it simply as a measure of natural kind terms' amenability to supporting successful inductive inferences.

⁴¹ One may be suspicious about projectibility on the ground of Popperian doubts about the success of our inductive inferences in general. But this would be a non-sequitur. One of the key motivations for natural kinds is that they are meant to support our inductive inferences and explanations, quite independently of whether one is realist, nominalist or constructivist about kinds (as clarified in the footnote above). Even a Popperian, with reservations about our ability to project into the future, would owe us an account of why inductive inferences about green emeralds have been successful up to date. And normally such an account would involve reference to natural kinds, quite independently of which view of natural kinds one favours (i.e. again realist, nominalist, or constructivist), and independently also of what view one endorses about the relation between natural kinds and dispositional properties, for example (e.g., whether kind membership can be identified with a set of necessary and sufficient dispositional properties).

⁴² Putnam (‘The meaning of “meaning”’, in Putnam, H.Mind, Language, and Reality: Philosophical Papers vol. 2, (Cambridge: Cambridge University Press, 1975), 269)Google Scholar famously defined the meaning of a word (say ‘water’) as a four-component vector, including: 1. Semantic marker (e.g. natural kind; liquid); 2. Syntactic marker (e.g. mass noun, concrete); 3. Stereotype (e.g. colorless; transparent;…); and 4. Extension (e.g. H₂O). Putnam identified the meaning of ‘meaning’ with the extension of the word.

⁴³ Recall that for Putnam, op. cit. note 42, 270, ‘although we have to use a description of the extension to give the extension, we think of the component in question as being the extension (the set), not the description of the extension’.

⁴⁴ See Putnam, op. cit., 231, on this point.

⁴⁵ Goodman's paradox does not just apply to predictions about future cases, but more in general to inductive inferences involving projections from examined cases to (past, present, and future) unexamined cases. See on this point Sober, E., ‘No model, no inference: a Bayesian primer on the grue problem’, in Stalker, D. (ed.) Grue, the New Riddle of Induction (La Salle, Illinois: Open Court Publishing Company, 1994), 193–223Google Scholar; and Israel, R. ‘Two interpretations of grue – or how to misunderstand the new riddle of induction’, Analysis 64 (2004), 335–9CrossRefGoogle Scholar.

⁴⁶ LaPorte, op. cit. note 1, 107.