Lewis’s Global Descriptivism and Reference Magnetism

ABSTRACT In ‘Putnam’s Paradox’, Lewis defended global descriptivism and reference magnetism. According to Schwarz [2014], Lewis didn’t mean what he said there, and really held neither position. We present evidence from Lewis’s correspondence and publications which shows conclusively that Lewis endorsed both.


The Obvious and the Abstruse Reading of 'Putnam's Paradox'
How does language hook onto reality? In 'Putnam's Paradox' [1984], David Lewis famously offered a two-part answer-global descriptivism plus reference magnetism. Global descriptivism says that the reference of all of our language's non-logical vocabulary is settled en bloc by a total theory, total in the sense that it is a term-introducing theory for every term in our language [ibid.: 224]. Our language's non-logical vocabulary refers to whatever is required to make our total theory come out true (or as close to true as possible). Global descriptivism cannot constitute the whole truth about reference, because there will be countless assignments of referents to the non-logical vocabulary that make total theory come out true if we allow assignments of ill-demarcated, entirely miscellaneous properties. So, radical indeterminacy results: this was the lesson that Lewis drew from Putnam's 'model-theoretic argument'. To avoid radical indeterminacy, Lewis added reference magnetism, which restricts the eligible referents of non-logical vocabulary to more-or-less natural properties, their boundaries demarcated by objective sameness and difference in nature (or as close to being so demarcated as possible) [ibid.: 227]. Lewis described reference magnetism as 'inegalitarianism' about properties because it discriminates between natural properties and what Lewis called 'gruesome' ones. He concluded that global descriptivism plus reference magnetism ensures determinacy of reference (or as close to determinacy as possible)-with a caveat.
In 'Putnam's Paradox', Lewis had granted, for the sake of argument, Putnam's outlook that beliefs and desires have content because the sentences expressing them do. Lewis's preferred philosophy of language was thought-first: before language can have content, beliefs and desires must. So Lewis added the caveat, 'It would have been better, I think, to start with the attitudes and go on to language ' [ibid.: 222]. Lewis referred the reader to 'New Work for a Theory of Universals' for his explanation of how 'the threat of radical indeterminacy in the assignment of content to thought is fended off' by restricting eligible contents of beliefs and desire to (more-or-less) natural ones [1983: 375]. Call the doctrine that the only eligible contents for thoughts are (moreor-less) natural properties 'content magnetism'.
Taken at face value, Lewis meant what he said in 'Putnam's Paradox'. He was a global descriptivist who adopted reference magnetism to fend off radical indeterminacy of reference, modulo the caveat that content magnetism is first required to fend off radical indeterminacy of content at the level of thought. Call this the 'obvious reading'.
An alternative reading of 'Putnam's Paradox', advanced by Wolfgang Schwarz, claims that Lewis 'did not actually believe what he was writing there ' [2014: 27]. In 'Against Magnetism', Schwarz alleges, 'there is no evidence that Lewis ever endorsed' reference magnetism, and 'Global Descriptivism, with or without magnetism, was not Lewis's own view' [ibid.: 17,29]. Call this the 'abstruse reading'.
Schwarz ascribes to Lewis content magnetism without reference magnetism, stating that 'objective naturalness plays essentially no role in Lewis's theory of language' but provides only a 'constraint on mental content ' [ibid.: 21, 26]. Schwarz grants that Lewis did 'defend magnetized Global Descriptivism in "Putnam's Paradox"', but maintains that Lewis didn't mean it because 'It flatly contradicts almost everything he wrote elsewhere on language ' [ibid.: 29]. Without further expounding the alleged contradiction, Schwarz offers a dilemma: either 'for a very brief period in the early eighties, Lewis gave up the brilliant account of language he had developed in great detail since the mid-sixties and replaced it by an utterly ridiculous alternative ' [ibid.] or, in 'Putnam's Paradox', 'Lewis defended a position he didn't actually endorse. This was not unusual for him; once he published an article under the pseudonym of his cat, Bruce, in which he argued against one of his own theories ' [ibid.].
This cannot be right. First, it was unusual for Lewis to publish defences of positions that he didn't endorse. His only pseudonymous article, 'Censored Vision' [LeCatt 1982], is relevantly unlike 'Putnam's Paradox'. Lewis wrote 'Putnam's Paradox' under his own name, not his cat's, much less his cat's pseudonym. Whenever Lewis referred back to 'Putnam's Paradox', he wrote in the first person. Referring back to 'Censored Vision', Lewis [1986: 289-90] called its author 'he' or 'LeCatt', and disagreed with LeCatt.
In this paper we provide decisive evidence against the abstruse reading, drawn from Lewis's correspondence as well as his publications. His letters have recently become available in Princeton's Firestone Library. If Schwarz were right, we would expect to discover Lewis distancing himself from reference magnetism and global descriptivism in correspondence. But, in several letters between 1983 and 2001, Lewis described himself as adhering to reference magnetism, global descriptivism, or both. We conclude by showing that Schwarz's dilemma is a false one. We make clear that 'Putnam's Paradox' does not contradict Lewis's other writings on language. There is no need to conclude that he either temporarily abandoned his account of language or didn't mean what he wrote in 'Putnam's Paradox'.

Correspondence: Reference Magnetism
Lewis originally sent Putnam a draft of 'Putnam's Paradox' in 1982, which Putnam then discussed in a paper at the December 1982 American Philosophical Association conference. Lewis sent another letter in reply, writing, 'I expect to agree completely with your statement that I need the natural properties for many other purposes than to secure determinacy (more or less) of reference' [Lewis-to-Putnam, 03/01/1983]. 1 Lewis told his long-term philosophical confidant J.J.C. Smart that Putnam's paper 'seems to concede that inegalitarianism about properties would provide the necessary constraints on reference, provided the discrimination among properties was not of our own making; but he finds inegalitarianism spooky' [Lewis-to-Smart 31/03/1983]. Lewis never wrote to Smart, Putnam, or anyone else that he didn't himself need inegalitarianism to secure determinacy of reference.
Further evidence that Lewis embraced reference magnetism is found in a letter to Paul Teller, replying to Teller's own suggestion that inegalitarianism is required for reference. Lewis wrote thus [Lewis-to-Teller, 16 April 1983]: Concerning your suggestion about naturalness as a constraint on reference: I'm not sure what you're proposing. It might be this: (1) that imperfect naturalness of some properties is to be understood in terms of definability from perfectly natural properties; then for words to have imperfectly natural referents is the same as for them to be definable from words, if such there be, with perfectly natural referents. … (1) is fine with me; it makes explicit something I've supposed, but left out because I didn't have details-a theory of simplicity of definitions -to offer.
Lewis proposed something very close to (1) in 'Putnam's Paradox' [1984: 228]. He could hardly have said that (1) [2001] where she had argued that no property is objectively more or less miscellaneous than another, because properties are only privileged subjectively. Lewis replied, 'Subjective privileging of properties requires fairly determinate content of your property-privileging thoughts (or language)' and 'Fairly determinate content of thought (or language) (except in certain extremely trivial cases) requires reference magnetism. Reference magnetism requires antecedently privileged properties' concluding, 'So at least some privileging of properties must be objective' [Lewis-to-McGowan, 20/09/2001]. He explained what he meant by 'extremely trivial cases'-either a world with only one thing, or a theory without non-logical vocabulary. Lewis wouldn't have done so unless he believed that reference magnetism constrains interpretations of theories with non-logical vocabulary in worlds with multiple things.

Correspondence: Global Descriptivism
In 'Putnam's Paradox', Lewis distinguished 'local' from 'global' descriptivism. The former explains how new expressions acquire reference via association with clusters of descriptions whose significance is already established. Lewis avowed a causal version of local descriptivism where reference of names is determined by causal descriptions [1984: 226-7, 1997: 339]. But he deemed local descriptivism 'disappointingly modest' because it doesn't explain how term-introducing descriptions acquire significance [1984: 223-4]. Global descriptivism is intended to take up the shortfall by explaining how terms in general acquire reference: they refer to whatever it takes to make total theory come true. As we shall see, contra Schwarz, Lewis endorsed both local and global descriptivism.
In 1988, Lewis wrote to Peter Forrest, 'My view is inegalitarian-descriptivist', explaining, 'Once the over gruesome interpretations are out by inegalitarianism, I would hope that fit can do the rest of the job. Fit is something like making ideal theory come true' [Lewis-to-Forrest, 14/11/1988]. This certainly shows Lewis avowing reference magnetism by appealing to inegalitarianism to rule out assignments of miscellaneous properties. Far from confirming Schwarz's claim that 'Global Descriptivism, with or without magnetism, was not Lewis's own view ' [2014: 29], it also strongly suggests a commitment to global rather than local descriptivism. In 'Putnam's Paradox', ideal theory is identified as 'empirically ideal theory, verified come what may' and 'ideal theory is the term-introducing descriptive theory to which global descriptivism applies ' [1984: 221, 224]. To Forrest, Lewis characterizes 'fit' as 'something like making ideal theory come true', so inegalitarian-descriptivism involves something like applying global descriptivism to ideal theory, as Lewis had described it in 'Putnam's Paradox'. The next line of the letter reads, 'I'm not sure how much farther we can go while joining Putnam in ignoring the difference between interpreting thought and interpreting language.' But that line doesn't confirm Schwarz's assertion that Lewis didn't believe what he wrote, or that Lewis didn't hold the view he had just called 'my view'. It simply gestures towards Lewis's content-magnetism caveat, acknowledged by the obvious reading, that language cannot have determinate reference unless thoughts have determinate content.
In 1989 Lewis, visiting Melbourne, met Barry Taylor to discuss a draft of Taylor's 'Just More Theory' [1991], which criticizes 'Putnam's Paradox'. The next day, Lewis sent a letter to Taylor. Putnam had claimed that ideal theory can't be false, because there are inevitably miscellaneous assignments of reference to make it come out true. Taylor convinced Lewis that Putnam hadn't been refuted because there was nothing that Lewis could say to force Putnam to concede that such assignments were unintended. But, Lewis replied, Putnam hadn't refuted his position either. Putnam hadn't undermined Lewis's conviction that most miscellaneous interpretations are unintended. So, Lewis reiterated, 'my position is that not all the models of ideal theory are intended' [Lewis-to-Taylor, 08/ 09/1989]. Lewis later wrote to Smart that while he'd have liked a 'win' against Putnam, he'd happily settle for a draw because [Lewis-to-Smart, 22/08/1991] Given a draw I am free to go on as before thinking that a permutation that turns 'cat' into the word for cherries is an incorrect interpretation of English; that not just any interpretation that makes epistemically ideal theory come true is a correct interpretation.
Lewis didn't disavow the position of 'Putnam's Paradox' even in the face of criticism, as we would expect if Schwarz were right. Lewis continued to own it as 'my position'.

The Place of Magnetized Global Descriptivism in Lewis's Philosophy of Language
Schwarz claims that it would be absurd to think that, during the early 1980s, Lewis had abandoned 'the brilliant account of language he had developed in great detail since the mid-sixties and replaced it by an utterly ridiculous alternative', magnetized global descriptivism, which 'flatly contradicts almost everything he wrote elsewhere on language' [Schwarz 2014: 29]. Schwarz doesn't explain the alleged contradiction, but presumably considers magnetized global descriptivism to be incompatible with Lewis's earlier account of reference. But, before 'Putnam's Paradox', Lewis didn't have one.
In the works that Schwarz considers the summum bonum of Lewis's philosophy of language, Convention and 'Languages and Language', Lewis argued that we use a language L iff by convention we are truthful and trusting in L. In Convention, Lewis admitted that 'Conventions of truthfulness pertain to whole sentences and leave the interpretations of parts of sentences underdetermined ' [1969: 198]. For early-Lewis, a 'grammar' includes an assignment of interpretations to words. Lewis thought that some grammars are simple and intuitive, while others aren't. But he admitted that he had no method for selecting the best one. He considered a method of evaluation based on Chomsky's psycholinguistic reflections-that the best grammar is one that enters into the psychological explanation of linguistic competence. But he found Chomsky's hypothesis 'speculative'. If it failed, Lewis conceded, 'we must look elsewhere for a privileged method of evaluation, or we must give up. To give up would be to accept Quine's thesis of the inscrutability of reference ' [ibid.: 199-200].
In 'Languages and Language', Lewis distinguished three interpretive steps: (1) assignment of contents to a subject's system of propositional attitudes, given their functional liaisons to sensory stimulations and behavioural outputs; (2) assignment of truth-conditions to the full sentences of the public language that the subject speaks, based on interpretation of the propositional attitudes of the population to which she belongs; (3) assignment of meanings to subsentential constituents of the public language, based on truth-conditions of full sentences.
Lewis's focus was (2), from attitudes to truth-conditions of sentences: for him, roughly speaking, to know the meaning of 'snow is white' is to know that this sentence is conventionally correlated with the belief that snow is white. With respect to (3), from sentences to the meanings of their parts, Lewis had dropped Chomsky's hypothesis that the best grammar enters into the psychological explanation of linguistic competence. Without other grounds for selecting the best grammar, Lewis felt 'inclined to share in Quine's doubts about the determinacy of the third step ' [1975: 21]. Lewis made the same claim in 'Radical Interpretation' with respect to interpreting speakers of a language M: 'the truth conditions of full sentences in M do not suffice to determine the rest of M: the parsings and meanings of the constituents of sentences ' [1974: 44]. So Lewis, following Quine, saw no royal road from the truth-conditions of sentences to the reference of their constituents. Because early-Lewis only discussed the truth-conditions of sentences, and admitted that he lacked an account of reference, it's difficult to see how the account of reference in 'Putnam's Paradox'-magnetized global descriptivism-can 'contradict', as Schwarz claims, what Lewis wrote elsewhere on language.
Schwarz states that 'the task of semantics, for Lewis, is not just to come up with any old grammar that delivers the right truth-conditions, but with a systematic, simple and intuitive grammar ' [2014: 20]. But in Convention Lewis already didn't think that these virtues provided a basis for determining the best grammar-hence his appeal to Chomsky's psycho-linguistic hypothesis. Except for systematicity, these virtues aren't even mentioned in 'Languages and Language', where Lewis's policy was to refrain (where possible) from saying anything about the best grammar of a language because of his doubts about the clarity and objectivity of our standards for evaluating grammars [1975: 20-1]. Such a policy could hardly have been in order if Lewis had held the standards of simplicity and intuitiveness, etc., to be clear and objective enough to select the best grammar. Nor could his pessimism about the determinacy of (3), from truth-conditions of sentences to the meanings of their parts, have been warranted.
In 1978 Gary Merrill sent to Lewis a draft of his 'Model-Theoretic Argument Against Realism' in which he sought, not for himself but on behalf of realists, to defang Putnam's model-theoretic argument by appealing to an ontology of relations [Merrill 1980: 72]. Lewis wrote back, 'I like your reply to Putnam', praising Merrill's 'requirement that the predicates be interpreted by genuine relations' [Lewis-to-Merrill, 11/ 10/1978]. Lewis continued, 'I also think that it might be OK to simply insist that you', the realist, 'do have the distinction between genuine relations and others, and are entitled to use it in saying what an admissible interpretation is'. As we've shown, Lewis previously lacked an account of admissible interpretations of sentence-constituents. We can see why Lewis liked Merrill's reply. It filled the gap that Lewis had acknowledged in his other writings on language.
Had early-Lewis held that settling the assignments of truth-conditions to sentences eo ipso settled the reference of subsentential constituents, then it wouldn't be far-fetched to consider his philosophy of language at odds with the magnetized global descriptivism of 'Putnam's Paradox' and his letters. They would yield potentially conflicting recipes for determining reference. The first would involve selecting (somehow) the best grammar for a language, based on conventionally established truth-conditions for its sentences. The second would involve finding an ideal theory and selecting the best grammar, based on a trade-off between making the theory true and the naturalness of the candidate interpretations. Since the two recipes select the best grammar on different bases, truth-conditions versus making ideal theory come true, they cannot be guaranteed to select the same grammar. But Lewis didn't think that settling the assignments of truth-conditions to sentences eo ipso settled the reference of their constituents. He never proposed the first recipe, but stressed that the truth-conditions of sentences do not determine reference, that interpretive step (3) doesn't follow from step (2). So, Lewis's proposal of the second recipe in 'Putnam's Paradox' doesn't conflict with what he had previously written. Appeal to making ideal theory come true supplements his earlier account, which had gone no further than assigning truth-conditions: choose an assignment of reference that makes ideal theory come out true, relative to the assignment of conventionally established truth-conditions. 2

Disclosure statement
No potential conflict of interest was reported by the authors.