Full Text (some footnotes omitted)
- In 1950, having Methods of Logic and a revision of Mathematical Logic in hand, I set my sights on a book of more broadly philosophical character. It proved in the fullness of time to be Word and Object1, and the fullness of time was nine years. I foresaw by 1952 that it would be a long pull and became impatient to make some of my philosophical views conveniently accessible meanwhile. Henry Aiken and I were with our wives in a Greenwich Village nightspot when I told him of the plan, and Harry Belafonte had just sung the calypso "From a logical point of view." Henry noted that this would do nicely as a title for the volume, and so it did.
- The book did nicely as well. In the course of its two editions and its many printings it sold nearly forty thousand copies in English and I have no notion how many in Spanish, Italian, Polish, German, and Japanese. Eight of the nine essays have reappeared also independently in one or more anthologies, and each in one or more translations. The first two, indeed, have been anthologized to extinction: twenty-four and twenty-five times respectively and in seven and six languages. I am much gratified and flattered by all this, and likewise by the readiness of my friends at Harvard University Press to take over the paperback rights and keep up the output.
- The time for revision is past. The book is dated, and its dates are 1953 and 1961. On the present occasion I have revised just a single page, one that contained mistaken criticism of Church and Smullyan. It is page 154, amid the tumultuous pages where most of the 1961 revision took place.
- But I shall improve the opportunity in this preface for a few caveats. One is that "On what there is" is nominalistic neither in doctrine nor in motivation. I was concerned rather with ascribing ontologies than with evaluating them. Moreover, in likening the physicists' posits to the gods of Homer, in that essay and in "Two dogmas," I was talking epistemology and not metaphysics. Posited objects can be real. As I wrote elsewhere, to call a posit a posit is not to patronize it.
- The holism in "Two dogmas" has put many readers off, but I think its fault is one of emphasis. All we really need in the way of holism, for the purposes to which it is put in that essay, is to appreciate that empirical content is shared by the statements of science in clusters and cannot for the most part be sorted out among them. Practically the relevant cluster is indeed never the whole of science; there is a grading off, and I recognized it, citing the example of the brick houses on Elm Street.
- Both that essay and the next, "The problem of meaning in linguistics," reflected a dim view of the notion of meaning. A discouraging response from somewhat the fringes of philosophy has been that my problem comes of taking words as bare strings of phonemes rather than seeing that they are strings with meaning. Naturally, they say, if I insist on meaningless strings I shall be at a loss for meanings. They fail to see that a bare and identical string of phonemes can have a meaning, or several, in one or several languages, through its use by sundry people or peoples, much as I can have accounts in several banks and relatives in several countries without somehow containing them or being several persons. It is usually convenient elsewhere in linguistics to distinguish homomorphs by meanings or history — sound (sonus) and sound (sanus), for example — but when we are philosophically concerned about meaning we had best not bury it. I hope this paragraph has been superfluous for most readers.
- Finally, some technical remarks about "New foundations." We see in pages 98-99 the superiority of ML over NF in respect of mathematical induction and the existence of the class of natural numbers. There remains, however, this related infirmity in ML: Rosser has shown that the class of natural numbers cannot be proved in ML to be a set, or element, if ML is consistent. We can still add an axiom to that effect, and indeed we need it for the theory of real numbers. But it is inelegant to have to add it.
- NF and ML can be further criticized for allowing self-membership, which beclouds individuation2. The glory of classes, over against properties, is their clear individuation3: they are identical if and only if they have the same members. This, however, is relative individuation4; classes are individuated only as clearly as their members. Under self-membership the individuation5 ceases to wind down.
- Russell's theory of types has an epistemological advantage over NF and ML: it lends itself to a more plausible reconstruction of the genesis of high-level class concepts6. From the theory of types to the set theories of Zermelo and von Neumann, in turn, a natural transition can be made7. NF is to be reckoned as an artificial alternative devised afterward for its convenience and elegance; and ML is another. The advantages are real, despite the above reservations.
- During the forty-odd years since NF was first published, much ingenious work has been done by Rosser, Benes, Specker, Orey, Henson, Jensen, Boffa, Grishin, and others in hopes of either deriving a contradiction or proving that the system is consistent if a more classical set theory is consistent. The problem is still open, but a number of curious and surprising relationships have been uncovered in the course of the search.
… Cambridge, Massachusetts
Footnote 1: See "Quine (W.V.) - Word & Object".
Footnote 6: See The Roots of Reference (La Salle, Ill.: Open Court, 1973), pp. 120ff ("Quine (W.V.) - The Roots of Reference").
Footnote 7: See Set Theory and Its Logic (Cambridge: Harvard, 1963, 1969), §§ 38, 43 ("Quine (W.V.) - Set Theory and its Logic").
Text Colour Conventions (see disclaimer)
- Blue: Text by me; © Theo Todman, 2019
- Mauve: Text by correspondent(s) or other author(s); © the author(s)