Hostname: page-component-78c5997874-4rdpn Total loading time: 0 Render date: 2024-11-16T18:25:18.675Z Has data issue: false hasContentIssue false

Decidability and essential undecidability

Published online by Cambridge University Press:  12 March 2014

Hilary Putnam*
Affiliation:
Princeton University

Extract

There are a number of open problems involving the concepts of decidability and essential undecidability. This paper will present solutions to some of these problems. Specifically:

(1) Can a decidable theory have an essentially undecidable, axiomatizable extension (with the same constants)?

(2) Are all the complete extensions of an undecidable theory ever decidable?

We shall show that the answer to both questions is in the affirmative. In answering question (1), the decidable theory for which an essentially undecidable axiomatizable extension will be constructed is the theory of the successor function and a single one-place predicate. It will also be shown that the decidability of this theory is a “best possible” result in the following direction: the theory of either of the common diadic arithmetic functions and a one-place predicate; i.e., of addition and a one-place predicate, or of multiplication and a one-place predicate, is undecidable.

Before establishing the main result, it is convenient to give a simple proof that a decidable theory can have an axiomatizable (simply) undecidable extension. This is, of course, an immediate consequence of the main result; but the proof is simple and illustrates the methods that we are going to use in this paper.

Type
Research Article
Copyright
Copyright © Association for Symbolic Logic 1957

Access options

Get access to the full version of this content by using one of the access options below. (Log in options will check for institutional or personal access. Content may require purchase if you do not have access.)

References

1 The terminology of this paper is that of Undecidable theories, Tarski, Mostowski, and Robinson, North Holland Publishing Co., Amsterdam (1953)Google Scholar. This work will be cited as “U.T.”.

A solution to problem (1) has erroneously been announced by Myhill, (Solution of a problem of Tarski's, this Journal, vol. 21 (1956), pp. 4951)Google Scholar. Actually Myhill has solved a related problem; but an answer to (1) does not follow from what he proves, as he asserts. (Cf. section 7., below).

(Except for section 7) attention in the present paper is confined to theories with a finite number of constants. Kreisel has previously solved problem (1) for theories which use an infinite number of constants. A statement of Kreisel's proof appears in his review of the article by Myhill just cited; see Mathematical reviews, vol. 17, no. 8 (09 1956), p. 815Google Scholar.

2 This is easily seen to be a reformulation of a problem in U.T., p. 18. (Cf. the discussion below). From now on, when the term ‘extension’ is used, ‘with the same constants’ will be understood.

3 This problem was suggested by G. Kreisel whose interest and infectious enthusiasm have led me to work on these questions.

4 U.T., p. 64. Also, a decision method for the elementary theory of the successor function and an arbitrary one-place predicate is given in section 3. below. This can, of course, be used to decide sentences of R.

5 Logical signs are used as names of themselves throughout the present paper, and not in their object-language use.

6 Craig's method is as follows (On axiomatizability within a system, this Journal, vol. 18 (1953) pp. 3032Google Scholar): let {A1, A2, …} be a recursively enumerable but not recursive set of axioms for a theory T. Replace each axiom Ai by the conjunctions whose members are Ai repeated n times, for each n such that n is the number of a proof that Ai belongs to the set {A1, A2, …} (or rather that the Gödel number of At belongs to the corresponding recursively enumerable set of integers). The recursiveness of the new axiom set follows from the fact that the set of pairs (n, A) such that A belongs to {A1, A2, …} and n is the number of a proof that this is the case (in a suitable formalism) is recursive.

7 ⊤ and ⊥ (“tee” and “eet”) are Quine's symbols for truth and falsity (Methods of logic, Harvard (1950Google ScholarPubMed)). They are always eliminated by elementary reduction techniques; or else the whole formula reduces to ⊤ or ⊥. The logical notation in the present paper follows Quine (e.g., in using juxtaposition for conjunction; in using both ~ and − for negation), except that more dots than are strictly necessary for punctuation are sometimes used to facilitate reading. Following U.T., identity is included in each elementary theory as a logical constant.

8 In the sequel, a single formula will also be called a “conjunction” (with one member).

8a Beiträge zur Algebra der Logik, insbesondere zum Entscheidungsproblem, Mathematische Annalen, vol. 9, pp. 136Google Scholar.

9 “Prime formulas” are formulas of the forms P(x + m), (x + m), x + m = y + n, x + my + n, where m, n, may be positive, negative, or zero.

10 If the procedure eventuates in a formula of the form (∃x)m(xy), this is reduced to ⊤. Formulas of the forms (∃x)(x = y + m) and x = x + m are also eliminated.

10a “Integers n, n + 1, n + 2, … n + 4 realize PPPP” if and only if the integer n satisfies the formula PPPP; i.e., if and only if P(n)P(n+ 1)(n + 2)P(n+3)P(n+4) is true. — And similarly in similar cases.

11 O rekursivnoj otdélimosti, Doklady Akadémii Nauh USSR, vol. 88 (1953), pp. 953956Google Scholar.

12 At the referee's suggestion, I stress the informal character of this question. The terms ‘immediately stronger,’ ‘just weak enough,’ etc., at the beginning of section 6. are also used informally.

13 U.T., esp. pp. 77–80.

14 By Theorem 6 of U.T. (given below in section 7.).

15 The method is in fact contained in the proof of Lemma 1 of section 3.; the only modification is as follows: formulas of the form (∃x)m(xy) are not reduced to ⊤ (as in fn. 10 above), but are instead replaced by (∃x)m+1(x = x).

16 In U.T. only the weaker statement is given (p. 49) that a theory is essentially undecidable if every recursive function is definable in it.