home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky sci.math:17408 sci.logic:2480
- Path: sparky!uunet!think.com!hsdndev!husc-news.harvard.edu!husc10.harvard.edu!zeleny
- From: zeleny@husc10.harvard.edu (Michael Zeleny)
- Newsgroups: sci.math,sci.logic
- Subject: Re: The Continuum Hypothesis: Must it be {True or False}
- Message-ID: <1992Dec24.233312.18834@husc3.harvard.edu>
- Date: 25 Dec 92 04:33:11 GMT
- References: <1992Dec14.200024.6435@nas.nasa.gov> <1992Dec24.034938.11339@smsc.sony.com> <1992Dec24.161747.18827@husc3.harvard.edu>
- Organization: The Phallogocentric Cabal
- Lines: 48
- Nntp-Posting-Host: husc10.harvard.edu
-
- In article <1992Dec24.161747.18827@husc3.harvard.edu> I wrote:
-
- >FLT is a \Pi^0_1 sentence. In any first-order, or otherwise complete
- >theory, undecidability of a \Pi^0_1 sentence requires that it be false
- >in some models of the theory, but true in others; for otherwise by
- >semantic completeness it would have been provable or refutable,
- >according as it were valid or contradictory. The question of truth
- >reduces to whether the sentence can be false in a standard model, --
- >iff not, it is by definition true. Now, by Tarski's definition of
- >satisfaction, falsehood in the models of the theory depends on the
- >existence therein of a witness set for the \Sigma^0_1 negation of the
- >original sentence; but in case of an arithmetical sentence, these
- >witnesses must be the integers. If they are standard integers, then
- >they have names in the language of the theory, and therefore the
- >original sentence receives an explicit refutation in the latter, in
- >contradiction to the hypothesis of undecidability. On the other hand,
- >if all the witnesses in all the relevant models are *non-standard*
- >integers, then _a fortiori_ the proposition in question is true in all
- >standard models, and so true _simpliciter_, albeit not valid. Hence,
- >if a \Pi^0_1 sentence is undecidable in a semantically complete theory
- >of Peano Arithmetic, wherein provability is coincident with validity,
- >then it is _ipso facto_ true.
- >
- >On the other hand, in second-order PA, all models are standard by
- >categoricity; in other words, the adoption of a full strength
- >induction axiom allows us to characterize the integers up to
- >isomorphism. But then the completeness argument, as distinct from
- >soundness alone, is likewise inapplicable for standard second-order
- >semantics. (All we can get is that undecidability in second-order PA
- >entails falsehood in a faithful (satisfying AC and every instance of
- >comprehension) Henkin model. Henkin semantics is non-categorical, and
- >generally is equivalent to the case of many-sorted first-order logic,
- >so I think that the first-order situation is recapitulated in this
- >case.) But the moral of this story is that there is a critical
- >difference between undecidability in a first-order theory, which by
- >L\"owenheim-Skolem theorems cannot characterize infinite structures up
- >to isomorphism, and a higher-order theory, which can do so.
-
- To continue, suppose a \Pi^0_1 sentence is undecidable in second-order
- PA; suppose further that it is not valid, i.e. there is a model of
- second-order PA,in which it is false. By categoricity, such a model
- would contain standard integer witnesses to its negation, which could
- be used in an explicit refutation of the sentence in question.
- Contradiction.
-
- cordially,
- mikhail zeleny@husc.harvard.edu
- "Nothing can be said truly of what does not exist."
-