The statement "ZFC $\vdash 0=1$" is independent of ZFC due to Goedel's second incompleteness theorem. That got me wondering, for what other statements $\phi$ is "ZFC $\vdash \phi$" independent of ZFC?

Now of course, for any statement $\phi$ for which "ZFC $\nvdash \phi$" is true, then "ZFC $\vdash \phi$" is independent of ZFC. So instead, I'll ask which "ZFC $\vdash \phi$" are independent of ZFC + Con(ZFC), and stronger theories?

(Of course, easy examples are Con(T) for T=ZFC+Con(ZFC) or stronger theory, but are there others?)

  • $\begingroup$ Independent does not mean "unprovable" but "neither provable nor disprovable". $\endgroup$
    – Asaf Karagila
    Nov 10, 2017 at 21:00
  • $\begingroup$ @AsafKaragila I'm aware. Does something in my question suggest that I meant unprovable? $\endgroup$ Nov 10, 2017 at 21:02
  • $\begingroup$ Read the first sentence of the second paragraph. It literally says that you mean "unprovable". $\endgroup$
    – Asaf Karagila
    Nov 10, 2017 at 21:02
  • 1
    $\begingroup$ And also, now you are asking for independence over $\text{ZFC}+\text{Con}(\text{ZFC})$, rather than merely independence over $\text{ZFC}$. These are not the same thing, and that change affects everything I have said so far. $\endgroup$ Nov 10, 2017 at 23:21
  • 1
    $\begingroup$ @JoelDavidHamkins It shouldn't be interpreted as"statements independent of (ZFC + Con(ZFC) and stronger theories) but as "statements independent of (ZFC + Con(ZFC)) and statements independent of (stronger theories)". Each theory gives a separate question. $\endgroup$ Nov 10, 2017 at 23:25

2 Answers 2


You seem to use "independent of $T$" to mean "unprovable in $T$", so I'll interpret the question that way (not as "neither provable nor refutable in $T$).

If $ZF\vdash\phi$ is true, then it can be proved in ZF and in fact in much weaker systems, just by taking a proof of $\phi$ in ZF and verifying that it is indeed a proof (and ends with $\phi$). Conversely, if $ZF\vdash\phi$ is false, then it can't be proved in ZF or in ZF + Con(ZF) or in any arithmetically sound theory. (My Platonism is showing here, as I take it for granted that ZF + Con(ZF) is arithmetically sound because it's true.) So $ZF\vdash\phi$ is unprovable in ZF + Con(ZF) iff it is false, i.e., if $\phi$ is unprovable in ZF.

  • 4
    $\begingroup$ @PyRulez: How do you what is true in true arithmetic? Is Con(ZFC) true? Is Con(ZFC+Con(ZFC)) true? is Con(ZFC+there is a measurable cardinal) true? Is Con(ZF+DC+AD) true? I can continue for more or less forever with these questions (and for each answer I would probably ask a follow up of "Why? How do you know that?"), at some point you're going to have to give me a better answer than just "true in true arithmetic". $\endgroup$
    – Asaf Karagila
    Nov 10, 2017 at 21:05
  • 6
    $\begingroup$ @AsafKaragila I do not see why anything more is needed. An arithmetic statement is true if it holds in the standard model, that is the standard interpretation of the term. Maybe I do not understand what you are asking. $\endgroup$ Nov 10, 2017 at 21:07
  • 3
    $\begingroup$ @AsafKaragila Of course. $\endgroup$ Nov 10, 2017 at 21:11
  • 3
    $\begingroup$ I agree with @NoahSchweber's last comment about Platonism, and I'd add that, when I say that an arithmetical statement is true, I mean exactly what most mathematicians who are not logicians mean by that. $\endgroup$ Nov 11, 2017 at 2:10
  • 5
    $\begingroup$ @Qfwfq: I would say it is not. Most non-logician mathematicians have an even firmer sense of the natural numbers "as a structure", and when they say "true" they mean "true in that structure". At the same time few know much of anything about formal logic. // Separately, phrases such as "true in PA" are somewhat painful to read, even if I know what is meant, because they mix truth with provability in a way that the two should not mix. $\endgroup$ Nov 11, 2017 at 18:40

I think you're asking what happens if you iterate adding consistency statements. There's a whole book about that, "Inexhaustibility" by Torkel Franzen, though it starts with arithmetic rather than ZFC. This article is also good:



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.