# Why should the underlying set of a model be a set?

In logic (especially in model theory and set theory), one defines a model $\mathcal A$ for a given signature $S$ to be given by a underlying set $A$ and some additional structure (by which I mean: for every constant symbol in $S$ an element of $A$, for every $n$-ary function symbol in $S$ a function $A^n\to A$, and so on). In set theory, one often considers models satisfying some set theory axioms (often one takes ZFC); these models have the form $(V, \in)$, where $V$ is called a universe and $\in$ a binary relation on $V$.

I find it philosophically unsatisfying that set-theoretic universes should be sets, since within a set-theoretic universe satisfying ZFC, this universe thinks that it is a proper class, if you know what I mean.

Why does one not allow models/universes to be proper classes?

Why are the set-theoretic universes sets?

• This answer may be relevant to your question: math.stackexchange.com/a/1368646/630 – Carl Mummert Aug 14 '16 at 16:05
• I wonder whether this question is a duplicate of math.stackexchange.com/q/56726/630 . I am not voting at the moment, to see if others agree. – Carl Mummert Aug 14 '16 at 16:07
• @Carl I don't think it's a duplicate of that question. – Stefan Mesken Aug 14 '16 at 16:11
• @CarlMummert Definitely strongly related. However the other question partly comes across as not understanding how a set theory can have a model at all, and is more specific to set theory. This question is more about why we insist models be sets more generally. – 6005 Aug 14 '16 at 16:19

One of the reasons is that the $\sf ZFC$ axioms are not equipped for handling classes properly. For example $\sf ZF$ proves that all the axioms of $\sf ZFC$ hold in the inner model $L$, which is a proper class. But we cannot define a truth definition for that class, as that would violate Tarski's theorem. And the proof that all the axioms hold in $L$ is not a single proof, but rather a schema of proofs.

Sure, we can handle some class models relatively okay (e.g. the surreal numbers, which have a relatively simple theory), but arbitrarily speaking? We want our foundational theory to have access to the truth definition of a structure. Otherwise, what sort of foundation does the theory provide us?

If you notice, almost all the statements about class models of set theory are meta-theoretic statements.

Yes, we could use class-theories like $\sf KM$ (Kelley-Morse) and $\sf NBG$ (von Neumann-Goedel-Bernays), and this might allow us to extend the reach of truth definitions for various class models; but certainly not for the entire universe, as that would contradict Tarski's theorem about the undefinability of the truth. So the universe of set theory will, essentially, always be an internal class.

• Thanks. But I already do not understand your first sentence: aren't the axioms of Morse-Kelley set theory enough to handle classes? – Treppe Aug 14 '16 at 16:14
• @Rene: I think you were looking at the wrong place. The general idea is that now the Levy-Montague reflection theorem is not a meta-theorem, but rather a theorem, when applied to statements involving only sets. So in particular, you get a uniform definition for a truth predicate. This is the same trick as proving that in full second-order arithmetic we can define a first-order truth predicate. – Asaf Karagila Aug 14 '16 at 17:06
• "A collection of natural numbers which is a proper class"? Doesn't MK, like NBG, have a rule that the intersection between a set and a class is a set? (That's what I got from Mendelson's brief description of it. anyway). – Henning Makholm Aug 14 '16 at 17:13
• @Henning: I think you're right (at least if you include the "Limitation of size" which appears in the Wikipedia entries). I misunderstood some things. My point, however, that the truth definition is accessible as a class, not as a set. In the sense that every parameter-free definition for that set has to appeal to proper classes. – Asaf Karagila Aug 14 '16 at 17:31
• @AsafKaragila: It's not even limitation of size -- $\text{set}\cap\text{class}=\text{set}$ is simply how the Axiom of Separation looks in NBG. I think the real point here is that what MK can define truth for is just formulas where all quantifiers are restricted to sets (that is, the language of ZFC), whereas MK still can't define truth for its own full language. So seen from MK, a class model of ZFC is still a restricted notion the same way considering only set models within ZFC is a restricted notion. – Henning Makholm Aug 14 '16 at 17:46