[FOM] ZFC, the Universe, and Replacement

rgheck rgheck at brown.edu
Mon Feb 1 19:00:46 EST 2010

Much of what Jeremy has been saying seems to focus on the question
whether the acceptance of ZFC depends upon one's being able in some
sense to characterize the universe, V, of sets. But this thought can
be questioned. as it was by George Boolos in such papers as "On 
Second-order Logic", "The Iterative Conception of Set", "Iteration 
Again", and especially "To Be Is To Be a Value of a Variable" and 
"Nominalist Platonism". The issue is also discussed in detail in Richard 
Cartwright's paper "Speaking of Everything", and there are related 
model-theoretic discussions in papers by Agustin Rayo, Gabriel Uzquiano, 
Van McGee, and Timothy Williamson, among others. (Sorry if I've left 
anyone out!)

In short, the response is that we should not suppose (a) that the 
motivation for ZFC somehow depends upon a characterization of the 
"totality" of sets and (b) we absolutely should not suppose that it 
depends upon the presumption that the "totality" so characterized (the 
"domain" of the theory) itself has to *be* a set. Jermey's emphasis on 
the claim that we have no good reason to believe in the existence of a 
"model" of ZFC, where this would be a *set* with certain properties, 
suggests rather strongly that the advantages he claims for ZC rest upon 
(a) and (b) or claims to the same effect.

It is an old idea, going back to the very origins of set-theory, that 
neither of these should be accepted, though Cantor's own reasons for 
rejecting (a) and (b) seem to have been a little peculiar, or at least 
to have had some peculiar elements. As regards (a), the thought is that 
a purely generative conception, as developed particularly in Scott's 
axiomatization of `stage theory', is adequate to motivate ZFC. As 
regards (b), the thought is that this is just wrong, for reasons 
Cartwright expresses particularly forcefully.

There are ways of developing some of these ideas, via notions like that 
of "indefinite extensibility", that can lead to a somewhat contructivist 
way of thinking, but there are plenty of people who think of the 
set-theoretic universe as being in some sense "open-ended" who do not 
have particularly constructivist leanings. Nor, I should add, am I 
actually saying that one has to think of the set theoretic universe as 
open-ended to reject (a) and (b). But it's a natural direction to take.

Somewhat ironically, later in his life Boolos expressed doubts about the 
extent to which a "generative" motivation could be given for 
replacement. He suggested, in "Iteration Again", I believe, that the 
intutions that drive replacement might better be understood as based 
upon a conception of set bound up with `limitation of size' (a very 
Cantorian idea). And he came to believe, toward the end of his life, 
that maybe replacement wasn't very well motivated at all. But his 
reasons were quite different and were ultimately based upon little more 
than revulsion at the very idea that there might be objects as large as 
those ZFC churns out. To quote:

...[J]ust what exactly is the matter with saying ZFC isn't correct 
because it tells us that there are [lots] of objects and there aren't 
that many objects? ...To be sure, one who says this may be asked how he 
knows there aren't. But the reply, "Get serious. Of course there aren't 
that many things in existence. I can't *prove* that there aren't, of 
course, any more than I can *prove* that therea aren't any spirits shyly 
but eagerly waiting to make themselves apparent when the Zeitgeist is 
finally ready to acknowledge the possibility of their existence. But 
there aren't any such spirits and there aren't as many things around as 
[that]. You know that perfectly well, and you also know that any theory 
that tells you otherwise is at best goofball."---that reply, although it 
does not *offer reasons* for thinking that there are fewer than [that 
many] objects in existence, would not seem to manifest any illusions 
that could be called metaphysical realist. ["Must We Believe in Set 
Theory", in *Logic, Logic, and Logic*, p. 145]

The remark about metaphysical realism isn't to the present point, but 
one gets the idea.

For those who are wondering, the specific number Boolos has in mind is 
the least ordinal \lambda for which \lamba = \aleph_\lambda. ZFC of 
course entails the existence of such an ordinal.


Richard G Heck Jr
Romeo Elton Professor of Natural Theology
Brown University

More information about the FOM mailing list