The virtual theory of classes is a partial counterfeit of set theory fashioned purely of logic(stlix)
The virtual theory of classes eventually gets merged with the real theory in such a way as to produce a combination which, though not strictly more powerful than the real theory alone would be, is smoother in its running(stlix)
There are advantages (and disadvantages) in holding with John von Neumann and perhaps Cantor that not all classes are capable of being members of classes. In theories that hold this, the excess vocabulary has come in handy for marking the distinction; classes capable of being members are called sets(stl3)
Ultimate classes are not being members in turn of further classes. We can know this technical sense of 'set' and still use the terms 'set' and 'class' almost interchangeably. For the distinction emerges only in systems that admit ultimate classes, and even in such systems the classes we have to do with tend to be sets rather than ultimate classes until we get pretty far out(stl3)
The natural attitude on the question what classes exist is that any open sentence determines a class. Since this is discredited, we have to be deliberate about our axioms of class existence and explicit about our reasoning from them; Intuition is not in general to be trusted here(stl5)
The distinction between set and ultimate class is not to be confused with that between real and virtual class. Ultimate classes, for theories that admit them, are real: they belong to the universe of discourse, they are values of quantifiable variables. The virtual theory of classes, on the other hand, does not invoke classes as values of variables; it talks much as if there were classes, but explains this talk without assuming them(stl20)
Willard Quine's virtual classes are precisely Kurt Godel's "notions", apart from a couple points of style. One point is that unlike Gödel Quine talks of them as if of classes, subjecting them to class operations and the like(stl21)
A motive for talking thus ostensibly and eliminably of classes and relations is compactness of expression(stl26)
The acceptance of classes as members of further classes adds very significantly to what can be said about numbers and other mathematical objects(stl29)
What constitutes them individuals is not inclassitude, but identity with their unit classes (or, what comes to the same thing, identity with their own sole members)(stl32)
Even the many theorems which, because they are governed by universal rather than existential quantifiers, we can prove without axioms or premises of existence, would lose all potential content and interest if classes were excluded(stl38)
We are committed to there being at least one class. For we are assuming the classical logic of quantification, which, as is well known, assumes there to be at least one thing(stl38)
The notation of the theory of classes suffices for the arithmetic of natural numbers, and also suffices for further sorts of numbers(stl119)
A certain axiom schema may not imply there are Infinite classes; however, a certain axiom schema may state that given certain ones, other infinite classes must accompany them(stl164)
For belonging to a class, fulfilling the membership conditions is not enough; existence is required(stl170)
A class is a set, that is, if no bigger than some set; so John von Neumann. But in the end he settles on a stronger axiom of replacement, whereby a class is a set if and only if it is not as big as the class of all sets(stl311)