Frege's definition of the empty set was not a raw extensional one: he did not simply write the partial string {} and say, "That's it: that's the empty set." His account was more intensional: "There is a set that contains all and only true contradictions," but since per his LNC-compliant logic he would say that there are no true contradictions, it follows that the set in question is empty (has no elements that we actually quantify over, here).

Frege's definition has been generalized (or at least adapted) to other qualifications: one that I saw was, "The empty set is the set of all ill-founded sets," but since modulo ZFC there is a "fact" that V = WF (for WF = the class of well-founded sets), such a set would be empty. So another could be, "There is a set of all disordered sets," or, "There is a set of all sets that don't have powersets," etc. Inside a world where the axioms say that all sets have property f, that world's empty set will be the set of all sets that don't have f.

But so from outside of any world as such, things might appear quite different. In the literature, they usually frame transworld witness so that the emphasis is on how elements of V can vary from world to world: some versions of V have more or different elements than other versions, V can be extended by forcing, etc. and it is variation of content on this relatively maximal level that is the concern of theorists. However, per the Fregean option, we can also look at how each world's empty set (as opposed to its universal set/proper class) has more or different elements between worlds. If V = L, for example, then 0 "doesn't exist," neither do zero dagger, zero pistol, or any zero section that might be proposed. So in the constructible universe, the empty set = zero = the set of all the zero sections, which is empty only in constructible or sufficiently similar worlds but is teeming with elements from the perspective of other worlds.

The question: but suppose we were trying to be inclusive of finitism or even ultrafinitism, or at least respectful of their degree of caution with respect to infinitary reasoning, and so we wanted to introduce infinitesimals in such a way that they did not have to be reciprocals of actually-infinite numbers. Bassler[16] uses the perspicuous term parafinite to navigate these waters, so let us speak of parafinitesimals instead. Now, the relative elements of a Frege-flavored empty set could often by styled as parts/aspects/etc. of emptiness or the number zero, which despite its elementarity is fraught with nontrivial effects (e.g. the indeterminacy and undefinability of divison by zero, or the divergence between the Conway argument for 00 = 1 and the quasi-model where the expression is undefined). Would describing these "parts"/"aspects" as parafinitesimally small, as in parafinitesimally close to zero without necessarily being equal/identical to it, be a way to introduce "logical infinitesimals" without properly offending finitistic sensibilities? After all, even by the strict Fregean introduction, we have that infinitesimals as examples of "actual infinity" are then nonexistent elements of an empty set in a finitistic world where the closest counterpart to "actual infinity" is the class of all sets (which following the resolutions to Russell's paradox, is usually not taken for a set, even by non-finitist set theorists). Without claiming that there are infinitely many axioms/models buttressing the appeal to the multiversal standpoint, can a finitist allow for a finite number of universes (perhaps even some where actual infinity is countenanced), such that as transworld witnesses they can see into parafinitesimal or even "properly" infinitesimal elements of an empty set?

  • 1
    Frege's definition of the empty set was: "There is a set that contains all and only true contradictions," Reference, please. There were not sets for F, but classes and classes are "specified" by concepts. Thus, the null class is specified by the concept "to be different from itself", a concept that has no objects instantiating it. Jul 24, 2023 at 5:55
  • @MauroALLEGRANZA at that time, they sometimes used the word "class" (or cognates) for sets, like you can read translations of Cantor where the phrasing is "the first number class" and "the second number class" when he's referring to alephs. But as for Frege: Jul 24, 2023 at 14:10
  • "Recall that Frege defined the number 0 as the number of the concept being non-self-identical, and that 0 thereby becomes identified with the extension of all concepts which fail to be exemplified." (SEP) I have seen the phrasing as "set of true contradictions" instead of "set of violations of the law of identity" and we normally allow the required conversion anyway (per classical/similar logic). Jul 24, 2023 at 14:11
  • Ok thanks; I'm quite sure that you cannot find a reference for the "set conception" in Frege :-) Jul 24, 2023 at 14:17
  • see Frege’s Analysis of Natural Numbers: The following series of concepts [derive from Frege’s analysis of numbers: C0 = [λx x≠x] ... Then Frege defines the finite cardinal numbers, as follows: 0 = #C0. Jul 24, 2023 at 14:26

2 Answers 2


You raise some intriguing technical points about the metaphysical status of infinitesimals and empty sets from a multiversal finitistic perspective. However, I think there are a few conceptual issues worth clarifying first:

  • Frege defined the empty set extensionally as the set with no elements, not intensionally as "the set of all contradictions" which could imply it has contradictory elements.

  • Infinitesimals are generally mathematical entities defined relative to a particular formal system or model, not objects that meaningfully "exist" in an absolute metaphysical sense across worlds.

  • Multiverse theories in physics and philosophy are highly speculative and contentious. Invoking them may obfuscate rather than clarify the core issues.

  • Ultrafinitism rejects even relatively small infinite numbers, so positing infinitesimals may still be objectionable from that stance.

My inclination would be to avoid metaphysically loading mathematical entities with too much "transworld existence" baggage in the first place. Formal systems are abstractions created to model certain aspects of reality, not fully describe its ultimate nature.

Perhaps a more straightforward way forward is focusing the discussion on which formal frameworks are most relevant and useful for different mathematical purposes from a finitistic perspective, without getting into metaphysical disputes over the ontological status of their entities across possible worlds. This could lead to more fruitful engagement.

In summary, while technical details matter, the bigger issue seems to be mutual understanding and bridging divergent philosophical assumptions. Establishing common ground on appropriate standards of rigor may be more productive than debates over the precise metaphysics of empty sets. But I'm happy to discuss further if you feel I'm missing key points!

  • I would like to note that Frege did use the "weird" definition of the empty set: "zero corresponds to the class of objects that are unequal to themselves." Maybe this was a new definition vs. something he wrote before, or maybe he changed to a new definition later, IDK. But yeah, he was a logicist and so wanted something special to characterize the empty set, and his method has been used in other ways, for transfinite cardinal types that a theory rules out: these end up as empty-set "elements" (which don't exist!). It's a weird picture, I know, and I don't accept it blindly, but... Jul 23, 2023 at 17:08

This is a linguistics problem. If we are defining infinite expressions using finite indicators (like π, α) then we cannot describe this as proof of infinity. Until now, neither infinite nor infinitesimals have been defined. But you can coverage your expressions with "{" and "}" characters.

  • I think part of the instability came from the subconscious oscillation between the 1/A concept of infinitesimals as reciprocals of infinite A's, and Euler's 0/0 intuition, but these "objects" (terms) have been defined in other ways since, with the same bases in some sense (mostly), so technically there are definitions of terms that count as infinite reciprocals, etc. For example, the nilpotence definition can vary across options. And we freely speak of hyperreal and surreal infinitesimals, thanks to Robinson and Conway. Jul 23, 2023 at 17:11
  • Ontolojik ile lojistik yapı sınıflandırmaları birbirlerine asimetriktir. Bahsettiğiniz şeyin gerçekliklerarası sistem transformasyonu sürecinin olup olmaması olarak algılıyorum. Gerçekliğimiz kapsamında bir cevher, yani bir küçük, tek başına bir sistem oluşturamaz. Bu nedenle sonlu bir kapsam içerisinde, teleolojik sabit olan omeganın cevheri böldüğü hakkında bir izdüşümüne sahip isek, en büyük pozitif tamsayının 1 olduğunu değerlendiremeyecek kadar omeganın işlevselliğini tanımlayamayız. Tüm çelişkileri kapsayan Frege'nin boş kümesi, işlevsel olmamasıyla sistem oluşturamaz...
    – fkybrd
    Jul 23, 2023 at 22:00

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .