There are several definitions of "finite", like Dedekind's and Tarski's (Thanks to A.K. for point out the latter - first time I've heard of it): From the Wikipedia entry on Finite Set:

(Richard Dedekind) Every one-to-one function from S into itself is onto.

(Alfred Tarski) Every non-empty family of subsets of S has a minimal element with respect to inclusion.

And there are more.

Do these distinctions matter when considering definitions of fundamental categories like topological spaces or measure spaces? For example,

In topology: Why do we require a topological space to be closed under finite intersection?

In measure theory: "A measure is continuous from above if [given measurable sets and closed under intersection] at least one set has finite measure," alternatively, in the definition of sigma-finite measures (same article).

Do the various definitions of finiteness lead to non-isomorphic or non-equivalent categories?

My question is not specifically about topology or measure theory, but these are basic definitions introduced at undergrad level, so I thought, better to understand the context via basic examples.

  • $\begingroup$ 1) For those of us not in the know, can you provide a reference to these definitions of "finite" per Dedekind or Tarski? 2) Not being familiar with either of these definitions, is there any example of a set which is "finite" in one definition and not in another? $\endgroup$
    – NKS
    Jan 23, 2013 at 4:54
  • $\begingroup$ @NKS, added Wikipedia links $\endgroup$ Jan 23, 2013 at 4:58
  • $\begingroup$ @NKS I believe there are (at least) 16 different definitions of 'finite set' and no two are equivalent. $\endgroup$ Jan 23, 2013 at 4:59
  • 3
    $\begingroup$ It should be pointed out that all these definitions are equivalent under the axiom of (countable) choice. $\endgroup$
    – Asaf Karagila
    Jan 23, 2013 at 7:47
  • $\begingroup$ @AsafKaragila, although I find Choice interesting, I am primarily interested in constructive mathematics. Green & Tao in their arithmetic progressions paper write that they carefully avoid Choice, like a plague of some sort. $\endgroup$ Jan 23, 2013 at 21:45

1 Answer 1


The various non-equivalent notions of finiteness you allude to refer to definitions of 'finite set'. The concepts of finiteness you mention that are relevant to analysis (in general) refer to a number being finite. It is common-place in analysis to assume the standard models for Peano Arithmetic and the real numbers. Both of these are categorical models (that means that any two models are isomorphic). Then, all the elements in a model of PA are finite and all the elements in $\mathbb {R}$ are finite as well. It is common to consider the extended real numbers $\mathbb {R} \cup \{\pm \infty\}$. Then $\pm \infty$ are by convention infinite numbers.

So, there is little (if any) relation between the different definitions of 'finite set' and the notion if measure theory of a set having finite measure or in topology for taking finite intersections. Finite in the former means a real number (and not $\pm \infty$) while in the latter it means 'perform intersection a finite number of times, counted using some natural number'.

  • $\begingroup$ I've seen this categorical defintion of Dedekind finite, no reference to sets: "An object A in a category C is Dedekind finite if all monos $j:A \to A$ are isomorphisms" mathoverflow.net/questions/66096/a-dedekind-pseudo-finite-set - correct? $\endgroup$ Jan 23, 2013 at 5:08
  • $\begingroup$ yes, that's a possibility. Still, this makes no difference for finiteness issues of natural numbers or real numbers. $\endgroup$ Jan 23, 2013 at 5:15
  • $\begingroup$ But what about for the spaces at issue, which are not numbers? Unions and intersection can be indexed by arbitrary functions right? $\endgroup$ Jan 23, 2013 at 5:20
  • 1
    $\begingroup$ I'm not an expert on finiteness but as far as I know this has no implications to analysis. You are talking about performing a certain binary operation a finite number of times. Finite in this case means counted by some natural number, and these are categorical so there is only one meaning to that. $\endgroup$ Jan 23, 2013 at 5:25
  • 1
    $\begingroup$ as far as I know in the presence of AC all these different notions of finite are equivalent. It is likely that in intuitionistic mathematics such issues as you describe are important. At least in standard analysis AC is accepted. $\endgroup$ Jan 23, 2013 at 5:42

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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