Though I've understood the logic behind's Russell's paradox for long enough, I have to admit I've never really understood why mathematicians and mathematical historians thought it so important. Most of them mark its formulation as an epochal moment in mathematics, it seems to me. To my uneducated self, Russell's paradox seems almost like a trick question, like a child asking what the largest organism on Earth is, and getting back an answer talking about a giant fungus.

"Well, OK..."

The set of all sets admits all sorts of unmathematical objects right? Chickens, spoons, 2 $\times$ 4s, what have you...I can't imagine a situation where a mathematician solving a concrete problem would invoke a set like that. The typical sets one sees in everyday mathematics are pretty well defined and limited in scope: $\mathbb{Z}$, $\mathbb{R}$, $\mathbb{C}$, $\mathbb{R}^n, \mathbb{Q}_p$, subsets of those things. They are all built off each other in a nice way.

But clearly important people think the result significant. It prompted Gottlob Frege, a smart and accomplished man, who Wikipedia tells me extended the study of logic in non-trivial ways, to say that the "foundations of [his] edifice were shaken." So I suppose Russell's paradox is "deep" somehow? Is there a practical example in mathematics that showcases this, where one might be led horribly astray if one doesn't place certain restrictions on sets, in say, dealing with PDEs, elliptic curves, functional equations? Why isn't Russell's paradox just a "gotcha"?

  • 2,362
  • 17
  • 22
  • 3,211
  • 5
  • 23
  • 24
  • 2
    I think the problem was that the set-theory back then allowed the existence of a set of all sets that did not contain itself, in the way that it did. And it obviously makes an inconcistency in the (naive) theory of sets. I am not sure about any direct practical significance this has, but it is not a good starting point for a mathematical theory. – Eivind Mar 01 '11 at 23:19
  • 34
    Russell's paradox highlighted the problem with unfettered Comprehension. Frege was trying to develop a logical system that required him to consider a kind of unrestricted comprehension, and Russell's example shows that the issue is with comprehension. But the idea is actually very clever: it lies behind the proof that given a set $A$ there is always $B\notin A$, and to some extent behind the proof of Cantor's Theorem that no set can be bijected with its power set. – Arturo Magidin Mar 01 '11 at 23:29
  • 37
    By the way, we *still do* take Russell's Paradox seriously, or at least most of us do. Sometimes you encounter passages like "we ignore set-theoretic difficulties" in math books and papers, but in my opinion this translates to "We can't be bothered to say only things that are completely correct. We are hoping that you are either so unsophisticated as not to notice or so sophisticated as to be able to fix what we've said more easily than we could." I.e., it is somewhere between lazy and irresponsible. – Pete L. Clark Mar 01 '11 at 23:57
  • 28
    I'd mildly disagree with Pete, in that ignoring set-theoretic issues doesn't have to be laziness or irresponsibility. It could mean the author thinks the exposition is deemed easier to digest when one ignores such issues -- there's perhaps a small implication that filling in such details isn't difficult so it's a tolerable lack of formality for what is seen as an expositional benefit. It's much like how all written proofs in journals are not formal proofs, but just enough details so that one could fill in a formal proof using twenty times the number of characters. – Ryan Budney Mar 02 '11 at 04:16
  • 1
    @Ryan: choosing to elide details is often a good expository choice. But "we ignore set-theoretic difficulties" doesn't mean (or so I had intended) leaving out details. It means that pretending that certain things are true when they actually are not. In my opinion, it is somewhat irresponsible to do that for any reason. And I think that some people who write in this way believe in principle that "getting it right" is just a fussy nuisance, but years of doing such ignoring have given them poor intuition about what is easily fixable... – Pete L. Clark Mar 02 '11 at 05:18
  • 1
    For instance, I have seen in a lecture someone attempt to take a direct sum indexed over all abelian groups (in a situation where it was not clear that any of the objects needed to be zero). I sort of cleared my throat and explained that that just didn't make any sense to me, and the speaker did not in fact know how to correct himself (or even really understand what the problem was). That's lazy or irresponsible, IMO. – Pete L. Clark Mar 02 '11 at 05:21
  • 1
    @Arturo Magidin: Despite what's already been said, I agree with the intent of the title question...shouldn't people have responded immediately 'Yes, of course that's a contradiction, so that means the described set (the set of all sets) just doesn't exist' (and so no need to worry further) ? There's no contradiction in -math-, just a contradiction in that particular statement. Yes, so comprehension can't be applied successfully arbitrarily, but that doesn't mean 'math' falls apart. – Mitch Mar 02 '11 at 19:16
  • 20
    @Mitch: First, the very notion that there was such a thing as a set that could not exist was **entirely new**; up to then, with unlimited comprehension, the idea was that *anything* could be a set. So in fact, Russell's paradox introduces an entirely new idea: that there are sets that just plain don't exist. You say, "fine, then let's move on". Okay, but the problem is that there was a lot of work that had been done or was ongoing, in which sets were used *without* checking if the thing described actually existed or not (cont...) – Arturo Magidin Mar 02 '11 at 20:08
  • 17
    @Mitch: (cont...) So, since there are some things we can describe but which turn out to *not* be sets, then anything that was done without worrying about whether they were sets or not has to be reviewed and checked. Dedekind's construction of the real numbers as Dedekind cuts... oops, need to check if those things exist or not; Ideals and modules in algebraic number theory... oops, need to check if those things exist or not; geometry arguments that talk about "all points" or things like that... oops, need to check if they actually exist or not. (cont...) – Arturo Magidin Mar 02 '11 at 20:10
  • 11
    @Mitch (cont...) Finally: remember the context. Some people were actively worried about the fact that arguments were leading to antinomies and contradictions, and there was an active attempt to clean up the foundations. Because a lot of *math* was built on arguments that involved ideas closely related to sets, all those arguments needed to be checked. Just because the Russell sets seems unnatural does not mean that "natural-looking sets" will be okay. (Goedel's sentence is unnatural, but there are "natural looking" sentences that are formally undecidable). – Arturo Magidin Mar 02 '11 at 20:12
  • 4
    @Mitch (I wonder if, when we get to three long comments, I shouldn't just post a community wiki answer instead...) This reminds me of a quote; it describes a castle on a cliff where, after every great storm, the spiders rush to rebuild their torn spiderwebs, afraid that if they don't the entire castle will fall down. The quote is taken from a book on the logical foundations of mathematics... It's likely that to most working mathematicians, Russell's paradox was ignored much like the problems with infinitesimals in calculus were; but for those who were interested in foundations, it was big. – Arturo Magidin Mar 03 '11 at 03:36
  • @Arturo Magidin It's not too late! This question is just begging for the expansive tour-de-force type answer you've become known for -- that's partly why I left it open ;-) – Uticensis Mar 03 '11 at 04:10
  • @Billare: Ah, flattery will get you everywhere. Anyway, I've repeated some of the stuff above, and added some other personal impressions. – Arturo Magidin Mar 03 '11 at 05:53
  • See also this MathOverflow answer: http://mathoverflow.net/questions/18659/set-theory-in-practice – Uticensis Mar 08 '11 at 23:32
  • Apparently I do not have *50 reputation* to be allowed to add a meta-comment. But I would like to say that I agree with Arturo Magidin and that I found a similar quote about the metaphor of the spiders on a review by Martin Davis about some book of mathematical logic. *On the banks of the Rhine a beautiful castle had been standing for centuries. In the cellar of the castle an intricate network of webbing had been constructed by the industrious spiders who lived there. One day a great wind sprang up and destroyed the webs. Frantically the spiders worked to repair the damage. For you see, they t – user25296 Feb 05 '15 at 08:58
  • I'm not sure if this has been mentioned, but there is also the fact that in classical logic -- the logic that most mathematicians use, knowingly or not -- we have the principle of explosion: from a contradiction, we can deduce anything. So if we admit the Russell set and Russell's paradox, how do we know that any theorems we've proven aren't just $r \in r \land r \notin r \vdash P$ in disguise. – Miles Rout Sep 12 '16 at 22:49

16 Answers16


Russell's paradox means that you can't just take some formula and talk about all the sets which satisfy the formula.

This is a big deal. It means that some collections cannot be sets, which in the universe of set theory means these collections are not elements of the universe.

Russell's paradox is a form of diagonalization. Namely, we create some diagonal function in order to show some property. The most well-known proofs to feature diagonalization are the fact that the universe of Set Theory is not a set (Cantor's paradox), and Cantor's theorem about the cardinality of the power set.

The point, eventually, is that some collections are not sets. This is a big deal when your universe is made out of sets, but I think that one really has to study some set theory in order to truly understand why this is a problem.

Mark Hurd
  • 688
  • 1
  • 8
  • 21
Asaf Karagila
  • 370,314
  • 41
  • 552
  • 949
  • 6
    Note that Cantor showed that the powerset of $X$ is strictly larger than $X$ itself before Russell came up with the paradox. You can view Russell's paradox as a statement saying that there are more subclasses of the universe than sets. In some sense the paradox has an even easier proof than Cantor's theorem, since you don't have to deal with an arbitrary map assigning a class to each set. The map is already given, it is the identity. So really what Russell showed is that the identity map of the universe is not onto the "collection" of all classes. – Stefan Geschke Sep 20 '11 at 10:22
  • 5
    I have studied a bit more history since writing this answer. Russell's paradox was actually proved by Zermelo, as I have learned not too long ago. – Asaf Karagila Dec 29 '11 at 11:30
  • @Stefan: I recall that Russell came up with his paradox specifically by considering Cantor's power set theorem in the context of Frege's *Grundgesetze der Arithmetik*. – MJD Mar 28 '12 at 14:00
  • 5
    About the well-known proofs by diagonalization, the halting problem is also a very important one. – Denis Jan 28 '14 at 13:09

My dad likes to tell of a quotation he once read in a book on philosophy of mathematics. He does not remember which book it was, and I have never tried to track it down; this is really hearsay to the fourth degree, so it may not even be true. But I think it is pretty apt. The quote describes a castle on a cliff where, after each a storm finally dies down, the spiders come out andrun around frantically rebuilding their spiderwebs, afraid that if they don't get them up quickly enough the castle will fall down.

The interesting thing about the quote was that it was attributed to a book on the logical foundations of mathematics.

First, note that you are looking at the problem from a perspective of someone who "grew up" with sets that were, in some way, carefully "built off each other in a nice way." This was not always the case. Mathematicians were not always very careful with their foundations: and when they started working with infinite sets/collections, they were not being particularly careful. Dedekind does not start from the Axiom of Infinity to construct the naturals and eventually get to the reals; and moreover, when he gives his construction it is precisely to try to answer the question of just what is a real number!

In some ways, Russell's paradox was a storm that sent the spiders running around to reconstruct the spiderwebs. Mathematicians hadn't been working with infinite collections/sets for very long, at least not as "completed infinities". The work of Dedekind on the reals, and even on algebraic number theory with the definitions of ideals and modules, was not without its critics.

Some mathematicians had become interested in the issues of foundations; one such mathematician was Hilbert, both through his work on the Dirichlet Principle (justifying the work of Riemann), and his work on Geometry (with the problems that had become so glaring in the "unspoken assumptions" of Euclid). Hilbert was such a towering figure at the time that his interest was itself interesting, of course, but there weren't that many mathematicians working on the foundations of mathematics.

I would think like Sebastian, that most "working mathematicians" didn't worry too much about Russell's paradox; much like they didn't worry too much about the fact that Calculus was not, originally, on solid logical foundation. Mathematics clearly worked, and the occasional antinomy or paradox was likely not a matter of interest or concern.

On the other hand, the 19th century had highlighted a lot of issues with mathematics. During this century all sorts of tacit assumptions that mathematicians had been making had been exploded. Turns out, functions can be very discontinuous, not just at a few isolated points; they can be continuous but not differentiable everywhere; you can have a curve that fills up a square; the Dirichlet Principle need not hold; there are geometries where there are no parallels, and geometries where there are an infinite number of parallels to a given line and through a point outside of it; etc. While it was clear that mathematics worked, there was a general "feeling" that it would be a good idea to clear up these issues.

So some people began to study foundations specifically, and try to build a solid foundation (perhaps like Weierstrass had given a solid foundation to calculus). Frege was one such.

And to people who were very interested in logic and foundations, like Frege, Russell's paradox it was a big deal because it pinpointed that one particular tool that was very widely used led to serious problems. This tool was unrestricted comprehension: any "collection" you can name was an object that could be manipulated and played with.

You might say, "well, but Russell's paradox arises in a very artificial context, it would never show up with a "real" mathematical collection." But then, one might say that functions that are continuous everywhere and nowhere differentiable are "very artificial, and would never show up in a 'real' mathematical problem". True: but it means that certain results that had been taken for granted no longer can be taken for granted, and need to be restricted, checked, or justified anew, if you want to claim that the argument is valid.

In context, Russell's paradox showed an entirely new thing: there can be collections that are not sets, that are not objects that can be dealt with mathematically. This is a very big deal if you don't even have that concept to begin with! Think about finding out that a "function" doesn't have to be "essentially continuous" and can be an utter mess: an entirely new concept or idea; and entirely new possibility that has to be taken into account when thinking about functions. So with Russell's, an entirely new idea that needs to be taken into account when thinking about collections and sets. All the work that had been done before which tacitly assumed that just because you could name a collection it was an object that could be mathematically manipulated was now, in some sense, "up in the air" (as much as those castle walls are "up in the air" until the spiders rebuild their webs, perhaps, or perhaps more so).

If nothing else, Russell's paradox creates an entirely new category of things that did not exist before: not-sets. Now you think, "oh, piffle; I could have told them that", but that's because you grew up in a mathematical world where the notion that there are such things as "not-sets" is taken for granted. At the time, it was the exact opposite that was taken for granted, and Russell's paradox essentially tells everyone that something they all thought was true just isn't true. Today we are so used to idea that it seems like an observation that is not worth that much, but that's because we grew up in a world that already knew it.

I would say that Russell's paradox was a big deal and wasn't a big deal. It was a big deal for anyone who was concerned with foundations, because it said "you need to go further back: you need to figure out what is and what is not a collection you can work with." It undermined all of Frege's attempt at setting up a foundation for mathematics (which is why Frege found it so important: he certainly had invested a lot of himself into efforts that were not only cast into doubt, but were essentially demolished before they got off the ground). It was such a big deal that it completely infuses our worldview today, when we simply take for granted that some things are not sets.

On the other hand, it did not have a major impact on things like calculus of variations, differential equations, etc., because those fields do not really rely very heavily on their foundations, but only on the properties of the objects they are working with; just like most people don't care about the Kuratowski definition of an ordered pair; it's kind of nice to know it's there, but most will treat the ordered pair as a black box. I would expect most of them to think "Oh, okay; get back to me when you sort that out." Perhaps like the servants living on the castle not worrying too much about whether the spiders are done building their webs or not. Also, much like the way that after Weierstrass introduced the notion of $\epsilon$-$\delta$ definitions and limits into calculus, and then re-established what everyone was using anyway when they were using calculus, it had little impact in terms of applications of calculus.

That rambles a bit, perhaps. And I'm not a very learned student of history, so my impressions may be off anyway.

Arturo Magidin
  • 356,881
  • 50
  • 750
  • 1,081
  • 5
    Excellent...thanks for expositing...that was my main concern, that I was having 20-20 hindsight, unaware of the conditions of thought at the time. – Mitch Mar 03 '11 at 15:13
  • 3
    But I take issue with the metaphor, in that it connotes that foundations are -totally- useless. Practically, it's easy to see that bridges are staying up (mostly) without them (the Babylonians did OK with mostly unproven facts, not needing the foundations of Euclid's axiom system). But one characteristic of modern mathematics is discovering the foundations under the foundations, at every level. My math history is rusty, but didn't $\epsilon-delta$ pave the way for modern (20th C) analysis? – Mitch Mar 03 '11 at 15:20
  • @Mitch: I agree with you that the metaphor stretches things past the breaking point in many ways. I suspect that if I were to manage to track it down it would end up being very much tongue-in-cheek. But it's true that for most people who *use* math foundations are about as much interest as scaffolding on buildings that have been completed. Nice to know they were there, but the fact that the building is shelter from the storm and does not appear to be falling down is all that matters, (even if the architect is very worried about that crack in wall of the basement). – Arturo Magidin Mar 03 '11 at 16:18
  • 1
    @Arturo Magidin You answer was as wonderful as I expected, thank you for it. But I disagree with a couple of things you write, like: "one might say that functions that are continuous everywhere and nowhere differentiable are 'very artificial'..." Don't these sort of functions arise in infinite sums of Fourier series, a field that was birthed from its applications to engineering? I think I was not very clear in what I was trying to say: I'm well aware that strange exceptions can lead to interesting mathematics. – Uticensis Mar 03 '11 at 20:34
  • @Billare:The point I was trying to make is that if you think about functions the way, say, Euler thought about functions, then everywhere continuous-nowhere differentiable functions will look artificial (remember also that Fourier's work was not initially accepted as valid, and it took some effort to get it to be accepted). Such functions look artificial if you have that point of view. Or, the characteristic function of the rationals: it's obvious once you see it, and once you accept the very general definition of function, but if you think of functions as "curves", it seems artificial. – Arturo Magidin Mar 03 '11 at 20:40
  • 1
    @Arturo Magidin (cont'd) But the common "set of all sets" story that is often used to relate Russell's paradox didn't appear that way to me, because it admitted unmathematical objects. What gave me substantial insight was Asaf's answer: The essential quality to pull away from the oft-told story is that one has to _check_, and not just assume that a collection of objects satisfies a formula. And, from another answer, I understand what distinguishes the "set of all sets" as pathological its self-referential nature; such sets might naturally arise in more interesting mathematical venues. – Uticensis Mar 03 '11 at 20:41
  • @Billare: Yes: the "big deal" was that Russell pointed out that there are things that are not "sets", and entirely new category of things, so to speak. (Since Asaf's answer gave you substantial insight, perhaps that should be the accepted answer? Mine is more of a ramble). – Arturo Magidin Mar 03 '11 at 21:16
  • @Arturo Magidin I debated which answer to accept for like an hour: Different individual posts gave me good insights, but were at the same time limited; yours came later, but collated everything important. A conundrum! I ultimately decided I wanted yours on top, for the benefit of future readers. But as you wish... – Uticensis Mar 03 '11 at 21:18
  • Arturo, I have heard the spider story in colloquium talks by Leo Harrington, the Berkeley logician. – JDH Mar 06 '11 at 11:50
  • @JDH: Interesting; thanks Joel! – Arturo Magidin Mar 06 '11 at 19:55
  • 1
    Morris Kline also uses the spider metaphor in _Mathematics: The Loss of Certainty_ – Matt Stephenson Mar 07 '12 at 18:47

The difficulty can be hard to see from a modern point of view, where set theory books have been written in a way that avoids the problem. So imagine that you had never learned any set theory, but you are familiar with the English word "set". You probably think, as mathematicians did at the time, that any well-defined collection of objects forms a set, and any set is a particular well-defined collection of objects. This seems like a perfectly reasonable idea, but Russell's paradox shows that it's actually inconsistent. Moreover, the "set" that we now associate with Russell's paradox -- the set of all sets that don't contain themselves -- seems like a perfectly well-defined set. After all, each particular set is either a member of itself or not, so the ones that are not seem to be a well-defined collection. This casts doubt on the English language term "set", which in turn seems to cast doubt on all mathematics done in English. That was why the paradox was so revolutionary -- because things people had been routinely using turned out to be inconsistent, and it seemed possible that more paradoxes might be found that would cast doubt on other mathematical objects.

The solution that mathematicians eventually adopted was to drop the normal English meaning of set ("any well-defined collection of things") and replace it with a different concept, the "iterative" concept of sets. But it took some time for that solution to even be proposed, much less accepted. Even now, a key argument that the new conception is consistent is that nobody has managed to find an inconsistency. If that seems less than completely certain, that's because it is.

Several other paradoxes were discovered around the same time. One of the more interesting ones in my opinion, which was discovered somewhat later, is "Curry's paradox". It consists of this sentence: "If this sentence is true then 0=1". If you use the normal natural-language proof technique for proving an if/then statement, you can actually prove that sentence is true (it's not hard - try it). But if that sentence is true, then 0=1 - and you can prove that too, in the normal mathematical way. The fact that the normal techniques we use in everyday mathematics can prove 0=1 is certainly paradoxical, since each of those techniques seems fine on its own. Like Russell's paradox, Curry's paradox both casts doubt on our informal natural-language mathematics, and shows that certain formal theories that we might wish were consistent are actually inconsistent.

Carl Mummert
  • 77,741
  • 10
  • 158
  • 292
  • 7
    Curry's paradox is a modern version of Epicurius' Liar's Paradox,which centers around the natural-language sentence:"I am Lying." If the sentence is true,then it's false and converse. The entire discipline of metamathematics has developed largely to tackle problems like this and although a lot of "real" mathematicians consider such problems mere dinner theater and nothing that effects mathematics in "the real world",I would seriously take issue with that if they are simultaneously accepting logic and set theory-at least implicitly-as their foundations. – Mathemagician1234 Aug 20 '11 at 21:30

The significance of Russel's paradox is not just philosophical. Russel's paradox implies a contradiction in the presence of the unbounded axiom of comprehension:

$\exists s. \forall e. e\in s \leftrightarrow \phi$

where $\phi$ is the logical formula typically containing $e$. This axiom “creates” the set $s$ of all elements $e$ satisfying $\phi$. In other words, it turns properties into sets.

Let $r$ be the set of elements $e$ satisfying $e\notin e$. Then $r$ satisfies: $\forall e. e\in r \leftrightarrow e\notin e$. Instantiate $e$ to $r$: $r\in r \leftrightarrow r\notin r$. $p:=r\in r$ for clarity. Use $(p\leftrightarrow (\neg p))\rightarrow \bot$ valid in intuitionistic propositional logic. We've got a contradiction.

If you suspect there is a gap in my proof, I formalized it in Coq, but the accompanying text is mostly in Russian. To prove a contradiction, I need building blocks, which I introduced via “Variable” — the universe “all_sets”, the relation “in_set”, and our honey “comprehension”.

P.S. Contradictions in pure Coq are not found. ;)

  • 2,362
  • 17
  • 22

Set theory's success was so great in the late nineteenth century that there were high hopes of making it a foundation for philosophy, let alone mathematics.

In particular, Frege had a simple ontology that solved several foundational problems in philosophy. For instance, "what is the number three?" To Frege, it was none other than the set of all triples. "What is the color blue?" It is the set of blue things, etc. In this way there is no need to create an ontological distinction between a property ("concept" if I understand the classical terminology correctly, which I may not) and its extension (the set of all things with that property) because there is a one to one correspondence between concepts and extensions.

This may seem pedantic, but it's a tremendously useful philosophical axe. Try to sit down and give a coherent definition of the difference between a "property" and a "thing" and what it means for a thing to have the property. It's not so simple. So 19th century set theory provided an answer to some ancient questions.

Unfortunately, the answer is wrong. The concept "S is a set and S does not contain itself" has no extension. This was Russell's original formulation of the paradox, I believe, and is why Frege was so upset by it.

  • 25,480
  • 3
  • 35
  • 61

Much of what you say makes sense philosophically, and there are also ways to make it precise mathematically: For example, one result in reverse mathematics is that mathematicians often need even less complex sets than they use.

But I think you are actually asking several different questions.

Why did mathematicians take Russell's paradox seriously?

Did they? Those that were dealing with foundations certainly did, but I'd be surprised if that claim was true for the average mathematician.

The set of all sets admits all sorts of unmathematical objects right?

Well, that depends on how you define "set." In naive set theory, the idea that a set can contain any given object is a common point of view; I think it appeals to some pre-existing linguistic intuition about the word "set." But as you say:

I can't imagine a situation where a mathematician solving a concrete problem would invoke a set like that.

This is correct for a lot of mathematics, but there also are a lot of areas where you can easily get into trouble. Cardinal and ordinal numbers are the first examples, historically speaking (BTW, the Burali-Forti paradox apparantly predates Russell's paradox). Category theory is a more complicated case. One isn't necessarily tempted to invoke a "set of all sets," but similar things which are just as problematic (and arguably related).

But clearly important people think the result significant. It prompted Gottlob Frege, a smart and accomplished man, who Wikipedia tells me extended the study of logic in non-trivial ways, to say that the "foundations of [his] edifice were shaken."

I think you are generalizing too much from this sentence. Russell first stated his paradox explicitly about Frege's system. This is because Frege was basically the first person who tried to reduce all of mathematics to a single formal system. And then it turned out that his system, which took years to work out, could in fact prove every statement and was thus useless, technically speaking. It's not hard to imagine he was a bit disappointed.

Is there a practical example in mathematics that showcases this, where one might be led horribly astray if one doesn't place certain restrictions on sets, in say, dealing with PDEs, elliptic curves, functional equations?

I would answer "no" because whenever set-theoretic difficulties come up, the problematic sets turn out to be "tricky" in exactly the way you describe. However, being "tricky" is not a particularly solid criterion, and one that different people might have different opinions about. So when in doubt, mathematicians prefer to delegate such questions to a specific formal system, usually ZFC. That is, the criterion that is actually used nowadays is: "Can this set be constructed in ZFC?" If it can't, that doesn't necessarily mean that a contraction can arise from talking about it; this is where things become more interesting.

  • 4
    "I'd be surprised if that claim was true for the average mathematician." The paradoxes led somewhat directly to David Hilbert's program to try to prove the consistency of mathematics in a finitary way, with the goal of banishing once and for all the doubts that the paradoxes raised. Does Hilbert count as an average mathematician? The separation between foundations and other mathematics was not particularly sharp until the mid 20th century. – Carl Mummert Mar 02 '11 at 15:17
  • Good point. The second of Hilbert's problems suggests to me that he had already been interested in foundations before, but maybe the same is true for most other mathematicians. Or at least most of the famous ones. – Sebastian Reichelt Mar 02 '11 at 17:31
  • 2
    @Carl: One thing about Hilbert, though, is that his interest in foundations, especially for someone of his stature, *was* somewhat outside the norm, wasn't it? – Arturo Magidin Mar 03 '11 at 03:25
  • 1
    I'm not enough of a historian to know with certainty what the norm was. Certainly other prominent mathematicians in the early 20th century were interested in foundations. Later the Bourbaki were also mathematicians formemost. Category theory and topos theory were founded by mathematicians. So I think mathematicians "of stature" are just as likely to be interested in foundations, but I can't say whether they are *more* likely. In the early 20th century it was still possible to know the majority of math, and foundations were more accessible than today, when they almost require specialization. – Carl Mummert Mar 03 '11 at 12:09

Partially quoting Terence Tao, "One problem with this axiom ('Universal specification' or 'axiom of comprehension') is that it creates sets which are far too 'large' - for instance one can talk of a set of all objects (a so called universal set). Since sets are themselves objects, this means sets are allowed to contain themselves, which is somewhat silly state of affairs.One way to informally resolve it to think of objects as being arranged in hierarchy. .... " -- Analysis1, 2nd ed., Hindustan Book Agency, pp47

  • 5,747
  • 6
  • 25
  • 36

Maybe this answer is just stating the obvious. If so I apologize in advance.

Often when one proves things in mathematics, one uses a proof by contradiction. This means that we assume the opposite of what we want to prove, make some deductions, and eventually deduce something absurd like $0=1$ or "$x\in X$ and $x \notin X$." The falsehood of this last statement then implies the truth of the original statement that we wanted to prove.

But you see now how this kind of logic breaks down if there is an inconsistency embedded in our everyday mathematics? Using proof by contradiction you could then "prove" any statement at all in the following way: "Assume by way of contradiction that A has a non-real eigenvalue. Let S be the set of all sets which do not contain themselves. [...] As this is a contradiction, A has only real eigenvalues." Even if it was in this case obvious that the contradiction had nothing to do with A and everything to do with the tacked-on Russell's paradox, this might not be obvious in a more complicated situation.

Dan Petersen
  • 6,524
  • 2
  • 23
  • 29
  • A more explicit explanation about the relationship between Russell's paradox and the law of excluded middle would be of interest here. – Muhammad Alkarouri Jan 04 '12 at 08:41
  • @Muhammad: If you are interested in the relation between Russell's paradox and the law of excluded middle, you might consider reading Hartry Field's paper "The Consistency of the Naive(?) Theory of Properties" (You can find a copy of it on his homepage). What I find of interest in his paper is the following (I quote him verbatim): "(i)...if we have a naive theory of properties in the background, we have all the advantages of proper classes without need of any 'set-like entities' beyond ordinary sets; (ii)...given this naive theory of properties, ordinary iterative set theory (ZF) is – Thomas Benjamin Feb 01 '14 at 16:15
  • a highly satisfactory theory; and (iii) there is no obvious need for any additional theory of 'naive sets' ." It should be noted that he proves the consistency of the naive theory of properties by weakening excluded middle (he resolves the semantic paradoxes by weakening excluded middle as well). – Thomas Benjamin Feb 01 '14 at 16:28

I think the key point in Russell's paradox is the self-reference. It was important as well in subsequent developments like Gödel's incompleteness theorems.

The possibility of self-reference is what in a way opens up a theory. Something that makes it richer. But also something that from an axiomatic point of view allows for the possibility of paradoxes.

Douglas Hofstadter has written a load of popular books on the subject.

  • 14,777
  • 2
  • 44
  • 81

Because it represented a contradiction in the current formulation of math. In math a contradiction will propagate and invalidate everything, you can use it to prove and disprove any statement. This paradox could not be solved by math, it required a reformulation of math.

  • 8,398
  • 3
  • 41
  • 91
  • 2
    My point was that the paradox -is not- a contradiction in the system. The fact that you can evaluate '$P \land \neg P$ ' to get '$F$' does not mean there is a contradiction in propositional logic, just that a contradiction implies falsehood there. I feel like (in hindsight obviously) that Russell's paradox is not a problem with the system (it still may be consistent), but a problem with the (natural) human idea that 'set' can applied to any collection. – Mitch Mar 03 '11 at 15:26
  • 1
    @Mitch: There is a contradiction in the system, see my answer. – beroal Mar 06 '11 at 21:05
  • @beroal i read your answer. what you are saying is that there is a contradiction in the system if the proposition that is proven false, is an axiom of that system. with that understanding, you could have just replied so directly instead of leaving the interpretation of your answer open to the reader – symbiont Dec 22 '19 at 11:50
  • @symbiont I did not intend to explain what is a contradiction in my answer. I supposed that the readers know that implying a contradiction is equivalent to implying absurd. – beroal Dec 22 '19 at 16:44
  • @beroal i'm probably just nitpicking. i think you gave a good answer though – symbiont Dec 22 '19 at 19:36

Instead of giving you an answer that expresses an opinion or a particular point of view (regardless of how widespread or not that point of view might be), I would like to give you a list of source material that might be helpful to you.

First, in order to understand why the mathematical and logical communities take Russell's paradox seriously, I would urge you read the SEP (Stanford Encyclopedia of Philosophy) article "Frege's Theorem and Foundations for Arithmetic" sections 1 and 2 (you can find SEP articles on the Web). Reading these sections will help you understand Frege's "Basic Law V" from which all the trouble sprang. After digesting these sections, I would urge you to read the rest of the article in order to appreciate the significance of Frege's work on the foundations of arithmetic irrespective of the problems with Basic Law V.

Second, find a copy of Jean van Heijenoort's book "From Frege to Godel: A Source Book in Mathematical Logic, 1879 to 1931". In it you will find (in translation) Russell's letter to Frege announcing his discovery of the paradox and Frege's reply to Russell. Pay particular attention to his (Russell's) conclusion that "under certain circumstances a definable collection [Menge] does not form a totality." This ties in to the notion of proper class as well as (to the discerning reader) the possibility of 'roads not taken'.

Third, I would urge you to read the following paper by Thoralf Skolem, "Investigations on a Comprehension Axiom without Negation in the Defining Propositional Functions". In this paper, Skolem proves the existence of "domains" (i.e. models) such that the axiom of comprehension was satisfied for elementary propositional functions $\phi$, that is $\phi$ being built from atomic propositions u$\in$v by use of conjunction and disjunction, provided only conjunction, disjunction, and u$\in$v are allowed in $\phi$. He also shows that for these models the axiom of extensionality also holds. This shows that there are consistent fragments of naive set theory, in answer to your second question.

Finally, I urge you to look up and read any and all articles by the following Authors: Olivier Esser, Ross Brady, Thierry Libert, Roland Hinnion, and Zach Weber, among others. These researchers are doing interesting work in naive set theory.

I hope you find this helpful. Happy reading.

Thomas Benjamin
  • 1,185
  • 8
  • 12

It's not too hard to show that given a logical contradiction, you can prove anything. Russell's paradox showed that a certain set of axioms were inconsistent. If followed then that these axioms could be used to prove any statement. The simplicity of Russell's paradox made it clear which axiom was problematic, and it's was removed from further treatments of the subject, and we hope that what remains is consistent.

The point is, if you have an inconsistent axiomatic system, then you can't really trust anything that you derive from it, because, well, everything is derivable. Take any mathematical fact that you think is important, $2\times 2=4$, $\{1, \dots, n\} \subseteq \{1, \dots, n+1\}$, etc. and you can prove the negation using the axioms that lead to Russell's paradox.

So it didn't really matter what Russell's paradox was per se (well, it did in the sense that it showed which axiom was problematic), so much as the fact that there was a paradox at all.

Also, keep in mind that this all happened before Gödel's incompleteness theorems. At the time, it was still believed by many mathematicians that it would be possible to, given any mathematical statement, algorithmically compute whether it was true or not under the accepted set of axioms. The fact that there was a hole in the currently accepted set of axioms was therefore a big deal.

  • 9,046
  • 5
  • 30
  • 50

I cannot say it for russels paradox itself, but for computer science analogons. There are Questions, which can fundamentally not be answered. Not because of too few knowledge or power, but because of logical "reason". With Undecidability, this seems to be "obvious" today. But: Back then, the mathematicans thought, they just has to provide the right set, that they could solve every problem in the world. That there would be no theoretically unsolvable problems.

This is a so huge fact, that it takes some time, to really acknowledge it without any trivialization..


While Russell's paradox may seem like a small technicality, it disproves one of the very foundations of mathematics, so it has some drastic implications.

For example: Using Curry's paradox, a variant of Russell's paradox, you can prove that $0 = 1$.

Discovering that not every definable collection is a set was like discovering that $π$ is not actually a number, and that while you can treat it like a number for the purposes of multiplying and dividing, $\sqrt[x]{π}$ is actually undefined, so every mathematical proof that uses $\sqrt[x]{π}$ could potentially be wrong, and would have to be reconsidered.

  • 1,230
  • 1
  • 16
  • 31
  • 1
    But sets *can* in fact contain themselves. Read about non-well founded set theory. Even more so, it is possible to have a universal set, read about New Foundations. And a universal set is most definitely an element of itself. – Asaf Karagila Apr 20 '15 at 16:07
  • 1
    The paradoxes of naive set theory all pointed in one direction: ***unrestricted comprehension*** is inconsistent. And so what we had to do was to restrict how formulas define new subsets. But nothing was problematic about sets having themselves as elements, or the set of all sets as a general concept. – Asaf Karagila Apr 20 '15 at 16:09
  • @Asaf: Okay. So, can sets contain themselves in ZFC? And how could I word this better, considering that I am writing as someone who knows very little about set theory, for someone who knows very little about set theory? Would it be correct to say "not everything is a set", or is there a better way of putting it? – Zaz Apr 20 '15 at 16:15
  • No, not in $\sf ZFC$. But there is other set theories outside of $\sf ZFC$. – Asaf Karagila Apr 20 '15 at 16:16

Mathematics must ultimately be based on a limited collection of rules and axioms of logic and set theory. If you don't stray too far, you can probably get away with ignoring most if not all the finer points.

The domain of discourse for most mathematicians is just a few well-defined sets of numbers or points in space, and sets constructed from them. When most mathematicians make generalizations, it is not usually about every object in the universe, but only about things like numbers or points. This greatly simplifies matters to the point that weird situations like Russell's Paradox never actually comes up in practice. For example, there is nothing paradoxical about a set $r$ such that $\forall x\in N:[x\in r \iff x\notin x].$ You just have $r\notin N$ and that's the end of it! Likewise, if you define $r$ as a subset of $N$ such that $\forall x:[x\in r \iff x\in N \land x\notin x].$ Then you get $r\notin N$ and $r\notin r.$

The above mentioned rules and axioms, however, have to cover every eventuality. If you are going to be making generalizations about every object in the universe -- somebody has to! -- you have to guard against beasties in the set-theoretic wilderness like Russell's Paradox.

Some form of restricted comprehension axiom prevents the construction of the so-called Russell Set. If you are always working with some underlying set(s), comprehension is automatically restricted. Problems of comprehension can only occur when you define those underlying sets. If they are consistently defined (a big IF), you can pretty much ignore any concerns about comprehension after that.

Dan Christensen
  • 13,194
  • 2
  • 23
  • 44

Fortunately, set theory isn't that important to modern math. But category theory is, and an analogue of Russell's paradox stops you from constructing what would otherwise be one of the most obvious things to construct, the category of categories.

In practice, I've seen a bunch of expositions that go "Let $\mathbf{Cat}$ be the category of all (small) categories..." with a bit of a wink and a nudge to get you on the right track. So I don't think this is a terribly important issue -- but it does give a natural place for a working mathematician to get accidentally involved in set theory.

By Russell's paradox proper, there's also no set of groups, rings, topological spaces, and so on, but this can be remedied by introducing proper classes, and as far as I can tell, you'd never want to topologize the class of topological spaces or something like that.

Edit: I realized your question was more historical. Russell came at a time of tremendous vagueness in math, and his work was largely an effort to eliminate such vagueness. It was surprising that such a project could possibly have nontrivial results. Frege, specifically, had been working on a massive treatise on logic that was totally destroyed by Russell's paradox, because it relied so heavily on sets whose elements could come from anywhere. If you're interested, Logicomix is a really good graphic novel about Russell's life and work -- addressed to non-mathematicians, but a good read and helps motivate issues like these.

Paul VanKoughnett
  • 5,104
  • 2
  • 27
  • 27
  • 23
    -1: "Fortunately, set theory isn't that important to modern math." Objection, your honor -- subjective and argumentative. After you get your Fields Medal, we will be much more willing to hear such categorical [!] pronouncements. Until then, please show some restraint. – Pete L. Clark Mar 02 '11 at 11:31
  • 5
    You're right that that was worded too strongly. However, I think it's a fair judgement, and one doesn't need to be a Fields Medalist to observe it. Cohen proved the independence of the continuum hypothesis from ZFC in 1963 -- what are the major developments in the field since then? Have they impacted other fields? My department employs maybe one set theorist, and I think it's fairly large -- is it different elsewhere? – Paul VanKoughnett Mar 03 '11 at 01:53
  • 4
    My department employs three set theorists, another model theorist and we have at least two postdocs in those fields. We don't have a very big department and there are only three grad students in those fields, so I fail to see your point in that last comment. – Asaf Karagila Mar 03 '11 at 06:57
  • 4
    @Paul: Set theory remains an active field for mathematical research. If you were at, say, Berkeley rather than Harvard (which is a rather small department, by the way) that would probably be more apparent. Certainly modern set theory is connected to model theory and general topology, for instance. If you want to learn more about the major developments of the field, you should ask a question or do some reading. Please be aware that a dismissive attitude to whole areas of mathematics is something that many professional mathematicians will find obnoxious and immature. – Pete L. Clark Mar 03 '11 at 08:17
  • 8
    @Pete Very respectfully, Prof. Clark, I don't think you are being all that charitable towards Mr. Van Koughnett. His appraisal of set theory's importance was but one sentence amongst the many insights he provided. Moreover, he used a quantifier in his appraisal: he included "that." I detected nothing in the tone of his answer that was unthoughtful. Now he could very well have been wrong: If so, your downvote was proper and you have set things right by providing good information. But to characterize him as being "dismissive" or undeservedly putting on the airs of a Fields Medalist is unfair. – Uticensis Mar 03 '11 at 12:15
  • @Paul VanKoughnett: "By Russell's paradox proper, there's also no set of groups...". Really? At least for those straightforward algebras, I thought those collections were sets (or small categories). – Mitch Mar 03 '11 at 15:11
  • 2
    @Mitch: Trivially, you can just take the free group on every set to get a proper class of groups. But even up to isomorphism you get a proper class: the free groups on two sets aren't isomorphic for sets of different infinite cardinalities (by cardinal arithmetic, I believe). The cardinals are a proper class, so groups are. The same argument can be reused in many other examples. – Paul VanKoughnett Mar 03 '11 at 18:41
  • @Pete L. Clark: Maybe I will ask a question. It's not that I'm not curious, but I honestly can't recall attending a lecture or reading a paper where deep set-theoretic concerns intruded on the material. By contrast, concepts from, say, algebraic topology pop up in all kinds of other places, and so I'd consider it a more healthy field. If you find this opinion obnoxious, that's your right, but I'm not alone in it and I've heard it echoed by professional mathematicians. – Paul VanKoughnett Mar 03 '11 at 18:52
  • 11
    @Paul, by this argument I can say that alg. top. is not common, as I have not encountered it in papers I read. – Asaf Karagila Mar 04 '11 at 09:12
  • 12
    @Paul: it is one thing to have opinions about the relative merits of various branches of mathematics -- probably we all do, to some extent or another. (When I was a graduate student, I brought a non-math friend into a party in the math department lounge. I had the idea that he should pose as a math student from another department who was studying "large cardinals", since no one there would want to follow up on it. He didn't really get the joke.) But it's something else to post this on an internet forum, where it will stay semi-permanently and be read by everyone... – Pete L. Clark Mar 04 '11 at 09:48
  • 10
    ...Given that you know that there are several prominent set theorists active on this site as well as plenty of mathematicians involved in graduate admissions at their institutions, what is your rationale for offering an unsolicited negative opinion about the current viability of set theory? That some other people at Harvard say the same things? Well, guess what...it's obnoxious when they say it too, but maybe they feel like they can afford to alienate portions of the math community. I wouldn't recommend that strategy to any undergraduate (or to anyone, but especially not an undergraduate). – Pete L. Clark Mar 04 '11 at 09:52
  • 5
    -1: I like set theory. – hot_queen Jan 31 '14 at 02:25