Igor Pak's blog

What is Combinatorics?


Do you think you know the answer?  Do you think others have the same answer?  Imagine you could go back in time and ask this question to a number of top combinatorialists of the past 50 years.  What would they say?  Would you agree with them at all?

Turns out, these answers are readily available.  I collected them on this page of quotes.  The early ones are uncertain, defensive, almost apologetic.  The later ones are bolder, prouder of the field and its status.  All are enlightening.  Take your time, read them all in order.

Why bother?

During my recent MIT visit, Jacob Fox showed me this blog which I found to be rather derogatory in its treatment of combinatorics and notable combinatorialists.  This brought me back to my undergraduate days in Moscow, reminded of the long forgotten but back then very common view of combinatorics as “second rate mathematics”.  In the US, I always thought, this battle has been won before my time, back in the 1980s.  The good guys worked hard and paved the road for all younger combinatorialists to walk on, and be proud of the field.  But of course the internet has its own rules, and has room for every prejudice known to men.

While myself uninterested in engaging in conversation, I figured that there got to be some old “war-time” replies which I can show to the Owl blogger.  As I see it, only the lack of knowledge can explain these nearsighted generalizations the blogger is showing.  And in the age of Google Scholar, there really is no excuse for not knowing the history of the subject, and its traditional sensitivities.

But while compiling the list of quotes linked above, I learned a few things.  I learned how tumultuous was the history of combinatorics, with petty fights and random turns into blind alleys.  I learned how myopic were some of the people, and how clever and generous were others.  I also discovered that there is a good answer to the question in the title (see below), but that answer is not a definition.

What do authorities say?

Not a lot, actually.  The AMS MSC (which I love criticizing) lists Combinaorics as 05, on par with about 70 fields, such as Number Theory (11), Geometry (51), Probability (60) and Computer Science (68).  It is also on the same level as  Nonassociative rings (17), K-theory (19) and Integral equations (51), which are perfectly fine areas, just much smaller.  It is one of the 32 categories on the arXiv, but who knows how these came about.

At the level of NSF, it is one of the 11 Disciplinary Research Programs, no longer lumped with “Algebra and Number Theory” (which remain joined at the hip according to NSF).  Around the country, Combinatorics is fairly well represented at the top 100 universities, even if breaking “top 10” barrier remains difficult.  Some are firmly in the “algebraic” camp, some in “probabilistic/extremal”, some have a lot of Graph Theory experts, but quite a few have a genuine mix.

This all reminded me of a story how I found out “What is mathematics“.  It started with me getting a Master of Arts degree from Harvard.  It arrived by mail, and made me unhappy.  I thought they made a mistake, that I should have been given Master of Sciences.  So I went to the registrar office, asked to see the director and explained the problem.  The director was an old lady, who listened carefully, and replied “let me check”.  She opened some kind of book, flipped a few pages and declared: “Yes, I see.  No mistake made.  Mathematics is an Art.”   Seeing my disappointed face, she decided to console me “Oh, don’t worry, dear, it’s always been that way at Harvard…”

What the experts are saying.

About the title question, I mean.  Let’s review the quotes page.  Turns out, a lot of things, often contradicting each other and sometimes themselves.  Some are cunning and ingenuous, while others are misleading or plain false. As the management maxim says, “where you stand depends on where you sit”.  Naturally, combinatorialists in different areas have a very different view on the question.

Few themes emerge.  First, that combinatorics is some kind of discrete universe which deals with discrete “configurations”, their existence and counting.  Where to begin?  This is “sort of” correct, but largely useless.  Should we count logic, rectifiable knots and finite fields in, and things like Borsuk conjecture and algebraic combinatorics out?  This is sort of like defining an elephant as a “large animal with a big trunk and big ears”.  This “descriptive” definition may work for Webster’s dictionary, but if you have never seen an elephant, you really don’t know how big should be the ears, and have a completely wrong idea about what is a trunk.  And if you have seen an elephant, this definition asks you to reject a baby elephant whose trunk and ears are smaller.  Not good.

Second theme: combinatorics is defined by its tools and methods, or lack of thereof.  This is more of a wishful thinking than a working definition.  It is true that practitioners in different parts of combinatorics place a great value on developing new extensions and variations of the available tools, as well as ingenuous ad hoc arguments.  But a general attitude, it seems, is basically “when it comes to problem solving, one can use whatever works”.  For example, our recent paper proves unimodality results for the classical Gaussian coefficients and their generalizations via technical results for Kronecker coefficients, a tool never been used for that before.  Does that make our paper “less combinatorial” somehow?  In fact, some experts openly advocate that the more advanced the tools are, the better, while others think that “term ‘combinatorial methods’, has an oxymoronic character”.

Third theme: combinatorics is “special” and cannot be defined.  Ugh…  This reminds me of an old (1866), but sill politically potent Russian verse (multiple English translations) by Tyutchev.  I can certainly understand the unwillingness to define combinatorics, but saying it is not possible is just not true.

Finally, a piecemeal approach.  Either going over a long list of topics, or giving detailed and technical rules why something is and something isn’t combinatorics.  But this bound to raise controversy, like who decides?  For example, take PCM’s “few constraints” rule.  Really?  Somebody thinks block designs, distance-regular graphs or coding theory have too few constraints?  I don’t see it that way.  In general, this is an encyclopedia style approach.  It can work on Wikipedia which is constantly updated and the controversies are avoided by constant search for a compromise (see also my old post), but it’s not a definition.

My answer, after Gian-Carlo Rota.

After some reading and thinking, I concluded that Gian-Carlo Rota’s 44 y.o. explanation in “Discrete thoughts” is exactly right.  Let me illustrate it with my own (lame) metaphor.

Imagine you need to define Russia (not Tyutchev-style).  You can say it’s the largest country by land mass, but that’s a description, not a definition.  The worst thing you can do is try to define it as a “country in the North” or via its lengthy borders.  You see, Russia is huge, spead out and disconnected.  It lies to the North of China but has a disconnected common border, it has a 4253 mile border with Kazakhstan (longer than the US-Canada border excluding Alaska), surrounding the country from three sides, it lies North-West of Japan, East of Latvia, South-West of Lithuania (look it up!), etc.  It even borders North Korea, not that this tiny border is much in use.  Basically, Russian borders are complicated and are a result of numerous wars and population shifts; they have changed many times and might change again.

Now, Rota argues that Combinatorics is similarly formed by the battles, and can only be defined as such.  It is a large interconnected field concentrated (but not coinciding!) around basic discrete tools and problems, but with tentacles reaching deep into “foreign territory”.  Its current shape is a result of numerous “wars” – the borderline problems are tested on which tools are more successful, and whoever “wins”, gets to absorb a new subfield.  For example, in its “war” with topology, combinatorics “won” graph theory and “lost” knot theory (despite a strong combinatorial influence).  In other areas, such as computer science and discrete probability, Rota argues there a lot of cooperation, a mutually beneficial “joint governance” (all lame metaphors are mine).  But as a consequence, if one is to define Combinatorics (or Russia), the historical-cultural approach would go best.  Not all that different from Sheldon’s approach to define Physics “from the beginning”.


In conclusion, let’s acknowledge that Combinatorics can indeed be defined in the same (lengthy historical) manner as a large diverse country, but such definition would be neither short nor enlightening, more like a short survey.  As Danny Kleitman writes, in practice this lack of a clear and meaningful definition of the subject “never bothered him”, and we agree.  I think it’s time to stop worrying about that.  But if someone makes blank general statements painting all of combinatorics in a certain way, this is just indefensible.

UPDATE (May 29, 2013)

I thought I would add a link to this article by Gian-Carlo Rota, titled “What ‘is’ mathematics?”  This was originally distributed by email on October 7, 1998.   For those too young to remember the Fall of 1998, Bill Clinton’s testimony was released on September 21, 1998, with its infamous “It depends on what the meaning of the word ‘is’ is” quote.  Rota’s email never mentions this quote, but is clearly influenced by it.

UPDATE (August 5, 2015)

Since this article was written, Russian borders changed again, and not in a way I could have imagined (or supported). I am guessing, Combinatorics boundaries also changed.  See e.g. our latest blog post titled The power of negative thinking on the use of Computability in Enumerative Combinatorics.