The Universe of Discourse

Tue, 10 Jan 2006

Foundations of Mathematics in the early 20th Century

Last fall I read a bunch of books on logic and foundations of mathematics that had been written around the end of the 19th and beginning of the 20th centuries. I also read some later commentary on this work, by people like W. V. O. Quine. What follows are some notes I wrote up afterwards.

The following are only my vague and uninformed impressions. They should not be construed as statements of fact. They are also poorly edited.

1. Frege and Peano were the pioneers of modern mathematical logic. All the work before Peano has a distinctly medieval flavor. Even transitionary figures like Boole seem to belong more to the old traditions than to the new. The notation we use today was all invented by Frege and Peano. Frege and Peano were the first to recognize that one must distinguish between x and {x}.

(Added later: I finally realized today what this reminds me of. In physics, there is a fairly sharp demarcation between classical physics (pre-1900, approximately) and modern physics (post-1900). There was a series of major advances in physics around this time, in which the old ideas, old outlooks, and old approaches were swept away and replaced with the new quantum theories of Planck and Einstein, leaving the field completely different than it was before. Peano and Frege are the Planck and Einstein of mathematical logic.)

2. Russell's paradox has become trite, but I think we may have forgotten how shocked and horrified everyone was when it first appeared. Some of the stories about it are hair-raising. For example, Frege had published volume I of his Grundgesetze der Arithmetik ("Basic Laws of Arithmetic"). Russell sent him a letter as volume II was in press, pointing out that Frege's axioms were inconsistent. Frege was able to add an appendix to volume II, including a heartbreaking note:

"Hardly anything more unwelcome can befall a scientific writer than that one of the foundations of his edifice be shaken after the work is finished. I have been placed in this position by a letter of Mr Bertrand Russell just as the printing of the second volume was nearing completion..."

I hope nothing like this ever happens to any of my dear readers.

The struggle to figure out Russell's paradox took years. It's so tempting to think that the paradox is just a fluke or a wart. Frege, for example, first tried to fix his axioms by simply forbidding (xx). This, of course, is insufficient, and the Russell paradox runs extremely deep, infecting not just set theory, but any system that attempts to deal with properties and descriptions of things. (Expect a future blog post about this.)

3. Straightening out Russell's paradox went in several different directions. Russell, famously, invented the so-called "Theory of Types", presented as an appendix to Principia Mathematica. The theory of types is noted for being complicated and obscure, and there were several later simplifications. Another direction was Zermelo's, which suffers from different defects: all of Zermelo's classes are small, there aren't very many of them, and they aren't very interesting. A third direction is von Neumann's: any sets that would cause paradoxes are blackballed and forbidden from being elements of other sets.

To someone like me, who grew up on Zermelo-Fraenkel, a term like "(z = complement({w}))" is weird and slightly uncanny.

(Addendum 20060110: Quine's "New Foundations" program is yet another technique, sort of a simplified and streamlined version of the theory of types. Yet another technique, quite different from the others, is to forbid the use of the ∼ ("not") operator in set comprehensions. This last is very unusual.)

4. Notation seems to have undergone several revisions since the first half of the 20th Century. Principia Mathematica and other works use a "dots" notation instead of or in additional to using parentheses for grouping. For example, instead of writing "((a + b) × c) + ((e + f) × g)", one would write "a + bc :+: e + fg". (This notation was invented by—guess who?—Peano.) This takes some getting used to when you have not seen it before. The dot notation seems to have fallen completely out of use. Last week, I thought it had technical advantages over parentheses; now I am not sure.

The upside-down-A (∀) symbol meaning "for each" is of more recent invention than is the upside-down-E (∃) symbol meaning "there exists". Early C20 would write "∃z:P(z)" as "(∃z)P(z)" but would write "∀z: P(z)" as simply "(z)P(z)".

The turnstile symbol $$\vdash$$ is Russell and Whitehead's abbreviation of the elaborate notation of Frege's Begriffschrift. The Begriffschrift notation was essentially annotated abstract syntax trees. The root of the tree was decorated with a vertical bar to indicate that the statement was asserted to be true. When you throw away the tree, leaving only the root with its bar, you get a turnstile symbol.

The ∨ symbol is used for disjunction, but its conjunctive counterpart, the ∧, is not used. Early C20 logicians use a dot for conjunction. I have been told that the ∨ was chosen by Russell and Whitehead as an abbreviation for the Latin vel = "or". Quine says that the $$\sim$$ denotes logical negation because of its resemblance to the letter "N" (for "not"). Incidentally, Quine also says that the ↓ that is sometimes used to mean logical nor is simply the ∨ with a vertical slash through it, analogous to ≠.

An ι is prepended to an expression x to denote the set that we would write today as {x}. The set { u : P(u) } of all u such that P(u) is true is written as ûP. Peter Norvig says (in Paradigms of Artificial Intelligence Programming) that this circumflex is the ultimate source of the use of "lambda" for function abstraction in Lisp and elsewhere.

5. (Addendum 20060110: Everyone always talks about Russell and Whitehead's Principia Mathematica, but it isn't; it's Whitehead and Russell's. Addendum 20070913: In a later article, I asked how and when Whitehead lost top billing in casual citation; my conclusion was that it occurred on 10 December, 1950.)

6. (Addendum 20060116: The ¬ symbol is probably an abbreviated version of Frege's notation for logical negation, which is to attach a little stem to the underside of the branch of the abstract syntax tree that is to be negated. The universal quantifier notation current in Principia Mathematica, to write (x)P(x) to mean that P(x) is true for all x, may also be an adaptation of Frege's notation, which is to put a little cup in the branch of the tree to the left of P(x) and write x in the cup.

[Other articles in category /math] permanent link