> I've never been able to construct a series of tokens in
    > which the meaning changes based on which of those you
    > use, and it's legal either way.

Here's a series of tokens where switching comma to full stop means different things and both are legal:

    1> 1,0.
    0
    2> 1.0.
    1.0

In the first line, the comma is an expression separator. Attempting to use a full stop instead results in a floating point number.

    f(X, Y) when X, Y -> ok.
    g(X, Y) when X; Y -> ok.

In this case, the guard in f is turned into a guard sequence by switching a comma to a semicolon. Both are legal.

Matt