For what it's worth, I wrote a book on this, and discussed it on SO.
The basic idea is to apply information theory to software. Examples of modern ideas that are consistent with it are DSLs, the so-called "KISS principle", and DRY, but it goes far beyond that. In a nutshell, the fewer editing changes needed to implement average functional requirements, the fewer chances there are to introduce mistakes, and it could be argued the redundancy (as a code) of source, leading to exposure to errors, is the primary if not only source of bugs.
ADDED: Algorithmic (or Kolmogorov) Information Theory expresses the information content of something as the length of the shortest program that can compute it. I have two problems with it.
The theoretical work that is done
with it goes off on abstract
tangents that have little practical
application.
I think what is even more important
than the length of a program is the
length of the sequence of edits that
produces it as a function of the
requirements. I don't know of any
theoretical work on this aspect of
it.
Basically, the idea is if a language is minimal with respect to a problem domain, then programs in that language correspond 1-for-1 with problem statements, so a program can never crash (i.e. satisfy no requirement). On the other hand, if programs contain more than that minimal amount of information (have redundancy), then there can be programs that don't match any possible requirement (i.e. they are simply wrong regardless of the requirement). The more redundant they are, the larger the space of invalid programs is compared to the valid ones. Any human errors in entering these programs is thus much more likely to produce an invalid program (that crashes or gets invalid answers).
That's why it is important to minimize the redundancy of the language w.r.t. the problem domain. That's what DSLs do, ideally. In my opinion, good programmers instinctively arrange their code so as to allow new or revised functional requirements to be implemented with the smallest number of edits. (Which is not at all the same thing as the shortest length, but does tend to correlate with length.)
ADDED: This is an example of what I'm talking about.