Will Wimsatt: Evolution and the stability of functional architectures

A puzzle about functional organization has gone largely unnoticed in our debates
about the nature of function. Although we recognize that systems evolve and acquire new
functions, no-one has systematically discussed whether there are constraints on how this is likely
to occur. This naturally suggests the widely discussed problem of evolutionary innovation,
but I am interested here in the complementary problem: things we already know about
epistatic co-dependencies in functional organization suggest that the common
conservation of organization under mutations or sexual recombination should be quite
remarkable. This arises either for Cummins style role function or on the selectionist account of
function. There are strong constraints on the addition of new functions imposed by conditions of
evolvability and generative entrenchment. Evolvability favors increases in robustness for
functionally important architectural features. And greater generative entrenchment produces more
constraints on what changes can be adaptive, yielding a rapidly declining probability that macromutations
of increasing size will work. These are both bound to affect the structure of functional
architectures and the character of functional innovations. If these constraints were violated
systematically and frequently, the flux of changes in functional architectures would give us
significant troubles even in individuating and identifying functions in complex organizations and
would make evolution virtually impossible. But taking account of these constraints is more
broadly revealing. Doing so gives a new perspective on the relation between selectionist and
causal role accounts of function