If we can articulate a goal, which is tantamount to recognizing a behaviour that fulfils that goal, then we can probably define, or create, a mechanism that achieves that goal, to within some tolerable error of observation. But not all that happens is teleological. When we say that we can provide a mechanistic explanation of an avalanche, there is the exchange of energy, but there is no goal, nothing to replicate, and no underlying mechanism.
We see mechanisms in coordinative systems. We also see them (I suspect) in Gibsonian perception/action machines. We bring them into being through the constraints of the experimental psychologists laboratory.
So the question necessarily arises: whose goals are we talking about? Mechanisms, or purposes, are probably necessary to identify the elements for whom the question of natural selection arises. The common structures of memetics and Darwinianism may point to commonality. Whose goals?
Enaction seems to provide a good language for talking about this important question.