Well, I didn't mean to enter into technicalities, just wanted to add a third option to those presented by Fodor (intensional process and laws of selection), because the idea that "selection for" may be intensional worries me, and I think everyone should be worried, while the non-existence of laws of selection is not a source of concern.

Fodor's "option b" for addressing the issue of coextensive traits is "disentangle them". He does this in some possible worlds, others do that in the laboratory. Either way, I agree with Fodor that it's unlikely that this procedure will solve all the problems. I'm even ready to accept that some coextensions may be nomic (e.g., governed by the laws of developmental biology), barring any possible-world disentanglement.

The option I offered is to double check the coextensive traits themselves. Did they get the attention of genetics, developmental biology, functional morphology, physiology, something else? If so, we may expect that these disciplines will tell us more about other things we could not know by mere reflection. Some of these things may be related to function and suggest an adaptive hypothesis. Another interesting possibility would be discovering that a trait is a developmental correlate of some other trait. Either way, we would be in a better position in order to evaluate whether an adaptive hypothesis would provide the right kind of explanation.

(An incidental point of some importance: this is how we might proceed if we are after a historical explanation of real-life traits. A perfectly legitimate but different endeavor is to assign fitness values to "disentangled traits" when modeling the patterns of natural selection.)

How could this option affect the "selection for" issue? Roughly, by replacing "coextension" with two (and, very likely, more) biological phenomena: developmental correlation and functional interaction. Taken together, the theories that biologists have built around these phenomena provide plausible grounds for justifying the distinction between traits that are "selected" and traits that are "selected for". Very roughly, to show how this can be, I'd like to provide yet another resume of Darwin's theory. 

In order not to be carried away by mentalistic intuitions, I suggest a temporary change of name: let's call the core of our theory "survival of the fittest". The theory says that in a certain kind of process of trait fixation the following happen: (1) the fittest in each generation possess one or more differential traits; (2) there is a process connecting the possession of those differential traits to improved chances of leaving offspring. To me it sounds extensional enough: although we get to know the details only post hoc, the only condition for being the fittest is the possession of a certain set (possibly a singleton) of differential traits. The traits and their interactions with organism and environment do the rest. Let's call these traits adaptations.

The theory does not require that every differential trait observed in the fittest is also an adaptation -- leaving ample room for the developmental biologist's generalizations. It requires, however, that all adaptations are involved in improving the chances of leaving offspring through their interactions with other parts of the organism -- ask the functional morphologist. Thus, a principled distinction between traits that undergo fixation (selected) and traits that undergo fixation and are also adaptations (selected for) seems tenable.