In recent days, I've been tussling here with Mark Rowlands over how one might interpret some of what he has written. This exchange highlights three different EC sorts of hypotheses that different in what quantifiers they invoke, namely,
1) All cognitive processes are extended.
2) Most cognitive processes are extended.
3) Some cognitive processes are extended.
Now, Mark vigorously denies that he has ever held 1), but instead only 3). And, he indicates that even I accept 3), so that we are really on the same side after all.
But, I only accept 3), insofar as I do, because it's just so hard to muster an argument that absolutely all cognitive processes are indeed internal. I've all along admitted that extended cognitive processes are possible, so it's hard to rule out any instances of that possibility.
Instead, I have resisted all the proffered instances of EC, e.g. Inga-Otto, Tetris, Kanzi, Rush Hour, the Globe Theatre, etc., etc., and the reasoning that has been offered in support of those instances.
Now, philosophers typically gravitate to the universal and existential quantifiers, but I think the more interesting versions of EC are those that involve other quantifiers, such as in 2). Now, to my knowledge, no advocates of EC has ever explicitly endorsed a version such as 2), rather than 1) or 3). 1), of course, entails 2), but even advocates of 3) seem to run arguments that, if they were good (which I generally think they are not), would suffice to establish something like 2). So, philosophers often "undersell" the strength of the arguments they offer. (To take a possible case, Clark suggests that we are natural born cyborgs. Maybe if we are perennial tool users and tool use leads to extended cognition, then something like 2) would follow.)
I think versions like 2) are the most interesting, since 1) is, well, obviously false as Rowlands will now help me argue and 3) is so weak that it does not challenge the scientific enterprise of intracranial cognition. It doesn't that much matter if there is some unique exotic case of one person with a brain implant in which cognition is extended. If most instances of cognition are intracranial, then there is perhaps enough there to study to warrant cognitive psychology going on in much the way cognitivists have been going for a couple of decades now.
So, to borrow a phrase from Sutton, et al., I think that versions of EC quantified along the lines of 2) form the rich middle ground.