C&C raise a potential objection to EM that some commentators have picked up on, and that it is the issue of portability and reliability (Adams and Aizawa 2001, this volume). The brain and body constitute a package of cognitive capacities that are portable in that they can always be brought to bear on a cognitive task; they form the constant cognitive core of an individual. The coupling of these core cognitive resources to the local environment is too contingent: the cognitive core can be too easily decoupled from its environment. Two conclusions can be drawn from this observation. First, it is the core cognitive resources that are of real interest to cognitive scientists, who are largely interested in the cognitive processes to be found in the brain. Second, the contingency of coupled cognitive resources shows that they are not really part of the reliable and portable cognitive resources that agents bring to bear on the world. (Menary, 2010, p. 7).Since Menary refers to two A&A papers, it kind of looks to me as though he thinks that we think that lack of portability and reliability counts against EC. But, I don't think we think this.
A&A defend some weak form of cognitivism. So, suppose we have a device that runs the right kind of program on the right kind of non-derived representations so that it counts as a cognitive system. Now, let the system be such that it repeatedly and spontaneously splits in half for one minute, then comes back together for one minute. It's one, then two, then one, then two. It's fine by me if the system is a cognitive system for one minute, then (probably) not a cognitive system for a minute. To my mind, it's having the right program on the right kind of representations that makes for the cognitive. Durability doesn't seem to me to matter too much. The wrinkle in this is the degree to which one has to accept some "historical" condition on content or having a mind. If you think that Swampman has no content and no mind, then you won't like the fast switching. But, if you are ok with Swampman having a mind, then the repeated connected/unconnected case should probably be a minded/unminded case. That's all pretty rough (especially for those who do not know about Swampman), but maybe good enough for blog work.