That wouldn't be a very interesting experiment

Several of Rupert's arguments (and some from Adams and Aizawa as well) rely on the fact that, particularly in an experimental setting, organismically bounded cognitive performance can differ wildly from purported examples of extended cognitive performance. For example, if I am asked to memorize a list of items, I can generally remember up to seven, and over a relatively short period of time I will tend to forget more and more of these items, as long as I am not concentrating on them. But give me a notepad, and I can "remember" a practically endless number of items, and short of losing the notepad or its suffering some kind of unlikely damage, I will probably never "forget" these items. Similarly, if I'm asked to read a book and then put it away, answering questions about its content, I will perform very differently than I will if I am allowed to keep the book and refer to it while answering questions. (On some accounts, in the right circumstances, a reader and a book together form an extended cognitive system, with many of the reader's beliefs about the subject matter of the book residing in the book itself.)

Since performance differs so widely between the two cases, the argument goes, skepticism is called for on a few grounds. The main idea, I think, is that the wide difference in performance suggests we have a difference in natural kind. They don't behave similarly under relevantly similar circumstances, so they're not the same kind of thing. They don't give similar results under similar experimental conditions, so the science which performs these experiments shouldn't treat them as being of a kind.

In a JoP article, if I recall correctly, Rupert even puts it this way. There simply is no interesting cognitive science experiment we can do on a person's memory if we allow for his written notes to count as "memory."

I agree (or, if I'm remembering my Rupert wrong, then I posit) that no particularly interesting cogsci experiment could be done on such a subject. But importantly, there's a distinction between a cogsci experiment and an interesting cogsci experiment. If I were to set up, right now, a study seeing how long it takes people to forget an item after memorizing it as part of a list of ten items, this also wouldn't be very interesting--for many such experiments have already been done and the results are fairly well estabalished. Nevertheless, despite the fact that it wouldn't be interesting, this doesn't make it fail to be an experiment studying a subject in cognitive science. On the contrary, it's a paradigmatic example of such a study.

Similarly, I'd argue, an experiment on subjects including their notebooks, allowing their notebooks to be counted as part of their "memory," may not be interesting--because we all know exactly what would happen beforehand, for example--but that doesn't mean it's not a study of cognitive systems.

But what about the wide difference in performance mentioned above? The relevance of this is unclear to me. Individuals do differ in their cognitive abilities, after all. Some people can remember more than ten items. A few rare individuals--not all of them with disorders such as autism!--can remember, apparently, a practically endless number of items in a list. Do studies of these subjects not count as studies of cognitive agents? Or course not. They're cognitive, despite the fact that they perform very differently than most other cognitive agents we've studied.

It seems to me that cognition is a problem that different systems solve in different ways. Most human organisms solve it in a way that brings along with it limitations like "only seven items" and so on. Some human organisms seem to solve it in some other ways. Other organisms solve it in still other ways. And the use of a notebook to record memories is yet another way to solve the cognition problem.

It is valuable for many practical and scientific reasons to study the ways human organisms tend to solve the cognition problem when denied the use of external resources. This tells us something about the general class of "brain-based" or "brain-bound" cognition methods. But this doesn't mean that's the only kind of cognition there is, of course!

But it should be asked, what's the use of a general category of "cognition" if the interesting scientific work gets done about only particular means of cognition? Aren't the various means the real kinds here, and isn't generalized "cognition" itself something of a red herring?

One suggestion here might be that it's too easy to be mislead by examples like the notebook used to memorize lists. That's a trivial task. But more complex tasks of various sorts, carried out with the aid of a notebook may yield more interesting correlations between important variables. And it doesn't seem implausible, at least, to think that, as we vary the complexity of tasks incrementally, and the amount and kind of external resources available, then in some range of complexity we'll find an interesting continuation of patterns that begin at the point of no resource availability. (In other words, we might find a function that aptly, using few variables, describes the subject's performance continuously all the way from a state of zero resource availability to a state of great resource availability, given sufficient tax complexity.)

That's just a suggestion--maybe the beginning of a suggestion of one way to experimentally confirm or disconfirm extended cognition, at least of certain types or given certain accounts.

Another suggestion is that even if extended cognition turns out not to be a good scientific kind, it may well be a conception that needs to be adhered to for humanistic, political or moral purposes. Perhaps what Otto is doing is quite discontinuous, empirically speaking, from what any of us are doing when we remember things. And yet perhaps interfering with Otto's notebook is best concieved of, for moral purposes, as interfering with his personal coherence in the same way that interfering with my memory would be interfering with my personal coherence.

Can good moral categories really fail to track the scientific ones in this way? I'm not sure.