the idea is that when the sentence in the notebook is deployed by Otto in the right sort of way, then and only then can it count as among Otto's beliefs. (Rowlands, 2009b, p.631).Now, I think I can agree with this. Where Rowlands, Clark, and I appear to disagree is on what counts as "the right sort of way". Clark, for his part, proposes that conditions of "trust and glue" constitute "the right sort of way". Cognitivists, however, think that the conditions involve something like figuring in computations of such and such a configuration. The distinction is hard to make out in any detail, since, well, we don't know that much about the mechanisms of the manipulation of symbols in an LOT. But, the general point should be clear. By contrast, it is not entirely clear to me just how restrictive an account of the processing condition Rowlands wants to place on sentence manipulation. In his other paper, "Extended Cognition and the Mark of the Cognitive," it seems to me that he did not want to place any conditions on character of symbol manipulations. Essentially, you just have to have the manipulation of symbols with non-derived content.
(Incidentally, Wilson and Clark's paper in the Robbins-Aydede Handbook of Situated Cognition also alludes to coupling "in the right sort of way".
Rowlands, M. (2009b). "The Extended Mind" Zygon, 33(3), 628-640.