Carruthers suggests that cognitive scientists are confused about the meaning of “metacognition,” citing our work as an illustrative example. In fact, we follow a standard definition of the term, adopted from Nelson and Narens (Reference Nelson, Narens and Bower1990). (This particular formulation appears in Anderson & Oates [Reference Anderson and Oates2007], but the definition is in widespread use. See, e.g., Dunlosky Reference Dunlosky, Hunt and Ellis2004; Dunlosky & Bjork Reference Dunlosky and Bjork2008; Dunlosky & Metcalfe Reference Dunlosky and Metcalfe2009; Metcalfe Reference Metcalfe1993; Metcalfe & Shimamura Reference Metcalfe and Shimamura1994.) The definition runs as follows:

This offers an information-theoretic characterization of metacognition that is neutral regarding the form that information takes, or the processing it undergoes. Thus, it is quite incorrect to say that cognitive scientists use the term “in two quite distinct ways, often without noticing the difference” (target article, sect. 5.1, para. 2). We use the term consistently in a way that leaves open the various ways in which such a relationship could be implemented. We are not confused about the difference between systems that involve “metarepresentations of [its] own first-order cognitive processes as such” (sect. 5.1, para. 2) and those that don't; rather, this distinction is not relevant to the definition of metacognition.

In fact, some of the processes in the systems we implement are indeed metacognitive in Carruthers' more restricted sense. To take just one example, mentioned by Carruthers: If an active logic system notices the presence of both P and ¬P in its knowledge base (KB), it will assert Contra(P, ¬P, t). That is a statement about – a metarepresentation of – the state of the KB at time t (i.e., that it contained that contradiction). Our systems can reason about this fact with that metarepresentation, and consequently take various control steps, the simplest of which is to refrain from using these premises in further deduction (Anderson & Perlis Reference Anderson and Perlis2005a). But other processes in active logic systems, and other of our metacognitive systems, effect such monitoring and control without explicit metarepresentations of this sort (see, e.g., Anderson et al. Reference Anderson, Oates, Chong and Perlis2006).

Of course, Carruthers is free to define his terms and circumscribe his interests as best serves his argument, and if this were merely a terminological dispute, we would not be submitting a commentary. But there is a more substantive point in the background, which potentially affects Carruthers' overall proposal. Carruthers writes: “Generally the term is used, as it has been throughout this article, to mean cognition about one's own cognition. Metacognition, in this sense, is inherently higher-order, involving metarepresentations of one's own first-order cognitive processes as such” (sect. 5.1, para. 2, emphasis in original). The implication seems to be that for something to be about another requires a higher-order metarepresentation. But we would like to suggest that this associates higher-order-ness with meta-ness and aboutness (if we can be forgiven the neologisms) in a way that is not necessary.

First, it is not clear that aboutness requires higher-order-ness. Surely a representation or a process can be about another without being at a different level, or in a different representational language. Indeed, can't a process (or representation) be about itself? (See, e.g., Perlis Reference Perlis1985; Reference Perlis1988; Reference Perlis1997; Reference Perlis2000; Perlis & Subrahmanian Reference Perlis, Subrahmanian, Gabbay, Hogger and Robinson1994.) It is a common bias, perhaps stemming from Tarski, that there must be a hierarchy of meta-languages, each standing back from the one it refers to. But Tarski adopted that approach to avoid technical difficulties in formal logic; it is not necessary a priori.

Second, it is not clear that meta-ness requires higher-order-ness. In related writings, we have suggested that representation requires only the following: tokens, whatever their form/content, that can be used to guide actions with respect to certain targets (Anderson & Perlis Reference Anderson and Perlis2005b; Anderson & Rosenberg Reference Anderson and Rosenberg2008). On these accounts, the information being used and manipulated during cognition is representational just in case it is used to guide behavior with respect to targets in various circumstances. Likewise, a metacognitive monitoring and control process represents a cognitive process, just in case it allows the metacognitive component to guide actions with respect to the cognitive process. Such monitoring and control is indeed (we maintain) cognition about cognition – is thus metacognition – without having to be/utilize higher-order representations of cognition as such.

As should be clear from the preceding, we have a somewhat different understanding of what the representational aboutness relation requires. This most definitely applies to self-representation as well (Anderson & Perlis Reference Anderson and Perlis2005b), although it is perhaps worth noting that the account of self-awareness we develop in the cited paper is – despite differences in the fundamental criteria for aboutness – nevertheless compatible with the “mindreading is prior” framework that Carruthers advocates.

So why might all of this matter to Carruthers? Because of Carruthers' understanding of what aboutness requires, he is driven to adopt a higher-order, meta-representational account of what having certain thoughts about another's thoughts (“mindreading”) requires. In contrast, the less restrictive option offered by us opens the door for a broader range of theories of what our responsiveness to the mental states of others requires. This would include, for instance, Shaun Gallagher's interesting, and interestingly different, interaction-based account of understanding self and others (Gallagher Reference Gallagher2004; Reference Gallagher2005). It would have been useful and instructive to see how this rather broader portrayal of the competing possibilities might have affected Carruthers' argument, discussion, and conclusions.