Human beings habitually, effortlessly, and for the most part unconsciously represent one another as persons. Adopting this personal stance facilitates representing others as unified entities with (relatively) stable psychological dispositions and (relatively) coherent strategies for practical deliberation. While the personal stance is not necessary for every social interaction, it plays an important role in intuitive judgments about which entities count as objects of moral concern (Dennett Reference Dennett1978; Robbins & Jack Reference Robbins and Jack2006); indeed, recent data suggest that when psychological unity and practical coherence are called into question, this often leads to the removal of an entity from our moral community (Bloom Reference Bloom2005; Haslam Reference Haslam2006).
Human beings also reflexively represent themselves as persons through a process of self-narration operating over System 1 processes. However, in this context the personal stance has deleterious consequences for the scientific study of the mind. Specifically, the personal stance invites the assumption that every (properly functioning) human being is a person who has access to her own mental states. Admirably, Carruthers goes further than many philosophers in recognizing that the mind is a distributed computational structure; however, things become murky when he turns to the sort of access that we find in the case of metacognition.
At points, Carruthers notes that the “mindreading system has access to perceptual states” (sect. 2, para. 6), and with this in mind he claims that in “virtue of receiving globally broadcast perceptual states as input, the mindreading system should be capable of self-attributing those percepts in an ‘encapsulated’ way, without requiring any other input” (sect. 2, para. 4). Here, Carruthers offers a model of metacognition that relies exclusively on computations carried out by subpersonal mechanisms. However, Carruthers makes it equally clear that “I never have the sort of direct access that my mindreading system has to my own visual images and bodily feelings” (sect. 2, para. 8; emphasis added). Moreover, although “we do have introspective access to some forms of thinking … we don't have such access to any propositional attitudes” (sect. 7, para. 11; emphasis over “we” added). Finally, his discussion of split-brain patients makes it clear that Carruthers thinks that these data “force us to recognize that sometimes people's access to their own judgments and intentions can be interpretative” (sect. 3.1, para. 3, emphasis in original).
Carruthers, thus, relies on two conceptually distinct accounts of cognitive access to metarepresentations. First, he relies on an account of subpersonal access, according to which metacognitive representations are accessed by systems dedicated to belief fixation. Beliefs, in turn, are accessed by systems dedicated to the production of linguistic representations; which are accessed by systems dedicated to syntax, vocalization, sub-vocalization, and so on. Second, he relies on an account of personal access, according to which I have access to the metacognitive representations that allow me to interpret myself and form person-level beliefs about my own mental states.
The former view that treats the mind as a distributed computational system with no central controller seems to be integral to Carruthers' (2009) current thinking about cognitive architecture. However, this insight seems not to have permeated Carruthers' thinking about metacognition. Unless the “I” can be laundered from this otherwise promising account of “self-knowledge,” the assumption of personal access threatens to require an irreducible Cartesian res cogitans with access to computations carried out at the subpersonal level. With these considerations in mind, we offer what we see as a friendly suggestion: translate all the talk of personal access into subpersonal terms.
Of course, the failure to translate personal access into the idiom of subpersonal computations may be the result of the relatively rough sketch of the subpersonal mechanisms that are responsible for metarepresentation. No doubt, a complete account of metarepresentation would require an appeal to a more intricate set of mechanisms to explain how subpersonal mechanisms can construct “the self” that is represented by the personal stance (Metzinger Reference Metzinger2004). As Carruthers notes, the mindreading system must contain a model of what minds are and of “the access that agents have to their own mental states” (sect. 3.2, para. 2). He also notes that the mindreading system is likely to treat minds as having direct introspective access to themselves, despite the fact that the mode of access is inherently interpretative (sect. 3.2). However, merely adding these details to the model is insufficient for avoiding the presumption that there must (“also”) be first-person access to the outputs of metacognition. After all, even with a complete account of the subpersonal systems responsible for the production and comprehension of linguistic utterances, the fixation and updating of beliefs, and the construction and consumption of metarepresentations, it may still seem perfectly natural to ask, “But how do I know my own mental states?”
The banality that I have access to my own thoughts is a consequence of adopting the personal stance. However, at the subpersonal level it is possible to explain how various subsystems access representations without requiring an appeal to a centralized res cogitans. The key insight is that a module “dumbly, obsessively converts thoughts into linguistic form and vice versa” (Jackendoff Reference Jackendoff1996). Schematically, a conceptualized thought triggers the production of a linguistic representation that approximates the content of that thought, yielding a reflexive blurt. Such linguistic blurts are proto-speech acts, issuing subpersonally, not yet from or by the person, and they are either sent to exogenous broadcast systems (where they become the raw material for personal speech acts), or are endogenously broadcast to language comprehension systems which feed directly to the mindreading system. Here, blurts are tested to see whether they should be uttered overtly, as the mindreading system accesses the content of the blurt and reflexively generates a belief that approximates the content of that blurt. Systems dedicated to belief fixation are then recruited, beliefs are updated, the blurt is accepted or rejected, and the process repeats. Proto-linguistic blurts, thus, dress System 1 outputs in mentalistic clothes, facilitating system-level metacognition.
Carruthers (Reference Carruthers, Evans and Frankish2009) acknowledges that System 2 thinking is realized in the cyclical activity of reflexive System 1 subroutines. This allows for a model of metacognition that makes no appeal to a pre-existing I, a far more plausible account of self-knowledge in the absence of a res cogitans.
