I
Carruthers argues that a consequence of the “mindreading is prior” account is that the mindreading faculty should have introspective access to perceptual and quasi-perceptual states. Two reasons are given. First, because the mindreading faculty must have access to perceptual inputs about the actions of others (targets), it “should be capable of self-attributing those percepts in an ‘encapsulated’ way, without requiring any other input” (target article, sect. 2, para. 4). But arguably the mindreading faculty would not have evolved unless it were able to predict the behaviors of others; and for this, the faculty must have access to non-perceptual beliefs about past and general facts about particular targets. If a mindreading animal, for example, observes a target orient its eyes toward food and thereby attributes sees-food, it is unlikely that it will be able to predict what the target will do unless it has access to relevant past and general facts about the target that are not perceptually available in the scene or hardwired into the faculty itself (e.g., that the target is a member of its own group/family, has recently eaten, has shared food in the past, etc.). By parity of reasoning, since the mindreading faculty must have access to non-perceptual beliefs, the faculty should be capable of self-attributing such thoughts in an “encapsulated” way without requiring any further input.
Carruthers argues that introspection of quasi-perceptual states is also likely on the “mindreading is prior” account because, being perceptual, they are (when attended to) “globally broadcast” to all concept-forming systems, including the mindreading faculty. But arguably attended-to thoughts are “globally broadcast” as well. In fact, Baars (Reference Baars1997) argues that attended-to thoughts, just as much as attended-to percepts, “create a vast access to perhaps all parts of the nervous system” (p. 59). For example, if one were to observe a subject looking at a red apple, one's mindreading faculty would likely infer that the subject sees the color of the apple; however, this default inference would have been prevented if, prior to observing the subject, the thought should have occurred to one that this subject had recently informed one that she was color-blind. It is quite plausible, therefore, that attended-to thoughts are also broadcasted to the mindreading factually, and by parity of reasoning, the faculty should be capable of self-attributing such thoughts in an “encapsulated” way without requiring any further input.
II
The standard argument in support of introspective access to thoughts runs as follows (see Goldman Reference Goldman2006, p. 230, for an example):
1. Sometimes we know what we think, and yet what we think is quite unrelated (for interpretative purposes) to any of the contents to which we have subjective access (such as the contents of perception, proprioception, episodic memory, or the contents of introspection regarding perceptual and quasi-perceptual states).
2. Therefore, it is unlikely in such cases that we know what we think as a result of an interpretation from the subjectively accessible contents rather than as a result of an introspective access to the thought itself.
This is an inductive argument to the best explanation. Hence, the more interpretatively irrelevant the subjectively accessible contents are to the self-ascribed thought, the more likely the process is introspection. Furthermore, because the processes involved in interpretation and introspection are unconscious, it is to be expected that the greater the degree of interpretative relevance the subjectively accessible contents bear to the self-ascribed thought, the greater the chances are that it will appear to the subject as if he knows what he thinks on the basis of introspection when it is actually the result of interpretation. I suspect that this is what is happening in the split-brain case that Carruthers describes. The subject has access to the contents of his perceptual and proprioceptive states that represent him as walking out of the testing van, and he probably has access to the contents of his memory which represent the location of the van to the house, and that the house has a Coke in it, and so on. All of this would likely allow him to interpret (at an unconscious level) himself as going to get a Coke from the house. Is it possible for it to always appear as if our knowledge of our own thoughts is introspective when it is interpretative? Yes, but it is unlikely the more interpretatively unrelated these contents are to the self-ascribed thought.
The preceding argument assumes that the belief in introspective thoughts is the result of an inductive argument and is not simply built into our pre-theoretic concept of the mind's epistemic access to itself. Carruthers disagrees. He argues that on the “mindreading is prior” account, it is to be expected that the mindreading system should represent this epistemic access as introspection and not interpretation, since doing so would “greatly simplify” the system's computational operations without any significant loss in reliability. However, if the mindreading system's concept of introspection is Carruthers' broad and negative definition (roughly, a reliable method for forming metacognitive judgments that is not interpretative and is different in kind from our access to other minds), then in order for the mindreading system to apply its concept of introspection to a subject, it will need to assume or have reason to believe that the subject's access to its mind is not interpretative. This would seem to undermine Carruthers' claim of there being a greater degree of simplification in the operations of a mindreading system that represented the mind's access to itself as introspection compared to one that represented it as interpretation only, since the former would require the possession and application of the concepts of introspection and interpretation, whereas the latter would require the possession and application of the concept of interpretation only. It is more plausible to suppose that the mindreading system's model of the mind's access to itself would simply be that of a reliable judgment-forming process that is different in kind from that used to form judgments about other minds. But such a model of the mind's access to itself would be neutral regarding whether it is introspection (as Carruthers defines it) or interpretation – at least, with respect to the kind of interpretivism that Carruthers defends in the “mindreading is prior” account, which also holds that the mind's interpretative access to its own thoughts is reliable and different in kind from its access to the thoughts of others.
