I agree with Carruthers that there is no piece of human cognitive architecture dedicated to introspection. But the right response is to abandon the search for one introspective metacognitive system, whether dedicated or coopted. We become aware of our states of mind by a variety of methods, which depend on a variety of systems.
Consider Hurlburt's experiments. Carruthers suggests that subjects employ their mindreading systems on their own behavior and circumstances when beeped. But he worries that not all cases can be handled like this, since there is nothing going on at the time of the beep that looks like evidence for one interpretation or another. To save the self-mindreading view, Carruthers is forced to conclude that subjects are basing their interpretation on their immediately prior behavior. This is a rather desperate expedient. People do not seem to report any awareness in these cases of their own behaviour or even, very often, their own states of mind. This supports the view that we put ourselves into a position to assert a state of mind by doing whatever we do to get into that state in the first place (Evans Reference Evans1982).
When Hurlburt (Reference Hurlburt1997) discusses the case of Donald, for instance, he does not show that Donald was interpreting himself at all when he was beeped. Donald noted, for instance, that his son had left the record player on again (Hurlburt Reference Hurlburt1997, p. 944). He reported attending to a fact about his environment. He was later brought to see the facts he noticed as evidence that he harbored unacknowledged anger toward his son. Donald went over his own transcript after beeping and interpreted his behaviour as he might anyone else's. This case helps Carruthers in one way, but it also shows something he misses about the beep cases. Donald did not report anything about his own mind or engage in any self-interpretation when he was beeped.
Many of Hurlburt's subjects do not, when beeped, report anything that looks like introspection or any other self-examination. They report thinking about the world, not about themselves. Carruthers mentions a subject who is wondering what her friend will be driving later; this is not a thought about one's own state of mind at all, and it is not clear why we need to call this interpretation. The subject is not self-interpreting. She is in a first-order state of wondering about the world, and being in that state is what lets her express it; she knows what she thinks by looking at the world, not by treating her deliberations as evidence for self-mindreading.
The assumption that ordinary deliberation must be accompanied by mindreading for one to report it is unnecessary and it gets Carruthers into trouble with those cases where subjects lack interpretative evidence from their own behaviour. Subjects report what they are thinking about, and often it is not themselves. If they were really engaged in self-mindreading, you would expect them to talk explicitly about their own beliefs and desires. But they often don't, which suggests that we should understand them as doing something other than either self-monitoring or self-interpretation. The absence of evidence for interpretation that Carruthers frets about is real, but it doesn't support the view he opposes. Rather, it supports the view that that we often know what we think by thinking about the world and not about ourselves. This supports the picture of self-attribution of belief that we find in Evans (Reference Evans1982, pp. 223–26), in which it is often just a matter of making up one's mind. Carruthers acknowledges that in cases of settled belief we can access our beliefs through memory. He reads Evans as showing that metacognitive access can arise through turning our mindreading capacities on our memory reports. But that is needlessly baroque. The simplest theory is that belief self-attributions are often just episodes of remembering.
I can assert a belief that p via the same procedure that would I go through in order to assert that p. This might be the result of working out what I believe via self-mindreading. But in other cases, when I state my belief that p, I am just remembering that p is true. Interpretative evidence is not needed. We do not have to assume that any interpretation is going on at all. In other cases, I put myself in a position to assert a belief by wondering if it is true. This is the way to handle the case of the woman who wonders what car she will go home in; we do not need to think of her as interpreting herself at all. Rather, the beeper leads her to say where her attention is focused, and it is focused on the world, so that is what she talks about. And attending to the world is not introspection or self-interpretation, even if it lets you say what you are thinking about.
Carruthers remains in needless thrall to the idea that metacognition needs a device that is directed at the mind. But when you self-attribute a propositional attitude, you are often not using an inward glance but an outward one: you are thinking about the world. Evans is concerned with this wider capacity to figure things out. When he talks of putting oneself in a position to report a belief, he is thinking of our abilities to deliberate about the world. Carruthers has isolated one way in which we may think about objects in the world – that is, we may treat them as things with minds, and we may look at our own behavior in that light too. But there is no reason to suppose that all our self-attributions come from self-mindreading. Introspection does not rely on any one system, neither an inner eye nor a mindreading device; it depends on all the ways one might think of states of affairs as believable or desirable. We can know our thoughts by looking at the world.
