Philosophy Compass 1/6 (2006): 617–630, 10.1111/j.1747-9991.2006.00043.x

The Ins and Outs of Introspection Philip Robbins* Washington University in St. Louis

Abstract

Introspection admits of several varieties, depending on which types of mental events are introspected. I distinguish three kinds of introspection (primary, secondary, and tertiary) and three explanations of the general capacity: the inside access view, the outside access view, and the hybrid view. Drawing on recent evidence from clinical and developmental psychology, I argue that the inside view offers the most promising account of primary and secondary introspection.

Everyone agrees that introspection involves representing one’s own mental events. It is also generally agreed that introspection is largely under voluntary control, rather than automatic; that it can be selectively directed or pointed, subject to the operation of attentional mechanisms; that it lacks a distinctive phenomenology of its own, over and above whatever phenomenology the introspected events might have; that it gives rise to self-attributions of mental events, typically first-person in form; and last but not least, that it affords epistemic access only to one’s own mind, not to anyone else’s. This minimal description of introspection is bedrock. Beyond it, senses of the term multiply. In a relatively narrow sense, introspection is a capacity to detect those mental states and processes occurring in oneself that are phenomenally conscious, that is, mental states and processes that have a characteristic “feel” to them: pains, itches, emotions, visual and auditory images, imaginings, rememberings, judgings, wonderings, and the like. The deployment of introspection gives rise to meta-consciousness, that is, awareness of being in or undergoing a particular conscious state or process ( Jack and Shallice; Lambie and Marcel; Schooler).1 In a wider sense, introspection is a capacity to detect one’s mental events, whether or not those events are phenomenally conscious. The class of mental events that are introspectible in this sense is arguably more extensive than the class of events that are introspectible in the narrow sense, since it includes propositional attitudes: beliefs, desires, intentions, hopes, and other mental states that plausibly lack phenomenal feel (Carruthers, “Simulation”; Lormand; Nichols and Stich; Robbins).2 In a still wider (and looser) sense, introspection is a capacity to detect one’s own mental events and dispositions, including personality traits (Wilson,“Knowing”; Strangers). © Blackwell Publishing 2006

618 . The Ins and Outs of Introspection

Correlatively, we can distinguish three introspective subcapacities. Primary introspection is a capacity to detect and categorize one’s conscious mental states and processes, such as pains, emotions, and imaginings. Secondary introspection is a capacity to detect and categorize one’s non-conscious mental states, such as beliefs and desires.3 Tertiary introspection is a capacity to detect and categorize one’s dispositional mental states, such as neuroticism and conscientiousness.4 Introspection in the narrow sense above covers just the first of these subcapacities, introspection in the intermediate sense covers the first two, and introspection in the widest sense spans all three. A full account of introspection will need to tell a plausible story about each strand of the general capacity, the first two strands especially. 1. Three Views of Introspection There are three main approaches to the mechanics of introspection. According to the first approach, introspection is a process that affords relatively direct access to one’s mind. It does so in virtue of operating on “inside information” about one’s internal psychological states – information that is very close to the source. In particular, the informational linkage between source and receiver is not mediated by inferential mechanisms of the sort that are plausibly involved in the attribution of mental states to other people, namely, mechanisms that process information based on observations of the target’s overt behavior and circumstances. On the first view, these third-person mindreading mechanisms play no role in introspection.5 Instead, the inferential machinery of introspection taps relatively directly into the machinery of mind itself. I’ll call this the “inside access” view of introspection. Note that there is no implication here that “inside” in “inside access” means inference-free. The idea is rather that the information channel between receiver (one’s introspection device) and source (one’s mind) is relatively direct, since it takes in no “outside information.” But inference is still involved. The exact character of the inference varies, depending on what version of the view is under consideration. Over the years, several possibilities have been defended. According to the phenomenological version, the machinery of introspection is sensitive to the qualitative features of mental states, and classification of one’s mental states proceeds via detection of those features (Goldman,“Psychology”). According to the representationalist-functionalist version, introspection is realized by a “monitoring mechanism” that reads off one’s mental states on the basis of information about their functional profile and representational content (Nichols and Stich; see also Armstrong).6 According to the neural version, introspective mechanisms detect and classify one’s mental states on the basis of information about their neural properties, analogously to the way that interoceptive mechanisms, like pain and hunger, detect one’s bodily states © Blackwell Publishing 2006

Philosophy Compass 1/6 (2006): 617–630, 10.1111/j.1747-9991.2006.00043.x

The Ins and Outs of Introspection . 619

on the basis of information about their physical properties (Goldman, Simulating). This general picture of introspection contrasts sharply with the “outside access” view. On the outside view, introspection consists largely of the processing of information that lies at a greater remove from the target and to which one enjoys no special access, such as information about one’s behavior and circumstances. As with the inside view, there are multiple variants. According to the “theory” theory, introspection exploits the same body of knowledge used to attribute mental states to other subjects, namely, a folk theory of psychology (Gopnik; see also Ryle). According to the “displaced perception” account, introspection uses perceptual information about the local environment to infer how the latter is currently being experienced (Dretske). Similarly, according to the “ascent-routine” account, subjects determine whether or not they believe a given proposition by judging the truth-value of the proposition and inferring accordingly: yes if the proposition is true, no otherwise (Gordon).7 There are substantial differences between these accounts, especially between the theory theory and its rivals. But all three accounts have the same basic structure. They all assume that the introspective judgment is grounded on inference from the observation of mind-external states of affairs, either behavioral or environmental. A third option for theorizing about introspection, the “hybrid” view, splits the difference between the inside and outside access views, incorporating elements of both (Bem). According to the hybrid view, introspection draws on both inside and outside information about the target. Motivating this view is the suspicion that in typical cases, neither form of information on its own suffices for detection and classification of one’s mental states. For introspection to work smoothly, it needs to draw on information from both internal monitoring or scanning (as per the inside access view) and behavioral and environmental observation (as per the outside access view). These views of introspection have different empirical consequences. Since the inside access view assumes that introspection does not implicate mechanisms of the sort that figure in third-person mindreading, it implies that the first capacity should be dissociable, both diachronically and synchronically, from the second (Nichols and Stich).8 Non–theory-theoretic versions of the outside view, such as the displaced perception account, are on a par with the inside view in this respect, since they too assume that introspection is functionally independent of third-person mindreading. In contrast to the inside view, the theory theory assumes that introspection and third-person mindreading are subserved by a single mechanism, so it precludes the possibility of the first capacity being dissociated from the second (Gopnik; Carruthers,“Simulation”; Frith and Happé). The hybrid view says that introspection implicates third-person mindreading mechanisms in an essential way, so it too rules out dissociation. Unlike the theory theory, © Blackwell Publishing 2006

Philosophy Compass 1/6 (2006): 617–630, 10.1111/j.1747-9991.2006.00043.x

620 . The Ins and Outs of Introspection

however, it leaves open the possibility that full-blown introspection comes on line after third-person mindreading does (but not before). Hence, unlike the theory theory, the hybrid view is consistent with diachronic dissociation, provided that the temporal direction of dissociation is reversed. In what follows, we’ll see how these predictions fare with respect to the empirical record. The conceptual and empirical geography of this area is complex. To do full justice to this complexity, we would need to consider each of the three types of introspection identified above (primary, secondary, and tertiary) at length. I will focus on the case of primary introspection: partly because this is the core form of introspection as traditionally conceived (and arguably the heart of the matter), and partly because the evidential picture is more detailed here than it is in the other two cases (see section 2). The treatment of secondary introspection will be more sketchy, and the case of tertiary introspection will largely be left to the side (see section 3). 2. Primary Introspection As noted earlier, the term “primary introspection” refers to the detection and classification of a specific class of mental events: phenomenally conscious states and processes, such as pains, emotions, imaginings, and rememberings. Of the three types of introspection identified above, the primary type is plainly the best candidate for explanation along the lines of the inside access view. Popular wisdom suggests that our own phenomenally conscious states are precisely those states to which each of us enjoys immediate, privileged, and relatively unfettered (and unfiltered) epistemic access – access of a higher and more intimate sort than we have to the phenomenal states of others. A natural way to test this view is to see whether there are individuals whose primary introspective capacity is intact despite defects in their capacity to attribute phenomenal states to others (synchronic dissociation), or whether introspection comes on line in normal individuals before third-person mindreading does (diachronic dissociation). Evidence of such dissociations would support the inside access view and undermine both the theory theory and the hybrid view. 2.1. THE CASE FOR DISSOCIATION

Both friends and foes of the inside view have appealed to clinical studies of autistic spectrum disorders, especially Asperger syndrome, to make their case (Carruthers, “Autism”; Frith and Happé; Nichols and Stich). Autism is a pervasive neurodevelopmental disorder that adversely affects social cognitive function, including capacities for social interaction, communication, and empathy. At the severe end of the autistic spectrum is Kanner syndrome, which is marked by low IQ as well as language deficits. At the mild end is Asperger syndrome, in which general intelligence and core language skills are spared. Across the spectrum one finds impairments of third-person © Blackwell Publishing 2006

Philosophy Compass 1/6 (2006): 617–630, 10.1111/j.1747-9991.2006.00043.x

The Ins and Outs of Introspection . 621

mindreading, as gauged by performance on various tests of the capacity to parse the social world in mentalistic terms. A paradigmatic example of such tests is the Sally-Anne task, which is used to assess understanding of the concept of belief. In this task, the subject observes a puppet show featuring two characters, Sally and Anne. Sally has a basket and Anne has a box; Sally puts her marble into the basket and leaves; while Sally is away,Anne removes the marble from the basket and puts it into the box. The subject is then asked where Sally thinks her marble is, or where Sally will look for it. To answer correctly, the subject needs to understand that Sally’s beliefs are distinct from her own, and to resist assimilating the former to the latter. Other tests of third-person mindreading focus on production and comprehension of deception, conversational pragmatics, understanding of figurative language (metaphor, sarcasm, and irony), recognition of emotions from facial expressions, and understanding stories involving bluff and double bluff (Baron-Cohen). Proponents of the inside access view argue that individuals with Asperger syndrome have no difficulty introspecting their mental states, despite their difficulty attributing such states to other people (Nichols and Stich). Since the topic of this section is primary introspection, let us consider this claim with specific reference to phenomenal mental states. There are several pieces of evidence that bear on the issue. The first comes from a clinical study of introspection using the “descriptive experience sampling” method developed by Russell Hurlburt (Hurlburt, Happé, and Frith). In this experimental paradigm, subjects are instructed to carry a small device that beeps at random intervals, and to respond to each beep by “freezing” the contents of their awareness and jotting down a description of it in a notebook. Three adults with Asperger syndrome participated in the study by sampling and recording their experiences for a week. All three of the subjects were able to produce introspective reports, albeit with varying degrees of success. This lends prima facie support to the dissociation hypothesis. On closer inspection, however, this interpretation of the results appears questionable. Here are three points that invite skepticism. First, the sample size (n = 3) is too small for the results of the study to carry much weight. Second, even within this small sample, proficiency at experience sampling co-varied with proficiency at third-person mindreading, judging from the subjects’ performance on false-belief tasks. Granted, false-belief tests do not tap the ability to attribute conscious mental states per se. But successful performance on these tests does correlate with the latter ability, and it does indicate some capacity for sophisticated third-person mindreading. This pattern of results militates against dissociation. Third, subjects’ reports were highly anomalous. In studies of descriptive experience sampling by normal subjects, the reports mention several different phenomenal state types, including emotional feelings, bodily sensations, visual images, and inner speech (Hurlburt; see also Hurlburt and Heavey). By contrast, the subjects © Blackwell Publishing 2006

Philosophy Compass 1/6 (2006): 617–630, 10.1111/j.1747-9991.2006.00043.x

622 . The Ins and Outs of Introspection

in the Asperger study reported visual experiences, visual imagery in particular, almost exclusively. On the assumption that the phenomenal experience of Asperger subjects is even remotely like that of normal individuals (that is, not restricted to the visual modality), this suggests that introspection is impaired in the disorder. Autobiographical reports from Asperger individuals, which refer to experiences across a range of sensory modalities, make this assumption hard to deny.9 Against this, one might read the autobiographical reports as evidence that primary introspection is intact in Asperger syndrome (Nichols and Stich). But this claim is hard to justify, for several reasons. First, since long-term autobiographical memory is notoriously unreliable even in normal individuals, and tends to be highly reconstructive and “theory-laden,” reports of this kind cannot be taken at face value. This goes double for reports from clinical subjects suffering from a range of cognitive deficits. Second, the reports in question were not produced by a systematic interview process, and partly as a result of this, they are difficult to interpret. In particular, they do not provide clear evidence that these subjects have normal meta-consciousness. What authors sometimes report as anomalous sensory experiences (e.g., insensitivity to pain), could just as well be accounted for in terms of defective introspection (that is, failure to identify and classify pains as such).10 Third, and most tellingly, there is independent experimental evidence of defects in emotion introspection in Asperger subjects. This last point is worth considering in detail. There is little doubt that Asperger subjects have difficulty recognizing the emotions of people around them. Evidence to this effect is sufficiently robust to motivate ongoing discussions of the “amygdala theory” of autism (Dziobek et al.; but see Castelli for qualifications). It appears that these individuals also have difficulty recognizing their own emotions. Indeed, a standard component of the symptomology of autistic spectrum disorders is “alexithymia,” understood as the inability to identify and describe one’s own emotions, or lack of emotional insight (Hill, Berthoz, and Frith). Alexithymia is assessed on the basis of self-reports generated using the Toronto Alexithymia Scale, which includes items such as “I am often confused about what emotion I am feeling” and “When I am upset, I don’t know if I am sad, frightened, or angry.” Subjects rate each item using a 5-point Likert scale (1 = strongly agree, 5 = strongly disagree). In a recent large-scale study, nearly half of the participants with autistic spectrum disorders (nearly all of them Asperger’s) ranked as severely impaired at emotion self-processing, whereas none of the normal participants fell into this category (Hill, Berthoz, and Frith). The results of this and related clinical studies (e.g., Berthoz and Hill) suggest that emotion introspection is impaired in Asperger syndrome. Another source of evidence bearing on the dissociation hypothesis comes from studies of meta-cognition – specifically, studies of meta-memory – in autistic children (Farrant, Boucher, and Blades). In one experiment, subjects © Blackwell Publishing 2006

Philosophy Compass 1/6 (2006): 617–630, 10.1111/j.1747-9991.2006.00043.x

The Ins and Outs of Introspection . 623

were given a simple memory test (memorizing a list of numbers) and then asked what strategy they had used to perform the task (“What did you do to help you to remember all the numbers?”). In order to answer the meta-memory question, subjects presumably had to recall from episodic memory how they had gone about memorizing the list of numbers, introspect that remembrance of thought processes past, and then report that introspection. Sample responses referred to strategies like cumulative rehearsal: “When you said ‘one,’ ‘six,’ ‘four,’ I remembered ‘one,’ ‘six,’ ‘four’ in my mind, then got the other number, then said it all together” (Farrant, Boucher, and Blades 118). Children with autism performed remarkably well on this task; in fact, there was no statistically significant difference between their performance and that of normal children. Nichols and Stich interpret this result as supporting their hypothesis that introspection is dissociable from third-person mindreading. But there are two problems with this interpretation of the data. First, the autistic children in this study performed unusually well on standard false-belief tasks, suggesting that their capacity for third-person mindreading may have been relatively intact. Hence, the evidence for dissociation here is weak at best. Second, in a separate study of meta-cognition in autistic children, Farrant and colleagues tested children on “recall readiness,” the ability to judge when one has successfully committed information to memory (Farrant, Blades, and Boucher). Here they found that children with autism performed significantly less well than normals. In contrast to the study described above, the subjects in this study were not tested on false-belief tasks. Had those tests been done, they might have shown that performance on the recall readiness task positively correlates with performance on tests of third-person mindreading – contrary to what the dissociation hypothesis predicts. One last piece of evidence bearing on the issue of synchronic dissociation comes from self-reports of individuals with Asperger syndrome regarding “private self-consciousness,” that is, the frequency with which one reflects on oneself as a subject of thoughts, feelings, and other mental states (Blackshaw et al.). The study in question used the Self-Consciousness Scale (Fenigstein, Scheier, and Buss), the relevant subscale of which – the Private Self-Consciousness Scale (PrSC) – includes the following items: • • • • •

I’m generally attentive to my inner feelings. I’m aware of the way my mind works when I work through a problem. I’m alert to changes in my mood. I’m often the subject of my own fantasies. I sometimes have the feeling that I’m off somewhere watching myself.

Blackshaw and her team found that Asperger subjects scored significantly higher on the PrSC than normals. Since PrSC scores presumably track introspective ability to some extent, this result suggests that introspection is relatively intact in Asperger syndrome. The Asperger subjects also scored lower than normals on the Projective Imagination Test, a test of third-person © Blackwell Publishing 2006

Philosophy Compass 1/6 (2006): 617–630, 10.1111/j.1747-9991.2006.00043.x

624 . The Ins and Outs of Introspection

mindreading ability in which subjects are shown a line drawing of a social situation and asked to describe the situation in mentalistic terms (e.g.,“What are the people in the picture thinking or feeling?”). These findings provide modest support for the dissociation hypothesis. Let us pause briefly to take stock. We’ve just reviewed a fair bit of evidence bearing on the hypothesis that primary introspection is dissociable from third-person mindreading. The overall pattern of results, though suggestive, is inconclusive. As things stand, the dissociation hypothesis is not clearly confirmed by the clinical record. To get more empirical traction on this hypothesis, we need to consider studies of introspection in normally developing individuals. Relative to the case of synchronic dissociation, evidence regarding diachronic dissociation is in short supply. But it is also more telling. For example, in a study of young children’s capacity to report on the imaginative pretenses of others, a group of three-year-olds were shown a scene from a television show in which the characters pretended to be on an airplane while sitting on a bench (Rosen, Schwebel, and Singer). The children were then asked whether the characters were thinking about being on a plane or sitting on a bench, and the vast majority (90%) of them gave the wrong answer. An earlier study testing the ability of three-year-olds to report on their own immediately past pretenses, by contrast, yielded the opposite result (Gopnik and Slaughter). The children were instructed to close their eyes and think of a blue dog, and then to repeat the exercise by closing their eyes and thinking of a red balloon. At that point they were asked what they had been thinking of first, a blue dog or a red balloon. The vast majority (80%) of the children answered correctly. Though it is possible that subjects in this study answered the experimenter’s question without recourse to introspection (e.g., by recalling the experimenter’s instructions), the data are suggestive. A similar pattern of results – that is, weak attribution of phenomenal states to others together with strong attribution of phenomenal states to self – is suggested by studies of visual perspective-taking in young children (Gopnik and Slaughter). In one study, the children were seated across a table from the experimenter and shown a drawing that looked either like a turtle standing on its feet or like a turtle lying on its back, depending on which side of the table one was seated at. The children were asked first about the apparent orientation of the turtle from their vantage point, and then about its apparent orientation from the vantage point of the experimenter. Three-year-olds performed poorly on this third-person mindreading task, with at most half of them answering correctly. But in a first-person variant of the task, the three-year-olds did much better. In this version of the task, the children looked at the drawing once, switched seats with the experimenter, and looked at it again. At that point they were asked how the turtle had looked to them before the perspective shift. This time, three-quarters of the children gave the right answer. © Blackwell Publishing 2006

Philosophy Compass 1/6 (2006): 617–630, 10.1111/j.1747-9991.2006.00043.x

The Ins and Outs of Introspection . 625

None of this bodes well for the theory theory or the hybrid view. But nor should it come as any great surprise. The idea that we typically detect and classify our phenomenal states by reflecting on our behavior is wildly counterintuitive. Indeed, it is doubtful that anyone has defended such a view.11 Rather, proponents of these views tend to focus on the self-attribution of beliefs, desires, and other (plausibly qualia-free) mental states: the domain of secondary introspection. In the domain of primary introspection, neither the theory theory nor the hybrid view is a serious contender. This much seems clear concerning the introspection of certain categories of phenomenal states, including conscious mental processes, such as episodic rememberings, and bodily sensations, such as pains and itches. Whether it carries over equally to the case of emotions, moods, and other types of phenomenal states, is open to question. 2.2. BEYOND DISSOCIATION

The results of our empirical analysis present us with a choice. Either we endorse the inside view, or we fall back on a version of the outside view which is consistent with the dissociation hypothesis. There are two main options: introspection as displaced perception (Dretske) and introspection via ascent routines (Gordon). Neither is promising as an account of primary introspection. Let’s take them in reverse order. The ascent-routine view is in the first instance an account of how we introspect beliefs, and by extension how we introspect other propositional attitudes (desires, intentions, and the like). It does not explain how we introspect phenomenal states and processes, and it offers no help with this explanatory problem. As an account of primary introspection, it is a non-starter. The displaced perception view is only slightly better off. According to this view, we introspect by inferring our experiential state from information about the world gained through perception. For example, we look outside the window and we see that the sky is blue, and we infer from the fact that the sky is blue that we are experiencing it as such. This story is fine as far as it goes, but it doesn’t go very far. For one thing, it doesn’t begin to explain how we introspect pains, itches, emotions, rememberings, and a whole host of other phenomenal states and processes for which there is no clear environmental correlate. For another thing, it fails to explain how the inference from perceptual belief to introspective belief is mediated, that is, it fails to specify the mechanism responsible for this transition in thought. Imagine that you look at an object on your desk (a coffee cup, say) and see that it is cylindrical. Why do you infer from the fact that the object in front of you is cylindrical that you are seeing it as such? Why don’t you infer instead that you are feeling it as such, that is, having a tactile experience of its shape? As Goldman (Simulating) notes, the displaced perception view is strangely silent about how we detect and classify the functional (vs. representational) dimension of our mental © Blackwell Publishing 2006

Philosophy Compass 1/6 (2006): 617–630, 10.1111/j.1747-9991.2006.00043.x

626 . The Ins and Outs of Introspection

states. In that sense, the view seems radically underspecified, and in need of filling-in by some variant of the inner access view. 3. Non-primary Introspection Secondary introspection is the detection and classification of one’s mental states that are widely (though far from universally) thought of as lacking distinctive qualia. The paradigm cases of this class of states are the propositional attitudes: beliefs, desires, intentions, hopes, and the like. Assuming there is nothing that it’s like to be in such states, what’s the best way to explain our capacity to attribute them to ourselves? As before, in the case of primary introspection, there are three main options: inside access, outside access, or a combination of the two. And as before, there are the corresponding predictions. The inside view implies the possibility of secondary introspection being dissociated from third-person mindreading, and non–theory-theoretic versions of the outside view do the same; the theory theory and the hybrid view rule out this possibility (though the hybrid view allows for backwards diachronic dissociation). In general, the empirical picture looks much the same here as it did in the primary case: weak on the clinical side (evidence of synchronic dissociation), stronger on the developmental side (evidence of diachronic dissociation). Though it might appear that individuals with Asperger syndrome can introspect their beliefs, desires, intentions, and the like normally despite difficulties attributing such states to others (Nichols and Stich), there is little evidence to back up this dissociation claim (Robbins). Descriptive experience sampling studies do not speak to the issue, since they do not address the ability to introspect attitudes, and autobiographical reports are suspect on methodological and conceptual grounds. Furthermore, there is evidence that children with autistic spectrum disorders have difficulty introspecting their intentions (Phillips, Baron-Cohen, and Rutter). In a study done with four-year-olds, subjects were instructed to shoot a toy gun at a row of targets after indicating which target they’d selected. The outcome of the shooting was manipulated to ensure that the actual outcome sometimes failed to coincide with the subject’s stated intention. The children were then asked which target they had intended to hit. Normal four-year-olds answered the question correctly, regardless of whether the actual outcome corresponded to their stated intention. Autistic children, on the other hand, always replied that the actual outcome was the intended one. The results of this study cast doubt on the idea of autism as a disorder in which secondary introspection is intact. As for the developmental relation between secondary introspection and third-person mindreading, direct evidence is hard to come by. But there is some data supporting dissociation. In particular, studies of young children’s understanding of the relation between seeing and knowing suggest that young children are better at reporting their own epistemic states than the epistemic states of others under similar conditions (Wimmer, Hogrefe, and © Blackwell Publishing 2006

Philosophy Compass 1/6 (2006): 617–630, 10.1111/j.1747-9991.2006.00043.x

The Ins and Outs of Introspection . 627

Perner). In one study, one group of three-year-olds were allowed to look inside a box, while another group were not allowed to look. The children were then asked whether or not they knew what was in the box. The majority of children answered correctly (those who looked said yes, those who didn’t look said no). Then the children observed another person either looking or not looking inside a box, after which they were asked whether or not that person knew what was inside. In this third-person version of the task, substantially fewer children got the right answer. This is bad news for the theory theory (Nichols and Stich), and equally bad news for the hybrid view. Once again, we’re left to choose between the inside view and a version of the outside view that is not disconfirmed by the dissociation data. We already noted some liabilities of the displaced perception view, in particular the fact that it cannot account for our ability to introspect the functional type of our mental states. Clearly, an account of secondary introspection that cannot explain how we introspectively distinguish between beliefs, desires, intentions, and other propositional attitudes will be out of the running. So that rules out Dretske’s story. Gordon’s ascent-routine story is problematic for a similar reason: namely, it cannot readily be extended to cover the case of attitudes other than belief (Nichols and Stich). It’s plausible enough, at least on its face, that we determine whether or not we believe the sky is blue by checking to see if the sky is blue and “answering” accordingly (yes if it is, no if it’s not). It’s totally implausible that we determine whether or not we hope the sky is blue, or whether we expect it to be blue, on this basis. That limitation of explanatory scope – in addition to a failure to specify how the inference from belief about the world to belief about one’s belief about the world is mediated (by what mechanism, if not some sort of monitoring or scanning device?) – seems fatal to the ascent-routine approach. So here too the inner access view prevails. Things look very different in the case of tertiary introspection, since the self-attribution of personality traits – states which are bound up with stable, long-term behavioral dispositions – is not plausibly driven by inside information. Relatedly, recent evidence suggests that third-person attribution of personality traits and third-person attribution of emotions are underpinned by distinct bits of neural circuitry, and that these capacities are functionally dissociable (Heberlein et al.). If this point carries over to first-person attributions, that would reinforce the idea that tertiary introspection and primary introspection are very different animals. For this and other reasons, it seems unlikely that there is a single, unified account of introspection that explains all three species of the genus.12 Notes * Correspondence address: Philosophy-Neuroscience-Psychology Program,Washington University in St. Louis, Campus Box 1073, 1 Brookings Drive, St. Louis, MO 63130, USA. Email: [email protected]. © Blackwell Publishing 2006

Philosophy Compass 1/6 (2006): 617–630, 10.1111/j.1747-9991.2006.00043.x

628 . The Ins and Outs of Introspection 1

The conceptual distinction between consciousness and meta-consciousness (also known as reflexive consciousness, reflective awareness, and meta-awareness) is empirically supported by evidence of dissociations between them, as in studies where people catch themselves in the act of “zoning-out” during reading (Schooler). Hence, as Williams James observed a long time ago, “we must never take a person’s testimony, however sincere, that he has felt nothing, as proof positive that no feeling has been there” (James 211). 2 For dissenting voices on the phenomenology of the attitudes, see e.g., Flanagan; Goldman, “Psychology”; Chalmers; Horgan and Tienson. 3 The distinction between primary and secondary introspection is a first-person analog of the distinction between the phenomenal and intentional stances (Robbins and Jack). To take the phenomenal stance toward something is to think of it as a locus of phenomenal experience, whereas to take the intentional stance toward something is to think of it as a locus of intentionality. Preliminary evidence from clinical studies of autism and psychopathy, and functional neuroimaging of normal individuals, suggests that these stances are functionally and neurally distinct. 4 This extension of the term “introspection,” though not uncommon in social and personality psychology, is rare in philosophy. Partly for that reason, little will be said about it here. 5 The same point applies to simulationist accounts of third-person mindreading, since on the inside access view, introspection does not implicate off-line simulation. But simulation-based mindreading does implicate introspection, since simulation routines are generally thought to bottom out in mental self-attribution (Goldman, Simulating). 6 See Jack and Shallice and Schooler for accounts of introspection in which the relevant information concerns exclusively functional and exclusively representational properties, respectively. 7 Though Gordon presents ascent routines as an alternative to introspection, what they really offer is an alternative to introspection as conceived on the inside access view. For present purposes, the ascent-routine story still qualifies as a story about how introspection works. 8 Note that to say of capacities X and Y that X is dissociable from Y is not (yet) to say that X and Y are doubly dissociable, that is, dissociable in both directions. Capacities X and Y are doubly dissociable just in case X is dissociable from Y and vice versa (Shallice). Thus, it’s at least conceptually possible that introspection is dissociable from third-person mindreading but not the other way around. Some proponents of simulation theory claim this as a matter of empirical fact (Goldman, Simulating). It is an interesting question, so far unaddressed in the literature, what theorists of this stripe should say about the fact that schizophrenic individuals who present with passivity symptoms (e.g., delusions of control and thought insertion) appear to have defective introspection but normal third-person mindreading (Nichols and Stich; Robbins). 9 For example, in her autobiography Gunilla Gerland writes: “All the time I was growing up, I suffered from an almost constant shudder down my spine . . . It was a constant torture . . . It was like cold steel down my spine”; and “If I was made to touch jewellery, I felt a sharp whistling metallic noise in my ears, and my stomach turned over. Like a note falsely electrified, that sound would creep from the base of my spine upwards until it rang in my ears, tumbled down into my throat and settled like nausea in my stomach” (54). 10 Gerland writes: “My insensitivity to pain was now as good as total . . . nothing hurt at all. And yet I felt – my actual feelings were not shut off – because when I was aware that I had injured myself somewhere, I could sense something, a non-pain, which branched out into my body from the place where the injury was. But the fact was, it didn’t hurt” (157). Since it cannot be assumed that Gerland is alert to the distinction between phenomenal consciousness and introspective meta-consciousness – and to the possibility that these capacities can and do come apart, even in normal cases (see note 1) – passages like these do not show that Asperger subjects are introspectively normal. Indeed, they can just as plausibly be read as evidence to the contrary. 11 Gopnik is no exception. Though she staunchly opposes perceptual models of introspection, she concedes that there are cases in which there is “genuine perception of a psychological state” (Gopnik 12). Talk of perception here is shorthand for talk of relatively direct, subjectively privileged epistemic access, along the lines of the inside view. 12 Many thanks to Murat Aydede, Tony Jack, Shaun Nichols, Adam Shriver, Jeff Zacks, and an anonymous referee for helpful comments and advice. I’m especially grateful to Shaun for the stimulus of his own work on the topic, and for lively and enlightening discussion of the issues over the past several years. © Blackwell Publishing 2006

Philosophy Compass 1/6 (2006): 617–630, 10.1111/j.1747-9991.2006.00043.x

The Ins and Outs of Introspection . 629

Works Cited Armstrong, D. M. A Materialist Theory of the Mind. London: Routledge, 1968. Baron-Cohen, S. “Theory of Mind and Autism: A Fifteen Year Review.” Understanding Other Minds: Perspectives from Developmental Cognitive Neuroscience. Eds. S. Baron-Cohen, H. Tager-Flusberg, and D. J. Cohen. 2nd ed. Oxford UP, 2000. 3–20. Bem, D. J. “Self-Perception Theory.” Advances in Experimental Social Psychology. Vol. 6. Ed. L. Berkowitz. New York:Academic Press, 1972. 1–62. Berthoz, S. and Hill, E. L. “The Validity of Using Self-Reports to Assess Emotion Regulation Abilities in Adults with Autism Spectrum Disorder.” European Psychiatry 20 (2005): 291–8. Blackshaw, A., P. Kinderman, D. J. Hare, and C. Hatton. “Theory of Mind, Causal Attribution and Paranoia in Asperger Syndrome.” Autism 5 (2001): 147–63. Bowler, D. M. “ ‘Theory of Mind’ in Asperger’s Syndrome.” Journal of Child Psychology and Psychiatry 33 (1992): 873–93. Carruthers, P. “Autism as Mind-blindness: An Elaboration and Partial Defence.” Theories of Theories of Mind. Eds. P. Carruthers and P. K. Smith. Cambridge: Cambridge UP, 1996. 257–73. ——. “Simulation and Self-knowledge: A Defense of Theory-theory.” Theories of Theories of Mind. Eds. P. Carruthers and P. K. Smith. Cambridge: Cambridge UP, 1996. 22–38. Castelli, F. “Understanding Emotions from Standardized Facial Expressions in Autism and Normal Development.” Autism 9 (2005): 428–49. Chalmers, D. The Conscious Mind: In Search of a Fundamental Theory. Oxford: Oxford UP, 1996. Dretske, F. Naturalizing the Mind. Cambridge, MA: MIT Press, 1995. Dziobek, I., S. Fleck, K. Rogers, O. Wolf, and A. Convit. “The ‘Amygdala Theory Of Autism’ Revisited: Linking Structure to Behavior.” Neuropsychologia 44 (2006): 1891–9. Farrant, A., M. Blades, and J. Boucher. “Recall Readiness in Children with Autism.” Journal of Autism and Developmental Disorders 29 (1999): 359–66. ——, J. Boucher, and M. Blades. “Metamemory in Children with Autism.” Child Development 70 (1999): 107–31. Fenigstein, A., M. F. Scheier, and A. H. Buss. “Public and Private Self-consciousness: Assessment and Theory.” Journal of Clinical and Consulting Psychology 43 (1975): 522–7. Flanagan, O. Consciousness Reconsidered. Cambridge, MA: MIT Press, 1992. Frith, U. and F. Happé. “Theory of Mind and Self-consciousness: What is it Like to be Autistic?” Mind and Language 14 (1999): 1–22. Gerland, G. A Real Person: Life on the Outside. London: Souvenir Press, 1997. Goldman, A. I. “The Psychology of Folk Psychology.” Behavioral and Brain Sciences 16 (1993): 15–28. ——. Simulating Minds: The Philosophy, Psychology, and Neuroscience of Mindreading. Oxford: Oxford UP, 2006. Gopnik, A. “How we Read our own Minds: The Illusion of First-person Knowledge of Intentionality.” Behavioral and Brain Sciences 16 (1993): 1–14. —— and V. Slaughter. “Young Children’s Understanding of Changes in their Mental States.” Child Development 62 (1991): 98–110. Gordon, R. “Simulation without Introspection or Inference from Me to You.” Mental Simulation: Evaluations and Applications. Eds.T. Stone and M. Davies. Oxford: Blackwell, 1995. 53–67. Heberlein, A. S., R. Adolphs, D. Tranel, and H. Damasio. “Cortical Regions for Judgments of Emotions and Personality Traits from Point-light Walkers.” Journal of Cognitive Neuroscience 16 (2005): 1143–58. Hill, E., S. Berthoz, and U. Frith. “Brief Report: Cognitive Processing of own Emotions in Individuals with Autistic Spectrum Disorder and in their Relatives.” Journal of Autism and Developmental Disorders 34 (2004): 229–35. Horgan, T. and J. Tienson. “The Intentionality of Phenomenology and the Phenomenology of Intentionality.” Philosophy of Mind: Classical and Contemporary Readings. Ed. D. Chalmers. Oxford: Oxford UP, 2002. 520–33. Hurlburt, R. T. Sampling Normal and Schizophrenic Inner Experience. New York: Plenum Press, 1990. ——, F. Happé, and U. Frith. “Sampling the Form of Inner Experience in Three Adults with Asperger Syndrome.” Psychological Medicine 24 (1994): 385–95. © Blackwell Publishing 2006

Philosophy Compass 1/6 (2006): 617–630, 10.1111/j.1747-9991.2006.00043.x

630 . The Ins and Outs of Introspection —— and C. L. Heavey. Exploring Inner Experience: The Descriptive Experience Sampling Method. Amsterdam: John Benjamins, 2006. Jack, A. I. and T. Shallice. “Introspective Physicalism as an Approach to the Science of Consciousness.” Cognition 79 (2001): 161–96. James,W. The Principles of Psychology. Vol. 1. New York: Holt, 1890. Lambie, J. A. and A. J. Marcel. “Consciousness and Emotion Experience: A Theoretical Framework.” Psychological Review 109 (2002): 219–59. Lormand, E. “Nonphenomenal Consciousness.” Noûs 30 (1996): 242–61. Nichols, S. and S. Stich. Mindreading: An Integrated Account of Pretence, Self-awareness, and Understanding other Minds. Oxford: Oxford UP, 2003. Phillips,W., S. Baron-Cohen, and M. Rutter. “Understanding Intention in Normal Development and in Autism.” British Journal of Developmental Psychology 16 (1998): 337–48. Robbins, P. “Knowing Me, Knowing You: Theory of Mind and the Machinery of Introspection.” Journal of Consciousness Studies 11 (2004): 129–43. —— and A. I. Jack. “The Phenomenal Stance.” Philosophical Studies 127 (2006): 59–85. Rosen, C., D. Schwebel, and J. Singer. “Preschoolers’ Attributions of Mental States in Pretense.” Child Development 68 (1997): 1133–42. Ryle, G. The Concept of Mind. Chicago: U of Chicago P, 1949. Schooler, J. “Re-representing Consciousness: Dissociations between Experience and Metaconsciousness.” Trends in Cognitive Sciences 6 (2002): 339–44. Shallice,T. From Neuropsychology to Mental Structure. Cambridge: Cambridge UP, 1988. Wilson, T. D. “Knowing When to Ask: Introspection and the Adaptive Unconscious.” Journal of Consciousness Studies 10 (2003): 131–40. ——. Strangers to Ourselves: Discovering the Adaptive Unconscious. Cambridge, MA: Harvard UP, 2002. Wimmer, H., G. Hogrefe, and J. Perner. “Children’s Understanding of Informational Access as a Source of Knowledge.” Child Development 59 (1988): 386–96.

© Blackwell Publishing 2006

Philosophy Compass 1/6 (2006): 617–630, 10.1111/j.1747-9991.2006.00043.x