1 The epistemic indispensability of introspection

Zoologists who study zebras pay close attention to the impact of zebras on their environments, the environment’s own impact on zebra populations, various correlates and indicators or zebra presence, and so on. But in constructing their theories of zebras, zoologists use not only observations of zebras’ causal traces, conditions of zebras’ presence, and zebra correlates. They also use observations of zebras themselves. Indeed, given that it is possible to observe zebras, it would be folly for zoologists to refuse to take into account observations of zebras in constructing their theories of zebras. When studying leptons, we must construct our theories without taking into account direct observations of leptons, since we cannot observe the leptons themselves. But given that we can observe zebras, it would be perverse to construct our theories thereof without taking zebra observations into account. More generally, whenever we can observe a type of phenomenon, it is perverse to insist on developing our understanding of it in complete disregard of our observations of tokens.

This general principle has immediate implications for our understanding of the mind. If some mental phenomena can be observed, it would be very odd indeed to insist on bracketing all such observation in developing our mature understanding of the mind. It is of course controversial whether we can observe any mental phenomena in something like the sense in which we can observe zebras. There is a tradition, going back at least to Locke, that considers introspection a kind of inner observation, whereby some mental phenomena are observed in much the same sense in which zebras are observed. This tradition has been under sustained attack for much of the twentieth century. Nonetheless, I have argued elsewhere (Kriegel 2011a, Ch. 1) that it is fundamentally correct: introspection does afford us a sort of observational contact with (some) conscious experiences.Footnote 1 If so, the following argument suggests itself:

(1):

When a phenomenon P is observable, any legitimate understanding of P must take account of observations of P.Footnote 2

(2):

Some mental phenomena—certain conscious experiences—are introspectively observable. So:

(3):

Any legitimate understanding of the mind must take account of introspective observations of conscious experiences

This is an argument for what one might call the epistemic indispensability of introspection.

It is possible to resist this argument by rejecting the observational model of introspection (Premise 2), but as noted I have defended it elsewhere. Another option is to deny the general principle that observation is crucial to our understanding of observable phenomena (Premise 1). It might be claimed, for example, that the principle holds only where the relevant type of observation is minimally trustworthy, and that introspective observation is not even that. One might be tempted to read Schwitzgebel (2011) as pressing this sort of challenge to the epistemic indispensability of introspection. A more accurate interpretation, however, is that Schwitzgebel concedes this epistemic indispensability but concludes from it that a scientifically respectable understanding of the mind must be beyond reach, given that introspection is so utterly untrustworthy. The main goal of this paper is to address this challenge from Schwitzgebel (§3). Before doing so, however, let me consider a more straightforward objection to the above argument: that introspection cannot be epistemically indispensable given that cognitive science has flourished without it (§2).

2 Introspection and cognitive-scientific practice

To my mind, there are two problems with the idea that cognitive science is doing fine without introspection. The first is that it is not doing fine. The second is that it is not doing without introspection. It is a familiar comment—though admittedly a controversial one—that cognitive science has met with limited success when it comes to phenomenal consciousness.Footnote 3 This is the sense in which cognitive science is not doing fine without introspection, and I will not belabor the point here. Let me focus rather on my claim that cognitive science is not doing without introspection to begin with.

To appreciate the enduring role of introspection in cognitive science, let us start with the distinction, often attributed to Reichenbach but present already in Bolzano (1837, §15), between the ‘context of discovery’ and the ‘context of justification.’ A dramatic example is provided by the German chemist Friedrich Kekulé’s discovery of the molecular structure of benzene. The evidence Kekulé (1865) cited in justification of his model of benzene had to do with isomers, derivatives, and so on. But a quarter-century later Kekulé recounted that his initial discovery was due to a daydream in which he ‘saw’ a snake biting its own tail. Clearly, in this case, the manner in which the model was discovered is irrelevant to the manner in which it is justified. Thus the contexts of discovery and justification can come apart, even if typically the gap between them is not this dramatic.

My claim in this section is that even if introspection has been purged from the context of cognitive-scientific justification, it certainly continues to underlie large segments of research in the context of cognitive-scientific discovery. In fact, significant portions of modern cognitive science strike me as based on introspective discovery paving the way to non-introspective justification. Often the scientist, being a reflective introspector, experiences an initial introspective insight into some psychological phenomenon, and on its basis forms a hypothesis; s/he then proceeds to devise experimental tasks that ingeniously use exclusively third-person measures (often reaction times) to generate non-introspective evidence for the introspectively formed hypothesis.

A fine example of this is Roger Shepard’s seminal work on imagery and the phenomenon of mental rotation (Shepard and Metzler 1971). That we use mental rotation of private images to compare shapes of objects imaged is of course what introspection teaches. The ingenuity in Shepard’s research was to devise an experimental paradigm in which pairs of similarly shaped but differently oriented three-dimensional ‘objects’ with varying angles of putative rotation were to be judged for similarity by subjects. The fact that subjects took longer to judge the shapes to be similar when the angle of putative rotation was greater suggested that these subjects were indeed engaged in mental rotation. Thus the purely ‘objective’ (read: third-person) measure of reaction time served to ratify what was already known on the basis of ‘subjective’ (first-person) introspective impression.Footnote 4

A more recent example is Vilayanur Ramachandran’s ingenious demonstration of number-color synaesthesia (Ramachandran and Hubbard 2001). Ramachandran used panels of numerals printed in a way that made it difficult to distinguish different numerals. Normal subjects took significantly longer to identify incongruent numerals than number-color synaesthetes, to whom the incongruent numerals presumably appeared incongruently colored. Of course, that number-color synaesthesia exists we know on the basis of synaesthetes’ introspective reportage since at least the nineteenth century (see Galton 1880). But Ramachandran’s reaction-time-based demonstration had the advantage of purging appeal to introspection in ratifying this knowledge.Footnote 5

From casual observation of cognitive-science conferences and colloquia, my impression is that this sort of gambit is pervasive in vision science and throughout cognitive (neuro-)psychology. Scientists often devise ingenious experimental designs that circumvent explicit appeal to introspection, but the original hunch underlying the research is founded on personal introspection.Footnote 6

It is an open question just how far cognitive science would get if it purged introspection not only from the context of justification but also from the context of discovery. Suppose cognitive science insisted, from its inception, not only on devising non-introspective justification of introspectively formed hypotheses, but also on exclusively non-introspective hypothesis formation. In fact, consider a possible world otherwise like ours but where cognitive scientists lack (and always have lacked) any introspective capacities. My own suspicion is that we would be shocked to find out just how skeletally poor the scientific understanding of the mind is in such a world.Footnote 7 If so, the role of introspection in our own scientific understanding of the mind is greatly underrated in the ‘official narrative’ about cognitive science. For the gap between the state of our knowledge and understanding and the state of knowledge and understanding in that counterfactual world is owed entirely to our implicit (‘unofficial’) use of introspection.

To conclude, the claim that cognitive science is doing fine without introspection is doubly problematic: not only for its insistence on the success of cognitive science when it comes to phenomenal consciousness, but also for its suppressed assumption that cognitive science conducts its business without appeal to introspection. I have presented a (somewhat subversive!) picture of cognitive science according to which appeal to introspection is rife in cognitive-scientific research, though at the level of discovery rather than justification. My contention is that introspection is not only epistemically indispensible, but also not really dispensed with in cognitive-scientific practice.

3 The reliability of introspection: schwitzgebel’s challenge

Whatever the case for the operative role of introspection in cognitive science, the greatest challenge to the legitimacy of that role is surely that introspection is simply feeble and unreliable. For if it is, it would be irrelevant whether or not introspection is dispensable or dispensed with. An understanding of the mind founded upon it would be thoroughly discredited, regardless of whether a better alternative is available. Thus Schwitzgebel (2011, p. 118) writes: “[I hold] that the introspection of current conscious experience is both (i) possible, important, and central to the development of a full scientific understanding of the mind and (ii) highly untrustworthy, at least as commonly practiced.” Here Schwitzgebel does not deny that appeal to introspection would be indispensable for anyone who wishes to produce a scientifically respectable understanding of the mind. Rather, he puts in question the attainability such a respectable understanding, in light of the ‘untrustworthiness’ of introspection. Call this sort of challenge to the use of introspection in understanding the mind Schwitzgebel’s Challenge.Footnote 8

To be sure, there is a long philosophical tradition of over-trusting introspection. In its strongest form, this tendency can be articulated as a conjunction of two converse theses, one asserting the perfect reliability of introspection and one the omnipotence of introspection. According to the first, introspection is infallible:

(II) If subject S introspects having phenomenology P, then S has P.Footnote 9

According to the second, introspection never misses anything that passes within its purview, rendering phenomenology self-intimating:

(SI) If subject S has phenomenology P, then S introspects having P.

The conjunction of II and SI casts introspection as perfectly trustworthy. We may call the conjunction introspective dogmatism (or perhaps introspective maximalism, since it portrays introspection as maximally powerful).

Unfortunately for all involved, introspective dogmatism (or maximalism) is highly implausible. Introspection is far from perfectly reliable and far from omnipotent. However, the fact that introspection is not maximally trustworthy does not show that it is thoroughly unreliable and/or entirely impotent.Footnote 10 For our present purposes, what matters is whether introspection could be shown to be minimally justificatory, that is, have the least demanding epistemic properties that would be needed for it to play a legitimate role in the context of justification (and not just discovery). This requires that we identify these minimal epistemic properties, concerning both reliability and potency.

On the side of reliability, plausibly what is required for introspection to be minimally justificatory is that it enjoy above-chance reliability. To a first approximation, we may formulate the claim as follows:

(ACR) If subject S introspects having phenomenology P, then S is more likely to have P than if S does not so introspect.

On the side of potency, meanwhile, being minimally justificatory would plausibly require that introspection enjoy non-negligible potency:

(NNP) If subject S has phenomenology P, then S is more likely to introspect having P than if S does not have P.

Let us call the conjunction of ACR and NNP introspective minimalism. I contend that introspective minimalism, or something very much like it, would resist extant criticisms of introspection, while also undergirding the legitimacy of appeal to introspection in justifying hypotheses about consciousness. Something stronger than minimalism may yet be true, but the truth of minimalism would suffice to legitimize introspective appeal. Importantly, minimalism is so weak that it is very likely to be indeed true. Thus to make the case for introspective minimalism is effectively to meet Schwitzgebel’s Challenge.

Suppose, for instance, that introspection turns out to be as trustworthy as our sense of smell, that is, as reliable and as potent as a normal adult human’s olfactory system.Footnote 11 Then introspective minimalism would be vindicated. Normally, when we have an olfactory experience as of raspberries, it is more likely that there are raspberries in our immediate environment (than if we do not have such an experience). Conversely, when there are raspberries in our immediate environment, it is more likely that we would have an olfactory experience as of raspberries (than if there are none). So the ‘equireliability’ of olfaction and introspection would support introspective minimalism. Such equireliability is highly plausible.

It is worth noting that introspective minimalism can be refined in various ways. Thus, ACR and NNP do not explicitly contain any quantifiers, suggesting that they are intended as doubly universal, applying to all subjects and all phenomenologies (all values of S and P). This may turn out to be too strong. Perhaps it would be wiser to restrict these claims to normal subjects normally circumstanced. For it may be that under conditions of cognitive overload, or in psychologically malformed subjects, introspecting a phenomenology does not increase the probability that the phenomenology is in fact present (and/or the presence of a phenomenology does not increase the probability that it be introspected).Footnote 12 Likewise, there may be reasons to exclude certain special types of phenomenology from ACR and/or NNP. That is, various ‘exemptions’ may need to be carved. For example, according to many phenomenologists, a person’s field of consciousness typically involves a ‘fringe’ or ‘margin’ that contributes to one’s overall experience very lightly and unimposingly.Footnote 13 Arguably, however, fringe phenomenology cannot in principle be introspected, since introspecting it would render it focal rather than fringe.Footnote 14 Likewise, consider the phenomenology of experiential immersion or engrossment, such as a basketball player experiences when ‘in the zone.’ This phenomenology of engrossment may also be non-introspectible, insofar as turning one’s introspective attention onto it would require taking a step back from it, so to speak, disrupting its characteristic feelings of rightness and flow.Footnote 15 More generally, there may be a class of phenomenologies whose very essence requires the absence of introspective attention; we may call these (doubtless sub-optimally) ‘elusive phenomenologies.’ If so, ACR and NNP would probably need to be restricted to the complement class of non-elusive phenomenologies.Footnote 16

Taking into account the just-discussed restrictions, we would obtain the following doubly refined thesis of introspective minimalism:

(RIM) For any (normally circumstanced) normal subject S and any non-elusive phenomenology P: If S introspects having P, then S is more likely to have P (than if S does not so introspect) & If S has P, then S is more likely to introspect having P (than if S does not have P).

Other restrictions may be called for upon closer examination. Still, the fully refined minimalist thesis would very likely be non-trivial yet in a position to undergird the legitimate scientific use of introspection not only in the context of discovery but also in the context of justification.

4 Objections and replies

One form of objection to introspective minimalism, refined or not, is to produce a counterexample. This would require identifying a circumstance involving a (normal) subject and a (non-elusive) phenomenology, such that the subject’s having that phenomenology does not probabilify her introspecting it and her introspecting it does not probabilify her having it. Certain intense emotions might serve here. When one is infuriated with X, one tends to be consumed by one’s fury. One’s rage thus does not dispose one to introspect one’s phenomenology of fury, which introspecting would require a certain degree of detachment. This puts in question any kind of self-intimation in the phenomenology of fury or rage.Footnote 17

There are several possible responses to this sort of objection. First, one may deem that in virtue of its consuming quality, the phenomenology of fury is precisely an elusive phenomenology in the above sense, and is therefore outside the scope of RIM. Secondly, while the phenomenology of fury may not dispose one to introspect it, it may still make it more probable that one introspect having it than if one did not have it at all.Footnote 18 Finally (And more controversially), it is worth remembering the distinction some introspectionists have drawn between direct and indirect introspection (Titchener 1912): the former occurs simultaneously with the introspected; the latter occurs later and involves an element of memory. Even if the phenomenology of fury does not dispose one to enjoy direct introspection of it, arguably it does dispose one to enjoy indirect introspection of it.Footnote 19

Another traditional objection targets the replication of introspective ‘findings.’ Introspective disagreements about standard phenomenal experience are rife, and there appears to be no procedure for moving beyond any initial disagreement. More than anything, this was the behaviorists’ most fundamental and most effective criticism of introspectionist psychology.Footnote 20

In addressing this objection, it is important to distinguish two claims: that introspective judgments are not replicable in principle, and that they are not replicable in practice. The in-principle claim is surely too strong to be plausible. For one thing, introspective agreements greatly outnumber the disagreements, we just take them for granted.Footnote 21 More speculatively, it is hard to imagine that lifelong phenomenal duplicates would encounter many introspective disagreements. If so, it may be the great variation in, and complexity of, individuals’ conscious life (hence of the phenomenological facts about them) that leads to phenomenological disagreements.

This complexity and variety does raise the specter of in-practice irreplicability, which poses a genuine challenge to introspective appeal in cognitive-scientific practice. There are two potential approaches to it that ought to be pursued, however. One, pursued vigorously by fin-de-siècle introspectionists, involves systematically honing subjects’ introspective abilities, in the hope that greater introspective competence would lead to eventual convergence of reports. The thought is that accurate and attentive introspection is actually fairly difficult, especially when subtle phenomenology is concerned, and introspective disagreements are often due to performance failure. Once we apply the performance/competence distinction to introspectors, and make sure to cull data from only the most competent introspectors,Footnote 22 introspective disagreements will be limited; or so it is predicted. Notoriously, however, in practice even highly demanding introspective training did not eradicate disagreements among introspectionist psychologists.Footnote 23 The other potential approach to the problem, then, is to focus on reports not by subjects with introspective expertise, but by subjects with particularly vivid phenomenology. The thought here is that the vivacity of experience would render it more readily and accurately introspectible, making otherwise subtle phenomenal nuances more manifest to subjects, who in consequence will encounter fewer introspective disagreements; or so, again, it is predicted. This approach was not pursued by introspectionists, and requires much preliminary work on criteria of phenomenal vivacity and other issues.Footnote 24 But it is worth trying all the same.Footnote 25

A very different objection is that introspective minimalism already represents a major departure from pre-theoretic ‘naïve’ optimism about introspection. For example, the folk view considers introspection easy: if one experiences a certain phenomenology, one is expected to effortlessly appreciate that this is so.Footnote 26 Likewise, the folk tend to be highly confident in their introspective impressions, apparently regarding introspection as relatively close to infallible, at least in ordinary circumstances. The folk view is thus much closer to introspective maximalism than minimalism. To that extent, introspective minimalism represents a revisionary approach to folk psychology here.Footnote 27

My response is twofold. First, I am sometimes tempted to question the account of folk psychology just sketched. The account casts introspective confidence statements as truth-apt descriptive ones: ‘I am highly confident that what I am feeling right now is frustration’ is taken to attempt to represent an observer-independent fact of the matter. But for many occurrences of such statements expressivism seems more accurate than descriptivism. Charles Siewert once pointed out to me that our privileged access to our phenomenology is a (‘the’?) central source of our sense of dignity as separate, inviolable, self-possessing individuals.Footnote 28 This is why telling someone what they really feel, overriding their own claims about what they feel, often seems morally and not just epistemically wrong. Conversely, being told what one really feels over one’s protestations tends to elicit moral rather than epistemological indignation. The full case for this would have to be prosecuted elsewhere, but my suspicion is that the ethos of first-person authority is rooted in this sort of respect for the inviolable dignity of the other. If so, the primary function of ‘I am highly confident that what I am feeling right now is frustration’ may be to assert one’s authority over one’s own internal life, that is, to demand dignity and respect. This suggests a picture of introspective confidence judgments as primarily in the business of demanding, not describing. That is, it suggests that despite their surface grammar, they should not be treated as descriptive truth-apt statements. It is wrong, then, to suppose that folk psychology assumes that introspection is easy and confidence-imbuing—though the surface grammar of the relevant segment of the folk-psychological discourse does encourage that interpretation.

However, suppose folk psychology is committed to a view of introspection that is, upon reflection, unreasonably sanguine. Suppose, that is, that it is in need to revision when it comes to its take on the trustworthiness of introspection. This by itself does not undermine the legitimacy of appeal to introspection in theorizing about the mind. As long as the revision is not so drastic as to put in question something like the refined thesis of introspective minimalism, introspection enjoys the minimal reliability and potency required for its legitimate use.

In fact, cognitive science can partly illuminate why our introspective grasp of our inner world can be expected to be considerably weaker than our perceptual grasp of the external world. It is well-established that much of our perceptual grasp of the external world relies on calibration of information from different perceptual modalities.Footnote 29 Our observation of our internal world, however, is restricted to a single source of information, and not the most powerful to begin with. I have speculated above on the equireliability of introspection and olfaction. Imagine a race of creatures whose perceptual contact with the external world is limited to a sense of smell (as powerful as ours). Its perceptual grasp on the external world would be doubly weaker than ours: it would lack information from more powerful senses (such as vision), and it would lack the ability to calibrate information from multiple sources. Its perceptual grasp on its environment would consequently be much shakier than ours. It would parallel, I contend, our introspective grasp on our own internal world.

5 Conclusion

To conclude, much of our disappointment with introspection may be owed to undue expectations. In fact, all that is needed for introspection to be cognitive-scientifically valuable is (a) that introspecting a phenomenology would make it more likely that one is having the phenomenology than not introspecting it, and (b) that having a phenomenology would make it more likely that one introspect the phenomenology than not having it. This minimal requirement for usefulness is very plausibly met by our introspective faculty. Given that introspection affords us observational contact with our mental life, this means that introspection is in fact epistemically indispensable for any mature understanding of the mind. For it would be perverse to attempt to understand a phenomenon we can observe with minimal reliability in disregard of our observation thereof.