The Unreliability of Introspection

Traditionally, many philosophers like to attribute special status to at least some kinds of knowledge that we have of our minds.  The purported reliability of introspection is often invoked by those who propose to construct a first-person science–a science based on private evidence delivered through introspection.  Even Daniel Dennett, a naturalist and skeptic about self-knowledge par excellance, says that we have authority on what it is like to be us.

As someone who thinks there is no such thing as first-person science based on private evidence, I am glad to notice and recommend some recent, fascinating pieces on the unreliability of introspection about conscious states.

In an unpublished paper that he will present at the upcoming PSA 2006 meeting, Eric Schwitzgebel argues that introspection of occurrent conscious states is unreliable in a wide range of circumstances.

In a paper forthcoming in Nous, Day Haybron focuses more specifically on our ignorance of our affective states.


  1. Well, I wouldn’t say that what Dennett claims implies that our authority about our beliefs is the last word on what’s true about our beliefs. I mean you should try to see what he means that we should treat the verbal reports as expressing heterophenomenological worlds: these worlds can be theorist’s fiction, nothing else (they can even be false; heterophenomenology has no method to say whether these worlds are true or false in any sense). To conclude, I don’t think Dennett would deny we can make mistakes in ascribing beliefs to ourselves. He seems to make this point several times in ‘Sweet Dreams’ (see especially his point about Mr. Clapgras).

  2. Anna-Mari

    Il dottore Piccinini,

    In your fascinating “Data from Introspective reports” you write as follows:
    “… A further upgrade towards a more scientific use of introspective reports would be the identification of mechanisms and processes responsible for generating introspective reports and of ways in which these mechanisms can malfunction. Neuroimaging studies, as well as lesion studies and other traditional techniques, are likely to help in this respect.In principle, knowing the neural correlates of introspection can help determine whether a behavior is or isn’t a genuine introspective report. “

    Hmmm… Just a brief question: How can one know, whether a neural mechanism is a neural correlate of “genuine” introspective report in the first place?

    G, am I missing something – again?


  3. gualtiero piccinini

    Great question.  Strictly speaking, you are not looking for the neural correlate of the reports themselves (which are outside the head) but the neural processes that generate the report.  There are two main relevant components of these processes:  the states/processes that are the object of the reports and the process of generating the report.  To find the neural correlates, you need to do what you normally do:  roughly speaking, look for what areas of the brain light up while the subjects are performing the relevant tasks.  In this case, look for what lights up while the subjects are performing the task that gives rise to the states/processes that are the object of the reports and also for what lights up while the subjects are performing the task of producing the reports.

    This is just a first approximation.  You need to control for confounders.  For example, you need to investigate whether the “introspective” reports are the outcome of some kind of genuine introspection (i.e., access to one’s mental states) or some other process, such as the kind of a priori theory-driven reports that Nisbett and Wilson 1977 talked about.

    Does this seem unreasonable?

  4. Marcin Milkowski

    Well, Dennett’s wording is ambiguous, I admit. Nevertheless, I’d look at it from the point of view of psychological experimental practice, i.e., verbal reports are not corrigible by any evidence to which the subject has no access to. Take optical illusion: there must be a tension between what I report and what is true about what I see (this is what optical illusion is about, right?). You cannot say I’m wrong in saying that I see [insert any description of optical illusion here]. On the one hand, I’m exactly right because this is how I experience this kind of illusions, but on the other hand, I’m of course wrong (I cannot see or know anything that is false). The important thing is not to equivocate these two senses of being right or wrong about experience. The first one is based only on our subjective authority, and the other one is fallible.
    Maybe Dennett could use some more vegeterian speak about subjective authority (the speak of fiction is simply misleading, I grant it).

  5. Anna-Mari

    Ciao G,


    “There are two main relevant components of these processes: the states/processes that are the object of the reports and the process of generating the report. To find the neural correlates, you need to do what you normally do: roughly speaking, look for what areas of the brain light up while the subjects are performing the relevant tasks.”

    – Si, si, pero… (perhaps) no.

    Let`s imagine: I am a neuroscientits, and you are my poor test subject in fMRI.

    Now, I say to you: “Introspect.” And you do, whatever you do. How can I _publicly_ know, whether you are introspecting or not. I can guess or I can use folk psychology, as you suggested in your paper. (I have always found that idea in your paper really promising and good, but I still do not see the light.) Sooner or later I just would have to ask you: “Did you introspect”?

    And then… the publicity of reports comes into the picture. As Frege once pointed out; I cannot feel your pain, and you cannot have my symphaties. Does it help, if you tell me about your pains or behave “painy way”?

    The problem, of course, is that the method of gathering data about the psychological states is not _public_ in case of introspection or introspecitive data. As far as I know, the publicity requirement (of scientific method) is about publicity of _report_, not public method of seeing or hearing things… ( I just cannot find the right word now, not “observation”… aargh, hope you`ll grasp the idea anyway???)

  6. gualtiero piccinini

    Again, good question.  As you know, according to my way of framing things, the method of gathering data about psychological states is always public, even in the case of introspection.  For the method of gathering data from introspective reports does not include the production of the introspective reports.  It begins after that.

    But this doesn’t answer your question about how do we know whether a subject is really genuinely introspecting.  The answer to that is (to a first approximation), because we have checked that her reports are really providing accurate information about her mental states, or at least because wee have checked under similar enough conditions.  For instance, Ericsson and Simon 1993 showed that certain kinds of concurrent reports correlate well with what their analysis of the subjects’ task would predict their subjects to be thinking about, as well as with what their eye movements predicted they were thinking about.  Conclusion:  The subjects were really introspecting, because there is independent evidence that they were thinking what they reported they were thinking.  Contrast this with some of the studies reviewed by Nisbett and Wilson 1977.  They showed that under some conditions, subjects reported having motivational states that there was independent evidence subjects did not have.  Conclusion:  The subjects were not really introspecting.  In principle, you could add imaging data as a further method of checking whether the subjects are really reporting on their mental states or something else is going on. 

  7. gualtiero piccinini

    I’m sorry, but I don’t know what you mean by saying that “all things are from nature or by nature”, or why this has anything to do with a first-person science based on private evidence.

Comments are closed.

Back to Top