Tuesday, March 29, 2016

Introspective Attention as Perception with a Twist

Tomorrow I'm off to the Pacific APA in San Francisco. Thursday 4-6 I'm commenting on Wayne Wu's "Introspection as Attention and Action". This post is adapted from those comments.

(Also Wed 6 pm I'm presenting my paper "A Pragmatic Approach to the Metaphysics of Belief" and Sat 6-9 I'm critic in an author-meets-critics on Jay Garfield's Engaging Buddhism. Feel free to stop by!)


What does introspective attention to one's perceptual experiences amount to? As I look at my desk, I can attend to or think about my ongoing sensory experiences, reaching judgments about their quality or character. For example, I'm visually experiencing a brownish, slightly complicated oblong shape in my visual periphery (which I know to be my hat). I'm having auditory experience of my coffee-maker spitting and bubbling as it brews the pot. What exactly am I doing, in the process of reaching these judgments?

One thing that I'm not doing, according to Wayne Wu, is launching a new process, distinct from the perceptual processes of seeing the hat and hearing the coffee-maker, which turns an "attentional spotlight" upon those perceptual processes. Introspective attention, Wu argues -- and I agree -- is a matter of applying phenomenal concepts in the course of ordinary perceiving, with the goal of arriving at a judgment about your perceptual experience -- doing so in a way that isn't radically different from the way in which, according to your goals, you can aim to categorize the things you perceive along any of several different dimensions.

I hope the following is a Wu-friendly way of developing the idea. Suppose you're looking at a coffee mug. You can do so with any of a variety of perceptual goals in mind. You can look at it with the aim of noticing details about its color -- its precise shade and how consistent or variable its color is across its face. You can look at it with the aim of noticing details of its shape. You can look at it with the aim of thinking about how it could effectively be used as a weapon. You can look at it with a critical aesthetic eye. Each of these aims is a way of attending differently to the mug, resulting in judgments that employ different conceptual categories.

You can also look at the mug with an introspective aim, or rather with one of several introspective aims. You can look at the mug with the aim of reaching conclusions about what your visual experience is as you look at the mug rather than with the aim of reaching conclusions about the physical mug itself. You might be interested in noticing your experience of the mug's color, possibly distinct from mug's real color. According to Wu, this is not a process radically different from ordinary acts of perception. Introspection your visual experience of the color or shape of the mug is not a two-stage process that consists of first perceiving the mug and then second of introspecting the experiences that arise in the course of that perceptual act.

The approach Wu and I favor is especially attractive in thinking about what the early introspective psychologists E.B. Titchener and E.G. Boring called "R-error" or "stimulus error". Imagine that you're lying on your stomach and an experimenter is gently poking the bare skin of your back with either one or two toothpicks. You might have one of two different tasks. An objective task would be to guess whether you are being poked by only one toothpick or instead by both toothpicks at once. You answer with either "one" or "two". It's easy to tell that you're being poked by two if the toothpicks are several centimeters apart, but once they are brought close enough together, the task becomes difficult. Two toothpicks placed within a centimeter of each other on your back are likely to feel like only a single toothpick. The objective task would be to guess how many toothpicks you are being poked with in reality.

An introspective task might be very similar but with a crucial difference: You are asked to report whether you have tactile experience of two separate regions of pressure or only one region. Again you might answer "one" or "two". This is of course not the same as the objective task. You're reporting not on facts about the toothpicks but rather on facts about your experience of the toothpicks. The objective task and the introspective task have different truth conditions. For example if two toothpicks are pressed to your back only 6 millimeters apart and you say "one" you've given the wrong answer if your task is objective but quite possibly the right answer if your task is introspective.

[Edwin G. Boring in 1961]

Here's the thing that Titchener and Boring noticed, which they repeatedly warn against in their discussions of introspective methodology: People very easily slide back and forth between the introspective task and the objective task. It's not entirely natural to keep them distinct in your mind over the course of a long series of stimuli. You might be assigned the introspective task, for example, and start saying "one", "one", "two", "one", "two", "two", "two" -- at first your intentions are clearly introspective, but then by the thirtieth trial you have slipped into the more familiar objective way of responding and you're just guessing how many toothpicks there actually are, rather than reporting introspectively. If you've slipped from the introspective to the objective mode of reporting, you've committed what Titchener and Boring call stimulus error.

For Wu's and my view, the crucial point is this: It's very easy to shift unintentionally between the two ways of deploying your perceptual faculties. In fact I'm inclined to think -- I don't know if Wu would agree with me about this -- that for substantial stretches of the experiment your intentions might be vague enough that there's no determinate fact about the content of your utterances. Is your "one" really a report about your experience or a report about the world outside? It might be kind of muddy, kind of in-between. You're just rolling along not very thoughtful of the distinction. What distinguishes the introspective judgment from the perceptual judgment in this case is a kind of minor thing about your background intentions in making your report.

Introspection of perceptual experience is perception with a twist, with an aim slightly different from the usual aim of reporting on the outside world. That's the idea. It's not a distinct cognitive process that begins after the perceptual act ends, ontologically distinct from the perceptual act and taking the perceptual act as its target.

When you know that your experience might be misleading, the difference can matter to your reporting. For example, if you know that you're pretty bad at detecting two toothpicks when they're close together and you have reason to think that lots of the trials will have toothpicks close together, and if your focus is on objective reporting, you might say: "Well, 50-50% -- might be one, might be two for all I know". For introspective reporting, in contrast, you might say something like "Sure feels like one, though I know you might well actually be touching me with two".

In visual experience, noticing blurriness is similar. Take off your glasses or cross your eyes. You know enough about the world to know that your coffee mug hasn't become objectively vague-edged and blurry. So you attribute the blurriness to your experience. This is a matter of seeing the world and reaching judgments about your experience in the process. You reach an experiential judgment rather than or in addition to an objective judgment just because of certain background facts about your cognition. Imagine someone so naive and benighted as to think that maybe actual real-world coffee mugs do in fact become vague bordered and blurry edged when she takes off her glasses. It seems conceivable that we could so bizarrely structure someone's environment that she actually came to adulthood thinking this. That person might then not know whether to apply blurriness to the outward object or to her experience of the object. It's a similar perceptual act of looking at the mug, but given different background concepts and assumptions in one case she reaches an introspective attribution while in the other case she reaches an objective attribution.

[image source]



  • "Introspection, What?" (2012), in D. Smithies and D. Stoljar, eds., Introspection and Consciousness (Oxford). My own positive account of the nature of introspection.
  • "Introspection". My Stanford Encyclopedia review article on theories of introspection.
  • "The Problem of Known Illusion and the Resemblance of Experience to Reality" (2014), Philosophy of Science 81, 954-960. Puzzlement and possible metaphysical juice, arising from reflections on the weird relation between objective and introspective reporting.

    (k)nowable said...

    Love this!

    Callan S. said...

    Well if you were cross eyed your whole life you might very well attribute everything as being blurry.

    With seeing, you have a straight eyed contrast to work with. Where's the contrast with the toothpicks on the back? Without a contrast I can't really see any difference between an introspection test and an objective test? Never mind the subject, what is the tester asking for? When the subject can only treat their introspection as their best objective evaluation, then what is the tester asking for? It seems a perception fault of the tester, really. Like expecting a small child to see something, when the relative perspective of a child has things in the way of seeing it that the adult does not. The 'Ah ha, you are having a stimulus error!' response seems more like an error itself, treating the testers own capacity to see whether it is one or two toothpicks and projecting that onto the subject. It seems a similar theory of mind problem as when a child thinks an adult who just came into the room knows how many jelly beans are in a bag the child just looked in. The adult isn't having a stimulus error when they don't know how many jelly beans there are. Why a stimulus error when you don't know how many toothpicks there are?

    Eric Schwitzgebel said...

    Thanks for the comment, Callan! Here are two possible stimulus errors in the toothpick case: 1. Saying "maybe two, not sure" when it feels like one spot of pressure but you're think the experimenter might really be putting two close together. Or saying "two" when you feel a single oblong patch of pressure that you know could only be caused by two toothpicks, if your task was to count subjectively felt regions of pressure. (Boring emphasizes the latter case.)

    Callan S. said...

    I'm not sure what's wrong with #1, Eric? Baring the use of any extra info (like seeing the scientist in a reflective surface), the subject is reporting as best they can. If they report "Maybe two, not sure" then that's their report. There's no error there as much as there's no error with the adult who just came into the room not stating the correct number of beans in the bag?

    If the subject is trying to guess when the scientist is doing 'Well, he's done 'one' a few times now, I'm guessing he's done two and I don't want to look too stupid, so I'll say two', I think that's another matter. More an ego error!

    Not sure the other part is fair - where they asked to count subjectively felt, separate regions of pressure? Then it's a matter of reporting separate regions. I could get two clusters of a dozen toothpicks each and press them inches apart and they would report two if I was asking for a report of regions. But a count of toothpicks is a different question.

    It might be worth considering a factor is the scientists incredulity that the person could be this blind to what is obvious to the scientist.

    Something like running one of those experiments where the scientist thinks he is conducting the experiment but is actually the subject. Perhaps some of the subjects who's back is being toothpicked is fed the answers as to how many toothpicks it is. Others are not (the control). Possibly also run a test with the toothpicked subject being fed the answer randomly, 50% chance each time (this subject knowing when they are fed and when they are left to their own devices). If the fed subjects seem to come out with no stimulus errors reported by the subject scientist and the control subjects are reported as having stimulus errors, then there's some evidence there for observational error in the subject scientist. The subjects being fed the answers are completely unnatural responses. They would, however, line up perfectly with the subject scientists clearer perception of what is occuring.

    chinaphil said...

    Very interesting. I note the close relationship to your paper on whether we really see circular things viewed from an angle as ellipses.

    I have a couple of comments on the language used. I think I agree with the position you describe you and Wu as sharing, but I'd prefer to say that I cannot introspect on perception, because I don't think we have any ability to access the bit that happens between light hitting our eyes and our brains registering it (linguistically or non-linguistically) as "red". As you put it, we can "look...with the aim of noticing details"; or we can attempt to use memory to reconstruct perceived details to which we did not pay attention at the time. But I'd always emphasize the impossibility of accessing the "black box" parts of many perceptual processes. I'd also emphasize the subjectivity of the process of looking for details: we can look at a mug's "colour" because colour is a salient concept for us. In part that is a biological phenomenon, but in part it's cultural - as I understand it, what we call "colour" is actually an amalgam of a number of different physical features - wavelength, saturation, brightness - that at the very least *could* be separately considered. The point here is that what you're calling introspection on perceptual experiences is not bringing us any closer to any unmediated qualia (if such a thing exists).

    The second language point is that I wouldn't call the mix up of introspective judgments and perceptual judgments "muddy". I think I agree with you about what's going on, but I see it as inevitable, and built into the language. The implication of the word "muddy" is surely that there is, or could be, a "clear" state in which we don't conflate our judgments about the outside world with our judgments about our own perceptions, but I just don't think there is - or if there is, it's the kind of thing that people train for years to achieve. It seems like the kind of thing fire-walkers do; or what Buddhist adepts sometimes claim to achieve. The lesson to draw from the toothpick experiment would be that it is extremely hard for people to apply a new concept consistently just minutes after they've been instructed in it by a scientist. Particularly when they're being poked with sticks at the same time - something to remember in your lecturing, I suppose! Poke students with sticks less. (I've been doing some training in my day job, and the urge to poke can sometimes be strong...)

    David Duffy said...

    Where does working memory fit in with this? In that attention to the task means one will also review immediate memories of the sensation before making the report. In which case both modes might be possible off the same event.