Tye’s new take on pain

By Manolo Martinez

[UPDATE] I have posted a quick summary of the imperative content strategy in dealing with affective phenomenology here.

Hello everyone.

This is my first post for Brains! So, yep: great to be here, etc. A quick caveat: I’m afraid this post makes a tl;dr response perfectly appropriate. I’ll try to be much more bloggy posty the next time around. Thanks a lot to those who bear with me in the meantime.

Tye and Cutter [T&C]’s Tracking Representationalism and the Painfulness of Pain, forthcoming in Philosophical Issues is a discussion of several new (and not so new) objections to representationalism based on its purported inability to accommodate the characteristically unpleasant phenomenology of pain. I think the paper does a great job of clarifying several assumptions of the theory, particularly about the underlying theory of mental content. I still believe, though, that so-called tracking representationalism -the combination, defended in this paper, of representationalism and what T&C call the tracking theory of intentionality [TTI]- is ill-suited for explaining the painfulness of pain -I have defended elsewhere that imperative contents are a much better choice for representationalism about pain. In this post I will only discuss one reason for thinking that TTI won’t do. Maybe in a subsequent post (if you guys are interested) I will defend the imperative-content strategy from T&C’s complaints.

According to TTI,

Tokens of a state S in an individual x represent that p in virtue of the fact that: under optimal conditions, x tokens S iff p, and because p. (p.2 [all page numbers from T&C forthcoming])

This is a version of indicator semantics along the lines made familiar by Stampe or Dretske. For TTI to count as wholly naturalistic the clearly normative notion of “optimal conditions” must be discharged -or, as Belnap puts it in an analogous context, a naturalist has to feel miserable under it has been discharged. T&C propose a broadly teleological understanding of optimal conditions:

 In the case of visual experience, we are in optimal conditions just in case we are in conditions of the sort that our visual system was designed to operate in by natural selection or by analogous processes in the course of ontogenic development. The same holds, mutatis mutandis, for other sorts of perceptual experience. (p.2)

The resulting theory, then, is a teleosemantics of sorts -as I say, very much in the spirit of other proposals on offer. Now,according to T&C

(Pain Content) The content of a pain experience of an individual A is something like: there is a bodily disturbance of (physiological) type d in location l, and d is bad for A to degree x. (p. 9)

The first conjunct in Pain Content is supposed to be identical with (or be the supervenience base for, depending on the account) the sensory phenomenology of pains -its perception-like component, that provides information about its location, volume, etc. The second conjunct stands in the same relation (identity/supervenience) to the characteristic unpleasantness of pain -its painfulness. We are now in a position to see the problem with the theory that is the target of this post:

I think it’s fair to say that there is a consensus in pain studies according to which the sensory and painfulness dimension of pain phenomenology can be modulated independently. For example, in pain asymbolia people feel adequately the sensory component of pain, but do not feel its painfulness, or only in a greatly diminished manner. Take the case of Fred, a subject with morphine-induced asymbolia. If T&C are right, Fred’s pain state represents correctly that [there is a bodily disturbance of (physiological) type d in location l] but incorrectly that [ d is bad for Fred to degree x] -in fact it’s bad for him with degree x’ >> x.To accommodate this case, we need that Fred’s pain state is tokened under optimal conditions for the sensory component but not under optimal conditions for the painfulness component. A fortiori, for this we need these two optimal conditions to be different. Now, can they be different?

Recall what T&C have to say about optimal conditions: they are the conditions under which our [pain] system was designed to operate. Now, what follows is a piece of empirical speculation but, I think, an extremely plausible one: the conditions under which the detection of bodily disturbances evolved were conditions under which these disturbances were apt to harm (i.e., bad for their bearers). Consider cuts. There are, to be sure, instances of these bodily disturbances that are not bad for their bearers: tumour-removal cuts, wound-stitching cuts and the like. But it’s overwhelmingly plausible that the detection of cuts did not evolve under pressure to signal these latter beneficial instances, but the much more numerous and important deleterious cuts. That is: the optimal conditions that fix the content of the sensory dimension of pain are conditions under which bodily disturbances are bad.

If this is so (as I say, an empirical speculation, but a seemingly safe one) T&C, after all, cannot explain asymbolia, because they cannot account for the possibility of a representation being veridical about the presence of a bodily disturbance but not veridical about the presence of badness. Under TTI, and their proposed way to unpack optimal conditions, the sensory dimension of pain cannot be separated from the painfulness dimension. Although, as they say, their proposal about optimal conditions is vague and schematic, the above seems to suggest that no account in the ballpark -that is, one that unpacks teleology in terms of selection history- will do. Unless concepts and a principle of compositionality are used to fix pain contents, of course, but this is  an option dismissed by T&C (rightly, I think.)

In conclusion, at the very least, the naturalistic credentials of T&C’s account have to be put on hold until a notion of optimal conditions is produced that is both naturalistic and able to deliver the required independence between sensory and painful components of pains.

10 Comments

  1. Colin Klein

    If there’s anything that’s get me beyond my long-time lurking on Brains, it’s this!

    So with the caveat that I’ve only read the part of the T&C paper that criticizes Manolo and I, a few thoughts. First, I should note in passing that T&C actually misrepresent my position—I actually hold the stronger position that there is *no* representational content to pain, and so that it is entirely exhausted by its imperative content. I give some reasons against representational content in my original paper on the topic. For this reason, I also disagree with Grahek’s characterization of pain asymbolia, but that’s a separate debate.

    The most interesting objection in T&C, IMO, is one that Adam Pautz has also pressed me on—to wit, that the imperative account can’t deal with varying intensities of pain. I don’t think this is either obvious or true. Manolo has given the same kind of answer that I think I want to give here:
    https://consciousnessonline.wordpress.com/2011/02/16/the-real-trouble-for-phenomenal-externalists-the-science-of-taste-smell-and-pain/

    But to re-argue for the same point: Ordinary language already makes a distinction between more and less urgent imperatives (“Close the door (if you could)!”, “Really, close the door!”, “Damnit, close the door now!”). Some languages actually build these distinctions into their grammatical structure. Those are of course crude distinctions, but there’s no reason not to think that phenomenological imperatives might not have more fine-grained content: “Don’t move your ankle [with priority P]”, say, where P specifies some function that assigns desirability to possible worlds. (I’ve got a paper that I’m R&R-ing right now that tries to spell this out in more detail using Hamblin’s semantics for imperatives. Putnam also sketches a framework like this in “The Mental Life of Some Machines”, though I don’t have it handy to get the reference.

    Why accept this sort of move? Well, as far as T&C go, it seems like they should, because the representationalist has to make this move too. Ordinary language only makes crude distinctions between (say) levels of color saturation. But no representationalist should accept—for this reason alone—that representationalism is inadequate to our experience. They should be happy to say that the relevant representations of intensity are more fine-grained. In some cases, we can and have introduce more technical vocabulary that captures more distinctions. Similarly, in the case of pain, many people have introduced various scales that capture the distinctions more precisely (the dolorimeter, the MPQ, the Wong-Baker faces, etc.)

    One interesting question, on which I don’t have a settled opinion, is whether pain intensity follows a ratio/interval scale, or something more like a merely ordinal scale. I suppose I’m actually inclined towards the latter (and so I’m against, say, Pautz’s claim that it makes sense to talk about one pain being twice as strong as another), but I do

  2. Interesting discussion. Just a few quick points.

    (i) Colin mentions in his comment an objection to the imperative view about degree that I’ve made and that is also made by my fellow UTexans Cutter and Tye. I make it in a note – note 36 for those interested – of my recent ‘Do Theories of Consciousness Rest on a Mistake?’ paper (on my website), along with a few other objections to Colin’s version of the imperative theory.) I agree with Colin that the objection is not obviously decisive – the imperativist can say something about degree. But one worry I still have is about ratio scaling. Donald Price, one of the leading pain researchers (appeal to authority!!), has a series of experiments on this. He concludes VAS (visual analogue scale) is a ratio scale. It’s especially hard to see how the imperativist might accommodate ratio judgments. In reply, at Consciousness Online, Manolo said his view can accommodate ratio judgments. But I’m not sure I quite understand the things he says there, like “Disturbance, stop with urgency U” and “Disturbance, stop with urgency Udouble” – at least I think more needs to be said about what this means. By contrast, in his comment Colin says he’s inclined to reject ratio judgments, so the imperativist needn’t account for them. But, again, there’s evidence from Price and others that such judgments make sense. (It’s generally – although not universally – believed that ratio scaling make sense in other modalities: for instance, when it comes to perceived loudness. Why couldn’t it make sense as regards pain intensity?)

    (ii) In the same place I raise the prima facie problem about degree, I raise some other objections to Colin’s version of the imperative theory (some but not all of which I think carry over to Manolo’s version, as I discussed at Consciousness Online) – I think the objection about degree is not the only one. But I won’t go over them here because they’re a bit far afield from Manolo’s post.

    (iii) More related to Manolo’s post: I think Manolo’s psychosemantic objection to Tye and Cutter’s (non-imperativist) theory of pain is interesting. Let me try out another potential psychosemantic objection and see what folks think (I brought it up with Cutter and he said he has to think more about it). As Tye and Cutter say, their account requires that the following be true of Mild in my Mild-Severe case: “the explanation for why Mild tokens F — a functional state defined by its role in producing mild avoidance behavior– when he undergoes [lesion] d is that d is, to a small degree, apt to harm him.” (They say to “harm” is to hinder the system (or one of its subsystems) from performing its function(s).) On Tye’s psychosemantics, this explanatory claim needs to be true, if F is to *represent* that d is apt to a small degree to harm him. But, to me at least, this explanatory claim is intuitively false. I’ll back that up in a later comment.

  3. To see why I think C&T’s explanatory claim (mentioned in my last comment) is false, suppose Mild undergoes d, but, while d is usually apt to harm him to small degree, it is not to a small degree apt to harm him in this case. Maybe in this case in reality it is not apt to harm him at all, or maybe it is to a *high* degree apt to harm him, b/c of environmental/context differences. E. g. he happens to come across some nasty bacteria which are likely to infect d and cause serious problems so that even though it is a mild lesion it is in this case to a *high* (not small) degree apt to harm him. There are many possibilities. But suppose, again, that d is still there and his pain system is working in accordance with design. Then F will be tokened in Mild all the same – even though in this case d is *not* to a *small* degree apt to harm him. So it seems to me that it is the mere presence of the bodily state d that explains particular tokenings of F and that the small-apt-to-harm level (even when present) is explanatory irrelevant to particular tokenings of F: for were d present and the small-apt-to-harm level absent, F would still result.

    I think we have to distinguish two explanatory claims, one correct and the other incorrect. The correct explanatory claim is this: the fact that d is generally apt to harm to a small degree members of Mild’s species explains the *general fact* that members of his species have a pain system that responds to d with functional state F involving mild avoidance behavior. But what Tye’s psychosemantics requires is that under optimal conditions *particular tokenings* of F are explained by the fact that d is apt to harm to a small degree. It is this explanatory claim I object to – for again, in cases where we keep d there but alter the harm level, F still results.

    Consider an analogy. Suppose all cases of some fruit have chemical property C. The majority of cases also have chemical property P, which is poisonous to a certain species. But some cases of the fruit lack P and so are not at all poisonous. The fact that the fruit is generally apt to harm the members of the species explains the general fact that they evolved taste systems that causally respond to all cases of C with a functional state F involving withdrawal. But now consider a particular tokening of F under optimal conditions in some individual member of the species. What explains this particular tokening? I think the answer is: just the presence in the mouth of the individual of a foodstuff with chemical property C. The fact that it is apt to harm is not explanatorily relevant to this *particular tokening*, even if it explains the general fact already noted. Indeed, had this been one of the lucky cases of the fruit that is *not* poisonous to the individual and so *not* apt to harm him, F (involving violent withdrawal) still would have resulted just the same.

    So I agree with Manolo that C&T’s intriguing proposal might face psychosemantic issues.

  4. Hi, Adam, just a quick doubt. Why can’t T&C unpack ‘optimal conditions’ as: those conditions under which the tokening of F helps explain that F was historically selected for.

    So, F is only tokened in optimal conditions if tokened in the conditions that explain its fitness-conduciveness: plausibly, only whenever d is apt to harm in a small degree.

    They would block your objection by pointing out that the particular tokening of F you focus on does not happen in optimal conditions. Am I missing something?

  5. Hi Manolo, thanks for the follow up. True, the tracking intentionalist only needs the ‘because’ claim to hold in optimal conditions. But the ‘because’ claim holds in an optimal case C only if certain counterfactuals are true about what happens in nearby cases C*, C**, etc – ‘because’ claims have cf consequences. I’m saying the relevant counterfactuals are not true. It does not matter whether C*, C**, etc. are deemed optimal or not.

    Analogy: suppose all or most red things just happen to be radioactive and so bad for us. So we make a simple light-detector (a spectrometer) that beeps when red things are present. So we might say *design conditions* require that red things be radioactive. Now suppose in a design condition C the machine beeps in the presence of a red and radioactive thing. In this case did the detector beep because the thing was radioactive/harmful or just because it was red? To answer this ‘because’ question, we need to consider a nearby case C* in which it is not radioactive/harmful (we take away the radioactive stuff in it) – yet it’d still be red. In C* the machine still beeps just the same. This is evidence that in design condition C the machine didn’t beep on this particular occasion *because* it was radioactive/harmful but just because it was red. And, contrary to what you suggest, it doesn’t matter that C* is not a design condition. So on a tracking theory beeping represents red, not bad.

    Two other points. (i) Explanation is an intentional notion, bringing in interests and background knowledge and so on, so relying heavily on that notion in the explication of (sensory) intentionality, as the tracking intentionalist does, might be circular.

    (ii) On tracking intentionalism, F represents badness iff under optimal conditions F occurs iff and because a bad thing is present. This means that in the case I described above, in which a creature eats a non-poisonous, non-bad case of the generally poisonous fruit but still gets functional state W involving withdrawal, conditions can’t be optimal – even though everything seems great. Otherwise W can’t represent badness. Likewise for my case in which F (involving mild withdrawal etc.) is tokened even though d happens to be harmful to a high (not small) degree b/c of bacteria. This adds to the difficulty of defining optimal conditions, for what’s the general non-circular, non-trivializing def. that gets you this result? I think yr friendly suggestion for a def. doesn’t work. (For one thing, in cases in which d is harmful to a high (not small) degree, F is still fitness conducive. Likewise even in the non-poisonous case the following condition still holds: most cases of the fruit are poisonous. And this condition generally explains why W is fitness conducive.)

    I present what I think are more serious problems in my objection-reply section here:

    https://consciousnessonline.files.wordpress.com/2011/02/the-real-trouble-adam-pautz.pdf

  6. Thanks for your reply, Adam. Yes, I see now your point. I think you are right.

    In earlier papers Tye has suggested that “Experiences represent various features by causally correlating with, or tracking, those features under optimal conditions.”

    I guess I wasn’t really reading the new version, and was more or less thinking about this other one. The “because” problem that you rightly point out does not appear with this other version, does it.

    Anyway, I agree with you that this kind of psychosemantics is too simplistic and, for example it’s unable to avoid widespread indeterminacy in content attributions, I myself prefer a teleosemantics in more or less Millikan’s spirit -although not quite: Millikan’s biosemantics is also subject to indeterminacy.

  7. Thanks very much, Manolo. Well, I actually agree with what Cutter and Tye say early on in their paper: tracking intentionalism or something like it might represent our best shot at actually coming up with a naturalistic theory of phenomenal consciousness. and i think many of the standard arguments against it fall short. (that’s why a lot of my work is devoted to it – i try to come up with new arguments against it!) i’ve argued (in my ‘simple view of consciousness’ paper that C&T briefly discuss) that the problems with a Millikan style theory of sensory representation might be even more serious than the problems with the straightforward tracking theory. but that’s a long story!

  8. Colin Klein

    A few more thoughts on pain intensity, in response to Adam’s comments.

    I’m not sure whether or not pain intensity is a continuous quantity. I’m actually inclined to think that it is, though that’s only a hunch. Either way, I think that the imperative theory can handle it by the strategy above. In my paper on phantom limb pains, I’ve argued that the content of an imperative includes a utility function defined over possible worlds (don’t take the pw talk too seriously; it’s a remnant of grad school). That function may be an ordinal or a cardinal function—if you want pain intensity to be a continuous quantity, you go for the latter.

    As for the question itself, I’ve been poking around the literature on the VAS scale, and I think it’s still ambiguous—this probably more b/c of issues in measurement theory than anything about pain itself. Most of the literature *assumes* that pain is a continuous quantity, and is concerned with showing that the VAS (or whatever) scale has ratio properties (and so is, given the assumption a good scale). The question is thus not really directed at pain itself, but at the properties of the instrument. Many people tend to additionally assume that if a scale has ratio properties then it must be measuring some underlying quantity. But I’m of the camp that thinks that this is a dubious move that rests more on Stevens’ instrumentalism than on any deep principles about psychology. There’s a nice short paper by Wayne Hall from 1981 where he brings up these issues in the context of pain scales.

    Adam does bring up a good point: other sensations seem to be continuous. But here bodily sensations might just be a distinct class, because they play (at least on my account) a distinct role: their purpose isn’t to inform, but to get an agent to act in appropriate ways. They could do this job simply by ranking possible outcomes in a commensurate way. If you’re both hungry and thirsty, it’s not biologically helpful to know that you’re twice as hungry as you are thirsty—you just need to know which problem to tackle first.

Comments are closed.

Back to Top