Wednesday 2 April 2014

Explaining Delusions (5)

This is a response to Max Coltheart's contribution to the blog, posted on behalf of Phil Corlett. 
Phil Corlett

Thanks Max.

I suppose if a two-factor explanation of polythematic delusions is an open question, then an aberrant prediction error explanation of monothematic delusions is likewise open.

I think our theories have more in common than perhaps we are allowing with this debate and our apparent disagreement. I think we are talking about different levels of explanation (Marr, 1977). My implementational and algorithmic explanations may be overlapping with your computational one, but it seems you disagree. This is curious and unfortunate, since prediction error is at the heart of both of our explanations.

We make our prediction error case using behavioral and functional imaging data – if people with monothematic delusions were to show aberrant prediction error signals that correlate with delusion severity (as we showed in endogenous and pharmacological settings as well as in healthy individuals with odd beliefs), I would be inclined to favor an aberrant prediction error explanation (Corlett et al., 2010). I suspect they might. Which is to say, I think aberrant prediction error can account for monothematic delusions.

I suppose we shan’t resolve this discussion unless we acquire these data in your head injured patients with monothematic delusions. You have described the phenomenology of these patients in detail, but as I suggested previously, there is a paucity of empirical study of these subjects – particularly with regards to the levels with which our prediction error model is concerned.

More troubling (to me at least) is the paucity of data from these monothematic delusional subjects with regards to the key predictions of two-factor theory. For example, there are some data on the skin conductance responses of Capgras sufferers to familiar faces (they support the notion that familiarity responses are reduced (Ellis et al., 1997)). To my knowledge though, no one has shown excessive familiarity in Fregoli delusion (the belief that strangers are your family members in disguise) nor have the predicted global deficits in skin conductance response in Cotard delusion (the belief that one is dead) been reported. In the absence of these data, are we to consider 2-factor accounts applicable only to Capgras?

Some further thoughts on polythematic delusions.

Thanks for introducing the distinction. As the apparent originator of the term, perhaps you can clarify some points for me?

It seems to me that your definition and examples of polythematic delusions conflates the number of beliefs (poly – many beliefs) with bizarreness (Cermolacce et al., 2010). Are all polythematic delusions bizarre by your definition? I ask because I am still not clear on what I would predict with regards a general belief evaluation deficit (Factor 2). I personally would predict a deficit in evaluating all beliefs, which might lead to the endorsement of numerous odd beliefs. This is the case when one considers healthy people with unusual ideas – they entertain numerous strange beliefs across numerous domains (some of which contradict one another (Pechey and Halligan, 2012)), I would consider this a belief evaluation dysfunction. In your extensive interactions with patients with monothematic delusions have you ever seen anything like this?

Since we are discussing challenging topics, what is the 2-factor stance on autochthonous delusions – beliefs that apparently arise completely de novo? Do these speak to the sub-personal nature of the 2-factors (Dennett, 1969)?

Now to respond to your responses: As I suggest above, I think we might be talking about different levels of explanation but we are essentially talking about the same things. As I suggested in my initial response, I feel that prediction error may be the implementation of Factors 1 and 2 – it signals that the parameters of our current model are inadequate and need to be updated (within model prediction error, factor 1) and it signals that the current model is inadequate and needs to be replaced (prediction error over models, your factor 2).

When you responded to Lisa’s question about prediction error, you suggested that we (myself, Paul Fletcher, Chris Frith) were wrong, that we thought there was a failure in prediction error and that the prediction error in 2-factor theory was intact/appropriate.

I think this is where some confusion arose.

Do you agree that our theories might have more in common than we are appreciating?

I suppose the only thing you might take issue with is that I think prediction error dysfunctions at the levels of factor 1 and factor 2 may be related to one another. I feel that perception and belief ought not be strictly separated and that perception might involve belief, or as Helmholz called it unconscious inference (Helmholtz, 1878/1971), a point that you allowed for but did not feel relevant to delusions. For me, this is critical, since aberrant prediction errors can arise from various junctures (e.g. a failure of top-down belief-based prediction, excessive bottom-up sensation).

In the spirit of our discussion though, I should like to finish with a respectful disagreement: I believe you suggested the third factor – generation of candidate explanations – not me. This arose when I raised the peculiar temporality of the factors: the brain regions mediating factors 1 and 2presumably get damaged simultaneously, but people wake up from their coma, see their spouse, lack familiarity, update to the imposter belief and then never update again. You suggested a third factor, candidate genesis, about which I sought clarification (e.g. is there anyone in whom this process is disrupted?). I don’t think this answered my original concern though; if belief evaluation is damaged at the same time as factor 1, why would the new imposter belief be formed? Why do they update their belief this one final time? I’m afraid I still don’t understand.


References

Cermolacce M, Sass L, Parnas J (2010) What is bizarre in bizarre delusions? A critical review. Schizophrenia Bull 36:667-679.


Corlett PR, Taylor JR, Wang XJ, Fletcher PC, Krystal JH (2010) Toward a neurobiology of delusions. Prog Neurobiol.


Dennett DC (1969) Content and Consciousness. London: Routledge and Kegan Paul.


Ellis HD, Young AW, Quayle AH, De Pauw KW (1997) Reduced autonomic responses to faces in Capgras delusion. Proc Biol Sci 264:1085-1092.


Helmholtz H, von. (1878/1971) The Facts of Perception. In: Selected Writings of Herman von Helmholtz (Kahl R, ed): Weslyan University Press.


Marr D, Poggio, T. (1977) From understanding computation to understanding neural circuitry. Neurosciences Res Prog Bull 204:301-328.


Pechey R, Halligan P (2012) Using co-occurrence to evaluate belief coherence in a large non clinical sample. PLoS One 7:e48446.

6 comments:

  1. Thanks Phil! This is a follow-up comment to what I wrote about Max's post.

    I really like your idea about level difference. I also agree that your view is different from what is usually taken to be two-factor account at some important issues.

    But, the relationship between your view and two-factor account is still not perfectly clear to me. Here is what you say about delusion maintenance, "abberrant prediction errors might re-evoke the representation of the delusion without definitively disconfirming it. This would drive preferential reconsolidation over and above any new extinction learning. The net effect would be a strengthening of the delusion through reconsolidation rather than a weakening by extinction" (Corlett et al. 2009, 3). Certainly, as you suggested, this is a prediction error driven explanation of maintenance. But, I am not sure still that this rules out two-factor account.

    To me, your model is suggesting that the factor that is responsible for delusion adoption (i.e. abnormal salience or experience) and the one responsible for delusion maintenance (i.e. abnormal(?) memory reconsolidation process) are caused by the same thing, namely, aberrant prediction error signals. But, isn't this perfectly consistent with the idea that the adoption-factor and maintenance factor are nonetheless two distinct factors? For example, the idea that my cough and fever are caused by the same event (e.g. viral infection) is perfectly consistent with the idea that cough and fever are two distinct "factors" of my flu. In response to two-factor theorists who claim that adoption-factor and maintenance-factor are distinct, you claim that those factors are caused by the same thing, namely, aberrant prediction error signalling. But, maybe, two-factor theorists can respond to you by saying that those factors are distinct although they have a common cause for the same reason that cough and fever are distinct although they can have a common cause.

    Maybe, at this point, we need to come back to the issue of levels. You do not posit any distinct factor other than aberrant prediction-error signals in the account of maintenance in the lower, implementation levels. But, you might be happy to talk about distinct factor for maintenance in the higher, computational levels.

    Does this make sense?

    Kengo

    ReplyDelete
  2. Thanks Kengo. That makes sense.

    I think perhaps we need to carefully define what we mean by factor. In your flu example, the symptoms, cough and fever, are effects of the virus.

    With regards to delusions, I believe formation and maintenance (cold and fever in your example) are caused by aberrant prediction errors (the virus in your example).

    Is that clear?

    So I think a single factor, aberrant prediction error, causes both delusion formation and delusion maintenance. First aberrant prediction error drives people to attend to and learn about things that stimuli, thoughts, percepts, emotions, events that would normally be ignored, disregarded or deemed irrelevant. These experiences are the impetus (or cause) of delusion formation.

    Once formed, delusions become a prior/expectation/schema through which all subsequent prediction errors are viewed - Karl Jaspers called them dominant apperceptive schema.

    I believe that every time the delusion is deployed, every time it does some explanatory work - as a result of subsequent prediction error - it is reactivated and reconsolidated, laid down more strongly.

    So, whilst delusion formation and maintenance are different processes or consequences, I believe (and have some very preliminary evidence) that they are driven by the same single factor - aberrant prediction error.

    I think this is a key point of departure from 2-factor theory.

    Best wishes,

    Phil

    ReplyDelete
    Replies
    1. This comment has been removed by the author.

      Delete
    2. That is really helpful. Thanks, Phil!

      Just a small point. You said that the delusion formation process and maintenance process are the effects of the same thing, namely, aberrant prediction error signals. But, those processes are at the same time the causes or the causally responsible factors of delusional beliefs, aren't they? It is perfectly coherent that an event E1 is the the effect of E2, and the cause of E3. For example, in the case where bad weather causes train delay, which in turns causes passengers' complaint, train delay is the effect of bad weather and the cause of passengers' complaint. I take your proposal to be something like this. Aberrant prediction error signals cause abnormal salience/memory reconsolidation, which in turn produce/maintain delusional beliefs. Thus, abnormal salience/memory reconsolidation are the effects of aberrant prediction error signals and the causes of delusional beliefs.

      So, presumably, the best way to capture the distinctive feature of your claim is to say that, in your account, the "ultimate" cause of delusional belief is the same thing, namely, aberrant prediction error signals. There are different psychological processes involved in the etiology of delusional belief, but all of them are "ultimately" driven by the same thing.

      Is this a fair understanding of your proposal?

      Kengo

      Delete
  3. If a single factor explains both the formation and the maintenance of the Capgras delusion, what is that single factor, and why does it give rise to this particular delusional belief ("My wife has been replaced by an impostor"? I agree that this issue represents a key point of departure from 2-factor theory. I do not believe a one-factor answer can be given to the question I ask.

    Best wishes,

    Max

    ReplyDelete
  4. I think we are agreed that the surprising lack of familiarity for ones's wife is the impetus for forming the Capgras delusion - it demands explanation.

    I suppose the point at which our accounts overlaps the most is your recent abductive inference argument, this explains how the particular explanation is arrived at - inference to the best explanation for an odd experience.

    It is unclear to me how this explanation is related to your two-factor theory.

    It seems to me that your abductive inference account has much in common with Brendan Maher's empiricist argument, that delusions represent a rational response to an irrational experience, which I agree with.

    Previously in our discussion, you allowed for Factor 2 to be a pre-morbid bias and not something damaged at the same time as Factor 1. So, you have already allowed for delusion in the absence of Factor 2 damage.

    The other point of departure between our accounts is that in prediction error theory, that perceptual disruptions (aberrant prediction errors) can also arise top-down, as a result of inappropriate belief learning or an inability to use priors to constrain experiences. Such top down damage, e.g. to VMPFC might cause spontaneous confabulation, a delusion-like phenomenon that we have already discussed.

    Best wishes,

    Phil



    ReplyDelete

Comments are moderated.