As a materialist I am often troubled by how rarely I use neuroscientific
studies in building or testing cognitive hypotheses. I’m often tempted to write
this off a just a methodology issue; BOLD doesn’t get us to representing
vehicles, EEG and MEG pick up amalgamated waves etc etc. But, still, why
shouldn’t I find what I’m looking for in such studies? That I don’t is
worrisome- especially being from Adelaide, home of U.T.
Place’s brain (link below). Now, 10 years after the ‘decade of the brain’ surely these
methods can answer the questions I’m interested in, after all isn’t cognitive
neuroscience soon supposed to replace every other cognitive science?
For a short time yesterday afternoon I thought I
had found a case where the methods of cognitive neuroscience were to become available
to test a genuinely cognitive hypothesis. Finally I’d have to find something
else to heckle my imaging friends with at conferences. But their collective
eyes had barely stopped rolling at me when I gave up my optimism.
What had gotten my excited was Manos Tsakiris’ (2010) explanation of the rubber
hand illusion. The rubber hand illusion is the experience of a model hand
(usually a rubber prosthesis) being one’s own hand and the experience of touch
coming from the rubber hand. It is induced by touching one’s real hand (which
is held out of sight) in synchrony with the rubber hand (which one sees). Now,
Tsakiris can explain this illusion by the hierarchical processing of three
comparisons. First, the shape and texture of the rubber hand are compared to a
stored body model. If there is a close enough match then the next comparison
can be performed. Second, the arrangement of the rubber hand is compared to a
representation of the current posture of the body. If the rubber hand is
arranged so as to be plausibly a part of the body processing moves onto the
next stage. Finally, the felt touch is compared to the seen touch for temporal
and directional synchrony. If they are similar enough then an experience
(representation) of two distinct events (seen and felt touch) is replaced by an
experience of a single event, the one touch both seen and felt. From this a
sense of the rubber hand as one’s own is elicited.
Now you won’t, of course, get a feel from this
of just how good this model is. But, trust me, it can explain a lot. What I
care about here is that we have an explanation with the usual cognitive psychology
ontology, i.e. representations and computations. Excellent. Now what Tsakiris
adds to this is specific (data driven) hypotheses as to the neural networks
necessary for each comparison in the hierarchy to be performed. Necessary for
the first comparison is the right temporal parietal junction (rTPJ). Necessary
for the second comparison is the anterior parietal cortex (APC). And necessary
for the final comparison is the posterior parietal cortex (PPC).
Great! So we should be able to selectively
impair each comparison by repeated transcranial magnetic stimulation (rTMS) over
each area. Not that I really know if this is possible, e.g. the areas could be
too deep or too big for rTMS to work. But, it seemed like it was worth me checking
out, as we should get different experiences when attempting to induce the
rubber hand illusion after selectively knocking out the rTPJ, APC or the PPC.
As the processing is hierarchical knocking out any of these areas should
prevent the illusion from being elicited. Where we will see a difference is in
some judgements the subject will be able to make.
If the rTPJ is knocked out, impairing the first
comparison then the subject should be unable to discern similarity between
their real hand and the rubber hand in terms of shape and surface features,
e.g. skin texture. If the APC is knocked out the subject should be able to make
this kind of similarity judgement, but be unable to judge the consistency between
the arrangement of the rubber hand and their current posture. Finally if the PPC
is knocked out the subject should be able to make the above two judgements, but
be unable to tell if the seen and felt touch are in synchrony.
Alright, the methods of cognitive neuroscience
can test predictions from a cognitive theory!
The problem is, of course, that each of these
brain regions is needed for more than one function. For example, while Tsakiris
assigns the role of comparing the seen and felt touches, i.e. the third
comparison above, to the PPC, the PPC in the left hemisphere is also necessary
for representing the spatial arrangement between body parts (Corradi-Dell’Acqua
et al. 2008). That is, at least some of the information
needed for the second comparison is represented in the lPPC. So, it seems, if
we knock out the PPC we will knock out both comparisons. So we won’t be able to
selectively impair them after all.
Damn
Oh well…
Next time.
Looking at the activity in different brain locations won’t do it. We need explicit neuronal models — theoretical constructs — of cognitive brain *mechanisms* (not black-box functionalism) that have the hypothesized structure and dynamics competent to perform our cognitive tasks. Then we can make predictions about what kind of cognitive performance to expect under specified stimulus conditions. For an example of this kind of approach, see *Seeing-more-than-is there* in “Space, self, and the theater of consciousness”, pp. 324-325, here:
https://people.umass.edu/trehub/YCCOG828%20copy.pdf
Glenn, interesting post. Your suggestion of testing this theory by TMS sounds good to me and not as useless as you seem to suggest. More generally, I am puzzled by your suggestion that this would be the first case of testing a cognitive hypothesis by cognitive neuroscience methods. Cognitive neuroscientists test cognitive hypotheses every day. You yourself cite evidence that various brain areas are involved in certain tasks. Presumably those claims are supported by cognitive neuroscientific evidence. And don’t forget about neurophysiological methods (usually not on humans, but still relevant), which many people would include as part of cognitive neuroscience.
thanks for your thoughts (and the paper!). I just wanted to add one more reason to worry about testing the model using TMS. Suppose selectively impairing one of the rTPJ, APC or PPC didn’t have the effect predicted by the model. This could falsify the whole model, or just the hypothesis regarding the neural localisation. In other words, the computational/representational account could still be true even if these predictions are not borne out.
As you might suspect, I am also worried that this might be a general concern.
And Guiltiero, there is a danger of us talking past each other here, I think I mean something more specific by ‘cognitive hypothesis’ than you. Perhaps we can get a grip on it this way, can you provide a specific example of a cognitive hypothesis being tested using the methods of cognitive neuroscience?