Talk:Similar philosophies

JS:

I think the comparison with O'Regan and Noë, and with other work in that spirit, is very helpful. That work raises various issues that it's worth distinguishing between, to see how the ideas here bear on them. You (AG) distinguish between a couple of related questions raised by O&N:

(1) Does the world (and action in it) replace the need for representations?

(2) Is there a one-to-one correspondence between types of visual episode (including experiences) and types of neural activation?

Some initial thoughts, without worrying here about O&N's exact commitments in that article:

(1) As you say, cases of virtual movement bring out how hard it would be to answer this with a straightforward 'Yes'. And even in cases of actual movement, your proposal does involve representations of a certain sort: "W, is a representation of sensory+motivational contexts and contingencies." So if O&N are taken to suggest a straightforward 'Yes', your proposals are importantly different. For one thing, your proposals are definitely not behaviorist in the philosopher's sense, whereas Noë has sometimes been accused of that (rightly or wrongly). It's one thing to say that representations represent things in terms of actions, another to say that actions replace representations.

On the other hand, you and O&N do agree that there's no need for a "a Cartesian reconstruction of the world", an internal model of a certain kind. I suppose it would be good to know more about exactly what's being denied here, to get a sense of what's distinctive in your proposal. For instance, one charitable way of understanding the idea of an internal model is as the idea that patterns of neural activation are isomorphic with the spatial structure of the represented world. It's not clear to me that your proposals deny that.

(2) Here's that quote again:

"There can therefore be no one-to-one correspondence between visual experience and neural activations. Seeing is not constituted by activation of neural representations. Exactly the same neural state can underlie different experiences, just as the same body position can be part of different dances." (p966)

I think there are a few ways to read this. But one thought is that, just by identifying what's going on in neural terms, we could not (even in principle) identify anything that corresponds one-to-one with a type of visual experience. The O&N-ish thought behind this might be that we get that sort of correspondence only once we take into account what a type of neural activation has been recruited for, in terms of actions in the world. (For more on this see e.g. Hurley and Noë, 'Neural Plasticity and Consciousness', ￼Biology and Philosophy 18: 131–168, 2003.)

That's a minimal reading, and it's not obviously contradicted by cases of virtual movement, but clearly it's still very controversial. Am I right that nothing you say requires even this much? To put that another way, am I right that you allow purely neural representations in following sense: the actions we make are causally responsible for the neural representations we acquire, but once those representations are in place, what they represent could in principle be read off them, using the theory, without our needing to know what actions they've been recruited for?

(I realize the 'in principle's here are a bit vague and hard to assess, but I hope they give the gist of the idea.)

AG: (1) Yes, I agree, there needs to be more clarity about exactly what a representation means. I fail to understand how the need for representation can be avoided altogether for a model of human vision. I agree completely with your last sentence. Representing the world in terms of contexts for actions is still a type of representation. I find it difficult to pin down exactly what type of representation O&N propose. In the model proposed here, $\mathbf{W}$ is all that is stored and it can certainly be described as a representation, in my view.

It is easy to say what is being denied in relation to a Cartesian reconstruction of the world. I discuss this briefly in relation to neurophysiological proposals about 3D representation in the brain. Evidence of a representation of space like the one that can be said to exist in V1 but in another coordinate frame (i.e. with an identifiably different origin and axes) would be a serious blow to the ideas set out here. There is also a great deal of psychophysical evidence, including much of the recent work from our lab that supports the contention that there is not a single Cartesian representation of space in the brain[1][2]

(2) I struggle here. "...we could not (even in principle) identify anything that corresponds one-to-one with a type of visual experience." I am at a loss if that is the case. That one-to-one correspondence is what, as I understand it, neuroscientists are trying work towards. I don't know what an argument 'in principle' of this sort would look like. I am all in favour of 'tak[ing] into account what a type of neural activation has been recruited for, in terms of actions in the world' but we are still left with only synapses and firing rates. There is nothing else, is there?

You ask about the claim that 'the actions we make are causally responsible for the neural representations we acquire, but once those representations are in place, what they represent could in principle be read off them... without our needing to know what actions they've been recruited for'. I would definitely agree with this. I think it is very important in making the step between action (which is what the representation was originally built for, during evolution) and perception. If I pick up an object and turn it around in my hand, I see the slant of its surfaces, its shininess, its colour, I feel its weight. I am not aware that all these judgements are ones that could have discriminated between possible actions nor that I may be making many (or all) of them in series rather than taking in and perceiving the object all at once.

But all of this information is stored in $\mathbf{W}$.

Returning to O&N's statement: 'Exactly the same neural state can underlie different experiences, just as the same body position can be part of different dances.' Because the theory here (and, I imagine, the belief of almost all neuroscientists) relies on the assumption that neural states ($\mathbf{W}$ and $\vec{r}$) are the only variables we have to play with, this statement seems as if it must be false. Nevertheless, there is a lot in what O&N say about the importance of sequences of movements, or a 'dance' as they describe it. If $\vec{r}$ were to jump around randomly or nonsensically, as it might in an epileptic fit, one can imagine these neural states not being compatible with perception. The idea of $\vec{r}$ moving along familiar paths, ones that are normally but not necessarily associated with movements in the real world, bears many similarities to O&N's notion of a 'dance'.

References

1. Koenderink, J. J., van Doorn, A. J., Kappers, A. M., & Lappin, J. S. (2002). Large-scale visual frontoparallels under full-cue conditions. Perception, 31(12), 1467-1476.
2. Svarverud, E., Gilson, S., & Glennerster, A. (2012). A demonstration of ‘broken’visual space. PloS one, 7(3), e33782.