How Perception Yields Knowledge (McDowell Lecture 2)
I’ve finished editing and commenting on my notes from the second McDowell lecture after a hectic weekend of travel and whatnot. I hope as many people like these as seemed to like the last ones (here). It was fascinating hearing McDowell talk and getting to grips with his position from my own Sellarsian/Brandomian perspective. What’s most interesting is that I’ve independently been hitting on some similar ideas to McDowell, but coming from a different, and I believe slightly more productive direction. I’m hoping to give a paper at the upcoming Dublin Sellars’ workshop which deals with these issues (‘Is there a TV in my head?: Content, Functional Mapping, and the Myth of the Given’). I’ve made the abstract available (here) for anyone who is interested (I’m free to give the same paper elsewhere, if you’ll invite me!), along with the abstract for my prospective paper at the Liverpool Thinking the Absolute conference (‘Absolute Spirit as a Work of Art’), which is on unrelated and much more speculative topics (here). Right, back to work then.
1. The question of this lecture is this: how does perception yield knowledge? The basic answer provided by the last lecture is Charles Travis': that it does so by placing our surroundings in view. However, this answer must be elaborating. McDowell is going to break with Travis’ approach in elaborating this idea. Travis thinks that perception places our surroundings in view simpliciter, but that this means perception has no particular content.
2. McDowell thinks that Travis’ view is correct for some cases of perceptual knowledge, but that there are cases (those of what Husserl would call ‘direct bodily presence’) that this account doesn’t properly account for. There are cases of indirect empirical knowledge (say, on the basis of inference) where the world is in view, but not because of the content of our experience on its own.
3. The problem for Travis is that he thinks any account of perceptual content would debar us from holding that its epistemic role is placing our surroundings in view. McDowell’s task is to show that we can retain this insight while nonetheless ascribing content to experience, and to show that there are cases in which the ascription of content is necessary to explain some cases of how the world is brought into view (or as Heidegger would say ‘disclosed’) by perception.
4. What McDowell wants to do is to draw a distinction between certain features of direct perceptual experience that only bring our surroundings into view in the same way that indirect perception, and other features of direct perceptual experience that provide us with knowledge that cannot be given otherwise. This is roughly the distinction between seeing that there is a pig in front of you (which can be indirectly as well as directly available) and seeing that the pig is a certain shade of pink, or has a certain surface texture (which are supposedly barred to indirect perception). McDowell thinks that this distinction is something to do with differences between the perceptual capacities that are involved in bringing the world into view in each case.
5. The conclusion of the first lecture was the negative claim that naive realism is compatible with the ascription of content to experience, whereas this lecture aims to prove is the positive claim that we must ascribe content to experience in order to have an adequate account of it. The real point around which the discussion turns is just how we conceive the epistemic character of experiences. McDowell’s argument against Travis is essentially that the relation between the epistemic character of experience and the phenomenal content of experience differs between kinds of experience, and that the incompatibility argument his position is based on depends upon treating experience as unitary in a way it is not.
6. It is a standard philosophical idea that experience never provides better than defeasible warrant about the environment, insofar as the possession of the relevant content is compatible with the fact that content misrepresents the world. This idea breaks down in two different ways: one that claims experience has no warrant, but accepts it has content, and one that claims experience has warrant, but no content. This is a straightforward split between internalist and externalist views of the nature of experience. McDowell wants to avoid both of these positions. Of an experience that purports to disclose the world, it’s possible that it succeeds doing so. The question is what possible role the subjective character of that experience can play in producing this representational success. McDowell seems to want to say that this role can vary between different kinds of perception.
8. What McDowell wants to argue is that there is sometimes conclusive warrant for beliefs about the features of our environment provided by the content of direct perception. An example of these are those perceptual cases discussed in the first lecture under the heading of ‘incompletely describable’ experiences, or those features of experience which are in some sense intrinsically inexhaustible by descriptions of them through that clauses. The detailed phenomenal features of our experience are describable (e.g., the pig’s shade of pink, or surface texture), and they may yield knowledge about features of our environment, but the way the content of our experience yields this sort of knowledge is distinct from non-phenomenal features.
9. To show this we have to go into detail about the nature of those capacities required to have contentful experiences and the way these capacities are tied together. For instance, McDowell thinks that one cannot activate one’s capacity to have an experience of redness (and thus have red things disclosed to one) without it being activated in conjunction with a capacity to have an experience of something that can display this quality, such as an experience of a rectangle. One can only have red things disclosed to one if one can have red things disclosed to one. This togetherness is thus a crucial feature of the way experiential content is tied to perceptual capacities. What’s interesting then is that even defective activation of these capacities, such as the hallucination of a red rectangle (which doesn’t disclose such a rectangle), will still display this sort of functional interconnection. Second order capacities for identifying aspects of our internal capacities can thus identify invariant features between defective and non-defective activations of capacities. This shows a disconnect between the defectiveness of first order capacities and second order capacities.
Q1. Functionalism Revisited: This stuff about togetherness and the complex interconnection between first order capacities and our second order capacities for assessing defects in the former is the most opaque part of this lecture. It’s not that it’s necessarily wrong, but that it invites us to tell a much more complicated story about both the possible cognitive architectures of perceiving beings and the actual cognitive architectures we as humans display. The togetherness example shows (or purports to show) that our perceptual capacities are not just defective/non-defective when considered in relation to the features of the environment that they’re supposed to track, but that they have internal relations to one another that allow for more nuanced forms of error/success. However, this leads into the point I made in my last post about between the two different styles of functional explanation on offer (abstracted from the environment and situated in the environment) and the way they are connected. It seems like McDowell should want to tell a seriously explanatory functionalist story about both sides of the purport/success divide and how they relate to one another, but he shies away from functionalism. Some suggested to me that to describe McDowell as a functionalist is perverse. Frankly, I think the linguistic lengths he goes to to avoid using the simpler language of functionality that is available here are what is perverse. To talk about capacities that can be judged to be defective is just to talk about functions. It’s pretty much that simple. If you’re going to go this far, why not go for the whole functionalist hog?
10. Someone can mistake defective exercises of their perceptual capacities for non-defective ones. This seems to imply that no exercise of a perceptual capacity can provide conclusive warrant in the way McDowell thinks. He thinks that this doesn’t apply to all kinds of experience though, because some of our perceptual capacities are second-order capacities for discriminating the features of our first order ones (including fine-grained descriptions of internal representational states (purport) such as shades and textures). Importantly, there are fallible second-order capacities for determining the fallibility of our first-order capacities, but the point is that the fallibility in each case is distinct. The kinds of modal reasoning (counterfactual variation) we are dealing with in each case are subtly different, and are related through functional considerations regarding the relations of the capacities. This is supposed to show that we can just *know* that there’s a pig bodily present before us, because our second-order capacities give us reason to think that the first-order capacities producing the experience are functioning properly. Moreover, these second-order capacities do so by discriminating features of the internal representational states of the first-order capacities, meaning that it is through paying attention to the way in which the content of our experience purports to represent that we see it succeeds in representing. Experiential content thus plays an important role in the process of bringing our surroundings into view in some cases.
Q2. Defeasibility, Monotonicity and Modality: McDowell wants to insist that he can have conclusive warrant, but this seems too strong. This is because he doesn’t seem to want the warrant to be conditional upon anything. He somehow wants it to be absolute and yet fallible. This is really strange. The way to see this is to think a bit about the nature of defeasible warrant, or non-monotonic reasoning (inferences for which ‘if P the Q’ holds good by default, but which can be invalidated by the addition of sets of additional premises). There is a useful distinction to be had here between global non-monotonicity and local non-monotonicity. The former is what we have in mathematics. It is precisely deductive, in such a way that absolutely no addition of further premises can undermine a good inference. The latter is more complicated, and very important, insofar as it is what is made explicit by the panoply of possible modal operators (e.g., historical necessity, physical necessity, political possibility, etc.). We might say that these latter kinds of reasoning are imprecisely deductive, insofar as they depend on a more or less implicit form of restriction very much like the more or less implicit forms of quantificational restriction we find outside of mathematical contexts. What gets restricted between the different localities is the potential defeasors that we are willing to consider, and thus which inferences are treated as defeasible in practice. Take the example of political possibility/necessity: someone might claim that if a political candidate is caught stealing, then they will not be elected. This is a monotonic inference in all but those reasoning contexts in which we are willing to ignore very outlandish possibilities (e.g., group mind control, the legalisation of theft, etc.). This kind of variation in potential defeasors is basically what we’d call restriction of the range of counterfactual variation in most hypothetical reasoning contexts.
The problem for McDowell is that if conclusiveness is interpreted as global monotonicity, then his thesis is simply false, whereas if it’s interpreted as local monotonicity, then it’s fine, but not quite in the way he means it. What McDowell seems to be getting at is that the ranges of defeasors for our first-order and second-order capacities are different, and thus we can assume that the second-order capacities work (by simply not considering the specific factors that would cause them to fail) while still actively reasoning about the fallibility of our first-order capacities. From the perspective of a subject who makes this assumption, then, there is a sense in which they can just know whatever their first-order capacities are telling them in certain cases, insofar as their second-order capacities can determine the correct functioning of their first-order capacities and thereby provide the beliefs they produce with warrant. This means that the relation between first-order and second-order perceptual capacities handles the relation between the abstracted representation (purport) and situated representation (success) in the proper functional terms, as long as it is handled right. The big problem seems to be that McDowell isn’t willing to become a full blown functionalist (a la Kant/Sellars), because he seems to think that this abandons the sort of common sense correctness of naive realism. He’s right to be afraid of this, but wrong to let this fear prevent him from pursuing the explanatory thread to the end. You have to get outside of the perspective of the subject in order to understand the nature of representational purport properly. If you don’t, you’re just re-importing a cartesian epistemic privilege on the subject’s behalf. You’re not taking second-order fallibility seriously. There could easily be third, fourth, fifth, etc., orders of perceptual capacities, related to one another in complex and overlapping ways. There is no upper limit, and we shouldn’t pretend that the first-person perspective defined by the highest factual order somehow reintroduces subjective certainty.
We end up with a bizarre hybridisation of folksy phenomenological wisdom and quasi-functionalist apology for it. The fundamental problem for McDowell is that he wants to explain the unity of the two sides of the subjective character he’s talking about in introspective phenomenological terms. It’s a matter of what it’s like to be in that state according to you, rather than what it is to be in that state, according to a dual functionalist account that ties together our internal and external causal economies. This is disastrous, as only the latter could in principle incorporate the externalist element McDowell is trying to incorporate in order to save naive realism. At the end of the day, it only works if we’re naive about how we’re able to describe our relation to the external world, which is to say, if we think we can genuinely explain this relation without engaging in natural scientific causal explanation. This naivety goes hand in hand with his collapsing of the second level of defeasibility into global monotonicity. The discursive space becomes one that is once more defined by my personal prejudices, or by *my* authority, and the point of the functionalist story about higher-order perceptual capacities should be precisely to take away precisely this kind of authority (a la Sellars).