Last year I started to try to take note of papers that engage with my work in some way (previous posts here, here, here, here, here, here, and here). The hope was to get some thoughts down as a reference point for future paper writing. So far not much in that department has been happening; with a 3 year old and a 1 month old it is tough to find time to write (understatement!) but I am hoping I can “normalize” my schedule in the next few weeks and try to get some projects off of the back burner. At any rate I have belatedly noticed a couple of papers that came out and thought I woud quickly jot down some notes.
The first paper is one by Joseph Gottlieb and came out in Philosophical Studies in October of 2017. It is called The Collapse Argument and makes the argument that all of the currently available mentalistic first-order theories of consciousness turn out to really be versions of the higher-order theory of consciousness. I don’t know Joseph IRL (haha) but we have emailed about his papers several times, though I usually get back him too late for it to matter on account of the 16 classes a year I have been teaching since 2015 (for anyone who cares: I am contractually obligated to teach 9 a year and in addition I teach another 7 as an adjunct (the maximum allowed by my contract)…sadly this is what is required in order for my family to live in New York! ) and I have blogged about his work here before (linked to above) but I really, really like this paper of his. First, I obviously agree with his conclusion and it is nice to see some discussion of this issue. I took some loose steps in this direction myself in the talk I gave at the Graduate Center’s Cognitive Science Speaker Series back in 2015. I thought about writing it up but then had my first son and then found out about Joseph’s paper, which is better than what I could have come up with anyway! I suppose the only place we might disagree is that I think this applies to Block’s first-order theory as well.
But even though I really like the paper there is a bit I would quibble about (but not very much). Gottlieb seems to take seriously my argument that higher-order theories are in principle compatible with phenomenological overflow but I am not sure I agree with how he puts it. He says,
As Richard Brown (2014) points out, HO theorists don’t need to claim that we are aware of our conscious states in all their respects. I might be aware that I am seeing letters (a fairly generic property) but not the identity of every letter I am seeing. In other words, I can be unaware of some of the information represented by the first- order state without the state itself being unconscious (ibid). What happens, then, is: I am phenomenally conscious of the entire 3 X 4 array, with representations of the identities of all the letters available prior to cuing. But only a small number (usually around four) ever get through, accessed by working memory. That’s overflow, and perfectly consistent with HO theory.
In the paper he is citing I was trying to make the point that the higher-order theories which deny overflow do not thereby also commit themselves to the existence of unconscious *states* which are doing heavy lifting. If the states are targeted by the appropriate higher-order representation then those states are conscious. Yet one may not represent all of the properties of the state and so, even though the state is conscious, there is information encoded in the state which you are not aware of (and so is unconscious). That unconscious information (that is to say, that aspect of the conscious state) is (presumably) what you come to be aware of when you get the cue in the relevant experiments. So it is a bit strange to see this part of the paper cited as supporting overflow (though I do think the position is compatible with overflow I wasn’t thinking of it in this way). But I think I see his point. On the higher-order view it will true to say that one has a phenomenally conscious experience of all of the letters and the details but only access a few (even though what it is like for one may not have all of the details, which is really what I think the overflow people mean to be saying).
This point, though, is I think they key difference between higher-order theories and Global Workspace theories (which is what Block is really targeting with his argument). The basic idea behind the higher-order approach is this. When one is presented with the stimulus all or most of the details of the stimulus are encoded in first-order visual states (that is, states which represent the details of the visual scene). Let’s call the sum-total representational state S. S represents all (or most) of the letters and their specific identifies. One can have S without being aware that one is in S. In this case S is unconscious. Now suppose that one comes to have a (suitable) higher-order awareness that one is in S. According to the higher-order theory of consciousness one thereby comes to have a phenomenally conscious experience of S and becomes consciously aware of what S represents. But since one’s higher-order awareness is (on the theory) a cognitive thought-like state, it will describe its target. Thus one can be aware of S in different ways. Suppose that one is aware of S merely as a clock-like formation of rectangles. Then what it is like for one will be like seeing a clock-like formation of rectangles. Being aware of S seems to keep S online and as one is cued one may come to have a different higher-order awareness of S. One may become aware of some of the details already encoded in S. One was already aware of them, in a generic way, but now one comes to be aware of the same details but just in more detail. Put more in terms of the higher-order theory, one’s higher-order thought(s) come to have a different content than they previously did. The first higher-order state represented you as merely seeing a bunch of rectangles and now you have a state that represents you as seeing a bunch of rectangles where the five-o’clock position is occupied by a horizontal bar (or whatever). Notice that in this way of thinking about the case there are no unconscious states (except the higher-order ones). S is conscious throughout (just in different respects) and it will be true that subjects consciously see all of the letters (just not all of the details).
I want to keep this in mind as I turn to the second paper but before we do I also like Gottlieb’s paper because it actually references this blog! I think this may be the first time my personal blog has been cited in a philosophy journal! I will have more to say about that at some point but for now: cool!
The second paper is by Nicholas D’Aloisio-Montilla and came out in Ratio in December 2017. It is called A Brief Argument for Consciousness without Access. This paper is very interesting and I am glad I became aware of it and D’Alosio-Montilla’s work in general. He is trying to develop a case for phenomenological overflow based on empirical work on aphantasics. These are people who report lack of the ability to form mental imagery. I have to admit that I think of myself this way (with the exception of auditory imagery) so I find this very interesting. But at any rate the basic point seems to be that there is no correlation between one’s ability to form mental imagery (as measured in various ways) and one’s ability to perform the Sperling-like tasks under discussion in the overflow debate. His basic argument is that if you deny phenomenological overflow then you must think that unconscious representations are the basis of subject’s abilities. Further, if that is the case then it must be because subjects form a (delayed) mental image of the original (unconscious) representation. But there is evidence that subject’s don’t form mental images and so evidence that we should not deny overflow.
I disagree with the conclusion but it is nice to see this very interesting argument and I hope it gets some attention. Even so, I think there is some mis-characterization of my view related to what we have just been talking about in Gottlieb’s paper. D’Alosio-Montilla begins by setting the problem up in the following way,
The reports of subjects [in Sperling-like tasks] imply that their phenomenology (i.e. conscious experience) of the grid is rich enough to include the identities of letters that are not reported (Block, 2011, p.1; Land- man et al., 2003; cf. Phillips, 2011b). As Sperling (1960, p.1) notes, they ‘enigmatically insist that they have seen more than they can … report afterwards’. Introspection therefore suggests that subjects consciously perceive almost all 12 items of the grid, even if they are limited to accessing the contents of just one row (Block 2011; Carruthers, 2015). The ‘overflow’ argument uses this phenomenon as evidence in favor of the claim that the capacity of consciousness outstrips that of access. Overflow theorists maintain that almost all items of the grid are consciously represented by perceptual and iconic representations (D’Aloisio-Montilla, 2017; Block, 1995, 2007, 2011, 2014; Bronfman et al., 2014; for further discussion, see Burge, 2007; Dretske, 2006; Tye, 2006).
This is a nice statement of the overflow argument and the claim that it is the specific identifies of the items of the grid which are consciously experienced but this way of framing the argument begs the question against the higher-order interpretation. The reports in question do not imply rich phenomenology because, as we have just discussed, subjects are correct that they have consciously seen all of the letters even if they are wrong that they consciously experienced the details. Because of this the higher-order no-overflow theorist can accept that there is no correlation between mental imagery ability and Sperling-like task performance and for pretty much the same reasons that the first-order theorist does: because there is a persisting conscious experience.
D’Aloisio-Montilla then goes on to give two objections to his interpretation of my account. He puts it this way,
A final way out for the no-overflow theorist might be to allow for a limited phenomenology of the cued item to occur without visual imagery (Brown, 2012, 2014; Carruthers, 2015). Brown (2012, p. 3) suggests that subjects can form a ‘generic’ experience of the memory array’s items while the array is visible, since attention can be thinly distributed to bring fragments of almost all items to both phenomenal and access consciousness. Phenomenology, for example, might include the fact that ‘there is a formation of rectangles in front of me’ without specifying the orientation of each rectangle (Block, 2014). However, there a still number of problems with an appeal to generic phenomenology. First, subjects report no shift in the precision of their conscious experience when they are cued to a subset of items that they subsequently access (Block, 2007; Block, 2011).
First, I would point out that my goal has always been to show that the higher-order theory of consciousness is both a.) compatible with the existence of overflow but also b.) compatible with no-overflow views and gives a different account of this than Global Workspace Theories (or other working memory-based views). So I am not necessarily a ‘no-overflow theorist’ though I am someone who thinks that i.) overflow has not been established but assumed to exist and ii.) even if there is overflow it is mostly an argument against a particular version of the Global Workspace theory of consciousness, not generally against cognitive theories of consciousness.
But ok, what about his actual argument? I hope it is clear from what we have said above that one would not expect subjects to report ‘a shift in precision’ of their phenomenology. One has a conscious experience (generic or vague in certain respects) but in so doing you help to maintain the first-order (detailed) state. When you get the cue you focus on the aspect of the state which you had only generically been aware of (by coming to have a higher-order awareness with a different content) but what it is like for you is just like feeling like you see all of the details and then focusing in on some of the details. No change in precision. But even so these appeals to the subject’s reports are all a bit suspect. I use the Sperling stimulus in my classes every semester as a demo of iconic memory and an illustration of how philosophical issues connect to empirical ones and my students seem to be mixed on whether they think they “see all of the letters”. Granted we only do 10-20 trials in the classroom and not in the lab (in Sperling they did thousands of trials) and these are super informal reports made orally in the classroom…but I still think there is a issue here. I have long wanted there to be some experimental philosophy done on this question. It would be nice to see someone replicate Sperling’s results but also include some qualitative comments from subjects about their experience. I almost tried to get this going with Wesley Buckwalter years ago but it didn’t go through. I still think someone should do this and that the results would be useful in this debate.
D’Aloisio-Montilla goes on to say,
Second, subjects are still capable of generating a ‘specific’ image – that is, a visual image with specific content – when the cue is presented. Assuming that the cued item is generically conscious on the cue’s onset, imagery would necessarily be implicated in maintaining any persisting consciousness of the cued item (whether gist-like or specific) throughout the blank interval. Thus, we can still expect to see a correlation between imagery abilities and task performance, because subjects can generate either (1) a visual image with specific phenomenology, or (2) a visual image with generic phenomenology (Phillips, 2011a; Brown, 2014). In any case, subjects who generate a specific phenomenology of the cued item should perform better than those who rely solely on a gist-like experience, and so Brown’s interpretation is also called into question.
But again this seems to miss the point of the kind of no-overflow account the higher-order thought theory of consciousness delivers. It is not committed to mental imagery as a solution. Subjects have a persisting conscious experience which may be less detailed than they experience it as.
Shesh that is a lot and I am sure there is a lot more to say about it but nap time is over and I have to go and play Dinosaur now.