Our next Mind & Language article symposium is on Philipp Koralus’ paper, “The Erotetic Theory of Attention: Questions, Focus and Distraction”, from the journal’s February 2014 issue, with commentaries by Felipe De Brigard, Christopher Mole, Catherine Stinson, and Sebastian Watzl.
In his paper, Philipp argues that the functional role of attention should be understood in terms of the relationship between questions and what counts as answers to those questions, showing how this account sheds light on numerous issues in the philosophical and psychological study of attention in perception, cognition, and action. Below, he summarizes the article in a brief précis, then responds to each of the commentaries.
Comments on this post will be open for at least a couple of weeks. Thanks to Philipp and all the commentators, and in advance to all of you for what I’m sure will be an excellent discussion. Thanks also to Sam Guttenplan, the other Minds & Language editors, and the staff at Wiley-Blackwell for their continued support of these symposia.
PS. You can learn more about Philipp and his work here and here.
***
The relevant links are below:
Thanks all for these thoughts, and to John and Robert for getting this going.
With the aim of getting the discussion going, I’d just like to pose one question as a way of drawing together some of Sebastian’s worries that also relates to the contrast between the erotetic theory and the cognitive unison theory discussed in Chris and Catherine’s remarks (I hope to come to Felipe’s comments at some point). Let me apologize in advance if I’ve missed the places where obvious replies are available in the text (I have read them somewhat quickly).
We now have two adverbial theories of what it is to F attentively, Chris’s and Philipps. My question is whether Chris’s account has one advantage in terms of slightly more generality in respect of some relevant F’s?
The idea is that there might be some tasks that are not goal-directed and yet can be done attentively (Sebastian has a putative list, some of which might fit the bill). I take it that Philipp’s notion of task is restricted to goal-directed actions, so task is a technical notion for him that is only a subset of all actions. On p. 33, he speaks of “every task…aims at some range of goals that can be cognitively monitored”. I don’t think such a claim is true if you replace “task” with “intentional action.” I agree that a useful apparatus can be thinking of such tasks in terms of questions/answers.
The issue is whether we can do things attentively without the task being goal-directed, and it seems that we can: I look at X attentively just to look. You can do this too: pick something and just have a look for as long as you want. There might be better examples, putative non-goal directed actions that can be attentively done.
I think Chris can handle such cases because his account of cognitive unison is not tied to monitoring goal completion. One’s looking is attentive because the relevant resources are brought to bear to serve looking.
I suppose one can come up with goals such that the erotetic apparatus can get a more natural grip on the case, but I wonder if that wouldn’t be a bit ad hoc. Accordingly, why isn’t Chris’s account more general on this point given that he has a less restricted notion of task?
A first thing I had in mind when I suggested that the class of tasks must either is the class of intentional actions or a superset (rather than a subset) of it was that I thought that this was the only way to make it somewhat plausible that tasks are the only things that can be done attentively. So, I took it to be charitable reading of Phillipe’s view of tasks that they are a (maybe improper) superset of intentional actions in order to make plausible the claim that these are all and only forms of behavior that can occur attentively.
In addition, I took it that it is quite common in the philosophy of action to equate intentional action with actions done for a reason (e.g. in Anscombe on Intention), and that is commonly supposed to imply that the agent has some goal (e.g. in Smith’s influential Humean Theory of Motivation). Classically, these intentional actions are also supposed to be acts of which we are in a position to have “non-observational knowledge.” I took it that cognitive monitoring is one major view for explaining why we are in a position to have such knowledge.
Of course, it’s ok to restrict the class of tasks more than that (it’s a technical notion, after all). But now the worry that Wayne raises above becomes pressing.
Thanks for starting the discussion, Wayne.
The idea is that my notion of a task is in not more restrictive than Chris’. He actually explains this really helpfully in his comments.
I actually mention the task of staring at a wall on p. 33 of my paper. I don’t see a problem at all. In the minimal sense I need, every action has success conditions. If I’m going to be staring at a wall just to look, I better be staring at the wall (continuously). If I’m attentively staring at a wall, I’ll be sensitive to when my gaze wanders and when it’s on target.
Thanks also to Bob for tuning in. The short answer is that the erotetic theory of attention has no in-principle commitment to making the question/answer process conscious in all cases. I’m staying neutral.
However, on a side note, it’s worth noting that there are theorists who want to deny that the cases described are cases of unconscious attention (this is sometimes taken to be obviously nuts, but I think it’s not so obvious). They do so by arguing that attention in the full sense involves a lower-level extraneous component in tandem with a top-down component. The idea would then be to say that cases like the ones described are explained by effects on the extraneous component working without the right involvement of the top-down component, so that they don’t involve attention in the full sense. If you adopt the erotetic theory of attention, you can say that the extraneous bottom-up component just is what I call “focus.” You could then say that all those results are results about focus but not about attention. Whether that’s what you would want to say is a different matter. The erotetic theory does not directly commit you to that. However, if you would like to resist the idea that attention can come apart from consciousness, as some theorists might, you could try to attribute the described data to effects of focus rather than attention and the erotetic theory of attention would give you a framework for formulating this sort of view.
Thanks Philipp for the response, though I’m not sure it addresses my worry and I didn’t see Chris’s elaboration of your account to do so either. In his commentary, Chris says, in response to cases like doodling or daydreaming: “The erotetic theory allows that these can indeed be tasks, whenever their subject’s pursuit of them is a *targeted attempt to make something the case.*” This response is plausible to me if the making something the case is a matter of goal direction. But I’m thinking of cases where there isn’t any obvious goal direction or some clear success/completion condition that lets the erotetic apparatus get a grip.
It seems to me at least initially plausible that there are cases of attentive behavior where this monitoring of task completion doesn’t get a grip. I was not thinking of a case where you decide to look at the wall in order to focus on it, to hold your gaze for 5 seconds, to not look elsewhere, etc. These teleological characterizations naturally admit of talk of success conditions where this can be fleshed out in terms of the erotetic theory.
But can’t we just do things where there isn’t anything more to be said about succeeding beyond the fact that one just does it? And in such cases, are there not cases where this doesn’t naturally give rise to talk of monitoring, even in a deflated sense (e.g. where this is subpersonal). So, it seems like I could just be doodling or staring at something, and indeed in an attentive way (process theorists would say that you are just thereby attending to what you are doodling or looking at). But I would not be bothered if something broke my gaze or my gaze drifted (or I stopped doodling or started in fact writing). There would be no sense that somehow, I left the task incomplete and connected to this, no sense that anything needs to be monitored.
So, perhaps I can turn this into a question: I am shifting my gaze around just looking at things, for no reason. I’m just going to have a look, I say. Perhaps I am bored, have time to pass but really have no goals or intentions beyond just looking. My gaze passes over an object and lingers on it for what happens to be 7 seconds (but again, the time isn’t at issue) and I move on without being bothered by doing so. Perhaps in that 7 seconds, I scan various features of the object, not for any reason, but just because that is how my attention shifts (using process language again). Can’t the lingering involve attentively looking at the object, even if in some minimal sense? If so, what is the erotetic analysis of the case? Or do you just deny that this would be attentive looking?
I would want to say that if you’re looking at something attentively but you’re not particularly bothered if your attention (not just your focus) gets successively drawn away, then you are just not particularly attached to sticking with one task. In the very broad sense of task at issue, I’d say you are shifting between different tasks successively. Task monitoring does not commit you to being bothered by task switches.
I’d suspect that so-called attention deficit disorder has something to do with having trouble staying motivated to stick with one task (and not with trouble assigning focus, for example). A recent study of parallel versus serial visual search in ADHD fits in nicely: Rejnen and Opwis (2008) in JoV say that their ” results suggest that children with ADHD show no specific deficits in selective attention but show poorer performance when external motivators are lacking,” since there was no specific problem with serial search for the ADHD participants. What they call selective attention, as involved in serial visual search, I would call focus.
Interesting. At the end of his response to Felipe De Brigard’s question about involuntary attention Koralus says “It could be that involuntary attentional capture is involuntary focus capture which then may or may not be followed by a conscious accommodation of a question. I could also be that both focus capture and question accommodation can happen without volition.” I’m unsure if this means that Koralus assumes that attention involves a consciously held question. There is plenty of evidence that attention can be directed by unseen cues – the transient stimulus that attracts exogenous attention can be quite unseen and yet the effect of attention is revealed by its effect on subsequently presented seen targets (e.g. McCormick, P.A. (1997). Orienting attention without awareness. Journal of Experimental Psychology: Human Perception and Performance, 23 (1), 168-180 – for a comprehensive review see Mulckhuyse, M & Theeuwes, J (2010) Unconscious attentional orienting to exogenous cues: A review of the literature. Acta Psychologica 134(3), 299-309). I’ve just been doing similar stuff using masked cues and it is quite clear that the subjects don’t know that they are attending. They have an explicit task (e.g. report the orientation of lines when they appear) but they don’t know that some lines are processed differently from others by virtue of occurring at cued locations. (I’ve also done this in blindsight where the subject was unaware of the cues and unaware of the target stimulus.)
Thanks to Bob for tuning in. The short answer is that the erotetic theory of attention has no in-principle commitment to making the question/answer process conscious in all cases. I’m staying neutral.
However, on a side note, it’s worth noting that there are theorists who want to deny that the cases described are cases of unconscious attention (this is sometimes taken to be obviously nuts, but I think it’s not so obvious). They do so by arguing that attention in the full sense involves a lower-level extraneous component in tandem with a top-down component. The idea would then be to say that cases like the ones described are explained by effects on the extraneous component working without the right involvement of the top-down component, so that they don’t involve attention in the full sense. If you adopt the erotetic theory of attention, you can say that the extraneous bottom-up component just is what I call “focus.” You could then say that all those results are results about focus but not about attention. Whether that’s what you would want to say is a different matter. The erotetic theory does not directly commit you to that. However, if you would like to resist the idea that attention can come apart from consciousness, as some theorists might, you could try to attribute the described data to effects of focus rather than attention and the erotetic theory of attention would give you a framework for formulating this sort of view.
I’ve been trying to post the following comment since yesterday, but it keeps disappearing into cyberspace.
There are several lines of discussion here to which I’d like to contribute. I hope it won’t be thought presumptuous if, in this comment, I restrict myself to a couple of places where the discussion raises apparent problems for my own view.
The first point that I’d like to address is the one de Brigard raises: Is the metaphysical distinction between processes and adverbial phenomena a real one?
This is a tricky issue because it takes us into the territory of meta-metaphysics; a territory in which it is hard to move without finding that one has trodden on toes. I am enough of a realist about metaphysical matters to think that the distinction is indeed real — at least as real as the distinction between objects and events (from which it is derived).
What, then, do I say about de Brigard’s supposedly problematic example of neural synchrony? I say very much the same thing that Philipp says, in his reply to commentators. Neural synchrony is a matter of certain events (neural firings) occurring over the same period, with the same rhythm, and with the same pattern of deviations from that rhythm. Since events are the sorts of entities that have rhythms, and objects are not, the thing that qualifies some occurrence as being an instance of neural synchrony is a property of events (rather than a property of objects). Neural synchrony therefore counts, quite straightforwardly, as an adverbial phenomenon (by the criterion given in §2.3 of Mole 2011).
A scientist studying neural synchrony might very well be interested in the process (or, more likely, the several processes) that bring synchrony about, and so might talk in a process-first sort of way. But when it comes to seeing how her theory of these processes does the work of explaining neural synchrony, the thing that we’ll need to explain will still, ultimately, be the instantiation of a manner of occurrence (viz. synchronousness). Realizing this might be of more than merely metaphysical importance, when, for example, it comes to understanding whether the plurality of the processes that bring synchrony about compromises the integrity of synchrony, as a single phenomenon with its own theoretical role to play.
I seem to be having more luck with posting comments now, so here’s another.
The second line of discussion to which I’d like to chip in concerns Koralus’s claim (to which Stinson is sympathetic) that the cognitive unison theory is unable to account for degrees of attention.
Koralus says “I think it’s intuitive that I can pay more or less attention to my reading, or to driving.” and that “Unless I’ve missed something, there is no sense [according to the cognitive unison theory] in which the same task is performed with more or less attention.” It would be a serious objection to the view if there were indeed no such sense, but I think that there is one.
The cognitive unison theory has a couple of things to say about performing the same task with more or less attention. (It says them, for those who want to read along, in §4.12 of Mole 2011.)
The first point is that we should think of our claims about degrees of attention as allowing for lots of pragmatic variation, with different standards of comparison being used in different contexts. Such claims are similar, in this regard, to our claims about degrees of skill. One can sensibly ask ‘Is this golfer more skilled than that one?’ or ‘Is this surgeon more skilled than that one?’, but, it seems to me, we would usually want to reject the question of whether some golfer is more skilled than some surgeon. The way to allow for that is to admit that the task in question plays a role in determining the scale relative to which our judgements of degree are made. The result of this is that it would be a mistake to think that any general account of skill would need to have some one account of degrees built into it. Similarly, it seems to me that it would be a mistake to build some one account of degrees into our theory of attention. The cognitive unison theory allows that there might be several different ways to answer the question of whether x is paying more attention that y. It gives a role to the attended task, in determining which of these ways we use. Casting the task in this role does not prevent comparisons about degrees of attention to the same task.
Although this goes someway to addressing the issue, the second of the things that the cognitive unison theory says about degrees might, in the present context, be the more crucial one. Here’s an analogy (from p. 86 of Mole 2011) which explains the point.
Suppose you are looking to recruit a new staff member, and that, in doing so, you produce a list of criteria that a candidate must satisfy in order to be qualified for the job. These criteria, you say, specify exactly what it will be for a person to count as ‘being qualified’. It is no objection at all to point out that these criteria are in some way violated by those people who are only partially qualified. Nor would anybody suppose that this was an objection, unless they were confused about the meaning of ‘partially’.
Similarly, it is no objection to the cognitive unison theory, thought of as a specification of what it is for a person to count as paying attention, to point out that people who are only partially attending fall short of meeting the conditions that it specifies. They can fall short to various degrees, and when they do it will be true to say of them that they are paying more or less attention, just as it can be true to say of the partially qualified candidates that some are more, and some less, qualified for the job.
A specification of what it is to be qualified for the job gives you an account of what it is to be partially qualified, even if that specification takes the form “X is qualified if and only if …”, rather than “X is qualified to degree n, to the extent that …”. Similarly, a theory of what it is to pay attention gives you an account of partial attention, even without being the ‘graded account’ that Stinson claims would be preferable. In this sense, the cognitive unison theory does allow that a task can be performed with more or less attention.
One of the areas where the erotetic theory of attention diverges from the cognitive unison theory is with respect to partial attention. In section 4.12.2 of Chris’ book, he writes “But there is still a sense in which the cognitive unison theory does not allow for cases of attention to be partial [his emphasis].” He continues, further down the page, “cases where something less than unison is instantiated are not cases of attention at all.” I think this the area where Catherine, Sebastian (elsewhere in print), and I share some worries. On the erotetic theory of attention, there is no reason not to expect cases of partial attention. Moreover, we don’t have to reconstruct a notion of different degrees of attention indirectly. It seems to me that this is a reason to prefer ETR.
I was a little puzzled by the reply to my comments. I wasn’t suggesting that logical formalisms are bad things, nor that my problem with them is that they’re hard to decode. What I was questioning was whether all the details of this particular formalism are really necessary, and what exactly it is that they add to the account. I also wasn’t suggesting that we throw out accounts based on probability distributions, nor that we mustn’t assign bits of one’s formalism with intuitive labels.
I agree that making it non-technical might make notions like congruence hand-wavy. I was suggesting that congruence could be formalized in a simpler way, without there having to be questions involved, and wonder what would be lost. As far as I can tell, nothing important would be.
So I’m not sure how to interpret comments like “I am not claiming to have shown that we must think of completion conditions for tasks in terms of questions. I am merely arguing that there is no clear reason we cannot do so, which means there is no clear reason to reject my view on such grounds.” Calling it an erotetic theory, and much of the paper being concerned with showing how task completion might be monitored via questions suggests at least that you want us to think of completion conditions in terms of questions, and that that’s one of the main points of the paper. Perhaps no real harm is done by making a formalism more complex that it need be, and I even suggested that if the complexities were ones that made the account more intuitive or accessible, then that would be a reason for keeping them. Where that’s not the case, there certainly are reasons for not adding unnecessary formal details. One is that it might make readers believe that question-answering is somehow integral to the view, when perhaps it isn’t. It’s not exactly radical to consider simplicity a virtue when evaluating theories.
So is question-answering integral to the view or not? If it is, then what exactly does it contribute beyond the streamlined version I suggested? The last line of your reply suggests the answer is: “task monitoring involves an epistemic aim.” But why do we need questions in order to do that monitoring? Even if we take questions to be minimal representations of ranges of acceptable answers, this involves positing a layer of representations that we then need some cognitive resources to encode, store, and retrieve. I realize that you don’t require them to be conscious representations in every case, but hiding them outside of consciousness doesn’t make them seem any less superfluous. Can’t we monitor task completion without explicit representations of the completion conditions?
I still don’t see much room for significant simplification, but maybe Catherine could work out a proposal. I take part of cognitive science to be in the business of giving accounts of the function or computational aim of cognitive systems. For example, on Marr’s view, the function of vision might be to map retinal inputs to object representations. I think that the function of attention is a form of task monitoring. Monitoring implies a minimal epistemic aim, at least the aim of monitoring *whether we have completed the task*. The minimal way of representing an epistemic aim seems to be in terms of a question (note again: a question is not an interrogative sentence, but in the realm of contents). Now, we sometimes talk of attentional *focus* as applying to objects, features, or positions, rather than applying to tasks. With the erotetic theory, we can make sense of this without saying that attention proper applies to anything but tasks. Here is how: if we take questions to have a distinguished domain, as we find in standard proposals in the semantics literature, we can define a functional role for *focus* in terms of a condition for what it takes to cognize something as an answer (mirroring a similar condition in linguistic interpretation). We say that one of the conditions for counting something as answering the question involved in our task monitoring is that what is picked out by *focus* must correspond to the domain of the question. This allows us to unify attention in cognitive control and the notion of attentional *focus* in perception. You attend to a task to the extent that you are sensitive to whether your corresponding monitoring question has been answered, and in order to count it as answered focus has to pick out something correspond to the domain of the question.
Thank you, all, for the comments, and apologies for my slow reply. I have really enjoyed this discussion. I just wanted to offer my two cents on two of the issues that some of you have been talking about. First, on the issue of process vs adverbial phenomena. As Chris said, having a full-fledged and in-depth discussion about this question would get us into metaphysical terrains that we may not want to get into. I guess all I wanted to add, after reading the replies, is that I am less of a realist than Chris is, and more of a “constructive empiricist” of sorts, when it comes to cognitive ontologies. I feel that thinking of attention as something akin to cognitive unison, not to be identified with a unique cognitive process, is a great hypothesis—indeed, a hypothesis that seems to fit the data better than many other alternatives in the offing. But I don’t think the evidence is strong enough (yet) to reject the view that there may be a possible unifying description of the processes underlying attentional and only attentional behaviors that may fit the data equally well. After all, neuroscientists keep coming up with novel and more comprehesive models of cognitive processes that go way beyond modeling single cell behavior, or even co-variation of BOLD signal, to capture the complex, tiered and convoluted dynamics of neural systems. So I guess I advocate for some sort of agnosticism.
The second point is about the issue of involuntary attention. In your reply, Phillip, you give two options to the case of the sound behind you while you’re reading. Option one, and I quote, “is that you accommodate the answer, adopting a new question (and thereby a new task) to which “There’s a loud noise behind you” would be a congruent answer, maybe, “what sort of thing going on behind me?” You might then end up further investigating because the information that there is a loud noise may only be a partial answer to your newly accommodated question.” But I find this option unsatisfactory. I am not asking how the theory can accommodate switching attention voluntarily from one task (i.e., reading) to another task (i.e., finding out the source of the bang). I am asking how the theory accommodates my involuntary and mandatory switch to the bang behind me from my voluntary effort to pursue the task of reading. I want to know how the theory accommodates my attentional shift PRIOR to my engaging in asking the question “what sort of think is going on behind me”. For I feel that I wouldn’t have been able to entertain that question if my attention hadn’t been captured, albeit involuntarily, by the bang. Maybe you want to say that what was captured involuntarily wasn’t my attention, but the “focus”. (In fact, you DO say that at the end of the paragraph: “It could be that involuntary attentional capture is involuntary focus capture which then may or may not be followed by a conscious accommodation of a question. I could also be that both focus capture and question accommodation can happen without volition.” ) Of course, you can say that, but I think that’s just re-naming the problem, as I also mention in my comments. Here “focus without volition” just renames the notion of “involuntary attention”, and the answer is, at best, circular.
The second possibility you offer is, and I quote again, is: “You could also refuse to be distracted and then exercise some top-down control on your focus assignment to bring it back to something that’s relevant to your task.” Sure, but this does not offer a solution to the problem. You are not telling us how the theory handles cases of involuntary exogenous attention, but how it handles cases of shifting attention back to voluntary, top-down control cases. That is a different issue.
At the end of that paragraph you acknowledge that “whether accommodating a new question (and task) is always something that requires the exercise of volition is a separate issue.” And I guess that this hits the needle in the head, because I just don’t think it can be a separate issue. I really do think this is an important issue the theory needs to face, if it is supposed to capture the phenomenon of attention in general.
Also, I noticed that Carolyn Dicey Jennings, over at the newapps blog, has started an interesting discussion on this issue. Here’s the link:
https://www.newappsblog.com/2014/05/the-problem-of-captured-attention.html