Yesterday, I suggested that we need a theory of attention that gives attention a central place in the mind and serves as a unified framework that integrates different approaches to attention, both in philosophy and in the empirical sciences. Today, I will sketch the priority structure framework and give some reasons why we should accept it.
I will first introduce the framework with a metaphor. Consider the various elements that characterize your current state of mind: you are thinking about attention; you see the various words in front of you, but also – somewhat dimly – the color of your window; in the background, you hear some music; you feel an urge for another cup of coffee and a lingering worry somehow colors the whole situation. Think of these elements of your mind as little stories that tell you about various aspects of the world and yourself. Attention is not another story, another element of your mind. Instead, attention concerns how the elements are arranged. Attention is the newspaper that contains the stories. It has the thoughts, perceptions and feelings as parts. Just like the newspaper is structured by the story placement, so your mind is structured by what is prioritized relative to what. The front page is just one place where a mental state can appear in our attention. What makes a mental state front page (what makes it occupy your attention at some time) is a relation to what is pushed to the back or into the small print. Priority structure is about how a mental life is formed from its parts. Attention consists in the activity of regulating those structures: pushing elements up and down in the priority ordering, or maintaining a structure for some period of time.
The priority structure framework shows how to integrate attention with other aspects of our mental life. How is seeing related to visual attention? Or thought to intellectual attention? On the one hand, attention depends on other aspects of the mind. In order to attend to an object, you must either perceive it, think about it, imagine it, or be, for example, worried about it. Generally, as I argue in Structuring Mind, it is impossible to attend to something without being in some other mental state that is about that thing. On the other hand, attending to something is also never identical to any of the mental states on which it depends (or any combination of them). Visually attending to something, for example, is never just an instance seeing that thing, since you have direct voluntary control over what you are visually attending to, while you have no direct voluntary control over what you are seeing. This pattern of both dependence and independence is best explained by a view that treats the other elements as parts of attention – parts, though, that even together do not exhaust the whole.
What more is there to the whole than its parts? We get a clue by thinking about how attention involves a comparison. That is what makes various metaphors attractive: attention as a spotlight that illuminates one part of a scene more than others; or as a limited resource or capacity that at each moment feeds one aspect of the mind more than others. From this we get the view that attention is constituted by a structure in which mental states are related or compared to one another.[1] The whole consists of the mental states (the parts) and their relations (the structure). It is possible to change the relations between the parts without changing the parts. With voluntary control, you can, for example, change from organizing your visual world around seeing this text to organizing it around seeing the color of your window. The parts are the same, but their relation has changed.
-
Sebastian Watzl, Structuring Mind: The Nature of Attention and How it Shapes Consciousness (OUP, 2017)
This gives us the skeletal version of the priority structure framework: attention is constituted by a structure of the mind, where mental states are related to each other. The fundamental notion is that relation. I call it the priority relation. The other aspect of the structure are the elements that occur in it. In my view, these are personal or subject level mental states, not aspects of sub-personal mental processing (though the personal level structure will be mirrored by the structure of that mental processing: here the view makes contact with empirical studies). I think that all aspects of our current mental life are parts of our priority structures. Indeed, I am inclined to think that what makes a state a state of the subject, as opposed to a state of her subsystems, is in part that it is caught by a single net of priority relations. The newspaper is what connects the stories.
With the priority relation, we can unify all the forms of attention. Here are some examples. Visual attention to a word or color consists in prioritizing seeing that word or color over other aspects of your visual state. Intellectual attention to an idea consists of prioritizing thoughts about that idea, relative to other thoughts, but also relative to perception, emotion or feeling. In these cases, attention has an external object. Attending to such objects gets explained in terms of the relative priority of mental states that are about those objects (the intentional directedness of attention gets explained in terms of the intentional directedness of the structure elements). But priority structures contain more information than what you attend to: it is one thing to attend to a person with a positive, loving emotion, and quite another if all your attention is directed at them in spite and hate. Both emotions may have the same intentional objects, but your attention differs depending on which of them is prioritized. What is prioritized may also be the aspects of the mind – experiences, goals and intentions – that are constitutive of an activity, like playing soccer: if they are, you engage in the activity attentively.
Priority structures have many shapes. Some of them are what I call spiky: one element is prioritized over everything else. Think of intense visual focus on a single object or location, adequately described as a spotlight on that thing. But attention can also be more like a lantern[2], dispersed through a whole scene and taking in what comes. Priority structures may be flat with everything in equal priority, like when you are “taking in” the whole surrounding world on a pleasant vacation. The latter is again different from being “on alert”, like in a sketchy neighborhood where your mind is attuned to something that isn’t yet there, the danger that could come from anywhere. Here fear organizes the perceptual world around itself (its thematic field, a notion from the phenomenologist Aaron Gurwitsch, which in Structuring Mind I show how to define – together with terms like the margin or the fringe of consciousness that are other structural “invariants”, as Gurwitsch calls them, of the character of lived experience).
What does it mean for one aspect of the mind to be priority related to another? I have come to think that we should not look for anything like a conceptual analysis or metaphysical reduction. We need to accept that attention is a basic feature of the mind that cannot be identified with anything else. In this respect, it is probably like knowledge or perception. Any attempt to define the relative priority of mental states will either be subject to unwelcome counterexamples, or quickly run into circles. We can, though, describe its functions, and we have a grip on how priority structures shape the field of consciousness. Tomorrow I will write about the function with a post on attention and agency. The day after, I will write about its reflection in conscious experience.
—
[1] Some of the metaphors treat attention as the cause of the structure, rather than as constituted by the structure: but such a view cannot explain the dependency noted above.
[2] Alison Gopnik, in her book The Philosophical Baby, argues that babies operate in lantern mode much more than adults.
Great post! Is priority always contrastive ie one mental state having priority over another. The act of moving from focused attention to diffuse attention doesn’t mean giving anything priority over your original focal target. Instead, it’s more like removing any priority relations at all. And yet it’s still very much an act of attention. Thanks, Tom
I think the priority relation is always comparative. But I allow equal priority as one kind of comparison. I call the primitive notion “weak priority”, in analogy to what in economics and decision theory is called “weak preference”. It can be interpreted roughly as “has at least as much priority as”. Weak priority between two mental states, therefore, allows that the two mental states have the same priority. If one mental state has priority over another, that is strict priority (which you can define by weak priority one way, but not the other). In an extreme case of diffuse attention every mental state has at least as much priority as any other.
Hi Sebastian,
First, a hearty congratulations on what I have come to consider the most important book on attention I’ve read in the past 5 or so years. I’ve assigned it to my post-docs and research programmer and we’re doing a chapter at a time together. Many of the insights central to the book are central to our own computational approach to attention, so I was very happy to see so many familiar ideas while reading.
One clarification I’m after is about “split attention.” More specifically, somewhere in the book you talk about moment-to-moment guidance of low-level attention, which sort of suggested to me that you may be making a distinction between what is instantaneously focal, and what has been very recently focal, with the latter making intelligible the notion of multiple foci or “split attention.” I don’t know if the former or the latter is the right way to read out what you mean by “split attention” or having more than one item as “central” in the priority structure. More clearly, I can’t tell if you take it to be possible to have more that one focus of attention at precisely the same time or not. Speaking purely from a computational perspective, it seems to me to be true that if focus is related to broadcast (in the global workspace sense), then having more than one focus of attention at any individual moment potentially hamstrings attention’s capacity to be an integrator, since each function computed by each active subnetwork of the mind that receives a complex multi-focal broadcast would have to prioritize which part to process. Even if that were possible in principle, there would then be the problem of figuring out how to bind all of the resultant computations together in a sensible way. So, my group has spent considerable effort resisting ideas about “split foci” for reasons like this, but I was unsure how you came down on these kinds of issues, since broadcast isn’t really something that’s treated at great length in the book.
Thanks – that makes sense.
Hi Paul,
great to hear that you are liking the book. I’m very happy that it finds some interest also in your field. I am really curious to, at some point, hear more your own approach and how it interacts with the priority structure framework.
To your specific question. An important idea to get on the table here is that of a priority system: that’s a set of mental elements that are all connected by relations of weak priority (you can walk on a priority path from each element to any other). Genuinely split attention, in my view, now can mean one of two things (there is also a third, dynamic kind, that you mention, where one is quickly switching between different cognitive threads). The first would be several elements that simultaneously are of top priority in a single priority system. These elements are connected to each other only because other elements are deprioritized relative to both of them. I think this happens, for example, in multiple object tracking. In cases like this, I think one shouldn’t expect interference between the top elements, since they are sustained by the same background structure.
The second type of split attention would be cases of several disjoint priority systems. Here the integrating role of attention would be seriously threatened. I talk about this type in Chapter 5 Sec. 2.4 of the book. When people talk about different attention systems for different sensory modalities, different hemispheres, or different tasks, one might interpret them in the disjoint attention systems sense. But I think that this is not the best interpretation of the evidence: the systems usually talk to each other at least in some tasks.
So I think I agree with you: if there were several disjoint subnetworks we’d get exactly the sorts of lack of integration that, in my view, attention is supposed to overcome. I think it’s in principle possible to have organisms or systems built with such disjoint systems, but it would have major disadvantages, and I don’t think there is evidence that it actually happens
Thanks for your blog…it is a crucial topic if we are to make progress.
In my work looking at how consciousness works, I have found that attention is key, but I am wary of just considering it as a prioritisation or filtering mechanism. For me, a key role of attention is to precisely designate, playing a similar role to naming a variable in software. Consciousness then arises when we pay attention not to the external world, but to the content of our own minds.
Thanks for the reply. If you drop me a note at paul.bello@nrl.navy.mil, I’d be happy to share a recent submission of ours that tries to make sense of multiple object tracking without invoking multifocal attention. You’ll also get an inkling of what our system is like. Might be a great way to start a more detailed discussion about parallels because I think there may be too many to cram into a simple email.
Peter: thanks for your comment. I think of the prioritization as quite different from filtering. Filters normally concern only what is let through, and throw away the rest. But for me the background is as important as the foreground. Attention relates those, compares and integrates them. I also think of filtering as a sub-personal information processing mechanism. While my type of priority structures are personal level. They can be implemented also by, for example, various forms of modulations. As for your “naming a variable” idea: while I would need to hear a bit more, I worry that it also neglects the background and concerns only “what is attended” (in this case named or designated, I suppose), while I think we need to think about the relation or comparison between “what is attended” and “what is unattended”
Paul: sounds good. I’ve just sent you an email.
Hi Sebastian,
Thank you for your posts. I was curious if you might say more about why, for you, the ‘background is as important as the foreground’, and perhaps about the wrongness of metaphors like ‘filters’, ‘spotlights’, etc. that neglect the background. Here’s what I think I understand: on your framework, experiences with low priority are not thereby unattended; they are nodes in a priority structure, and so are attended, even if peripherally. But I was curious if you could say just a bit more about the significance of the point. Some of your comments above seem to suggest the answer might have to do with attention’s role in ‘integrating’ and ‘comparing’ information throughout the priority structure. Here, I think your (otherwise really nice) newspaper analogy ceases being helpful. The newspaper orders information by priority, but in what sense is the low priority information (e.g., on the back page) ‘as important’ as the information on high priority information, apart from the fact that it is deemed newsworthy enough to make it in at all (and so still of higher priority than other information)? Is there some distinctive role for low priority experiences in cognition? I find issues in this area the hardest to understand, but also potentially the most interesting part of the view.
(If you discuss this at more length in your book, feel free to refer me to it. I have the book, but I have been jumping around, and may have just forgotten where you go into this.).
Hi Aaron,
thanks a lot for your question! When I wrote, in my reply to Peter, that what is de-prioritized is “as important” as what is prioritized, I meant primarily that it is as important for attention (since the foreground is defined relative to the background), not that it is as important to the person (I think that what is important to the person, in most cases and if things go well, is what is in her foreground). You also asked about the distinctive role of low priority: I think they need not have such a role, but they can have one. Sometimes, for example, the fact that something is de-prioritized relative to something else can be what sustains the priority of the latter, and may color the priority element in a distinctive way (I think that Gurwitsch’s thematic fields have that function, and can be defined in this way. I talk about them in the chapter on Phenomenal Structure).
I would say this way of looking at attention is incorrect. Attention is not a mental process of ideation or based in thought but simply a state of observation of every thought feeling and emotion as and when they arise. Attention is a capacity of the brain that is not based in the thought. In the framework described in this article it separates the thinker and the thought, but this is an illusion. They are one and the same.
Hi Anonymous Thinker: I agree that attention need not be based in thought. Only some forms of attention are, like when you are intensely thinking about some problem. But much of our attention involves perception, feeling and emotion. You write that we should think of attention as a state of observation of thought, feeling and emotion as and when they arise. I completely agree that we often *can* do that. The Roman philosopher Marcus Aurelius for example connects this form of attention to happiness and the good life, when he writes “those who do not observe the movements of their own minds must of necessity be unhappy”. I think that we sometimes say that someone isn’t *really* paying attention, when we think that they are paying attention in the wrong way (like “you aren’t really paying attention!” to someone who isn’t listening to us, but instead listens to someone else). So, insofar as one thinks that what observing one’s own mind, introspective attention – as I would call it, is a good form of attention, one might then think of that as “really paying attention”. But I think that someone who is in the grip of anger and attends to the person she is angry at is still paying attention. Or someone who is paying attention to a test on a computer screen, as they are done in many cognitive psychology experiments. I think we can connect these more simple forms of attention with the one you are interested in.
Also: I don’t see how the framework separates thinker and thought.