Attention Is Cognitive Unison_ An Essay in Philosophical Psychology-Oxford University Press (2011)
Read from April 29th, 2020 to May 4th, 2020
the conversation about the metaphysics of the psychology of attention has been more or less dormant for the century since philosophy and psychology split. This is not because philosophers and psychologists have been ignoring one another. The two disciplines have been talking, but for the last century they have not been talking about attention.
Philosophers, by ignoring the branch of psychology that has attention-related phenomena as its central explananda, have tended to operate with a false picture of the psychologists’ project. According to this false picture psychology gives us a theory in which cognition is thought of as a set of computations, realized somewhere in the frontal lobes, while the perceptual processes and the action-executing processes, located nearer to the back of the head, orchestrate the inputs and outputs for these frontal, cognition-constituting computations.
Psychologists, for their part, have been investigating attention without regard to philosophy, and so have spent the last century blunting ever sharper empirical tools against what is fundamentally a metaphysical problem
This book presents a single unified theory of attention, intended to apply to attention in all its forms.
The theory says that the relationship of attention to the cognitive processes executed in various parts of the brain is analogous to the relationship between unison and the individual performances of the members of an orchestra. Just as there is no place in the orchestra where unison sits, so, according to this theory, there is no place in the brain where attention is located. A similar point applies to particular cognitive activities: Just as there are no particular melodies that all instances of unison involve, so there are no particular patterns of neural activity that we should expect to find in all instances of attention. The fact that a person is paying attention does not, by itself, entail anything specific about which cognitive processes are taking place or where.
If attention is cognitive unison, then facts about specific cognitive processes do not, by themselves, entail facts about attention. Just as the execution of a particular activity on the part of a musician may be unison constituting on one occasion and not on another, depending on how it is related to the things that his fellow musicians are doing, so a cognitive process that plays a role in attention in one context may, in a different context, be instantiated without attention. In this way the cognitive unison theory entails that no particular cognitive activity is sufficient for attention, or even a very reliable indicator of it. The
Metaphysics, as we have it now, gives us tools that can usefully be deployed on foundational issues in psychology. The development of these tools has, perhaps inevitably, upped the entry requirements for the metaphysical debates. Those debates now tend to be incomprehensible to those not familiar with the issues and with the accompanying vocabulary. And by now few psychologists are interested enough in what has been going on to think that the entry price to the metaphysical debates is worth paying. The result is that most psychologists think that, once we have all agree to be physicalists, there are no interesting metaphysical questions about mental entities to be asked. ‘Metaphysical’, they think, is just a show-off’s word for ‘spooky’.
By ‘metaphysical’ all that I mean here is ‘basic’ in the sense of ‘basic even from the point of view of physics’.
attention is ‘the taking possession by the mind, in clear and vivid form, of one out of what seem several simultaneously possible trains of thought’ (James 1890: 381).
His chapter on attention includes, under the title ‘To How Many Things Can We Attend at Once?’, a discussion of experiments into the ability that we now call ‘subitizing’ (that is, the ability to see straightaway whether one, two, or three stimuli are present, without having to go through any procedure of counting them)
Disagreements about which phenomena should be taken as the paradigmatically attention-involving ones were widespread at that time, as they were in the decades before and after.
James’s Principles came at a time that was crucial to psychology’s development as a science, and it was unusually influential. At the time when James was writing there were at least three different ways of approaching the topic of attention, each of which had its own advocates.
There were those who took the most fundamental thing about attention to be its involvement in willed action, and in the coordination of bodily movements. There were those who took the most fundamental thing about attention to be its involvement in perception, and the achievement of ‘sensory clearness’. And there were those who took the most fundamental thing about attention to be its involvement in the direction of thinking, and in the coordination of one’s internal monologue.
When Bradley titled his 1886 essay ‘Is There Any Special Activity of Attention?’ it was in order to urge that the question be answered in the negative.
When it comes to naming attention-constituting processes, Bradley, like James, provides a list of just two: ‘redintegration’ (that is, the following of an association between two ideas) and ‘blending’ (the forming of such an association). He writes that ‘in attention there is either no activity at all beyond the common processes of redintegration and blending, or, if the activity exists, itself is not attention’ (1886: 316).
In taking it that the explanation should proceed by identifying the processes that constitute employment these imagined scientists would be making a metaphysical mistake: a mistake about the sort of thing that employment is.
The most authoritative scientific rejection of that positivistic basis for behaviorism was given by Donald Broadbent in his 1958 book Perception and Communication; a book that also set the agenda for almost all thinking about attention in the decades that followed. It came at a crucial time for psychology’s development. The year before Broadbent’s book had seen the publication of B. F. Skinner’s Verbal Behaviour (1957), in which Skinner made an ambitious attempt to apply a straightforwardly behaviorist explanatory approach to distinctively human aspects of cognition. Thanks to Noam Chomsky’s vituperative review of Skinner’s book, published in 1959, the failure of Skinner’s project was conspicuous and remains famous. That failure made a major contribution to the abandonment of behaviorism and to the corresponding rise of cognitive psychology. But the shift away from behaviorism that followed was not owing to Skinner’s failure alone. It also had much to do with
Théodule Ribot’s 1889 book, Psychologie de l’attention, had offered a strikingly behaviorist approach to attention, which Ribot asserts with characteristic rhetorical flamboyance: Are the movements of the face, the body, and the limbs and the respiratory modifications that accompany attention, simple effects, outward marks, as is usually supposed? Or are they, on the contrary, the necessary conditions, the constituent elements, the indispensable factors of attention? Without hesitation we accept the second thesis. (25)
The Second World War had lent a new urgency to questions about such things as a person’s capacity to pay attention to multiple radar screens for prolonged periods. The research that such questions prompted revealed interesting effects. These cried out for experimental study, and for a theory to explain them. The rehabilitation of attention as a topic for scientific study was called for, but that rehabilitation required the breaking of the public-observability constraints that governed the positivism-influenced varieties of behaviorism that were dominant in psychology at the time. It required the rejection of the behaviorists’ claim that any properly scientific theory should be given using only terms that refer to publicly observable entities.
The most authoritative scientific rejection of that positivistic basis for behaviorism was given by Donald Broadbent in his 1958 book Perception and Communication; a book that also set the agenda for almost all thinking about attention in the decades that followed. It came at a crucial time for psychology’s development.
The year before Broadbent’s book had seen the publication of B. F. Skinner’s Verbal Behaviour (1957), in which Skinner made an ambitious attempt to apply a straightforwardly behaviorist explanatory approach to distinctively human aspects of cognition. Thanks to Noam Chomsky’s vituperative review of Skinner’s book, published in 1959, the failure of Skinner’s project was conspicuous and remains famous.
That failure made a major contribution to the abandonment of behaviorism and to the corresponding rise of cognitive psychology. But the shift away from behaviorism that followed was not owing to Skinner’s failure alone. It also had much to do with the increasingly detailed articulation of a scientifically respectable alternative to the behaviorist paradigm. Chomsky’s own book, Syntactic Structures, published the same year as Verbal Behaviour, and Donald Broadbent’s Perception and Communication, published the year after, together established an alternative to the behaviorist framework. In their different ways, and drawing on different conceptual resources from the then emerging field now known as ‘informatics’, they suggested a framework that would enable psychologists to study cognitive phenomena while avoiding both the methodologically unsatisfactory method of introspection and the explanatorily inadequate method of mentioning nothing but publicly observable behaviors and stimuli.
Broadbent’s distinctive contribution was to show how our understanding of a subject’s internal processing architecture could be disciplined by the then-new strategy of importing into psychology the intellectual resources used in designing information technologies.
Broadbent’s distinctive contribution was to show how our understanding of a subject’s internal processing architecture could be disciplined by the then-new strategy of importing into psychology the intellectual resources used in designing information technologies. Whereas Chomsky employed the theory of computations considered in abstraction from the practicalities of their implementation. Broadbent employed the intellectual resources that are used by communications engineers to negotiate those practicalities.
Information technology was progressing dramatically at the time when Broadbent and Chomsky were writing. It was progressing rapidly, with the invention in September 1958 of the integrated circuit.
But at a much early stage in the discussion, before these broader methodological themes have been broached, he writes: Perhaps the point of permanent value which will remain in psychology if the fashion for communication theory wanes, will be the emphasis on problems of capacity. [. . .] The fact that any given channel has a limit is a matter of central importance to communication engineers, and it is correspondingly forced on the attention of psychologists who use their terms. (1958: 5) This introduction of the notion of capacity limitations into discussions of perception was, as Broadbent predicted, hugely and permanently influential. It was from claims about capacity, thought of in information processing terms, that all theories of attention in the decades following Broadbent would be built.
According to Broadbent’s own capacity-based theory of attention there is a single attentional bottleneck at which capacity limitations are especially pertinent, a bottleneck that arises because the information coming in from the senses is processed by two systems operating in series. The first system, Broadbent thought, has a large capacity for information processing. The second has a much smaller capacity. The bottleneck produced by the connection of the two is the locus of attention in the sense that, although all stimuli are automatically subjected to processing by the first, large-capacity system, the stimuli that make it through the bottleneck into the small-capacity system count, ipso facto, as stimuli to which attention is paid. This two-systems-and-a-bottleneck picture was intended as a communication-theoretic rendering of the everyday idea that simple features of one’s environment, such as the fact that there are people talking in the next room, come to one’s awareness involuntarily, whereas the details of these things, such as the content of the conversation that is taking place, can be detected only for one or two of the things that are going on, with the question of which things have their details detected depending on the focus of one’s attention.
Current philosophers tend to regard this fixation on natural language as a phase that philosophy did well to get out of, and understandably so, but the bad reputation of ordinary language philosophy is in some ways undeserved.
Ryle had treated heed concepts as belonging in the class that he dubbed ‘mongrel categoricals’ (1949: 135–138). By this he meant that statements employing heed concepts not only tell us what is taking place, but also situate that event relative to a set of dispositions and hypotheticals. Ryle provides some examples that make the point clear. ‘That bird is migrating’ is one of his paradigm cases: The description of a bird as migrating has a greater complexity than the description of it as flying in the direction of Africa, but this greater complexity does not consist in its narrating a larger number of incidents.
‘it is migrating’ can serve as an explanation of ‘it is flying south’, whereas ‘he is attending’ cannot explain the fact that he is playing the piano. At most, it explains a different fact: that his piano playing is successful.
In White’s work, the idea that the explanation of attention cannot proceed by identifying which processes take place becomes the claim that ‘attention’ ‘does not name any specific activity; it indicates the circumstances in which the activity occurs and thus signifies what, on this occasion, it amounts to or is a form of’ (1964: 6).
Simply to say that someone is attending, or paying attention, gives us no more clue as to what activities he is engaged in than simply to say that he is practising. What ‘attending’ tells us is that his activities and energies, whatever they are, are directed to and focused on something which occupies him. (1964:
In Bradley’s treatment this is because claims about attention tell us, not about which processes are taking place, but about their immediate causal context: The attentionconstituting processes are those that are ‘prompted by interest’ and that ‘bring about engrossment’ (1886: 316). In Ryle’s treatment it is because claims about attention tell us something about processes and also tell us something hypothetical: Intellectual processes constitute attention when they occur in a context such that similar processes would have taken place in a range of counterfactual situations (1949: 135–136). In White’s treatment, the idea is that claims about attention do not tell us which processes are taking place. Instead they tell us how those processes relate to the agent’s other activities and to his goals.
Our theory of attention should put us in a position from which we can answer both of the following questions: 1. What is attention? 2. What is it for something to be done attentively?
Phenomena like combustion I will call ‘process-first’ phenomena. Phenomena like haste I will call ‘adverbial phenomena’.
There are various different taxonomic principles that might be used to carve up the set of all objects into categories. Some of the resulting taxonomies will be useful for various purposes. Some will be completely useless. Among the useful taxonomies are those that group objects on the basis of the substance from which they are made. Since there is more than one way to individuate substances, there is more than one way to give a substance-based taxonomy of objects. One substance-based taxonomy groups all plastic things into one set, all wooden things into another, and so on. A different taxonomy might divide up the plastic things more finely, grouping all the PVC into one set, all the acrylic things into another, and so on. There are lots of ways that a taxonomy of objects might be given. Suppose now that we are given the set of objects and asked to find a way of specifying the set of all paperweights. If we are allowed to use only the classificatory resources of a substance-based taxonomy then we will not be able to do it. The problem we face is not that paperweights are ‘multiply realizable’, so that there are many substances from which a paperweight could be made. It is true that paperweights can be realized in multiple substances, but it is not this that stops us from delineating the paperweights with the resources of a substance taxonomy: The problem of multiple realizability could be solved simply by specifying a union of several sets. What stops us from being able to delineate the paperweights using the distinctions provided by a substance taxonomy is that being made of the same substance as a paperweight does not suffice for being a paperweight. That makes it impossible to rule in all of the paperweights without also ruling in some of the nonpaperweights.
One sort of taxonomy, useful for several purposes, is the taxonomy that groups events according to the processes that they instantiate.
The set of instances of an adverbial phenomenon cuts any taxonomy by process (as reflected by the fact that the things that can be done hastily can also be done nonhastily). The set of instances of a process-first phenomenon cuts any taxonomy by manner of performance (as reflected by the fact that imitations of the manner in which combustion occurs need not themselves be instances of combustion). The distinction that we are interested in can therefore be made precise if we can find a way to distinguish between taxonomies by process and taxonomies by manner of performance. And now our philosophical task is relatively straightforward: A distinction between these different sorts of taxonomic principles, at least in the case of taxonomies of events, is easy to draw. The
taxonomies that classify events on the basis of process and taxonomies that classify events on the basis of manner of occurrence.
A taxonomy is a taxonomy on the basis of process if the taxonomy classifies events on the basis of the having or gaining of a property by an object. A taxonomy is a taxonomy on the basis of manner if the taxonomy classifies events on the basis of the having or gaining of a property by an event.
the psychology of memory. The processes responsible for short-term memory are probably very different, cognitively speaking, from the processes responsible for long-term memory; and the processes responsible for episodic memory are probably very different, cognitively speaking, from those responsible for semantic memory, and so on (Baddeley 1990, chapter 3; Schacter, Wagner, and Buckner 2000). It follows from this that memory is not a natural kind of cognitive process. But that fact does not present a problem for the attempt to explain memory via the identification of cognitive processes. We can explain memory perfectly well by the piecemeal identification of various processes corresponding to its various instances. Adverbial
There is an excellent and long-standing body of empirical work suggesting that attention is involved in the performance of tasks that require people to search as quickly as possible through a crowded visual display in order to find a target that is defined by certain combinations of features.
There are [. . .] some explanatory consequences of the hypothesis that there is a single general kind of attention of which perceptual and sensational attention, and conscious thought and imagination are all subspecies. It is a familiar truth about attention that any one of these kinds of attention can interrupt any one of the others. Perceptual attention can be interrupted by conscious thought; conscious thought can be interrupted by external events which capture the thinker’s attention; either of these two subspecies of the occupation of attention can be interrupted by imagination; and so on. What we have here is not merely some family resemblance between varieties of conscious states, 42 Attention Is Cognitive Unison but apparently some form of competition for the exclusive use of a single limited faculty of attention. The familiar facts about attention are explained if there is a single, suitable high-level resource, drawn upon either by perception, conscious thought or imagination, a resource with access to some of its own recent states and to memory representations generated by its own previous states. I do not say that it is absolutely impossible to explain any one of these facts in some other way. I conjecture, though, that other explanations will be ad hoc. (66–67)
Christopher Peacocke, in his 1998 discussion of attention and self-knowledge, identifies a feature that common sense attributes to attention and that may require the truth of something like the process-first account if it is to be explained in a parsimonious way.
Allport, Antonis, and Reynolds’s 1972 ‘disproof of the single channel hypothesis’ for attention showed, rather surprisingly, that the performance of an attentiondemanding task, such as the auditory shadowing of speech,2 was not at all detrimental to the concurrent performance of certain other attentiondemanding tasks, such as the sight-reading of piano music. The advocate of the process-first account, if he is to maintain his favored explanation of interference, must now maintain that the resources that remain available after the resources required by shadowing have been used up are sufficient for sight-reading. If we call the resources required by auditory shadowing ‘AS’, then the process-first account is now committed to the claim that AS + SR ≤ T.
If we consider tired or distracted thinkers we can find cognitive activities that are prompted by interest but that are not attentive. We can also find activities that are not prompted by interest but that are attentive. On a hot and muggy night the sleepless traveler finds that his attention keeps turning listlessly to the movements and sounds made by the fly that is circling around his ceiling.
In Ryle’s words: To describe someone as now doing something with some degree of some sort of heed is to say [...] that he is actually meeting a concrete call and so meeting it that he would have met, or will meet, some of whatever other calls of that range might have cropped up, or may crop up. (1949: 141) White’s view contrasts with Ryle’s in being less concerned with what the thinker might have done had things been different and more concerned with what the thinker actually did. Instead of claiming that an agent’s attending to something is a matter of how his activities relate to what he was disposed to do, White claimed that an agent’s attending is a matter of how his actual activities relate to the various other things that he is currently doing: Simply to say that someone is attending, or paying attention, gives us no more clue as to what activities he is engaged in than simply to say that he is practising. What ‘attending’ tells us is that his activities and energies, whatever they are, are directed to and focused on something which occupies him. (1964: 7) When we met these views at the end of chapter 1, it was mentioned that White has an explanation-based argument against Ryle’s disposition-based approach.
In general, the attention with which someone does something cannot explain his doing of that very thing.
White himself states his positive claims quite concisely: Degrees of attention are not to be explained as more or less intense engagement in one specific activity, but rather as concentrating more or fewer of our activities on the one object. Starving travellers are said to talk food, think food, live food.
Full attention to X consists not only in the range of activities that are centred on X but in the absence of activities concerned with things other than X. Hence, the common privative force of the notion of attention; one minds one’s own business by not minding that of other people. (1964:
Let α be an agent, let τ be some task that the agent is performing, and call the set of cognitive resources that α can, with understanding, bring to bear in the service of τ, τ’s ‘background set’. α ’s performance of τ displays cognitive unison if and only if the resources in τ’s background set are not occupied with activity that does not serve
A subject’s ‘tasks’, as these are to be understood here, are the things that the subject is in the business of doing and that she is active with. To specify the tasks in which an agent is engaged, we adopt the agent’s point of view on her own activities. Normal human tasks are such things as making a cup of tea, following a conversation, or looking for the car keys. They are activities with natural descriptions of a sort that the subject would typically accept as a description of her goal. Not
But an agent must have some understanding of what it is that she is about. What is required, in order for τ to count as a task of α’s, is for it to be the case that α has some understanding of τ on the basis of which her performance of τ is guided.
What about the buzzing fly?
it is guidance by understanding that gives the cognitive unison theory its official definition of ‘task’: A task is defined as an activity of the agent’s the execution of which is under the guidance of the agent’s understanding of that activity. An activity fails to be a task if it is not agent involving, if the agent has no understanding of that activity, or if the agent has an understanding but his understanding does not guide the way in which the activity is performed.
It counts looking at a chair as a task of the agent’s (because the way in which the episode of an agent’s looking develops is guided by the agent’s understanding of looking).
Seeing a chair typically is not a task of the agent’s, because episodes of mere chair seeing typically proceed without the agent deploying his understanding of seeing. Tripping over a chair typically is not a task either.
One’s attention can be caught involuntarily by a sudden sound at the window. If the way in which that episode of looking to the window unfolds is guided by one’s understanding of what it is that one is doing in looking to the window, then the looking is a task, although not a task that one performs voluntarily. Because it is a task the looking is something that can be done attentively. If, on the other hand, one is distracted by a sound at the window without the processing of that sound being integrated into any understanding-guided activity, then the case is one where the sound is merely distracting and is not itself attended.
It is important that the notion of a task be relatively undemanding, since it must be undemanding enough to allow infants and animals, in so far as we think of these as paying attention, to have tasks.
Imagine that we have a pair of thinkers, each of whom is asked to prove Pythagoras’s theorem, and each of whom, in response to this request, utters just the same sentences and produces just the same diagram. Imagine, also, that these sentences and diagrams do indeed represent a proof of the theorem. From the description of the case given so far, we cannot tell whether either, or neither, or both of these thinkers understand the theorem or its proof. The two thinkers may produce matching responses, but it may be that one of the thinkers understands the theorem while the other performs parrot fashion, or just strikes lucky in the course of bluffing. Since there is no limit to the sorts of things that might be learned by rote, and no limit on the luck that a bluffer might happen to have, there are no particular responses for which we can say that the thinker with understanding will produce those responses while the thinker without understanding will fail to produce them. What the thinker who lacks understanding lacks is not a cognitive state that enables him to make an appropriate response (he may have such a state if he has been appropriately drilled). What he lacks is a capacity to redeploy the cognitive basis of his performance in a way that enables him to deal with alternatives.2 If the thinker really does not understand at all, then the cognitive basis of his performance will not equip him to deal with alternatives at all, not even with trivial variations, although some other cognitive state, imparted by some other piece of drilling, might. It may be that if we say ‘prove the Pythagorean theorem’, then the thinker without understanding knows, because he has been drilled, which moves to make, but, because these moves are not founded in any kind of understanding, he may be stumped if we say ‘show that the square of the hypotenuse is equal to the sum of the squares on the other two sides’. A thinker with a little more understanding might be able to deploy that same cognitive basis so as to respond appropriately to the alternative formulation of the question, but this thinker may still be stumped if we ask him to how to calculate the distance between any pair of points on a plane, given their Cartesian coordinates. A different thinker might have been able to answer that question purely on the basis of her understanding of the theorem. But, in a setting where the theorem is normally put to some quite different use than the calculation of distances between points, even this last thinker might count as failing to really understand the theorem. All thinkers who understand Pythagoras’s theorem will, on account of their understanding, be able to deal with some alternatives. Some of those thinkers will be able to deal with lots of alternatives. Others will be able to deal with only a few.
in Wittgensteinian terms: Understanding is that which enables one to deal with novel situations, but the particular range of novel situations with which one is equipped to deal will depend on the sort of interaction that one typically has with world. Whether the cognitive state responsible for one’s going on is a state that enables one’s going on to be a going on with understanding depends on one’s form of life. Questions about whether some intellectual capacity on the part of an agent counts as an understanding therefore need to be relativized to a specific range of novelty appropriate to the agent’s form of life.
A cognitive process, in this sense, is a process that operates on representations that encode their contents for the agent of the task: In order for a process to count as cognitive, there must be an agent-level contentful state whose content is directly determined, at least in part, by the content of the representations on which that process operates. More precisely, a process is a cognitive process if and only if it is an immediate consequence of the fact that this process is operating on a representation with the content . . . ξ . . . that the agent is in a position to believe that . . . ξ . . . , or to wonder whether . . . ξ . . . , or to remember . . . ξ . . . , or see that . . . ξ. . . , or to have some other attitude toward the content of the representation in question.
The set of resources that must operate without processing distracters in order for the subject’s performance to be counted as an instance of cognitive unison is the set of resources that potentially serve the attended task.
The relationship between the subject’s understanding of a task and the background set for that task is worth emphasizing because it is this feature of the cognitive unison view that gives the theory its account of why some tasks are such that, as one’s understanding of them grows, one can complete them without attention. It is also crucial to the theory’s account of why it is typically harder to give sustained attention to familiar, well-understood tasks than it is to give attention to tasks that are poorly understood. The thought in both cases is that giving one’s attention to a well-understood task involves marshaling a large set of resources, just because the task is so well understood, whereas a part of what it is for a poorly understood task to be poorly understood is that the set of resources that one can bring to bear is small and therefore brings itself into unison relatively easily.
Complex tasks create problem cases for the theory that equates attention with simple cognitive unison. They are problematic when the agent performing them has what we might call ‘superfluous competence’. 4.9.1 Superfluous Competence
It arises whenever a thinker’s understanding of a task equips him to perform that task in two ways that differ in the cognitive resources that potentially serve them.
an agent counts as performing a superordinate task attentively if he is performing a subordinate task attentively, as part of a strategy for performing the superordinate task.
Even if one does not want to revert to a process-first theory, in which attention is identified with storage in working memory, or with the operation of any other particular processes, one might still want one’s theory of attention to pick out those processes as making some special contribution to attention. The cognitive unison view does not do any such picking out by itself, but the view can nonetheless accommodate the fact that those processes do make a special contribution to attention by showing how that special status can be derived from contingent features of our psychology.
The resources in question might have a special role because they are resources that are heavily in demand in the performance of the sorts of tasks that creatures like us are typically engaged by. If so then those resources will be ones that occur in lots of background sets, and so the facts about which tasks engage those resources will have a special role in determining which of our tasks can be performed with unison. A resource might also come to have a special role on account of being particularly good at marshaling the support of other resources to the task that it serves. If so then it will be a resource that has a special role in bringing about unison and in maintaining
Working memory might also have a special role in human attention for the second reason. It might, that is, be unusually influential in determining the activities by which other cognitive resources are engaged, so that, once a task is being served by working memory, other resources are more likely to be recruited to the service of that task, with the result that a task that is served by working memory is more likely to be performed by resources that operate in unison. Nilli Lavie and her collaborators, although operating with a somewhat different picture of attention from that which I have been advocating here (Lavie, 1995), have observed a number of empirical effects that suggest working memory does indeed play such a role. Lavie finds that the attention of subjects in her experiments is much more likely to remain fixed on a task if their working memory is available to serve that task, even when the task itself makes relatively few memory-related demands. When working memory is taken up with something irrelevant, attention becomes much easier for irrelevant stimuli to capture (Lavie and De Fockert 2005).
Something similar may perhaps be true of the resources that are involved in visual focalization and saccade targeting—processes whose connection to attention has been established in a series of experiments by Tirin Moore and his collaborators (see, especially, Moore and Armstrong 2003). Given the importance of vision to the cognitive lives of creatures like us, it seems likely that detailed visual processing of the stimuli that are relevant to a task might be a good recruiter of other resources to the processing of those stimuli, in a way that the processing of the locations of auditory stimuli is not.
The Privative Character of Cognitive Unison The condition that the cognitive unison theorist claims must hold over the background set if the subject is to count as attentive is a negative one: It is the absence of any irrelevant processing. An alternative approach would be to offer a more positive portrayal of attention, as requiring that some quantity or proportion of the background set be actively employed on the task in hand. It is not obvious how we should choose between a privative view and these more positive, threshold-setting alternatives.
We do not think of children as great payers of attention, but this is not because they fail to attend. It is because their attention is easily caught and so is highly vulnerable to distraction. The state of attentiveness is one that children struggle to remain in, not one that they struggle to attain.
In general, the capacity for inattentive engagement with a task is a sophisticated achievement. We see it when adults perform tasks that they have mastered. We see it hardly at all in children and animals, and only rarely when adults attempt tasks with which they are unfamiliar.
Because the facts about which objects one is paying attention to are grounded in facts about which tasks one is performing, one cannot simply comply with an instruction to pay attention to this chair without taking on some task in which the chair has a role. The task need not involve active manipulation of the chair, of course. It may be a simple perceptual task, such as getting visually acquainted with the chair, or detecting any changes that the chair might undergo. In this way the cognitive unison view collapses a distinction that Alan White, from whose suggestions the view started, was careful to draw. This is the distinction between what White called ‘agent attention’ and what he called ‘spectator attention’.
The subject can simultaneously meet both sets of requirements because, since the background sets of these tasks do not intersect, there is no conflict between the requirements of unison for each of them. In this way the cognitive unison theory
Tasks with intersecting background sets can therefore be simultaneously given attention if the resources in the intersection are quiescent. If the resources in the intersection are recruited to either task (as they might be if a remarkable event occurs in one or other of the stimulus streams), then, in all but a very few cases, the background set of the other task will no longer be operating in unison. When that happens the other task will cease to be attentively performed. It is for this reason that divided attention between such tasks is possible but not entirely stable.
tasks are simultaneously given full attention. It must instead say that for these instances of divided attention, although both tasks may be performed at the same time, they are not both attended at the same time. Attention is paid now to one, now to the other.
agree with the phenomenology of dividing attention
But what process controls cokngext switchijng?
The attention given to a task with a relatively large background set (such as the task of standing perfectly still) can be said to be a greater degree of attention than the attention given to a task with a relatively small background set (such as the task of blinking at regular intervals). The attention given to a task whose performance demands activity from a great many of the resources in its background set (such as the watching of a gripping scene) can be said to be a greater degree of attention than the attention given to the task when some of the resources in the background set are quiet (as in our earlier example where one gives attention to the scenery through which one’s train is passing).
Even when all the resources in a background set are working in the service of a task, they may not be working very hard.
Finally, it may be that, as we see in the case of children, a subject pays close attention to one thing but is nonetheless highly susceptible to distraction from it. In such a case, we may say that the attention paid is of a lesser degree than the attention that is paid by the subject who resists distraction.
the question of how to avoid causation-based objections came to be one of the main topics of discussion in the philosophy of mind
A whole distributed sprawl of events converged upon and contributed to the match’s lighting. It is a somewhat arbitrary and somewhat pragmatic matter how we choose to partition this sprawl of influences into figure (the ‘cause’) and ground—(the ‘background conditions’).
We cannot move to ‘the striking of the match by a man in a three-piece caused the match to light’, because a partitioning that includes the man and his suit needlessly includes background conditions that are unnecessary for the production of the lighting.
With this in mind, we can see that ‘the underlying processes caused the behavior’ can be shown to be compatible with ‘the cognitive unison caused the behavior’ as long as (1) the unison-citing claim differs from the process-citing claim only in the generosity of its figure/ground partitioning, and (2) the added generosity of the unison-citing claim avoids letting in background conditions that are unnecessary for the production of the attention-relevant behaviors.
In general what one needs to do, in order to bring it about that there are such things as swerves in response to such things as unexpected hazards, is to bring it about that the cognitive resources that drivers can bring to bear in the service of their driving are not occupied elsewhere. Ensuring that there is unison is a more reliable strategy for bringing it about that hazards are avoided than is any strategy of ensuring that any particular hazard-detection or swervecoordinating resources are active. For the purposes of explaining hazard avoidance, the unison-citing explanation therefore has a generality that the process-citing explanation does not enjoy.
The cognitive unison theory has negative consequences for psychology only because it places a limit on the sort of explanatory work that can be accomplished by psychological research that is concerned with identifying and describing the individual attention-constituting processes. Taking account of this limit requires us to make some explanatory reconstruals of certain psychological research projects. It is the job of this chapter to negotiate these reconstruals, and to note their methodological consequences.
Since it is when we attempt to put process-identifying research to explanatory work that metaphysical commitments are incurred, the psychologist who remains anxious to preserve the Wundtian ideal of a metaphysics-free methodology might attempt to disavow explanatory aspirations, claiming instead that accurate description is her research’s only goal. Such a claim could not be maintained for long. Psychology depends at every stage on nondeductive modes of inference. Such inferences require that claims with explanatory import are at stake. A claim with no explanatory aspirations cannot be supported by an inference to the best explanation. Nor, plausibly, can enumerative induction be a source of reasons to believe a theory unless something is assumed about the explanation of the class of phenomena over which the induction generalizes (Harman 1965). If the desire to be free of metaphysics led the psychologist to abandon explanation altogether it would leave her with a methodology that was inferentially constrained to the point of fruitlessness.
We can bring to light these metaphysical differences by using a particular case to illustrate the different explanatory construals that different theorists place on research that identifies an attention constituting process. Elliot Freeman’s work on the role in attention of ‘integrative horizontal connections’ in early visual cortex provides a clear example. Freeman’s research shows, using a combination of behavioral and electroencephalographic studies, that these early integrative processing systems play a role in implementing certain instances of attention that determine the way in which patterns of collinearity are used when parsing the basic contours of a visual scene (Freeman, Sagi, and Driver 2001; Freeman 2005; Khoe, Freeman, Woldorff and Mangun. 2006). How is such a finding to be taken as a contribution to the explanation of attention? One construal of Freeman’s discovery is as the identification of one particular species in the overarching genus of attention processes. Another is as the discovery of a particular effect that the central attention mechanisms can produce. A third construal takes Freeman to have identified a new member of the loosely united family of attention processes. A fourth takes the same discovery to be a contribution to the more or less heterogeneous scientific theory with which our mistaken commonsense talk about ‘attention’ should be replaced. A fifth takes the discovered process to be a new component of attention—a piece of the jigsaw that will give us a picture of attention itself only when all the pieces are viewed together (and viewed in the right configuration). A sixth construal takes Freeman’s discovery to be the discovery that the attentiondefining principles—which might be taken to be functional principles, or might be taken to be more abstract mathematical principles—are instantiated at a particular locus, early in the visual processing stream. This list of possible construals is by no means exhaustive, but it is long enough for us to see the point: Each of these different construals of the explanatory import of Freeman’s identification of an attentionconstituting process takes a slightly different view of the metaphysical relationship between attention and its underlying processes. Each therefore has a slightly different understanding of attention’s metaphysical status. Some of these construals will seem natural to some psychologists, and others will seem natural to others. A psychologist like John Duncan, who advocates a family resemblance view of the processes that constitute attention, will favor the interpretation of Freeman’s results as the identification of one member in an open-ended, potentially quite heterogeneous family (Duncan 2006). Psychologists like Claus Bundesen and Thomas Habekost, who advocate the idea that there is a distinctive set of mathematical principles governing various different instances of attention at various different levels of neural organization, will favor an interpretation of Freeman’s results as the discovery of a new instance in which these principles operate (Bundesen and Habekost 2008). A psychologist who thinks, as John K. Tsotsos, Laurent Itti, and Geraint Rees (2005: xxx) do, that there is a particular functional role that all instances of attention play will think that Freeman has described one of the processes that occupy that functional role. A psychologist who thinks, as Eric Knudsen (2007) does, that attention is a complex process built out of several simpler components will favor an interpretation of Freeman’s work as giving us an account of one of these components. And so on.
The first feature, on which philosophers have traditionally focused, is multiple realizability: Instances of employment can be constituted by manual labor, vigilant observation, or administrative organization; instances of haste can be constituted by tea makings, breakfasts, marriages,
Reynolds and Duncan, together with Bob Desimone have been responsible for developing this biased competition model of attention (Desimone and Duncan 1995; Reynolds and Desimone 2000). Reynolds and Desimone summarize the central claims of the model like this: Multiple stimuli appearing together activate populations of neurons that compete with one another, possibly through mutual inhibition. When attention is directed to one of the stimuli, this causes an attentional feedback signal to be directed to the neural population activated by the attended stimulus. This feedback biases the competition in favor of the attended stimulus, enabling it to propagate its signal forward to the next cortical area. (Reynolds and Desimone 2000: 234)
As with the selectivity of unison, the selectivity of a competition is not owing to limitations in the competing processes: There is no need to postulate any limitations in the number or in the ability of the competitors in order to explain the fact that a competition has just one winner. An architecture in which different processes compete for the control of processing resources at various levels throughout the nervous system is precisely the sort of architecture in which states of unison could emerge, and could do so without there being any one attention-explaining process to bring these states about, or any one bottleneck in capacity through which all and only the attended things have passed.
The cognitive unison theory’s construal of research within the biased competition school is therefore a positive one. The theory does not give the biased competition school exclusive rights to the modeling of attention. There may be several different routes by which states of cognitive unison can be brought about and maintained. But the research from the biased competition school has, at least, given us a detailed account of the anatomy and mechanism of one of these routes, and has provided an abstract model that explains and predicts many of the attention-related effects that are revealed by neural and behavioral observations of humans and nonhuman primates. If the cognitive unison theory gives us the correct account of what attention is, then the biased competition model may give us the correct account of how many of the instances of this attention-realizing unison come about, and of how they get maintained.
Competition (especially when accompanied by a biasing feedback signal) is a way to bring about unison across a hierarchically organized network competing processes. It is these processes, when operating in unison, that constitute attention, not only the top-down processes that generate the signal by which they are biased.
The two deepest mysteries of the mind are the mystery of consciousness and the mystery of content. The first of these mysteries—the so-called ‘hard problem’—arises from our failure to see how to give an intellectually satisfactory account of the fact that the mental states of creatures like us are experienced by their subjects. The second mystery—‘Brentano’s problem’—arises from our failure to see how to produce an intellectually satisfactory answer to the question of how any naturally arising state could have a property that our beliefs, desires, hopes, and inquiries do have: the property of being essentially about actual or possible things, propositions, or states of affairs, rather than merely encoding information derived from those things, propositions, or states of affairs.
Many psychologists and some philosophers think that the theory of attention will contribute to solving the first of these mysteries. A smaller group, this one composed mostly of philosophers, thinks that the theory of attention will contribute to solving the second.
My own view is that the two mysteries are indeed intimately connected but that the theory of attention enables us to solve neither of them.
The mystery originates from the fact that mental states, at least in cardinal cases, are states that have contents, and that have these contents essentially. Consider, as an example, my belief that Belfast is farther east than Dublin. That, obviously enough, is a mental state that has as its content the proposition that Belfast is farther east than Dublin. The content is essential to the belief in the sense that, if the belief had not had that content, it would have been a different belief, and if it had not had any content, it would not have been a belief at all.
The cognitive unison theory can allow for attention without consciousness in cases of collective agency
In developmental psychology, and in some philosophical works prompted by it, a number of writers have converged on the idea that a theory of joint attention should contribute to removing the mysteries that surround our capacity for knowledge of other minds (Moore and Dunham 1995; Eilan, Hoerl, McCormack and Roessler 2005).
Austin, James, H. (2009), Selfless Insight: Zen and the Meditative Transformations of Consciousness (Cambridge, MA: MIT Press).