On Automaticity and Control

[Update: given really helpful discussion below, I’ve put a summary of some of the extended comments below]

I want to begin with a distinction that I think is fundamental to our thinking about the mind and yet is not adequately characterized by philosophers or psychologists (the latter will beg to differ). This is the distinction between automaticity and control. These concepts are like that of attention: they identify fundamental features of mind, seem to be exemplified in simple phenomena that we grasp intuitively, and yet are invoked theoretically without adequate analysis. We’ll return to attention in detail in a later post, but I believe you need to understand control and automaticity before you can fully understand attention. The ideas of automaticity and control are also central to our understanding of agency as well. After all, action is an experssion of the agent’s control and yet actions can also be done automatically.

There are related terms that creep up in the vicinity, some mundane, some quasi technical. These include: mere happenings, involuntary, intentional, stimulus driven, goal-directed, voluntary and so on. I propose to ignore these concepts in favor of speaking just of automaticity and control, although I do not claim that the latter pair covers everything invoked in the prior list. How then should we characterize automaticity and control?

Let me mention how these ideas figure in our thinking about the mind by noting a difference between actively looking for something and something’s popping out to you. You look for a friend in a crowd, but unfortunately, she is wearing clothing of a similar nondescript color as others. She’s effectively camouflaged. So, you visually scan, trying to find her. This is effortful and something that involves your control: how you scan seems to be deliberate, voluntary (you see me invoking the related concepts here…no analysis yet!). If your friend were instead wearing a bright, neon pink jacket, you would locate her easily. Here, she might visually pop-out, grabbing your attention automatically we might say. You can’t help but notice her.

Hopefully, that example is enough to acknowledge that in a simple mundane task (visual search), we can divide two types of phenomena in terms of automaticity and control.

What have psychologists to say about this? An important discussion is that of Walt Schneider and Richard Shiffrin (Walt is a member of the CNBC and in psychology at Pitt). In a seminal 1977 paper in Psychological Review (“Controlled and Automatic Human Information Processing: 1. Detection, Search, and Attention”), they provide definitions of controlled and automatic processes. On automatic processes, they suggest:

 An automatic process can be defined…as the activation of a sequence of nodes with the following properties: (a) The sequence of nodes (nearly) always becomes active in response to a particular input configuration, where the inputs may be externally or internally generated and include the general situational context. (b) The sequence is activated automatically without the necessity of active control or attention by the subject (2).

 They then define control as follows: “[a] controlled process is a temporary sequence of nodes activated under control of, and through attention by, the subject” (2). As this definition will not be adequate for philosophical purposes, we’ll need to look elsewhere.

John Bargh has written quite a bit on automaticity and the notion is quite prevalent in certain areas of social psychology. Gordon Logan has a detailed theory of automaticity as well, but let’s focus on Schneider and Shiffrin’s ideas. The core idea seems to be this: we define automaticity in terms of control and then control in terms of attention. Since I wanted to use automaticity and control to understand attention (recall our visual search case), this analysis doesn’t help me. Moreover, I think automaticity and control have a more essential connection to agency. That’s the starting point of how I think about the notions.

An important step is to realize that we shouldn’t talk about a process as being automatic or controlled in itself. Rather, a process can be said to be automatic or controlled in respect of the features that it exemplifies. Thus, we relativize application of the concepts in respect of features: a process can be controlled in respect of feature F while automatic in respect of feature G. This is a proposal about how we should regiment certain theoretical terms in cognitive science. The payoff is that it allows for characterizations of process that plausibly exemplify both automaticity and control. Agency is one such type of process: many things that we do have automatic and controlled features, say reaching for that specific cup. Where I intentionally reach for that cup, I exert a type of agentive control, but many aspects of my reaching are (thankfully!) automatic: the kinematics of the reach, the precise grip I form, the sequence of joint rotations, etc.

The following definitions then are roughly stated but for our purposes, suffice to tie the ideas to agency in light of the relativization just mentioned (a finessed definition can be found in the paper linked below). I am sure you will come up with compelling counterexamples, but let’s start with the proposals, and perhaps discuss them in comments. These definitions concern processes in which the agent is in some sense a participant (they are attributed to the agent). They can be adapted to talk about control and automaticity in brain regions (exercise for later). Thus:

 

Process P is controlled in respect of its feature F iff P’s having/being F is a result of the agent’s intending that there be a P having F

Process P is automatic in respect of its feature F iff P’s having F is not controlled in respect of F.

 

A fully automatic process is one that does not exemplify any controlled features.

Going back to our reaching case, the behavior of reaching for that cup, an action process, is controlled in respect of its exemplifying that action property (namely being a reaching for that cup) precisely because the agent intended to reach for that cup. Of course, we need the standard qualifiers for causal deviance, temporal variables etc., but let’s assume the necessary qualifiers for sake of argument. Other aspects of the action can be automatic to the extent they are not intended (controlled). So, the process involving a precise rotation of the hand, a precise grip, a precise kinematics is automatic.

The proposals exemplify two virtues: (a) they allow that control and automaticity are defined in opposition to each other (automaticity is the absence of control) and (b) they allow us to attribute both to the same process.

I think once we have an understanding of these ideas, we can actually make some headway on a variety of issues including agency, schizophrenia, and attention. I’ll discuss each of those in later posts. You might have your own phenomena that you think the notions should apply to so I would be pleased to hear about whether the basic ideas might apply to your domain.

Undoubtedly, philosophy being what it is, I’m sure there are refinements. It would be enough for me, however, that we consider automaticity and control as a set of concepts that are worthy of philosophical analysis.

A couple of papers: For some philosophical applications of these ideas, I am quite taken with Galen Strawson’s discussion of mental action (even if I disagree with him!). My response deploying the notions noted above is published in this volume and a penultimate version can be downloaded here.

 

Update: I’ve benefited from all the discussion, but I want to highlight one specific thread about control raised by Joshua Shepherd, Ellen Fridland and Wayne Christensen. I hope this will summarize one of the many interesting points readers have raised. You can read through the comments if you are interested, but to roughly summarize:

What has become apparent is that there are many things that we might call control. Josh brings up the notion of skilled behavior as an exemplification of a high level of control, and we can take the elite pianist as an example. Ellen noted different notions of control (Josh as well), and Wayne Christensen in one of his posts emphasizes the dynamic connections between these. There’s lots here, but I think it raises at least one challenge which I put as follows:

Take the elite pianist. Note that there are different forms of control that are being implemented, different things that are aptly called control, and perhaps with each, some relevant correlated sense of automaticity. Provide adequate analyses of these that is sensitive both to conceptual conditions and to known empirical phenomena. The goal here is to provide an adequate theoretical characterization that is then generalizable to other forms of skilled and unskilled behavior, that helps us make sense of the acquisition and loss of skills, of performance etc. Further, see if it can be generalized to philosophical questions about volition and agency.

 

49 Comments

  1. Wayne Wu

    Let me add how you might modify the definitions to use the notions of automaticity and control in cognitive science (here, we can think of top-down versus bottom-up). Basically, choose a control structure, use some analog of intention, say a representation of a goal, and then define control (top-down) in terms of that control structure using the previous definitions as as model. I’ll come back to an example of this in considering top-down versus bottom-up attention when I discuss attention.

  2. Hi Wayne,

    Really glad to see you raising these issues here. As I think you know, I definitely agree that automaticity and control deserve more philosophical attention, and I’ve learned a lot from your recent papers on these (and related) topics.

    A few initial thoughts (I might be slow on follow-up comments since I’m moving and things are a bit upside down).

    1] I find the distinction between automaticity and control interesting, but less than mandatory. So I wonder what you think about examples of behavior that seem highly controlled in part because important aspects of the behavior are automatized (skilled behavior, basically).

    2] I wonder how exactly you cash out the ‘in result of’ when you tie a controlled feature of a process to an agent’s intention. How fine-grained are the content of intentions, I wonder (not just a question for you, of course)? Do motor schemata and the way an agent sequences motor schemata when executing an intention (e.g., an intention to swing a golf club) count as a part of what the agent intends?

    A lot more to say, but perhaps that’s enough for now. Cheers.

  3. Wayne Wu

    Hi Josh:

    Thanks for these thoughts. (1) is an interesting case. I’m going to borrow a line from the paper I link to above which I owe to my last piano teacher, the gist of which is that sometimes being skilled is precisely to maximize automization. So, maybe being skilled supervenes on the right sort of mix of control and automaticity as I’ve defined. This allows for the “skilled because of automization” point you make.

    My teacher’s point was that one practices so that one can play as one wants. Part of the point, I think, is that much of what a musician does is automatic, how one strikes a note, the sequence of notes. If I aim to play a piece beautifully, there are things I don’t want to have to think about, if only getting my fingers to play the right sequence. Its the nuances that I want to control, and every time I play the piece, this might be slightly different, say emphasizing one voice over another, say in a multi voice Bach fugue. I’m not that good, but pianists who are genuinely skilled exert skill by having this sort of control overlayed on lots of automaticity.

    It is a good question how to understand skills but I’m inclined to see it as a more complex property, of which automatization and control (as defined) are part. I know you have more to say about that (Ellen Friedland has also worked on this quite a bit…perhaps I can coax her to join the conversation).

    (2) is a good question and potential challenge. I’m inclined to think that we don’t simply intend to do F, but intend to do F by way of G. So that’s a bit of more fine-grainedness for you. I don’t think we intend all the way to motor schemata, e.g. the sequence of movements we make, when playing a piece we’ve memorized. So, there’s more to flesh out here that depends on how we think of intentions. But I suspect many would agree that intentions, at least subject level states that I think of as typical products (though not necessarily) of practical reasoning, are not so fine-grained as to represent motor schemata.

    On swinging the golf club, the process having that property, specified at that coarse grain, is what is controlled. But the precise kinematics is not, so that commits me to something about the content of intention. Alternatively, you might show me that on a plausible theory of intention, the kinematics are represented in intention, in which case, my theory will drive me to say that they are controlled. I take it that in some sense, this is empirical and also conceptual (see the Sea of Science exchange with Brendan and James).

    How’s that for a first pass?

    • Hi Wayne,

      Thanks for the (speedy) reply! That’s an attractive line about skill. I’m not sure I disagree, but perhaps it’s worth noting that in characterizing skill in this way you divorce skill and control. Some might resist, claiming that skilled action is the paradigm of control: that the very merging of what you’re calling automatic and controlled features of processes into one skilled process is in fact control. So I guess I’m pointing to an alternate way of understanding control – one that does not begin by dividing automaticity and control. I’m not endorsing this alternate way (at least not here, especially since I haven’t spelled out how this alternate way would look in any detail), and I don’t know who else might. I just wanted to raise the issue to see what you thought.

      Regarding the content of intention, I agree there is (and should be) a large empirical component to any correct answer. You’re also right that theory might push us one way or another. How robust is the personal/sub-personal distinction we’re working with (I take it you have something like this distinction in mind when you refer to subject level states)? Does the phenomenology of intending play any role in fixing a good story about the content of an intention (and should it)? Etc.

      A thought in reaction to what you say: say that over time an agent develops a set of ways to execute familiar intentions (swinging a golf club, or whatever). The acquisition of a familiar intention sets off the sequencing of familiar motor schemata, involving the precise kinematics of the activity. I take it perceptual feedback will vary the specifics of the motor schemata from case to case, but this might just as well be true regarding the subject level content of the particular intention. So I’m not seeing a strong causal reason to distinguish the rough-grained property (e.g., ‘I shall swing this golf club’) and the more fine-grained kinematics. There are other good reasons, no doubt. I should think more about what they might be. So I wonder: why do you find the distinction attractive?

  4. Wayne Wu

    Hi Josh:

    My speedy replies are due to my trying to do something other than write lectures! This is more interesting.

    On the first point, it would be interesting to see how skill=control is spelled out. I haven’t thought about it as much as I probably should, but I’m inclined to the weaker supervenience claim. I think we need to leave room for control precisely for the purposes of acquiring skill. This is why many psychologists are happy to tie consciousness and effort to control. I think some of them might think that control entails consciousness. In the paper I link to above, I note how many different characterizations of automaticity are out there, done roughly the way Fodor originally characterized modularity, in terms of a set of features. The previous proposals above try to regiment. Still, there’s lots to sort out, and I hope this post and work by you, Ellen and others will push the envelope.

    Anyway, I suspect we don’t disagree too much on this.

    I’ve no strong feeling about where to draw the boundary so that certain properties of a process are automatic and certain ones are controlled (your second point). Let the previous points simply mark an intuition. I guess I do think there is some rough boundary, since there are too man properties a process exemplifies that I think some have to be automatic. But where to draw the line…?

    It is said that when we reach for an object, there’s an initial acceleration phase and then deceleration, about two thirds of the way to the target object I think (this is from memory…work done originally by Marc Jeannerod perhaps?). I take it that my reach’s having that property is automatic, and so I was thinking of things of this sort in respect of kinematics.

    Maybe a concrete example would help if we drill down further on this point (i.e. specification of the different contents at issue, at different grains). Ok, back to writing lectures!

  5. Hi Wayne and Josh,

    Thanks Wayne for encouraging me to weigh in. I have more thoughts on how to understand control when it comes to skill than how to individuate intentions, so I’ll stick with the former.

    Recently, I’ve been developing an account of skill where I say that there are (at least) four levels of control in a skilled action that we ought to do justice to when formulating a theory of skill. The first level is what Wayne is calling control simpliciter and what I call strategic control: the intentional, personal-level, often conscious kind of control that will most readily be accepted by everyone as real control. Basically, these are the goals and strategies involved in skilled actions.

    The other 3 levels include top-down influences on attention that develop over time with the intentional practice of a skill. The development, also through intentional practice, of appropriate repertoires of motor routines, and then the even lower level kinematic processes that change over time and practice but not ever through any explicit attention to those processes in practice.

    I call them all control because they are sensitive and responsive in appropriate ways to changing environmental cues and to the relevant variables of an action instantiation. I think these are all different and that they are also different from what Wayne is calling control and what I call strategic control but I don’t think there is any principled reason to call them all control–my reason is just that I can’t think of a better word. They certainly exhibit many characteristics of adaptive, diachronic learning processes and so they seem controlled and intelligent to me. But certainly, they are automatic. They are intelligent and automatic and I know Wayne thinks these two things aren’t mutually exclusive. So the disagreement seems to be whether to call them controlled or not–not whether or not they exhibit cognitive characteristics.

    To me, it seems that the most important aspect if thinking about the four levels of control is that we take them seriously in order to try to understand many neglected areas of cognition. Maybe calling these kinds of processes controlled will encourage other philosophers to take them seriously as genuinely interesting and authentically cognitive. That’s a pragmatic issue, of coarse.

    Another thought is that maybe we could think about agreeing on a terminology (if the phenomena are more or less acceptable to us all) in order to avoid unnecessary confusions in debates going forward. I’m open to negotiations!

    • Wayne Wu

      Hi Ellen:

      Thanks for that. With your comment and Josh’s, I’m beginning to see something that is probably worth negotiating more carefully than I have, namely talk of “control”. I suspect that the term really has many different meanings as you both bring out.

      For my application of the distinctions, my own emphasis is on automaticity. It’s useful to draw on it to understand different facets of attention, as I hope to bring out a little bit in a later post, but also to understand one basis of positive symptoms in schizophrenia, as I will bring out in the next post. Given that focus, I define automaticity negatively, relative to the absence of something that is plausibly a type of control if anything is, namely an intention-influenced process of the right sort of influence. Call this “control”.

      At this point, I think you and Josh are correct to note that there are ways in which “control” gets out of control, so to speak. We can think of skills, or the various adaptive and learning processes you note. Perhaps the way forward is setting out in detail the different types of processes that one is apt to speak of as control, a taxonomy of sorts, and then to have an assessment of how to think of them. Many of these, as you suggest, might come out as automatic under my account of automaticity.

      I suppose one further distinction that might make a difference is the personal/subpersonal one, but we all know that this distinction itself is need of further clarification.

      Still, the point is that perhaps laying out the different forms of control is the first step, and doing so with an eye to psychological phenomena as well (learning, acquisition of skills etc.). I do think that’s a useful project worth pursuing, and I know you see it that way as well.

  6. Hi Wayne,

    Another thought that I had about control was that it is very much like attention: in that there is the most commonly accepted kind–“folk attention” and with empirical study and philosophical care, we come to see that there isn’t just one kind of attention but many. And, I think the same may go for control. That isn’t at all meant as an objection to your distinction since I see it as useful to contrast intentional control and automaticity and I think this is still possible while doing justice to other kinds of processes that seem to deserve to be called “controlled” too.

    And, of course, I think that starting with something like a rough taxonomy is a good idea since it’ll help us to distinguish at least some important features of different kinds of control. In the end, I think it will be the features of these processes that will be most interesting for theoretical understanding and empirical progress.

    Also, I just want to plug Wayne Christensen, John Sutton, and John Michael who have also been thinking about these issues (skill and control) and they focus a lot on the same kind of control that you’ve been considering, Wayne. They call it cognitive control. I’ll try to get them to weigh in.

    • Wayne Wu

      Hi Ellen:

      It will be interesting to discuss attention in light of the issues that are arising in this discussion I am committed, by way of publishing certain claims!, to there being “one kind” of attention, though there are ways to modify this: selection for action (e.g. automatic versus controlled!). Perhaps there is something that unifies many of the things that we would want to call control.

      Definitely agree that identifying certain repeatable properties of psychological processes that might fall under the title “control” would be of most use in thinking about this further in a fruitful way. So, some sort of clear taxonomy of such processes again would be useful (your rough list above, which I’m sure you spell out elsewhere, would be a start).

      I guess if I had to add a label to the control at issue in the definition above, it would be “agentive” control, which would cover bodily and purely mental processes. But this would just be terminology. It’s this contrast with agency that I think useful to think about automaticity. Certainly, one test of the usefulness of the conceptual regimentation would be to see that it helps make sense of certain phenomena (e.g. auditory hallucinations and attention…promissory note!). I would certainly now add the caveat that there can be other notions of control!

      Maybe we need a concrete example where the different sorts of processes can be easily identified…

  7. Hi Wayne,

    Interesting stuff – as Ellen says, John Sutton, John Michael and myself have been working along similar lines. Here’s a link to a paper by me, John Sutton and Doris McIlwain on cognitive control in skilled action. It’s long and covers a lot of ground, but the idea you present above is one of the key points in the paper – see esp. section 2.3.3. So I’m looking forward to reading some of your papers!
    https://waynechristensen.wordpress.com/2013/08/15/cognitive-control-in-skilled-action/

    Cheers,
    Wayne

    • Wayne Wu

      Thanks Wayne for that link. It’s a substantial piece of work, and I hope to look at it more carefully at some point. I think the question of skills is an important topic, and a quick perusal of the early sections emphasizes to me lots of work that has been done. I think these issues will probably arise again when I try to post on some of the other announced topics. Clearly, I may want to think more about these points.

      I wonder if it might be useful to separate some questions beyond the plea above for a taxonomy. This links to my first post and exchange in the comments there with James and Brendan.

      As I noted, I’m really interested in the idea of agentive control which for me is just another way to talk about action. I take the “standard” view in identifying intentions as causally essential (the focus on intentions in particular can be relaxed). There’s the philosophical question, “What is action?”, and there is an empirical question about the processes of action. Cognitive scientists focus on the latter, though they typically also have a different conception of action than one that immediately engages the philosophical question, or so I have found (I’m thinking that their focus is more on the control of motor movements and modulation of such movements).

      There might be something similar going on here. I am interested in agentive control (action). I notice that actions have automatic features and yet involve control. I like the gloss by Schneider and Shiffrin, but for my own theoretical reasons, cannot make a fundamental appeal to attention. So, you get the above. They answer the questions: what is a controlled process? What is an automatic process? So, the answer is something like: a process P is controlled iff it satisfies condition C. Perhaps this answer can be arrived at largely a priori.

      Consider a different question: what is the process of control? Imagine that in attacking this question, the answer amounts to unpacking condition C in terms of identifying different additional processes that are causally related to P or to the establishment of abilities to P etc. Here, perhaps the empirical has a stronger purchase on how we go about answering the question.

      I hope that distinction is clear enough. The issue then is that questions about skill might concern processes of control such that in unpacking it, we understand what control of a process comes to (e.g. agentive control).

      That’s perhaps too crude to be too satisfying, but it’s a gesture at another level of regimentation that is probably good to have one the table, i.e. different questions the theory of control and automaticity aims to answer.

      • Thanks Wayne, I see that I was being a bit hasty – the paper I linked to is about the relations between cognitive and automatic control in skilled action, but it isn’t directly addressed to the nature of agentive control. I’d agree though that unpacking how control works is going to help clarify what agentive control is. I agree with Ellen that we need to distinguish multiple kinds of control, and we also need to distinguish different kinds of automaticity. Shiffrin and Schneider conceptualize automaticity in terms of processes that occur independently of cognitive control, and the Stroop task paradigmatically exemplifies this: word reading occurs *despite* the task instructions and intentions of the subjects. But this isn’t what’s happening with the kind of automaticity that you describe with your piano playing example. Here, the automaticity of technique is highly sensitive to intentions. The *focus* of cognition may be on higher order musical features – the musical message, etc. – but control is profoundly influencing the automatic action processes that occur. It’s helping to select and shape in detailed ways (‘parameterize’) the motor control structures involved in performing the action.

        One conclusion that has been drawn from this is that automaticity in Shiffrin and Schneider’s sense is actually quite rare: for the most part ‘automatic’ processes *don’t* occur independently of higher control. Pashler & co make the point in relation to the pop-out effect in visual search that you describe: true bottom-up pop-out in visual search is uncommon – people are very good at ignoring stimuli that aren’t part of their ‘search set’ (‘Attention and performance’, Annual Review of Psychology 2001). Hommel also has a nice discussion of these kinds of phenomena in ‘Consciousness and control: not identical twins’ (Journal of Consciousness Studies, 2007).

        Cheers,
        Wayne

        • Wayne Wu

          Hi Wayne

          I think we are largely in agreement. In the paper I link to above, I speak of a “simple connection” between automaticity and control where automaticity is the absence of control. In that respect, to the extent that one endorses this simple connection and holds that there are multiple forms of control, then I suppose there will be multiple forms of automaticity. There is the worry, lurking in the background of this exchange, and which Ellen brings out, about a proliferation of terminology. But so long as we are clear…

          You raise a good point in the last paragraph about pop-out, worth emphasizing. I’m not sure that automatic processes are rare (see my comment to Gary), but I agree that many psychological processes that we think are automatic are influenced by cognition. As you note, there was and in fact still is a debate about this among attention theorists. Stephen Yantis argued 20 years ago that perhaps sudden onset stimuli were the only ones that truly grab attention, with many pop-out effects in visual search being influenced by one’s goals and indeed, where the stimulus (feature singleton) was not goal- or task-relevant, the subject could effectively ignore it. This was surprising to me when I first learned of it (another pair of individuals in this debate whose work is worth looking at are Charles Folk and Jan Theeuwes).

          There’s some interesting work by Yarbus on eye movement that is relevant here, and I’ll post on that when I get to attention.

          What I don’t want to get lost here is the idea of relativizing control or automaticity relative to some feature, independent of how we unpack the idea of control and, if we endorse the simple connection, of automaticity. A given process, then, can have both controlled and automatic features. Given the pervasive influence of our goals and intentions (which you allude to in your comment), it’s not surprising that control insinuates itself into many processes.

          When you say: “The *focus* of cognition may be on higher order musical features – the musical message, etc. – but control is profoundly influencing the automatic action processes that occur. It’s helping to select and shape in detailed ways (‘parameterize’) the motor control structures involved in performing the action”, it seems to me that an empirical goal is to spell out the underlying mechanisms. I’d be interested in seeing this spelling out.

          • Yes, I think we are mostly in agreement, and some of what I was worried about can be covered nicely by your account. There are a few niggling worries though connected to the points Josh raised.

            One of the things I had in mind in wanting to distinguish forms of automaticity is distinguishing between automaticity that approximates the classical view, and the kind of automaticity present in elite skills like high level piano.

            So at the strong end of the scale for automaticity there are processes like word reading. If you focus on a word it’s pretty much impossible not to read it – given the appropriate stimulus, the process is obligatory. But overt actions as a whole are rarely if ever automatic in this sense. Utilization syndrome is what you get when action really is stimulus-driven. In the case of Stroop, it takes effort but subjects do suppress the conflicting response and name the printed colour.

            As far as overt actions go, habits, routines, and everyday skills come closest to the classical model of automaticity. Arguably, they involve much more cognitive control than is often recognised, but they do often involve extended, endogenously structured motor programs that can run off with relatively light higher control. Susceptibility to action slips is one consequence of having an action production process with a high degree of endogenous structuring and light cognitive guidance.

            But highly developed skills like elite classical piano are really quite different. They aren’t organised to operate with strong endogenous structuring and relatively minimal higher control. To the contrary, elite performers actively resist routinization so as to prevent performance from becoming dull and wooden. There is a tremendous amount of automaticity, but it is maintained in an open state so as to be exquisitely sensitive to higher control.

            To a first approximation this is all fine for you – we just need to be careful when we specify which aspects of the action production processes are controlled and which aren’t. But Josh’s points about grain of intention and the way that skill enhances control do raise some tricky issues.

            First, consider indirect forms of control. It’s hard to directly control word reading when focusing on a word, but it’s easy to exert indirect control through the control of gaze. Consequently, reading sentences and longer texts isn’t obligatory – voluntary control is required to maintain the reading process.

            Moreover, in the normal case you intend to read the words. Stroop creates an odd situation where word reading produces a conflict with the task requirements, but in normal reading you direct your eyes at the text so that word reading will happen.

            This is where Josh’s point about the grain of intention is important. In forming action intentions specifications for action are generated – in Christensen, Sutton & McIlwain we call this an action gist. The action gist might directly specify some of the implementation parameters (e.g., you reach for the coffee with your left hand so that you can then pick up your phone with your right hand) and leave others for automatic control, like detailed kinematic parameters. But as well as direct specification of some implementation parameters, there can be more global specifications. You don’t directly control the detailed kinematics of reaching, but your reaching movements usually happen pretty much as you intend them to. When movement doesn’t go as intended there is enhanced attention, correction, and sometimes ongoing effortful learning to bring your motor system more into alignment with your intentions.

            Furthermore, the effect of skill is to enhance this control – an elite pianist has far more control over performance than a bedroom plunker. So in this larger sense it’s not clear there is an oppositional relation between control and automaticity. Narrowly, some feature F of a process is directly controlled or it isn’t, but more broadly it can be the case that F isn’t subject to direct specificatory control, but is subject to broader forms of control.

            I don’t have developed views on all this, and I do think your account is getting at a valuable notion of automaticity. But maybe it will be necessary to distinguish different kinds of automaticity after all, in order to take into account the different ways that control can exert influence.

            Cheers,
            Wayne

          • Wayne Wu

            Hi Wayne:

            Since I see lots of agreement, let me now simply formulate the challenge (or a challenge). If Josh and Ellen are still following they can weigh in:

            Take the elite pianist. Note that there are different forms of control that are being implemented, different things that are aptly called control, and perhaps with each, some relevant correlated sense of automaticity. Provide adequate analyses of these that is sensitive both to conceptual conditions and to known empirical phenomena. The goal here is to provide an adequate theoretical characterization that is then generalizable to other forms of skilled and unskilled behavior, that helps us make sense of the acquisition and loss of skills, of performance etc. Further, see if it can be generalized to philosophical questions about volition and agency.

            You, Ellen and Josh are already doing this, I take it. I think it’s a great project. There ought to be a conference!

            One question: by the classical view of automaticity, is this what I’ve been calling the “simple view”?

  8. Hi Wayne,

    Thanks for doing this!

    Forgive me if my question would be answered if I was more familiar with your work, but I was wondering what your views on animal cognition are, particularly with respect to which animals intend things and which animals don’t.

    Since your account of automaticity is defined in terms of intentions, it seems important to figure out which animals on this planet are intenders and which ones aren’t. When I observe my cat chasing a laser pointer, I am inclined to understand pretty much the entire gamut of the behavior in terms of automatic processes: automatic orienting response to red dot, automatic perception of it’s ecological significance, automatic engagement of motor programs, etc. And if this is true of cats, it will almost certainly make it true of many other animals, making it seem like in nature automaticity is the rule and control is the exception.

    It certainly seems possible to take a kind of intentional stance towards the cat and view the laser-chasing in terms of the cat intending to catch the red dot, but it seems like there must be a fact-of-the-matter about whether the cat is *really* an intender or whether it’s just massively convenient for us to treat cats *as* intenders. The same goes for every other creature, including those borderline cases like insects or fish.

    So, in a nutshell, my question boils down to: (1) do you think there is a fact-of-the-matter about which animals are really intenders and (2) If the answer to (1) is yes, what scientific methodology could we use to figure out these facts and apply them to the phylogenetic spectrum such that we have a definitive answer?

    Thanks!

    • Wayne Wu

      Hi Gary:

      Thanks for your note. I think you are right that automaticity is the rule rather than the exception. Even if we restrict ourselves to human behavior, the same is true. Control, however we end up defining it (see comments above), is fairly specialized. One of the issues arising from above is how specialized it is, depending on what one means by control!

      One way to respond to the question of animals is to note that the definition of control offered above can be relaxed a bit. For example, one could speak of goal representations rather than intentions (which is one kind of goal representation), and this might allow for many different sorts of mental states that would cast the net more broadly so more non-human animals would be capable of control, as defined, and thus, agents of a sort.

      People also speak of motor intentions, and these strike me as a different sort of beast which I suspect any creature that has motor capacities likely must have. But here, it seems better to speak of motor plans or schemata, though the issue veers towards the terminological.

      I have perhaps an overly philosophical conception of intention which ties it, in principle, to practical reasoning. They are states that are the products of practical reasoning, though they do not need to always arise as a result of explicit reasoning. That’s a gesture at a theory that needs to be explicated further (which alas, I haven’t done). But let’s say that you grant that point, which might be too restrictive. Then one possible attack is to identify not intentions per se, but capacities for reasoning about what to do. But whatever you take intentions to be, sure, I do think there is a fact of the matter about it, but in the end, how one transforms this to an empirically tractable question (e.g. “Does creature X have intentions?”) will depend on what we take intentions to be. My overly sublimated conception (perhaps!) suggests that the question will turn to criteria for attributing capacities for certain kinds of reasoning.

      This isn’t going to be a satisfying answer to your question, but hopefully a sketch of a direction I would go.

  9. Hi All,

    A lot of good stuff here. For the record, I just finished a dissertation primarily on control (some other stuff too), and I had the sense while writing that almost no one was working actively on the important issues. So I’m pretty happy to see all this discussion. I have too much to say for a blog post, so I’ll go against my instinct and be as brief as possible.

    Ellen: How do you explicate the relations between the four levels? Answering that might take time, so maybe what I’d want to know first is whether the first level is necessary and/or sufficient for an agent to exercise or possess control in some sense, or whether the other levels are necessary to capture what we usually mean when we speak of control.

    Wayne W: You said, ‘the point is that perhaps laying out the different forms of control is the first step, and doing so with an eye to psychological phenomena as well (learning, acquisition of skills etc.).’ I agree. In work of my own (in progress), I try to isolate what I think the many forms of control have in common – cognitive control, attentional control, motor control, and more. It turns out to be difficult to say with any precision just what control is, but I’m hoping the work will make some progress on that front, and also that it will help bring some theoretical unification to the ways cog sci talks about control. One thing this discussion underscores for me (and you seem to agree, Wayne) is that talk of control can quickly get hung up in terminological disputes, or in some swamp where people talk past each other by failing to be explicit about what they really want out of a discussion of control (or what they think a theory of control should do for them, etc.). I don’t have an easy solution to that problem.

    Wayne C: Really glad to see this paper exists! Looking forward to reading it.

    • Wayne Wu

      Hi Josh:

      Let’s say that “control” in cognitive science turns out to pick out many different phenomena, where there is no interesting shared property each type exemplifies. I suppose the interesting sort of control among those will be the one that helps me understand agentive control (i.e. action), where “control” is used more in some philosophical sense. This is what I meant by the (b) type questions in my first post.

      Perhaps at some point, I would drop the notion of “control” at the level of analysis of cognitive science. It is just a term after all. I suspect that what you call cognitive, attentional and motor control are all going to be part of understanding, mechanistically, agentive control.

      The goal of other published work of mine (to be self-centered here!) is to identify a basic structure to agentive control (Many-Many Problem, attention as selection for action, intention as structuring solutions to the Many-Many Problem). That picture is largely a priori driven, but if it fleshes out a structure for agentive control, it also fleshes out a structure to fit the other sorts of control you note.

      The invocation of skills, which you began here!, is a helpful one, something that needs to be part of the picture of action. It would be a problem for my more arm chair account of action if it did not allow for skilled action.

  10. James Genone

    H again Wayne,

    Great discussion here. This might be naive on my part (and apologies if you’ve addressed it in the paper you linked to), but it strikes me that the process of forming intentions can be subject to automaticity and control, especially given the emphasis on intentions being the result of practical reasoning (which some, at least, would consider a skill). So I’m wondering: might there be a threat of circularity in your definitions?

    • Wayne Wu

      Hey James

      Thanks. Interesting point. I think that deliberation, like any action, is subject to the division between control and automaticity (as you note). Since we are also raising issues about the philosophical significance of the distinction, I have wondered if the debates of volition and belief might benefit from the deployment of these ideas…a thought I’ve always meant to explore, but time…

      Would it be enough to say that intentions themselves do not imply control? Might this skirt any worries about circularity that seem to be in the offing? After all, some intentions might arise spontaneously, i.e. automatically, without deliberation. I wouldn’t say that intentions have to arise from practical deliberation, but have a more complex relation to it. So the analysis of control via intentions is not the analysis by something that itself implies or presupposes control.

  11. James Genone

    Hi Wayne,

    If you grant that intentions don’t require (any?) control, then those that aren’t controlled at all will be automated. But automaticity is defined in terms of lack of control. Maybe enough can be said about intentions independently of whether or not they are automated or controlled to provide some additional clarity to the notions of control and automaticity (introducing some non-vicious circularity). Perhaps the notions of intention and control could even be jointly defined.

    My own hunch (and I’m somewhat ignorant of work on intention) is that automaticity is frequently important for forming intentions (as you suggested, we often don’t have time to deliberate). On the other hand, it seems to me that there is usually some sense in which we can be described as controlling the formation of an intention (even if the details are automated). I’m reminded here of Frankfurt’s regress of higher-order volitions and desires: perhaps there’s some level at which control gets exerted in the very forming of a higher-order intention, without requiring any further prior intentions? And maybe this could be expressed in some kind of reflexive version of your definition?

    • Wayne Wu

      Hi James:

      Thanks for the response (a nice respite from writing lectures…though I really need to write some more!).

      I don’t think the analysis of intention needs to advert to the notions of automaticity and control, nor do I think the concept of intention entails claims about automaticity or control. So that’s why I don’t see an actual circularity.

      It is true that actual intentions can be characterized by these notions, say this intention’s being fully automatic, this other due to control in light careful deliberation by the subject. But this isn’t a matter of the analysis of the notion.

      So, yes, I take it that as you note, enough can be said about intentions that don’t appeal to control and automaticity, and indeed, I think if one looks at canonical discussions of intentions, I suspect that those notions will be absent. I could be wrong, of course, and perhaps the ideas will creep up in the analysis of intention. That would require reinvestigation of certain phenomena with the additional tools focus on automaticity/control might avail one.

      I’d have to refamiliarize myself with the Frankfurt case to have a comment on that!

  12. James Genone

    Thanks Wayne, this helps me to understand your view. I think some theories of intention (e.g. Grice) identify forming an intention with something like willing or deciding, and what I had in mind was that these accounts presuppose some notion of control in the analysis of forming intentions. I may well be wrong about this, or you might just want to reject such accounts. The sketchy idea in the second paragraph of my last post was meant to be a friendly suggestion for avoiding any disagreement here, by building possibility of control and intention arising simultaneously, on the model of what Searle called ‘intention-in-action’.

    • Wayne Wu

      Hi James:

      “by building possibility of control and intention arising simultaneously, on the model of what Searle called ‘intention-in-action’.”

      I’m definitely on board with that, and not simply because I was Searle’s student!

  13. Hi Wayne,

    Since in your introduction you spoke of “formulating philosophical questions such that it is clear whether we can get an empirical purchase on the questions at issue”, I hope it’s fair to ask whether you see a possibility of getting an empirical purchase on the controlled/automatic distinction. Your definition in terms of intent is not exactly operational, so the answer isn’t obvious.

    • Wayne Wu

      Hi Bill:

      I think the question, What is automaticity, control?, is more on the conceptual end of the continuum of questions, where the continuum concerns the amount of empirical import. So, it’s more of a conceptual than empirical question. In the quote, I had in mind more attempts by scientists to engage with some traditional questions, without a sufficiently deep understanding of those questions.

      Where I see the empirical purchase in automaticity and control comes out a little in the exchange above, which is buried now! Basically, in looking at mechanisms that implement forms of control, once we specify those. What has come out is the possibility of many different things we might call (plausibly) control and how disparate these might be!

  14. Hi All,

    Briefly, though Wayne hasn’t mentioned it much here, I’d like to point those following the thread to Wayne’s ‘Confronting Many-Many Problems’ paper, in Nous (2011). It has some nice discussion of agentive control.

    To Wayne Christensen: I do have a few papers on relevant issues in the works, none published yet. A few are almost ready for submission, and I’ll e-mail you about them.

    To James Genone and also Wayne Wu: This issue of forming intentions (and whether we do) is one of my favorite. Wayne has a nice paper on this, and I’m writing a paper on deciding that treats it as an intentional action over which agents can exercise control. A growing number of folks (including, I think, Wayne) doubt this can be done. but maybe not: on the account I’m developing, there is no problem with claiming that control and intention arise simultaneously – the trick is just to see where in the process of deliberation and decision these notions do work.

    • Wayne Wu

      Interesting issue on decisions, Josh. I am less sure about how much “control” one can exert over beliefs or decisions as products of explicit reasoning. I agree that forming them can be intentional actions, so put in my terms, the question then is what aspects of the formation of belief or decision can be controlled, and which of those might matter to the philosophical debates in respect of volition and belief/decision. Open questions, be interesting to see what you find!

  15. Wayne Wu

    On different notions of control, I leave it open for commenters to post their notions and brief analyses of them. I’d be interested in hearing about other ideas. Perhaps the concert pianist provides a good trope to ground things.

  16. Josh Weisberg

    What a great discussion! Thanks to Wayne and all.

    I wonder about the transition between controlled and automatic processes. I’ll use a guitar playing example, rather than piano, because I suck at piano! When I first learn a guitar part, I have to consciously, effortfully, attentively, etc. practice the part. But as I progress, more and more of the sequence of music becomes “automated” (scare quotes just in case this doesn’t fit everyone’s criteria). During this process, I can roll along in some parts, but in the hard bits, I need to “exert control” (wanted to say “attend” but let’s leave that out for now). Eventually, the whole sequence becomes “automated.” My conscious mind is free to focus on other things, like trying to sing with the guitar part or attending to some of the subtler nuances of performance, like dynamics and tempo.

    Does focusing on this gradual transition help with some of the questions here? It seems a bit off to me to say that I no longer have an intention to play the automatic bit. Rather, the intention to do so is no longer conscious/focused upon/etc. But maybe that’s not right. I’m a little unsure how to track intentions: it gets pretty messy pretty fast. (Hence the morass of the action literature!)

    My predilection is to say that in the automatic process, I am no longer conscious of the sequence of actions required to do what I am doing. I may still be conscious of the action in other ways–as what I want to do now, as playing this bit of the song, etc.–but not as “first do a G chord, then do a C chord, then do that trilly thing…” Interestingly, if I have trouble, I usually need to get the start of a sequnece right and the I kind of “speed through” relying on “muscle memory.” But in the transition, I may still have to exert control over the first few steps.

    But I take it using consciousness to do work here is fraught with trouble, and not just because of nonhuman animals. Another worry, though, about Wayne’s proposal is that to get the right intentions, we may end up with “intentions we’re conscious of”. So consciousness re-appears. But maybe not. Anyway, these are more musings than a question, though I do wonder about intentions in the transition from controlled to automatic.

    Cheers!

    Josh Weisberg

    • Wayne Wu

      Thanks Josh for those thoughts.

      I have left out consciousness and attention, and I’ve not done the issue of intention sufficient justice. You bring this out nicely in your post and the others have pushed on this as well.

      I do think that we need to bring attention back into play here if we are going to tackle the challenge above seriously. So, I think your intuitions in the guitar case are exactly right. Definitely: let’s start talking about attention in thinking about how one exerts control, and how attention is different once things become automated. This requires being clear what attention is (three posts from now, though maybe I should move that up!).

      I think attention is just selection for action (I have a functional conception of it, and attention can be unconscious!). Briefly: I think the way you deploy attention is going to vary in the transition you describe, so how you select items to guide action. I’ll try to spell this out in the next post on attention rather than doing it here, but cryptically, I suspect that there is a way in which attention is somewhat scattered even if directed as you first learn, and then becomes more focused and less scattered as you acquire expertise. Ok, cryptic, I know.

      Consciousness I really want to avoid as much as possible in that it might be more obscuring than clarifying in this context (I share your hesitation). I think of its involvement really as the target of explanation rather than something we can use to explain. Certainly, in the psychological discussions, it was often brought up in the context of control–so conscious controll–but I think we can (I hope we can) clarify lots without bringing it in at the get go.

      Clearly, the issue of intentions needs to be thought out more, and I think that the theory of intention would benefit from thinking about the dynamics of attention in extended, skilled activity. Some of Elizabeth Pacherie’s earlier work might be useful here (see her paper in Cognition). I suspect that the contents of intention will change, in a way, as one acts in an extended way, and the level of grain might also differ. In learning a piece, I might have intentions to play specific notes in specific sequences (using demonstratives here for those specific notes, in my intention) but later simply intend to play this passage. So, perhaps there is a coarsening of intentions in the transition from controlled to automatic. Just my two-cents for now.

    • Wayne Wu

      Oh, to add one more thing: if there are different notions of automaticity/control, then there might be a different story of transition for each. This makes the explanatory/theoretical task that much more complicated, but it’s good to know the challenges ahead.

  17. Josh Weisberg

    Great, Wayne. Thanks for the reply! (And you really must not want to write those lectures! ;>) )

    I think that’s a very interesting suggestion about “coarsening” intentions. Simon’s “chunking” idea comes to mind here.

    I look forward to reading the rest of your posts and checking out the papers.

    Cheers

  18. That’s a nice challenge Wayne, sounds like a good project 🙂

    On the issue of awareness, intentionality and skill – FYI, this is an interesting study of a pianist (Gabriela Imreh) preparing to perform a piece:
    Chaffin, R., & Logan, T. (2006). Practicing perfection: how concert soloists prepare for performance. Advances in Cognitive Psychology, 2(2-3), 113–130. https://www.ac-psych.org/?id=2&rok=2006

    Broadly speaking it supports the points you’re (Josh and Wayne W) both making about changing patterns of awareness during skill learning. As familiarity with the piece improves the focus tends to shift from basic technical issues such as fingering to higher order musical features like interpretation and expression (see figure 2 particularly). There’s a coarsening of awareness and attention in some respects, but also worth emphasising that there’s an increase in the resolution of awareness and attention in others – better awareness and control of the overall performance and it’s key musical features.

    Also FYI, the skills research group I work with wrote a review of Chaffin’s book:
    Geeves, A., Christensen, W., Sutton, J., & McIlwain, D. (2008). Review: Roger Chaffin, Gabriela Imreh & Mary Crawford, Practicing Perfection: Memory and Piano Performance. Empirical Musicology Review, 3(3), 163–172. https://kb.osu.edu/dspace/handle/1811/34109

    This is quite a nice description of expertise in pianists:
    “In order to deliver their utmost, they [concert pianists] must remain focused on the musical message, on the emotional qualities of the work, on the overall structure of the composition, and not on the notes. The work that these experts have put in, innumerable hours of training over a period of years, even decades, enables them to concentrate on sound quality and expression, forgetting about technique and difficulty. Instead of delivering a routine performance fixed by repetitive practice, musicians are able to react flexibly to the environment. They are able to modify tone, tempo, and use of pedal, for example, to adapt to the acoustics of the hall. They can follow a spontaneous urge, deciding onstage to play a phrase with more flamboyance or introspection. This is achieved by listening to their fantasy. Once the goal is set and the sound imagined, they act. A high-level command is issued, eliciting a set of complex movements. There is no time for thought to be given to the ‘what’ or ‘how’ of creating this desired effect. This is musical expertise.”
    (Adina Mornell, concert pianist & professor, University of Music and the Dramatic Arts, Graz. In Wulf Attention and Motor Skill Learning 2007, p. 140.)

    The flexibility and sensitivity to the situation described are key things to note – marks of a very high degree of control. In some respects this highly developed situational sensitivity is even more evident in jazz – e.g. Lawrence Kart’s description in the liner notes of the Miles Davis album Filles de Kilimanjaro:
    “Listen, for example, to the interplay between Davis and Williams, and then between Shorter and Williams, on “Petits Machins” and the two takes of “Tout De Suite.” At first, in terms of volume and polyrhythmic detail, Williams seems more aggressive alongside Shorter, but the byplay between them is just that – a brilliant, heady, almost puppyish joust in which both participants are always there but their identities are never really at stake. Not so, though, opposite Davis, where Williams often seems to be reading Miles’ mind, even finishing the trumpeter’s phrases before he gets a chance to play them. It is the drummer who shapes most of “Petits,” and on the alternative take of “Tout,” with its sibylline trumpet solo, the empathy between William and Davis is so complete, their taste for counterpunching foxiness so much of a piece, that a DNA test seems to be called for.”

    Cheers,
    Wayne

  19. Josh Weisberg

    Also, that era of Miles’ group is among the most unbelievable–Tony Williams, Wayne Shorter, Herbie Hancock, Ron Carter, and Miles. Among the best examples of a hive-mind I know of!

  20. Ellen Fridland

    Such a great discussion here! Sorry to miss out on the meat of it. Moving to a different country, suprising to no one except me, is actually very hectic! I’d really like to keep the discussion about control going. And I’d like to second Wayne’s idea of a conference. Any ideas on who could host?

    • Wayne Wu

      One possibility is an APA session, but a full blown conference or short workshop might be nice. I would vote for something a bit interdisciplinary, though the participants should be thought out so that there would actually be a nice exchange between empirical and philosophical sides. But it would be fun.

      Ellen, you know people in the field more than I do, so I turn the question back to you or Wayne (indeed, your shout out on FB brought some of us together who did not know of the others work!).

      Good luck on your move!

  21. A conference would be a lot of fun. Ellen or Wayne or whomever, if you have some ideas about this e-mail me, maybe we could cook something up, or maybe I could help you cook something up (or don’t: this is just to say I’m willing to help).

    • Wayne Wu

      One issue is to come up with a set of 2-3 clear questions that aren’t overly ambitious but still weighty and substantive, to constrain things. So, maybe we can use this forum just to hash out ideas for questions. What are the questions that we need to address in respect of automaticity/control/skills?

  22. Josh Weisberg

    A conference would be awesome. Maybe something at the SPP or SSPP (or European equivalents)? An interdisciplinary thing would be great.

    I’m personally into the question of how the conscious/nonconscious distinction maps to/differs from the automatic/non-automatic distinction(s). But I’m happy with anything in this area–there is lots of stuff here!

  23. I’m not great at the non-ambitious questions. Some questions emerging (for me) from the above discussion: what is the relation between automaticity and skilled behavior? What is the role of intention in control? Riffing off of Josh W, is there a role for consciousness (and perhaps more specifically, conscious attention) in the exercise of control or in skilled behavior?

Comments are closed.