Hi Eva, your idea of equating melody with sequential proprioception may be more than an analogy. Maybe our most fundamental sense of sequence is proprioception because we need it for motion & coordination, and our perception of melody is a beneficial side effect of this level of perception when applied to the auditory modality? But when combined with the propensity to dance, it further facilitates temporally coordinated proprioception?
···
On 12 May 2019, at 10:54, Eva de Hullu (eva@dehullu.net via csgnet Mailing List) csgnet@lists.illinois.edu wrote:
[Eva de Hullu 2019-05-12_09:51:36 UTC]
In following this discussion about anticipatory postural control, I think an important difference between PCT and the other theories mentioned is that in PCT, causality is circular, closed loop. On the first proposed level, there is only perception of here-and-now intensities. I think one of the reasons why other theories struggle so much to fit observations into their theory, is because of their effort to model closed-loop events into linear causal models. In that way, you’ll need difficult models with prediction and anticipation, that can be explained by PCT through the perceptual hierarchy.
How could we go from a here-and-now perception to the perceived “anticipatory behavior”?
I think we could have a better understanding through the perceptual hierarchy, where levels are build from the bottom up, layer by layer, into increasing complexity. In answering these kind of questions, we could use some help from the proposed hierarchy.
Going up in the hierarchy, the perception of time emerges when transitions become a way of making sense of changing configurations. The sequence level is several steps higher, and at this level prediction must take place.
Imagine a melody, where each note is followed by another note at a certain pace and pitch. When you hear a melody for the first time, there are no references yet for this melody, and you’ll not experience the sense of predicting the next note. But after listening to it carefully, you’ll have created references for the melody, and with these references you’ll have the sense of being able to predict what follows. If you hear a familiar melody, your references and perceptual input will match if they are equal. If the melody changes from your expectations, you’ll find that you can no longer predict what’s next; without references for the melody, you’ll not recognize the song.
Following this, the anticipatory postural control could just concist of references in the form of sequential perceptions. In a melody, the sequence is a way of organizing single notes (categories) in such a way that they are no longer just single notes, but also a melody. A movement such as reaching, is not just a single pose, but also a sequence of poses. If you look at what happens from the perspective of circular causality, interpreting the steps in the sequence as (existing) references, I believe PCT is perfectly able to explain what happens.
Eva
PS I’m conscious that I might be repeating or missing some of the things already said here. I waited a bit too long submitting this answer, so excuse me if I’ve not properly addressed or incorporated other replies.
On Sat, May 11, 2019 at 9:00 PM Richard Marken csgnet@lists.illinois.edu wrote:
[Rick Marken 2019-05-11_11:58:37]
On Fri, May 10, 2019 at 8:05 PM Heather Broccard-Bell random.information.consultant@gmail.com wrote:
HB-B: I think this does make sense, and would be interesting to build. I am curious about a ‘critical experiment’ that would distinguish the PCT model from the mainstream account… although it does seem to be one of these phenomena that people have described without couching it in any particular theoretical framework, so this may not be possible.
RM: I think the APA phenomenon is implicitly couched in a particular theoretical framework; that of an open-loop causal model of behavior. The implication is that the APA is an output that is being produced open-loop in anticipation of a future perturbation. This can be tested because my model assumes that APAs are actually controlled variables that are used to control another controlled variable – the event called “reaching”. So applying an appropriately gentle, varying force disturbance to the postural change being made during the period of the APA should be resisted. This resistance should be detectable as appropriate variation in the force exerted by the animal during the APA period – the force used to produce the APA. It should also be detectable as consistency of the APA on different trials with different disturbances; this could be determined by analysis of visual records to see if the variation in the APA over trials is what would be expected based on variatoin in teh disturbances.
RM: I think there might be a way to create a “portable demo” of behavior that seems to involve an APA. I’ll see if I can think of one.
Best
Rick
On Fri, May 10, 2019, 17:05 Richard Marken csgnet@lists.illinois.edu wrote:
[Rick Marken 2019-05-10_17:04:34]
On Fri, May 10, 2019 at 12:50 PM Heather Broccard-Bell random.information.consultant@gmail.com wrote:
HB: I kind of think a lot of the problem is semantic. Some seem to be upset by the term “prediction” or “anticipation”, and I think what Rick/Henry suggest is a reasonable explanation for a mechanism for what most people would call anticipation/prediction. It can’t be magic :))
RM: The only time I get upset by terms like “prediction” and “anticipation” is when they give the wrong impression of how the PCT model explains some behaviors. This usually happens when it “looks like” a behavior involves “prediction” or “anticipation”. For example, the behavior called “catching a fly ball” appears to involve “prediction” or “anticipation”. You can see it in the movements of the fielder in my baseball catch simulation (https://www.mindreadings.com/ControlDemo/CatchXY.html). The fielder seems to be making “anticipatory field adjustments” (AFAs?) in anticipation of being in the proper position to catch the ball. But we have found that the fielder’s behavior can be accounted for quite well by a model that involves no prediction at all; the fielder in the ball catching demo gets to the correct position by controlling two present time perceptual variables: vertical and horizontal optical velocity.
RM: The PCT model I propose to account for the postural adjustments that seem to anticipate future disturbances (APAs) also involves no prediction. My model involves present time control of an event perception that I call “reach”; I didn’t define it very well because I don’t have a good idea of what it is but this event would be a temporal pattern of proprioceptive sensations – sensed angles of the limbs, center of gravity, etc. This temporal pattern is a perceptual variable, just as a temporal pattern of phonemes is a perceptual variable. The “reach” control system I proposed uses variations in posture (including APAs), limb angles and muscle forces to bring this perceptual variable to the reference state (that I called “reach”) while protecting it from disturbances that can affect the perception at any point during its production. I’m conceiving this control as being analogous to controlling for saying a particular word, as described in the “Control of Sequence” chapter of B:CP. In that model of word production there is also no prediction involved; the perceptual function just “constructs” a word based on the sequence of phonemes (analogous to the postural adjustments occurring during a “reach”) that are produced by lower level systems. It doesn’t predict what phoneme is coming next.
RM: In order to test this model we really have to come up with a clearer definition of the proposed controlled variable – which I suppose could be called “reachness”. And it really would be good to have a working model of this behavior, if for no other reason than to show how a model that controls a variable that is defined over time can be controlled. I believe that the most important way that PCT differs from other control models of the behavior of living systems is that it assumes that ALL behavior is control; And since control always involves the control of perception that means that simple behaviors involve the control of simple perceptions and complex behaviors involve the control of complex perceptions – and these are usually perceptions that are defined over sometimes considerable amounts of time. This is one of the main reasons why Bill proposed a hierarchy of perceptions. It was to account for the obviously hierarchical nature of behavior – behaviors “using” other behaviors to achieve their purposes; for example, using behaviors like running, throwing, catching and hitting to achieve the purpose of playing baseball – and for the fact that the behaviors that use other behaviors to achieve their purposes involve control of more complex perceptions than the behaviors that are being used; the perception controlled when “playing baseball” is much more complex (and defined over a much longer time period) than the perceptions controlled when running, throwing, catching and hitting.
HB-B: We know that both humans and cats fail to show APAs if they are taken by surprise.
RM: Yes, I noticed that. From the perspective of my model, that would result from not setting the reference for “reach” or " bilateral arm movement" or whatever the event is. This happens a lot in my sport, racquetball. The ideal way to take a forehand shot is to do a lot of APAs that get you into position to take it off the back wall. But sometimes I get a shot that comes at me so fast that I don’t have time to set up for a “proper forehand shot” – no APAs needed – so I just do what I can to move the racket to the ball like a beginner.
RM: I hope this makes some sense because I think it would really be a great area for research on control of higher level perceptual variables – variables that are defined over time.
Best
Rick
So yes, some environmental information serves to remove the surprise, kicking the control systems into gear. Is that (those) control system(s), dealing with the appearance of the target often used in the cat experiments to signal the upcoming stimulus, totally divorced from the “paw bat” ones? Should they be coupled? What happens if the expected stimulus never arrives, but the target remains? I guess there is fatigue at some point. What is the relevant question to ask?
In populations of upper motor neurons in motor and premotor cortex, there is often activity seconds before any movement is observed. What does that mean for control systems? That’s obviously not just delay. Those populations get input from… like everwhere. Motor cortex is bonkers!
On Fri, May 10, 2019, 11:59 Richard Marken csgnet@lists.illinois.edu wrote:
[Rick Marken 2019-05-10_11:53:31]
RM: Well, this has certainly been an interesting thread so far. I’m happy to see that there is a kind of consensus that the PCT explanation of apparent “anticipatory” behavior is that it is likely to be actions aimed at controlling a “higher level” controlled variable. The landslide was started by Henry Yin who said:
HY: There’s no conflict between negative feedback and anticipatory behavior. The real question is what kind of CV do you need to produce such behavior.
RM: I think this is the right question: what kind of CV would an organism be controlling that would result in your seeing such apparently “anticipatory” behavior. Martin suggests that the CV is a sequence:
MT: [Here is a] quick and dirty PCT approach. I start with the intention to bat at the target, which would set a reference for a sequence that could crudely be taken as “store energy, set trigger, wait, release energy when required”, or in a more general sense “Prepare, wait, execute”.
RM: And Erling agrees:
EJ: I rather like Martin’s proposal for controlling a prior stage in a sequence.
RM: And I agree that the CV is something like a sequence, but not like the one Martin suggests. In the case of the an anticipatory postural adjustments (APAs) seen in the cat’s reaching behavior (described in the Sheppen and Drew paper), it seems to me that the controlled variable is the “reach” itself – an event – like a “golf swing”. An event differs from a sequence only inasmuch as it is a continuously changing sequence of lower level perceptions that is defined over time. The “reach” requires making postural changes over time that get the paw to the intended destination while maintaining balance. So I see the APAs as the initial components of this “reach” event.
RM: I see these APA’s as being analogous to the initial muscular movements – for example, contraction of the diaphragm - that are involved in producing a word. These movements are not outputs produced in anticipation of a future disturbance but are the outputs in a control loop that is aimed at controlling a perception of the word. The word itself, like reach, is the controlled variable – a variable that is defined over time.
RM: Because “reach” is defined over time it looks like the initial postural adjustments that are made to control this perception are anticipatory. But I believe that they are actually just like the outputs in any control loop – continuously acting to keep the controlled variable in a reference state, protected from disturbances. If this is the case – if APAs are actually the initial outputs of a control loop controlling a variable that is defied over time – then disturbances to these APA’s will be resisted. Heather mentions some data that is consistent with this view:
NB-B: I suspect that the attached (Aruin, 2003) paper goes some way toward reconciling the “invariant, pre-programmed” feedforward conceptualization with the actual phenomenon: namely, it’s not really invariant (or, well, it’s not invariant in that there is a pre-programmed set of muscles activated in a pre-programmed way).
RM: I haven’t read the whole article carefully but apparently what Arium did was look to see whether variations in the initial the position of the body affected the nature of the APAs. And indeed it did. I see these variations in the initial position of the body as disturbances to a controlled variable (in Arium’s case it was “bilateral arm movement” rather than “reach”) and these disturbances are compensated for by the variations in the APAs.
RM: Of course, the idea that APAs are the outputs in a control loop controlling a variable defined over time has to be subjected to far more tests. And control models should be built to show that control of a higher level perception, like “reach”, can lead to behavior that “looks” anticipatory. So there’s some general ideas for PCT research that would be a great place to start for researchers interested in doing research on purpose!
Best
Rick
–
Richard S. Marken
"Perfection is achieved not when you have nothing more to add, but when you
have nothing left to take away.�
–Antoine de Saint-Exupery
–
Richard S. Marken
"Perfection is achieved not when you have nothing more to add, but when you
have nothing left to take away.�
–Antoine de Saint-Exupery
–
Richard S. Marken
"Perfection is achieved not when you have nothing more to add, but when you
have nothing left to take away.�
–Antoine de Saint-Exupery
