[From Bruce Abbott (2018.03.05.1840 EST)]
[From Erling Jorgensen (2018.03.05 1255 EST)]
Bruce Nevin 2018-03-05_09:06:46 ET
BN: … The following is in Bill’s review of the ms. of Casting Nets and Testing Specimens:
In summary, it sounds as if people only tae [take] action when their internal standards are threatened. I think you should make it clear that they also, at the same time, are adjusting internal standards as a way of baringing [bringing] patterns of perception into being. Many of these variations have no external causes: those are not just reactions to disturbances, but represent creative purposive acts demanded by higher levels in the system for reasons having nothing to do with fending off disturbances …
Bill to Phil, 4 December 1987 (*Dialogues *p. 381)
EJ: A distinction I have often used in my own understanding of PCT is that between compensatory tracking versus pursuit tracking. It seems to me that disturbance-resistance is compensatory tracking, i.e., counteracting the effects of forces pushing a perception away from its preferred setting(s). By contrast, (but only a slight contrast), a change in perceptual value simply by virtue of a changing reference standard is an instance of pursuit tracking.
EJ: As I say, the distinction is a small one. It does parallel the understanding that there are two entry points from outside a given control loop, from above and from below. This is to say that there are two key mechanisms – under normal operating conditions – for changing the value of a perceptual signal. One is from above: change the reference signal, and the perception will subsequently follow it, by means of pursuit tracking. The other is from below: insert a disturbance to the perceptual variable, and the perception will start to be affected, until it is counteracted by the effects of compensatory tracking.
EJ: This binary classification leaves aside the effects of reorganization, which can also change perceptual signals, whether by creating new perceptual input functions, or by adjusting gain or slowing or other parameters. It does tend to support the intuition that a paradigmatic way to study perceptual control systems is by way of tracking tasks, whether set up as pursuit or compensatory. This discussion has helped me to see that the former is aligned with reference-following, and the latter aligned with disturbance-resistance.
I like the clarity of your distinction, so I hope what I am about to say doesn’t muddy the waters too much!
When I worked in the Glass Science department at the Owens-Illinois Glass Co.’s Technical Center back in the early 70s, we had a small programmable glass furnace that employed a PID (proportional-integral-derivative) controller. One could program the reference temperature of the PID to follow a given profile over time. As the reference temperature changed, the PID controller varied the electrical power to the furnace to keep the error between the reference value and sensed temperature small. In this way the actual furnace temperature “pursued” the changing reference value.
If the program called for a steady temperature, then the controller would defend against disturbances to that temperature (such as the introduction of a cold sample). Operated in this way, the controller performed compensatory tracking. But note that such disturbances could occur during pursuit tracking as well, and the controller again would try to compensate for them while continuing to pursue the changing reference level.
But one does not necessarily have to conceive pursuit tracking as necessarily involving a changing reference. In the TrackAnalyze demo provided with LCS III, an onscreen target changes position smoothly but unpredictably during the experimental run, and the participant’s job is to keep the cursor aligned with the target. Bill Powers stated that the reference value to be attained in this task is zero difference between target and cursor position; the participant is asked to maintain this reference value during the run. To keep the difference at zero while the target moves, one must of course pursue the target with the cursor. The movements of the target can be viewed as disturbances to target position. The difference between this task and the usual compensatory tracking task comes down to whether the disturbance acts on the position of the target or on the position of the cursor. The actual control system involved is the same either way. But if all we are doing in this task is to keep the difference between the two positions at zero against disturbances, then this is really no different than the usual compensatory tracking task!
A way to recover your nice distinction between pursuit and compensatory tracking is to suggest that the instructions to the participant are effectively to set her internal reference value for the cursor to the position of the target, wherever that may be at any given moment.
So why did Bill suggest treating the reference as a zero difference between target and cursor? I suspect it is because the target seen on the screen is “out there” in the environment, whereas reference signals are supposed to be inside the organism. If you say that the current target position is to be treated as the reference position, then this makes it seem (incorrectly) that the reference is part of the external environment rather than inside the participant. By specifying the reference as “zero difference,” Bill made it clear that the reference is not external. Yet it makes no difference to the analysis which description one adopts – zero difference or target position, and by adopting target position as the reference, one makes it clear that the cursor is then “pursuing” a moving target (actually an internal reference set to target position) and not merely compensating for a nonzero difference.
Bruce