Controlled Variables are Perceptual Variables

[bnhpct]

Adam, I can only agree with much of what you have said. Rick, though Adam may not believe it, I can attest that you have become much more considerate in recent years, more aware of the effect as well as the avowed intent of what you write, and that is very encouraging.

I have to agree with you, Rick, that naturalistic observation is is essential to PCT, to identify the variables that we humans perceive and control and figure out how we organize them in levels or other categorizations. Personally, I must agree with this, because without this my field does not exist. Not just observation of humans of course, but other living things are of less immediate concern in my field.

Rick, you say the mathematics of control are not relevant to such observation. Adam, you may be aware of a way that poor grasp of the mathematics of control misdirects naturalistic observation or distorts our attribution of levels or categories of perception. If you see such problems, please tell us.

The mathematics of control are directly relevant to another PCT project to which you both have made many valuable contributions, the building of computer programs that simulate control systems and demonstrate control phenomena. This has been limited almost entirely to the lowest level of the perceptual hierarchy.

@rsmarken

RM: Perhaps this is a good place to point out that in actual PCT research controlled variables are rarely described in terms of physical variables. The variables {v1, v2, …, vn} that are the basis of most of the controlled variables that have been identified by PCT research are actually perceptual variables themselves: such as the length of the lines that make up the sides of the rectangle controlled in my “What is Size” demo or the colored shapes that make up the components of the program controlled in my “Program Control” demo

There’s a confusion of the two projects in this statement. These are demonstrations of phenomena, as is the simulation of catching a fly ball. On your website, demos and and research are properly in different places. The two projects intersect. Demonstrations and simulations are not in themselves research, but as a test-bed they contribute to research. Research into e.g. baseball-catching includes testing different assumptions as to what is controlled by comparing the relative success of simulations based on those assumptions. However, knowledge of the computer code can be overinterpreted.

@rsmarken

(yes, it is a program that is controlled in that demo – not a sequence or sequence of sequences; there is no sequence that is maintained in that demo when you are controlling a program).

The behavior of the “Program Control” demo is produced by computer code structured as if/then contingencies, and therefore you know that what appears to the subject is produced by one if/then ‘program’ or the other. For that reason, you say that the subject is controlling a program. However, the user cannot perceive the code that generates those appearances on the screen, the user only perceives the appearances. The same appearances on the screen could be coded as two repeating sequences, randomly alternating. One sequence complies with the programmatic description “if the shape is circle, the next color is blue; else, the next color is red”, and the other does not. True, the user has been instructed in program terms, but in actual execution I think you will find that it resolves to sequences. The contingency word “else” opens an infinite universe but experience shows that this universe contains only one shape, a square. The recitation of the program becomes “if circle, blue. If square, red.” An initial boggle is taking these to be simultaneous, e.g. “if it’s a circle, the circle is blue”, so after being tricked by that a few times the sequentiality becomes salient. Two sequences are OK “Circle, then blue; square, then red.” Anything else requires a press. So now the task is to recognize the sequences that require a press: “Circle-red-press; square-blue-press”. This reorganization is difficult to avoid, because the program statement makes salient that which requires no action. The cat is not focusing on the empty mousehole, the cat is focusing on the mouse which memory and imagination tell him to be prepared to snatch from that spot.

@rsmarken

RM: It is these kinds of definitions of controlled variables – definitions in terms of lower level perceptual variables of which the controlled variables are a function – that are sought in PCT research. The question of the correspondence between controlled variables and their environmental correlates is rarely asked because, except possibly for the lowest level perceptual variables – intensities and sensations – there is no environmental correlate of these perceptions variables. The definition of controlled variable in terms of lower level perceptions is enough to give one an idea of what aspects of the physical environment – the environment described by the models of physics and chemistry – are actually controlled

Bill’s 1979 analysis of “getting ready to back a stick-shift car out of a driveway” would seem to qualify as “actual PCT research” by this criterion, but only because Table 1 on p. 179 of LCS focuses on control actions and therefore is limited to the perceived variables that the successive actions directly affect. Bill’s table refers to door, seat, belt, mirror, window, clutch, key, keyhole, starter, starter noise, engine, engine noise, lever, stick-shift car, and driveway. We need neither avoid the phenomenological rabbit-hole of their “really real” existence nor get lost in it. These are perceptions that Bill remembered controlling as he wrote the lines in that table, and they are perceptions that each of us can remember and imagine controlling as we read those words. That accord is as good as it gets; and it had better be because it’s all we’re going to get. The angles, distance relationships, etc. are not more real just because they are immediately affected by measurable motor outputs. They, too, are perceptual constructs.

I agree with your framing that Bill’s table suggests a project for research. If Bill framed it as a project, he didn’t do so in this chapter in Ozer’s book. (He used it there to hammer yet again on the folly of counting ‘behaviors’.) Note that the whole table describes control of a sequence which has as its goal being in the car on the road at the foot of the driveway, which is the first step of another sequence. In this post from last April we revived the link to the spreadsheet you started, and I again raised the question how we are going to include sequence perceptions in this naturalistic investigation. Here’s my 2019 Manchester paper on that.

We need to seek these kinds of variables for PCT research as well as angles, distance relationships, etc. Bill suggested that Miller, Galanter, & Pribram (1960) presents a good collection of program perceptions. It’s at that link and in archive.org.