The core methodology of PCT is The Test for the Controlled Variable (or “for the Controlled Quantity”) . “The Test” has been described in a number of places, but questions and requests for clear definition still recur in CSGnet email traffic, and answers given, only to be lost to sight and forgotten. Because it is in the very bedrock of PCT, it seems essential to place a careful and comprehensive definition of The Test front and center where everyone can easily find it and refer to it.
Bill Powers published two formal specifications of The Test.
- “The nature of robots[:] Part 3: A closer look at human behavior,” BYTE 4(8), August, pages 110, 112 [pp. 10-11 of the online PDF].
- Behavior: The control of perception, pp. 232–246.
Our late colleague Phil Runkel “rephrased” these two authoritative descriptions in one comprehensive definition in his last book, People as living things (in the last pages of Chapter 7). Because not everyone has that book (though we all ought to), I present that definition of The Test here so that participants in this forum can easily access it and refer to it.
Marken’s study of intention and the study by Robertson and colleagues of the defense of the self-concept both demonstrate uses of the Test for the Controlled Quantity, which is the basis for method in all PCT research. Marken’s purpose was to show how purpose or lack of purpose can be discerned in the relation between environmental events and the perception the participant brought about of the movements of a line on the computer screen. In Table 7–1, we can look for the connection (the correlation) between the position of a line and the disturbances given it by the environment. In every trial by either participant, when we look for the line having the lower correlation with the environmental disturbance, that line turns out to be the one whose position the participant intended to control. The other line follows the disturbances to a much greater degree. The correlation of the other line with the disturbance given the chosen line was not high—did not approach 1.0—because the position of the other line was also connected to the arrow keys. The correlations between the intended line and the disturbance were small, averaging –.05 for one participant and .09 for the other. That pattern exemplifies The Test, which tells us to look for a variable (one perceivable by the person) that is acting as if a purposeful influence is acting on it. It is often clearer, actually, to say this in the negative: we look for a perceived variable that is not behaving as it would in a nonliving environment. We look for a perceived variable that is not behaving as it would if there were not a purposeful influence acting on it.
Robertson and colleagues wanted to test the opposition of action to disturbance in maintaining a perception very high in the neural hierarchy—the self-concept. They did that by having the “experimenter” present the participant with an idea that they thought would contradict or disturb the participant’s self-concept. That is the core idea of The Test: disturb the presumed perceived quantity and see whether the person opposes the disturbance.
Notice how different The Test is from traditional psychological research. Traditionally, psychologists have looked for strong correlations between input (the “independent variable”) and output (the “dependent variable”). In Marken’s experiment, that would be the correlation between the disturbance (input) and the position of a line the participant can act upon (output). In Table 7–1, the correlations with the line-position the participant does act upon are those in the columns headed “Intended line”; contrary to the traditional assumption, they are not strong, but are very small. Under PCT, they are of course predicted to be small; the intended line is the line moved purposely by the participant and not abandoned to the influences of the environmental disturbances. The correlations with the line-position the participant ignores, shown under “Other line,” are much larger, also as predicted.
I turn now to a more formal and detailed description of The Test for the Controlled Quantity. The procedure is contained in the following nine steps. …
- Select a variable that you think the person might be maintaining at some level. In other words, guess at an input variable. Examples: light intensity, sensation of skin temperature, admiration in another person’s voice. (Powers often speaks of the input quantity, because one usually looks for an amount or degree of some variable—such as temperature—the perception of which is controlled.)
- Predict what would happen if the person is not maintaining the variable at a preferred level.
- Apply various amounts and directions of disturbance directly to the variable.
- Measure the actual effects of the disturbances.
- If the effects are what you predicted under the assumption that the person is not acting to control the variable, stop here. The person is indeed not acting to control it; you guessed wrong about the variable.
- On the other hand, if the effect is markedly smaller than the predicted effect, look for what the person might be doing to oppose the disturbance. Look for a cause of the opposition to the disturbance which, by its own varying, can counterbalance variations in the input quantity (such as pulling as necessary on the handle of your umbrella to keep the wind from carrying it off). That cause may be caused by the person’s output. You may have found the feedback function.
- Look for the way by which the person can sense the variable. If you can find no way by which the person can sense the variable (the input quantity), stop. People cannot control what they cannot sense.
- If you find a means of sensing, block it so that the person cannot now sense the variable. If the disturbance continues to be opposed, you have not found the right sensor. If you cannot find a sensor, stop. Make another guess at an input quantity.
- If all the preceding steps are passed, you have found the input quantity, the variable the person is controlling.
When you find the controlled variable, you can then usually make a very good guess about the nature of the internal standard controlling it. But describing the internal standard in precise words is not of first importance. The important thing, both for further experimentation and in practical affairs, is to have found how to disturb the controlled variable and how to avoid doing so.
Sometimes, both in research and in everyday life, you can ask people to adopt temporarily an internal standard that you describe to them. If they have themselves freely chosen to comply with your request and if you can describe the internal standard clearly and objectively (as in the examples of research I have given so far), you are off and running. At other times, you may not be able to persuade the person to adopt the internal standard you have in mind. Even if the person is willing, the person may not understand your request sufficiently well. In that case, you must start from scratch and use all the steps of The Test to discover what variable the person is indeed controlling.
To use The Test, you must make a guess about an internal standard and then change something in the environment that the person senses. If you succeed in changing the thing—that is, if the person does not act to maintain it the way it was—then you have guessed wrong. If the person does act against the change you try to make, then you have guessed right, or at least you are on the right track. You know something about the person you did not know before. But you may have guessed wrong about the aspect of the change, the input quantity, that you think the person is wanting to maintain. You will discover that fact if later steps in The Test go wrong. Then you have to guess again. You will, however, be ahead of the game, because you know that the input has something to do with the change you tried to make in the environment.
It is easy for an onlooker, watching someone ward off a threat to a controlled variable, to make a wrong guess about the variable the person is trying to maintain. Cries of “No! No!” or “I won’t do it!” or “You think you’re pretty clever, don’t you?” or a stony silence—those are all good indicators that some variable is being disturbed, but poor indicators of what the variable might be like. For example, what perception might a person be defending who said to you, “Don’t talk to me that way!”? Here are some possibilities:
- I’ve been trying to be helpful to you, and now you tell me I’ve been actually been doing you harm. That’s exactly opposite my intent, and it hurts me to hear that; I don’t want to hear that.
- I don’t want people to hear you speaking disrespectfully to me.
- That’s a frivolous way to talk, and I want the people here to believe you are taking this seriously.
- You sound desperately discouraged; please don’t give up hope. I want to hear optimism.
- You talk as if I have done something bad! I am not going to think of myself as a bad person! I don’t want to hear you telling me I am a bad person!
Once you have made a guess, you can then hunt for something you can do that might disturb that variable. Then you have to be careful about interpreting the person’s reaction. The person might want you to stop talking out of fear that you will disturb the variable you have hit upon, or simply because you are distracting the person’s attention from a task the person wants to resume.
Even a simple physical action can be perplexing. You move through a crowded hotel lobby. You step aside to avoid someone and find yourself pushing against a third person. The third person makes a quick contrary shove that opposes your push. Is the person simply trying not to fall over, is the person maintaining his manliness, or is the person wanting to communicate an antipathy toward physical contact with strangers?
It is rarely possible, in the natural setting, to hit upon a good guess at the first try. Narrowing the possibilities requires several tries, sometimes a good many. People do, of course, learn a good deal about the internal standards of others after making wrong guesses for several months or years. Still, people can live together for decades, giving careful attention every day to evidences of disturbance, and still be surprised at the reactions of family members. People who claim, “I know what you are thinking!” after brief acquaintance are being fatuous; so are those who say, “Well, you ought to have known what I was thinking!”
Sometimes we are not sure whether a person is intending to control a variable. Sometimes, after we have “defined” a variable in such a way that we can recognize changes in it (for example, the brightness of light on a page or the number of people talking at once in a conversational group) and have tried to alter it, we find that the person pushes back, but not skillfully. That is, the person seems to show poor control. The person may be trying to control that variable, or the person’s effect on it may be a side-effect of the person’s intent to control a different variable. In a message to the CSGnet on 16 October 2000, Rick Marken said this:
[If] The Test tells you that a variable is not being controlled very well, then there are at least three possible reasons for this finding:
- The variable, as defined, is not a controlled variable. This is the default hypothesis when a variable fails The Test. The next step is to try a different definition of the possible controlled variable and Test again.
- The variable, as defined, is a would-be controlled variable: the behaving system [for example, a human] is trying to control this variable but [has not yet found an effective way to do it]. This might be our hypothesis if we have reason to believe that most control systems of this type do control this variable.
- The variable, as defined, is a controlled variable: the behaving system … is not controlling it very well, because there is a conflict. This might be another hypothesis if we have reason to believe that most control systems of this type do … control this variable.
Sometimes people new to The Test worry that they might do damage to the people they want to Test. It is essential, in carrying out The Test, to make sure, when you take an act you think will disturb a controlled variable, that you do not move (or speak) so strongly that the person will be unable to counteract what you do (or say). If you move or speak too strongly, you will not discover what you want to discover. Knocking the person over with a bulldozer does not tell you anything useful about the ability of the person to stand upright. The Experimenter with the rubber bands always wants to keep the amplitude of the disturbances small enough so that the Controller can easily maintain control. The piano teacher always wants to keep the bad news about the pupil’s fingering small enough so that the pupil can quickly rectify the faults.
I know a couple of people who were one day talking about perceptual control while they were whizzing along a freeway at about 60 miles per hour. The passenger offered to demonstrate to the driver how people cope with disturbances. He took hold of the side of the steering wheel and pulled down, very gently at first, then more strongly, while the driver, of course, resisted that disturbance and kept the car going along in its lane. Then the passenger gradually, slowly released the wheel. You can see that the passenger in that Test certainly had no wish to exceed the driver’s ability to keep the car in its lane. That is a good example to keep in mind when you are thinking of the degree of disturbance you want to apply. Just imagine that you might be killed if you pull too hard on the steering wheel.
[Excerpted, with permission, from People as Living Things: The psychology of perceptual control by Philip J. Runkel, which may be purchased from Living Control Systems Publishing or from other sources such as Amazon. This book is recommended to every serious student of PCT.]