Academic journal article The Psychological Record

On the Accuracy and Reliability of Predictions by Perceptual Control Theory: Five Years Later

Academic journal article The Psychological Record

On the Accuracy and Reliability of Predictions by Perceptual Control Theory: Five Years Later

Article excerpt

Accuracy and reliability are such common features of human behavior that we often overlook our many successes and notice the rare times when we fail. We move food from the plate and into our mouths so easily and precisely that we take the results for granted, noticing instead the rare slips between plate and lip. We routinely maneuver ourselves and our conveyances (such as horses, bicycles, airplanes, and automobiles) from home, to work or play, and back, often over long distances and with no mishaps. The probability of a successful trip is remarkably high; we notice, instead, the rare exceptions.

No matter how vividly they occupy our attention, catastrophic failures while eating, or while traveling, occur only once in many thousands, or millions, of successes. Accurate, reliable results are the rule, even though our behavior occurs in a world that varies in ways we cannot precisely predict. We cannot know in advance the weight and balance of the food on our spoon during each bite, or the states of all of the muscle fibers in our own hands and arms. We cannot know in advance the moment-by-moment conditions of the road, the weather, other drivers, or our own bodies. What we know in advance is the result we intend to accomplish. In the face of inevitable variability in the world, our own actions also vary just as is needed, and we produce the unvarying results that are hallmarks of intentional behavior.

Here, I demonstrate a way to model and simulate intentional behavior and I show that in the long term the model's quantitative predictions are as accurate and reliable as the behavior of the person. I use the model from perceptual control theory (PCT), a theory that explains how people achieve consistent results in a variable world. PCT and some of its applications are described in a foundational book (Powers, 1973) and in edited works and anthologies (Hershberger, 1989; Marken, 1990, 1992; Powers, 1990, 1992; Rodrigues & Lee, 1994). In this paper, I do not attempt to compare PCT with any other major theory of behavior, but I discuss some common misconceptions about PCT and purposive behavior.

Predictions

Data from studies of people in groups are often called "nomothetic." Nomothetic data may allow us to predict things like the average score for a group, or the proportions of people in certain groups who will perform certain classes of actions. For example, we might predict the proportion of first-year university students with a certain score on a standard entrance examination who will join at least one university-recognized social organization and will also maintain a grade average of "B" or higher. However, we cannot predict with certainty whether a particular student with that score will join an organization, or maintain a "B" average, or do both. In contrast, "idiographic" predictions, like those in PCT, pertain to specific individuals. As I apply it here, PCT yields quantitative predictions of the movements a specific person will use to create and maintain (control) an intended result.

General Method

I used the behavioral task and modeling procedures reported by Powers (1989), Bourbon, Copeland, Dyer, Harman, and Mosley (1990), and Bourbon and Powers (1993). I describe them below. Predictions were run on 8 June 1988, at Stephen F. Austin State University, Nacogdoches, Texas. Predicted data were collected during a presentation at the annual meeting of the Control Systems Group, Durango, Colorado, on 31 July 1993.

The Subject and the Task

I was the subject. Figure 1A shows the experimental arrangement; Figure 1B shows the causal interactions among environmental variables; Figure 1C shows the results when I did one run of Condition 1 when no disturbance affected the cursor. The task was pursuit tracking. I used a control handle (h) to keep a cursor (c) aligned with a target (t) comprising two marks on the computer screen. Every 1/30 sec during a run, the program (1) took new values of the target (7) and the random disturbance (d) from two time-indexed records created in 1988 at the start of Condition 1; (2) sampled the position of the handle and converted it to a value scaled to the height of the computer screen; and (3) plotted the cursor and target at new positions on the screen. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.