Start the demonstration by pressing the "Run" button. Keep the lower line (the cursor) directly under the upper line (the target) by moving the mouse to compensate for the disturbances to the cursor. The disturbances are time varying waveforms. The word "Repeat" that is printed during the trial marks the point where the disturbance waveform repeats. A graph of the results appears after 1 minute of tracking. Press the "Run" button at any time to restart the tracking task.
The responses (mouse movements) that keep a stimulus (like cursor position) under control in a tracking task occur in a closed loop. The stimulus in this loop (the controlled variable) is both a cause and an effect of responses. This means that Stimulus-Response (S-R) explanations of control won't work. An S-R interpretation of control says that something about the stimulus (such as how far the cursor is to the left of right of the target) causes you to respond in just the right way so that the cursor stays close to the target.
In the Nature of Control demonstration it was shown that the correlation between cursor (S) and mouse movements (R) was nearly 0.0. This suggests that cursor movements are not the cause of the mouse movements that keep the cursor on target. However, the S-R correlation could be low because we not are measuring the aspect of S that is the actual cause of R. Rather than attempting to find the aspect of S that causes R, the present demonstration tests whether any aspect of S can be found that causes R. If some aspect of S does cause R then if variations in R on two different occasions are the same, the variations in S on these two occasions should be the same as well.
In this demonstration you produce identical variations in R during two successive periods of a tracking task because the disturbance to the controlled variable (S) is the same during these two periods. A graph of the results of the experiment will be displayed automatically as soon as you have completed both periods of the tracking task. The graph will remain on the screen until you start the tracking task again by pressing the "Run" button.
The left side of the graph shows the results for the human subject (you); the right side of the graph shows the results for a control model doing the same task. Note that variations in R during the first (black squares) and second (red squares) periods of the experiment are virtually identical for both subject and model. The R1- R2 correlation between response variations during these two periods is typically on the order of .99. Note that the variations in S during the first (black squares) and second (red squares) periods of the experiment are not identical for both subject or model. Indeed, the S1-S2 correlation between stimulus variations during these two periods is typically far lower than the correlation between response variations during the same periods. The S1-S2 correlation is typically close to 0.0 and rarely more than .3, certainly not high enough to account for the nearly perfect correlation between responses (typically > .99) on repeated trials.
The results of this demo are so surprising that some have suggested that the high correlation between R1 and R2 results from the fact that the subject is simply repeating the mouse movements made during the first part of the experiment from memory. This would presumably explain why there is a high correlation between R1 and R2 when there is a low correlation between S1 and S2. The subjust is matching second phase mouse movements (R2) to those made in the first phase (R2) without even watching the cursor. You can test the plausibility of this explanation by closing you eyes after 30 seconds of tracking (which is when the same disturbance repeats) and trying to imitate the mouse movements you just made. If the "repeated movements from memory" explanation is correct you should again obtain a high correlation between R1 and R2 while there is a low correlation between S1 and S2. And cursor variations before and after the 30 second period (S1 and S2) should also stay close to the target, as they did when you were actively controlling cursor position. You will see that the "repeated movements from memory" explanation fails because the correlation between R1 and R2 will be quite low and, while the correlation between S1 and S2 will also be low, you will be able to see from the graph that the stimulus traces, S2, in the second phase (with movements made form memory) stay nowhere near the target. The results of this experiment shows that there is no aspect of the stimulus input to a control system that seems to serve as the cause of the responses that control that input. In fact, cursor movements (S) do cause responses (R) but since responses also cause cursor movement the causal connection between S and R is "masked by the simultanous causal connection between R and S. The lesson of this demonstration is that S-R causality cannot help us understand the phenomenon of control. When behavior involves control -- as it does in this tracking question -- what we want to understand is not the cause of the responses but, rather, the variable that is under control. We will see in the Mindreading demo that determining what a person is controlling is not always obvious.
This demonstration works best when you are able to control the cursor well. The better your control of the cursor, the lower will be the correlation between stimulus variations during the two periods of the tracking task.