C.43:5.3 - Comparing explanations of an organism’s response
A researcher observes rapid corrections of a person’s hand movement and wants to distinguish a response to delayed visual position from a response that also predicts the consequences of recent motor commands. The result sought is an explanation of the obtaining control, not a recommendation that the person perform the researcher’s algorithm. Use C.28.CM:4.3–4.6 to construct serious alternative mechanisms, derive their different consequences and qualify the evidence that could distinguish them.
Consider two deliberately simplified hypotheses for a hand moving along one axis toward a target at zero. Positive positions and commands are rightward. The display supplies an earlier position y; let c be the net displacement that the second model predicts from the motor commands issued since that displayed position was sampled. The hypotheses share a corrective gain k = 0.5 and a next command expressed as displacement over the same short interval. Their command-to-movement model, observation delay and absence of other corrective inputs are stipulated conditions of this comparison, to be qualified before interpreting an experiment.
| Hypothesized mechanism | How it constructs the next command | With displayed position y = 2 cm and predicted intervening displacement c = -3 cm |
|---|---|---|
| Response to the delayed visual position alone | Use the displayed deviation: u = -k*y | Request -1 cm, a further leftward correction. |
| Prediction using the visual position and recent commands | Estimate current position as x_hat = y + c, then use u = -k*x_hat | Estimate -1 cm and request +0.5 cm, a rightward correction. |
The display is the same, but the proposed information-to-action relations predict opposite signs. The difference follows from the intermediate state estimate, not from calling one model predictive. With no predicted intervening displacement, both command -1 cm; that condition alone cannot distinguish them. The useful experimental contrast therefore combines a controlled display delay with a qualified history of intervening movement, while observing the direction and timing of the ensuing correction. Increasing delay without constructing those different predictions would leave the research question unanswered.
Suppose a rightward correction is observed under the contrasting conditions. It disagrees with the visually driven model’s leftward prediction and is compatible with the second model. Before attributing the response to internal prediction, examine how the commands and observations were established and what other information the person could use. A third model that uses current proprioceptive information can also produce a rightward correction. That alternative requires a further discriminating condition or leaves the internal mechanism unresolved; merely fitting the second model cannot remove it. Conversely, an inaccurate command-to-movement model can defeat this particular predictor without defeating prediction as a whole class of explanation.
The common Method has supplied a connected research result: two proposed observation-and-action arrangements, a condition under which their consequences differ, an observable contrast and the remaining alternative that changes the inference. The subject investigation determines which relations actually obtain. This application uses the same control construction with an evidential conclusion instead of an instruction to the organism.