Flexibility

Flexibility is not a virtue in a model. Although a theorist may be reluctant to specify a model completely until more data are available, a model that is capable of accounting for many different patterns of behavior is much less useful for predicting outcomes than one that is inflexible. Formally specified theories differ in the number of patterns of behavior that can be predicted, and the ones with the more limited range of predictions are to be preferred. This is a basis for seeking simple, parsimonious theories.

For example, in one case a scalar timing model with two free parameters was judged to be superior to one with five free parameters (Church and Gibbon, 1982).

Quantitative fits are much more satisfactory than qualitative fits that are based on a judgment that the general shape of the observed functions corresponds to the predictions of the theory. A problem with qualitative fits is that they are overly flexible.

With many potential sources of variability in the clock, memory, and decision processes, scalar timing theory has considerable flexibility. Additional flexibility comes from the use of different assumptions regarding sampling from memory and number of thresholds, and the use of different assumptions for different procedures (Brunner et al., 1997). Scalar timing theory has a large number of parameters that can be varied, but typically only a few parameters are varied to account for a large number of observed data points.

0 0

Post a comment