Is Intermittent Updating of Probability Estimates Due to Motor Costs?

A decade ago an excellent article with an educational reminder of some forgotten history was published by Gallistel and colleagues.¹ The main method was to observe the trial by trial reports of participants estimating the Bernoulli parameter while observing a sequence of "coin flips" with intermittent changes in the underlying, true, parameter. The main conclusion was that participants were not doing trial-by-trial updating, because their reports showed a step-and-hold pattern of updating. What Julia Schirmeister investigated was a possible motor confound, identified by Forsgren in their rebuttal.²

It takes more effort in the traditional administration of the task to update one's estimate than to leave it be. So, a reasonable strategy might be to tolerate a little bit of imprecision and save yourself some work (especially if you are an undergraduate participating for course credit and have a goal to minimize time on task as much or more than accuracy). While the basic data are in her thesis Julia has continued to refine her analysis. A manuscript is under review, but you can get a recent take on it by looking at her poster she recently presented at the 2024 Cognitive Neuroscience Society meeting that was held in Toronto.