Abstract
The evaluation of the perceptual threshold is arguably one of the most important and most difficult aspects of perception study, particularly in visual perception, where it is a pillar of the experimental process. The choice of the most appropriate adaptive procedure, and the selection of the proper parameters, are difficult but key aspects of the experimental protocol. For instance, Bayesian methods such as QUEST, require the a priori choice of a family of functions, as well as the specification of multiple parameters. Importantly, the choice of an ill-fitted function or parameters will induce costly mistakes and errors in the experimental process. We study the application of a new adaptive procedure, ZOOM (Zooming Optimistic Optimization of Models), which efficiently solves this problem. ZOOM is a state-of-the-art method of statistical learning, that uses two dual concentration inequality to form a K-nary tree and identify the perceptual threshold of interest. In previous works, ZOOM has been shown to have strong mathematical guarantees that are missing from many of its alternatives. Here we investigate the benefits of using ZOOM to identify perceptual threholds. To that end, we performed a wide range of simulations, covering many possible experimental protocols, as well as one real experiment with random dots kinematograms. Our results show that compared to existing approaches, ZOOM can be applied to any arbitrary psychometric problem, and is robust to the choice of its parameters, which do not need to be manually chosen using heuristics. Depending on the setting, ZOOM is up to three times more accurate than its counterparts, in particular for difficult psychometric functions or when the parameters of the other methods were not properly chosen. Given these advantages and its ease of use, we argue that ZOOM can improve the process of many psychophysics experiments.