by Prof. Dr. Miguel Ángel García Pérez
The psychometric function describes how performance on a perceptual task varies with stimulus magnitude. Typically, the data follow a sigmoidal path and the analysis involves fitting a convenient and largely arbitrary function to extract shape descriptors (location and slope) that presumably capture sensory components of performance such as absolute thresholds or just noticeable differences. This practice is justified on the grounds of an implicit model that purportedly separates out the sensory components of performance, thus eliminating the contaminating influence of decisional and response components. A close look at data collected with a broad range of psychophysical methods nevertheless reveals peculiarities that are incompatible with such a model, suggesting that decisional and response components are still present in the data and, thus, undermining the interpretation of the a psychometric function as a pure reflection of the sensory determinants of performance. This talk discusses these peculiarities and presents a process model of performance that accounts for all of them. The model includes sensory, decisional, and response components captured by distinct parameters whose estimation allows effective separation of these components. A formal analysis of several classes of psychophysical methods further reveals that some of them confound these components in an inextricable way, which makes them unsuitable for research. Other types of method allow a complete separation, but this can only be achieved when data are analyzed properly. Practical recommendations for good practice will be given that summarize all these points.
All the first webinars here.