Discrimination indexes and models for studying sensory functioning in aging

Abstract

A popular paradigm in experimental psychophysics has subjects estimate sensation magnitude by assigning numbers to stimuli in some way. While it is typical to analyze the central tendency (e.g. means and slopes) of the subjects' psychophysical. functions, there is often a greater need to analyze the internal consistency of these functions. A subject who gives increasing mean responses across increasing stimulus intensities and also gives highly consistent responses within stimulus intensities is showing superior sensory discrimination. We propose new discrimination indexes, based on measures of association and lack-of-fit, that summarize monotonic regressions of the subject's data, as well as non-metric and metric-sensitive measures related to Kendall's coefficient of concordance. We use these indexes in quadratic spline regression models for cross-sectional age trends in sensory discrimination, with covariates included to adjust for task demands and gender differences.Because such data are potentially affected by increasing variability with age, we describe a method to assess this and adjust for it using reweighted least squares.