In some experimental situations, the psychometric function underlying performance may not be stable, but instead may shift along the stimulus axis in response to changes in attention, learning, or task difficulty. When this occurs, the measured threshold may be influenced and the slope of the measured function will be inaccurately shallow. With commonly used experimental procedures, it is difficult to know whether a shallow psychometric function slope is a true reflection of the sensory process, or is a result of “averaging” a highly variable underlying function. Here, a new method is described of estimating psychometric function slope from the variability in two interleaved adaptive tracks, consulted on alternate trials, that is resistant to the effects of shifting performance levels. Further, a mechanism is described for assessing the likelihood that a threshold was, in fact, stable over the course of its measurement. Computer simulations are reported as well as verification of the method in measurements of human performance on a psychophysical task. Several conditions of externally imposed variability were simulated to establish the ability of these procedures to identify unstable functions and produce accurate slope estimates. The procedures worked well for thresholds shifting by as little as 4 dB if the variation did not occur too rapidly. The procedure and associated analyses are recommended as a relatively “free” means of calculating slope and quantifying threshold reliability with little extra experimental effort.
ASJC Scopus subject areas
- Arts and Humanities (miscellaneous)
- Acoustics and Ultrasonics