Abstract
One of the most precise methods to establish psychometric functions and estimate threshold and slope parameters is the constant stimuli procedure. In this procedure the distribution of stimulus values presented to the participants is specified before the experiment. This distribution is composed of a large range of stimulus values, which enables to fully develop the psychometric function, but requires a large number of trials. Adaptive procedures (e.g., non-parametric staircases, Bayesian procedures...) enable reliable threshold estimation while reducing the number of trials by concentrating stimulus presentations around observers’ supposed threshold. Here, the stimulus value for the next trial depends on observer’s responses to the previous trials. One recent substantial improvement of Bayesian procedures is to also estimate the slope (related to discrimination sensitivity). The Bayesian Quest+ procedure (Watson, 2017), a generalization and extension of Watson and Pelli’s (1983) Quest procedure, includes this refinement. Surprisingly, this procedure is barely used. Our goal was to assess the efficacy of this procedure empirically, in four yes-no tasks evaluating size (Expt. 1A and 1B), orientation (Expt. 2), or temporality (Expt. 3). We compared points of subjective equivalence (PSEs) and discrimination sensitivity obtained in 91 adult participants in total with the Quest+, constant stimuli and simple up-down staircase procedures. While PSEs did not differ between procedures, sensitivity estimates obtained with the 64-trials Quest+ procedure were overestimated (i.e., just noticeable differences, or JNDs, were underestimated). Correlations between PSEs and between JNDs of the Quest+ and constant stimuli procedures were significant in three experiments (r range: .52–.85, p range: <.001–.024). Overall, this study empirically confirmed that the Quest+ procedure can be considered as a method of choice to accelerate PSE estimation, while keeping in mind that sensitivity estimation should be handled with caution.