Challenging standard practices in adaptive psychophysics

Abstract

One of the most cited manuscripts in behavioral hearing research is Levitt [J. Acoust. Soc. Am. 49, 467–477. (1971)]; it is typically the only reference cited in the procedures section of studies reporting thresholds obtained via adaptive tracking. This is problematic because Levitt (1971) informs only one among many parameters used in threshold estimation, and the relationship between Levitt’s transformed up-down method and the proportion correct targeted by the procedure is commonly disrupted by the selection of the remaining parameters [Vis. Res. 38, 1861–1881. (1998)]. We explored the evidence supporting standard practices and found that step size, stopping criterion, and threshold estimation from track data have limited theoretical motivation and can introduce systematic bias and error into threshold estimates as shown in behavioral and simulation studies. As a result, we propose an approach to developing and describing methods for behavioral threshold estimation that offer improved efficiency, reliability, reproducibility, and are no less consistent with Levitt (1971). [Work supported by NIH NIDCD DC015051.]