The invasion of psychoacoustics by inferential statistics

The basic results that form the psychoacoustic foundation of our current knowledge of auditory perception, collected over nearly 50 years, were generated without any use of null-hypothesis significance testing (inferential statistics). The common use of inferential statistics in psychoacoustic articles published in JASA did not occur until the 1980s. Yet, today it is almost impossible to publish a psychoacoustic article in JASA, or anywhere else, without using inferential statistics. In addition, recent controversies concerning replication in the social and behavioral sciences are beginning to encroach on the field of psychoacoustics. In this presentation, a case will be made that several elements of the experimental design inherent to many, if not most, psychoacoustic experiments often make the use of inferential statistics unnecessary. These aspects (many derived from the framework of the Theory of Signal Detection) include careful experimental control of stimuli and response acquisition, ratio-scale measurements, experienced subjects, large number of trials per subject, within-subject (repeated measures) experimental design, and strong theoretical context (e.g., an ideal observer). Some scenarios of experimental design and data presentation will be considered, suggesting that inferential statistical analysis might not be needed in some psychoacoustic research. [Work supported by NIDCD and Facebook Reality Labs.]