Contribution
On the validity and reproducibility of small-N psychoacoustical threshold measurements
* Presenting author
Abstract:
Over several decades, the psychoacoustical community has often published experimental threshold data measured on rather small numbers of participants. Many classical papers tested as few as only three or four participants. This has especially been the case when all subjects were normal hearing, when detection or discrimination thresholds were investigated, e.g., in classical masking experiments, and when those 3-4 subjects showed thresholds with very small inter-individual variability.Lately, it can be observed in several branches of science that many researchers believe that only large-N studies can provide valid and reproducible results. Therefore, this paper investigates the reproducibility of psychoacoustical threshold measurements by analyzing a large pool of threshold data acquired from repetitions of classical (mostly small-N) experiments. The data have been gathered by 134 groups of three students each who performed some experiments as part of their assignment in a psychoacoustics course. These experiments covered a wide range of topics, including spectral and/or temporal masking, BMLD, modulation detection and modulation masking, CMR, etc. We present a quantitative analysis of the differences between the patterns of mean thresholds in the original studies and the small-N repetitions, the intra-subject stability across repeated threshold measurements within one experiment, and the inter-subject variability.