Related Experiment Videos
Response distributions in intensity resolution and speech discrimination
1Research Institute for Language and Speech, University of Utrecht, The Netherlands.
The Journal of the Acoustical Society of America
|November 20, 1998
Summary
This study tested the assumption of equal response distributions for calculating sensitivity (d-prime). While distributions were unimodal, variances differed, but this is unlikely to significantly impact d-prime calculations.
Area of Science:
- Auditory perception
- Psychophysics
- Signal detection theory
Background:
- The accuracy of signal detection theory (d-prime) relies on assumptions about response distributions.
- Specifically, it assumes equal, Gaussian distributions for stimuli responses.
Purpose of the Study:
- To empirically test the assumption of equal, Gaussian response distributions for d-prime calculation.
- To investigate these distributions using auditory stimuli: pure tones and stop consonants.
Main Methods:
- Directly measured response distributions using a non-numerical magnitude estimation technique.
- Subjects rated stimulus positions on a quasi-continuous scale.
- Tested stimuli included 1000-Hz tones varying in level and interpolated stop consonants (/p/, /t/, /k/).
Main Results:
- Response distributions were generally unimodal for both tone and consonant stimuli.
- A key finding was that the variances of these distributions differed across stimuli.
- Despite differing variances, distributions remained sufficiently unimodal.
Conclusions:
- The assumption of equal variances in response distributions for d-prime calculation is often violated.
- However, the observed differences in variance are unlikely to pose serious problems for estimating d-prime.
- The findings support the continued use of d-prime in auditory research despite potential violations of strict assumptions.