Operating Characteristics from Yes-No and Forced-Choice Procedures

Abstract

Operating characteristics were obtained from yes-no (YN) and two-alternative forced-choice (2AFC) experiments in auditory signal detection. Four listeners used a six-point scale of confidence ratings with each of these psychophysical methods. An extreme rating (1 or 6) indicated high confidence that the signal was or was not presented (in YN) or that the signal occurred in the first- or second-observation interval (in 2AFC). Note that response bias, occasionally exhibited by listeners in 2AFC, and which affects the proportion of correct judgments when conventional binary decisions are made, is of lesser concern when confidence ratings are used and operating characteristics obtained. The signal was a sinuosid of 1000 cps presented for 0.1 sec against a continuous background of noise. Testing was conducted at E/N0's of 7.9 and 15.8. The values of the detectability parameters of the operating characteristics obtained in 2AFC were generally very close to predicted values based upon performance in YN. It would be hazardous, of course, to predict d′ in YN on the basis of performance in 2AFC, since it is not always safe to assume that the operating characteristic to be obtained in YN will have unit slope when plotted on normal-normal coordinate paper.