Detection of a Noise Signal of Varying Duration

Abstract

The ability of human observers to detect a thermal-noise signal (S) as a function of the duration of S presented in a thermal-noise background has been studied. Subjects were asked to correctly identify, in a two-alternative forced-choice situation, the time interval in which S occurred. Signal-to-noise ratios in dB based on a steady-state voltage (S/N), were varied according to the subject's performance on a block of 12 trials. The S/N was varied in 1-dB steps until a 75% correct-detection level was maintained. This Block Up and Down, Two-Interval, Forced-choice (BUDTIF) procedure is described in detail. Threshold S/N's obtained decreased linearly with increase in log duration from 1.0 to 100 msec, then approached minimum of −9.4 dB at 1000-msec duration. When these same threshold S/N's are converted to increments [(S+N)/N] and expressed in dB a linear decrease with log time from 1.0 to 47 msec is seen with an abrupt discontinuity, then a further nearly linear decrease from 47 to 1000 msec. The question is raised whether S/N or [(S+N)/N] is the effective stimulus to the subject.