On the efficiency of 2IFC threshold measurement

A tone intensity-discrimination experiment compared the efficiencies of two threshold-tracking rules used with two-interval, forced-choice (2IFC) methodology. Half of the time, a two-right, one-wrong (“2-1”) level-change rule [H. Levitt, J. Acoust. Soc. Am. 49, 467–477 (1971)] tracked thresholds with a binomial probability of 0.707. The remainder of the data were collected under a five-right, two-wrong (“5-2”) rule which tracked 0.736. Stimuli were 1000 Hz tone bursts (100 msec) presented binaurally over headphones at 70 dB SPL. A fixed step size for changes in ΔI (2 dB) was maintained by adding the increment and standard channels in quadrature. Thresholds were determined for sessions of 40, 55, 70, 85, 100, and 115 2IFC trials. Six thresholds were taken at each session size, for each of the two level-change rules. Thresholds were computed as the mean level of ΔI revisited. Sweat factors [M. M. Taylor and C. D. Creelman, J. Acoust. Soc. Am. 41, 782–787 (1967)] were calculated and averaged across the seven subjects for each of the 12 experimental conditions (session-size x rule). The 2-1 rule generated smaller sweat factors, averaged across sessions, and was thus the more efficient rule. An analysis of variance yielded the additional finding that psychophysical efficiency was independent of session size.