A behavioristic analysis of Stevens magnitude estimation method

A procedure independently proposed by Oyama and Lewis for treating the data of magnitude estimation is discussed in relation to the Graham-Stevens controversy on the psychophysical methods. In this procedure, the critical value of the stimulus needed to make the subject report numbers equal to or greater than each integral number is sought, instead of the median or geometrical mean of numbers reported for each level of stimulus intensity. Stevens’ magnitude estimation method has no special advantage over the other cross-modality matching methods, and the subjective magnitude of number should be estimated if one wants to know real relations between stimulus intensities and sensory magnitudes in various modalities.     

Efficient estimation of sensory thresholds

Laboratory computers permit detection and discrimination thresholds to be measured rapidly, efficiently, and accurately. In this paper, the general natures of psychometric functions and of thresholds are reviewed, and various methods for estimating sensory thresholds are summarized. The most efficient method, in principle, using maximum-likelihood threshold estimations, is examined in detail. Four techniques are discussed that minimize the reported problems found with the maximum-likelihood method. A package of FORTRAN subroutines, ML-TEST, which implements the maximum-likelihood method, is described. These subroutines are available on request from the author.     

The measurement of loudness using direct comparisons of sensory intervals

Subjects were required in each trial to directly compare two pairs of tones and indicate which pair of tones had the greater loudness difference. Ten 1200 Hz tones differing only in intensity were employed. Subjects made binary comparisons among the 45 tone pairs which can be formed from the set of ten tones. The subjects' binary comparisons of the tone pairs were found to satisfy the transitivity and monotonicity requirements of a positive difference structure. These comparisons of loudness intervals were used to construct a rank order of loudness difference. A loudness scale was constructed from a nonmetric analysis of the rank order of loudness difference for the 45 tone pairs and indicated that loudness was a power function of sound pressure with an exponent of 0.26.