It is usually assumed that, in magnitude estimation and production procedures, subjects assign numbers to sensation magnitudes on ratio scales, i.e., according to ad hoc units. However, S. S. Stevens and others have found that units imposed by the experimenter produce response biases, and Hellman and Zwislocki were able to explain a strong response bias in some of their data by assuming that their subjects tended to pair numbers and loudness magnitudes on an absolute scale. This implies that, through common usage, numbers become associated with psychological magnitudes on a one‐to‐one basis, and a natural unit develops. Some subsequent experiments confirmed this hypothesis. Present experimental results of loudness scaling performed by several groups of subjects with varying degrees of sophistication reveal an unexpectedly small variability of the median scale units used by the groups and asymmetrical learning effects.

This content is only available via PDF.