Both on technical grounds and from experience in industrial noise measurements, it is recognized that the decibel scale, used as a loudness scale, does not yield numerical values proportional to the loudness sensation. In many spheres of work this occasions great inconvenience and there is need for an agreed rule for converting the decibel observations of practical noise measurements into loudness values on a linear loudness scale. A survey has therefore been made of published data with a view to deducing the most probable law for the relation between stimulus and sensation for a 1000‐cycle pure tone. Relations derived from mental estimates of loudness, from data on monaural‐binaural listening and from data on the properties of tone combinations are discussed. The evidence indicates that not only does the nerve‐impulse theory of audition provide an explanation of the faculty of making mental estimates of relative loudness and of the monaural‐binaural listening phenomena, but that stimulus‐sensation relations deduced by the two methods are in substantial quantitative agreement. On theoretical grounds it is not thought that tone combination data are admissible for this purpose and an important divergence between these and monaural‐binaural data appears to exist. Consequently the suggested relation put forward is not felt to be the most probable law. For the final relation derived in the present paper from the available data, the number 100 is assigned to the loudness evoked by an intensity of 100 db above 0.0002 dyne per sq. cm reference pressure. For practical industrial noise measurements, the relation is expressed sufficiently nearly by a simple power law between the limits of 30 and 115 db. Over this range, the loudness (L) is given byL = d5×10−8, wheredis the number of decibels above reference pressure. In addition to its being more in conformity with introspective judgments of relative loudness, this simple rule is thought to be preferable to the suggested curve because it is easily remembered, accurately definable and easily used with the aid of a slide rule.