2.4 Magnitude estimation
An advantage of the method of magnitude estimation is that no ceiling effects show up and it has (theoretically) an infinite resolution5 (e.g. Zwicker and Fastl 1999). In the procedure, pairs of sounds are presented. The first sound A is called anchor sound and the second sound B test sound. Throughout an experiment the anchor sound is kept constant and the test sound is varied. To a psychoacoustic magnitude of sound A, e.g. its loudness, a numerical value, e.g. 100, is assigned. The task of the subject is to assign to the test sound B a numerical value which represents the relation in the psychophysical magnitude (loudness) between sound A and sound B. If for example sound B is perceived 20 percent softer than sound A the subject should give the response 80. By magnitude estimates a direct relation of psychophysical magnitudes is obtained which is of advantage for cost/benefit analysis. Intra-individual as well as inter-individual differences of magnitude estimates usually are within 10 percent. However, sometimes the choice of the anchor sound may influence magnitude estimation significantly. Therefore it is suggested to use at least two anchor sounds, one with a large value of the psychophysical magnitude in question and the other with a small value. The psychophysical methods mentioned so far all have their advantages and disadvantages: Random access and semantic differential give more qualitative descriptions of sound quality. For quantitative assessment of sound quality, methods like category scaling and magnitude estimation are recommended. While traditional category scaling is confined to five step or seven step scales, magnitude estimation in principle has an infinite resolution. However, in practice, effects of the frame of reference as well as influences of the choice of the anchor sound have to be taken into account.