Noncombination of pitch and loudness in multidimensional scaling.

Carvellas and Schneider [J. Acoust. Soc. Am 51, 1839--1848 (1972)], using magnitude estimates of the dissimilarity of sinusoidal tones in a multidimensional scaling program, found that their data fit a city-block and a Euclidean metric equally well, but other research has indicated that subjects do not combine dimensions of sinusoidal tones to arrive at an overall estimate of stimulus similarity. Yet multidimensional scaling seems to require subjects to combine dimensions of similarity. In this leter a model is put forward to demonstrate that subjects need not combine dimensions directly in order to generate dissimilarity judgments that are very close to either the city-block or Euclidean metrics. Thus it is possible that while subjects do not combine the dimensions perceptually they can adopt a cognitive strategy that leads one to believe they do.