OLFACTORY PERCEPTUAL SPACE MODELS COMPARED BY QUANTITATIVE METHODS

The long time constants of the olfactory system associated with the numerous judgements required by paired comparison designs produce expensive experiments in terms of time and manpower. Hence, when odor quality perception studies are reported, they tend to receive extended scrutiny. This report applies newer methods of multidimensional scaling to data reported by M. Woskow in 1964. The analysis includes quantitative methods which compare the results of the scaling methods used. The INDSCAL algorithim by D. Carroll and J. Chang and the POLYCON algorithm by F. Young were compared together with Woskow's original factor analysis. The INDSCAL solution was not remarkably differenct from Woskow's original results. The POLYCON solution agreed closely with INDSCAL on two of three dimensions in the three dimensional solution. The INDSCAL method offers a description of individual differences in perception, and indicated that four of Woskow's 20 subjects may have been substantially anomalous perceivers of the odorants used. The quantitative methods used to compare different solutions based on the same data can be extended to solutions based on different sets of data. This report proposes a method whereby perceptual models may be combined to generate a composite perceptual space based on several different data sources, be they experiments, laboratories, or replications. © 1979 Information Retrieval Ltd, 1 Falconberg Court, London WIV 5FG, England.