Tests of human olfactory function: principal components analysis suggests that most measure a common source of variance

Abstract

It is not known whether nominally different olfactory tests actually measure dissimilar perceptual attributes. In this study, we administered nine olfactory tests, including tests of odor identification, discrimination, detection, memory, and suprathreshold intensity and pleasantness perception, to 97 healthy subjects. A principal components analysis performed on the intercorrelation matrix revealed four meaningful components. The first was comprised of strong primary loadings from most of the olfactory test measures, whereas the second was comprised of primary loadings from intensity ratings given to a set of suprathreshold odorant concentrations. The third and fourth components had primary loadings that reflected, respectively, mean suprathreshold pleasantness ratings and a response bias measure derived from a yes/no odor identification signal detection task. In an effort to adjust for potential confounding influences of age, gender, smoking, and years of schooling on the component structure, a matrix of residuals from a multiple regression analysis, which included these variables, was also analyzed. A similar component pattern emerged. Overall, these findings suggest, in healthy subjects spanning a wide range, that (1) a number of nominally distinct tests of olfactory function are measuring a common source of variance, and (2) some suprathreshold odor intensity and pleasantness rating tests may be measuring sources of variance different from this common source.