Human odor detectability: New methodology used to determine threshold and variation

Abstract

Current ambiguity concerning the related issues of optimal means for measurement of odor sensitivity and the functional properties of the olfactory system hinders progress in basic and applied research on the human sense of smell. To address these needs, we selected n-amyl acetate (nAA) as a test odorant and developed a methodology in which participants (Ps) receive multiple presentations each session of several concentrations. Yes-no responses as to whether odor was detected are analyzed using binomial statistics, with the probability that a given proportion of yes responses (or greater) would occur by chance alone being treated as the inverse of detectability. Over the course of multiple sessions, this information is also used to maximize the collection of data in the peri-threshold region. Surprisingly, data collected over as many as 14 sessions were fit well by a single logistic regression model relating probability and concentration. Threshold concentrations, defined as those corresponding to a probability of 0.05, varied from 7.11 to 167.53 p.p.b. (v/v) for 11 Ps. Our approach and findings, if shown to be representative of other combinations of Ps and odorants, could accelerate the pace of research in human olfaction by providing a comprehensive operational definition of the limit of the olfactory system to detect odorant molecules. © Oxford University Press 2003; all rights reserved.