Development of Psychophysical Methods for Estimating Human Odour Detection Thresholds

Abstract

Human odour detection thresholds play a significant role in both fundamental and applied science. The psychometric function (PF) – quantifying the relationship between sensory response and physical stimulus intensity – provides an appropriate approach for acquiring threshold estimates in modern psychophysics. Fitting a conventional PF with intercept (midpoint threshold) and slope (rate of increase) parameters requires a substantial quantity of data. This approach is therefore uncommon in odour threshold estimation because of the high susceptibility of olfactory receptors to adaptation and fatigue. The American Society for Testing and Materials (ASTM) provide two standard procedures – E679 and E1432 – for odour threshold estimation. The data collection method outlined by these two procedures is proven effective for circumventing complications arising from prolonged testing. However, the theoretical underpinnings pertaining to the threshold calculation methods have been questioned openly in the literature. This thesis assessed the ASTM procedures for estimating thresholds using 10 chemical odorants, which are common in food and beverage products, for a large group of judges. ASTM E679, which has been queried for ignoring bias effects caused by guessing, was found effective for group threshold estimates but not individual estimates. ASTM E1432, which prescribes fitting a conventional PF based on five to eight replications, was found unreliable for estimating individual thresholds. A new threshold calculation method was then proposed to address the shortcomings with these standards: fitting the PF with the slope parameter fixed at a specific value. By circumventing estimation of the slope parameter, the number of replications required for fitting a PF reduced considerably. Justification for use of a constant slope hinges on empirical observations of only a slight difference in the slope parameters of the conventional functions across individuals. The fixed-slope PF was first implemented using the slope of the group PF (the group-slope PF), and then developed further using the slope that gives the best overall goodness-of-fit (the optimal-slope PF). Both the existing and developed methods were finally assessed for reliability and concurrent validity using an independent dataset. Favourable results were obtained for the optimal-slope PF. Overall, this thesis describes the development of a novel method for odour threshold estimation, based upon existing standards – ASTM E679 and E1432. The final developed method – the optimal-slope PF – was demonstrated to be the best candidate for producing reliable and valid threshold estimates within the practical constraints of olfactory data collection. Development of this method has both fundamental and applied importance. In addition to enhancing accuracy in measuring olfactory sensitivity, application of this method helps align odour threshold estimation with an established psychophysical framework.