Determining odour detection thresholds: Incorporating a method-independent definition into the implementation of ASTM E679

ASTM international standard practice E679 prescribes the use of a 3-alternative forced-choice (3AFC) method with an ascending concentration series for measuring odour detection thresholds. It recommends obtaining an individual's threshold by geometrically averaging the concentrations at which the judge's detection response reverses from incorrect to correct. The legitimate reversal point is defined by a stopping rule. Previous researchers have identified some methodological flaws with this calculation approach, relating to the use of stopping rules and the method's unconventional definition of thresholds. The current study aims to empirically investigate these issues in more depth. Thresholds for three odorants were obtained for 100 judges. 1,8-Cineole, isobutyraldehyde and beta-damascenone were tested in accordance with the data collection procedure of ASTM E679. Initially, thresholds were estimated by the ASTM-based method with four different stopping rules. These estimates were subsequently compared against thresholds estimated by an alternative method; that of fitting psychometric functions. The theoretical grounds on which the latter method is based, and the clear connection between performance and stimulus concentration that it demonstrates, provide a thorough comparison of both conventionally-defined thresholds and method-independent thresholds against the ASTM-based method. Comparisons between the thresholds resulting from the various estimation methods demonstrated (1) the choice of stopping rule had a significant effect on threshold estimates (P < .05); (2) each stopping rule measured a specific level of performance, indicating a useful correspondence between the stopping rules and the psychometric function. These findings provide supporting evidence for the continued use of the ASTM-based method: albeit, if the stopping rule is standardized. One of the tested stopping rules (Stop 3) provided threshold estimates consistent with those estimated for d' = 1.0, and is therefore recommended for future research. (C) 2012 Elsevier Ltd. All rights reserved.