Effect of calculation methods and presence of false positives on perception thresholds
Abstract
Perception thresholds are commonly used in food science to discriminate between those compounds that are more or less odour-active. It is not rare to use values from other studies and apply them to our data, however, we observed that the procedure to evaluate and calculate perception thresholds is not standardised. The aim of this study was to explore the influence of calculation methods on threshold values. To do so, we used a 3-AFC sensory procedure to determine the orthonasal and retronasal detection values of 26 compounds, amongst them some compounds that usually contribute to the aroma of strawberry and caramel (e.g. vanillin, Furaneol and damascenone). The thresholds were calculated by either Best Estimate Threshold (BET) method or by logistic regression. Moreover, an algorithm for the removal of false positives was applied to adjust the assessors' raw responses (raw and adjusted data). The results showed significant effects (p<0.05) from the different methods but also from the removal of false positives. Statistical analyses, like PCA, showed a linear dependence between the four calculation procedures, the highest values being obtained from BET/adjusted data, followed by logistic regression/adjusted data and BET/raw data, and logistic regression/raw data as the lowest. A similar effect of calculation method and presence of false positives was observed for both orthonasal and retronasal detection thresholds. These results proved that calculation methods and data treatments, such as the removal of false positives responses, have a non negligible effect on final threshold values and, for this reason, that special attention must be paid when using values from the literature in order to not under- or overestimate the odour activity of a compound in a food.