How to map the affective semantic space of scents

Abstract

The investigation of the semantic space associated with subjective affective experiences or feelings linked to odour perception has recently emerged. Because of the specificity of the emotional effects of odours, the terms derived from traditional models of emotion are unlikely to optimally account for odour-associated feelings. In this study, sets of terms derived from two traditional models, basic emotions and valence by arousal by dominance dimensional emotions, were compared with a recently elaborated olfaction-specific set of terms (Geneva Emotion and Odour Scale; GEOS). Three main criteria were considered: (1) the feeling's intensity reported in response to odours; (2) the inter-rater agreement concerning the reported feelings; and (3) the power to discriminate feelings evoked by various odorous substances. The evidence strongly suggested that the set of terms proposed by GEOS outperformed the terms derived from the two classical models in measuring the subjective affective experience elicited by odours. These results are interpreted with respect to a good correspondence between the functions of olfaction and the meaning conveyed by GEOS terms.

Keywords:

• Emotion
• Feeling
• Olfaction
• Odour
• GEOS

Acknowledgements

This research was supported by the National Center of Competence in Research (NCCR) for the Affective Sciences, financed by the Swiss National Science Foundation (51NF40-104897) and hosted by the University of Geneva, and by grants from Firmenich, SA, to DS and PV.

Notes

¹ Stricto sensu, basic emotions share a common set of characteristics that consist of specific neural, bodily, expressive, and feeling components (Izard & King, 2009; Matsumoto & Ekman, 2009). Consequently, using terms describing emotions without bringing arguments about the “basic” nature of those emotions, as described earlier, means that one refers to the concept of basic emotions too roughly, without fully considering the underlying assumptions carried by this theory, omitting many required criteria for an emotion to be considered as basic.

²Traditionally, only three scales are used to assess the three dimensions of arousal, valence, and dominance. For instance, participants are requested to evaluate their feelings on continuum scales varying from relaxed to excited, unpleasant to pleasant, and without control to under control, respectively. We decided to represent each dimension by two scales and not one in order to equalise the dimensionality of the three models. Indeed, comparing six-scale questionnaires with a three-scale one could have been criticised from a statistical point of view and would have automatically biased the result of the comparisons (e.g., a decrease in dimensionality will automatically reduce the inter-rater agreement because the Cronbach's α increases as a function of the number of items).

³City-block (Manhattan) distance is simply the mean difference across dimensions. In most cases, this distance measure yields results similar to the simple Euclidean distance. However, in this measure, the effect of single large differences (outliers) is dampened because they are not squared. Ward's method uses an analysis of variance approach to evaluate the distances between clusters. In short, this method attempts to minimise the sum of squares of any two (hypothetical) clusters that can be formed at each step. In general, this method is regarded as very efficient; however, it tends to create clusters of small size (see Electronic Statistics Textbook for further details; StatSoft, Inc., 2011).