*4.4. Comparison and Relationship of Methods*

To compare the conscious methods of sensory analysis with the unconscious method, the overall liking scores (from hedonic scales) were used as a reference method. This comparison was performed to determine the method that best represents the variance observed and to provide the most accurate consumer insights.

The conscious methods, CATA emotions and emojis, explained more variation in the consumer data—67.8% and 66.0%, respectively. The emojis model also explained the variation for the cultural differences (*p* < 0.05), which could not be explained by the other two models. The unconscious method (FER) explained quite a low level of variation within the data (8.8%). Only two terms were relevant in the model, which were 'surprised' and 'disgusted'. This outcome suggests that self-reported methods can be used alone with a hedonic scale to explain the variability in consumer response. In contrast, facial emotion recognition should be used in combination with other techniques for a better estimation of variance across consumer responses.

The MFA plot, shown in Figure 4, explains the relationship between the hedonic scales and conscious and unconscious methods of sensory analysis for the six yogurt types. The plot shows an association of positive terms with the overall liking scores and shows negative terms to be opposite to liking. In a beer tasting study [22], unconscious biometric responses were successfully integrated with conscious sensory responses for predicting consumer liking of beers. However, in another study by van Bommel et al. [38] where implicit facial expressions were compared with the explicit self-reported emotions, little overlap was found between methods, which were not directly comparable. A sensory study of organic herbal infusions [8] also found similarities between the conscious responses chosen by the consumers (EsSense Profile®) and unconscious responses from their facial expressions and successfully characterized emotional responses. Negative emotions were better identified by FaceReaderTM and positive emotions were better displayed with an EsSense Profile®. However, there are differing responses in the literature, as another study, involving a comparison of self-reported and implicit responses for beer, found liking scores to have a higher discrimination compared to facial expression and also found self-reported emotional responses to display the highest discrimination compared to the other two methods [39]. Another study with juices also found that a higher correlation of negative emotions from facial expression analysis was observed with hedonic liking measurements [40].

The price perception in this study for the most liked yogurt (cookies) and least liked yogurt (berries) was directly related to the liking predicted by all three sensory methods. The plain plant soy yogurt and the reference dairy yogurt were similarly liked by the participants, indicating that liking does not necessarily depend on the source of protein. However, the different food textures of the yogurts evoked different emotions in our study. Reports have shown that emotions are more linked to the intrinsic properties of a food product and can vary accordingly, even if the liking scores are the same. The relation between the emotions and liking is not straightforward [41], also confirmed in our study with yogurts. Further, familiarity or neophobia is an important parameter that plays an significant role in the representation of emotions across cultures [42]. In a study with Greek yogurts, different textures produced varying consumer responses and liking [43]. Hence, it is important to understand the emotions linked to the tasted yogurts, as in our case, to understand food choice. Studies have shown that different cultures understand the same set of emotions, expressing these differently [33]. Other structural and compositional factors may also affect liking scores, which were not explored in detail in this study.
