HomeEducationDoctorate (PhD & EngD)For current candidatesPhD infoUpcoming public defencesPhD Defence Daisuke Kaneko | Methods to assess food-evoked emotions across cultures

PhD Defence Daisuke Kaneko | Methods to assess food-evoked emotions across cultures

Methods to assess food-evoked emotions across cultures

The PhD defence of Daisuke Kaneko take place (partly) online.

The PhD defence can be followed by a live stream.

Daisuke Kaneko is a PhD student in the research group Human Media Interaction (HMI). Supervisor is prof.dr. J.B.F. van Erp from the Faculty of Electrical Engineering, Mathematics and Computer Science (EEMCS).

Traditionally, consumer science groups and food companies use self-report measures, in which people indicate how much they like a certain food after having tasted it, to predict consumers’ future buying behaviors. However, it has been suggested that food-evoked emotion may improve prediction of individuals’ food choice. It may reveal previously unknown product attributes that go beyond traditional sensory and acceptability measurements, and can be a valuable source of information for product development and marketing. Although many research groups acknowledge that the evaluation of food-evoked emotions is probably important to predict food choice behavior, and although many different emotion-association questionnaires and measures have been developed and used, it is still unclear how to best assess food-evoked emotions.

This thesis investigated what (type of) measures are used to evaluate food-evoked emotions and to what extent these measures are able to capture food-evoked emotions. Also, we explored the added value of the different measures and their combination, all of these specifically in a cross-cultural context, where the same foods can be either novel or familiar, and where measures other than explicit, verbal measures are expected to be especially valuable. A new cross-cultural food image database and pictorial emotional response tool were developed. This thesis contributed to the methodology to more comprehensively and reliably measure food related emotions in cross-cultural contexts. To achieve the goals, this thesis consists of two sections, ‘Measures to evaluate food-evoked emotion’ and ‘Measures of food-evoked emotion in cross-cultural context’.

In the first part of the thesis, the following topics are addressed; what (type of) measures are currently used to evaluate food-evoked emotion (Chapter 2), to what extent explicit (‘conscious’) and implicit (‘unconscious’) measures are able to capture food-evoked emotion by tasting drinks (Chapter 3), whether combining implicit measures in a model improves the sensitivity (Chapter 4), and how emotional context affects pleasantness of novel and familiar drinks as measured using both explicit and implicit measures (Chapter 5).

The literature review in Chapter 2 shows that previous studies predominantly rely on explicit ratings and questionnaires or a combination of redundant explicit measures to assess food-evoked emotions. We recommend researchers to use a combination of physiological, behavioral, and cognitive measures that covers all three emotional processing levels by providing a 3 x 3 complementary toolbox table.

The study in Chapter 3 provides an overview of the sensitivity of simultaneously measured explicit and implicit responses to regular drinks and a vinegar solution. It shows that besides explicit measures of valence (pleasantness) and arousal, most implicit measures are sensitive to large differences in affective experience (regular drinks and the vinegar solution), where sip size had the largest discriminative power. HR was sensitive even to subtle differences in affective experience as indicated by a correlation between explicit arousal ratings and HR. Chapter 4 shows that a predicting model trained by using implicit responses (especially sip size and HR) to the vinegar solution as a ground truth high arousal-low valence drink, and to regular drinks as low arousal-high valence drinks, result in quite accurate estimates of explicit arousal scores assigned to subtle different regular drinks. In contrast, the traditional approach of using explicit valence and arousal ratings as a ground truth to train the model failed to provide sophisticated predictions.

The study described in Chapter 5 examines the effect of emotional context on the pleasantness of novel and familiar foods using different types of measures. It shows that familiar soups were rated more positively than novel soups in a negative emotional condition whereas both soups were equally rated in a positive emotional state. The effect of emotional state was robust over at least one week. Besides explicit ratings, implicit behavioral measures (sip size and willingness-to-take-home) indicate similar patterns of effects of soup, state, and session.

The second part of the thesis addresses measures of food-evoked emotion in a cross-cultural context. It describes the development and validation of the food image database CROss-CUltural Food Image Database (CROCUFID)” (Chapter 6), the development and validation of a language independent tool with Emoji to evaluate food-evoked emotions (so called EmojiGrid) (Chapter 7), to what extent this EmojiGrid can reliably assess food experience across cultures (Chapter 8), and the potential of explicit and implicit measures to cross-culturally assess food-evoked emotions (Chapter 9).

Chapter 6 introduces a new food image database (CROCUFID) to facilitate cross-cultural study in food research to meet the demand for non-western and low valence food images. Validation data for the CROCUFID food images were obtained from a total of 805 participants (ranging in age between 18 and 70) with different cultural background (United Kingdom, United states, and Japan). CROCUFID now contains more than 980 standardized food and non-food images, covering some of Asian and Western cuisines. The database is still growing through contributions of researchers around the world.

In Chapter 7 and 8, a new graphical tool with Emojis called “EmojiGrid” is introduced and validated through cross-cultural studies. The study in Chapter 7 shows that each of 16 self-designed Emojis conveys the intended degrees of valence and arousal using a correlation analysis between these Emojis and a conventional graphical tool called SAM scale. Also, this chapter shows that the EmojiGrid is an intuitive and efficient graphical tool to capture the affective dimensions of valence and arousal for food image stimuli. Chapter 8 further explores the potential of EmojiGrid for cross-cultural studies in respondents with different mother tongues (respondents from Germany, the Netherlands, United Kingdom and Japan). This was done through an online experiment including 60 food images covering a wide range of valence. The study shows that as expected, the EmojiGrid produces the nomothetic U-shaped relation between valence and arousal ratings on the same food image set for all four different countries, where the pattern of the Asian country (Japan) differs from the other, western countries. The EmojiGrid appears to be a valid and language-independent self-report tool for cross-cultural research on food-related emotions.

Chapter 9 investigates the potential of implicit measures to capture food-evoked emotions in across cultures. Participants from Thailand (representative for Middle Response Style (MRS) in explicit measures) and the Netherlands (representative for an Extreme Response Style (ERS) in explicit measures) were presented with food images and drinks from their own and other culture. In addition, images of universal and low valence (disgusting) food were shown. This study shows the expected explicit response bias in Thai and Dutch participants, while implicit measures of HR (in case of food images) and sip size (in case of drinks) followed the prior expectations of genuine food experience. The study suggests that physiological responses may complement self-reports in that they are insensitive to cultural response bias that could lead to misunderstanding across cultures.

The thesis ends with Chapter 10, discussing the main findings of the presented studies as a whole, as well as the important aspects from a commercial point of view. We emphasize that besides explicit measures, it is recommended to combine implicit (neuro)physiological and behavioral measures to better understand food-evoked emotions. Also, we newly developed a validated food image database (CROCUFID) and an intuitive language-independent affective evaluation tool (EmojiGrid) to facilitate cross-cultural studies in food research. Throughout the thesis, implicit measures and combining implicit measures show their potential to capture food-evoked emotions, especially arousal, without asking explicitly. Among them, HR and sip size may be the most promising implicit responses to measure emotions evoked by food experiences. However, new devices that are able to remotely measure implicit (neuro)physiological and behavioral responses without interrupting individuals’ daily life are eagerly anticipated for further investigations on food-evoked emotions.