The impact of evoked consumption contexts and appropriateness on emotion responses

Food Quality and Preference 32 (2014) 277–288

Contents lists available at ScienceDirect

Food Quality and Preference journal homepage: www.elsevier.com/locate/foodqual

The impact of evoked consumption contexts and appropriateness on emotion responses Betina Piqueras-Fiszman a,b,⇑, Sara R. Jaeger c a

Department of Engineering Projects, Universitat Politécnica de Valencia, Camino de Vera s/n 46022, Valencia, Spain Department of Experimental Psychology, University of Oxford, South Parks Road, OX1 3UD Oxford, UK c The New Zealand Institute for Plant & Food Research Ltd., Mt Albert Research Centre, Private Bag 92169, Victoria Street West, Auckland 1142, New Zealand b

a r t i c l e

i n f o

Article history: Received 9 April 2013 Received in revised form 3 September 2013 Accepted 6 September 2013 Available online 17 September 2013 Keywords: Emotion responses Consumption context Appropriateness Consumer research Methodology

a b s t r a c t Sensory and consumer scientists are seeing the need to explore broader measures, such as product-driven emotions, to be able to fully understand and satisfy consumers’ demands. However, to date, the consumer methodologies used in food-related emotion research have neglected a crucial element linked to emotions: the consumption context. This is a concern, because consumers’ emotional responses to food are elicited by many other sources apart from the sensory properties of the food itself. The present research aimed to fill in this gap of knowledge by exploring how evoked consumption contexts affect responses obtained from an emotion questionnaire. Two food stimuli (apple and chocolate brownie) and three different consumption contexts (a ‘weekend breakfast’, an ‘afternoon snack’, and ‘after a special dinner’) were used to vary the level of product-context appropriateness. A within-subject design was used in which 157 participants first completed an emotion questionnaire (36 terms, checklist format) in response to each of the evoked consumption contexts (presented in written format). A week later, the same participants completed the same emotion questionnaire while imagining themselves eating an apple or chocolate brownie (shown as images) without any given context (control condition), and then under the three different contexts (random monadic sequential order). When the chocolate brownie was the food stimulus, 20 of 36 emotion responses were significantly different from context to context (including the no-context condition). Conversely, only 9 significant differences were established when apple was the food stimulus. It was also found that frequency of use of positive emotion terms was higher in those consumption contexts that were considered more appropriate, whereas negative emotion terms were more frequently selected when the context was perceived as less appropriate for the product. Taken together, these findings highlight how the context in which a food is consumed (even if imagined) can affect consumers’ emotion responses towards it. More broadly, they demonstrate the reciprocal effect that the context and the product have on the elicited emotion responses, which are not situationally invariant. This research also explored how psychological traits related to emotional eating affected the emotion responses and found that in general higher emotional eaters selected a higher number of emotion terms (positive and negative). This study represents a needed advance in food-related emotion research, and upon replication of the results with other product-context combinations, could offer an easy-to-implement methodology that should more closely match the emotion responses that consumers would elicit in real-life situations while eating and drinking. Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction In the last 5 years, great interest has been focused on the impact that foods and beverages can exert on our emotions/feelings (e.g., Manzocco, Rumignani, & Lagazio, 2013; Schifferstein & Desmet, 2010; Schifferstein, Fenko, Desmet, Labbe, & Martin, 2013) and how this relates to food acceptance/liking. In sensory and con⇑ Corresponding author at: Department of Experimental Psychology, University of Oxford, South Parks Road, OX1 3UD Oxford, UK. Tel.: +44 7982974940. E-mail address: betina.piqueras-fi[email protected] (B. Piqueras-Fiszman). 0950-3293/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.foodqual.2013.09.002

sumer research, questionnaire methods are being developed to obtain measures of emotions/feelings elicited by (or associated to) food (e.g., Cardello et al., 2012; King & Meiselman, 2010; King, Meiselman, & Carr, 2010; King, Meiselman, & Carr, 2013; Moskowitz, Silcher, Beckley, Minkus-McKenna, & Mascuch, 2005; Thomson, Crocker, & Marketo, 2010). Regarding the methodological research pertaining to emotion qustionnaires, King et al. (2013) have demonstrated the impact of several parameters of the EsSense ProfileÒ questionnaire format (i.e., type of questionnaire, order of emotions listed, and position of emotions with respect to an overall acceptability question) on hedonic and emotion responses;

278

B. Piqueras-Fiszman, S.R. Jaeger / Food Quality and Preference 32 (2014) 277–288

the difference in response when testing a product by different means; the impact of number of samples; and the impact of time of day when the emotion test is conducted. Other researchers have looked at individual differences (Jaeger & Hedderley, 2013), what it feels like completing one of these emotion-related surveys (Jaeger, Cardello, & Schutz, 2013), or whether the emotion responses provided are reliable (stable) for repeated presentations (Cardello et al., 2012). However, to date, the influence of evoking a consumption context during an emotion task on the participants’ emotion responses has not been explored. The contextual characteristics of eating/ drinking occasions have a greater impact our enjoyment of the food than we are aware of (Köster, 2003; e.g., Bell, Meiselman, Pierson, & Reeve, 1994; de Graaf et al., 2005; Edwards, Meiselman, Edwards, & Lesher, 2003; Hersleth, Monteleone, Segtnan, & Næs, 2012; King, Meiselman, Hottenstein, Work, & Cronk, 2007; Petit & Sieffermann, 2007; Stroebele & De Castro, 2004). Moreover, consumption context is inextricably linked to how we feel (Richins, 1997). It sets up our mood and mindset and may give rise to an array of emotions (either positive, negative, or neutral) that very likely will modulate our perception of the food, and our liking and enjoyment of the food and consumption experience. Desmet and Schifferstein (2008) introduced the concept of ‘source of food emotion’ to emphasize that food emotions can be elicited by five types of sources: (1) sensory properties of the food itself, (2) experienced consequences, (3) associated consequences, (4) personal or social meanings, and (5) behavior of others involved. It is worth noting that four of these sources could be categorized as pertaining to the context. Trying to discern sources from consequences may be challenging, but perhaps not really necessary, since this combination of factors plays an important (and intrinsic) role in emotion elicitation. Therefore, this reciprocal ‘halo’ effect should be taken into account in emotion-related consumer research. It is also relevant to consider the effect of the context/situation in consumer research because in many occasions, the choice of a certain food product over another is driven mainly by a planned context of consumption, and consequently, we tend to enjoy it more in that situation than in another. It is evident that certain products fit better in specific situations of consumption. Take, for instance Köster’s (2003) example of a chocolate-covered ice-cream on a stick, which is normally a highly appreciated product globally: while it may fit in a family dinner situation, or even in a friends-out situation, it could be quite unacceptable in a formal business lunch. Thus, the same product may fit one situation and not another, and while this difference in appropriateness has been proved to affect consumers’ reported acceptance ratings toward the product (Schutz, 1994), it has not been explored yet with emotion data. Important questions have not been addressed yet, such as: To what extent are the emotions reported solely food-driven and to what extent do food-elicited emotions depend on consumption context? What is the contribution of the consumption context to the emotions rated? Can these two drivers of emotion elicitation be easily distinguished? What effect does the appropriateness of a given consumption situation have on emotion responses of foods? Here we attempt to address these questions. Thus, the aim of this study was to investigate the impact of different consumption situations on consumers’ emotion responses to two different generic food products with different emotional consequences (Thomson et al., 2010). The specific objectives were: (1) To compare the emotion responses evoked by foods in different conditions: under certain evoked consumption contexts and without context (control condition); (2) To investigate whether the emotion profiles of the foods imagined to be consumed under certain contexts follow the emotion profiles elicited by the contexts alone; (3) To explore whether the appropriateness and frequency of consumption of the foods in those contexts exert any impact

on the emotion profiles; and (4) To explore how psychological traits, namely those related to emotional eating behaviors, affect the emotion responses. Note that, purely for simplification purposes, the expressions ‘‘emotion association’’ or ‘‘emotion response’’ will be used to refer to the affective phenomena/feeling-related concepts mentioned throughout this paper, being aware that they may not be strictly emotions. The emerging debate about what exactly are the responses elicited by emotion questionnaires (Jaeger et al., 2013) is necessary and requires attention, but it is not the focus of this work. Nevertheless, in this manuscript we do not assume that the responses obtained with explicit protocols, such as emotionscoring surveys, give valid reflections of inner thoughts or purely experienced emotions or feelings, that is why we decided to refer to them as associations. 2. Materials and methods 2.1. Participants A total of 157 participants (53.5% were females; 42% were 20–35 years old, 46% were 36–50 years old, and 12% were 50+ years old) were recruited at the Universitat Politécnica de Valencia (Valencia, Spain) from a managed panel by a research group at the Department of Food Technology, or from a range of short courses given at this university. After being told the protocol of the study, they gave their informed voluntary consent to participate. No previous information was given about the purpose of the study, and after completing the study they were remunerated for participation. 2.2. Emotion questionnaire Though there are several methods to measure emotional responses (both implicit and explicit) to consumer goods, to date, self-reported questionnaires remain a simple and popular approach among practitioners in sensory and consumer research. In this study, the questionnaire used was based on the EsSense ProfileÒ wordlist developed by King and Meiselman (2010), translated to Spanish, and reduced, since some of the meaning of the translated terms would be contained in other terms of the list (merry) or would not be suitable (tame and aggressive), as found in pilot work, where these terms elicited confusion among participants. The final list containing 36 items (listed in alphabetical order) was checked by five Spanish people to ensure that they understood all the terms and found them potentially applicable for the purpose of the study (Richins, 1997). A checklist version of the questionnaire was chosen for this study since it has been demonstrated that it is an acceptable approach for measuring emotion-related responses (Meiselman, King, & Carr, 2013) and it was considered to be not too demanding/tiresome, considering the objectives of this study (i.e., each participant having to complete an emotion questionnaire several times). See the Appendix section for the translated wordlist. 2.3. Selection of contexts and products An underpinning premise of this study was that the appropriateness of consuming a given product in one context or another affects the eating experience (Cardello & Schutz, 1996) as it may affect its emotions elicited too. The contexts and the products used throughout this study were selected with the purpose of creating product-context combinations that would vary in terms of appropriateness in consumers’ minds. We studied two food products in three different consumption situations. The two products, apple

B. Piqueras-Fiszman, S.R. Jaeger / Food Quality and Preference 32 (2014) 277–288

and chocolate brownie, were shown to participants as images (a Granny Smith apple with a slice cut apart, and two overlapped bite-sized squared portions of a dense chocolate brownie, respectively, over a white background in a size similar to real) with the aim of creating boundaries regarding participants’ imagined product and to limit possible associations with a specific brand and product emotional conceptualization). Moreover, chocolate brownie and apples would be expected to differ in their emotional consequence (higher vs. lower) based on pilot work from author SRJ who found very different EsSense profiles for chocolate and apple and this difference was expected to extend to the products selected here. Three different common consumption contexts (flexible, in the sense that participants were able to adapt the contexts to their own personal situations) were defined. They differed in the type and amount of food consumed, in the time of the day, in the location, in a possible simultaneous activity, in the social setting, in the person’s state of mind and physical condition, and in the recurrence of the moment/situation (Bisogni et al., 2007; cf. Hein, Hamid, Jaeger, & Delahunty, 2010; Hein, Hamid, Jaeger, & Delahunty, 2012). The evoked contexts were: (i) Breakfast on a weekend morning; (ii) Afternoon break snack on a weekday; and (iii) After a special dinner at home in good company. These were presented to participants in written format, since it has been previously demonstrated that participants are able to hold a vivid image of the occasion conveyed while performing a task (Hein et al., 2010; Hein et al., 2012). Now, there are a couple of points that we would like to clarify about our approach. First, by collecting these emotion responses/associations from consumers by means of showing them food images as stimuli, we are not implying that people would normally experience the same emotions when thinking about a food than when actually consuming it, but there is evidence that they do report the same (or very similar) emotion responses in both cases (Cardello et al., 2012). Second, by evoking contexts we are not suggesting that people experience (or even report) the same emotions when thinking about a consumption context than when participating in it for real, but nevertheless, evoking consumption scenarios is a tool widely used in research that helps the respondents to ‘‘frame’’ their responses. 2.4. Pretest 2.4.1. Pretest procedure Seeking to use a within-subjects design for data collection whereby each participant would be required to imagine several contexts for one product, a pretest (computer-based) with 40 participants was performed to ensure they would be able to do this easily. This pretest consisted in asking participants to imagine themselves eating either product (i.e., an apple or, as was initially planned in this pretest, a piece of milk chocolate) under each of the three selected consumption contexts mentioned in Section 2.3 (except for the dinner context, which in a first instance was evoked as ‘after a romantic dinner with the tea/coffee’ in this pretest). Half of the participants carried out this task imagining they were consuming the chocolate, while the other half imagined they were consuming apple (both groups were balanced in terms of sex and age range), and the contexts were presented in random sequential order. After each of these tasks, a question regarding the ease in imagining the three contexts was asked. To explore whether participants could still imagine the consumption context, even in the event of not finding it very appropriate or not being a frequent consumer of that product in that given context, questions about the appropriateness of the given product in each occasion, and

279

about if they regularly consumed that product were included too. 9-point scales were provided below the questions regarding the easiness and the appropriateness (with labels anchored at 1 = ‘not at all’, and 9 = ‘very’), and a ‘yes/no’ question was asked to determine if they considered themselves as regular consumers of that product in that context. T-tests were performed to compare the mean appropriateness and easiness scores of among the contexts. 2.4.2. Pretest results The results for those participants who completed the apple questionnaire showed that they did not usually eat apples. However, they did find it moderately appropriate for breakfast (M = 5.3, SD = 2.6) and as an afternoon snack (M = 5.2, SD = 2.3), t(19) = 0.18, p = 0.86; but not for after a romantic dinner (M = 1.9, SD = 1.5), t(19) = 4.71, p < 0.0001; and t(19) = 6.05, p < 0.0001, compared to the breakfast and afternoon contexts, respectively. Tentatively, they rated apples as moderately appropriate due to lack of regular consumption (several participants mentioned that they ate other fruits instead). Importantly, regardless of the fact that they were not frequent consumers of apples in those contexts, they were able to imagine themselves in those consumption contexts relatively easily (M = 6.0, SD = 2.6; M = 6.2, SD = 2.1; and M = 4.2, SD = 2.4, respectively for the breakfast, afternoon, and romantic dinner occasions). Regarding the results for chocolate in this pretest, 4 out of 20 subjects consumed chocolate (similar to that of the image) in the breakfast and afternoon occasions, and 11 out of 20 in the dinner occasion. Their mean ratings for appropriateness of consumption increased with the time of the day (breakfast occasion: M = 4.2, SD = 1.2; afternoon: M = 5.7, SD = 1.7, t(19) = 5.10, p < 0.0001; and dinner: M = 7.5, SD = 1.9, t(19) = 7.40, p < 0.0001, t(19) = 6.09, p < 0.0001, respectively). Similar results were observed for the ratings concerning the easiness to imagine those occasions (M = 5.3, SD = 2.2; M = 6.2, SD = 1.7; and M = 7.4, SD = 1.8, respectively). Given that several participants commented that they would eat chocolate after dinner as a dessert (but not after dinner with tea/ coffee) and not necessarily in that format, for the main study the image stimulus of the piece of chocolate was changed to an image of brownie squares. In addition, the description for the dinner situation was also modified to: ‘‘Think about a special dinner occasion at home (in good company) when you’re having something as dessert’’ since participants also mentioned that if they were not in a couple they felt bad for having to imagine a romantic situation. These changes to be implemented in the main study were expected to provide the same pattern of results for the other two occasions in terms of easiness to imagine and appropriates of consumption. 2.5. Main study Fig. 1 shows a diagram of the empirical procedure, which is described in detail in the sections that follow. A within-subject design was used across context conditions (participants provided their responses for all the contexts in random monadic sequential order) and between-subject design across the products. One group, N = 76, was shown the apple as the food stimulus (51.3% were females; 47% were 20–35 years old, 42% were 36–50, and 11% were 50+ years old), while the other group, N = 81, was shown the chocolate brownie instead (55% were females, 38% were 20–35 years old, 49% were 36–50, and 13% were 50+ years old; both groups were balanced in terms of sex and age range, chi-squared tests results were all p > 0.09). Participants completed the study in a quiet room seated individually and using a computer.

280

B. Piqueras-Fiszman, S.R. Jaeger / Food Quality and Preference 32 (2014) 277–288

Fig. 1. Diagram of the entire procedure followed.

2.5.1. Comparing emotion responses of products under evoked contexts and without context (control condition) To address the first objective of the study, participants were presented with the four conditions (no context and the three contexts described earlier) and for each condition they were asked to complete the emotion questionnaire. This stage consisted of the following steps (Fig. 1, stage B): (1) Warm-up (no context) task. Participants were first presented with a warm-up task to get acquainted with the procedure and avoid first-position effects. In this task they were shown an image of a handful of walnuts over a white background and asking them to complete the questionnaire. The instructions were as follows: ‘‘Imagine you are eating walnuts similar to those shown in the picture. How do you feel? Please select all of the words that apply’’. (2) No-context task (control condition). During this second step, the procedure was identical to that of the warm-up task, but while having to imagine that they were eating either an apple or a chocolate brownie (shown as images). A question about their frequency of consumption of that product was included at the end. This was phrased: ‘‘Approximately, how frequently do you consume this product?’’ (scale anchors: 1 = ‘never’; 2 = ‘rarely’; 3 = ‘sometimes’; 4 = ‘commonly’; 5 = ‘very often’). (3) Context tasks. Next, all the three conditions with consumption contexts were presented in random order across participants with the emotion questionnaire, with 3-min breaks in between. The context descriptions were phrased (using apple for the example): (i) Breakfast context: ‘‘Imagine it is a typical breakfast occasion on a weekend morning. Try to imagine that an apple similar to that shown in the picture is available to you and you decide to have some’’; (ii) Afternoon context: ‘‘Think about a typical afternoon occasion when you are eating a snack. Try to imagine it is an apple similar to that shown in the picture’’; and (iii) Dinner context: ‘‘Think about a special dinner occasion at home (in good company) when you’re having something as dessert. Try to imagine it is an apple similar to that shown in the picture’’. 2.5.2. Comparing emotion responses of the contexts alone with those of the products under the contexts The second objective of the study was to obtain the emotion profile of the contexts alone to be able to compare them with those

obtained involving the products. For this, one week earlier, all the participants (N = 157) were invited to complete a ‘context alone’ task, in which they were presented with the three contexts described before (in random sequential order) and for each of them they were asked to complete the emotion questionnaire (Fig. 1, stage A). So, for instance, for the dinner context the instructions given were: ‘‘Think about a special dinner occasion at home in good company when you’re having something as dessert. How do you feel? Please select all of the words that apply’’. 2.5.3. Exploring the effect of product-context appropriateness on emotion responses After completing the emotion questionnaires of Stage B (cf. Fig. 1), the participants were asked about the perceived appropriateness of eating that product in each of the contexts and about the frequency of product consumption. The questions were phrased: ‘‘How appropriate would it be to eat this product in this occasion?’’ (with 9-pt scales, labels anchored at 1 = ‘not at all’, and 9 = ‘very’), and ‘‘Approximately, how frequently do you consume this product in the occasion that you imagined?’’ (scale anchors: 1 = ‘never’; 2 = ‘rarely’; 3 = ‘sometimes’; 4 = ‘commonly’; 5 = ‘very often’). 2.5.4. Exploring the effect of emotional eating behaviors on emotion responses To explore whether individual differences regarding emotional eating behaviors (and not emotional traits, or eating behaviors, individually) affect the emotion responses, after completing the tasks described above (stage B), all the participants were asked to complete the emotional section of the Dutch Eating Behavior Questionnaire (DEBQ; Van Strien, Frijters, Bergers, & Defares, 1986). This section (referred here as DEBQ-e) consisted of 13 questions concerning emotional attitudes toward eating (e.g., Do you have the desire to eat when you are irritated? When you are feeling lonely? When you are worried?) This questionnaire was used since it has displayed good internal consistency and factorial validity, and good validity for food consumption (Jansen et al., 2010). 2.6. Data analyses In order to investigate the effect of the contexts on the emotion responses (first objective), Cochran Q-tests were performed on each of the 36 emotions to explore differences between the frequency of mention of each emotion among the contexts (including the no-context condition). Post-hoc pairwise comparisons between

B. Piqueras-Fiszman, S.R. Jaeger / Food Quality and Preference 32 (2014) 277–288

the contexts were made using the Marascuilo procedure (Levine, Stephan, Krehbiel, & Berenson, 2011). Although literature suggests that this procedure is very conservative, the level of significance was set nonetheless at 5%. In addition, a multiple factor analysis (MFA; Bécue-Bertaut & Lê, 2011; Bécue-Bertaut & Pagès, 2008) was performed to compare the configurations between the four context conditions for the two foods. For this analysis, only those emotions mentioned by more than 5% of the participants across the four different conditions were considered (Bécue-Bertaut, Álvarez-Esteban, & Pagès, 2008). To check whether the difference between the two products’ emotion configurations within each context was significant, the frequencies of emotion elicitation of the products were compared within each context using Friedman’s test. Finally, a logistic regression was carried out to determine which emotions varied significantly as a function of the product, the context, and/or their interaction. To address the second objective, the relation between the emotion responses evoked by the contexts alone and those evoked by the products within the contexts was investigated by performing McNemar tests on all 36 emotions. Another MFA was performed in order to compare the integrated configuration of the emotions in relation to each of the three evoked contexts. Separate tests were performed on each product category (context-alone, apple, and chocolate brownie). Only those emotions mentioned by more than 5% of the participants across the three different contexts were analyzed. The frequencies of emotion elicitation of the products within each context and the contexts alone were compared using Friedman’s test to explore whether the differences between the two products’ emotion configurations within each context and that of the context alone were significant. Multiple pairwise comparisons were done using Bonferroni correction. A two-way ANOVA (with Tukey’s test) was performed on the appropriateness ratings considering product, context, and their interaction as explanatory variables to examine their effect on the appropriateness ratings and how these differed across contexts within a product and within the contexts between the products (third objective). To explore the effect that the appropriateness of consuming a product in different contexts has on the emotion ratings, participants within each product were divided into three different groups according to the appropriateness ratings (of consuming that product in each context) they reported, which were either low (L; 1–3); medium (M; 4–6), or high (H; 7–9). Although this split creates uneven groups of participants, it was selected over a triadic split because having same range of scores was expected to facilitate the interpretation of these results. The effect of appropriateness of consumption was explored by performing a logistic regression on the emotion responses, considering the three appropriateness groups as the explanatory variable. In instances where the number of participants were very unbalanced, the results were interpreted with caution (i.e., 8 subjects found the brownie for the dinner context to be inappropriate, while 53 perceived it as being very appropriate; in contrast, 42 found the apple inappropriate for the dinner context, while only 16 thought it was very appropriate). Regarding the ratings about the frequency of consumption in those contexts, these were quite low (M = 2.4 and 2.2, for chocolate brownie and apple, respectively across all contexts), therefore, these data were not further considered in the analyses. However, all the subjects considered themselves as being regular consumers of the products (M = 3.2 and 3.7 respectively, though apparently not in the contexts provided). Having this condition met, the data from these respondents was considered valid. Finally, to explore the impact of the participants’ emotional eating behavior on their emotion responses (fourth objective) the participants were segmented into three groups according to their average DEBQ-e score. The split was performed by characterizing

281

participants as being either ‘low emotional eaters’ (L; M < 2), ‘medium emotional eaters’ (M; 2 6 M < 3), and ‘high emotional eaters’ (H; P3), for the same reason as described for the appropriateness ratings. The numbers of participants in these groups were: 32, 30, and 14 (respectively) in the apple study, and 25, 41, and 15 for the chocolate brownie study. Then logistic regression was performed on the 36 emotions, considering the levels of emotional eating as the explanatory variable. Sex and age groups were analyzed with post-hoc tests to check the relationship with their emotional eating behavior. For all statistical analyses, the significance level was set at 5%. The data analyses were performed using XLStat 2012 (Addinsoft, NY, USA) and FactoMineR (Lê, Josse, & Husson, 2008) in R language (R Development Core Team 2007, Vienna, Austria).

3. Results 3.1. Relationship between the emotion responses linked to the evoked consumption contexts and the products alone (no-context condition) This section addresses the first objective of this research, which was to explore the effect of the evoked contexts (and control condition) on the emotion responses. Table 1 summarizes the results obtained by comparing the frequency of the 36 emotions terms reported under the three contexts and under the no-context condition, for the apple and the chocolate brownie. The chocolate brownie, on average, evoked significantly more emotion associations than apple did (M = 16.2 vs. 11.3, p < 0.05 for no-context, M = 16 vs. 10.7 for breakfast, and M = 20.6 vs. 12.5 for the dinner context, p < 0.01) except for, the afternoon context (M = 12.9 vs. 12, p = 0.89, Table 1 bottom row). More importantly, the frequencies of 9 emotion terms associated with the apple were significantly different from context to context, while in the case of the chocolate brownie, this figure ascended to 20, 15 of which were significantly more mentioned under the evoked dinner context. In contrast, for the apple, the differences were more evenly distributed among the four conditions. These first results seem to suggest that the emotion associations reported in response to the chocolate brownie are more context-dependent than those reported in response to the apple. The results of logistic regression analysis (considering the products, the contexts, and their interaction as explanatory variables) highlighted which of these emotion associations varied significantly depending mainly on the product (bored, p < 0.001; calm, p < 0.01; energetic, p < 0.05; guilty, p < 0.0001; happy, p < 0.01; worried, p < 0.05), on the context (active, p < 0.05; affectionate, p < 0.0001; calm, p < 0.05; energetic, p < 0.001; friendly, p < 0.001; loving, p < 0.05; pleased, p < 0.03), and on the interaction between the product and the context (active, p < 0.05; good, p < 0.01; guilty, p < 0.05; pleased, p < 0.01; satisfied, p < 0.05). Fig. 2a shows the partial individuals plot of the MFA (first two dimensions accounting for 84.2% of the variance, 62.5% and 21.7%, respectively). Focusing on the position of the context’s coordinates, it is apparent that the three contexts are dispersed across the MFA 2-dimensional space, which indicates that the emotion configurations reported under each of them are quite different (as shown in Table 1). Dinner was the context that most differed from the others (opposed along the first dimension) and that contributed the most to the variance observed (72.2% of dimension 1). Focusing now on the representation of the emotion terms on the space, Fig. 2b shows a selection of the most relevant terms for the purpose of this analysis to ease the reading of the graph. It could be observed that the dinner context was strongly associated with specific emotions, namely affectionate and loving, when imagined with the two foods (hence the two foods contributing equally to dimension 1, shown in Fig. 2a), and that both the chocolate

282

B. Piqueras-Fiszman, S.R. Jaeger / Food Quality and Preference 32 (2014) 277–288

Table 1 Frequency (%) of emotion terms selected for the apple and chocolate brownie under each context and under the no-context condition with Cochran’s Q test results to detect significant differences. Apple

Chocolate brownie

Emotion

No.-cnt

Breakf.

Aftern.

Dinner

Emotion

No-cnt

Breakf.

Aftern.

Dinner

Active⁄⁄ Adventurous Affectionate⁄⁄⁄ Bored Calm⁄⁄ Daring Disgusted Eager Energetic⁄⁄⁄ Enthusiastic Free Friendly⁄⁄⁄ Glad Good⁄ Good-natured Guilty Happy Interested Joyful Loving⁄⁄ Mild Nostalgic Peaceful Pleasant Pleased⁄ Polite Quiet Satisfied⁄ Secure Steady Tender Understanding Warm Whole Wild Worried Average number

31.6b 5.3 2.6a 21.1 11.8a 4.0 2.6 6.6 32.9b 9.2 10.5 7.9a 15.8 51.3ab 0.0 0.0 19.7 5.3 10.5 4.0a 11.8 2.6 10.5 34.2 13.2ab 4.0 0.0 34.2b 10.5 11.8 2.6 4.0 5.3 4.0 1.3 4.0 11.3A

25.0ab 9.2 6.6a 26.3 18.4ab 9.2 7.9 9.2 21.1ab 4.0 14.5 2.6a 19.7 43.4ab 1.3 2.6 19.7 2.6 14.5 6.6ab 17.1 2.6 6.6 23.7 6.6a 2.6 1.3 18.4a 9.2 4.0 4.0 2.6 9.2 4.0 4.0 4.0 10.7A

22.4ab 2.6 4.0a 26.3 32.9b 4.0 5.3 9.2 17.1a 7.9 11.8 6.6a 21.1 59.2b 4.0 0.0 21.1 4.0 6.6 4.0a 19.7 1.3 14.5 25.0 22.4b 2.6 1.3 32.9ab 10.5 5.3 1.3 9.2 5.3 6.6 1.3 2.6 12.0A

10.5a 4.0 27.6b 21.1 19.7ab 5.3 6.6 4.0 7.9a 5.3 7.9 21.1ab 27.6 38.2a 4.0 0.0 25.0 6.6 15.8 15.8b 11.8 5.3 7.9 38.2 18.4ab 7.9 2.6 34.2b 13.2 5.3 7.9 4.0 10.5 4.0 2.6 2.6 12.5A

Active⁄ Adventurous Affectionate⁄⁄⁄ Bored⁄⁄ Calm⁄ Daring⁄ Disgusted Eager Energetic⁄⁄⁄ Enthusiastic⁄ Free Friendly⁄⁄⁄ Glad⁄⁄ Good Good natured⁄⁄ Guilty⁄⁄⁄ Happy⁄ Interested⁄ Joyful⁄⁄⁄ Loving⁄⁄⁄ Mild Nostalgic Peaceful Pleasant⁄⁄⁄ Pleased⁄⁄ Polite⁄⁄⁄ Quiet Satisfied⁄⁄⁄ Secure Steady Tender⁄⁄⁄ Understanding Warm Whole Wild Worried Average number

19.8a 7.4 8.6a 1.2a 11.1ab 12.3ab 1.2 3.7 43.2b 19.8ab 9.9 28.4ab 49.4ab 56.8 1.2ab 18.5ab 56.8ab 1.2a 7.4a 11.1a 16.0 1.2 9.9 42.0ab 40.7b 4.9ab 1.2 46.9b 7.4 4.9 7.4ab 4.9 16.0 1.2 3.7 4.9 16.2B

32.1a 3.7 12.3a 2.5ab 24.7ab 9.9ab 2.5 3.7 48.1b 13.6ab 11.1 14.8a 39.5ab 58.0 4.9ab 13.6a 50.6ab 2.5ab 18.5ab 11.1a 19.8 1.2 12.3 33.3a 32.1ab 1.2a 1.2 44.4b 11.1 3.7 4.9a 2.5 16.0 3.7 1.2 7.4 16.0B

18.5a 1.2 4.9a 7.4b 14.8a 6.2a 1.2 7.4 33.3ab 11.1a 7.4 14.8a 34.6a 45.7 1.2a 30.9b 40.7ª 2.5ab 9.9a 8.6a 14.8 3.7 8.6 32.1a 22.2a 2.5a 4.9 27.2a 4.9 2.5 2.5a 2.5 13.6 1.2 6.2 11.1 12.9A

25.9a 7.4 43.2b 1.2a 18.5ab 18.5b 0.0 2.5 23.5a 25.9b 11.1 40.7b 56.8b 60.5 9.9b 8.6a 60.5b 7.4b 23.5b 35.8b 16.0 3.7 8.6 56.8b 40.7b 12.3b 1.2 51.9b 13.6 2.5 16.0b 4.9 27.2 1.2 1.2 2.5 20.6B

⁄⁄⁄

Indicates significant differences between the frequencies of the terms elicited under each context/condition at a significance level of p < 0.001; ⁄⁄at p < 0.01; ⁄at p < 0.05. Values within rows with different lowercase superscripts are significantly different according to Marascuilo paired comparisons test (at p < 0.05). Values in the last row with different uppercase superscripts indicate that the frequencies of apple and chocolate brownie are significantly different within each context, according to Friedman’s test (at p < 0.05).

Fig. 2. Representation of the first two dimensions of the MFA space showing (a) the four context conditions as mean points (breakfast, afternoon, dinner, and no-context) and the partial individuals representing the emotion configurations of the contexts associated with the apple and the chocolate brownie; and (b) a selection of the most relevant emotion terms.

B. Piqueras-Fiszman, S.R. Jaeger / Food Quality and Preference 32 (2014) 277–288

brownie and the apple under the breakfast and no-context conditions was associated with energetic. However, under the afternoon context apple was associated with good, peaceful, and calm, while the chocolate brownie would seem to make them associate it with guilty, an observation that hints at an important effect of the interaction between the context and the food being consumed. In fact, the differences of the emotion configurations between the products varied as a function of the context (though the RV coefficient of the products was 0.86, close to 1, which means that they have similar global configurations). For instance, while for dinner both the apple and the brownie contributed equally to dimension 1 (49.6% and 50.4%, respectively), their individual coordinates separate along dimension 2 (see Fig. 2a), which means that other emotion associations correlated with dimension 2 influenced their configuration (mainly those associated with apple, Fig. 2b). Something similar is observed for the emotion configurations of both products when no context was evoked, while for breakfast and afternoon the difference between the configurations of the foods seemed to be smaller. 3.2. Relationship between the emotion responses elicited under the contexts alone and with the products So far our results show that emotion responses differed significantly from context to context, but it remains unclear what emotions were elicited by the consumption contexts alone and how they are related to those of apple and chocolate brownie under the same contexts. The partial individual’s plot of the MFA for the three contexts (Fig. 3a) answers this question. Clear differences can be seen in the configurations of the emotions reported without product (context alone), with the apple, and with the chocolate brownie, although it can be inferred that the chocolate brownie elicited a closer emotion profile to the context alone than the apple, for which coordinates are further apart along dimension 2 in the three contexts. The majority of the emotion associations for these two conditions (chocolate brownie and context alone) were spread along dimension 1 (Fig. 3b). The breakfast context itself elicited emotion associations such as free and energetic, afternoon eager and worried, and dinner daring, affectionate, loving, and tender (Fig. 3b). In contrast, the associations evoked by an apple or a choc-

283

olate brownie consumed in these same contexts were quite different: imagining themselves eating an apple in the breakfast context elicited associations like adventurous and eager, in the afternoon pleased and calm, and for the dinner affectionate. The configuration of the emotions associated with the chocolate brownie was somewhat more similar to that without the product, with some changes, such as guilty being elicited when consumed in the afternoon and polite in the dinner context (Fig. 3b). Table 2 shows the results of McNemar tests, performed to compare the participants’ emotion responses reported in each context without the food products (that is, the context itself) and with the food products. These results support the relations obtained from the MFA (Fig. 3), revealing that 13 out of 36 emotions differed significantly from evoking only a consumption context to evoking a consumption context with the chocolate brownie, while 20 emotions out of the 36 differed for the case of the apple. Focusing on the differences within the contexts, those participants who had responded to the chocolate brownie stimulus used significantly differently 9 emotions from one breakfast context to the other breakfast context, 6 in the case of the afternoon context and 6 for the dinner context. For instance, let’s look at the most significantly different cases. Breakfast (on its own) was not associated with guilty (not being selected at all), but 13.6% of the respondents did elicit this association with eating chocolate brownie for breakfast. As for afternoon itself, calm, guilty, and happy were selected by 33.3%, 3.7%, and 16% of the respondents, while the proportion of frequency of these terms were very different when the chocolate brownie was involved as part of the afternoon snack (see Table 1 for the frequency data of these terms). As for the dinner context, loving was selected by 66.6% of the respondents, which nealy doubles the frequency of selection of that term when the chocolate brownie was part of this context (35.8%). Regarding those who had the apple as stimulus, they used significantly differently 13, 2, and 16 emotions, respectively for breakfast, afternoon and dinner (note that these differences are also reflected in Fig. 3a). To mention just some examples, bored was associated with the breakfast itself only by 2.6% of the respondents, while this proportion increased to 26.3% when apple was involved in this context. Similarly, for this group of participants, dinner

Fig. 3. Representation of the first two dimensions of the MFA space showing (a) the three contexts as mean points (breakfast, afternoon, and dinner) and the partial individuals representing the emotion configurations of the contexts associated with the apple, the chocolate brownie, and the context alone; and (b) a selection of the most relevant emotion terms.

284

B. Piqueras-Fiszman, S.R. Jaeger / Food Quality and Preference 32 (2014) 277–288

Table 2 McNemar results highlighting those emotion terms which differed significantly in frequency of occurrence between being selected under the contexts alone and being selected with either food stimulus (apple or chocolate brownie). Black cells indicate the emotion terms that were significantly different at a significance level of p < 0.001; dark grey cells at p < 0.01; light grey cells at p < 0.05; and white cells indicate non-significant differences (p > 0.05). Apple

Active Adventurous Affectionate Bored Calm Daring Disgusted Eager Energetic Enthusiastic Free Friendly Glad Good Good natured Guilty Happy Interested Joyful Loving Mild Nostalgic Peaceful Pleasant Pleased Polite Quiet Satisfied Secure Steady Tender Understanding Warm Whole Wild Worried

Chocolate brownie

Breakfast

Afternoon

Dinner

Breakfast

Afternoon

Dinner

0.043 0.344 0.118 0.000 0.004 0.727 0.031 1.000 0.185 0.004 0.005 0.000 0.001 0.073 0.125 0.500 0.000 0.219 0.052 0.007 0.230 1.000 0.092 0.017 0.388 0.008 1.000 0.332 0.096 1.000 0.070 0.039 0.454 1.000 1.000 1.000

0.248 1.000 0.508 0.230 1.000 1.000 0.125 0.096 0.503 1.000 0.664 0.002 0.851 0.362 0.625 0.125 0.267 0.625 0.035 0.289 1.000 0.219 0.549 0.856 0.332 0.219 0.219 0.122 0.267 1.000 0.375 0.754 0.180 0.727 1.000 0.065

0.118 0.508 0.000 0.000 0.029 0.000 0.063 0.453 0.146 0.000 0.267 0.017 0.000 0.362 0.727 1.000 0.000 0.065 0.004 0.000 0.115 1.000 0.180 0.243 0.036 0.004 0.500 0.110 0.001 0.625 0.000 0.013 0.001 1.000 0.500 0.500

0.839 0.375 0.332 0.500 0.003 0.125 0.500 1.000 0.035 0.078 0.003 0.327 0.087 0.281 1.000 0.001 0.851 0.125 0.031 0.027 0.050 1.000 0.118 0.442 0.327 0.008 1.000 0.585 0.804 0.688 0.057 0.063 0.815 1.000 1.000 0.031

0.007 0.625 0.688 1.000 0.001 0.453 1.000 0.013 0.023 0.344 1.000 0.302 0.093 0.296 0.625 0.000 0.001 0.625 0.549 1.000 1.000 0.453 0.210 0.052 0.064 1.000 0.727 0.701 0.180 0.125 1.000 0.109 0.146 0.500 0.219 0.629

0.027 0.774 0.110 1.000 0.332 0.424 1.000 0.688 0.003 0.608 0.549 0.701 0.597 0.720 1.000 0.016 0.572 0.344 0.405 0.000 0.096 1.000 0.549 1.000 0.430 0.210 1.000 0.511 0.286 1.000 0.000 0.021 0.064 0.125 1.000 1.000

was associated with affectionate, enthusiastic, glad, happy, loving, secure, and warm by a significantly higher number of participants (the frequencies being 64.5%, 26%, 63.1%, 72.4%, 64.5%, 29.5%, 41.8% and 35.5%, respectively) than when apple was part of the dinner context (see Table 1). The term bored was the only term that inverted this pattern for this context, not being selected at all for the dinner context itself. These results, obtained by measuring directly the emotions elicited by the contexts alone, are in line with the effects shown in Table 1 and support the notion of a reciprocal effect of the context and the food being consumed on the associated emotions. 3.3. Effect of the appropriateness on the emotion responses To be able to better understand several of the results observed, such as the differences between the products’ emotion configurations within and between contexts, focus is now directed to the product-context appropriateness ratings. The results of the ANOVA revealed that the appropriateness of consuming an apple differed significantly from the afternoon to the breakfast context (M = 6.4 vs. 4.5, p < 0.0001), and from the afternoon to the dinner context (M = 3.8, p < 0.001), but not from the breakfast to the dinner context (p = 0.177). For the chocolate brownie, it was only the dinner context (M = 6.7) which was significantly more appropriate than the breakfast (M = 4.6, p < 0.0001) and the afternoon contexts (M = 4.5, p < 0.0001). These results confirm again that the products and consumption contexts varied in terms of their appropriateness, and additionally explain some of the differences mentioned in the

previous sections. For instance, as seen in Fig. 2a, the difference in emotion configuration between the apple and the chocolate brownie was large (and highly significant) for the dinner context, which makes sense considering that the appropriateness of eating apple or a chocolate brownie was significantly different for this occasion (M = 3.7 for apple vs. 6.7 for chocolate, p < 0.0001 according to Tukey’s test). The difference in appropriateness ratings for the afternoon context (M = 6.3 vs. 4.5, respectively, p < 0.05) did not seem to be reflected in the emotion configurations of the two products elicited under this contexts, since they were quite similar (not significantly different, p = 0.89). As for the breakfast context, the difference in appropriateness ratings between the apple and the chocolate brownie was non-significant (M = 4.5 vs. 4.6, respectively, p = 0.99), but nevertheless their emotion configurations did differ significantly (p < 0.01). Why this lack of correspondence in results for the afternoon context and the breakfast? Well, the fact that a consumption context is deemed similarly appropriate for two products does not necessarily mean that all the emotions associated with them will be similarly elicited, and vice-versa. Grouping the participants according to how appropriate they found eating the products in each context (low-L, medium-M, and high-H, see Table 3) enabled additional understanding of the differences of elicitation of some emotions between the contexts and products. Note that, since the frequency ratings were not related at all with the appropriateness ratings, the possibility of obtaining these groups as an effect of differences in the use of the scale can be ruled out. Results from logistic regressions revealed that several emotions were associated differently between

285

B. Piqueras-Fiszman, S.R. Jaeger / Food Quality and Preference 32 (2014) 277–288 Table 3 Summary of participants who took part in the main study, divided according to their context-product appropriateness ratings. Apple appropriateness Level Low (1–3) Medium (4–6) High (7–9)

Breakfast N M (SD) 32 2 (0.9) 27 5.2 (0.8) 17 8.1 (0.9)

Chocolate brownie appropriateness Afternoon N M (SD) 16 2.1 (0.9) 14 5.5 (0.7) 46 8.1 (0.9)

Dinner N 42 18 16

the groups of participants. In the case of the apple, 9 emotion associations differed significantly in their frequency under the breakfast context, 12 under the afternoon context, and 10 under the dinner context. Several differences were also observed between the three different groups for chocolate brownie, the number of emotions that differed significantly among the groups was 8, 11, and 7 under the breakfast, afternoon, and dinner contexts, respectively. What is interesting is that of those emotion associations, participants in appropriateness groups H and M reported significantly more positive emotions, while the participants in group L reported significantly more negative emotions for all the contexts. Fig. 4a and b represent the spider plots of the data (as proportions) of participants in appropriateness groups L and H for the afternoon context ordered by the frequency of group L. Fig. 4a represents the apple data, while Fig. 4b represents the chocolate brownie data. An interesting observation from Fig. 4a is that, for the apple, bored was the emotion term with the highest frequency for appropriateness group L (75% of the total mentions), while good was the term with the highest frequency for the group H (76%). Similarly, disgusted was the third most selected term by group L (25%) while the other groups did not select that emotion. It is worth noting as well that the frequency of guilty was significantly higher for group L for the chocolate brownie for all of the contexts (only afternoon shown, Fig. 4b), and particularly higher for the afternoon context (which was the most selected term, 60% of mentions), while for the other appropriateness groups (i.e., M and H) the most selected terms were good and happy. 3.4. Effect of emotional eating behavior on the emotion responses Fig. 5a and b represent the spider plots of the proportioned frequency data only of the two groups of participants characterized as being either low emotional eaters (L), or high emotional eaters (H)

M (SD) 1.8 (0.8) 4.9 (0.7) 7.6 (0.9)

Breakfast N M (SD) 31 2.1 (0.9) 28 5.1 (0.8) 22 7.7 (0.8)

Afternoon N M (SD) 30 1.8 (0.9) 31 5.0 (0.8) 20 7.8 (0.9)

Dinner N 8 20 53

M (SD) 1.9 (0.8) 5.2 (0.9) 8.1 (0.9)

according to the DEBQ-e for the afternoon context (rest of plots not shown). By comparing Fig. 5a (apple) and Fig. 5b (chocolate brownie) it can be noticed that the differences between the three groups was larger for the chocolate brownie than for the apple stimulus, suggesting that the chocolate brownie is a product that discriminates to a greater extent the levels of consumers’ emotional eating behavior, compared with apple. In addition, for the chocolate brownie, the highly emotional eaters checked a larger number of emotion words and did so more frequently (e.g., joyful, tender, loving), even the negative emotion words, than the other two emotional eating groups across all the conditions. Logistic regression revealed that the number of emotion associations that differed significantly in their frequency between the three groups was 10, 8, 3, and 5 under no context, breakfast, afternoon, and dinner conditions, respectively (those for afternoon indicated in Fig. 5b). It is interesting to observe that while some emotion associations were highly mentioned for all the contexts and groups (e.g., happy, good, satisfied, pleasant), the frequency of other emotions varied noticeably from context to context, especially for group H (e.g., mild for under no-context, guilty under the afternoon context, and joyful, polite, daring, and affectionate under the dinner context. For the apple, the distinction between the emotional eating groups is not so clear: the number of associations selected and their frequency was not necessarily linked to the participants’ level of emotional eating behavior. In fact, the results from the logistic regression showed that the numbers of emotions significantly different among the groups were 3, 0, 3 and 4 under no context, breakfast, afternoon, and dinner conditions, respectively (indicated in Fig. 5a). In brief, these results seem to be in line with those from Jaeger and Hedderley (2013), who found that emotion intensity toward foods varied as a function of individual emotional traits. However, in this study it was observed that this only held true for the chocolate brownie, as opposed to the apple, which suggest that the

Fig. 4. Emotion profiles (% responses) of the products as reported by participants segmented according to how appropriate they found eating the focal product in the evoked afternoon context (on a 9-pt scale for appropriateness, L = 1–3 and H = 7–9; only groups L and H are shown to make the graphs easier to read). Emotions terms with ⁄⁄⁄ are significantly different among the three groups according to logistic regressions at a significance level of p < 0.001; ⁄⁄ at p < 0.01; and ⁄ at p < 0.05. (a) Apple; and (b) chocolate brownie.

286

B. Piqueras-Fiszman, S.R. Jaeger / Food Quality and Preference 32 (2014) 277–288

Fig. 5. Emotion profiles (% responses) of the products as reported by participants segmented according to their emotional eating behavior (measured with the DEBQ-e, on a 5-pt scale for emotional eating, L = <2 and H = P3; only groups L and H are shown to make the graphs easier to read). Emotions terms with ⁄⁄ are significantly different among the three groups according to logistic regressions at a significance level of p < 0.01; and ⁄ at p < 0.05. (a) Apple; and (b) chocolate brownie.

emotions elicited by an apple may not depend on the emotional eating behavior of the participants, regardless of the consumption context.

4. Discussion Emotion-related responses to foods are being studied with increasing frequency in consumer and sensory science, as are context-related responses; however, the combination of both measures has remained unaddressed until now. In this research study we aimed to bridge this gap and, additionally, address in an encompassing way several relevant issues surrounding the measurement of emotion responses under evoked consumption situations. We have considered a variety of interrelated variables and have investigated their effect on emotion responses toward foods. These variables were: (1) The emotional ‘footprint’ that a food per se has; (2) The dimensions of consumption of the evoked contexts; (3) The appropriateness between the foods and the contexts; and (4) The level of emotional eating behavior of the participants. Consideration of these variables is important since how we feel while eating depends on the context we happen to be in, on how appropriate we find to eat that food in that given context, and on our emotional state when eating. As seen throughout this paper, the emotion profiles reported under the three evoked contexts and under no context were very different, and the frequency of eliciting several emotion associations differed significantly from context to context, with the dinner context generally eliciting most emotion associations for the two foods. Therefore, it can be concluded that emotions associated to products are not situationally stable. Furthermore, the number of emotion associations that varied across contexts was different for apple and chocolate brownie (9 vs. 20, respectively). In addition, it was also observed that the difference between the two products varied depending on the context evoked. These results suggest that the context is an important determinant of the emotion word responses associated to the products. In fact, by looking at the emotion configurations of the contexts alone, we were able to explain some of the emotions elicited by the foods under those contexts. For instance, apparently the evoked afternoon context was associated mainly with worried and eager, the breakfast active, and, as expected, the dinner context was associated with terms such as

affective (Fig. 3b). These emotion associations were also among the most elicited for the products, and according to the NcNemar test (Table 2) a large number of the emotions elicited with the products did not differ significantly from those evoked by the context alone (23 and 16 for the chocolate brownie and apple, respectively), a fact that confirms that the context in which a food is consumed can be a strong determinant of the elicited emotion associations. Altogether, the results demonstrate that the elicitation of those associations seems to be caused by the reciprocal effect of both the product and the contexts in which they were imagined to be consumed. The situational/contextual dependence of consumer responses (to date limited to preference or acceptance measures; e.g., Cardello & Schutz, 1996; Schutz, 1994) to food products implicitly entails an appraisal of product-context appropriateness. In this study we measured the participants’ appropriateness perception of the products imagined to be consumed in each of the evoked contexts. The results shed light on some of the emotion results that otherwise would remain difficult to understand. For instance, differences in perceived appropriateness may be the reason why the apple and the chocolate brownie had different emotion profiles for the dinner context, since the appropriateness ratings for these two products under this context were significantly different. In addition, most of the negative emotions elicited under certain contexts could also be explained by taking into account the appropriateness ratings: for example, the chocolate brownie was associated with guilty under the afternoon context (for which it was considered on average as medium appropriate, M = 4.5), while under the dinner context it was associated with affection and politeness, among other positive emotions. Moreover, segmenting participants according to their appropriateness ratings helped to better understand the reason for the elicitation of negative emotion associations, since significant differences were observed among the appropriateness groups (i.e., those who found a product very appropriate to a context elicited more positive words, and with more frequency, while the opposite was observed for those who found it inappropriate). Taken together, these findings highlight the idiosyncratic nature of elicited emotion profiles, which may largely depend on the particular consumption habits of consumers (even when they belong to the same geographical region). Hence it could be said that fully predicting the influence that a certain consumption context might have on a product’s emotion profile would be difficult. Therefore,

B. Piqueras-Fiszman, S.R. Jaeger / Food Quality and Preference 32 (2014) 277–288

in future food-related emotion measurement involving contexts we recommend taking into account the emotion profiles of the contexts alone and the consumers’ product-context appropriateness appraisal, since in combination these types of information have proved very valuable. At this point relevant questions have not been discussed in detail, such as: When no consumption context is evoked, do consumers base their emotional responses on what they are feeling at that precise moment while completing the questionnaire? Or do they base their responses on what they would feel normally when consuming that food stimulus? Not providing a context to respondents may lead to uncertainty on how to interpret results or inferences drawn from inaccurate data. In fact, around 40% of participants wrote additional comments (spontaneously), and among the comments, several of those in the no-context condition were of the type: ‘‘What am I supposed to feel while eating an apple?’’, ‘‘I’m worried because I’m waiting for an important call’’, ‘‘Why would I feel polite while eating this food?’’ This does not mean that the participants’ responses under no context are wrong, but rather that the results should be interpreted with caution, since they might be irrelevant to the purpose of the task (Jaeger et al., 2013). This research has also explored the effect of one individual trait of participants on the emotion profiles obtained under the evoked consumption contexts; particularly, we explored the emotional eating behavior of participants, by means of using the emotion section of the DEBQ (Van Strien et al., 1986). We expected it to have a certain impact (and help explain) the emotion profiles observed, since previous research has demonstrated that the subjects’ general emotional character is an explanatory variable of the heterogeneity in emotional evaluation of certain foods (Jaeger & Hedderley, 2013). Accordingly, we found that differences in emotional eating behavior of participants can affect the frequency of emotion associations elicited under different contexts: the more emotional eaters selected more emotion terms (positive and negative) and more frequently; however, this was observed more clearly for the chocolate brownie than for the apple, which suggests that their level of emotional eating did not affect the frequency of elicitation of emotions. Regarding the tool to measure the emotion associations, we are aware that the study of emotions is quite paradoxical: Emotions are implicitly spontaneous and merely asking about them as ‘‘how do you feel?’’ (mostly with close-ended questions) could very likely induce them to report emotion-related responses that may be somewhat ‘‘artificial’’ (or uncertain as to what they are, as mentioned in the Introduction). Nevertheless, studying these responses can be very relevant, and to date surveys are still an easy and valuable approach if the results are carefully interpreted. In this study, we created an emotion questionnaire based on the wordlist of the EsSense ProfileÒ since it is a tool widely applied among food-related emotion researchers; however, we would hypothesize that similar results would be observed if the approach proposed here was applied to other food-related emotion techniques. Having said this, evoking consumption contexts using non-survey (or indirect) methods would be worth exploring too. In addition, giving participants the chance to write any comments or thoughts generated additional insights about the conditions and the reasons that underlie food emotions (though we included open ended questions mainly as a support tool). For instance, a participant in this study reported for the chocolate brownie under the breakfast context: ‘‘I really like brownie, but I find it less satisfactory to eat it for breakfast, because it’s the moment when I just wake up and I don’t enjoy food as much’’, and another participant wrote: ‘‘Eating an apple reminds me of when I’m dieting, I find that very boring’’. Comments like these demonstrate that food-related emotions can be built upon much more than on the sensory properties of the foods, such as associations, past experiences, and the contexts. Additionally,

287

some participants also suggested other terms not included in the list, such as unsatisfied or rewarded, facts which hint at the need of capturing prevalent feelings or emotions that consumers might experience throughout their interaction with a given product by developing a customized well-focused questionnaire (Richins, 1997; Thomson & Crocker, 2013). Regarding the presentation of the food stimuli as images, as mentioned earlier, we believe that it was an easy and convenient representation of the products that would possibly elicit emotion associations closer to tasted foods than food names would (regardless whether they are really associations or experienced emotions). Replicating the study with tasted foods or food names would be necessary to confirm this hypothesis. Concerning the evoked contexts/occasions, notwithstanding their representativeness for popular situations, more research should be performed to determine whether similar emotion profiles would be elicited in the real contexts to ensure external validity. However, our work was not aimed at checking the external validity, though that it is something that should be addressed in future studies. At the moment, what we have found is that the elicited emotion profiles are ‘‘put in context’’ and that the results make sense, hence our belief that these associations could possibly be closer to what they would experience in those situations, at least that is what we can rely on to date. Food developers, for instance, a beer company, might want to evoke a common consumption context, like a pub, say, with the aim of eliciting emotion associations that their consumers might feel in their real life, instead of having the participants respond focusing on the product alone. 5. Conclusions While sensory and consumer research has become more and more aware of the importance of gathering information regarding how a food/beverage makes consumers feel, little has been done to study the effect of the consumption context in emotion elicitation. The research presented here bridged this gap of knowledge and demonstrated that the same food can be associated with different emotion profiles depending on the context where it is imagined to be consumed in, and that the effect that the contexts may exert varies as well depending on the food in question. Additional pieces of information obtained, such as the participants’ product-context appropriateness appraisal and their emotional eating behavior, helped to better understand several underlying reasons of emotion responses. After all, one aim of food-related emotion research is to study which emotions would consumers normally experience in response to eating and tasting a food/beverage in their lives, and, if possible, what types of conditions are responsible for these emotions. We believe that this study represents a needed advance in this area of research and upon replication of the findings with different products, contexts, and/or experimental techniques, could be an easy-to-implement methodology to address this purpose (also when constructing the lexicon for a questionnaire) and possibly help obtain more accurate (faithful) emotion responses, closer to those that consumers would feel in real life. However, further research will also be needed to validate and corroborate the hypothesis that the emotion responses elicited under certain evoked contexts are closer to those experienced in the real contexts in totally natural conditions. Acknowledgements The authors would like to thank Puri García from the Department of Food Technology (Universitat Politécnica de Valencia) for helping recruit participants for this study, the Ministerio de Educación (Spain) for the PhD scholarship awarded to author BPF, and to

288

B. Piqueras-Fiszman, S.R. Jaeger / Food Quality and Preference 32 (2014) 277–288

the New Zealand Institute for Plant & Food Research Ltd. for contributing financially to data collection. Appendix A. Spanish translation of terms included in the questionnaire.

Terms in English

Translation in Spanish

Active Adventurous Affectionate Bored Calm Daring Disgusted Eager Energetic Enthusiastic Free Friendly Glad Good Good natured Guilty Happy Interested Joyful Loving Mild Nostalgic Peaceful Pleasant Pleased Polite Quiet Satisfied Secure Steady Tender Understanding Warm Whole Wild Worried

Activo Aventurero Afectuoso Aburrido Calmado Atrevido Repugnado Impaciente Energético Entusiasmado Libre Amigable Alegre Bien Bondadoso Culpable Feliz Interesado Jovial Cariñoso Apacible Nostálgico Pacífico Agradable Complacido Amable Reservado Satisfecho Seguro Firme Tierno Comprensivo Cálido Íntegro Rebelde Preocupado

References Bécue-Bertaut, M., Álvarez-Esteban, R., & Pagès, J. (2008). Rating of products through scores and free-text assertions: Comparing and combining both. Food Quality and Preference, 19, 122–134. Bécue-Bertaut, M., & Lê, S. (2011). Analysis of multilingual labeled sorting tasks: Application to a cross-cultural study in wine industry. Journal of Sensory Studies, 26, 299–310. Bécue-Bertaut, M., & Pagès, J. (2008). Multiple factor analysis and clustering of a mixture of quantitative, categorical and frequency data. Computational Statistics and Data Analysis, 52, 3255–3268. Bell, R., Meiselman, H. L., Pierson, B. J., & Reeve, W. G. (1994). Effects of adding an Italian theme to a restaurant on perceived ethnicity, acceptability, and selection of foods. Appetite, 22, 11–24. Bisogni, C., Falk, L. W., Madore, E., Blake, C., Jastran, M., Sobal, J., et al. (2007). Dimensions of everyday eating and drinking episodes. Appetite, 48, 218–231.

Cardello, A. V., Meiselman, H. L., Schutz, H. G., Craig, C., Given, Z., Lesher, L. L., et al. (2012). Measuring emotional responses to foods and food names using questionnaires. Food Quality and Preference, 24, 243–250. Cardello, A. V., & Schutz, H. G. (1996). Food appropriateness measures as an adjunct to consumer preference/acceptability evaluation. Food Quality and Preference, 7, 239–249. de Graaf, C., Cardello, A. V., Kramer, F. M., Lesher, L. L., Meiselman, H. L., & Schutz, H. G. (2005). A comparison between liking ratings obtained under laboratory and field conditions: The role of choice. Appetite, 44, 15–22. Desmet, P. M. A., & Schifferstein, H. N. J. (2008). Sources of positive and negative emotions in food experience. Appetite, 50, 290–301. Edwards, J. S. A., Meiselman, H. L., Edwards, A., & Lesher, L. L. (2003). The influence of eating location on the acceptability of identically prepared foods. Food Quality and Preference, 14, 647–652. Hein, K. A., Hamid, N., Jaeger, S. R., & Delahunty, C. M. (2010). Application of a written scenario to evoke a consumption context in a laboratory setting: Effects on hedonic ratings. Food Quality and Preference, 21, 410–416. Hein, K. A., Hamid, N., Jaeger, S. R., & Delahunty, C. M. (2012). Effects of evoked consumption contexts on hedonic ratings: A case study with two fruit beverages. Food Quality and Preference, 26, 35–44. Hersleth, M., Monteleone, E., Segtnan, A., & Næs, T. (2012). Consumers’ acceptance and probability of buying dry cured ham in two different meal contexts – effects of intrinsic and extrinsic product attributes. Presentation at the 5th European Conference on Sensory and Consumer Research, 9–12 September 2012. Bern, Switzerland. Jaeger, S. R., Cardello, A. V., & Schutz, H. G. (2013). Emotion questionnaires: A consumer-centric perspective. Food Quality and Preference, 30, 229–241. Jaeger, S. R., & Hedderley, D. I. (2013). Impact of individual differences in emotional intensity and private body consciousness on EsSense ProfileÒ responses. Food Quality and Preference, 27, 54–62. Jansen, A., Nederkoom, C., Roefs, A., Bongers, P., Teugels, T., & Havermans, R. (2010). The proof of the pudding is in the eating. The validity of the DEBQ external eating scale and the critical role of depression. Appetite, 55, 164. King, S. C., & Meiselman, H. L. (2010). Development of a method to measure consumer emotions associated with foods. Food Quality and Preference, 21, 168–177. King, S. C., Meiselman, H. L., & Carr, B. T. (2010). Measuring emotions associated with foods in consumer testing. Food Quality and Preference, 21, 1114–1116. King, S. C., Meiselman, H. L., & Carr, B. T. (2013). Measuring emotions associated with foods: Important elements of questionnaire and test design. Food Quality and Preference, 28, 8–16. King, S. C., Meiselman, H. L., Hottenstein, A. W., Work, T. M., & Cronk, V. (2007). The effects of contextual variables on food acceptability: A confirmatory study. Food Quality and Preference, 18, 58–65. Köster, E. P. (2003). The psychology of food choice: Some often encountered fallacies. Food Quality and Preference, 14, 359–373. Lê, S., Josse, J., & Husson, F. (2008). FactoMineR: An R package for multivariate analysis. Journal of Statistical Software, 25, 1–18. Levine, D. M., Stephan, D. F., Krehbiel, T. C., & Berenson, M. L. (2011). Statistics for managers (6th ed.). Boston, MA: Prentice Hall. Manzocco, L., Rumignani, A., & Lagazio, C. (2013). Emotional response to fruit salads with different visual quality. Food Quality and Preference, 28, 17–22. Moskowitz, H. R., Silcher, M., Beckley, J., Minkus-McKenna, D., & Mascuch, T. (2005). Sensory benefits, emotions and usage patterns for olives: Using Internet-based conjoint analysis and segmentation to understand patterns of response. Food Quality and Preference, 16, 369–382. Petit, C., & Sieffermann, J. M. (2007). Testing consumer preference for iced-coffee: Does the drinking environment have any influence? Food Quality and Preference, 18, 161–172. Richins, M. L. (1997). Measuring emotions in the consumption experience. Journal of Consumer Research, 24, 127–146. Schifferstein, H. N. J., & Desmet, P. M. A. (2010). Hedonic asymmetry in emotional response to consumer products. Food Quality and Preference, 21, 1104–1110. Schifferstein, H. N. J., Fenko, A., Desmet, P. M. A., Labbe, D., & Martin, N. (2013). Influence of package design on the dynamics of multisensory and emotional food experience. Food Quality and Preference, 27, 18–25. Schutz, H. G. (1994). Appropriateness as a measure of the cognitive-contextual aspects of food acceptance. In H. J. H. MacFie & D. M. H. Thomson (Eds.), Measurement of food preferences (pp. 25–50). London: Blackie Academic. Stroebele, N., & de Castro, J. M. (2004). Effect of ambience on food intake and food choice. Nutrition, 20, 821–838. Thomson, D. M. H., & Crocker, C. (2013). A data-driven classification of feelings. Food Quality and Preference, 27, 137–152. Thomson, D. M. H., Crocker, C., & Marketo, C. (2010). Linking sensory characteristics to emotions: An example using dark chocolate. Food Quality and Preference, 21, 1117–1125. Van Strien, T., Frijters, J. E. R., Bergers, G. P. A., & Defares, P. B. (1986). The dutch eating behavior questionnaire (DEBQ) for assessment of restrained, emotional, and external eating behavior. International Journal of Eating Disorders, 5, 295–315.