Variability of Affective Responses to Odors: Culture, Gender, and Olfactory Knowledge

Chemical Senses, Feb 2013

Emotion and odor scales (EOS) measuring odor-related affective feelings were recently developed for three different countries (Switzerland, United Kingdom, and Singapore). The first aim of this study was to investigate gender and cultural differences in verbal affective response to odors, measured with EOS and the usual pleasantness scale. To better understand this variability, the second aim was to investigate the link between affective reports and olfactory knowledge (familiarity and identification). Responses of 772 participants smelling 56–59 odors were collected in the three countries. Women rated odors as more intense and identified them better in all countries, but no reliable sex differences were found for verbal affective responses to odors. Disgust-related feelings revealed odor-dependent sex differences, due to sex differences in identification and categorization. Further, increased odor knowledge was related to more positive affects as reported with pleasantness and odor-related feeling evaluations, which can be related to top-down influences on odor representation. These top-down influences were thought, for example, to relate to beliefs about odor properties or to categorization (edible vs. nonedible). Finally, the link between odor knowledge and olfactory affect was generally asymmetrical and significant only for pleasant odors, not for unpleasant ones that seemed to be more resistant to cognitive influences. This study, for the first time using emotional scales that are appropriate to the olfactory domain, brings new insights into the variability of affective responses to odors and its relationship to odor knowledge.

A PDF file should load here. If you do not see its contents the file may be temporarily unavailable at the journal website or you do not have a PDF plug-in installed and enabled in your browser.

Alternatively, you can download the file locally and open with any standalone PDF reader:

Variability of Affective Responses to Odors: Culture, Gender, and Olfactory Knowledge

CamilleFerdenzi 1 2 S. Craig Roberts 0 3 AnnettSchirmer 6 SylvainDelplanque 1 2 SezenCekic 2 4 ChristellePorcherot 5 IsabelleCayeux 5 DavidSander 1 2 DidierGrandjean 2 4 0 School of Biological Sciences, University of Liverpool , Crown St., Liverpool L69 7ZB , United Kingdom 1 Department of Psychology, Laboratory for the Study of Emotion Elicitation and Expression (E3 Lab), FPSE, University of Geneva , 40 Bld du Pont d'Arve, 1205 Geneva , Switzerland 2 Swiss Center for Affective Sciences, University of Geneva , 7 rue des Battoirs, 1205 Geneva , Switzerland 3 Present address: School of Natural Sciences, University of Stirling , Stirling FK9 4LA , United Kingdom 4 Department of Psychology, Neuroscience of Emotion and Affective Dynamics Laboratory (NEAD Lab), FPSE, University of Geneva , 40 Bld du Pont d'Arve, 1205 Geneva , Switzerland 5 Firmenich SA , route des Jeunes 1, PO Box 239, 1211 Geneva 8 , Switzerland 6 Department of Psychology, Faculty of Arts and Social Sciences, National University of Singapore , Lower Kent Ridge Road , Singapore 119099 , Singapore Emotion and odor scales (EOS) measuring odor-related affective feelings were recently developed for three different countries (Switzerland, United Kingdom, and Singapore). The first aim of this study was to investigate gender and cultural differences in verbal affective response to odors, measured with EOS and the usual pleasantness scale. To better understand this variability, the second aim was to investigate the link between affective reports and olfactory knowledge (familiarity and identification). Responses of 772 participants smelling 56-59 odors were collected in the three countries. Women rated odors as more intense and identified them better in all countries, but no reliable sex differences were found for verbal affective responses to odors. Disgust-related feelings revealed odor-dependent sex differences, due to sex differences in identification and categorization. Further, increased odor knowledge was related to more positive affects as reported with pleasantness and odor-related feeling evaluations, which can be related to top-down influences on odor representation. These top-down influences were thought, for example, to relate to beliefs about odor properties or to categorization (edible vs. nonedible). Finally, the link between odor knowledge and olfactory affect was generally asymmetrical and significant only for pleasant odors, not for unpleasant ones that seemed to be more resistant to cognitive influences. This study, for the first time using emotional scales that are appropriate to the olfactory domain, brings new insights into the variability of affective responses to odors and its relationship to odor knowledge. Introduction The perception of odors is frequently associated with affective responses that are prone to interindividual variation. Studies on olfactory abilities (such as identification and sensitivity) usually report two major factors of interindividual variability: gender and culture (e.g., Herz 2009). Here, one of our main aim was to investigate gender and cultural variability specifically in verbal affective responses to odors, using a tool developed recently to evaluate peoples feelings related to odor perception (emotion and odor scales [EOSs]; Chrea etal. 2009; Ferdenzi etal. 2011). In terms of odor detection, there is no conclusive empirical evidence for gender differences (for a review see Doty and Cameron 2009). Nevertheless, where gender differences exist, women are usually more sensitive to odors compared with men (e.g., Koelega 1994; Kobal etal. 2001). For example, women do seem to be more easily sensitized to odors (Dalton et al. 2002), which might explain why they suffer from odor intolerance more than men (Nordin etal. 2004). Studies using verbal reports revealed that, in everyday settings, women care about olfaction more than men do, compared with other sensory modalities (Herz and Inzlicht 2002; Havlicek et al. 2008). Furthermore, women are better at identifying and memorizing odors of various origins, such as food or body odors (Schleidt etal. 1981; Doty etal. 1984; Larsson et al. 2003). Differences in measured and verbally reported olfactory skills may be considered reliable because they have been reported repeatedly and because they are present early in development (Richman etal. 1995; Mallet and Schaal 1998; Choudhury etal. 2003; Ferdenzi etal. 2008). In terms of emotional responses to odors, women report more frequent evocations of emotional memories by odors and stronger feelings of happiness, sadness, well-being, and reduction of stress as a consequence of smelling odors (Martin etal. 2001). Using presentation of real odors, several studies have confirmed these gender differences in various aspects of olfactory emotional responses. Women give lower hedonic ratings than men to the unpleasant odor of pyridine (Olofsson and Nordin 2004) and to human body odors (Doty etal. 1975, 1982). This greater affective reactivity to odors is also expressed in greater electrophysiological responses in women relative to men (Olofsson and Nordin 2004; Pause et al. 2010). However, not all odors are associated with a greater female affectivity: some odors trigger more positive or negative hedonic responses in one gender, depending on geographical location. Indeed, gender differences seem to depend also on cultural factors, as shown in the National Geographic Smell Survey (Wysocki etal. 1991). Unfortunately, such differences remain difficult to explain and have barely been exploredsince. To our knowledge, studies using a more comprehensive approach of odor-related affective feelings (i.e., using a larger variety of affective measures than just hedonicity) have not investigated interindividual variability (e.g., Desmet and Schifferstein 2008; Churchill and Behan 2010; King and Meiselman 2010). Such approaches are nonetheless needed to fully investigate gender and cultural differences in affective responses to odors. Gender and cross-cultural studies in odor perception, in turn, have mainly focused on odor hedonicity, which is only a limited aspect of odor-related affective feelings (Delplanque etal. 2012). Two major conclusions can be drawn from these studies. On one hand, geographic variation in hedonic ratings has been found in two major large-scale studies (Pangborn et al. 1988; Wysocki etal. 1991), with higher pleasantness attributed to odors encountered more frequently or to odors contained in products that have more positive connotations (e.g., an odor present in candies vs. medication). On the other hand, a certain degree of consensus can also be found between cultures; for example, there seem to be convergent negative evaluations of the odors of decaying organic matter, feces, and body odors in European, Asian, and American populations (Schleidt et al. 1988; Schaal et al. 1997). Further investigation with a wider field of affective responses than simply hedonics is now enabled by the EOSs (Chrea etal. 2009; Ferdenzi etal. 2011). To better understand gender and culture variation of affective responses to odors, it seemed crucial to investigate the link between these responses and perceivers knowledge about the odors. Indeed, there is some evidence that odor knowledge, that is, odor identification (naming) and familiarity (feeling of knowing), 1) differ as a function of perceivers sex and culture, and 2)is linked to intensity and valence of affective response to odors. For example, positive relationships between familiarity and pleasantness of odors have been recurrently shown (Jellinek and Kster 1983; Engen 1988; Rabin and Cain 1989; Distel etal. 1999), as well as positive relationships between identification and pleasantness (Ayabe-Kanamura etal. 1998b; Distel and Hudson 2001; Herz 2003; Djordjevic etal. 2008; Rouby etal. 2009). Moreover, shifting the valence of the context of odor presentation has been shown to dramatically change the affective valence of an odor. Herz and von Clef (2001) reported lower hedonic evaluations of the mixture butyric/isovaleric acids when it was labeled vomit than when it was labeled parmesan cheese. Similarly, different beliefs associated to a given odor are strong modulators of how this odor will be perceived (e.g., hazardous vs. healthy attribution; Dalton 1996) and of the perceivers physiological response (e.g., stimulating vs. relaxing attributions; Campenni etal. 2004). Although other factors might be influential (e.g., physicochemical properties of odorous molecules; Khan etal. 2007), top-down influences are thus believed to be an important factor for the determination of odor affective tone. We propose to investigate this further in this study (which constitutes our second main aim), using more sophisticated measures of affective responses to smells. In this study, we investigated verbal affective response to odors, using not only classical hedonic ratings but also, and especially, the newly developed EOSs, comprising three published versions for two European countries (United Kingdom and Switzerland) and one Asian country (Singapore). These scales include more than 30 different affective terms, organized in 6 to 7 main categories of feelings and meant to be rated for their perceived intensity resulting from the perception of odors. These affective terms were selected because they were evaluated by participants belonging to each culture as the most relevant of a large set of adjectives (see Chrea etal. 2009; Ferdenzi etal. 2011). They encompass terms related to happiness/wellbeing, energy, sensuality/desire, and disgust in the three countries, and several other culture-specific categories (see Materials and methods). We first decided to investigate the effects of gender and country on affective ratings, familiarity, and identification, with the hypothesis that women might report stronger verbal affective responses to odors than men. Second, we explored some aspects of the link between odorrelated affective feelings (hedonic ratings and EOS ratings) and odor knowledge (familiarity and identification). Materials and methods Participants The participants were recruited from the general public, in a Science Fair in Geneva (Nuit de la Science; N=151 females and 59 males, mean age standard deviation=37.8 12.1years), in the World Museum of Liverpool (N = 207 females and 144 males, aged 32.3 13.8 years), and in the Science Center of Singapore (N = 124 females and 87 males, aged 30.0 9.0years). Participants had spent most of their life in the countries where the experiment took place (or in one of the adjacent countries with the same language, e.g., France for the Swiss sample, and Ireland for the British sample). The experiment was performed in the official languages of the countries, namely French in Geneva and English in Liverpool and Singapore. Informed written consent was obtained prior to participation. Committees on Research Ethics of the University of Geneva, the University of Liverpool, and the National University of Singapore approved the study. Affective ratings The Geneva, Liverpool, and Singapore EOSs (GEOS, LEOS, and SEOS; Chrea et al. 2009; Ferdenzi et al. 2011) were used to measure affective responses to odors of participants belonging to the respective cultures. The scales consist of 36 or 37 affective terms organized in 6 or 7 categories (cf. Appendix A). Happiness/well-being, energy, sensuality/desire, and disgust are categories common to the three countries; soothing/peacefulness is common to the two European countries and several categories such as sensory pleasure (Geneva), nostalgia and hunger/thirst (Liverpool), intellectual stimulation, spirituality, and negative feelings (Singapore) are country specific. The procedure used to develop the scales and consisting in identifying the most relevant terms among a wide range of emotions sensu stricto (see Scherer 2005), moods, personality traits, and attitudes, is described in detail in Chrea etal. (2009) and Ferdenzi etal. (2011). A total of 56 odorous stimuli were used in Liverpool and Geneva, and 59 were used in Singapore (Appendix B). They represented a large range of everyday odors including: 1)as many pleasant as unpleasant odors, 2)a high proportion of familiar odors to elicit affective reactions linked to autobiographical memories (including culture-specific odors, such as durian in Singapore), and 3) odors related to various contexts (food: sweet, savory, fruits, spices, drinks, vegetables; and nonfood: cosmetic, household, woody, plants, animals, floral, medicine). The odorous substances, provided by Firmenich SA, Geneva, were diluted in odorless dipropylene glycol to obtain similar subjective intensities (see Delplanque etal. 2008; Chrea etal. 2009). Pen-like devices (Sniffin Sticks) were filled with 7mL of each diluted solution and coded with a three-digit number. To limit olfactory fatigue and test duration, each participant evaluated a subset of seven or eight odors (eight subsets in total, see Appendix B). During data collection, the odors were presented in random order. Procedure The participants took part in a 20-min session either under a tent outside (Geneva) or in a well-ventilated room (World Museum of Liverpool, Science Center of Singapore). After having smelled each of the seven or eight odors, respondents were asked to rate the intensity of their feelings with the help of the different affective terms. They were presented with the affective terms on a computer interface and gave their answers using a visual analog scale labeled from not at all to extremely, subsequently translated into a 0200 score. For each odor, affective ratings were followed by familiarity, pleasantness, and intensity ratings on similar scales and, in Liverpool and Singapore only, by free odor identification. In total, each odor was evaluated by 2032 participants in Geneva, 4146 in Liverpool, and 2428 in Singapore. Score computation Familiarity, pleasantness, and intensity raw scores were used (comprised between 0 and 200) in the odor-based analyses (scores averaged by odor, see Statistical analyses). In the rater-based analyses (raw scores, see Statistical analyses), the scores were transformed into categorical variables with three modalities (0 for scores comprised between 0 and 66, 0.5 for scores between 67 and 133, and 1 for scores between 134 and 200). The latter transformation was used because the scores did not follow a normal distribution but, based on visual inspection, a trimodal distribution. Identification scores were computed as follows (data available in Liverpool and Singapore only). For each odor, participants received the score of 0 when they gave no answer or a wrong answer (e.g., banana for soya bean), 0.5 when they gave an answer that was almost correct (e.g., tau hway, which is a kind of tofu pudding made of soya bean curd, for the odor of soya bean, or orange for the odor of grapefruit), and 1 when they gave the correct answer (e.g., soya bean or soya bean milk for the odor of soya bean). For the odor-based analyses, the percentages of correct answers by odor were computed, by summing the scores of all participants having evaluated this odor, dividing it by the number of participants, and multiplying it by100. EOS affective responses of each participant to each odor were summarized by using the factor scores, namely the coordinates of a given odor rated by a given participant on the factor formed by a group of affective terms (i.e., a category of feelings such as energy) (M-Plus v.6). All obtained factor scores were then shifted by +100 to obtain positive values only. It must be kept in mind that these summarized (factor) scores are not based on exactly the same individual terms in the three countries (although the categories of feelings bear the same title, cf. Appendix A). In addition, three different affective scores were attributed to each odor for some of the odor-based analyses. Namely, they are the average factor scores of the participants who: 1) successfully identified the odor (i.e., correct and almost correct answers), 2) misidentified the odor (wrong answers), and 3) did not identify the odor (no answer). We considered these average affective scores to be meaningful only when their computation was based on the scores of at least five participants: therefore, odors with insufficient number of participants in at least one of these three subgroups were removed from the analysis (for example, not enough participants provided an incorrect identification for the odor of peppermint, and not enough participants successfully identified the odor of fig). Out of 56 odors, this represented 25 odors in Liverpool, 14 in Singapore, and 38 when the data of both countries were pooled. Statistical analyses Rater-based analyses The first series of analyses consisted in testing the effects of gender and country on the olfactory variables (with R v.2.13.1; see Here, raw data (nonaveraged) were used. Because the EOS affective factor scores had a gamma rather than a normal distribution, we used a general linear modeling (GLM) procedure taking into account the gamma distribution. This GLM investigated the main effects of gender and country on EOS scores while controlling for the main effects of odor, familiarity, and intensity. A similar procedure was used to test the interactions gender by country and gender by odor: in these cases, the analyses controlled for the main effects of gender, country when applicable (so that only the effect of the interaction per se is tested), and also for odor, familiarity, and intensity. Finally, these main effects and interactions were also tested on familiarity, intensity, pleasantness, and identification while controlling for the main effects of odor only. Odor-based analyses This second series of analyses used scores averaged by odor to investigate the links between odor knowledge (familiarity and identification) and affective responses (pleasantness and EOS ratings) (with Statistica v.9). First, we correlated pleasantness ratings with familiarity ratings and percentage of identification (Pearson correlation coefficients). This was performed separately for two groups of odors: those below and those above the median of the average pleasantness rating. We tested the difference between correlation coefficients for both groups of odors. Second, a similar approach was conducted for EOS affective ratings, and in addition, we conducted repeated-measures analyses of variance (ANOVAs), with identification (correct, wrong, not identified) and EOS affective categories (energy, disgust, etc.) as between-odor factors. To qualify the differences between the three groups of identification, repeated-measures ANOVAs with identification as between-odor factor were run for each EOS affective category separately and were followed by post-hoc Tukey honestly significant difference tests. To refine this question at the odor level, we first computed the EOS score difference between the conditions correct identification and no identification. This was done for each odor, and for each of the five affective categories found to significantly vary as a function of identification in the previous ANOVAs. We then performed a cluster analysis (Wards method on City-block [Manhattan] distances) on 22 eligible odors (i.e., for which average scores were available in all retained EOS categories). With this analysis, different patterns of influence of identification on EOS affective ratings can be identified. Results Gender and country differences (rater-based analyses) Women rated the odors as significantly more intense than men did and they had better identification scores than men (ps < 0.00025; all probabilities were Bonferroni corrected, i.e., divided by four as there were four tests for gender, country, gender by country, and gender by odor effects). Women did not differ from men for familiarity and pleasantness ratings. These effects were not odor- or country-dependent (no significant gender by odor/country interactions). On the contrary, men gave higher EOS affective ratings than women on happiness/well-being, sensuality/desire, and energy (ps<0.00025; no significant difference on disgust), and the direction of these differences was maintained even when we conducted the GLM without controlling for other variables (odor, intensity, familiarity). However, the significant gender by country interactions, obtained for happiness/well-being, sensuality/desire (ps<0.00025), energy but also disgust (ps<0.0025), tell us that the gender differences in favor of men are exclusively present in the Swiss sample (Bonferroni-corrected posthoc contrasts; no significant gender differences in the other countries). For disgust only, there was also a significant gender by odor interaction (p < 0.0025) suggesting that the effect of gender was odor-dependent (for detailed results by odor, see Figure1). Finally, there were significant country differences on almost all olfactory variables. Pleasantness, familiarity, and intensity (ps < 0.00025), but not identification, were significantly lower in Singapore than in both other countries and higher in Geneva than in both other countries (except pleasantness that did not differ between Liverpool and Geneva; post-hoc contrasts). There were significant country differences for all four EOS affective categories (p < 0.0125 for disgust and ps < 0.00025 for the others). The largest cultural differences are illustrated in Figure2. Link between olfactory knowledge and pleasantness (odor-based analyses) The correlations between familiarity and pleasantness were not significant for unpleasant (U) odors (Geneva: rU=0.36; Liverpool: rU=0.39; Singapore: rU=0.33; see N and ps in Figure 3), but significantly positive for pleasant (P) odors (Geneva: rP=0.71; Liverpool: rP=0.79; Singapore: rP=0.80; Figure 3). Subsequent experiments in four additional geographic areas replicated this pattern (see Supplementary material 1). The correlation coefficients for unpleasant and pleasant odors significantly differed in Geneva (p < 0.05), Liverpool (p < 0.05), and Singapore (p < 0.01; one-tailed tests justified by previous work allowing to predict the direction of these differences; Delplanque etal. 2008). Unsurprisingly, the percentage of identification (correct or not) and the percentage of correct identification were highly correlated to familiarity (r=0.85 and 0.66 in Liverpool, and 0.83 and 0.53 in Singapore, respectively, ps<0.001; no identification data in Geneva). We thus tested whether there also was an asymmetry between unpleasant (U) and pleasant (P) odors for identification-pleasantness correlations. It was the case, but the unpleasant-pleasant difference was significant only in Liverpool (identification: rU = 0.10 and rP = 0.63, coefficients difference significant at p < 0.05, one-tailed; correct identification: rU=0.18 and rP=0.69, difference significant at p<0.05). In Singapore, although the pattern was similar, that is, correlations were lower for unpleasant odors, there was not significant unpleasant-pleasant difference (identification: rU=0.31 and rP=0.55; correct identification: rU = 0.23 and rP = 0.58; coefficients difference not significant). For both countries, rsP were significant at ps <0.003, and rsU were not significant (ps > 0.094) (see Supplementary material2). Note that the group of pleasant odors was rated as more familiar on average than the group of unpleasant odors (Geneva: 131 vs. 86, t54 = 6.81, p < 0.001; Liverpool: 117 vs. 69, t54=6.24, p<0.001; Singapore: 98 vs. 57, t57=6.35, p<0.001). Pleasant odors also triggered verbal labels more often than unpleasant odors (Liverpool: 51% vs. 38%, t54 = 3.87, p < .001; Singapore: 54% vs. 37%, t57 = 4.17, p < .001), but not significantly more correct labels. Note also that in this whole section, sex-separated analyses showed same results as for the whole group. Link between olfactory knowledge and EOS ratings (odor-based analyses) As for pleasantness ratings, EOS affective ratings tended to show in many cases a similar asymmetry between unpleasant (U) and pleasant (P) odors. This asymmetry was again characterized by 1)significant links (ps <0.05) with familiarity/identification/correct identification for P odors and nonsignificant links (ps > 0.05) for U odors, and 2)stronger correlation coefficients for P than for U odors (one-tailed tests). Correlations were below 0 for negatively connoted affective categories and generally positive for the other affective categories. If many affective categories met the statistical rules enunciated above, it must be noted that not all did in all countries, which makes the U-P asymmetry for EOS affective ratings a tendency rather than a rule (see Supplementary material 2). Again, sexseparated results did not differ from the wholegroup. Further, we compared EOS affective ratings of participants who successfully identified (correctly or almost correctly), misidentified (wrong answer), or did not identify (no answer) the odors. There were highly significant interactions (ps<0.001) between identification (correct, wrong, not Figure 3 Pearson correlation (r) between pleasantness and familiarity for unpleasant and pleasant odors in Geneva, Liverpool, and Singapore. ***p<0.000167 (i.e., 0.001/6, Bonferroni correction), ns: not significant or p > 0.0083 (i.e., 0.05/6, Bonferroni correction). identified) and EOS affective categories (energy, disgust, etc.), when considering the data from Liverpool (F12,288 = 5.44; 7 affective categories, 25 odors), Singapore (F12,156 = 4.01; 7 affective categories, 14 odors), and from both countries together (F6,222 = 7.77; 4 common affective categories, 38 odors). ANOVAs per affective category revealed that identification groups significantly differed for five EOS categories (Figure 4A): happiness/well-being (F2,74 = 7.95, p < 0.001), nostalgia (F2,48 = 5.76, p < 0.01), intellectual stimulation (F2,26 = 3.58, p < 0.05), energy (F2,74 = 6.69, p < 0.01), and disgust (F2,74 = 4.86, p < 0.05). Post-hoc tests showed that compared with unidentified odors, correctly identified odors received significantly lower disgust ratings and significantly higher ratings on the other four EOS affective categories. According to these post-hoc tests, the wrong identification category always had an intermediate position (not significantly different from at least one of the other identification groups according to post-hoc tests); it was, therefore, not included in the subsequent cluster analyses and interpretations. To interpret these average effects further, with a finer approach at the odor level, we ran a cluster analysis on the variation of EOS affective scores (between correctly identified and unidentified), which allowed us to characterize five types of patterns. The number of clusters (five) was set based on visual determination of the inflection point on the plot of linkage distances. Cluster 1, constituted by the odors of tangerine, caramel, strawberry, and peppermint, was characterized by particularly large variations in happiness/well-being (score variation: +32) and intellectual stimulation (+30), the variations in energy (+18), nostalgia (2), and disgust (8) being more moderate. Cluster 2 (lavender, grapefruit) was characterized by large variations in happiness/well-being (+63), nostalgia (+42), energy (+39), and small variations in intellectual stimulation (+6) and disgust (6). In Cluster 3 (laundry soap, civet, eucalyptus, cigarette smoke, cheese), no noticeable variations were found: disgust +7, nostalgia 0, energy 2, intellectual stimulation 2, happiness/well-being 3. Cluster 4 (beer, coffee, shampoo, cream strawberry, floral strawberry) was mainly characterized by variation in nostalgia (+23) and less by variations in happiness/well-being (+14), energy (+12), intellectual stimulation (+3), and disgust (7). Finally, large variation in disgust (27) was the main characteristic of Cluster 5 (clove, fried shallot, cucumber, dirty socks, fire smoke/smoked ham, beef), variations in intellectual stimulation (+8), happiness/well-being (+5), nostalgia (+5), and energy (2) being more limited. To sum up, the results indicate that, although affective feelings triggered by some odors remain unchanged regardless of whether the odor is correctly identified (Cluster 3), affective feelings elicited by other odors are affected: either intellectual stimulation and happiness/well-being are increased by correct identification (Cluster 1), or nostalgia, energy, and happiness/well-being are increased (Cluster 2), or nostalgia only is increased (Cluster 4), or disgust only is decreased (Cluster 5). Discussion In this study, we investigated the variability of affective responses to odors, measured by classical hedonic scales, but also, and especially, by the culture-specific EOSs developed recently (Chrea et al. 2009; Ferdenzi et al. 2011). The first aim of our study was to investigate culture and gender differences in odor-induced reported feelings. First, we found that the odors were less familiar, less intense, and less pleasant in Singapore than in the European Figure 4 Average EOS affective scores (mean standard error of the mean) of participants who successfully identified (correctly or almost correctly), wrongly identified, and did not identify the odors. Averages were computed on 25 odors for affective categories specific to Liverpool (L), on 14 odors for categories specific to Singapore (S), and on 38 odors for categories common to both countries (L + S). (A) Affective categories that were significantly influenced by identification; (B) Affective categories that were not (repeated-measures ANOVAs per category). countries (when controlling for the factor odor). This could be due to differences in the olfactory environment of Europe and Asia, with the former being less odorous than the latter. If true, Singaporeans may be more habituated to olfactory experiences and may hence be less sensitive. It cannot be excluded though that these differences could be due to the fact that the odorous substances used in this study were manufactured in a European country and chosen by European experimenters, despite the great effort to choose universal odors (e.g., peppermint, strawberry, caramel) and odors specific to Singapore (e.g., pandan, soy, durian). Regarding the EOS ratings, cultural differences were found to be highly odor-dependent: among others, strawberry elicited less positive affective responses (sensuality/desire, energy, and soothing/peacefulness) in Switzerland than in the other countries, and durian elicited much less disgust in Singapore (see Figure2), where it is a very popular fruit (in western countries, its unfamiliar odor is often described as decaying matter; see also Ferdenzi etal. 2011). Second, we found that women rated the odors as more intense than men did, independently of their nationality, which is consistent with some studies on odor threshold detection (Koelega 1994; Kobal et al. 2001; Dalton et al. 2002). Women in the three countries were also found to be better than men at correctly identifying odors, which is an unanimous finding in the literature (e.g., Doty et al. 1984; Larsson et al. 2003) and could be related to gender differences in verbal proficiency. Women indeed have better verbal proficiency/access to semantic knowledge (e.g., Larsson etal. 2003), which is a significant predictor of odor identification (Larsson etal. 2000). However, we found that men expressed more intense feelings to odors. Specifically, they reported stronger feelings related to happiness/well-being, sensuality/desire, and energy. This effect is particularly surprising because 1)one could have hypothesized that womens better identification abilities would enhance their emotional responses to odors (as expected from our results), 2) in other studies, women had more extreme hedonic responses to odors (Doty et al. 1975, 1982; Olofsson and Nordin 2004), and 3)more generally, women tend to verbally report more intense emotions than men (for a review, see Brody and Hall 2008). However, the mens stronger affective verbal responses were only due to the Swiss sample. This could be related to the fact that the proportion of men was lower in Geneva than in the two other countries specifically for the age groups 3050years old (25% men, vs. 38% and 47% in Liverpool and Singapore; chi square on the sample sizes: 2 = 5.10, p < 0.05 and 2 = 9.60, p < 0.01, respectively) and over 50years old (14%, vs. 49% and 40% in Liverpool and Singapore; 2=8.23, p<0.01 and 2=1.89, p<0.20, respectively); there was no difference for the 1630year olds (39% men, vs. 41% and 38% in Liverpool and Singapore; chi square on the sample sizes: 2=0.08, p>0.70 and 2=0.03, p>0.80, respectively). Indeed, intensity of both positive and negative experienced emotions is likely to decrease with age (Fernandez-Ballesteros etal. 2010). Thus, the gender difference we found in favor of men might be confounded with an effect of age, which is an aspect of odor-related affective responses that deserves attention in future research. As suggested in the National Geographic Smell Survey (Wysocki et al. 1991), we also found that the direction of gender differences depended on the odors evaluated. Especially, we found that women had stronger reactions than men on the EOS category disgust for human/animal odors (body odor, dirty socks, leather), food odors related to milk (cheese, yogurt), sulfuric products (fried shallot, onion), and for smoky odors (fire smoke, cigarette smoke; Figure1a), whereas men were more repelled by vegetal and floral odors (cucumber, vetyver, magnolia etc.; Figure 1b). Although enhanced negative reactions to human odors in women compared with men have already been described in other studies (breath, axillary sweat: Doty 1986; Stevenson and Repacholi 2003), there is no pre-existing evidence of mens greater disgust reactions to particular odors. These differences might be due, at least in part, to the extent to which participants of each sex correctly identified unpleasant odors. Indeed, among the largest differences between the sexes (see Figure1), odors that were rated more disgusting by women were also generally better identified by them (body odor, cheese: 35% more correct answers in women than in men), and similarly for men (beef, rum, durian: 13% more correct answers in men). For other odors, how they were categorized during identification might play an important role (independently of the correctness of identification). For example, compared with women, men provided more often negative (incorrect) terms to qualify the odor of honey (such as chemicals/ammonia/urine: 40% of their answers vs. only 20% of womens answers). This link between identification and affective response to odors was explored further in our study. Our second aim was indeed to investigate the link between affective variables (pleasantness and EOS scales) and odor knowledge measured by familiarity (feeling of knowing) and identification (explicit verbal associations). Overall, we found that higher familiarity and more frequent or correct identification were associated to more positive and less negative affects. This result, reported in the literature for the classical pleasantness scale (e.g., Distel etal. 1999; Herz 2003; even in young subjects: Bensafi etal. 2007) was confirmed here, with both a pleasantness scale and a finer measure of odor-related feelings and proved to be specifically true for pleasant odors. The odors forming the pleasant group were also found to be more familiar and to trigger more identification attempts (correct or not) than the odors of the unpleasantgroup. The positive relationship between odor knowledge and hedonic/affective response to odors were probably the consequence of top-down modulation of odor perception, like in studies that manipulated the valence and the availability of verbal information attached to the odor and showed a modulation of hedonic ratings in line with the connotation of the verbal association (e.g., Herz 2003). Here, we assume (also because we noticed that during data collection) that participants tried to identify the odor before starting the affective ratings, even if the identification question was presented after, and that it might have influenced the subsequent ratings. The same might have occurred for familiarity, which represents prior knowledge of the odor in a wider sense than identification (identification is the odor-specific semantic knowledge, probably corresponding to high familiarity, whereas familiarity is the feeling of knowing not systematically associated to a precise odor name; Larsson 1997). These top-down influences have been interpreted before according to the organization of odors in an associative verbal network related to individuals past experience, and to the fact that connotation of the odor source (more than the pure olfactory sensation) or of the context in which the odor was encountered in the past drives emotional responses to the odor (Herz 2003). However, these influences were asymmetrical. The positive link between odor knowledge (familiarity/identification) and affective ratings (pleasantness, EOS) was revealed only for pleasant odors, not for unpleasant ones. This pattern has been obtained previously in Japanese, German, and Mexican samples (Ayabe-Kanamura etal. 1998a), more recently in a Swiss sample (Delplanque et al. 2008), and negative odors were shown to be less affected than positive ones by cognitive influences (Herz 2003). According to the latter author, the negative odors could have been more quickly and more superficially analyzed due to the unpleasant experience of smelling them, consequently limiting the depth of their cognitive treatment. Alternatively, perception of unpleasant odors may be based more on bottom-up mechanisms, where for example hedonicity of monomolecular odorous compound could depend on its physicochemical properties (e.g., Khan et al. 2007). This would generate a more stable negative sensory sensation and this would make these indicators of potential threats (e.g., odor of spoiled food) less malleable to cognitive influences and more able to maintain a good (and adaptive) level of alertness. Even if complex mixtures of compounds were used, such a phenomenon might be in play for the odors of Cluster 3 (see Results), whose EOS affective ratings, unlike other odors used in our study, were unaffected by identification. This cluster involves some of the most unpleasant odors of the set (civet, cigarette smoke, cheese): they might have very strong intrinsic perceptual properties making them resistant to elaborated cognitive influences. Note that it cannot be excluded that several very positive odors, such as laundry soap and eucalyptus also comprised in Cluster 3, also have particularly resistant perceptual properties. Future research might provide insights into the neurophysiological bases of such variability in susceptibility to cognitive influences. As mentioned in the Results section, the link between odor knowledge and our newly developed EOS affective categories varied as a function of the category. The cluster analysis provided an illustration of this, showing that the influence of identification on emotional response forms different patterns for different odors. By opposition to Cluster 3 where no influence of identification was found, in Cluster 5 for instance correct identification of the odors (clove, fried shallot, cucumber, dirty socks/cheese, fire smoke/smoked ham, beef) generated an important decrease of the disgust ratings specifically. It has probably more to do with a categorization process, in edible versus nonedible odor source. These odors were all food odors that were of rather low intrinsic pleasantness. It is very likely that only once they were categorized as edible (through correct odor identification) did people feel less reluctant toward them. A similar result was found by Herz and von Clef (2001), and de Araujo et al. (2005) provided evidence that such differences in semantic information provided during odor perception triggered differences in brain activation patterns, namely in the orbito-frontal cortex. It is believed that this brain region provides a top-down signal to the piriform cortex, an important substrate for the perception of odors as perceptual wholes (or objects, see Stevenson and Wilson 2007; Gottfried 2010), thus contributing to build odor representations according to individuals own experience and expectations. Accordingly, the different clusters we presented in the Results section may correspond to different kinds top-down signals related with odor identification. For example, the common beliefs that lavender and citrus odors, two odors typically associated with aromatherapy products, have relaxing and stimulating positive effects might have been activated when the correct odor names were accessed. This has undoubtedly affected the feeling representation in the direction of higher energy and higher well-being. Similarly, activation of top-down signals related to personal past experiences might have had a role in nostalgia modulation in Cluster 4. Additional information about the content of associated memories would be useful here to understand why the odors of beer, coffee, shampoo, and strawberry specifically triggered more nostalgia when they were correctly recognized. To summarize, this study brought new evidence of culture and gender differences in odor perception (and especially in odor-related affective feelings, using the new EOSs. It also attempted to relate these differences to variations in odor knowledge. Women had more accurate semantic knowledge than men, which was however not translated into stronger affective responses to odors. Rather, there were some odordependent sex differences for the EOS disgust category, in favor of women or of men according to how correctly identified the odors were and to the valence of the odor category chosen to identify the odor. We further showed that, when including all participants, increased odor knowledge was generally related to more positive affects (pleasantness, EOS ratings), certainly due to top-down influences on odor representation. The advantage of using a measure such as the EOS, finer than the classical pleasantness scales, is that we were able to distinguish between groups of odors hypothetically affected by different kinds of topdown influences: especially, cognitive influences related to beliefs about odor properties, and related to categorization (especially edible vs. nonedible). Finally, the link between odor knowledge and olfactory affect was asymmetrical. It was significant only for pleasant odors, not for unpleasant ones that seemed to be more resistant to elaborated cognitive influences. To conclude, this study, for the first time using emotional scales that are specific (and thus appropriate) to the olfactory domain, brings new insights into the variability of affective responses to odors and its relationship to odor knowledge. Supplementary material Supplementary material can be found at http://www.chemse. Funding This work was supported by Firmenich and the National Center of Competence in Research (NCCR) Affective Sciences financed by the Swiss National Science Foundation. Acknowledgements The authors wish to thank Andrew Giger and Mike Graham for enabling the study to be conducted in the Science Center in Singapore and in the World Museum in Liverpool, respectively. Pearlene Lim (in Singapore) and Ruth Cox (in Liverpool) were of great help in analyzing the odor identification responses. Nadine Gaudreau, William Jolly, Olivier Rosset, and Dajana Kapusova are thanked for their valuable technical help. References Larsson M. 1997. Semantic factors in episodic recognition of common odors in early and late adulthood: a review. Chem Senses. 22(6):623633. Appendix A Categories and terms of theEOSs GEOSGeneva EOS (36 terms) 1. (Happiness-Well-being)* 3. (Sensuality-Desire)* 5. (Soothing-Peacefulness) LEOSLiverpool EOS (37 terms) 3. (Sensuality-Desire)* 5. (Soothing-Peacefulness) 7. (Hunger-Thirst) SEOSSingapore EOS (36 terms) 1. (Happiness-Well-being)* 3. (Sensuality-Desire)* attracted, feeling awe, happiness, pleasant, pleasantly surprised*, well-being in a good mood, pleasantly surprised* dirty*, disgusted*, nauseous, repelled, sick*, uncomfortable, unpleasant*, unpleasantly surprised attracted, desire*, in love*, lustful, romantic*, sensual*, sexy*, to feel intimacy clean, energetic*, refreshed*, rejuvenated, revitalized*, stimulated comforted, dreamy, drowsy, meditative, peaceful, protected, relaxed, soothed nostalgic, sentimental famished, salivating, thirsty comforted, happiness, pleasant, pleasantly surprised*, relaxed, well-being dirty*, disgusted*, horrible, irritated, sick*, uncomfortable, unpleasant*, unpleasantly surprised* admiration, adoring, charmed, desire*, in love*, romantic*, sensual*, sexually aroused, sexy* energetic*, refreshed*, revitalized* amusement, fascinated, interesting religious feeling, spiritual feeling * Terms and categories common to the three countries. Appendix B Odorous substances (Firmenich SA) and their concentrations (in volumevolume percentage) Sulfury, onion (sclarymol) Wood 1 (Agarwoodsmoke) 20 Aniseed (anethol) Butternut popcorn Mushroom (carbinol) Rotten egg (sulfox) Wood 2 (Firsantol) Dirty socks (isovaleric acid) 1 Fried shallot, onion Lily of the valley Strawberry 1 (Cream strawberry) aIn Singapore only Body odor, sweat Incense 2 (Chinese incense) Wintergreen (methyl-salicylate) Wood 3 (Wolfwood) Laundry soap (Ariana) Dynascone

This is a preview of a remote PDF:

Camille Ferdenzi, S. Craig Roberts, Annett Schirmer, Sylvain Delplanque, Sezen Cekic, Christelle Porcherot, Isabelle Cayeux, David Sander, Didier Grandjean. Variability of Affective Responses to Odors: Culture, Gender, and Olfactory Knowledge, Chemical Senses, 2013, 175-186, DOI: 10.1093/chemse/bjs083