Identifying flavor preference subgroups. Genetic basis and related eating behavior traits

Appetite 75 (2014) 1–10

Contents lists available at ScienceDirect

Appetite

journal homepage: www.elsevier .com/locate /appet

Research report

Identifying flavor preference subgroups. Genetic basis and related eatingbehavior traits q

0195-6663/$ - see front matter � 2013 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.appet.2013.11.020

Abbreviations: A, additive genetic variance component; C, common environment variance component; CA, citric acid; CP, capsaicin; CR, cognitive restraint; CSFfor Sweet Foods; D, genetic dominant variance component; DZ, dizygotic; E, unique environment variance component; EE, emotional eating; FNS, Food Neophobia Ssubgroup 1, the basic; G2, subgroup 2, the adventurous; GF, General Food questionnaire; GHI, General Health Interest ; GWA, Genome Wide Association; HTAS, HeTaste Attitude Scales; HWE, Hardy Weinberg Equilibrium; LMS, Labeled Magnitude Scale; MAF, minor allele frequency; MZ, monozygotic; PKD1L3, polycystic kidne1-like 3 protein coding gene; PROP, 6-n-propylthiouracil; RSE, the Rosenberg Self-Esteem Scale; SF, Specific Food questionnaire; SNP, single nucleotide polymTAS1R1, taste receptor type 1 member 1 protein coding gene; TFEQ, Three-Factor Eating questionnaire; UE, Uncontrolled Eating.

q Acknowledgements: This work was supported by the University of Helsinki funds, the Finnish Food Research Foundation, the Academy of Finland [Grant numbers205585 to J. K. and 266592 to K. S.], the Academy of Finland Centre of Excellence in Complex Disease Genetics to J. K. and the DIOGENES (Diet, Obesity and Genessupported by EU [contract number FP6-513946]. We thank Givaudan Company for providing the strawberry aroma used in the study, the twins for participating,Siltala and Eero Vuoksimaa for their excellent assistance in data collection.⇑ Corresponding author.

E-mail address: [email protected] (O. Törnwall).

Outi Törnwall a,⇑, Karri Silventoinen b,c, Tero Hiekkalinna d,e, Markus Perola d,e,f, Hely Tuorila a,Jaakko Kaprio c,g,e

a Department of Food and Environmental Sciences, Agnes Sjöbergin katu 2, 00014 University of Helsinki, Finlandb Department of Social Research, Unioninkatu 37, 00014 University of Helsinki, Finlandc Department of Public Health, Hjelt Institute, Mannerheiminetie 172, 00014 University of Helsinki, Finlandd Department of Chronic Disease Prevention, National Institute for Health and Welfare, Mannerheimintie 166, 00270 Helsinki, Finlande Institute for Molecular Medicine Finland, Tukholmankatu 8, 00014 University of Helsinki, Finlandf The Estonian Genome Center, University of Tartu, Tiigi 61b, 50410 Tartu, Estoniag Department of Mental Health and Substance Abuse Services, National Institute for Health and Welfare, Mannerheimintie 166, 00270 Helsinki, Finland

a r t i c l e i n f o

Article history:Received 26 April 2013Received in revised form 29 November 2013Accepted 30 November 2013Available online 19 December 2013

Keywords:LikingFood neophobiaPungencySournessHeritabilityLinkage

a b s t r a c t

Subgroups based on flavor preferences were identified and their genetic and behavior related character-istics investigated using extensive data from 331 Finnish twins (21–25 years, 146 men) including 47monozygotic (MZ) and 93 dizygotic (DZ) pairs, and 51 twin individuals. The subgroup identification (hier-archical and K-means clustering) was based on liking responses to food names representing sour, umami,and spicy flavor qualities. Furthermore, sensory tests were conducted, a questionnaire on food likes com-pleted, and various eating behavior related traits measured with validated scales. Sensory data includedintensity ratings of PROP (6-n-propylthiouracil-impregnated filter paper), hedonic and intensityresponses to sourness (orange juice with and without added citric acid, 0.42%), pungency (strawberryjelly with and without added capsaicin 0.00013%) and umami (‘mouthfeel flavor’ taste solution). Ratingsof liking of 41 general food names were categorized into salty-and-fatty, sweet-and-fatty, fruits and vege-tables and fish foods. Subgroup differences (complex samples procedure) and the genetics underlying thesubgroups (structural equation modeling) were investigated. Of the resulting two groups (basic, n = 140,adventurous n = 152; non-grouped n = 39), the adventurous expressed higher liking for sour and spicyfoods, and had more tolerance for capsaicin burn in the sensory-hedonic test. The adventurous were alsoless food neophobic (25.9 ± 9.1 vs. 32.5 ± 10.6, respectively) and expressed higher liking for fruits and veg-etables compared to the basic group. Genetic effects were shown to underlie the subgroups (heritability72%, CI: 36–92%). Linkage analysis for 27 candidate gene regions revealed suggestively that being adven-turous is linked to TAS1R1 and PKD1L3 genes. These results indicate that food neophobia and genetic dif-ferences may form a barrier through which individual flavor preferences are generated.

� 2013 Elsevier Ltd. All rights reserved.

, Cravingcale; G1,alth and

y diseaseorphism;

100499,), projectand Mari

2 O. Törnwall et al. / Appetite 75 (2014) 1–10

Introduction

Food preferences are formed due to the effects of environment(exposure and experience) and genetic predispositions (e.g. taste,preferences and neophobic reaction to foods) which together playa central role in determining food selection and diet quality (Birch,1999; Story, Neumark-Sztainer, & French, 2002). Although individ-ual diversity exists in flavor responses to foods, individuals tend todiffer more dramatically in terms of what they like or dislike,than they differ in their sensory perceptions (Moskowitz, 1985).Surprisingly, flavor preferences have been rarely used in marketresearch in identification of consumer clusters, while more atten-tion has been paid to demographics, attitudinal and food-relatedmotivations (Gao et al., 2011; Logue & Smith, 1986; Contento,Michela, & Goldberg, 1988; Verdurme & Viaene, 2003).

Honkanen, Olsen, and Myrland (2004) demonstrated the appli-cability of preference data on segmenting consumers, using 1168Norwegian teenagers. They identified four distinctive subgroupsvarying in general liking of foods, particularly fish. Later on,Honkanen (2010) found six preference based subgroups in 1081Russian adults from which the largest group (various food lovers)showed high preference for all foods. Another valuable approachin consumer studies have been the use of psychographic measuressuch as food related lifestyles (Wycherley, McCarthy, & Cowan,2008) and personality dimensions, e.g. food neophobia (Henriques,King, & Meiselman, 2009). Among 1037 British consumers, Wycherleyand others identified six distinct life-style groups from which two(adventurous and rational) together accounted for nearly half of thepopulation. Both groups showed interest and reacted positively tothe attributes such as taste and quality of specialty foods (special,exclusive and quality products). Henriques and others used foodneophobia (reluctance to/avoidance of novel foods) as the basisof subgrouping and studied its effect on acceptability of novelfood items in 389 North American consumers. They concludedthat neophobics and neophilics perceived sensory characteristicssimilarly, but showed a different degree of liking for a food. Asmany studies have shown, clustering is a major improvement overundifferentiated approaches to ‘the consumer’ (MacFie, 2007). Inaddition, it can be helpful for nutrition educators who need under-standing of food choice motives to better tackle unhealthy eatingbehaviors.

In the context of the FinnTwin12 study (Kaprio, 2006; Kaprio,Pulkkinen, & Rose, 2002), investigating genetic and environmentaldeterminants of young adults’ health, a data set including flavorpreferences as well as behavioral and personality related traitswas collected. From these data, the heritability of food neophobia(Knaapila et al., 2007, 2011), astringency perception (Törnwallet al., 2011) and the preference for sourness and pungency (oralsensory burn) (Törnwall et al., 2012; Törnwall, Silventoinen, Kapri-o, & Tuorila, 2012) have previously been investigated. The evidencefor the role of genes in less well-studied traits such as umami,sourness and pungency is limited (Dotson, Babich, & Steinle,2012). At present, the objective was to obtain a coherent pictureof flavor preferences among young adults to the extent that thepresent data allowed. Subgroups were identified based on avail-able preference data (sour, umami and pungency), and subse-quently, the sensory, genetic and behavior related characteristicswere explored among the obtained subgroups.

Methods

Respondents

A total of 331 adult Finnish twins (146 men and 185 women,mean age 22.0 years, range from 21 to 25 years) participated in

the study, including 47 monozygotic (MZ) and 93 dizygotic (DZ)complete twin pairs and 51 twin individuals without their co-twin.The data were collected during 2008–2009 as part of the fourthwave of the longitudinal FinnTwin12-study (Kaprio, 2006; Kaprioet al., 2002). A total of sixteen participants were excluded fromperforming a specific sensory test as they were or had been usingthyroid medication (6-n-propylthiouracil, PROP), or reported al-lergy to citrus fruits (sourness), strawberries or chili (pungency).The study protocol was approved by the Ethics Committee of Hel-sinki University Hospital District. The respondents gave their writ-ten informed consent in the beginning of testing.

Sensory stimuli

PROP-impregnated filter paper was used for PROP-testing as de-scribed earlier (Keskitalo et al., 2007; Zhao, Kirkmeyer, & Tepper,2003). Umami sample was prepared on the preceeding or theday of testing by adding 0.5 g of Mouthfeel Flavor aroma powder(0.1–0.5% p137074, Givaudan, Switzerland) to 250 ml of tap water.The umami aroma was stirred well with water allowing the aromapowder to dissolve properly. Orange juice spiked with 4.20 g/L cit-ric acid (CA) vs orange juice without CA were used for sournesstesting. For pungency, strawberry flavored jelly spiked with0.0013 g/L capsaicin (CP) vs jelly without capsaicin were used.Samples were prepared as described previously (Törnwall,Silventoinen, Keskitalo-Vuokko, et al., 2012; Törnwall, Silventoinen,Kaprio, et al., 2012). All samples were served at room temperature.

Overview of data collection

Respondents were invited to the twin research unit located inHelsinki, Finland, for a 1-day assessment. The day started with sen-sory testing followed by other health related assessments (e.g.interview and neuropsychological tests) according to the Finn-Twin12 study protocol (Kaprio, 2006; Kaprio et al., 2002). Data col-lection scheme related to the present study is presented in Fig. 1.Upon arrival to the research unit, a ‘‘Home questionnaire’’ was re-turned and the participants were given both written and oralinstructions concerning the sensory tests. Sensory evaluations (4tests: PROP, umami, sourness and pungency) were performed inthe morning after an overnight fast (12 h) in an undisturbed class-room-type environment with divider screens for privacy. The orderfor sensory tests (Fig. 1) was chosen to minimize the interaction ef-fects between the tests (e.g. pungency test was completed last dueto the long lasting perception). For blood collection and weight andheight measurements, performed by a research nurse, the partici-pants were divided into two groups to streamline the testing pro-tocol. The blood was drawn (2 � 10 ml) from the first group inbetween the PROP and umami tests and from the second groupafter the pungency test. Other questionnaires; Specific Food ques-tionnaire (SF) and General Food questionnaire (GF) were com-pleted depending on available time, in between and after thesensory testing.

Sensory procedure

PROP intensity was evaluated in the beginning of sensory test-ing. To distinguish the taste of PROP from that of the filter paper,respondents first tasted pure filter paper and after this PROP-impregnated filter paper. Before and in between the stimulations,subjects rinsed their mouths with tap water. Both filter paperswere held in the mouth for 10 s. The rating was done after a shortbreak (PROP intensity builds up after a short delay) using a LabeledMagnitude Scale (LMS) by Green et al. (1996). The verbal labels andtheir positions in the line from bottom to up were: ‘‘barely detect-able’’ (2 mm), ‘‘weak’’ (7 mm), ‘‘moderate’’ (20 mm), ‘‘strong’’

Fig. 1. Data collection scheme representing the order of the sensory tests. PROP = 6-n-propylthiouracil.

O. Törnwall et al. / Appetite 75 (2014) 1–10 3

(42 mm), ‘‘very strong’’ (59 mm) and ‘‘the strongest imaginable’’(120 mm). The rationale for selecting LMS was that it is particu-larly suitable for sensations with broadly defined perceptual qual-ities (Green et al., 1996) such as PROP.

Pleasantness and intensity of umami were rated based on tast-ing the umami aroma solution. A 9-point scale ranging from 1 to 9was used. In the pleasantness scale, the values 1, 5 and 9 were la-beled as ‘‘very unpleasant’’, ‘‘not unpleasant nor pleasant’’ and‘‘very pleasant’’, respectively. In the intensity scale only the endpoints (1 and 9) were labeled as ‘‘no taste’’ and ‘‘extremely strongtaste’’, respectively. Sourness and pungency pleasantness andintensity were rated based on tasting first the untainted referencesample and then the spiked sample (orange juice vs. orange jui-ce + CA and strawberry jelly vs. strawberry jelly + CP). A 9-pointscale was used for the rating (-4 to 4) in which the pleasantnesswas labeled as ’’less pleasant than the reference sample (-4)’’, ’’sim-ilar to the reference sample’’ (0), ’’more pleasant than the referencesample’’ (4), and the intensity scale points labeled as; ’’less intensethan the reference sample’’ (-4), ’’similar to the reference sample’’(0) and ’’more intense than the reference sample’’ (-4). In betweenthe tests and after every sample respondents rinsed their mouthswith water. The test was conducted under supervision, and a timerwas used to reassure the time for keeping each sample in themouth. Further details are described in original publications(Törnwall, Silventoinen, Keskitalo-Vuokko, et al., 2012; Törnwall,Silventoinen, Kaprio, et al., 2012).

Questionnaires

Home questionnaire was sent to the participants prior the testday and data on education, smoking (never, have tried or quit, occa-sional and daily), alcohol consumption and drunkenness frequency(how often alcohol used and how often drunk per week/moth/year) were collected. The ‘Specific Food’ (SF) questionnaire wasused to measure liking responses to the names of foods and bever-ages reflecting four flavor qualities: 13 umami items, 21 sour itemsand 10 pungent items. In addition, the liking of pungent sensationwas determined by asking the pleasantness of ‘‘mild’’, and ‘‘extre-mely strong’’ pungency caused by spices in foods (part of the SFquestionnaire, not measured by sensory tests). The ‘General Food’(GF) questionnaire included 41 Finnish foods for which the likingratings were given. A labeled 7-point scale was used for liking withalternatives ‘‘very unpleasant’’, ‘‘fairly unpleasant’’, and ‘‘slightly

unpleasant’’, ‘‘neither pleasant nor unpleasant’’, ‘‘slightly pleasant’’,‘‘fairly pleasant’’ and ‘‘very pleasant’’, respectively.

Questionnaires were completed during the course of the testingday to assess 7 behavior or personality related traits: food neopho-bia, self-esteem, General Health Interest, Craving for Sweet Foods,cognitive restraint, uncontrolled and emotional eating. Food neo-phobia was measured by using The Food Neophobia Scale (FNS)developed by Pliner and Hobden (1992). General Health Interest(GHI) and Craving for Sweet Foods (CSF) were determined withHealth and Taste Attitude Scales (HTAS), developed by Roininen,Lähteenmäki, and Tuorila (1999). Both FNS and subscales of HTASinclude a balanced set of positive and negative statements (10statements in FNS, 8 for GHI and 6 for CSF; e.g. in FNS: ‘‘I don’t likenew foods’’ and in HTAS: ‘‘I always follow a healthy and balanceddiet’’). Statements were rated on a 7-point scale from 1 = ’’stronglydisagree’’ to 7 = ’’strongly agree’’. The Rosenberg Self-Esteem Scale(RSE), (Rosenberg, 1989), comprised of 11 statements such as ‘‘Attimes, I think I am no good at all’’ rated by a 4 point scale(1 = strongly disagree, 4 = strongly agree). A revised version of theThree-Factor Eating questionnaire (TFEQ-R18) was used (Karlsson,Persson, Sjöström, & Sullivan, 2000) to measure three eatingrelated behavior dimensions: Cognitive Restraint (CR, consciousrestriction of food intake to control body weight), Uncontrolled Eat-ing (UE, tendency to overeat) and Emotional Eating (EE, inability toresist emotional cues). TFEQ included 5 statements for CR, 9 for UEand 3 for EE rated on a 4-point scale with various response options.

DNA collection and genotyping

Blood samples were subjected to DNA extraction using an auto-mated Autopure LS purification instrument (Qiagen) at theNational Institute for Health and Welfare, Helsinki. The DNAsamples were genotyped aiming to provide data for the upcominggenetic studies within Finntwin12-project (Kaprio, 2006). Singlenucleotide polymorphism (SNP) were used as markers and thegenotyping was completed in the Sanger Institute (Sanger, Hinx-ton, UK) by Illumina 670 K Bead Chip platform (similar to the com-mercial 660w chip: www.illumina.com). The markers were asdelivered by the chip and thus included in the present study basedon their position. Basic QC was done to the genotyping data at theSanger Institute as part of their high throughput protocols. On re-ceipt of the data, we checked for minor allele frequency (>1%),genotyping success rate per SNP and per individual (>95%), HWE

Table 1Composite variables (in italics) quantifying liking of different food groups, based onitems of Specific Food questionnaire (SF) and General Food (GF) questionnaire.apleas = Cronbach’s a.

Specific Food items apleas General Food items apleas

UMAMI Salty-and-fatty foods(n = 328)

0.63

Cheese items (n = 307) 0.79 Fried potatoes or chipsCheddar cheese Salty snacks (e.g.

popcorn)Parmesan cheese PizzaEmmental cheese Hamburgers

4 O. Törnwall et al. / Appetite 75 (2014) 1–10

(p > 1 � 10–6), gender, and heterozygosity. In addition, to checkwhether any individuals were unexpectedly related to each other,multidimensional scaling plot (using a pairwise-IBS matrix) withonly one member of each known family was created. After the ped-igree was reassured to be correctly formulated, the basic filters(MAF, genotyping success, HWE) were reapplied to the data. Onlyone twin per MZ pair was genotyped. However, monozygositywas confirmed by genotyping both twins for multiple polymorphicmarkers at the Paternity Testing Unit, National Institute for Healthand Welfare. In separate studies we have genotyped both twins ofMZ pairs (n = 77) and find no differences between twins for SNPs.

Cheese spread Meat dumplingsUnripened cheese Sweet-and-fatty foods

(n = 330)0.84

Salad items (n = 330) 0.67 Ice-creamCucumber PastryTomato DessertsLettuce Chocolate

Pulse and corn (n = 326) 0.69 Fruits and vegetables(n = 330)

0.83

Corn Vegetarian dishesPeas Fresh vegetables

SOUR Boiled vegetablesSour fruits and berries (n = 328) 0.82 Fruits

Redcurrant BerriesCranberry Fish (n = 330) 0.76Currant juice Other fishLingonberry Salmon or Rainbow troutRhubarb TunaLemon

Sour dairy items 0.78Natural sourmilkNatural yogurtButtermilk

Less sour fruits and berries(n = 328)

0.63

StrawberryPeachOrangeBanana

Data analysis

Processing of the ratings

Body mass index was calculated and categorized according toWorld Health Organization (2013). The individual liking ratings(SF-questionnaire) were subjected to factor analysis within eachflavor category (umami, sour and pungent) and different groupsemerged: cheese items, salad items, pulse and corn, sour fruits andberries, sour dairy items, less sour fruits and berries, and spicy foodsand spices. New composite variables (see Table 1) were formed asmeans of the ratings. In the forthcoming text, these composite vari-ables are referred to as SF-variables. Composite variables were alsoformed from the GF ratings (see Table 1) according to Knaapilaet al. (2011) and referred later in the text to as GF-variables. Cron-bach’s alpha (a) was used to estimate the reliability of the vari-ables. The composite variables for FNS (a = 0.87), RSE (a = 0.89),HTAS (GHI: a = 0.91, CSF: a = 0.83) and TFEQ (CR: a = 0.83 , UE:a = 0.85 and EE: a = 0.80) were formed as described previously(Karlsson et al., 2000; Pliner & Hobden, 1992; Roininen et al.,1999; Rosenberg, 1989).

BlueberryPUNGENTSpicy foods and spices (n = 314) 0.92

Chili spiceSpicy foodTabascoSpicy mustardMexican foodIndian foodChili con carne

Preference based subgroups

SF-variables (cheese items, salad items, pulse and corn, sourfruits and berries, sour dairy items, less sour fruits and berries,and spicy foods and spices) and pleasantness of mild and extre-mely strong pungency caused by spices were used in Hierarchicaland K-means clustering to classify the respondents to different fla-vor preference subgroups: 9 variables in total, 3 for each flavorquality (sour, pungent and umami). The variables were chosen asthey were considered to provide general measures of flavor prefer-ences (not product specific as in the case with sensory variables)and were composed of several data points (robust measures).Pleasantness of pungency caused by spices (mild, and extremelystrong) were also included so that all the flavor qualities were bal-anced for the number of variables (3 in each). Hierarchical cluster-ing (Ward’s method, Squared Euclidean distance) was used todetermine the optimum number of clusters: a clear demarcationpoint for the clusters was determined from the agglomerationcoefficients and the Dendrogram plot. K-means method with 2cluster solution was used for the final classification of respondents.

To explore the differences between the subgroups in back-ground data, the goodness-of-fit statistics (v2, p-values) were used.For sensory, personality and food preference data the paired struc-ture of the data (clustering in twin pairs) was corrected by usingcomplex samples procedure, general linear model. Fixed factors in-cluded 2 respondent groups (2 levels) and gender (2 levels). Finally,logistic regression analysis (complex samples) was performed toassess the effects of explanatory variables for the subgrouping.The analyses were performed by PASW statistical package (PASWStatistics 18).

Genetic modeling

The genetic modeling of twin data based on structural equationmodeling allows the decomposition of trait variance into four la-tent variance components: additive genetic (A), genetic dominant(D), common environment (C) and unique environment (E). Thelast, E, includes the random measurement error. As our data onlyconsisted of twins reared together (no adopted twins or other rel-atives included), the C and D components could not be modeledsimultaneously due to their confounding nature (Neale & Gardon,1992). The modeling relies on the assumption that monozygotic(MZ) twins share the same DNA-sequence whereas dizygotic(DZ) twins share, on average, half of genes identical by descent.The second assumption is that the environment influences bothMZ and DZ twins equally. The best fitting model is searched fromalternatives: ACE/ADE (full model), AE, CE, E by comparing the sim-pler models to the full model applying the goodness-of-fit statistics(v2, p-values). In the present paper, the ADE-model was selectedbased on the within-pair correlation patterns and then comparedto the model fits of AE and E. The binary cluster membership var-iable generated by the K-means clustering was used and the anal-ysis completed by Mx statistical package (version 1.7).

O. Törnwall et al. / Appetite 75 (2014) 1–10 5

Linkage and association analysis

Linkage analysis was used to examine whether being adventur-ous was linked to some known chemosensory genes. In the test ofassociation conditional on linkage, we explored whether the sameallele segregates between the families. The nature of this analysiswas explorative and readily available Genome Wide Association(GWA)-data were utilized to examine whether links to flavor genescould be found.

After randomly removing one MZ twin per family (MZ twins aregenetically identical, therefore only one MZ twin needed for theanalysis), 226 twin individuals, provided with genotyping datawere subjected to qualitative linkage. The genes of interest werechosen mainly based on current knowledge on gene regions func-tioning in taste (umami, sour and bitter), taste bud development,chemesthesis (pungent) and food reward (palatable food): Adleret al., 2000; Calò et al., 2011; Bautista et al., 2005; Bufe, Hofmann,Krautwurst, Raguse, & Meyerhof, 2002; Chandrashekar et al., 2000;Felsted, Ren, Chouinard-Decorte, & Small, 2010; Ishimaru et al.,2006; Kim et al., 2003; Pérez et al., 2002; Raliou et al., 2009; Silver,Clapp, Stone, & Kinnamon, 2006; Thatcher, Doherty, Orvisky,Martin, & Henkin, 1998. Originally, there were 31 genes of interest,but four had to be excluded (TAS2R20, TAS2R41, TAS2R44 andTAS2R46) due the fact that the genetic markers on 670k GWA-chipdid not cover all the areas of interest. Thus, altogether 27 previ-ously published candidate genes including 163 directly genotypedSNPs were analyzed. Each gene area was extended by 500 kb up-and downstream in order to catch the genetic variation also inthe promoter region and in the 30 end. The genes, gene-areas andSNP-markers are listed in the Supplemental Table 1. The PSEUDO-MARKER software (Hiekkalinna et al., 2011) was used for perform-ing likelihood-based linkage and/or linkage disequilibrium analysisbecause it has shown to outperform other family-based associationmethods (Hiekkalinna et al., 2012). Multiple corrections were notapplied to any of the p-values obtained with the analysis.

Results

Respondent subgroups ‘basic’ and ‘adventurous’

Two subgroups (G1 and G2) emerged including 140 and 152respondents, respectively (n = 39 not clustered due to partiallymissing data). The subgroups were identified by using the SF-vari-ables which clearly separated the groups (Fig. 2). The respondentsin G2 exhibited higher liking especially towards pungent but alsotowards sour flavor qualities compared with G1. Thus, the groupswere named as ‘basic’ (G1) and ‘adventurous’ (G2). Over 90% ofthe respondents in each group were on their secondary stage ofeducation. The basic group had more women (63.6%) whereasthe adventurous comprised of almost equal proportions of men(52%) and women (48%) respondents (Table 2). In general, the sub-groups were fairly unifrom apart from the alcohol consumption anddrunkenness frequency which were the two distinctive backgroundfactors between the groups. The basic respondents used alcoholless and got drunk less frequently than the adventurousrespondents.

Hedonic responses to sensory stimuli (umami, sour and pun-gent) confirmed the subgroup differences (Fig. 3a). Basic respon-dents perceived particularly the pungent sample as less pleasantrelative to the reference, whereas the adventurous had more toler-ance for the sensation. With the more sour CA-juice, the pleasant-ness responses between the groups were marginally different(F1,177 = 3.7; p = 0.055). Differences in the intensity ratings of sourand pungent stimuli (Fig. 3b) or PROP and umami were not ob-served. Interestingly however, the adventurous included more

PROP super-tasters (n = 14, intensity rating >100) than the basicgroup (n = 8), whereas the non-taster respondents (intensity rating<20) were similar in both groups (basic n = 29, adventurous n = 27,respectively). Intensity rating cut-offs for tasters and non-tasters ofPROP were determined according to Prescott, Ripandelli, andWakeling (2001).

Personality and eating behavior

Among the personality and eating behavior traits, the groupswere distinct in food neophobia (Table 3). The adventurous wereclearly less food neophobic (F1,178 = 29.1; p < 0.001) compared tothe basic respondents. The adventurous also showed higher likingfor the GF-variable: fruits and vegetables (Table 3) compared to thebasic group (F1,178 = 13.6; p < 0.001).

Gender specific differences

Within the subgroups, gender differences were significant inbackground variables, SF, GF and sensory variables. However, sim-ilar trends were observed. That is, if a lower mean value was ob-served within the basic women respondents compared to themen, a lower value was also observed in women within the adven-turous, and vice versa. In general, the background variablesshowed as expected, that men consumed alcohol more and gotdrunk more often compared to the women. SF and GF-variablesindicated that women preferred sour dairy items: F1,178 = 10.1;p < 0.01 (SF), less sour fruits and berries: F1,178 = 15.9; p < 0.001(SF), pulse and corn F1,178 = 6.6; p = 0.01 (SF), fruits and vegetables:F1,178 = 56.8; p < 0.001 (GF) and sweet-and-fatty foods F1,178 = 6.0;p = 0.02 (GF). The men in turn, favored spicy foods and spicesF1,178 = 6.2; p = 0.01 (SF), and extremely strong pungency caused byspices F1,178 = 16.1; p < 0.001 (part of SF), and salty-and-fatty foods:F1,178 = 11.1; p < 0.01 (GF). The sensory tests showed that womenrespondents perceived umami and pungency as less pleasant(F1,178 = 6.4; p = 0.01, F1,177 = 11.8; p < 0.01), and PROP and pun-gency as more intense (F1, = 4.3; p = 0.04 and F1,177 = 11.0;p = 0.01) compared to the men. Women also showed greater ten-dency for Cognitive Restraint (F1,178 = 42.7; p < 0.001), EmotionalEating (F1,178 = 71.0; p < 0.001), General Health Interest(F1,178 = 29.4; p < 0.001) and Craving for Sweet Foods (F1,178 = 45.3;p < 0.001), whereas the men were more self confident by Rosen-berg Self-Esteem scale (F1,178 = 8.6; p = 0.004).

Predicting the preference subgroups

All the variables (background, SF, GF, behavior, sensory vari-ables) were included in the logistic regression analysis. The vari-ables were omitted if they did not significantly contribute to themodel (Wald v2 test for model effects, p > 0.05). In the final model,pleasantness of pungency (by a sensory test) and liking of fruits andvegetables (GF) predicted respondents belonging to adventurouswith higher odds (OR = 1.57, CI 1.32–1.86 and OR = 1.4, CI 1.01–2.03, respectively) whereas being assigned to the basic groupwas best predicted with food neophobia FNS (OR = 0.94, CI 0.92–0.97). Overall, the group membership was explained up to 69%by these three predictors.

Genetic modeling, linkage and association

Genetic modeling of the twins was conducted to test whetherthe separation of the respondents to different groups could bedue to genetic variability. The cluster membership was used to cal-culate the within correlation patterns for MZ and DZ twins:r(MZ) = 0.72 (CI: 0.30–0.93) and r(DZ) = 0.32 (CI: 0–0.6). As theMZ correlation was more than twice the DZ correlation, the genetic

Fig. 2. Preference characteristics of the respondent clusters (Specific Food composite variables which were used in the cluster analysis). G1 = basic subgroup,G2 = adventurous subgroup. M = males, F = females. �p < 0.05, ��p < 0.001.

Table 2Characteristics of respondent segments: body mass index, residence, smoking anddrinking habits. M = and F = females.

Basic n = 140,51 M, 89F %

Adventurous n = 152,79 M, 73F %

SexMales 36.4 52.0Females 63.6 48.0

Body mass index (kg/m2)a

Underweight (<18.5) 2.9 3.3Normal range (18.5–24.99) 72.1 70.4Overweight (P25.00) 19.3 21.7Obese (P30.00) 5.0 4.6

Region of residenceSouthern Finland 50.0 48.0Western Finland 25.0 30.0Eastern Finland 10.7 12.5Northern Finland 14.3 9.2

SmokingNever tried 16.4 15.1Not smoking (have tried or

quit)44.3 45.4

Occasionally (not daily) 10.7 12.5Daily (1–20 or more smokes/

day)28.6 27.0

Alcohol consumptionb

Never 2.9 3.3Rarely (once every couple of

months or less)12.8 5.9

Fairly often (once/month toonce/week)

63.6 56.6

Frequently (couple times/week or more)

20.7 34.2

Drunkennessc

Never 7.1 4.6Rarely (once every couple of

months or less)45.0 34.2

Fairly often (once/month toonce/week)

47.9 56.6

Frequently (couple times/week or more)

4.6

a Classification according to World Health Organization (WHO).b v2 = 9.2, p < 0.05.c v2 = 10.5, p < 0.05.

6 O. Törnwall et al. / Appetite 75 (2014) 1–10

effects were assumed to be additive, thus the model chosen as thestarting point was ADE. It was then tested if the model could besimplified by dropping the D, or, A and D components. The resultsin Table 4 showed that AE-model provided the best fit for the data,that is, ADE-model could be simplified to AE but not further to E asthe latter resulted in significant difference between E and ADE. Nosignificant differences between the gender in the groups werediscovered (v2 = 383.76, df = 284, p = 0.98). Genetic effects (inAE-model) explained 72% (CI: 36–92%) and specific environmentaleffects 28% (CI: 1–64%) of the variation underlying the subgroups.

As the genetic component emerged, it was further hypothesizedthat being adventurous may be linked to known flavor perceptiongenes. Fig. 4 represents the single point linkage analysis results inthe areas where the highest estimated lod scores (logarithm ofodds above 1.8) were observed. The threshold was chosen accord-ing to Lander and Kryglag (1995), who stated that a lod score of 1.9can be interpreted as suggestive linkage result. For these locitested, under recessive pseudomarker analysis model, the scoreswere 1.86 (TAS1R1: rs2235564 and rs6577584) and 1.88 (PKD1L3:rs12102451) showing linkage in umami and sour receptor genes inchromosomes 1 and 16, respectively. The two SNPs in TAS1R1 genearea were in perfect linkage disequilibrium (D0 1.0, r2 1.0), thusrepresenting the same linkage signal. However, given the generallyaccepted threshold for genome-wide significance, lod >3.0, the re-sults should be taken as suggestive. The test of association condi-tional on linkage was also performed but did not showsignificant results. The rest of the results are presented in the Sup-plemental Table 1.

Discussion

With data comprised of flavor preferences, sensory and behav-ioral dimensions, we were able to elucidate the architecture of foodpreferences in young adults. Two even respondent subgroups wereidentified differing in their reported liking for spicy and sour foods.The division was confirmed by sensory-hedonic responses to par-ticularly pungent stimuli. The adventurous represented a groupof young adults who favored spicy and sour foods, and showedmore tolerance for the perceived pungency compared to the basicgroup. Only a limited number of previous studies, such as those of

Fig. 3. (a–b) Responses to sourness and pungency obtained by sensory experiments (panel 3a hedonic ratings, panel 3b intensity ratings) for two respondent clusters (meanand std. error of mean). G1 = basic subgroup, G2 = adventurous subgroup. M = males, F = females. ��p < 0.001.

Table 3Respondent subgroups: personality, eating behavior and preference (significant differences between groups bolded). M = and F = females.

Trait Theoretical range Basic Adventurousn = 140, 51 M, 89F n = 152, 79 M, 73FMean (SD) Mean (SD)

PersonalityFood neophobiaa,** 10–70 32.5 (10.6) 25.9 (9.1)Self esteemb 10–40 31.2 (5.6) 31.9 (5.2)

Eating behaviorGeneral Health Interesta 1–7 4.1 (1.3) 4.2 (1.4)Craving for Sweet Foodsa 1–7 4.3 (1.3) 4.1 (1.4)Cognitive Restrainta 0–100 33.0 (20.8) 32.7 (21.1)Uncontrolled Eatinga 0–100 33.9 (16.4) 36.8 (17.7)Emotional Eatinga 0–100 29.5 (25.5) 25.6 (22.4)

PreferenceSalty-and-fatty foodsc 1–7 5.9 (0.8) 6.0 (0.9)Sweet-and-fatty foodsa 1–7 6.0 (0.8) 6.0 (1.0)Fruits and vegetablesa,** 1–7 5.8 (0.9) 6.1 (0.8)Fisha 1–7 5.4 (1.3) 5.6 (1.2)

a Basic n = 139.b Basic n = 138, Adventurous n = 151.c Basic n = 137.

** p < 0.001 between basic and adventurous.

O. Törnwall et al. / Appetite 75 (2014) 1–10 7

Honkanen et al. (2004), Honkanen (2010), Logue and Smith (1986),have studied preference patterns using a range of different foods.These studies were focused merely on bland foods (e.g. chicken,pork, meatballs, cereal, hamburgers, fuits, dairy products etc.)and had no sensory data available. Thus, no attention was paid tospecific chemosensory qualities. Interestingly, the groups identi-fied in the present study showed large differences in liking of foodswith specific flavor qualities (e.g. sour fruits and berries and spicyfoods and spices), but showed no differences in the liking of mostbland foods (salty-and-fatty foods, sweet-and-fatty foods, and fish).

Table 4Model fit statistics (v2 = chi-square, df = degrees of freedom) for univariate models forsubgrouping. AE-model (bolded) provides the best fit.

Model v2 df Models compared Dv2 Ddf p

Saturated 380.225 276ADE 383.898 287AE 383.898 288 AE vs ADE 0 1 1

AE vs Saturated 3.673 12 0.99E 397,112 289 E vs ADE 13.214 2 0.001

ADE-model includes additive genetic (A), genetic dominant (D) and unique envi-ronment (E) components. AE-model including components A and E, and E-modelincluding component E.

Thus, information of specific flavors of foods deepen our insightsinto individual preferences and reveal aspects of food choicebehavior that otherwise remain concealed.

Examination of the personality and eating behavior relatedtraits between the subgroups showed a distinct difference onlyin food neophobia scores. Even though HTAS and TFEQ subscaleshave shown their usefulness as predictors of dietary behavior(Lähteenmäki & Tuorila, 1995; Zandstra, De Graaf, & Van Staveren,2001), they do not appear to be closely linked to the specificflavor preferences measured in the present study. It is possiblethat HTAS and TFEQ behavioral traits are overridden by foodneophobic/neophilic tendency. Food neophobia was an interestingmarker for the adventurous, as the respondents in this groupscored lower on food neophobic scale compared to the basicrespondents (FNSG2: 25.9 ± 9.1 vs. FNSG1: 32.5 ± 10.6). Plinerand Hobden (1992) reported an average score of 34.5 for FNSand Knaapila et al. (2011) a score of 31.2. Thus, whereas the basicgroup seems to represent the general population, the adventur-ous are more neophilic. Low food neophobia scores provide a log-ical explanation for elevated liking of spiciness, preference forfruits and vegetables and more tolerant sensory hedonic responsesto pungency among the adventurous. Previously, food neophobiahas been shown to steer responses to unfamiliar foods (Arvola,

Fig. 4. Recessive pseudomarker single-point linkage analysis results from genes with the highest lod scores (lod > 1.8) showing suggestive linkage in umami (TAS1R1) andsour (PKD1L3) genes. Locations of the single nucleotide polymorphisms (SNPs) are given in x-axis. In total, the analysis included 27 gene areas, with 163 SNP-markers. Wepresent here only the lod scores higher than 1.8, all the results are provided in the supplemental Table 1. lod = logarithm of odds. rs-number = the SNPs identification number.

8 O. Törnwall et al. / Appetite 75 (2014) 1–10

Lähteenmäki, & Tuorila, 1999; Pliner, Lahteenmaki, & Tuorila,1998), and to affect the degree of product acceptance (Henriqueset al., 2009) as well as liking and use of fruits and vegetables(Cooke, Carnell, & Wardle, 2006; Knaapila et al., 2011; Wardle,Carnell, & Cooke, 2005). As food neophobia affects the overallquality of a diet (Falciglia, Couch, Gribble, Pabst, & Frank,2000), and fruits and vegetables are linked to healthful eatingand disease risk reduction (Van Duyn, 2000), health-wise benefitsmay be pronounced in the adventurous group. Based on their lowfood neophobia scores, we may also speculate these individualsto be more sensation seeking, extrovert and open to new experi-ences than the basic respondents, as these personality featureshave been negatively associated with food neophobia (Knaapilaet al., 2011; Pliner & Hobden, 1992).

Gender differences were found in variables that differed acrossthe subgroups. Women preferred the sour food items (sour dairyitems, and less sour fruits and berries) whereas the men were in fa-vor of spicy foods and spices, extremely strong pungency caused byfoods, and pungent stimulus (capsaicin spiked jelly). These resultscorroborate previous findings, showing that women exhibit higherpreference for foods associated with healthiness (Kähkönen &Tuorila, 1999), whereas men tend to prefer foods with strong andrich flavors (Frewer & Trijp, 2006; Logue & Smith, 1986). The un-equal distribution of men and women in the subgroups was nothowever, the cause of observed subgroup differences. The trendswere similar for both genders in all relevant variables, and the gen-der was not a significant predictor of the subgroups in the regres-sion model.

Genetic differences between the individuals seem to contributeto the obtained clusters. The genetic effects explained 72% (CI: 36–92%) of the variation underlying the subgroups, whereas 28% (CI:1–64%) was due to environmental effects. The confidence intervalsare fairly wide causing limitations to the precision of the heritabil-ity estimate. However, the result is convincing for several reasons.First, strong heritabilities (67–78%) have been reported for foodneophobia (Cooke, Haworth, & Wardle, 2007; Knaapila et al.,2007, 2011) and second, the preference of sour foods (e.g. fruitsand berries) and spicy foods have also been defined as partiallyinherited characteristics (Törnwall, Silventoinen, Keskitalo-Vuokko,et al., 2012; Törnwall, Silventoinen, Kaprio, et al., 2012; Faust,1974). Furthermore, in our logistic regression model, it was namelyfood neophobia, pleasantness of pungency and liking of fruits andvegetables that explained the subgroups up to 69%. Thus, thesepartially genetically regulated traits may indeed underlie theindividual preference patterns and in combination lead to theformation of two distinct subgroups.

Recent literature (Knaapila et al., 2012; Pirastu et al., 2012) sug-gests that the differences in chemosensory perception are partiallydue to genetic variability. The present study showed genetic con-tribution underlying the subgroups, thus we hypothesised thatgenes regulating chemosensory perceptions could also contributeto the respondent’s alignment to a particular group. Loci on geneareas coding for umami (TAS1R1) and sour (PKD1L3) taste recep-tors (chromosome 1 and 16, respectively) were found to be linkedwith being adventurous. Interestingly, the subgroups respondeddifferently to sourness, the adventurous favoring sour foods andadded citric acid in sensory hedonic testing, whereas sensoryintensity evaluations showed very small differences between thesubgroups. The linkage results were not significant in a genome-wide sense and we did not have parental genotypes available tosubstantiate our results by increasing the statistical power, thuswe conclude these findings to be suggestive and requiring replica-tion. They, however, support the existence of genetic componentsbehind these sensory traits. While previous literature confirmsthe involvement of TAS1R1 in umami perception (Raliou et al.,2009) and shows that sensitivity to monosodium glutamate is af-fected by the receptor genotypes (Raliou et al., 2011), the role ofPKD1L3 in taste bud cells is still under controversy (Nelson et al.,2010). Nonetheless, genes regulating umami and sour taste mayhave relevance in the acceptability of many foods hence affectingnutrition. Preference for umami has shown to influence food intakeand protein metabolism (Kondoh, Mori, Ono, & Torii, 2000; Mori,Kawada, Ono, & Torii, 1991) and putatively, function as a biologicalsignal for protein (Prescott, 2001). Naturally sour products such asfruits, berries and vegetables, in turn, are known for their healthbenefits and epidemiologic evidence exists of their protective rolein cancer and coronary heart diseases (Van Duyn, 2000).

The fact that large differences were observed in the hedonic rat-ings of sour and pungent stimuli, but next to none in the intensityevaluations of the subgroups, is noteworthy and supported by pre-vious literature (Small et al., 2003). Thus, processes underlying thesensory and hedonic properties are indeed separate. Different he-donic outcomes could be a result from variance in the perceptualqualities of the stimulus or alternatively, the perception might beidentical but the processing of the sensory signals in reward cir-cuits differ across individuals (Loney, Blonde, Eckel, & Spector,2012). In addition to the receptor genes, variation in taste re-sponses can depend on many other genes involved in differentstages of taste processing (Bachmanov et al., 2011). Furthermore,genes regulating reward can moderate brain responses to food(Felsted et al., 2010) and thus strongly influence taste preferencesand food intake.

O. Törnwall et al. / Appetite 75 (2014) 1–10 9

Conclusions

Two subgroups were identified in a young twin population dif-fering in their liking of particularly sour and pungent foods. Foodneophobia and genetic variability were the likely ‘building stones’leading to a particular flavor preference pattern in a subgroup.These results may partially explain individual food choices andhelp understand human eating behavior and food choices.

Appendix A. Supplementary material

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.appet.2013.11.020.

References

Adler, E., Hoon, M. A., Mueller, K. L., Chandrashekar, J., Ryba, N. J., & Zuker, C. S.(2000). A novel family of mammalian taste receptors. Cell, 100(6), 693–702.

Arvola, A., Lähteenmäki, L., & Tuorila, H. (1999). Predicting the intent to purchaseunfamiliar and familiar cheeses. The effects of attitudes, expected liking andfood neophobia. Appetite, 32(1), 113–126.

Bachmanov, A. A., Bosak, N. P., Floriano, W. B., Inoue, M., Li, X., Lin, C., et al. (2011).Genetics of sweet taste preferences. Flavour and Fragrance Journal, 26(4),286–294.

Bautista, D. M., Movahed, P., Hinman, A., Axelsson, H. E., Sterner, O., Högestätt, E. D.,et al. (2005). Pungent products from garlic activate the sensory ion channelTRPA1. Proceedings of the National Academy of Sciences of the United States ofAmerica, 102(34), 12248–12252.

Birch, L. L. (1999). Development of food preferences. Annual Review of Nutrition,19(1), 41–62.

Bufe, B., Hofmann, T., Krautwurst, D., Raguse, J., & Meyerhof, W. (2002). The humanTAS2R16 receptor mediates bitter taste in response to b-glucopyranosides.Nature Genetics, 32(3), 397–401.

Calò, C., Padiglia, A., Zonza, A., Corrias, L., Contu, P., Tepper, B., et al. (2011).Polymorphisms in TAS2R38 and the taste bud trophic factor, gustin gene co-operate in modulating PROP taste phenotype. Physiology & Behavior, 104,1065–1071.

Chandrashekar, J., Mueller, K. L., Hoon, M. A., Adler, E., Feng, L., Guo, W., et al. (2000).T2Rs function as bitter taste receptors. Cell, 100(6), 703–711.

Contento, I. R., Michela, J. L., & Goldberg, C. J. (1988). Food choice amongadolescents. Population segmentation by motivations. Journal of NutritionEducation, 20(6), 289–298.

Cooke, L., Carnell, S., & Wardle, J. (2006). Food neophobia and mealtime foodconsumption in 4–5 year old children. International Journal of BehavioralNutrition and Physical Activity, 3(1), 14.

Cooke, L. J., Haworth, C. M., & Wardle, J. (2007). Genetic and environmentalinfluences on children’s food neophobia. The American Journal of ClinicalNutrition, 86(2), 428–433.

Dotson, C. D., Babich, J., & Steinle, N. I. (2012). Genetic predisposition and tastepreference. Impact on food intake and risk of chronic disease. Current NutritionReports, 1(3), 175–183.

Falciglia, G. A., Couch, S. C., Gribble, L. S., Pabst, S. M., & Frank, R. (2000). Foodneophobia in childhood affects dietary variety. Journal of the American DieteticAssociation, 100(12), 1474–1481.

Faust, J. (1974). A twin study of personal preferences. Journal of Biosocial Science,6(1), 75–91.

Felsted, J. A., Ren, X., Chouinard-Decorte, F., & Small, D. M. (2010). Geneticallydetermined differences in brain response to a primary food reward. The Journalof Neuroscience, 30(7), 2428–2432.

Frewer, L., & Trijp, H. V. (2006). Understanding consumers of food products.Woodhead Publishing Ltd.

Gao, Z., House, L. O., Gmitter, Fred G., Jr., Valim, M. Filomena, Plotto, A., & Baldwin,E. A. (2011). Consumer preferences for fresh citrus: impacts of demographic andbehavioral characteristics. International Food and Agribusiness ManagementReview, 14(1).

Green, B. G., Dalton, P., Cowart, B., Shaffer, G., Rankin, K., & Higgins, J. (1996).Evaluating the ‘Labeled Magnitude Scale’ for measuring sensations of taste andsmell. Chemical Senses, 21(3), 323–334.

Henriques, A. S., King, S. C., & Meiselman, H. L. (2009). Consumer segmentationbased on food neophobia and its application to product development. FoodQuality and Preference, 20(2), 83–91.

Hiekkalinna, T., Göring, H. H., Lambert, B., Weiss, K. M., Norrgrann, P., Schäffer, A. A.,& Terwilliger, J. D. (2012). On the statistical properties of family-basedassociation tests in datasets containing both pedigrees and unrelated case–control samples. European Journal of Human Genetics, 20, 217–223.

Hiekkalinna, T., Schäffer, A. A., Lambert, B., Norrgrann, P., Göring, H. H., &Terwilliger, J. D. (2011). PSEUDOMARKER. A powerful program for jointlinkage and/or linkage disequilibrium analysis on mixtures of singletons andrelated individuals. Human Heredity, 71(4), 256–266.

Honkanen, P. (2010). Food preference based segments in Russia. Food Quality andPreference, 21(1), 65–74.

Honkanen, P., Olsen, S. O., & Myrland, Ø. (2004). Preference-based segmentation. Astudy of meal preferences among Norwegian teenagers. Journal of ConsumerBehaviour, 3(3), 235–250.

Ishimaru, Y., Inada, H., Kubota, M., Zhuang, H., Tominaga, M., & Matsunami, H.(2006). Transient receptor potential family members PKD1L3 and PKD2L1 forma candidate sour taste receptor. Science Signaling, 103(33), 12569.

Kähkönen, P., & Tuorila, H. (1999). Consumer responses to reduced and regular fatcontent in different products. Effects of gender, involvement and healthconcern. Food Quality and Preference, 10(2), 83–91.

Kaprio, J. (2006). Twin studies in Finland 2006. Twin Research and Human Genetics,9(06), 772–777.

Kaprio, J., Pulkkinen, L., & Rose, R. J. (2002). Genetic and environmental factors inhealth-related behaviors. Studies on Finnish twins and twin families. TwinResearch, 5(05), 366–371.

Karlsson, J., Persson, L., Sjöström, L., & Sullivan, M. (2000). Psychometric propertiesand factor structure of the Three-Factor Eating questionnaire (TFEQ) in obesemen and women. Results from the Swedish obese subjects (SOS) study.International Journal of Obesity and Related Metabolic Disorders. Journal of theInternational Association for the Study of Obesity, 24(12), 1715.

Keskitalo, K., Knaapila, A., Kallela, M., Palotie, A., Wessman, M., Sammalisto, S., et al.(2007). Sweet taste preferences are partly genetically determined. Identificationof a trait locus on chromosome 16. American Journal of Clinical Nutrition, 86(1),55–63.

Kim, U., Jorgenson, E., Coon, H., Leppert, M., Risch, N., & Drayna, D. (2003). Positionalcloning of the human quantitative trait locus underlying taste sensitivity tophenylthiocarbamide. Science, 299(5610), 1221.

Knaapila, A., Hwang, L., Lysenko, A., Duke, F. F., Fesi, B., Khoshnevisan, A., et al.(2012). Genetic analysis of chemosensory traits in human twins. ChemicalSenses, 37(9), 869–881.

Knaapila, A., Silventoinen, K., Broms, U., Rose, R. J., Perola, M., Kaprio, J., et al. (2011).Food neophobia in young adults. Genetic architecture and relation topersonality, pleasantness and use frequency of foods, and body mass Index—Atwin study. Behavior Genetics, 41(4), 512–521.

Knaapila, A., Tuorila, H., Silventoinen, K., Keskitalo, K., Kallela, M., Wessman, M.,Peltonen, L., Cherkas, L. F., Spector, T. D., & Perola, M. (2007). Food neophobiashows heritable variation in humans. Physiology & Behavior, 91(5), 573–578.

Kondoh, T., Mori, M., Ono, T., & Torii, K. (2000). Mechanisms of umami tastepreference and aversion in rats. The Journal of Nutrition, 130(4), 966S–970S.

Lähteenmäki, L., & Tuorila, H. (1995). Three-Factor Eating questionnaire and the useand liking of sweet and fat among dieters. Physiology & Behavior, 57(1), 81–88.

Lander, E., & Kryglag, K. (1995). Genetic dissection of complex traits. Guidelines forinterpreting and reporting linkage results. Nature Genetics, 11, 241–247.

Logue, A., & Smith, M. E. (1986). Predictors of food preferences in adult humans.Appetite, 7(2), 109–125.

Loney, G. C., Blonde, G. D., Eckel, L. A., & Spector, A. C. (2012). Determinants of tastepreference and acceptability. Quality versus hedonics. The Journal ofNeuroscience, 32(29), 10086–10092.

MacFie, H. (2007). Consumer-led food product development. Woodhead PublishingLtd.

Mori, M., Kawada, T., Ono, T., & Torii, K. (1991). Taste preference and proteinnutrition and L-amino acid homeostasis in male Sprague–Dawley rats.Physiology & Behavior, 49(5), 987–995.

Moskowitz, H. R (1985). New directions for product testing and sensory analysis offoods. Food and Nutrition Press.

Neale, M. C., & Cardon, L. R. (1992). Methodology for genetic studies of twins andfamilies. Doprdrecht, Germany: Kluwer Academic Publishers BV.

Nelson, T. M., LopezJimenez, N. D., Tessarollo, L., Inoue, M., Bachmanov, A. A., &Sullivan, S. L. (2010). Taste function in mice with a targeted mutation of thepkd1l3 gene. Chemical Senses, 35(7), 565–577.

Pérez, C. A., Huang, L., Rong, M., Kozak, J. A., Preuss, A. K., Zhang, H., et al. (2002). Atransient receptor potential channel expressed in taste receptor cells. NatureNeuroscience, 5(11), 1169–1176.

Pirastu, N., Robino, A., Lanzara, C., Athanasakis, E., Esposito, L., Tepper, B. J., et al.(2012). Genetics of food preferences. A first view from silk road populations.Journal of Food Science, 77(12), S413–S418.

Pliner, P., & Hobden, K. (1992). Development of a scale to measure the trait of foodneophobia in humans. Appetite, 19(2), 105–120.

Pliner, P., Lahteenmaki, L., & Tuorila, H. (1998). Correlates of human food neophobia.Appetite, 30(1), 97–115.

Prescott, J. (2001). Taste hedonics and the role of umami. The future of MSG. FoodAustralia, 53(12), 550–554.

Prescott, J., Ripandelli, N., & Wakeling, I. (2001). Binary taste mixture interactions inPROP non-tasters, medium-tasters and super-tasters. Chemical Senses, 26,993–1003.

Raliou, M., Boucher, Y., Wiencis, A., Bézirard, V., Pernollet, J., Trotier, D., et al. (2009).Tas1R1–Tas1R3 taste receptor variants in human fungiform papillae.Neuroscience Letters, 451(3), 217–221.

Raliou, M., Grauso, M., Hoffmann, B., Schlegel-Le-Poupon, C., Nespoulous, C., Débat,H., et al. (2011). Human genetic polymorphisms in T1R1 and T1R3 tastereceptor subunits affect their function. Chemical Senses, 36(6), 527–537.

Roininen, K., Lähteenmäki, L., & Tuorila, H. (1999). Quantification of consumerattitudes to health and hedonic characteristics of foods. Appetite, 33(1), 71–88.

Rosenberg, M. (1989). Society and the adolescent self-image (rev ed.). WesleyanUniversity Press.

10 O. Törnwall et al. / Appetite 75 (2014) 1–10

Silver, W. L., Clapp, T. R., Stone, L. M., & Kinnamon, S. C. (2006). TRPV1 receptors andnasal trigeminal chemesthesis. Chemical Senses, 31(9), 807–812.

Small, D. M., Gregory, M. D., Mak, Y. E., Gitelman, D., Mesulam, M. M., & Parrish, T.(2003). Dissociation of neural representation of intensity and affectivevaluation in human gustation. Neuron, 39(4), 701.

Story, M., Neumark-Sztainer, D., & French, S. (2002). Individual and environmentalinfluences on adolescent eating behaviors. Journal of the American DieteticAssociation, 102(3), S40–S51.

Thatcher, B., Doherty, A., Orvisky, E., Martin, B., & Henkin, R. (1998). Gustin fromhuman parotid saliva is carbonic anhydrase VI. Biochemical and BiophysicalResearch Communications, 250(3), 635–641.

Törnwall, O., Dinnella, C., Keskitalo-Vuokko, K., Silventoinen, K., Perola, M.,Monteleone, E., et al. (2011). Astringency perception and heritability amongyoung Finnish twins. Chemosensory Perception, 4(4), 134–144.

Törnwall, O., Silventoinen, K., Kaprio, J., & Tuorila, H. (2012). Why do some like ithot? Genetic and environmental contributions to the pleasantness of oralpungency. Physiology & Behavior.

Törnwall, O., Silventoinen, K., Keskitalo-Vuokko, K., Perola, M., Kaprio, J.,& Tuorila, H. (2012). Genetic contribution to sour taste preference.Appetite.

Van Duyn, M. (2000). Overview of the health benefits of fruit and vegetableconsumption for the dietetics professional. Selected literature. Journal of theAmerican Dietetic Association, 100(12), 1511–1521.

Verdurme, A., & Viaene, J. (2003). Consumer beliefs and attitude towards geneticallymodified food. Basis for segmentation and implications for communication.Agribusiness, 19(1), 91–113.

Wardle, J., Carnell, S., & Cooke, L. (2005). Parental control over feeding and children’sfruit and vegetable intake. How are they related? Journal of the American DieteticAssociation, 105(2), 227.

World Health Organization (2013). Global database on Body Mass Index. <http://apps.who.int/bmi/index.jsp?introPage=intro_3.html> Retrieved 25.04.13.

Wycherley, A., McCarthy, M., & Cowan, C. (2008). Speciality food orientation of foodrelated lifestyle (FRL) segments in great Britain. Food Quality and Preference,19(5), 498–510.

Zandstra, E., De Graaf, C., & Van Staveren, W. (2001). Influence of health and tasteattitudes on consumption of low-and high-fat foods. Food Quality andPreference, 12(1), 75–82.

Zhao, L., Kirkmeyer, S. V., & Tepper, B. J. (2003). A paper screening test to assessgenetic taste sensitivity to 6-n-propylthiouracil. Physiology & Behavior, 78(4),625–633.