The Food Choice Questionnaire (FCQ) was developed through factor analysis of responses from a sample of 358 adults ranging in age from 18 to 87 years. Nine factors emerged, and were labelled health, mood, convenience, sensory appeal, natural content, price, weight control, familiarity and ethical concern.
The Food Choice Questionnaire (FCQ) was developed through factor analysis of responses from a sample of 358 adults ranging in age from 18 to 87 years. Nine factors emerged, and were labelled health, mood, convenience, sensory appeal, natural content, price, weight control, familiarity and ethical concern.
The Food Choice Questionnaire (FCQ) was developed through factor analysis of responses from a sample of 358 adults ranging in age from 18 to 87 years. Nine factors emerged, and were labelled health, mood, convenience, sensory appeal, natural content, price, weight control, familiarity and ethical concern.
Development of a Measure of the Motives Underlying the Selection of Food: the Food Choice Questionnaire ANDREW STEPTOE and TESSA M. POLLARD Department of Psychology, St Georges Hospital Medical School, London J ANE WARDLE Imperial Cancer Research Fund Health Behaviour Unit, Institute of Psychiatry, London A number of factors arethought to inuencepeoples dietary choices, including health, cost, convenienceandtaste, but thereareno measuresthat addresshealth- relatedandnon-health-relatedfactorsinasystematicfashion. Thispaper describes thedevelopment of amultidimensional measureof motivesrelatedtofoodchoice. The Food Choice Questionnaire (FCQ) was developed through factor analysis of responses from a sample of 358 adults ranging in age from 18 to 87 years. Nine factors emerged, and were labelled health, mood, convenience, sensory appeal, natural content, price, weight control, familiarity and ethical concern. The questionnairestructurewasveriedusingconrmatoryfactor analysisinasecond sample (n=358), and testretest reliability over a 2- to 3-week period was satisfactory. Convergent validity wasinvestigated by testingassociationsbetween FCQ scales and measures of dietary restraint, eating style, the value of health, health locus of control and personality factors. Dierences in motives for food choiceassociated with sex, ageand incomewerefound. Thepotential usesof this measurein health psychology and other areas arediscussed. 1995 Academic Press Limited I N1onic1ioN Concernabout foodchoicesthat mayhaveadverseeectsonhealthiswidespread in thedeveloped world, and is embodied in documents such as Healthy People2000 (1992) and The Health of the Nation (1992). Recommendations to restrict salt and fat intakeandincreasecomplexcarbohydrateandbreconsumptionarecentral tenets in public health nutrition guidelines (Cannon, 1992). Nevertheless, fat consumption remainshighinthewesternworld, andisincreasingincountriesundergoingindustrial development (Lands et al., 1990; Trichopoulou & Efstathiadis, 1989). Eective modication of dietary patterns depends on an understanding of the factors governing food choice. Recent years have witnessed growing interest in studying the attitudes and beliefs associated with healthy eating. Measures of ThisresearchwassupportedbyGrant L209252013fromtheEconomicandSocial ResearchCouncil, U.K. Wearegrateful to Emma Winstanley for her assistancein data collection. Addresscorrespondenceto: Dr A. Steptoe, Department of Psychology, St GeorgesHospital Medical School, Cranmer Terrace, London SW170RE, U.K. 01956663/95/060267+18 $12.00/0 1995 Academic Press Limited 268 A. STEPTOE ET AL. attitudes relevant to theoretical frameworks such as thehealth belief model and the transtheoretical model of changehavebeen developed (Kristal et al., 1990; Smith & Owen, 1992; Trenkner et al., 1990). For example, Glanz et al. (1993) havedescribed measures of psychosocial factors inuencing fat and bre consumption, including items related to beliefs in thelinks between diet and disease, perceived benets and barriers to behaviour change, social support, social norms, motivation and self- ecacy. This work holds the promise of leading to improved dietary modication programmes (McCann et al., 1990). However, health is clearly not the only factor peopletakeinto account when choosingtheir food, and a focus on health may lead to exclusiveemphasis on a set of motives that areof limited signicancefor many people. I t is therefore important to explore the role of other inuences on food choice. I t has long been recognized that food availability and cultural factors are dominant in food selection. Cultural inuences lead to dierences in the habitual consumption of certain foods and in traditions of preparation, and in certain cases can lead to restrictions such as exclusion of meat and milk from the diet (Lau, Krondl &Coleman, 1984). Foodisafocusof social interaction, andtheconsumption of prestige foods may becomean index of social status(Sanjur, 1982). Thesystem of provision, including food production and manufacture, marketing, delivery and sale, has been shown to havea major impact on what peopleeat (Fine& Leopold, 1993). At theindividual level, tasteor sensory appeal, likes and dislikes, and sheer habit areall relevant (Krondl & Lau, 1982; Rozin, 1984; Parraga, 1990). Tastemay beparticularly important in selection of high fat diets, sincefats areresponsiblefor the texture and aroma of many foods (Drenowski, 1992). On the other hand, healthy diets may be consumed for non-health reasons such as concern about appearance (Cockerham, Kunz & Lueschen, 1988). Weight control is a major determinant of food choicefor individuals concerned about their body weight. The growth in environmental awareness over the past two decades has led to concerns about the use of natural ingredients and packaging that may have an impact on purchasing decisions. Thereis also evidencethat stress and negativeemotions may inuence food selection and consumption (McCann, Warnick & Knopp, 1990; Wardle, 1987a). All these factors indicate that health is only one of many con- siderations relevant to food choice. More eective implementation of health pro- motionstrategiesmaydependontherecognitionof thestatusof healthincomparison with other motives in theselection of food. Multidimensional Measures of Food Choice Eorts to develop multidimensional measures of factors related to food choice at the individual level, including both health and non-health motives, have been limited. Repertory grid approaches have been attempted, but these are time-con- sumingandpatternstendnot tobestableacrossindividuals(Bell et al., 1981; Tuorila & Pangborn, 1988). Ratings of dierent foods on dimensions such as pleasure, health, tradition, convenience, familiarity, prestigeand priceweredescribed by Lau et al. (1984) and by Rappaport and coworkers (1992), but in neither case was a formal set of measures developed. A more systematic method was developed by MichelaandContento(1986) inastudyof 5- to13-year-oldchildren. A seriesof foods wereratedonanumber of dimensionsincludinghealthfulness, taste, convenienceand social inuence. Substantial variations in the intraindividual correlations between 269 MOTI VES UNDERLYI NG THE SELECTI ON OF FOOD evaluativeratings and consumption of thesesamefoods werefound, and clusters of subjects characterized by dierent motivational patterns wereidentied. ThisapproachhasbeenextendedbyWardle(1993) who includedratingsof liking (taste) andhealthfor arangeof foodsinastudyof mothersandtheir adult children. Two indices wereconstructed for each individual based on thecorrelation between liking ratings and consumption frequency (taste index) and between healthiness ratings and consumption frequency (health index). Thetasteindex was consistently higher than the health index, although there was a good deal of variation across individuals in both. The health index was higher in older women than either their adult sons or daughters. The advantage of this method is that it circumvents the need for subjects to give introspective reports on their motives, but it has the drawback of being time-consuming and inappropriatefor large-scaleresearch. More conventional questionnaire methods have been developed for the in- vestigation of eating disorders, including the restraint scale (Polivy, Herman & Warsh, 1978), the Three Factor Eating Questionnaire (Stunkard & Messick, 1985) and theDutch Eating Behaviour Questionnaire(Van Strein et al., 1986). However, thesemeasures aremoreconcerned with cognitiverestraints on food intakeand the circumstances surrounding excessiveconsumption than with thefactors inuencing foodchoice. An18-itemReasonsfor EatingScalewasdevisedbyHarmatz andKerr (1981), and responses from110 students were factor analysed by Williams, Spence and Edelman (1987). Six factors emerged, highlighting theimportance of aect on eating, as well as sensory appeal, habit and pleasurein food preparation. However, the questionnaire was developed for the investigation of obesity, and is limited in scope. TheNutrition Attitudes Survey, a measureof attitudes relevant to low-fat diets, has been described by Hollis et al. (1986). Four factors emerged fromthesurvey of 357 adult volunteers, and included helpless and unhealthy, food exploration, meat preference andhealth consciousness. Thehelplessand unhealthy and meat preference factors were both associated with high meat consumption and with elevated low density lipoprotein concentration in the blood, while the health con- sciousness factor was associated with lower meat consumption. The helpless and unhealthy factor identied by Hollis et al. combines items about convenience (e.g. many days, becauseI min a hurry, I eat whatever is handy), mood and aect (I eat morewhen I feel down), and lowself-ecacy (when it comes to food, I have no will-power). Thehealth consciousness factor predominantly indexes willingness to change (to avoid heart disease, I would be willing to alter my eating habits) rather than concerns about health that currently govern food choice. The general applicability of themeasureis limited by theomission of items concerning sensory appeal and cost. Evaluationof thisliteraturesuggeststhat thedevelopment of amultidimensional questionnaireto assess theperceived importanceof dierent factors in food choice might bevaluableboth in theinvestigation of food consumption and in health and consumer psychology more broadly. The assessment of dierent factors within the samemeasureallows direct comparisons to bemadeabout therelativeimportance of dimensionssuchashealth, price, sensoryappeal andconvenience. Wehypothesized that several distinct factors associated with food choicewould beidentied as well as thefour listed above, sinceconcern with weight control, thedegreeof familiarity of the food, and mood have all been cited as relevant in previous studies. I n this paper, wedescribethedevelopment of themeasureand identication of dimensions 270 A. STEPTOE ET AL. through factor analysis, its validation in an independent sampleusingconrmatory factor analysis, testretest reliability over a 2- to 3-week interval, and associations with other variables such as sex, age, income, eating style, social desirability biases and health values. The relationship between responses to the questionnaire and patterns of food consumption is a separate issue related to the applications rather than validation of theinstrument, so will bepresented elsewhere. S1inx 1 I n Study 1, a preliminary food choice questionnaire of 68 items was generated through consideration of existing literature and discussion with nutritionists and health psychologists. I tems covered various motives that have been identied in other research, includingcommonlyrecognizedfactorssuchashealth, sensoryappeal, convenienceof purchaseand cost, together with areas of potential signicancesuch asenvironmental concernsandtheextent to which foodconformsto thepreferences of family and peer group. Measures of dietary restraint, emotional eating, social desirabilityresponsebiases, andhealthvaluewerealsoadministered, andassociations with food choicefactors evaluated as evidenceof convergent validity. Demographic characteristics including marital status, occupation, education, income and eating habits (vegetarian etc.) werealso collected. Mr1non Subjects Questionnaireswerepostedto105universitystudents, 90employeesof auniversity library, and 635 peopleselected at randomfromtheelectoral register of a London borough; 68%of thestudents, 64%of thelibrary employeesand 40%of theLondon residents returned completed questionnaires, giving a sample of 358. The sample included 220 women and 138 men, and ranged in age from1887 (mean 341, SD 152). Thebody mass index of participants averaged 237(SD 32). No respondents stated that they suered froma chronic diseaserequiringa special diet; 409%were married or living with a partner, 493% were single, and 98% were separated or widowed; 64 subjects (179%) reported that children lived at homewith them; 61% of thesamplewerein full or part-timeemployment, 232%werestudents, 89%were unemployed or homemakers and 67% were retired. I n terms of education, 108% had no formal qualications whilea further 156%left school havingcompleted the General Certicate of Secondary Education (GCSE, equivalent to tenth grade); 384%hadcompletedhighschool withA level qualications, and352%haduniversity degrees. Thereported annual incomeof subjectswasdistributed asfollows: <5000, 345% (mainly students); 500010000, 107%; 1000015000, 181%; 1500020000, 124%; 2000030000, 136%; >30000, 107%. Materials Preliminary food choicequestionnaire(FCQ) The preliminary questionnaire was designed to assess a wide range of con- siderations that might betaken into account by individuals when choosing what to 271 MOTI VES UNDERLYI NG THE SELECTI ON OF FOOD eat (see Table1). Subjects were asked to endorsethe statement I t is important to methat thefood I eat on a typical day . . . for each of the68 items by choosing between four responses: not at all important, a littleimportant, moderately important and very important, scored 1 to 4. Dietary restraint and eating style EatingstylewasassessedwiththeDutchEatingBehaviour Questionnaire(DEBQ) (Van Strein et al., 1986). This measureconsists of 33items concerningeatinghabits that assess threefactors: restrained eating, indexing restraint in theconsumption of foodanddietingbehaviour; emotional eating, whereitemsconcerneatingwhenupset or in negativemoods; and external eating, concerned with disinhibition and eating in responseto thesight or smell of food. Scoreson each scalecould rangefrom15, with higher scores indicating greater restraint, sensitivity to emotional conditions andsensitivityto external cues. Thescalehasadvantagesover other similar measures (Wardle, 1986), and has been shown to berobust in theU.K. population (Wardle, 1987b). The internal consistency (Cronbach ) scores in the present sample were 092, 095and077for therestraint, emotional eatingandexternal scalesrespectively. Valueof health The value that individuals place on good health was assessed using the Health as a Value scale developed by Lau, Hartman and Ware (1986). This four-item questionnaire consists of statements such as I f you dont have your health, you dont have anything, and responses were scored on a six-point scale where 1= strongly disagreeto 6=strongly agree. Ratings across thefour items wereaveraged to producescoresin therange16. TheHealth asaValuescaleisareliablemeasure that has been widely used in health research. Social desirability Social desirability biases were assessed using the ten-item reduction of the MarloweCrownesocial desirability scaledeveloped by Strahan and Gerbasi (1972), inwhichhigher scoresreect greater tendenciestowardsproducingsociallyfavourable responses. Rrsii1s ScaleConstruction and Factor Analysis The 68 items of the FCQ were factor analysed with varimax rotation. Various solutionswereconsidered, but thestructurethat appearedbest to combineecological sensewith parsimony involved ninefactorsthat together accounted for 495%of the variance, with Eigen values rangingfrom124to 172. I tems with a severely skewed distribution and thosewhich did not load clearly on a singlefactor werediscarded. I n order to develop a relatively short questionnaire, a maximumof six was set on the number of items included on each scale, and the highest loading items were selected. This procedureresulted in theretention of 36 items. Table1summarizesthefactor analysisperformed on the36itemFCQ. Thenine factors accounted for 652%of thevariance. Factor 1 consists of six health-related statementsandisthereforelabelledhealth(Cronbach=087). Factor 2iscomposed 272 A. STEPTOE ET AL. T:nir 1 Food ChoiceQuestionnaire items and factor loadings I t is important to methat thefood I eat on a typical day: Loading Factor 1 Health 22. Contains a lot of vitamins and minerals 077 29. Keeps mehealthy 075 10. I s nutritious 075 27. I s high in protein 072 30. I s good for my skin/teeth/hair/nails etc 068 9. I s high in breand roughage 066 Factor 2 Mood 16. Helps mecopewith stress 079 34. Helps meto copewith life 079 26. Helps merelax 078 24. Keeps meawake/alert 060 13. Cheers meup 060 31. Makes mefeel good 057 Factor 3 Convenience 1. I s easy to prepare 082 15. Can becooked very simply 081 28. Takes no timeto prepare 076 35. Can bebought in shops closeto whereI liveor work 065 11. I s easily availablein shops and supermarkets 059 Factor 4 Sensory Appeal 14. Smells nice 080 25. Looks nice 072 18. Has a pleasant texture 070 4. Tastes good 053 Factor 5 Natural Content 2. Contains no additives 081 5. Contains natural ingredients 072 23. Contains no articial ingredients 071 Factor 6 Price 6. I s not expensive 087 36. I s cheap 087 12. I s good valuefor money 076 Factor 7 Weight Control 3. I s lowin calories 087 17. Helps mecontrol my weight 079 7. I s lowin fat 074 Factor 8 Familiarity 33. I s what I usually eat 079 8. I s familiar 079 21. I s likethefood I atewhen I was a child 066 Factor 9 Ethical Concern 20. Comes fromcountries I approveof politically 087 32. Has thecountry of origin clearly marked 079 19. I s packaged in an environmentally friendly way 043 I temnumbers refer to theorder in which statements werepresented in thenal 36 itemFood ChoiceQuestionnaire. Thefull factor analysis matrix is availableon request fromtheauthors. 273 MOTI VES UNDERLYI NG THE SELECTI ON OF FOOD T:nir 2 I ntercorrelations (100) between Food ChoiceQuestionnairefactors Sensory Natural Weight Health Mood Convenience appeal content Price control Familiarity Mood 34 Convenience 14 27 Sensory appeal 19 32 5 Natural content 59 28 5 22 Price 20 14 32 4 9 Weight control 38 21 7 2 31 14 Familiarity 9 34 29 13 8 13 5 Ethical concern 37 25 12 13 39 22 9 10 p<0001. of six items concerningstress, copingand mood, and is consequently labelled mood (=083). Factor 3hasveitemsandconcernseaseof foodpurchaseandpreparation, and is therefore considered to be a convenience factor (=081). Factor 4 consists of four statements related to appearance, smell and taste, and can be regarded as indexing sensory appeal (=070). Factor 5 includes three items related to the use of additivesandnatural ingredients, andislabellednatural content (=084). Factor 6has threeitems associated with cost of food, and indexespriceas amotivein food selection (=082). Factor 7 consists of threeitems related to consumption of low calorie food and is labelled weight control (=079). Factor 8 is also composed of three items, and these are associated with familiarity (=070). Factor 9 has three items concerned with environmental and political considerations and is labelled ethical concern (=070). Scoresoneachscalewerecomputedbyaveragingunweightedratingsfor individual items, so couldrangefromaminimumof 1to amaximumof 4. Theintercorrelations between the scales of the FCQ are shown in Table 2. A number of signicant associations wereobserved. Themost prominent was between food choicemotives related to health and to natural content (r=059). There were also moderate correlations between health and mood, ethical concern and weight control, between mood, sensoryappeal andfamiliarity, andbetweenconvenienceandprice. However, noneof theseremaining correlations implied morethan 14%shared variance. Theassociations between theFCQ and other measures such as theDEBQ and Health as a Value questionnaire are discussed later in the section on convergent validity. S1inx 2 Study 2was designed to assess thereplicability of thenine-factor FCQ in a new community sample, and test the reproducibility scores over a 2- to 3-week period. I n addition, theassociations between theFCQ and two pertinent personality traits, 274 A. STEPTOE ET AL. neuroticism and openness to experience, were evaluated. The internal subscale of the Multidimensional Health Locus of Control (MHLOC), Health as a Value and thesocial desirability measurewerealso administered. Mr1non Subjects A postal survey was carried out with 400 students and 641 London residents. Replies were received from135 (340%) students and 223 (348%) residents, giving a sampleof 358. Themean agewas 305(SD 143) with a rangeof 1789years, and the sample included 184 women and 174 men. The mean body mass index was 238kg/m 2 (SD 35), and no participants reported sueringfroma chronic disease involvinga special diet. Theproportion of married subjects was 663%, while292% were single and 45%were divorced or widowed. The number of respondents with children livingat homewas 54(151%), a similar proportion to that found in Study 1. The proportion of respondents in full or part-time employment was 447%, 409%werestudents, 7%wereunemployed or homemakers and 64%wereretired. Educationally, 109%had no formal qualications, and 123%had achieved GCSE level. Theproportion with A levels(high school completion) was510%while258% had degrees. I n terms of annual incomedistribution, 501%reported an incomeof less than 5000; 500010000, 78%; 1000015000, 116%; 1500020000, 99%; 2000030000, 104%; >30000, 101%. Two weeks after receipt of thequestionnaires, a repeat questionnaire(consisting of theFCQ and a food frequency questionnaire) was sent out. Thefood frequency data arenot described in this report. A total of 245(684%) of subjects returned the second questionnairewithin a 48-day period. Theaveragenumber of days between completion of thetwo questionnaires was 197 (SD 52). Materials I nadditionto the36-itemFCQ, healthasavalueandsocial desirabilitymeasures described earlier, additional questionnaires wereas follows. Personality Two factors fromthe NEO Five-Factor I nventory (FormS) wereadministered (Costa& McCrae, 1991). Weselected neuroticismaspotentially relevant in thelight of itsrelationshipwithhealthrisk, stressandcoping, andtheopennessto experience factor aspotentially relevant to willingnessto eat awiderangeof foods. Each factor consists of twelve items rated on a ve-point scale, and scores could range from0 to48withhigher ratingsrepresentinggreater neuroticismandopennesstoexperience. Thereliability of themeasurehas been extensively evaluated, and it has been used widely inpersonality research, counsellingandhealth psychology (Costa& McCrae, 1992). Locus of control Respondents beliefs in their ability to inuence their own health status were assessed with theinternal health locus of control scale(FormB) fromtheMHLOC 275 MOTI VES UNDERLYI NG THE SELECTI ON OF FOOD (Wallston, Wallston & DeVellis, 1978). This scale consists of six items (e.g. I am directly responsible for my health), and subjects responded on a six-item scale ranging from 1=strongly disagree to 6=strongly agree. Ratings were averaged to produce a score in the range 16, with higher scores reecting greater perceptions of internal control over health. Data Analysis Conrmatory factor analysis was carried out with structural equation modelling using the EQS program (Bentler, 1989), with the generalized least squares (GLS) normal theory estimation method. Because 2 t indices aresensitiveto samplesize, models with good t may show large 2 values when the sample size is large. We thereforeusedtheratio of chi-squaredto degreesof freedomassuggestedby Marsh, Balla and McDonald (1988), with values under ve indicating reasonable t. I n addition, the non-normed t index (NNFI , Bentler & Bonett, 1980) and the com- parativet index (CFI , Bentler, 1990) wereused to evaluatemodel t. Rrsii1s Conrmatory Factor Analysis Thesimplenine-factor model in which each itemof theFCQ loaded on a single factor provided a good t for thedata collected in Study 2 (n=358), 2 =997, df= 594, p<0001; NNFI =0991; CFI =0991 (Fig. 1). All parameters estimates were signicant at p<0001. Allowingfactorsto intercorrelategaveamodest improvement in t. These analyses indicate that the FCQ factor structure identied in Table 1 was conrmed in theindependent Study 2 sample. TestRetest Reliability and I nternal Consistency I t can beseen that all correlations between scores at thetwo administrations of each FCQ scalewere>070 (Table3), suggesting that thereliability of thescales is acceptable. The mean scores on the two occasions were signicantly dierent for two factors, healthand mood, t (245)=242and 298respectively, ps<0025. I n each case, averageratingswereslightly reduced on thesecond occasion. I ntercorrelations betweenthefactorsweresimilar to thosedetailedfor Study1inTable2. Theinternal consistencyof theFCQfactorswashigh, withCronbach scoresasfollows: health= 081, mood=083, convenience=084, sensory appeal=072, natural content=086, price=083, weight control=085, familiarity=072, ethical concern=074. Associations with Sex, Ageand I ncome A comparison was made of body mass index and of scores on the FCQ scales in Studies 1 and 2, including sex as a factor and covarying for age. No signicant dierences between studies wereobserved. Consequently, thesamples fromStudies 1 and 2 werecombined for investigating dierences by sex, ageand income. Themultivariateanalysis of variancecomparing men and women with ageas a covariateshowedanoverall eect of sex, F(9,651)=143, p<0001(Table4). Separate 276 A. STEPTOE ET AL. Ficir 1. Summary of conrmatory factor analysis model for theFood ChoiceQues- tionnaire. Rectangles represent themeasured variables, with itemnumbers corresponding to items in theFCQ (seeTable1). Largecircles arelatent constructs (factors) and small circles areresidual variances (100). Parameter estimates arestandardized (100). T:nir 3 Testretest reliability of theFood ChoiceQuestionnaire(n=245) Scale Time1 Time2 Mean Standard Mean Standard Correlation deviation deviation coecient Health 283 072 277 070 0814 Mood 211 073 201 077 0771 Convenience 275 080 274 079 0830 Sensory appeal 299 063 294 065 0729 Natural content 247 086 248 086 0811 Price 283 080 279 079 0773 Weight control 238 088 237 084 0814 Familiarity 175 068 180 077 0714 Ethical concern 185 078 181 076 0801 univariate tests indicated signicant dierences on seven of the nine factors, F(1,659)=992to579, p<0001, withonlysensoryappeal andfamiliaritynot diering between thesexes. For all theother scales, theratings for women weresignicantly 277 MOTI VES UNDERLYI NG THE SELECTI ON OF FOOD T:nir 4 AverageratingsontheFoodChoiceQuestionnaireinmenandwomen. Studies 1 and 2 combined (n=706) Scale Men Women Mean Standard Mean Standard deviation deviation Health 264 077 301 062 Mood 200 073 221 074 Convenience 263 076 287 073 Sensory appeal 292 063 300 060 Natural content 227 084 264 083 Price 262 082 288 075 Weight control 201 081 260 079 Familiarity 179 071 179 071 Ethical concern 177 074 206 076 higher than for men. Eects were particularly prominent for health and weight control, with mean dierences of 036 and 059 respectively. There were a number of signicant correlations between FCQ scales and age. For both women and men, signicant positivecorrelations werefound between age and natural content (r=022 and 023, p<0001), familiarity (r=030 and 018, p<0001) and ethical concern (r=012 and 020, p<0025). I n addition, positive correlations in women were seen between age and health (r=022, p<0001) and sensory appeal (r=018, p<0001) as factors inuencing food choice. Among men, mood (r=017, p<0005) and weight control (r=025, p<0001) were positively cor- related with age. Neither sex showed any association between convenience or price and age. The signicance of these eects was unchanged when social desirability was taken into account using partial correlations. The impact of income on FCQ scales was evaluated in the expectation that peoplewith larger incomes would beless inuenced by pricein their food choices. The combined samples of Studies 1 and 2 were divided into three income groups: low (<5000, n=270), moderate (500020000, n=228) and high (>20000, n= 148). Multivariate analysis of variance with income as a grouping factor and age, sex and social desirability as covariates produced a signicant overall incomeeect, F(18,1266)=119, p<00001. Univariate analyses indicated that the income groups diered on three factors price, sensory appeal and familiarity, F(2,640)=311 to 904, p<005(Table5). An orderly relationship between incomeand theimportance of pricein food selection was apparent, with less emphasis beingplaced on priceby better-o individuals. Familiarity showed the reverse eect, being rated as more important by lower incomegroups. Sensory appeal was rated as less important by thelow-incomethan either themoderate- or high-incomegroups. Tests of Convergent Validity Scores on theshortened social desirability scalecorrelated with two of theFCQ scales: health (r=019, p<0001) and natural content (r=016, p<0001). This factor 278 A. STEPTOE ET AL. T:nir 5 Motives for food choicein relation to income Scale Lowincome Mediumincome High income Mean Standard Mean Standard Mean Standard deviation deviation deviation Health 282 073 296 067 271 073 Mood 213 075 213 072 202 072 Convenience 278 072 283 073 262 080 Sensory appeal 287 064 a 302 059 b 301 058 b Natural content 235 086 264 082 242 086 Price 312 070 a 279 069 b 209 065 c Weight control 230 088 242 082 227 083 Familiarity 183 070 a 179 068 b 164 070 c Ethical concern 191 078 204 076 179 072 I n each row, cells sharing thesamesuperscript werenot signicantly dierent on post hoc tests. T:nir 6 Valueof health and motives for food choice Scale Lowhealth value High health value Mean Standard Mean Standard deviation deviation Health 272 071 296 070 Mood 198 074 225 074 Sensory appeal 286 063 306 058 Natural content 231 082 261 086 Weight control 223 084 243 084 was thereforetaken into account when assessingrelationships with other measures. I norder to test theconvergent validity of thescales, anumber of apriori predictions weregenerated concerning associations with other questionnaires. The Health as a Value measure was administered in both studies, with the expectation that scores would correlate with the FCQ health score. Health value scores averaged 446 in women and 423 in men, t=317, df=710, p<0005. The associations with FCQ factors wereanalysed by dividingthepopulation by median split into high and lowvalueof health sub-groups. Multivariateanalysisof variance withhealthvaluesub-groupandsexasbetween-subject factorsandsocial desirability as thecovariatewas then carried out. Therewas no signicant interaction between health value and sex, but a main multivariate eect of health value was observed, F(9,647)=527, p<0005. Univariate eects were signicant for the FCQ health, mood, sensory appeal, natural content and weight control scales, F(1,647)=781 to 187, p<001 (Table6). As expected, respondents who placea high valueon health had signicantly greater scores on theFCQ health scale. However, they also rated mood, sensory appeal, natural content and weight control as moreimportant. The 279 MOTI VES UNDERLYI NG THE SELECTI ON OF FOOD value of health was not related to the importance placed on convenience, price, familiarity or ethical factors in determining thechoiceof food. Sincepreoccupationwithweightcontrol iscentral todietaryrestraint, wepredicted that the weight control factor would be correlated with the restraint scale of the DEBQ. Restraint scores werelowamongmen (mean 219, SD 080), so analyses of this factor wereconned to women. Usabledatawereavailablefrom214of the220 women in Study 1, and they weredivided by median split into thosewith low(mean 213, SD 050) and high (mean 353, SD 051) restraint scores. Multivariateanalysis of theFCQ with ageand social desirability as covariates showed a signicant eect of restraint groupoverall, F(9,189)=139, p<0001. Univariateeectsweresignicant only for two scales. As predicted, FCQ weight control scores were higher in the restrained than unrestrained women, mean 294, SD 064 vs. 212, SD 061, F(1, 197)=888, p<0001. I n addition, pricewasconsidered lesssignicant for high- than for low-restrained women, mean 272, SD 079 vs. 294, SD 068, F(1,197)=469, p<005. FCQ weight control ratings also correlated with DEBQ emotional eating scoresamongwomen(r=024, p<0001). I t wasalso foundthat theFCQ moodscale was positively correlated with both the emotional eating (r=038, p<0001) and external eating scales of theDEBQ. I n relation to locus of control, it was predicted that subjects with high internal MHLOC scores would be more concerned with health and the use of natural ingredients than those with less strong beliefs about personal responsibility for health. These predictions were fullled for both women and men with signicant correlationsbetweeninternal LOC scoresandFCQhealth(r=040and029, p<0001) and natural content (r=037 and 016, p<005). Controlling for social desirability, all correlationsin women remained signicant, asdid theassociationsbetween FCQ health and internal LOC in men. However, thecorrelation in men between internal LOC and FCQ natural content was no longer signicant after social desirability had been taken into account. Neuroticismas a personality disposition is associated with proneness to anxiety and stress responses. I t was therefore predicted that neuroticism might correlate positivelywiththeFCQ moodfactor. Theassociationwasnot signicant for women, but was conrmed for men (r=020, p<001). The openness to experience factor fromtheNEO was predicted to correlatenegatively with familiarity, and this eect wasconrmedfor bothwomenandmen(r=033and025respectively, p<0001). Theseeects remained signicant after controlling for social desirability biases. DiscissioN Our attempt to develop a brief questionnaire to assess perceived inuences on food selection at the individual level identied nine distinct factors. Several of the factors are similar to dimensions described by other investigators, including healthfulness, tasteor sensoryappeal, price, convenience, andtraditionor familiarity (Drenowski, 1992; Lau et al., 1984; Rappaport et al., 1992). Mood or aect and concern with weight control havealso emerged in previouswork (Holliset al., 1986; Williams et al., 1987). I n contrast, some potential inuences on food choice, such as how lling or satisfying the food is and how much it is liked by family and friends, did not emergeas consistent factors in theseanalyses. TheFCQ was shown to have high internal consistency in both studies. The conrmatory factor analysis 280 A. STEPTOE ET AL. suggested that the factors are robust, and adequate short-term stability has also been established. Sinceanumber of theFCQscaleswereintercorrelated, thepossibilityof asmaller group of higher order factors being useful was explored. A three-factor solution accounted for 58% of the variance, with the FCQ health, natural content, weight control and ethical concern loading on factor 1, convenienceand priceon factor 2, and mood, sensory appeal and familiarity on factor 3. However, wedo not favour this approach for two reasons. Firstly, someindividual FCQ scales did not load on single factors: mood and familiarity shared high loading on two or three factors. Secondly, it may be more useful to be able to investigate a wider range of specic motives related to food choice, rather than broad dimensions. I n terms of individual scales, the health factor contains items related to the prevention of chronic disease (e.g. high in bre and roughage) and to general nutrition and well-being (e.g. nutritious). The preliminary questionnaire ad- ministered in Study 1 contained several other items related to nutrition and health (e.g. easy to digest, part of a balanced diet) that loaded on the health factor, but lessstrongly than thesix itemsselected for thenal inventory. Thehealth factor also included theitemgood for my skin/teeth/hair/nailsetc. Thisisconsistent with evidencethat concern for appearancemay predict healthy dietary choices (Hayes & Ross, 1987). I t is interesting that the itemlow in fat did not load on the health but on the weight control factor. I t is possible that the association of dietary fat with weight is a consequenceof thesamplecontaining a high proportion of young adults, and that in a middle-aged group more conscious of cardiovascular disease risk, fat restrictionwouldhavebeenlinkedwithhealthandchronicdiseaseprevention. The sex dierence in ratings on the health scale indicates that women pay more attention to this factor than do men. The result is consistent with other studies of health-related behaviours and beliefs, which typically showhealthier dietary choices and more positive attitudes towards the health benets of salt and fat restriction and increased bre intake in women than men (Wardle & Steptoe, 1991). The importance of health as a reported motive for food choice increased with age in women but not men. Convergent validity for the health scale was provided by the signicant andpositiveassociationswithhealthasavalueandinternal health-related locus of control. Themood scalecontains items related to general alertness and mood, as well as to relaxation and stress control. The emergence of this factor suggests that mood and stress may play a rolein determining not only thequantity of food consumed, but also the selection of foodstus (Wardle, 1987a). Convergent validity for the moodfactor wasevaluatedbyitsrelationshipwiththeemotional eatingandexternal eatingscalesof theDEBQ. Theseeectswereseeninwomenbut not men, for whom DEBQ scores were very low, suggesting that there may be sex dierences in the relationshipbetweenaect-relatedfoodselectionanddietarydisinhibition(Grunberg & Straub, 1992). Among men, the mood factor was associated with neuroticism, implying that anxiety-proneindividuals may bemorelikely to beinuenced by the desire to maintain emotional well-being through eating than are emotionally more stablemen. Theconveniencefactor includesitemsrelatedbothtothepurchaseandpreparation of food, whilesensory appeal involves smell, tasteand appearance. Rappaport et al. (1992) found that health motives werenegatively correlated with convenience, while health and pleasure(sensory appeal) wereunrelated. Neither ndingwas conrmed 281 MOTI VES UNDERLYI NG THE SELECTI ON OF FOOD in thisanalysis, sincehealth wasnot signicantly associated with convenience, while health and sensory appeal showed asmall but reliablepositivecorrelation (Table2). The discrepancy with earlier ndings may have resulted from the very dierent measurement methods employed. I n the comparison of income groups, the low income group rated sensory appeal as less important than did better o subjects. Peoplewith less disposableincomemay not beableto taketasteinto consideration as much as wealthier individuals, and may haveto set other priorities. The natural content scale reected concern with the use of additives and the selection of natural ingredients. The correlation between the health and natural content factors was high (r=059 in Study 1 and r=063 in Study 2), suggesting a strong association between the two. Nevertheless, it would appear that general health-promotingaspectsof nutritionareperceivedasdistinct fromconcernsrelated to toxins and theingestion of unnatural non-foods added for cosmetic reasons. I t can be argued that additives and articial ingredients are frequently incorporated intofoodsaspreservatives, andmaythereforehavebenecial propertiesinpreventing consumptionof foodsthat havedecayed. I t wouldbeinterestingthereforetodiscover whether respondents in cultures with a less jaundiced opinion of thefood industry would viewnatural ingredients and theabsenceof additives so positively. Priceis an obvious inuence on food choice. Thecost of food is a much more important element inselectionamongpeoplewithlowincomescomparedwiththose that arebetter o (Table5). Pricewas also rated as moreimportant amongwomen than men. I n the U.K., women typically have responsibility for food shopping for the household, so may be more aware of budgetary limitations than are men with their morespontaneous pattern of food purchases. Another interesting observation is that women showing dietary restraint were less inuenced by price than the unrestrained. I t may bethat desireto eat lowcaloriefood outweighs considerations of cost for restrained eaters. I t was anticipated that weight control would emerge as a signicant factor in foodchoiceinthelight of theextensiveresearchondietaryrestraint andtheprevalent cultural preferencefor thin bodies. A signicant association was observed between weight control and mood as factors inuencing food choice, and this is consistent with previous research (Wardleet al., 1992). Thecorrelation of weight control with natural content (r=031 in Study 1 and r=042 in Study 2) reects the fact that manypeoplepractisingcaloricrestrictionfavour natural foodssuchasrawvegetables over prepared dishes. As predicted, scores on the FCQ weight control scale were higher in restrained than unrestrained women. They wereunrelated to income, but werehigher among respondents who stated that they valued health highly. The eighth factor to emerge in the factor analysis was labelled familiarity. I t includeditemsconcerninghowimportant it isfor thepersonto eat their accustomed diet, rather than beingadventurous in food choices. A positiveassociation between familiarity and mood was observed, suggesting that peoplewhosedietary selection is inuenced by the need to regulate stress responses also prefer familiar foods. Familiarity was one of the few factors not to dier on average between men and women. I t was, however, related to age, since it appears that older people are less adventurous in their choices. Associations were also observed between familiarity and income, with signicantly higher familiarity ratings among people with lower incomes. I t may bethat asincomesincrease, peoplearelessbound to buy only food they know about, and can aord to takegreater risks with food selection. For the nancially less well o, theconsequences of an unfortunatechoicearemoresevere, 282 A. STEPTOE ET AL. and people will be more likely to select what they know they like. Corroborative evidence concerning the familiarity scale is provided by the negative correlations withtheopennesstoexperiencefactor ontheNEOFive-Factor PersonalityI nventory. High scores on this personality factor are thought to reect openness to new experiences and broad interests, and might thereforeberelevant to adventurousness in food choice. Ethical concernemerged as an independent factor inuencingfood choice. I tems related to environmental and political issues loaded on this factor. I t is notablethat ethical concern was not correlated with social desirability scores at a statistically acceptablelevel, so endorsing theethical concern items was not simply a reection of presenting a set of motives that was presumed to be socially acceptable. I n the present sample, ratingsof ethical concernincreasedwithage, andwerehigher among women than men. Thissuggeststhat themalestudentswho constituted themajority of theyounger men in thestudy population wereparticularly unaected by ethical issues in relation to food choice. Sensory appeal, health, convenienceand pricearethemost important factorson average(Table4), with theveother factors being typically endorsed less strongly. Health is certainly not more important than other factors on average, and this supportstheargument that amultidimensional approach to motivesgoverningfood choice is appropriate. The variation in the relative importance of dierent factors for dierent segments of thepopulation may makeit possibleto createproles for distinct groups. For example, price was the most important factor for the lowest incomegroup, andsensory appeal for thebetter o (Table5). Appropriatestrategies for health promotion may perhaps bedeveloped for sectors with dierent priorities. I f, for instance, convenience takes precedence over health, then education and information about healthy food that is also readily available and easy to prepare might beof greater valuethan messages emphasizing health alone. Thelimitations of this study must berecognized. Data wereonly availablefrom the segment of the sample who returned the questionnaire, so may represent the viewsof individualswithaninterest infoodselection. Thequestionnaireisconcerned with thefactors that areperceived as relevant to food choice, and thesefactors do not necessarily reect actual dietary selection behaviour. Thecultural limitations of this approach to the investigation of food choice should also be recognized. The nine factors that emerged in this study were endorsed by members of a society in whichawiderangeof safeproductsarereadilyavailableandwell labelled. Awareness of environmental issues is relatively high, and a substantial proportion of food is importedsoissuessuchascountryof originaresalient. Consumersarenot restrained by actual shortages, or by seasonal variations to the extent that might be the case in rural, less auent or less developed societies. Dierent factors might assume greater importance in other cultures, in societies that produce most of their own food, or in populations exposed to an unpredictable food supply. Nevertheless, within western urban populations, the FCQ provides the opportunity to assess a broad rangeof factors perceived as relevant to food selection. RrrrrNcrs Bell, A. C., Stuart, A. M., Radford, A. J . & Cairney, P. T. (1981) A method for describing food beliefs which may predict personal food choice. J ournal of NutritionEducation, 13, 2226. 283 MOTI VES UNDERLYI NG THE SELECTI ON OF FOOD Bentler, P. M. (1989) EQS: Structural Equations Program Manual. Los Angeles: BMDP Statistical SoftwareI nc. Bentler, P. M. (1990) Comparative t indexes in structural models. Psychological Bulletin, 107, 238246. Bentler, P. M. & Bonett, D. G. (1980) Signicancetests and goodness of t in theanalysis of covariancestructures. Psychological Bulletin, 88, 588606. Cannon, G. (1992) Food and Health: TheExperts Agree. London: Consumers Association. Cockerham, W. C., Kunz, G. & Lueschen, G. (1988) On concern with appearance, health beliefs, andeatinghabits: areappraisal comparingAmericansandWest Germans. J ournal of Health and Social Behavior, 29, 265270. Costa, P. T. & McCrae, R. R. (1991) NEO PI -R Professional Manual. Odessa, U.S.A.: Psychological Assessment Resources I nc. Costa, P. T. & McCrae, R. R. (1992) Bibliography for theRevisedNEO Personality I nventory and NEO Five-Factor I nventory. Odessa, U.S.A.: Psychological Assessment Resources I nc. Drenowski, A. (1992) Nutritional perspectives on biobehavioral models of dietary change. I n: K. K. DeRoos (Ed.), Promoting Dietary Changein Communities. Pp. 96112. Wash- ington: Fred Hutchinson Cancer Research Center. Fine, B. & Leopold, E. (1993) TheWorld of Consumption. London: Routledge. Glanz, K., Kristal, A., Sorensen, G., Palombo, R., Heimendinger, J . & Probart, C. (1993) Development and validation of measures of psychosocial factors inuencing fat- and bre-related dietary behavior. PreventiveMedicine, 22, 373387. Grunberg, N. E. & Straub, R. O. (1992) The roleof gender and tasteclass in the eects of stress on eating. Health Psychology, 11, 97100. Harmatz, M. G. & Kerr, B. B. (1981) Over-eatingbehaviour: amulti-causal approach. Obesity and Metabolism, 1, 134139. Hayes, D. & Ross, C. E. (1987) Concern with appearance, health beliefs, and eating habits. J ournal of Health and Social Behavior, 28, 120130. Healthy People2000 (1992) Boston: J ones and Bartlett. Hollis, J . F., Carmody, T. P., Connor, S. L., Fey, S. G. & Matarazzo, J . D. (1986) The Nutrition Attitudes Survey: associations with dietary habits, psychological and physical well-being, and coronary risk factors. Health Psychology, 5, 359374. Kristal, A. R., Bowen, D. J ., Curry, S. J ., Shattuck, A. L. & Henry, H. J . (1990) Nutrition knowledge, attitudes, and perceived normsascorrelates of selectinglow-fat diets. Health Education and Research, 5, 467477. Krondl, L. & Lau, D. (1982) Social determinants in human food selection. I n: L. Barker (Ed.), The Psychobiology of Human Food Selection. Pp. 139151. Chichester: Ellis Horwood. Lands, W. E. N., Hamazaki, T., Yamazaki, K., Okuyama, H., Sakai, K., Goto, Y. & Hubbard, V. S. (1990) Changing dietary patterns. American J ournal of Clinical Nutrition, 51, 991993. Lau, D., Krondl, M. & Coleman, P. (1984) Psychological factors aectingfood selection. I n: J . Galler (Ed.), Nutrition and Behavior. Pp. 397415. NewYork: PlenumPress. Lau, R. R., Hartman, K. A. & Ware, J . E. (1986) Health as a value: methodological and theoretical considerations. Health Psychology, 5, 2543. McCann, B., Warnick, G. & Knopp, R. (1990) Changes in plasma lipids and dietary intake accompanying shifts in perceived workload and stress. Psychosomatic Medicine, 52, 97108. McCann, B. S., Retzla, B. M., Dowdy, A. A., Walden, C. E. & Knopp, R. H. (1990) Promoting adherence to low-fat low-cholesterol diets: review and recommendations. J ournal of theAmerican Dietetic Association, 90, 14081414. Marsh, H. W., Balla, J . R. & McDonald, R. P. (1988) Goodness-of-t indicesin conrmatory factor analyses: theeects of samplesize. Psychological Bulletin, 103, 391410. Michela, J . & Contento, I . (1986) Cognitive, motivational, social andenvironmental inuences on childrens food choices. Health Psychology, 5, 209230. Parraga, I . M. (1990) Determinants of food consumption. J ournal of the American Dietetic Association, 90, 661663. Polivy, J ., Herman, P. C. & Warsh, S. (1978) I nternal andexternal componentsof emotionality in restrained and unrestrained eaters. J ournal of Abnormal Psychology, 87, 497504. 284 A. STEPTOE ET AL. Rappaport, L., Peters, G., Hu-Corzine, L. & Downey, R. (1992) Reasons for eating: an exploratory cognitiveanalysis. Ecology of Food and Nutrition, 28, 171189. Rozin, P. (1984) The acquisition of food habits and preferences. I n: J . D. Matarazzo et al. (Eds), Behavioral Health: A Handbook of Health Enhancement and Disease Prevention. Pp. 590607. NewYork: Wiley-I nterscience. TheHealth of theNation (1992) London: HMSO. Sanjur, D. (Ed.) (1982) Social andCultural Perspectivesof Nutrition. EnglewoodClis: Prentice Hall. Smith, A. M. & Owen, N. (1992) Associations of social status and health-related beliefs with dietary fat and bredensities. PreventiveMedicine, 21, 735745. Strahan, R. & Gerbasi, K. C. (1972) Short, homogenous versions of the Marlowe-Crowne social desirability scale. J ournal of Clinical Psychology, 28, 191193. Stunkard, A. J . & Messick, S. (1985) Thethreefactor eatingquestionnaireto measuredietary restraint, disinhibition and hunger. J ournal of Psychosomatic Research, 29, 7184. Trenkner, L. L., Rooney, B., Viswanath, K., Baxter, J ., Elmer, P., Finnegan, J . R., Graves, K., Hertog, J ., Mullis, R., Pirie, P. & Potter, J . (1990) Development of a scale using nutrition attitudes for audiencesegmentation. Health Education Research, 5, 479487. Trichopoulou, A. D. & Efstathiadis, P. P. (1989) Changes of nutrition patterns and health indicators at thepopulation level in Greece. American J ournal of Clinical Nutrition, 49, 10421047. Tuorila, H. & Pangborn, R. M. (1988) Prediction of reported consumption of selected fat- containing foods. Appetite, 11, 8195. Van Strien, T., Fritjers, J . E. R., Bergers, G. P. A. & Defares, P. B. (1986) Dutch eating behaviour questionnaire for assessment of restrained, emotional and external eating behaviour. I nternational J ournal of Eating Disorders, 5, 295315. Wallston, D. S., Wallston, K. A. & DeVellis, R. (1978) Development of themultidimensional health locus of control (MHLC) scales. Health Education Monographs, 6, 160170. Wardle, J . (1986) The assessment of restrained eating. Behaviour Research and Therapy, 24, 213215. Wardle, J . (1987a) Compulsive eating and dietary restraint. British J ournal of Clinical Psy- chology, 26, 4755. Wardle, J . (1987b) Eatingstyle: avalidationstudyof theDutchEatingBehaviour Questionnaire in normal subjectsand women with eatingdisorders. J ournal of Psychosomatic Research, 31, 161169. Wardle, J . (1993) Food choices and health evaluation. Psychology and Health, 8, 6575. Wardle, J . & Steptoe, A. (1991) The European Health and Behaviour Survey: rationale, methodsandresultsfromtheUnitedKingdom. Social ScienceandMedicine, 33, 925936. Wardle, J ., Marsland, L., Sheikh, Y., Quinn, M., Fedoro, I . & Ogden, J . (1992) Eatingstyle and eating behaviour in adolescents. Appetite, 18, 167183. Williams, A., Spencer, C. P. & Edelman, R. J . (1987) Restraint theory, locus of control and thesituational analysis of bingeeating. Personality and I ndividual Differences, 8, 6774. Received 22 December 1994, revision 15 February 1995