The development and validation of Spanish

1 downloads 0 Views 134KB Size Report
was corroborated in a cross validation study with a second sample of college ..... across groups of: (a) covariance matrices, (b) factor form (number of factors), (c) factor ... que punto cada comentario describe como usted piensa o se siente en ...
Behaviour Research and Therapy 38 (2000) 1125±1138

www.elsevier.com/locate/brat

The development and validation of Spanish versions of the State and Trait Food Cravings Questionnaires p Antonio Cepeda-Benito a,*, David H. Gleaves a, M. Carmen FernaÂndez b, Jaime Vila b, Tara L. Williams a, Jose Reynoso a a

Department of Psychology, Texas A&M University, College Station, TX 77843-4235, USA b Departmento de PsicologõÂa, Universidad de Granada, Spain Received 3 March 1999; received in revised form 12 July 1999

Abstract Objective: We developed and tested the psychometric properties of Spanish versions of the Trait and State Food Cravings Questionnaires (FCQ±T and FCQ±S respectively). Method: The instruments were translated and adapted to Spanish and administered to undergraduate students from a Southern university in Spain (N = 271). The data were analyzed using con®rmatory factor analysis to compare the factor structure of the English and Spanish versions of both questionnaires. Results: The factors structure of both questionnaires obtained excellent ®t indices across their Spanish versions with the one exception that some factors of the FCQ±S were more highly intercorrelated among the Spanish sample than the American. Discussion: This study supports the conceptualization of food cravings as universal multidimensional motivational states that can be reliably measured and supports the use of the Spanish versions of the FCQ. 7 2000 Elsevier Science Ltd. All rights reserved.

In response to the often expressed dissatisfaction with the ``unscienti®c'' and ``imprecise'' de®nition and measurement of food cravings (e.g., Booth, Conner, & Gibson, 1989; Michener & Rozin, 1994; Gendall, Joyce & Sullivan, 1997; Gendall, Sullivan, Joyce, & Bulik, 1997; Weingarten & Elston, 1990), Cepeda-Benito and colleagues (Cepeda-Benito, Gleaves, Williams, Portions of this paper were presented at the annual convention of the Association for Advancement of Behavior Therapy, November 1998, Washington, DC. * Corresponding author. Tel.: +1-409-845-8038; fax: +1-409-845-4727. E-mail addresses: [email protected] (A. Cepeda-Benito), [email protected] (D.H. Gleaves). p

0005-7967/00/$ - see front matter 7 2000 Elsevier Science Ltd. All rights reserved. PII: S 0 0 0 5 - 7 9 6 7 ( 9 9 ) 0 0 1 4 1 - 2

1126

A. Cepeda-Benito et al. / Behaviour Research and Therapy 38 (2000) 1125±1138

& Erath, 1999; Williams, Gleaves, Cepeda-Benito, Samudio, & Harrison, 1998) developed and validated two multi-factorial, food craving instruments: the Food Craving Questionnaire-Trait (FCQ±T) and the Food Craving Questionnaire-State (FCQ±S). These instruments are consistent with the theory that food cravings can arise from or be expressed as physiologicallyor psychologically-mediated processes (Weingarten & Elston, 1990). The construction of the questionnaires was in accord with recent multidimensional measures of drug cravings (e.g., Ti€any & Drobes, 1991; Ti€any, Singleton, Haertzen, & Henning®eld, 1993). Cepeda-Benito et al. (1999) found the FCQ±T and FCQ±S to have excellent internal consistency, with the FCQ±T also showing good three-week test±retest reliability. The FCQ±S had lower test±retest reliability, but this ®nding was congruent with the expectation that psychological states change as a function of time and circumstances and was also demonstrated in an experimental manipulation (Cepeda-Benito et al., 1999: Study 2). Using con®rmatory factor analyses, the authors also obtained results supporting the notion that cravings have multidimensional features (Ti€any, 1990; Weingarten & Elston, 1990). That is, the analyses yielded excellent ®t indices for nine and ®ve factor solutions for the FCQ±T and FCQ±S respectively (Cepeda-Benito et al., 1999: Study 1). The factor structure of these questionnaires was corroborated in a cross validation study with a second sample of college students (CepedaBenito et al., 1999: Study 3). As support for the construct validity of the instruments, the FCQ±T subscales correlated signi®cantly with other trait eating-related measures; i. e., the hunger and disinhibition scales of the Eating Inventory (Stunkard & Messick, 1985, 1988), and the FCQ±S scores of participants who completed the questionnaire after a 12 h food deprivation period were considerably higher on all ®ve dimensions than the craving reports of subjects who ate breakfast just before completing the questionnaires (Cepeda-Benito et al., 1999: Study 2).

1. Cross-cultural research, eating disorders, and purpose of the study Cross-cultural investigations are important because human behavior and their controlling variables can be common (emic or universal), di€erent (etic, or culture-speci®c), or have both emic and etic aspects across di€erent cultures (Matsumoto, 1996). Therefore, cross-cultural research allows investigators to make inferences about the truth of human behavior and act accordingly. Conducting cross-cultural research on eating-related phenomena is particularly important given that some aspects of Western culture are thought to play a pivotal role in the nature of eating disorders and that many nations and cultures are increasingly adopting Western values (Pate, Pumariega, Hester, & Garner, 1992). The tenet that eating-disorders are culturally-speci®c originally stemmed from the common ®nding that anorexia and bulimia were largely restricted to young, white or Jewish females from achievement-oriented, upper-middle class families from Western countries (e.g., Bruch, 1978; Pierce, 1990). Whereas new evidence suggests that eating disorders are becoming more prevalent across other social classes and cultures than previously thought (Crago, Shisslack & Estes, 1996; Pate et al., 1992; Raich et al., 1992), the hypothesis that culture plays a fundamental role in the etiology and development of eating disorders continues to be sound. That is, studies consistently ®nd that exposure to and assimilation of Western (American)

A. Cepeda-Benito et al. / Behaviour Research and Therapy 38 (2000) 1125±1138

1127

values are clearly associated with the prevalence of eating disorders in non-Western cultures (Crago et al., 1996; Pate et al., 1992). Given the relevance of cross-cultural research on eating-related phenomena, the construct of food cravings in other cultures warrants study. If the construct generalizes to di€erent cultures, we would want to know how universal or culture-speci®c the relationships between food cravings and other eating-related variables are. As a ®rst step towards the achievement of these goals, the present study developed Spanish versions of the FCQ±T and FCQ±S and tested for their measurement equivalence with the English (USA) versions. The establishment of equivalence of measurement is perhaps the most important methodological issue in crosscultural research (Lonner & Malpass, 1994). Eating disorders research from Spanish speaking countries is scarce. The few available studies indicate that many eating disorder ®ndings generalize to Spanish speaking populations. For the most part, this research shows that body image disturbance and self-esteem are also associated with eating psychopathology in Spanish speaking countries (e.g., Martõ nez & Spinetta, 1997; Raich, Torras Claraso & Mora Giral, 1997). Other Spanish research has also shown that body image disturbances and low self-esteem are associated with the adoption of the Western ideal of thinness and what the achievement slimness symbolizes, i.e., success, control, attractiveness, competitiveness (Raich et al., 1992; Toro, Cervera, & PeÂrez, 1989). To our knowledge, there has only been one previous attempt to study food cravings in a Spanish speaking population. Mahaluf, Nader, Correa and Vargas (1991) interviewed Chilean pregnant women using a ``semi-structured opinion poll'' of 60 questions and identi®ed six distinct features of pregnancy-speci®c food cravings or antojos. The qualitatively identi®ed features were (1) pregnancy-dependent, (2) a sense of urgency, (3) problems delaying the craving, (4) the need to eat big amounts of food, (5) a sense of special satisfaction after eating the desired food, and (6) indisposition if the craving could not be satis®ed. It should be noted that the predominant meaning of the word ``antojo'' is ``whim''. Antojo is also used to name behaviors and actions that are motivated by capriciousness rather than by reason or need. Thus, the present study of cravings clearly broadens the range of eating-related phenomena that has been investigated with Spanish speaking populations to date.

2. Method 2.1. Participants and procedures Spanish participants were 271 volunteer undergraduate psychology students from the Universidad de Granada, a university in Southern Spain. Seventy ®ve percent of the participants were female. Ages ranged from 17 to 39 with a mean age of 20 (SD=2.82) years. The American participants were the 217 undergraduate introductory psychology students who are described in Cepeda-Benito et al., (1999). Sixty six percent of the American participants were female. Ages ranged from 18 to 44 with a mean age of 21 (SD=2.9). American students responded to posted and verbal announcements that requested volunteers for a psychology experiment. Those who agreed to participate received extra credit towards

1128

A. Cepeda-Benito et al. / Behaviour Research and Therapy 38 (2000) 1125±1138

their psychology courses (see Cepeda-Benito et al., 1999). Spanish participants were recruited through verbal announcements at the beginning of di€erent class periods. Participants anonymously completed the questionnaires during their regular class periods. Participants received no tangible compensation for their participation. All administration sessions were conducted in group format, with all participants completing the 88 and 60 item original versions of the FCQ±T and FCQ±S, respectively. Additionally, subjects completed a brief demographic sheet where they indicated their sex, age, and the time at which they last ate something.

2.2. Measures The FCQ±T was derived from a total of 88 statements that were generated using 10 theoretical dimensions of trait food cravings. Participants were asked to indicate how frequently each statement ``would be true for you in general'' using a six point scale that ranged from ``Never'' or ``Not Applicable'' to ``Always''. The FCQ±S was derived from a total of 60 statements representing seven dimensions of state food cravings. Participants were asked to indicate the extent to which they agreed with each statement ``right now, at this very moment'' using a ®ve point Likert scale that ranged from ``Strongly Agree'' to ``Strongly Disagree''. Cepeda-Benito et al. (1999) performed Con®rmatory Factor Analysis (CFA) and speci®ed a priori a 10-dimensional model for the FCQ±T, and a 7-dimensional model for the FCQ±S. For both instruments, the authors examined goodness of ®t as well as factor loadings and modi®cation indices to determine if items needed to be eliminated. The ®nal version of the FCQ±T contained nine scales and 37 items. The overall alpha for the FCQ±T was 0.97, and subscale alphas ranged from 0.81 to 0.94. Test±retest reliability for the FCQ±T was r = 0.88, and test±retest reliability for the subscales was good ranging from r = 0.72 to r = 0.88. The ®nal version of the FCQ±S contained ®ve scales and 15 items. The overall alpha for the FCQ± S was 0.94 and subscale alphas ranged from 0.82 to 0.88. As expected, the test±retest reliability for the state version was lower than that for the trait version (r = 0.56, with test±retest reliabilities for the subscales ranging from r = 0.40 to r = 0.63). The nine and ®ve factor models of the FCQ±T and FCQ±S, respectively, were tested and the results indicated good ®ts for both models. The authors also found support for the discriminant validity of both models as con®dence intervals (+ or ÿ2 standard error) around the factor correlations did not contain 1.0 for either instrument (see Cepeda-Benito et al., 1999). That is, all factors were independent of each other and there were no two factors that should have been combined. The original 88 and 60 items of the FCQ±T and FCQ±S, respectively, were independently translated into Spanish by the ®rst and last authors of the present article, both of whom are bilingual in Spanish and English. The two translations of each questionnaire were highly similar and easily consolidated into a single Spanish translation per questionnaire. The Spanish questionnaires were then translated back into English by an experienced and ¯uent Spanish-asa-second-language instructor whose native language is English. The back translations were then compared to the original English questionnaires item by item. Discrepancies were identi®ed and adjustments to the Spanish translation were resolved by discussion. Finally, the

A. Cepeda-Benito et al. / Behaviour Research and Therapy 38 (2000) 1125±1138

1129

Table 1 Comparing models for English and Spanish versions of the FCQ±T Model

Hypothesis

w2

df

GFIa

NFIb

TLIc

CFId

RMSEAe

1 2 3

Equal covariance matrices Equal no. of factors Equal factor structure Model 3-2 di€erence (LX invariant) Modi®ed structure model Model 4-2 di€erence Equal factor correlations Model 5-4 di€erence (LX+PH invariant) Equal error matrices Model 5-4 di€erence (LX, PH, and TD invariant)

927 1641 1711

703 1186 1223

0.91 0.80 0.79

0.99 0.97 0.97

0.98 0.99 0.99

1.0 0.99 0.99

0.037 0.040 0.041

1685

1221

0.79

0.97

0.99

0.99

0.040

1727

1257

0.78

0.97

0.99

0.99

0.040

2140

1294

0.76

0.96

0.99

0.98

0.053

4 5 6

Dw 2

Ddf

p

70

37

< 0.01

44

35

0.14

41

36

0.23

413

37

< 0.01

a

GFI = Goodness of Fit Index. NFI = Normed Fit Index. c TLI = Tucker-Lewis Index. d CFI = Comparative Fit Index. e RMSEA = Root Mean Square Error or Approximation. b

questionnaires were administered to a small group of Spanish college students and their feedback was incorporated into the ®nal Spanish translations. 2.3. Statistical methods 2.3.1. Di€erences in factor structure CFA was conducted with the generalized least squares (GLS) method (for recommendations for GLS see Anderson & Gerbing, 1988; Browne, 1982, 1984). For ®t indices, we examined the w 2 statistic, the Goodness-of-Fit index (GFI; JoÈreskog & SoÈrbom, 1993), the Normed-Fit Index (NFI; Bentler & Bonett, 1980), the Tucker-Lewis index (TLI; see Marsh, Balla & McDonald, 1988), the Comparative Fit Index (CFI; Bentler, 1990), and the root mean square error of approximation (RMSEA; Browne & Cudeck, 1993, Finch & West, 1997). Values of the GFI, NFI, TLI, and CFI range from zero to 1.00, with a value close to 1.00 indicating a better ®t (e.g., Byrne, 1989; Mulaik et al., 1989). For The RMSEA, values of less than 0.05 are considered a close ®t and less than 0.08 an adequate ®t; values of greater than 0.10 suggest room for improvement in the model (Finch & West 1997). The TLI, CFI, and RMSEA have been found to be una€ected by sample size (Bentler, 1990; Fan, Thompson & Wang, 1999; Marsh et al., 1988). To test for factorial invariance across the American and Spanish groups, we tested a sequence of multi-sample, ``stacked'' measurement models as described by Bollen (1989), Byrne (1989), JoÈreskog and SoÈrbom (1993, and Lomax (1983). Using only the 37 and 15 items of the ®nal versions of the FCQ±T and FCQ±S respectively, we sequentially tested the equivalence

1130

A. Cepeda-Benito et al. / Behaviour Research and Therapy 38 (2000) 1125±1138

Table 2 Factors (scale a's for Spanish version), items, and item factor loadings (standardized loadings from the LAMBDA X matrix) for Food Cravings Questionnaire±Trait for both English and Spanish versions/samples (when di€erent, loadings for the Spanish version appear in parentheses)a Having Intentions and Plans to Consume Food (0.57) 5 Sin duda alguna, las ganas de comer me hacen pensar en como voy a conseguir lo que quiero comer (Food cravings invariably make me think of ways to get what I want to eat) 18 Cada vez que deseo comer algo en particular, me pongo a hacer planes para comer (Whenever I have cravings, I ®nd myself making plans to eat) 23 Cuando se me antoja una comida, normalmente intento comerla tan pronto como pueda (When I crave certain foods, I usually try to eat them as soon as I can) Anticipation of Positive Reinforcement that may Result from Eating (0.72) 9 Como para sentirme mejor (I eat to feel better) 10 Algunas veces, mi vida parece prefecta cuando como lo que me apetece (Sometimes, eating makes things seem just perfect.) 15 Cuando como lo que deseo me siento mejor (Eating what I crave makes me feel better) 24 Comer lo que me apetece mucho me sienta estupendamente (When I eat what I crave I feel great) 37 Comer me alivia (When I eat food, I feel comforted) Anticipation of Relief from Negative States and Feelings as a Result of Eating (0.72) 16 Cuando como lo que deseo me siento menos deprimido (When I satisfy a craving I feel less depressed) 19 El comer me tranquiliza (Eating calms me down) 21 Despues de comer no tengo tantas ansiedades (I feel less anxious after I eat) Lack of Control over Eating (0.84) 2 Cuando tengo deseos intensos de comer, una vez que me pongo a comer no puedo parar de comer (When I crave something, I know I won't be able to stop eating once I start) 3 A veces, cuando como lo que se me antoja, pierdo control y como demasiado (If I eat what I am craving, I often lose control and eat too much) 22 Si tengo la comida que deseo, no puedo resistir la tentacioÂn de comerla (If I get what I am craving I cannot stop myself from eating it) 25 No tengo la fuerza de voluntad de resistir mis deseos de comer las comidas que se me antojan (I have no will power to resist my food crave) 26 Una vez que me pongo a comer tengo problemas en dejar de comer (Once I start eating, I have trouble stopping) 29 Si me dejo llevar por la tentacioÂn de comer pierdo todo mi control (If I give in to a food craving, all control is lost) Thoughts or Preoccupation with Food (0.80) 6 No hago mas que pensar en la comida (I feel like I have food on my mind all the time) 8 A veces me encuentro pensativo preocupado con comida (I ®nd myself preoccupied with food) 27 Por mucho que lo intento, no puedo parar de pensar en comer (I can't stop thinking about eating no matter how hard I try) 28 Gasto demasiado tiempo pensando en lo proÂximo que voy a comer (I spend a lot of time thinking about whatever it is I will eat next)

0.66 0.66 0.57 0.68 0.66 0.64 0.50 0.71 0.68 0.86 0.76 0.77 0.80 0.73 0.63 0.82 0.71 0.77 0.76 0.82 0.74

A. Cepeda-Benito et al. / Behaviour Research and Therapy 38 (2000) 1125±1138

1131

Table 2 (continued ) Having Intentions and Plans to Consume Food (0.57) 30 A veces me doy cuenta de que estoy sonÄando despierto y estoy sonÄando en comer (I daydream about food) 31 Cada vez que se me antoja una comida sigo pensando en comer hasta que como lo que se me antojo (Whenever I have a food craving, I keep on thinking about eating until I actually eat the food) 32 Cuando tengo muchas ganas de comer algo estoy obsesionado con comer lo que deseo (If I am craving something, thoughts of eating it consume me) Craving as a Physiological State (0.65) 11 Se me hace la boca agua cuando pienso en mis comidas favoritas (Thinking about my favorite foods makes my mouth water) 12 Siento deseos intensos de comer cuando my estoÂmago esta vacõ o (I crave foods when my stomach is empty) 13 Siento como que mi cuerpo me pidiera ciertas comidas (I feel as if my body asks me for certain foods) 14 Me entra tanto hambre que mi estoÂmago se siente com un pozo sin fondo (I get so hungry that my stomach seems like a bottomless pit) Emotions that may be Experienced Before or During Food Cravings or Eating (0.62) 20 Siento deseos de comer cuando estoy aburrido(a), enfadado(a), o triste (I crave foods when I feel bored, angry, or sad) 33 A menudo deseo comer cuando siento emociones fuertes (My emotions often make me want to eat) Cues that may Trigger Food Cravings (0.76) 1 Cuando estoy con alguien que esta comiendo me entra hambre (Being with someone who is eating often makes me hungry) 34 Cada vez que voy a un banquete termino comiendo maÂs de lo que necesito (Whenever I go to a bu€et I end up eating more than what I needed) 35 Para mi es difõ cil resistir la tentacioÂn de tomar comidas apetecibles que estaÂn a mi alcance (It is hard for me to resist the temptation to eat appetizing foods that are in my reach) 36 Cuando estoy con alguien que se pasa comiendo, yo tambieÂn me paso (When I am with someone who is overeating, I usually overeat too) Guilt from Cravings and/or for Giving in to Them (0.84) 4 Detesto no poder resistir la tentacioÂn de comer (I hate it when I give into cravings) 7 A menudo me siento culpable cuando deseo ciertas comidas (I often feel guilty for craving certain foods) 17 Cuando como algo que deseo con intensidad me me siento culpable (When I eat what I am craving I feel guilty about myself)

0.90 (0.40) 0.65 0.71 0.61 0.57 0.63 0.63 0.75 0.77 0.68 0.62 0.72 0.74 0.68 0.66 (98) 0.72

a

Instructions: Hemos escrito una lista de comentarios que gente ha hecho acerca de sentimientos, pensamientos, deseos, tentaciones, y antojos relacionados con la comida y el comer. Por favor, utilizando los nuÂmeros que hay debajo de cada descripcioÂn, escriba en el espacio de la izquierda la frecuencia con la que usted se siente asi, o hasta que punto cada comentario describe como usted piensa o se siente en general: Nunca (1), Raramente (2), Algunas veces (3), A menudo (4), Casi Siempre (5), Siempre (6). In a recent cross-validation of the English version we added two additional items to the ``Emotions'' scale: Cuando estoy muy estresado me entran deseos fuertes de comer and Me entran deseos fuertes de comer cuando estoy disgustado (a) (see Cepeda-Benito et al., 1999: Study 3).

across groups of: (a) covariance matrices, (b) factor form (number of factors), (c) factor structure (loadings), (d) correlations among the three factors, and (e) error matrices. To determine if constraining matrices to be invariant across the groups led to a signi®cant

1132

A. Cepeda-Benito et al. / Behaviour Research and Therapy 38 (2000) 1125±1138

decrease in model ®t, we performed w 2 di€erence tests comparing the constrained and unconstrained models (see Bollen, 1989; Byrne, 1989; JoÈreskog & SoÈrbom, 1989). 2.3.2. Correlation analyses To further test the validity of the FCQ±T and the FCQ±S, we correlated time since eating with each of the factor scores from both craving questionnaires. Time since eating was calculated by subtracting the time at which each participant last ate from the time at which the questionnaire was being administered. If the FCQ±T measures trait or typical characteristics of cravings within an individual, FCQ±T scores should be consistent across time and circumstances. Conversely, FCQ±S scores should be positively associated with time since eating (Hill, Weaver & Blundell, 1991).

3. Results The results of the sequence of models with the trait version are presented in Table 1. Results suggested the covariance matrices for the two groups were very similar (w 2 [703, N = 473]=927.56, GFI=0.94, NFI=0.97; TLI=0.99; CFI=1.0; and RMSEA=0.037). The model for the same number of factors also ®t well (w 2[1186, N = 473]=1641.29, GFI=0.80; NFI=0.97; TLI=0.99; CFI=0.99; and RMSEA=0.040). When we constrained the LX matrix to be invariant, there was a signi®cant loss in ®t (delta w 2 (37, N = 473)=70.13, p < 0.05). However, examination of modi®cation indices suggested that this was due to only two items. When we allowed these items to be estimated separately in the two groups, the ®t was much better and not signi®cantly poorer than the previous unconstrained model (delta w 2 [35, N = 473]=44.00, p = 0.14. When we then constrained the Phi matrix (factor correlations) to be invariant across groups, there was not a signi®cant loss in ®t (delta w 2 [36, N = 473]=41.95, p = 0.23), suggesting that the factor correlations were very similar across groups. Factor loadings are presented in Table 2. Constraining the error matrices to be invariant did lead to a signi®cant decrease in ®t (delta w 2 [37, N = 473]=413.58, p < 0.05). Because this is a very restrictive assumption, we did not pursue di€erences further at this time. Inter factor correlations for the FCQ±T ranged from 0.18 to 0.84, with an mean iter-factor correlation of 0.63 (SD=0.13). None of the con®dence intervals around the factor correlations (+ or ÿ2 standard errors; see Anderson & Gerbing, 1988) contained 1.0, supporting the discriminant validity of the model (that the items measured separate factors). The overall alpha for the Spanish version of the FCQ±T was 0.92. Individual alphas for three of the scales were somewhat low, ranging from 0.57 to 0.65, and the remainder six alphas ranged from 0.72 to 0.84 (see Table 2). The results for the same sequence of models for the state version of the instrument indicated the American and Spanish samples were quite similar (see Table 3). A test of equal covariance matrices suggested a good model ®t (w 2 (120, N = 475)=326.26, NFI=0.98; CFI=0.99; and RMSEA=0.085). The model for the same number of factors also had a good ®t (w 2 [160, N = 475]=375.24, GFI=0.89; NFI=0.98; TLI=0.99; CFI=0.99; RMSEA=0.075. When we constrained the factor loading to be invariant across groups, there was a signi®cant loss in ®t (delta w 2 [15, N = 475]=61.54, p < 0.01). Examination of modi®cation indices suggested that

A. Cepeda-Benito et al. / Behaviour Research and Therapy 38 (2000) 1125±1138

1133

Table 3 Comparing models for English and Spanish versions of the FCQ±State Model Hypothesis

w2

1 2 3

326.26 120 0.96 375.24 160 0.89 436.78 175 0.88

0.98 0.98 0.98

0.98 0.99 0.99

0.99 0.99 0.99

0.085 0.075 0.080

384.38 171 0.90

0.98

0.99

0.99

0.073

433.11 181 0.90

0.98

0.99

0.99

0.077

392.52 177 0.90

0.98

0.99

0.99

0.072

449.99 192 0.89

0.98

0.99

0.99

0.075

Equal covariance matrices Equal no. of factors Equal factor structure (LX invariant) Model 3-2 di€erence Modi®ed structure model Model 4-2 di€erence Equal factor correlations matrices (LX+PH invariant) Model 4-3 di€erence Modi®ed model Model 6-4 di€erence Equal error matrices (LX, PH, and TD invariant) Model 7-6 di€erence

4 5 6 7

df

GFIa NFIb TLIc CFId RMSEAe Dw 2

Ddf p

61.54 15

< 0.01

9.14 11

0.61

56.16 10

< 0.01

8.14

6

0.22

57.47 15

< 0.01

a

GFI=Goodness of Fit Index. NFI=Normed Fit Index. c CFI=Comparative Fit Index. d TLI=Tucker±Lewis Index. e RMSEA=root mean square error or approximation. b

the decrease in ®t was due to three items. We then allowed those to be estimated separately for the two groups and the model improved. The di€erence between this model and the unconstrained model was not statistically signi®cant (delta w 2 [11, N = 475]=9.14, p = 0.61). We then constrained the factor loading matrix (Phi) to be invariant across groups. This initially led to a signi®cant decrease in ®t (delta w 2 [10, N = 475]=56.16, p < 0.01). Examination of the modi®cation indices suggested that there were four factor correlations that seemed to di€er across groups. That is, factor correlations for the Spanish state questionnaire were considerably large (range=0.73±0.97, Mean=0.84, SD=0.09). However, only one of 10 con®dence intervals around the factor correlations (+ or ÿ2 standard errors) contained 1.0 (``Desire'' and ``Lack of Control''), providing partial support for the discriminant validity of the model. When we allowed the four factor correlations to be estimated separately, the ®t improved, and the ®t of this model was not signi®cantly di€erent from the unconstrained model (delta w 2 [6, N = 475]=8.14, p = 0.22). Factor loadings are presented in Table 4. As with the trait version, constraining the error matrix (TD) to be invariant, led to a decrease in ®t (delta w 2 [15, N = 475]=57.47, p < 0.01), and we did not further investigate di€erences here. The overall alpha for the Spanish version of the FCQ±S was 0.94. Individual alphas for the scales ranged from 0.74 to 0.88 (see Table 4). These alphas are fairly high considering that each scale contained only three items. For the Spanish sample, time since eating ranged from 0 to 17.25 h, with a mean of 3.6 (SD=3.00), a median of 2.5 h, and a mode of 2 h. That is, the length of food deprivation for

1134

A. Cepeda-Benito et al. / Behaviour Research and Therapy 38 (2000) 1125±1138

Table 4 Factors (scale a's for Spanish version), items, and item factor loadings (standardized loadings from the LAMBDA X matrix) for Food Cravings Questionnaire±State for both English and Spanish versions/samples (when di€erent, loadings for the Spanish version appear in parentheses)a An Intense Desire to Eat (0.88) 1 Ahora mismo, tengo un deseo intenso de comer una o varias comidas en particular (I have an intense desire to eat one or more speci®c foods) 2 Ahora mismo tengo un antojo por una o varias comidas en particular (I'm craving one or more speci®c foods) 3 Ahora mismo me urge comer una o varias comidas en particular (I have an urge for one or more speci®c foods) Anticipation of Positive Reinforcement that may Result from Eating (0.82) 4 Si pudiera comer una o varias comidas en particular me sentirõ a perfectamente (Eating one or more speci®c foods would make things seem just perfect) 5 Estoy seguro de que si comiera lo que deseo, mi humor mejorarõ a (If I were to eat what I am craving, I am sure my mood would improve) 6 Comer una o varias comidas en particular me harõ a sentir maravillosamente (Eating one or more speci®c foods would feel wonderful) Anticipation of Relief from Negative States and Feelings as a Result of Eating (0.77) 7 Si comiera algo no me sentirõ a tan deÂbil y aletargado (If I ate something I wouldn't feel so sluggish and lethargic) 8 Me sentirõ a menos antipaÂtico e irritable si pudiera satisfacer mis deseos de comer (Satisfying my craving would make me feel less grouchy and irritable) 9 Me sentirõ a maÂs alerta si pudiera satisfacer mis deseos de comer (I would feel more alert if I could satisfy my craving) Lack of Control over Eating (0.74) 10 Si tomara una o varias comidas en particular no podrõ a parar de comerlas (If I had one or more speci®c foods, I could not stop eating it) 11 Mi deseo de comer una o varias comidas en particular puede maÂs que yo (My desire to eat [one or more speci®c foods] seems overpowering) 12 Ahora mismo, se que voy a seguir pensando en lo que me apetece hasta que lo consiga (I know I'm going to keep on thinking about one or more speci®c foods until I actually have it) Craving as a Physiological State (0.81) 13 Tengo hambre (I am hungry) 14 Si pudiera comer algo ahora mismo, mi estoÂmago no se sentirõ a tan vacio (If I ate right now, my stomach wouldn't feel as empty) 15 Me siento deÂbil por no comer (I feel weak because of not eating.)

0.87 0.82 0.81 0.70 (0.95) 0.93 (0.20) 0.85 0.85 0.96 (0.53) 0.87 0.65 0.66 0.82 0.92 0.83 0.75

a

Instructions: Hemos escrito una lista de comentarios que gente ha hecho acerca de sentimientos, pensamientos, y deseos, tentaciones, y antojos relacionados con la comida y el comer. Por favor, utilizando los nuÂmeros que hay debajo de cada descripcioÂ, escriba en el espacio de la izquierda hasta que punto usted esta de acuerdo con que usted se siente asõ ahora mismo. Por favor note que algunas preguntas se re®eren a una o varias comidas en particular (comidas que le apetecerian ahora), mientras que otras se re®eren a comidas que le apetecen en general: Absolutamente nada de acuerdo (1), Casi nada de acuerdo (2), Neutral (3), Un poco de acuerdo (4), Muy de acuerdo (5).

the entire sample could be characterized as short, as 90% of the participants were included by a time-since-eating range of 1 to 7 h in a positively skewed distribution. All of the correlations between time since eating and the FCQ±T factors were below 0.08 and none were statistically

A. Cepeda-Benito et al. / Behaviour Research and Therapy 38 (2000) 1125±1138

1135

signi®cant. Conversely, the correlations between time since eating and the FCQ±S factors ranged between 0.19 and 0.35, and all were statistically signi®cant. That is, the data are consistent with the FCQ±T and the FCQ±S being a trait and a state measure, respectively.

4. Discussion This study found that the factor structures for the Spanish and English versions of the FCQ±T and FCQ±S to be very similar. That is, there were no major, factor structure di€erences for the two versions. For the FCQ±T, the groups were strikingly similar in most parameters, suggesting measurement equivalence across Spanish and American, college populations (see Little, 1997). Thus, these data support the generalizability of the conceptualization of cravings as a multidimensional motivational construct (Ti€any, 1990). Although we found moderate-to-high correlations among the factors, the data supported their discriminant validity. That is, none of the con®dence intervals around the factor correlations contained 1.0 (see Anderson & Gerbing, 1988), suggesting that the factors are indeed distinct constructs. Furthermore, the ®nding of high inter-factor correlations is not surprising because it suggests the presence of a higher-order unifying factor (food cravings). Thus, these results suggest that while total FCQ±T scores could be used as a general measure of trait cravings, individual FCQ±T factor scores could be useful in providing discriminant information in future research. It is possible that the FCQ±T could be used to identify and di€erentiate craving pro®les between speci®c populations. These data could then be used to test how di€erent pro®les are associated with the assessment, treatment, and prognosis of eating disorders. For example, we could hypothesize that among anorexia nervosa patients, those typed as restricting would score lower in the ``Lack of Control'' dimension than would those of the binge-eating/purging type (see also Gendall, Joyce, Sullivan & Bulik, 1998; Overduin & Jansen, 1996; van der Ster Wallin, Norring & Holmgren, 1994). For the FCQ±S, there was also considerable support for the factorial similarity across samples. Four out of ten factor correlations did, however di€er, with the correlations being consistently higher in the Spanish group. The fact that a few parameters di€ered makes the term ``partial invariance'' more appropriate to describe the structure equivalence results for the FCQ±S (see Byrne, 1989). Although only one of the inter-factor correlation intervals for the FCQ±S contained 1.0 (``Lack of Control/Desire''), the inter-factor correlations of the Phi matrix were nevertheless high (range=0.73±0.97). However, it should also be noted that the factors in the Phi matrix are theoretically measured without error and, thus, higher than correlations between factor scores that are generated by simply adding item scores. In the present case, the correlations between the simple factor scores for the FCQ±S ranged from 0.62 to 0.76. Therefore, for practical use in most research situations, the FCQ±S factors may not be so highly inter correlated. Spanish versions of both instruments demonstrated good overall internal consistency, suggesting the presence of higher-order factors (i.e. food cravings). The individual scale alphas were almost surprisingly high for the FCQ±S given that each scale had only three items (and alpha is a€ected by the number of items on a scale). On the trait version, the alpha for the

1136

A. Cepeda-Benito et al. / Behaviour Research and Therapy 38 (2000) 1125±1138

``Having Intentions and Plans to Consume Food'' factor was perhaps problematically low. However, this was a sub-scale that had only three items and the moderately high factor loading for all three items suggested adequate measurement of the latent factor. Nevertheless, if future research attempts to revise the questionnaire, adding items to those with the fewest items may improve their reliability. It is unclear why the structure equivalence was more strongly supported for the FCQ±T than the FCQ±S. However, the answer might be related to the ``Desire'' factor of the FCQ±S. That is, the FCQ±T does not measure the ``Desire'' dimension of craving, and this factor was involved in three of the four correlations that di€ered between the Spanish and American sample. Therefore, it is possible that the ``Desire'' dimension is a more general measure of cravings for Spaniards than for Americans. Nevertheless, the overall results indicated ``partial invariance'' and partial discriminant validity for the factor structure of the FCQ±S. As with the FCQ±T, the results provide support for the multidimensional nature of cravings as measured by the FCQ±S. Further evidence for the construct validity of the instrument was provided by the di€erential correlation patterns between time since eating and each of the FCQ instruments. These patterns of correlations provide further support for the construct validity of the instruments. While FCQ±T factor scores did not correlate with length of time since eating, FCQ±S factor scores were consistently associated with time since eating. That is, the FCQ±T was conceptualized as a trait measure and as such should be less in¯uenced by temporary and transient states than the FCQ±S; which was developed to measure state dependent cravings. These data are also consistent with the ®nding that a 12 h deprivation interval signi®cantly increased self-reported state cravings but no substantial changes in trait cravings (CepedaBenito et al., 1999: Study 2). Although the correlation between time since eating and the FCQ± S factors could be characterized as ``small to medium'' (i.e., 0.10 to 0.30, Cohen, 1988), they are sizable if we take into consideration that the range for time since eating was rather restricted and positively skewed (i.e., longer food-deprivation periods were under-represented). To our knowledge this study represents the most comprehensive cross-cultural examination of food cravings to date. We feel this type of research is timely as it is becoming increasingly evident that eating disorder phenomena are no longer limited to rich Western nations (Pate et al., 1992). The FCQ±T and FCQ±S have the potential of furthering cross-cultural eatingrelated research. Their established measurement equivalence should facilitate comparative research among Spanish and English speaking countries. The natural next step in this line of research should include examinations of the extent to which food cravings are associated with other eating-related variables, and investigations of whether these relationships are emic or etic.

Acknowledgements This research was supported by the Texas A&M University International Research Travel Assistance Grant Program and Psychology Department and the Departamento de PsicologõÂ a of the Universidad de Granada. We thank the Faculty, undergraduate assistants and participants from both departments for their generous help and facilitation of this research. We would also like to thank Lisa Cepeda for her assistance in translating the instruments.

A. Cepeda-Benito et al. / Behaviour Research and Therapy 38 (2000) 1125±1138

1137

Portions of the data reported here were presented at the annual meeting of the Association for the Advancement of Behavior Therapy, November 1998, Washington DC.

References Anderson, J. C., & Gerbing, D. W. (1988). Structural equation modeling in practice: A review and recommended two-step approach. Psychological Bulletin, 103, 411±423. Bentler, P. M. (1990). Comparative indexes in structural models. Psychological Bulletin, 107, 238±246. Bentler, P. M., & Bonett, D. G. (1980). Signi®cance tests and goodness of ®t in the analysis of covariance structures. Psychological Bulletin, 88, 591±606. Bollen, K. A. (1989). Structural equations with latent variables. New York, NY: John Wiley. Booth, D. A., Conner, M. T., & Gibson, E. L. (1989). Measurement of food perception, food preference, and nutrient selection. Annals of the New York Academy of Sciences, 561, 226±242. Browne, M. W. (1982). Covariance structures. In D. M. Hawkins, Topics in applied multivariate analysis (pp. 72± 141). Cambridge, England: Cambridge University Press. Browne, M. W. (1984). Asymptotically distribution-free methods for the analysis of covariance structures. British Journal of Mathematical and Statistical Psychology, 37, 62±83. Browne, M. W., & Cudeck, R. (1993). Alternative ways of assessing model ®t. In K. A. Bollen, & J. S. Long, Testing structural equation models (pp. 136±162). Newbury Park, CA: Sage. Bruch, H. (1978). The golden cage: The enigma of anorexia nervosa. Cambridge, MA: Harvard University Press. Byrne, B. M. (1989). A primer of LISREL: Basic applications and programming for con®rmatory factor analytic models. New York: Springer-Verlag. Cepeda-Benito A., Gleaves D.H., Williams T.L., Erath S.A. 1999. The development and validation of the State and Trait Food Cravings Questionnaires. Manuscript submitted for publication. Cohen, J. (1988). Statistical Power Analysis for the Behavioral Sciences (2nd ed.). Hillsdale, NJ: Erlbaum. Crago, M., Shisslak, C. M., & Estes, L. S. (1996). Eating disturbances among American minority groups: A review. International Journal of Eating Disorders, 19, 239±248. Fan, X., Thompson, B., & Wang, L. (1999). E€ects of sample size, estimation methods, and model speci®cation on structural equation modeling ®t indexes. Structural Equation Modeling, 6, 56±83. Finch, J. F., & West, S. G. (1997). The investigation of personality structure: Statistical models. Journal of Research in Personality, 31, 439±485. Gendall, K. A., Joyce, P. R., & Sullivan, P. F. (1997). Impact of de®nition on prevalence of food cravings in random sample of young women. Appetite, 28, 63±72. Gendall, K. A., Joyce, P. R., Sullivan, P. F., & Bulik, C. M. (1998). Food cravers: Characteristics of those who binge. International Journal of Eating Disorders, 23, 353±360. Gendall, K. A., Sullivan, P. F., Joyce, P. R., & Bulik, C. M. (1997). Food cravings in women with a history of anorexia nervosa. International Journal of Eating Disorders, 22, 403±409. Hill, A. J., Weaver, C. F., & Blundell, J. E. (1991). Food craving, dietary restraint and mood. Appetite, 17, 187±197. JoÈreskog, K. G., & SoÈrbom, D. (1993). LISREL 8 User's reference guide. Chicago: Scienti®c Software, International. Little, T. D. (1997). Mean and covariance structures (MACS) analyses of cross-cultural data: Practical and theoretical issues. Multivariate Behavioral Research, 32, 53±76. Lomax, R. G. (1983). A guide to multiple-sample structural equation modeling. Behavior Research Methods & Instrumentation, 15, 580±584. Lonner, W. J., & Malpass, R. S. (1994). Psychology and culture. Boston: Allyn and Bacon. Mahaluf, J. Z., Nader, A. N., Correa, E. D., & Vargas, J. V. (1991). Craving in pregnancy Ð Initial de®nition. Revista de Psiquiatria Clinica, 28, 118±125. Marsh, H. W., Balla, R., & McDonald, R. P. (1988). Goodness-of-®t indices in con®rmatory factor analysis: The e€ect of sample size. Psychological Bulletin, 103, 391±410. Martõ nez, E. T., & Spinetta, M. (1997). Behavior therapy in Argentina. The Behavior Therapist, 20, 171±174.

1138

A. Cepeda-Benito et al. / Behaviour Research and Therapy 38 (2000) 1125±1138

Matsumoto, D. R. (1996). Culture and Psychology. Paci®c Grove, CA: Brooks/Cole Publishing. Michener, W., & Rozin, P. (1994). Pharmacological versus sensory factors in the satiation of chocolate craving. Physiology and Behavior, 56, 419±422. Mulaik, S. A., James, L. R., Alstine, J. V., Bennett, N., Lind, S., & Stilwell, C. D. (1989). Evaluation of goodnessof-®t indices for structural equation models. Psychological Bulletin, 10, 430±445. Overduin, J., & Jansen, A. (1996). Food cue reactivity in fasting and non-fasting subjects. European Eating Disorders Review, 4, 249±259. Pate, J. E., Pumariega, A. J., Hester, C., & Garner, D. M. (1992). Cross-cultural patterns in eating disorders: A review. Journal of the American Academy of child and Adolescent Psychiatry, 31, 802±809. Pierce, K. (1990). A feminist theoretical perspective on the socialization of teenage girls through ``Seventeen'' magazine. Sex Role, 23, 491±500. Raich, R. M., Rosen, J. C., Deus, J., PeÂrez, O., Requiena, A., & Gross, J. (1992). Eating disorder symptoms among adolescents in the United States and Spain: A comparative study. International Journal of Eating Disorders, 11, 63±72. Raich, R. M., Torras, Claraso J., & Mora, Giral M. (1997). AnaÂlisis estructural de variables in¯uyentes en la aparicioÂn de alteraciones de la conducta alimentaria. Psycologia Conductual, 5, 55±70. Stunkard, A. J., & Messick, S. (1985). The Three-Factor Eating Questionnaire to measure dietary restrain, disinhibition and hunger. Journal of Psychosomatic Research, 29, 71±83. Stunkard, A. J., & Messick, S. (1988). Eating Inventory Manual. San Antonio: Hardcourt Brace & Company. Ti€any, S. (1990). A cognitive model of drug urges and drug-use behavior: Role of automatic and nonautomatic processes. Psychological Review, 97, 147±168. Ti€any, S. T., & Drobes, D. J. (1991). The development and initial validation of a questionnaire of smoking urges. British Journal of Addiction, 86, 1467±1476. Ti€any, S. T., Singleton, E., Haertzen, C. A., & Henning®eld, J. E. (1993). The development of a cocaine craving questionnaire. Drug and Alcohol Dependence, 34, 19±28. Toro, J., Cervera, M., & PeÂrez, P. (1989). Body shape, publicity and anorexia nervosa. Social Psychiatry and Psychiatric Epidemiology, 23, 132±136. van der Ster, Wallin G., Norring, C., & Holmgren, S. (1994). Binge eating versus nonpurged eating in bulimics: Is there a carbohydrate craving after all? Acta Psychiatrica Scandinavica, 89, 376±381. Weingarten, H., & Elston, D. (1990). The phenomenology of food cravings. Appetite, 15, 231±246. Williams, H., & Gleaves, D. (1998). The development of a food craving questionnaire. Poster presented at the Annual Conference of the Eastern Psychological Association, February.