The Family Fat Talk Questionnaire: Development and psychometric properties of a measure of fat talk behaviors within the family context

Body Image 12 (2015) 44–52

Contents lists available at ScienceDirect

Body Image journal homepage: www.elsevier.com/locate/bodyimage

The Family Fat Talk Questionnaire: Development and psychometric properties of a measure of fat talk behaviors within the family context Danielle E. MacDonald a,b,∗ , Gina Dimitropoulos b,c , Sarah Royal a,b , Andrea Polanco a , Michelle M. Dionne a a

Ryerson University, Toronto, Ontario, Canada University Health Network, Toronto, Ontario, Canada c University of Toronto, Toronto, Ontario, Canada b

a r t i c l e

i n f o

a b s t r a c t

Article history: Received 26 March 2014 Received in revised form 3 October 2014 Accepted 5 October 2014 Keywords: Fat talk Body image Family Psychometrics Exploratory factor analysis Confirmatory factor analysis

Fat talk has been well studied in female peer groups, and evidence suggests it may also be important in family contexts. However, no instrument exists to validly assess fat talk within the family. The purpose of this study was to develop a measure of fat talk within families and to establish its psychometric properties in young adult women. In Study 1, the Family Fat Talk Questionnaire (FFTQ) was developed and exploratory factor analysis suggested a 2-factor structure (“Self” and “Family” fat talk), and strong internal consistency. Study 2 confirmed its 2-factor structure using confirmatory factor analysis. Study 3 demonstrated the construct validity of FFTQ scores, including significant correlations with related constructs and predictable gender differences. Study 4 demonstrated the stability of FFTQ scores over two weeks. Therefore, the FFTQ produces valid and reliable scores of fat talk behaviors both exhibited and observed by young adult women within the family context. © 2014 Elsevier Ltd. All rights reserved.

Introduction Sociocultural messages about beauty often permeate social interactions and patterns of communicating (Smolak & Levine, 2001). The term fat talk was coined to describe negative bodyrelated conversations that occur between female adolescents (Nichter & Vuckovic, 1994). More specifically, fat talk has been defined as a normative, back-and-forth conversation pattern in which one or more girls/women makes disparaging comments about her own body (e.g., “I’m so fat!”), which leads the other girls/women involved to either negate the comments (e.g., “No you’re not!”) or to similarly disparage themselves (“No, I’m so fat!”; Nichter, 2000). Fat talk appears to serve numerous functions including the management of interpersonal relationships, the strengthening of emotional connections to peers, eliciting reassurance about one’s weight, preventing peer rejection (Nichter, 2000; Nichter & Vuckovic, 1994), as well as facilitating upward and downward social comparisons within female peer groups (Bailey & Ricciardelli, 2010). Males also engage in fat talk; however, the content of their conversations differs from women’s conversa-

∗ Corresponding author at: Department of Psychology, Ryerson University, 350 Victoria Street, Toronto, Ontario, Canada M5B 2K3. Tel.: +1 416 458 5840. E-mail address: [email protected] (D.E. MacDonald). http://dx.doi.org/10.1016/j.bodyim.2014.10.001 1740-1445/© 2014 Elsevier Ltd. All rights reserved.

tions (Engeln, Sladek, & Waldron, 2013). Fat talk is correlated with body dissatisfaction in both adolescent girls and women (Sharpe, Naumann, Treasure, & Schmidt, 2013), and body dissatisfaction increases immediately following experimental exposure to fat talk (Stice, Maxfield, & Wells, 2003). This latter finding suggests a temporal relationship between fat talk and state body dissatisfaction. Research has also shown positive correlations between fat talk, body shame, and restrained eating (MacDonald Clarke, Murnen, & Smolak, 2010; Royal, MacDonald, & Dionne, 2013). Although most research has focused on fat talk within peer groups, fat talk may also occur in, and have important implications within, the family context. Parental overvaluation of appearance and achievement of a low body weight may contribute to body dissatisfaction and under- or overeating in children (e.g., Keery, Boutelle, van den Berg, & Thompson, 2005; Kluck, 2008, 2010). Additionally, mothers who discuss weight may be more likely to have daughters with disordered eating (Fulkerson, McGuire, Neumark-Sztainer, Story, French, & Perry, 2002; Keery et al., 2005; Neumark-Sztainer, Bauer, Friend, Hannan, Story, & Berge, 2010). Furthermore, negative comments about appearance and appearance-related teasing by both parents and siblings is related to weight reducing practices, body dissatisfaction, low self-esteem, depression, and disordered eating in adolescent girls and young women (Eisenberg, Berge, Fulkerson, & Neumark-Sztainer, 2012; Keery et al., 2005; Kluck, 2010).

D.E. MacDonald et al. / Body Image 12 (2015) 44–52

Despite these findings, there is no published measure that adequately assesses fat talk in the family context. There are three validated measures of fat talk in peer contexts (i.e., EngelnMaddox, Salk, & Miller, 2012; MacDonald Clarke et al., 2010; Royal et al., 2013), but these measures do not query about fat talk within the family. Although the Parental Influence Questionnaire (Abraczinskas, Fisak, & Barnes, 2012) and Caregiver Eating Messages Scale (Kroon Van Diest & Tylka, 2010) assess parental influence on body image and eating behaviors, neither assesses family fat talk. Neither of these measures focuses on the body parts that are the specific targets of fat talk discussion, and the latter focuses primarily on eating-related messages. Furthermore, research on peer fat talk shows that both sides of the fat talk conversation are important (Salk & Engeln-Maddox, 2011), but neither measure assesses the respondent’s behaviors. Given the described relationships between negative comments and teasing about appearance from family members and elements of psychological wellness such as body dissatisfaction, restrained eating, and eating disorder symptoms, a psychometrically sound measure of fat talk that is specific to the family context is needed in this area. The Current Study Accordingly, the first goal of this study was to develop a measure of family fat talk by adapting a psychometrically sound measure of peer fat talk for undergraduate women – the Fat Talk Questionnaire (FTQ; Royal et al., 2013) – to be appropriate for use within the family. The second goal was to establish the family version of the FTQ’s preliminary psychometric properties in young adult women (35 and younger), including its factor structure, internal consistency, construct validity, and temporal stability. We chose to focus only on young women for this preliminary psychometric investigation (a) to be consistent with previous research on fat talk, (b) because research has indicated that the nature of fat talk differs by gender, and (c) because we expected that fat talk specifically within the family context might differ between older and younger women given their different roles within the family. Ethics approval was obtained from the university Research Ethics Board for all four studies reported within this paper. Study 1: Development and Exploratory Factor Analysis The goal of Study 1 was to develop the Family Fat Talk Questionnaire (FFTQ) items and to examine its internal factor structure using exploratory factor analysis among young adult women. Method Questionnaire development. The FFTQ items were developed by adapting the FTQ items (Royal et al., 2013) to be appropriate for use within the family. The FTQ is a 14-item self-report scale that asks respondents to indicate the frequency with which they engage in various fat talk behaviors when they are with similarweight female peers (e.g., “When I am with one or several close female friends, I complain that my stomach is fat”). Items are rated on a 5-point scale ranging from Never to Always. FTQ scores are computed by summing the responses for the 14 items. The FTQ consists of a single factor, and the items were found to be internally consistent (˛ = .94) and temporally stable (r = .90) over two weeks in undergraduates (Royal et al., 2013). Construct validity in undergraduates has been shown using significant correlations with measures of body dissatisfaction, body shame, body surveillance, restrained eating, social physique anxiety, and peer fat talk; a nonsignificant correlation with socially desirable responding; and significant differences in frequency of fat talk between men and women (Royal et al., 2013).

45

The first three authors developed the FFTQ items. The first and third authors were female PhD candidates in clinical psychology who specialize in eating disorders and have previous experience with scale development in the area of fat talk. The second author is a female PhD-level social worker who specializes in family issues related to eating disorders. We chose to adapt the FTQ items rather than develop new items because the original FTQ was originally developed using a rigorous qualitative methodology, and its content was reduced from a comprehensive list of 62 items. Additionally, the FTQ’s psychometric properties in undergraduate women were rigorously investigated and very strong. The rigorous item development is a strength of the FTQ, and as such, we chose to adapt these items as we expected that the content of typical fat talk comments would be similar between peer and family contexts. Additionally, by adapting the FTQ, there is correspondence between the peer and family fat talk constructs as assessed by these measures. The 14 items of the FTQ were adapted in two ways to assess both sides of fat talk conversations within the family: First, to assess the respondent’s own behaviors when interacting with her family members in the past year; and second, to assess behaviors that the respondent observed her family members engaging in during the past year. For example, the FTQ item “When I’m with one or several close female friend(s), I complain that I am fat” was adapted to include both of the following: “When I’m with my family members, I complain that I am fat,” as well as “When I’m with my family, I hear them complain that they are fat.” We elected to adapt the items in both ways because of the social nature of fat talk (Nichter, 2000) and because both sides of the conversation are important (Salk & Engeln-Maddox, 2011), particularly in the family context. Items were preceded by these instructions: “We are interested in the comments you say out loud when you are with your family members over the last year. We are also interested in the comments your family members made about their bodies over the last year. We define family broadly to include parents, siblings, partners, etc. Please keep this in mind when filling out the following questions.” We added this one-year timeframe (which the FTQ does not use) to ensure that participants reflected upon current behaviors, as it is possible that family fat talk may change as individuals pass through different developmental phases and as relationships with family members evolve over time. Items were rated on a 5-point scale from 1 (Never) to 5 (Always). After FFTQ items were developed, we independently consulted with two experts who are clinical psychologists specializing in eating disorders and have experience in body image scale development—neither was involved with the present study. Both experts agreed that the family fat talk construct had been comprehensively surveyed in the FFTQ items. Participants. Participants were female undergraduates (N = 278) who were recruited from the undergraduate psychology research participant pool. Ages ranged from 17 to 35 years (M = 19.1, SD = 2.6). The sample was ethnically diverse. The most common ethnicities represented were Caucasian (45.5%), East Asian (12.3%), South Asian (10.8%), mixed ethnicity (8.3%), Southeast Asian (7.9%), and Black (6.9%). Participants had a mean body mass index (BMI) of 22.2 kg/m2 (SD = 4.2). Measures. Preliminary Family Fat Talk Questionnaire (FFTQ). The 28item preliminary FFTQ, described above, was given to participants. Demographics questionnaire. Basic demographic information was collected, including age, gender (to confirm that they were female), ethnicity, and self-reported height and weight. Body mass

46

D.E. MacDonald et al. / Body Image 12 (2015) 44–52

Table 1 Factor loadings for exploratory factor analysis. Item

Factor 1

2

1. When I’m with my family members, I complain that my arms are too flabby. 2. When I’m with my family members, I complain that my body is out of proportion. 3. When I’m with my family, I complain that I am fat. 4. When I’m with my family, I complain that I should not be eating fattening foods. 5. When I’m with my family, I complain that my clothes are too tight. 6. I criticize my body compared to my family members’ bodies. 7. When I’m with my family members, I complain that I feel pressure to be thin. 8. When I’m with my family members, I complain that I’m not in shape. 9. When I’m with my family members, I hear them complain that their arms are too flabby. 10. When I’m with my family, I hear them complain about the proportion of their bodies. 11. When I’m with my family, I hear them complain that they are fat. 12. When I’m with my family, I hear them complaining that they should not be eating fattening foods. 13. When I’m with my family, I hear others complain that their clothes are too tight. 14. When I’m with my family members, I hear them criticize their bodies compared to their family members’ bodies. 15. When I’m with my family members, I hear them pressure each other to be thin. 16. When I’m with my family members, I hear others complain that they are not in shape.

−.10 .06 −.13 .00 .04 .09 .11 .06 .66 .79 .79 .68 .68 .73 .59 .72

.68 .63 .95 .61 .66 .64 .63 .68 −.02 .00 .01 −.06 .06 .02 .11 −.05

Note: Bolded values indicate significant factor loadings. N = 278.

index [BMI = weight (kg)/height (m2 )] was calculated for each participant. Procedure. The study was conducted online using Qualtrics software. Informed consent was obtained electronically prior to completing the questionnaires, and an electronic debriefing form appeared on the computer after the questionnaires explaining the purpose of the study. Participants were awarded partial course credit (1%) as compensation. Results Item analysis and reduction. In terms of data cleaning and missing data, 0.5% of total item-level data were missing. Data appeared to be missing at random, with no participants systematically missing multiple item responses, and no items systematically missing responses from numerous participants. Missing values were estimated using the mean of the other items focusing on the same target (i.e., respondent or family member; Tabachnick & Fidell, 2001). Initial item analysis was conducted on the 28 items and did not reveal any items with low item-total correlations (r < .30). Initial Cronbach’s alpha was very high, ˛ = .95, indicating redundancy between the items. In terms of item reduction, it was decided a priori that it was most desirable to conduct the item deletion such that both items from pairs with matching content (i.e., respondent’s behavior and family’s behavior) should be either retained or deleted, in order to preserve consistency in the reduced scale. The item reduction strategy involved examining inter-item correlations, and deleting one item from each pair with high or moderately high inter-item correlations. This was done in a stepwise manner such that the two items with the highest inter-item correlation were examined with respect to item content, and one item was selected for deletion based on a face-valid assessment of which of the two items appeared to be a better overall fit with the construct. This process resulted in the deletion of 12 items, leaving a total of 16 items. Cronbach’s alpha was .90 for these 16 items. Exploratory factor analysis. An exploratory factor analysis (EFA) using principal axis factoring was conducted on the retained 16 items, using an oblique rotation (promax), because we expected any emerging factors to be correlated. The Kaiser–Meyer–Olkin measure revealed appropriate sampling adequacy, KMO = .92. Barlett’s test of sphericity, 2 (120) = 2124.78, p < .001, indicated that the inter-item correlations were sufficiently large to conduct an

EFA. Parallel analysis was utilized to determine the eigenvalue cutoff for retaining factors, as this is recommended as being a more rigorous threshold than the commonly used Kaiser’s eigenvalue criterion of 1 (Fabrigar, Wegener, MacCallum, & Strahan, 1999). O’Connor’s (2000) SPSS syntax for parallel analysis was utilized, which runs 1000 iterations of random data eigenvalues in order to set the eigenvalue threshold for the 95th percentile of a random sampling distribution, based on the present sample characteristics (i.e., Cases = 278, Variables = 16). Two components emerged from the EFA, and both of their eigenvalues exceeded the corresponding eigenvalues produced by the parallel analysis (Criterion 1: 1.52; Criterion 2: 1.40). Combined, the two extracted factors explained 49.8% of the variance (Factor 1: eigenvalue = 6.11, 38.22% variance explained; Factor 2: eigenvalue = 1.85, 11.55% variance explained). The significance level for factor loadings was set at r = .46, p < .001, which was determined by doubling the critical value for a Pearson correlation using the selected alpha level and sample size of N = 278 (Stevens, 2002; a very conservative p-value was selected due to the large sample size). The pattern matrix was examined and revealed that the first eight items loaded onto Factor 2, and the latter eight items loaded onto Factor 1. No items loaded onto both factors. See Table 1 for item-factor loadings. In order to further validate the fit of this solution, the EFA was recomputed with a forced 2-factor solution. As expected, the results of the forced solution (eigenvalues, variance explained, and factor loadings) were completely identical to the unconstrained EFA. The results of Study 1 therefore indicate that the FFTQ has two factors, each with eight items. The item content clearly distinguished the factors, with Factor 1 reflecting fat talk exhibited by the respondent’s family members (“Family” subscale) and Factor 2 reflecting the respondent’s behavior (“Self” subscale). Cronbach’s alphas and subscale scores (comprised of item totals) were computed for the applicable items (Family: ˛ = .89, M = 19.9, SD = 6.6; Self: ˛ = .88, M = 17.9, SD = 6.6). Study 2: Confirmatory Factor Analysis The goal of Study 2 was to confirm the two factor structure of the FFTQ obtained in Study 1 using confirmatory factor analysis. Method Participants. Participants were 174 women recruited online using Cognilab/Mechanical Turk. Amazon Mechanical Turk is a crowdsourcing online labor market which allows recruitment of

D.E. MacDonald et al. / Body Image 12 (2015) 44–52

individuals (“workers”) to complete “microtasks” for a nominal payment, determined by the task “requester” (see Mason & Suri, 2012 for a detailed description of this platform). Mechanical Turk is only available to requesters in the United States (though the “workers” are sourced globally), but the Cognilab system is available to Canadian researchers and is directly integrated to recruit participants from Mechanical Turk’s participant pool. Research recruitment using Mechanical Turk is becoming increasingly popular, and recent evaluations of this method indicate that high quality data can be obtained using this platform quickly and with minimal expense (Buhrmester, Kwang, & Gosling, 2011; Mason & Suri, 2012; Paolacci & Chandler, 2014). Further, reliable and valid data on body image have been gathered via Mechanical Turk (Gardner, Brown, & Boice, 2012). Participants were women ranging in age from 19 to 35 (M = 27.9, SD = 4.0). The mean BMI (computed from self-reported height and weight) was 27.9 (SD = 7.6). The sample was ethnically diverse, with the most common reported ethnicities being South Asian (50.3%), Caucasian (30.1%), Southeast Asian (6.4%), and Black (3.5%), with the remainder of the sample comprising various other ethnicities. Procedure. Study 2 was advertised on Cognilab/Mechanical Turk as “Psychology research study related to body image and social behaviors (females only, approx. 5 mins).” If interested, participants could click on the task, which brought them to a consent form. Participants were paid $0.25 for their participation. After consent, they were taken to the task, which consisted of the 16-item FFTQ derived in Study 1 (Cronbach’s alpha in the present sample: Self, ˛ = .90; Family, ˛ = .93), followed by the demographic questionnaire. Gender was included in the demographic questionnaire in order to screen out male respondents. Results The confirmatory factor analysis was conducted using MPlus version 7 (Muthén & Muthén, 2012). Missing data occurred in 0.0% of the FFTQ items, though a small number of participants failed to report demographic variables (e.g., ethnicity), which were kept as missing. Individuals who failed to report age and/or gender were excluded. Based on the results of the EFA, a 2-factor solution was specified, with one latent factor for the Self subscale and one latent factor for the Family subscale. Because the distributions for some of the items were slightly skewed, we used Maximum Likelihood with Robust Estimation (MLR), which adjusts model fit indices to accommodate non-normality (Muthén & Muthén, 2012). Model fit was considered good if the root-mean square error of approximation (RMSEA) < .06, and both the comparative fit index (CFI) and the Tucker-Lewis index (TLI) > .95 (Hu & Bentler, 1999). In addition, because the subscales consist of item pairs matched in content and which differed only slightly with respect to the individual referred to by the item (i.e., respondent or family member), we estimated error correlations for item pairs across factors. For example, “When I’m with my family, I complain that my clothes are too tight,” which loads on the Self subscale, is a mirror image of the item “When I’m with my family, I hear others complain that their clothes are too tight,” which loads onto the Family subscale. When the wording of pairs of items are highly similar, it is often the case that errors for these items are highly correlated, reflecting measurement error that is due to a shared method effect (Brown, 2003; Kenny & Kashy, 1992). Thus, we planned a priori to freely estimate the cross-factor error covariances for the eight pairs of items. Because robust estimation was used, we conducted a scaled chi-square difference test to compare model fit (Satorra, 2000). The results showed that the two factor model including error covariances among the eight pairs of items provided a good fit to the data, scaled 2 (95) = 152.58, p < .001, RMSEA = .06, CFI = .96,

47

Table 2 Factor loadings for confirmatory factor analysis. Item

Standardized factor loadings Factor 1 (Family)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

Factor 2 (Self)

R2

.63 .73 .81 .75 .77 .67 .74 .76

.39 .53 .65 .57 .60 .46 .54 .58 .49 .61 .73 .60 .68 .60 .60 .58

.70 .78 .85 .78 .83 .78 .77 .76

Note: Bolded values indicate significant factor loadings. R2 reflects variance explained in each observed item. N = 174.

Table 3 Correlations between latent factors derived by confirmatory factor analysis.

Self factor with family factor Item 1 with Item 9 Item 2 with Item 10 Item 3 with Item 11 Item 4 with Item 12 Item 5 with Item 13 Item 6 with Item 14 Item 7 with Item 15 Item 8 with Item 16

Correlation

Standard error

p

.63 .38 .09 .005 .44 .26 .19 .33 .38

.08 .08 .10 .11 .10 .09 .11 .10 .10

<.001 <.001 .39 .96 <.001 .004 .07 .001 <.001

Note: N = 174.

TLI = .95. Tables 2 and 3 summarize the parameter estimates of this model, including the standardized factor loadings and variance explained in each observed item (Table 2), and the correlations between the latent factors as well as the error correlations (Table 3). Similar to Table 1, all items loaded highly on their respective factors, and the pattern of factor loadings was consistent with those observed in the EFA. Thus, these findings largely confirm the factor structure of the FFTQ indicated by the EFA in Study 1. Study 3: Construct Validity and Associated Psychometric Properties The goal of Study 3 was to examine the construct validity (i.e., convergent, divergent, and known groups validities) of the FFTQ in an undergraduate female sample. Convergent validity was examined using correlations between the FFTQ subscales and variables that should theoretically be related to the family fat talk construct. First, we hypothesized a positive correlation between family and peer fat talk. It was predicted that individuals who engage in more fat talk with family members would also be involved with more fat talk conversations with their peers, reflecting a general tendency to talk negatively about one’s body. Secondly, because fat talk involves negative bodyrelated conversations, we hypothesized that family fat talk would be positively related to body dissatisfaction, as this reflects the content of fat talk conversations. Thirdly, because the nature of fat talk involves objectifying one’s own body (i.e., speaking about it as if it is an object to be observed), we predicted that those who reported higher fat talk with family members would also report higher selfobjectification, defined as the extent to which a woman takes an outsider’s perspective of her own body via body surveillance, as

48

D.E. MacDonald et al. / Body Image 12 (2015) 44–52

well as body shame which is conceptualized as a consequence of body surveillance (McKinley & Hyde, 1996). Fourth, because fat talk is by definition a social behavior, we predicted that individuals who engage in more fat talk would have more body-related anxiety in social situations, reflected by positive correlations between family fat talk and social physique anxiety (i.e., anxiety about one’s body being viewed and evaluated by other people; Hart, Rejeski, & Leary, 1989). Finally, given the well-established relationship between body dissatisfaction and eating concerns, we predicted that individuals who engage in more family fat talk might also report higher restrained eating, which is characterized by efforts toward dietary control. Consistent with these hypotheses, previous research on the original FTQ found positive associations between peer fat talk and each of these constructs (e.g., Royal et al., 2013), suggesting that relationships between these variables and family fat talk are also likely to exist. Moreover, although we hypothesized that measures of each of these constructs would be correlated with both Self and Family subscales, given that these related constructs all reflect individual qualities, we expected stronger correlations with the Self subscale, and more modest correlations with the Family subscale. This would support that the subscales measure independent but related aspects of family fat talk. Finally, we also examined the relationship between family fat talk and respondent BMI on an exploratory basis given the measure’s relevance to body weight. However, because Royal et al. (2013) did not find a correlation between peer fat talk and BMI, we did not make any specific hypotheses about the directionality of this possible relationship. Discriminant validity was assessed using social desirability (i.e., presentation of a positive image of oneself to others; Crowne & Marlowe, 1960). Royal et al. (2013) found a nonsignificant relationship between the FTQ and social desirability, suggesting that participation in fat talk is a unique social behavior discrepant from a general motivation to be viewed in a socially desirable manner. This finding is important because fat talk may be a normative social behavior (Britton, Martz, Bazzini, Curtin, & LeaShomb, 2006). Thus, we did not expect an association between family fat talk and social desirability in the current study. Known groups validity was investigated by comparing scores between women and men. The FFTQ focuses on body image concerns that may be particularly relevant to women, and several studies have also demonstrated that men engage in less fat talk behavior than women (Martz, Petroff, Curtin, & Bazzini, 2009; Payne, Martz, Tompkins, Petroff, & Farrow, 2011; Royal et al., 2013). It was hypothesized that men would report less engagement in family fat talk conversations than women (Self subscale), but that both genders would perceive observing similar levels of fat talk from family members (Family subscale). Method Participants. Participants were female (n = 83) and male (n = 50) undergraduates. Ages ranged from 17 to 34 years of age (M = 20.3, SD = 3.3). The sample was ethnically diverse. The most common ethnicities represented were Caucasian (48.0%), mixed ethnicity (12.8%), East Asian (10.4%), South Asian (8.8%), Southeast Asian (8.0%), and Black (6.4%). Participants had a mean self-reported BMI of 22.7 kg/m2 (SD = 4.0). All subsequent data from Study 3 is reported on the female sample only, except for the known groups validity analyses. Measures. Family Fat Talk Questionnaire (FFTQ). The 16-item FFTQ was used. In the female sample, Cronbach’s alphas were ˛ = .88 for the

Self subscale and ˛ = .90 for the Family subscale. Respondents were also asked to indicate which family members they were thinking about when responding to FFTQ and were provided the following options: mother(s), father(s), brother(s), sister(s), partner(s), child(ren), and other. They could endorse as many family members as they wished. Fat Talk Questionnaire (FTQ). As described previously, the FTQ is a 14-item scale measuring fat talk behavior within female peers groups on a 5-point scale, ranging from 1 (Never) to 5 (Always). Scores are computed by totaling the items, with higher scores indicating greater engagement in peer fat talk. Previous research has reported excellent psychometric properties in undergraduate women, including internal consistency, test–retest reliability over two weeks, and construct validity (Royal et al., 2013). Cronbach’s alpha in the female sample was ˛ = .94. Body Shape Questionnaire (BSQ). The BSQ was administered to assess women’s concerns about their bodies. Its 34 items are rated on a 6-point scale, ranging from 1 (Never) to 6 (Always). Scores are computed by totaling the items, with higher scores reflecting greater body shape concerns. The BSQ has demonstrated good construct validity with female students, and scores have differentiated female students from women diagnosed with bulimia (Cooper, Taylor, Cooper, & Fairburn, 1987). Cronbach’s alpha in the female sample was ˛ = .98. Objectified Body Consciousness Scale (OBCS). The OBCS is a 24item self-report questionnaire of objectification of one’s own body, using a 7-point Likert scale ranging from 1 (Strongly Disagree) to 7 (Strongly Agree). The OBCS has a total score and three subscales: Surveillance, Body Shame, and Appearance Control Beliefs (eight items each). The Surveillance and Body Shame subscales were used in this study, which are computed by totaling the respective subscale items, with higher scores indicating greater body surveillance and body shame, respectively. The OBCS has demonstrated adequate internal consistency and good construct validity with undergraduate women (McKinley & Hyde, 1996). In the present sample, Cronbach’s alpha for the Body Shame subscale was ˛ = .70. Initially, Cronbach’s alpha for the Surveillance subscale was low, ˛ = .61. The deletion of one item (i.e., Item 5) improved the internal consistency to ˛ = .71; accordingly, this modified 7-item version of the subscale was used in the validation analyses. Revised Restraint Scale (RS). The RS contains 10 items that examine restrained eating (chronic dieting; Herman & Polivy, 1980). The RS does not assess “successful dieting” (i.e., caloric restriction leading to weight loss; Stice, Fisher, & Lowe, 2004), but rather reflects cognitive and behavioral efforts to manage eating, which may be punctuated by bouts of disinhibition (Heatherton, Herman, Polivy, King, & McGree, 1988). It has two subscales (in addition to a total score): Concern for Dieting, and Weight Fluctuations. Half of the items are scored on a 4-point scale (0–3), and half on a 5-point scale (0–4); scale anchors vary by item. The RS has been used extensively with undergraduates, and previous research has demonstrated good internal consistency, 1-week and 4-week test–retest reliability, and construct validity in female students (e.g., Allison, Kalinsky, & Gorman, 1992). Cronbach’s alpha in the female sample was ˛ = .82 for the total score. Social Physique Anxiety Scale (SPAS). The SPAS is a 12-item self-report questionnaire of social physique anxiety, measured on a 5-point scale ranging from 1 (Not at All) to 5 (Extremely Characteristic). Scores are computed by summing the items, with higher scores indicating greater social physique anxiety (Hart et al., 1989). The SPAS has shown good internal consistency, 8-week test–retest

D.E. MacDonald et al. / Body Image 12 (2015) 44–52

reliability, and construct validity in undergraduates (Hart et al., 1989), and positive correlations with peer fat talk have also been reported in undergraduates (Royal et al., 2013). Cronbach’s alpha in the female sample was initially low, ˛ = .51. Accordingly, we removed three items (i.e., 1, 5, and 8), resulting in a more acceptable Cronbach’s alpha of ˛ = .75; this modified 9-item scale was used in the validity analyses. Marlowe-Crowne Social Desirability Scale (SDS). The SDS is a 33-item questionnaire examining the extent to which respondents endorse behaviors that are socially acceptable but unlikely, thus assessing a response set oriented toward social desirability (Crowne & Marlowe, 1960). Items are scored on a dichotomous scale, with options being either False or True. Responses indicative of social desirability receive one point, and responses are summed. The SDS has shown good internal consistency and construct validity in undergraduate students (Crowne & Marlowe, 1960). Cronbach’s alpha in the female sample was initially, low ˛ = .64. Accordingly, we removed seven items (i.e., 4, 13, 16, 17, 18, 19, and 27), resulting in a more acceptable Cronbach’s alpha of ˛ = .70; this modified 26-item scale was used in the validity analyses. Demographics questionnaire. The demographic measure described in Study 1 was administered to gather demographic variables and self-reported height and weight. Procedure. Participants were recruited from the undergraduate psychology research participant pool, and completed the study online using Qualtrics software. The study was described as an investigation of the relationships between eating, body image, and social behaviors. They were first presented with the consent form followed by the questionnaire battery, which was presented in the same order for all participants. The debriefing form was presented to participants electronically at the end of study. Participants received partial course credit (1%) as compensation. Individuals were excluded if they participated in Study 1. Results Descriptive statistics. Females had a mean Self subscale score of 17.5 (SD = 6.6) and mean Family subscale score of 18.1 (SD = 6.6). In terms of missing data, 0.9% of total item-level data were missing. More data were missing from the Family subscale, but there were no missing data trends with respect to specific items. Where individuals were missing data from one to two items on a given subscale, the missing items were replaced with the subscale mean (prorated to the number of items available; Tabachnick & Fidell, 2001). In one instance, a participant was missing the entire Family subscale; these data were kept as missing. The 83 female participants reported fat talk behaviors from the following family members: 80 mothers, 52 fathers, 37 sisters, 37 brothers, five partners, zero children, and 27 “other” family members. Convergent and discriminant validity. In the female sample, Pearson r correlations were computed between scores on the FFTQ subscales and measures of peer fat talk, body dissatisfaction, objectified body consciousness (body surveillance and body shame), social physique anxiety, and restrained eating to establish convergent validity, as well as BMI. As predicted, the Self subscale was significantly correlated with all of these measures, except for body surveillance which was nonsignificant. The Family subscale was significantly correlated with all of these measures except body surveillance, restrained eating, and BMI. These latter correlations were also all more modest than those with the Self subscale. Additionally, the Self and Family subscales were significantly but

49

Table 4 Construct validation measures and correlations with FFTQ (female data). Variable

Correlation with FFTQ

Fat Talk Questionnaire Body Shape Questionnaire OBCS: Surveillancea OBCS: Body Shame Revised Restraint Scale Social Physique Anxiety Scaleb Social Desirability Scalec Body Mass Index

Self subscale

Family subscale

.59*** .63*** −.20 .44** .43*** .50*** −.16 .25*

.31** .27* .06 .22* .15 24* .14 −.17

M (SD)

33.6 (12.5) 99.4 (40.6) 25.0 (7.0) 29.7 (8.5) 14.5 (6.6) 26.6 (6.7) 17.5 (4.0) 22.5 (4.1)

Note: OBCS, Objectified Body Consciousness Scale. N = 83. a Modified 7-item scale. b Modified 9-item scale. c Modified 26-item scale to improve internal consistency. * p < .05. ** p < .01. *** p < .001.

modestly correlated (r = .34, p = .002), indicating that the two subscales represent two independent but related constructs. In terms of discriminant validity, neither Self nor Family subscale scores were correlated with social desirability. See Table 4 for correlations between the FFTQ and construct validation measures, as well as means and standard deviations of these measures, in the female sample. Known groups validity. FFTQ scores were compared between female and male participants using multivariate analysis of variance. There was a significant multivariate effect of gender, Wilks’  = .87, F(2, 127) = 9.61, p < .001, partial 2 = .13. Examination of univariate effects indicated gender differences only on the Self subscale. Females reported significantly higher mean Self subscale scores than males (Females: M = 17.6, SD = 6.5; Males: M = 12.8, SD = 5.1), F(1, 128) = 19.33, p < .001, partial 2 = .13. Family subscale scores were not significantly different between genders (Females: M = 18.1, SD = 6.6; Males: M = 16.6, SD = 6.7), F(1, 128) = 1.53, p = .22. These findings indicate that, consistent with hypotheses, women were significantly more likely than men to engage in fat talk behaviors with family members, but that men and women perceived their family members to engage in fat talk to a similar degree. Study 4: Test–Retest Reliability The purpose of Study 4 was to determine the temporal stability of the FFTQ using an undergraduate sample. It was predicted that frequency of engagement in fat talk in the family environment would be consistent across a 2-week period. Method Participants. Forty-four female undergraduate students participated in Study 4. They were entered in a draw to win one of two $10 gift cards to a restaurant as compensation. The sample included participants whose ages ranged from 19 to 31 years (M = 21.2, SD = 2.4). The sample was ethnically diverse. The most commonly represented ethnicities were Caucasian (65.9%), Southeast Asian (11.4%), East Asian (4.5%), South Asian (2.3%), Arab (2.3%) and mixed ethnicity (6.8%%). The average self-reported BMI was 21.4 kg/m2 (SD = 3.6). Procedure. Two researchers attended two upper-level psychology courses at Time 1 and described the study verbally to each class. Participants were told that we were conducting research on body image and the family, and were not informed that the current study was interested in temporal stability or that the same

50

D.E. MacDonald et al. / Body Image 12 (2015) 44–52

questionnaire would be administered on both occasions. Interested participants provided written consent prior to participating. They included a unique identification code on their FFTQ questionnaires, composed of the first two letters of their mother’s maiden name and last two digits of their phone number, which permitted linking of data from Times 1 and 2 while providing adequate anonymity, a procedure that has been successfully employed in other studies (e.g., Royal et al., 2013; Tylka, Bergeron, & Schwartz, 2005). Two weeks later, the experimenters returned to the same classes and participants completed the FFTQ and a demographics questionnaire, followed by debriefing about the purpose of the study. Results For the 44 participants who completed the study at both Time 1 and Time 2, 0.0% of item level data were missing. However there were 26 additional participants who completed the questionnaire at Time 1 but not at Time 2; these participants were excluded. FFTQ subscale scores were significantly correlated between Time 1 and Time 2: Self, r = .80, p < .001; Family, r = .76, p < .001. Furthermore, the mean scores were compared using paired samples t tests. There were no significant differences between administrations for the Self subscale t(40) = −0.55, p = .58, or the Family subscale, t(40) = 1.70, p = .10. Thus, scores were stable over two weeks. Internal consistency of the Self subscale was ˛ = .80 at Time 1, and ˛ = .86 at Time 2, and for the Family subscale was ˛ = .84 at Time 1, and ˛ = .87 at Time 2. Discussion The goal of the current study was to develop a measure of family fat talk and to establish its factor structure and psychometric properties in young adult women. The 16-item FFTQ was adapted from a measure of peer fat talk, with matched-content items assessing both sides of the fat talk conversation – statements made by the respondent, and statements the respondent has observed being made by her family members. Exploratory factor analysis suggested a 2-factor structure, one factor reflecting the respondent’s behaviors and the other reflecting family member behaviors, and confirmatory factor analysis confirmed this factor structure. The factors were named the “Self” and “Family” subscales, and both subscales had strong internal consistency. Convergent validity was demonstrated by showing that scores on both subscales were associated with measures of related constructs, including peer fat talk, body dissatisfaction, body shame, and social physique anxiety, indicating that individuals who report engaging in and hearing fat talk in the family are also more likely to report negative body image and related concerns. Scores on the Self, but not the Family subscale were also related to restrained eating and BMI, indicating that women who engage in greater fat talk in the family context are more likely to have higher body weight, and are also more likely to engage in efforts to control their weight. Collectively, these findings support the convergent validity of the FFTQ. Additionally the subscale scores were also modestly but significantly correlated with one another, demonstrating that they represent related but separate constructs. Finally, given that each of the aforementioned constructs used for convergent validity represent individual behaviors/qualities, the fact that the significant correlations were stronger for the Self compared to the Family subscale is in further support of its construct validity. Interestingly, neither subscale was related to body surveillance, suggesting that participation in family fat talk and the tendency to closely observe one’s own body may not be associated. In contrast, peer fat talk has been shown to be associated with body surveillance (Royal et al., 2013), which supports that family and peer fat talk are independent constructs.

However, it is important to note that the Cronbach’s alpha for the body surveillance subscale was unusually low in our sample, which may have attenuated its correlations with other constructs. Discriminant validity was demonstrated by showing that neither subscale correlated with social desirability. In addition, women reported engaging in more fat talk in the family compared to men, but both genders perceived observing similar amounts of fat talk from their family members. Finally, scores on both subscales were reliable over a 2-week period. Accordingly, FFTQ scores can be considered a valid and reliable measure of family fat talk in young adult women. This finding is important because research on fat talk indicates that this is a common social behavior, particularly for young adult women, and that there may be a number of negative consequences of fat talk, such as increased body dissatisfaction (e.g., Stice et al., 2003). Moreover, research suggests that negative body-related conversations specifically in the family context may predict body dissatisfaction (Helfert & Warschburger, 2011), weight reducing practices (Eisenberg et al., 2012; Kluck, 2010), and the development of eating disorders (Neumark-Sztainer et al., 2007) in girls and young women. The development of the FFTQ provides a psychometrically sound instrument by which fat talk in the family can be measured in future research on this topic with young adult women.

Strengths, limitations, and future directions Strengths of this research include the four study structure and the methodological rigor within each study, which permitted the comprehensive assessment of the FFTQ’s psychometric properties. Large samples in Studies 1 and 2 were informed by methodological recommendations about sample size for factor analysis (Mundfrom, Shaw, & Ke, 2005). The samples in all four studies were ethnically diverse, which increases the generalizability of the findings and is an important strength particularly given that body image and eating disorder research often reports data on predominantly Caucasian samples. Moreover, the use of a non-undergraduate sample in Study 2 to successfully confirm the factor structure derived with undergraduates in Study 1 suggests that the FFTQ’s structure is likely stable across different groups of young women (though additional research is needed for confirmation). Finally, this study replicates and extends the original Fat Talk Questionnaire study (Royal et al., 2013), from which the FFTQ was developed, by showing that the respondent’s own fat talk behaviors (i.e., Self subscale scores) fall on a single factor. The addition of the Family subscale is an important extension because it captures the respondent’s perception of the other side of a reciprocal social behavior. As such, the development of the FFTQ was both theoretically driven and methodologically rigorous, and the results support the psychometric properties for its use with young women to assess fat talk within the family. Additional strengths include that the measure is brief and easy to administer, understand, complete, and score. Despite its strengths, the study has a number of limitations. First, we investigated the structure of the FFTQ only in young women (aged 35 and younger). This decision was made because we predicted that family behaviors may vary as women age and assume different family roles and as relationships with other family members change. It is not yet known whether the FFTQ is a valid measure of family fat talk in middle aged or older women. Another important limitation is that although we assessed “both sides” of the conversation, these perspectives were only as perceived and self-reported by the respondent. Collateral information from other family members is important to confirm these findings, particularly with respect to the validity of the Family subscale scores. It was unfortunately not feasible to assess multiple members of the same family in the

D.E. MacDonald et al. / Body Image 12 (2015) 44–52

current research, but this is an important area for future study. For example, it would be interesting to have the FFTQ completed by mother–daughter dyads to determine whether both parties’ perspectives on family fat talk converge. Finally, although participants reported the family members to whom they were referring when completing the FFTQ, we did not assess family fat talk with a specific family member. For example, it might be interesting to assess women’s fat talk conversations specifically with their mothers. We did not restrict our research question to mothers given the preliminary nature of this research, and we could not examine fat talk with mothers specifically because although the majority of respondents endorsed mothers on the family member list, other family members were also reported. We could also not compare individuals who did versus did not endorse mothers in the family list in Study 3, as the latter category included only three respondents. In addition, the FFTQ specifies that individuals consider fat talk that they have heard or participated in within the family context only in the past year. This timeframe was used in order to ensure respondents were reporting current fat talk behaviors, as we expected that fat talk might differ during different developmental periods. However, it is also possible that fat talk in the family at specific developmental periods (e.g., adolescence) may be most important in terms of impacting body image. Future research might investigate exposure to fat talk in the family context during a specific developmental period (e.g., ages 12–16). For example, longitudinal research examining whether exposure to fat talk in early adolescence predicts body image concerns or disordered eating in late adolescence would help to illuminate whether family fat talk plays a role in the development of such concerns. In terms of other methodological limitations, counterbalancing the order of questionnaires in Study 3 could have better controlled for order effects. Cronbach’s alphas were low on the measures that were administered last (i.e., SPAS, SDS, and OBCS). These measures typically produce more acceptable alphas, and suggest that participants may have been tired and may have begun to respond more randomly or inattentively. Counterbalancing the measures would have helped to control for this. Additionally, this body of research but particularly Study 2 (given its online recruitment source) could have benefitted from validity questions to screen out inattentive or random responders. Finally, there are a number of other populations for which family fat talk might be relevant, but which the present research does not address. For example, research shows that men engage in fat talk but the content of their conversations differ from women’s conversations (Engeln et al., 2013). It would be interesting to study family fat talk in fathers, sons, and brothers, to understand how their fat talk patterns compare to mothers, daughters, and sisters. It is likely, however, that the targets of men’s fat talk conversations focus on different body concerns than those assessed by the FFTQ items (e.g., muscularity and leanness rather than thinness; Engeln et al., 2013). Additionally, given the well-established role of fat talk in adolescent peer groups (Nichter, 2000) as well as developmental changes in family relationships, it would be interesting to explore the psychometric properties of the FFTQ in adolescents, and compare the findings to adult women. Finally, it is likely that fat talk in the family context is also relevant to women with eating disorders, both potentially in terms of women who are acutely affected by an eating disorder, as well as in terms of barriers to recovery for women who are in treatment or working on maintaining recovery. For example, understanding whether eating disorder patients hear or participate in fat talk in the family may allow clinicians to educate family members on the effects that these conversations may have on their family member’s recovery, and to help them learn more recovery-oriented ways of communicating. The specific nature and impact of family fat talk on women with eating disorders is a yet

51

unstudied area. If the FFTQ is valid in this population, this measure would permit future research on family fat talk in women with eating disorders. Conflict of interest None of the authors have any conflicts of interest or funding sources to declare. Acknowledgements The authors would like to express their gratitude to Jeffrey Wardell for his statistical consultation in Study 2, and Elizabeth Wong and Rachel Bar for data collection assistance in Study 4. References Abraczinskas, M., Fisak, B., & Barnes, R. D. (2012). The relation between parental influence, body image, and eating behaviors in a nonclinical female sample. Body Image, 9, 93–100. http://dx.doi.org/10.1016/bodyim.2011.10.005 Allison, D. B., Kalinsky, L. B., & Gorman, B. S. (1992). A comparison of the psychometric properties of three measures of dietary restraint. Psychological Assessment, 4, 391–398. http://dx.doi.org/10.1037/1040-3590.4.3.391 Bailey, S. D., & Ricciardelli, L. A. (2010). Social comparisons, appearance related comments, contingent self-esteem and their relationships with body dissatisfaction and eating disturbance among women. Eating Behaviors, 11, 107–112. http://dx.doi.org/10.1016/j.eatbeh.2009.12.001 Britton, L. E., Martz, D. M., Bazzini, D. G., Curtin, L. A., & LeaShomb, A. (2006). Fat talk and self-presentation of body image: Is there a social norm for women to self-degrade. Body Image, 3, 247–254. http://dx.doi.org/10.1016/j.bodyim.2006.05.006 Brown, T. A. (2003). Confirmatory factor analysis of the Penn State Worry Questionnaire: Multiple factors of method effects? Behavior Research and Therapy, 41, 1411–1426. http://dx.doi.org/10.1016/s0005-7967(03)00059-7 Buhrmester, M., Kwang, T., & Gosling, S. D. (2011). Amazon’s Mechanical Turk: A new source of inexpensive, yet high-quality, data? Perspectives on Psychological Science, 6, 3–5. http://dx.doi.org/10.1177/1745691610393980 Cooper, P. J., Taylor, M. J., Cooper, Z., & Fairburn, C. G. (1987). The development and validation of the Body Shape Questionnaire. International Journal of Eating Disorders, 6, 485–494. http://dx.doi.org/10.1002/ 1098-108x(198707)6:4<485::AID-EAT2260069405>3.0.CO;2-O Crowne, D. P., & Marlowe, D. (1960). A new scale of social desirability independent of psychopathology. Journal of Consulting Psychology, 24, 349–354. http://dx.doi.org/10.1037/h0047358 Eisenberg, M. E., Berge, J. M., Fulkerson, J. A., & Neumark-Sztainer, D. (2012). Associations between hurtful weight-related comments by family and significant other and the development of disordered eating behaviors in young adults. Journal of Behavioral Medicine, 35, 500–508. http://dx.doi.org/10.1007/s10865-011-9378-9 Engeln, R., Sladek, M. R., & Waldron, H. (2013). Body talk among college men: Content, correlates, and effects. Body Image, 10, 300–308. http://dx.doi.org/10.1016/j.bodyim.2013.92.001 Engeln-Maddox, R., Salk, R. H., & Miller, S. A. (2012). Assessing women’s negative commentary on their own bodies: A psychometric investigation of the Negative Body Talk Scale. Psychology of Women Quarterly, 36, 162–178. http://dx.doi.org/10.1177/0361684312441593 Fabrigar, L. R., Wegener, D. T., MacCallum, R. C., & Strahan, E. J. (1999). Evaluating the use of exploratory factor analysis in psychological research. Psychological Methods, 4, 272–299. http://dx.doi.org/10.1037/1082-989X.4.3.272 Fulkerson, J. A., McGuire, M. T., Neumark-Sztainer, D., Story, M., French, S. A., & Perry, C. L. (2002). Weight-related attitudes and behaviors of adolescent boys and girls who are encouraged to diet by their mothers. International Journal of Obesity and Related Metabolic Disorders, 26, 1579–1587. http://dx.doi.org/10.1038/sj.ijo.0802157 Gardner, R. M., Brown, D. L., & Boice, R. (2012). Using Amazon’s Mechanical Turk website to measure accuracy of body size estimation and body dissatisfaction. Body Image, 9, 532–534. http://dx.doi.org/10.1016/j.bodyim.2012.06.006 Hart, E. A., Rejeski, W. J., & Leary, M. R. (1989). The measurement of social physique anxiety. Journal of Sport and Exercise Psychology, 11, 94–104. http://dx.doi.org/10.1037/t07809-000 Heatherton, T. F., Herman, C. P., Polivy, J., King, G. A., & McGree, S. T. (1988). The (mis)measurement of restraint: An analysis of conceptual and psychometric issues. Journal of Abnormal Psychology, 97, 19–28. http://dx.doi.org/10.1037/0021-843X.97.1.19 Helfert, S., & Warschburger, P. (2011). A prospective study on the impact of peer and parental pressure on body dissatisfaction in adolescence girls and boys. Body Image, 8, 101–109. http://dx.doi.org/10.1016/j.bodyim.2011.01.004 Herman, C. P., & Polivy, J. (1980). Restrained eating. In A. J. Stunkard (Ed.), Obesity (pp. 208–225). Philadelphia, PA: Saunders.

52

D.E. MacDonald et al. / Body Image 12 (2015) 44–52

Hu, L.-t., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1–55. http://dx.doi.org/10.1080/10705519909540118 Keery, H., Boutelle, K., van den Berg, P., & Thompson, J. K. (2005). The impact of appearance-related teasing by family members. Journal of Adolescent Health, 37, 120–127. http://dx.doi.org/10.1016/j.jadohealth.2004.08.004 Kenny, D. A., & Kashy, D. A. (1992). Analysis of the multitrait-multimethod matrix by confirmatory factor analysis. Psychological Bulletin, 112, 165–172. http://dx.doi.org/10.1037/0033-2909.112.1.165 Kluck, A. S. (2008). Family factors in the development of disordered eating: Integrating dynamic and behavioral explanations. Eating Behaviors, 9, 471–483. http://dx.doi.org/10.1016/j.eatbeh.2008.07.006 Kluck, A. S. (2010). Family influence on disordered eating: The role of body image dissatisfaction. Body Image, 7, 8–14. http://dx.doi.org/10.1016/ j.bodyim.2009.09.009 Kroon Van Diest, A. M., & Tylka, T. L. (2010). The Caregiver Eating Messages Scale: Development and psychometric investigation. Body Image, 7, 317–326. http://dx.doi.org/10.1016/j.bodyim.2010.06.002 MacDonald Clarke, P., Murnen, S. K., & Smolak, L. (2010). Development and psychometric evaluation of a quantitative measure of “fat talk”. Body Image, 7, 1–7. http://dx.doi.org/10.1016/j.bodyim.2009.09.006 Martz, D. M., Petroff, A. B., Curtin, L., & Bazzini, D. G. (2009). Gender differences in fat talk among American adults: Results from the psychology of size survey. Sex Roles, 61, 34–41. http://dx.doi.org/10.1007/s11199-009-9587-7 Mason, W., & Suri, S. (2012). Conducting behavioral research on Amazon’s Mechanical Turk. Behavioral Research Methods, 44, 1–23. http://dx.doi.org/10.3758/s13428-011-0124-6 McKinley, N. M., & Hyde, J. S. (1996). The Objectified Body Consciousness Scale: Development and validation. Psychology of Women Quarterly, 20, 181–215. http://dx.doi.org/10.1111/j.1471-6402.1996.tb00467.x Mundfrom, D. J., Shaw, D. G., & Ke, T. L. (2005). Minimum sample size recommendations for conducting factor analyses. International Journal of Testing, 5, 159–168. http://dx.doi.org/10.1207/s15327574ijt0502 4 Muthén, L. K., & Muthén, B. O. (2012). Mplus user’s guide (7th ed., pp. 197–200). Los Angeles, CA: Muthén & Muthén. Neumark-Sztainer, D., Bauer, K. W., Friend, S., Hannan, P. J., Story, M., & Berge, J. M. (2010). Family weight talk and dieting: How much do they matter for body dissatisfaction and disordered eating behaviors in adolescent girls? Journal of Adolescent Health, 47, 270–276. http://dx.doi.org/10.1016/j.jadohealth.2010.02.001 Neumark-Sztainer, D. R., Wall, M. M., Haines, J. I., Story, M. T., Sherwood, N. E., & van den Berg, P. A. (2007). Shared risk and protective factors for overweight and disordered eating in adolescents. American Journal of Preventive Medicine, 33, 359–369. http://dx.doi.org/10.1016/j.amepre.2007.07.031 Nichter, M. (2000). Fat talk: What girls and their parents say about dieting. Boston, MA: Harvard University Press.

Nichter, M., & Vuckovic, N. (1994). Fat talk: Body image among adolescent girls. In N. Sault (Ed.), Many mirrors: Body image and social relations (pp. 109–131). New Brunswick, NJ: Rutgers University Press. O’Connor, B. P. (2000). SPSS and SAS programs for determining the number of components using parallel analysis and Velicer’s MAP test. Behavior Research Methods, Instruments, & Computers, 32, 396–402. http://dx.doi.org/10.3758/BF03200807 Paolacci, G., & Chandler, J. (2014). Inside the turk: Understanding Mechanical Turk as a participant pool. Current Directions in Psychological Science, 23, 184–188. http://dx.doi.org/10.1177/0963721414531598 Payne, L. O., Martz, D. M., Tompkins, K. B., Petroff, A. B., & Farrow, C. V. (2011). Gender comparisons of fat talk in the United Kingdom and the United States. Sex Roles, 65, 557–565. http://dx.doi.org/10.1007/s11199-010-9881-4 Royal, S., MacDonald, D. E., & Dionne, M. M. (2013). Development and validation of the Fat Talk Questionnaire. Body Image, 10, 62–69. http://dx.doi.org/10.1016/j.bodyim.2012.10.003 Salk, R. H., & Engeln-Maddox, R. (2011). “If you’re fat, then I’m humongous!”: Frequency, content, and impact of fat talk among college women. Psychology of Women Quarterly, 35, 18–28. http://dx.doi.org/10.1177/0361684310384107 Satorra, A. (2000). Scaled and adjusted restricted tests in multi-sample analysis of moment structures. In R. D. H. Heijmans, D. S. G. Pollock & A. Satorra (Eds.), Innovations in multivariate statistical analysis (pp. 233–247). London: Kluwer Academic Publishers. Sharpe, H., Naumann, U., Treasure, J., & Schmidt, U. (2013). Is fat talking a causal risk factor for body dissatisfaction? A systematic review and meta-analysis. International Journal of Eating Disorders, 46, 643–652. http://dx.doi.org/10.1002/eat.22151 Smolak, L., & Levine, M. P. (2001). A two-year follow-up of a primary prevention program for negative body image and unhealthy weight regulation. Eating Disorders, 9, 313–325. http://dx.doi.org/10.1080/106402601753454886 Stevens, J. P. (2002). Applied multivariate statistics for the social sciences (4th ed.). Hillsdale, NJ: Erlbaum. Stice, E., Fisher, M., & Lowe, M. R. (2004). Are dietary restraint scales valid measures of acute dietary restriction?: Unobtrusive observational data suggest not. Psychological Assessment, 16, 51–59. http://dx.doi.org/10.1037/1040-3590.16.1.51 Stice, E., Maxfield, J., & Wells, T. (2003). Adverse effects of social pressure to be thin on young women: An experimental investigation of the effects of “fat talk”. International Journal of Eating Disorders, 34, 108–117. http://dx.doi.org/10.1002/eat.10171 Tabachnick, B. G., & Fidell, L. S. (2001). Using multivariate statistics. Needham Heights, MA: Allyn & Bacon. Tylka, T. L., Bergeron, D., & Schwartz, J. P. (2005). Development and psychometric evaluation of the Male Body Attitudes Scale (MBAS). Body Image, 2, 161–175. http://dx.doi.org/10.1016/j.bodyim.2005.03.001