Participants
In order to participate, students had to be at least 17 years of age. A total of 330 students (Mage= 18.620, SDage = 1.33) from a U.S. southeastern university completed the survey (Table 1). Students identified as white (69.4%), black (15.5%), Hispanic or Latinx (5.1%), Asian (4.4%) or other (4%). Most students were freshman (80.4%), female (60.4%), and identified as straight (91.2%). Height and weight were self-reported (MBMI = 23.8; Min = 16.64; Max = 41.6).
Procedure
All data was collected through an online survey using Qualtrics. Initially, participants were informed that the study was about fashion. Participants reported basic demographic information (age, year of college, race, gender, sexual orientation, and height and weight; Table 1) and questions about fashion. The questions about fashion were to maintain the impression that the study was about fashion, thereby hopefully reducing socially desirable responding.
Next, participants viewed 22 images of models and answered questions about each. Whether the participant saw a thin or plus size model was randomized per image. For example, all participants saw the same clothing for the first image, but approximately half saw a thin model and the other half saw a plus size model. Models wearing the same outfit could then be compared, ensuring that the actual clothing item was controlled for. Participants then filled out virtual measures, and then were debriefed on the study’s true purpose and asked to provide consent to use their data. The university’s institutional review board approved the study.
Measures
Model Rating Scale. A simple 3-item scale was constructed in order to assess participants’ attitudes towards the models (Table 2). There were also three questions that focused on the clothing rather than the model, in an effort to create the appearance that the scale was focused on fashion rather than just the model. Thus, these questions were not used for data analysis. The three questions of interest centered on perceptions of the model, using a 5-point Likert scale (1 = Strongly Disagree to 5 = Strongly Agree, with higher scores indicating more approval of the model). The scale has a minimum of 3 points and a maximum of 15 points. The scale had excellent internal consistency with a Chronbach’s alpha of .91, and each of the three questions individually had excellent internal consistency (Table 2).
Questions were designed based on forms of weight stigma established throughout the literature; for the first question, “is this model an appropriate choice for this dress,” past research has established that people of overweight or obesity status are more likely to be seen as less capable for a job position [2, 13]. Women who were overweight were seen as having a less professional appearance [14]. This stigma is apparent even when facial attractiveness is taken into account [15]. Therefore, the first question was designed to assess whether plus size models are seen as capable at modeling clothing in comparison to thin models. Participants were asked whether the model was an appropriate choice for this dress.
The second question, “does the model seem confident,” focused on trait-attribution, as negative traits are more likely to be assigned to people who are overweight or obese [3]. The second question asked whether the model seems confident, as individuals with higher adiposity are more likely to be seen as having lower self-regard [13]. The third question focused on physical attractiveness. In a study using undergraduate males, BMI was found to be a strong determinate for attractiveness [16]. Therefore, the third question, “does the model look attractive in this dress,” asked whether the model looks attractive.
Anti-Fat Attitudes Questionnaire (AFAQ) is a 13-item questionnaire developed to assess attitudes regarding fat [17], and has been found to be a robust way to measure weight bias [18]. Participants rated their level of agreement on a 10-point Likert scale (0 = very strongly disagree; 9 = very strongly agree), with higher scores indicating greater bias. The scale is divided into three subscales; the seven-item Dislike subscale measures overall bias of higher weight individuals. The three-item Fear of Fat subscale measures concern for gaining weight oneself. The three-item Willpower subscale measures the perception that higher weight can be controlled through motivation. A confirmatory factor analysis has supported the division of three scales [19]. Each subscale has established reliability and validity (Dislike α = .70, Fear of Fat α = .85, & Willpower α = .75). In the current investigation, each scale had good or excellent internal consistency (Dislike α = .93, Fear of Fat α = .89, & Willpower α = .86).
Sociocultural Attitudes Towards Appearance Questionnaire-4 (SATAQ-4) is a 22-item questionnaire that measures attitudes embodiment of social attitudes on appearance, with a subscale on both athletic ideas and thin ideals, with each subscale consisting of 5 items [20]. There are also three subscales focusing on a source of the pressure: family, peers, and media, each consisting of 4 items. All questions use a 5-point Likert scale, with 1 = Definitely Disagree to 5 = Definitely Agree, and higher scores symbolizing higher internalization or pressure. Each subscale had good validity and good reliability; Athletic (α = .87 - .92), Thin (α = .86 - .92), Family Pressure (α = .85 - .90), Peer Pressure (α = .88 - .90), and Media Pressure (α = .94 - .96) [20]. In the current investigation, each scale had good or excellent internal consistency (Athletic α = .91, Thin α = .81, Family Pressure α = .90, Peer Pressure α = .94, and Media Pressure α = .96).
Body Appreciation Scale (BAS) contains ten items that assesses participants’ positive feelings towards their body, using a 5-point Likert scale (from “never” to “always”) [21]. Higher scores indicate more body appreciation. The BAS has established reliability and excellent validity (α = .94) [21]. In the current study, the scale had excellent internal consistency (α = .95).
Statistical Analyses
Data was analyzed using SPSS 27. A paired-samples t-test was used to determine whether students rated plus size models differently in comparison to thin models, ratings for each image was summed and averaged across model types. Correlations were used to examine relationships among thin and plus size model ratings and weight bias, sociocultural attitudes toward weight, and body appreciation (Table 3). In order to examine variable predictions of plus size model ratings beyond participant thin rating patterns, hierarchal multiple regressions were run, with thin models in the first block and variables of interest in the second block. Each regression was calculated with thin model ratings in block 1, and a single AFA or SATAQ subscale in block 2, and plus size models ratings as the predictor (Table 4).
T-tests were also used to examine differences in model ratings based on gender, race, and BMI. Race was divided among groups that are more likely to endorse the thin ideal (European and Asian Americans) or not as likely (African Americans)[7, 12, 22]. Individuals who identified as Latinx were not included in this analysis, as data for this group in terms of idealization of the thin figure may be complex and depend on several factors, such as identification with ethnic background, and the current study did not control for these factors [22]. Hierarchal multiple regressions were also used to examine key participant demographics, both with thin model ratings in block 1, and either gender, race, or BMI (separately) at block 2, and plus size model ratings as the predictor (Table 5).
Additionally, mediational analyses were conducted. First, it was examined whether fat dislike possibly accounted for differences in ratings between thin- and plus-size models. Additionally, gender was included as a covariate to examine whether this pattern was observed across both males and females. Second, the relationship between gender and plus size model ratings was examined further to explore whether AFA Dislike and AFA Willpower both potentially mediated the relationship. Thin model ratings were included for these models as a covariate to account for overall response patterns. Bootstrapping was utilized for indirect effects across all mediational analyses as this technique is robust to nonnormal data [23].