Development and validation of a measure of curvy ideals internalization

Abstract

Several sociocultural female body ideals exist — thin, muscular/athletic, and, more recently, curvier ideals, which research specifically suggests are more prevalent among Black women. Two validated measures assess women’s desire for curvier bodies, but neither assess certain facets of curvy ideals (e.g., thick vs. slim-thick) separately. We developed and validated the Curvy Ideals Internalization (CII) Scale, to be used alone or alongside existing measures of appearance ideal internalization. Focus groups among racially/ethnically diverse women informed initial items. A sample of 897 White (37.1%), Black (34.2%), and biracial Black and White (28.7%) U.S. women completed the initial 37-item CII to determine factor structure, narrow the item pool, and examine validity and reliability. A separate sample (N = 366) of U.S. Black, White, and biracial women completed the CII to confirm the factor structure. The final CII has eleven items, with factors assessing thick/curvy ideal internalization and facets of slim-thick ideal internalization: thin waist and large breast size. The CII has adequate internal consistency, test-retest reliability, construct validity, and factorial validity. The CII is appropriate for use among Black, White, and biracial women to assess internalization of curvier body ideals and needs to be validated in more diverse samples.

1. Introduction

Etiological theories regarding the onset and maintenance of body dissatisfaction and disordered eating — the dual pathway model and the tripartite influence model — identify internalization of unrealistic body ideals as a key risk factor for the development of eating pathology (Keery et al., 2004; Schaefer et al., 2017; Stice & Bearman, 2001; Stice et al., 2002; Thompson et al., 1999). More specifically, the dual pathway model posits that pressures to achieve unrealistic cultural appearance ideals in combination with appearance ideal internalization (i.e., acceptance of the cultural beauty ideal as one’s own personal standard of beauty) increase an individual’s risk for body dissatisfaction (Stice & Van Ryzin, 2019). Body dissatisfaction, in turn, increases risk for dietary restriction and negative affect, which subsequently increase eating disorder risk (Stice & Van Ryzin, 2019). The tripartite influence model specifies that sociocultural pressures to achieve unrealistic beauty ideals come from peers, parents, significant others, and the media (Keery et al., 2004; Schaefer et al., 2017). These sociocultural pressures become problematic when they are internalized, ultimately increasing risk for body dissatisfaction and disordered eating (Keery et al., 2004; Stice & Bearman, 2001; Stice et al., 2002; Stice & Van Ryzin, 2019; Thompson et al., 1999).

In part due to stereotypes regarding the demographics of eating disorders, which presume that eating pathology is primarily experienced by White women (Sonneville & Lipson, 2018), early investigations of appearance ideal internalization largely focused on the thin ideal. The thin ideal is characterized by a slim physique with a slender waist and little body fat and became prominent among White women in Western mainstream media during the 1990s. Indeed, the two most widely used measures of appearance ideal internalization, the Sociocultural Attitudes Towards Appearance Questionnaire (Heinberg et al., 1995; Thompson et al., 2004) and the Ideal Body Stereotypes Scale (Stice et al., 2008), were each developed to exclusively assess constructs relevant to the thin ideal. In the last 30 years, research on the thin ideal has proliferated, with numerous investigations of the dual pathway model and tripartite influence model providing empirical support for the role of thin ideal internalization in the development of body image disturbance and eating pathology (Thompson & Stice, 2001).

However, evidence also suggests the importance of considering internalization of alternative appearance ideals as potential risk or protective factors for body image and eating disturbance. For example, research examining internalization of the muscular or athletic ideal suggests that this body ideal is highly relevant to both men and women and is associated with body dissatisfaction and disordered eating across genders (Schaefer et al., 2015, 2017; Stefanile et al., 2019). Further, existing research suggests that Black women, who are underrepresented in research on body image and eating disturbance, may prefer body types not adequately captured within traditional measures of appearance ideal internalization (Hernández et al., 2021; Hunter et al., 2021; Overstreet et al., 2010). Therefore, it is critical that research identifies and assesses diverse appearance ideals that may have deleterious or protective effects on body image and eating behaviors.

1.1. Body image ideals among Black women

On average, Black women generally endorse higher levels of body satisfaction and lower levels of restrictive eating compared to their Asian, White, and Latina peers (Burke, et al., 2021; Rodgers et al., 2017), leading some researchers to suggest that cultural differences in appearance ideal internalization may contribute to this reduced risk (Frederick et al., 2022; Rakhkovskaya & Warren, 2014; 2016). Extant research suggests that Black women may value different beauty ideals than those historically promoted by mainstream media or presented within traditional measures of appearance ideal internalization (Chin Evans & McConnell, 2003; Guan et al., 2012; Warren et al., 2005). Indeed, Black women endorse lower levels of thin ideal internalization than White women, and thin ideal internalization is less predictive of negative outcomes (e.g., body image disturbance) in Black women than White, Latina, and Asian women (Burke et al., 2021; Rakhkovskaya & Warren, 2014). In contrast with the traditional thin ideal, research suggests that many Black women prefer curvier bodies than White women (Overstreet et al., 2010; Petersons et al., 2007; Webb et al., 2013). These curvier bodies may be “thick” – primarily denoting a voluptuous lower body with ample buttocks, hips, and thighs (Gentles-Peart, 2018; Hughes, 2018), or “slim-thick”¹– denoting an hourglass-shaped body possessing large breasts, hips, and buttocks, but a slim waist (Wilfred & Lundgren, 2021). Consistent with this preference, magazines with a strong Black readership frequently include larger models than magazines with predominately White readers (Shoneye et al., 2011). Curvier body types have also gained in popularity in mainstream media in the past decade, glamorized by predominantly Black or biracial celebrities like Nicki Minaj, Beyoncé, the Kardashians, and Lizzo (Hunter et al., 2021).

While curvy ideals may be especially relevant for Black women; research suggests that curvy ideals may also be prevalent among other communities of color. For example, prior research has found that Latina women prefer curvier and larger bodies than do White women (Franko et al., 2013; Hernández et al., 2021; Viladrich et al., 2009). Additionally, Indian women may prefer curvier figures than the thin ideal typically portrayed in Western media (Raman, 2015). Additionally, in a qualitative study of Native American women’s body image, when asked what celebrity had the ideal body, many of the young women selected celebrities with curvier figures (Naegele & Cook, 2017). Overall, research examining internalization of curvier and larger body ideals among women of color is limited, especially outside of Black and Latina populations. More research is needed to understand whether curvy body ideals are particularly prevalent among Black, Indigenous, and women of color.

However, it is unknown whether internalization of a curvier body ideal is protective or problematic for Black women, or women more broadly (Hernández et al., 2021; Hughes, 2018). In one respect, curvy ideals may result in relatively decreased risk for body dissatisfaction or disordered eating via a reduced emphasis on extreme thinness or low overall body fat. Alternatively, it is possible that ideals emphasizing rare combinations of features (e.g., large breasts and buttocks combined with a thin waist) may be unattainable for many women, potentially promoting body dissatisfaction and problematic weight/shape control behaviors aimed at achieving this alternative ideal (e.g., combinations of dieting, exercise, use of waist trainers, cosmetic surgery, etc.).

1.2. Existing curvy ideal measures

Despite increasing attention on curvy beauty ideals among women, very few peer-reviewed articles have been published on the topic, potentially due to an absence of validated measures. Very recently, attempts to address this assessment gap resulted in two measures being developed. Hunter et al. (2021) developed the Curvy Ideal Silhouette Scale (CISS), a matrix comprised of 25 computer-generated images of female bodies that systematically vary on dimensions of curviness (i.e., waist-to-hip ratio) and thinness (i.e., body fat percentage). Respondents indicate the images that best represent their actual and ideal bodies. Discrepancy scores between the actual and ideal images yield a curviness discrepancy index (CDI) and fat discrepancy index (FDI). Although the CDI was significantly positively correlated with the FDI, and negatively correlated with body image disturbance, body dissatisfaction, and disordered eating among a sample of U.S. women, discrepancy scores (i.e., the difference between one’s actual and ideal body) may be best conceptualized as a measure of curviness or thinness dissatisfaction, rather than a pure measure of internalization. Although participant selections of ideal bodies reflecting a curvy ideal on the CISS may be conceptualized as a measure of curvy ideal internalization, this approach has not yet been validated for use with the CISS, and therefore cannot currently be recommended.

Since then, Hernández et al. (2021) published the Hourglass Body Shape Ideal Scale (HBSIS), a six-item measure assessing women’s desires for an hourglass body shape. The initial 27-item pool for the measure was first developed with a focus group of racially and ethnically diverse undergraduates (Hernández et al., 2021). Notably, the researchers and focus group members reduced the initial item set to ten items prior to collecting validation data in a large, diverse sample of women. Psychometricians typically recommend generation and testing of a large pool of potential items, then using statistical and theoretical guidance to determine the final items to include in a questionnaire (Haladyna, 2004). Thus, it is possible that the researchers may have prematurely constrained the item pool to fit with preconceived notions about the construct, potentially eliminating pertinent items in the process. Further, although internalization has been conceptualized as a cognitive process (Kelman, 2006; McCauley, 1989; Schaefer et al., 2017), the final HBSIS includes behavioral manifestations of the desire and drive for a curvy appearance (e.g., “It is important for me to emphasize my breasts when I dress”) that emphasize clothing choices rather than actual body types and may be less likely to translate into disordered eating psychopathology.

Importantly, both the CISS and HBSIS measure internalization of a “slim-thick” figure, in which a larger bust, hips, and buttocks are paired with a slim waist (Wilfred & Lundgren, 2021). They do not assess whether women have internalized a “thick” body ideal, which research suggests differs most notably in being bigger in all body areas without the need for a small waist (Gentles-Peart, 2018; Hughes, 2018). No existing measures assess internalization of a thick body ideal. However, initial qualitative research supports the existence of the thick body ideal (Gentles-Peart, 2018; Hughes, 2018) and its distinction from a slim-thick body ideal. Due to the lack of a thick body ideal measure, there is no quantitative research to determine whether women differ on the degree to which they internalize these distinct curvy body ideals (e.g., thick vs slim-thick), what the correlates and consequences of internalizing these differing body ideals may be, and whether there are racial differences in internalizing a thick versus a slim-thick body ideal.

1.3. Current study

The purpose of the current study was to develop and validate a new measure of curvy body ideal internalization, the Curvy Ideals Internalization (CII) scale. As the curvy body ideal may be most pertinent to Black women, a key goal of the current study was to ensure adequate input from this group to help address calls for more culturally sensitive measures (Burke et al., 2017; Burnette et al., 2020; Kempa & Jones Thomas, 2000; Lowy et al., 2021; Rodgers et al., 2018). A valid measure of Black women’s body ideal internalization is necessary to better understand cultural appearance ideals and related pressures to achieve those body ideals among this group. Further, as biracial/multiracial individuals represent one of the fastest growing groups within the U.S. Census (2020) but have received very limited attention in the body image and eating disorders literature (Burke, et al., 2021), examination of how such a measure operates within biracial Black women was an additional goal of the current investigation.

Given the focus of this study on women of color, we specifically endeavored to include individuals of color including Black women, within the study team. The authors included one biracial Black and White woman, one Black Latina woman, three White women, and one White man. The two co-authors who led the focus group identified as Black Latina and White non-Hispanic/Latina. The research assistants who reviewed and commented on the initial CII item pool and reviewed the surveys prior to their administration included seven women (three Black, one Asian, one biracial Black/Asian, and two White) and two White men. All members of the research team, unless otherwise noted, did not identify as Hispanic/Latinx. This research idea was originally spurred by a collaboration between the first author, who specializes in body image and eating disorders, and the second author, who specializes in diverse social identities and intergroup contact.

2. Study 1: Curvy Ideals Internalization Scale development

The purpose of Study 1 was to develop a large item pool to describe body ideal internalization among racially and ethnically diverse women.

2.1. Methods

2.1.1. Participants

Ten undergraduate women, nine of whom identified as women of color², were specifically recruited from organizations with student of color representation at a small liberal arts college in the Northeast U.S. to take part in two focus groups exploring diverse women’s perspectives on beauty ideals. Emails were sent to these groups requesting participants join a 30–45-minute Zoom discussion on diverse contemporary beauty ideals. Participants were not compensated for their participation.

2.1.2. Procedure

Following consent procedures, focus group facilitators asked open-ended questions to generate discussion of prominent appearance ideals, including appearance ideals centered around different curvy body ideals. Groups were conducted on Zoom due to COVID-19 and were recorded and transcribed, removing participant identification. Researchers opened the group with the following script to generate conversation:

Today we will be discussing various beauty ideals. In general, body image research typically focuses around the thin or athleticthin beauty ideals, and we are looking for more information on other beauty ideals that exist such as curvy, thick, and slim-thick. We want you all to feel comfortable in this space and to discuss things freely. We hope to speak as little as possible and are just here to guide the conversation when needed.

Following the initial prompt, researchers asked open-ended questions, asked participants to describe beauty ideals portrayed in the media and from other sources in their life (e.g., family, community settings, peers, etc.), whether certain body ideals were more valued than others depending on one’s race or ethnicity, and how they would define thick, curvy, and slim-thick.

2.2. Results

Focus group participants described both a thick body ideal and a slim-thick body ideal as being separate body ideals that were pertinent to women of color. Most of the focus group participants used slim-thick as their primary terminology for hourglass-shaped or curvy bodies. Four major themes were addressed in the focus groups, guided primarily by researchers’ questions and by the flow of participants’ responses. Sample quotations are provided for each theme in Table 1. First, participants described the thick body ideal (Theme 1), which emphasized thick arms, thick thighs, and a thicker waist. In contrast, participants equated curvy, slim-thick, and hourglass, and explained that this body ideal was largely described by a thin waist and large breasts (Theme 2). A third theme described body ideals present among women of color, noting that Black and Latinx communities were accepting of larger and curvier bodies, and did not idealize a thin body ideal. Lastly, (Theme 4) the focus group participants discussed the context-dependent nature of their own body image ideals, which shifted to thinner and slim-thick bodies when they were immersed in White-dominated spaces, such as their predominantly White college environment. Women also noted the prevalence of diet and exercise regimes on social media that purported to achieve a slim-thick body ideal.

Table 1.

Example Quotations from Focus Group Members (N = 10).

Body Ideal Theme Discussed	Descriptions
Thick Body Ideal	• [The] thick [body ideal] has some arms, some tummy, some thighs... you’re just a little thicker everywhere. Something more realistic, I guess. • It’s really centered around like having a really big butt and I would hope thighs that match but that’s not always required these days. I guess boobs are included now too but I feel like even if you are a little smaller on top, as long as you have a pretty big butt and thighs you are considered thick. • Whenever I think of thick I think of like they have like a bit of meat in all parts of their bodies, like they have like you know, a little bit of a tummy, pretty thick thighs, a big butt • If you’re thick you’re just a little thicker everywhere, you have bigger arms. • Thick honestly is more associated with ... butt and thighs, because you can be thick and have no boobs.
Slim-Thick/Curvy/Hourglass Body Ideal	• I feel that something that really contributed to that trend [the slim-thick body ideal]... [was] Kim Kardashian. It was only after her arrival and her fame that the beauty ideal of the hourglass, but like a really thick, slim-thick hourglass, really became popular. • The Kim K body type is different from the thick body type • To me slim-thick has the flat stomach • I would say that like what’s been glorified at least on my timeline is like a big butt, a very snatched [slim] waist, um and ya that’s pretty much it. You know basically like an hourglass figure or a pear-shaped figure. • I feel like I see a lot of athletic slim-thick so like a large butt, but abs and a thin waist and muscly arms, but not too muscly cause then people think you look like a guy, but kind of like that. • Curvy is ... having ... boobs, Is ...the thing ... you have to have, that’s how hourglass figures form. • What you look [like] on top ... differentiates curvy from thick a lot of the time.
Body ideals in communities of color	• BIPOC people have been pushed in the direction toward Kim Kardashian by media, but if you compare it to what it was in the late 90’s [and] early 2000’s... it was more acceptable to have a little bit of a stomach as long as you had a big butt, big boobs... but the shape itself looked a lot different than what we see now. • Being a Latina, growing up ...is inclusive in that way [in being accepting of larger bodies]. But also like if you are too skinny ... everyone’s worried and your parents are like “oh my god you need to eat more... it’s very like frowned upon. But you know everyone embraces the other side. But then there’s also that part where it’s like “oh, you’re too big” or whatever. • Whereas the Black community is like if you are like skinny, that’s not idealized. It’s more so, slim-thick, thick and even I feel like I don’t know I feel like the Back community is more open to you know plus size people than the White community • Being in a predominantly Black community, I don’t really concern myself with being larger.
Context-dependent body ideals among women of color	• I feel like the culture here [at a predominantly White college] is really thin girls in like Lululemon leggings all the time, working out all the time. And it makes you feel really uncomfortable sometimes. I definitely am more conscious of my body here than I am at school.then I am at home [in the Black community]. • When I was home, a lot of the kids I went to school with were Hispanic or Black. And like it was accepted, ... your body type. But here [at a predominantly White college]... a lot of the girls are thinner, so... I feel a little more uncomfortable with my body here than I do at home. • I went to a ... predominantly Black ... high school, and I was really skinny at the time.... I was kinda just teased a lot for being so skinny... then I gained weight towards the end of high school and I was okay, that was kinda cool now. ... but then I got to college, and I was still gaining weight and it was really weird being next to people that were so skinny. ... Especially if you want to participate in ...hookup culture, and you do not necessarily fit ... the skinny ideal... and we do not have that many men of color on campus, so. You kinda feel this pressure to... These pressures from high school still exist at the same time, alongside the ideal of wanting to be super skinny to ... keep up with the White girls. So it’s ... weird... and ... switches out when it’s not convenient for me.

Note. BIPOC = Black, Indigenous, and people of color

2.3. Study 1 discussion

Two focus groups with women of color discussed diverse body ideals including describing curvy, thick, and slim-thick body ideals. The focus group participants identified that thick differed from slim-thick, which they equated with curvy and hourglass ideals. In particular, the women identified that the slim-thick body ideal had a slim waist, whereas a thick body was generally bigger everywhere and did not need to have a slim waist. Additionally, most women reported that the slim-thick body ideal also had large breasts. The women noted that both body ideals were more common among women of color, whereas a thin or athletic ideal was more common among White women. The body features that focus group participants described as thick, slim-thick, curvy, and hourglass informed CII item development.

3. Study 2: Curvy Ideals Internalization Scale initial validation

Based on focus group responses, researchers (3^rd and 4^th authors) developed a 37-item pool using the SATAQ-4 internalization item structure to capture participants’ descriptions of the curvy/thick ideal. The purpose of Study 2 was to identify a subset of items from the initial item pool that performed well, to identify the factor structure of the CII using EFA, and to assess the CII’s construct validity, internal consistency, and test-retest reliability. The CII’s construct validity was assessed by examining correlations between the CII and well-validated measures of appearance orientation, appearance satisfaction, perceived sociocultural body image pressures, body appreciation, and internalization of thin and muscular body ideals.

3.1. Construct validity

Most of the scales used to assess construct validity were selected as part of a larger study on Black and (Black/White) biracial women’s body image, and therefore do not represent an exhaustive selection of potential measures that would provide evidence of convergent and discriminant validity. Additionally, we were mindful to limit survey length to minimize participant burden and maximize data quality (Kost & Correa da Rosa, 2018; Liu & Wronski, 2017; Toepoel & Lugtig, 2018). We hypothesized that thick and curvy ideal internalization (assessed by the CII) would be significantly correlated with the discrepancy between one’s current and ideal level of curviness, and that the correlation between the CII and the discrepancy between women’s current and ideal curviness would be larger than the correlation between the CII and the discrepancy between their current and ideal adiposity. We also hypothesized a significant positive association between the CII and appearance orientation and thin and muscularity ideal internalization. Lastly, we hypothesized that the CII would significantly positively correlate with measures of parental influence on dieting behaviors and pressure from the media, family, peers, and partners to achieve a slim body ideal. The relationship between the CII and body appreciation and appearance satisfaction were considered exploratory.

3.2. Methods

3.2.1. Participants

The recommendation for EFA sample size is a minimum of ten participants per item (Tabachnik & Fidell, 2007). To yield a large and diverse sample, the study was conducted online via the recruitment platform CloudResearch, using Prime Panels, as part of a larger body image study on Black and (Black/White) biracial women’s body image.³ A total of 2,486 participants followed the CloudResearch link to begin the study on SurveyMonkey. Of those, 176 did not consent, 42 did not fill out any additional items after consenting, and 472 failed the bot check question, yielding 1,796 participants. Inclusion criteria were: monoracial Black, monoracial White, or biracial Black and White racial identity, U.S. resident, correctly answering at least two attention check questions, and non-male gender identity. A total of 117 participants with other racial/ethnic identifications, and 758 participants who consented but did not complete the CII, were removed, yielding 921 participants. We included cisgender women (n = 836, 93.2%), transgender women (n = 14, 1.6%), and those who identified as gender nonbinary (n = 22, 2.5%) or agender (n = 25, 2.8%), as the CloudResearch sample all identified as women prior to recruitment, and we aimed to take an inclusive approach. Additionally, research has found similar levels of appearance comparisons, body monitoring, thin ideal internalization (Strübel et al., 2020) and body dissatisfaction (Wilcheck-Aviad et al., 2020) among cisgender and transgender women, and both objectification theory (Brewster et al., 2019; Comiskey et al., 2020; Strübel et al., 2020) and the tripartite influence model (Strübel et al., 2020) have been supported among transgender women, suggesting similarities across cis and transgender women’s body ideals and body dissatisfaction. Removal of six transgender men and ten cisgender men yielded a sample of 905. Finally, removal of eight duplicate responses yielded a final sample of 897 White (n = 315, 37.1%), Black (n = 291, 34.2%), and biracial Black and White (n = 244, 28.7%) women. Age ranged from 18–87, M(SD)_Age = 39.56 (15.52).

Participants were contacted approximately 2–3 weeks after completing the initial survey to complete a second round of surveys containing the CII to assess the CII’s test-retest reliability. Participants completed informed consent and were compensated for their participation. Of the 428 participants who began the second survey, 201 were removed for the following reasons: lack of consent, lack of responses following consent, duplicate responses, inability to match participants, and male gender identity. This left 227 participants (34.8% White; 33.5% Black; 31.7% biracial) who completed both the first and second survey administration. Participant demographics are reported in Table 2.

Table 2.

Demographic characteristics for Studies 2–4.

Demographic		Study 2 N = 897	Study 2 Test-retest Sample n = 227	Study 3 N = 366
Age M(SD) - years		39.56 (15.52)	45.27 (17.20)	37.06 (11.84)
BMI M(SD) - kg/m2		–	28.40 (7.95)	31.18 (8.73)
		n (%)	n (%)	n (%)
Race	White	315 (37.1)	79 (34.8)	226 (61.7)
	Black/African-American	291 (34.2)	76 (33.5)	94 (25.7)
	Biracial (Black & White)	244 (28.7)	72 (31.7)	46 (12.6)
Gender Identity	Cisgender woman	796 (8.45)	223 (98.2)	360 (98.4)
	Transgender woman	13 (1.5)	1 (0.4)	6 (1.6)
	Gender nonbinary	21 (2.5)	3 (1.3)	0 (0.0)
	Agender	20 (2.4)	0 (0.0)	0 (0.0)
Sexuality	Heterosexual	718 (84.5)	199 (87.7)	295 (80.6)
	Bisexual	94 (11.1)	19 (8.4)	55 (15.0)
	Lesbian	32 (3.8)	7 (3.1)	10 (2.7)
	Asexual	6 (0.7)	2 (0.9)	6 (1.6)
Education Level	Less than HS degree	30 (3.5)	8 (3.5)	2 (0.5)
	HS degree or equivalent	196 (23.1)	52 (22.9)	36 (9.8)
	Some college	229 (26.9)	62 (27.3)	61 (16.7)
	Associate’s degree	118 (13.9)	46 (20.3)	44 (12.0)
	Bachelor’s degree	168 (19.8)	41 (18.1)	156 (42.6)
	Graduate degree	109 (12.8)	18 (7.9)	67 (18.3)
Income	$0–24,999	229 (26.9)	72 (31.7)	42 (11.5)
	$25,000–49,999	244 (28.7)	67 (29.5)	123 (33.6)
	$50,000–74,999	158 (18.6)	47 (20.7)	82 (22.4)
	$75,000–99,999	86 (10.1)	20 (8.8)	71 (19.4)
	$100,000–149,999	78 (9.2)	16 (7.0)	33 (9.0)
	> $150,000	55 (6.5)	5 (2.2)	15 (4.1)

Note. HS = High School; BMI = Body Mass Index

3.2.2. Measures

3.2.2.1. Thin and muscular ideal internalization.

The SATAQ-4 (Schaefer et al., 2015) is a well-validated measure of internalization of thin and muscular body ideals and perceived pressures from media, family members, and peers to conform to those body ideals. Items are rated on a 5-point scale from 1 (Definitely Disagree) to 5 (Definitely Agree). The SATAQ-4 includes five subscales, two of which were included in the current study: Thin Ideal Internalization (e.g., “I want my body to look like it has little fat.”) and Muscular/Athletic Ideal Internalization (e.g., “I spend a lot of time doing things to look more muscular”). Subscale items are averaged, with higher scores representing greater internalization of the thin or muscular/athletic body ideal, respectively. Convergent and discriminant validity, internal consistency, and test-retest reliability have been established among primarily White undergraduate female samples (Schaefer et al., 2015). An examination of measurement invariance among White and Black women found that the original factor structure was an adequate fit in both samples (Burnette et al., 2020). Additionally, Burnette et al. (2020) reported similar and adequate Cronbach’s α for Black and White women on the SATAQ-4, with Black women scoring significantly lower than White women on thin ideal internalization, muscular/athletic ideal internalization, and media pressures. In the current study, McDonald’s ω =.894 and Cronbach’s α =.892 for the Thin Ideal Internalization subscale, and ω =.795 and α =.784, for the Muscular/Athletic Ideal Internalization subscale.

3.2.2.2. Perceived sociocultural pressures.

The Perceived Sociocultural Pressure Scale (PSPS; Stice & Bearman, 2001) assesses pressure to be thin from family, peers, the media, and dating partners (e.g., “Family members tease me about my weight.”). The 10-item Likert-type scale is rated from 1 (None) to 5 (A Lot); items are averaged for each subscale. The scale has demonstrated predictive validity, internal consistency, and test-retest reliability (Stice et al., 2002). Prior research has supported validity and internal consistency of the PSPS in Black women (Rogers Wood & Petrie, 2010) In the current study, McDonald’s ω and Cronbach’s α both were.881 for the Family Pressure subscale, ω =.826 and α =.818 for the Peer Pressure subscale, α =.871 for the Partner Pressure subscale, and α =.858 for the Media Pressure subscale (McDonald’s ω requires a three-item minimum).

3.2.2.3. Multidimensional Body-Self Relations Questionnaire.

(MBSRQ; Cash, 2000) The MBSRQ is a 69-item measure with ten subscales, three of which were used in the current study. Appearance Orientation assesses investment in appearance (e.g., “Before going out, I usually spend a lot of time getting ready.”), Appearance Evaluation assesses satisfaction with overall appearance (e.g., “I like my looks just the way they are.”), and Body Areas Satisfaction assesses satisfaction with various body parts, including face, hair, muscle tone, height, and weight. The MBSRQ is scored on a Likert-type scale, with the Appearance Orientation and Appearance Evaluation subscales rated from 1 (Definitely Disagree) to 5 (Definitely Agree) and Body Areas Satisfaction subscale items rated from 1 (Very Dissatisfied) to 5 (Very Satisfied). Items are averaged after reverse scoring. The MBSRQ demonstrated adequate internal consistency (Cronbach’s α ranging from.70-.91) and 1-month test-retest reliability in initial validation samples (r =.71-.94; Cash, 2000). Prior research has supported the validity and internal consistency among Black women for these three MBSRQ subscales (Falconer & Neville, 2000; Smith et al., 1999). In the current study, Appearance Orientation ω =.851 and α =.856, Appearance Evaluation ω =.870 and α =.869, and Body Areas Satisfaction ω =.870 and α =.892.

3.2.2.4. Body appreciation.

The Body Appreciation Scale-2 (BAS-2; Tylka & Wood-Barcalow, 2015) is a 10-item measure assessing body appreciation (e.g., “I take a positive attitude towards my body). Participants rate Likert-type items from 1 (Never) to 5 (Always). Items are averaged; higher scores indicate greater body appreciation. The BAS-2 had adequate internal consistency, test-retest reliability, convergent validity, and discriminant validity in female samples, and CFA supported a unidimensional structure (Tylka & Wood-Barcalow, 2015). The BAS-2 has demonstrated adequate validity and internal consistency among Black women (Boutté et al., 2022), as has the original BAS (Cotter et al., 2015). In the current study, ω =.960 and α =.959.

3.2.2.5. Parental influence.

The Parental Influence Questionnaire (PIQ; Abraczinskas et al., 2012) is a 16-item measure developed to assess parental influences on dieting and weight and shape concern, rated on a 5-point Likert scale from 1 (Strongly Disagree) to 5 (Strongly Agree). The PIQ has two subscales, Direct Influence (e.g., “My parents watched closely what I ate”), and Modeling (e.g., “My parents complained about their weight”). The PIQ factor structure, convergent and discriminant validity, and internal consistency were supported in the validation sample (Abraczinskas et al., 2012). In the current study, ω =.929 and α =.924 for the Direct Modeling subscale and ω =.923 and α =.922 for the Modeling subscale.

3.2.2.6. Curvy ideal figural rating scale.

At the time of data collection, the CISS (Hunter et al., 2021) was the only scale assessing a desire for a curvy figure. The CISS is a computer-generated visual 5 ×5 grid of women’s bodies assessing the discrepancy between women’s actual and ideal body fat and curviness, each on a 5-point scale. The CISS curviness discrepancy index (CDI), calculated by current - ideal curviness, was significantly correlated with measures of body image disturbance, body dissatisfaction, and disordered eating, and was correlated with the fat discrepancy index (FDI), calculated by current - ideal body fat (Hunter et al., 2021). A positive CDI score reflects a desire for a curvier figure, whereas a negative CDI score reflects a desire for fewer curves. Similarly, a positive FDI score reflects a desire for more body fat and a negative FDI score reflects a desire for less body fat. Of note, the FDI was not correlated with measures of family pressure to be thin, body dissatisfaction, or disordered eating, and the CDI was not significantly correlated with family pressure to be larger in the validation sample, of which 41% were Black women (Hunter et al., 2021), suggesting potential concerns with the construct validity of the CISS. In the validation sample, the CISS had adequate 2-week test-retest reliability (Hunter et al., 2021).

3.2.3. Procedure

The study was determined to be exempt from IRB review by the ethical review board as per 45 CFR 46.104(d)(2). Participants were recruited by CloudResearch to complete a survey entitled “Women’s Mental Health and Diversity” on SurveyMonkey and were paid upon survey completion. After providing informed consent, participants provided demographic information, and completed the CII and other measures.

3.2.4. Data analysis

3.2.4.1. Factor structure.

An EFA with principal axis factoring and oblique Promax rotation was used to examine the factor structure of the initial CII item pool. Oblique Promax rotation was selected because it is expected that underlying factors would be correlated (Tabachnick & Fidell, 2007). Bartlett’s test of sphericity and the Kaiser-Meyer-Olkin measure of sampling adequacy assessed the factorability of the initial CII item pool, with a statistically significant Bartlett’s test of sphericity and Kaiser-Meyer-Olkin measure ≥ 0.60 indicating factorability (Tabachnick & Fidell, 2007). Prior to conducting EFA, multivariate outliers were identified and removed from analysis and items were analyzed for skewness and kurtosis, with transformations applied to any non-normal items (Tabachnick & Fidell, 2007). Items were considered acceptable with skewness statistics < |1| and kurtosis < |2| (Tabachnick & Fidell, 2007).

The number of factors selected to be retained in the CII was determined using both theoretical and statistical guidance (Tabachnick & Fidell, 2007). Statistical guidance included parallel analysis, visual examination of the scree plot, and the Kaiser-Guttman criterion of retaining factors with eigenvalues > 1.0 (Guttman, 1954; Kaiser, 1960; Tabachnick & Fidell, 2007; Watkins, 2006), which provide a maximum number of acceptable factors. Theoretical guidance was used to reduce factors, if necessary, such that factor structures aligned with hypothesized constructs and resulted in meaningful factors, in which items were conceptually related (Tabachnick & Fidell, 2007).

Item analysis was conducted to select a final item pool that best fit the factors. Items with low inter-item correlations (rs ≤.30), low primary factor loadings (≤.40), or with a secondary factor loading of ≥.30 were also removed (Tabachnick & Fidell, 2007). Finally, items suggesting a personal desire for the body ideal (e.g., “I want a big butt”) were retained over similar items reflecting awareness of the body ideal (e.g., “Big butts are attractive.”) when items had similar factor loadings. Similarly, items reflecting cognitive rather than behavioral features were selected for similar items, to retain consistency with the SATAQ-4R and current conceptualizations of body ideal internalization as a cognitive construct (Schaefer et al., 2017) when items had similar content and factor loadings. Inter-item correlations > .80 were considered to measure an overlapping facet of CII, so one of any such item pairs was removed to yield a final, more parsimonious item set.

3.2.4.2. Reliability.

The CII’s internal consistency was assessed using McDonald’s ω and Cronbach’s α, with values above.70 considered acceptable (Hayes & Coutts, 2020).

3.2.4.3. Construct validity.

The CII’s convergent validity was examined using Pearson product-moment correlations between the CII and measures of appearance satisfaction (MBSRQ-AE, MBSRQ-BAS), appearance orientation (MBSRQ-AO), internalization of thin and muscular/athletic appearance ideals (SATAQ-4R), parental influences on dieting (PIQ), sociocultural pressures for thinness (SPSP), body appreciation (BAS-2), and the CISS curvy and body fat discrepancies (CISS-CDI, CISS-FDI). Fisher’s r to Z transformations were used to assess whether the hypothesized differences in strength between correlations were statistically significant.

3.2.4.4. Test-retest reliability.

Intraclass correlation coefficients (ICCs) were used to calculate the test-retest reliability of the CII Global and subscale scores at the first and second survey administration; ICC > 0.70 indicates high test-retest reliability (Cohen, 1988).

3.3. Results

3.3.1. Item development

The original item pool was developed by using focus group participants’ descriptions of physical features associated with thick, curvy, hourglass, and slim-thick, and applied those features to the structure of the Sociocultural Attitudes Towards Appearance Questionnaire-4 (SATAQ-4; see Schaefer et al., 2015; 2017). For example, a focus group participant discussed a slim-thick ideal as having “bigger breasts ... a small waist and then bigger hips” and slim thighs. Research assistants incorporated those descriptions into questionnaire items, such as “I want to have a flat stomach with a thin waist and big breasts,” and “Big butts are attractive.” The original item pool included items assessing beauty ideals described during the focus groups. These included a desire to look thick, a desire to look curvy, a desire to have a big butt, big thighs, a slim waist, a flat stomach, thin arms, medium or small breasts, and large breasts. Some items assessed combinations of these features (e.g., “It is important for me to have a flat stomach with a big butt and big thighs”). Other items assessed only single features (e.g., “I want big breasts”) to avoid potential problems associated with double-barreled items. Following initial item development, items were reviewed and edited by three body image experts (1^st, 4^th, and 5^th author) to ensure that items conceptually fit with the construct of body ideal internalization. Items were also reviewed by a social and racial identity expert (2^nd author) to ensure that they were appropriate for women with diverse racial and social identifications. The current measure was developed to match the item structure, scaling, scoring, and interpretation of the SATAQ-4R. This allows for comparison of internalization scores across measures of thin, muscular/athletic, and curvy appearance ideal internalization. The CII was intended to be used either alone or alongside the SATAQ-4R. The initial CII item pool is included as Supplemental Material.

3.3.2. Factor structure

Bartlett’s test of sphericity was significant, χ² = 27,220.22, df = 703, p < .001, and the Kaiser-Meyer-Olkin value was.95, indicating that CII items were appropriate for factor analysis. Multivariate outliers were removed from subsequent analysis. Analyses were conducted among the remaining 895 participants. Initial item-analysis suggested the removal of seven items (Items 15, 31, 32, 33, 35, 36, and 37) from the 37-item pool due to low inter-item correlations (< .30; Tabachnick & Fidell, 2007) prior to EFA. None of the remaining items lowered the scale’s internal consistency, so no additional items were removed prior to EFA.

With the remaining 30 items, parallel analysis identified a maximum of four factors, four factors had eigenvalues above 1.0, and visual examination of the scree plot and the variance accounted for by each factor suggested that the first two factors accounted for the greatest proportion of variance in the solution (55.00%), with a smaller proportion of the variance accounted for by the third (5.83%) and fourth factors (5.12%). The initial factor solution is presented in Supplementary Materials Table 1, along with removed items and their reasons for removal. Examination of the pattern matrix from the four factor-solution generated by parallel analysis showed that Item 6 (“I want my body to look very thick.”) did not load on any factors (factor loadings ranged from −.024-.232), and Item 23 (“I spend a lot of time doing things to have a flat stomach with a thin waist and big breasts”) cross-loaded on Factors 2 (.414) and 3 (.471) with relatively low loadings on both factors; consequently Items 6 and 23 were removed. Items on Factors 1–3 were conceptually distinct: Factor 1 included items related to desiring a large butt and thighs and a desire to look thick or curvy (Thick/Curvy), Factor 2 included items related to desiring a small waist and flat stomach (Thin Waist), and items that loaded onto Factor 3 reflected a desire for large breasts (Breast Size). Items that loaded onto Factor 4 (Items 2, 8, 11, and 14) overlapped conceptually with Factor 1. Thus, these items were removed and the EFA was re-run with the remaining items. Parallel analysis identified a maximum of three factors, three factors had eigenvalues above 1.0, and visual examination of the scree plot and the variance accounted for by each factor suggested that the first three factors accounted for the greatest proportion of variance in the solution (46.59%, 12.95%, and 7.44%, respectively). All items loaded onto only one factor each, with factor loadings ranging from.590-.951 (this three-factor solution is presented in Supplementary Materials Table 2).

A smaller subset of items (3–5 Items per factor) was selected from these 24 items for the final CII to create a brief, parsimonious item set based on items with the highest factor loadings using the following guidelines. First, behavioral items were removed to retain consistency with the SATAQ-4R’s cognitive conceptualization of the internalization construct (Items 4, 12, 16, 19, 20, and 26). Second, items that tapped into liking a body shape but did not assess the personal desire for that body shape were also removed to retain conceptual consistency with the SATAQ-4R internalization construct (Items 1, 3, and 13). For item pairs with correlations > .80, suggesting the items assessed overlapping concepts (Tabachnick & Fidell, 2007), the item with the higher factor loading was retained. Only one item pair had inter-item correlations higher than.80 (Items 17 and 18): Item 17 was retained because it had a higher factor loading (.730 vs.712). Additionally, given the conceptual overlap between Items 21 and 22 and high correlation (r =.75), Item 22 was removed, because it had a lower factor loading (.620 vs.646). Lastly, Item 34 was removed due to low conceptual overlap and lower factor loading (.541) with the other items in the factor.

The final 12-item, 3-factor solution is presented in Table 3. Factor 1, Thick/Curvy body ideal internalization, accounted for 46.37% of the total variance in the CII, Factor 2, Breast Size, accounted for 15.89% of the total variance in the CII, and Factor 3, Thin Waist, accounted for 12.15% of the variance, totaling 74.41% of the total variance accounted for by the three CII factors. The factors were moderately correlated with one another: rs ranged from.41-.48, suggesting that the factors represent related but not overlapping constructs. We hypothesized that a higher order Global CII included all three subscales.

Table 3.

Rotated Factor Pattern Matrix, with Eigenvalues and Percent Variance Explained for the CII subscales in Study 2.

Original Item Number	Item	Thick/Curvy	Breast Size	Thin Waist
5	It is important for me to look thick.	.895	−.003	−.043
7	I want my body to look very curvy.	.861	.000	−.097
9	I think a lot about looking thick.	.685	.037	.13
10	I think a lot about looking curvy.	.89	−.032	−.008
17	I want a flat stomach with a big butt and big thighs.	.631	.049	.108
21	I want to have a flat stomach with a thin waist and big breasts.	−.041	.902	−.022
28	I want big breasts.	−.009	.946	−.048
29	It is important for me to have big breasts.	.117	.776	−.059
30	I spend a lot of time thinking about having big breasts.	−.031	.564	.292
24	I like exercises that make my stomach flatter.	.057	−.048	.775
25	It is important for me to have a small waist.	−.046	.131	.641
27	I like exercises that make my waist smaller.	.001	−.073	.929
	Eigenvalues	5.56		1.91
	Percent variance explained	46.37		15.89

Note. CII = Curvy Ideals Internalization; Extraction Method: Principal Axis Factoring; Rotation Method: Promax with Kaiser Normalization. The rotation converged in three iterations.

3.3.3. Internal consistency

McDonald’s ω for the Global CII =.890 and Cronbach’s α =.894. The Thick/Curvy subscale ω =.900 and α = and.900, the Thin Waist subscale ω =.836 and α =.825, and the Breast Size subscale ω =.896 and α =.889.

3.3.4. Construct validity

Intercorrelations between the CII Global scale, the CII subscales, and convergent measures are presented in Table 4, along with means and SDs.

Table 4.

Scale and Subscale Means, Standard Deviations, Possible Scale Ranges, and CII Correlations with Validation Measures in Study 2.

Scale/Subscale	M(SD)	Scale Range	CII Global	Thick/Curvy	Breast Size	Thin Waist
CII Global	2.92 (0.92)	1–5	1.0
CII Thick/Curvy	2.72 (1.17)	1–5	.844***	1.0
CII Breast Size	2.77 (1.18)	1–5	.809***	.466***	1.0
CII Thin Waist	3.44 (1.09)	1–5	.688***	.377***	.447***	1.0
SATAQ-4 Thin-Ideal	2.96 (0.94)	1–5	.508***	.305***	.421***	.559***
SATAQ-4 Muscular Ideal	2.56 (1.08)	1–5	.581***	.490***	.453***	.424***
MBSRQ-AO	3.56 (0.73)	1–5	.352***	.218***	.238***	.451***
MBSRQ-AE	3.33 (0.93)	1–5	.167***	.266***	.066*	−.011
MBSRQ-BAS	3.36 (0.87)	1–5	.206***	.261***	.042	.042
PIQ Direct Influence	2.36 (0.88)	1–5	.346***	.259***	.332***	.223***
PIQ Modeling	2.48 (1.07)	1–5	.449***	.382***	.364***	.302***
CISS-CDI	0.07 (1.42)	−4–4	.120***	.158***	.040	.062
CISS-FDI	−0.43 (1.34)	−4–4	.093**	.169***	.055	−.069*
PSPS Media	2.99 (1.43)	1–5	.206***	.062	.211***	.277***
PSPS Family	2.14 (1.22)	1–5	.361***	.265***	.346***	.240***
PSPS Peers	2.16 (1.15)	1–5	.411***	.300***	.366***	.319***
PSPS Partners	2.72 (1.18)	1–5	.381***	.273***	.340***	.303***
BAS-2	3.69 (0.97)	1–5	.162***	.239***	.059	.033
Age	39.56 (15.52)	≥18	−.380***	−.475***	−.186***	−.160***

Note. CII = Curvy Ideals Internalization; SATAQ-4 = Sociocultural Attitudes Towards Appearance-4; MBSRQ = Multidimensional Body Self-Relations Questionnaire; AO = Appearance Orientation; AE = Appearance Evaluation; BAS = Body Areas Satisfaction; PIQ = Parental Influence Questionnaire; CISS = Curvy Ideal Silhouette Scale; CDI = Curviness Discrepancy Index; FDI = Fat Discrepancy Index; PSPS = Perceived Sociocultural Pressures Scale; BAS-2 = Body Appreciation Scale-2

^***p < .001

^**p < .01

^*p < .05

3.3.4.1. Curviness and fat discrepancy.

Consistent with expectations, the CII Global scale was significantly positively correlated with the CISS-CDI and CISS-FDI. However, contrary to expectations, there was no difference in the strength of association between the CII Global scale’s correlation with the CISS-CDI and CISS-FDI, Fisher’s z = 0.59, p =.56. The Thick/Curvy subscale was significantly positively correlated with the CISS-CDI and CISS-FDI. In contrast, the Breast Size and Thin Waist subscales were not significantly associated with the CISS-CDI and the Breast Size subscale was also not significantly associated with the CISS-FDI. The Thin Waist subscale was significantly negatively associated with the CISS-FDI.

3.3.4.2. Appearance orientation.

In line with hypotheses, the CII Global scale and all three subscales were significantly positively correlated with the MBSRQ-AO subscale.

3.3.4.3. Thin and muscular ideal internalization.

As hypothesized, we found medium to large positive correlations between the CII Global scale and its subscales and the SATAQ-4 internalization subscales (rs from.31-.58). There were some differences across CII subscales in the strength of correlations with the SATAQ-4 internalization subscales. The Breast Size subscale demonstrated significantly stronger positive correlations than the Thick/Curvy subscale and a significantly weaker positive correlation than the Thin Waist subscale did with Thin Ideal Internalization. Additionally, the Thick/Curvy subscale was more strongly correlated with Muscular Ideal Internalization than the Thin Waist subscale was, and the Thick/Curvy subscale had a stronger positive correlation with Muscular Ideal Internalization than it did with Thin Ideal Internalization. In contrast, the CII Thin Waist subscale had a stronger correlation with Thin Ideal internalization than it did with Muscular Ideal internalization. The strength of the correlations between Breast Size and the two SATAQ-4 Internalization subscales did not differ.

3.3.4.4. Pressures to achieve a slim body ideal.

Partially supporting hypotheses, the CII Global scale and its subscales were also significantly positively correlated with pressures to achieve a slim body ideal, measured using the PIQ and PSPS, with one exception: the Thick/Curvy subscale was not significantly correlated with media pressures to achieve a slim body ideal. There were some differences across CII subscales in the strength of correlations on measures assessing sociocultural pressures to be thin. Specifically, the Breast Size subscale demonstrated significantly stronger positive correlations than the Thick/Curvy subscale did with PSPS - Family and Media Pressures and PIQ - Direct Influence. The Breast Size subscale also had a significantly higher positive correlations than the Thin Waist subscale with the PSPS - Family Pressure and PIQ - Direct Influence subscales. Additionally, the Thick/Curvy subscale was significantly more strongly correlated than the Thin Waist subscale was with PIQ Modeling; however, the Thick/Curvy subscale had a lower positive correlation than the Thin Waist subscale did with the PSPS - Media Pressure subscale.

3.3.4.5. Body appreciation.

The BAS-2 was positively correlated with the CII Global and the Thick/Curvy subscale but was not significantly correlated with the Breast Size and Thin Waist subscales.

3.3.4.6. Appearance satisfaction.

The CII Global scale and Thick/Curvy subscale were significantly positively correlated with the MBSRQ-AE and MBSRQ-BAS; however, the Thin Waist subscale was not significantly associated with either measure of appearance satisfaction, and the Breast Size subscale was only significantly positively correlated with the MBSRQ-AE.

3.3.5. Test-retest reliability

Test-retest reliability (22 ± 8 days) for the CII was good; The ICC for Global CII =.86, Thick/Curvy ICC =.86, Breast Size ICC =.84, and Thin Waist ICC =.78.

3.4. Study 2 discussion

Using EFA, three factors were identified for the CII: Factor 1 assessed women’s internalization of thick and curvy body ideals, Factor 2 assessed women’s internalization of a body ideal with large breasts, and Factor 3 assessed women’s internalization of a body ideal with a small waist. Factors 2 and 3 correspond with the construct of a slim-thick body ideal (Wilfred & Lundgren, 2021). Although participants identifying with a thin ideal would score highly on the Thin Waist subscale and may also internalize larger breasts, both larger breasts and a slender waist were integral parts of the slim-thick body ideal described by focus group participants in Study 1. Thus, we felt that these factors, assessing internalization of a slim-thick body ideal, were important to retain. Correlations between the CII and well-validated measures of appearance orientation, appearance satisfaction, perceived sociocultural body image pressures, body appreciation, and internalization of thin and muscular body ideals supported the validity of the measure. Lastly, the CII demonstrated excellent test-retest reliability and internal consistency.

4. Study 3: Confirmatory factor analysis and group differences

The current study sought to reproduce the lower-order CII factors identified using EFA in Study 2 and to test whether Global CII represented a higher-order factor, hypothesized based on theory and the moderate correlations between the CII subscales. Lastly, Study 3 tested whether CII factors differed by racial group.

4.1. Methods

4.1.1. Participants and procedure

The study was determined to be exempt from IRB review by the ethical review board per 45 CFR 46.104(d)(2). Participants were recruited via Amazon’s Mechanical Turk (MTurk; www.mturk.com; N = 366) for monetary compensation for a study called: “Diverse Women’s Opinions About Advertising Effectiveness,” as part of a larger study on Black and (Black/White) biracial women’s body image. Online human intelligence platforms, like MTurk, have been recommended for reaching typically hard-to-reach populations (Scharmer et al., 2020), such as women of color. This sample was restricted to a 98% approval rate, > 18 years old, U.S. location, non-male gender identity, and White, Black, or biracial (Black and White) racial identity.

Of the 1516 individuals who accessed the survey, 1150 were removed for the following reasons: lack of consent, lack of responses following consent, failed attention or bot check items, and duplicate or improbable responses (e.g., reporting height of 7’4” and weight of 50 lbs., or self-reported race that did not match with biological parents’ reported races). This yielded a final sample of 366 participants (61.7% White; 25.7% Black; 12.6% biracial). Participant demographics are reported in Table 2. After providing informed consent, participants completed questionnaires, including the CII, with one bot check and three attention check questions throughout the survey.

4.1.2. Data analysis

4.1.2.1. Factor structure.

The CFA used weighted least squares meanand variance-adjusted estimator (WLSMV), as WLSMV is a robust estimation method for ordinal data (Li, 2016). Model fit was determined by examining the following fit indices: χ² values, the comparative fit index (CFI), Tucker-Lewis Index (TLI), root mean square error of approximation (RMSEA), and standardized root mean square residual (SRMSR; Credé & Harms, 2015). Traditional guidelines (Hu & Bentler, 1999) suggest that CFI values above 0.90 represent acceptable model fit, with.95 or higher representing good model fit. RMSEA values of 0.06 or less, and SRMR values of 0.08 or less, indicate good model fit (Bentler, 1990; Hu & Bentler, 1999). However, Credé and Harms (2015) caution against the strict use of cut-off scores to determine model fit, noting that the cut-off scores are arbitrary and also were not developed for use with higher-order models. Thus, models with lower RMSEA and SRMR and higher CFI and TLI were considered better fitting than those with higher RMSEA and SRMR and lower CFI and TLI, but cut-off scores were not used to determine whether the model fit was acceptable or not (Credé & Harms, 2015). Item performance was examined via factor loadings, R², and examination of modification indices (Byrne, 2012). The CFA was conducted in MPlus, version 1.8.6(1) for Mac (Muthén & Muthén, 2007).

Based on guidelines presented by Credé and Harms (2015), the CFA models tested whether data supported the presence of higher-order factor structure, by comparing a higher order model to several lower order models.

4.1.2.2. Norms and group differences by racial identification.

Racial differences in CII subscales were tested combining samples from Study 2 and Study 3. ANOVAs tested whether there were significant differences between participants’ racial identification and continuous demographic characteristics (age and BMI) and χ² for categorical demographic characteristics (education, income, and sexual orientation). Any demographic characteristic that differed across racial groups was included as a covariate in subsequent analyses. ANCOVAs with Bonferroni-corrected post hoc comparisons were conducted to assess mean differences across racial group, controlling for any covariates which differed by racial group.

4.2. Results

4.2.1. Factor structure

The CFA initially identified that the model with best fit (based on fit indices) was not the higher-order model, but a more parsimonious oblique first-order model, in which the three subfactors were correlated, but in which there was no Global CII factor (see Credé & Harms, 2015 for visual illustrations of the alternate lower-order models). Based on focus group participants’ descriptions of the thick versus the slim-thick beauty ideals, a CFA with the Thin Waist and Breast Size factors combined into one Slim-Thick factor was also examined. However, the oblique first-order model, with three correlated subfactors remained a better fit⁴.

Modification indices were high for Slim Waist by Item 21 (“I want to have a flat stomach with a thin waist and big breasts.” MI = 74.18), Thick/Curvy by Item 30 (“I spend a lot of time thinking about having big breasts.” MI = 40.32), and Breast Size by Item 25 (“It is important for me to have a small waist.” MI = 39.17) indicating significant cross-loadings. Rather than respecify the model to freely estimate Item 21, it was removed, both in the interest of parsimony and to avoid model overfitting (Brown, 2015; Byrne, 2012). This resulted in improved model fit indices. Modification Indices still indicated significant cross-loading for Items 25 (MI = 38.23) and 30 (MI = 36.28). Because removal of Items 25 and 30 would result in two 2-item factors, which can result in decreased internal consistency (Ziegler, Kemper, & Kruyen, 2014), the CFA was re-run, adding these cross-loadings, sequentially. Allowing Item 25 to load on Thin Waist and Breast Size significantly improved model fit, as did allowing Item 30 to load on the Thick/Curvy factor. The final model, omitting Item 21 and allowing for cross-loading of Items 25 and 30, was a generally good fit, χ²(39) = 173.07, RMSEA =.098[.083,.113], CFI =.978; TLI =.969; SRMR =.039. Thus, we recommend the final CII include the Thick/Curvy subscale with original item numbers 5, 7, 9, 10, 17 and 24, the Thin Waist subscale with original item numbers 24, 25, and 27, and the Breast Size subscale with original item numbers 28, 29, and 30 (see Table 3 Supplementary material for final CII measure and scoring instructions). Standardized factor loadings and R² values for the final 11-item questionnaire are reported in Table 5.

Table 5.

Standardized Factor Loadings, Significance Values, and R² Values of Final CII items.

Initial Item Number	Item	Standardized Factor Loadings
	Thick/Curvy Subscale	Loading	SE	R 2
CII 5	It is important for me to look thick.	.890***	.016	.793
CII 7	I want my body to look very curvy.	.768***	.026	.590
CII 9	I think a lot about looking thick.	.905***	.016	.819
CII 10	I think a lot about looking curvy.	.869***	.018	.754
CII 17	I want a flat stomach with a big butt and big thighs.	.755***	.028	.570
	Thin Waist Subscale
CII 24	I like exercises that make my stomach flatter.	.718***	.038	.516
CII 25	It is important for me to have a small waist	.645***	.041	.555
CII 27	I like exercises that make my waist smaller.	.941***	.036	.885
	Breast Size Subscale
CII 28	I want big breasts.	.840***	.019	.706
CII 29	It is important for me to have big breasts.	.970***	.014	.941
CII 30	I spend a lot of time thinking about having big breasts.	.668***	.034	.646

Note. CII = Curvy Ideals Internalization

^***p < .001

4.2.2. Group differences by racial identification

Participants’ age differed significantly across racial groups, F (2, 1,254) = 78.58, p < .001, with White participants significantly older, M (SE)_AGE = 44.04 (0.59), than both Black, M(SE)_AGE = 34.50 (0.68), and biracial participants, M(SE)_AGE = 33.96 (0.80). Education differed significantly across racial groups, χ² (10) = 30.55, p < .001, as did income, χ² (10) = 34.61, p < .001, and sexual orientation, χ² (8) = 24.34, p < .001, so these were also included as covariates in subsequent ANCOVAs. BMI did not differ significantly across racial groups, p =.09.

After controlling for age, income, education, and sexual orientation, participants’ Thick/Curvy subscale scores differed significantly across racial groups, F (2, 1,249) = 12.05, p < .001. Post hoc comparisons showed that White women, M(SE) = 2.72 (0.045), had significantly lower Thick/Curvy internalization than Black women, M (SE) = 3.05 (0.051), p < .001, and biracial women, M(SE) = 2.97 (0.059), p =.004. Black women’s Thick/Curvy internalization did not significantly differ from biracial women’s, p =.88. After controlling for covariates, there were significant differences across racial groups on the Breast Size subscale, F (2, 1,249) = 6.08, p =.002. Post hoc comparisons showed that White women, M(SE) = 2.82 (0.053), scored significantly higher than Black women, M(SE) = 2.53 (0.060), p =.002, but did not differ significantly from biracial women’s scores, M(SE) = 2.71 (0.071), p =.67. Black women’s Breast Size internalization did not significantly differ from biracial women’s, p =.17. Lastly, participants’ Thin Waist scores did not differ significantly across racial groups, after controlling for covariates, F (2, 1,249) = 1.17, p =.31, with non-significant differences between White women, M(SE) = 3.51 (0.046), and Black women, M(SE) = 3.44 (0.052), p =.89, between White women and biracial women, M(SE) = 3.56 (0.061), p = 1.00, and between Black and biracial women, p =.42.

4.3. Study 3 discussion

The current study compared a higher order model with more parsimonious orthogonal and oblique first order models using CFA and supported an oblique first order model in which the three subfactors of the CII were correlated. CFA did not support a higher order model with a global CII scale. The final scale has 11 items, and supported inclusion of the Thick/Curvy, Thin Waist, and Breast Size subscales.

There were main effects of race on thick/curvy body ideal internalization, with Black and biracial women scoring higher on this subscale than White women, supporting prior research reporting preferences for larger buttocks and thighs among women of color (Capodilupo & Kim, 2014; Overstreet et al., 2010; Hunter et al., 2021; Webb et al., 2013). Additionally, there were main effects of race on women’s internalization of a body ideal with large breasts, with White women scoring significantly higher than Black women. Biracial women’s Breast Size scores were in between Black and White women’s but did not differ significantly from either group. In contrast, women from all racial groups had similar internalization of a body ideal with a small waist.

These findings were aligned with focus group participants’ descriptions of the thick body ideal as largely specific to communities of color, and not particularly relevant among White women. In contrast to the thick body ideal, focus group participants noted that both White women and women of color coveted a slim-thick body, which aligned closely with the CII Thin Waist and Breast Size constructs and was supported by a lack of significant differences in the desire for a small waist across groups.

5. Discussion

We developed and validated a measure of curvy body ideals internalization, the CII, in two large samples of White, Black, and biracial (Black and White) adult women (Ns = 897; 366) residing in the U.S. In Study 1, women of color were recruited to take part in focus groups to help inform initial item generation of the CII. Study 2 evaluated the factor structure, test-retest reliability, and construct validity of the CII in a large diverse female sample. Study 3 confirmed the factor structure and examined racial differences in CII subscales in a separate diverse sample of women. The CII demonstrated strong construct validity, including significant positive correlations with measures of appearance satisfaction, appearance orientation, parental influence to diet, sociocultural pressures to diet, and internalization of the thin and athletic/muscular ideals. Additionally, the current study used psychometric theory to refine the item selection from a larger item pool using EFA and item analysis (cf. Haladyna, 2004; Tabachnik & Fidell, 2007), rather than significantly reducing the item pool prior to conducting EFA. The CII was designed to be used either on its own or in conjunction with the SATAQ. When used on its own, the CII provides a valid measure of curvy body ideals, which may facilitate future research directly examining the risk factors and consequences of this experience. In addition, the CII can be used alongside the SATAQ to facilitate research comparing the prevalence and correlates of three distinct forms of internalization. The current measure was validated among two large, diverse samples of women, a notable strength. Lastly, the CII identified subscales reflecting idealization of a thick body ideal, with the Thick/Curvy subscale, and a slim-thick body ideal, with the Thin Waist and Breast Size subscales, a distinction which has not been explored in existing measures of the desire for a curvy body (Hernández et al., 2021; Hunter et al., 2021).

Across both focus groups, participants noted that the slim-thick body ideal was not realistic or achievable, and was commonly coveted across all racial groups, whereas the thick ideal was more often seen among communities of color and was a more realistic body ideal. This was supported by findings of significant differences across racial groups for the thick/curvy internalization, with fewer differences across racial groups evident in breast size internalization, and no group differences in thin waist internalization. Differences between thick and slim-thick idealization were also evident in the strengths of correlations between the CII subscales and other measures.

For example, the Thick/Curvy subscale had a higher correlation with muscular ideal internalization than with thin ideal internalization, suggesting that women who internalize a thick body ideal with larger lower body features may also desire muscularity more than they do thinness. In contrast, the correlations between the CII Breast Size subscale and both SATAQ-4 internalization subscales were similar, suggesting that women who desire a slim-thick upper body may desire a lean and toned figure, more generally. Additionally, the Thin Waist subscale had higher correlations with thin ideal internalization than with muscular/athletic ideal internalization, suggesting a greater association with desire for overall thinness. Overall, the focus group and CII subscales suggest that among individual women, especially women of color, there may be greater diversity and nuance in body type desired, with some women preferring more body fat overall, and some preferring body fat in some parts of their bodies (e.g., breasts, buttocks) but not in other areas (e.g., waist). Additionally, women in our focus group noted that their body ideal internalization shifted based on the contexts in which they lived (e.g., in schools or communities of color vs. predominantly White environments), suggesting the need for more variable and inclusive measures to assess women of color’s current body ideal internalization.

The current study also provided insight into whether internalization of curvier body ideals is potentially beneficial or harmful. There is more evidence supporting Thick/Curvy subscale as a positive influence than the Breast Size and Thin Waist subscales. In particular, the Thick/Curvy subscale was significantly correlated with body appreciation and body area satisfaction, whereas the Breast Size and Thin Waist subscales were not. Additionally, the Thin Waist subscale was not significantly correlated with appearance evaluation, and the Thick/Curvy subscale had a significantly stronger positive correlation with appearance evaluation than did the Breast Size subscale. Furthermore, the Breast Size and Thin Waist subscales were significantly correlated with media pressures to achieve a thin ideal, while the Thick/Curvy subscale was not. Future research should examine whether the Thick/Curvy, Breast Size, and Thin Waist subscales may differentially predict eating disorder symptoms and facets of positive body image.

Although the thick/curvy ideal internalization was associated with fewer negative constructs in the current study, even pressure to be thick or curvy may be harmful for those who do not fit that body type (Hughes, 2018). For example, a few women in our focus group noted pressures to be thick/curvy when they were in Black or Latina communities. Thus, for women who do not fit the sociocultural body ideal that predominates in their main social contexts, they may experience body-related pressure and insecurity. Thus, internalizing a curvier body ideal may be protective in some contexts and problematic in others, depending on women’s actual body type and the sociocultural pressures in their environment.

The CII also provides potential clinical utility in assessing a construct related to the understudied body image concerns of women with curvy body ideals, which may be associated with under-researched problematic behaviors. For example, while dietary restriction and excessive exercise are common behaviors among women who have internalized thin and athletic ideals, women who have internalized a slim-thick body ideal may use specific clothing to try to amplify their curves (e.g., Spanx, waist trainers, corsets), which can cause harm with regular use. Some women also seek more extreme body modifications, such as cosmetic surgery, including breast and buttocks implants (Abel de la Peña et al., 2006; Roy et al., 2021). Women who attempt to alter their body shape using less drastic means may engage in specific exercise routines to attempt to augment their buttocks (e.g., weighted squats, lunges, etc.), while simultaneously working to slim their waists (often via dietary restriction, diet pills, abdominal workouts, and cardiovascular exercise). This was echoed by women in the focus groups who described diets and exercise regimes promoted by social media influencers to achieve the slim-thick body ideal. Moreover, women of color may feel both pressures to achieve slim and muscular physiques and thick or slim-thick physiques depending on the contexts in which they spend time.

Recent research from the 2012–2013 National Epidemiologic Survey on Alcohol and Related Conditions III (NESARC-III) found that non-Hispanic Black women represented 2.8% of those with anorexia nervosa, 8.7% of those with bulimia nervosa, and 8.6% of those with binge eating disorder (Coffino et al., 2019). However, these prevalence rates may not capture women with a range of disordered eating and exercise behaviors aimed at achieving a thick or slim-thick body ideal who may not meet diagnostic criteria for anorexia, bulimia, or binge eating disorder. Thus, measure development assessing a range of harmful thick and slim-thick body ideal-related behaviors is also needed. Research has found low rates of treatment utilization among non-Hispanic Black (Coffino et al., 2019; Pike et al. 2001) and Hispanic/Latinx individuals who have eating disorders (Alegria et al., 2007; Coffino et al., 2019; Higgins Neyland & Bardone-Cone, 2019). Failure to address and better understand women of color’s specific eating and body image psychopathology may contribute to their treatment underutilization.

The current study must be considered alongside its limitations. First, we did not collect the specific racial and ethnic backgrounds from focus group participants, which was a notable researcher oversight. While the participants primarily described themselves as coming from Black and Latinx communities, and most participants provided race and ethnicity information subsequently, we did not have information from our full sample of focus group attendees. Additionally, samples for Study 2 and 3 were collected online, both due to COVID-19 and the need to recruit large samples of women of color for factor analysis (Tabachnick & Fidell, 2007). However, research suggests that MTurk samples yield data equal or superior to student samples, especially regarding motivation for participation and diversity of population demographics (Buhrmester et al., 2011; Paolacci et al., 2010). Some recent research has called the validity of MTurk data into question (e.g., Burnette et al., 2021; Kennedy et al., 2020), suggesting that participants may be completing questionnaires as quickly as possible without answering honestly or may be dishonest about their identities to be able to complete higher paying studies. The current study’s MTurk sample followed recommendations by Burnette et al. (2021) to screen data to better ensure data quality and validity and did not advertise differential pay rates based on participants’ race. While we cannot guarantee the validity of the MTurk sample in Study 3, the fact that CFA results largely supported EFA results from a sample with considerable data quality screening by CloudResearch suggests that these samples were largely consistent and reporting honestly when completing the CII and other measures.

We did not limit participants’ age, so participants in the current study were older than the average college-aged sample, in which body image concerns are typically elevated (Eisenberg et al., 2011; Filipova & Stoffel, 2016; Hudson et al., 2012). It is possible that women’s desire for a curvy or thick body may change over time. Given the lack of validated measures of the desire for a curvier body until recently, data have not been published assessing whether age-related changes are evident in thick/curvy body ideal internalization via longitudinal research. Indeed, the CII and its subscales and the HBSIS (Hernández et al., 2021) were significantly inversely correlated with age. However, recent research has identified that women in midlife also experience body image dissatisfaction, often associated with changes in their bodies and sociocultural emphases of youthfulness in defining attractiveness (Hofmeier et al., 2017; Jackson et al., 2014; McGuinness & Taylor, 2016; Pruis & Janowsky, 2010; Rodgers et al., 2016; Samuels et al., 2019). Additionally, recent research has found an increase in eating disorders among women in middle age and later life (Samuels et al., 2019), and midlife women with higher body dissatisfaction also are at greater risk of significant depressive symptomatology (Jackson et al., 2014). Lastly, one study reported that although younger women had greater thin ideal internalization and experienced greater societal influence on their body image, older and younger women’s body dissatisfaction did not differ (Pruis & Janowksy, 2010). Thus, research suggests that body image concerns are prevalent in women as they age and are associated with other mental health problems. Future research is needed both to examine changes in thick/curvy body ideal internalization over time in longitudinal research and to determine the psychometric properties of the CII in younger samples.

A third limitation of the current study was our sample’s diversity. While we had two large samples of Black and biracial women, the CII was not validated among women from other racial and ethnic backgrounds, as data collection was part of a larger study investigating Black and biracial women’s body image. Thus, the CII currently only has demonstrated validity and reliability among Black, biracial (Black and White), and White women, which represents a limitation to its broad application among women of color from other racial and ethnic backgrounds. Research suggests that Latina women also endorse lower levels of thin ideal internalization than White women, and thin ideal internalization is less predictive of negative outcomes (e.g., body image disturbance) among Latina populations (Burke, Schaefer, et al., 2021; Rakhkovaskaya & Warren, 2014). In contrast with the traditional thin ideal, research also suggests that many Latina women prefer curvier and larger or thicker body types than White women (Franko et al., 2013; Hernández et al., 2021; Viladrich et al., 2009). Thus, validating the CII among Latina samples, other racial/ethnic groups, and multiracial women represents a next important step. Additionally, the study included participants who identified as transgender, gender nonbinary, and agender; however, the sample was predominantly cisgender women, and findings are likely to generalize only to cisgender Black, biracial, and White women. Larger samples of transgender, gender nonbinary, and agender participants are necessary to better understand whether they internalize thick or slim-thick body ideals. Furthermore, research is necessary to better understand the body ideals internalized by non-cisgender individuals of color, and to examine whether race and gender identity intersect in determining individuals’ body ideal internalization. Lastly, validation outside of the U.S. is also important, given that culture significantly influences body image ideals (Guan et al., 2012; Lowy et al., 2021).

Finally, although we assessed a range of body image and sociocultural dieting and appearance pressure measures, we did not compare the CII scale to measures of disordered eating. Future research should examine the association between the CII and disordered eating measures. Based on the mixed results in the current study, more research is necessary to understand whether CII subscales represent protective or risk factors. In particular, we found that internalization of a thick body ideal was more strongly associated with positive body image and less strongly associated with eating disorder risk factors than slim-thick internalization, as represented by the Thin Waist and Breast Size subscales. Furthermore, there were racial differences in thick body ideal internalization and breast size internalization on the CII but no racial differences in internalization of a thin waist. Further research is necessary to determine under what circumstances and for whom internalizing a thick or curvy body may be beneficial or harmful, and whether the Thick/Curvy, Breast Size, and Thin Waist subscales differentially predict eating disorder symptoms. Hunter et al. (2021) found that wanting a curvier figure than one’s perceived current figure was associated with decreased eating disorder pathology. However, Hernández et al. (2021) reported significant positive correlations between the desire for an hourglass figure and measures of weight and shape concern, food preoccupation and restriction, and eating shame. These findings suggest nuance likely exists in whether curvy ideal internalization is protective or problematic, especially for women of color. Examining the relationships between the CII subscales and measures of disordered eating among diverse groups of women may also help clarify this question.

Although assessing measurement invariance is recommended when there are mean differences in endorsement of items across groups to determine whether differences also exist in factor structure across those groups, sample size limitations precluded the examination of measurement invariance across racial groups in the current study. In particular, Brown (2015) and Byrne (2012) note the importance of cross-validation with an independent sample to avoid rejection of the null hypothesis (H₀ = all parameters are the same across groups) due to chance, made likely by the high sample sizes needed for CFA. It was not feasible to collect a third separate large sample of Black and biracial women for the current study due to the time and cost already required to collect the two datasets for Studies 2 and 3; thus, measurement invariance was not able to be assessed in the current study and remains an important future direction.

Despite these limitations, the current study had notable strengths: we (1) validated the CII in two large, diverse samples of women; (2) included assessment of a thick body ideal not included in existing curvy ideal measures (Hernández et al., 2021; Hunter et al., 2021); (3) the CII is directly comparable to SATAQ Internalization subscales; and (4) the CII enables assessment of what appear to be distinct larger body ideals (i.e., thick and slim-thick) providing greater nuance in assessing women’s larger body ideals.

Data Availability

Data will be made public upon publication.