• Mike C. Parent and Bonnie Moradi, Department of Psychology, University of Florida.

  • We thank Lisa Seymour for assistance with data collection.

Address correspondence and reprint requests to: Mike C. Parent, Department of Psychology, University of Florida, P.O. Box 112250, Gainesville, FL 32611-2250. E-mail:


The present study undertakes the first factor analysis of the Conformity to Feminine Norms Inventory (CFNI) conducted since the instrument's development. Confirmatory factor analysis using data from 243 women offered mixed support for the original 84-item CFNI structure and pointed to ways to modify the structure and reduce the length of the measure; these modifications resulted in the nine-factor CFNI-45. The CFNI-45 demonstrates superior data-model fit and comparable reliability coefficients relative to the original form of the measure. The CFNI-45 subscales also yielded high correlations with corresponding original form subscales, supporting use of the CFNI-45 as an efficient measure of the original constructs. Potential uses of the CFNI-45 in research and practice that address the role of feminine gender-role conformity in women's experiences are discussed.

Gender roles are culturally enforced rules that outline appropriate behaviors for women and men (Brehm, Miller, Perlman, & Campbell, 2002). The study of gender-role conformity—the extent to which individuals agree with or abide by the gender expectations set upon them by their culture—has flourished in recent years. However, the preponderance of research has focused on the experiences of men. Numerous instruments have been developed to measure men's gender-role conformity and the conflict or stress associated with such conformity (e.g., Masculine Gender Role Stress Scale: Eisler & Skidmore, 1987; Male Role Norms Inventory: Levant et al., 1992; Conformity to Masculine Norms Inventory: Mahalik et al., 2003; Gender Role Conflict Scale: O’Neil, Helms, Gable, Vogel, & Wrightsman, 1986; Male Role Norms Scale: Thompson & Pleck, 1986). These instruments have been used to identify correlates of men's masculine gender-role conformity, including relationship satisfaction, reluctance to seek psychological help, and psychological health (e.g., Berger, Levant, McMillan, Kelleher, & Sellers, 2005; Burn & Ward, 2005; Good et al., 2006; Levi, Chan, & Pence, 2006).

By comparison, there is a dearth of parallel instrumentation and research on women's gender-role conformity that examines feminine norms as an interconnected set of norms that constitute how the dominant culture construes femininity. As a result, researchers wishing to examine women's conformity to feminine norms have often used measures that assess related constructs, but do not assess feminine-norm conformity directly (Kite, 2001). For example, researchers have used instruments that assess individual differences on instrumentality and expressiveness (e.g., Bem Sex Role Inventory [BSRI]: Bem, 1974; Personal Attributes Questionnaire: Spence, Helmreich, & Stapp, 1974), that focus primarily or exclusively on attitudes about appropriate rights and roles for women and men (e.g., Feminine Role Norms Scale: Lefkowitz, Shearer, Gillen, & Espinosa-Hernandez, 2009; Femininity Ideology Scale: Levant, Richmond, Cook, House, & Aupont, 2007; Hypefemininity Scale: Murnen & Byrne, 1991; Attitudes Toward Women Scale: Spence & Helmreich, 1978), or that measure perceived stress associated with violating feminine gender role norms (e.g., Feminine Gender Role Stress Scale: Gillespie & Eisler, 1992). These constructs may be related to conformity to feminine norms, but do not reflect respondents’ personal conformity to such norms directly.

Alternatively, researchers have focused on conformity to single manifestations of feminine norms such as domestic cleanliness, thinness, concern over physical appearance, and modesty. Women's adherence to these norms has been linked with body dissatisfaction (Cahill & Mussap, 2007; Tiggemann, 2006), ratings of a hypothetical woman's perceived masculinity and femininity (Harris & Sachau, 2005), emotional distress during treatment for breast cancer and melanoma (Carver et al., 1998; Lichtenthal, Cruess, Clark, & Ming, 2005), concern over evaluation by others (Rudman & Glick, 2001), and women's ratings of their own competency (Berg, Stephan, & Dodson, 1981). As such, conformity to specific feminine norms appears to be related to important aspects of women's experiences and well-being. But assessing specific norms as disparate constructs ignores how the set of interconnected feminine norms together can function to oppress women (Frye, 1983).

From a feminist perspective, strict adherence to feminine norms (and masculine norms) can limit women's (and men's) potential by limiting the range of socially acceptable choices and behaviors available to them. In addition, feminine and masculine norms reflect and reinforce presumed gender differences that afford greater value, power, and privilege to men and masculinity than to women and femininity. For example, feminine norms of relational orientation, modesty, appearance focus, and domesticity (Mahalik, Morray et al., 2005) reflect relinquishing or sharing power, whereas masculine norms of drive to win, power over women, violence, and self-reliance (Mahalik et al., 2003) reflect asserting power in one's own life and over others. These gender-role norms create a bind for women, that is, women who adhere to feminine norms may be rewarded for satisfying social expectations, but relegated to low social status, and women who violate feminine norms may be treated with contempt, but allowed some access to higher social status (Cuddy, Fiske, & Glick, 2004; Fiske, Cuddy, Glick, & Xu, 2002; Frye, 1983).

Given the centrality of analyses of power and oppression in feminist therapy, attention to conformity to feminine norms also is a central aspect of such therapy. Specifically, articulations of feminist therapy principles routinely include attending to the implications of women's gender-role socialization for clients’ experiences and presenting concerns (e.g., Worell & Johnson, 2004). Empirical evidence also underscores attention to gender-role socialization as an integral aspect of feminist therapists’ work with women (Chester & Bretherton, 2001; Moradi, Fischer, Hill, Jome, & Blum, 2000). Additionally, the American Psychological Association's (APA, 2007) Guidelines for Psychological Practice with Girls and Women (hereafter Guidelines) suggest that women's gender-role socialization occurs throughout the life span, that the enactment of gender roles may occur beneath awareness, and that such enactment can have deleterious implications for women's mental health, task performance, and aspirations.

Thus, direct assessment of feminine-norm conformity is important for advancing feminist research and practice with women. But such advancements are limited to the extent that most available measures do not assess adherence to feminine norms directly or assess narrow manifestations of such norms. To address these conceptual and measurement gaps, Mahalik, Morray et al. (2005) developed the Conformity to Feminine Norms Inventory (CFNI) as a measure of women's conformity to a range of feminine norms salient in North American society. As such, the CFNI is a promising tool for research and clinical work. In the present study, we built on Mahalik, Morray et al.'s (2005) foundational work to evaluate the factor structure of the CFNI and explore directions for its refinement. Specifically, we sought to evaluate the multidimensional conceptual model of feminine norms underlying the CFNI and to broaden the utility of the CFNI by optimizing its length and improving its psychometric properties.

Development of the CFNI

Mahalik, Morray et al. (2005) began development of the CFNI by identifying feminine gender-role norms reflected in the existing literature. They focused on U.S. European American, middle- and upper-class, heterosexual construals of feminine norms, reasoning that such dominant cultural norms of femininity define the standards to which American women of various cultural backgrounds are held. Next, five focus groups of women (undergraduate students, graduate students, and community members; 30 European American, 1 Hispanic/Latina, 1 biracial; 22 heterosexual, 6 lesbian, 4 bisexual) were asked to identify cultural messages about “how women are supposed to act, think, and feel” (Mahalik, Morray et al., 2005, p. 419). Two of the researchers (demographic characteristics not described) listened to audiotapes of the focus groups, identified norms that emerged from the focus groups, and incorporated those norms into the list of norms identified from the literature review. The list of norms was sorted into higher order clusters, and the researchers worked with two additional focus groups of women and men (graduate students; demographic characteristics not described) who met weekly over an 8-month period to refine those clusters and create corresponding items. Resultant items were then piloted with four small groups of college women (20 to 30 people per group, demographic characteristics not described) who offered their reactions and responded to the items. The aforementioned two focus groups discussed the pilot samples’ reactions as well as the internal consistency reliability estimates that emerged from the pilot data and further refined items to improve their readability, reliability, and content validity. At the end of this process, 12 feminine norms—Relational, Sweet and Nice, Thinness, Put Others First, Look Young, Sexy, Modesty, Domestic, Care for Children, Romantic Relationship, Sexual Fidelity, and Invest in Appearance—were identified. Twelve items were created for each norm, resulting in an initial pool of 144 items.

Following this process, an exploratory factor analysis (EFA) was conducted using data from a sample of predominantly European American heterosexual undergraduate women. Mahalik, Morray et al. (2005) indicated that they examined various possible factor structures—beginning with a 12-factor solution reflecting the feminine norms that emerged from the literature review and focus groups—and then reduced the number of factors to which the EFA was constrained in progressive steps. Based on this process, Mahalik, Morray et al. (2005) first selected a 10-factor solution as the most conceptually interpretable, but noted some problematic items in this factor solution as well. Specifically, items for Look Young, Put Others First, and Sexy did not load clearly on any specific factor or were dispersed across factors. Because these three factors did not emerge as distinct norms, Mahalik, Morray et al. (2005) deleted the items intended for these factors and reexamined the factor structure. They selected an eight-factor solution as the final model. But the process by which the eight-factor model was selected was not detailed, leaving some questions about optimal structure: importantly, how did a nine-factor solution (i.e., the initial 12 factors minus the 3 deleted factors) fit the data?

Based on the final eight-factor solution, items were retained in the measure if they had a factor loading of at least .40 on one factor and did not cross-load over .30 on another factor. These criteria led to the retention of 84 items. Most items loaded onto their intended factor, with the exception that items intended for the Relational and Sweet and Nice factors loaded together onto a single Nice in Relationships factor. This pattern again raises the question about how a nine-factor solution that allowed these items to load onto separate factors would have fit the data.

The final version of the CFNI proposed by Mahalik, Morray et al. (2005) was an eight-factor, 84-item measure that reflected Nice in Relationships (sample item: “I believe that my friendships should be maintained at all costs”), Thinness (sample item: “I would be happier if I was thinner”), Modesty (sample item: “I always downplay my achievements”), Domestic (sample item: “It is important to keep your living space clean”), Care for Children (sample item: “Taking care of children is extremely fulfilling”), Romantic Relationship (sample item: “Having a romantic relationship is essential in life”), Sexual Fidelity (sample item: “I would feel guilty if I had a one-night stand”), and Invest in Appearance (sample item: “I spend more than 30 minutes a day doing my hair and makeup”). This eight-factor CFNI, with 84 items that subsume nine and drop three of the originally intended feminine norms, is the version currently available for use.

With their sample of predominantly European American heterosexual women, Mahalik, Morray et al. (2005) found promising initial psychometric support for the CFNI. Specifically, CFNI subscale scores had low to moderate intercorrelations, supporting the multidimensionality of feminine norms as distinct but related constructs. Furthermore, according to Ponterotto and Ruckdeschel's (2007) matrix for interpretation of Cronbach's alpha, CFNI subscale items yielded internal consistency reliabilities that were fair to excellent. Two-to-three week test-retest reliability coefficients also were strong across subscales. In terms of validity, CFNI scores were correlated as expected with conceptually related constructs. For example, Nice in Relationships, Care for Children, Romantic Relationship, Sexual Fidelity, and CFNI total scores were correlated positively with scores on the BSRI Feminine Identity scale (Bem, 1974). Similarly, Modesty, Sexual Fidelity, and CFNI total scores were correlated negatively with BSRI Masculine Identity scores. Additionally, Domestic, Romantic Relationship, Sexual Fidelity, and CFNI total scores were correlated positively with Passive Acceptance subscale scores of the Feminist Identity Composite (Fischer et al., 2000). Finally, CFNI Thinness scores were correlated positively with Drive for Thinness, Bulimia, and Body Dissatisfaction subscales of the Eating Disorders Inventory-2 (EDI-2; Garner, 1991), and Nice in Relationships and Care for Children norms were associated negatively with scores on the Ineffectiveness and Interpersonal Distrust subscales of the EDI-2. Thus, with predominantly European American heterosexual women, the CFNI demonstrated initial promise as a measure of conformity to multiple feminine norms.

Since its development, the CFNI has been used to examine the relations of predominantly European American or Caucasian heterosexual college women's conformity to feminine norms with vocational, identity, and mental health-related variables. For example, with a sample of primarily European American college students, Tokar, Thompson, Plaufcan, and Williams (2007) found that CFNI total scores were related positively with learning experiences in Artistic, Social, and Realistic (i.e., working with objects or outdoors) career domains. Furthermore, CFNI total scores mediated the relations of gender with these learning experiences, such that women reported greater conformity to feminine norms than did men; and conformity to feminine norms, in turn, was associated with greater Artistic, Social, and Realistic learning experiences. This pattern of findings suggests that feminine-norm conformity may help explain some women's channeling into certain career-related learning experiences and pursuits. In another study with predominantly Caucasian, heterosexual, middle- or upper-middle class college women, Hurt et al. (2007) found that feminist self-identification was related negatively with CFNI Thinness, Invest in Appearance, and Romantic Relationship subscale scores, each of which in turn was related directly or indirectly with greater body concerns, eating disorder symptoms, depression, and lower self-esteem (although the significance levels of indirect effects were not reported in this study). Thus, use of the CFNI with predominantly European American or Caucasian heterosexual college students is yielding support for the notion that conformity to feminine norms may play an important role in women's life experiences and mental health. As such, attention to psychometric refinement of the CFNI and improvements to facilitate its use seem warranted.

Need for Reexamination of the CFNI

The CFNI is an important contribution to the literature because it assesses conformity to feminine norms directly and multidimensionally. However, there has been no psychometric evaluation of the CFNI beyond the initial instrument development study. Examining the conceptual model underlying the CFNI, evaluating its psychometric properties, and exploring optimization of subscale and overall instrument length, each can advance the utility of the CFNI and can facilitate its broader use in research and clinical work with women. This study aims to address these needs.

The first aim of this study is to revisit the conceptual model and factor structure of the CFNI. Specifically, as noted previously, a nine-factor solution to CFNI data was not described in the construction of the measure. A nine-factor solution could potentially provide a superior model for the CFNI, because the items in the collapsed Nice in Relationships factor were originally intended to measure two separate constructs. Thus, in this study, we compare the fit of an eight-factor solution with that of a nine-factor solution that splits the Nice in Relationships items into (a) Sweet and Nice and (b) Relational factors.

The second objective of this study is to optimize subscale length using item retention criteria recommended to yield clear indicators of underlying constructs. Specifically, the 84-item length of the CFNI may limit its use in research and practice when participant fatigue and clinical efficiency are of concern. Additionally, the length of the CFNI was shaped by retention of items that met absolute cutoffs for factor loadings (greater than .40) and cross-loadings (less than .30; Mahalik, Morray et al., 2005). These criteria may have been liberal in that strong and clear indicators of each construct (i.e., items with high factor loadings) were retained along with items that were less clear indicators (i.e., items with loadings near .40). In fact, retained items varied considerably in the magnitude of their factor loadings (from .40 to .86). Also, scale development recommendations suggest elimination of items with loadings that are less than .15 higher than any cross-loadings to ensure retention of items that are distinctive indicators of their factors (Worthington & Whittaker, 2006). The absolute loading and cross-loading cutoffs used in the CFNI's development, however, suggest that items with smaller loading to cross-loading discrepancies may have been retained. Thus, the item-retention criteria used in the development of the CFNI resulted in subscales that varied substantially in length (from 7 to 18 items) and included strong and distinct indicators for each factor along with some relatively weaker and structurally more ambiguous indicators. It may be possible to create a more parsimonious instrument by trimming items that are weaker indicators of their intended constructs while retaining the psychometric strengths of the original CFNI.

In order to accomplish these objectives, we conducted confirmatory factor analyses (CFA) of the CFNI. CFA is useful for evaluating the structural stability of a measure once initial structure has been established using EFA (for overviews of CFA and related concepts, see Weston & Gore, 2006, and Worthington & Whitaker, 2006). To address the first aim of the study, we used CFA to examine data-model fit for the eight-factor structure of the CFNI compared with a nine-factor structure that separates the Sweet and Nice factor from the Relational factor. To address the second aim of the study, we used recommended criteria to retain the strongest items and reexamine data-model fit for the abbreviated CFNI.



Analyses were conducted using data from 243 undergraduate women from a Canadian university that serves a student body of approximately 25,000 and is located in a suburban area of a mid-sized city in central Canada. Participants’ ages ranged from 18 to 49 years (M = 20.12, SD = 4.43, Mdn = 19.0; two participants did not report their age). In terms of race/ethnicity, most participants identified as White (70%), and the remaining participants identified as Asian/Asian American (20%), Aboriginal/Native American (2%), Hispanic/Hispanic American (less than 1%), African/African American (less than 1%), biracial/multiracial (2%), or other race/ethnicity (4%); two participants did not identify their race/ethnicity. Sexual orientation was assessed using a free-response item allowing for flexibility in self-identification. Thirty-three participants chose not to disclose a sexual orientation or provided responses that could not be meaningfully coded due to apparent misunderstanding of the item (e.g., “in a long-term relationship”). Of those who did provide a codeable response, 1% identified as unsure or questioning, 1% identified as being attracted about equally to both sexes, 8% identified as being attracted to both sexes but primarily to men, and 91% identified as being attracted exclusively to men.


All participants completed the CFNI (Mahalik, Morray et al., 2005). The CFNI has 84 items that are rated on a 4-point scale ranging from 0 (strongly disagree) to 3 (strongly agree). Appropriate items were reverse scored. Subscale scores were calculated by summing responses for items on the subscale, and total CFNI scores were calculated by summing responses for all items. Although the conventional approach is to calculate CFNI subscale and scale scores by summing item ratings, we also present descriptive data for averaged item ratings to facilitate interpretation of scores across subscales with different numbers of items. As indicated in Table 1, means, standard deviations, and Cronbach's alphas obtained in our study were similar to those found by Mahalik, Morray et al. (2005) during construction of the CFNI. The present sample's subscale averages fell in the middle of the 0 to 3 continuum of possible scores, with Modesty (M = 1.44, SD = .34) and Nice in Relationships (M = 2.16, SD = .31) subscales anchoring the low and high end of subscale averages. Subscale intercorrelations in the present sample were generally low (.00 to .33, most in the .00s and .10s) and paralleled the range of correlations that Mahalik, Morray et al. (2005) obtained (−.01 to .41, most in the .00s and .10s).

Table 1. 
Descriptive Statistics and Bivariate Correlations for the 84-Item Conformity to Feminine Norms Inventory (CFNI) Eight-Factor and Nine-Factor Model Subscales
Subscale12345678910Current StudyMahalik et al. (2005)
MavgSDavgMsumSDsumα MsumSDsumα
  1. aThe Relational and Sweet and Nice subscales were computed separately in the nine-factor structure, and they were combined into the Nice in Relationships subscale in the eight-factor structure. Means and standard deviations with the subscript “avg” reflect values based on averaging item ratings and are offered to facilitate interpretation across subscales with varying numbers of items; means and standard deviations with the subscript “sum” reflect values based on summing item ratings and are offered because prior studies typically report summed scores for the CFNI.

  2. *p < .05. **p < .01.

 1. Relationala          2.13.3717.042.970.76   
 2. Sweet and Nicea0.45**         2.18.3621.773.590.72   
 3. Invest in Appearance0.15*0.01        1.91.5213.393.660.8112.013.750.82
 4. Domestic−       1.90.4715.173.790.8214.103.100.84
 5. Romantic Relationship0.110.120.28**0.13      1.50.4413.513.970.7915.643.730.77
 6. Modesty0.030.21**−0.020.08−0.19**     1.44.3412.963.070.7313.023.670.82
 7. Sexual Fidelity0.110.33**−*    2.01.6120.086.070.8920.225.550.85
 8. Thinness−**0.030.18**0.07−0.15*   1.69.6118.546.700.9020.876.180.90
 9. Care for Children0.22**0.33**−0.15*0.130.22**−0.010.33**−0.11  1.97.5423.616.450.9226.516.240.92
10. Nice in Relationshipsa0.82**0.88***0.15*0.27**0.000.33** 2.16.3138.805.600.8139.485.970.84
11. CFNI: Total0.43**0.59**0.31**0.34**0.49**0.26**0.52**0.40**0.58**0.60**1.86.21156.0618.010.87162.7318.260.88


A total of 250 women were recruited to participate in this study and received credit in their Introduction to Psychology course in exchange for participation. Before beginning the survey, participants received and signed informed consent forms indicating that they would be completing a questionnaire about femininity. Administrations took place with 15 to 25 participants at a time, in a classroom large enough to accommodate each participant at a separate table. To provide fair opportunity for earning course credit, men were offered a different questionnaire (see Parent & Moradi, 2009). After completing the survey, participants were debriefed and given contact information for the researchers, should they have had any follow-up questions about the study. Based on preliminary data screening (detailed below), one participant with substantial missing data and six others identified as multivariate outliers were removed from the data set, yielding a final sample size of 243.


Suitability for Confirmatory Factor Analysis

We first identified missing data points in our data set. Sixteen participants had missing data, and one of these cases was missing more than 20% of a CFNI subscale. This participant was dropped from the data set. For the remaining participants, 12 had one missing data point and 1 participant each had two, three, and four missing data points. We replaced missing data using the individual participant's mean for nonmissing subscale items. Given the low volume of missing data and the acceptable reliability of the CFNI subscales, casewise mean substitution was an acceptable approach for missing data replacement (Schafer & Graham, 2002).

Researchers have suggested that sample sizes of 200 are typically adequate for CFA (e.g., Kline, 2005; Quintana & Maxwell, 1999) and that complex models with greater degrees of freedom (such as the models in this study) require smaller sample sizes in order to achieve higher power than do more simple models (MacCallum, Browne, & Sugawara, 1996). Thus, we determined that the present sample size was sufficient to test the factor structure of the CFNI.

The data met guidelines for univariate normality (Weston & Gore, 2006). With regard to multivariate normality, six cases had Mahalanobis distances that were significant at p < .001, indicating that they were multivariate outliers (Tabachnick & Fidell, 1996). Thus, we removed these cases from the data set. With the removal of these six cases, and the one participant with substantial missing data, our final data set contained 243 participants.


CFA was conducted using maximum likelihood estimation. We followed Kline's (2005) and Worthington and Whittaker's (2006) recommendations regarding fit index reporting for model evaluation. As such, we present the chi-square statistic with degrees of freedom, the mean residual square error of approximation (RMSEA) with 90% confidence interval (CI), the standardized root mean square residual (SRMR), and the comparative fit index (CFI). For sample sizes of less than 500, RMSEA and SRMR values below .10 and CFI values above .90 are typically interpreted as indicative of good fit (Weston & Gore, 2006), but it is important to emphasize that these values are conventions. Numerous researchers have strongly averred that rigid cutoff values should not be adopted across all models and all situations. Indeed, the use of cutoffs as “golden rules” has been argued against by psychometricians (e.g., Marsh, Hau, & Wen, 2004). Thus, we present conventional guidelines for evaluating model fit, without claim that any single indicator justifies the categorization of a model as having good or poor fit.

CFA of the CFNI

We first compared the fit of the original eight-factor model of the CFNI with a nine-factor model that split the Nice in Relationships items into two factors that reflected being nice to others (Sweet and Nice) and the primacy of maintaining friendships (Relational). In both the eight-factor and the nine-factor models, items were constrained to load onto their intended factor and factors were allowed to intercorrelate. A test of the eight-factor model yielded mixed results, with RMSEA and SRMR suggesting acceptable fit, but CFI suggesting poor fit, χ2(3,374, N = 243) = 5,359.84, p < .001, RMSEA = .049, 90% CI: .047, .052, SRMR = .0743, CFI = .75. We compared this to the nine-factor model, which demonstrated marginally superior fit, χ2(3,366, N = 243) = 5,255.28, p < .001, RMSEA = .048, 90% CI: .046, .051, SRMR = .0732, CFI = .77. All factor loadings were significant. Thus, for both models, RMSEA and SRMR were within acceptable ranges, but CFI was lower than the conventional cutoff. Factor loadings for the eight-factor model are presented in Table 2, and loadings for the separated Sweet and Nice and Relational factors in the nine-factor version are presented in Table 3 (loadings on all other factors in the nine-factor model remained the same as those in the eight-factor model). Latent variable intercorrelations are presented in Table 4 and suggest low to moderate relations among CFNI factors.

Table 2. 
Eight-Factor Solution for the 84-Item Conformity to Feminine Norms Inventory (CFNI), Proposed by Mahalik et al. (2005)
ItemMahalik et al. LoadingCurrent StudyItemMahalik et al. LoadingCurrent StudyItemMahalik et al. LoadingCurrent Study
  1. Note. NiceRel = Nice in Relationships; Appearance = Invest in Appearance; RomRela = Romantic Relationship; SexFid = Sexual Fidelity; Child = Care for Children. For interpretability, item numbers reflect numbering on the 84-item CFNI.

  2. Fit Indexes: χ2(3,374, N = 243) = 5,359.84, p < .001, RMSEA = .049, 90% CI: .047, .052, SRMR = .0743, CFI = .75 (see text for definitions of statistical abbreviations).

Table 3. 
Sweet and Nice Factor and Relational Factor Loadings in the Nine-Factor Solution for the 84-Item Conformity to Feminine Norms Inventory
ItemNew SubscaleFactor LoadingUniqueness
  1. Note. All other factors were the same as those in the eight-factor model.

  2. χ2(3,366, N = 243) = 5,255.28, p < .001, RMSEA = .048, 90% CI: .046, .051, SRMR = .0732, CFI = .77 (see text for definitions of statistical abbreviations).

NiceRel33Sweet and Nice0.610.63
NiceRel53Sweet and Nice0.600.64
NiceRel84Sweet and Nice0.550.70
NiceRel69Sweet and Nice0.520.73
NiceRel75Sweet and Nice0.490.76
NiceRel13Sweet and Nice0.480.77
NiceRel79Sweet and Nice0.480.77
NiceRel38Sweet and Nice0.460.79
NiceRel63Sweet and Nice0.430.82
NiceRel1Sweet and Nice0.360.87
Table 4. 
Latent Factor Correlations
  1. Note. Latent factor correlations below the diagonal are for the Conformity to Feminine Norms Inventory (CFNI) 84-item version, with row 10 reflecting the Nice in Relationships factor in the eight-factor model. Correlations above the diagonal are for the nine-factor CFNI-45, with columns 1 and 2 reflecting the Relational and Sweet and Nice factors in the nine-factor model.

  2. *p < .05. **p < .01.

 1. Relational0.44**0.22*0.050.12−0.090.05−0.010.21*
 2. Sweet and Nice0.58***0.020.38**−0.030.39**
 3. Appearance0.19***−0.02−0.110.14−0.15*
 4. Domestic0.**−0.050.06−0.020.21**
 5. Romantic Relationship0.120.160.32**0.20*−0.18*0.040.21**0.19*
 6. Modesty−0.010.16−0.030.04−0.23*−
 7. Sexual Fidelity0.140.39**−−0.16*0.32**
 8. Thinness−0.02−0.010.21**0.010.22**0.10−0.14*−0.05
 9. Care for Children0.27**0.40**−0.140.17*0.23**0.020.36**−0.08
10. Nice in Relationships0.**−0.020.40**

Several criteria pointed to the superior interpretability of a nine-factor model. First, we examined the chi-square difference test. This test indicated that the nine-factor model was a significantly better fit to the data than the eight-factor model, χ2Diff(8) = 104.56, p < .001. There is disagreement in the literature, however, regarding whether models with different numbers of latent variables are classifiable as nested, and thus whether the chi-square difference test is a legitimate test of model improvement (Brown, 2006). As such, we also present the Akaike Information Criterion (AIC) and the Expected Cross-Validation Index (ECVI) to compare models. These fit indexes account for both data-model fit and model parsimony (Brown, 2006) and can be used to compare competing similar but non-nested models. Lower scores on the AIC and ECVI indicate a superior model. On these criteria, the nine-factor model (AIC = 5,663.28, ECVI = 23.40, 90% CI: 22.61, 24.23) demonstrated marginal superiority over the eight-factor model (AIC = 5,751.84, EVCI = 23.77, 90% CI: 22.96, 24.61).

Strength of factor loadings supported the nine-factor model as well. Specifically, underspecification of the number of factors can lead to relatively weaker factor loadings (Brown, 2006). Indeed, in the eight-factor model, factor loadings for Nice in Relationships items were relatively weaker than factor loadings for items on other subscales (see Table 2), with 15 of 18 loadings falling below .50. In contrast, in the nine-factor model, 9 of the 18 factor loadings for the separate Sweet and Nice factor and Relational factor surpassed .50.

Finally, from a conceptual standpoint, the nine-factor model was justifiable because two separate themes seemed to be present in Nice in Relationships items. Specifically, items on the Relational factor (e.g., “I make it a point to get together with my friends regularly”) focused on the value placed on, and effort devoted to, friendships, making them conceptually and semantically different from items on the Sweet and Nice factor (e.g., “I always try to make people feel special”), all of which pertained to being nice without mention of friendships. Thus, both empirical and conceptual evidence supported interpretation of a nine-factor model for the CFNI, which is also consistent with the originally intended separate factors for (a) Sweet and Nice and (b) Relational. As such, we adopted the nine-factor model in moving forward with length optimization.

Examining a Short Form of the CFNI

We examined the potential for developing an abbreviated version of the CFNI using the strongest latent variable indicators. Specifically, we examined factor loadings as a criterion for item retention. Factor loadings represent the relative contribution of each item to explaining its latent variable. Thus, our first step was to trim items that contributed relatively less to the explanation of their intended latent variable or factor. To this end, we considered the five highest-loading items on each factor for retention. A limit of five items per subscale satisfies recommendations for minimum numbers of indicators per factor (three items per factor is a typical minimum; see Worthington & Whittaker, 2006) while keeping Cronbach's alpha reliability coefficients, which tend to increase as the number of items increases, within acceptable ranges. Through this process, we eliminated 39 items to form the CFNI-45.

To evaluate the stability of our factor loadings and resultant item elimination decisions, we compared our factor loadings against the loadings obtained by Mahalik, Morray et al. (2005; details obtained from J. R. Mahalik, personal communication, January, 26, 2009). Some variability in factor loadings is expected across our CFA of the 84-item CFNI and Mahalik, Morray et al.'s (2005) EFA, because the factor loadings that Mahalik, Morray et al. (2005) reported were from an EFA of 118 items (that is, after items for the unsupported factors were deleted from the initial 144 item pool, but before low-loading items for the eight retained factors were deleted). As such, factor loadings for the final 84 CFNI items might have shifted slightly after Mahalik, Morray et al. (2005) eliminated poorer items. Nevertheless, consistency between the rank order of items based on Mahalik, Morray et al.'s (2005) loadings and our loadings would provide additional support for our item elimination decisions to form the CFNI-45. As indicated in Table 2, there was substantial correspondence across the two samples in the within-subscale rank order of items based on loading magnitudes, and items eliminated based on the present CFA were largely those with the lowest loadings in Mahalik, Morray et al.'s (2005) EFA.

Using the abbreviated 45-item measure, we again examined the fit of the nine-factor model to the data. The CFNI-45 nine-factor model surpassed or approximated cutoffs for acceptable fit on all indicators, χ2(909, N = 243) = 1,384.55, p < .001, RMSEA = .044, 90% CI: .042, .051, SRMR = .0647, CFI = .89. All factor loadings for this model were significant. Factor loadings for the CFNI-45 are presented in Table 6. Latent variable covariances are presented in Table 4 and suggest low to moderate relations among CFNI factors. Cronbach's alphas for subscale items were close to the values for the original subscale items, and they ranged from .68 to .89, falling in the fair-to-excellent ranges according to Ponterotto and Ruckdeschel's (2007) matrix for interpreting Cronbach's alpha. Also, correlations between the original and abbreviated subscale scores ranged from .87 to .97 (see Table 5), indicating substantial consistency in the constructs measured by the original 84-item CFNI and the CFNI-45.

Table 6. 
Factor Loadings for the Conformity to Feminine Norms Inventory (CFNI)-45
ItemFactor LoadingUniquenessItemFactor LoadingUniquenessItemFactor LoadingUniqueness
  1. Note. NiceRel = Former Nice in Relationships items; Appearance = Investment in Appearance; RomRela = Romantic Relationships; SexFid = Sexual Fidelity; Child = Involvement with Children. (R) = Item was constrained to new Relational subscale. (S) = Item was constrained to new Sweet and Nice subscale. All factor loadings are significant at p < .001. For interpretability, item numbers reflect numbering on the 84-item CFNI.

  2. Fit Indexes: χ2(909, N = 243) = 1,384.55, p < .001, RMSEA = .044, 90% CI: .042, .051, SRMR = .0647, CFI = .89 (see text for definitions of statistical abbreviations).

NiceRel35 (R)0.600.64Domestic70.740.45SexFid470.800.36
NiceRel48 (R)0.570.67Domestic160.740.45SexFid290.770.40
NiceRel44 (R)0.560.68Domestic510.690.52SexFid390.750.43
NiceRel18 (R)0.550.70Domestic680.690.53SexFid780.750.43
NiceRel82 (R)0.510.74Domestic320.590.65SexFid650.720.48
NiceRel53 (S)0.660.57RomRela310.810.35Thinness370.850.28
NiceRel84 (S)0.590.66RomRela500.790.38Thinness30.840.29
NiceRel33 (S)0.580.66RomRela740.540.71Thinness640.830.31
NiceRel69 (S)0.500.75RomRela580.520.73Thinness770.690.53
NiceRe175 (S)0.440.80RomRela670.480.77Thinnes710.660.57
Table 5. 
Descriptive Statistics and Bivariate Correlations for the Conformity to Feminine Norms Inventory (CFNI)-45
  1. aCorrelations between corresponding CFNI-45 and original form subscales.

  2. *p < .05. **p < .01.

 1. Relational         2.02.4610.112.300.690.95
 2. Sweet and Nice0.31**        2.23.4311.152.150.680.92
 3. Invest in Appearance0.18**−0.02       1.84.60
 4. Domestic0.030.090.04      2.11.5110.542.540.820.94
 5. Romantic Relationship0.13*0.14*0.27**0.18**     1.68.55 8.422.740.760.94
 6. Modesty−−0.04−0.14*    1.32.38 6.611.900.720.87
 7. Sexual Fidelity0.050.29**−   1.75.78 8.733.900.870.96
 8. Thinness0.000.000.18**−0.010.17**0.12−0.15*  1.70.75 8.493.750.880.95
 9. Care for Children0.16*0.31**−0.150.18**0.23**0.040.27**−0.06 2.08.5910.422.960.890.96
10. CFNI-45: Total0.42**0.51**0.38**0.38**0.55**0.18**0.46**0.39**0.51**1.86.2483.6810.81 0.790.96


Conformity to cultural norms of femininity is posited to play an important role in women's lives across a range of domains, including mental health, relationships, and work (e.g., Brown & Brodsky, 1992; Philpot, Brooks, Lusterman, & Nutt, 2002; Worell & Johnson, 2004). Also, in a societal context that construes power hierarchically, prescriptive feminine norms serve to constrain and disempower women. Thus, assessment of conformity to such norms is important for feminist research and practice with women. Researchers have studied conformity to feminine norms by using measures that approximate, but do not directly assess, this construct (e.g., measures of instrumentality and expressiveness, measures of attitudes about appropriate rights and roles for women) or by using measures that focus on specific norms without considering them as part of a broader set of interconnected norms that reflect societal construal of femininity. Thus, the CFNI represents an important advancement in research on women's experiences and well-being, because it assesses conformity to feminine norms directly and multidimensionally. This study contributes to the literature on women's conformity to feminine norms by providing a psychometric evaluation and abbreviation of the CFNI. Overall, our results indicated some support for the CFNI, but also pointed to concerns about factor structure. We used our data to inform factor realignment and item reduction, leading to the development of the more parsimonious nine-factor, 45-item CFNI-45.

The present data were consistent with the overall multidimensional structure of the CFNI as proposed by Mahalik, Morray et al. (2005), but also highlighted potential problems with the Nice in Relationships factor. Specifically, items for this factor were originally intended to measure two separate constructs related to (a) niceness and (b) the importance of having and maintaining friendships; but these two constructs were collapsed into a single factor in the eight-factor model (Mahalik, Morray et al., 2005). This study's comparison of the eight-factor model with a nine-factor model in which Relational and Sweet and Nice factors were separated suggested the superiority of the nine-factor model. Specifically, the chi-square difference test, AIC, ECVI, and improvements observed in factor loading ranges offered empirical support for the nine-factor model over the eight-factor model; but neither model met all fit index criteria for acceptable fit. Furthermore, conceptual analysis of the subscale items indicated clear differentiation between Relational items and Sweet and Nice items. Thus, the nine-factor model had empirical and conceptual advantages over the eight-factor model, and item reduction to develop the CFNI-45 proceeded using the nine-factor model.

The CFNI-45 is approximately half the length of the original 84-item CFNI, yet it retains all nine factors. Reliability coefficients for CFNI-45 subscale items were comparable to the original subscale item reliabilities. This consistency is noteworthy given that Cronbach's alpha values are associated positively with number of subscale items (e.g., Ponterotto & Ruckdeschel, 2007). Fit indices for the CFNI-45 provided tentative support for the nine-factor structure, and the data-model fit for the CFNI-45 was superior to that of the eight- and nine-factor original forms of the CFNI. However, it is important to highlight that, while the CFNI-45 absolute fit indices (i.e., RMSEA, SRMR) met criteria for acceptable fit, CFI did not. The large number of indicators in the model should be considered in interpreting the relatively poorer CFI compared to RMSEA and SRMR values because simulation studies have found that, in correctly specified models, CFI tends to decline whereas absolute fit indices tend to improve with increases in the number of variables in the model (Kenny & McCoach, 2003). Still, the stability of the nine-factor solution for CFNI-45 data should be examined across other samples.

Furthermore, correlations between corresponding CFNI-45 and original CFNI subscales were in the high .80s and .90s, indicating substantial consistency in the underlying constructs assessed by the two versions. Generally low latent factor and subscale intercorrelations for the CFNI-45 paralleled those observed for the original CFNI and are consistent with a multidimensional conceptualization of feminine norms. Also, at the subscale level, the items deleted to form the CFNI-45 (i.e., low loaders in our data) corresponded well with low-loading items in the original CFNI development study, suggesting stability in CFNI factor structure across the two samples. These correspondences between the two forms suggest tentatively that validity evidence accumulated for the original CFNI may be found to be applicable to the CFNI-45. Still, further research is needed to evaluate the discriminant and convergent validity of CFNI-45 scores across diverse samples. Such research could test the independence of CFNI-45 scores from impression management and their convergence with other measures of feminine norms.

Another important avenue for further research on the validity of the CFNI-45 (and its original from) is to test the cultural generalizability of the norms captured by the measure; this approach is particularly important because available research with the CFNI-45 has been conducted primarily with respondents from the majority culture. To address this limitation, cross-cultural research on the content of feminine norms could reveal some norms that are applicable across cultures and other norms that are culture specific (e.g., Adler, 1993; Williams & Best, 1990). For example, Williams and Best (1990) found cross-country variability in the perceived valence and differentiation of masculine and feminine characteristics and even found that characteristics viewed as masculine in some countries were viewed as feminine in other countries (e.g., “arrogant” was perceived as a male-associated in most countries but as female-associated in Nigeria). Such cross-cultural research on the content of gender norms is particularly important in light of evidence that items on gender role–related measures vary in their meaningfulness across cultural groups (Gibbons, Hamby, & Dennis, 1997). Similarly, cross-cultural research that tests the structural properties of the CFNI-45 and evaluates relations of CFNI-45 subscale scores with other variables can reveal whether the feminine norms captured in the measure are related to each other and to other variables similarly across cultures (both within North American cultural groups and cross-national cultural groups). Longitudinal research with immigrant populations whose cultural gender norms diverge from those of the cultural majority could reveal developmental processes in acquisition of majority cultural femininity norms. Such research on the validity of the conceptual framework underlying the CFNI-45 across diverse populations is important for elucidating how women of diverse backgrounds experience feminine norms as well as for identifying culture-specific feminine norms that could be incorporated to broaden the utility of the CFNI-45 across groups.

We hope that the CFNI-45 will be employed as an efficient tool for researchers who wish to examine women's conformity to feminine gender norms. Specifically, the CFNI-45 can help to facilitate assessment of feminine gender-role adherence. In the past, researchers interested in women's feminine gender-role conformity have relied on instruments that approximate gender-role conformity or developed study-specific instruments. Use of the CFNI-45 can advance the literature by allowing researchers to assess women's feminine gender-role conformity directly rather than relying on proxy measures. Thus, use of the CFNI-45 can facilitate research on the role of women's conformity to feminine norms in many domains, including career decision making, responses to violence, reproductive and parenting choices, interpersonal relationships, mental health, and well-being. The shorter length of the CFNI-45 also makes it more suitable for inclusion in survey studies where multiple measures may be administered and participant fatigue or boredom is a concern.

The CFNI-45 also might be useful in counseling and therapy to attend to the potential roles of conformity to feminine norms in women's lives and presenting concerns. Indeed, the APA's Guidelines suggest that women's gender role socialization can have deleterious implications for women's mental health, task performance, and aspirations, and that enactment of gender roles may occur beneath awareness. The Guidelines highlight the utility of directly assessing and bringing to light conformity to feminine norms and the lower societal power and privilege associated with these norms (relative to masculine norms, in a societal context that construes power hierarchically). Indeed, empirical evidence indicates that attention to gender role socialization is an integral aspect of feminist therapists’ work with women (Moradi et al., 2000). In fact, Moradi et al. (2000, p. 289) found that “Reframing clients’ definitions of their problems to include the impact of socialization” was one of the top three most-often used therapy behaviors for self-identified feminist therapists when working with women clients. This emphasis was also present in Chester and Bretherton's (2001, p. 539) survey of Australian feminist-identified therapists, who rated “Has knowledge of effects of sex-role stereotyping on women” and “Encourages client to free herself from sex-role stereotypes” as essential to being a feminist therapist. The CFNI-45 may facilitate implementation of this critical aspect of the Guidelines and feminist therapy theory and practice.

Specifically, the CFNI-45 can serve as a brief and efficient tool to assess and discuss the content of feminine gender norms and their internalization with clients. For example, the CFNI-45 can be introduced to the client as a tool to measure the degree to which each of the norms is salient for the client, and to explore the potential costs and benefits that maintaining or changing patterns of feminine-norm conformity might have for the client. (This approach is similar to Mahalik, Talmadge, Locke, and Scott's [2005] suggestions for use of the CMNI with men in counseling.) For example, at the individual level, strong adherence to the Relational norm may have the benefit of maintaining a strong social network of friends, but preoccupation with the dissolution of that network may also be exhausting and interfere with other aspects of functioning. Similarly, strong adherence to the Modesty norm may be adaptive in some social groups, but may interfere with negotiating one's worth in workplace settings. At the societal level, prescribing relational and modesty norms to women and interpersonal dominance and self-reliance norms to men can serve to disempower women. Discussing the meaning and implications of CFNI scores with the client in this manner can help to contextualize interpretations and empower the client to consider the potential benefits of flexibility in enacting feminine norms across contexts. Using such procedures, the CFNI-45 might facilitate analysis of women's feminine-norm adherence in therapy and educational interventions, and it may be employed to investigate the progress and outcomes of such interventions, making such uses of the CFNI-45 worthy of empirical investigation.

Although the present findings point to the promise of the CFNI-45, we also acknowledge the limitations of the present study, and we hope that future research can build on our findings. First, as noted previously, CFI values did not rise to the conventional cutoff of .90 to indicate good fit, missing that cutoff by .01 for the nine-factor CFNI-45. However, the CFI of the short form model was an improvement over that of the original model, and other indicators of fit improved for the CFNI-45 as well. Thus, further attention to the structural properties of the CFNI-45 is warranted. Also, participants in the present study were predominantly White undergraduates from a single university in Canada, and thus the results of the present study are not immediately generalizable to other populations. We hope that the CFNI-45, with its abbreviated length, will facilitate examination of the psychometric properties of the CFNI with women of diverse age, racial/ethnic, sexual orientation, socioeconomic, and other backgrounds. Also, as stated previously, it is important to consider that feminine norms are culturally bound constructs, and exploration of cross-cultural stability and variability of these norms is needed. Similarly, feminine norms evolve across time. Thus, attention to shifts in feminine norms across cohorts and time periods is needed. The CFNI-45 can be used as a basis for such investigations and as a base for future modifications to capture the evolving nature of feminine norms. As additional norms are identified across time and cultures, or as current norms recede, these temporal and cultural variations can be incorporated into the CFNI. This modular approach to the CFNI also can reduce the need for a multitude of study-specific instruments measuring similar constructs.