**4. Results**

#### **4.1 Analytic approach**

Data analysis took place in two distinct phases. Due to the number of variables present in the stereotype matrix, we conducted a series of exploratory factor analyses (EFAs), paring down the number of variables to more manageable factor scores. These factor scores were then used to chart differences among clusters of variables found in the data. Once factor scores were calculated, a multivariate analysis of variance (MANOVA) was conducted to determine how factor scores from the stereotype matrix differed based on the race and sexual orientation of the hero. A principal component analysis was conducted on the factors and utilized a direct oblique rotation with Kaiser Normalization due to the perceived interrelatedness of the stereotype variables.

The stereotype matrix asked participants about 71 different traits that the person in the cape may demonstrate, each of which was rated on a seven-point Likert scale ranging from Strongly Disagree to Strongly Agree. During initial EFA tests, physical traits – such as blonde hair, dark skin, and dark hair – were dominating the factor

structure and clouding interpretation. Thus, they were removed from further analyses. Then, through an iterative process, the number of factors was reduced from 71 variables to 35 variables. This iterative process involved removing significantly crossloading variables individually and subsequently rerunning the model to see how the factor structure was affected. This process was repeated until there were no significant cross-loadings, and the factor structure was easily interpretable. The 35 variables supported a four-factor structure. To further support the use of a four-factor structure, a parallel analysis was conducted utilizing Patil, Surendra, Sanjay, and Donavan's [50] web-based engine. Comparing the mean eigenvalues of the web-based parallel analysis to the total eigenvalues of the SPSS analysis, a four-factor structure was further supported.

The four-factor structure possessed some minor cross-loadings but was deemed easily interpretable. The four factors were as follows: Positive Traits, Machismo Traits, Social Status Traits, and Socially Undesirable Traits. Variables that loaded onto each factor were calculated into a single factor score. These factor scores indicate the degree to which participants ascribed stereotypes of that factor to the hero image. Positive scores indicate that participants believe the hero possesses a trait, with higher scores being indicative of greater trait ascription (i.e. strongly agree that a hero possesses this trait). Negative scores indicate that participants believe that the trait is not true of the hero, with more extreme scores being indicative of greater trait denial (i.e. strongly disagree that a hero possesses this trait).

A MANOVA was conducted examining the race of the hero (White Black Latinx Arab Native American Asian) and the sexual orientation of the hero (Heterosexual Gay) in the context of the four factors outlined previously: positive traits, machismo traits, socially undesirable traits, and social status traits. A test for equality of variances, Box's M (100, 122861.79) = 1.656, *p* < = .001, indicated that assumptions of the normality of the data were violated. However, the MANOVA is considered robust to this assumption as long as group sizes are greater than 30 [51]. Thus, the factorial MANOVA was considered cautiously interpretable when using Pillai's Trace due to its robustness, especially when dealing with unequal sample sizes.

Pillai's Trace for each of the IV's indicated differences based on the race of the hero (*p* < .001, partial η<sup>2</sup> = .039) and no significant difference based on the sexuality of the hero (*p* = .842, partial η<sup>2</sup> = .004). However, there was indication of an interaction between the hero's race and sexuality (*p* = .011, partial η<sup>2</sup> = .022). Due to the statistically significant results of the MANOVA, additional interpretation of the main effects and interaction effects were warranted. The race of the hero supported differences in trait ascription with positive traits (*p* = .017, partial η<sup>2</sup> = .037), machismo traits (*p* = .012, partial η<sup>2</sup> = .040), and socially undesirable traits (*p* < .001, partial η<sup>2</sup> = .066). The interaction of the hero's race and sexuality supported differences in trait ascription in regard to positive traits (*p* = .050, partial η<sup>2</sup> = .026) and social status traits (*p* = .035, partial η<sup>2</sup> = .028).

With regard to race, *post hoc* examinations did not reveal statistically significant differences among the races of hero when it came to positive traits, though the White hero was trending toward receiving more negative beliefs about possessing positive traits in comparison to the Black hero (Tukey's HSD *p* = .083). In terms of machismo, the White hero was considered to have more machismo-esque traits than the Asian hero (*p* = .03). In terms of socially undesirable traits, the Arab hero was considered to have more undesirable traits than the Asian (*p* = .014), Black (*p* = .001), and Native American heroes (*p* = .001).

To get an idea differences along participant identities, the data file was split across a number of participant demographics. The original factorial MANOVA was

conducted again within split subsets of the participant demographics. The MANOVAs were conducted based on race (White Latinx Black Other Races), gender (Male Female), and sexual orientation (Heterosexual/Heterosexual Non-Heterosexual). While the research team recognizes that the experiences of each of these identities and their intersections are extremely varied, and that the current analyses may imply a sense of homogeneity, these groups were formed based on the number of participants for each identity being analyzed. Results of these analyses should be used to offer insight into future studies as opposed to being interpretable in their current state.

Participant race seems to contribute to differences in hero interpretation based on examination of Pillai's Trace. Although most of the participant data did not retain enough power for significance, we did see that White participants significantly differed in how they ascribed hero traits based on hero race (*p* = .008, partial η<sup>2</sup> = .055). Latinx participants did not differ significantly in how they ascribed traits based on hero race or sexuality. Black participants also did not demonstrate differences in trait ascription based on hero race or sexuality. Participants of other racial groups in aggregate also did not differ in how they ascribed traits based on hero race and sexuality.

Men did not demonstrate significant differences in how they ascribed traits based on examination of Pillai's Trace. However, women were significantly different in how they ascribed hero traits based on hero race (*p* = .001, partial η<sup>2</sup> = .050). Though they did not differ in the ascription of traits based on hero sexuality, there were differences among women in terms of how hero race and sexuality interacted to form their trait ascriptions (*p* = .029, partial η<sup>2</sup> = .031).

Heterosexual/heterosexual participants demonstrated differences in how they ascribed traits on the basis of hero race (*p* < .001, partial η<sup>2</sup> = .041). However, nonheterosexual participants did not demonstrate significant differences in how traits were ascribed to a hero on the basis of race or sexuality.

## **5. Discussion**

The present study had three main goals: (1) to test the hypothesis that participants ascribe stereotypical characteristics to original, fictional characters based on race and sexuality; (2) to examine whether utilizing images developed for research, rather than proprietary images, would be a feasible methodology for testing for implicit biases and stereotype ascription; and (3) to seek evidence of ongoing racial and sexuality stereotypes within the current zeitgeist. The current study offers mixed support for the first goal, strong support for the second, and moderate support for the third.

The results indicate that a relatively diverse sample of participants significantly differed in how stereotypical traits were ascribed to heroes based on race. While trait ascription did not vastly differ as a result of hero sexuality, intersections of race and sex did contribute to differences in how people saw the hero. In this study, the White and Arab heroes – specifically the gay Arab hero – received the greatest endorsement of negative traits. While it is not within the purview of this data to understand the exact causality of this difference, initial hypotheses may be that the White hero was something of an acceptable target, with participants being more willing to ascribe the negative traits to the White image. Furthermore, the Arab hero having a higher endorsement of undesirable traits compared to other heroes of color may lie in the overwhelming portrayal of Arab characters in media as villainous. There also appears to be a significant interaction in the identities of the gay Arab hero. Future studies

should explore these differences further. Thus, our initial assumption that participants would differ in their perceptions of an original fictional character based on race and sexuality were moderately supported by the results.

The images of the heroes (see Appendix C, **Figures 1C**–**6C**) were created entirely in-house, offering a wholly original, nonproprietary tool in the examination of media effects. The results of this study indicate that such images can be used effectively to gauge how individuals may ascribe stereotypical traits and demonstrate implicit biases in their responses to free-response items. Due to the novel nature of these images, they were devoid of the history, context, and other potential confounds that more well-known comic book images may fall prey to. Thus, original images such as the images utilized in this study have great utility for future research, suggesting that goal two is strongly supported.

As for exploration of racial and sexual stereotypes, this study offers a degree of support that participants ascribe traits differently based on hero race and race × sexual identity intersections. This is increasingly important as the current media landscape moves towards the inclusion of intersectionality within media portrayal. With the increased presence of intersecting identities in media, it will be crucial for researchers to gauge the dynamic differences in consumer perceptions. Future work can further explore these differences with clear factor structures and deliberate study design.

## **6. Implications of current study**

Previous research has shown that media influences can affect self-perceptions. The success of the comic book industry clearly indicates the popularity of this genre within mainstream media entertainment. This critical not only for adults but also for the early impacts that they may have on children. The present study demonstrates that there are underlying imagery trends that should be taken into consideration when these heroes are created or portrayed through movies, television, or print. Through the application of stereotypical attributes to the superhero images, we could potentially see an effect on self-perception. This result should stand as a caution to creators and consumers as the presentation of superheroes of color are being introduced into the mainstream comic book media.

As the comic book media superhero genre [1, 8] continues to grow, it is important for researchers to incorporate these images into the evaluations of media. The present research supports previous work that shows the prevalence of stereotypes in media [34, 35, 39], as well as specifically in comic book media [29, 52]. Therefore, it is vital to evaluate these comic book heroes from a critical race and sexual orientation perspective. The findings of this study may also be used in furthering efforts to create appropriate diversity and equality among the comic book superhero genre.

## **7. Future directions**

Future studies should consider introducing heroes with intersectional race, sex, and sexual orientations. The intersections of these identities may influence stereotype ascription. Future studies may also want to consider utilizing free-response methodology for stereotype ascription as in Niemann et al. [49]. That may allow for further insights into the current state of stereotype ascription in American society. By

pursuing other methods of inquiry within the realm of stereotypes and media effects, we can achieve a greater understanding of the myriad influences that comic books and related media have on how we view others and ourselves. We plan to build upon this study and are willing to work alongside any scholars interested in the use of our superhero imagery.
