Female Disempowerment Disguised as a Halloween Costume

Objective :We explore the relationship between gender stereotypes and North American Halloween costumes. Method (Study 1): Extending Nelson's analysis of gender-markers in mass-produced children’s Halloween costumes, Study 1 explored gender-typing in children’s costumes (n = 428), also adding a sample of adult’s costumes (n = 428) from major retailers, coding for character archetypes (heroes, villains, and fools), active-masculinity/passive-femininity, and for degree of disguise. Results (Study 1): Compared to boys’/men’s costumes, girls’/women’s costumes represented more ornamental feminine-passivity. Method (Study 2): Ornamental feminine-passivity was explored in an additional sample of baby girls’ (n = 161), child girls’ (n = 189), teen girls’ (n = 167), and women’s (n = 301) costumes, coded for character archetypes and markers of infantilization and sexualization. Results (Study 2): In addition to age differences in character archetypes, women’s costumes were most likely to be sexualized (especially heroes), girls’ and teenage young women’s costumes were most likely to combine both infantilization and sexualization, and baby girls’ costumes were least likely to incorporate either gender-markers. Conclusion : Costumes reinforce gender stereotypes differentiating boys/men and girls/women and the ways in which girls/women are stereotyped varies across the lifespan. Patterns are discussed with regard to how gender stereotypes embedded in holiday traditions reinforce messages of disempowerment for women and girls.


INTRODUCTION
Holidays serve to bind a culture together, maintain heritage, and reinforce traditional values that often include stereotypic gender roles, as well [1].With the growth of consumerism, holidays have also evolved into "industries" with mass marketed specialty merchandise [2] For Halloween in North America, commercially-produced costumes account for the largest share of such holiday consumer spending, and the child-oriented costume market has been dramatically expanding to meet growing demand for adult-sized costumes [3,4] Children and adults alike may face pressures to abide by gender role expectations in their costume choices [5].Our research explores gender typing in Halloween costumes marketed to both children and adults, recognizing gender socialization pressures as a lifespan issue [6,7].
Amongst the scant academic studies of Halloween costuming and gender, Ogletree, Denton, and Williams attributed children's gender typed costume choices to their internalized gender role schemas [8].However research suggests there are external influences on gender typed choices such as product advertisements and even characteristics of the manufactured items themselves [9].Gender-targeted products inform consumers of traits and values that are expected of their gender, create gendered demand, and encourage consumers to be attracted to the product itself to fulfill those expectations.The notion that inanimate commercial objects contribute to gender socialization has been recently highlighted in a report by the American Psychological Association [10], implicating clothes, dolls, toys, and other gender-targeted merchandise as contributing to the sexualization of girls.Despite ongoing debate over which came firstgender-typed merchandise or consumer demand for it-data has been accumulating to show that commercially produced products serve as proximal influences on consumer attitudes and demand [11,12].
Arguing that "the commercial marketplace plays a major role" in children's costuming choices (p.138), Nelson examined gendered messages on package labels of over 400 ready-to-wear costumes and costume patterns sold in North America [13].When she collected her sample (the 1990's), costumes were not gender-segregated by merchandisers on the sales floor, so Nelson identified implicit gender markers used by marketers as a proxy for explicitly gender-labeled, gender-targeted packaging (e.g.pastel vs. primary colors, fashion vs. athletic shoes, etc.).Nelson then used Klapp's [14] symbolic interactionist model of character archetypes to classify costumes into masculine and feminine versions of the good (heroes), the bad (villains), and the silly (fools).Merchandisers exploit such familiar archetypes to generate consumer demand for their products [15].Nelson classified heroes into three subtypes: conventional (e.g.warriors, princesses), superheroes who possess supernatural powers (e.g.comic book super-humans), and exemplars of prosocial conformity (e.g.pioneer boy, Puritan girl).She classified villains into three types: symbols of Death (e.g.ghosts), monsters (e.g.werewolves), and anti-heroes (e.g.pirates, witches).She classified fools who conjure perceptions of simplicity into two types: humans (e.g.clowns) or non-humans (e.g.objects or animals).Nelson's coding revealed costumes marketed to boys to be more evenly dispersed across these eight possible costume character archetypes (see Table 1).In contrast, the majority of costumes marketed to girls (over 70%) were concentrated in just two of the eight categories: conventional heroes (i.e. in the feminine form of beauty queens and princesses) and non-human fools (e.g.baby animals, bugs).She also informally observed that boys' costumes portrayed a range of "active-masculine" traits involving agency; in contrast, girls' costumes emphasized "passive-feminine" traits involving passive display of beauty and winsome charm, what she also refers to as "ornamental feminine passivity."Such gendered themes have been recognized variously in feminist social science as resulting in an association of boys and men with activity and agency and girls and women with passivity and display [16,17].Macmillan et al. noted that males are socialized toward an agonic ideal of power through physical expressions, accomplishment, and power over others [5].In contrast, women are socialized toward indirect power through attracting attention from others.Kirsch and Murnen [18] suggest this is a common cultural pattern of boys objectifying girls for their appearance and girls reciprocating with self-objectification, so that "girls treat themselves as an object; they learn to be observed and recognize the utility in their appearance" (p.20).
Contemporary Halloween masquerade is a lighthearted vestige of its historic more serious purpose: in mediaeval times, a successful costume served to disguise the wearer in order to protect them from evil spirits lurking on October 31 st [9].Today's more playful function of costuming is to disguise the wearer's usual way of being as a liberation from their regular self [19].The creative freedom that comes from masquerade is an integral part of costume play, and some argue it can provide a rare opportunity to be freed from everyday practical limitations and cultural norms [20].As Nelson [21] noted, "Children's Halloween costumes suggest a flight of imagination that remains largely anchored in traditional gender roles, images, and symbols… of what women/girls and men/boys are capable of doing even within the realm of their imaginations (p.143).
So, how might gender-typed and gender-targeted commercial costumes restrict this masquerade function to fit within the confines of everyday gender role norms?Nelson [13] anecdotally observed that girls' costumes provided less coverage and accentuated the wearer's own physical attractiveness as the child herself (rather than her costume character) became the featured subject.It seems that girl's costumes had to fulfill demands for feminine ornamental display-she must not simply be a princess, she must display herself as pretty, in her princess costume.Similarly, as Macmillan et al. observed of women's costuming preferences, full disguises are less desirable, in particular because anonymity does not allow the wearer to get "credit" for the ornamental display she feels compelled to exhibit [5].This is in marked contrast to achievement contexts, where women report downplaying display demands with clothing that conceals (or disguises) the body to regulate other's sexual objectification and their own self-objectifying bodymonitoring [22,23].It is also in contrast to what how boys approach their costuming.Research shows that men report a greater preference for Halloween costumes that disguise with masks and report feeling more freedom to portray different identities or roles while wearing costumes [5,24].Nelson [13] mentioned a tendency for boys' costumes to be more elaborate and fantastical, especially the gruesome villains.When it comes to the function of boys' and men's everyday garments, they already enjoy a distinct advantage over that of girls' and women's, as their clothing is more likely to favor function, comfort, durability, and coverage [25].It is thus reasonable to suspect that if a costume's function is to disguise and conceal so that the wearer can masquerade, this function may be more available in boys' and men's costumes.

STUDY 1 -RESEARCH QUESTIONS
To explore patterns in a contemporary sample of costumes, we extended Nelson's [13] study to include costumes marketed to children and also adults.In Study 1, we compared the representation of girls', boys', women's, and men's costumes across the Klapp costume character archetypes, on Nelson's active-masculinity/passive-femininity, and for degree of disguise.In Study 2, we examined an additional sample of girls' and women's costumes with regard to Nelson's construct of ornamental feminine passivity.
Based on Nelson's [13] distributions of the Klapp character archetypes of hero, villain, and fool, we expected boys' costumes to be "less singular in the visual images they portrayed" (p.141), and therefore dispersed across all the Klapp subtypes.We also expected girls' costumes to be concentrated in subcategories of conventional heroes (e.g.attractive princesses, beauty queens, and brides) and non-human fools (e.g.adorable ladybugs and baby animals).We expected the pattern of archetypes in men's costumes to be similar to boys', and in women's costumes to be similar to girls'.Throughout this section, we made note of any differences and similarities in our patterns compared to Nelson's.Finally, within our sample, we compared proportions of costumes falling into three major character archetypes to determine if there were statistically significant gender differences between boys' and girls' costumes, and then between women's and men's costumes (although Nelson did not run such statistical tests).
Next, we explored active-masculinity/passive-femininity.She noted that girls' costumes tended to be on the passivefeminine end of this dichotomy, characterized as erotic, ornamental, and "concentrated in the narrow realm of beauty" (p.141), and boy costumes tended to be on the active-masculine end, characterized by themes of agency, warriors, and repellent or "blood-curdling" masculinity (p.142).Based upon Nelson's informal observations, we expected boys' and mens' costumes would display more active-masculinity than would girls' and women's costumes.
We then explored the additional dimension of disguise.We expected boys' and men's costumes would be more disguising (i.e. more functionally masquerading) than would girls' and womens'.We also expected a positive correlation between active-masculinity and disguise, because active-masculine costumes would provide more coverage in support of the costume's masquerade function, and feminine-passive costumes would provide less coverage in support of display concerns.

STUDY 1 -METHOD
Costume Sample.Our costume sample came from two major nationwide online costume "superstores" in the USA, operating out of "big box stores" and seasonal "pop-up stores," but also accessible online via a standard online search for "Halloween Costumes," offering a large selection, and nationwide shipping.We diverged from Nelson's method of visiting brick-and-mortar costume and sewing stores to view packaging labels for several reasons.First, internet store sales have dramatically overtaken in-store sales in the United States [26].Second, online catalogs reflect the gendertargeted marketing boom [27] of "conveniently" directing the costumer to gender-appropriate browsing via explicitly gender-tagged product links where they will find the product displayed in vivid photographic imagery by a "gender appropriate" model.Third, with a decline in home sewing [28] and an increase in the proliferation of inexpensive outsourced product manufacturing [29], the mass-produced costume market has grown substantially.
Our sample of costume images was initially collected in 2010 and 2011 (about a decade after Nelson's study) and printed off onto a costume master file.We matched Nelson's [13] sample size of 428 children's costumes and added 428 adults' costumes, for a total of 856 costumes.Using pre-labeled gender links "boys' costumes," "girls' costumes," "women's costumes," and "men's costumes," we selected every other unique and non-repeating costume (up to the total of 233 for girls, 195 for boys, 233 for women, and 195 for men).Matching Nelson's sample size made for easier comparisons, and it kept our costume selection to a manageable size.With the costume master as a reference, coders were able to revisit costume images online through the end of 2013, as needed, with the sample of costumes still accessible, and most still available for purchase, at the time of write up.
Coding Procedure.A mixed-gender group of eight undergraduate psychology majors (two men and six women) worked with the authors to code all costumes in the selected sample.Coding assistants were given operational definitions of each code and a few practice trials for training purposes.Then a pair of assistants was assigned a subset of costumes to code (about 200 costumes per pair), each applying codes separately from each other.After each coder in the pair completed independent coding of their set, they met to count and reconcile any discrepancies.Discrepancies were reconciled through discussion between the coding pair, referring back to costume images and operational definitions.Interrater agreement was calculated as the proportion of matching codes to total codes for each category after all coders had completed coding the entire costume sample (but before the coding pairs met to reconcile to 100% agreement).

Coding of Costume Gender and Age.
The internet stores organized costumes by age and gender on separate links for boys', girls', men's, and women's costumes, typically available in sizes small, medium, large, and extra-large.For occasional instances when more than one age or gender of model was pictured in a single product image or when a costume was incorrectly linked by age or gender due to a web link error (e.g. an image of a woman displayed on the boys' costume link), coders replaced that costume with the next eligible costume on the website.
Coding of active-masculinity/passive-femininity.We coded each costume for active-masculinity/passive-femininity, based on a set of five indicators derived from Nelson's descriptions.Coders rated costumes on a three point scale (0 = not at all, 1 = somewhat, and 2 = great deal) for how agentic (interrater agreement = 86%), aggressive (interrater agreement = 86%), eroticized (interrater agreement = 90%), fearsome (interrater agreement = 99%), and winsome (interrater agreement = 88%) each costume was.The agentic dimension was operationally defined as the display of self or costume in action, using a skill, and/or in an alert, confident, or triumphant stance.The aggressive dimension was operationally defined as the display of warrior traits of physical strength, preparation for fighting, violence, or weapon use.The eroticized dimension was operationally defined as an eroticized, seductive, or sexually suggestive display.The fearsome dimension was operationally defined as a frightening, grotesque, or visually repellent display.The winsome dimension was operationally defined as an adorable, charming, cute, innocent, or submissive display.After coders reconciled their ratings, a final scale score was calculated as an average of these five ratings (reverse coding eroticized and winsome), with high indicating more active-masculinity (and low indicating more passive-femininity), putting Nelson's dichotomy on a single continuous scale (inter-item scale reliability for the five item scale was Alpha = .63).
Coding for Disguise.Coders separately rated costumes on four indicators of disguise.On a 3 point scale (0 = not at all, 1 = somewhat, and 2 = great deal), coders rated each costume for masking of the face (interrater agreement 99%), skin display (interrater agreement 90%), tightness (interrater agreement 81%), and coverage (interrater agreement 84%).Masking of the face was operationally defined as display of any full or partial masks or facial applique that concealed the wearer's face.Skin display was operationally defined as exposure of the wearer's arms, legs, or torso.Tightness was operationally defined as fabric that was body shape revealing, clinging, and form fitting.The coverage dimension was operationally defined as the incorporation of a sufficient quantity and thickness of fabric, to provide the wearer overall bodily coverage, drape, or insulation against the elements.After coders reconciled their ratings, a final scale score was calculated as an average of these four ratings (reverse coding the skin display and tightness scores), with high indicating more disguise (inter-item scale reliability for the four item scale was Alpha = .83).

STUDY 1 -RESULTS AND DISCUSSION
Costume Character Archetypes.In view of Nelson's findings, we expected boys' costumes to be more evenly dispersed across all the Klapp subtypes and girls' costumes to be concentrated in the subtypes of conventional heroes and non-human fools.We expected a parallel pattern in comparing men's costumes to women's costumes.Table (1) displays the number and percentage of costumes in each character classification, alongside those from Nelson's [13] original sample.
Our sample of boys' costumes was generally dispersed across the costume character archetypes, however there was a notable increase in conventional heroes (from 26% in Nelson's sample to 42% in our sample of boys' costumes).The conventional heroes consisted of famous licensed protagonists from film and television merchandising, a media marketing trend that has recently become a standard marketing strategy [30].The proportion of superheroes and exemplars was about the same as in Nelson's study, as was the overall requirement that boy heroes represent strong and powerful characters such as boxers, race car drivers, and any character scheme promising a "show of force" or to be "ready for battle."Boys' villains costumes remained distinctly less friendly than girls' villains, with boys' villains described as "evil" and "slayer," paired with violent or repellent props (e.g.weapons, oozing blood operated with a hand pump) and anti-heroes made up of licensed film and television "bad guys".The category with the greatest dip in boy costumes (compared to Nelson) was in the category of fools-with human (e.g.clowns) or nonhuman (e.g.animals) fool costumes for boys.As expected, the largest proportion of girls' costumes (67% in our study compared to 72% in Nelson's study) fell into the categories of conventional heroes and non-human fools.As in Nelson's study, girls' heroes included princesses, beauty queens, and fashionable divas, at the expense of superheroes (i.e., boys were 5 times more likely to be superheroes in our study and in Nelson's), and girls' fools were typically dainty and winsome winged creatures or bashful baby animals.There was an increase, compared to Nelson, in the proportion girls' villain costumes, specifically in symbols of death and monsters popularized by television and movies [31,32].Similar to Nelson's sample, girls' villain costumes were accompanied with friendly descriptors such as "sparkle," "adorable," and "pretty." The largest proportion of men's costumes were conventional heroes, most typically representing strength, ruggedness, and accomplishment.As with boys', men's villains were paired with violent imagery and film and television "bad guys".The unexpected pattern in men's costumes was in the proportion of fools-which was substantially higher than was found in women's costumes.Fools dropped out in our sample of boys, but they remained for men and girls.However, the men's fool costumes were different from the "adorable" girls' fool costumes, with men's fool costumes often outrageously comical or politically controversial, incorporating cross dressing, mockery of women, ageism (e.g.exposing faux "aging" female body parts), intentionally vulgar phallic symbols, or sexual innuendo (not for sexual appeal, as was emphasized in women's fool costumes).These costumes offered men an excuse for both sexual display and for "playing" dress up, with promises to "be the life of the party."Men's costumes seemed to emphasize masculine pressures toward same-sex approval and one-upmanship [33].
As expected, women's costumes were not as dispersed across categories as were men's costumes.A higher proportion of women's costumes were heroes, in comparison to men's costumes, with the largest proportion (59%) falling into the single category of conventional heroes.In addition to the standard beauty queens and divas, women's hero costumes included masculine occupational heroes Nelson had noted in boys' costumes (e.g.firefighters, police, etc.), however these women heroes were highly sexualized in revealing clothing and high heels inconsistent with the uniform of the job.Sexualized outfits were repeated in all of women's superhero costumes (e.g."Sexy Robin/Batman Costume") and in standard "pin-up" girl costumes, as well.Women's and girls' villains included more overt themes of sex-work and seduction.Fools represented less than one in ten women's costumes compared to one in four girls' costumes, and women's fools emphasized sexiness rather than the winsomeness found in girls' fools.Across all types of women's costumes, there were messages encouraging competitive self-sexualization for attention [34], often with accompanying textual descriptions promising "all eyes" will be on her.
We compared the proportions of the character archetypes to test for statistically significant gender differences, comparing the proportions of boys versus girls costumes, and then the proportions of women's versus men's costumes.In the children's sample, the proportion of hero costumes was higher for boys than for girls, X 2 (n = 428, df = 1) = 4.40, p. < .05,and the proportion of fool costumes was higher for girls than for boys, X 2 (n = 428, df = 1) = 8.13, p. < .01.In the adults' sample these patterns were unexpectedly flipped, with the proportion of hero costumes being higher for women than for men, X 2 (n = 428, df = 1) = 20.02,p. < .001and the proportion of fool costumes being higher for men than for women, X 2 (n = 428, df = 1) = 24.79,p. < .001.There were no statistically significant gender differences in the proportions of villains in the children's sample, X 2 (n = 428, df = 1) = .014,NS, nor in the adult sample, X 2 (n = 428, df = 1) = 1.81,NS.
Disguise.As anticipated, boys' costumes offered more disguise (M = 1.65;SD =.26) than did girls' costumes (M = .99;SD =.41), t (426) = 19.28,p. < .001(Cohen's d = .68).Likewise, men's costumes offered more disguise (M = 1.51;SD =.39) than did women's costumes (M = .27;SD =.43), t (426) = 31.13,p. < .001(Cohen's d = .83).Boys' and men's costumes often completely covered the wearer with a full head mask and full body suit.This was rarely the case for girls' and women's costumes, with women's costumes having the least disguise.In our sample, boys were offered more liberation from the "regular" self in the act of Halloween masquerade, without being obligated to literally possess featured traits [19].Combining across all costumes, we found the expected positive correlation between active-masculinity and disguise, r (n = 856) = .704,p. < .001;thus, regardless of gender or age, the more activemasculinity displayed in the costume, the more disguise it afforded.

STUDY 1 -CONCLUSION
Despite the promise of progress in the ten years passing since Nelson's [13] study and the potential for greater choice in an expanded costume market, children are still being presented with the same gender-typed costume options.We found very similar gender typing in the character archetypes of children's costumes to what Nelson reported.Boys' costume choices represented a wider range of characters of agency and girls costume choices were more restricted to pretty heroes and darling fools.For our adult costume sample, we found more divergent patterns, with women's costumes predominantly incorporating a heightened emphasis on display through sexualized heroes, and men's costumes offering options to be agentic heroes and clever fools.We found statistically significant gender differences in costume archetypes for both children and adults.Medium effects sizes were evident in the differences between girls' and boys' options, and medium to large effect sizes were evident in the differences between women's and men's options.Moreover, the two dimensions were related to each other across the entire sample, suggesting that costumes providing greater active masculinity also provided the wearer more opportunity to masquerade.Inversely, costumes high in passive femininity offered less opportunity for playful disguise, hindering what is the express purpose of costume masquerade-to be incognito.Girls' and women's costumes emphasized the wearer herself as the winsome or sexualized feature of that costume.
The "ornamental passive femininity" that Nelson described of girls' costumes generally emphasized winsomeness, as she made only minor reference to eroticizing.With the inclusion of a sample of women's costumes, we found the eroticizing or sexualizing to be far more common.To a lesser degree, we also saw an emphasis on the sexualizing of girls' costumes, too; however, for them the ornamental value also emphasized innocent sweetness, or what is referred to as infantilizing.It seems that these two themes of sexualizing and infantilizing are in need of closer attention.
As feminist psychologists have noted [36,37], infantilization and sexualization are at the very foundation of sexism against women in contemporary Western culture.The pervasiveness of these themes, acting separately or in tandem, contributes to perceptions of their normativity while they erode the agency, instrumentality, and independence of girls and women [38].There has been significant and growing research attention paid to sexual objectification, culminating in APA's [10] report on the sexualization of girls.Little attention has been paid to infantilizing stereotypes of girls and women, with the largest share of the literature to be found in the work of evolutionary psychologists' who view female neoteny as a biologically determined adaptive female sex-linked trait [39,40].Feminist scholarship is more likely to view such associations as stereotypes serving to justify a status quo that prevents adult women from being viewed as mature and competent [41].

STUDY 2 -RESEARCH QUESTIONS
We collected an additional sample of costumes for Study 2, selecting only girls' and women's costumes, so that we could explore the pattern of two unique subtypes of Nelson's construct of ornamental feminine passivity in across age groups.First, we explored the qualitative nature of the costume characters, classified into Klapp's three major character archetypes, as in Nelson [13] and in our Study 1.Given previous findings, we expected the largest proportion of costumes would again be heroes (i.e.feminine beauty queens, princesses, etc.).We also tested for statistically significant age differences in the proportions of the archetypes in girls versus women's costumes.Based on results of Study 1, we anticipated a higher proportion of girls' costumes would be fools, compared to women's costumes.
Next, we explored infantilization and sexualization.Infantilization of girls and women involves viewing them as immature and equating femininity with vulnerability and submission [37,42].Infantilizing is the foundation of the "school girl" motif [12], "romper craze" [43], "princess effect" [44], and "girl power," when are set apart from ordinary power [45].Infantilization is also evident in the fashion media with its strategic fetishizing of infantilized poses by models, putting their hands over or in their mouths, covering their eyes, tilting their heads, twirling their hair, or gazing in a state of mental drift as if lost in a daydream [37,46,47].Clothing styles, fabrics, and decorations also contain infantilizing elements that resemble doll clothes [25,48].Nelson [13] did not explicitly code for infantilizing themes; however, some of the infantilizing elements she observed included the use of the color pink (hence the "Pink Dragon" in her title), bashful gestures (with female costume models depicted in a non-agentic stance), or additions of bows (e.g. to suggest an animal costume female).She also noted how infantilization of girls' villain costumes "emphasized their winsome rather than wicked qualities, to neutralize the malignancy" (p.142).
Sexualization emphasizes seductive ornamental display with clothing that has become eroticized [10,49,37,50].Sexualization has a new potency as the images of mainstream fashion have coopted themes from the commercial sex industry, such as symbols of bondage, sadomasochism, and sexual exploitation [12,51,52].Termed "The Lolita Effect" or the "pornographization" of the culture, the "stigmata of prostitution" is brought into the everyday life of women and girls [12,25,52,53].Sexualizing clothing and products are increasingly available for girls of all ages [50, 54 -56], however due to the taboo of sexuality being associated with youth, infantilizing elements are often added to dilute the overt sexuality products marketed to younger girls' [37].Nelson [13] did note occasions, particularly for villains, in which costumes emphasized "erotic" themes (e.g."trim at the top of the ribcage served to create the suggestion of a bosom," and "the names of villains emphasized the erotic side of their villainy e.g.Echantra," p. 141).
We explored the patterns of infantilizing and sexualizing across the four age groups of baby girls', child girls', teen girls', and women's costumes.We expected to find both infantilizing and sexualizing themes to some extent in all age groups.With little research on infantilizing, it was unclear how it would vary across age groups, however, we did assume sexualization would be more frequent in costumes for teen girls and women.Furthermore, based on observations of Goodin et al. [37] that girls' clothing incorporates infantilizing elements as a "cover" for sexualization (to diffuse resistance that might occur if clothing is exclusively sexualizing), we expected that the combination of infantilization and sexualization would be most common in children's and teen's costumes, targeted at an age group where overt sexualization is increasing, but still taboo.
Finally, we developed a descriptive inventory of the symbols of sexualization and infantilization evident in our sample of girl Halloween costumes.Visual symbols or cues are important units of meaning [57], including with regard to gender [58].We developed a thematic coding framework with examples, not unlike the extracts use in narrative qualitative codes [59].Cultural symbols are strategically developed and applied by product designers [60], thus the inventory of symbols we develop in one domain, Halloween costumes, is likely to inform our understanding of other female-targeted merchandise and gender socialization via objects in other domains.

Costume Sample
Costumes were selected from two major nationwide costume "superstores" accessible online, using the same criteria as in Study 1, but different stores.This sample of costume images was initially collected in 2011 (a year after the data collection for Study 1), printed off into an image master file that coders used as a reference, revisiting images on the websites as necessary through the end of 2013.We sampled every other costume to generate a sample of 161 baby girls,' 189 child girls', and 167 teen girls' costumes, and every third costume to generate a sample of 301 women's costumes.Most costumes sampled were still accessible and available for purchase at time of write up.

Coding Procedure
A mixed-gender group of four undergraduate psychology majors (one man and three women) worked with the authors to code all costumes in the selected sample.Coding assistants were given operational definitions of each code and practice trials for training purposes.Then each pair of assistants was assigned a subset of costumes to code (about 250 costumes per pair), each applying the codes separate from each other, so that each costume was coded by independently by two different individuals.After each coder in the pair completed the independent coding of their set, they met to count and reconcile discrepancies.Discrepancies were reconciled through discussion between the two coders in the coding pair, while referring back to the costume on the website and the operational definitions.Interrater agreement was calculated as the proportion of matching codes to total codes for each category after all coders had completed coding the entire costume sample, but before the coding pairs met to reconcile to 100% agreement.

Coding of Costume Gender and Age.
The online Halloween stores had already organized costumes on separate links by gender and by age into girls' (further subdivided by the stores into baby/toddler girls' for ages 0 to 4, child girls' for ages 4 to 14, teen/"tween" girls' for ages 14 to 18) and women's costumes.For any rare instances of costume images picturing more than one model in the product image, duplicating exactly a costume already coded, or incorrectly linking at costume by gender or age due to web link error (e.g. an image of an adult man depicted under baby girls' costumes link), coders replaced that costume with the next eligible costume on the website.
Coding of the Costume Character Archetypes.Costume character archetypes were coded into the three primary character archetypes originally defined by Klapp's [14]: Heroes, Villains, and Fools as in Nelson [13].This procedure was identical to that used in Study 1 (interrater agreement was 94%).

Coding of Disempowerment via Infantilization and Sexualization.
Coders rated the costumes separately for presence or absence of infantilization and sexualization in each image, with operational definitions based on Goodin et al [37] and Jeffreys [25].Infantilization was defined as display of self or costume with an emphasis on bashfulness, charm, harmlessness, innocence, immaturity, sweetness, and/or winsomeness.Interrater agreement was 88%.Sexualization was defined as display of self or costume with an emphasis on display of seduction, sexual appeal, and allusions to sex work.Interrater agreement was 91%.From these codes, costumes were higher order coded into four types of disempowerment: 1) no disempowerment (neither infantilized nor sexualized), 2) infantilized only, 3) sexualized only, or 4) a combination of both.
Developing the Inventory of the Symbols of Infantilizing and Sexualizing.All costumes were reviewed again by the authors in a thematic analysis of the symbols of infantilization and sexualization [61].The authors separately and independently reviewed costumes coded as infantilized, making lists of the symbols indicative of infantilizing and then reviewed costumes coded as sexualized, making lists of the symbols indicative of sexualizing.Next, the authors met, and through discussion, merged their lists to generate a single higher order organizational scheme for the symbols they identified.The authors classified the symbols according to what dimension of the costumes each the symbol was applied: Cosmetics (face or body makeup), embellishments (extra decorations or specialized cuts in the costume), fabrics (the types of material used to make the costume), poses (how the model was posed for the image), motifs (common familiar cultural themes or genres drawn from), and language (descriptors used in the costume name and/or description).Labels were then generated for each set of symbols to describe the message communicated by each set of symbols.This type of qualitative coding, shifting from examples, to category headings, and generating higher order and mid-level classifications, is part of the flexible coding method afforded by thematic analysis [61].The resulting inventory is organized by costume dimension, message, and examples of symbols (with the examples akin to narrative "extracts" or quotes in qualitative interviews).This coding was for descriptive rather than quantitative purposes as a means to catalog visual data in a systematic way, therefore themes are not listed in order of frequency and more than one element could and did often appear in a single costume.

STUDY 2 -RESULTS AND DISCUSSION
Costume Character Archetypes.In view of the patterns found in Nelson and Study 1, we expected the largest proportion of this sample of costumes would be heroes.Table (2) displays the number and percentage of each costume character archetype by age.Given the pattern of results of Study 1, we anticipated a higher proportion of girls' costumes (totaled across baby, child, and teen) would be fools, compared to women's costumes.It is notable that the proportions of heroes, villains, and fools for girls was virtually identical to the proportions for girls in Study 1 (a difference of only two percent) and the proportions in the women's costume sample were also virtually identical to what was found in Study 1 (off by only 1 percent).As expected the most common character type was heroes, for girls and women.The sample of girls' costumes included heroes valued for their attractiveness (e.g."pretty pink princess").Women's hero costumes emphasized their attractiveness value (e.g.Sexy Snow White), even when depicting more traditionally masculine heroes (e.g."glamour cowgirl", "hottie firefighter", "pin-up sailor" and "scandalously sexy baseball player").Just as in women's everyday wear [25], these costumes de-emphasized function (e.g.tutus instead of pants on pirates, and high heels instead of work boots on a firefighter).Girls' villains emphasized sweetness over harmfulness (e.g."precious pirate" or "darling witch") and women's villains emphasized eroticism over loathsomeness (e.g."bewitch" your admirers, "conjure" attention).Girls' fools conjured attributions of sweetness (e.g."sweet lil' bee" or "melt hearts candy") and the less common women's fools featured sexuality rather than sweetness; so, the child girl's "cuddly lion" costume became "sexy cuddly lion."Although some women's fools included sexual innuendo (e.g. a pixie is recast as "pixie lust," the ubiquitous girls' ladybug is recast as "light me up" lady bug), they did not include the vulgar or unappealing visual displays common of the men's fool costumes in Study 1).Overall, echoing Nelson [13] and Study 1, these costumes marketed to girls and women emphasized pretty and sexualized feminine display over skilled or foreboding masculine competence.
Based on the results of Study 1, we anticipated a higher proportion of girls' costumes would be fools, compared to women's costumes.We compared the proportions of each of the three costume character archetypes available in girls' costumes (as a combined category) compared to what was available in women's costumes (see Table 2).Heroes were more common in women's costumes than in girls' costumes, X 2 (n = 816, df = 1) = 20.90, p. < .001.There were no age differences in the proportions of villains in women's versus girls' costumes, X 2 (n = 816, df = 1) = .002,NS.As predicted, we found fools were more common in girls' than women's costumes, X 2 (n = 816, df = 1) = 32.63,p. < .001.Disempowerment via Infantilization and Sexualization.Only 12% of female costumes were free from both infantilizing and sexualizing forms of disempowerment.Fig. (1) displays the patterns of disempowerment as a function of age (overall 65.2% for baby girls', 95.2% for child girls', 93.9% for teen girls', and 92.7 for women's costumes).We tested for age differences in likelihood to be infantilized, sexualized, or both using the Kruskal-Wallis (H) test, a nonparametric omnibus analysis of variance test with no assumptions of normality, producing a chi-square value that enables the comparison of more than two independent non-randomly assigned groups and Mann Whitney U post-hoc comparisons.In these tests, there are four non-randomly assigned groups (baby girls' vs. infant girls' vs. child girls' vs. women's costumes), therefore any significant H tests were followed up with Mann Whitney U post-hoc tests, to determine which groups differ significantly.The independent variable was costume age group (baby girl, child girl, teen girl, and women), and three separate tests were run with each of the three mutually exclusive dependent variables: first with the dependent variable being the proportion of costumes that were disempowered in an exclusively infantilizing manner, second with dependent variable being the proportion of costumes that were disempowered in an exclusively sexualizing manner, and third with dependent variable being the proportion of costumes that were disempowered with a combination of infantilization and sexualizaton.(A fourth test on age differences in proportion of costumes not disempowered would have been redundant and was not run.)Across the whole sample, 20% of costumes contained disempowerment exclusively in the form of infantilization (without sexualization).There were statistically significant age differences in rates of exclusive infantilizing, H (df = 3, n = 816) = 185.66,p. <. 001.Mann Whitney U post hoc tests revealed that baby girls' costumes were significantly more likely to be infantilized (56.1%) than child girls' (20.6%), and both were significantly more infantilized than teen girls' (6.2%) and women's (7.0%) costumes, all at the p. < .001level.
Over the entire sample, 33% of costumes were exclusively sexualized (without infantilization).We found age differences in exclusive sexualization.As predicted, sexualization was more likely in the older age groups of teen girls' and women's costumes than in baby girls' and child girls' costumes, H (df = 3, n 816) = 263.34,p. < 001.Mann Whitney U post hoc tests showed that, with the exception that child girls' and teen girls' costumes did not differ from each other, all ages significantly differed from each other at the p. < .05level, with baby girls' displaying the least (1.8%), then child girls' (15.9%) and teen girls'(21%), then women's costumes (66.8%) displaying the most exclusive sexualization Overall, 35% of costumes included a combination of both infantilization and sexualization.We expected the combination of both infantilization and sexualization to be most common in children's and teen's costumes, because the infantilizing serves as a "cover" for the sexualization [37].H (df = 3, n 816) = 206.49,p. < .001.As predicted, this combination occurred at equally high levels for child girls' (58.7%) and teen girls' (66.7%) costumes (which again, did not differ from each other) but differed significantly (at the p. < .001level) from baby girls' (7.3%) and women's (18.9%) costumes (which did not statistically significantly differ from each other).The proportion of costumes with a combination of sexualization and infantilization was about as common as the proportion that were exclusively sexualized (but more common than the proportion only infantilizated).Moreover, the observation that the proportion of child girls' costumes with a combination of sexualization with infantilization did not differ statistically significantly from the proportion in teen girls' costumes may be a disturbing indicator of a normativity of sexualize girls.
Inventory of the Symbols of Infantilizing and Sexualizing.We generated a list of the examples of the disempowering symbols of sexualization and infantilization, organized according to thematic dimensions of the costume that emerged in our coding, displayed in Table (3).Based on the themes that emerged in our coding, we found divergent but parallel dimensions.The cosmetics that accompanied infantilized costumes were often theatric, clownish, or doll-like, the embellishments were decorative, the fabrics were innocent, the poses submissive, the descriptors were patronizing, and the motifs were borrowed of youth and school days.For example, infantilized symbols in cat costumes included drawn on whiskers and noses, fuzzy pink ears on a headbands with bows, pink tutus and stockings, Mary Jane style shoes, and furry tails, with models turning their knees inward, and text referring to playfulness and kittens.Sexualized costumes were accompanied with glamorous, designer, or fashion makeup, revealing embellishments, eroticized fabrics, seductive poses, suggestive language, and the motifs borrowed from sex work and high fashion.For example, sexualized symbols in cat costumes included false eyelashes, eyeliner, and red lipstick, black satin ears, lace-up corsets and miniskirts or tight bodysuits, black fishnet thigh-high stockings, high heels or boots, and leather-look tails, with models turning her legs in a wide stance, her hips shifted to one side, and text describing her as flirty and frisky.The interplay of these themes across columns was more common than their separation, consistent with the patterns displayed in Fig.

STUDY 2 -CONCLUSION
The fairly substantial infantilizing in girls' costumes can impact their sense of efficacy [37] as they make a "mockery" of female power [21].Perhaps societal symbols to disempower girls and women (whether conscious, covert, or institutionalized) only need to be engaged once they actually can pose a threat to the status quo, so the older she is, the more these markers are applied [62,63].
Though exclusive sexualization was not evident in the majority of girls' costumes, there were still substantial proportions, especially for child and teen girls' costumes.This is of concern when it appears in girls as young as 4 -14, since sexualizing contexts have been shown to inhibit healthy maturation, leading girls to become body conscious while 0.

Only Infantalized
Only Sexualized they are still learning to understand the meaning and implications of their sexuality [44, 64 -66].Girls and women choose sexualized display in part because they are choosing from available options, and marketing techniques to promote these images [50].Girls report that their choice of costume is affected by a desire (or what they are told should be their desire) for attention (Lamb & Brown, 2006 [50] and are encouraged to flaunt sexuality, even if they do not feel intrinsic interest in doing so [12,44].This supports Goodin and colleagues [37] observation of this insidious combination of disempowering messages directed at children. Sexualizing and infantilizing images are of concern in feminist psychology [36,47], and they were quite prominent in Study 2. With repeated pairing over time, the one may come to imply the other, so that infantilizing can essentially become eroticized, creating a confusing combination [67], encouraging a tendency to view children as sexually seductive [68].The general emphasis on appearance and the lack of agentic role models further contributes to the inhibition of creativity and efficacy in girls and women [69].Some postmodern feminist thought points to women's choice to dress in an infantilized or sexualized way as a means to rebel or reclaim these displays [44,70].Certainly some measure of these hyper-normative costumes may have been poking fun at it, perhaps dismantling normative power [71].Some girls and women may choose to use Halloween as an opportunity to embrace their sexuality, or to rebel against traditional modesty.However, some research suggests another perspective.One study by Erchull and Liss [72] found that women "believing that sex is a personal source of power was related to greater experiences of body surveillance and unwanted body evaluation (p.50)." Overall, in Study 2, it appeared that girls and women had very similar pressures of intertwining infantilizing and sexualizing to wrestle with, starting quite young.Future research might focus on developing a standardized scoring system for all media to assess the degree of infantilizing and sexualizing in a range of products, including everyday clothing items, toys, magazine images, movies, and literature [73].Marketers have been found to overwhelmingly favor thin, young, white models [74].Sizeism, ageism, and racism may add layers to sexualization and infantilization, and further potential for threats to girls' and women's self-esteem [75].

CONCLUSION
Dressing in gender-stereotyped garments is one way in which children and adults participate daily in creating disparate gendered environments, reinforcing conventional gender roles [37, 45 -49].Clothing is public part of our gender performance that can be judged by others [76,77].Feminist psychology must remain vigilant and attentive to the fact that powerful sources of gender role restriction and gender typing remain even in a seemingly mundane activity such as Halloween costuming and not to discount their influence despite so many cultural advances for women and girls [38,50].
Creative play is essential to growth and it opens up possibilities to reinvent norms [55].An overarching theme of Halloween is creative expression, which itself inspires originality, imagination, and risk taking [78].Creative behaviors are influenced by external factors such as cultural values [79], gender role schemas [1 -6], and restricted choice in commercially produced products [10].In their book Packaging Girlhood: Rescuing our Daughters from Marketers' Schemes, Lamb and Brown [80] argue that girls are particularly disadvantaged in the commercial marketplace compared to boys.Similarly, we found boys and men still enjoyed more creative freedom in playful disguise while girls and women were held in the binds of display concerns.
As millions wear Halloween costumes annually [3], costume designs and choices are likely to have some impact on how we think about gender.The relationship between individually held beliefs and commercially produced gender stereotypes is an important matter for continued future research.We must continue to work on media literacy in order to resist the power of commercially produced gender socialization, encourage alternative media outlets with positive gender images, develop peer discussions, and demand more counter-stereotypic media images.

ETHICS APPROVAL AND CONSENT TO PARTICIPATE
Not applicable.

HUMAN AND ANIMAL RIGHTS
No humans/animals were harmed in the process of conducting this research.