Gender Stereotyping and Under-representation of Female Characters in 200 Popular Children’s Picture Books: A 21st Century Update
Mykol C. Hamilton, David Anderson, Michelle Broaddus, and Kate Young. The authors wish to thank undergraduate assistants Nate Olson and Ashley Vinsel for their invaluable assistance.
RELEASED: Feb. 19, 2007
DANVILLE, KY—Abstract. Gender stereotyping and under-representation of girls and women have been documented in children’s picture books in the past, in the hope that improvements would follow. Most researchers have analyzed award winning books. We explored sexism in top selling books from 2001 and a 7-year sample of Caldecott award winning books, for a total of 200 books. There were nearly twice as many male as female title and main characters. Male characters appeared 53% more times in illustrations. Female main characters nurtured more than male main characters did, and they were seen in more indoor than outdoor scenes. Occupations were gender stereotyped, and more women than men appeared to have no paid occupation. Few differences were found between Caldecott award books and other books. A comparison of our book sample to 1980s and 1990s books did not reveal reduced sexism. The persistence of sexism in picture books and implications for children and parents are discussed.
Johnny and Susie involves a brave boy and a helpless girl, and Foxtown Four is about four boys, one king, and three male dragons. In fact, the books are imaginary, but if they were real, why would it matter? Who cares if girls are cast in passive roles, boys in active ones, or if male characters outnumber female characters? They’re just books.
Does Sexism in Picture Books Matter?
Second, experimental research strongly suggests that gender bias in picture books is harmful to children. Schau and Scott (1984) reviewed 21 studies on the effects of sexist vs. nonsexist children’s instructional materials (e.g., male versus female characters; sexist versus nonsexist generic pronouns), and discovered a consistent tendency for sexist materials to strengthen children’s biases. Dependent variables in the 21 studies included occupational interests, gender-typed attitudes, and interest in traditional versus nontraditional activities. In one study (Ashton, 1978) 3-5 year old children read gender-biased or -unbiased children’s picture books. Children who read biased books later made more stereotypic toy choices. Based on these and other studies, Tognoli, Pullen, and Lieber (1994) concluded that gender bias in children’s books gives boys a sense of entitlement and lowers girls’ self-esteem and occupational aspirations. Moreover, Weitzman, Eifler, Hokada, and Ross (1972) argued that the dearth of female characters teaches both sexes that girls are less worthy than boys. Other researchers have concluded that children’s literature provides girls and boys with standards of masculinity and femininity (Peterson & Lach, 1990), offers socially sanctioned behavioral models that children may imitate (St. Peter, 1979), and presents a basic model for understanding oneself and others (Rachlin & Vogt, 1974).
Most of the research that documents the extent of bias in children’s books has focused on picture books for young children rather than on chapter books or other reading material for older children. Parents read picture books to their children when the children are very young. At ages 3-5 children begin the processes of actively learning to distinguish the sexes and of forming gender stereotypes (Powlishta, Serbin, & Moller, 1993). Therefore the gender roles and the numbers of female and male characters portrayed in the books probably have serious effects on these children’s gender role development and self-image.
In sum, evidence suggests that sexism in children’s picture books affects children in ways that feminists and parents with varied politics consider negative. But is the potential harm realized? If only a small minority of picture books have sexist content, harmful effects would be limited.
Sexism in Picture Books: How Much?
More recentstudies have confirmed that gender bias lingers on. Barnett (1986) stated “The higher representation of males has been found not only in broad samplings of children’s literature but in award-winning books as well” (p. 343). In their literature review, Tognoli et al. (1994) traced 12 studies conducted from 1972 to 1993 that analyzed sexism in a wide variety of children’s literature samples including Caldecott winners and runners-up, Newbery winners, non-award-winning books, and even French children’s picture books. Despite the studies’ different dependent variables and book samples, every one of the 12 studies revealed some type of gender bias. A recent study by Anderson and Hamilton (2005) showed that gender bias in children’s picture books is not limited to bias against girls and women. Portrayals of one important role, that of parent, slight men in several ways. For example, fathers were under-represented and portrayed as relatively stoic actors who took little part in the lives of their children.
Might changes in women’s roles and opportunities over the decades have led to a steady rise in the female-to-male character ratio and to less stereotyped portrayals of the sexes? In fact, the pattern of changes has varied by decade and by study. Clark, Guilmain, Saucier, and Tavarez (2003) documented the inconsistency of changes in Caldecott winners and runners-up from the 1930s to the 1960s in their aptly titled article: Two Steps Forward, One Step Back. They found that stereotyping and under-representation rose and fell repeatedly and that they were inversely related. In decades in which there was more stereotyping of female characters (1930s and 1950s) there was actually better representation of female characters, and conversely, in the 1940s and 1960s, female characters were more stereotyped but under-representation was not as severe.
Other researchers have found other patterns. For example, Nilsen (1978) documented a continued downward trend in the visibility of female characters in Caldecott books from the 1950s to the 1970s, and Kortenhaus and Demarest (1993) found a steady decrease in stereotyping in award-winning books from the 1940s to the 1970s.
Several researchers have found that the 1980s was a decade of improvement in both portrayal and representation of female characters (e. g., Collins, Ingoldsby, & Dellmann, 1984; Dougherty & Engel, 1987; Kinman & Henderson, 1985; Williams, Vernon, Williams, & Malecha, 1987). However, the improvements were relatively weak in that female characters continued to be under-represented, with as many as two male characters for every female character, and three to one ratios of employed men to employed women. For example, Allen, Allen, and Sigler (1993) found only a slight shift toward equal representations of male and female characters in their comparison of Caldecott winners and honor books published from 1938-1940 to books published from 1986-1988.
Award-Winners vs. Non-Award-Winners
Other researchers have seen the value of studying non-award winning books available to children on the shelves of local libraries. Barnett (1986) studied a total of 1,537 picture books by dividing the children’s picture book area at a local public library into 18 sections and having each of 18 readers look at all the books in her or his section. McDonald (1989) analyzed 22 Caldecott winners and 19 non-award-winning books randomly selected from the shelves at a local public library and a university library. Another unique sampling method was used by Tognoli et al. (1994). They randomly chose 200 books from the shelves of three libraries in New York City. The researchers looked at 100 books published before 1980 and 100 published after 1980.
Although, as we have just seen, several studies have moved beyond the exclusive study of award winning books, only Tepper and Cassidy (1999) contrasted award winning and popular books within the same study.
Researchers have consistently documented the absence of a relationship between author sex and title/central character sex. For instance, Tognoli et al. (1994) found that, although more men than women had authored the pre-1980 books they analyzed and the reverse was true for the post-1980 books, title roles in both samples were dominated by male characters or by characters of ambiguous sex, regardless of author sex. In other studies, Heintz (1987) and Kolbe and Lavoie (1981) found no relationship between author sex and central character sex.
The Current Study
The present study was designed to assess gender bias in 200 top-selling children’s picture books. We thought that it was important to explore a sample of both award winners and popular books, given that the one study that compared the two samples revealed some differences in the degree of sexism (Tepper & Cassidy, 1999). We gathered a large sample of books and studied a wide range of variables, that had been examined in many of the earlier studies. We looked at gender representation in pictures and characters; characters’ behaviors, settings, and personality; and the relationship between author sex and character sex. The following hypotheses were tested.
1. There would be fewer female than male child characters overall and adult characters overall, fewer female title characters, fewer female main characters, and fewer pictures of female than male characters.
2. We tested the relationship between sex of author and sex of title and main characters.
3. Female main characters would more often be portrayed indoors than outdoors; the reverse would hold true for male characters. Female main characters would be portrayed as passive, be rescued by another character, and behave in a nurturing or caring manner more often than would male characters. Male main characters would be portrayed as active, rescue another character, and behave assertively or aggressively more often than would female characters.
4. The occupations of the predominant adult male and female characters would more often be gender traditional than gender nontraditional, male characters would inhabit a wider range of occupations than would female characters, and more female than male characters would be portrayed as not having an occupation outside the home.
5. We explored possible differences between Caldecott winners and non-award-winning books in the numbers of female and male title characters, main characters, overall characters, and illustrations.
6. We performed exploratory analyses to test for changes in representation and stereotyped portrayals since the 1980s.
The designation “characters overall” refers to all characters, both primary and secondary.
Our sample included the 30 Caldecott Medal winners and honor books for 1995-2001. However, in order to draw from a larger and more representative collection of widely read picture books, we located 155 best-selling children’s books that had not won Caldecott or Newbery awards but were listed as top sellers in 1999-2001 by the New York Times, Amazon.com, Barnes and Noble, or Publishers Weekly. Also included were nine additional best-selling Little Golden Books and three non-overlapping books from the 2001 New York Public Library list of books everyone should know and the 2001 Funorama.com top 10 picture books. All of these designations translate into relatively large sales volumes. Of course, there is considerable overlap among these collections. Our total sample consisted of 200 books.
Survey Instrument. The survey instrument consisted of 22 items. The first three were the book’s title, year of publication, and author. The other items were frequency counts, classifications, and fill-in-the-blanks.
Following Weitzman et al. (1972), for each book we counted the numbers of female and male title characters, main characters, and characters overall, as well as the numbers of each sex represented in illustrations. Characters could be human, animal, or other (plants, robots, etc.). Title characters were classified as male if a name commonly seen as masculine was used in the title, as female if the name was commonly recognized as feminine. We did not analyze title characters for books with both a female and a male title character or with a gender-neutral title character, as there were very few of either. In addition, we analyzed the numbers of characters overall for both children and adults, as did Heintz (1987).
We combined variables used by Weitzman et al. and Williams et al. (1987), and analyzed how often male and female main characters were portrayed as active or passive, found indoors or outdoors, rescued or were rescued by another character, and nurtured or were nurtured. Raters counted the number of scenes in which a main character was active or passive and the number in which the character’s action had taken place outdoors or indoors. The character was then classified according to which behavior or setting was predominant. Counts were made for the rescue and nurture variables.
We compared the range of occupations for main characters of each sex, but we also classified each adult character as having a traditional, non-traditional, or gender-neutral occupation, or no occupation.
We drafted an instrument to code the above variables. An annotated version of the instrument provided detailed coding information. For example, coders were supplied with a list of stereotypical men’s occupations and stereotypical women’s occupations.
To refine the instrument, a male professor, a female professor, two female students, and one male student read a sub-sample of the books, and the results were compared. Over several iterations of this process the items were amended and the coding rubric was refined until the readers coded the books consistently. Two female students then coded all of the items for each book.
Cohen’s kappa tests were used to assess the inter-rater reliability of the two readers for nominal variables such as sex of title character and traditional vs. non-traditional jobs. All Cohen’s kappa values indicated either good or excellent agreement. When the readers’ responses differed, that book was not used in the analysis.
For frequency data, reliability was measured with Pearson product-moment correlation coefficients. Of 12 variables, eight had acceptable r values (greater than .70). For the other four variables, we analyzed only the books on which there was a less than 10% discrepancy between the two raters’ counts.Most of the disagreement fell among subjective items or items for which it was very difficult to obtain accurate counts. For example, it was a challenge to count female and male characters when the characters appeared in crowd scenes or when the illustrations were very small. The larger the numbers of character illustrations in a book, the more dissimilar the male-female character frequency counts were, in both absolute and proportional terms. When frequency variable responses differed by under 10%, raters’ responses were averaged. At times it was hard to determine whether even a large, well-defined picture was intended to depict a male or a female, but if the character also appeared in the text, then we identified sex through pronouns or other verbal cues. If the character did not appear in the text, then we made a subjective judgment based on visual cues such as clothing, body shape, or facial features. Dougherty and Engel (1987) recommend considering a picture neutral unless it contains very clear sex cues. We used a different criterion. If we believed that children were likely to interpret a character as male, we coded it as male. A strong case can be made for our less stringent categorization procedure, based on people’s tendency to assume that a person or animal is male in the absence of strong cues that it is female. Such a tendency has been shown in studies on the People = Male bias (Cole, Hill, & Dayley, 1983; Hamilton, 1991) and the Animal = Male bias (Lambdin, Greer, Jibotian, Wood, & Hamilton, 2003). Hamilton et al. and Cole et al. found that when a person is referred to without any cues about sex, that person is generally believed to be male. Lambdin et al. performed three studies on children’s and adults’ perceptions of stuffed animals. Gender-neutral animals were nearly always referred to as he and seen as male by both children and adults, as were masculine animals. Even quite feminine-appearing stuffed animals were referred to as he and labeled as male by many participants, both children and adults.
We did not want to give as much weight to each depiction of a female or male character in a crowd as we as we did to female or male characters who appeared alone or in a small group. Therefore, if more than six depictions appeared in a picture, we assigned the picture the sex of the majority. If there were equal numbers of female and male characters on a page, we did not code the page for sex.
Student t-tests were used to analyze interval and ratio data. Nominal data were analyzed with chi-square tests. An alpha level of .05 was used.
There were 75 male and 42 female title characters for a ratio of 1.8:1, X2 = 9.31 (1, N = 117), p < .002. For main characters, the frequencies were 95 male main characters and 52 female main characters, again a ratio of 1.8:1, X2 = 12.58 (1, N = 147), p < .001. The mean number of boy characters per book was .78 (SD = 1.05) and7=87the mean of girl characters was .63 (SD = 1.03), yielding a 1.2:1 ratio, t = -1.90 (199), p < .003. The adult character means were 2.34 for male adults (SD = 3.51) and 1.33 for female adults, once again a 1.8:1 ratio, t = -4.27 (184), p < .001. And finally, the mean number of male pictures per book was 42.90 (SD = 42.1) vs. a mean of 28.10 female pictures (SD = 41.16), yielding a 1.5:1 ratio, t = -5.44 (198), p < .001.
Exploratory Hypothesis 2
A test of exploratory Hypothesis 2 revealed that male authors accounted for a greater number of male than female title characters and main characters. Male authors wrote books with more male (f = 52) than female (f = 19) title characters, a ratio of close to 3 to 1, X2 (1, N = 71) = 15.34, p < .0001. Female authors did not favor one sex over the other; they produced 20 female title characters and 25 male title characters, X2 (1, N = 45) = .56, p = .46. Male authors also wrote about more male (f = 57) than female (f = 21) main characters, again a ratio of nearly 3 to 1, X2 (1, N = 78) = 16.62, p = .000. Female authors did not feature significantly more male (f = 33) than female (f = 23) main characters in their books, X2 (1, N = 56) = 1.78, p = .18. In addition, there were somewhat fewer female authors (N = 84) than male authors (N= 101), further adding to the over-representation of males in title character and main character roles.
Hypothesis 3, regarding the ways in which female versus male main characters were portrayed, was confirmed for nurtures/cares for another character and marginally supported for is seen indoors versus outdoors, but disconfirmed for rescues another character, is rescued by another character, passivebehaviors, and assertive/aggressive behaviors. As hypothesized, female main characters (M = .70, SD = .92) were more than three times more likely than were male main characters (M = .21, SD = .49) to perform nurturing or caring behaviors, t(125) = 3.89, p < .0001. Slightly more female main characters were found indoors than outdoors (59% versus 41%; 27 indoors versus 19 outdoors), whereas male main characters tended to be found outdoors more often (43%, indoors, 57% outdoors; 37 indoors versus 50 outdoors), X2 (1, N = 133) = 3.15, p = .08. Neither sex, contrary to the hypothesis, was more likely than the other to be portrayed as active or as passive. In fact, both male and female characters were portrayed more often as active than as passive; 86% of male and 79% of female characters were shown in active roles [female characters: 37 active, 10 passive, X2 (1, N = 47) = 15.52, p < .0001; male characters: 72 active, 12 passive, X2 (1, N = 84) = 42.86, p < .0001).
The three aspects of Hypothesis 4, which concerned occupations, were tested with chi-square analyses. As predicted, more of both female and male adults’ occupations were traditional than were non-traditional. Of 23 female adult characters shown with an occupation, 21 had stereotypically feminine occupations and only 2 had nontraditional occupations, X2 (1, N = 23) = 15.70, p < .0001.Of the 37 male adult characters with occupations, 33 were shown in stereotypically masculine occupations, and 4 were shown in nontraditional occupations, X2 (1, N = 37) = 22.72, p < .0001.
In addition and as hypothesized, male characters had a significantly broader range of occupations (32 different jobs) than did female characters (12 different jobs), X2 (1, N = 71) = 4.36, p = .04. Also as predicted, female characters were significantly more likely to show no evidence of an occupation outside the home (82 of 109; 75%) than male characters were (57 of 111; 51%), X2 (1, N = 220) = 13.48, p < .0001.
Exploratory Hypothesis 5
Only one marginally significant difference was found between the Caldecott sample and the popular book sample. Although both types of picture books had more male than female characters overall, the interaction between book type and degree of under-representation of female characters approached significance. Caldecott books under-represented female characters to a greater extent, with a ratio of 2.6 to 1, than did other popular books, with a ratio of 1.5 to 1 (Caldecott: male character M = 3.79, SD = 6.39; female character M = 1.47, SD = 1.77. Other popular books: male character M = 2.98, SD = 3.07; female character M = 2.00, SD = 2.19), F(1, 183) = 3.60, p = .06. There were no significant differences in gender representation between the Caldecott and non-award-winning books in title characters, main characters, or illustrations.
Exploratory Hypothesis 6
Analyses of changes in representation and portrayals of the sexes since the 1980s were based on non-statistical comparisons of our percentages and ratios to data from earlier studies, a method used in several previous time comparison studies (e. g., Collins , 1984; Dellman-Jenkins, Florjancic, and Swadener, 1993; and Kortenhaus and Demarest; 1993). Because there were no significant differences between our Caldecott and non-Caldecott samples on any of the variables relevant to this hypothesis, we combined our Caldecott and non-Caldecott books. In addition, we used 1980s and 1990s studies regardless of whether the research involved Caldecotts, non-Caldecotts, or both.
Title Characters. We found a ratio of 1.8 male title characters to every female character in our sample. Ratios in earlier studies ranged from 1.6:1 to 2.4:1. Five of the seven ratios fell within .2 of ours. Dellman-Jenkins et al. (1993) found a ratio of 1.6:1, Allen et al’s. (1993), Collins et al’s. (1984), and Kortenhaus and Demarest’s (1993) Caldecott sample (we were unable to combine the two samples for these comparisons) were all at or near 2:1 (2:1, 2:1, 2.1:1,respectively). Only Turner-Bowker (1996) and Kortenhaus and Demarest’s non-Caldecott sample showed ratios substantially different from ours (2.4:1 and 2.3:1, respectively). We find these mixed results inconclusive on the variable of title characters and, therefore, tentatively conclude that there has been no increase in the proportion of female characters since the 1980s.
Main characters. There was no clear pattern of change for main character. Our ratio was 1.8 male main characters for every female main character. Four of the seven measures of this variable for 1980s and 1990s books had ratios within.2 points of ours (Collins et al. (1984) found a ratio of 1.7:1; Kortenhaus and Demarest’s (1993) Caldecott ratio was 1.9:1, and their non-Caldecott ratio was 1.8:1; Williams et al.’s (1987) ratio was 1.6:1). The other three studies revealed closer to equal representation of the sexes than we found. Allen et al. (1985) reported a ratio of 1.2:1, Turner-Bowker’s (1996) ratio was = 1.4:1, and Dellman-Jenkins et al. (1993) actually found a reversed ratio (1 male character for every 1.3 female characters). No pre-2000 study showed a worse ratio of female to male main characters than we found in our study.
Pictures. No clear pattern was revealed in the comparison of 21st century books to books from the 1980s and 1990s. Our book sample’s male-to-female picture ratio was 1.5:1. Of the seven comparison analyses, two showed a distinctly higher male to female ratio of pictures than we found. Allen et al. (1985) found a ratio of 2.3:1, the ratio in the Dellman-Jenkins et al. (1993) study was 2.1:1. The other five studies revealed ratios within .2 points of ours: Collins, et al. (1984) found a ratio of 1.4:1; for Dougherty and Engel (1987) it was 1.7:1; Kortenhaus and Demarest’s (1985) Caldecott sample ratio was 1.4:1 and their non-Caldecott sample ratio was 1.5:1; Turner-Bowker’s (1996) ratio was 1.7:1; and Williams et al. (1987) found a ratio of 1.7:1.
Portrayals. In our study, portrayal analyses were made on main characters only. Earlier studies analyzed combinations of main characters, all characters, and pictures. Although this circumstance does not make for perfect comparisons between our results and those of the earlier research, it can reveal trends in gender portrayals.
In our book sample, about equal percentages of male main characters (86%) and female main characters (79%) were portrayed as active. Earlier researchers found varying results. Kinman and Henderson (1985) discovered a strong bias toward active male characters (84% active), as did we. But only 10% of female characters were portrayed as active in their sample. Williams et al. (1987) reported somewhat less bias than did Kinman and Henderson; 94% of male characters were classified as active, and 57% female characters were classified as active. Oskamp, Kaufman, and Wolterbeek (1996) reported 69% active male characters, and s, 77% active female characters—a slightly higher proportion of active female characters than active male characters. It appears that there may have been a positive change from the 1980s to the 1990s, then possibly a slip back for the 2001 book sample, though the numbers of studies for such a three-way comparison are small, and our sample did include a small number of books from before 2000.
Assertive/aggressive. There was no significant gender-related difference in our sample for assertive/aggressive behaviors. The two earlier studies we examined showed conflicting results. The ratio for Williams et al. (1987) was 4:1—four aggressive boys for every aggressive girl, but in contrast, Oskamp et al.’s (1996) ratio was 1:2. The sample for the latter study was quite small, however, as actual counts of 1 and 2 yielded the 1:2 ratio. Again, perhaps there has been a positive change, but the data do not allow for a strong conclusion.
Male and female characters were equally likely to perform rescue behaviors in our book sample, as was true for Williams et al. (1987). Oskamp et al. (1996) reported rescue behaviors, but again their counts were too small to be very meaningful (the three rescue behaviors yielded a 2:1 ratio). It is thus with little confidence that we can say that since the 1980s boys and girls have been equally likely to engage in rescue behaviors.
Female characters in our sample were much more likely than male characters to perform nurture behaviors; the ratio was 3.3:1. It appears that since the 1980s and 1990s, stereotyping on this dimension has slid back to pre-1980s levels. We found more bias than was found in studies of 1980s and 1990s books. Kinman and Henderson (1985), Oskamp et al. (1996), and Williams et al. (1987) found ratios of 2.7:1, 1.7:1, and 3:1, respectively.
In our book sample, 57% of male characters and 43% of female characters were portrayed outdoors. Tognoli et al. (1994) found that 61% of male characters and 48% of female characters were shown outdoors. Collins et al. (1984) did not give numbers, but reported a higher percentage of male than female characters outdoors. Allen et al. (1985), in contrast, found that 41% of male and 51% of female characters were portrayed outdoors. We interpret the comparison of our findings to these inconsistent pre-2000 data as an indication that there has been no change in recent decades.
Our results showed that about twice as many adult male as adult female characters were visibly employed outside the home (49% and 25%, respectively). Oskamp et al. (1996) found a 3:1 ratio. This comparison is encouraging, but with only one study to contrast to ours, we cannot be sure that the apparent reduction in stereotyping is real.
In our book sample men were shown in a wider range of occupations than were women; 32 different jobs were shown for men and 12 for women, a 2.7:1 ratio. These numbers demonstrate strong stereotyping, but earlier studies uncovered an even narrower range of job roles for women. Allen et al. (1985) reported men in 80 different jobs and women in only 20, a 4:1 ratio. Tognoli et al. (1994) reported 79 jobs for men and 22 for women, a 3.6:1 ratio. Thus there has been some improvement in the portrayals of women’s and men’s job range, but equity has not been reached.
Female characters are still under-represented in children’s picture books. Our data indicate that this claim is true not just for prize-winning books, but also for other books that today’s children and parents are likely to purchase. In our sample of 200 popular books, close to twice as many books had male title characters as had female title characters (75 versus 42) and had male main characters as had female main characters (95 vs. 52). There were also close to twice as many male as female adult characters overall per book (means of 2.34 and 1.33, respectively). For child characters overall the means were not as lopsided, with 24% more male than female child characters. Boys appeared in 53% more pictures than did girls.
In our sample of books, the disparity between numbers of female and male main and title characters is accounted for by author sex, such that under-representation of female characters is due to male authors alone. Our finding differs from those of Tognoli et al. (1994), Heintz (1987), and Kolbe and Lavoie (1981), who showed that in books published over several decades, female and male authors were equally likely to slight female characters. We view our result as reflective of an improvement; women writers may have stopped writing more about boys and men than about girls and women. We say “may have stopped” because, although for women authors the differences in numbers of male vs. female title characters (25 and 20, respectively) and male vs. female main characters (33 and 23) are not statistically significant, the raw numbers do not necessarily indicate that female authors no longer favor male characters. It would be of value to explore author sex differences further, to see whether there actually has been a reduction in women authors’ tendency to write about male characters and to compare men and women authors’ bias in portrayals of the sexes.
Another fact that contributes to under-representation of female characters in picture books is that so-called boys’ books, as well as boys’ movies and comics and music, outsell girls’ (Rider, 2000). If we lived in an egalitarian society that placed high values on girls and women and traditionally feminine qualities, then stories about girls and women, whether the characters played traditional, nontraditional, or mixed/neutral roles, would sell as well as books about boys and men. On the other hand, if it is action that sells books about boys, a broader representation of girls in active, outdoor roles could make girls’ books more popular among both girls and boys.
We should be aware that there may be more subtle ways in which the sexes are portrayed stereotypically. Perhaps authors consciously or unconsciously resort to subtle sexism because blatant sexism no longer passes unnoticed. Prejudice and stereotyping tend to go underground when their more overt forms become less socially acceptable, as, for example, in symbolic racism (Sears, Peplau, & Taylor, 1991). Many years ago Goffman (1976) found subtle visual sexism in his examination of gender bias in advertising. Such cues as height of placement of male and female models, posture differences, eye/head aversion, exaggerated differences in height and musculature, and occupation of space distinguished men from women. We found instances of such visual sexism in several picture books. For instance, on the title page of Pigs at Odds (Axelrod, 2000), the male character holds the car door open for the female character, who looks up at him while clutching a purse. On page 6 of In Daddy’s Arms I Am Tall (Steptoe, 2001), the black-suit-clad father walks the mother, who is wearing white lace, to church. He looks down at her; she looks down at the ground. These and other subtle forms of sexism should be formally explored in children’s picture books.
Occupational stereotyping has not gone underground. Men and women were much more likely to have traditional than non-traditional jobs. Men were seen in more than nine times as many traditional as non-traditional jobs and women were portrayed in traditional jobs over 10 times as often as they were portrayed in nontraditional jobs. For example, the predominant female adult character in Clifford's First School Day (Bridwell, 1999) is a teacher; she is a stewardess in Alligator Tales (Smeeton, 2001), a librarian in Hopping Hens Here! (Gilkow, 2000), and a maid in Mr. Willowby's Christmas Tree (Barry, 2000). Predominant female characters in other books appeared as nannies, servants, nurses, dressmakers, quiltmakers, and dancers.
In addition, whereas only a small proportion of men or women clearly had jobs outside the home, over two-thirds of those who did were men. Although women’s range of jobs has increased in relation to men’s in the past two decades, men were still shown in a much wider range of jobs than women were (2.7:1).
One important goal of our study was to test whether Caldecott books are representative of the books read to children today. The fact that we found only one borderline significant difference in the degree to which Caldecott vs. other picture books under-represent female characters suggests that Caldecott books may represent the overall category of children’s picture books well. We believe, however, that researchers should continue to explore this comparison. Comparisons within and across studies between Caldecott and other books have yielded inconsistent results. For example, Nilsen (1978), in her analysis of Caldecott books, found very different patterns of change over time than did Clark et al. (2003) in their examination of non-award-winners. Tepper and Cassidy (1999) studied both categories and found several differences between their Caldecott and non-Caldecott samples. Continued systematic research might reveal whether there are any consistent and noteworthy differences in the gender bias of contemporary examples of the two groups of picture books.
Our measures of change since the 1980s reveal that we have entered a period of leveling out in the representation of girls and women, and that the trend is mixed for portrayals. The male-to-female ratios of title and main characters and pictures remain poor, between 1.5:1 and 1.8:1. There is weak evidence that there is less gender stereotyping in aggressive and active behaviors and in one aspect of employment than there was in the previous 20 years, but women continue to be less likely than men to be visibly employed outside the home, and they continue to have a narrower range of jobs. There is no change in the portrayal of female vs. male characters outdoors or indoors; there is a continued slight tendency for boys and men to be seen outdoors more than girls and women are. Nurturant behaviors are even more likely now than in the 1980s and 1990s to be performed exclusively by girls and women.
In conclusion, modern children’s picture books continue to provide nightly reinforcement of the idea that boys and men are more interesting and important than are girls and women. Every hypothesis we tested concerning numbers of male vs. female characters was confirmed, as were all hypotheses about occupational stereotyping. Two of the five behavioral measures we assessed showed stereotypical differences between the sexes. With attention to the persistent problems inherent in children’s picture books, parents, teachers, and librarians can choose selectively for balanced portrayals of gender roles until the time when authors and publishers provide us with such balance.