Signorella & Liben (1985) Assessing Children's Gender-Stereotyped Attitudes

The article below was originally published in Psychological Documents, which changed names to Social and Behavioral Sciences Documents in 1986 and then ceased publication in 1988.
The citation for the article is: Signorella, M. L., & Liben, L. S. (1985). Assessing children's gender-stereotyped attitudes. Psychological Documents, 15, 7. (Ms. No. 2685). The measure described in this article is listed in C. A. Beere. (1990). Gender roles: A handbook of tests and measures (pp. 228-230).New York: Greenwood.
For more information, contact M. L. Signorella.

Assessing Children's Gender-Stereotyped Attitudes

Margaret L. Signorella and Lynn S. Liben

Pennsylvania State University

Synopsis. Theoretical and methodological reasons for distinguishing attitudes about gender stereotypes from knowledge of gender stereotypes are outlined. Critical features of a gender attitude measure are identified and incorporated into a measure of children's gender-stereotyped attitudes (Gender-Stereotyped Attitude Scale for Children [GASC]). The GASC was given to three samples of first and second-grade children (Ns = 60, 109, 110), and to one sample of kindergarten, second-grade, and fourth-grade children (N = 130). For a variety of activities and occupations that had been judged by adults as masculine, feminine, or neutral, children were asked "Who can...." or "Who can be a...." Children who gave above the median number of "both men and women" responses were classified as low in stereotyping of attitudes, while children who gave at or below the median number of "both men and women" responses were classified as high in stereotyping of attitudes. Results concerning (1) item analyses, (2) test validity, and (3) age and sex comparisons are reported. First, for all four samples, the GASC had high internal consistency, with coefficients ranging from .83 to .95. The individual items in the measure discriminated between children high and low in stereotyped attitudes. Children tended to give stereotyped responses to the masculine and feminine items and unstereotyped ("both men and women") responses to the neutral items, indicating that children were able to use the "both" category correctly. Second, test validity was demonstrated by the finding that stereotyping classification on the GASC was positively related both to stereotyping classification on a trait measure of stereotyping, and to memory for gender-related pictures. Third, there were no significant sex differences in any of the four samples. In the sample that covered a relatively wide age range, fourth graders and second graders gave significantly more "both men and women" responses than did kindergartners (ps < .01). The importance of distinguishing between attitudes and knowledge in interpreting past work and in planning for future investigations was discussed.

INTRODUCTION

Children's gender stereotypes have been of interest for both theoretical and practical reasons. Cognitive theories of gender identity, for example, link the acquisition and persistence of gender typing to children's stereotyped schemas about the characteristics and behaviors of females and males (e.g., Bem, 1981; Kohlberg, 1966; Martin & Halverson, 1981). Indeed, most theories of gender role development attempt to explain how children come to learn and enact gender-stereotyped characteristics (e.g., see Huston, 1983). Similarly, researchers and practitioners have been concerned with the effects of restrictive gender stereotyping on educational goals, career choices, and self-concept. For example, there has been interest in documenting the degree of stereotyping in children's career choices (e.g., Bacon & Lerner, 1975), as well as in designing intervention programs to alter the stereotyping in children's views of men's and women's roles (e.g., Guttentag & Bray, 1976).
Two related, but separable issues are implicit in the theoretical and applied concerns about gender stereotyping. The first issue concerns children's understanding of which activities and behaviors most people in our culture assign to either women or men. We will refer to this aspect as knowledge of gender stereotypes. Just as children may be expected to become increasingly knowledgeable across a wide variety of cognitive and social domains, children may be expected to become increasingly knowledgeable about gender stereotypes with age. The second issue concerns the degree to which children agree with the gender stereotypes that exist in our culture. We will refer to this aspect as attitudes toward gender stereotypes. Just as developing individuals can increasingly come to appreciate a range of perspectives and become capable of hypothetical reasoning (e.g., Inhelder & Piaget, 1969; Perry, 1970), individuals may be expected to become increasingly flexible in their attitudes about gender stereotypes as they develop. Based on the cognitive-developmental approach to gender identity, in particular, one would expect to find a peak in stereotyping of gender attitudes around kindergarten (e.g., Kohlberg, 1966; Kohlberg & Ullian, 1978), followed by increasing flexibility in attitudes with age (e.g., Rebecca, Hefner, & Oleshansky, 1976; Robinson & Green, 1981).
These two aspects of gender stereotyping--knowledge and attitudes--have often been confused in developmental research (for one of the exceptions, see Carter & Patterson, 1982). It is important to recognize that the two aspects demand somewhat different assessment procedures. The focus of the present paper is on the assessment of children's gender-stereotyped attitudes. Although past researchers have designed scales to measure children's gender attitudes, the necessary features of a scale of this kind have not been made sufficiently explicit. In the present paper we will first review some of the issues that should be considered in building such a scale. Specifically, we will consider: (1) the type of response options, (2) the number of response options, (3) the type of question asked, and (4) the content of the scale items. Second, because little information (apart from reliability) is available for prior scales, we present detailed information about a particular scale (Gender- Stereotyped Attitude Scale for Children [GASC]) developed to incorporate the desired features.
Probably the single most critical feature distinguishing a measure of attitudes from one of knowledge is the type of response options. When one is interested in children's knowledge of gender stereotypes, it is appropriate to ask children to assign items either to males or to females. Forced choice procedures of this kind maximize the likelihood of eliciting stereotyped responses (Beere, 1979, Ch. 4). As would be expected with a measure of knowledge, most developmental studies in which only "male" or "female" responses were permitted report age-related increases in gender-stereotyped responses (Albert & Porter, 1983; Best, Williams, Cloud, Davis, Robertson, Edwards, Giles, & Fowles, 1977; Coker, 1984; Edelbrock & Sugawara, 1978 ["sex-role discrimination" measure]; Gettys & Cann, 1981; Leahy & Shirk, 1984; Masters & Wilkinson, 1976; Nadelman, 1974; Reis & Wright, 1982; Thompson, 1975; Williams, Bennett, & Best, 1975; but not Kuhn, Nash, & Brucken, 1978).
In contrast, when one is interested in assessing attitudes, the procedure must allow the child's attitudes room to operate. This may be accomplished by giving children the explicit option of responding that an activity can be done by both males and females. Results of studies that provide a "both men and women" option are generally consistent with the two age-related predictions of cognitive developmental theories noted earlier, that is, an increase in stereotyping up to kindergarten, followed by a decrease in stereotyping thereafter. Specifically, all four studies that included a "both" option with children younger than kindergarten found the predicted peak in stereotyping at kindergarten (Archer, 1984; Flerx, Fidler, & Rogers, 1976; Schau, Kahn, Diepold, & Cherry, 1980; Urberg, 1982). Furthermore, virtually all studies that included a "both" response with kindergarten through elementary school children reported decreases in stereotyping (increases in "flexibility") of attitudes with age (Bacon Lerner, 1975; Carter & Patterson, 1982; Cummings & Taebel, 1980; Garrett, Ein, & Tremaine, 1977; Kierscht & Rice, 1981; Marantz & Mansfield, 1977; Meyer, 1980; Shepard & Hess, 1975; Umstot, 1980; Urberg, 1982). The exceptions to this generalization tend to be studies with narrow age ranges, which simply find no age effects (Jones & McBride, 1980; Lerner, Benson, & Vincent, 1976; Zuckerman & Sayre, 1982).
The pattern of results reported above suggest that at a minimum, an attitude measure must include a "both men and women" option in addition to the "men" and "women" options included in knowledge measures. A second concern is whether the number of response options should be limited to these three, or should instead include additional choices as well. Some past assessment procedures have included additional response options. Garrett et al. (1977) used a Likert-type response format which permitted one of five responses: "only women;" "mostly women, a few men;" "women and men;" "mostly men, a few women;" "only men." They reported from pilot testing that younger children in particular had more difficulty with the intermediate "mostly...a few..." options. Since the children's errors tended to be random or toward the neutral option, Garrett et al. (1977) argued that such problems would only serve to lessen the chances of finding their hypothesized decrease in stereotyping from early to late elementary school. For other research questions, however, the additional noise in the data may present a more serious problem. Attempts to correlate the stereotyping of attitudes with other measures would be hindered by the presence of random error. In addition, finding a predicted increase in. stereotyping up to kindergarten would be difficult if younger children use the neutral option when confused. Thus, to facilitate use of stereotyping measures with a wider age range, it would be preferable to give children only three choices: "men," "women," and "both men and women."
Giving children a clear neutral choice is also important for theoretical reasons. If a Likert-type response format is used, an average score of neutral may be obtained from a variety of response patterns. A child may have given consistently neutral responses, or a mixture of stereotyped and neutral responses. However, the prediction of increasing flexibility with age, for example, is a prediction that age-related increases will occur in the frequency with which children make the unstereotyped "both men and women" response. It is appropriate, then, to focus on children's tendency to use the "both" category in responding to the question of which sex can participate in an activity.
Third, the type of question used to elicit children's responses will also affect the operation of their attitudes in the response. When Carter and Patterson (1982) asked their subjects who "usually" does an activity, the subjects gave mostly stereotyped responses and showed age-related increases in such responses (i.e., knowledge). In contrast, when the children were asked to indicate who "can" do the activities, they used the "both men and women" category and were increasingly flexible with age. In sum, asking children who can (or should [e.g., Shepard & Hess, 19751) do an activity and allowing them a "both men and women" response will be effective in eliciting children's attitudes.
Finally, it is important to consider the content of items to be included on a gender attitude scale designed for children. The pool of potential items includes activities, occupations, and traits. Previous research on knowledge of stereotypes has shown that even kindergarten children are almost as knowledgeable as adults with respect to gender assignment of activities and occupations (e.g., Gettys & Cann, 1981; Nadelman, 1974). In contrast, preschoolers have relatively incomplete knowledge of trait stereotypes (e.g., Albert & Porter, 1983; Williams et al., 1975) and thus could not be tested meaningfully with trait items. Because many research questions (e.g., critical age comparisons in testing cognitive-developmental theory) require testing children younger as well as older than kindergarten, an activity/occupational stereotyping measure would be useful in more situations than would be a trait measure.
The four considerations outlined above were central to the development of the gender attitude scale (GASC) described here. Various activities and occupations were named, and children were asked to indicate who can participate in them. Children were given the three response options described above ("men," "women," and "both men and women"). On the basis of the number of "both men and women" responses given, children were classified as high or low in stereotyping of their attitudes.
The questionnaire containing masculine, feminine, and neutral activities and occupations was given to four samples of children. Three samples consisted of first and second graders, while the fourth had kindergartners, second graders, and fourth graders. To evaluate the stereotyping measure, item analyses and information on test validity are reported. The responses of boys and girls were compared to assess the presence and reliability of sex differences in stereotyped attitudes. Finally, using the sample with the wider age range, it was possible to test the hypothesis derived from cognitive-developmental theory that children become less stereotyped in their attitudes with age.

METHOD

Subjects

Sample 1-2 A. The subjects were 36 first graders (19 boys and 17 girls) and 24 second graders (13 boys and 11 girls) from a predominantly white, suburban school in southwestern Pennsylvania (Liben & Signorella, 1980).
Sample 1-2 B. The subjects were 51 first graders (26 boys and 25 girls) and 58 second graders (30 boys and 28 girls) from a predominantly white, suburban school in southeastern Michigan.
Sample 1-2 C. The subjects were 57 first graders (20 boys and 37 girls) and 53 second graders (29 boys and 24 girls) from a predominantly white, rural school in central Pennsylvania.
Sample K-2-4. The subjects were 42 kindergartners (20 boys and 22 girls), 40 second graders (21 boys and 19 girls), and 48 fourth graders (25 boys and 23 girls) from a predominantly white, rural school in southeastern Michigan (Signorella & Liben, 1984, Study 1).

Stereotyping Measures

Activities and occupations. Two issues guided the selection of the items for the GASC. First, since this measure was intended to tap children's stereotyped attitudes, the items had to reflect shared beliefs in this culture about the groups "men" and "women" (see Ehrlich, 1973, Ch. 2; Sherif, 1976, Ch. 10, for definitions of stereotypes). Thus, the items were chosen on the basis of consensus in adult judges' ratings, rather than on the basis of Census data (e.g., Garrett et al., 1977). The neutral items were, therefore, those consistently judged neutral. Second, the masculine as well as the feminine items included both household activities and outside occupations.
Included in the measure were 14 masculine, 14 feminine, and 7 neutral items. The complete list of masculine and feminine items is contained in Table I in the Results section. The neutral items were: work a hard puzzle, like to go to the beach, ride a bicycle, like to do things outside, like to go to the movies, play cards, and act in a play. The items chosen came from one of three sources. Some of the items had been included previously in measures used by Nadelman (1970, 1974) and Flerx et al. (1976). The remaining items were chosen from a list of activities and occupations given to 27 female and 16 male college students. The subjects were asked to rate each item as masculine, feminine, or neutral. Those items designated as stereotyped had been placed in one category significantly more frequently than in the other two.
Traits. To provide information on test validity, an additional stereotyping measure, taken from Williams et al. (1975) was administered to Sample K-2-4. This measure was originally a measure of knowledge, as children in Williams et al. were only given two choices for their answers: men or women. To convert this scale to an attitude measure, children were also given a "both men and women" response option (see Procedures). Each of 12 masculine and 12 feminine traits was represented by a word or phrase that would be understandable to kindergartners. The masculine items were: gets into fights, is sure of themselves, is a messy person, works hard in getting ahead, goes on adventures, can work by themselves, makes careful plans, brags about things they have done, says bad words, makes up most of the rules, talks loudly, is the strong person. The feminine items were: is the gentle person, daydreams, gets very upset and excited, is always changing their mind, talks a lot, buys silly things, is always fussing, is the complaining person, is soft hearted, always says thank you, does everything just right, is the shy person.

Procedure (1)

Sample 1-2 A. The children were tested individually by a female experimenter. The procedure was adapted from Flerx et al. (1976). Children were told, "I've got a list of activities here, and I'd like you to tell me whether you think these things can be done by men, by women, or by both men and women. Let's practice on a few. Can you show me who can be a Daddy? Show me who can be a Mommy. Show me who can have two legs." To give their answers, children were instructed to point to one of three line drawings of figures, mounted on cardboard: two men, two women, and a man and a woman. The positions of the three sets of figures were varied between children. Children were asked, for each of the 35 items, to "Show me who can..." The order of presentation of items was randomized between children.
Sample 1-2 B. The children were tested in small groups by a female experimenter. An attempt was made to group the children by reading level, based on teacher recommendations. Each child was given a booklet with mimeographed stick figures. Each item in the stereotyping measure was typed in large letters above a line containing two stick-figure men (trousers and no hair), two stick-figure women (skirts and long hair), and a stick-figure man with a stick-figure woman. The order of items was randomized, but each child received the same random order of items. Children were given the same instructions and practice items as in Sample 1-2 A, except that the children were asked to show who could do each item by circling or crossing out the appropriate pair of figures. After making sure on the practice items that all children understood the directions and were filling out the sheet properly, the children were led through the list of 35 items, each prefaced by "Show me who can.... "
Sample 1-2 C. The children were tested individually by one of three female experimenters using the procedure described for Sample 1-2 A.
Sample K-2-4. The children were tested individually by either a male or female experimenter. Both the GASC and the trait stereotyping measure were administered using the procedure described for Sample 1-2 A. The order of presentation of the measures was randomized, as was the order of presentation of items within each measure.

RESULTS AND DISCUSSION

Each child's score on the stereotyping measure was the number of "both men and women" responses given to the stereotyped (masculine and feminine) items. Internal consistency reliabilities (as measured by Kuder-Richardson formula 20) for the four samples, in order, were .83, .86, .95, .91.
For each sample, the median stereotyping score was determined. The medians for Samples 1-2 A, 1-2 B, and K-2-4 were 7, 9, and 6, respectively, out of a possible 28.(2) The median for Sample 1-2 C was 12.5, out of a possible 26. Children were classified as high (< Mdn) or low (> Mdn) in stereotyped attitudes. Stereotyping classification did not differ as a function of experimenter in Samples 1-2 C or K-2-4, so the data from the experimenters were combined.

Item Analysis

To examine whether or not high and low stereotype children responded differently to the individual items in the stereotyping measure, the number of "both" responses given by the two groups in each sample was compared (see Table 1). This procedure is comparable to computing item-total correlations. Because the patterns of responses for the three grades in Sample K-2-4 were the same, the data were combined. As can be seen in Table 1, with few exceptions the items consistently discriminated between children of differing attitudes.

Table 1
Percentages of High and Low Stereotype Children Giving a "Both Men and Women" Response to the Masculine and Feminine Items
	1-2 A		1-2 B		1-2 C		K-2-4
	HS	LS	HS	LS	HS	LS	HS	LS
(n =)	(30)	(30)	(52)	(57)	(55)	(55)	(69)	(61)
ITEM TYPE Masculine
Doctor	57	37	52	18	75	24	47	12
Mow the lawn	37	13	46	7	91	67	36	2
Carry their own suitcase on a trip	70	60	92	72	95	80	51	28
Cook in a restaurant	70	30	88	52	--^a	--^a	52	10
Fly a plane	70	10	58	14	80	29	45	6
Fix a car	3	0	27	0	38	4	14	0
Play football	3	0	15	0	40	4	5	2
Go fishing	47	3	46	5	89	36	34	8
Drive a truck	40	3	52	4	75	24	44	1
Shovel snow	43	3	71	21	89	45	32	5
Be an umpire	20	2	33	4	51	7	21	4
Fix a sink	3	0	35	2	36	4	17	3
Collect garbage	7	10	31	4	85	24	23	6
Race a car	7	0	10	5	75	11	27	4
Feminine
Cook in a kitchen	27	3	37	9	89	22	23	1
Set the table	43	3	48	12	85	25	27	1
Wash dishes	43	13	35	11	91	20	21	1
Hug other people a lot	87	50	87	82	96	91	49	30
Take care of children	63	23	83	49	96	47	36	14
Use a sewing machine	7	0	15	7	53	5	6	0
Bake cupcakes	20	0	35	2	71	16	15	1
Make a hat	73	33	79	42	--^a	--^a	38	12
Ballet dancer	20	0	52	12	35	13	27	3
Nurse	17	0	31	1	60	15	21	2
Teacher	93	47	94	67	98	71	57	16
Secretary	33	27	63	30	78	24	29	5
Telephone operator	80	47	73	18	85	31	30	5
Clean the house	37	3	31	9	65	15	19	2
Note. The following abbreviations were used: HS = High Stereotype, LS = Low Stereotype. ^aThis item was eliminated from the measure.

Neutral Items

The numbers of "both" responses to the masculine, feminine, and neutral items were determined, and correlations computed among these three totals for the four samples (see Table 2). Because the pattern of correlations was the same for all three grades in Sample K-2-4, the data were combined. If children are discriminating between stereotyped and gender-neutral items in the use of the "both" category, then there should be a significantly greater association between responses to masculine and feminine items than between the responses to masculine and neutral items or to feminine and neutral items. To test the significance of the differences in the correlations, the procedure described by Steiger (1980) for dependent correlations was used. First, an overall test was done on each of the correlation matrices obtained from the samples to determine if any significant differences were present in a matrix. In all four cases, a significant value was obtained, (3, N = 60) = 40.9, (3, N = 109) 146, (3, N = 110) = 219, (3, N = 130) = 207, and for Samples 1-2 A, 1-2 B, 1-2 C, and K-2-4, respectively. Next, pairwise comparisons were made within each matrix. For all four samples, there were significant (p < .01) differences between the masculine-feminine and masculine-neutral correlations, t (57) = 3.22, t (106) = 4.47, t (107) = 6.58, and t (127) = 5.16, for Samples 1-2 A, 1-2 B, 1-2 C, and K-2-4, respectively. Likewise, the feminine-neutral correlations were significantly lower than the masculine-feminine correlations (p < .01) for all four samples, t (57) = 3.99, t (106) = 4.11, t (107) = 7.52, and t (127) = 5.16, for Samples 1-2 A, 1-2 B, 1-2 C, and K-2-4, respectively.
The significantly smaller correlations with the neutral items are not surprising given the distribution of responses to those items. Even though the children in all four samples were responding in a fairly stereotyped fashion on the masculine and feminine items, the neutral items tended to elicit "both" responses. For example, the median numbers of "both" responses (out of 7) to the neutral items were 6, 6, 7, and 5.5, for Samples 1-2 A, 1-2 B, 1-2 C, and K-2-4, respectively. Thus, it appears that children are able to use the "both" category in responding to these questionnaire items. It is unlikely, then, that children are using the "both" category just to be agreeable or because of ignorance.

Table 2
Correlations Among Masculine, Feminine, and Neutral Items.
	Sample
Item Pair	1-2 A	1-2 B	1-2 C	K-2-4
M-F	.65	.74	.85	.78
M-N	.25	.44	.47	.46
F-N	.19	.46	.43	.49
Note. The following abbreviations were used: M = Masculine, F = Feminine, N = Neutral.

Sex and Age Comparisons

Two methods were used to compare the stereotyping levels of the various age and sex groups. First, children were cross-classified according to sex (female, male), grade (1, 2 or K, 2, 4), and stereotyping of attitudes (high [< Mdn], low [> Mdn]). To test for sex and grade differences, a sex X grade X stereotyping log-linear analysis was done on these frequencies. For Samples 1-2 A, 1-2 B, and 1-2 C there were no significant associations. For Sample K-2-4 there was a grade X stereotyping interaction, (2, N = 130) = 6.35, p < .05. Kindergartners were more likely to be classified as highly stereotyped, while fourth graders were more likely to be classified as low in stereotyping.
For the second analysis, sex X grade ANOVAs were done on the number of "both" responses. The means are displayed in Table 3. The same pattern of results was obtained. That is, there were no significant sex or grade effects in Samples 1-2 A, 1-2 B, and 1-2 C. In Sample K-2-4 there was a main effect for age, F(2, 124) = 7.08, p < .01. Newman-Keuls tests indicated that fourth and second graders gave significantly more "both" responses than did kindergartners (ps < .01). The interaction of sex and grade was not significant, F(2, 124) = 2.04, p = .13. (See also the footnote to Table 3.)

Table 3
Mean Number of "Both Men and Women" Responses to the Masculine and Feminine Items as a Function of Grade, Sex, and Sample
	Sample 1-2 A	Sample 1-2 B	Sample 1-2 C	Sample K-2-4^a
Kindergarten
Girls				5.9 (4.2)
Boys				4.8 (4.3)
First Grade	8.4 (4.1)	10.9 (7.5)	13.1 (6.1)
Girls	6.0 (4.5)	7.6 (5.1)	12.8 (6.4)
Boys
Second Grade
Girls	7.3 (3.2)	9.8 (4.9)	11.9 (7.4)	8.4 (5.9)
Boys	9.5 (5.4)	9.8 (4.8)	15.2 (7.6)	7.8 (6.0)
Fourth Grade
Girls				8.5 (7.2)
Boys				12.1 (7.8)
Note. Standard deviations are in parentheses. ^aThere were two typographical errors in the original paper on the means for Sample K-2-4 (the kindergarten girls’ mean was 6.0 instead of 5.9, and the means for the fourth grade girls and boys were reversed). These errors have been corrected in this table.

As predicted, then, children became less stereotyped in their attitudes with age, as has been found with other attitude measures (e.g., Garrett et al., 1977). Although these changes are consistent with cognitive-developmental theory, it is not yet possible to identify the underlying mechanism(s) of change (e.g., see Carter & Patterson, 1982; Leahy & Shirk, 1984; Tremaine, Schau, & Busch, 1982). It does seem safe to conclude, however, that the age differences are not due to cohort effects. A cohort effect would be more likely to result in greater rather than reduced stereotyping with age in cross-sectional comparisons, given the trend toward decreased rigidity of gender-stereotyped attitudes in recent years (e.g., Huston-Stein & Higgins-Trenk, 1978). Support for the hypothesis of age rather than cohort effects can be found in a study by Lerner, Vincent, and Benson (1976). Elementary school girls who had been given a measure of gender-stereotyped attitudes were retested one year later. There were significant decreases in the stereotyping of their attitudes.
There were no significant sex-related differences in any of the samples tested, unlike some of the previous work on children's gender-stereotyped attitudes. Although much of the research on gender-related attitudes has been conducted with girls only, those studies that do include both sexes and that find sex differences report that girls are less stereotyped than boys (Archer, 1984; Cummings & Taebel, 1980; Flerx et al., 1976; Frost & Diamond, 1979; Garrett et al., 1977; Payne, 1981; Shepard & Hess, 1975; Urberg, 1982). The consistency of the present results across the four samples suggests that previous reports of greater stereotyping in boys must be considered a function not only of the sex of subjects but also of the items included in the questionnaire.
Traits. As noted in the Method, the children in Sample K-2-4 were given an additional attitude measure composed of traits (12 masculine & 12 feminine, Williams et al., 1975) to provide a comparison with the GASC. The median number of "both" responses was 10. Children were classified as high (< 10) or low (> 10) in the stereotyping of their attitudes. Children's stereotyping classification on the traits was related to their stereotyping classification on the GASC,

(1, N = 130) = 20.65, p < .01. This highly significant association occurred because 69% of the children received the same classification on both measures.
A sex (female, male) X grade (K, 2, 4) X stereotyping (high, low) log-linear analysis revealed a grade X stereotyping interaction,

(2, N 130) = 6.07, < .05. As with the GASC, kindergartners were more likely to be classified as highly stereotyped while fourth graders were more likely to be classified as low in stereotyping. Comparable results were obtained from a sex X grade ANOVA on the number of "both" responses. There was a main effect for grade, F (2, 124) = 3.97, p < .05. Fourth graders (M = 12.5) gave significantly more "both" responses than did kindergartners (M = 7.8). Second graders (M = 10.2) did not differ significantly from the other two grades.
Validity

The construct validity of the GASC was demonstrated in two ways. First, the results of the age comparisons in Sample K-2-4 using this measure conform to the expectations of cognitive-developmental theory. Specifically, less stereotyping with age was observed in that fourth graders were more likely to be classified as less stereotyped and kindergartners were more likely to be classified as more stereotyped. It is interesting to note that less stereotyping with age was observed on the trait measure as well. These age trends support our contention that the precise nature of the question posed in the assessment procedure will determine whether attitudes or knowledge are being measured, given that the items are the same and familiar to most subjects.
Second, classification of children as more or less stereotyped on the GASC is related to children's memories for pictures of traditional and nontraditional sex-role behavior. Highly stereotyped children showed better recognition (Liben & Signorella, 1980 (Sample 1-2 A]) and recall (Signorella & Liben, 1984, Study I (Sample K-2-4] and Study 2) for traditional versus nontraditional pictures. Less stereotyped children either showed no differential memory for traditional versus nontraditional pictures (Liben & Signorella, 1980; Signorella & Liben, 1984, Study 2) or better memory for nontraditional pictures (Signorella & Liben, 1984, Study I).
The convergent validity of this measure was demonstrated by the comparison in Sample K-2-4 to the responses to the trait measure. There was a significant tendency for children to receive the same classification on both the GASC and the trait measure.

Conclusions

In the present paper we have argued for making a distinction between knowledge of stereotypes and attitudes toward stereotypes. There are two major reasons why such a distinction is important. First, the course of development for knowledge may be expected to differ from the course of development for attitudes. Second, the relationship between behavior and knowledge may differ from the relationship between behavior and attitudes. Each of these issues is discussed in turn.
With respect to developmental change, knowledge of gender stereotypes increases with age whereas stereotyping of attitudes decreases with age. Knowledge of stereotypes, particularly about activities and occupations, reaches ceiling levels rather quickly during childhood. In addition, Ruble (1983) has concluded that knowledge of stereotypes has changed little in the last decade. Thus, the most interesting questions about knowledge of stereotypes will concern how early these stereotypes are acquired by children and to what other kinds of early learning they are related (e.g., gender constancy). In contrast to knowledge, both historical and developmental changes in gender-related attitudes have been observed. Little information is known, either, about the reasons for these age-related changes, or about the relationship between ontogenetic changes and historical changes.
The second reason for distinguishing between attitudes and knowledge is that the two aspects will probably have different relationships with behavior. We have argued that there will be little individual variation in knowledge in children, particularly from kindergarten. In our four samples, for example, children were rarely "wrong" in their assignment of activities to the stereotyped categories. If behavior in gender-related situations is of interest, then only attitude measures can reflect individual differences in acceptance of gender stereotypes.
We believe that the failure to distinguish clearly between knowledge and attitudes may be responsible in part for inconsistent findings in past research relating children's gender stereotyping to cognitive processes. A relationship between individuals' stereotyping and memory has been reported in some but not all studies. As noted earlier, we (Liben & Signorella, 1980; Signorella & Liben, 1984) have found a relationship between stereotyping of children's gender-related attitudes and patterns of memory for gender-relevant material. In contrast, Cann and Newbern (1984) failed to find a relationship between 6- and 8-year-olds' stereotyping and memory. However, Cann and Newbern used a stereotyping measure with a large knowledge component which, given the argument presented earlier, would presumably have obscured rather than clarified the relationship between individuals' stereotyped schemes and memory.
In addition to the two kinds of stereotyping measures discussed thus far (i.e., knowledge, most commonly measured by a choice between what males do versus what females do, and attitudes, most commonly measured by providing a "both men and women" choice as well), some investigators have used measures that assess individuals' own preferences for stereotypic activities. Conceptually, it might be expected that there would be some relationship between an attitude measure, such as we have developed, and a preference measure because both measures appear to tap a flexibility component. In other words, both measures may reflect children's willingness to endorse stereotypic feminine and masculine items. Subjects who can accept both types of items for themselves and for people in general may then demonstrate other types of flexibility (e.g., in memory for traditional and nontraditional sex roles).
Indeed, such a relationship between preferences and attitudes has been demonstrated in adults. Helmreich, Spence, and Holahan (1979), for example, have found that responses to the Attitudes Toward Women Scale (a measure of attitudes toward behaviors in which women and men might engage) are related to individuals' own preferences for engaging in stereotyped activities. To our knowledge, there has been only one study in which children's own preferences for gender-stereotyped activities have been related to memory for stereotypic and nonstereotypic material. Specifically, Martin and Halverson (1983) gave kindergarten children a measure of preference for stereotyped activities ("sex-role preference" from Edelbrock & Sugawara, 1978). When memory responses from a cued recall task were scored in a manner similar to that used by Liben and Signorella (1980), a relationship was found between preferences and types of pictures recalled, paralleling the results reported by Liben and Signorella between attitudes and memory. Unlike the findings of Signorella and Liben (1984), however, Martin and Halverson did not find a relationship between individuals' preferences and mnemonic reconstructions. Additional research is needed to determine whether the difference in findings from the two studies is due to the relatively more restricted variability in preferences reported by Martin and Halverson, or to the difference between an attitude and a preference measure, or to yet some other procedural difference between the two studies.(2)
In summary, we have presented information on a measure of children's gender-stereotyped attitudes. The measure was shown to be internally consistent, and the individual items discriminated between children who were high or low in the total number of unstereotyped responses. There were no significant sex-related differences in stereotyping of attitudes in any of the four samples. In the sample with the wider age range, children were less likely with age to be classified as highly stereotyped. Children who were high versus low in stereotyping also showed qualitative differences in their responses to another stereotyping measure, and in their memory for stereotyped pictures. Finally, we have suggested that a clear distinction between attitudes and knowledge may help in understanding results from past studies relating memory to gender stereo typing, and may suggest interesting-directions for future work.

ACKNOWLEDGEMENTS

This paper is part of a continuing, collaborative effort; thus, order of authorship is arbitrary. We would like to thank the administrators, teachers, parents, and students at Elm Grove School, Peters Township School District, McMurray, Pennsylvania; Isbister Elementary School, Plymouth School District, Plymouth, Michigan; Dundee Community Schools, Dundee, Michigan; and the Bellefonte School District, Bellefonte, Pennsylvania, for their cooperation. The help of Michael Blum, Wesley Jamison, Michael Jimenez, Nancy Martinelli, Ann Merriwether, Sharon Paynter, and Karen Takach in data collection and tabulation is also appreciated.

ADDRESSES and AFFILIATIONS

Corresponding author: Margaret L. Signorella, Penn State McKeesport, 4000 University Drive, McKeesport, PA 15132-7698; e-mail: msignorella@psu.edu. Lynn S. Liben is at Penn State’s University Park Campus.

REFERENCES

Albert, A. A., & Porter, J. R. (1983). Age patterns in the development of children's gender-role stereotypes. Sex Roles, 9, 59-67.

Bacon, C., & Lerner, R. M. (1975). Effects of maternal employment status on the development of vocational-role perception in females. Journal of Genetic Psychology, 126, 187-193.

Beere, C. A. (1979). Women and women's issues: A handbook of tests and measures. San Francisco: Jossey-Bass.

Bem, S. L. (1981). Gender schema theory: A cognitive account of sex typing. Psychological Review, 88, 354-364.

Best, D. L., Williams, J. E., Cloud, J. M., Davis, S. W., Robertson, L. S., Edwards, J. R., Giles, H., & Fowles, J. (1977). Development of sex-trait stereotypes among young children in the United States, England, and Ireland. Child Development, 48, 1375-1384.

Cann, A., & Newbern, S. R. (1984). Sex stereotype effects in children's. picture recognition. Child Development, 55, 1085-1090.

Carter, D. B., & Patterson, C. J. (1982). Sex roles as social conventions: The development of children's conceptions of sex-role stereotypes. Developmental Psychology, 18, 812-824.

Cummings, S., & Taebel, D. (1980). Sexual inequality and the reproduction of consciousness: An analysis of sex-role stereotyping among children. Sex Roles, 6, 631-644.

Edelbrock, C., & Sugawara, A. 1. (1978). Acquisition of sex-typed preferences in school-aged children. Developmental Psychology, 14, 614-623.

Ehrlich, H. J. (1973). The social psychology of prejudice. New York: Wiley.

Flerx, V. C., Fidler, D. C., & Rogers, R. W. (1976). Sex role stereotypes: Developmental aspects and early intervention. Child Development, 47, 998-1007.

Frost, F., & Diamond, E. E. (1979). Ethnic and sex differences in occupational stereotyping by elementary school children. Journal of Vocational Behavior, 15, 43-54.

Garrett, C. S., Ein, P. L., & Tremaine, L. (1977). The development of gender stereotyping of adult occupations in elementary school. Child Development, 48, 507-512.

Gettys, L. D., & Cann, A. (1981). Children's perceptions of occupational sex stereotypes. Sex Roles, 7, 301-308.

Guttentag, M., & Bray, S. (1976). Undoing sex stereotypes. New York: Harper & Row.

Helmreich, R. L., Spence, J. T., & Holahan, C. K. (1979). Psychological androgyny and sex-role flexibility: A test of two hypotheses. Journal of Personality and Social Psychology, 37, 1631-1644.

Huston, A. C. (1983). Handbook of Child Psychology: Vol. 4. Socialization, personality and social behavior (4th ed., pp. 387-467). New York: Wiley.

Huston-Stein, A., & Higgins-Trenk, A. (1978). The development of females: Career and feminine role aspirations. In P. B. Baltes (Ed.), Life-span development and behavior (Vol. 1, pp. 258-297). New York: Academic Press.

Inhelder, B., &.Piaget, J. (1969). The early growth of logic in the child. New York: Norton.

Jones, L. M., & McBride, J. L. (1980). Sex-role stereotyping in children as a function of maternal employment. Journal of Social Psychology, 111, 219-223.

Kail, R. V., & Levine, L. E. (1976). Encoding processes and sex-role preferences. Journal of Experimental Child Psychology, 21, 256-263.

Kierscht, M. S., & Rice, P. L. (1981). Androgyny and the young child. Journal of Genetic Psychology, 138, 303-304.

Kohlberg, L. (1966). A cognitive-developmental analysis of children's sex-role concepts and attitudes. In E. E. Maccoby (Ed.), The development of sex differences (pp. 80-173). Stanford, CA: Stanford University Press.

Kohlberg, L., & Ullian, D. Z. (1978). Stages in the development of psychosexual concepts and attitudes. In R. C. Friedman, R. M. Richert, & R. L. Vande Wiele (Eds.), Sex differences in behavior (pp. 209-222). Huntingdon, NY: Robert E. Krieger.

Kuhn, D., Nash, S. C., & Brucken, L. (1978). Sex role concepts of two- and three-year-olds. Child Development, 49, 445-451.

Lerner, R. M., Benson, P., & Vincent, S. (1976). Development of societal and personal vocational role perception in males and females. Journal of Genetic Psychology, 129, 167-168.

Lerner, R. M., Vincent, S., & Benson, P. (1976). One-year stability of societal and personal vocational role perceptions of females. Journal of Genetic Psychology, 129, 173-174.

Liben, L. S., & Signorella, M. L. (1980). Gender-related schemata and constructive memory in children. Child Development, 51, 11-18.

Marantz, S. A., & Mansfield, A. F. (1977). Maternal employment and the development of sex-role stereotyping in five- to eleven-year old girls. Child Development, 48, 668-673.

Martin, C. L., & Halverson, C. F. (1981). A schematic processing model of sex typing and stereotyping in children. Child Development, 52, 1119-1134.

Martin, C. L., & Halverson, C. F. (1983). The effects of sex-typing schemas on young children's memory. Child Development, 54, 563-574.

Masters, J. C., & Wilkinson, A. C. (1976). Consensual and discriminative stereotypy of sex-type judgments by parents and children. Child Development, 47, 208-217.

Meyer, B. (1980). The development of girls' sex-role attitudes. Child Development, 51, 508-514.

Nadelman, L. (1970). Sex identity in London children: Memory, knowledge, and preference tests. Human Development, 13, 28-42.

Nadelman, L. (1974). Sex identity in American children: Memory, knowledge, and preference tests. Developmental Psychology, 10, 413-417.

Payne, B. D. (1981). Sex and age differences in the sex-role stereotyping of third- and fifth-grade children. Sex Roles, 7, 135-143.

Perry, W. D. .(1970). Forms of ethical and intellectual development in the college years: A scheme. New York: Holt, Rinehart, & Winston.

Rebecca, M., Hefner, R., & Oleshansky, B. (1976). A model of sex-role transcendence. Journal of Social Issues, 32(3), 197-206.

Reis, H. T., & Wright, S. (1982). Knowledge of sex-role stereotypes in children aged 3 to 5. Sex Roles, 8, 1049-1056.

Robinson B. E., & Green, M. G. (1981). Beyond androgyny: The emergence of sex-role transcendence as a theoretical construct. Developmental Review, 1, 247-265.

Ruble, T. L. (1983). Sex stereotypes: Issues of change in the 1970s. Sex Roles, 9, 397-402.

Schau, C. G., Kahn, J. H., Diepold, J. H., & Cherry, F. (1980). The relationships of parental expectations and preschool children's verbal sex typing to their sex-typed toy play behavior. Child Development, 51, 266-270.

Shepard, W. 0., & Hess, D. T. (1975). Attitudes in four age groups toward sex role division in adult occupations and activities. Journal of Vocational Behavior, 6, 27-39.

Sherif, C. W. (1976). Orientation in social psychology. New York: Harper Row.

Signorella, M. L., & Liben, L. S. (1984). Recall and reconstruction of gender-related pictures: Effects of attitude, task difficulty, and age. Child Development, 55, 393-405.

Steiger, J. H. (1980). tests for comparing elements of a correlation matrix. Psychological Bulletin, 87, 245-251.

Thompson, S. K. (1975). Gender labels and early sex role development. Child Development, 46, 339-347.

Tremaine, L., Schau, C. G., & Busch, J. (1982). Children's occupational sex-typing. Sex Roles, 8, 691-710.

Umstot, M. E. (1980). Occupational sex-role liberality of third-, fifth-, and seventh-grade females. Sex Roles, 6, 611-617.

Urberg, K. A. (1982). The development of the concepts of masculinity and femininity in young children. Sex Roles, 8, 659-668.

Williams, J. E., Bennett, S. M., & Best, D. L. (1975). Awareness and expression of sex stereotypes in young children. Developmental Psychology, 11, 635-642.

Zuckerman, D. M., & Sayre, D. H. (1982). Cultural sex-role expectations and children's sex-role concepts. Sex Roles, 8, 853-862.

FOOTNOTES

1)A summary of the items and the procedure for administering the measure is available.
(2)The actual median for Sample K-2-4 is 7. However, because of the large number of ties at the median, 6 was used to provide more equal numbers of high and low stereotype children.
(3)Four other memory studies included measures of children's own preferences for stereotyped activities (Coker, 1984; Kail & Levine, 1976; Nadelman, 1970, 1974). The memory measures in these studies did not, however, include nontraditional material along with the traditional (stereotyped) items. Nadelman was not able to relate individual preferences to memory since virtually all children expressed same-sex preferences, although children did show better memory for same-sex items. Using the Nadelman materials, Coker found a significant correlation between preference and memory for boys but not for girls. Kail and Levine found that children with same-sex preferences encoded words along a masculine-feminine dimension, while children with cross-sex preferences did not. Because these studies did not use nontraditional memory items, they do not provide a clear test of the flexibility hypothesis. They do, however, show the expected relationship between preferences and memory.