|
|
Measuring beliefs about adolescent personality and behavior.
Christy M. Buchanan; Grayson N. Holmbeck.
Journal of Youth and Adolescence, Oct 1998 v27 i5 p607(2)
Author's Abstract: COPYRIGHT 1998 Plenum Publishing Corporation A measure that assesses individuals' expectations and stereotypes concerning adolescents was developed. Initially, 123 adult respondents answered open-ended questions about the personality and behavior of typical and observed adolescents. based on their responses, a close-ended scale was developed where respondents rate the degree to which 65 traits and behaviors are characteristic of typical adolescents. Refinement and testing of this scale was done with a sample of 361 college students and 112 parents of young adolescents; the final number of items after such refinement was 44. The subscales demonstrated adequate internal consistency, test-retest reliability, and construct validity. The multifaceted nature of the measure and absence of stereotypical language, along with its flexibility for obtaining comparable information on other age groups provide advantages over the small number of previously used measures.
Full Text: COPYRIGHT 1998 Plenum Publishing Corporation
INTRODUCTION
Parents' beliefs about their children and about parenting, and parents' expectations concerning the content, timing, and causes of children's behavior, are important factors to consider in understanding parenting (Goodnow and Collins, 1990; Smetana, 1994). Although the links between generalized parenting attitudes and parenting behavior are indirect and complex (Holden and Edwards, 1989; Miller, 1988; Sigel, 1986), as are links between attitude and action more generally (Fazio, 1986), data suggest that parents' ideas about their children influence parenting and parent-child relationships. Despite the general interest in this area, only a handful of studies have focused on beliefs about adolescence (Boxer et al., 1984; Buchanan et al., 1990; Freedman-Doan et al., 1993; Holmbeck and Hill, 1988; Offer et al., 1981; Stoller et al., 1996).
These studies have attempted to describe the beliefs parents and others hold, and determinants of these beliefs. based on research with parents of younger children, investigators have argued that parents' expectations about adolescence have implications for how parents interpret their children's behavior as those children become adolescents, and therefore for the evolving relationship between parents and adolescents. Past research also suggests that expectations about adolescence differ by both parent and child gender (Boxer et al., 1984; Buchanan et al., 1990; Holmbeck and Hill, 1988), and depending on the child's pubertal status and the current parent-child relationship (Freedman-Doan et al., 1993).
Unfortunately, further investigation and integration of ideas and findings in this field is currently hindered by the lack of a standard scale to measure expectations. The current study was undertaken as a first step in providing such a measure. A standard scale with adequate psychometric properties would allow investigators to tackle a wide range of questions about the predictors and consequences of beliefs about adolescence held by parents and other adults who work with adolescents.
There are several ways to conceptualize beliefs about a developmental period such as adolescence (Buchanan et al., 1990; Collins, 1990). Of interest in this study were category-based and target-based expectations concerning personality and behavior.(3) Individuals have category-based expectations for adolescents' personality and behavioral attributes, or expectations about attributes that are thought to pertain to adolescents as a category of people. Category-based beliefs are presumably influenced by societal stereotypes that exist about the category, and ought to be discernible among parents and nonparents alike, given their shared nature. Individuals also hold target-based expectations, or expectations for any one adolescent's personality and behavior, regardless of what other adolescents are like.
Parents' expectations for the behavior of their own children may be influenced by their category-based expectations, but should also be influenced by personal experience with that child (Buchanan et al., 1990). American society has a strong stereotype, often conveyed through the media, of adolescence as a difficult time, involving such attributes as rebellion (e.g., Gelman, 1990), parent-child conflict (e.g, Meltz, 1995), risk-taking behavior (e.g., Kantrowitz, 1990), irresponsibility (e.g., Isbister, 1996), and peer pressure.
Each of the studies that have addressed people's endorsement of such stereotypes has used a different measure, developed for that particular study, with little attention to testing the reliability and validity of the measure itself. And although previous measures ask about many of the most common stereotypes about adolescents, no one has assessed the range of attributes assigned to adolescents as a category. Given the potential importance of parents' and others' beliefs about adolescence, our aim was to develop a sound measure that could be used across studies.
The advantages of the measure described in this paper over the measures reported in previous studies include the following:
(1) It was developed based on individual's responses to an open-ended questionnaire asking them to describe, in their view, both the "stereotypical" and the "average" adolescent, so that the descriptors were not limited by the expectations and biases of the investigators.
(2) It is more multidimensional than previous measures, assessing expectations about a wide range of traits and behaviors, including positive as well as negative ones.
(3) It is not obvious that what is being assessed is endorsement of stereotypes, hopefully reducing experimenter demands; and
(4) It is possible for investigators to adapt the scale to assess beliefs about other developmental periods, or about specific individuals, in a way that expectations can be easily compared across target groups or individuals.
In relation to the last point, McCauley et al. (1980) argue that stereotypes are probabilistic - that an individual holds a stereotype of a particular group if they think that group has characteristics at a different rate than does a different group of people. For example, a person might say that adolescents fight with their parents. But unless that person claims that adolescents are more likely to fight with their parents than are younger children, this claim does not reflect a stereotype of adolescents, per se. Thus, it maybe important to assess expectations about individuals in other developmental periods, as well as expectations about adolescents, in order to fully understand a person's behavior toward adolescents.
In the first phase of the study, we generated a pool of descriptors that are applied to adolescents, based on the responses of college students and parents. Frequently mentioned items from this pool were used in the second phase, in which new groups of college students and parents rated the extent to which they believed the descriptors apply to adolescents. Data from the second phase were used to assess internal reliability, test-retest reliability, and construct validity of the measure. Demographic predictors of parents' and young adults' expectations were also examined. As noted, college students were included as participants in both phases of measure development. Even though our ultimate interest is in how the beliefs of parents and other adults influence their interactions with adolescents, we believed it would be acceptable to include young adult college students in measure development.
As young adult members of society, we expected their responses to mirror, at least in part, the ideas of the broader society in which they live, as well as their own very recent experiences and impressions. In sum, the aim of this study was to generate valid and reliable scales of descriptors that could be used as indicators of individuals' beliefs about adolescent personality and behavior. We expected that if, indeed, individuals' beliefs were influenced by societal stereotypes, that adolescents as a group would be rated more negatively in many respects (e.g., more rebellious, more emotional, less friendly) than would younger children (e.g., elementary school aged children) rated using the same scales. In addition, we wanted to see how the endorsement of stereotypes about adolescents differed by demographic factors, including gender, socioeconomic status, and family structure.
Previous research, for example, suggests that mothers are more likely than fathers to have negative expectations about adolescents; whether this represents more stereotyping of adolescents or whether mothers' more negative expectations occur across different age groups is unclear. Little attention has been paid to the influence of other demographic variables on such beliefs, so we did not have a priori predictions for these indicators.
METHOD
Phase One
The aim of the first phase of the study was to generate a wide range of descriptors that people use to describe adolescents. Two samples were used:
(1) 62 college students recruited through a course at a Midwestern university, and
(2) 61 parents recruited through a parenting class at a community center in California.
Students and parents who agreed to participate answered two open-ended questions about adolescent personality: 1. Please describe the personality of the stereotypical American adolescent. (Include both positive and negative descriptions.) 2. based on your own observations of adolescents, please describe what you believe to be the personality of the average American adolescent. (Include both positive and negative descriptions.) In two additional questions, adolescent "behavior" was substituted for "personality"; thus, participants were asked to generate descriptions of both the personality and the behavior of the stereotypical adolescent and of the typical adolescent in their own experience. Lists of descriptors were generated based on participants' responses.
Descriptors retained for further study were selected using two criteria: 1. The descriptor was mentioned by four or more participants in response to any one question about stereotypical or observed personality and behavior in either sample. Thirty-six descriptors fitting this criterion were chosen.
Six additional eligible items were not retained; descriptors that mentioned specific activities such as "playing sports" and being "into cars" were excluded, because we judged them as more likely to be influenced by individual taste than developmental status. Several participants described adolescents as "going to school," but because "going to school" is a fact of life (for most) more than a judgment that might be disputed by some or endorsed by others, it also did not seem to fit with other descriptors. The three other excluded items ("wants things right away," "quiet," "shy"), each mentioned by four participants in one pilot sample, were inadvertently left off the list.
2. The descriptor was mentioned three times in either sample in response to any of the four questions, and was judged by the investigators to be either a "stereotypical" trait or behavior from the standpoint of the literature on adolescence, or a positive descriptor not yet represented by the more commonly mentioned items. Twelve descriptors were added based on this criterion. In addition to the 48 items chosen based on participants' responses, 17 items were added by the investigators to increase our confidence that the scale would not leave out any traits or behaviors that, based on both academic and popular literature (see Buchanan et al., 1992), seemed to be an important part of our characterizations of adolescence (e.g., emotional, irritable, impulsive, withdraws from parents).
A few positive descriptors were also among this group of 17 items (e.g., considerate, generous) with the continued aim of balancing negative descriptors with positive ones. The final set of items numbered 65. Phase Two A questionnaire was generated using the 65 descriptors derived from Phase One. Because we were interested in stereotypes (i.e., views of adolescents versus children from other developmental periods; McCauley et al., 1980) as well as absolute beliefs about adolescents, we asked participants about the extent to which each descriptor was an accurate descriptor separately for adolescents and for elementary school children, using the following format: "For each characteristic or behavior, please mark the probability that a typical adolescent [elementary school child] possesses that characteristic or displays that behavior. (Another way to think about this is to mark the proportion of adolescents [elementary school children] that you believe possess the characteristic or display the behavior.) Each descriptor was then presented next to a scale ranging from "0% (or definitely NO)" to "100% (or definitely YES)." Half of the subjects answered the questions first about adolescents; the other half of the subjects answered first about elementary school children. Also included in the questionnaire were scales from previous studies measuring beliefs about adolescence (Buchanan et al., 1990; Holmbeck and Hill, 1988).
Combined, these scales consisted of 18 statements about adolescents such as "Adolescence is a stormy and stressful time" and "Adolescents prefer to talk to peers over parents." Respondents were asked to indicate their agreement with each statement on a 7-point Likert scale ranging from Strongly Agree to Strongly Disagree. The questionnaires also included measures of the respondent's own adolescent experience. For each trait and behavior on which respondents rated the typical adolescent and the typical elementary school child, college students also rated the extent to which that trait or behavior was characteristic of themselves during their adolescence.
Using a scale where 1 = not at all well and 10 = very well, they answered the question "How well would each of the following characteristics describe your own personality and behavior as an adolescent? This scale was not used with parents because of concerns about the length of the questionnaire for this group. In contrast, both college students and parents rated eight adjectives (e.g., "challenging," "easy") summarizing their adolescent experience on a scale ranging from 1 = strongly agree to 4 = strongly disagree. Questionnaires were administered to 361 college students and 112 parents (77 mothers and 35 fathers, from 82 families). Using college students allowed us to assess "societal" ideas about adolescents in a sample large enough to use principal components analysis for scale development (Floyd and Widaman, 1995; Tabachnick and Fidell, 1996). College students were recruited through introductory psychology classes at an undergraduate institution in the Southeast. In exchange for participation, they received course credit. For test-retest reliability calculations, a subset of the students (n = 151) filled out the questionnaire twice, with Time 1 occurring approximately 4 weeks prior to Time
2. Parents were recruited from a middle school in the same Southeastern city. Letters describing the study were sent to parents of young adolescents (6th and 7th graders), so as to assess expectations in a group of parents who would be anticipating the adolescent period for a particular child. Parents returned a form indicating whether they were willing to participate. Parents who were willing were mailed consent forms, questionnaires, and a stamped return envelope. The 82 families who participated represent approximately 14% of the families with 6th or 7th graders attending that middle school. College student participants were primarily white (89%), and from never divorced families (84%). This sample was 53% female. Parent participants were also largely from intact marriages (71% of mothers, 91% of fathers), and white (89% of mothers, 94% of fathers). Among mothers, 52% reported family income of over 50,000; 15% reported family income of less than 30,000 (1% did not report income). The distribution of family income among fathers was almost identical to that of mothers. The most common educational level among mothers was a college degree (47%), with 21% of mothers reporting more than a college degree and 32% less than a college degree. Thirty-seven percent of the fathers had a college degree, with 31% reporting more than that and 31% less. All mothers and fathers had graduated from high school. RESULTS Defining Categories of Beliefs About Adolescents A principal components analysis was conducted using responses to the adolescent descriptors from the 361 college students and the 77 mothers (total n = 438). Fathers were excluded from this analysis because they represented largely nonindependent cases from the mothers in the sample. Oblimin rotation of factors was used because the factors were expected to be intercorrelated. An analysis using all 65 descriptors resulted in 13 factors that accounted for 64.5% of the variance, using the criterion of eigenvalues greater than one.
The first 9 of the 13 factors were easily interpretable; however, the remaining four factors consisted of small numbers (1-3) of moderate to high loading items that were not readily interpretable. Therefore, our next step was to drop all items that did not load above .40 on one of the first 9 factors. Twenty items were eliminated by this criterion. Table I. Item Loadings for 9 Factors Representing Sets of Beliefs About Adolescent Personality and Behavior(a) Factor 1: Risk-Taking/Rebellious Takes risks .79 Tests limits .73 Rebellious .71 Reckless .71 Stubborn .68 Rude .61 Impulsive .53 Restless .50 Selfish .45 Factor 2: Friendly Friendly .79 Generous .72 Fun-loving .71 Considerate .67 Caring .53 Factor 3: Problem Behaviors Uses alcohol .92 Sexually active .91 Smokes .84 Uses drugs .82 Dates .61 Factor 4: Classic Adolescent Behaviors Listens to music .79 Into clothes .69 Materialistic .64 Faddish(b) .43 Factor 5: (Not) Social Spends time with friends -.70 Social -.56 Sad(b) .46 Factor 6: Internalizing Awkward .66 Anxious .65 Insecure .64 Confused .61 Emotional .52 Depressed .41 Factor 7: (Not) Active Active -.83 Adventurous -.79 Ambitious -.65 Energetic -.62 Factor 8: Conforming Conforms to peers .83 Easily influenced by friends .71 Faddish .49 Distractible .42 Factor 9: Upstanding/Prosocial Interested in school .84 Inquisitive .69 Intelligent .61 Helpful .60 Hardworking .59 Honest .58 a based on college student and mother report (n = 438). b Item ultimately dropped from scale.
A second principal components analysis using oblimin rotation was conducted using the remaining 45 items. Ten factors had eigenvalues greater than one, and they accounted for 65.4% of the variance in these items. All 45 items loaded above .40 on one of the first 9 factors; no item loaded above .40 on the 10th. The items and their loadings for the 9 interpretable [TABULAR DATA FOR TABLE II OMITTED] factors are presented in Table I. Only one item - "faddish" - loaded above .40 on more than one factor. Because it loaded most highly on Factor 8 ("Conforming"), it was included with the other items on Factor 8 and dropped from Factor 4 ("Classic Adolescent Behaviors"). With two exceptions, the subscales were highly reliable (above .70) in both the derivation sample (i.e., students and mothers combined) and in each subgroup. The first exception was Classic Adolescent Behaviors, where the alpha for mothers alone was .47. We thus examined the values on the 3 items for this subscale individually for each mother. This examination revealed that 2 of the mothers clearly did not answer these items similarly - for these 2 mothers, at least 2 items in the scale diverged by over 80% in the extent to which they were attributed to adolescents (e.g., 10% of adolescents were said to be "into clothes" while 100% were said to "listen to music"). These 2 "outliers" were therefore dropped from the dataset, reducing the total n for mothers to 75, and the total number of families represented to 81.(4) The second exception to good internal reliability occurred on the "Social" subscale. Dropping the item "sad" increased reliability from an abysmal .29 to an acceptable .70. Consequently, the item of sad was excluded from this subscale. Table II displays Cronbach's alpha coefficients for all 9 subscales after the two outlier cases were dropped. All subscales have acceptable to high reliability.(5) [TABULAR DATA FOR TABLE III OMITTED] [TABULAR DATA FOR TABLE IV OMITTED]
The final item pool for all subscales consisted of 44 items. For further analyses, the subscales for "(Not) Social" and "(Not) Active" were created so that the high end of the scale meant "Social" and "Active" respectively. Correlations among the nine subscales can be found in Table III. The correlations range from -.02 to .65, although 79% of the correlations are less than or equal to .40. [TABULAR DATA FOR TABLE V OMITTED] Test-Retest Reliability of Subscales Table IV shows correlations between the scores on the 9 Time 1 subscales and their respective subscales from Time 2 for the 151 college students who participated in the study twice. All correlations are equal to or above .55, except Social, at .36. Applying the Subscales to Elementary School Children To determine if the subscales could be applied to preadolescents, internal reliabilities were computed based on descriptors as students and parents rated them for elementary school aged children (see Table V). The majority of the alphas were acceptable. The lowest occurred for Conforming, because "distractible" did not correlate as well with the other items on this scale as it had when adolescents were rated. For all subgroups (i.e., students, mothers, fathers), reliability of Conforming would have improved if "distractible" were eliminated from the scale. Two other items did not fit as well onto their respective subscales when describing elementary school students as they had for descriptions of adolescents, although the internal reliabilities of the subscales remained high. First, the item "dates" did not correlate as well with the other items on the "Problem Behaviors" subscale for elementary school children as it had for adolescents. Apparently, "dating" and other "adult-like behaviors" such as drinking, using drugs, smoking, and having sex mean somewhat different things during the elementary years but similar things during adolescence. Secondly, the item "ambitious" did not correlate as well with the other items on the Active subscale for elementary school children as it did for adolescents. Perhaps "ambitious" is a term that individuals tend to apply to older individuals more so than they do to the young. Table VI. Correlations Between the 9 Subscales Describing Adolescents and Holmbeck and Hill's (1988) "Storm and Stress" Scale Students Mothers Fathers Subscale (n = 361) (n = 75) (n = 35) Active .03 -.08 .01 Classic Adolescent Behaviors .25(b) .38(b) .23 Conforming .23(b) .48(b) .51(b) Friendly .02 -.21 -.16 Internalizing .27(b) .44(b) .41(a) Problem Behaviors .17(b) .35(b) .21 Risk-Taking/Rebellious .35(b) .55(b) .34(a) Social .20(b) .15 .13 Upstanding/Prosocial -.09 -.33(b) -.31 a p [less than] .05. b p [less than] .01. Test-retest reliability for the subscales used to describe elementary school students ranged from .42 to .65; only two correlations were less than .50.
Although slightly lower overall than the Time 1-Time 2 correlations for the adolescent subscales, these correlations were still substantial, indicating good stability in answers across a four-week period. Construct Validity To assess convergent validity, we examined associations between the beliefs measure developed for the current study and items that previous investigators had used to assess people's beliefs (Buchanan et al., 1990; Holmbeck and Hill, 1988). The Buchanan et al. (1990) scales, as created in that study, did not have adequate internal reliability in the current samples. Thus, we created one scale from the 9 items used by Holmbeck and Hill (1988; Cronbach's alpha = .82 in our sample) for validation purposes. This scale indicates a person's endorsement of adolescence as a time of "storm and stress." Correlations among our nine subscales and the Holmbeck and Hill (1988) "storm and stress" scale are shown in Table VI. The correlations indicate scale validity in that the "storm and stress" scale correlates as one would predict with our current scales describing adolescents as risk taking and rebellious, internalizing, and influenced by friends and peer culture. A belief in the existence of storm and stress among parents was negatively correlated with a description of adolescents as "upstanding and prosocial" (hardworking, interested in school, etc.). Storm and stress did not correlate either positively or negatively with descriptions of adolescents [TABULAR DATA FOR TABLE VII OMITTED] as "friendly" or as "active," indicating that these are independent beliefs. We also investigated the construct validity of our subscales by comparing respondents' descriptions of adolescents with their descriptions of elementary school children. We expected that if the subscales tapped into dimensions of adolescent character that reflected society's stereotypes, that college students and parents would report higher levels of rebellion, problem behaviors, classic adolescent behaviors, sociality, internalizing, and conformity, and lower levels of friendliness and upstanding/prosocial qualities when rating adolescents than when rating elementary school children. We did not have a prediction about ratings of being "active." Paired t-tests were conducted comparing ratings of adolescents vs. elementary school children among the three groups of respondents (college students, mothers, and fathers) separately. In line with our predictions, there were significant differences in the ratings of adolescents and elementary school children on all subscales by students, mothers, and fathers alike (see Table VII). All differences were in line with our predictions; the difference with regard to activity indicated a belief in all subgroups of participants that elementary school children are more active (active, ambitious, adventurous, energetic) than are adolescents. Clearly, our participants' beliefs fit with society's stereotypes of adolescents.
In order to examine the extent of stereotyping, we used the method suggested by McCauley et al. (1980). We divided the probability of an adolescent having a trait by the probability that an elementary school child has the same trait (e.g., the probability that an adolescent is active divided by the probability that an elementary school child is active). Scores above or below 1.00 indicate that the individual holds different beliefs about the two age groups, and the size of the deviation from 1.00 is a measure of how strong a particular stereotype is. Reflecting the mean differences shown in Table VII, the strongest stereotypes concerned what we have labeled "classic adolescent behaviors," "problem behaviors," and "internalizing." The weakest stereotypes occurred for the positive traits (with adolescents seen as less positive than elementary school children) - friendliness, activity, and upstanding/prosocial nature. Beliefs About Adolescents as Related to Demographic Factors We compared demographic groups on mean endorsement of traits and behaviors for adolescents and elementary school children separately, and on their level of "stereotyping" of the adolescent age group.
Although there were several differences between demographic groups in their views of both adolescents and elementary school aged children, there were fewer group differences in the extent to which they stereotyped of adolescents (i.e., saw adolescents more negatively than they did elementary school children, as indicated by the ratio scores described above). Because of the focus in this study on views of adolescents as a unique group, we concentrate on demographic differences in stereotypes below. Gender Gender differences among college students were examined using multivariate analysis of variance (MANOVA), with the nine subscales measuring beliefs about adolescents as the dependent variables and gender of student as the independent variable. The multivariate effect was significant (F[9, 339] = 3.56, p = .000). Univariate F tests indicated that female students were more likely than males to have stereotypes of adolescents as less active ([M.sub.females] = .89, [M.sub.males] = .92, F = 4.33, p [less than] .05) and less upstanding or prosocial ([M.sub.females] = .91, [M.sub.males] = .96, F = 5.39, p [less than] .05), and more internalizing ([M.sub.females] = 1.61, [M.sub.males] = 1.32, F = 16.87, p [less than] .001) and conforming ([M.sub.females] = 1.23, [M.sub.males] = 1.12, F = 7.56, p [less than] .01) than school-aged children. Gender comparisons among parents were restricted to the 29 mother-father pairs where both parents filled out a questionnaire. A repeated measures MAN OVA with 9 subscales as dependent variables, and parent gender as the within-subject independent variable was conducted. A significant interaction between subscale and parent gender emerged (F [8, 21] = 2.88, p [less than] .05). Univariate paired t-tests indicated that the only significant gender differences occurred for the traits of "internalizing" and "upstanding/prosocial." Mothers were more likely than fathers to stereotype adolescents as internalizing ([M.sub.mothers] = 1.54, [M.sub.fathers] = 1.29, t = -2.56, p [less than] .05) and not upstanding or prosocial ([M.sub.mothers] = .85, [M.sub.fathers] = .97, t = 3.49, p [less than] .01). Ethnicity and Family Structure Ethnic comparisons could be conducted only among college students, because there were too few minority parents who participated. Similarly, the effects of marital status could not be examined among fathers because most participating fathers were in first marriages. Thus, MANOVAs were used to predict the nine beliefs subscales with student ethnicity (minority vs. white), and, separately, with family structure among both students and mothers. No effects of either ethnicity or family structure emerged. Parental Education and Income College students did not report on their parents' education or income; thus, analyses in this section are restricted to the parent sample. Correlations were conducted between both parental education and family income, and each of the stereotypes subscales. Where they were related, higher levels of education and income were associated with more negative beliefs and more stereotyped notions about adolescence. More highly educated mothers held a stronger stereotype of adolescents as internalizing than did less educated mothers (r = .44, p [less than] .01). More highly educated fathers were more likely to rate adolescents as risk taking and rebellious (r = 51, p [less than] .01) than less educated fathers; fathers with higher income were more likely to rate adolescents as risk taking and rebellious (r = .45, p [less than] .01) and less likely to rate them as social (r = -.40, p [less than] .05) than fathers with less income. Age/Parental Status We compared the stereotypes of college students to those of parents, to get some insight into possible effects of age or parental status on one's beliefs about adolescence as an age group. Because of the gender differences reported above, we conducted the comparisons within each gender group. Among males, there was a significant multivariate effect of age/parental status on stereotypes about adolescence (F[9,183] = 2.36, p [less than] .05). Univariate tests revealed that college males were more likely than fathers to say that adolescents were especially involved in problem behaviors ([M.sub.fathers] = 9.87, [M.sub.male] student = 17.85, F = 4.81, p [less than] .05). College males were also less likely to stereotype adolescents as unfriendly ([M.sub.fathers] = .90, [M.sub.male] student = .99, F = 6.29, p [less than] .05). Multivariate tests also indicated a significant effect of age/parental status on beliefs among females (F[9, 256] = 4.61, p [less than] .001). Like their male counterparts, college females were more likely than mothers to say that adolescents were especially involved in problem behaviors ([M.sub.mothers] = 11.40, [M.sub.female] student = 22.17, F = 12.47, p [less than] .001). College females were less likely, however, to stereotype adolescents as risk-taking and rebellious ([M.sub.mothers] = 1.37, [M.sub.female] student = 1.14, F = 16.21, p [less than] .01).
Beliefs About Adolescence and One's Own Adolescent Experience Using college students' ratings of their own adolescence, we created the same nine subscales that emerged from their descriptions of the typical adolescent. In addition, we created two subscales that emerged from a factor analysis of the 8 items used to summarize their experience during adolescence (e.g., my adolescence was challenging, my adolescence was fun). The first subscale indicated that adolescence had been "difficult" (it was challenging, not easy, difficult, not smooth, stressful, and turbulent); the second indicated that adolescence had been "fun" (it was exciting and fun). The internal reliabilities were .84 and .81 respectively among college students. The correlation between the two factors was -.21. Nine correlations were computed, comparing the college students' scores for each "typical adolescent" subscale and the respective subscale rating their own adolescence. The magnitude of the correlations ranged from .25 (for social) to .52 (for internalizing) and all were significant at p [less than] .01, using a one-tailed test. In addition, students' ratings of their own adolescence as "difficult" were correlated with expectations that adolescents were internalizing (r = .31, p [less than] .01), conforming (r = .12, p [less than] .05), risk taking and rebellious (r = .24, p [less than] .01), and involved in classic adolescent behaviors (r = -.15, p [less than] .01). Students' ratings of their own adolescence as "fun" were correlated with the belief that adolescents were active (r = .11, p [less than] .05), not internalizing (r = -.23, p [less than] .05), friendly (r = .14, p [less than] .14), social (r = .14, p [less than] .05), and upstanding and prosocial (r = .16, p [less than] .05). Internal reliabilities for mothers' ratings of the scale items about one's own adolescence as difficult or fun were high (.87 and .72 respectively); fathers' were slightly lower, but still acceptable (.67 and .70 respectively). As with college students, there were several significant correlations between mothers' ratings of their own adolescence and their beliefs about the typical adolescent. For mothers, recalling their own adolescence as difficult was related to beliefs that the typical adolescent would be internalizing (r = .34, p [less than] .01), conforming (r = .39, p [less than] .01), and risk-taking and rebellious (r = .30, p [less than] .05). Recollections of adolescence as "fun" were related to mothers' beliefs that adolescents in general were not internalizing (r = -.24, p [less than] .05), not conforming (r = -.26, p [less than] .05), and not risk-taking and rebellious (r = -.31, p [less than] .05). There was only one significant correlation between fathers' recollections of their own adolescence and their beliefs about the typical adolescent. Fathers who recalled their adolescence as fun were more likely to rate adolescents as active (r = .45, p [less than] .01).
DISCUSSION
Our aim in this study was to develop a measure of beliefs about adolescence that was reliable and valid, and that could be used to measure category-based beliefs about adolescence in an absolute sense (i.e., without regard to one's beliefs about other age groups), and in a relative or stereotypical sense (i.e., in comparison to beliefs about other age groups). The measure we developed allows parents and other individuals to report the extent to which 44 traits and behaviors apply to adolescents, and from these 44 items, nine characteristics of adolescents can be measured (refer back to Table I for a list of specific items on each scale). All subscales except Classic Adolescent Behaviors and Social had high internal reliability and exhibited good stability across time among a variety of samples: college students at each of two time periods, mothers, and fathers. The internal reliability of the Classic Adolescent Behaviors (.56 among mothers) and Social (range from .68 to .70 for mothers and students) subscales was not as high as it was for the other scales, although it was probably adequate. When subscales were applied to descriptions of elementary school students, the internal reliabilities fell somewhat, but most were still adequate and many were quite high. The small drop in reliabilities is not surprising given that the subscales were created based on individuals' descriptions of adolescents. With respect to construct validity, subscales that tapped into some of the less desirable facets of our society's stereotype about adolescence (e.g., internalizing, risk taking and rebelliousness, influence of peer culture) correlated positively and significantly with a previous measure of beliefs in adolescence as a time of storm and stress (Holmbeck and Hill, 1988), indicating convergent validity. In addition, endorsement of each subscale for adolescents as compared to elementary school aged children was in line with what one would predict based on society's stereotypes. Among college students, mothers, and fathers alike, adolescents were rated as more internalizing, conforming, risk taking and rebellious, social, involved in problem behaviors, and influenced by peer culture and less active, friendly, and upstanding/prosocial, than elementary school aged children. In general, we found little evidence for more or less extreme stereotyping based on demographic characteristics. The demographic factor that had the most links to stereotyping was gender of the respondent. Among both college students and parents, females held stronger negative stereotypes of adolescence than did males. This was especially true for perceptions of adolescents as internalizing and not upstanding or prosocial. This gender difference is in line with previous findings indicating that females hold more negative views of adolescence than do males (Buchanan et al., 1990; Holmbeck and Hill, 1988), and may reflect more negative personal experiences during adolescence for females than males. Research indicates that puberty is a more negative experience for females than it is for males (Dorn et al., 1988; Duncan et al., 1985), and that a variety of problems (e.g., depression, low self-esteem, eating disorders) become more common for females during this time (Attie and Brooks-Gunn, 1989; Nolen-Hoeksma and Girgus, 1994; Petersen et al., 1991; Simmons and Blyth, 1987). Ethnicity and family structure were not related to the extent to which individuals stereotyped adolescents, but higher education and affluence predicted more negative stereotypes. Higher levels of education may be associated with increased exposure to traditional psychological theories about storm and stress during adolescence (but see Holmbeck and Hill, 1988, for evidence that taking a course in adolescent development lowers beliefs in storm and stress); they may also be related to increased exposure to popular sources of information that convey the negative stereotypes (e.g., newspapers and magazines). The predominance of white, educated, middle- to upper-class respondents in our sample limits our ability to draw strong conclusions about these demographic factors, however.
Further study of ethnic, economic, and educational influences on views of adolescence is much needed. Younger respondents, who themselves are close in time to their own adolescence, view adolescents more positively on some characteristics (rebelliousness, friendliness), but are more likely to say that adolescents exhibit problem behaviors, than are parents. Late adolescents may have a more realistic view of the extent to which adolescents today are engaging in behaviors such as sex, drinking, and drugs, than do parents, who have reason to hope that many adolescents do not participate in such behaviors. Yet late adolescents may be reluctant to rate adolescents too negatively overall, since they have so recently been counted as a member of that category. Other studies have found that parents' and college students' expectations about adolescence are related to their own experiences during adolescence - or at least their recollections of those experiences (Boxer et al., 1984; Holmbeck and Hill, 1988).
We found the same among college students and mothers. Among fathers, there were few links between beliefs about adolescence and personal recollections of what their own adolescence was like. Our sample of fathers is small, and mainly limited to white, affluent, married men, but the lack of correspondence between personal experiences and beliefs about adolescents in general among this group raises the question of what are the important influences on fathers' beliefs. As we have noted, a small number of previous studies examined the beliefs about adolescence held by adults (Boxer et al., 1984; Buchanan et al., 1990; Freedman-Doan et al., 1993; Holmbeck and Hill, 1988). The measure described in this study differs from those used in the earlier studies in several important ways. First, it was developed based on individual's responses to an open-ended questionnaire describing their view of a "stereotypical" and an "average" adolescent so that the descriptors were not limited by the expectations and biases of the investigators. Second, it is more multidimensional than previous measures. A variety of positive as well as negative characteristics are measured. It is possible that parents' and others' behavior toward adolescents is determined by the constellation of different beliefs, both positive and negative, and not simply by the extent to which they endorse certain negative stereotypes. Third, the format of the measure has some advantages. In the current measure, because stereotypical language about adolescents is not used (e.g., "storm and stress," "identity crisis"), it should be less obvious to respondents that what is being assessed is endorsement of stereotypes. In turn, this ought to reduce the possibility of respondents' answering either in accordance with or contrary to what they believe the experimenter expects them to answer. Another advantage of the format is the flexibility it gives investigators to make direct comparisons: for example, to compare beliefs about adolescents with beliefs about other age populations, or beliefs about adolescents in different historical periods, or beliefs about one's own child vs. other adolescents.
Finally, this measure has been studied more extensively than any of the earlier measures; in this paper, we have documented acceptable levels of internal reliability, test-retest reliability, and construct validity. The availability of this measure will allow investigators to pursue questions of how parents' cognitions concerning the typical adolescent relate to their expectations for their own children (target-based expectations), their beliefs about how influential they can be as a parent during adolescence, and their parenting practices and the developing parent-child relationship during adolescence.
Financial support of this research was provided to the first author by the Research and Creative Activities Fund (supported by the Z. Smith Reynolds Foundation), the Summer Stipends Program, and the Department of Psychology at Wake Forest University. 3 We are not including other potentially important beliefs, such as beliefs about appropriate timetables for specific adolescent behavior (e.g., Dekovic et al., 1997; Feldman and Quatman, 1988), beliefs about influences on personality and behavior, or beliefs about the ability of parents to affect personality and behavior (e.g., Bugentai et al., 1989; Freedman-Doan et al., 1993). 4 One of the mothers dropped was the sole parent participant from the family; the other mother's spouse had participated and was retained in the dataset for later analyses involving fathers. Thus, the number of families decreases by only one. 5 Because of the relatively low reliability of the Classic Adolescent Behaviors subscale for mothers, and because faddish loaded on both Classic Adolescent Behaviors and Conforming, we also computed the reliability for a subscale we have called "Peer Culture," which combines items from these two scales. The internal reliability for Peer Culture was .68 for mothers and above .75 for all other groups. Results indicate that the Peer Culture scale has psychometric properties equal to or surpassing its component scales. Thus, investigators may want to consider combining these scales.
REFERENCES
Attie, I., and Brooks-Gunn, J. (1989). Development of eating problems in adolescent girls: A longitudinal study. Develop. Psychol. 25: 70-79.
Boxer, A. M., Solomon, B., Offer, D., Petersen, A. C., and Halprin, F. (1984).
Parents' perceptions of young adolescents. In Cohen, R. S., Cohler, B. J., and Weissman, S. H. (eds.), Parenthood: A Psychodynamic Perspective. Guilford Press, New York. Buchanan, C. M., Eccles, J. S., and Becker, J. B. (1992).
Are adolescents the victims of raging hormones?: Evidence for activational effects of hormones on moods and behavior at adolescence. Psychol. Bull. 111: 62-107. Buchanan, C. M., Eccles, J. S., Flanagan, C. F., Midgley, C., Feldlaufer, H., and Harold, R. D. (1990).
Parents' and teachers' beliefs about adolescents: Effects of sex and experience. J. Youth Adolesc. 19: 363-394. Bugental, D. B., Blue, J., and Cruzcosa, M. (1989).
Perceived control over caregiving outcomes: Implications for child abuse. Develop. Psychol. 24: 532-539. Collins, W. A. (1990).
Parent-child relationships in the transition to adolescence: Continuity and change in interaction, affect, and cognition. In Montemayor, R., Adams, G. R., Gullotta, T. R. (eds.), From Childhood to Adolescence: A Transitional Period. Sage, Newbury Park, CA. Dekovic, M., Noom, M. J., and Meeus, W. (1997).
Expectations regarding development during adolescence: Parental and adolescent perceptions. J. Youth Adolesc. 26: 253-272. Dorn, L. D., Crockett, L. J., and Petersen, A. C. (1988).
The relations of pubertal status to intrapersonal changes in young adolescents. J. Early Adolesc. 8: 405-419. Duncan, P. D., Ritter, P. L., Dornbusch, S. M., Gross, R. T., and Carlsmith, J. M. (1985).
The effects of pubertal timing on body image, school behavior, and deviance. J. Youth Adolesc. 14: 227-235. Fazio, R. H. (1986).
How do attitudes guide behavior? In Sorrentino, R. M., and Higgins, E. T. (eds.), Handbook of Motivation and Cognition: Foundations of Social Behavior. The Guilford Press, New York. Feldman, S. S., and Quatman, T. (1988).
Factors influencing age expectations for adolescent autonomy: A study of early adolescents and parents. J. Early Adolesc. 8: 325-343. Floyd, F. J., and Widaman, K. F. (1995).
Factor analysis in the development and refinement of clinical assessment instruments. Psychol. Assess. 7: 286-299. Freedman-Doan, C. R., Arbreton, A. J. A., Harold, R. D., and Eccles, J. S. (1993).
Looking forward to adolescence: Mothers' and fathers' expectations for affective and behavioral change. J. Early Adolesc. 13: 472-502. Gelman, D. (1990, Summer/Fall). A much riskier passage. Newsweek Special Issue, pp. 10-16. Goodnow, J. J., and Collins, W. A. (1990).
Development According to Parents: The Nature, Sources, and Consequences of Parents' Ideas. Erlbaum, Hillsdale, NJ. Holden, G. W., and Edwards, L. A. (1989). Parental attitudes toward child rearing: Instruments, issues, and implications. Psychol. Bull. 106: 29-58. Holmbeck, G. N., and Hill, J. P (1988). Storm and stress beliefs about adolescence: Prevalence, self-reported antecedents, and effects of an undergraduate course. J. Youth Adolesc. 17: 285-306. Isbister, H. (1996, July 18). Ads portray teen-agers as lazy and irresponsible [Letter to the editor]. Winston-Salem Journal, p. A10. Kantrowitz, B. (1990, Summer/Fall). The dangers of doing it. Newsweek Special Issue, pp. 56-57. McCauley, C., Stitt, C. L., and Segal, M. (1980). Stereotyping: From prejudice to prediction. Psychol. Bull. 87: 195-208. Meltz, B. F. (1995, June 12). Growing strains. Winston-Salem Journal, p. D2. Miller, S. A. (1988). Parents' beliefs about their children's cognitive development. Child Develop. 59: 259-285. Nolen-Hoeksma, S., and Girgus, J. S. (1994).
The emergence of gender differences in depression during adolescence. Psychol. Bull. 115: 424-443. Offer, D., Ostrov, E., and Howard, K. I. (1981).
The mental health professional's concept of the normal adolescent. AMA Arch. Gen. Psychiat. 38: 149-153. Petersen, A. C., Sarigiani, P. A., and Kennedy, R. E. (1991).
Adolescent depression: Why more girls? J. Youth Adolesc. 20: 247-271. Sigel, I. E. (1986).
Reflections on the belief-behavior connection: Lessons learned from a research program on parental belief systems and teaching strategies. In Ashmore, R. D., and Brodzinsky, D. M. (eds.), Thinking About the Family: Views of Parents and Children. Lawrence Erlbaum Associates, Hillsdale, NJ. Simmons, R. G., and Blyth, D. A. (1987).
Moving into Adolescence: The Impact of Pubertal Change and School Context. Aldine de Gruyter, Hawthorne, NY. Smetana, J. G. (ed.). (1994).
Beliefs about parenting: Origins and developmental implications. New Directions for Child Development, No. 66, Jossey-Bass, San Francisco. Stoller, C. L., Offer, D., Howard, K. I., and Koenig, L. (1996).
Psychiatrists' concept of adolescent self-image. J. Youth Adolesc. 25: 273-283. Tabachnick, B. G., and Fidell, L. S. (1996).
Using Multivariate Statistics (3rd ed.) Harper and Row, New York. Christy M. Buchanan, Assistant Professor, Department of Psychology, Wake Forest University, Winston-Salem, North Carolina 27106. Received Ph.D. from the University of Michigan. Research interests include family relationships at adolescence, and family and biological influences on adolescent development. To whom correspondence should be addressed. Grayson N. Holmbeck, Associate Professor, Department of Psychology, Loyola University of Chicago, Chicago, Illinois 60626. Received Ph.D. at Virginia Commonwealth University. Research interests include family relationships during adolescence and adolescent development in chronically iii and physically disabled children.