Daily Companionship in Late Childhood and Early Adolescence: Changing Developmental Contexts Reed Larson University of Illinois, Urbana-Champaign
Maryse H. Richards Loyola University of Chicago LARSON, REED, and RICHARDS, MABYSE H . Daily Companionship in Late Childhood and Early
Adolescence: Changing Developmental Contexts. GHILD DEVELOPMENT, 1991, 62, 284-300. The study empltws t^9^s^|mg;|j|i); data to exttmiiw age: d^^^mmm in ^ ^^Ef^^ i ^ j ^ g ^ t ^ of children's a ^ : ^ i i g i i ^ ^ i « B e n t s ' daily experience with »^eir femUies, ftfehSs, ^ri^iaBie. Kirtieipants l a i ^ s ^ i ^ carried electronic pagers for 1 week and reported their companionship, location, and affect at random times when signaled by the pagers. Findings show a dramatic decline in amount of time spent with family, with older students reporting half as much time with their families as younger students. Among boys, this family time was replaced by time spent alone; among girls, by time alone and with friends. Affect reported when with family became less positive between the fifth and seventh grade, but was more positive again in the ninth grade for boys. Affect with friends became more favorable across this age period; affect when alone did not vary. These age differences suggest changes in adolescents' daily opportunities for cognitive growth, emotional development, and social support.
The objective of this article is to chart age differences in the quantity and quality of time children and young adolescents spend with different companions. Theorists contend that individuation from parents and deepening of relationships with peers are central developmental projects of adolescence (BJos, 1967,- Gilligan, 1982,- Havighurst, 1984). Therefore, we expect adolescents to spend less time with tJheir families and more time with their friends than preadolescents. Accompanying age changes in emotional states experienced when with family and friends might also be expected. This article evaluates these age trends in daily interaction with family, with friends, and also alone, employing time-sampling data from Midwestern fifth to ninth graders. A large body of research substantiates theorists' predications that entry into adolesence is accompanied by a shift in the relative importance of family and friends. Age changes include: (a) a shift in "associative orientation" from family to friends (Bow-
erman & Kinch, 1959); (b) increased emotional detachment from parents, with changes occurring somewhat earlier for girls (Steinberg & Silverberg, 1986); (c) increased conflict with and emotional distance from parents, particularly around the occurrence of puberty (Steinberg, 1981, 1987); (d) increased behavioral independence from parents (Feldman & Quatman, 1988; Simmons & BIyth, 1987; Smetana, 1988); (e) increased influence of peers over certain domains of daily life (Berndt, 1979,1989; Hartup, 1983); and (f) increased intimacy with friends, especially among girls (Berndt, 1982;^ Hunter & Youniss, 1^2), and decreased intimate disclosure to parents (Buhrmester & Furman, 1987). Closeness and mutual valuation in relationships with parents do not necessarily decline in adolescence (Blyth, Hill, & Thiel, 1982; Hunter & Youniss, 1982; Smetana, 1988); nonetheless, this research clearly shows that closeness with friends increases and emotional autonomy from parents increases, with the timing of these trends varying somewhat by gender.
This research was supported by NIMH grant 1 ROl MH38324, awarded to the authors. We are grateful to Mary Ghandler for assistance with the data analyses. Gorrespondence should be addressed to Reed Larson, Division of Hnman Development and Family Studies, University of Illinois, 1105 W. Nevada, Urbana, IL 61802. [Child Development, 1991, 62, 284-300. © 1991 by the Society for Research in Child Development, Inc. All rights reserved. 0009-3920/91/6202-.0003$0LOO]
Largon and Richards
285
Little research, however, evaluates how these psychological and emotional changes are manifested in a child's hour-to-hour experience with family and friends. A timebudget study by Montemayor and Brownlee (1987) found a decline between sixth to seventh grade and eighth to twelfUi grade in amount of time spent with fathers, but not mothers. In a questionnaire study by Buhrmester and Furman (1987), eighth graders rated their frequency of interactions with mothers, fathers, and teachers as lower than did flfth graders, and rated their frequency of interaction with same-sex friends (girls only) and opposite-sex friends as greater; ratings of time with grandparents and a closest sibling did not differ by age. In another questionnaire study, Crockett, Losoff, and Petersen (1984) found eighth-grade girls to report much more time talking on the phone than sixth-grade girls. These limited findings suggest a shift in time from family members to friends but do not provide precise information on the overall trends or on the quality of afFect experienced with these different companions.
often with family members and more often alone or with friends.
There is little research regarding the time young adolescents spend by themselves. Being alone is a frequent alternative for adolescents to being with family members and friends (Montemayor, 1982) and may present distinct liabilities and opportunities for development (Larson, 1990). Projective data collected by Coleman (1974) and questionnaire data obtained by Marcoen, Goossens, and Caes (1987) suggest that adolescents may find being alone less frightening than do children, although Marcoen et al. did not find that adolescents reported any greater desire to be alone than did children.
In addition to testing the two hypotheses, we include exploratory analyses aimed at providing fuller understanding of the age differences. First, we conduct tests of quadratic age trends along with tests of linear trends in order to detect nonmonotonic age patterns. Family relationships, for example, have sometimes been found to show nonlinear age trends across this period (Hill & Holmbeck, 1986; Steinberg, 1981). Second, we evaluate age differences in time with specific subcategories of companions, such as mothers, extended family members, and same-sex chums, since prior research suggests that patterns may vary for each. Last, we examine whether trends in time with family and friends are related to shifts in where adolescents spend time. We expected that youth might spend more time away from home across this age period and that this might be interrelated with changes in companionship (see Montemayor & Van Komen, 1980).
The present study employs time-sampling data obtained using the Experience Sampling Method (ESM) in which participants carry electronic pagers and report on their activities, companionship, and internal states at random times when signaled by the pagers (Csikszentmihalyi & Larson, 1987; Larson & Csikszentmihalyi, 1983). In an earlier set of papers on these data, we focused on age differences in the activities these youths reported (Larson & Richards, 1989). It was found that older participants—the adolescents—-spent less time playing, watching TV, and participating in sports than preadolescents and spent more time talking (particularly girls) and listening to music. Within many activities, age trends in companionship were noted, with older participants reporting doing the activity less
In this article, we reanalyze these data with the focus on companionship. Two central hypotheses are tested: first, that early adolescence is associated with decreased time with family and increased time with friends, and second, that average affect with family becomes more negative and affect with friends more positive across this age period. Since past theory and research suggest gradual, cumulative age changes, we test the hypotheses by evaluating linear age trends. Also, since sex differences have been found frequently in the work cited above and in many other studies of adolescent relationships (e.g., Cilligan, 1982; Montemayor & Van Komen, 1980; Savin-Williams, 1979; Youniss & Smollar, 1985), we decided tliat boys and girls should not be grouped in the same analyses but rather should be considered separately. By making this choice, we elect not to evaluate the comparison between sexes but rather to focus on the trends within each sex.
Method Sample The participants in this study were 483 randomly selected flfth to ninth graders (ages 9-15) from four suburban Chicago school neighborhoods. Two of the neighborhoods were in a working-class suburb, two in a middle- and upper-middle-class suburb. All four were composed almost exclusively
286
Child Development
of people of European ancestry. The school systems these youths attended did not subject them to major changes in school structure or cohort until ninth grade, at which point students from several schools were brought together in large high schools. Participants took part in one of eight waves, occurring over a period of 2 years. Sample selection was stratifled to obtain equal numbers of (a) boys and girls, (b) from each neighborhood, (c) in each season of the year between the beginning of the flfth grade and the winter of the ninth grade. The flnal sample of 483 includes 70% of the 688 randomly selected individuals initially invited to participate. Twenty-four percent of those invited decided not to participate or failed to obtain permission from their parents. Six percent began the study but failed to provide usable data. The overall participation rate was higher during the school year (75%, compared to 53% for the summer). Sample loss was approximately equal among boys and girls but was somewhat greater for boys in the older grades (flfth to ninth: 24%, 23%, 33%, 42%, 37%). A survey of the entire flfth to eighth grade in two of the schools indicated that students who declined to participate did not differ in self-esteem or in the socioeconomic prestige of their parents' jobs. Participation rates were somewhat higher among students in single-parent families and somewhat lower among students with remarried parents. A complete description of the sample is provided in Larson (1989a). Procedures At the beginning of the week of participation, each student was given an electronic pager and a booklet of self-report forms. Their instructions were to carry the pager and booklet with them for a week and fill out one self-report form immediately after receipt of each pager signal. Seven signals were sent each day, with one at a random time within every 2-hour block between 7:30 A.M. and 9:30 P.M. Nearly all participants started the study on a Tuesday or Wednesday. At the end of the 7-day period, the pagers and booklets were collected; the students were asked to take part in an interview and flll out a series of questionnaires, and they were paid $8.00 for their participation. In total, the students responded to 18,022 signals, providing an average of 37.3 self-reports per person. To calculate the students' rate of response to the pager signals, we discounted times when the students
went to bed early or slept late since they had been encouraged to turn the pager off in such instances. We also estimated that approximately 6% of the pager signals were not received due to mechanical failure of the pager. Discounting these factors, the students filled out self-reports for approximately 86% of the signals sent to them. Analyses of additional data indicated that missed reports occurred across a wide range of activities and situations, but may have been more common during sports, church services, family trips, and socializing away from home. Larson (1989a) calculated adjusted estimates of the students' allocation of time to different activities, taking these differential rates of loss into account, and found that the adjustments made only slight differences in estimates of how time was spent. Hence, the missed self-reports appear to have only a minor influence on the time budget estimates. Measures On each self-report form, participants responded to a series of items asking about their situation and state just before they received the signal. Companionship.—Companionship was determined by responses to a fixed-response item asking, "Who were you with (or talking to on the phone)?" Fifteen nonexclusive choices could be checked. Responses have been coded into four superordinate categories—family, friends, alone and other— which are composed of 13 subcategories. Family includes times the student marked being with: (1) family group (any constellation of nuclear family members that includes at least one parent), (2) mother only, (3) father only, (4) sibling(s) only, and (5) extended family (one or more extended relatives with or without the presence of nuclear family members). Friends includes times with: (6) same-sex friend, (7) same-sex group, (8) mixed-sex group (two or more friends, of whom at least one is not of the respondent's sex), (9) opposite-sex friend, and (10) friends and family (any combination of people that includes at least one friend and one family member). This last subcategory was included with friends because analyses from this and a previous ESM study of adolescents (Csikszentmihalyi & Larson, 1984) suggest that friends typically set the activity and tone in these situations: activities and moods more closely resemble times with friends than times with family. Alone includes: (11) alone, others nearby (student marked "alone, other people nearby"), and (12) all alone (student marked "alone, no one
Larson and Richards 287 around"). Instances when a person reported talking on the phone, but marked being alone, were coded according to the phone companion because mood states in this situation suggest they should be coded as such (see Larson, 1979, 1990). Others is composed of only one subcategory; (13) others (includes teachers, coaches, and other adults, as well as co-workers at a job and children being cared for by the student; also includes times when students were talking on the phone and failed to indicate with whom and times when they were with friends and others). Data for an additional category, times when the students marked that they were "in class," are excluded from the analyses for reasons explained in the analysis section below.
validity (Larson, 1989b). Since there were substantial differences in individuals' mean responses to these items, including a decline in mean affect with age (Larson & Lampman-Petraitis, 1989), we have standardized values on this scale so that differences between contexts can be examined considering only within-individual variation. This standardization was computed by the equation, z^ =- (Xy - meanj)/SDj, in which Xfj was the raw score for a given selfreport, mean^ was the person's mean for the scale, and SD^ was the person's standard deviation for the scale. Hence, each person's z-scores are distributed around a mean of 0.0 with a standard deviation of 1.0, with positive scores indicating affect that is more positive.
The reliability of the students' companionship reports is demonstrated by consistency in individuals' reports from the first to the last half of the week. Each person's sequence of self-reports was split at the midpoint in the week and the percentage of time reported in each companionship category computed separately for the first and second half. These percentages were highly correlated (time alone; r[483] = .51; time with family; r[483] = .53; time with friends: r[483] = .49), suggesting that individuals were quite stable in patterns of social companionship across the week.
Data Analyses
Location.—On the ESM self-report form, students were asked the open-ended question, "Where were you?" Responses to this item were coded into 68 categories (interrater agreement = 99%). For the purposes of this report, these categories have been collapsed into three: (a) at home, not in bedroom; (b) at home, in bedroom; (c) away from home. Bedroom is considered a separate category because previous findings show it to have particular importance as a refuge of privacy for adolescents (Parke & Sawin, 1979). Activity.—On the ESM form, students reponded to a question asking what they were doing. Responses were coded into 127 categories, with an interrater agreement of 94%. For analyses here, we have grouped these into 11 categories. Affect.—Students rated their mood at each signal on a set of 7-point semantic differential items. A factor analysis study indicated that three of these items (happyunhappy, cheerful-irritable, friendly-angry) form a scale of affect, which has strong internal reliability (alpha = .75) and construct
Exclusion of class time.—In order to control for summer/school year differences and focus on discretionary time, we have excluded occasions when students reported being "in class" (times in school when students were not in class were included). For participants during the school year, class time constituted an average of 30.7% of their self-reports, a percentage that did not vary as a function of grade or sex, F(9,391) = 1.15, p = .32. Class time accounted for only .9% of the summer participants' time. Season and social class.—With this exclusion, summer participants did not differ in the proportion of time they reported spending with family, friends, or alone. A three-way MANOVA with grade, sex, and season as the factors indicated significant differences in the percentage of time individuals reported being in each of these three superordinate companionship categories, F(9,1134) = 2.08, p = .03. These seasonal differences, however, were quite small, attributable to slightly more time (4%) spent with friends and less time spent with family in the spring. There also were no significant interactions of season with grade and sex. Hence, season was not included in the main analyses of the article. Preliminary analyses comparing students from the working-class and middle-class communities showed no significant main effects or interactions, hence this social class variable was also not included in the analyses reported below. Analysis strategy.—The hypotheses are tested using aggregate scores for each person as the unit of analysis, following the recommendations of Larson and Delespaul (1990). For analyses of quantity of time.
288
Child Development
these scores are percentages (e.g., the percentage of time spent alone); for the investigation of affect, these scores are means (e.g., mean affect when alone). These scores were analyzed using multivariate analysis of variance (MANOVA). Planned univariate twotailed t-tests, evaluating the significance of linear and quadratic polynomial grade trends, are examined only when the multivariate F is significant. For evaluation of the hypotheses, we employ an alpha of .05; for the exploratory analyses (which include evaluation of the quadratic terms) we use an alpha of .01.
Results Age Differences in Daily Companionship Our first hypothesis predicted a shift in allocation of time from family to friends across this age period. We evaluated this hypothesis by conducting MANOVAs in which the percentages of time a student spent with family, friends, and alone were the dependent variables and grade was the factor. A highly significant effect for grade was found for both boys, F(12,708) = 3.79, p < .001, and girls, F(12,7H) = 4.17, p < .001. These flndings show a dramatic decline in time with family across this age period (Fig. 1). In flfth grade, both genders spent close to half (47.8%) of their nonclass waking hours with their families. In ninth grade, this flgure was closer to one-quarter (28.1% for boys and 26.5% for girls). The univariate linear grade trend for family time was highly significant for both boys, t = 6.04, p < .001, and girls, t = 5.37, p < .001. A significant quadratic term for girls, t = 2.86, p = .005, reflected a steeper drop for them in family time between the fifth and sixth grades. Diminished time witii family among older students was replaced, not by time with peers, but by time spent alone, particularly among boys. The amount of time boys spent with friends did not differ significantly across this age period, while the amount of time they spent alone increased by 16.4% between the fifth and ninth grades, linear t = 4.79, p < .001. With age, boys spent less time with their families and more time alone. Girls spent more time with friends and alone with age. The amount of time girls spent with friends increased by 13.5% from the flftii to ninth grades, linear t = 3.29, p < .001. Amount of time spent alone showed both a linear trend, f = 2.18, p = .03, and a quadratic trend, t = 2.87, p = .004, indicat-
ing that amount of time alone for girls increased substantially between the fifth and seventh grade and then leveled off in the subsequent grades. As is evident in Figure 1, the time that older girls no longer spent with their families was spent alone in the sixth and seventh grade and then was increasingly devoted to friends in the eighth and ninth grades. Variations by specific companions.— We next evaluated whether these age trends could be attributed to specific comi^nionship subcategories by recomputing the MANOVAs employing the percentages of time in the first 12 of the 13 subcategpries as the dependent variables. Since the distributions of these percentages were skewed toward 0, a square-root transformation was used to normalize them for the MANOVAs. As before, these MANOVAs yielded a significant effect for grade both for boys, F(48,908) = 1.72, p - .002, and girls, F(48,924) = 1.83, p < .001. In examining the univariate trends, we employ an alpha of p < .01, since these analyses are exploratory and not hypothesis driven. The univariate significance tests indicate that the decline with age in amount of time with family can be attributed to differences in specific family subcategories (Table 1). In particular, both boys and girls in the older grades spent much less time with family groups. A signiflcant linear age difference also occurred in amount of time spent with siblings for boys and in the amount of time spent wi6i extended relatives, particularly for girls. It is notable that no age difference occurred in the amount of time participants spent with their mothers or faAers only. These univariate tests also suggest why net amount of time with friends did not increase with age for boys and does increase for girls. Boys in the older grades reported signiflcantly more time with one person of the opposite sex (Table 2). This trend, however, is counteracted by a close-to-slgniflcant age decrease in amount of time with friends and family jointly. For girls significant and close-to-significant linear age increases occurred for all subcategories of friends, except friends and family together. The most dramatic trend was in amount of time spent in mixed-sex groups, which was more than double in the ninth grade. The increase in time alone across this age period appears to be most attributable to age differences in amount of time spent "all alone." For boys, time alone with no one else nearby is more than three times greater
Larson and Richards
289
BOYS Mean Percent of Time
100%
100%
80%
-80%
60%
60%
With Friends Alone
40%
40%
20%
20%
With Family
9th
Girls Mean Peroent of Time
100%
100%
80%
-80%
60%
60%
With Friends Alone
40%
40%
20%
20%
With Family
5th
FIG. 1.—Daily companionship x grade (note that time in class and a small percentage of time spent with others are excluded).
in the ninth grade, as compared to the flfth, a trend that is highly significant. For girls, this trend is close to significant. The linear trend for time spent "alone others nearby" was also signiflcant for boys, but not for girls, who showed a close-to-signiflcant quadratic trend, ( = 2.17, p = .03, reflecting a peak in this category in the seventh grade. Variations related with location.—It was also anticipated that age trends in com-
panionship might vary as a function of location; adolescents were expected to spend less time with their families as a correlate of spending less time at home. The data, however, do not support this interpretation. First, we examined age trends in the amount of time spent in the three locations, irrespective of companionship. MANOVAs were conducted using the percent of time in two of the locations ("not at home" and "at
a.
Maryse H. Richards Loyola University of Chicago LARSON, REED, and RICHARDS, MABYSE H . Daily Companionship in Late Childhood and Early
Adolescence: Changing Developmental Contexts. GHILD DEVELOPMENT, 1991, 62, 284-300. The study empltws t^9^s^|mg;|j|i); data to exttmiiw age: d^^^mmm in ^ ^^Ef^^ i ^ j ^ g ^ t ^ of children's a ^ : ^ i i g i i ^ ^ i « B e n t s ' daily experience with »^eir femUies, ftfehSs, ^ri^iaBie. Kirtieipants l a i ^ s ^ i ^ carried electronic pagers for 1 week and reported their companionship, location, and affect at random times when signaled by the pagers. Findings show a dramatic decline in amount of time spent with family, with older students reporting half as much time with their families as younger students. Among boys, this family time was replaced by time spent alone; among girls, by time alone and with friends. Affect reported when with family became less positive between the fifth and seventh grade, but was more positive again in the ninth grade for boys. Affect with friends became more favorable across this age period; affect when alone did not vary. These age differences suggest changes in adolescents' daily opportunities for cognitive growth, emotional development, and social support.
The objective of this article is to chart age differences in the quantity and quality of time children and young adolescents spend with different companions. Theorists contend that individuation from parents and deepening of relationships with peers are central developmental projects of adolescence (BJos, 1967,- Gilligan, 1982,- Havighurst, 1984). Therefore, we expect adolescents to spend less time with tJheir families and more time with their friends than preadolescents. Accompanying age changes in emotional states experienced when with family and friends might also be expected. This article evaluates these age trends in daily interaction with family, with friends, and also alone, employing time-sampling data from Midwestern fifth to ninth graders. A large body of research substantiates theorists' predications that entry into adolesence is accompanied by a shift in the relative importance of family and friends. Age changes include: (a) a shift in "associative orientation" from family to friends (Bow-
erman & Kinch, 1959); (b) increased emotional detachment from parents, with changes occurring somewhat earlier for girls (Steinberg & Silverberg, 1986); (c) increased conflict with and emotional distance from parents, particularly around the occurrence of puberty (Steinberg, 1981, 1987); (d) increased behavioral independence from parents (Feldman & Quatman, 1988; Simmons & BIyth, 1987; Smetana, 1988); (e) increased influence of peers over certain domains of daily life (Berndt, 1979,1989; Hartup, 1983); and (f) increased intimacy with friends, especially among girls (Berndt, 1982;^ Hunter & Youniss, 1^2), and decreased intimate disclosure to parents (Buhrmester & Furman, 1987). Closeness and mutual valuation in relationships with parents do not necessarily decline in adolescence (Blyth, Hill, & Thiel, 1982; Hunter & Youniss, 1982; Smetana, 1988); nonetheless, this research clearly shows that closeness with friends increases and emotional autonomy from parents increases, with the timing of these trends varying somewhat by gender.
This research was supported by NIMH grant 1 ROl MH38324, awarded to the authors. We are grateful to Mary Ghandler for assistance with the data analyses. Gorrespondence should be addressed to Reed Larson, Division of Hnman Development and Family Studies, University of Illinois, 1105 W. Nevada, Urbana, IL 61802. [Child Development, 1991, 62, 284-300. © 1991 by the Society for Research in Child Development, Inc. All rights reserved. 0009-3920/91/6202-.0003$0LOO]
Largon and Richards
285
Little research, however, evaluates how these psychological and emotional changes are manifested in a child's hour-to-hour experience with family and friends. A timebudget study by Montemayor and Brownlee (1987) found a decline between sixth to seventh grade and eighth to twelfUi grade in amount of time spent with fathers, but not mothers. In a questionnaire study by Buhrmester and Furman (1987), eighth graders rated their frequency of interactions with mothers, fathers, and teachers as lower than did flfth graders, and rated their frequency of interaction with same-sex friends (girls only) and opposite-sex friends as greater; ratings of time with grandparents and a closest sibling did not differ by age. In another questionnaire study, Crockett, Losoff, and Petersen (1984) found eighth-grade girls to report much more time talking on the phone than sixth-grade girls. These limited findings suggest a shift in time from family members to friends but do not provide precise information on the overall trends or on the quality of afFect experienced with these different companions.
often with family members and more often alone or with friends.
There is little research regarding the time young adolescents spend by themselves. Being alone is a frequent alternative for adolescents to being with family members and friends (Montemayor, 1982) and may present distinct liabilities and opportunities for development (Larson, 1990). Projective data collected by Coleman (1974) and questionnaire data obtained by Marcoen, Goossens, and Caes (1987) suggest that adolescents may find being alone less frightening than do children, although Marcoen et al. did not find that adolescents reported any greater desire to be alone than did children.
In addition to testing the two hypotheses, we include exploratory analyses aimed at providing fuller understanding of the age differences. First, we conduct tests of quadratic age trends along with tests of linear trends in order to detect nonmonotonic age patterns. Family relationships, for example, have sometimes been found to show nonlinear age trends across this period (Hill & Holmbeck, 1986; Steinberg, 1981). Second, we evaluate age differences in time with specific subcategories of companions, such as mothers, extended family members, and same-sex chums, since prior research suggests that patterns may vary for each. Last, we examine whether trends in time with family and friends are related to shifts in where adolescents spend time. We expected that youth might spend more time away from home across this age period and that this might be interrelated with changes in companionship (see Montemayor & Van Komen, 1980).
The present study employs time-sampling data obtained using the Experience Sampling Method (ESM) in which participants carry electronic pagers and report on their activities, companionship, and internal states at random times when signaled by the pagers (Csikszentmihalyi & Larson, 1987; Larson & Csikszentmihalyi, 1983). In an earlier set of papers on these data, we focused on age differences in the activities these youths reported (Larson & Richards, 1989). It was found that older participants—the adolescents—-spent less time playing, watching TV, and participating in sports than preadolescents and spent more time talking (particularly girls) and listening to music. Within many activities, age trends in companionship were noted, with older participants reporting doing the activity less
In this article, we reanalyze these data with the focus on companionship. Two central hypotheses are tested: first, that early adolescence is associated with decreased time with family and increased time with friends, and second, that average affect with family becomes more negative and affect with friends more positive across this age period. Since past theory and research suggest gradual, cumulative age changes, we test the hypotheses by evaluating linear age trends. Also, since sex differences have been found frequently in the work cited above and in many other studies of adolescent relationships (e.g., Cilligan, 1982; Montemayor & Van Komen, 1980; Savin-Williams, 1979; Youniss & Smollar, 1985), we decided tliat boys and girls should not be grouped in the same analyses but rather should be considered separately. By making this choice, we elect not to evaluate the comparison between sexes but rather to focus on the trends within each sex.
Method Sample The participants in this study were 483 randomly selected flfth to ninth graders (ages 9-15) from four suburban Chicago school neighborhoods. Two of the neighborhoods were in a working-class suburb, two in a middle- and upper-middle-class suburb. All four were composed almost exclusively
286
Child Development
of people of European ancestry. The school systems these youths attended did not subject them to major changes in school structure or cohort until ninth grade, at which point students from several schools were brought together in large high schools. Participants took part in one of eight waves, occurring over a period of 2 years. Sample selection was stratifled to obtain equal numbers of (a) boys and girls, (b) from each neighborhood, (c) in each season of the year between the beginning of the flfth grade and the winter of the ninth grade. The flnal sample of 483 includes 70% of the 688 randomly selected individuals initially invited to participate. Twenty-four percent of those invited decided not to participate or failed to obtain permission from their parents. Six percent began the study but failed to provide usable data. The overall participation rate was higher during the school year (75%, compared to 53% for the summer). Sample loss was approximately equal among boys and girls but was somewhat greater for boys in the older grades (flfth to ninth: 24%, 23%, 33%, 42%, 37%). A survey of the entire flfth to eighth grade in two of the schools indicated that students who declined to participate did not differ in self-esteem or in the socioeconomic prestige of their parents' jobs. Participation rates were somewhat higher among students in single-parent families and somewhat lower among students with remarried parents. A complete description of the sample is provided in Larson (1989a). Procedures At the beginning of the week of participation, each student was given an electronic pager and a booklet of self-report forms. Their instructions were to carry the pager and booklet with them for a week and fill out one self-report form immediately after receipt of each pager signal. Seven signals were sent each day, with one at a random time within every 2-hour block between 7:30 A.M. and 9:30 P.M. Nearly all participants started the study on a Tuesday or Wednesday. At the end of the 7-day period, the pagers and booklets were collected; the students were asked to take part in an interview and flll out a series of questionnaires, and they were paid $8.00 for their participation. In total, the students responded to 18,022 signals, providing an average of 37.3 self-reports per person. To calculate the students' rate of response to the pager signals, we discounted times when the students
went to bed early or slept late since they had been encouraged to turn the pager off in such instances. We also estimated that approximately 6% of the pager signals were not received due to mechanical failure of the pager. Discounting these factors, the students filled out self-reports for approximately 86% of the signals sent to them. Analyses of additional data indicated that missed reports occurred across a wide range of activities and situations, but may have been more common during sports, church services, family trips, and socializing away from home. Larson (1989a) calculated adjusted estimates of the students' allocation of time to different activities, taking these differential rates of loss into account, and found that the adjustments made only slight differences in estimates of how time was spent. Hence, the missed self-reports appear to have only a minor influence on the time budget estimates. Measures On each self-report form, participants responded to a series of items asking about their situation and state just before they received the signal. Companionship.—Companionship was determined by responses to a fixed-response item asking, "Who were you with (or talking to on the phone)?" Fifteen nonexclusive choices could be checked. Responses have been coded into four superordinate categories—family, friends, alone and other— which are composed of 13 subcategories. Family includes times the student marked being with: (1) family group (any constellation of nuclear family members that includes at least one parent), (2) mother only, (3) father only, (4) sibling(s) only, and (5) extended family (one or more extended relatives with or without the presence of nuclear family members). Friends includes times with: (6) same-sex friend, (7) same-sex group, (8) mixed-sex group (two or more friends, of whom at least one is not of the respondent's sex), (9) opposite-sex friend, and (10) friends and family (any combination of people that includes at least one friend and one family member). This last subcategory was included with friends because analyses from this and a previous ESM study of adolescents (Csikszentmihalyi & Larson, 1984) suggest that friends typically set the activity and tone in these situations: activities and moods more closely resemble times with friends than times with family. Alone includes: (11) alone, others nearby (student marked "alone, other people nearby"), and (12) all alone (student marked "alone, no one
Larson and Richards 287 around"). Instances when a person reported talking on the phone, but marked being alone, were coded according to the phone companion because mood states in this situation suggest they should be coded as such (see Larson, 1979, 1990). Others is composed of only one subcategory; (13) others (includes teachers, coaches, and other adults, as well as co-workers at a job and children being cared for by the student; also includes times when students were talking on the phone and failed to indicate with whom and times when they were with friends and others). Data for an additional category, times when the students marked that they were "in class," are excluded from the analyses for reasons explained in the analysis section below.
validity (Larson, 1989b). Since there were substantial differences in individuals' mean responses to these items, including a decline in mean affect with age (Larson & Lampman-Petraitis, 1989), we have standardized values on this scale so that differences between contexts can be examined considering only within-individual variation. This standardization was computed by the equation, z^ =- (Xy - meanj)/SDj, in which Xfj was the raw score for a given selfreport, mean^ was the person's mean for the scale, and SD^ was the person's standard deviation for the scale. Hence, each person's z-scores are distributed around a mean of 0.0 with a standard deviation of 1.0, with positive scores indicating affect that is more positive.
The reliability of the students' companionship reports is demonstrated by consistency in individuals' reports from the first to the last half of the week. Each person's sequence of self-reports was split at the midpoint in the week and the percentage of time reported in each companionship category computed separately for the first and second half. These percentages were highly correlated (time alone; r[483] = .51; time with family; r[483] = .53; time with friends: r[483] = .49), suggesting that individuals were quite stable in patterns of social companionship across the week.
Data Analyses
Location.—On the ESM self-report form, students were asked the open-ended question, "Where were you?" Responses to this item were coded into 68 categories (interrater agreement = 99%). For the purposes of this report, these categories have been collapsed into three: (a) at home, not in bedroom; (b) at home, in bedroom; (c) away from home. Bedroom is considered a separate category because previous findings show it to have particular importance as a refuge of privacy for adolescents (Parke & Sawin, 1979). Activity.—On the ESM form, students reponded to a question asking what they were doing. Responses were coded into 127 categories, with an interrater agreement of 94%. For analyses here, we have grouped these into 11 categories. Affect.—Students rated their mood at each signal on a set of 7-point semantic differential items. A factor analysis study indicated that three of these items (happyunhappy, cheerful-irritable, friendly-angry) form a scale of affect, which has strong internal reliability (alpha = .75) and construct
Exclusion of class time.—In order to control for summer/school year differences and focus on discretionary time, we have excluded occasions when students reported being "in class" (times in school when students were not in class were included). For participants during the school year, class time constituted an average of 30.7% of their self-reports, a percentage that did not vary as a function of grade or sex, F(9,391) = 1.15, p = .32. Class time accounted for only .9% of the summer participants' time. Season and social class.—With this exclusion, summer participants did not differ in the proportion of time they reported spending with family, friends, or alone. A three-way MANOVA with grade, sex, and season as the factors indicated significant differences in the percentage of time individuals reported being in each of these three superordinate companionship categories, F(9,1134) = 2.08, p = .03. These seasonal differences, however, were quite small, attributable to slightly more time (4%) spent with friends and less time spent with family in the spring. There also were no significant interactions of season with grade and sex. Hence, season was not included in the main analyses of the article. Preliminary analyses comparing students from the working-class and middle-class communities showed no significant main effects or interactions, hence this social class variable was also not included in the analyses reported below. Analysis strategy.—The hypotheses are tested using aggregate scores for each person as the unit of analysis, following the recommendations of Larson and Delespaul (1990). For analyses of quantity of time.
288
Child Development
these scores are percentages (e.g., the percentage of time spent alone); for the investigation of affect, these scores are means (e.g., mean affect when alone). These scores were analyzed using multivariate analysis of variance (MANOVA). Planned univariate twotailed t-tests, evaluating the significance of linear and quadratic polynomial grade trends, are examined only when the multivariate F is significant. For evaluation of the hypotheses, we employ an alpha of .05; for the exploratory analyses (which include evaluation of the quadratic terms) we use an alpha of .01.
Results Age Differences in Daily Companionship Our first hypothesis predicted a shift in allocation of time from family to friends across this age period. We evaluated this hypothesis by conducting MANOVAs in which the percentages of time a student spent with family, friends, and alone were the dependent variables and grade was the factor. A highly significant effect for grade was found for both boys, F(12,708) = 3.79, p < .001, and girls, F(12,7H) = 4.17, p < .001. These flndings show a dramatic decline in time with family across this age period (Fig. 1). In flfth grade, both genders spent close to half (47.8%) of their nonclass waking hours with their families. In ninth grade, this flgure was closer to one-quarter (28.1% for boys and 26.5% for girls). The univariate linear grade trend for family time was highly significant for both boys, t = 6.04, p < .001, and girls, t = 5.37, p < .001. A significant quadratic term for girls, t = 2.86, p = .005, reflected a steeper drop for them in family time between the fifth and sixth grades. Diminished time witii family among older students was replaced, not by time with peers, but by time spent alone, particularly among boys. The amount of time boys spent with friends did not differ significantly across this age period, while the amount of time they spent alone increased by 16.4% between the fifth and ninth grades, linear t = 4.79, p < .001. With age, boys spent less time with their families and more time alone. Girls spent more time with friends and alone with age. The amount of time girls spent with friends increased by 13.5% from the flftii to ninth grades, linear t = 3.29, p < .001. Amount of time spent alone showed both a linear trend, f = 2.18, p = .03, and a quadratic trend, t = 2.87, p = .004, indicat-
ing that amount of time alone for girls increased substantially between the fifth and seventh grade and then leveled off in the subsequent grades. As is evident in Figure 1, the time that older girls no longer spent with their families was spent alone in the sixth and seventh grade and then was increasingly devoted to friends in the eighth and ninth grades. Variations by specific companions.— We next evaluated whether these age trends could be attributed to specific comi^nionship subcategories by recomputing the MANOVAs employing the percentages of time in the first 12 of the 13 subcategpries as the dependent variables. Since the distributions of these percentages were skewed toward 0, a square-root transformation was used to normalize them for the MANOVAs. As before, these MANOVAs yielded a significant effect for grade both for boys, F(48,908) = 1.72, p - .002, and girls, F(48,924) = 1.83, p < .001. In examining the univariate trends, we employ an alpha of p < .01, since these analyses are exploratory and not hypothesis driven. The univariate significance tests indicate that the decline with age in amount of time with family can be attributed to differences in specific family subcategories (Table 1). In particular, both boys and girls in the older grades spent much less time with family groups. A signiflcant linear age difference also occurred in amount of time spent with siblings for boys and in the amount of time spent wi6i extended relatives, particularly for girls. It is notable that no age difference occurred in the amount of time participants spent with their mothers or faAers only. These univariate tests also suggest why net amount of time with friends did not increase with age for boys and does increase for girls. Boys in the older grades reported signiflcantly more time with one person of the opposite sex (Table 2). This trend, however, is counteracted by a close-to-slgniflcant age decrease in amount of time with friends and family jointly. For girls significant and close-to-significant linear age increases occurred for all subcategories of friends, except friends and family together. The most dramatic trend was in amount of time spent in mixed-sex groups, which was more than double in the ninth grade. The increase in time alone across this age period appears to be most attributable to age differences in amount of time spent "all alone." For boys, time alone with no one else nearby is more than three times greater
Larson and Richards
289
BOYS Mean Percent of Time
100%
100%
80%
-80%
60%
60%
With Friends Alone
40%
40%
20%
20%
With Family
9th
Girls Mean Peroent of Time
100%
100%
80%
-80%
60%
60%
With Friends Alone
40%
40%
20%
20%
With Family
5th
FIG. 1.—Daily companionship x grade (note that time in class and a small percentage of time spent with others are excluded).
in the ninth grade, as compared to the flfth, a trend that is highly significant. For girls, this trend is close to significant. The linear trend for time spent "alone others nearby" was also signiflcant for boys, but not for girls, who showed a close-to-signiflcant quadratic trend, ( = 2.17, p = .03, reflecting a peak in this category in the seventh grade. Variations related with location.—It was also anticipated that age trends in com-
panionship might vary as a function of location; adolescents were expected to spend less time with their families as a correlate of spending less time at home. The data, however, do not support this interpretation. First, we examined age trends in the amount of time spent in the three locations, irrespective of companionship. MANOVAs were conducted using the percent of time in two of the locations ("not at home" and "at
a.
