Mixed-Gender Groups, Dating, and Romantic Relationships in Early Adolescence


Requests for reprints should be sent to Jennifer Connolly, Department of Psychology, York University, 4700 Keele St., Toronto, Ontario, Canada M3J 1P3. E-mail: connolly@yorku.ca


This study examined dating-stage and developmental-contextual models of romantic relationships during early adolescence. Same-gender friendships, affiliation with mixed-gender groups, dating, and romantic relationships were investigated in a sample of 1,284 young adolescents of diverse ethnocultural backgrounds. Data were collected cross-sectionally in Grades 5 through 8, as well as longitudinally in the fall and spring of an academic year. Consistent with a stage model, affiliation with mixed-gender groups and dating were qualitatively distinct activities that were sequentially organized and facilitated the progression from same-gender friendships to dyadic romantic relationships. The results also provide insights on how the developmental context may alter stage pathways: Dating activities were incorporated with mixed-gender affiliations, group-based romantic stages showed more stability than other stages, and the ethnocultural context influenced romantic timing. Finally, results indicated that participation in romantic activities heightened adolescents' future interest in having a romantic relationship.

One of the most striking social changes of adolescence is the emergence of romantic relationships (Collins, 2003). By late adolescence, most North American teenagers have had at least one romantic relationship (Carver, Joyner, & Udry, 2003). Before dyadic relationships emerge, young adolescents begin to participate in mixed-gender groups that provide opportunities for cross-gender associations and for dating (Connolly & Johnson, 1996; Feiring, 1999). Traditional dating-stage theories have hypothesized that these early romantic activities form a developmental sequence that leads progressively to romantic relationships (Feinstein & Ardon, 1973; McCabe, 1984). Despite their theoretical importance, empirical support for the claim that these activities represent a stage-developmental pathway to romantic relationships is scarce. In addition, contemporary views of adolescence highlight the possibility of more fluid growth patterns than proposed by classical developmental theories, as well as the importance of social context in shaping development (Brown, 1999; Lerner & Simi, 2000). The goal of this research was to examine early adolescent romantic activities from both traditional and contemporary perspectives. Characteristics of mixed-gender affiliative and dating activities as well as their developmental significance to the sequential emergence of romantic relationships in an ethnically diverse sample of young adolescents were explored.

In the early years of adolescence, having a romantic relationship is far from typical (Carver et al., 2003; Feiring, 1996). In contrast to older youths, less than 20% of young adolescents report a current boyfriend or girlfriend and this percentage increases only to 33% among 15- and 16-year-olds (Connolly & Johnson, 1996; Feiring, 1996). Although young adolescents are most often not in a dyadic romantic relationship, it is inaccurate to infer that romantic activity is limited at this time. Young adolescents are intensely preoccupied with romantic issues (Simon, Eder, & Evans, 1992), and they are fully aware of the dynamics of romantic relationships (Connolly, Craig, Goldberg, & Pepler, 1999). Moreover, young adolescents begin to affiliate in mixed-gender peer groups, and interactions with other-gender peers begin to occur (Broderick & Rowe, 1968; Montgomery & Sorell, 1998).

The significance of early romantic activities has been the subject of considerable speculation (Collins, 2003). Adopting a stage-developmental approach, traditional dating theories suggest that romantic activities are sequentially organized so that adolescents move progressively along a continuum of increasingly intimate romantic contact (Feinstein & Ardon, 1973; McCabe, 1984). This progression provides adolescents with a sequenced introduction to romantic relationships. Stage models suggest that two levels of romantic activities precede dyadic romantic relationships. In the transition from same-gender friendships, adolescents join mixed-gender groups. Next in the sequence, some of the adolescents in the group begin to form pairs who date as part of the group activities. The final stage is the initiation of dyadic romantic relationships that can function outside of the supportive structure of the peer group.

This developmental perspective raises important questions about the forms of early romantic activities and their contributions to the emergence of romantic relationships. From the perspective of a classic, or hard-stage, developmental model (Flavell, 1971), romantic stages should be composed of qualitatively distinct behavioral clusters and emerge in a consistently progressive sequence as one stage replaces the preceding one. Stage theories have been highly influential in accounting for cognitive features of adolescent growth. However, the balance of evidence suggests that hard-stage models are less successful in accounting for social development (Kohlberg & Ryncartz, 1990). Instead, social growth follows a soft-stage model in which structural clusters of behavior are evident, but consistently progressive movement through the stages is not obligatory (Kohlberg & Ryncartz, 1990). Indeed, stage stability or backward movement might occur at certain points of development. The possibility that romantic changes may not be fully captured by a lock-step sequence of stages was first alluded to in Dunphy's (1963) ethnographic analyses of adolescent cliques, crowds, and romantic relationships. Based on his observations, Dunphy concluded that individual and contextual influences could create variability in the ordering of developmental stages. He suggested that peer group stages could also be viewed as phases of development because adolescents' progression could entail considerable fluidity of growth as they pass in and out of peer groups and romantic involvements.1

Developmental-contextual theories of adolescent romance have built on the observations of Dunphy (1963) and describe romantic relationships within the broader framework of adolescents' interpersonal connections (Brown, 1999; Connolly & Goldberg, 1999; Connolly, Furman, & Konarski, 2000). Most recently, the peer group has been highlighted for its promotion of romantic relationships (Connolly et al., 2000). As adolescents join socially with groups of mixed-gender peers, the possibility of romantic relationships is increased by exposure to potential romantic partners as well as by observation of peer models of romantic cross-gender interactions.

In addition to providing a context for romantic relationships, developmental attributes of the peer group may influence the sequencing of early romantic stages. For example, forward progression may be less evident for romantic stages embedded in mixed-gender groups than for other stages. The transition to adolescence is a period of rapid change, and young adolescents seek peer connections to obtain a stable point of reference during this time (Brown, 1999; Newman & Newman, 2001). Because of this, we suggest that early romantic stages that concurrently promote participation in mixed-gender groups are likely to be more attractive than romantic stages that potentially draw adolescents away from these groups, such as dyadic romantic relationships. Second, affiliative and dating stages may coexist within adolescents' repertoires of romantic experience rather than replace one another. The growth of peer networks in adolescence has been characterized by an expansion of complementary relationships rather than replacement of preexisting relationships (Furman & Buhrmester, 1992). Hence, we suggest that romantic activities in early adolescence are incorporated into preexisting activities rather than abruptly replacing them. Finally, the gender composition of adolescents' peer groups may influence the timing of romantic development. Young adolescents with many other-gender peers in their networks move more quickly into romantic relationships than do adolescents with small other-gender networks (Feiring, 1999). Consistent with this, girls in single-sex schools move less quickly into adolescent activities than do girls in coed schools (Caspi, Lynam, Moffitt, & Silva, 1993; Mael, 1998). Thus, we suggest that schools that limit access to other-gender peers might delay romantic involvement in the early adolescent years. This effect may be most noticeable among girls because they typically report larger peer networks and more other-gender friends than do boys (Connolly & Johnson, 1996; Feiring, 1999).

Early romantic activity may also be influenced by the sociocultural context of adolescents' families. Comparisons of youths in North America, Europe, and Asia reveal differences in the expected transition into many adolescent activities including dating (Feldman & Rosenthal, 1991). Asian cultures emphasize interdependence among family members more than do Western cultures (Marcus & Kitayama, 1991) and do not encourage their children's early participation in heterosexual activities (Feldman & Rosenthal, 1990). Reflecting these familial values, adolescents of Asian descent endorse an older age at which dating should ideally begin than do their contemporaries of western European backgrounds (Feldman & Rosenthal, 1991). It is likely that transitions into the earliest phases of romantic activity will occur later for adolescents of Asian descent than for adolescents of other ethnic backgrounds.

An intriguing aspect of Dunphy's (1963) theory of peer groups is the proposal that the popular adolescents in mixed-gender groups provide models of romantic conduct for romantically naive adolescents. From a developmental-contextual perspective, adolescents who are not currently in romantic relationships are nonetheless exposed to the possibility of romantic interactions through participation in mixed-gender peer groups. This may well increase their interest in having a romantic relationship, and mixed-gender groups in which dating occurs may be especially facilitative of this interest. When investigating this link between romantic activity and romantic interest, puberty may be important. Pubertal maturation has been associated with dating and romantic activities (Gargiulo, Attie, Brooks-Gunn, & Warren, 1987). There has been speculation that this linkage is indirect and is mediated by peer influences (Buchanan, Eccles, & Becker, 1992; Dornbusch et al., 1981). Conclusive evidence of the independent roles of puberty and peer groups on romantic interest is, however, not yet available.

In this study we examined affiliations with mixed-gender groups, dating, and romantic relationships among young Canadian adolescents of Asian, European, and Caribbean backgrounds. To investigate developmental patterns, we invited participants enrolled in Grades 5 to 8 to take part in the study and we collected longitudinal data in the fall and spring of the school year. Consistent with dating-stage theories, we expected that mixed-gender group affiliations and dating activities would form distinct clusters of behavior and that changes from one stage to another would follow a sequenced course of development leading progressively to dyadic romantic relationships. However, incorporating contemporary developmental theories, we predicted more fluidity in romantic changes than would be expected in traditional stage theories. We expected that dating would coexist with mixed-gender affiliations rather than fully replace them. We also expected that romantic stages embedded in mixed-gender peer groups would show less change than other stages. As well, less romantic activity should be evident for Asian Canadian adolescents and for adolescents in single-sex schools. Finally, we tested the notion that one important outcome of participating in mixed-gender peer groups is an increase in the romantic interest of young adolescents who do not have a current romantic relationship.



The sample initially included 1,375 students (675 boys, 700 girls). Of these, 1,284 students also participated in the spring. The average age of the sample was 12.28 years (SD=1.11 years) and ranged from 9.31 to 14.84 years. In the fall there were 238 students in Grade 5, 276 in Grade 6, 417 in Grade 7, and 444 in Grade 8. In each school, Grades 7 and 8 had larger enrollments than Grades 5 and 6, and this is reflected in their proportion of the total sample. The students were enrolled in six elementary schools (Grades 1–8) that were located in a large south-central Canadian city. Four of the schools were coeducational and were part of the public education system. Two schools, which were independently administered, were single sex: one for boys and one for girls. The single-sex schools were located in close proximity to each other and school administrators took advantage of this to offer, on a voluntary basis, joint educational and social opportunities (e.g., dramatic productions, dances) for their students. A description of the research was mailed to the adolescents' families and signed parental consent was required. Student participation in the research averaged 78% of the available student population and was representative of the overall student body. The students who remained in the study at Time 2 did not differ demographically from those who did not continue. Slightly more boys (5%) than girls (2%) were absent at the second testing time, χ2(1)=8.98, p<.01.

A demographic questionnaire was included in the research protocol and adolescents were asked about the ethnicity or race with which they and their families identified. Family structure and parental education was also assessed. These sample characteristics are portrayed in Table 1.

Table 1. 
Demographic Characteristics of the Sample
 Percentage of the Sample
Ethnocultural background
 European Canadian70%
 Asian Canadian14%
 Caribbean Canadian6%
 Mixed heritage5%
Family structure
Single parent15%
Parent education
 High school or less
 Some university

To evaluate acculturation, the adolescents were asked about language use in their homes. English was the primary language for 96% of the Caribbean Canadians and 92% of the Asian Canadians. To quantify variation in socioeconomic circumstances, adolescents were assigned a socioeconomic status (SES) value using the Blishen scale (Blishen, Carroll, & Moore, 1987), which indexes Canadian occupations and income. Values for this scale can range from 15 to 100. Based on paternal occupations, a sample mean of 60.76 (SD=19.65) was obtained. This value is typical of lower-middle- and middle-class families. The three ethnic groups differed on SES, F(2, 1007)=13.27, p<.001. The mean for Caribbean Canadian youths was lower than the one for Asian Canadians and European Canadians (M=49.31, SD=2.52 vs. Ms=62.70 and 65.46, SDs=2.30 and 2.11).


Romantic activities

To assess participation in romantic activities, we administered a Dating Questionnaire in the fall and spring. Students were told that the questionnaire concerned love, dating, and romantic relationships among adolescents. The adolescents were given a list of eight items describing social activities that “adolescents could spend time doing after school and on weekends.” They were asked to endorse each item as either true or not true for themselves. The items were developed from two scales of heterosocial involvement (Broderick & Rowe, 1968; Silverberg & Steinberg, 1990). Further information was obtained from discussion groups about early adolescent social experiences conducted with three groups of boys and girls enrolled in a school not included in the current study. One item asked about participation in same-gender activities. Three items described activities that take place in mixed-gender groups but are not organized with reference to dating. Four items described activities in which there is a high level of romantic intention or explicit reference to dating. For these items, dating was defined as spending time or going out with a girl (or a boy) whom the adolescent liked, loved, or had a crush on. The eight items are listed in the Appendix.

The Dating Questionnaire also asked the adolescents if they had a current boyfriend or girlfriend and to indicate how long they had been together. They were asked how often they saw their boyfriend or girlfriend outside of school. The response options were: daily, a few times a week, once a week, once a month or less. If they did not have a current relationship, they were asked if they had recently (in the last 6 months) or ever had a boyfriend or girlfriend. At Time 1, 268 adolescents (20%) indicated they had a current romantic relationship, 79 (6%) indicated they had a recent relationship, and 317 (24%) indicated they once had a boyfriend or girlfriend. The modal length of the current relationship was 8 weeks (M=15.89 weeks, SD=32.81). Most adolescents reported after-school contact with their current boyfriend or girlfriend either weekly (23%) or a few times a week (41%). Daily after-school contact was reported by 23% of the adolescents and monthly contact was reported by 13% of the adolescents.

Romantic interest

To assess adolescents' interest in having a romantic relationship, the Dating Questionnaire included an item in which the adolescents indicated whether they would like to have a girlfriend or boyfriend in the near future. Only adolescents who did not have a current relationship were asked to respond. Possible responses included: “I don't care much about girl(boy)friends right now,” scored 1; “I'd like to have a girl(boy)friend, but it's not that important right now,” scored 2; and “I would really like to have a girl(boy)friend right now,” scored 3. The fall–spring reliability correlation was r(926)=.85, p<.001.

Pubertal maturation

In the fall and the spring, adolescents completed the Pubertal Development Scale (Petersen, Crockett, Richards, & Boxer, 1988), which describes changes in secondary sex characteristics. For both boys and girls these included growth spurt, appearance of body hair, and changes in skin. The girls were asked about breast growth and the boys were asked about facial hair and voice changes. Each item was rated on a scale ranging from 1 (has not yet started) to 4 (seems completed) and the average score across the items was computed. Scores ranged from 1 to 4 for both boys and girls. Responses were missing for 63 adolescents in the fall and 72 in the spring. The alpha coefficients were .78 in the fall and .76 in the spring. The fall–spring correlation was r(1212)=.83, p<.0001.


Romantic Activity in Early Adolescence

As a first step, we determined the percentage of adolescents who endorsed each activity on the Dating Questionnaire. These percentages, averaged across fall and spring, are shown in Table 2. All romantic activities were endorsed, with the rate of positive endorsement averaging 48% across the seven items. Most adolescents participated in some romantic activities and very few reported only same-gender interactions.

Table 2. 
Participation in Romantic Activities in Early Adolescencea
 Percentage of
Endorsing Item
Item Factor
  • Note. N=1,284.

  • a

    Averaged across fall and spring data.

  • b

    b Standardized factor loadings of confirmatory factor analysis of fall Dating Questionnaire items.

Same-gender activities only6%  
Mixed-gender affiliative activities
 Hang around with boys and girls82%.58 
 Go to clubs, groups, or sports activities with boys and girls76%.56 
Go to dances or parties with boys and girls74%.72 
Dating activities
 Go out with group of boys and girls at night27% .67
 Go out with a boy (girl) and a couple of girls (boys)31% .67
 Go on dates with a boy (girl) in a group24% .66
 Girls and boys go on dates26% .62
Have a boyfriend (girlfriend) now21%  

Factor Structure of Early Adolescent Romantic Activities

Our next step was to consider the qualitative distinctiveness between the items that measure affiliation in a mixed-gender peer group and items in which dating is a significant component. To test this, we carried out a confirmatory factor analysis in which the affiliative items were constrained to load on one factor and the dating items were constrained to load on a second factor. Using the fall data, the results indicate an excellent fit with the hypothesized structure (n=1,284), χ2(13)=17.08, p=.20, root mean square error of approximation (RMSEA)=.02, adjusted goodness-of-fit index (AGFI)=.99, non-normed fit index (NNFI)=.99. The standardized factor loadings are shown in Table 2. This analysis was repeated with the spring data (n=1,284) and the results again confirmed the fit of the two-factor model, χ2(13)=34.54, p=.001, RMSEA=.04, AGFI=.98, NNFI=.98.

We next considered whether the factor loadings replicated across time. Using the spring data, two models were tested. In the first model the structural assignment of items to factors was constrained to replicate the fall model, and in the second model the factor loadings on the items were also constrained to replicate the fall model. Testing the difference in fit of the two models resulted in a nonsignificant delta chi-square, Δχ2(4)=4.368, p>.30, indicating that the spring data produced an acceptable fit to the fully constrained model.

Finally, we examined whether the two-factor model was equally suitable across gender, grade, ethnic group, and school. To accomplish this, confirmatory factor analyses were computed separately for boys and girls, for each grade, for each single-sex school, and for each ethnic group. In each case, the factor structure and factor loadings obtained for the total sample in the fall were used to predict the structure and loadings of each subgroup, also using fall data. The results of these analyses confirmed the validity of the two-factor model across gender, ethnic, grade, and school subgroups. Consequently, we computed two scores, mixed-gender affiliations and dating, by summing the number of true responses for the items loading on each factor. The alpha coefficients, averaged across fall and spring, were .65 and .76, respectively. The correlations between fall and spring scores were .60 and .70.

Developmental Patterns of Romantic Activities

Mixed-gender affiliative and dating activities

To consider developmental patterns in adolescents' affiliative and dating activities, a repeated-measures multivariate analysis of variance (MANOVA) was computed. Time of year and type of romantic activity were within-subjects factors, and grade, gender, ethnicity, and school type (coed, single-sex boys, single-sex girls) were between-subjects factors. Because the sample differed on SES, this variable was included as a covariate. Its effect was nonsignificant; therefore, it is excluded from the analyses reported here. For this MANOVA, as well as all subsequent analyses, a conservative alpha was employed (p<.01) to avoid interpreting trivial differences arising from the large sample size. The mean scores, broken down by grade and time of year, are shown in Table 3.

Table 3. 
Developmental Changes in Adolescents' Participation in Affiliative and Dating Activities
  Affiliative ActivitiesDating Activities
  1. Note. Affiliative score range=0 to 3; dating score range=0 to 4. N=1,220 at both times with complete data.

Grade 5Fall1.71 (1.12)0.36 (0.85)
Spring1.83 (1.10)0.47 (0.91)
Grade 6Fall2.00 (1.04)0.59 (1.00)
Spring2.16 (1.01)0.64 (1.07)
Grade 7Fall2.38 (0.92)1.15 (1.33)
Spring2.54 (0.83)1.48 (1.44)
Grade 8Fall2.57 (0.76)1.39 (1.38)
Spring2.64 (0.69)1.57 (1.43)

Significant multivariate main effects were found for: type of romantic activity, F(1, 1234)=9.18, p<.01; time of year, F(1, 1234)=450.68, p<.0001; grade, F(3, 1234)=89.50, p<.0001; and ethnicity, F(2, 1234)=35.18, p<.0001. Significant multivariate interactions with time were found for grade, F(3, 1234)=8.01, p<.0001; school type, F(2, 1234)=10.14, p<.001; and ethnicity, F(2, 1234)=5.25, p<.0001. The main effect for type of romantic activity indicated that mixed-gender affiliations were reported more often than dating at both time points.

To explore the Grade × Time interaction, post hoc Sheffe tests were conducted to compare the affiliation and dating scores from the fall to the spring, separately for each grade. Grade 5 students did not show increases in either romantic activity. Grade 6 students increased in mixed-gender affiliations but not in dating. Grade 7 students increased in both affiliation and dating. Grade 8 students increased in dating but not affiliation. Post hoc tests were also conducted to compare grades in the fall and spring separately. The pattern of differences was identical at both times. Mixed-gender affiliation was lowest in Grade 5, higher in Grade 6, and highest in Grades 7 and 8, which did not differ. Dating in Grades 7 and 8 was greater than in Grades 5 and 6, whereas the two lower and two higher grades did not differ from each other.

Comparable post hoc analyses were conducted to explore the Ethnicity × Time interaction. Considering changes from the fall to spring for each ethnic group, the European Canadians increased both in mixed-gender affiliations (fall M=2.30, SD=.03 vs. spring M=2.41, SD=.04) and dating (fall M=1.05, SD=.04 vs. spring M=1.26, SD=.04). Caribbean Canadian adolescents increased in mixed-gender affiliations from the fall (M=2.26, SD=.12) to the spring (M=2.62, SD=.09), but not in dating from the fall (M=1.01, SD=.13) to the spring (M=1.23, SD=.14). Asian Canadian adolescents remained constant for both affiliations (fall M=1.90, SD=.04 vs. spring M=2.07, SD=.02) and for dating (fall M=.49, SD=.12 vs. spring M=.57, SD=.09). We also considered differences between the ethnic groups, separately for the fall and spring. The pattern of results was identical at the two time points. The Asian Canadian adolescents reported less mixed-gender affiliations and less dating than either the European Canadian adolescents or the Caribbean Canadian adolescents. The scores of the European Canadian and Caribbean Canadian adolescents did not differ.

The post hoc analyses of the School × Time interaction indicated that students in the coed schools increased from the fall to the spring both in mixed-gender affiliations (fall M=2.30, SD=.03 vs. spring M=2.43, SD=.03) and in dating (fall M=.93, SD=.04 vs. spring M=1.13, SD=.04). Boys in the single-sex school increased in dating from the fall (M=1.17, SD=.08) to the spring (M=1.35, SD=.09). Girls in the single-sex school did not increase in either type of romantic activity from the fall to the spring.2

Romantic relationships

We also examined developmental patterns in dyadic romantic relationships. We found that 20% of the adolescents had a boyfriend or girlfriend in the fall and 22% reported one in the spring, an increase that was not significant, t(1220)=2.18. The percentage of adolescents with a boyfriend or girlfriend did not differ across grade, 19%, 17%, 22%, and 21%, respectively, χ2(3)=2.32, or between genders, χ2(1)=2.32. There were significant ethnic differences. Adolescents from European Canadian and Caribbean Canadian families were more likely to have a boyfriend or girlfriend than were Asian Canadians, average fall–spring percentages of 21% and 24% vs. 10%, χ2(2)=13.23, p<.001.3

Romantic Stage Sequencing

To examine romantic stage development, changes from the fall to the spring in adolescents' romantic activities were determined and loglinear techniques were applied. To do this, adolescents were assigned to one of four mutually exclusive categories, based on the Dating Questionnaire. These were: (1) same-gender friendships for adolescents who indicated same-gender friendships only and did not endorse any mixed-gender or dating items, (2) mixed-gender affiliations for adolescents who responded positively to at least one of the mixed-gender items but did not endorse any dating activities or report having a romantic relationship, (3) dating for adolescents who reported at least one dating item but did not have a romantic relationship, and (4) dyadic romantic relationships for adolescents who reported having one.4 Although this method of stage assignment can obscure variation in reports of multiple romantic activities, its use was considered reasonable in the current study as it allows for an evaluation of broad patterns of change in romantic stages. Adolescents were assigned to a fall and a spring category, based on their responses at each time. In the fall, 7% of adolescents were classified as same-gender, 43% as affiliative, 28% as dating, and 22% as dyadic romantic relationships. The percentages in the spring were 5%, 40%, 31%, and 24%, respectively.

The 4 × 4 contingency distribution of adolescents' spring romantic stage, as a function of their fall romantic stage, was determined, and conditional probabilities of each spring category were calculated for each fall category. These are shown in Table 4. The first step in the loglinear analysis was to determine the overall association between fall and spring stages (Tabachnick & Fidell, 1989). The result indicated that adolescents' romantic activities in the spring were strongly associated with their activities in the fall, χ2(9)=952.93, p<.001.

Table 4. 
Conditional Probabilities of Romantic Stages in the Spring, for Each Fall Romantic Stage
 Spring Romantic StageTotal N (Fall)
Fall romantic stage
 Same sex.
 Romantic dyad.
Total N (spring)625234022911,278

The next step was to determine whether grade, gender, school, or ethnicity moderated this association (Tabachnick & Fidell, 1989). The fully saturated loglinear model was first calculated in which spring romantic stages were predicted from the main effects of fall stage, grade, gender, school, and ethnicity as well as all interaction terms. Backward selection procedures were used to determine which effects were essential (i.e., significantly altered the goodness of fit) relative to the saturated model. Models were computed that progressively excluded the five-, four-, three-, and two-way interaction terms and all were nonsignificant, χ2(215)=.02, ns; χ2(480)=39.19, ns; χ2(756)=225.46, ns; and χ2(828)=274.44. We then tested the significance of the main effects by eliminating each from the model. Only the exclusion of fall romantic stages resulted in a significant loss of fit, χ2(174)=778.74, p<.001. Based on these findings, analyses of sequential changes from fall to spring were conducted for the total sample.

The final step in the loglinear analysis was to assess the predicted sequences of romantic stages from fall to spring. To do this, the conditional probabilities of fall-to-spring changes were compared with Bonferroni-adjusted two-way chi-square analyses (Tabachnick & Fidell, 1989). To assess sequential stage changes, we compared the likelihood of moving to the next stage to moving two stages in the sequence (i.e., skipping a stage). To assess progressive upward changes, we compared the likelihood of changing to the next stage with changing to the preceding stage. To assess stage stability, we compared the likelihood of staying in a stage rather than shifting to another stage. The comparisons that were carried out and the chi-square results are shown in Table 5. In the analyses, the expected frequencies were derived from the actual distribution of romantic categories in the fall because our results indicated that their base rates were not equal.

Table 5. 
Comparison of Changes in Romantic Stage From Fall to Spring
 χ2 Valuesa
  • Note. χ2(df=1).

  • a

    Comparison of conditional probabilities for each pair in subset.

  • ***


Assessing sequential change
 1. Same gender to affiliative vs. same gender to dating34.03***
 2. Affiliative to dating vs. affiliative to dyadic30.80***
Assessing forward movement 
 1. Affiliative to dating vs. affiliative to same-gender0.15
 2. Dating to dyadic vs. dating to affiliative35.75***
Stability of group stages
 1. Affiliative to affiliative vs. affiliative to dating172.46***
 2. Affiliative to affiliative vs. affiliative to same-gender12.26***
 3. Dating to dating vs. dating to dyadic18.53***
 4. Dating to dating vs. dating to affiliative149.83***
Stability of other stages
 1. Same gender to same gender vs. same gender to affiliative212.27***
 2. Dyadic to dyadic vs. dyadic to dating267.30***

As shown in Table 5, there is evidence of sequential organization of the stages. One- versus two-stage shifts could be evaluated for adolescents who were classified as same gender in the fall and who changed either one stage to affiliative activities or two stages to dating. One- versus two-stage shifts could also be evaluated for adolescents who were in affiliative activities in the fall and changed to either dating or a romantic dyad in the spring. For each comparison, a one-stage shift was more likely than a two-stage shift.

Partial evidence for progressive change was also found. Adolescents in the dating stage in the fall were less likely to move to the preceding stage in the spring (affiliative) than they were to change into the following stage. However, comparison of the shifts for adolescents in affiliative activities in the fall indicated that they were as likely to move to dating activities as they were to move to same-gender friendships. This nonsignificant finding is counterintuitive given the change evident in percentages shown in Table 4. The failure of this difference to be significant reflects our use of the more conservative actual rather than expected base rates for mixed-gender activities.

Along with support for these general trends for progressive and sequential stage changes, the results also provided evidence for some modification of the stage theory view of romantic stages. As shown in Table 4, there is evidence of stage stability for romantic stages embedded in mixed-gender groups. Adolescents reporting affiliative or dating activities in the fall were more likely to remain in that stage than to move to a different stage in the spring. This differed from the relative instability of the same-gender friends stage. Adolescents reporting same-gender friendships in the fall were more likely to move into affiliative activities in the spring than to remain at the same level. This pattern, however, did not hold for dyadic romantic relationships. Adolescents in a romantic relationship in the fall were more likely to again report being in a relationship in the spring than to have moved back to dating.

Romantic Activities and Romantic Interest

The final analyses explore whether participation in affiliative or dating activities stimulates interest in having a romantic relationship among adolescents who do not currently have one. Hierarchical multiple regression analyses were employed to determine whether romantic interest in the spring could be predicted from mixed-gender affiliative and dating activities in the fall, controlling for puberty.

First, developmental changes and individual differences in romantic interest were examined with a repeated-measures ANOVA in which grade (dichotomized into lower and higher grades), school, ethnicity, and gender were between-subjects factors and time was a within-subjects factor. Romantic interest increased significantly from the fall to the spring (M=.99, SD=.65 vs. M=1.04, SD=.68), F(1, 941)=43.95, p<.0001, and was higher in the more advanced grades (M=.74, SD=.65 vs. M=1.03, SD=.64), F(1, 941)=5.80, p<.01. It also differed across ethnic groups, F(2, 940)=11.80, p<.001. Asian Canadian adolescents expressed less romantic interest than did European Canadians or Caribbean Canadians, who did not differ from each other (M=.74, SD=.68 vs. M=1.04, SD=.61 and M=.86, SD=.59). There were no differences in romantic interest across gender.5

The result of the regression analysis was significant and yielded an R=.51, F(7, 980)=47.85, p<.001. Dating in the fall contributed to romantic interest in the spring, controlling for age, ethnicity, puberty, school, and fall romantic interest, β=.24, t(987)=7.66, p<.001. Affiliation with mixed-gender groups was not a significant predictor, after controlling for dating. There were no significant interaction effects.


Romantic activities in the early adolescent peer group have been highlighted by both dating-stage and developmental-contextual theories as critical for the emergence of romantic relationships. The results of this study enlighten our understanding of the nature of these activities and how they are linked to romantic relationships. Mixed-gender affiliative and dating activities were found to be common among the young adolescents of our sample. Moreover, they were influential in heightening young adolescents' romantic interest as well as leading to romantic relationships. Supporting a stage-developmental model, we found that affiliative and dating activities were distinct behavioral clusters, which formed stages that were located in the transition between same-gender friendships and romantic relationships. Modifying a stage-theory perspective, we found that dating activities complemented rather than replaced pre-existing activities and that dating stages embedded in peer groups showed greater stage stability and less progressive change than stages not so embedded. Reinforcing the moderating role of context, the adolescents' ethnocultural background was linked to romantic timing.

In early adolescence, dyadic romantic relationships are uncommon (Carver et al., 2003), yet exclusive participation in same-gender friendships is no longer the norm. As adolescents begin to include romantic relationships in their social networks, activities in the mixed-gender peer group are an important interactional forum. Most young adolescents in our study participated in group activities with other-gender peers. More often, these activities were not explicitly focused on dating but instead brought boys and girls together in settings in which heterosocial interaction might occur but is not obligatory. From our study we can only be certain that adolescents report being present in these groups. We cannot discern whether some of these groups are ones they are required to attend, such as those in a church or sports setting. Likewise, we cannot infer the degree to which the adolescents actually interacted with other-gender peers when in these groups. Nonetheless, we speculate that mixed-gender groups are important because they are easily available to young adolescents who can take part at their own comfort level. Some adolescents may participate in mixed-gender interactions only indirectly by observation whereas others are more active. In view of our findings that mixed-gender groups increase romantic interest as well as romantic involvement, it is possible that both interactional and observational experiences contribute to the process of romantic development at this time. Clearly, an understanding of the ongoing group process would enhance our appreciation of the role of mixed-gender groups on romantic development.

A central question of this study was the sequencing of romantic activities in early adolescence. Dating-stage theorists have argued that romantic activities form a developmental progression that leads sequentially from same-gender friendships to dyadic romantic relationships. In this study we considered whether a stepwise sequence of stages fully captured early adolescent romantic growth. Our results provide support for many elements of this perspective. However, consistent with our understanding of the salience of peer groups in early adolescence, we also found support for a more fluid model of romantic growth that allows for overlapping activities, as well as stability of certain stages.

Consistent with theories of dating stages (Feinstein & Ardon, 1973; McCabe, 1984), mixed-gender affiliations and dating formed distinct clusters of activities, and both were developmentally typical of early adolescence. As well, evidence was found for a sequence in which adolescents moved from same-gender friendships to mixed-gender affiliations, then to dating, and finally to a romantic relationship. Of equal importance, the sequence of changes revealed stage-like patterns of development. Changes were more often from less intense to more intense levels of romantic activity, and these changes typically involved a shift to an immediately succeeding step in the sequence rather than two or more steps. Overall, these stage-sequencing findings are striking. However, it is important to bear in mind that the dichotomous scaling of the romantic items, which formed the basis for stage classification, may obscure adolescent variability within each stage. Romantic scales that allow for greater variability within and between stages might yield different findings, and this is an important direction for future research.

Full support for a stage model of romance was also qualified by findings inconsistent with this approach. First, adolescents did not cease to participate in mixed-gender activities once dating had begun. Instead, both activities co-occurred and were part of the overall pattern of heterosocial interaction. This pattern is consistent with other aspects of adolescents' social development in which it is typical for new relationships to take their place alongside pre-existing relationships (Furman & Buhrmester, 1992). The general picture of co-occurring romantic activities was true for all ages, although the proportions of affiliative and dating activities shifted between the younger and older portions of our sample. Mixed-gender affiliations increased most for the youngest adolescents, whereas dating increased most for the oldest portion of the sample. Adolescents in the mid-range grades reported increases in both affiliative and dating activities. These age-differential patterns highlight the changing focus of adolescents' romantic intent across the early adolescent period.

Also counter to the predictions of a stage model, romantic growth was not always to successive stages. Changes to a preceding romantic stage sometimes occurred, and there was considerable stability of group-based romantic stages. These findings mirror the fluid nature of romantic growth in early adolescence and suggest a picture of gradual movement toward more intense or prolonged romantic involvement tempered with shifts to less intense stages that are centered in peer groups. Evidently, early adolescent romantic stages do not always follow a lock-step sequence of progressively forward movement. Instead, romantic development might be more aptly described by the Piagetian notion of decalage in which change into a new stage proceeds by small steps forward and then backward into the comfort of the familiar stage (Woodward, 1979). Similarly, romantic change involves gradual ventures into more intense romantic involvement from a strong base in the protective setting of the peer group. Such fluidity of romantic movement characterizes romantic involvement at later stages of the life cycle. Romantic relationships of older adolescents are highly embedded in their overall peer networks (Connolly et al., 2000). Similarly, recently divorced adults indicate that a significant challenge is to re-establish connections with their social networks as a first step in the process of finding a new partner (Levinger, 1983). Perhaps the stages of romantic growth seen in these young adolescents are evident at other points in the life span as well.

The findings revealed a consistent pattern of more limited romantic activity among Asian Canadian adolescents than among other adolescents. Differential patterns of romantic activities in three ethnic groups suggest that romantic participation is embedded in the broader cultural environment created by ethnocultural origins. Families differ in the level of autonomy they encourage among their adolescents, and Asian families typically endorse a later start for many adult activities (Feldman & Rosenthal, 1991). Consistent with this, the adolescents from Asian Canadian families in our sample were more likely to report same-gender friendships, less interested in romantic relationships, and less likely to participate in any romantic activity. Despite these differences in levels of participation, Asian Canadian adolescents were similar to the other adolescents in the sequential patterning of romantic stages. Recently, there has been considerable speculation about romantic activity timetables and variation in developmental trajectories (Feiring, 1999). At question is whether romantic activities initiated later in adolescence follow a sequence similar to those seen with an initiation in early adolescence. Our finding of similar stage sequences for all subgroups of our sample, despite differences in timing and frequency, suggests that romantic development may unfold in a similar fashion whenever it is initiated. Finally, it is of interest that these ethnic group differences were evident even though the groups were similar on our measure of acculturation. It is possible that the use of English in the home is not a sensitive measure of acculturation for Asian youths in a Canadian context. Many of these youths' families came originally from Hong Kong and therefore would be familiar with the English language. However, it is possible that ethnic differences in romantic activities persist even among highly acculturated groups because these activities continue to be shaped by traditional family values.

The inclusion of single-sex schools in the study allowed for an exploration of peer group gender composition on the timing of romantic development. In general, adolescents in single-sex and coed schools differed little in their patterns of romantic development. One finding of note was that girls in the single-sex school did not increase in their romantic activities over the academic year. Associating with a larger group of boys is important in initiating romantic development (Feiring, 1999) and can promote an earlier entry into adolescent misbehavior (Caspi et al., 1993). Because we did not find a similar pattern of delay for the single-sex boys' school, our results support the view of Caspi et al. (1993) that adolescent debut into adult-like activities is often initiated by boys. Nonetheless, caution must be taken in interpreting our single-sex findings. It would be important to verify these results in other schools and other settings. It is possible that single-sex environments in which mixed-gender interactions were actively discouraged might yield more highly differential findings.

Our findings were notable for the general absence of gender differences. The boys and girls were similar in their romantic activities and in their developmental sequencing. This similarity of boys' and girls' romantic activities is of interest because adolescent girls typically participate in larger peer networks, have more other-gender friends, and report higher levels of support from their romantic partners than do boys (Connolly & Johnson, 1996; Feiring, 1996, 1999). It is important to note that the adolescents in the current study are younger than those most typically studied. Our data suggest that gender differences might emerge only in later adolescence when relationship patterns have had an opportunity to coalesce.

A significant limitation of this study is the lack of precision about stage classification that results from the use of a dichotomous scale to assess romantic activities. For example, the absence of frequency data for same-sex friendships might underlie the small number of adolescents classified at this stage. The method of classifying adolescents into romantic stages also obscures meaningful variability among them. Within each stage adolescents may differ a great deal in the extent to which they engage in any of the stage behaviors. Yet our methodology does not allow for an assessment of these “phases within stages.” Further research using a measure of romantic activities refined to take into consideration the quality and frequency of interaction, as well as the context of interaction with same- and other-sex peers, is crucial to establish the validity of the stages suggested in the current research.

This study focused on the experiences of adolescents living in an industrialized Western environment. Developmental pathways to romance in non-Western cultures might be different, depending on the importance attached to peer group activities, the influence of the family on relationships outside of the home, and exposure to Western media (Lerner & Simi, 2000). Adolescents growing up in cultures that minimize adolescents' ability to choose freely their friends and romantic partners might follow different routes to romantic relationships, and the results of this study should not be generalized without further empirical evaluation.

In summary, this study is a first attempt to explore the romantic activities of young adolescents and their role in the emergence of romantic relationships. The results articulate the characteristics of romantic life at this point in the life cycle and advance our understanding of the developmental trajectory to romantic relationships. Future research might now be directed to the relational processes that advance these activities and the developmental trajectories of later adolescence.


Dating Questionnaire

Please answer these questions about dating, love, and romantic relationships among teenagers. Dating is going out or spending time with girls (boys) you like, love, or have a crush on.

1. Boys and girls can spend time together in many ways. Answer True or False in the boxes beside each of the sentences below to describe the types of ways you spend time together with girls (boys) after school and on weekends.

• I only spend time with other boys (girls)TF
• I hang around with both boys and girlsTF
• I go to dances or parties where both boys and girls are thereTF
• I go to clubs, groups or sports activities where both boys and girls areTF
• I meet a group of boys and girls at nightTF
• I go out with a another boy (girl) and a couple of girls (boys)TF
• I go on dates with a girl (boy), but with a groupTF
• I go on dates with a girl (boy), just the two of usTF

2. Do you have a girl(boy)friend right now?

• Yes, I have one right now. We have been going out for________weeks.
• No, but I had a girl (boy) friend since the beginning of school. We went out for________weeks.
• No, but I've had one before.
• No, I've never had a girl(boy)friend.

3. How often do you spend time after school or on weekends with your current girl(boy)friend:

• Once a day
• A few times a week
• Once a week
• Once a month or less

4. If you do not have a girl(boy)friend right now, would you like to have one in the near future?

• I don't care much about girl(boy)friends right now.
• I'd like to have a girl(boy)friend but it's not that important right now.
• I would really like to have a girl(boy)friend right now.


  1. 1The term phase captures movement in both directions. We use the term stage to maintain consistency with dating theories while acknowledging the limitations of this term.

  2. 2Cross-school comparisons were not computed because any significant school effect could not be separated by gender.

  3. 3Cross-school comparisons were not computed because any significant school effect could not be separated by gender.

  4. 4Initial inspection of the data indicated that only 7 adolescents in the dating or dyadic relationship stage did not report preceding romantic activities. Consequently, classification was based on at least one behavior from the preceding classifications as well as the assigned one. Eleven adolescents reported atypical patterns and they were excluded from the analysis.

  5. 5Cross-school comparisons were not computed because any significant school effect could not be separated by gender.