Skip to main content
Open AccessReplication

Revisiting the Romeo and Juliet Effect ()

Reexamining the Links Between Social Network Opinions and Romantic Relationship Outcomes

Published Online:https://doi.org/10.1027/1864-9335/a000181

Abstract

We conducted a replication and extension of Driscoll, Davis, and Lipetz’s (1972) classic longitudinal survey of the Romeo and Juliet effect, wherein they found that increases in parental interference were linked to increases in love and commitment. Using the original measures, 396 participants were followed over a 3–4 month period wherein they reported love, commitment, trust, and criticism for their partners as well as levels of perceived interference from friends and family. Participants also completed contemporary, validated measures of the same constructs similar to those often implemented in studies of social network opinion. Repeating the analyses employed by Driscoll and colleagues, we could not find evidence for the Romeo and Juliet effect. Rather, consistent with the social network effect (Felmlee, 2001), participants reporting higher levels of interference or lower levels of approval reported poorer relationship quality regardless of outcome measured. This effect was likewise evident in a meta-analysis.

The “Romeo and Juliet effect” was coined by Driscoll, Davis, and Lipetz (1972) when they discovered that couples who reported an increase in parental interference in their romantic relationship also evidenced an increase in love over the same 6-month period. Since the original study, replications of the effect have been elusive. Few studies find support for anything approximating the effect (Felmlee, 2001; Parks, Stan, & Eggert, 1983; Sprecher, 2011). Instead, most find the “social network effect” (Felmlee, 2001) whereby disapproval from one’s social network – whether family or friends – leads to declines in romantic relationship quality (see Allan, 2006; Parks, 2007).

Nonetheless, much like the story, the lore of the Romeo and Juliet effect persists, and the finding continues to be cited in popular culture, blogs, textbooks, and articles (e.g., DeWall, Maner, Deckman, & Rouby, 2011; Fisher, 2004; Miller, 2011). Reis (2011), in his review of the history of relationships research, touted the effect as “ever-popular” (p. 219). A limitation of the counterevidence for the Romeo and Juliet effect is the fact that none of the follow-up studies used the original scales.

One essential difference between the original study and subsequent follow ups has been the operationalization of social network opinions. Driscoll and colleagues (1972) focused on whether couple members had communicated to one another that they felt the other’s parents were interfering in their relationship. In contrast, with few exceptions (e.g., Johnson & Milardo, 1984), the majority of subsequent studies examined network approval on a single continuum, such that if a party scored low on approval indices then they were considered disapproving of the relationship. Although interference is a behavioral manifestation of disapproval – a particularly active, direct form of disapproval – it may not be fair to compare studies measuring perceived approval, or the lack thereof, as equivalent to studies assessing interference.

Further, in the original study, both members of the couple completed interference scales about each respective set of parents only. Subsequent studies have assessed perceived network opinions from various sources, including perceived global assessments of network opinions (combining all parents, friends, society, etc.), asking about one source (i.e., parents, friends, or siblings), or asking separate questions for multiple sources. Assessing multiple network sources may have important implications for the Romeo and Juliet effect. Felmlee (2001) found that parental disapproval led to a decreased likelihood of break up but only when friends approved. In fact, some research suggests that friend opinions may carry more weight than that of parents (Etcheverry & Agnew, 2004), at least in Western cultures (MacDonald & Jessica, 2006). Likewise, other research has suggested that approval from other sources within the network can serve as a buffer to the presence of disapproving or interfering sources (Wright & Sinclair, 2012). Thus, the Romeo and Juliet effect, if it exists, might be less about the disapproval of their family and more about the approval of their allies. Even Romeo and Juliet had the friar and the nanny.

The current study addresses the gaps between the Driscoll and colleagues (1972) study and subsequent studies. First, we assess both parent and friend opinion. We also included the scales from Driscoll et al., which have not been used since, and administered more recent, validated measures for the same constructs. Including both original and contemporary measures allows us to address whether failures to replicate the effect were due to measurement differences.

The Romeo and Juliet hypothesis asserts that interference is linked to greater love and commitment for one’s partner. Therefore, using the original scales, if the effect exists we would expect to find this same relationship in a contemporary sample. It could be the case that this effect is in fact limited to the effects of increases in interference from parents. In which case it is not exactly competing with the social network hypothesis, but rather occurs in certain circumstances. Whereas when it comes to the contemporary assessment of network approval, we anticipate replication of the social network effect such that the perceived approval from friends and family (and increases in that approval) is linked to greater love and commitment.

Method

Participants

We recruited participants from Amazon’s Mechanical Turk (mTurk) crowdsourcing database. As many mTurk workers did not meet the basic criteria of being in a romantic relationship, we ended up with 1,602 participants attempting the screening survey. Due to time limitations, we were only able to keep the survey open for a month, at the end of that month we had 976 who met the study criteria of (1) were in a relationship lasting more than 6 weeks, (2) were not dating another person, and (3) their friends, partner’s friends, parents, and partner’s parents were currently aware of their relationship, and (4) they completed the survey responsibly. Additional eliminations for invalid email addresses, non-consent for recontact, or closed mTurk accounts resulted in 718 eligible participants completing wave one in July/August who were then contacted in November/December for wave two. Of the 458 participants who returned, 396 provided usable data. Participants were paid $1 for the wave 1 and $2 for the wave 2 (and were entered into a gift card drawing).

Married participants made up 48.5% of the sample, the remainder were in various stages of dating. Of participants returning for wave 2, only 19 were broken up. Approximately half (50.8%) of the relationships had a duration of 4 years or less. Ages ranged from 18 to 70, with an average of 31.58 (SD = 9.78). The sample was predominantly female (70.7%) and Caucasian (76.3%, 5.1% Asian or Pacific Islander, 0.5% Hispanic or Latino, 4.5% African-American, 0.5% American Indian, 12.9% Multiracial, or multiethnic). There were some demographic (e.g., age) differences between those who returned and those who did not, as well as some differences on variables of interest. However, all differences between wave 1 and 2 samples had very small effect sizes. We report all data exclusions, sample differences, measures, and how we determined our sample sizes in supplemental materials. All materials, data, and the preregistered design are available at: https://osf.io/6wxgf/.

Materials

Driscoll and Colleagues Measures

Participants first completed all of the original scales included in the Driscoll and colleagues study. Item order was randomized within scale. Unless otherwise noted, participants responded on a Likert scale of 1 (= not at all) to 6 (= extremely).

Social Network Interference

Six items assessed interference for each social network source (own friends, own parents, partner’s parents, partner’s friends). For example, participants were asked “How often has your romantic partner communicated to you that your parents are a bad influence?” They were asked the same questions about their communication with their partner “How often have you communicated to your romantic partner that his/her friends interfere?” Participants responded on a 5-point Likert scale where 1 = not at all and 5 = all the time. As in the original study, scores for both sets of parents were combined to an overall index of parental interference (α = .90 for both waves). Asking the questions about friends was an extension of the previous study, but again both sets were combined and the 12 items had a reliability of α = .91 in both waves.

Love

Driscoll and colleagues used four items to assess love (e.g., “how much do you love your partner”). Reliability was α = .84 in wave 1 and .89 in wave 2.

Commitment

A single item was originally used to assess commitment, “How committed are you to your marriage (or to marrying your current partner)?”

Additional Measures

Driscoll and colleagues also included five items to assess trust of the partner (e.g., “how dependable is your partner”) and six items to assess “criticalness” (e.g., “how critical are you of your partner”). Reliability was .89 in wave 1 and .92 in wave 2 for trust. Reliability was .75 for criticism at both waves.

Contemporary Measures

After finishing all of the original scales, participants completed the contemporary survey. Item order on each index was randomized. All relationship quality indices were responded to on a 9-point Likert scale with 1 = not at all true and 9 = definitely.

Social Network Opinion Scale

Eight items were compiled from an array of studies examining social network opinions. Four items assessed approval and four reverse-scored items assessed disapproval were administered for each of the four sources (i.e., friends, parents, partner’s parents, partner’s friends) and used the same 5-point Likert response format as the original interference scale. Parents and partner’s parents items were combined (α = .91 in wave 1 and .93 in wave 2) as were friends and partner’s friends items (α = .93 in wave 1 and .95 in wave 2).

Love

The 15-item Hatfield and Sprecher (1986) Passionate Love Scale was administered. Reliability was α = .89 in wave 1 and .92 in wave 2.

Commitment

We used the 10-item (five reversed) Lund (1985) Commitment scale. Reliability was α = .91 in wave 1 and .92 in wave 2.

Additional Measures

The 2-item Perceived Criticism Measure (Hooley & Teasdale, 1989) was already part of Driscoll and colleagues’ criticism measure, so we kept Driscoll et al.’s scale. However, for trust, the 17-item (four reversed) Rempel et al. (1985) scale was used. Reliability was α = .94 in wave 1 and .95 in wave 2.

Scores were averaged across all measures such that higher scores indicate higher levels of the respective factor. Also, when difference scores were computed, they were computed such that higher scores indicated an increase in that component.

Known Differences Between Current and Original Study

Unlike the original study, the present study is not a part of an on-going longitudinal marital intervention initiative with couples.1 Thus, the primary difference between the two studies is the sample. The online administration format and timeframe also differed in our study. Due to time constraints, we cut the follow-up window in half from what originally was used based on the recommendations of the project reviewers.

Results

Driscoll and colleagues conducted correlational analyses between the variables at Time 1 and Time 2 plus correlations between difference scores in parental interference and differences in each the relationship quality indices. Following Driscoll et al., correlations were computed for the total sample as well as for dating and married samples separately (Table 1 ). Table 2 includes the correlations between the difference scores. Note, Time 2 parental and friend interference were correlated at .51 overall, .54 for daters, and .53 for those married, all p’s < .001.

Table 1. Correlations between original social network interference measures and relationship quality indices across Time 1 and Time 2
Table 2. Correlations between difference scores in interference and relationship quality using original measures

Contrary to the Romeo and Juliet effect, higher interference was consistently linked with lower relationship quality (e.g., lower love, trust, commitment; higher criticism). Increases in parental interference were not related to increases in love and commitment over time. However, as was found by Driscoll and colleagues, increases in interference were linked to reductions in trust and increases in criticism among daters.

We repeated these analyses for the contemporary measures. The results are presented in Tables 3 and 4 . Note, Time 2 parental and friend approval were correlated at .69 overall, .73 for daters, and .64 for those married, p < .001. Across analyses, measures, samples, and sources, higher levels of social network approval were linked to higher relationship quality. Further increases in approval, particularly friend approval, were linked to increases in relationship quality.

Table 3. Correlations between social network opinion measures and contemporary relationship quality indices across Time 1 and Time 2
Table 4. Correlations between difference scores in approval and relationship quality using contemporary measures

Measures of parental interference negatively correlated with measures of parental approval at −.46 to −.57 across the different administrations. Correlations between measures of friend interference and friend approval ranged from −.52 to −.78.

Meta-Analysis

To examine consistency of effects, a meta-analysis of 22 studies, including the present data, was conducted to assess how peer and family networks correlate with romantic relationship outcomes (see Appendix). To be included, studies needed to employ assessments of social network opinions and commitment or love. We searched PsycINFO and Scopus using relevant terms and used citation searches to find articles that referenced the Driscoll et al. study or other highly cited social network studies (e.g., Felmlee, 2001; Sprecher & Felmlee, 1992) in July and August of 2013. Reference lists from published review papers and meta-analyses were searched for eligible studies. We also reviewed archived abstracts from past SPSP conference programs and submitted a call for papers on various listservs for unpublished data and manuscripts under review or in press.

The 22 eligible studies included 17 published articles, one dissertation, data from three unpublished datasets, and data described in the current article. For all studies, we calculated the Fisher’s Z statistic which was then converted to the Hedges’ g effect sizes for measures of relationship commitment and love. For studies with more than one outcome measure for a given construct, we calculated the effect size for each measure first, then calculated an average Hedges’ g effect for each study outcome. All meta-analyses were conducted using Comprehensive Meta-Analysis (v.2) software developed by Biostat (Borenstein et al., 2005).

We examined the extent to which network type: friend, family, or combined social network were associated with love (Table 5 , Figure 1 ) and commitment (Table 6 , Figure 2 ) by calculating a weighted mean effect size for each network type. Random effects models and the Q statistic are reported for each network type. Eleven studies provided data on the overall effect of network approval on love. Network approval was moderately and positively associated with love 0.49 (p < .001, 95% confidence interval [CI] = 0.26–0.72). Correspondingly, friend approval 0.40 (p < .05, 95% CI = 0.09–0.71), family approval 0.32 (p < .01, 95% CI = 0.10–0.55), and combined network approval 1.02 (p < .01, 95% CI = 0.39–1.65) were positively associated with love. Those who reported approval from their social networks reported higher ratings of love for their romantic partner.

Table 5. Random effects models of network type as a predictor of romantic love
Figure 1. Forest plot of effect sizes for romantic love. Each row represents a network type, with the square the reported effect size (Hedges’ g) and the bar representing a 95% CI of the effect size. The diamond represents the overall estimated effect size and the distribution of plausible effects sizes for a 95% CI (width of diamond).
Table 6. Random effects models of network type as a predictor of relationship commitment
Figure 2. Forest plot of effect sizes for commitment. Each row represents a network type, with the square representing the reported effect size (Hedges’ g) and the bar representing a 95% CI of the effect size. The diamond represents the overall estimated effect size and the distribution of plausible effects sizes for a 95% CI (width of diamond).

Similar results were found with commitment. Network approval was moderately and positively associated with commitment 0.62 (p < .001, 95% CI = 0.50–0.74). Likewise, friend approval 0.70 (p < .001, 95% CI = 0.54–0.86), family approval 0.56 (p < .001, 95% CI = 0.39–0.84), and combined network approval 0.63 (p < .001, 95% CI = 0.43–0.84) were all positively associated with commitment. Those who reported approval from their social networks reported higher levels of commitment.

Discussion

Using the scales employed by Driscoll and colleagues as well as contemporary measures we found no evidence for the Romeo and Juliet effect. Instead, with both the original and contemporary measures we found consistent support for the social network effect such that the greater the approval (and lower the interference or disapproval) the better the relationship fared. Likewise perceived increases in social network support corresponded to increases in a number of positive aspects of the relationship (e.g., love, commitment, trust) and decreases in criticism. This finding was observed in our replications of the original study and in a meta-analysis of the accumulated literature.

Limitations

The present study was not a perfect replication of the original study. As noted above, and documented in the preregistered proposal, there were differences in sample and administration. Future studies may wish to specifically recruit couples, particularly those experiencing conflict, from the community for an in-person survey over a 6–8 month timeframe. Additional samples may also be considered. Although evidence seems to favor that the majority of romantic relationships are harmed by interference, it may be possible to find couples who thrive despite network disapproval. Studies could investigate, then, how these couples differ from those who do not endure. Also, future research employing online samples may want to investigate additional tactics for increasing return response rates, especially if intent on following participants for a longer duration.

Implications

Even taking the limitations into account, the current study makes a number of contributions. First, by employing the original measures we were better able to draw comparisons to the Driscoll et al. study. Second, we were able to show that although the more infamous findings of the original study did not recur – interference did not correspond to increased love or commitment – we did replicate the two lesser known findings that interference is linked with poorer outcomes on trust and criticism. Third, we were able to demonstrate that there is overlap between measures of interference and network approval, which supports making comparisons between the original study and the subsequent studies, albeit with caution. The measures were not identical and further work should parse the dimensions of network opinion (e.g., approving vs. disapproving, active vs. passive). Fourth, by using diverse relationship quality indices, examining multiple network sources, and employing meta-analysis, we were able show the remarkable consistency of the social network effect – even on the original scales employed by Driscoll and colleagues.

Conclusion

When romanticizing the story of Romeo and Juliet we tend to overlook the fact that, in the end, even Romeo and Juliet confirmed the social network effect. Although they tried to stay together despite network disapproval, their relationship ultimately ended. Though few have a relationship end so dramatically, the accumulated evidence makes clear that relationship love and commitment is threatened, not strengthened, by the lack of support of others.

Note From the Editors

A commentary and a rejoinder on this paper are available (Driscoll, 2014; Wright, Sinclair, & Hood, 2014; doi: 10.1027/1864-9335/a000203).

1Although we have individual instead of dyadic data, it is important to note that the original authors found no differences by couple member regarding the existence of the effect. In fact, the couple scores were simply combined into a single index of parental interference based on this rationale (and evidence of intracouple homogeneity on indices). Accordingly, as the dyadic nature of the data was not integral to the effect being found initially we do not think the lack of dyadic data threatens the comparison.

References

Papers Included in the Meta-Analysis

Blair, K. L., & Holmberg, D. (2008). Perceived social network support and well-being in same-sex versus mixed-sex romantic relationships. Journal of Social and Personal Relationships, 25, 769–791.

Blaney, A., & Sinclair, H. C. (2013). Who are Romeo and Juliet? Identifying individual differences among those who exhibit the Romeo & Juliet effect. Poster presented at the 2013 Society for Personality and Social Psychology conference in New Orleans, LA. Unpublished dataset.

Bryan, L., Fitzpatrick, J., Crawford, D., & Fischer, J. (2001). The role of network support and interference in women’s perception of romantic, friend, and parental relationships. Sex Roles, 45, 481–499.

Bryant, C. M. (1996). Subcultural variations in the social network support of Latinos, African Americans, and Anglos: What is the association between the development of heterosexual relationships and the support of friends and family members? (Doctoral dissertation, University of Texas at Austin).

Bryant, C. M., & Conger, R. D. (1999). Marital success and domains of social support in long-term relationships: Does the influence of network members ever end? Journal of Marriage and the Family, 61, 437–450.

Driscoll, R., Davis, K. E., & Lipetz, M. E. (1972). Parental interference and romantic love: The Romeo & Juliet effect. Journal of Personality and Social Psychology, 24, 1–10.

Etcheverry, P. E., & Agnew, C. R. (2004). Subjective norms and the prediction of romantic relationship state and fate. Personal Relationships, 11, 409–428.

Etcheverry, P. E., Le, B., & Charania, M. R. (2008). Perceived versus reported social referent approval and romantic relationship commitment and persistence. Personal Relationships, 15, 281–295.

Knobloch, L. K., & Donovan-Kicken, E. (2006). Perceived involvement of network members in courtships: A test of the relational turbulence model. Personal Relationships, 13, 281–302.

Le, B., Buniva, E. J., Kozakewich, E., & Etcheverry, P. E. (2008). Social network relationship approval and commitment to long-distance relationships. Unpublished manuscript.

Lehmiller, J. J., & Agnew, C. R. (2006). Marginalized relationships: The impact of social disapproval on romantic relationship commitment. Personality and Social Psychology Bulletin, 32, 40–51.

Lehmiller, J. J., & Agnew, C. R. (2007). Perceived marginalization and the prediction of romantic relationship stability. Journal of Marriage and Family, 69, 1036–1049.

Lewis, R. A. (1973). Social reaction and the formation of dyads: An interactionist approach to mate selection. Sociometry, 26, 409–418.

Mitchell, H., Sinclair, H. C., & Barkley, T. (2005). The Romeo & Juliet effect revisited: Re-examining the influence of familial versus peer (dis)approval on romantic relationship quality. Poster presented at the 2005 Society for Personality and Social Psychology conference, New Orleans, LA. Unpublished dataset.

Parks, M. R., Stan, C. M., & Eggert, L. L. (1983). Romantic involvement and social network involvement. Social Psychology Quarterly, 46, 116–131.

Parks, M. R., & Eggert, L. L. (1991). The role of social context in the dynamics of personal relationships. In W. H. Jones & D. Perlman (Eds.), Advances in personal relationships: A research annual, Vol. 2. Advances in personal relationships (pp. 1–34). Oxford, UK: Jessica Kingsley.

Sinclair, H. C. (2007). Investigating effects of social opinion on romantic relationships, Paper presented at the 2007 Southeastern Psychological Association conference in New Orleans, LA. Unpublished dataset.

Sprecher, S. (1988). Investment model, equity, and social support determinants of relationship commitment. Social Psychology Quarterly, 51, 318–328.

Sprecher, S., & Felmlee, D. (1992). The influence of parents and friends on the quality and stability of romantic relationships: A three-wave longitudinal investigation. Journal of Marriage and the Family, 54, 888-900.

Zak, A., Coulter, C., Giglio, S., Hall, J., Sanford, S., & Pellowski, N. (2002). Do his friends and family like me? Predictors of infidelity in intimate relationships. North American Journal of Psychology, 4, 287–290.

Zhang, S., & Kline, S. L. (2009). Can I make my own decision? A cross-cultural study of perceived social network influence in mate selection. Journal of Cross-Cultural Psychology, 40, 3–23.

The authors would like to thank Sining Wu and Alison Patev for their assistance on the project, as well as the Advanced Social Psychology lab. We greatly appreciated the reviews received from Richard Driscoll, Susan Sprecher, Diane Felmlee, and anonymous reviewers on various stages of the project. Designed research: H. C. S., K. B. H., B. L. W.; Analyzed data: H. C. S., K. B. H., B. L. W.; Wrote paper: H. C. S., K. B. H., B. L. W. The authors declare no conflict-of-interest with the content of this article. This project received funding from the Center for Open Science (see osf.io/6wxgf/), Psi Chi, and the Mississippi State University Office for Research and Economic Development.

H. Colleen Sinclair, Department of Psychology, Mississippi State University, Social Science Research Center, 255 Lee Blvd., Mississippi State, MS 39762, USA, +1 662 325-4177,