Nice to know you: Positive emotions, self–other overlap, and complex understanding in the formation of a new relationship

Abstract

Based on Fredrickson's ((1998). What good are positive emotions? Review of General Psychology, 2, 300–319.; (2001). The role of positive emotions in positive psychology: The broaden-and-build theory of positive emotions. American Psychologist, 56, 218–226) broaden-and-build theory and Aron and Aron's ((1986). Love as expansion of the self: Understanding attraction and satisfaction. New York: Hemisphere) self-expansion theory, it was hypothesized that positive emotions broaden people's feelings of self–other overlap in the beginning of a new relationship. In a prospective study of first-year college students, we found that, after 1 week in college, positive emotions predicted increased self–other overlap with new roommates, which in turn predicted a more complex understanding of the roommate. In addition, participants who experienced a high ratio of positive to negative emotions throughout the first month of college reported a greater increase in self–other overlap and complex understanding than participants with a low positivity ratio. Implications for the role of positive emotions in the formation of new relationships are discussed.

At such moments, you realize that you and the other are, in fact, one. It's a big realization. Survival is the second law of life. The first is that we are all one.

Joseph Campbell

Keywords:

• Positive emotions
• relationships
• self–other overlap
• broaden-and-build

Acknowledgements

This research was supported by the National Institute of Mental Health (MH059615) and an award from the American Psychological Association and the John Templeton Foundation (2000 Templeton Positive Psychology Prize) to Barb Fredrickson. We would like to thank Phoebe Ellsworth, Kareem Johnson, Michael Cohn, and the rest of the University of Michigan Positive Emotions and Psychophysiology lab for comments on earlier versions of this article.

Notes

Notes

1. We note that this study was not designed to test the accuracy of the participant's judgments of their roommate. Indeed, previous studies have shown that people in positive relationships actually have a positive bias when judging their close other (Murray & Holmes, 1997). We cannot test for accuracy because we did not gather information from the roommates themselves to compare to the participants’ judgments. So our concept of complex understanding is not the degree of accuracy when judging traits, but rather the degree to which the participant employs an attributional strategy in judging the roommate that would most likely mimic the roommate's own attributional strategy in judging themselves (Aron et al., 1991).

2. The original study also contained an experimental manipulation. Between Times 1 and 2, participants logged on to a secure website every evening for 28 days and were prompted to describe an event from the past day (following the positive emotion measures we included in our analyses). Participants were randomly assigned at the beginning of the study to either find positive meaning in each day's events, or just describe their events with no specific meaning attached. There were no significant interactions between this experimental manipulation and any of our dependent measures, so these experimental manipulations are not discussed further.

3. We also asked the participants if they had previously known their roommate before coming to college. Twenty-two participants did state that they had known their roommate before college. We included these participants in these analyses to retain power. These participants did differ from the participants who did not previously know their roommate on self–other overlap at Time 1, t = 6.8, p < 0.01, and at Time 2, t = 2.45, p < 0.05, however, excluding these participants did not change any of the reported analyses.

4. We did not find significant gender differences on any of our main dependent variables, and adding gender to the regression equations did not change the significance of any of the beta weights. Therefore, the reported analyses collapse across gender.

5. The positive and negative subscales were created using a principle component analysis on the emotion reports at Time 1 (DES). Surprise and sympathy are not usually included in subscales of positive emotions; however, these items clustered together statistically with the positive subscale in this sample, so we included them in this positive emotions subscales for Times 1 and 2. However, the principal component analysis for the daily emotion reports revealed that sexuality, and not suprise clustered with the other positive emotion subscale items. Therefore, the positive emotion subscales for daily reports include sexuality and not surprise.

6. The positive to negative ratios are slightly skewed towards positivity because each emotion report contained 11 positive items and only 8 negative items. The skew of this ratio does not affect the correlations; however, when we split the sample along the Losada line of 2.9, we include in the High Positivity group some people who may have been in the Low Positivity group had there equal numbers of items (i.e., if someone felt “equal” amounts of emotions during the day, then their ratio should be 1, whereas their ratio would actually be 11/8 or 1.38). Including possible Low Positivity participants in our High Positivity sample works against our hypotheses, so our tests are more conservative. In addition, if we multiply the positivity ratios by 8/11 to center the ratios around 1, we lose no significance in the statistical analyses. After splitting the groups into High and Low positivity, the sample sizes become very different. This is consistent with previous findings using the Losada line (see Fredrickson & Losada, 2005). Plus, the Losada line is mathematically derived from other samples, which excludes the possibility that we split the sample a posteriori into subsamples that would confirm our hypotheses.

7. We chose to split the sample at 0.5 SD above the mean of the index for two reasons. First, we wanted the high relationship building group to only include participants who exhibited increased self–other overlap and complex understanding at Time 2 controlling for these variables at Time 1. Therefore, this group should have standardized residuals well above 0. Second, splitting the sample at 1 SD above the mean of the index includes only 10% of the sample, so in the interest of including more people in the high relationship building group, we split the sample at 0.5 SD.