Romantic (versus other) events and momentary affect: Immediate and lagged within-person associations among college students

Abstract

Given that affect is highly responsive to experiences representing current goals and values, and young adulthood reflects a period in which romantic relationships become increasingly important, this study explored the links between everyday romantic relationship events and momentary affect among young adult college students. Romantic events were then directly compared to academic and family events—two other salient life domains for these students—as predictors of current and subsequent momentary affect. Drawn from an ecological momentary sampling study designed to assess substance use, participants in dating relationships (N=130) completed four reports per day for 28 days (totaling 10,318 reports). Multilevel models tested within-person associations between positive and negative romantic events (broadly defined) as predictors of positive (e.g., happy, excited) and negative (e.g., sad, lonely) affect in the moment and beyond. Analyses included both event occurrence and event intensity models, facilitating event comparison. Models accounted for day-level effects and several relevant individual and relationship controls. Results indicated that positive romantic events were associated with immediate and lasting increases in positive affect and immediate (but not always lasting) decreases in negative affect, whereas negative romantic events were associated with immediate and lasting changes in both positive and negative affect. When significant, direct comparisons indicated that romantic events were associated with larger changes in concurrent and subsequent affect than academic or family events. Findings highlight the powerful role that young adults’ romantic relationships play in their emotional well-being, particularly in comparison to other developmentally and environmentally salient life domains.

Our intimate relationships are often the contexts in which we experience our most intense emotions (Berscheid & Ammazzalorso, 2003), and emotional experiences lay the foundation for long-term well-being. Corresponding to the emotional components of well-being (Diener, 2000), the moment-to-moment experiences of positive and negative affect have important implications for long-term health and happiness, both directly and indirectly. More frequent and more intense negative affect and affective fluctuation have been associated with depression and other psychopathology in the general population (Houben et al., 2015) and among young adults (Chaplin, 2006). Negative mood states are also known to increase cortisol reactivity (Jacobs et al., 2007) as well as the likelihood of maladaptive coping behaviors such as heightened substance use (Bresin et al., 2018; Testa et al., 2019). Conversely, experiencing positive affect is associated with numerous markers of health and success in longitudinal and experimental studies (Lyubomirsky et al., 2005). Emotion dynamics also play critical roles in the development and maintenance of romantic, and other interpersonal, relationships (Schoebi & Randall, 2015). Negative emotional states can make problem-solving and perspective-taking more difficult, decreasing the ability of couples to effectively manage conflict, or can even promote connection by signaling the need for social support or areas of improvement (Fischer & Manstead, 2008; Fredrickson, 2001). Therefore, understanding how romantic experiences are closely associated with affect throughout daily life, and how these associations compare to events representing other life domains, is important for understanding personal and relational health and well-being.

The emotional experiences of young adults attending college, in particular, present a pressing concern, given the documented increases in occurrence and severity of mental health issues among this population (Oswalt et al., 2020). Notably, developmental theories suggest that young adulthood represents a developmental period in which romantic relationships become of increasing importance compared to adolescence (Arnett, 2006), making the everyday interactions with a romantic partner increasingly influential in young adult college students’ emotional well-being throughout daily life. Despite ample evidence of the links between romantic relationship experiences and mental health among young adults (Mirsu-Paun & Oliver, 2017), how relationship experiences correspond to affective fluctuations in the moment-to-moment lived experiences of young adults’ daily lives has not been fully examined.

Positive and Negative Affective Experiences

The documented contrasting links between positive versus negative affective experiences and subsequent health can be explained by theories postulating that the two serve distinct functions and operate within distinct systems (Fischer & Manstead, 2008; Fredrickson, 2001), and therefore have distinct implications for romantic partners. Positive emotions such as happiness and joy are thought to encourage environmental and social engagement (Frederickson, 2001) and are linked to more effective problem-solving. Negative emotions such as sadness or anger are thought to encourage narrow mindsets (Fredrickson, 2001), yet sadness in particular may also serve prosocial functions through encouraging support from others when expressed (Fischer & Manstead, 2008) and facilitating relatively more effective forms of communication during relationship conflict (versus anger; Sanford, 2007). Furthermore, because these affective states operate somewhat separately, their connections to various daily experiences may differ. Indeed, theory and evidence indicate stronger links between positive events and positive affect and between negative events and negative affect (Altermatt, 2015; Flook, 2011).

Well-supported theories of emotion state that both positive and negative affective states are highly responsive to experiences related to an individual’s current goals and values (Frijda, 1986). Events more closely aligned with life domains that the individual deems are important and immediately relevant (e.g., relationships, academics, or family) will be more closely connected to that individual’s affect, especially in the moments those events are experienced. Such events would also have a longer lasting effect on affect, with lingering affective changes following the event. This may be particularly true for close social connections such as intimate relationships. Theory indicates that, because intimate relationships fulfill the basic human needs of belonging and closeness (Baumeister & Leary, 1995), it follows that positive experiences in these relationships would signal these needs being met, increasing happiness and well-being, and that negative experiences may signal these needs are in threat of not being met, decreasing happiness and well-being (Schoebi & Randall, 2015). For instance, arguments that involve an intimate partner, representing the romantic relationship domain, are likely of greater importance than events involving an acquaintance, and therefore would be expected to carry greater emotional weight, corresponding to increases in sadness and decreases in happiness. If the event is pleasant or positive, such as having a fun date or a particularly validating conversation with a romantic partner, positive affect will likely increase and negative affect will likely decrease.

Momentary Affect and Romantic Relationship Events

Positive and negative affect can fluctuate widely in response to various events experienced throughout daily life, particularly when those events involve a romantic partner (Berscheid & Ammazzalorso, 2003). Evidence for such dynamic associations between romantic relationship events and affect among partnered individuals have been previously documented on several scales. For example, in a 10-year longitudinal study assessing young adult couples’ conflict interactions in a laboratory, women’s depressive symptoms were higher at time points when the couple displayed relatively more negative conflict behaviors (i.e., psychological aggression and withdrawal) and lower at time points when the couple displayed relatively more positive conflict behaviors (i.e., engagement) (Laurent et al., 2009). On a daily level, diary studies have indicated that negative affect, such as sadness, distress, and anxiety, is higher on days when participants experience relatively more negative romantic relationship events (e.g., conflict with partner, stressful interactions), and this has been found among adolescents (Rogers et al., 2018) and married adults (Bolger & Schilling, 1991). Similarly, positive sexual interactions (i.e., pleasurable and intimate) with a romantic partner have been associated with higher positive affect and lower negative affect the following day (Kashdan et al., 2018), also indicating lingering changes in affect following specific romantic events.

Such results indicate the long-term and daily within-person connections between everyday romantic relationship events and fluctuations in positive and negative affect. However, people sometimes experience great emotional variability, and sometimes numerous affect-provoking events, throughout a single day (Stone et al., 1996). Therefore, even daily assessments could still mask critical moment-to-moment associations between relationship events and affect. Momentary-level data collection, or ecological momentary assessment (EMA) designed to capture participants’ lived experiences (Shiffman et al., 2008), is critical for understanding which factors are most closely related to positive and negative affect in any given moment (Diener, 2000). Studies have used such within-person momentary designs to assess links between affect and minor daily stressors broadly (e.g., Jacobs et al., 2007), or affect during couple interactions in a laboratory (Laurent et al., 2009). One relatively small EMA study of 44 participants was designed to assess emotional inertia in response to romantic conflict versus intimacy collected reports 4 times per day for 4 weeks (Luginbuehl & Schoebi, 2019). The results regarding emotional reactivity indicated that when participants reported experiencing an intimate moment since the previous diary, cheerfulness was higher and negative affect, including sadness, was lower compared to the previous diary; when conflict was reported, cheerfulness was lower and negative affect was higher compared to the previous diary. This study also indicates changes in affect following specific romantic events, but was not able to capture affect in the moments the events were occurring, leaving unknown the full extent of affective fluctuation surrounding these specific romantic events. Thus, more studies are needed to fully understand how everyday romantic events are linked to affect as these events are being experienced and in the ecologically valid context of everyday life. In the current study, we draw from an EMA study, which collected multiple momentary reports throughout the day across several (~28) days, to assess the within-person associations between romantic events and affect on a momentary basis, both in the moments the events are experienced and in the moments that follow.

Importantly, there are numerous experiences one can have with a romantic partner, both positive and negative throughout a day. Most studies exploring everyday romantic events in partnered individuals’ well-being have focused on specific events or behaviors, such as arguments with a spouse (Bolger & Schilling, 1991) or withdrawing during a conflict interaction (Laurent et al., 2009), and therefore do not account for the multitude of other relationship events that could have occurred throughout the day. A broader approach of assessing any event represented by the romantic relationship domain can provide a more general view of how romantic relationship experiences are linked to affect throughout daily life.

Romantic versus Other Events in Young Adults’ Lives

Given that affective states are highly responsive to experiences related to individuals’ current goals and values (Frijda, 1986), the multifaceted nature of young adults’ lives when attending college indicates that their goals and values are likely tied to several different life domains. Thus, affective fluctuations likely correspond to many types of daily events that represent these many life domains. While romantic relationships are highly salient components of partnered young adults’ lives (Arnett, 2006), it remains unclear how their associations with everyday well-being compare to two other critical domains: academics and family. For college-based young adults, daily events related to their schooling (e.g., failing or passing an exam) are expected to represent a life domain of high importance and salience, and therefore be closely linked to their affective states. Academic factors in scholarly environments have received much attention as analytical outcomes of emotional and psychological adjustment in students (e.g., Bruffaerts et al., 2018), but less attention has been paid to academic events as analytical predictors of affect or within-person affective fluctuations. One study did find that academic failures and successes have been linked to daily changes in elementary school children’s affect (Altermatt, 2015), indicating that academic events are linked to affect in daily life. Yet, it remains unclear how these findings translate to more emotionally developed and independent college students. Similarly, family also represents a life domain characterized by interpersonal relationships, and therefore, similar to romantic relationships, also represents the potential to fulfill the needs of belonging and closeness (Baumeister & Leary, 1995). Family contexts indeed have great influence over health and well-being, and continue to be influential through the transition from adolescence to young adulthood (Tsai et al., 2013). For instance, a longitudinal study of Dutch students found that closer parental bonds predicted greater general well-being from adolescence through young adulthood (Van Wel et al., 2002). However, within-person accounts of how academic- and family-related daily events are linked to momentary affect for those attending college—and therefore those who may have become more independent from their families—have not been explicitly nor simultaneously examined.

As real-time predictors of momentary affect, understanding the relative importance of everyday events across different life domains can provide information that is directly applicable to addressing college students’ mental health and well-being. Some evidence indicates romantic relationships may be stronger predictors of long-term health and well-being than other life domains, including work and family (Glenn & Weaver, 1981, Vaillant, 2003), while other evidence indicates well-being is predicted comparably well by relationship status and closeness of parental bonds (Van Wel et al., 2002). However, direct comparisons are rare. Thus, a second aim was to explore whether (positive and negative) academic and family events are also associated with concurrent and subsequent momentary affect, and then to directly compare them to romantic events as predictors of affect.

An Alternate Model: Event Intensity

Previous studies have provided clear evidence that the occurrence of a salient event has important implications for immediate and long-term affective well-being. The strength of affective responses to events may also depend on the intensity of that event. Events can vary in the degree to which they are experienced as positive or negative, with event intensity (e.g., highly negative versus mildly negative) being comparably or more closely tied to emotional outcomes than mere frequencies of events (Lazarus & Folkman, 1984; Nezlek et al., 2008). Studies on stressors have documented the utility of measuring the intensity of these events when assessing associations with health and well-being outcomes (e.g., Sarafino & Ewing, 1999); when measuring uplifting events, while both frequency and intensity were related to positive affect, the intensity appraisal was more strongly related (Maybery et al., 2006).

Notably, when using more general assessments of events across repeated reports, the intensity of the events captured could vary greatly, both within and between life domains. For instance, negative academic events may be more commonly experienced as less intense than negative romantic events, or some romantic events may be more intense than others. A partner’s infidelity may be more saddening than a small disagreement over dinner; exchanging initial “I love you”s may be more joy-inducing than acing a practice quiz in class. Accounting for event intensity can thus serve as a sort of equalizer when assessing what role events play in affective response, particularly when comparing events across life domains using repeated measures. Thus, conducting additional tests with event intensities can capture these important nuances in event experiences while also facilitating comparisons across different life domains, and can thus further our understanding of relevant experiences and affect in college students’ daily lives.

The Current Study

Drawing from the EMA portion of a larger study of young adult college students’ health behaviors including substance use, the current study assesses the role of romantic events in momentary affect, and how they compare to academic and family events as predictors of affect. Such associations were assessed both in the moments they are experienced (concurrent) and the moments following (subsequent), and using both event occurrence (main) and event intensity (alternate) models. Due to their direct connections to long-term outcomes and the mental health challenges facing young adult college students today, we prioritized positive affect that encompasses a wide spectrum of happiness and excitement (Lyubomirsky et al., 2005) and negative affect related most closely to sadness, depression, and loneliness (Oswalt et al., 2020). The study was guided by the following research questions and hypotheses:

RQ1: For partnered young adult college students, are positive and negative romantic events associated with changes affect in the moments they are experienced (concurrent associations)?

We predict that positive romantic events will be associated with increases in positive affect and decreases in negative affect, and that negative romantic events will be associated with decreases in positive affect and increases in negative affect (Laurent et al., 2009; Luginbuehl & Schoebi, 2019; Rogers et al., 2018).

RQ2: Are academic and family events also associated with affect, and if so, how do they compare to romantic events as predictors of concurrent momentary affect?

We predict that academic and family events will be associated with momentary affect (Frijda, 1986), but that romantic events will be associated with relatively greater fluctuations (Glenn & Weaver, 1981).

RQ3: Do changes in affect from romantic and other events carry over beyond the moments they are experienced (subsequent associations)?

We predict that these events will be associated with lingering changes in affect in the following report (Frijda, 1989).

Important covariates are also considered. Given the focus on substance use in the broader project, the models account for substance use indicators at the momentary and person level. Additionally, individuals differ substantially in their affective variability and interpersonal experiences (e.g., Eid & Diener, 1999; Flook, 2011). Neuroticism has consistently been associated with more frequent negative interpersonal experiences and to higher and more frequent negative mood states (e.g., Bolger & Schiller, 1991). Our analyses account for individual variability in emotional dynamics by including neuroticism as a person-level covariate and by the multilevel structure of the analytical models. Gender, which has also been found to correspond with affectivity and interpersonal experiences (e.g., Flook, 2011), was also included as a person-level control. Lastly, relationship characteristics such as duration, and relationship quality indicators such as perceived supportiveness from a partner, can factor into how people experience and respond to their romantic partners (Blumenstock & Papp, 2019; Bradbury & Karney, 2019). Because these factors may influence how romantic events relate to affect, they were also included as person-level and cross-level controls.

Methods

Data is from the first wave of data collection from an ongoing study of college students’ daily substance use behaviors and health with a focus on prescription drug misuse (Papp et al., 2020). The University of Wisconsin – Madison Institutional Review Board approved all procedures, and the study design and original analysis plan was registered on the Open Science Framework (see Author Note for link).

Participants and Procedure

From Fall 2017 to Fall 2019, freshmen and sophomores aged 18–21 were recruited from a university in the Midwestern United States, and oversampled for those who endorsed recent (past 3 months) misuse of a prescription medication (355 participants total; 300 endorsed recent misuse). Recruitment was continuous and included mass emails, flyers across campus and the local area, and targeted ads on social media and student newsletters. Recent misuse (or not) was determined via confidential phone screen procedure (see Papp et al., 2020).

Participation included two in-person lab visits with 28 days of EMA reporting in between (three participants’ reporting periods were abbreviated due to scheduling conflicts or device issues). The first lab visit consisted of informed consent procedures, questionnaires (including relationship characteristics and neuroticism items), EMA device disbursement (iPod Touch 6^th generation), and EMA training. The second visit consisted of additional questionnaires and receiving payment (up to USD$250, including a bonus incentive of $36 for timely EMA report completion). The current sample includes only participants who reported being in a dating relationship (N =130); 84 (64.6%) identified as female and the rest identified as male; 107 (82.3%) had endorsed recent misuse at screening. The sample mostly identified as Caucasian/White (107; 82.3%), with 6 (4.6%) identifying as Asian, and the other 12.4% identifying as either American Indian or Alaskan Native, Black or African American, Native Hawaiian or Pacific Islander, Other, or multiple races.

To capture a reasonably representative sample of the participants’ lived experiences throughout their daily lives (Shiffman et al., 2008), the EMA portion of the study included four time signals each day to prompt report completion. Device alarms were programmed to randomly signal within four time periods (morning = 8:00 AM to 11:30 AM; afternoon = 11:30 AM to 3:00 PM; evening = 3:00 PM to 7:00 PM; and night = 7:00 PM to 11:00 PM). Participants were also instructed to self-initiate an EMA report if they were intending to misuse a prescription drug. Thus, participants could complete a report at any time and reports were not connected to alarm signals. Whether prompt- or self-initiated, participants submitted a report by entering the app and selecting the option to begin a report. This was to ensure the reporting process was identical for each EMA report.

The app was designed specifically for this study by the university’s technology department in collaboration with the study team. All other functionality of the device beyond the app was made inaccessible to the participant (e.g., internet, music). The EMA reports were brief (~2 minutes on average) and included questions about participants’ location and others with them, substance use and intentions to misuse a prescription drug (sedative, tranquilizer, stimulant, or pain medication), mood, and events they may be experiencing at that moment. If participants endorsed intentions to misuse a prescription drug, a follow-up report was sent 15 minutes later to assess actual misuse behavior; the follow-up reports are not included in the current study. For more information about the development, utility, and full-sample response rates regarding the app, see Papp et al., 2020.

Measures

Momentary Positive and Negative Affect.

In each EMA report, participants were asked to rate their current affect based on emotion words (drawn from the PANAS-X; Watson & Clark, 1999) via the prompt “Please indicate to what extent you feel this way right now, that is, at the present moment.” For each of the items, one main word was presented, followed by related words in parentheses. The item for negative affect was “Sad (blue, downhearted, alone, lonely).” The item for positive affect was “Jovial (happy, joyful, delighted, cheerful, excited, enthusiastic, lively, energetic).” Thus, these single items represent an overall assessment of negative affect using several words associated with sadness and loneliness (but not of fear, anger, etc.), and of positive affect using several words related to happiness and excitement (but not of pride, confidence, etc.). Responses ranged from 1 (very slightly or not at all) to 5 (extremely).

Positive and Negative Events: Occurrence and Intensity.

In each EMA report, participants were asked “Are you currently experiencing any of the following events,” followed by a checklist that included romantic, academic, and family events. To allow participants to interpret for themselves what constituted an event in their lives, no detailed examples were given. For each event that was selected, participants were then asked to rate their experience of that event on a scale from −3 (very unpleasant) to 3 (very pleasant), with 0 indicating neutral (Jacobs et al., 2007). Positive and negative event rating variables follow Jacobs et al., 2007 and other studies assessing hassles and uplifts by the intensity appraisals (e.g., Maybery, 2006). If an event was rated as pleasant (1 to 3), it was considered a positive event (coded 0/1); if it was rated as unpleasant (−3 to −1) it was considered a negative event (coded 0/1). If the report indicated the event did not occur, both positive and negative events were coded as 0. Neutral events (rating of 0) were not considered positive or negative events and were also coded as 0. This resulted in six moment-level predictor variables for the current study: positive and negative variables for each of the three event types (romantic, academic, and family). Each report received a code for each event type. For event intensity (alternate model), the actual ratings (1–3) were used. For negative event ratings, the absolute value was taken so higher values indicated more negativity.

Neuroticism.

The Neuroticism scale from the Big Five Inventory (John & Srivastava, 1999) included 8 items. Prompted with “I see myself as someone who…”, items included “is depressed, blue”, “can be moody”, and “remains calm in tense situations”, and were rated on a scale of 1 (Disagree strongly) to 5 (Agree strongly); items were reverse coded so higher values indicated higher neuroticism. Scale reliability was good in this sample (Cronbach’s alpha = .84).

Relationship Length and Partner Support.

Participants who indicated they were currently in a dating relationship were asked the length of that relationship (converted to months). Using items from the Partner Support and Partner Strain scales (Walen & Lachman, 2000), an overall score for partner support was computed. The four items for support, answered on a 4-point Likert-type scale (1 = a lot; 4 = not at all) asked about partner understanding, care, reliability, and openness to talking about worries. The four items for strain, answered on a similar 4-point Likert-type scale (1 = often; 4 = never), asked about frequency of partner criticism, demandingness, let-downs, and annoyance. Items from the support scale were reverse-scored. All 8 items were summed so higher values indicated higher support. Reliability was adequate for this control variable (Cronbach’s alpha = .61).

Analyses

Multilevel modeling with maximum likelihood estimation was used to account for the nested nature of the data (e.g., moments within person) (Bolger & Laurenceau, 2013). Participants sometimes reported multiple romantic relationship events on a single day, indicating the need to account for the day level. Thus, 3-level multilevel models were used. For RQ1, two models were tested, one with concurrent positive affect and the other with concurrent negative affect as the outcome variable. Both models included positive and negative romantic events as predictors. Level-1 (moment level) predictors included positive and negative romantic events as time-varying slope predictors, person-mean centered and allowed to vary randomly at the person level. To adjust for time slope and artifacts from changes over time, a linear time slope across all data points from the participant (ReportNumber) is also included in level 1, and was centered at the sample expected midpoint (i.e., 56) (Bolger & Laurenceau, 2013). Affect from the previous report was also included (PreviousAffect). Given the sample was chosen based on recent prescription misuse, a variable for misuse intentions is also included (RxMisuseInt). Here is the level 1 model for RQ1 (“Pos”‘=positive; “Neg”=negative):

$$\begin{gathered} {\text{Affect}}_{\text{tij}}={\text{π}}_{0\text{ij}}+{\text{π}}_{1\text{ij}}*\left({\text{ReportNumber}}_{\text{tij}}\right)+{\text{π}}_{2\text{ij}}*\left({\text{RxMisuseInt}}_{\text{tij}}\right)+{\text{π}}_{3\text{ij}}*\left({\text{PreviousAffect}}_{\text{tij}}\right) \\ +{\text{π}}_{4\text{ij}}*\left({\text{PosRomanticEvent}}_{\text{tij}}\right)+{\text{π}}_{5\text{ij}}*\left({\text{NegRomanticEvent}}_{\text{tij}}\right)+{\text{e}}_{\text{tij}} \end{gathered}$$ Level 1

The second level (day level) adjusted for variability in affect between days. Count variables for the number of positive and negative events within a single day were entered as level-2 intercept controls and were group-mean centered.

$$\begin{gathered} {\text{π}}_{0\text{ij}}={\text{β}}_{00\text{j}}+{\text{β}}_{01\text{j}}*\left({\text{PosRomanticEventCount}}_{\text{ij}}\right)+{\text{β}}_{02\text{j}}*\left({\text{NegRomanticEventCount}}_{\text{ij}}\right)+{\text{r}}_{0\text{ij}} \\ \begin{array}{c}{\text{π}}_{1\text{ij}}={\text{β}}_{10\text{j}}\\ {\text{π}}_{2\text{ij}}={\text{β}}_{20\text{j}}\\ {\text{π}}_{3\text{ij}}={\text{β}}_{30\text{j}}\\ {\text{π}}_{4\text{ij}}={\text{β}}_{40\text{j}}\\ {\text{π}}_{5\text{ij}}={\text{β}}_{40\text{j}}\end{array} \end{gathered}$$ Level 2

The third level (person level) accounted for nesting within person and included participants’ proportion of positive and negative romantic relationship events across the reporting period, as well as person-level controls of gender, neuroticism, partner support, relationship length, and recent prescription drug misuse at screening. Continuous variables were grand-mean centered.

$$\begin{gathered} {\text{β}}_{00\text{j}}={\text{γ}}_{000}+{\text{γ}}_{001}\left({\text{RxScreen}}_{\text{j}}\right)+{\text{γ}}_{002}\left({\text{Gender}}_{\text{j}}\right)+{\text{γ}}_{003}\left({\text{Neuroticism}}_{\text{j}}\right) \\ +{\text{γ}}_{004}\left({\text{RelationshipLength}}_{\text{j}}\right)+{\text{γ}}_{005}\left({\text{PartnerSupport}}_{\text{j}}\right) \\ +{\text{γ}}_{006}\left({\text{ProportionPosRomanticEvents}}_{\text{j}}\right)+{\text{γ}}_{007}\left({\text{ProportionNegRomanticEvents}}_{\text{j}}\right)+{\text{u}}_{00\text{j}} \\ \begin{array}{l}{\text{β}}_{01\text{j}}={\text{γ}}_{010}\\ {\text{β}}_{02\text{j}}={\text{γ}}_{020}\\ {\text{β}}_{10\text{j}}={\text{γ}}_{100}\\ {\text{β}}_{20\text{j}}={\text{γ}}_{200}\\ {\text{β}}_{30\text{j}}={\text{γ}}_{300}\\ {\text{β}}_{40\text{j}}={\text{γ}}_{400}+{\text{u}}_{40\text{j}}\\ {\text{β}}_{50\text{j}}={\text{γ}}_{500}+{\text{u}}_{50\text{j}}\end{array} \end{gathered}$$ Level 3

For RQ2, to compare romantic relationship events to academic and family events as predictors of concurrent positive and negative affect, the corresponding variables for academic and family events were added to the four models from RQ1. This included the level-1 predictor variables for academic and family events, the level-2 control variables for number of academic and family events that day, and the level-3 control variables for proportion of academic and family events for that participant throughout the reporting period. Multivariate hypothesis testing was used in the HLM program to test whether coefficients for romantic events were significantly different from the coefficients for the academic events and for the family events at p < .05.

To test the subsequent associations (RQ3), the outcomes variables were affect at the following report (t+1); affect at the current report (t) was included. The same six event variables were included as predictors. To account for any events occurring at the subsequent report, moment-level variables for all events at the following report (t+1) were also included.

Two additional models were also run. First, to fully adjust for individual difference variables in the model results, we ran cross-level moderation models with all five individual difference characteristics as slope moderators of the six event types to see if the result patterns changed. Second, given the importance of event intensity in affective outcomes, we ran alternate models using event intensities. Here, the event intensity ratings were used (versus the dummy coded variables), and the mean intensity rating for the event types was entered as the person-level control (versus proportions of the event types).

Missing Data

MLM procedures account for missing data at level 1 without removing data that is not complete (e.g., differing response rates from participants); more weight is given to participants who provide more data and more reliable estimates (Bolger & Laurenceau, 2013). As is common with EMA data, most missing data involved failing to submit an entire report (i.e., no data is available for that report); only 0.4% of the submitted reports included partial (i.e., any missing) information of the key study variables. Tests of the level 1 model indicated data were missing at random (Little’s MCAR test, p = .125). The level 2 variables were aggregated directly from the level 1 data, and all 130 partnered participants provided complete responses to the relationship questions and other key variables. Thus, there is no missing data at levels 2 and 3. To test whether the associations differed for participants who completed more versus fewer reports, we conducted a robustness check to compare the findings based on those who completed 75% or more of the reports versus those who completed fewer.

Results

The EMA completion rates for this sample were high overall. Participants reported on an average of M = 27.1 days (SD = 6.4, range 4–35), and completed an average of M = 3.01 reports on these reporting days (SD=1.1, range 1–9), for a total of M = 79.4 reports (SD = 27.7, range=10–129). The completion rate, calculated by dividing the number of submitted reports by the number of expected reports based on assigned reporting days, was M = 0.72 (SD = 0.24, range 0.09–1.15). Because participants could self-initiate reports and continue completing reports beyond their scheduled 28 days (until they returned their devices), some completion rates exceeded 1.0. Intentions to misuse a prescription drug were endorsed in 134 (< 0.02%) reports.

Descriptive results for the events (Table 1) indicated that participants reported academic events most frequently and family events least frequently. Most of the romantic and family events were rated as positive, whereas about half the academic events were rated as positive.

Table 1.

Descriptive Information for Events (Counts and Ratings) Reported, by Event Type and Measurement Level

	Positive Events		Negative Events		Total
Total number of events reported	N		N		N
Romantic	718		187		905
Academic	991		869		1860
Family	329		130		459
Total	2038		1186		3224
Number of events reported per participant	M(SD)	Range	M(SD)	Range
Romantic	5.52(8.1)	0–71	1.44(4.3)	0–40
Academic	7.62(11.5)	0–67	6.68(8.2)	0–39
Family	4.35(6.4)	0–52	1.00(3.2)	0–31
Event intensity	M(SD)	Range	M(SD)	Range
Romantic	2.44(0.7)	1–3	1.90(0.8)	1–3
Academic	1.59(0.7)	1–3	1.85(0.9)	1–3
Family	2.15(0.8)	1–3	2.02(0.8)	1–3

Note. Drawn from 10,318 momentary reports submitted by N = 130 partnered participants.

Results for RQ1 indicated that, accounting for day-level effects and relevant controls, romantic events predicted changes in concurrent momentary affect. Positive romantic events were associated with increases in positive affect (B = 0.60) and decreases in negative affect (B = −0.20); negative romantic events were associated with decreases in positive affect (B = −0.59) and increases in negative affect (B = 0.74) (all ps < .001).

Key level 1 results from both the concurrent (RQ2) and subsequent (RQ3) models are presented in Table 2; full results available in supplemental materials. Results for RQ2 indicated that academic and family events were also predictors of concurrent momentary affect. All six event types were statistically significantly associated with changes in affect in the hypothesized ways, with one exception: positive academic events were unrelated to changes in both positive and negative affect.

Table 2.

Effects of Romantic, Academic, and Family Events Predicting Concurrent (t) and Subsequent (t+1) Momentary Affect in Daily Life: Key Level 1 Results from Main and Alternate 3-Level Multilevel Models

	Concurrent Affect (t)				Subsequent Affect (t+1)
Key Level 1 Fixed Effects	Positive		Negative		Positive		Negative
	B	p	B	p	B	p	B	p
Main Models: Event Occurrence
Intercept	2.848	<0.001	1.507	<0.001	2.717	<0.001	1.480	<0.001
Report number (time)	0.001	0.082	0.001	0.219	0.001	0.115	0.001	0.220
Rx misuse intentiona	0.008	0.935	0.058	0.490	−0.042	0.577	0.159	0.050
Previous affect (control)	−0.202	<0.001	−0.194	<0.001	−0.259	<0.001	−0.212	<0.001
Positive eventb
Romantic	0.669	<0.001	−0.196	<0.001	0.137	0.014	−0.034	0.344
Academic	0.058	0.173	−0.043	0.088	0.111	0.006	−0.032	0.299
Family	0.366	<0.001	−0.098	0.014	0.135	0.007	−0.067	0.090
Negative eventb
Romantic	−0.529	<0.001	0.867	<0.001	−0.280	0.003	0.285	0.012
Academic	−0.418	<0.001	0.243	<0.001	−0.096	0.047	0.141	0.003
Family	−0.451	<0.001	0.724	<0.001	−0.204	<0.001	0.208	0.005
Alternate Models: Event Intensity
Intercept	2.823	<0.001	1.503	<0.001	3.407	<0.001	0.937	<0.001
Report number (time)	0.096	<0.001	0.015	0.078	0.001	0.123	0.001	0.222
Rx misuse intentiona	0.008	0.926	0.044	0.596	−0.068	0.368	0.150	0.066
Previous affect (control)	−0.287	<0.001	−0.271	<0.001	−0.263	<0.001	−0.213	<0.001
Positive event intensityc
Romantic	0.247	<0.001	−0.082	<0.001	0.069	0.002	−0.012	0.320
Academic	0.065	0.014	−0.026	0.076	0.070	0.007	−0.009	0.629
Family	0.190	<0.001	−0.068	<0.001	0.067	0.001	−0.042	0.014
Negative event intensityc
Romantic	−0.385	<0.001	0.525	<0.001	−0.155	<0.001	0.212	<0.001
Academic	−0.228	<0.001	0.138	<0.001	−0.058	0.034	0.089	0.003
Family	−0.238	<0.001	0.342	<0.001	−0.110	<0.001	0.111	0.003

Note. Rx = prescription. Bold font indicates significant hypothesized effects at p < .05. N = 130 partnered participants. Full model results available in supplementary materials.

^aAccounting for focal construct of broader study (coded 0/1).

^bDummy coded (0/1).

^cScale of 1–3 for pleasantness (positive) and unpleasantness (negative).

When comparing the concurrent associations of romantic events to those of other events, the coefficients for romantic events were all descriptively larger, and 5 differences were statistically significant, offering partial support for our RQ2 predictions that romantic events would be most strongly associated with momentary affect. Specifically, compared to positive academic events, positive romantic events were associated with larger increases in positive affect (χ² = 75.1, p < .001), and larger decreases in negative affect (χ² = 13.2, p < .001). Compared to negative academic events, negative romantic events were associated with larger increases in negative affect (χ² = 14.8, p < .001); no differences were found for decreases in positive affect (χ² = 0.82, p > .50). Compared to positive family events, positive romantic events were associated with larger increases in positive affect (χ² = 16.7, p < .001), and larger decreases in negative affect (χ² = 3.92, p = .045). No significant differences were found in affective changes for negative romantic versus negative family events (χ² = 0.23 and 0.45, ps > .50).

Results for RQ3, accounting for events at the subsequent time point, indicated lingering changes in affect following some but not all events. All positive events were associated with continued increased positive affect at the following report, yet none were associated with lasting changes in negative affect. All negative events were associated with increased negative affect at the following report and decreased positive affect at the following report.

When comparing the subsequent associations of romantic events to those of other events, while effects for romantic events were descriptively larger, hypothesis tests indicated no statistically significant differences in estimated affective changes when comparing romantic events to other events (ps > .07). The alternate models indicated the same results except for one: negative romantic events predicted larger increases in negative affect (χ² = 6.02, p = .014).

For the cross-level moderation models that fully adjusted for individual differences, results of the key hypothesized associations were substantively identical except that negative family events no longer predicted subsequent positive affect (p = .12). Full results available in supplementary material. The robustness checks tested whether the results differed based on EMA completion rates. For the concurrent models (RQ2), the pattern of results was the same when the models described above were conducted using only the 61 participants who did not complete at least 75% of their expected reports. For the subsequent models (RQ3), negative academic events were no longer predicters of changes in positive affect (p = .34), and negative academic (p = .59) and negative family (p = .11) events were no longer associated with changes in negative affect. We suspect this is likely due to a relatively small effect size and the reduction in power from the subsample size, rather than an indicator of differing results between completion rates.

The alternate models using event intensity ratings indicated the same overall patterns of results, with two differences—both involving event types that were not significant predictors of affect in the event occurrence models. Specifically, increases in intensity of positive academic events predicted increases in concurrent positive affect (B = .065, p = .014), and increases in intensity of positive family events predicted decreases in subsequent negative affect (B = −0.042, p = .014). See Table 2 for key level 1 results; full results can be found in Table S2.

Figure 1 shows the full-model-based estimates for concurrent and subsequent positive and negative affect (with moderators) indicating reported affect prior to the event (intercept; t-1), when each event type occurred (concurrent; t), and at the report following the event (t+1).

Figure 1.

Note. Estimated momentary affect from the full multilevel models (with cross-level moderators), by event type, at the previous report (intercept; t-1), at the report during which the event was reported (t), and at the following report (t+1). The thin, solid, horizontal bars indicate the model intercepts.

Discussion

The current study provides a moment-level account of how romantic relationship experiences correspond to well-being in the daily lives of young adult college students, and the relative importance of romantic events compared to events representing other salient life domains (i.e., academics and family). Results indicated robust associations between romantic events and changes in both concurrent and subsequent affect, based on a rigorous analysis of moment-level events and affect as they were experienced. Notably, the study included affective experiences closely related to the mental health challenges reported by college students today—depression and loneliness (Oswalt et al., 2020)—and affective experiences closely aligned with positive outcomes long-term—happiness and excitement (Lyubomirsky et al., 2005).

As expected, positive romantic events were related to increases in positive affect and decreases in negative affect in the moment, with sustained changes in positive affect (but not negative affect) found at the following report. Negative romantic events were not common (averaging just over 1 per person), yet they were powerful predictors of changes in affect, related to immediate and prolonged decreases in positive affect as well as increases in negative affect. This aligns with previous research indicating that positive relationship events such as intimate moments or sexual experiences are associated with higher positive affect (Kashdan et al., 2018; Luginbuehl & Schoebi, 2019), and that negative relationship events such as conflict or maladaptive conflict behaviors are associated with lower positive and higher negative affect (Bolger & Schilling, 1991; Laurent et al., 2009; Luginbuehl & Schoebi, 2019).

Findings also align with theory indicating that events tied to important and immediately relevant goals and values will be closely tied to emotional reactions (Frijda, 1986), and that romantic relationships are likely highly important for young adult college students because this developmental period is characterized by growing emphasis and experience in romantic relationships (Arnett, 2006) and because romantic relationships represent a life domain that offers fulfilling the needs of belonging and closeness (Baumeister & Leary, 1995). The findings also offer potential explanations for the connections between relationship quality and overall health and well-being (Bradbury & Karney, 2019). Low quality relationships characterized by frequent negative events could result in greater and more sustained instances of negative affect; highly tumultuous relationships, characterized by frequent highly positive and highly negative events, could contribute to the emotional variability that is consistently associated with negative psychosocial outcomes (Chaplin, 2006; Houben et al., 2015). While a strong emotional reaction to a salient event is an adaptive response when appropriately matched with valence and degree (Frijda, 1986; Luginbuehl & Schoebi, 2019), a relationship characterized by numerous negative events could take a lasting toll on psychological well-being through chronic experiences of heightened negative affect throughout daily life. Conversely, high quality relationships that afford frequent and highly positive experiences with romantic partners may provide avenues for emotion regulation following negative events from other life domains such as school or family (Schoebi & Randall, 2015). This can be directly translated into useful information for mental health professionals on college campuses who may seek to uncover sources of unhealthy affective fluctuations or avenues for affective support. Such professionals may be aided by assessing young adults’ relationship status, and whether a romantic partner may be a potential source of emotional support or of emotional strain.

Similar to romantic events, most academic and family events were also reliably tied to affective states in the moments they were experienced and the following moments, indicating these are also important life domains that play a role in college students’ health and well-being, as has been seen in previous studies (Altermatt, 2015; Bruffaerts et al., 2018; Tsai et al., 2013; Van Wel et al., 2002). However, positive academic events were not associated with changes in concurrent affect. The study also expanded the previous literature by including and directly comparing romantic events to other events as predictors of momentary affect. Such comparisons can be useful as they point to important targets of intervention for those who work on college campuses. With limited time and resources, mental health professionals must choose which arenas to prioritize when it comes to addressing potential antecedents to mental health issues or detriments in well-being. Our findings also indicated that the effects for romantic events were statistically significantly larger than all those for academic events except one, whereas effects for family events were all descriptively smaller than those for romantic events, but these differences were only statistically significant for changes in positive affect. Indeed, positive romantic events were associated with the greatest increases in positive affect, and negative romantic events were associated with the greatest increases in negative affect, indicating their immediate importance. However, direct comparisons indicated romantic events related to larger changes in negative affect compared to academic events and positive family events. Negative romantic and family events did not differ in their associations with increases in negative affect. Notably, because the alternate analyses accounted for event intensity and indicated similar results, the findings cannot be completely explained by romantic events merely being experienced as relatively more intense than the other events. Another potential interpretation, drawing from theories of emotion (Frijda, 1986), is that interpersonal life domains, whether romantic or familial, are highly relevant to emotional states in daily life (Baumeister & Leary, 1995), potentially more so than academics and other achievement-oriented domains (Glenn & Weaver, 1981). Such an interpretation supports the notion that our needs for social connection may be more important for our overall health and happiness than our needs for achievement or status.

Positive and negative events were differentially associated with affect, supporting the view that positive and negative arise from different underlying systems (Fischer & Manstead, 2008; Flook, 2011; Fredrickson, 2001). Further, previous empirical and theoretical work has indicated that positive events are more strongly associated with positive affect while negative events are more strongly associated with negative affect (Altermatt, 2015; Fischer & Manstead, 2008; Flook, 2011). Results from the concurrent models indicated immediate affective fluctuations during all events (except positive academic) for both positive and negative affect, suggesting some system overlap, yet the long-term associations only indicated such cross-over for negative events—positive events were generally only associated with subsequent changes in positive affect, whereas negative events were associated with subsequent changes in both positive and negative affect. This aligns more closely with previous work indicating that negative relationship behaviors are more strongly connected to several (but not all) markers of well-being compared to positive relationship behaviors (Rivers & Sanford, 2018). The much larger effects found for negative events suggest that positive experiences might not fully counteract the detrimental effects of negative experiences on affect, at least not instantaneously. However, for young adults seeking to cope with their emotions following negative events, having a highly positive experience with a partner—whatever that may be for them—may provide the most effective, immediate relief via increases in positive affect and decreases in negative affect.

Several analytical strengths contribute to the interpretation of the results and offer analytical takeaways. The statistical models optimized the rich moment-level data, accounting for multiple levels of variation and interdependence as well as important person-level characteristics. The models accounted for individual differences in affectivity and relationship experiences by including neuroticism and participants’ average event ratings across the reporting period as person-level controls and cross-level moderators. Neuroticism was a significant intercept control in all models, and a significant cross-level moderator for several of the event intensity slopes (see supplemental material), reiterating the importance of accounting for participants’ overall negative affectivity as when assessing fluctuations in both positive and negative affect (e.g., Bolger & Schilling, 1991). The models also accounted for potentially confounding relationship characteristics, indicating the associations were not limited to specific types of relationships (e.g., new or old, good or bad). The models also accounted for day-level variance (Bolger & Laurenceau, 2013) and number of events experienced across that day. Findings thus echo previous associations between day-level fluctuations in affect and specific types of romantic events (Bolger & Schilling, 1991; Kashdan et al., 2018; Laurent et al., 2009; Rogers et al., 2018), while also suggesting that moment-level associations may be more informative than day-level associations when it comes to predicting affect throughout daily life.

We also tested both event occurrence models and event intensity models. For the majority of the associations, the overall conclusions were the same across models regarding how experiencing certain events relate to both concurrent and subsequent affect. For romantic events, there were no substantive differences in the associations with affect. However, for positive academic and family events, the findings suggest an added predictive value of event intensity.

Limitations and future directions should be noted. First, while the broad approach to romantic events was a strength by offering a more general view of the associations between everyday events and affect, it necessarily precluded the ability to discern which types of events may be more or less strongly associated with affect in any given moment, or whether any specific event will have a greater association than any other specific event for a given individual. When comparing, for example, failing out of an important class versus a minor disagreement with a romantic partner, the academic event will likely be of greater emotional salience. While such differences in weight were accounted for by the intensity models, the broad nature of the event measures limits our understanding of how any specific event will relate to momentary affective experiences. Assessing and comparing multiple, specific types of events could be a fruitful avenue for future research and clinical applications.

Second, while prescription misuse tendencies were controlled for analytically at the person and moment level (and neither variable was reliably associated with the outcomes), the sample was primarily of students who had recently misused a prescription medication at screening, which could limit the generalizability of the findings as prescription drug misuse is associated with other substance use and lower academic achievement for some (McLarnon et al., 2012). Prescription drug misuse has been reported by college students representing numerous demographic characteristics (McLarnon et al., 2012), and the recent misuse group did not differ demographically from the non-misuse group in the overall sample (Papp et al., 2020), lending some support to the applicability of the findings reaching beyond solely college students who misuse prescription drugs. Nevertheless, a next important step would be to replicate these findings in more representative samples of partnered college students.

Additionally, while the theoretical foundation suggests directionality, the correlational nature of the data cannot rule out alternative directional explanations of the findings (e.g., momentary affect leading to more positive or negative romantic events Schoebi & Randall, 2015). Key measures in the current study were self-reported and collected in the same report, which was a strength for investigating momentary associations, but came with the drawback of introducing potential for response biases. To reduce participant burden, EMA measures were necessarily brief, and while used in previous studies and high in face validity, have not all been subject to rigorous psychometric testing. The event measures were also fairly broad, introducing another limitation by trading specificity for generality. The affect measures represented a portion of the positive and negative affect spectrums; other affective experiences (e.g., anger, pride) could be an important avenue for future research. Partner status was only assessed at the first lab visit, and thus precluded the ability to assess change in partner status during the reporting period. Also, the current study only included one partner among dating individuals; future studies including both or all romantic partners could account for interdependence inherent in these relationships, or including uncommitted individuals who engage in romantic experiences such as via dating apps, furthering our understanding of these associations beyond partnered individuals. Lastly, the analyses could have been underpowered. Sample size for the overall study was determined by the larger goals of capturing prescription drug misuse and not relationship status. The robust effects found for momentary events and affect indicated sufficient power to detect large effects, though effects of more moderate sizes, such as the comparisons of family events or subsequent affect, may have been undocumented.

In conclusion, the current study provides further evidence that romantic relationship experiences play critical roles in emotional well-being, corresponding to the highest highs and lowest lows throughout daily life (Berscheid & Ammazzalorso, 2003; Bradbury & Karney, 2019). For young adult college students, moment-to-moment affective states are very closely tied to experiences within their romantic relationships throughout daily life, and romantic experiences may correspond to greater and longer lasting emotional fluctuations than events representing other critical domains of young adult college students’ lives.