Daily Social Interactions and Momentary Loneliness: The Role of Trait Loneliness and Neuroticism

Abstract

Objectives

Loneliness has been linked to poor mental and physical health outcomes in later life. Little is known about how daily social interactions relate to older adults’ everyday experiences of loneliness. This study examined the dynamic associations between social interactions and the momentary feelings of loneliness in older adults’ daily lives. We further examined whether individual differences in trait loneliness and neuroticism influenced the extent to which daily social interactions were related to moment-to-moment changes in loneliness.

Method

Participants were 317 community-dwelling older adults (aged 70–90 years) who reported their social interactions and momentary feelings of loneliness 5 times daily for 14 consecutive days using smartphones.

Results

Having more frequent, more pleasant, and in-person social interactions, as well as interactions with family and friends specifically, significantly predicted lower momentary loneliness a few hours later. Higher levels of momentary loneliness, in turn, predicted less likelihood of engaging in these types of social interactions subsequently. In addition, older adults with higher (vs lower) traits of loneliness and neuroticism experienced greater decreases in momentary feelings of loneliness after having more frequent or pleasant social interactions, or interactions with family members.

Discussion

These results expand our understanding of the dynamic associations between daily social interactions and loneliness in later life and provide insights to inform future research, including the possibility of behavioral interventions that target social interactions to reduce the risk for loneliness.

Loneliness is an aversive internal experience that arises in response to a perceived discrepancy between desired and actual social relationships (Perlman & Peplau, 1982). Reflecting the subjective experience and suffering from feeling socially disconnected, loneliness is distinct from social isolation, which is an objective state of having minimal social contact or social connections (Donovan & Blazer, 2020). A recent meta-analysis of studies conducted in 29 countries found that approximately 28.5% of older adults (age 60 years and older) reported feeling lonely (Chawla et al., 2021). Loneliness in later life has been linked with increased risk of cardiovascular disease, depression and anxiety disorders, functional limitations, cognitive impairment, Alzheimer’s disease and dementia, and all-cause mortality (see Holt-Lunstad et al., 2015; Lara et al., 2019; Ong et al., 2016; Perissinotto et al., 2012 for reviews). Given the health implications of loneliness, it is critical to better understand factors that could help prevent or mitigate loneliness effectively.

Because loneliness is the subjective response to perceived social deficits, it is reasonable to expect that the feeling of loneliness will vary as a function of individuals’ social interaction experiences. Indeed, studies that captured social interactions in day-to-day life found evidence that individuals’ trait levels of loneliness were negatively associated with the quantitative aspects of daily social interactions, such as the amount of time spent with others and the quality of social interactions (e.g., positive social interactions, meaningfulness of social interactions; Hawkley et al., 2007; Wheeler et al., 1983; Wolf & Davis, 2014). However, much of the research to date has treated loneliness as a static, individual difference variable (i.e., trait loneliness or chronic loneliness) rather than an experience or feeling that may vary from moment-to-moment and across different contexts for the same person. As a result, little is known about how older adults’ daily social interactions may relate to fluctuations in their everyday experiences of loneliness. The present study aimed to address this knowledge gap by examining the dynamic associations between different types of social interactions that older adults have in their daily lives and the moment-to-moment fluctuations in their feelings of loneliness captured by ecological momentary assessments (EMAs) in naturalistic environments.

Momentary Loneliness and Daily Social Interactions

Recently, researchers have begun to examine the way that state or momentary experiences of loneliness fluctuate within people over time and suggest that contextual factors may influence the momentary variation in loneliness. For example, a recent EMA study found that older adults experienced higher levels of loneliness in situations when they were alone compared to being with others (Compernolle et al., 2021). Older adults also reported more momentary loneliness during time periods when they watched television compared to times when they were not watching television (Fingerman et al., 2021). These studies provide evidence that in addition to its trait-like elements, loneliness can also be a transitory experience with links to individuals’ social contexts or momentary activities. Surprisingly, few studies have examined how momentary loneliness is related to daily social interactions, especially among older adults. Studies among young or middle-aged adults provide some evidence that loneliness can both influence and be influenced by changes in individuals’ daily social interactions. For example, having more frequent and better-quality social interactions was found to predict lower loneliness feelings at the daily level for a sample of adults (Mean_Age = 49 years; Kuczynski et al., 2021), supporting the predictive effects of both the quantity and quality of daily social interactions on momentary loneliness.

Although the directionality of the association between social interactions and loneliness is often conceptualized as social interactions (or lack thereof) predicting loneliness, bidirectional effects―such that individuals may experience loneliness and subsequently seek and engage in interactions―have also been examined. Findings are mixed regarding whether momentary loneliness predicts increases or decreases in subsequent social interactions. The evolutionary theory of loneliness posits that loneliness serves as a signal that one’s social connections are frayed and thus motivates actions to repair and maintain social connections by engaging in more social interactions (Cacioppo et al., 2014). Consistent with this argument, a recent study among college students found that feelings of loneliness drove people to engage in more social interactions subsequently if they had no social contact in the preceding time period (Reissmann et al., 2021).

Ironically, however, although loneliness signals the need for social connection, it is also associated with increased sensitivity to social threats and fear of rejection (Cacioppo & Hawkley, 2005; Cacioppo et al., 2009), which may trigger social withdrawal responses and avoidance of social interactions. One daily diary study found that following increases in momentary loneliness, individuals (Mean_age = 36 years) spent less time interacting with others the next day, supporting the hypothesized social withdrawal response to loneliness (Arpin et al., 2015). Our study aimed to extend previous literature by examining the bidirectional associations between different features of social interactions (quantity, quality, and partner types) and moment-to-moment changes in older adults’ feelings of loneliness in daily life. Beyond the within-person associations between social interactions and momentary loneliness, we also addressed the important question of who would benefit more from different types of social interactions by testing the role of individual differences in two psychological traits―trait loneliness and neuroticism―in moderating these within-person associations.

The Moderating Role of Trait Loneliness

Prior research suggests that individuals differ in their abilities to benefit from social interactions as a source of emotion regulation. According to the hypervigilance hypothesis, lonely individuals, relative to nonlonely individuals, are more likely to perceive social interactions as threatening, derive less pleasure from social interactions, and cope with interpersonal stress in a passive, isolative fashion (Cacioppo et al., 2009).

Previous laboratory-based research found evidence that individuals who were higher in trait loneliness showed weaker reward-related brain activity in response to pleasant pictures of people compared to those who were lower in trait loneliness, suggesting that lonely individuals were less rewarded by positive social stimuli (Cacioppo et al., 2009). However, recent evidence from studies that examined individuals’ emotional responses to real-life social interactions suggests that lonely individuals may have stronger emotional responses to social interactions in real life than their nonlonely counterparts. For example, chronically lonely individuals (those who are high in trait loneliness) reported greater boosts in feelings of enjoyment on days with more positive interpersonal events than usual, relative to those who were not lonely (Wolf & Davis, 2014). Another EMA study among adolescents found that those with high (vs low) trait loneliness reported higher levels of momentary loneliness when alone, but showed greater decreases in momentary loneliness when they had intimate company (van Roekel et al., 2018).

In summary, past theories and research support that the trait of loneliness plays an important role in influencing individuals’ emotional responses (including loneliness) to their social environment or experience. However, previous studies have mostly focused on young adults, and lab-based and observational studies have produced contradictory findings. Therefore, more research is needed to investigate whether trait loneliness would alter the degree to which older adults’ momentary loneliness is influenced by their daily social interactions.

The Moderating Role of Trait Neuroticism

Based on previous research, the personality trait of neuroticism is another important psychological factor that may influence the effects of daily social interactions on momentary loneliness. Neuroticism tends to be positively associated with trait loneliness in general (including among older adults; Buecker et al., 2020; Schutter et al., 2020; Wagner et al., 2016); a meta-analysis found the correlation to be 0.38 (95% confidence interval [CI] 0.35–0.42), and there appears to be strong shared genetic variance between neuroticism and trait loneliness, potentially related to negative affectivity (Abdellaoui et al., 2019; Spithoven et al., 2019). Importantly, however, neuroticism and trait loneliness are both conceptually and empirically distinct (Abdellaoui et al., 2019). Most notably, whereas neuroticism captures individual differences in the tendency to be emotionally unstable, easily stressed, and sensitive to potential threats in general (McCrae & Costa Jr., 1996), trait loneliness is characterized by a tendency to feel lonely and sensitivity to negative social stimuli more specifically (Cacioppo et al., 2009; Hawkley & Cacioppo, 2010). Thus, in any given moment involving social situations, trait loneliness and neuroticism may function similarly, but there is reason to examine whether and how they may influence the connection between state loneliness and social interactions differently given that trait loneliness is more specifically and closely related to social stimuli than trait neuroticism.

Personality theories posit that personality traits such as neuroticism can shape individuals’ perception and interpretation of a given social situation (e.g., social interactions) and affect their reaction to that situation (Barrett & Pietromonaco, 1997; McCrae & Costa Jr., 1996; Nezlek et al., 2011). Individuals with high neuroticism are hypersensitive to emotional and social cues and may thus experience stronger emotional responses to social interactions in general. There is a small but growing body of evidence suggesting that momentary emotional responses to both positive and negative social interactions could be amplified among individuals with high neuroticism. For example, individuals with high (vs low) neuroticism have been found to be prone to heightened distress in response to daily interpersonal conflicts (Mroczek & Almeida, 2004; Suls et al., 1998), but also benefited more emotionally (e.g., increased positive affect or decreased negative affect) from positive interactions with close social partners, such as family and friends (Chui et al., 2014; Mueller et al., 2021; Shackman et al., 2018). However, existing studies have focused on how trait neuroticism influences individuals’ general emotional responses to social interactions, and no study has examined whether trait neuroticism could also influence the moment-to-moment changes in older adults’ feelings of loneliness in response to their daily social interactions.

Present Study

The present study has two aims. First, this study examined the bidirectional within-person associations between social interactions and momentary loneliness using fine-grained, intensive longitudinal data (i.e., EMAs) collected in older adults’ natural environments. We focused on three major features of social interactions―quantity, quality, and partner types―given previous evidence linking each of these features with the health and emotional well-being of older adults (Huxhold et al., 2013; Mejía & Hooker, 2015). To provide a holistic picture of older adults’ daily social experiences, we also assessed older adults’ daily social activities (both in-person and online social activities). We expected that relative to having no social interactions, having frequent social interactions/activities and high-quality (more positive and less negative) social interactions would predict lower levels of momentary loneliness subsequently. Aging theories and research also suggest that with increasing age, people prefer to interact with close social partners, such as family and friends, rather than with other peripheral partners; and also perceived these interactions as more emotional meaningful (Charles & Carstensen, 2010). Therefore, we expected that having social interactions with family and friends would predict lower levels of momentary loneliness subsequently relative to having no social interactions. We also investigated the potential reciprocal effects of momentary loneliness on subsequent changes in social interactions, but did not provide a specific hypothesis given contradictory findings from previous research regarding whether momentary loneliness would relate to increases or decreases in subsequent social interactions (Arpin et al., 2015; Reissmann et al., 2021).

Second, we addressed the research question of who may benefit more from social interactions in terms of reducing their momentary loneliness by testing the moderating roles of two psychological factors that have been shown to relate to individuals’ emotional responses to social experiences―trait loneliness and neuroticism (Cacioppo et al., 2009; McCrae & Costa Jr., 1996). Based on theories and evidence suggesting that individuals who are higher in trait loneliness or neuroticism are more emotionally reactive to social interactions than their counterparts (Cacioppo et al., 2009; McCrae & Costa Jr., 1996; Mueller et al., 2021), we predicted that older adults higher (vs lower) in trait loneliness or neuroticism would experience lower momentary loneliness after having frequent social interactions/activities, positive social interactions, or interactions with close partners; and they would experience higher momentary loneliness after having negative social interactions.

Method

Participants and Procedure

Data were drawn from the EMAs of the ongoing Einstein Aging Study (EAS), collected between May 2017 and March 2020. Participants in EAS were recruited via systematic random sampling from New York City Registered Voter Lists for Bronx County. Introductory letters were mailed to potential participants with valid addresses and phone numbers. Follow-up phone screens were conducted to determine eligibility (i.e., English-speaking community-residing individuals who are ambulatory and aged ≥70 years; see Zhaoyang, Sliwinski et al., 2021 for more details) and to enroll those who agreed to participate. The final sample included 317 older adults who provided valid data during the 14-day formal EMA session (see Table 1 for sample demographics). Power analysis conducted for the grant proposal supporting this project demonstrated that this sample size provides sufficient power (>80%) to detect small-sized effects in all analyses at alpha of 0.5 level.

Table 1.

Descriptive Information on Sample Demographics and Key Study Variables

	N	Mean or %	SD	Range
Sample demographics
Sex (female)	317	67.5%
Age (years)	317	77.45	4.83	70.40–90.60
Race/ethnicity	316
White	145	45.9%
Black	127	40.2%
Hispanic White	31	9.8%
Hispanic Black	9	2.9%
Asian	4	1.3%
College degree or higher	317	46.4%
Currently employed	317	8.5%
Currently married	317	33.4%
Living alone	317	53.6%
Person-level study variables
Trait loneliness	316	1.34	0.46	1–3
Neuroticism	316	2.42	0.72	1–5
EMA study variables
Momentary loneliness	18,339	12.71	18.91	0–100
Any social interactions	18,340	77.2%
Pleasant social interaction	14,158	84.4%
Unpleasant social interaction	14,158	0.81%
Interaction with family members	14,155	38.1%
Interaction with friends	14,155	12.9%
Interaction with weak ties	14,155	11.4%
In-person social activity	18,334	36.4%
Online social activity	18,334	12.8%

Note: EMA = ecological momentary assessment; SD = standard deviation.

Following the initial phone-screening assessment, eligible participants completed the consent process and were invited to attend a visit to the research clinic. During this visit, they completed assessments on demographic and psychosocial characteristics and 1.5 hr of training on the study protocol and use of study smartphones (six participants completed this training remotely after the onset of the pandemic, whereas all other participants completed both training and formal sessions of EMA before the onset of the pandemic). The day after the training, the participants began a 2-day practice session of EMA followed by a 14-day formal EMA session using the study smartphones. The EMA protocol involved self-initiated wake-up and end-of-day surveys and four quasi-randomly beeped surveys (with approximately 3.5 hr interval) each day. After the EMA session, participants returned the study smartphones, and the data were downloaded from the smartphones. Participants who completed all aspects of data collection received $160 as compensation.

This study used data from baseline assessments of demographics and psychosocial characteristics and the EMAs (beeps and end-of-day surveys only; morning surveys did not include questions on loneliness and social interactions) on social interactions and momentary loneliness. The 317 participants provided 18,363 EMAs over the 14-day period of the formal EMA session (14,801 beep surveys and 3,562 end-of-day surveys). On average, participants completed 13.65 days of EMAs (standard deviation [SD] = 1.32, range = 3–14 days), 83.38% of all assigned beep surveys, and 80.26% of all assigned end-of-day surveys. Missing data analyses revealed that compliance rates of beep surveys were not significantly associated with any demographics listed in Table 1; the compliance rates of end-of-day surveys were only positively associated with education (γ = 0.22, p < .000). In addition, the compliance rates of any surveys were not associated with baseline trait neuroticism or loneliness.

Measures

Ecological momentary assessments

Momentary loneliness, social interactions and social activities were measured by EMAs five times each day. Descriptive information on key study variables are presented in Table 1.

Momentary loneliness

At each EMA survey, participants were asked, “Right now, do you feel lonely?” and could choose their response using a slider scale with end-points “not at all” (0) to “extremely” (100).

Social interactions

The quantity, quality, and partner types of social interactions were assessed at each EMA survey. Quantity: participants were asked whether they had any social interactions (defined as talking or spending time with someone in-person, by phone/computer, or by texting) since the last survey (i.e., in the past 3.5 hr). A binary variable was created to indicate whether participants had any social interactions in the past 3.5 hr or not. Quality: if participants responded having any social interactions, they were then asked to select whether their most recent social interaction was “pleasant,” “unpleasant,” “neutral,” or “both pleasant and unpleasant” on a single-choice question. Two binary variables were created to indicate whether any “pleasant” or “unpleasant” social interactions occurred in the past 3.5 hr, respectively (1 = yes, 0 = no social interaction). Partner types: participants were also asked to select the partner(s) involved in their most recent social interaction from a list including spouse/partner, children, other family members, friends, neighbors, acquaintances, strangers, and others. Based on past research (Zhaoyang et al., 2018; Zhaoyang, Scott et al., 2021), three binary variables were created to indicate whether any social interaction involving only family, friends, or weak ties (e.g., neighbors, acquaintances, strangers, and others) occurred in the past 3.5 hr, respectively (1 = yes, 0 = no social interaction).

In-person socializing and online social media activity

At each EMA survey, participants were asked what activities they had engaged in since the last survey (i.e., in the past 3.5 hr). Participants could select all activities that they had done from a list, including in-person socializing and using online social media (e.g., Facebook). Two binary variables were created to indicate whether any in-person socializing or online social media activity were reported (1 = yes, 0 = no).

Baseline questionnaire

Trait loneliness

The three-item UCLA loneliness scale was used to assess participants’ traits of loneliness (Hughes et al., 2004). Three items asked how often participants felt they lacked companionship, left out, and isolated from others, respectively. Responses were on a three-point scale (1 = hardly ever or never, 2 = some of the time, and 3 = often). An average score was created to indicate the mean levels of the trait-level loneliness (Cronbach reliability α = 0.80).

Trait neuroticism

The Big Five Personality Inventory 44-item scale (BFI-44, John & Srivastava, 1999) was used to measure the personality traits of neuroticism, extraversion, openness, conscientiousness, and agreeableness. Participants were asked to rate on a five-point Likert scale to what degree each of the 44 statements describes them (1 = disagree strongly, 5 = agree strongly). Although planned analyses focused on neuroticism, average scores were created for each personality trait for use in exploratory analyses (Cronbach reliability αs = 0.75, 0.76, 0.78, 0.79, and 0.75 for neuroticism, extraversion, openness, conscientiousness, and agreeableness, respectively).

Covariates

The following demographic and social context variables were assessed by baseline questionnaires and included as person-level covariates: age (in years), sex (male = 0, female = 1), race/ethnicity (1 = White, 2 = Black, 3 = Hispanic White, 4 = Hispanic Black, 5 = Asian, and 6 = other), education (1 = college degree or higher, 0 = below college), employment (1 = currently employed, 0 = not employed), living status (1 = living alone, 0 = not living alone), and marital status (1 = married, 0 = not married, including never married, divorced, widowed, or separated). In addition, the momentary status of solitude (1 = currently being alone, 0 = currently with others) and negative affect (anxious/tense, depressed/blue, frustrated, 0 = not at all, 100 = extremely, Watson & Clark, 1994) reported at each EMA, both of which have been linked with momentary loneliness (Compernolle et al., 2021; Meng et al., 2020; van Roekel et al., 2018), were included as momentary-level covariates.

Data Analysis

Multilevel modeling (MLM) was used to examine the within-person associations between social interactions and momentary loneliness (Singer & Willett, 2003). The data were structured hierarchically with EMAs nested within persons. Analyses predicting momentary loneliness as a continuous outcome were conducted using SAS PROC MIXED with restricted maximum likelihood. All models included a random intercept, a random slope of the within-person effect of social interactions, and an autoregressive variance–covariance structure for residuals (AR [1]) to account for the interdependence of outcome measures.

Analyses were conducted in a series of steps. First, within-person associations between social interactions that occurred since the last EMA (i.e., in the past 3.5 hr) and current feelings of loneliness were examined in several models, each of which tested the predictive effect of one feature of social interactions. A binary variable was used to capture whether participants had a certain type of social interaction in the past 3.5 hr (e.g., pleasant interactions) in comparison with having no social interactions. Following previous research (Sliwinski et al., 2009; Zhaoyang et al., 2020), we included the person-level mean of the same social interaction variable in the same model to separate between-person effect from the within-person effect of this type of social interaction on current feelings of loneliness. This approach keeps the meaningful unit of the binary variable of having social interaction (vs not) and thus provided more interpretable results compared with a person-mean-centered approach. In addition, each model also included day-level and hour-level trends (hours of a day, days of the study) and person- and momentary-level covariates (person-mean centered).

Second, we conducted time-lagged analyses to examine the extent to which effects of social interactions on momentary loneliness endured beyond the approximately 3.5-hr window between two EMA surveys within a day. Specifically, social interactions reported on time t−1 (i.e., occurring between the past 3.5 and 7 hr) was added to each of the models in step one to predict momentary loneliness feelings reported on time t. To differentiate and account for the concurrent effect, social interactions reported on time t were retained in the lagged models. Next, we examined the potential reciprocal effects of momentary loneliness on subsequent social interactions by testing multilevel logistic models using SAS PROC GLIMMIX, which predicted the occurrence of a certain type of social interaction (binary outcome) reported on time t from momentary loneliness reported on time t−1. Finally, a cross-level interaction between the binary variable of the occurrence of a certain type of social interaction and a person-level continuous moderator (e.g., trait loneliness or neuroticism; grand-mean centered) was added to each model in step one to test the moderation hypotheses; simple slopes were estimated for significant interaction effects.

Results

Descriptive Information

As shown in Table 1, older adults in our sample reported relatively low levels of trait loneliness (M of average score = 1.34 on a three-point scale; M of sum score = 4.02) and momentary loneliness across the 14-day study period on average (M = 12.71 on a 100-point scale). The intraclass correlation for the momentary loneliness score was 0.75, suggesting that momentary loneliness varied more at the between-person level (75%) than at the within-person level (25%), but there was still considerable variance in loneliness feelings from occasion to occasion for each person. In terms of social interactions, participants reported having social interactions on 77.2% of all EMA occasions. The majority (84%) of reported social interactions were rated as “pleasant” and only 0.8% of interactions were rated as “unpleasant.” In terms of social interaction partners, 38.1% of all reported interactions involved only family members, 12.9% involved only friends, and 11.4% involved only peripheral partners such as acquaintances, neighbors, or strangers. Participants also reported having in-person or online social activities in 36.4% and 12.8% of all EMA surveys, respectively. Supplementary Table 1 includes the correlations between study variables.

Within-Person Associations Between Social Interactions and Momentary Loneliness

The results from MLMs predicting current momentary loneliness feelings from social interactions that occurred in the past 3.5 hr (see Table 2) demonstrated that momentary loneliness was significantly lower when older adults reported having any type of social interactions, pleasant social interactions, interactions with family or friends, or in-person social activities in the past 3.5 hr relative to occasions when they had no social interactions or social activities in the past 3.5 hr. These predictive effects of social interactions held even after controlling for the effects of individual differences in demographics and situational covariates, including solitude status and negative affect. In contrast, having unpleasant social interactions, social interactions with weak ties, or online social activities in the past 3.5 hr did not significantly predict current levels of momentary loneliness.

Table 2.

Main Effects of Different Types of Social Interactions Predicting Momentary Loneliness

	Outcome = Momentary loneliness
	Fixed effects
Predictors = Social interactions since last EMA survey	Est.	SE	p	95% CI
Having any social interactions (vs no interaction)	−0.110	0.032	.001	(−0.173, −0.047)
Having pleasant social interactions (vs no interaction)	−0.102	0.034	.003	(−0.169, −0.035)
Having unpleasant social interactions (vs no interaction)	−0.078	0.226	.730	(−0.534, 0.377)
Having interactions with family (vs no interaction)	−0.113	0.037	.002	(−0.186, −0.041)
Having interactions with friends (vs no interaction)	−0.099	0.045	.029	(−0.188, −0.010)
Having interactions with weak ties (vs no interaction)	−0.079	0.060	.193	(−0.198, 0.040)
Having in-person social activities (vs no social activity)	−0.102	0.036	.005	(−0.174, −0.030)
Having online social activities (vs no social activity)	−0.002	0.049	.963	(−0.100, 0.095)

Notes: Outcome of momentary loneliness was square root transformed to reduce the skewness and the transformed momentary loneliness had a range of 0–10. Each social interaction predictor was examined in a separate model, which included person-level mean of the same social interaction variable, day- and hour-level trends (for detrending purpose), sex, age, race/ethnicity, marital status, living status, employment, education, and momentary level solitude status and negative affect. EMA = ecological momentary assessment; SE = standard error; CI = confidence interval.

Time-lagged analyses were then conducted by adding social interactions reported at time t−1 (occurred between the past 3.5 and 7 hr; lagged effect) to the models to predict momentary loneliness reported at time t. This lagged analysis did not find any significant lagged effects of social interactions on current momentary loneliness (results were thus not presented in tables).

We then tested the reciprocal effects of momentary loneliness on subsequent social interactions using lagged analyses (see Table 3). The results suggest that greater momentary loneliness at time t significantly predicted lower likelihood of having any type of social interactions, pleasant social interactions, interactions with family, or in-person social activities in the next 3.5 hr (reported at time t+1). These predictive effects of momentary loneliness held after controlling for individual differences in demographics and situational covariates.

Table 3.

Lagged Effects of Momentary Loneliness Predicting Different Types of Social Interactions

	Predictor = Momentary loneliness reported at time t
	Fixed effects
Outcome = Social interactions reported at time t+1	Est.	OR	SE	p	95% CI
Having any social interactions (vs no interaction)	−0.057	0.945	0.027	.035	(−0.110, −0.004)
Having pleasant social interactions (vs no interaction)	−0.071	0.931	0.028	.012	(−0.127, −0.016)
Having unpleasant social interactions (vs no interaction)	0.080	1.083	0.285	.780	(−0.479, 0.638)
Having interactions with family (vs no interaction)	−0.082	0.922	0.038	.032	(−0.156, −0.007)
Having interactions with friends (vs no interaction)	0.004	1.004	0.041	.923	(−0.076, 0.083)
Having interactions with weak ties (vs no interaction)	−0.054	0.947	0.044	.221	(−0.142, 0.033)
Having in-person social activities (vs no social activity)	−0.055	0.946	0.025	.029	(−0.105, −0.006)
Having online social activities (vs no social activity)	−0.061	0.941	0.057	.282	(−0.173, 0.050)

Notes: Multilevel logistic regression was used to predict the binary outcome. For better model convergence and interpretation, predictor of momentary loneliness was rescaled to 0–10 range. Each model also included momentary loneliness at time t+1 (concurrent effect), person-level means of momentary loneliness, day- and hour-level trends (for detrending purposes), sex, age, race/ethnicity, marital status, living status, employment, education, and momentary level solitude status and negative affect. SE = standard error; CI = confidence interval; OR = odds ratio.

Moderating Effects of Trait Loneliness

We next examined whether the within-person effects of social interactions on momentary loneliness varied across individuals with different levels of trait loneliness. Results (see Table 4) demonstrated that the individual differences in trait loneliness significantly moderated the within-person associations between social interactions and momentary loneliness for five out of eight social interaction variables: occurrence of any social interactions, pleasant social interactions, social interactions with family, social interactions with weak ties, and in-person social activities. The simple slopes for these interactions demonstrated a consistent pattern: The within-person effects of having any social interactions, pleasant social interactions, interactions with family, interactions with weak ties, and in-person social activities in the past 3.5 hr significantly predicted lower levels of momentary loneliness for older adults with higher trait loneliness (defined as 1 SD above the sample mean), but not for those with lower trait loneliness (defined as 1 SD below the sample mean; simple slopes were also tested for participants scored “1” at the trait loneliness scale to handle the out-of-range issue and revealed the same patterns of results). In other words, older adults scored higher in trait loneliness benefited more from having these types of social interactions or social activities with regard to reducing their momentary loneliness feelings in daily life, compared with those scored lower in trait loneliness, supporting our hypothesis.

Table 4.

Summary of Significant Moderation Effects of Trait Loneliness on the Within-Person Associations Between Social Interactions and Momentary Loneliness

	Outcome = Momentary loneliness
	Est.	SE	p	95% CI
Predictor = Social interactions; Moderator = Trait loneliness
Having any social interactions × Trait loneliness	−0.228	0.063	.000	(−0.350, −0.105)
Simple slope―Higher trait loneliness	−0.201	0.040	.000	(−0.280, −0.122)
Simple slope―Lower trait loneliness	0.008	0.045	.859	(−0.080, 0.096)
Having pleasant social interactions × Trait loneliness	−0.260	0.066	.000	(−0.391, −0.130)
Simple slope―Higher trait loneliness	−0.208	0.043	.000	(−0.293, −0.124)
Simple slope―Lower trait loneliness	0.031	0.047	.506	(−0.061, 0.124)
Having interactions with family × Trait loneliness	−0.260	0.075	.001	(−0.406, −0.114)
Simple slope―Higher trait loneliness	−0.228	0.050	.000	(−0.325, −0.130)
Simple slope―Lower trait loneliness	0.011	0.050	.820	(−0.087, 0.110)
Having interactions with weak ties × Trait loneliness	−0.278	0.121	.022	(−0.516, −0.040)
Simple slope―Higher trait loneliness	−0.184	0.075	.014	(−0.331, −0.037)
Simple slope―Lower trait loneliness	0.071	0.090	.427	(−0.105, 0.248)
Having in-person social activities × Trait loneliness	−0.240	0.078	.002	(−0.393, −0.088)
Simple slope―Higher trait loneliness	−0.213	0.050	.000	(−0.312, −0.114)
Simple slope―Lower trait loneliness	0.008	0.051	.875	(−0.092, 0.108)

Notes. Higher trait loneliness was defined as 1 SD above the sample mean; lower trait loneliness was defined as 1 SD below the sample mean. Outcome of momentary loneliness was square root transformed to reduce the skewness and the transformed momentary loneliness had a range of 0–10. Each interaction effect was examined in a separate model, which included person-level mean of the same social interaction variable, sex, age, race/ethnicity, marital status, living status, employment, education, and momentary level solitude status and negative affect, as well as day- and hour-level trend (for detrending purposes). SE = standard error; SD = standard deviation; CI = confidence interval.

Moderating Effects of Trait Neuroticism

Similarly, we examined whether the personality trait of neuroticism moderated the within-person effects of social interactions on momentary loneliness. The results revealed that individual differences in neuroticism significantly moderated three within-person effects of social interactions on momentary loneliness: any social interactions, pleasant social interactions, and social interactions with family (see Table 5). The simple slopes for these interactions also demonstrated a consistent pattern: The within-person effects of having any social interactions, pleasant social interactions, and interactions with family members significantly predicted lower levels of momentary loneliness 3.5 hr later for individuals with higher neuroticism (defined as 1 SD above the sample mean), but not for those with lower trait neuroticism (defined as 1 SD below the sample mean). In other words, older adults with higher neuroticism benefited more from having frequent and pleasant social interactions, as well as more interactions with family members with regard to reducing their momentary loneliness in daily life, compared with individuals with lower neuroticism, supporting our hypothesis.

Table 5.

Summary of Significant Moderation Effects of Trait Neuroticism on the Within-Person Associations Between Social Interactions and Momentary Loneliness

	Outcome = Momentary loneliness
	Est.	SE	p	95% CI
Predictor = Social interactions; Moderator = Trait neuroticism
Having any social interactions × Trait neuroticism	−0.133	0.042	.002	(−0.214, −0.051)
Simple slope―Higher trait neuroticism	−0.205	0.044	.000	(−0.291, −0.119)
Simple slope―Lower trait neuroticism	−0.014	0.044	.745	(−0.100, 0.071)
Having pleasant social interactions × Trait neuroticism	−0.133	0.044	.002	(−0.219, −0.047)
Simple slope―Higher trait neuroticism	−0.198	0.046	.000	(−0.289, −0.107)
Simple slope―Lower trait neuroticism	−0.006	0.046	.892	(−0.096, 0.083)
Having interactions with family × Trait neuroticism	−0.152	0.046	.001	(−0.242, −0.062)
Simple slope―Higher trait neuroticism	−0.220	0.049	.000	(−0.317, −0.124)
Simple slope―Lower trait neuroticism	−0.001	0.049	.980	(−0.098, 0.095)

Notes: Higher trait neuroticism was defined as 1 SD above the sample mean; lower trait neuroticism was defined as 1 SD below the sample mean. Outcome of momentary loneliness was square root transformed to reduce the skewness and the transformed momentary loneliness had a range of 0–10. Each interaction effect was examined in a separate model, which included person-level mean of the same social interaction variable, sex, age, race/ethnicity, marital status, living status, employment, education, and momentary level solitude status and negative affect, as well as day- and hour-level trend (for detrending purposes). SE = standard error; SD = standard deviation; CI = confidence interval.

We also conducted exploratory analyses to test whether the other four personality traits (extraversion, openness, conscientiousness, and agreeableness) moderated any within-person effects of social interactions on momentary loneliness and found no significant moderating effects for these traits.

Discussion

Loneliness is a prevalent phenomenon in later life and has substantial impact on older adults’ health, well-being, and longevity (Holt-Lunstad et al., 2015; Lara et al., 2019; Ong et al., 2016; Perissinotto et al., 2012). This study examined the dynamic associations between older adults’ daily social interactions and momentary feelings of loneliness, and further investigated whether individual differences in trait loneliness and neuroticism could influence the extent to which people benefited from distinct types of social interactions with regard to reducing their momentary loneliness in daily life.

Our study found evidence to support that older adults’ momentary feelings of loneliness can both influence and be influenced by their daily social interactions. Relative to having no social interactions, the occurrence of any social interactions, pleasant social interactions, interactions with family and friends, or in-person social activities all significantly predicted lower momentary loneliness a few hours later. This set of findings suggests that these types of social interactions may help mitigate momentary loneliness for older adults. At the same time, however, we found that momentary loneliness, in turn, predicted lower likelihood of having any social interactions, pleasant social interactions, interactions with family, or in-person social activities subsequently. This finding supports the social withdrawal hypothesis: Although loneliness signals the need for social connection, the feeling of loneliness is also associated with hypersensitivity to social threats and fear of rejection (Cacioppo et al., 2009; Hawkley & Cacioppo, 2010) and may in fact trigger social withdrawal responses and avoidance of social interactions (Arpin et al., 2015). Together, our study established the bidirectional, temporal order between the moment-to-moment changes in social interactions and loneliness feelings and highlighted the detrimental consequences of lacking social contact in later life: Having no social interactions (vs having social interactions) was related to greater momentary loneliness, which in turn led to further withdrawal from social interactions. These results provide support for the concept that there is a vicious cycle of social avoidance and feelings of loneliness in day-to-day life that could contribute to the persistence of older adults’ chronic loneliness and social isolation in the long term (Cacioppo & Hawkley, 2005; Cacioppo et al., 2014).

Another contribution of our study is that we addressed the important question of who would benefit more from daily social interactions by testing the moderating role of trait loneliness and neuroticism. We found robust evidence that older adults with higher trait loneliness benefited more from having daily social interactions than those with lower trait loneliness: Occurrence of any social interactions, pleasant social interactions, interactions with family, interactions with weak ties, and in-person social activities all predicted lower levels of momentary loneliness 3.5 hr later for those with higher (but not lower) trait loneliness. These findings support previous loneliness theories and research suggesting that trait level or chronic loneliness would make people hypersensitive and more reactive to socially relevant stimuli (Cacioppo et al., 2009; Wolf & Davis, 2014). Our finding is also in correspondence with other EMA studies that examined individuals’ general emotional responses (e.g., positive or negative affect) to real-life social interactions and demonstrated that lonely individuals benefit more emotionally from positive social interactions than their nonlonely counterparts (van Roekel et al., 2018; Wolf & Davis, 2014). Particularly, our finding regarding weak ties highlights that even social contacts with peripheral partners could bring emotional benefits to those who are chronically lonely, and thus suggests that intervention to increase social interactions of any kind may be helpful for this subpopulation.

Additionally, we found that having any social interactions, pleasant interactions, and interactions with family members predicted lower levels of momentary loneliness 3.5 hr later for individuals with higher (but not lower) neuroticism. Individuals who are higher in neuroticism are characterized as being hypersensitive to negative stimuli in general and prone to have stronger emotional responses to interpersonal experiences (McCrae & Costa Jr., 1996). Recent research also suggests that individuals with a more negative disposition (e.g., greater neuroticism) often rely on close partners to regulate their chronically elevated distress and would derive much larger emotional benefits from the company of close partners such as family and friends than nonneurotic people (Chui et al., 2014; Mueller et al., 2019; Shackman et al., 2018). In summary, our findings regarding the moderating roles of trait loneliness and neuroticism underscore the importance of considering individual differences for future research and interventions aimed to understand and prevent loneliness in later life.

There are several limitations of this study that present promising avenues for future research. First, few unpleasant social interactions were reported in our study (less than 1% of EMAs). Although this is consistent with aging theories and research that older adults are more likely to avoid interpersonal conflicts than younger adults (Birditt et al., 2020; Charles & Carstensen, 2010), the lack of variability in reported negative interactions may partially contribute to the nonsignificant findings observed with negative social interactions in this study. Future research may benefit by including more days of EMAs and more questions on negative social interaction to thoroughly understand its influence. Second, our study defined social interactions as “talking or spending time with someone in person, by phone/computer or by texting,” but did not clearly capture the channel of each reported social interaction episode (e.g., in-person, telephone, or online). Recent evidence suggests that older adults may prefer in-person social interaction over other channels of social interactions, but it is still an open question whether social interactions taking place over other channels can ever compensate for the lack of in-person interaction (Hülür & Macdonald, 2020). It is possible that in-person social interactions would have different effects on loneliness feelings compared with online or telephone interactions. Future research is needed to further examine whether the associations between daily social interactions and loneliness differ across distinct social interaction channels. Third, the majority of data for our study were collected before the height of the initial wave of the COVID-19 pandemic in the United States (98% data were collected before February, 2020) and thus did not capture the increasing popularity and importance of digital social interactions following the onset of the global pandemic for older adults. We had data from a question about online social activities but this only assessed the usage of social media broadly and did not capture the purpose of using social media (e.g., communication/interactions vs other noncommunication purposes). Although we found no significant association between online social activities and momentary loneliness in this study, this finding needs to be replicated in future research with more data on digital or online social interactions during and postpandemic. Fourth, although our study provided strong evidence to support the association between momentary loneliness and social interactions, it is important to note that participants’ daily social interactions and/or loneliness may also be influenced by other factors, such as their schedules or plans for the day or their environment. Finally, although the sample of older adults in this study was diverse in race/ethnicity, our findings may not generalize to older adults who do not live in metropolitan areas, do not speak English, or do not use smartphones or internet in their daily lives; our sample, recruited from New York City, may have different social interaction patterns than other populations (e.g., more opportunities for social interactions, more interactions with weak ties, and a higher likelihood of engaging in online social activities).

In conclusion, this study contributes to the literature on social interactions and loneliness by providing strong evidence to support bidirectional associations between daily social interactions and momentary loneliness: Having frequent, pleasant, and in-person social interactions and interactions with family and friends predicted lower momentary loneliness a few hours later; momentary loneliness, in contrast, predicted less likelihood of engaging in social interactions subsequently. This study also demonstrated that older adults with higher (vs lower) traits of loneliness and neuroticism benefited more from having frequent, pleasant social interactions and interactions with family members in terms of mitigating their momentary feelings of loneliness in daily life. These findings improve our understanding of the dynamic associations between daily social interactions and loneliness in later life and provide insights for future loneliness or social isolation interventions to target certain types of social interactions, as well as certain groups of older adults (e.g., those are higher in trait loneliness or neuroticism) who may benefit the most from these interventions.

Funding

This work was supported by the National Institute on Aging at National Institutes of Health (R03 AG067006 to R. Zhaoyang, P01 AG003949 to M. J. Sliwinski, R01 AG060933 to S. B. Scott, and T32 AG049676 to K. D. Harrington).

Conflict of Interest

The authors have no conflict of interests.