Introduction

The development and subsequent maintenance of meaningful social connections is essential for wellbeing (Cacioppo & Patrick, 2008). A breakdown, or loss, of these social ties increases risk for loneliness, defined as the distressing emotion that accompanies a perceived discrepancy between one’s desired and actual quality of social relationships (Peplau & Perlman, 1982; Pinquart & Sörensen, 2001). From an evolutionary perspective, transient loneliness is theorised to promote social connection and/or deter individuals from engaging in actions detrimental to their social group, thereby ensuring survival of genes (Cacioppo & Patrick, 2008). For approximately 10–30% of the adult population, however, loneliness develops into a chronic state (Steed et al., 2007; Theeke, 2009, 2010). Given the strong link between chronic loneliness and physical and mental health (Beutel et al., 2017; Hawkley & Cacioppo, 2010; Malcolm et al., 2019), it is imperative to broaden our understanding of mechanisms causing and/or maintaining these feelings.

The model of loneliness, pioneered by Cacioppo and Hawkley (2009) and advanced by Qualter et al. (2015), provides an account of how loneliness affects cognition and behaviour in a maladaptive manner to maintain distress. This model proposes that the aversive feeling of loneliness signals that a social connection is either broken or under threat. A consequence of this is that individuals preferentially attend to social information over and above non-social information. Although this may help them identify additional opportunities to reconnect with others (i.e., cognitive reaffiliation process; Qualter et al., 2015), if the process does not function as intended, it produces ‘cognitive biases’: attention (e.g., preferential allocation of attention towards stimuli indicative of rejection over stimuli indicative of (re)connection), confirmation (e.g., misinterpretation of others’ social behaviour(s) as hostile and/or rejecting, as opposed to ambiguous or neutral), and memory (e.g., tendency to retain/recall previous social situations of isolation and/or rejection). The initial aversiveness associated with loneliness also motivates individuals to withdraw from social situations (i.e., behavioural reaffiliation process; Qualter et al., 2015). Although this may help individuals to observe social situations from a safe distance, assess their level of threat, and respond accordingly, if this process does not function as intended, it increases the risk of ‘behavioural confirmation processes’ (i.e., maladaptive behaviours, such as awkward, inappropriate, and/or avoidance responses that produce more negative social interactions and/or provide evidence to support maladaptive cognitive biases). A self-reinforcing loop of maladaptive cognitions and behaviours that maintain loneliness is consequently formed (Cacioppo & Hawkley, 2009; Hawkley & Cacioppo, 2010).

The processes proposed to cause and/or maintain chronic loneliness have received some empirical support in correlational studies (Spithoven et al., 2017). For instance, loneliness is associated with greater initial fixation on stimuli indicative of social threat (Bangee et al., 2014; Qualter et al., 2013) and faster differentiation between social and non-social threats (Cacioppo, Balogh et al., 2015a; Cacioppo et al., 2016), greater expectations of rejection (Qualter et al., 2013) and misinterpreting the intent of others as hostile (Lau & Kong, 1999; Prinstein et al., 2005), and a pattern of withdrawal from social situations (Watson & Nesdale, 2012). As support for these associations has been largely correlational, it is unclear whether these processes play a causal role in maintaining feelings of loneliness (Zagic et al., 2022a, b). Experimental paradigms that enable the target process to be manipulated in opposing directions, whilst measuring changes in lonely affect, are needed to determine the role of maladaptive cognitions and behaviours in causing and/or maintaining loneliness.

Using varying experimental procedures, maladaptive cognitions and behaviours have been found to play causal roles in the development and maintenance of different emotional conditions, such as anxiety disorders (Hofmann, 2007), mood disorders (Scherrer & Dobson, 2009), and post-traumatic stress disorders (Ehring et al., 2009). Focusing on the role of maladaptive cognitions specifically, these studies have shown that cognitive biases related to attention, interpretation, and/or memory result in increased emotional distress. For instance, a speech task has often been used to heighten feelings of anxiety in socially anxious individuals while measuring changes in underlying cognitive mechanisms that occur pre- and post-induction (i.e., catastrophic automatic thoughts and negative expectations of one’s personal appearance; Vassilopoulos et al., 2014). A similar approach has recently been applied to loneliness in which three studies used different approaches to induce feelings of loneliness (for a review, see Pels & Kleinert, 2017). Two of the induction procedures were successful in inducing loneliness (Cacioppo et al., 2006; Rotenberg & Flood, 1999) and one was not (Hu, 2009). The study by Hu (2009) attempted to induce loneliness by asking participants to recall and write down one of their loneliest experiences, which yielded non-significant effects. In contrast, the study by Rotenberg and Flood (1999) randomised participants into one of three mood inductions: lonely, sad, or neutral. Participants randomised into the loneliness induction were presented two items from the revised UCLA Loneliness Scale (i.e., “I lack companionship” and “I am no longer close to anyone”), asked to think about times they felt that way (two-minutes per item), and then recall aloud for one-minute the experience that made them feel most lonely. The results indicated that the loneliness mood induction led to greater levels of loneliness than the sad or neutral mood inductions (Rotenberg & Flood, 1999). Although this paradigm was able to induce loneliness, it did not control for baseline levels of depression or social anxiety (i.e., despite these being linked to loneliness; Cacioppo et al., 2015; Fung et al., 2017), did not examine whether the induction had effects on general negative affect, and did not examine changes in processes proposed to maintain loneliness.

Only one study to date has been able to successfully induce loneliness and measure changes in mechanisms proposed to maintain loneliness (Cacioppo et al., 2006). This study randomised participants into a “high lonely” or “low lonely” induction. Participants in the high lonely group were guided from relaxation to hypnosis and asked to think of a time during which they felt lonely and re-experience those feelings (e.g., “Think of a time in which you felt isolated. You felt lonely. Perhaps you felt like you just didn’t belong – that you had no friends”), whereas participants in the low lonely group were asked to think of a time when they felt a sense of belongingness and to re-experience those emotions (e.g., “Think of a time in which you felt a sense of belonging. Perhaps you were a member of a group. Perhaps you had a best friend with whom you felt you could share anything.”). The results indicated that participants in the high lonely group, relative to participants in the low lonely group, reported significantly greater increases in feelings of loneliness and fear of negative evaluation (i.e., cognitive bias) with large effect size (Cohen’s d = 2.12; Cacioppo et al., 2006). Although promising, baseline levels of depression and social anxiety were not controlled, and the induction process led to significant changes in most outcome variables assessed (e.g., shyness, sociability, negative mood, positive mood, anxiety, anger, optimism, self-esteem, social skills, and social support), suggesting the effects were general to negative affect and not specific to loneliness. A more targeted approach for inducing lonely mood is therefore needed.

Social identification refers to the extent to which an individual subjectively perceives that a group positively informs their self-identity (Postmes et al., 2013). It has been linked to greater symptom improvement among people with anxiety and depressive disorders (Cruwys et al., 2014), social anxiety disorder (Meuret et al., 2016), and eating disorders (McNamara & Parsons, 2016). Interventions designed to promote fit between an individual and group, and therefore social identification, have been shown to decrease feelings of loneliness and increase feelings of social connectedness (Haslam et al., 2016), the latter of which is defined as having reciprocally meaningful social contact (O’Rourke & Sidani, 2017). The notion of promoting social fit continues to gain interest in the extant literature, such that research has identified ‘comparative’ (i.e., degree to which members of a group are perceived to be more similar to one another than they are to other groups; Turner, 1985) and ‘normative’ fit (i.e., discrepancy between one’s expectations for a group and their reality; Oakes et al., 1991) as factors influential in social identification. Given the link between social identification and loneliness, and these constructs being inherently subjective (i.e., can be conceptualised as a cognition), the current series of studies aimed to test a novel paradigm to induce transient feelings of loneliness by manipulating social identification (i.e., comparative and normative fit) and measure subsequent changes in strength of belief of lonely thoughts, thereby providing causal evidence to support the models of loneliness (Cacioppo & Hawkley, 2009; Qualter et al., 2015). We tested a mock online chat paradigm in which participants entered two different chat conditions: one designed to increase feelings of loneliness by promoting social discrepancies, and the other designed to decrease feelings of loneliness by promoting social identification. Changes in negative affect (loneliness, depression, and anxiety) and the strength of belief in lonely thoughts were examined before and after the mood induction. It was hypothesised that the loneliness induction would lead to greater feelings of loneliness and stronger belief in lonely thoughts, and that the social connectedness paradigm would lead to weaker feelings of loneliness and weaker belief in lonely thoughts.

Method

Participants

Participants were undergraduate students enrolled in an introductory Psychology unit at Macquarie University, NSW, Australia, who signed up to the study in exchange for course credit. To reduce social desirability responses, participants were informed the study was about understanding students’ experiences of studying during the COVID-19 lockdowns. Participants read general information about the study, and exclusion criteria, prior to signing up for a testing session. A total of 60 participants was recruited, however, one participant from the experimental group was removed due to having incomplete data, leaving a total of 59 participants in the final sample (Experimental = 29, Control = 30). Participants had a mean age of 24.3-years (SD = 10.2; range = 18–58 years), were predominantly female (55.9%), and lived on average with at least three other people (52.5%). Participants were excluded (n = 0) if they reported clinically significant levels of current psychological distress, as assessed by a clinician administered (DZ) severity rating scale (i.e., “In the past two weeks, did you feel depressed, sad, empty, or hopeless? [rated 7 or more out of 10]” and “Over the last several months, have you been continually worried or anxious about a number of events or activities in your daily life? [rated 7 or more out of 10]).

Outcome Measures

Positive and Negative Affect

Affect was assessed using selected items from the 60-item expanded version of the Positive and Negative Affect Schedule (PANAS-X; Watson & Clark, 1994) that measured the key constructs of lonely, depressed, and anxious affect. Participants were asked to rate the extent to which they felt the following emotions “right now”: lonely (i. alone, ii. lonely, iii. socially disconnected), depressed (i. sad, ii. blue, iii. down), and anxious affect (i. afraid, ii. timid, iii. nervous, iv. ashamed). Each question was rated on a five-point scale (1 = “very slightly or not at all” to 5 = “extremely”, range: 10–50) and a mean score was calculated for each of the three emotional constructs. Although the PANAS-X has been shown to have strong psychometric properties (Watson & Clark, 1994), there is no psychometric data available for the adapted version used in the current study. The internal consistency for the three emotional constructs at baseline were: lonely affect (a = 0.80), depressed affect (a = 0.89), and anxious affect (a = 0.83).

Loneliness

Loneliness was measured using the three-item UCLA Loneliness Scale (UCLA-LS-3; Russell et al., 1980). Each question was rated on a three-point scale: 1 = hardly ever; 2 = some of the time; 3 = often. The items were summed to provide a total score. The scale has been shown to have sound psychometric properties (Hughes et al., 2004). The reliability in the current sample was adequate (a = 0.61).

Depression

Depression was measured using the Patient Health Questionnaire-9 (PHQ-9; Kroenke et al., 2001). The PHQ-9 is designed to assess the severity of depressive symptoms. Participants were required to rate the extent to which they had experienced a range of depressive symptoms over the past two weeks across nine questions rated on a four-point scale: 0 = not at all; 1 = several days; 2 = more than half the days; 3 = nearly every day. The items were summed to provide a total score (range: 0–27). The PHQ-9 has been shown to have good reliability (Cronbach’s α = 0.89), sensitivity (88%), and specificity (88%) for major depression (Kroenke et al., 2001). It has been validated across several populations (Beard et al., 2016; Patrick & Connick, 2019; Wang et al., 2014), including university students (Kim & Lee, 2019). The reliability in the current sample was sound (a = 0.83).

Social Anxiety

Social anxiety was measured using the Mini-Social Phobia Inventory (Mini-SPIN; Connor et al., 2001). The Mini-SPIN is a self-rated, three-item scale designed to screen for Social Anxiety Disorder. Each question was rated on a five-point scale: 0 = not at all; 1 = a little bit; 2 = somewhat; 3 = very much; 4 = extremely. The items were summed to provide a total score. The Mini-SPIN has been shown to have good reliability (Cronbach’s α = 0.83; Wiltink et al., 2017), sensitivity (88.7%), and specificity (90%) for social anxiety disorder (Connor et al., 2001). It has been validated across several populations, including university students (Mörtberg & Jansson Fröjmark, 2019). The reliability in the current sample was sound (a = 0.75).

Social Isolation

The Lubben Social Network Scale-6 (LSNS-6; Lubben et al., 2006) measures perceived social support received by family and friends and was used to assess social isolation. The LSNS-6 has two subscales relating to: (1) people connected by birth, marriage, or adoption (e.g., “How many relatives do you see or hear from at least once per month?”); and (2) friends (e.g., “How many friends do you see or hear from at least once a month?”). Each subscale consists of three-items related to the participant’s social network size and perceived supports. Items are rated on a six-point Likert scale from 0 (none) to 5 (nine or more). A total score is obtained by summing each of the items. The LSNS-6 has been shown to have sound psychometric properties (e.g., Cronbach’s α = 0.83; Lubben et al., 2006; Myagmarjav et al., 2019). The reliability in the current sample was sound (a = 0.77).

Lonely Automatic Thoughts

The Lonely Automatic Thoughts Scale (LATS), a self-report rating scale, was purpose built for this study. Items were generated by two psychologists (DZ, VW) based on common automatic negative thoughts reported by lonely clients. These items were then cross-checked with automatic negative thoughts reported in five clinical interviews of older adults (mean age = 70.8) with severe levels of self-reported loneliness. The interviews were conducted by trained Psychologists using the Anxiety Disorders Interview Schedule for DSM-5 (ADIS-5; Brown & Barlow, 2014) at the Centre for Emotional Health at Macquarie University. This resulted in a 15-item self-reported scale that measured the extent to which individuals believe a series of lonely thoughts pertaining to the absence of a supportive social network, that they are not understood by others, and beliefs they are socially rejected by others (Table 1). Participants were required to rate the extent to which they believed each thought “right now” on a five-point Likert scale, rated from 1 (not at all) to 5 (all the time). Higher scores on the LATS indicated higher levels of cognitive content associated with self-reported loneliness (range:15–75). The 15-item questionnaire demonstrated good reliability in the current sample (a = 0.92).

Table 1 Items of the loneliness automatic thoughts scale
Full size table

Social Identification Scale (SIS)

A five-item questionnaire was primarily utilised to discourage participants from identifying the true aim of the study. A secondary purpose of this questionnaire was to assess whether the chat paradigm had successfully manipulated normative and/or comparative fit. Participants were asked to rate the extent to which they agreed with the following statements about their experiences during each of the online chats: (1) “I got along with the other students” (normative fit), (2) “My academic experience during COVID-19 was similar to theirs” (comparative fit), (3) “My social experience during COVID-19 was similar to theirs” (comparative fit), (4) “I was able to add valuable information to the conversation” (normative fit), and (5) “I felt socially connected with the other students” (normative fit). Participants were required to rate these questions after each chat. Each question was rated on a five-point scale: 0 = not at all, 1 = a little bit, 2 = somewhat, 3 = a lot, 4 = completely true. The reliability in the current sample was good (a = 0.92).

Online Group Chat Paradigm

Two online chat conditions were utilised to induce the desired affect by manipulating comparative and normative fit (Cruwys et al., 2020). The loneliness induction chat was designed to elicit transient feelings of loneliness, whereas the social connectedness induction chat was designed to reduce transient feelings of loneliness. Across both conditions, two confederates (DZ and a trained research assistant) played the role of two other ‘students’ at Macquarie University.

Participants were instructed that they would be entering two separate meetings on Zoom where they were required to discuss the challenges related to their studies and social life that they had encountered during COVID-19 with other students from Macquarie University. While in the Zoom meeting, they were advised that the lead researcher (DZ) would be changing their display name to their initials to ensure anonymity. Accordingly, participants were advised not to disclose personal information and were required to turn off their video and audio during the Zoom meeting and to only use the chat feature in Zoom when interacting with the other ‘students’. They were also advised that the lead researcher would be present during the chat to ensure that everyone was participating. The participants were not aware that the other ‘students’ were confederates played by the lead researcher and trained research assistants. The effectiveness of the deception was assessed during the debrief by asking participants whether they suspected they were interacting with confederates and not other students at Macquarie University.

The content of the chat in each condition roughly followed a predetermined script, with minor adjustments made to incorporate idiosyncratic responses by the participant. The average length of both chat conditions was approximately 10-minutes. Both chats started with standard comments made by the confederates to introduce themselves and then to provide comment on their experiences during COVID-19 lockdowns along the following themes: their experience with the transition to online learning, social interactions with peers (online and in-person), and general coping ability during the lockdowns. The confederates either responded to participants by emphasising similarities in experiences, or discrepancies from the participant, therefore manipulating the chat towards the desired effect.

Loneliness Induction

The loneliness induction script aimed to increase the sense of disconnection (and loneliness) in the participant by cultivating a conversation where the confederates appeared more similar to each other (and less similar to the participant). For example, the confederates ‘discovered’ through the chat that they came from similar backgrounds to each other, studied similar degrees, had similar interests, and had similar experiences during COVID-19, however these were deliberately reported to be different to the participant. The loneliness induction script was also designed such that the confederates interacted more with each other, and less with the participant, as the online meeting progressed. Confederates were required to adhere to this script, however, if the participants deviated from the topic of conversation, the confederates were instructed to respond minimally and attempt to redirect the conversation back to the script.

Social Connectedness Induction

The social connectedness induction script aimed to increase the sense of connection (and decrease loneliness) between the participant and the confederates. In this script, the confederates ‘discovered’ through the chat that they were from similar backgrounds to the participant, studied similar degrees, had similar interests, and had similar experiences during COVID-19. The confederates aimed to include the participant in the online conversation and express an interest in the participant. As above, confederates redirected the conversation back to the script if any deviations arose.

Procedure

The study was granted ethics approval by Macquarie University’s Human Research Ethics Committee (52,022,969,837,211). Students expressed their interest in the study by responding to advertisements on the course online recruitment platform. Students were informed that the study was designed to examine student’s experiences of online learning during the COVID pandemic lockdowns via a group online chat with other undergraduate students. They were sent a link to a Zoom meeting on the day of their session. During the testing session, participants were asked additional screening questions to assess their eligibility. This included whether they were currently experiencing clinically significant levels of psychological distress, substance abuse, or suicidality. If eligible, participants were sent a link, via the chat feature on Zoom, to an online survey on Gorilla (https://gorilla.sc/). After providing informed consent, participants were required to answer questions about their demographics (age, gender, living arrangement, cultural background), affect (PANAS-X), loneliness (UCLA-LS-3), depression (PHQ-9), social anxiety (Mini-SPIN), social network size (LSNS-6) and lonely automatic thoughts (LATS). At the end of the survey, participants were randomised via a 1:1 ratio into either the experimental or control arm that comprised of two chat inductions. The randomisation process was conducted by the randomisation function available in Gorilla. Participants randomised into the experimental arm first entered a loneliness induction chat, and then after completion of post-induction measures, entered the social connectedness chat. This arm is subsequently referred to as the loneliness arm. Participants randomised into the control arm completed two separate social connectedness induction chats (i.e., did not receive the loneliness induction chat). This arm is subsequently referred to as the social connectedness arm. After completing each chat condition, participants completed the SIS, PANAS-X, and LATS. Participants were debriefed and advised of the true nature of the study once they had completed both chat conditions.

Data Analysis

A sample size of 60 was required to provide power of 0.8 (α = 0.05, two-tailed) for detecting a medium effect (Cohen’s d = 0.5). Data was analysed by comparing changes in outcome variables from baseline to after chat one and after chat two between groups (i.e., loneliness versus social connectedness) using a series of Pearson correlations, paired t-tests, and repeated measures ANCOVAs, controlling for baseline levels of loneliness, depression, social anxiety, and social network size. These outcomes were identified as covariates given their strong association with loneliness in the extant literature (Cacioppo et al., 2015b; Fung et al., 2017). Data was analysed using IBM SPSS Statistics (Version 28.0).

Results

Sample Characteristics at Baseline

Participants’ mean self-reported level of loneliness on the UCLA-LS-3 was 5.2 (SD = 1.6; moderate severity). The mean score for depression on the PHQ-9 was 9.3 (SD = 5.4; mild severity). Participants’ mean level of social anxiety (Mini-SPIN) was 4.3 (SD = 2.9; non-clinical). The average size of participants’ social network (LSNS-6) was 16.8 people (SD = 5.9; range = 4–30). These outcome variables were included in the analyses as covariates. After controlling for baseline levels of loneliness (UCLA-LS-3), depression (PHQ-9), social anxiety (Mini-SPIN), and size of social network (LSNS-6), repeated measures ANCOVA indicated a significant Time x Condition interaction for the SIS (F(1, 53) = 99.97, p < 0.001, ηp2 = 0.654). Paired t-tests indicated that participants in the loneliness arm had a significant increase on scores on the SIS from after the loneliness induction chat (mean = 5.1, SD = 3.3) to after the social connectedness induction chat (mean = 17.1, SD = 3.3; t(28) = -13.88, p < 0.001, 95% CI = -13.69 to -10.17), whereas participants in the social connectedness arm had non-significant changes in scores of the SIS from after the first social connectedness induction chat to after the second social connectedness induction chat (t(29) = -1.00, p = 0.33, 95% CI = -1.53 to 0.53). Independent samples t-tests indicated no significant differences at baseline between groups on lonely, depressed, and anxious affect (p’s < 0.05). There were also no significant differences between groups on loneliness (UCLA-LS-3), depression (PHQ-9), social anxiety (Mini-SPIN), size of social network (LSNS-6), and belief ratings of lonely automatic thoughts (LATS; Table 2). Across the sample, participants rated belief in each of the lonely thoughts from 0 (not at all) to 4 (all the time). The thoughts with the highest mean belief rating related to perceived unavailability of a supportive social network (i.e., item 5, 8, and 12–15) compared to other automatic thoughts (e.g., negative evaluation or rejection; Table 3).

Table 2 Baseline participant outcome data
Full size table
Table 3 Individual item mean (standard deviation) scores on LATS across time
Full size table

A series of Pearson correlations was conducted to examine the linear relationship between outcome variables. Belief ratings of lonely thoughts positively correlated with lonely affect (r = 0.74, p < 0.001), depressed affect (r = 0.57, p < 0.001), anxious affect (r = 0.26, p = 0.048), loneliness (r = 0.55, p < 0.001), symptoms of depression (r = 0.59, p < 0.001), symptoms of social anxiety (r = 0.41, p < 0.001), and size of social network (r = 0.55, p < 0.001). Lonely affect negatively correlated with social network size and positively correlated with all other variables, whereas depressed and anxious affect positively correlated with all variables except for social network size. The remaining correlations are described in Table 4.

Table 4 Pearson correlations among outcome variables
Full size table

Positive and Negative Affect

After controlling for baseline levels of loneliness (UCLA-LS-3), depression (PHQ-9), social anxiety (Mini-SPIN), and size of social network (LSNS-6), repeated measures ANCOVA indicated non-significant main effects of Time for lonely (F(1, 53) = 0.01, p = 0.99, ηp2 = 0.000), depressed (F(1, 53) = 1.58, p = 0.21, ηp2 = 0.029), and anxious affect (F(1, 53) = 2.07, p = 0.14, ηp2 = 0.037). A significant Time x Condition interaction was observed for lonely (F(1, 53) = 10.14, p < 0.001, ηp2 = 0.161) and depressed affect (F(1, 53) = 3.80, p = 0.03, ηp2 = 0.067), but not anxious affect (F(1, 53) = 1.41, p = 0.25, ηp2 = 0.026; Fig. 1), and follow-up analyses were conducted to further explore these effects.

Fig. 1
figure 1

Change in affect and belief rating of lonely automatic thoughts between groups. (a) Belief rating of lonely thoughts over time (b) Lonely affect over time (c) Depressed affect over time (d) Anxious affect over time

Full size image

Paired t-tests indicated that for participants in the loneliness arm there were significant increases in lonely affect from baseline (mean = 2.44, SD = 0.96) to after the lonely induction chat (mean = 2.77, SD = 0.87; t(28) = -2.25, p = 0.03, 95% CI = -0.64 to -0.03), and non-significant changes in depressed affect (t(28) = 0.79, p = 0.44, 95% CI = -0.20 to 0.46). This indicates that the loneliness induction chat specifically induced lonely affect (and not depressed or anxious affect). Further a significant reduction in lonely (t(28) = 6.09, p < 0.001, 95% CI = 0.73 to 1.47) and depressed affect (t(28) = 3.77, p < 0.001, 95% CI = 0.23 to 0.78) was observed from after chat one (lonely) to after chat two (social connectedness induction) indicating that the subsequent social connectedness chat resulted in a significant decrease in lonely and depressed affect.

Paired t-tests indicated that for participants in the social connectedness arm there were significant reductions in lonely (t(29) = 3.06, p = 0.01, 95% CI = 0.14 to 0.69), depressed (t(29) = 4.77, p < 0.001, 95% CI = 0.32 to 0.81), and anxious affect (t(29) = 4.39, p < 0.001, 95% CI = 0.20 to 0.55), from baseline to after the social connectedness induction (chat one). A significant reduction in lonely (t(29) = 2.14, p = 0.04, 95% CI = 0.01 to 0.33), depressed (t(29) = 2.44, p = 0.02, 95% CI = 0.02 to 0.27), and anxious affect (t(29) = 2.11, p = 0.04, 95% CI = 0.003 to 0.20) was observed from after the first social connectedness induction to after the second social connectedness induction (chat two; see Table 5 for means and standard deviations).

Table 5 Mean (standard deviation) scores on lonely, depressed, and anxious affect, and belief rating of lonely automatic thoughts, across time
Full size table

Lonely Thoughts

After controlling for baseline levels of loneliness (UCLA-LS-3), depression (PHQ-9), social anxiety (Mini-SPIN), and size of social network (LSNS-6), repeated measures ANCOVA indicated a non-significant main effect of Time for belief rating of lonely automatic thoughts (F(1, 53) = 1.14, p = 0.32, ηp2 = 0.021). A non-significant Time x Condition interaction was also observed (F(1, 53) = 2.67, p = 0.07, ηp2 = 0.048; Fig. 1). A series of repeated measures ANCOVAS were conducted to examine the LATS at the item level to see if particular lonely thoughts changed overtime. These were all associated with non-significant differences (p > 0.05).

Discussion

Models of psychopathology emphasise the role of cognitive biases in maintaining distress (e.g., depression and social anxiety; Beck, 1967; Rapee & Heimberg, 1997). Similarly, maladaptive cognition is hypothesised to maintain chronic loneliness (Cacioppo & Hawkley, 2009; Qualter et al., 2015). In this study, we examined the impact of manipulations of perceived fit on affect and belief ratings of lonely automatic thoughts. Although the online group chat paradigm utilised in the current study was associated with specific increase in lonely affect, this was associated with non-significant changes in belief ratings of lonely automatic thoughts. Therefore, it remains unclear whether maladaptive cognition affects loneliness. This null finding contrasts previous research that has emphasised the importance of targeting maladaptive cognition in the treatment of loneliness (Zagic et al., 2022c) and empirical research that has demonstrated that cognitive restructuring of lonely automatic thoughts alleviates loneliness (Zagic et al., 2022). There are several potential reasons for the observed null effect.

One such account is that maladaptive cognition may be a partial, but not necessary cause of loneliness, particularly given that models of loneliness hypothesise that several different processes in combination cause and/or maintain loneliness (Cacioppo & Hawkley, 2009). The nature of the paradigm was such that it attempted to manipulate feelings of loneliness by eliciting change in comparative and/or normative fit. Although our data analysis indicated that the experimental paradigm was able to achieve this manipulation for the group who first received the loneliness induction followed by the social connectedness induction, (i.e., significant Time x Condition interaction on the SIS), there was no change for the group on receiving the second social connectedness induction. As comparative and normative fit was not measured prior to the chat (that was no feasible) we are unsure if the experimental chat also led to reductions in comparative and normative fit in the loneliness chat condition. Importantly, the SIS was purpose built for the study (primarily to maintain deception of the study aims) and lacks psychometric validation, so therefore it is difficult to determine whether these processes were adequately manipulated and future research with a psychometrically validated measure of comparative and/or normative fit is needed. Similarly, loneliness in the current study was assessed using only three items (i.e., UCLA-LS-3), which was associated with low internal consistency (a = 0.61), emphasising the need for a more appropriate measure of loneliness in replications of the current experimental paradigm. The average scores for the UCLA-LS-3 fell into the moderately severe range (mean = 5.2, SD = 1.6), despite participants having self-reported relatively large social network sizes (mean = 16.8, SD = 5.9), emphasising the subjective nature of loneliness but also identifying the potential of the former to have led to ceiling effects, whereas the latter may have offset the effects arising from the experimental manipulation.

However, a more likely interpretation is the distinction between transient versus chronic loneliness (Qualter et al., 2015). The nature of the experimental paradigm in the current study was such that participants interacted with confederate for approximately 10-minutes, and as such, it likely only elicited transient increases in loneliness. Negative automatic thoughts related to the perceived unavailability of a supportive social network, therefore, may be more relevant to chronic than transient loneliness. Furthermore, items of the LATS were worded in such a way that they captured chronic loneliness, and therefore, it is unlikely to have been impacted by the transient experimental manipulation of the current study, especially as the individual items within the LATS may reflect underlying schema. In fact, the items generated as part of the LATS itself were associated with limitations, namely that there was unintended overlap between some items of the LATS and those observed in the UCLA-20. Moreover, the current version of the LATS blends items that may reflect automatic negative thoughts (e.g., “I have no one to talk to”) versus items that are more like factual statements (e.g., “I am often rejected by others”). Future research in a larger sample of participants is needed to examine the impact of a more sustained induction of loneliness, or the same induction paradigm but in lonelier participants (i.e., the group mean on the UCLA-LS-3 for participants fell into the moderately lonely range), using a more appropriate measure of cognition in loneliness.

Furthermore, the specific maladaptive cognitions that underlie loneliness are not yet well established, and as such, the LATS measured cognitions thematically related to unavailability of a supportive social network, that individuals are not understood by others, that individuals are socially rejected, and that one cannot cope with being alone, which may have resulted in too broad a measurement. An alternative approach could have been to assess changes on specific cognitions, which may have been more informative, and is another direction for future research once a more psychometrically sound measure of cognition in loneliness is designed.

Loneliness is operationalised as lying on a continuum of quality of social connections, which extends from poor (i.e., loneliness) to strong (i.e., social connectedness). For instance, scales for loneliness correlate negatively with scales for social connectedness and high levels of social connectedness have a negative relationship with loneliness (Satici et al., 2016). This implies that if inducing loneliness should increase belief rating of lonely automatic thoughts, inducing social connectedness should decrease belief rating of lonely automatic thoughts. The results from the study, namely that there were non-significant changes in belief ratings of lonely automatic thoughts, were not consistent with this hypothesis. Instead, the social connectedness induction chat led to significant reductions in lonely and depressed affect (trend effect for anxious affect; p = 0.06) for participants in the loneliness arm, and significant reductions in lonely, depressed, and anxious affect for participants in the social connectedness arm. Therefore, the social connectedness induction chat led to significant reductions in general negative affect, not lonely affect, and as such, the non-significant effect for changes in belief ratings of lonely automatic thoughts may reflect the fact that this social connectedness chat had broad effects on affect, instead of specific effects on lonely affect.

Recent research suggests that loneliness is linked to depression (Courtin & Knapp, 2017; Hawkley & Cacioppo, 2010), and social anxiety (Lim et al., 2016; Meltzer et al., 2013), with these disorders being maintained by similar processes (Beck, 1967; Rapee & Heimberg, 1997). Despite these associations, previous loneliness induction paradigms have not controlled for baseline levels of depression and social anxiety (Cacioppo et al., 2006; Rotenberg & Flood, 1999). The current study fills this gap by providing evidence for specificity of effect, such that inducing lonely affect by manipulating comparative and normative fit leads to significant increases in lonely affect, but not depressed or anxious affect. This also provides evidence to support the notion that loneliness is distinct from depression and social anxiety (Cacioppo et al., 2015b; Fung et al., 2017). Future research should continue to examine the distinctness of loneliness from related constructs like depression and/or social anxiety to better inform effective treatment.

The findings of the current study should be interpreted considering their limitations. Firstly, the study recruited undergraduate students enrolled in an introductory Psychology unit (mean age of 24.3-years) who were not experiencing significant levels of psychological distress, and therefore the results lack generalisability. Future research should replicate the current experimental paradigm across the lifespan, and in fact, the effectiveness of the induction paradigm may be augmented with the use of a lonelier sample. This is particularly relevant given that the loneliness induction chat was not able to elicit significant changes in lonely automatic thoughts. Secondly, cognition was assessed using a measure that lacked standardisation and psychometric evaluation, which was also worded in such a manner that would not detect transient changes in cognition, and this should also be addressed in future research by incorporating a measure of cognition in transient loneliness that is psychometrically sound. Thirdly, the paradigm utilised in the current study was such that any changes in affect and cognition were likely transient, and therefore, future research is needed to determine whether similar effects are observed when more chronic feelings of loneliness are induced. Lastly, comparative and normative fit were assessed by a measure developed by the study authors, making it unclear whether the experimental paradigm was able to successfully manipulate these constructs. A more appropriate measure should be used in future research.

Despite these limitations, the findings of the current study have important clinical implications. They provide preliminary data to support a novel paradigm of inducing transient feelings of lonely affect by manipulating comparative and normative fit (controlling for baseline depression and social anxiety), and therefore, a potential means by which for future research to empirically examine cognitive and behavioural mechanisms underlying loneliness. Research of this nature is imperative to advance our understanding of the mechanisms involved in causing and/or maintaining loneliness as previous research has been largely correlational (Zagic et al., 2022a, b.). However, these effects should be interpreted with caution as changes in lonely affect in the current study were likely transient, and non-significant changes in lonely automatic thoughts were observed. Future research is needed to address the limitations described above and examine whether there are ways of augmenting the effectiveness of the induction paradigms described in the current study (e.g., utilising a lonelier sample of participants). Furthermore, findings of the current study have significant clinical implications, such that manipulating comparative and normative fit so that individuals are made to think they have similar backgrounds, interests, and experiences to confederates, provides a benefit for general affect, with these effects cumulating over time. This provides a promising new avenue for targets in the treatment of loneliness.

Conclusion

Experimentally manipulating comparative and normative fit to either emphasise differences or similarities between an individual and group members is an effective paradigm for increasing lonely affect or decreasing general negative affect, respectively. This provides a novel approach by which to empirically examine the cognitive and behavioural processes proposed to underlie loneliness (Cacioppo & Hawkley, 2009; Qualter et al., 2015). Future research is needed to explore ways in which the effectiveness of paradigm may be augmented.