The positive and negative emotion functions related to loneliness: a systematic review of behavioural and neuroimaging studies

Abstract Loneliness is associated with high prevalences of major psychiatric illnesses such as major depression. However, the underlying emotional mechanisms of loneliness remained unclear. We hypothesized that loneliness originates from both decreases in positive emotional processing and increases in negative emotion processing. To test this, we conducted a systematic review of 29 previous studies (total participants n = 19 560, mean age = 37.16 years, female proportion = 59.7%), including 18 studies that included questionnaire measures of emotions only, and 11 studies that examined the brain correlates of emotions. The main findings were that loneliness was negatively correlated with general positive emotions and positively correlated with general negative emotions. Furthermore, limited evidence indicates loneliness exhibited negative and positive correlations with the brain positive (e.g. the striatum) and negative (e.g. insula) emotion systems, respectively, but the sign of correlation was not entirely consistent. Additionally, loneliness was associated with the structure and function of the brain emotion regulation systems, particularly the prefrontal cortex, but the direction of this relationship remained ambiguous. We concluded that the existing evidence supported a bivalence model of loneliness, but several critical gaps existed that could be addressed by future studies that include adolescent and middle-aged samples, use both questionnaire and task measures of emotions, distinguish between general emotion and social emotion as well as between positive and negative emotion regulation, and adopt a longitudinal design that allows us to ascertain the causal relationships between loneliness and emotion dysfunction. Our findings provide new insights into the underlying emotion mechanisms of loneliness that can inform interventions for lonely individuals.


Introduction
Loneliness refers to an indi vidual's subjecti ve perce ption of unsatisfied social needs and relationships (Cacioppo et al., 2015 ).Loneliness is a pr e v alent condition that affects populations of all ages (Cacioppo et al., 2015 ;Loades et al., 2020 ;Van As et al., 2022 ).Substantial evidence indicates close associations between high loneliness le v els and v arious major physical and psychological illnesses, such as dementia, cancer, cardiovascular diseases, major depression, and anxiety disorders (Hawkley and Cacioppo, 2010 ;Ernst et al . 2021 ;Shen et al., 2022 ;Liang et al., 2023 ).Ther efor e, high rates of loneliness pose significant health threats and burdens to both individuals and modern society (Holt-Lunstad et al., 2015 ).It is widel y consider ed that loneliness originates from and exacerbates negative emotions about social relationships and inter actions, whic h cause lonely individuals to withdraw and disconnect from social networking (Cacioppo et al., 2015 ).Howe v er, the association between loneliness and positive emotions still remains lar gel y unclear.Elucidating the r elationships between loneliness and both positive and negative emotions is fundamental to building an affective framework of loneliness, which conceptualizes the de v elopment of loneliness based on both r e w ar d and pun-ishment systems .T his knowledge ma y then inform interventions targeted at reducing feelings of loneliness.
The association between high loneliness and negative emotions has been long established (Masi et al., 2011 ;Cacioppo et al., 2014 ).Pr e vious r esearc h sho w ed that negati ve and de pressi ve emotions were longitudinally and positively related to high loneliness le v els (Cacioppo et al., 2010 ).For example, according to a 5-year longitudinal study, loneliness pr ospectiv el y pr edicted future onset of de pressi ve symptoms (Cacioppo et al., 2010 ).Moreover, in a longitudinal study, it was discov er ed that among adults aged 50 years and older, elevated baseline loneliness scores were pr ospectiv el y associated with gr eater se v erity of depr essiv e symptoms at 12-year follow-up (Lee et al., 2021 ).Consistent with these findings, a negative emotion model of loneliness was pr e viousl y pr oposed, whic h posited that individuals with high loneliness levels tend to exhibit automatic negative biases to w ar ds social interactions and relationships (Hawkley and Cacioppo, 2010 ).As a consequence, despite desiring social connections, these individuals choose to avoid and withdraw from social occasions, ultimately becoming socially isolated (Holt-Lunstad et al., 2015 ;Bzdok and Dunbar, 2020 ).This model is supported by existing evidence Figur e 1: T he bivalence emotion model of loneliness .T he bivalence emotion model of loneliness suggests that changes in both positive and negative emotion systems can lead to loneliness .Furthermore , different br ain r egions ar e involv ed in pr ocessing v arious emotional aspects of loneliness .For instance , the striatum and VMPFC handle the positive emotions, while the insula and amygdala are responsible for processing the negative emotions.Existing evidence indicates a closer connection between loneliness and negative emotions, whereas the connection to positive emotions is less clear.This figure was created using materials obtained from BioRender.com,with full license granted.
On the other hand, the relationship between loneliness and positive emotions has received less attention.It is wellestablished that human decisions and behaviours are governed by both negative and positive emotion systems (Cardi et al., 2015 ).T hus , it is reasonable to presume that loneliness also has both positive and negative emotion components (Fig. 1 ).Previous work indicated that lonely individuals may exhibit blunted positive emotions, which is linked with a reduced capacity to perceive positive cues and enjoy positive experiences within social settings (Lieberz et al., 2021 ).According to this view, loneliness arises from a combined influence of reduced positive emotions and heightened negative emotions, particularly in social contexts (i.e. a dual emotion model).On the other hand, it could also be that lonely individuals show normal le v els of positiv e emotional pr ocessing (Wong et al ., 2019a ), yet they withdraw from social interactions due to excessive negative emotions, as outlined before (i.e. a negative emotion model).To our knowledge, existing literature has not provided a direct answer to these alternative frameworks.
While pr e vious e vidence on the neur oima ging mec hanisms of loneliness is limited (Wong et al., 2022 ), it is likely that loneliness is linked to altered functioning of brain circuitries involved in emotion processing, emotion regulation, and social cognition functions.Based on extensive previous literature, the positive emotion system includes the v entr omedial pr efr ontal cortex (VMPFC) and the striatum (Doré et al., 2017 ).The negative emotion system includes the amygdala and insula (Shiba et al., 2017 ).The emotion regulation system includes the later al pr efr ontal cortex (LPFC) (Hooker et al., 2010 ;Townsend et al., 2013 ), the anterior cingulate cortex (ACC) (Ochsner et al ., 2012 ), and the hippocampus (Doré et al., 2018 ;Barch et al., 2020 ), whereas the social cognition system includes the dorsomedial pr efr ontal cortex (DMPFC) (Eic khoff et al., 2016 ).Among these regions, the striatum and VMPFC ar e mainl y r esponsible for positiv e emotional pr ocessing (Hiser and Koenigs 2018 ;Filimon et al., 2020 ), the amygdala and insula ar e mainl y involv ed in negativ e emotional pr ocessing (Goldin et al., 2008 ;Shiba et al., 2017 ;Stew ar d et al., 2022 ), the LPFC and hippocampus are involved in top-down emotion regulation, and the DMPFC is involved in social cognition (Townsend et al., 2013 ;Lieberman et al., 2019 ;Arbula et al., 2021 ).These systems have been explained by extensive literature that largely agrees on the cor e br ain cir cuitries inv olved in emotion reacti vity (positi ve and negativ e), emotion r egulation, and social pr ocessing (e.g.Oc hsner et al ., 2012 ;Roy et al ., 2012 ;Etkin et al ., 2015 ;Whittaker et al ., 2018 ;Ferrari et al ., 2016 ;Martin et al ., 2017 ).Nevertheless, it should be noted that any categorization of regions into functional systems is ine vitabl y subject to contention, and these systems mer el y serv e as a provisional guide for our review of the loneliness-related neur oima ging liter atur e r ather than impl ying definitiv e functions of an y br ain r egion.As of now, it is still unclear how past neur oima ging studies on loneliness can inform whether loneliness is associated with altered functions of both positive and negative emotion systems, or whether it is primarily related to negative emotional function changes.
Ther efor e, this systematic r e vie w synthesized pr e vious behavioural and neuroimaging studies on the association between loneliness and both positive and negative emotions to clarify the difference between the 'dual emotion model' and the 'negative emotion model' fr ame works of loneliness.In the pr esent context, loneliness le v els wer e measur ed by differ ent v ersions of the UCLA loneliness scale (ULS).The original version of the ULS comprises 20 items with four-point Likert scales, and assesses the indi vidual's percei ved subjecti ve feelings of social isolation (Russell, 1996 ).Subsequent r esearc h has v alidated short versions of the ULS such as ULS-8, ULS-6, and ULS-3 across populations and cultures (Hays and DiMatteo, 1987 ;Russel 1996 ;Hughes et al ., 2004 ;Neto, 2014 ;Lin et al ., 2022 ).

Methods
We conducted this r e vie w following the Pr eferr ed Reporting Items for Systematic Re vie ws and Meta-Anal yses (PRISMA) guidelines (P a ge et al., 2021 ).This r e vie w pr otocol was pr e-r egister ed at PROS-PERO, International pr ospectiv e r egister of systematic r e vie ws (r efer ence number CRD42023445805).The PRISMA c hec klist can be found in Supplementary Table S1 .Mor eov er, we conducted a compr ehensiv e r e vie w of the r efer ence lists of r ele v ant articles to identify additional studies .T he inclusion of identified studies was verified by two authors (R.S. and Q.L.).The entir e liter atur e selection pr ocess is depicted in Fig. 2 .We only included studies that met the following criteria: (i) use of the ULS (ULS-20, ULS-3, ULS-6, ULS-8) to assess loneliness (Hays and DiMatteo, 1987 ;Russel 1996 ;Hughes et al ., 2004 ;Neto, 2014 ) Beck et al ., 1988 ) or task paradigms, (iii) publication in English, (iv) inclusion of at least 10 human participants (Lam et al., 2021 ), and (v) provision of quantitative results on the relationship between loneliness and emotions.Animal studies, liter atur e r e vie ws, and confer ence abstr acts wer e excluded.We also excluded studies involving samples of special c har acteristics (e.g. military soldiers) or major physical illnesses (diabetes) ( Supplementary Table S2 ).

Search str a tegy and study selection
The searc h r esulted in a total of 2027 articles of interest (1048 articles from Web of Science, 463 articles from PsycINFO, and 516 from PubMed).After eliminating 568 duplicate articles, in the first stage of screening, the titles and abstracts of the remaining 1459 studies were evaluated by two authors (Q.L. and R.S.).Studies that met any of the exclusion criteria as outlined before were screened out.A total of 1011 studies were excluded during the first stage of screening.The full text of the remaining 448 studies was further assessed to r esolv e an y r emaining uncertainty.Disa gr eements during this phase wer e r esolv ed thr ough discussions between the two authors (Q.L. and R.S.).Finally, 29 studies published between 2010 and 2023 were chosen to be reviewed in this study (Fig. 2 ).Among these, 18 studies provided findings only on the association between loneliness and behaviour al/questionnair e measur es of emotions, 11 studies provided findings only on the association between loneliness and neur oima ging substr ates of emotions only, and six studies provided findings on the association between loneliness and both behaviour al/questionnair e and neur oima ging measures of emotions.

Da ta extr action
Data extraction was independently performed by two authors (Q.L. and R.S.).For each eligible article, we extracted the following information: author(s), year of publication, participant characteristics (dia gnosis, a ge, sex r atio, and sample size), mood measur ement, loneliness measur ement, and main findings (Tables 1  and 2 ).

Quality assessment
The articles that met the eligibility criteria were assessed on quality using the BIOCROSS e v aluation tool (Wirsching et al., 2018 ), specifically designed to evaluate the quality of biomarker-based cross-sectional studies .T he BIOCROSS tool comprises ten items grouped into five domains: 'Study rationale', 'Design/Methods', 'Data analysis', 'Data interpretation', and 'Biomarker measurement'.Each item has three 'issues to consider' that e v aluate the quality of the study, and scores of 0, 1, and 2 are assigned depending on the number of issues reported.On the basis of the obtained score, the quality of the studies was classified as low, moderate, or high, with scores of ≤6 indicating low quality, scores of 7-12 indi-  cating moderate quality, and scores of ≥13 indicating high quality.
The results of the quality assessment are presented in Tables 1  and 2 and Supplementary Table S3 .

Characteristics of included studies
Out of 2027 records in total, 29 papers were included in this systematic r e vie w.Eighteen studies pr ovided onl y behaviour al r esults (four tested the relationship between loneliness and positive emotion, se v en tested the relationship between loneliness and negativ e emotion, se v en tested the r elationship between loneliness and both positive and negative emotions).Five studies provided onl y neur oima ging r esults (one tested the r elationship between loneliness and positive emotions, one tested the relationship between loneliness and negative emotions, three tested the relationship between loneliness and both positive and negative emotions).
Six studies provided both behavioural and neuroimaging results (one tested the relationship between loneliness and positive emotions, two tested the relationship between loneliness and negative emotions, three tested the relationship between loneliness and both positive and negative emotions).Among the 11 studies that pr ovided neur oima ging r esults, two studies r eported onl y the participants' gray matter measurements, one study reported only the participants' white matter connectivity, four studies reported participants' task-based activity and connectivity, three studies reported participants' resting-state functional connectivity, and one stud y re ported both white matter and task-based connecti vity.

Study quality assessment
The BIOCROSS tool was used to assess the quality of the 29 studies .Among these , fiv e studies wer e r ated as high quality ( ≥13 score), while 24 studies were rated as having moderate quality (7-12 score).No study was rated as low quality ( < 7 score).

Positive emotions
A total of 10 studies investigated the association between loneliness and positive emotions.Among these, one study was conducted in adolescents (Baytemir and Yildiz, 2017 ), four studies were conducted in young adults (Durak and Senol-Durak, 2010 ;Ditc he v a et al ., 2018 ;Satici 2019 ;Tang et al ., 2022 ), one was conducted in middle-aged adults (Ben-Zur 2012 ), and four were conducted in older adults (Steptoe et al ., 2011 ;Neto 2014 ;Davidson et al ., 2022 ;Donizzetti and Capone 2023 ).All 10 studies reported a negative relationship between loneliness and positive emotion among healthy individuals, although one study additionally found that this relationship may be reversed following social stress (as detailed next).
Among adolescents , Ba ytemir and Yildiz ( 2017 ) ( n = 263, mixedsex) sho w ed that loneliness as measured using the ULS-8 was negativ el y associated with gener al positiv e emotions as measured using the Positive and Negative Affect Scale (PANAS).This was the only study in adolescents that tested and demonstrated that loneliness was associated with reduced general positive emotions.
Among young adults, two studies also reported a negative relationship between loneliness as measured using the ULS-8 (Satici 2019 , n = 280, mixed-sex) or the ULS-20 (Durak and Senol-Durak, 2010 , n = 481, mostly females), and general positive emotion as measured using the PAN AS.A thir d study (Tang et al ., 2022 , n = 177, mostly females) measured participants' momentary positive emotions, and also found that loneliness (measured using ULS-20) was negativ el y associated with positive emotions such as excitedness and cheerfulness .T he fourth study measured participants' momentary positive emotions both at baseline and after acute stress induction using the Trier Social Stress Test (Ditcheva et al ., 2018 , n = 196, mixed-sex).This study again found that loneliness (measured using ULS-20) was negativ el y associated with momentary general positive emotion as measured using the PANAS.Ho w e v er, it was further found that following acute stress, lonelier individuals exhibited a greater increase of positive emotions compared to less lonely individuals, whereas loneliness was associated with less increase of positive emotions following the nostr ess contr ol condition.This finding highlights an intricate interplay between trait loneliness and situational social a ppr aisal on position emotional states.
Among middle-aged adults, one study (Ben-Zur 2012 , n = 196, mixed-sex) used a sample of married, div or ced, and wido w ed individuals, and sho w ed that loneliness as measured using the ULS-20 was negativ el y associated with dispositional optimism.
Among older adults, similar to the pr e vious r esults in younger samples, two studies both showed that loneliness as measured using the ULS-6 (Neto 2014 , n = 1154, mostly females) or the ULS-20 (Donizzetti and Capone 2023 , n = 1301, mixed-sex) was negativ el y associated with gener al positiv e emotions as measured using the PANAS.The latter study additionall y r e v ealed a negativ e r elationship between participants' loneliness le v el and pastmonth wellbeing.Mor eov er, a lar ge-sample stud y (Ste ptoe et al ., 2011 , n = 4258, mixed sex) re ported a negati v e corr elation between loneliness as measured using the ULS-3, and momentary positive emotions such as happiness and excitedness.Last, one recent study (Davidson et al ., 2022 , n = 428, mostly males) sho w ed that loneliness measured using the ULS-20 was negativ el y associated with past-week happiness level.

Summary
Ov er all, past studies generated consistent results that loneliness was associated with reduced positive emotions in both sexes .T he sample sizes of the studies were mostly large.Most studies examined gener al positiv e emotions using the PANAS scale, while a small number of studies examined specific positive emotions suc h as ha ppiness , excitedness , and life satisfaction.Most studies examined positive emotions within the past week, while the remaining studies looked at momentary or dispositional positive emotions.Ho w e v er, all the existing studies were cross-sectional, thus the direction of the loneliness-positive emotion relationship r emains unclear.Also, ther e was a paucity of studies that mined specific positiv e emotions, particularl y positiv e social emotions such as a feeling of unity and connectedness .Furthermore , it remains to be tested how the relationship between loneliness and positive emotions may depend on situational factors such as the social environment and stress level.
Among adolescents , Ba ytemir and Yildiz ( 2017 ) ( n = 263, mixed-sex) sho w ed that loneliness as measured using the ULS-8 was positiv el y associated with anxiety and de pressi ve symptoms as measured using the Social Anxiety Scale for Adolescents and Childr en's Depr ession Inv entory, r espectiv el y.This was the only study in adolescents that tested and demonstrated that loneliness was associated with increased anxiety and depression.
Similarly, in y oung adults, tw o studies found that loneliness as measured using the ULS-20 or ULS-8 was positively correlated with de pressi ve symptoms in the past 2 weeks (Durak and Senol-Durak, 2010 ; n = 481, mostly females; Shi et al ., 2023 , n = 4191, mixed-sex).Mor eov er, one study r eported a positiv e corr elation between loneliness as measured using ULS-20, and social anhedonia le v el (Tan et al ., 2020 , n = 824, mostly females).Last, one study sho w ed a positive correlation between loneliness as measured using ULS-8, and gener al negativ e emotions in the past week (Satici 2019 , n = 280, mixed-sex).
In middle-aged adults, it was again sho w ed that loneliness as measured using the ULS-20 was positively correlated with depression, anxiety, and stress symptoms (Velotti et al ., 2021 ; n = 1323, mostly females).Ho w ever, another study only reported a significant positive correlation between loneliness as measured using ULS-6 and stress symptoms, but not with depression or anxiety symptoms (Mik oc ka-Walus et al ., 2021 , n = 2110, mostly females).The difference in findings might be caused by the latter study specifically testing the relationship between loneliness and negative symptoms under the COVID-19 pandemic.Furthermore, one study sho w ed that loneliness as measured using ULS-20 was positiv el y corr elated with gener al negativ e emotions (Ben-Zur 2012 ; n = 196, mixed-sex), and another study sho w ed that loneliness as measured using ULS-8 was positively associated with distress emotions to the COVID-19 pandemic (Wegner and Liu, 2023 , n = 1354, mostly females).
Among older adults, similar to the findings in younger samples, it was shown that loneliness as measured using ULS-3 was positiv el y corr elated with depr essiv e symptoms in the past week (Steptoe et al ., 2011 ; n = 4258, mixed-sex), while loneliness as measured using UCLA-20 was positively correlated with social anxiety le v el in the past week (Hoffman et al ., 2021 , n = 120, mixed-sex).Two other studies both found that loneliness as measured using ULS-6 or ULS-20 was positiv el y associated with gener al negativ e emotions in the past week (Neto 2014(Neto , n = 1154, mixed-sex;, mixed-sex;Serra et al., 2021 , n = 61, mixed-sex).Last, one stud y re ported that loneliness as measured using ULS-20 was positiv el y associated with momentary negative emotions (Badal et al ., 2022 , n = 22, mostly females).

Summary
A considerable body of evidence supports a positive relationship between loneliness and negative emotions, particularly in adults and older individuals.Most of the studies have included relatively large sample sizes ( n > 200).Wher eas man y studies included samples mor e r epr esentativ e of females, se v er al studies used mixedsize samples and obtained similar findings .T he association between loneliness and (dis)stress emotions was consistently reported, while the association between loneliness, depression, and anxiety symptoms was mostly found.Notably, while almost all studies assessed r elativ el y short-term negative moods or symptoms, one study (Tan et al ., 2020 ) reported a significant positive correlation between loneliness and long-term social anhedonia, which indicates that loneliness may be linked with a dispositional de pressi ve trait.It is also worth noting that there was no study that measur ed negativ e emotions using behaviour al par adigms, which could be a direction for future studies .Moreo ver, the lack of longitudinal or intervention studies precluded any inferences on the direction of causality.

T he associa tion of loneliness and emotion-rela ted br ain circuitries
A total of 11 studies examined the association between loneliness and brain circuitries associated with emotional functions.All these studies measured loneliness using ULS-20.The positive emotion system includes the VMPFC and the striatum (Doré et al., 2017 ).The negative emotion system includes the amygdala and insula (Shiba et al., 2017 ).The emotion regulation system includes the LPFC and the hippocampus (Hooker et al., 2010 ;Townsend et al., 2013 ), while the social cognition system includes the DMPFC (Townsend et al., 2013 ;Lieberman et al., 2019 ;Arbula et al., 2021 ;Li et al. , 2022 ).W ithin the 11 studies, four studies examined the correlates of loneliness and the positive emotion system (Cacioppo et al ., 2009 ;Sin et al ., 2018 ;Inagaki et al ., 2016 ;Lieberz et al ., 2022 ), whereas two articles investigated the correlates between loneliness and the negative emotion system (Wong et al ., 2016 ;Wong et al ., 2019b ).Furthermore, seven studies tested the correlates of loneliness and the emotional regulation system (Cacioppo et al ., 2009 ;Tian et al ., 2017 ;Ohtsubo et al ., 2020 ;Shao et al ., 2020 ;Dong et al., 2021 ;Lieberz et al ., 2022 ;Liu et al ., 2016 ), whereas one study tested the correlates of loneliness and the social cognition system (Cacioppo et al ., 2009 ).It should be noted that we were unable to r e vie w the r elationship between loneliness and certain k e y r egions involv ed in emotion r egulation functions, suc h as the parietal cortex, because existing studies did not report findings on those regions.It is also very important to note that while we based the r e vie w of the neur oima ging findings on the pr oposed emotion systems, one should always be cautious when making inverse infer ences about psyc hological pr ocesses fr om neur oima ging findings.

Positive emotion system
Five studies examined the relationship between loneliness and the positive emotion system.Four of these studies emplo y ed the task-based fMRI design (Cacioppo et al ., 2009 ;Inagaki et al ., 2016 ;Ohtsubo et al., 2020 ;Lieberz et al ., 2022 ), while the fifth study measured participants' grey matter volume (Sin et al ., 2018 ).
Four studies were conducted in healthy young adults.One recent study (Lieberz et al ., 2022 , n = 82, mixed-sex) discov er ed that higher loneliness le v el was corr elated with gr eater social anxiety as assessed using the Liebowitz Social Anxiety Scale.In this study, participants who were classified as either high-or low-loneliness performed a social gambling paradigm in which their gambling c hoices wer e followed by either human or computer feedback.The findings sho w ed that high-loneliness participants gav e mor e pleasant ratings to negative human feedbacks, and sho w ed reduced striatal activity but enhanced striatum-hippocampus functional connectivity when receiving negative human versus computer feedbac ks, compar ed to low-loneliness participants.Another study (Cacioppo et al ., 2009 , n = 23, all females) sho w ed that loneliness le v el was negativ el y r elated to v entr al striatal activities when viewing pleasant social versus pleasant non-social pictorial stimuli.A third study (Inagaki et al ., 2016 , n = 31, mixed-sex) sho w ed that lonelier individuals reported reduced feelings of connection with close others, and higher v entr al striatal activity when vie wing ima ges of close others v ersus str angers.Last, Ohtsubo et al . ( 2020 ) ( n = 22, mostly females) examined the role of the VMPFC function in loneliness .T his study administered a 'commitment signaling task', which measured participants' evaluation of social exchange signals .T his study found that high loneliness was associated with lo w er activity in the VMPFC when e v aluating low-cost commitment signals .T his result indicates that lonely individuals may derive less pleasure from relatively ambiguous social exchange signals.
The fifth study tested the relationship between loneliness level and grey matter volume in older depressed patients (Sin et al ., 2018 , n = 52, mixed-sex).They found that loneliness le v el was positiv el y associated with left striatal grey matter volume in single-episode patients, but the relationship was r e v ersed among multiple-episode patients (i.e.negativ e r elationship).In this study, patients with multiple episodes on average had experienced 2.7 de pression e pisodes, while single-e pisode patients had experienced only one de pression e pisode .T he moderation of lonelinessstriatal volume relationship by depression episode number may r e v eal pr ogr essiv e depr ession-r elated br ain pathology that affects the coping mechanism of the positive emotion system to loneliness traits.It was previously proposed that the association between loneliness and brain patterns may show a nonlinear trajectory as lonely individuals progressively become increasingly depr essed, suc h that upregulation of neural structure or function may manifest during earlier stage of depression that represent compensatory processes to the negativ el y biased emotions (Shao et al ., 2020 ).Ho w e v er, suc h upr egulation may r e v erse dir ection as the compensatory process becomes exhausted in the long term among patients with multiple de pression e pisodes that are known to cause neurotoxicity and cellular deterioration (Belleau et al ., 2019 ).
Collectiv el y, the existing evidence generally indicates a negativ e r elationship between loneliness and the functional activity of the br ain positiv e emotion system including the striatum and the VMPFC.Ho w e v er, the findings were not entirely consistent regarding the sign of the relationship (Inagaki et al ., 2016 ).Also, findings on the relationship between loneliness and structural volume of the positive emotion system were still lacking and inconclusive.In gener al, e vidence is lac king for healthy middle-aged and older adults .Moreo ver, the sample sizes of existing studies were mostly modest, and only cross-sectional observational data were a vailable .

Negative emotion system
Two studies examined the relationship between loneliness and the negative emotion system.One of these studies emplo y ed the task-based fMRI design (Wong et al ., 2016 ), whereas the other study (Wong et al ., 2019b ) constructed structural brain networks for gr a ph theory anal ysis .Among these studies , one (Wong et al ., 2019b ) looked at the role of insula in the negative emotion system and the other one (Wong et al ., 2016) studied the involvement of the amygdala.

Insula
Among adolescents, Wong et al ., ( 2019b ) ( n = 40, mixed-sex) measur ed the positiv e and negativ e tr ait affect of siblings via the 20item Chinese Affect Scale, and found that loneliness le v el was negativ el y r elated to sibling-av er a ge positiv e affect, and positiv el y related to sibling-average negative affect.In addition, the siblings' mean negative affect levels were positively correlated with local efficiencies of white-matter networks in the insula.Furthermore, the local structural network efficiency of the insula contributed to mediating the relationship between loneliness and trait negative affect.These findings supported the relationships of loneliness and both positive and negative trait affect, and implicated the insula as a core region of the negative emotion system where structur al network c hanges may under pin the association between loneliness and increased negative emotion levels.Ho w ever, given this was a cross-sectional study, the causal relationship could not be established.

Amygdala
Among depressed and healthy older adults, Wong et al . (2016) ( n = 31, mixed-sex) looked at the role of the amygdala in the negative emotional systems associated with loneliness .T his study examined the relationship between loneliness and white-matter structural network connectivity strengths, and found that loneliness was negativ el y associated with the connectivity between the amygdala and the superior frontal gyrus .T his finding indicated that loneliness could be linked with r educed r egulation of the amygdala by dorsal pr efr ontal r egions, whic h may r esult in am ygdala hyperacti vity and increased negati ve emotion bias.
In summary, limited existing evidence supports a link between loneliness and the white-matter structural networks of the negative emotion system, particularly regarding the insula and amygdala.It is possible that higher local efficiency but reduced regulatory connectivity from the dorsal prefrontal cortex collectively result in enhanced activity of the negative emotion system in lonelier individuals.Ho w e v er, the e vidence is too limited to draw a conclusion, and no causal relationship could be established.

Emotion regulation system
A total of se v en studies explor ed the r elationship between loneliness and emotion regulation systems.Of these studies, one was conducted in adolescents (Wong et al ., 2019b ), four in young adults (Cacioppo et al ., 2009 ;Tian et al ., 2017 ;Dong et al ., 2021 ;Liu et al ., 2016 ), one in healthy adults (Lieberz et al ., 2022 ), and one in MDD patients (Shao et al ., 2020 ).Among these studies, five studies (Cacioppo et al ., 2009 ;Tian et al ., 2017 ;Wong et al ., 2019b ;Dong et al., 2021 ;Liu et al ., 2016 ) examined the role of the LPFC in emotion regulation.The sixth study (Lieberz et al ., 2022 ) examined the contribution of the hippocampus in emotion regulation.The last study (Shao et al ., 2020 ) looked at the involvement of the anterior cingulate cortex.

Lateral pr efr ontal cortex
Among the five studies that examined the association between loneliness and the LPFC, one examined task-based brain activities (Cacioppo et al ., 2009 ), two examined resting-state functional connectivity (Tian et al ., 2017 ;Dong et al ., 2021 ), one examined brain grey matter volume (Liu et al ., 2016 ), and one looked at participants' white matter tracts (Wong et al ., 2019b ).Cacioppo et al . ( 2009 ) ( n = 23, all females) sho w ed that loneliness le v el was negativ el y r elated with right v entr olater al pr efrontal cortex activities during viewing unpleasant social versus unpleasant non-social pictorial stimuli.In the same vein, another study (Dong et al ., 2021 , n = 57, mixed-sex) found that among a sample of internet gaming addicts, higher loneliness was associated with r educed r esting-state functional connectivity between the left dorsolateral prefrontal cortex (DLPFC) and cortical sensorimotor areas.
Mor eov er, Tian et al ., ( 2017 ) ( n = 30, all males) found that loneliness was associated with weaker resting-state top-down functional projection from the dorsal attentional network including the superior frontal gyrus and parts of the inferior frontal gyrus, to the v entr al attentional network including parts of the inferior frontal gyrus (among others).Another study (Liu et al ., 2016 , n = 405, mixed-sex) investigated participants' brain grey matter v olume, and sho w ed that high loneliness w as associated with incr eased gr a y matter volume in the left DLPFC.T he fifth study (Wong et al ., 2019b , n = 40, mixed-sex) measured the trait affect of siblings via the 20-item Chinese Affect Scale, and found that the siblings' mean negative affect levels were positively correlated with local efficiencies of white-matter networks in the LPFC, which contributed to mediating the relationship between loneliness and trait negative affect.
In summary, multimodal neural evidence consistently sho w ed an association between loneliness le v el and LPFC structur al and functional changes.Notwithstanding, the affective implication of these changes is unclear, with the only task-based imaging study indicating a r ele v ance to negativ e social emotion r egulation.Also, while functional imaging studies sho w ed negative correlations between loneliness and LPFC activities and connectivity strengths, structur al ima ging studies sho w ed positiv e r elationships between loneliness and LPFC grey matter volume and white matter connectivity efficiency.The discrepancy in these findings could be due to structural compensatory processes in the emotion regulation network due to increased demand of regulating negative emotions among lonely individuals.

Hippocampus
One study described pr e viousl y (Lieberz et al ., 2022 , n = 82, mixed-sex) discov er ed that high-loneliness participants sho w ed enhanced striatum-hippocampus functional connectivity when recei ving negati ve human versus computer feedback when performing a gambling task, compared to low-loneliness participants.The increase in functional connectivity between the striatum and the hippocampus may indicate potentiated negative emotion regulation using positive prospects among lonely individuals.

Anterior cingulate cortex
One study (Shao et al ., 2020 , n = 27, mostly females) sho w ed that higher loneliness le v el was associated with incr eased r estingstate functional connectivity between the anterior cingulate cortex and the inferior parietal cortex in depressed patients .T his result again indicates potentiated regulation of the emotion control network (i.e.anterior cingulate cortex) on the default mode network (DMN) region (i.e.inferior parietal cortex) during spontaneous self-r efer ential affectiv e pr ocessing.

Summary
Pr eliminary e vidence fr om single studies indicates that in both healthy and depressed individuals, loneliness was positiv el y associated with functional connectivity of the hippocampus and the anterior cingulate cortex with other brain regions involved in emotion and self-r efer ential pr ocessing.This connectivity increase may represent potentiated efforts of lonely individuals in r egulating negativ e emotions .T hese r esults ar e opposite to the decreased LPFC connectivity found in lonely participants as described earlier, which may depend on the precise brain circuitries, participant c har acteristics, and whether the connectivity was measured during task performance or resting state.

Social cognition system
Only one study (Cacioppo et al ., 2009 ) explored the role of the DMPFC in the social cognition system related to loneliness .T his study ( n = 23, all females) sho w ed that loneliness le v el was negativ el y r elated to DMPFC activities when viewing pleasant social versus pleasant non-social pictorial stimuli.This could indicate reduced social cognition and perspective taking in lonely individuals, particularly to positive social signals.

Discussion
To the best of our kno wledge, this w as the first systematic r e vie w exploring a bidirectional emotional processing model of loneliness.A total of 29 studies meeting our r esearc h criteria indicated significant relationships between loneliness and the processing of positive and negative emotions .Beha viourally, a total of 10 studies consistently sho w ed negativ e r elationships between loneliness le v els and positiv e emotions in gener al, but e vidence was lacking for whether loneliness is associated with decreases in positive social emotions specifically.Position social emotions such as lo ve , empathy, and social connectedness form the motivational bases for interpersonal relationships (Zaki, 2020 ), deficits in whic h ar e commonl y observ ed in psyc hopathology (Blair, 2018 ).A total of 14 studies sho w ed positive relationship between loneliness and negative emotions, which were exclusively obtained using questionnair e measur es of negativ e emotions .T hus , it remained unclear whether loneliness is associated with negative emotions measured using behavioural paradigms .Furthermore , five studies that investigated the relationship between loneliness and the brain positive emotion system mostly reported a negative r elationship, wher eas limited evidence from only two studies reported a possibly positive relationship between loneliness and the br ain negativ e emotion system suc h as the insula and the amygdala.Ho w e v er, the sign of the relationship (positive or negative) was not entir el y consistent, and the sample sizes of those studies wer e r elativ el y modest.Mor eov er, se v en studies in total sho w ed a consistent relationship between loneliness and the brain emotion r egulation system, particularl y in the pr efr ontal cortex, but the directional implication of the findings was unclear.Last, only one study reported a negative relationship between loneliness and functional activities of the brain social cognition system.One common shortcoming of the existing studies is the lack of longitudinal design, which precluded making causal inferences.

T he rela tionship betw een positi v e emotions and loneliness
Studies that examined the association between loneliness and questionnair e measur es of positiv e emotions yielded consistent results with studies that examined the association between loneliness and brain circuitries implicated in positive emotion processing.Specificall y, all the behaviour al studies r eported a significant negativ e r elationship between self-r eported loneliness le v el on one hand, and general or specific positive emotion on the other hand.In terms of the neur oima ging findings, most existing studies sho w ed reduced functional activity in the positive emotion system (striatum and VMPFC) among lonely individuals, although ther e wer e also exceptions indicating a r e v erse (i.e.positiv e) r elationship (Inagaki et al ., 2016 ).Viewed together, it appeared that lonely individual's reduced positive emotion function might be due to their blunted functional le v el of the br ain positiv e emotion system.Such a causal relationship needs to be directly tested by neuromodulation and longitudinal studies.Relevant to this, past r esearc h indicates that functional activity of the striatum and the VMPFC during a stressful experience predicted subsequent level of positive emotion state (Yang et al ., 2018 ), and dampened the relationship between high life stress and low positive emotion state (Nik olov a et al ., 2012 ).Given loneliness is intimately related to high str ess le v els, it might be that r educed striatum and VMPFC activity r enders lonel y individuals mor e vulner able to low positiv e emotion.
As noted before, one study reported a positive relationship between loneliness le v el and v entr al striatal activity when viewing images of close others versus strangers (Inagaki et al ., 2016 ).The reason for this result was unclear.The authors interpreted this finding as reflecting an increase in desire for social connectedness due to unmet social needs.It is possible that the sign of relationship (i.e.positive) reported in this study differed from that reported in the other studies (i.e .negative , Cacioppo et al ., 2009 ;Ohtsubo et al., 2020 ;Lieberz et al ., 2022 ) due to the nature of the task par adigm used.Ina gaki et al. used a task involving viewing photogr a phs of the participant's friend, but the task otherwise did not incor por ate an explicit social interaction element.By contrast, the other studies all emplo y ed tasks inv olving non-personal stimuli with explicit social interaction processes.It could be that vie wing photogr a phs belonging to the participant's own friends elicited a greater desire for social interaction in lonely individuals compared to encountering social stimuli of str angers, whic h was reflected as increased ventral striatal activity (Inagaki et al ., 2015 ).
It is also worth noting that in the study by Lieberz et al . ( 2022 ), lonel y participants gav e mor e pleasant r atings but sho w ed lo w er striatal activity to negative human feedback when performing a gambling task.Given the task paradigm involved a risky decisionmaking process, it could be that the lo w er striatal activity repr esented r educed learning about negati ve social outcomes, gi ven the important role of the striatum in reinforcement learning (e.g.Atallah et al ., 2014 ).In this regard, it could be that lonely individuals sho w ed a r educed ca pacity to utilize social information effectiv el y to guide decision-making, which may arise from their difficulties in perceiving social cues (Kanai et al ., 2012 ) or general perceptual decision-making (M ąka et al ., 2023 ).
These studies had se v er al limitations.First, the studies mostl y included young and older adults, while few studies included adolescents and middle-aged adults.Emotion issues related to loneliness during adolescence and middle adulthood deserve special attention as loneliness is a pr e v alent experience during these life stages (Lee et al., 2019 ;Surkalim et al . 2022 ;Pengpid and Peltzer, 2021 ).Adolescence is a critical stage for establishing social and emotional styles that can influence mental well-being (Blakemore , 2019 ).During adolescence , the brain is highly plastic and under goes r a pid de v elopment, making it vulner able to the influence of negative social environments (Lamblin et al., 2017 ;Vijayakumar et al., 2018 ).Similarly, emotions related to social relationships play a critical role in shaping the mental health of middle-a ged adults (Oshio, 2014 ).Furthermor e, br ain network functions have been found to show consider able differ ences between adolescence and adulthood (Tian et al., 2016 ).For example, one study r e v ealed significant changes in functional connectivity pattern of cortical networks such as the DMN from adolescence to middle adulthood (Tian et al., 2016 ).Ther efor e, it is important to study the relationship between loneliness and positive emotions, and the associated neural patterns, at each distinct developmental sta ge, particularl y during adolescence and middle adulthood.
Another issue was that most existing r esearc h focused on examining the relationship between loneliness and positive emotions in general, while evidence was lacking on whether loneliness is associated with changes in positive social emotions specifically.Positive social emotions were defined as positive feelings related to social interactions and interpersonal relationships, typically involving experiences of connection, cooperation, friendship, lo ve , and social support with others (J anko wski and Takahashi, 2014 ; Zaki, 2020 ;v an Tilbur g, 2023 ).While positiv e emotions in gener al and positive social emotions show certain correlations with each other, they have been found to dissociate from each other (Koush et al., 2019 ;Scharnowski et al., 2020 ;Lam et al., 2021 ;Atzil et al., 2023 ), and implicate only partially overlapping neural circuitries (Koush et al., 2019 ;Scharnowski et al., 2020 ).For example, while both positive emotions in general and positive social emotions implicate brain regions such as the striatum and DLPFC, the latter may additionally implicate the DMPFC and the subgenual anterior cingulate cortex (Koush et al., 2019 ;Scharnowski et al., 2020 ).Ther efor e, futur e studies should measur e positiv e social emotions alongside positive emotions in general, using both questionnaires and task paradigms.

T he rela tionship betw een nega ti v e emotions and loneliness
T he beha vioural studies included in this systematic r e vie w consistently sho w ed a significant positiv e corr elation between loneliness and general negative emotions such as depression, anxiety, and stress .T hese r esults ar e consistent with the emerging body of evidence indicating a critical role of loneliness in precipitating depression, anxiety, and stress (Varma et al., 2021 ;Wang et al., 2022 ).Most existing r esearc h supported a bidirectional relationship between loneliness and negative emotions.Specifically, loneliness could make individuals more susceptible to experiencing negative emotions (Heinrich and Gullone, 2006 ;Erzen and Çikrikci 2018 ), and ele v ated le v els of negativ e emotions can r esult in a lac k of interest in social interactions and emotional fatigue, thereby increasing the likelihood of feeling lonely (Heinrich and Gullone, 2006 ;Wolters et al., 2023 ).The very limited neuroimaging evidence also indicates loneliness may be associated with increases in functional le v el of the brain negative emotion system, specifically the insula and amygdala.Ho w ever, further studies need to pr ovide mor e conclusiv e e vidence supporting this association.
Ho w e v er, similar to the studies on positive emotions, existing r esearc h had mostl y focused on the r elationship between loneliness and gener al negativ e emotions, while evidence is lacking on whether loneliness is associated with changes in negative social emotions specifically.Although general negative emotions such as fear and anger are related to negative feelings to w ar ds social contexts, it has been shown that these types of negative emotions can dissociate (Cândea and Szentagotai-T ȃtar, 2018 ).Also, while gener al negativ e emotion pr ocessing was pr e viousl y found to be performed by brain circuitries including the amygdala, anterior cingulate cortex and insula, negative social emotions may additionally implicate regions of the brain social network such as the DMPFC, superior temporal gyrus and the temporoparietal junction (Lam et al., 2021 ).Ther efor e, futur e studies should specifically examine the relationship between loneliness and negative social emotions.
Another issue was that existing studies tended to r el y on questionnaire assessments rather than using task paradigms to assess negative emotions.While questionnaires focus on assessing an individual's subjective negative feelings and emotions, task paradigms tend to measure an individual's behavioural tendencies such as sensitivity and avoidance to negative stimuli.A compr ehensiv e e v aluation of negativ e emotions should encompass both subjective experiences and behavioural tendencies, with the latter being assessed using task paradigms.

T he rela tionship betw een emotion regula tion system and loneliness
We r e vie wed 3 studies (Wong et al., 2019b ;Shao et al., 2020 ;Lieberz et al., 2022 ) which assessed the relationship between loneliness and the brain emotion regulation circuitry in relation to negative emotion regulation.These findings indirectly supported an association between loneliness and negative emotions.Ho w e v er, other studies did not determine whether the emotion regulation circuitry pertained to positive or negative emotion regulation, making the inter pr etation of their findings ambiguous .T he emotional regulation system associated with loneliness particularly concerned the pr efr ontal cortex.For instance, e vidence indicated a negativ e corr elation between loneliness and the activ ation of the v entr olater al pr efr ontal cortex (VLPFC) specificall y during watc hing negative social stimuli (Cacioppo et al., 2009 ).Additionally, loneliness was found to be associated with r educed r esting-state functional connectivity of the VLPFC and other cortical networks (Tian et al., 2017 ;Dong et al., 2021 ).Ho w e v er, e vidence fr om structur al ima ging studies r e v ealed that in healthy individuals, higher le v els of loneliness were associated with increased gray matter volumes in the DLPFC.This structural volume increase could be a consequence of compensatory responses to reduced functioning of the LPFC, as observed among individuals with mood dysfunctions (Liu et al., 2016 ).This speculation needs to be tested using longitudinal design with both structural and functional imaging analyses.

T he rela tionship betw een emotion regula tion system and loneliness
Given loneliness originates from insufficient social relations, it is surprising that only one neuroimaging study revealed reduced functional activity of the brain social cognition system, specifically in the DMPFC, among lonely individuals while processing pleasant social versus pleasant nonsocial stimuli (Cacioppo et al ., 2009 ).The DMPFC is implicated in both self-r efer ential and social processing (Meyer and Lieberman, 2018 ).Indeed, it was proposed that the self-r efer ential function of the DMPFC might form the basis for this region to support social cognition and emotions, as one understands the mind of others thr ough self-r eflection (Yeshurun et al ., 2021 ).It has also been shown that the DMPFC is specifically involved in theory of mind processes based on a ppr aising social inter actions (Wa gner et al ., 2016 ).Inter estingl y, the social stim uli used in Cacioppo et al . ( 2009 ) also extensiv el y portr ayed social interactions .T hus , the finding suggested that the reduced DMPFC activity may account for lo w er capacity to derive social pleasure through empathizing with others in lonely individuals.Clearly, mor e r esearc h is needed to replicate the DMPFC activity decrease related to loneliness, and to examine potential alterations in other regions implicated in social cognition, such as the temporoparietal junction.

Other neur oima ging findings related to loneliness
Besides those r e vie wed her e, ther e exist se v er al r ecent neur oima ging studies which were not included in the present review due to non-focus on positive or negative emotions (Spreng et al ., 2020 ), being a r e vie w r ather than original inv estigation (Lam et al ., 2021 ), focus on social position rather than loneliness per se (Baek et al ., 2022 ), or being published after our search end date (Baek et al ., 2023 ).Ne v ertheless, these studies r eported r ele v ant findings that are consistent with those presented here.Spreng et al . ( 2020 ) revealed that loneliness measured by a single question was most pr ominentl y associated with altered resting-state functional connectivity, white-matter tract and grey matter volumes in the DMN, including k e y r egions suc h as the DMPFC.Lam et al . ( 2021 ) systematicall y r e vie wed the neur oima ging liter atur e of loneliness, and concluded that loneliness was associated with functional and structur al alter ations in the medial and lateral PFC, insula, amygdala, hippocampus, and v entr al striatum.Baek et al . ( 2022 ) re ported that indi viduals who occupy mor e centr al positions in the social network exhibited greater neural similarity to their peers in the DMN such as the DMPFC, while Baek et al . ( 2023 ) further sho w ed that lonely individuals display ed neural dissimilarity to peers within similar DMN r egions.Giv en the DMN is considered as a common neural substrate underlying both self-and other -processing (Y eshurun et al ., 2021 ), more research is critically needed that investigates the association of loneliness and affectiv e pr ocessing in the br ain acr oss intr a-and inter-individual contexts.

T he biv alence emotion model of loneliness: limitations and future directions
On the basis of the evidence review ed, w e propose a bivalence emotion model of loneliness, which posits that changes in both positi ve and negati ve emotion systems play significant roles in precipitating loneliness .T he existing findings gener all y indicate that loneliness is negativ el y associated with positive emotions, but positiv el y associated with negativ e emotions .T his pattern was observ ed acr oss studies that measur ed emotions using questionnaires, and those that examined brain cir cuitries inv olved in emotion processing and regulation.Ho w ever, several critical gaps existed regarding this model.First, evidence for the negative relationship between loneliness and positive emotions needs to be established among adolescents and middle-aged adults.Second, further r esearc h needs to test whether loneliness is associated with positive and negative emotions in general, or positive and negative social emotions in particular, by employing both questionnaires and paradigm measures of emotions .T hird, more research needs to be conducted on the relationship between loneliness and emotion regulation functions measured using questionnaires or beha vioural paradigms .Importantly, the regulation of positive and negative emotions needs to be differentiated.Fourth, futur e neur oima ging r esearc h should sim ultaneousl y examine structural and functional data, as well as their relationships, to elucidate whether structur al incr ease could be a compensatory response to functional impairment in lonely individuals.Neur oima ging r esearc h should also sim ultaneousl y assess behaviour al/questionnair e measur es of emotions and neur al patterns, as well as their associations, to str engthen infer ences about psychological meanings of the neural findings.Finally, longitudinal studies are urgently needed to establish the causal relationship between positive and negative emotion functions and lone- We used the following k e y terms to search the MED-LINE/PubMed/PsycINFO online database to identify pertinent articles published up to 4 May 2023: (lonel * ) AND ('positive emotion * ' OR 'negative emotion * ' OR 'positive affect * ' OR 'negative affect * ' OR rew ar d * OR punishment * OR reinforcement * ).

Table 1 :
Summary of studies that tested the association of questionnaire measures of emotion and loneliness.

Table 2 :
Summary of neur oima ging studies which tested the association of emotion and loneliness.

and the positi v e emotion system, the emotional regulation system amd the social cognition system
9Abbr e viation: Corr: corr ection; FDR: false discov ery r ate; FWE: famil y-wise err or; GRF: Gaussian r andom field; LSAS: Leibowitz social anxiety scale; MDD: major depr essiv e disorder; MRI: ma gnetic r esonance ima ging; PA: positive affect.