Social Reward Processing in Depressed and Healthy Individuals Across the Lifespan: A Systematic Review and a Preliminary Coordinate-based Meta-analysis of fMRI Studies

Abstract

Background:

Social rewards (e.g., social feedback, praise, and social interactions) are fundamental to social learning and relationships across the life span. Exposure to social rewards is linked to activation in key brain regions, that are impaired in major depression. This is the first summary of neuroimaging literature on social reward processing in depressed and healthy individuals.

Method:

We screened 409 studies and identified 25 investigating task-based fMRI activation during exposure to social stimuli in depressed and healthy populations across the lifespan. We conducted a systematic review followed by an Activation Likelihood Estimation (ALE) analysis of three main contrasts: a) positive social feedback vs. neutral stimuli; b) negative social feedback vs. neutral stimuli; c) positive vs. negative social feedback. We also compared activation patterns in depressed versus healthy controls.

Results:

Systematic review revealed that social rewards elicit increased activation in subcortical reward regions (NAcc, amygdala, ventral striatum, thalamus) in healthy and depressed individuals; and decreased activation in prefrontal reward regions (medial prefrontal cortex, orbitofrontal cortex) among depressed persons. Our meta-analysis showed, in both depressed and healthy individuals, increased cluster activation of the putamen and caudate in response to negative social stimuli vs. positive stimuli. We also found increased cluster activation in the inferior frontal gyrus (IFG) and the medial frontal gyrus (MFG) in healthy controls vs. depressed individuals, in response to negative social stimuli.

Conclusions:

Processing of social stimuli elicits activation of key brain regions involved in affective and social information processing. Interventions for depression can increase social reward responsivity to improve outcomes.

Humans live social lives, and their actions are driven by anticipated social outcomes (1). Social rewards – praise, approval, or positive experiences with others – drive social learning processes and interpersonal relationships. The value of social rewards and the motivation to exert effort is affected by the emotional valence attributed to others. Social reward anticipation is stronger for significant others (2,3), those with perceived high social status (4), and those judged as attractive (5,6).

Processing of social rewards elicits changes in brain activation and can be measured during exposure to rewarding social stimuli in fMRI tasks (e.g., viewing positive social feedback, receiving visual or verbal praise). Animal studies suggest that reward structures are activated during social interaction (7-10). In healthy adults, exposure to social and non-social rewards (e.g., money) engages reward structures (11-14) including the orbitofrontal cortex (OFC) (15-17), ventral tegmental area (VTA) (18), ventral striatum (12-14,16), and nucleus accumbens (NAcc) (19-21). Rewarding, positively valenced stimuli, particularly faces, typically activate a wider network of salience and default mode network (DMN) regions, including the middle frontal gyrus (MFG), dorsal anterior cingulate (dACC), insula, superior temporal gyrus (STG), and posterior parietal cortex (22-24). Activation in these regions, in addition to reward circuitries contributes to integrated social views of ourselves and others (25-27).

Two recent meta-analyses, investigating social reward processing in healthy adults using a paradigm called the Social Incentive Delay Task (SID), found that anticipation of social rewards engaged the ventral tegmental area, ventral striatum, and anterior insula (28). Response to social rewards was associated with a large set of regions, including the ventromedial frontal and orbitofrontal cortices, anterior cingulate cortex, amygdala, hippocampus, occipital cortex and brainstem (29). These meta-analyses excluded tasks other than the SID as well as individuals with psychiatric presentations. Our study extends this work by focusing specifically on studies of major depression and examining a broader range of social reward fMRI tasks to determine which unique brain regions are involved in processing of social emotionally-valenced stimuli.

Animal (10,30) and human studies (31) suggest that social rewards may protect against psychiatric distress. Social rewards may be especially valuable for depressed individuals who exhibit reward dysfunctions, including reduced expectation of positive outcomes, disrupted reward learning, and reduced willingness to exert effort to obtain rewards (32-34). Depression severity is linked with reduced activation in reward circuitries during exposure to social rewards (35,36).

Increased engagement in rewarding social experiences may serve as an efficacious treatment target in psychotherapy for major depression (37). Increased social interactions with significant others led to greater increase in behavioral activation and reduction of depression severity in psychotherapy for late life depression (38). Conversely, low perceived social support predicts poor outcomes in psychotherapy (39-41). In depression, abnormal activation of regions involved in social reward processing (e.g. subgenual nucleus accumbens, insula, ventrolateral prefrontal cortex) contributes to negative representations of self and others (42,43). This negative cognitive bias and resulting decreased responsiveness to social rewards may also contribute to persistent depression (44,45).

The need for social rewards is universal and highly motivating, yet most reward literature focuses on processing of monetary rewards. These studies have consistently shown that depressed individuals show deficits in anticipation and response to monetary rewards, with reduced activation in the reward system (46). Neural pathways responsible for processing of monetary rewards are also involved in processing of social information (46). This is the first fMRI meta-analysis of social reward tasks in healthy and depressed individuals across the lifespan. Since imaging research in social rewards is a relatively new area, who adapted a broad focus, including studies across age groups. For the purpose of this study, we used a broad definition of social reward, focusing on positively or negatively valanced social stimuli (e.g., human faces), and we considered exposure to social rewards during both anticipation and response to social stimuli.

We aimed to identify regions activated during exposure to social rewards in social reward fMRI tasks. We hypothesized that increased activation in key regions of the reward system (i.e., NAcc, ventral striatum, and orbitofrontal cortex) in response to positive social feedback compared to neutral stimuli or negative social feedback. We also hypothesized that depressed individuals will show decreased activation in key subcortical and prefrontal reward regions in response to social stimuli.

Materials and Methods

Study Selection

We completed pre-registration of this project in PROSPERO (https://www.crd.york.ac.uk/prospero/; Record #CRD42021213628). We conducted a systematic search in Pubmed and Web of Science. We searched PubMed using the following keywords: ((face) or (facial) or (interpersonal) or (social)) AND ((reward) OR (incentive) OR (feedback)) AND (fMRI) AND (depression) NOT (EEG) NOT (autism) NOT (schizophrenia). We searched in Web of Science using: Topic=(((face) or (facial) or (interpersonal) or (social)) AND ((reward) OR (incentive) OR (feedback)) AND (fMRI) AND (depress*)) NOT Topic=(EEG OR autism OR schizophrenia).We also reviewed references of review articles found. Articles were selected based on the following inclusion criteria: a) task-based fMRI study with a social component; c) reported results included whole brain; c) study population included individuals with depressive symptoms; d) task was administered at least at one single time point; e) participants not taking medication or on a stable dosage of antidepressants or anti-anxiety medication.

Exclusion criteria included: a) clinical population with primary disorder other than depression and anxiety or personality disorder; c) populations with cognitive impairments; d) use of positron emission tomography (PET) scan or electroencephalogram (EEG); e) resting state functional connectivity studies; f) Region of Interest (ROI) analysis. Studies had to report main contrasts for activation in an fMRI task, during anticipation of or response to socially valanced stimuli (e.g., human faces), as well as corresponding coordinate-based activation clusters to be included in the meta-analysis.

Study rating

Two independent raters with training and experience in neuroimaging and EEG research in depression reviewed each available abstract (authors NS, LWV, DP, and KL). In the case of a disagreement, the four raters discussed and reached consensus. The full-length articles selected were then reviewed by two independent raters and discussed in consensus meetings. Prior to consensus meetings, there was an inter-rater agreement of 95.1%.

Based on the systematic search described above, we initially identified 409 relevant studies. When more than one paper was published from the same dataset, only one publication reporting the main contrasts of interest with the largest sample investigated was included. After removing 127 duplicates, we screened 282 abstracts. We excluded 216 abstracts due to lack of depressed population (n=102) or animal sample (n=7), reports of resting-state MRI only (n=29) or no MRI report (n=8), a duplicate with a changed title (n=1), non-social reward tasks without valenced or facial stimuli n=29), poster conference abstracts (n=4), or if a publication was a review article or there was no report of data (n=37). Then, 66 full-text records were assessed for eligibility. Fourty-one records were excluded due to: a) use of resting state only design, monetary reward task or executive function task (n=11); b) study population that did not meet inclusion criteria above (n=15); c) no report of task-based activation map or coordinates (n=8); or d) reported ROI data only (n=5); or e) duplicate removed due to identical sample and task (n=1).Twenty-five eligible records were included (Figure 1).

Figure 1.

Flow Diagram for inclusion/exclusion of studies

Characteristics of Included Studies

Study characteristics are reported in Table 1. The studies included were published between 2005 and 2023. Studies were from the United States (n = 13), Canada (n = 2), Europe (n = 3), the United Kingdom (n = 4), China (n=2), Australia (n = 1). In most studies (86%; n =25), data was collected at a single time point. Three articles included a treatment component and only pre-treatment measures were included in the analyses, such that studies were only included if patients were scanned before they began a treatment protocol. In studies that included a diagnostic evaluation, diagnosis was established using the Structured Clinical Interview for DSM Disorders (SCID; 14 studies) (35,36,42,47-57), the Kiddie Schedule for Affective Disorders and Schizophrenia (K-SADS; 7 studies) (2,58-63), the Mini International Neuropsychiatric Interview (MINI; 1 study) (6), Composite International Diagnostic Interview-Venus (CIDI-V; 1 study) (3), or Screen for Child Anxiety Related Disorders (SCARED; 1 study) (64). Depression severity was assessed in the majority of studies using the Beck Depression Inventory (BDI; 12 studies) (65), followed by the Hamilton Depression Rating Scale (HAM-D; 3 studies) (66), and the Mood and Feelings Questionnaire (MFQ; 2 studies) (67) (see Table 1 for complete list of clinical scales) (see Table 1 for complete list of clinical scales).

Table 1.

Study Sample Characteristics.

Study	Overall N (% Female)	Group N	Age Mean (SD)	Clinical Measures Mean (SD)
Davey et al., 2011	N = 39 (64%)	Young Adults; MDD = 19* Young Adults; HC = 20*	18.9 (2.5)	BDI	MDD = 33.9 (11.3) HC = 3.1 (3.6)
Frey & McCabe, 2020	N = 43 (72%)	High Depression (HD) = 21	22.8 (NR)	BDI	HD = 26.1 (9.6) LD = 1.9 (1.3)
		Low Depression (LD) = 22		RSAS	HD = 18.6 (6.4) LD = 5.8 (4.3)
Greening et al., 2013	N = 36 (67%)	MDD = 18 HC = 18	27.3 (11.5)	BDI	MDD = 24.6 (9.8) HC = 1.6 (2.4)
He et al., 2019	N = 41 (49%)	Subclinical MDD = 21	19.5 (1.7)	BDI	Subclinical MDD = 16.0 (8.0) HC = 7.8 (5.8)
		HC = 20		SDS	Subclinical MDD = 0.5 (0.04) HC = 0.4 (0.06)
He et al., 2022	N = 47	Subclinical MDD = 22 HC = 25	19.4 (1.5)	BDI	Subclinical MDD = 16.3 (8.6) HC = 7.0 (4.9)
				SDS	Subclinical MDD = 0.5 (0.03) HC = 0.4 (0.05)
				SPSRQ (Reward)	Subclinical MDD = 35.3 (4.6) HC = 33.7 (3.1)
Healey et al., 2014	N = 27 (52%)	Late Adolescents (LA); MDD = 9; Anxiety = 3; HC = 15	20.4 (0.8)	CES-D	LA = 12.6 (14.5)
				RSAS	LA = 7.2 (7.2)
Keedwell et al., 2005	N = 24 (67%)	MDD = 12	NR	BDI	MDD = 33.5 (11.2) HC = 1.4 (2.0)
		HC = 12		GHQ	MDD = 33.5 (11.2) HC = 1.9 (1.2)
Kumar et al., 2015	N = 22 (64%)	MDD = 12 HC = 10	32.8 (12.5)	BDI	MDD = 25.2 (9.1) HC = 1.9 (3.9)
Laurent & Albow, 2013	N = 22 (100%)	MDD = 11 HC = 11	24.1 (4.1)	CES-D	MDD = 24.2 (9.4) HC = 7.5 (6.2)
Morgan et al., 2015	N = 19 (100%)	MDD = 19	43.9 (8.7)	SCID (depressive episodes)	MDD = 5.0 (5.0)
Morgan et al., 2019	N = 56 (56%)	Children; High Risk for MDD = 25	6.8 (0.8)	MFQ	High Risk = 40.5 (NR) Low Risk = 38.6 (NR)
		Children; Low Risk for MDD = 31		SCARED	High Risk = 9.7 (NR) Low Risk = 7.1 (NR)
Murrough et al., 2015	N = 38 (55%)	MDD = 18 HC = 20	33.4 (NR)	MADRS	MDD = 29.9 (6.8) HC = NR
Nagy et al., 2021	N = 61 (64%)	MDD +Childhood Maltreatment (CM) = 21 MDD = 19 HC = 19	33.3 (NR)	BDI	MDD + CM = 23.1 (5.7) MDD = 21.5 (3.3) HC = 4.3 (2.9)
				BAI	MDD + CM = 21 (NR) MDD = 18 (NR) HC = 3 (NR)
Penton-Voak et al., 2020	N = 36 (67%)	Subclinical MDD; Training = 19	22.0 (4.0)	BDI-ii; tx end	Intervention = 14.4 (5.7) No Intervention = 16.4 (7.0)
		Subclinical MDD; No Training = 17		HAM-D; tx end	Intervention = 11.7 (5.6) No Intervention = 14.8 (5.1)
				PANAS (negative); tx end	Intervention = 13.5 (3.4) No Intervention = 16.6 (6.1)
Perini et al., 2019	N = 30 (100%)	Adolescents; NSSI = 30 Adolescents; HC = 30	16.15 (0.8)	CDRS-R	NSSI = 45.7 (13.4) HC = 22.2 (4.9)
Quevedo et al., 2018	N = 81 (51%)	Adolescents; MDD = 43 Adolescents; HC = 38	14.6 (1.7)	CDRS	MDD = NR HC = NR
Schaefer et al., 2006	N = 23 (30%)	MDD = 9	32.35 (8.9)	HAM-D	MDD = 23.4 (7.2) HC = 0.4 (0.6)
		HC = 12		BDI	MDD = 24.3 (8.5) HC = 0.1 (0.3)
				PANAS (positive)	MDD = 2.1 (0.8) HC = 3.6 (0.7)
Schwartz et al., 2019	N = 15 (47%)	Adolescents; MDD + Anxiety = 15	15.3 (2.4)	MFQ	MDD + Anxiety = 12.0 (13.5)
				SCARED	MDD + Anxiety = 15.7 (10.2)
Seitz et al., 2023	N = 118 (78%)	PTSD = 25 Somatic Symptom Disorder = 29 MDD = 32 HC = 32	31.1 (11)	BDI	PTSD = 25.0 (11.3) Somatic Symptom Disorder = 16.1 (9.7) MDD = 32.7 (9.9) HC = 5.2 (4.6)
Sharma et al., 2016	N = 86 (51%)	Bipolar MDD = 24 Unipolar MDD = 24 HC = 38	38.6 (12.1)	BDI	BPD = 22.4 (7.9) MDD = 25.0 (8.7) HC = 2.4 (4.8)
Silk et al., 2014	N = 48 (71%)	Adolescents; MDD = 21 Adolescents; HC = 27	15.5 (1.7)	NR	NR
Silk et al., 2017	N = 48 (75%)	Adolescents; MDD = 20 Adolescents; HC = 28	14.6 (1.8)	NR	NR
Whittle et al., 2012	N = 30	Adolescents; MDD = 30	17.3 (0.4)	CES-D	MDD = NR
Wonch et al., 2016	N = 45 (100%)	Postpartum Depression Mothers = 28	29.9 (1.1)	EPDS	Postpartum Depression = 8.3 (0.8) HC = 3.1 (1.1)
		HC Mothers = 17		State-Trait Anxiety Index	Postpartum Depression = 44.9 (1.5) HC = 27.8 (2.0)
Yttredahl et al., 2018	N = 38 (100%)	MDD = 19 HC = 19	29.7 (11.05)	HAM-17	MDD = 14.8 (3.0) HC = NR

Note: BAI = Beck Anxiety Index; BDI = Beck Depression Inventory; BPD = Bipolar Major Depressive Disorder; CDRS(-R) = Children’s Depression Rating Scale (-Revised); CES-D = Center for Epidemiological Studies Depression Scale; EPDS = Edinburgh Postpartum Depression Scale; HAM-17 = Hamilton Depression Rating Scale (17-Item); HAM-D = Hamilton Depression Rating Scale; HC = Healthy Control; MADRS = Montgomery-Asberg Depression Rating Scale; MDD = Major Depression Disorder; MFQ = Mood and Feelings Questionnaire; NR = not reported; PANAS = Positive and Negative Affect Schedule; RSAS = Ryerson Social Anxiety Scale; SCARED = Screen for Child Anxiety Related Disorders; SCID = Structured Clinical Interview for DSM-V; SDS = Zung Self-Rating Depression Scale; SPSRQ = Sensitivity to Punishment / Sensitivity to Reward Questionnaire

^*Smaller N included in MRI analysis (MDD = 17; HC = 19)

Data Analysis

Aim 1: Systematic Review

We conducted a systematic review of the 25 studies. We summarized findings from all included studies. First, we reviewed the clinical characteristics of available samples and the components of the fMRI tasks used across studies. Second, we conducted a synthesis of findings on activation patterns during exposure to social reward stimuli. We complied results on activation patterns in three major networks involved in social reward processing: the reward system (prefrontal and subcortical reward structures), the salience network, and the default mode network (DMN). See Figure 2 for a conceptual diagram of key networks identified in the systematic review that correspond with the synthesis of findings (individual regions displayed at peak Harvard-Oxford Cortical and Subcortical Atlas coordinate locations for visualization purposes).

Figure 2.

Social Reward-Related Networks Implicated in Systematic Review

Aim 2: Meta-Analysis

Activation Likelihood Estimation (ALE)

We conducted quantitative meta-analysis of social reward-based activation using the ALE technique in BrainMap’s GingerAle software package (version 3.0.2; http://www.brainmap.org/ale/), with a random effect algorithm to assess convergence across 3-D coordinates of brain regions involved in processing socially rewarding stimuli across multiple studies. GingerALE uses input foci (i.e., input x, y, and z coordinates) to create study-specific activation maps based on the Gaussian probability distributions around each study’s foci, also accounting for between-subject variance, then uses the activation maps to find probability-based agreement in activation across studies. All coordinates were transformed into MNI space for analysis and visualization of results in standard space.

To identify convergence of social reward activation and significant ALE clusters across studies, we used a standard cluster-forming threshold of p < 0.01 and a cluster-level family-wise error (FWE) correction of p < 0.05 for cluster-level inference (68). The significant clusters are contiguous voxels, derived from the activation maps, that exceed the p-value threshold and the cluster-forming threshold minimum (69). These clusters were compared to an empirically-derived null distribution of clusters (70). The ALE analysis was conducted on a stimulated random data set that included the same number of studies, subjects, and foci as our entered data set. The FWE, then, tracked the distribution of ALE scores across 5000 permutations and set the ALE value such that only 5% of the distribution exceeds that value. This analysis yields a cluster map and ALE score that represents the degree of activation convergence across foci.

We divided the 25 studies included in the meta-analysis into three task-based subgroups and two group-based subgroups based on the most common fMRI contrasts (Table 2). The task-based analyses focused on social reward task contrasts that identified patterns of activation for valenced social stimuli relative to neutral stimuli. Task contrasts included: 1) positive social stimuli vs. neutral stimuli (k=9 (2,6,42,47,48,57,58,63)); 2) negative social stimuli vs. neutral stimuli (k=6 (2,6,42,48,53,63)) ; 3) positive vs. negative social stimuli (k=5 (36,49,51,71)). The group-based analyses focused on activation in response to social rewards or negative social stimuli in participants with depression vs. healthy control comparison participants. Group contrasts included: 1) depressed vs. healthy control activation for positive social stimuli (k=3 (48,53,71)); 2) depressed vs. healthy control activation for negative social stimuli (k=3 (48,54,71)). We considered any exposure to social rewards or negative social stimuli and did not separate studies examining anticipation from those examining feedback or consumption of socially valanced stimuli given that many tasks in our sample did not include an anticipation phase and thus separation of task phases was not possible.

Table 2:

Social Reward fMRI Contrast Categories

Study	Group Contrast	Type of Task	Task Contrast	Reason for Exclusion from ALE Analysis
Davey et al., 2011	MDD vs. HC; Young Adults	Passive; Acceptance or rejection of high vs. low rated faces, ambiguous feedback control condition	Positive Faces vs. Control Faces
			All Faces vs. Fixation
Frey & McCabe, 2020a	MDD; High Depression vs. Low Depression	Performance; Social learning of association between names & valenced faces	Social Reward
			Social Aversion vs. Neutral Faces
Frey & McCabe, 2020b	Serotonin vs. Dopamine Depletion vs. Placebo	Performance; Social learning of association between names & valenced faces	Social Reward Prediction	Findings account for neurotransmitter depletion status.
			Social Aversion Prediction
			Social Reward vs. Social Aversion
			Fearful Faces vs. Neutral Faces
Greening et al., 2013	MDD vs. HC	Performance; Facial emotion recognition, displayed with valenced distractor cues	Happy Face Target & Neutral Distractor
			Fearful Face Target & Neutral Distractor
			Neutral Target & Happy Distractor
			Neutral Target & Fearful Distractor
He et al., 2019	Subclinical MDD vs. HC	Performance; Monetary Incentive Delay Task with social evaluation manipulation	Response: Happy vs. Control
			Response: Sad vs. Control
			Anticipation: Sad vs. Control
He et al., 2022	Subclinical MDD vs. HC	Performance; Social judgement, likelihood of acceptance followed by acceptance or rejection	“Yes” anticipation followed by Acceptance
			“Yes” anticipation followed by Rejection
			“No” anticipation followed by Acceptance
			“No” anticipation followed by Rejection
Healey et al., 2014	MDD & Anxiety & HC; Late Adolescents	Passive; Social judgement, acceptance or rejection of high vs. low rated faces, ambiguous feedback control condition	Mutual Liking vs. Received Liking
			Positive Feedback vs. Ambiguous Feedback
Keedwell et al., 2005	MDD vs. HC		Happy Mood Condition
		Passive; Mood induction via audio recording, mood-congruent facial display	Sad Mood Condition
			Happy vs. Sad Mood Condition
Kumar et al., 2015	MDD vs. HC	Performance; Monetary Incentive Delay Task with social evaluation manipulation	Response to Reward; Stress vs. No Stress Condition
Laurent & Albow, 2013	MDD vs. HC; Mothers	Passive; Viewing valenced faces of own vs. other infant	Distressed Faces; Own vs. Other Infant
Morgan et al., 2015	MDD; Mothers	Passive; Viewing video recordings of own vs. other child, positive vs. negative recordings	Positive Faces; Own Child vs. Stranger
			Negative Faces; Own Child vs. Stranger
Morgan et al., 2019	MDD; High Risk vs. Low Risk; Children	Passive; Viewing valenced faces of own vs. other mother	Happy Faces; High Risk vs. Low Risk	Findings account for depression risk status.
Murrough et al., 2015	MDD vs. HC	Performance; Facial emotion recognition task	All conditions
Nagy et al., 2021	MDD + Childhood Maltreatment; MDD; HC	Performance; Facial recognition paradigm, match valenced test faces to central target face, shape matching control condition	Face-matching vs. Shape-matching
Penton-Voak et al., 2020	Sub-Clinical MDD; Intervention vs. No Intervention	Performance; Sex discrimination of valenced faces	Happy Faces vs. Sad Faces	Control: fixation / rest instead of neutral stimulus condition.
			Happy Faces vs. Fearful + Sad Faces
			Happy Faces vs. Rest
Perini et al., 2019	NSSI vs. HC; Adolescents	Performance; Facial emotion recognition of self vs. other, acceptance & rejection in other condition	Anticipation: Self vs. Other
			Response: Self vs. Other
Quevedo et al., 2018	MDD vs. HC; Adolescents	Performance; Emotional Self-Other Morph-Query (ESOM-Q) Task	Self vs. Other
Schaefer et al., 2006	MDD vs. HC	Passive; Social situation viewing, social interaction and facial stimuli, appetitive control condition	Group * Scan Time * Stimulus Condition	Findings account for scan time (pre- vs. post-tx).
Schwartz et al., 2019	MDD + Anxiety; Adolescents	Performance; Probe response task with probe displayed with valenced or neutral faces	Treatment Main Effect	Findings account for effect of treatment (behavioral therapy).
			Treatment Main Effect * Face Emotion * Probe Location
Seitz et al., 2023	Mixed Sample: MDD, PTSD, Somatic Sensory Disorder, HC	Performance; Social Incentive Delay Task	Social Reward Anticipation vs. Neutral Cue
Sharma et al., 2016	MDD vs. BPD vs. HC	Performance; Sex discrimination followed by valenced facial feedback	Happy Face vs. Angry Face
			Happy Face vs. Angry Face * Group * Depression Severity
Silk et al., 2014	MDD vs. HC; Adolescents	Performance; Chatroom task, social acceptance & rejection	Acceptance > Rejection
Silk et al., 2017	MDD vs. HC; Adolescents	Passive; Listening to audio clips of maternal praise or criticism	Criticism vs. Neutral
			Praise vs. Neutral
Whittle et al., 2012	MDD; Adolescents	Passive; Viewing video clips of positive, aggressive, and neutral behavior, own vs. other mother	Positive vs. Neutral Behavior
			Aggressive vs. Neutral Behavior
Wonch et al., 2016	Postpartum Depression vs. HC	Passive; Viewing valenced faces of own vs. other infant	Familiar Face vs. Unfamiliar Face
			Unfamiliar Face
			Non-Infant Stimuli
			All Conditions
Yttredahl et al., 2018	MDD vs. HC	Passive; Viewed highly rated “dating profiles” in scanner, followed by acceptance or rejection	Rejection vs. Neutral
			Acceptance vs. Neutral

Note: ALE = Activation Likelihood Estimate; BPD = Bipolar Major Depressive Disorder; HC = Healthy Control; MDD = Major Depression Disorder; NSSI = non-suicidal self injury

Results

Aim 1: Systematic Review

Sample Characteristics

The characteristics of the 25 included studies included in the ALE analysis are presented in Table 1. All studies (k=25) included depression symptoms; a subgroup of these studies (k=19) also included healthy control comparison samples of non-depressed individuals.

Seven studies included adolescent samples (2,58,60-64) and one study included children (59). Three studies included samples of new mothers (3,35,54). Two studies included subclinical depression (42,49) ¹, and five studies included mixed clinical samples with anxiety (58,64), non-suicidal self-injury (60), PTSD and somatic sensory disorder (57), and bipolar depression (36).

Types of Social Reward Tasks

Eleven studies were passive tasks, in which participants viewed valenced images (k=7 (3,6,47,52,54,58,59)), viewed valenced video recordings (k=2 (35,63)), or listened to valenced audio recordings (k=2 (2,71)). Fourteen studies had performance-based tasks, in which participants were actively making responses in the MRI scanner. The types of social reward tasks are specified in Table 2. Common tasks included emotion recognition tasks (k=8 (36,48,51,53,55,56,60,61)), social judgement tasks (k=2 (49,62)), and incentive delay tasks with a social component (k=4 (42,50,57,64)). Valenced cues or feedback in performance-based fMRI tasks primarily included positive and negative social stimuli in the form of emotional faces (k=19), but also included tasks with acceptance or rejection (k=5 (6,42,49,58,62)) and criticism or praise (k=2 (2,50)).

Five studies included a familiar vs. non-familiar component, in which activation in response to known vs. unknown face stimuli were contrasted (3,35,59-61). One study included a stress manipulation in which a reward processing task was performed under a social evaluative stress or non-stress condition (50). One study contrasted the processing of face vs. shape stimuli (56). These task variations did not encompass the main task or group contrasts and thus were not included in the ALE analysis.

Activation Patterns during Exposure to Social Feedback Across Studies

Convergent data suggest that both depressed individuals and healthy controls show activation of the prefrontal regions of the reward system (medial prefrontal cortex [mPFC], orbitofrontal cortex [OFC]; Figure 2a) in response to social rewards. This activation pattern was observed during passive viewing of socially rewarding stimuli (i.e., happy faces relative to sad, angry, or neutral faces) (47,71) and active tasks, during which participants received positive social feedback for successful task performance (36,57). Increased mPFC activation was also observed in adolescents with social anhedonia during receipt of positive social feedback (58), and individuals reporting high stress levels during an incentivized performance task under induced social stress (50). However, others found that depressed individuals showed reduced activity of the dmPFC and vlPFC (48), as well as the middle temporal gyrus (61) in response to negative social cues, compared to healthy controls. Those with subthreshold depression also showed reduced activity of the mPFC during anticipation of positive feedback (49).

Rewarding positive social stimuli were followed by reduced activation in the subcortical regions of the reward system (amygdala, nucleus accumbens [NAcc], ventral striatum, insula, thalamus; Figure 2b) among depressed individuals. Studies included passive tasks that involved viewing positive emotionally valenced faces and social learning tasks in individuals with depression (3,53), risk of depression (59), and depressed patients with history of childhood maltreatment (56), compared to healthy controls. Individuals with high depression symptoms showed reduced activation compared to those with low depression symptoms during prediction of happy faces, in the superior parietal lobe/precuneus, the right insula, supramarginal gyrus. This group also showed reduced activation of the ventral striatum in response to unexpected positive social stimuli (i.e acceptance) (53). Further, those with depression showed reduced activation in the amygdala (49) as well as the caudate, compared to healthy controls (48).

Findings are mixed regarding activation in the salience network (dorsal anterior cingulate cortex [dACC] and anterior insula; Figure 2c) during exposure to social stimuli. Four studies suggested that compared to healthy controls, depressed individuals show increased salience network activation during negative and positive social stimuli (5,6,49,57). In contrast, two other studies found reduced activation in regions of the salience network during receipt of social feedback in performance tasks among individuals with subclinical depressive symptoms (42) and those with a history of non-suicidal self-harm (60), compared to healthy controls. Another study showed that compared to depressed mothers, non-depressed mothers showed increased activation of the salience network (i.e. dACC) in response to familair social stimuli (i.e. own baby vs others’ baby) (54).

Finally, several studies found activation of regions of the default mode network (DMN; VMPFC, medial and superior temporal cortex, hippocampus, parahippocampal gyrus, posterior cingulate cortex; Figure 2d) during response to social rewards (57). Reduced DMN activation was observed in depressed mothers when viewing pictures of their own infant, relative to an unknown infant (35). Increased activation was found among individuals with subclinical depression in anticipation of negative social feedback on a performance task, compared to healthy controls (42). Further, depressed adolescents, compared to controls, showed increased activation in the bilateral amygdala, subgenual anterior cingulate, left anterior insula and left nucleus accumbens in response to negative social stimuli (i.e. rejection) (62).

Aim 1: Meta Analysis

We performed an ALE analysis on five contrasts: three task-based contrasts and two group contrasts. The task-based contrasts were: 1) positive social stimuli vs. neutral stimuli; 2) negative social stimuli vs. neutral stimuli; 3) positive vs. negative social stimuli. The group contrasts were: 1) depressed vs. healthy control activation for positive social stimuli; 2) depressed vs. healthy control activation for negative social stimuli.

Positive vs. Negative Social Feedback

Six studies included a total of 204 participants (139 depressed and 65 healthy controls). Participants’ ages ranged from 11 to 33 years. Depression severity in the depressed samples was quantified using the BDI in all studies included in this analysis.

The ALE analysis with 30 foci (individual MNI coordinates) from six studies that contrasted positive social stimuli relative to negative social stimuli indicated a significant cluster of activation in the caudate and putamen (Figure 3a). This cluster had a maximum ALE value of 0.0118 and included four foci from four distinct studies that reported increased activation for positive relative to negative social stimuli (Table 3a). The four studies contributing to the cluster had an MDD sample (49,51), subclinical MDD sample (49), and a combined sample of depressed and healthy control participants (36,62).

Figure 3a.

ALE: Positive > Negative Task Contrast Right Caudate / Putamen Figure 3b1. ALE: Negative Stimuli, MDD < HC Group Contrast Cluster 1: Medial Frontal Gyrus Figure 3b2. ALE: Negative Stimuli, MDD < HC Group Contrast Cluster 2: Inferior Frontal Gyrus

Table 3a:

ALE Cluster Results (Positive > negative Task Contrast)

Cluster Location	Cluster Size (mm3)	Coordinates (Center)			Coordinates (Peak)			Peak Z- score	ALE Score (10−2)	Contributing Studies
		x	y	z	x	y	z
Right Caudate / Putamen	3520	12.6	13.4	−3.2	12	16	2	4.47	1.17	1 foci	He et al., 2022 Social judgement, Acceptance > Rejection, Subclinical MDD
										1 foci	Sharma et al., 2016 Valenced facial feedback, Happy > Angry Faces, MDD & HC combined sample
										1 foci	Silk et al., 2014 Chatroom task, Acceptance > Rejection, MDD & HC combined sample

Note: ALE = Activation Likelihood Estimate; HC = healthy control; MDD = Major Depressive Disorder All coordinates in Montreal Neurological Institute (MNI) space; ALE cluster derived from cluster-forming p < 0.01 and family wise-error (FWE) cluster threshold p < 0.05; Number of permutations = 5000

Depressed vs. Healthy Controls for Negative Social Stimuli

Three studies included a total of 82 participants (41 depressed and 41 healthy controls). Participants’ ages ranged from 16 to. 59 years. Depression severity in this sample was quantified using the BDI and the CES-D. The ALE analysis with 11 foci from three studies that contrasted activation for negative social stimuli in depressed vs. Control participants indicated two significant clusters of activation in: 1) medial frontal gyrus (MFG) and cingulate gyrus; 2) interior frontal gyrus (IFG). The MFG cluster had an ALE maximum value of 0.0098 and included two foci from two distinct studies (Table 3b). The IFG cluster had an ALE maximum value of 0.0112 and also included two foci from two distinct studies (Table 3b). The studies contributing to the clusters (48,54,71) reported decreased activation in the MFG and IFG for negative stimuli in depressed participants relative to healthy controls.

Table 3b:

ALE Cluster Results (Negative Stimuli, MDD < HC Group Contrast)

Cluster Location	Cluster Size (mm3)	Coordinates (Center)			Coordinates (Peak)			Peak Z- score	ALE Score (10−2)	Contributing Studies
		x	y	z	x	y	z
Left Medial Frontal Gyrus (MFG)	4424	−4	37.5	31.1	−2	38	34	4.60	0.98	1 foci	Greening et al., 2013 Facial emotion recognition, Neutral Face with Negative Distractor, MDD < HC
										1 foci	Laurent & Albow, 2013 Distressed Faces, Own vs. Other Infant, MDD < HC
Left Inferior Frontal Gyrus (IFG)	4120	−42.9	42.8	−9.6	−44	42	−10	4.78	1.12	1 foci	Greening et al., 2013 Facial emotion recognition, Neutral Face with Negative Distractor, MDD < HC
										1 foci	Keedwell et al., 2013 Negative mood induction with sad faces; MDD < HC

Note: ALE = Activation Likelihood Estimate; HC = healthy control; MDD = Major Depressive Disorder All coordinates in Montreal Neurological Institute (MNI) space; ALE cluster derived from cluster-forming p < 0.01 and family wise-error (FWE) cluster threshold p < 0.05; Number of permutations = 5000

Discussion

We conducted a systematic review to synthesize available evidence on patterns of brain activation during social reward exposure. In a meta-analysis, we then identified key brain regions activated in response to positive and negative social stimuli in both depressed and healthy controls. We found that in both groups, exposure to positive social stimuli, relative to negative stimuli is linked to increased activation in the caudate and putamen. Further, compared to depressed individuals, healthy controls had increased activation of the the medial frontal gyrus (MFG) and the interior frontal gyrus (IFG) in response to negative social stimuli. Our results extend previous meta-analyses on brain activation during social reward exposure in non-depressed individuals (28,29).

Our task-based results suggest that positive social stimuli relative to negative social stimuli was associated with a cluster of increased activation in the caudate and putamen – both part of the dorsal striatum. The dorsal striatum has a key role in reward processing and reward learning (72). Our results align with previous studies demonstrating striatal activation during reward processing and suggest that positive social stimuli may elicit higher striatal activation compared to negative stimuli. This finding may in part be due to the broader definition of social stimuli in our meta analysis, which included a wide range of stimuli across studies. Specifically, many studies used happy/sad faces as social reward stimuli, which likely elicit activation in brain regions responsible for face processing and show deficits in depression (73). Future research can determine whether these activation patterns vary as a function of the type of social stimuli used.

Our group results indicate activation in prefrontal reward and DMN network regions, specifically in the MFG and IFG, that differs between depressed individuals and healthy controls. The IFG is involved in processing of social information on self and others, and may be particularly crucial for the evaluation of valence for social information (74). The MFG plays a role in self referential processes, emotion dysregulation, decision making and motivation to pursue rewards. Our results suggest that perhaps these regions are deactivated in depressed individuals due to impairments in these cognitive processes (75). Major depression is characterized with reduced motivation and reward anticipation as well as deficits in reward learning. Further, depressed individuals also show reduced capacity to learn and evaluate social information in their environment.

Previous reward studies have primarily focused on anticipation and receipt of monetary rewards. Findings suggest that processing of social and monetary rewards may recruit similar brain regions. For example, anticipation of social and monetary rewards engage the ventral tegmental area, ventral striatum, anterior insula and supplementary motor area (28). Our results regarding insular activation in response to social stimuli are in line with previous reports on processing of monetary rewards. Meta-analytic evidence suggests increased insular activation during anticipation of rewards and losses in healthy controls (76,77), and rewards consumption in depressed individuals (46). Further, a recent meta-analysis of the Social Incentive Delay (SID) task in healthy adults showed increased insular activation in response to positive social rewards in the SFG, among other regions, and increased IFG and insular activity in response to negative social feedback (29). Our findings suggest that these patterns are not specific to the SID task, but rather emerge with other social reward tasks and present among healthy and depressed individuals.

Our systematic review suggested that social stimuli elicit activation in critical brain regions in the reward system, the salience network, and the default mode network (DMN). Exposure to social feedback was most consistently associated with increased activation in prefrontal reward regions (medial prefrontal cortex [mPFC], orbitofrontal cortex [OFC]) in depressed and healthy individuals (36,47-50,57,58,61,71). These dopamine-dependent regions are critical for the learning of associations between valenced stimuli and their corresponding outcomes and play a role in the higher-order cognitive processes of reward prediction, reward learning, and flexible, goal-directed behaviors (78).

Studies showed decreased activation in subcortical reward regions (amygdala, nucleus accumbens [NAcc], ventral striatum, thalamus) during social reward exposure in individuals with depression and other psychiatric presentations, compared to healthy controls (3,48,49,51,53,55,56,64). Network-level deficits in subcortical reward regions among depressed individuals may be linked to a negative cognitive bias that reduces the impact of positive, rewarding feedback and can contribute to persistent depressive symptoms (3). Findings are in line with well-documented deficits in processing of monetary rewards in depression (77,79). Simplified interventions that target subcortical reward processes and promote positive emotion regulation strategies may rescue this abnormal network activity (51,64).

Social reward exposure is associated with activation in the DMN. Regions of the DMN (VMPFC, portions of the ACC, medial and superior temporal gyri, amygdala, hippocampus, parahippocampal gyrus) are critical for episodic memory and self-referential representations of emotionally valenced stimuli. Reduced activation may be specific to familiar, self-relevant social stimuli (35). Depressed individuals may be more likely to internalize negative information from those close to them, such as loved ones (35,59).

Regions of the salience network are also critical for social reward processing, but findings are mixed. Compared to healthy controls, depressed individuals (5,6,49,57) showed increased activation during exposure to social feedback, while those with subclinical depression (42) and non-suicidal self-harm (60) demonstrated reduced network activation. Further research is needed to determine associations between processing of social rewards in this network and mood symptoms.

There are several limitations to our study. First, our results are limited by the relatively small sample and the high variability in types of tasks used. Thus, we were not able to test differences in activation as a function of task types (e.g., passive viewing of faces or performance), as well as effects specific to task phase (i.e., anticipation vs receipt). However, the sample size of studies is comparable to other similar task-based fMRI meta analyses (80,81). Relatedly, since imaging social reward research is relatively new, we adopted a lifespan approach and included studies across age groups to summarize all current evidence on this topic. While this allows for a broad and more inclusive review, we could not separate analyses based on age groups. Second, even though our search was focused on depressed populations, our sample took a combined approach and inlcuded studies that collapsed across depressed participants and healthy controls in reporting their results, which affected our results and highlights the need for further research on social rewards specifically in depressed populations that are more impacted by deficits in reward processing. Third, studies included a single MRI scan and thus we could not examine longtituinal effects on social reward processing. Fourth, most studies utilized depression as a binary category (depressed versus healthy controls) rather than a continuous variable. Thus, we were unable to meta-analyze the association between depression severity and social reward processing.

In summary, our meta-analysis is the first to investigate social reward processing in healthy and depressed individuals across the lifespan. While this study is preliminary, our findings suggest differential activation patterns and network engagement for positive and negative social stimuli. Our results suggest that social positively valenced stimuli elicits higher activation in key reward regions (i.e. putamen and caudate), relative to negative social stimuli. Further, our results suggest that non-depressed individuals show higher activation in the Inferior Frontal Gyrus (IFG) and the Medial Frontal Gyrus (MFG), compared to those with depression, in response to negative social stimuli. These findings are aligned with previous studies showing impairments in decision making and motivation to pursue rewards in depressed individuals.

Social reward processing may be a potential treatment target for psychotherapeutic interventions for major depression. As opposed to monetary reward, social rewards exist in our natural environment, are freely available, and can be increased through therapeutic interventions. For example, increased exposure to social rewards has been shown to be beneficial in psychotherapy for depression (38). Human social reward research is relatively new and further empirical evidence is needed to determine its role on neural and behavioral levels. Future studies should investigate neural activation during both anticipation and response to social rewards. Standardized neuroimaging social reward tasks could provide an objective measure for changes in social reward processing during psychological interventions. Promising task components include social feedback from a highly valued familiar individual and varying types of valenced feedback. Findings from longitudinal data could guide development of interventions targeted to increase social reward processing and improve depression.

Highlights.

• The need for social rewards is universal and highly motivating. Social reward exposure engages reward structures and is associated with mood fluctuations in depressed and healthy individuals

• We conducted a systematic review and a coordinate-based meta-analysis of studies using fMRI social rewards tasks

• Our review indicated that social rewards elicit increased activation in subcortical regions of the reward networks (NAcc, amygdala, ventral striatum, thalamus) in both healthy and depressed individuals; and decreased activation in prefrontal reward regions (medial prefrontal cortex, orbitofrontal cortex) among depressed persons.

• Our meta-analysis showed that in both depressed and healthy individuals, there was higher striatal activation during exposure to positive relative to neutral social stimuli

• Healthy controls, compared to depressed individuals, showed increased activation in regions of the default mode and salience networks in response to negative versus positive social stimuli.

• Social reward responsivity is linked with activation in key brain regions and may serve as a promising treatment target in interventions for depression.