This review focuses on psychosocial adjustment in adults with histories of social withdrawal in childhood and/or adolescence (hereafter “childhood social withdrawal”). Social withdrawal is a behavioral pattern in which children remain alone more often than age-mates in contexts which present the opportunity for peer interaction, such as freeplay at childcare or recess at school (Gazelle and Rubin, 2010; Rubin and Chronis-Tuscano, 2021; Rubin et al., 2009; Spangler and Gazelle, 2009). Social withdrawal is an “umbrella term” for solitary behavior in social contexts, although children demonstrate this behavior for multiple reasons (Asendorpf, 1990a; Rubin et al., 2009; Spangler and Gazelle, 2009). We reviewed publications with a common focus on children who exhibited socially withdrawn behavior regardless of underlying affect, motivation, or social cognition, because most reviewed publications identified children as socially withdrawn without identifying the reason or motivation for withdrawal.

The most common reason for childhood social withdrawal is social anxiety (Coplan et al., 2013). Social anxiety is conceptualized as children's anticipatory worry about how they will be received by peers and whether they will interact competently (social evaluative concerns). Children who display social withdrawal due to social anxiety are conceptualized as experiencing social approach-avoidance conflict (Asendorpf, 1990a), sometimes referred to as “conflicted shyness,” in which they want to interact with peers (normative social approach motivation) but are impeded from engaging in normative rates of interaction by their social anxiety and social evaluative concerns (high social avoidance motivation). Two reviewed publications identified children as anxious withdrawn (Goodwin et al., 2004; Jakobsen et al., 2012), but most reviewed publications identified children as socially withdrawn without identifying the reason or motivation for withdrawal. Although a minority of children included in these publications may have demonstrated social withdrawal due to social disinterest (a.k.a. unsociability, low social approach motivation, Asendorpf, 1990a), most children in these publications should have demonstrated social withdrawal due to social anxiety. Another potential reason for social withdrawal is “avoidance” (having high social avoidance motivation paired with low social approach motivation). However, none of the reviewed publications examined adult outcomes of childhood avoidant withdrawal, with the exception of one study of both conflicted and avoidant shyness (Tang et al., 2022). For this study we review results for “conflicted shyness” for the sake of consistency with the most probable reason for withdrawal in other reviewed publications.

This article reviews long-term longitudinal studies in which children were identified as socially withdrawn among peers in general, including with familiar peers such as classmates (Coplan et al., 1994; Gazelle and Ladd, 2003; Gazelle and Rubin, 2010), rather than only with strangers or in novel situations. Although some children who demonstrate social withdrawal among familiar peers also display similar behavior when meeting unfamiliar peers (Gazelle et al., 2005), other children withdraw from unfamiliar peers but not familiar peers (Gazelle and Faldowski, 2014). Therefore, our focus on studies of children identified as withdrawing from peers in general should capture those who withdraw from familiar peers (and may also withdraw from unfamiliar peers), while minimizing inclusion of studies of those who only withdraw from unfamiliar peers. This is important because we conceptualize withdrawal from familiar peers as a manifestation of social evaluative concerns linked with social learning processes (Coplan et al., 1994; Gazelle and Faldowski, 2014; Gazelle and Rubin, 2010), whereas withdrawal from unfamiliar peers only is conceptualized as a temperamentally-based manifestation of wariness in the face of novelty (Kagan et al., 2007). It was necessary to use a broad definition of social withdrawal to ensure inclusion of relevant available research, while nonetheless distinguishing between studies of social withdrawal vs. behavioral inhibition (but allowing inclusion of studies which identify children who demonstrate both).

Reviewing studies of social withdrawal from peers in general was both practically necessary given the state of the literature, and unique relative to previous meta-analyses. Because most reviewed studies identified children who withdrew from peers in general without distinguishing between familiar and unfamiliar peers, it was not possible to examine withdrawal from familiar vs. unfamiliar peers as a factor in meta-analyses. Furthermore, meta-analyses of social anxiety and other anxiety outcomes (including in adulthood) in children who are wary of the unfamiliar (behaviorally inhibited children) have already been conducted (Clauss and Blackford, 2012; Sandstrom et al., 2020), whereas a meta-analysis of anxiety and other psychosocial outcomes in adults with childhood social withdrawal had not been conducted prior to the current investigation.

We reviewed 31 studies which shared a common focus on children who exhibited social withdrawn behavior among peers in general, despite the variety of withdrawal-related terms used. This included studies that identified children not only as socially withdrawn (16 studies) or anxious withdrawn (two studies, Goodwin et al., 2004; Jakobsen et al., 2012), but also with other constructs related to social withdrawal (solitary/solitude, shy, socially inhibited, and socially anxious/phobic) if they referred to children's social behavior in general, rather than only with strangers. Two studies assessed youths' solitude in relation to peers in general (Jaffee et al., 2002; Monk et al., 2023). Seven studies assessed youth shyness as self-consciousness in situations of perceived social evaluation, rather than exclusively in situations of social novelty (Caspi et al., 1989, 1988; Crum et al., 2006; Ensminger et al., 2002; Fothergill and Ensminger, 2006; Kerr et al., 1996; Tang et al., 2022). Another two studies met this definition of shyness and are described in the text, although it was not possible to include them in meta-analyses of variable-oriented analyses because they reported person-oriented analyses (Schmidt et al., 2017; Tang et al., 2017). Four studies assessed social inhibition as uneasy reactions (hesitant approach, fearful, anxious, reserved, and reticent) in social-evaluative or unfamiliar situations (Asendorpf et al., 2008), including with familiar classmates (Asendorpf et al., 2008; Bohlin and Hagekull, 2009; Kagan et al., 1962). Similarly, one study assessed child and adolescent social anxiety disorder (SAD or social phobia) in relation to peers in general (Fichter et al., 2009), rather than exclusively in relation to novel social partners or situations. SAD identifies children who avoid or endure with distress at least some social situations with peers because they are afraid of embarrassing themselves, when this avoidance or distress is sufficiently intense to interfere with their lives (APA., 2022; Gazelle and Rubin, 2010; WHO, 1982). We use the term “social withdrawal” or “social withdrawal and related constructs” to refer to all these terms in studies which share these common criteria.

According to a Developmental Science perspective (Bergman et al., 2000; Lerner et al., 2015; Magnusson and Stattin, 2006), individual development is influenced by the interplay of multi-level systems over time, ranging from those within the individual (e.g., their biological sex), to those in their proximal social context (e.g., interactions and relationships with parents and peers), to broader ecological contexts (e.g., cultural context), and generational time (generation). In this set of meta-analyses, we expect childhood and adolescent social withdrawal to predict adult psychosocial outcomes due to several proximal interpersonal processes. Therefore, to provide a framework for interpreting results, we propose that three interpersonal process mechanisms link childhood vulnerability to adult psychosocial adjustment. Additionally, drawing from “cascade” models, we propose that difficulties in one domain of development (e.g., interpersonal) can cascade or spill over to influence other domains over time (e.g., educational, Masten and Cicchetti, 2010). Additionally, we make the secondary assumption that the extent to which childhood and adolescent social withdrawal predicts adult psychosocial outcomes and the nature of those adult outcomes, may be influenced by systems at both more proximal (within-person) and more distal levels ranging from biological sex to broader context (generation, culture).

This set of meta-analyses address the following research questions: Does childhood social withdrawal predict (1) delayed achievement of adult developmental milestones for participation in families and society, risk for (2) internalizing patterns, including: (2a) adult social withdrawal, (2b) social anxiety sx/dx, (2c) other anxiety sx/dx, (2d) depressive sx/dx, (3) substance use, (4) physical ill-health, and (5) intergenerational difficulties?

We expected childhood social withdrawal to predict delays in adult developmental milestones for participation in families and society, internalizing problems and particularly social anxiety, and intergenerational difficulties in mothers and their offspring. We expected these adult outcomes to stem from the proposed interpersonal processes and developmental cascades. However, we did not expect childhood social withdrawal to predict adult substance abuse because substance abuse is not consistent with the overall over-controlled pattern exhibited by these individuals, nor the reduced likelihood that they would be introduced to substance abuse in social contexts. Additionally, we did not advance a specific hypothesis about adult physical health outcomes because the connection between interpersonal processes and health outcomes was less clear, especially given that the health outcomes investigated to date are diverse.

We expected social withdrawal across a broad range of childhood and adolescence to predict adult psychosocial maladjustment (Fichter et al., 2009). Therefore, we did not advance hypotheses about whether social withdrawal earlier vs. later in development (e.g., in childhood vs. adolescence) would differentially predict adult outcomes. Moreover, most publications included in the meta-analyses were not designed to address this question. Likewise, adult adjustment outcomes were intended to capture patterns of adaptation relevant across development in early to middle adulthood. As such, we also did not advance hypotheses about differential outcomes at different points of adult development.

We also test the secondary research question of whether the relation between childhood social withdrawal and each form of adult psychosocial adjustment was heterogenous, and therefore likely to be moderated (strengthened or weakened) by other (unspecified) factors. If results suggest moderation, although the extent of evidence available was insufficient for meta-analyses of moderated effects, we will review available evidence from individual studies for moderation by factors such as biological sex. Although this review of moderation would not have a meta-analytic basis, it would result from a systematic review of the literature and be intended to provide researchers with guideposts for moderators which have received some support in extant literature.

We followed PRISMA guidelines (Page et al., 2021) for search, selection, and data extraction procedures. Peer-reviewed empirical journal articles and a book published before April 11^th, 2023 were located with searches of a university library catalog, PsychINFO, PubMed, Web of Science, and Google Scholar databases. Keywords for the initial literature search were “social withdrawal,” “anxious withdrawal,” “anxious solitude,” “longitudinal,” “childhood,” and “adult outcomes.” The references of these publications then served as sources of additional articles. Theses, dissertations, non-empirical publications, and retrospective longitudinal studies were not reviewed. Also, studies using person-oriented trajectories as predictors (n = 2 studies) were not included in meta-analyses (but are described in the text) because person-and variable-oriented results could not be combined in weighted means. Articles were selected by analyzing definitions of social withdrawal and related constructs (shyness, solitude, inhibition, wariness, SAD/social phobia) to ensure that they were not operationalized only in reference to “new” or “novel” social partners or situations. Table 1 lists complete inclusion and exclusion criteria.

Of the 18 articles initially selected for review, 16 assessed childhood “social withdrawal,” and two “anxious withdrawal.” To enlarge the evidence base, a second search added the keywords “shyness,” “solitary/solitude,” “inhibition,” “wariness,” “social anxiety,” “social phobia,” “adolescence,” and “peer.” This search yielded 15 additional articles featuring the constructs “shyness” (seven articles), “inhibition” or “inhibition/overcontrol” (four articles), solitude/solitary (two articles), or “SAD” or “social phobia” (two articles). Studies were selected if the withdrawal-related construct definition referred to familiar peers (i.e., classmates) or peers in general (i.e., not only strangers and novel situations). We reviewed only studies that reported specified forms of psychosocial adjustment (Table 1) for adults with childhood social withdrawal.

Guided by inclusion and exclusion criteria (Table 1), the second author (in consultation with first author) screened publications for eligibility based on (1) titles and abstracts, and then (2) the full text of publications. The second author extracted data onto a pre-defined form, including: study information (authors, publication year, social withdrawal construct name, and definition); sample characteristics (size, age at childhood and adult assessments, biological sex proportions, country, and generation/year at first assessment); social withdrawal or related construct assessment (assessment name, informant); and relevant results for adults with childhood social withdrawal (but not aggressive-withdrawal) with notes on control variables and moderated and mediated effects. When publications lacked data, the second author attempted to contact corresponding authors via email. Four authors responded with the necessary information, one author was unreachable, and one author had passed away and his co-authors were not able to provide data. Therefore, it was not possible to include two eligible studies in meta-analysis for insufficient data (Copeland et al., 2014; Schwartzman et al., 2009), but they are nonetheless listed in tables. Thus 33 eligible publications were identified, and we were able to obtain sufficient information to calculate effect sizes for 31 of these publications. Figure 1 summarizes the PRISMA-guided literature search process.

To obtain a uniform effect size statistic for each meta-analysis, each relevant result (summary statistic) from each study was converted to a Fisher's z statistic (see statistical supplement). Fisher's z was chosen because: (1) Pearson's r was the effect size most frequently reported in studies, and its transformation into Fisher's z ensures a normal distribution; (2) Pearson's r and Fisher's z can both capture the link between childhood social withdrawal and adult psychosocial adjustment. Therefore, other effect size statistics were first converted into Pearson's r (Borenstein et al., 2021; Cumming, 2013) and then Fisher's z. Finally, other statistics were first converted to Cohen's d (Thalheimer and Cook, 2002), then Pearson's r (Borenstein et al., 2021; Cumming, 2013), and then Fisher's z. Then after meta-analyses were performed with Fisher's z values, results were converted back to Pearson's r for interpretability (Borenstein et al., 2021). Pearson's r values were interpreted as small (≥0.1), medium (≥0.3), or large (≥0.5) effects (Cohen, 1988). When a study reported multiple effects in the same domain of adult psychosocial adjustment, their mean effect size was entered into meta-analysis.

To generate a weighted mean effect size for each form of adult psychosocial adjustment, clustered effects meta-analyses with bootstrapping were performed in the software platform R (R Core Team, 2019) with the “metafor” (Viechtbauer, 2010) and “boot” (Canty and Ripley, 2024) packages. To account for dependency among multiple publications based on the same longitudinal sample, these correlated hierarchical effect structures (Hedges et al., 2010) were explicitly modeled as clusters (see statistical supplement) in “metafor.” Additionally, to account for the relatively small number of studies available for each meta-analysis, we used bootstrapping (see statistical supplement) (Davison and Hinkley, 1997), which has been shown to control Type 1 error rates while preserving power under these conditions (Davison and Hinkley, 1997; Joshi et al., 2022). Separate meta-analyses were computed for each form of adult psychosocial adjustment: (1) adult developmental milestones; (2) internalizing difficulties (including 2a-d): (2a) social withdrawal; (2b) SAD symptoms (sx)/diagnoses (dx); (2c) other anxiety disorder sx/dx; (2d) depressive sx/dx; (3) substance use sx/dx; (4) physical health; and (5) intergenerational difficulties.

Before computing a weighted mean effect for each outcome, the valence (+ or – sign) of some individual study effects were reversed for consistency with one another when needed. Specifically, effects were recoded to have the same valence (e.g., higher values suggest greater maladjustment).

All eligible studies are listed in Tables 2–6. Those eligible studies entered into meta-analyses are marked with an ^*in the reference section. A few eligible studies could not be entered into meta-analyses either because multiple publications reported effects for the same adult outcome for the same sample (a duplicate effect not entered into meta-analyses is marked with an × in Table 2), or because insufficient information was available to calculate an effect size [marked with an φ in Table 3 (other anxiety, depression)].

Each meta-analysis was assessed for publication bias statistically via Egger's regression and Begg's rank correlation tests (see statistical supplement) of asymmetry in funnel plots computed in Excel with the Meta-Essentials workbook. Egger's regression (t) and Begg's rank correlation (z) tests were not significant, indicating no significant deviation from symmetry in funnel plots (i.e., no significant bias; p > 0.05 for t tests; p > 0.10 for z tests). Specifically, funnel plots were not significantly asymmetric for adult (1) developmental milestones (t = 2.22, p = 0.08; z = 1.05, p = 0.29), (2a) social withdrawal (t = 2.27, p = 0.26; z = 0.52, p = 0.60), (2b) SAD sx/dx (t = 0.94, p = 0.52; z = 1.57, p = 0.12), (2c) other anxiety disorder sx/dx (t = −0.20, p = 0.86; z = 0.00, p = 1.00), (2d) depressive sx/dx (t = 0.20, p = 0.86; z = −0.49, p = 0.96), (3) substance use (t = 0.30, p = 0.78; z = −0.75, p = 0.45), (4) physical health (t = 1.10, p = 0.35; z = 0.49, p = 0.62), and (5) intergenerational difficulties (t = 0.78, p = 0.49; z = 0.73, p = 0.46). Therefore, findings indicate no significant deviation from symmetry in statistical tests, supporting the absence of significant publication bias (Lin and Chu, 2018) and other threats to reliability (such as unreliable results from small studies, data irregularities, and chance results, Sterne and Harbord, 2004). These results support the validity of meta-analytic results.

To examine the potential for moderation of effect sizes included in meta-analyses, we analyzed heterogeneity among effect sizes for each outcome. Analyses suggest significant heterogeneity among effect sizes for adult (1) developmental milestones (Q = 66.71, p < 0.001), (2a) social withdrawal (Q = 5.42, p < 0.01), (2b) SAD sx/dx (Q = 8.04, p < 0.001), (2c) other anxiety sx/dx (Q = 92.93, p < 0.001), (2d) depressive sx/dx (Q = 58.08, p < 0.001), (3) substance use (Q = 157.95, p < 0.001), and (4) physical health (Q = 257.07, p < 0.001). No significant heterogeneity was found for intergenerational difficulties (Q = 3.59, p = 0.46), perhaps because all publications on intergenerational difficulties were derived from a single longitudinal sample. Heterogeneity for most forms of adult adjustment suggests that moderated meta-analyses might be fruitful in the future. At present, moderated meta-analyses were not possible because of the limited number of studies available for review (i.e., < 4 effect sizes per level of moderator, Pearcey et al., 2021). Given this support for moderation in most adult outcomes, in addition to reporting meta-analytic results, we review evidence for moderation from individual studies when available.

We identified 33 eligible empirical publications reporting on prospective longitudinal studies of psychosocial adjustment in adults with childhood social withdrawal published over the past 61 years from 1962 to 2023. Of these studies, it was possible to obtain sufficient information to calculate effect sizes for 31 studies (listed in Tables 2–6, references marked with ^*in reference section) which were based on 14 long-term longitudinal samples.

Meta-analytic results (Tables 2–6) revealed that child social withdrawal significantly predicted: (1) delayed adult developmental milestones (r = 0.20, p < 0.05); more adult (2) internalizing tendencies (r = 0.24, p < 0.01), including: (2a) social withdrawal (r = 0.29, p < 0.0001); (2b) SAD sx/dx (r = 0.47 p < 0.0001); (2c) other anxiety sx/dx (r = 0.17, p < 0.05); and (2d) depressive sx/dx (r = 0.17, p < 0.01); and (3) intergenerational difficulties (r = 0.32, p < 0.0001). However, adults with childhood social withdrawal were less likely than others to abuse substances (r = −0.16, p < 0.01). Effects were of medium size (≥0.3) for adult SAD sx/dx and intergenerational difficulties, whereas effects were of small size (≥0.1) for adult developmental milestones, internalizing tendencies, social withdrawal, other anxiety sx/dx, depressive sx/dx, and substance abuse (Tables 2–6, Cohen, 1988).

Meta-analytic results also suggest that childhood social withdrawal was not significantly related to physical health (r = 0.13, ns; Table 5). However, this result may be influenced by biological sex and heterogeneity of physical health outcomes.

Tables 2–6 each list the following for the adult outcome identified at the top of the table: number of publications reporting on the outcome, number of unique samples on which those publications were based, and combined sample size across those unique samples (at the top of the table), and weighted mean clustered effect size (r) across all publications for the specified form of adult adjustment, 95% confidence interval, test statistic, and significance level (at the bottom of the table), and for each study reporting results for a particular adult outcome (in each row within the table): reference (first author and year only), sample size and biological sex percentage, child predictor construct, assessment name and informant for child predictor, young person's age in years at childhood and adult assessments, country from which sample was drawn, name of the longitudinal sample (for shared samples); calendar year at first assessment, mediator and/or moderator effects (if reported), form of adult adjustment, effect size (Pearson's r between childhood social withdrawal and adult outcome) for each relevant adult adjustment result from each publication.

In this section we provide a detailed account of each meta-analysis, including characterizing the evidence-base for each meta-analysis: the number of publications, number of unique longitudinal samples, and combined sample size on which each meta-analysis was based. We also describe specific types of adult psychosocial adjustment included in each meta-analysis.

When available, we highlight odds ratios and statistical controls for individual studies, and describe developmental trajectories that could not be included in meta-analyses. Given support for moderated effects for all adult outcomes (except intergenerational effects) in preliminary analyses, we also review patterns that suggest moderation by biological sex (and generation for adult developmental milestones) in individual study results.

Meta-analytic results support childhood social withdrawal as a medium-sized risk factor for adult SAD sx/dx and intergenerational difficulties in mothers with childhood histories of social withdrawal and their offspring; as well as a small-size risk factor for delayed adult developmental milestones, social withdrawal, other anxiety disorders, and depression. Meta-analytic results also support childhood social withdrawal as a small-size protective factor for substance abuse. Each of these forms of adjustment in adults with childhood social withdrawal is discussed in turn in the following sections.

Analyses of heterogeneity of effect sizes among individual studies contributing to each form of adult adjustment (except for intergenerational effects) suggests that the relation between childhood social withdrawal and adult outcomes was likely to be moderated by other factors, but available information was insufficient for moderated meta-analyses. Therefore, we also review available evidence from individual studies for moderation by biological sex (and generation for adult developmental milestones).

Consistent with the proposed missing out mechanism, meta-analytic results support childhood social withdrawal as a small-size risk factor for delayed adult developmental milestones for participation in families and society. The modest size of meta-analytic effects for adult developmental milestones may be due to their heterogeneity (e.g., employment, romantic relationships), and potential biological sex and generational moderation. Many individual studies indicated that childhood social withdrawal predicted delays in adult developmental milestones specific to biological sex, and the pattern of findings across studies with samples from different generations suggests that generational shifts may have occurred in some of these patterns over time.

In earlier generations, the pattern of individual study findings suggests that American women and men with childhood histories of social withdrawal, compared to their non-withdrawn same-sex counterparts, demonstrated different occupational delays (for women abbreviated or no work history, for men delayed entry into a stable career), and men demonstrated delayed romantic partnerships and procreation. In recent generations, both women and men with childhood histories of social withdrawal demonstrated similar delays in entering the work force and adult financial milestones, and men continued to demonstrate delayed romantic partnerships. Therefore, individual study patterns suggest potential moderation by biological sex and generation in which women as well as men now experience similarly delayed adult occupational and financial milestones. This potential generational shift suggests that childhood social withdrawal was detrimental for women in recent generations who were expected to take initiative in occupational and financial domains based on changed cultural norms. This pattern is also consistent with the possibility that more change in women's roles may have occurred in Western countries at a societal level than in heterosexual romantic and familial relationships. Nonetheless, because recent studies were conducted abroad, study of contemporary American adults with childhood social withdrawal is needed.

The pattern of results across investigations also suggests that the link between childhood social withdrawal and certain adult developmental milestones demonstrated some cultural specificity within Western countries. In particular, men with childhood social withdrawal demonstrated less occupational stability and achievement in the US (Caspi et al., 1989, 1988), but not Sweden (Kerr et al., 1996). This pattern may reflect greater demands for men's self-assertion in American vs. Swedish workplaces (Kerr et al., 1996). These patterns suggest that adults with childhood social withdrawal may find American workplaces inaccessible or that their competitive culture is incompatible with their interaction style. Although recent research supports delayed workforce entry for adults of both sexes with childhood withdrawal in Germany and Canada (Asendorpf et al., 2008; Denissen et al., 2008; Schmidt et al., 2017), further research is needed to examine whether such occupational delays occur for contemporary American adults.

Despite these occupational patterns, women with vs. without childhood social withdrawal did not demonstrate lower educational attainment in the US, but attained less education in Sweden (Kerr et al., 1996) and Canada (Serbin et al., 1998, 2011), and demonstrated poor educational achievement in Canada (Serbin et al., 1998, 2011). Therefore, despite the competitiveness of American work culture, the range of educational opportunities available in the US may support educational parity for adults with childhood histories of social withdrawal. As more studies become available from diverse cultures in the future, more nuanced meta-analytic results may be obtained by separate moderated meta-analysis of specific types of adult developmental milestones.

Consistent with expectations, childhood social withdrawal was a medium-size predictor of adult SAD sx/dx, and small-sized predictor of adult social withdrawal, other anxiety (e.g., GAD, OCD, specific phobias, panic disorder, and agoraphobia), depressive sx/dx, and all of these internalizing difficulties combined. These long-term longitudinal linkages between childhood social withdrawal and subsequent adult internalizing psychopathology reflect more than homotypic continuity. For childhood social withdrawal, which is not a diagnosis of psychopathology, to predict adult psychopathology it must translate into significant distress and impairment in functioning (i.e., life interference, Gazelle and Rubin, 2010). Socially withdrawn children and adolescents may experience distress when they experience significant interpersonal stress (e.g., peer exclusion and victimization) and/or lack of support (e.g., low maternal sensitivity). These patterns are compatible with diathesis (vulnerability)-stress processes in which individuals with childhood social withdrawal may develop psychopathology in the context of interpersonal stress, or experience mental health in the context of interpersonal support and low interpersonal stress (Gazelle and Rubin, 2019). Importantly diathesis-stress models of the development of psychopathology should nonetheless be informed by a broader Developmental Science perspective to provide insight on developmentally-relevant stressors (e.g., peer difficulties) and vulnerable periods of development (e.g., school transitions, puberty). Much is left to be understood about the circumstances under which individuals demonstrate continuity vs. change in social withdrawal from childhood and adolescence to adulthood and develop internalizing psychopathology or relative mental health by adulthood.

Parenting moderated the link between child social withdrawal and adult internalizing psychopathology in one reviewed publication. Low maternal sensitivity at age 4 strengthened (moderated) the relation between childhood shyness/inhibition and adult social anxiety and depression (Table 3, Bohlin and Hagekull, 2009). In the future, other interpersonal influences (e.g., peer relations) on the development of internalizing psychopathology should be investigated in adults with childhood social withdrawal.

With few exceptions, internalizing problems occurred in both women and men with childhood social withdrawal. Although extant research indicates that women compared to men experience higher rates of internalizing problems overall (Albano and Krain, 2005), results of individual studies included in our meta-analysis indicated that adults of both sexes who experienced childhood social withdrawal were at risk for adult internalizing psychopathology. This lack of biological sex differences suggests that childhood patterns of social withdrawal identify individuals who, perhaps regardless of their biological sex, have a long-term tendency to respond to stress in ways that render them vulnerable to internalizing difficulties (e.g., rumination).

Nonetheless, women and men with childhood social withdrawal may develop internalizing problems through somewhat different mechanisms. The only two instances of gender moderation of risk for adult internalizing psychopathology in individual reviewed publications involved moderation of indirect or mediated effects. This may have occurred because girls and women may weight their relationships more strongly in their self-appraisals and, therefore, are more likely to develop internalizing problems when they experience interpersonal difficulties (Albano and Krain, 2005), whereas boys and men may weight their achievement at school and work more strongly in their self-appraisals.

Consistent with the proposed interpersonal stress and relationship quality mechanisms, meta-analysis revealed that childhood social withdrawal was a medium-size predictor of intergenerational difficulties in mothers and their offspring. Women with childhood social withdrawal demonstrated unresponsive or intrusive parenting of their offspring, and their offspring responded to them with non-compliance, defiance, and aggression (Table 6). This parenting pattern may reflect a general tendency to enact ineffective strategies for pursuing social goals. Offspring's responses may suggest both withdrawn and externalizing behavior, but it is not clear whether these behaviors generalize beyond interactions with their mother. These patterns suggest that relationship difficulties are central to adult maladjustment and intergenerational transmission of risk in women with childhood histories of social withdrawal. However, as these intergenerational investigations were exclusively conducted with the urban, low-income, French Canadian CLP sample, it will be important to investigate whether they generalize to other contexts and samples.

Meta-analytic results support childhood social withdrawal (controlling for childhood aggression) as a small-size protective factor against adult substance use. All studies included in the meta-analysis reported that childhood social withdrawal predicted less adult substance use.

Meta-analytic evidence that childhood social withdrawal protects against adult substance use is important because it contradicts suggestions that anxious individuals might use substances to self-medicate (Turner et al., 2018). Rather, childhood social withdrawal may protect adults from substance use by reducing the likelihood that they are introduced to such behavior via social situations (e.g., parties) and friends (Foster et al., 2018; Martin-Storey et al., 2011), consistent with the missing out mechanism. Additionally, adults with childhood social withdrawal likely tend to have over-controlled tendencies, and may dislike the prospect of losing control via substance use.

One study not included in the meta-analysis (due to person-oriented results that could not be combined with variable-oriented results from other studies in a weighted mean) found that childhood shyness was related to increased risk for substance use (Tang et al., 2017). However, this finding was likely due to no control for childhood aggression. Therefore, aggressive-withdrawn children (Gazelle and Ladd, 2003; Ladd and Burgess, 1999) may be at risk for substance use. This pattern resembles that in Foster et al. (2018), in which the non-shared variance between childhood social withdrawal and externalizing problems reduced risk for adult alcohol problems, whereas their shared variance (perhaps negative affect) increased risk for adult alcohol problems.

Meta-analytic results did not suggest that childhood social withdrawal predicts adult physical health. However, the health conditions investigated in available publications were heterogeneous. The pattern of individual study results suggest that childhood social withdrawal predicted higher BMI and more dental visits in adults, headaches and pregnancy difficulties in women, but fewer adult injuries in men (Table 5). Social withdrawal may have reduced men's inclination to engage in risky behavior, resulting in fewer injuries. Therefore, adults with childhood social withdrawal may be less likely to engage in adult health risk behaviors (i.e., physical injury for men and substance use for both sexes). In the future, as more studies become available, moderated (biological sex) meta-analyses conducted separately by type of health condition will be warranted.

This is the first systematic review and meta-analysis of psychosocial adjustment in adults with childhood social withdrawal. This set of meta-analyses of prospective longitudinal studies contributes to current knowledge by supporting childhood social withdrawal as a risk factor for multiple forms of adult psychosocial maladjustment, but a protective factor against adult substance abuse.

The focus of this set of meta-analyses of childhood social withdrawal among peers is unique, as previous meta-analyses on the most closely related topics have focused on behavioral inhibition to unfamiliar people, objects, and situations as a risk for subsequent anxiety (Clauss and Blackford, 2012; Sandstrom et al., 2020). We reviewed longitudinal studies of adjustment in adults with childhood social withdrawal and related constructs in relation to peers in general, rather than exclusively in relation to unfamiliar social partners. We imposed as much clarity as possible on the operationalization of social withdrawal by the theoretically-motivated selection of studies which had assessed social withdrawal in relation to peers in general, including familiar peers such as classmates (not only unfamiliar social partners), to emphasize functioning among familiar peers. We encourage future investigators to assess children's withdrawal in relation to familiar peers because this is most likely to impact psychosocial development (Asendorpf, 1990b; Gazelle and Faldowski, 2014). We also encourage assessment of anxious and solitary behaviors, in addition to socially anxious affect and self-conscious thoughts. These methods may yield stronger, more consistent future findings.

The number of publications and longitudinal samples upon which they were based, and combined sample size across those publications was greater for some adult outcomes than others (Tables 2–6). Therefore, to ensure the accuracy of results, we conducted statistical tests with methods that are well-suited to small samples (i.e., bootstrapping, Joshi et al., 2022). Importantly, effect sizes and the significance of effects appear to correspond to the nature of each outcome rather than the number of publications, number of longitudinal samples, or combined sample size for each outcome. For instance, meta-analysis revealed a highly significant medium size effect for adult social anxiety based on 3 publications derived from 3 longitudinal samples and a combined sample size of 1,419 adults, and this was reinforced by a highly significant small size effect for the umbrella construct of internalizing difficulties (which included social anxiety) which was based on 12 publications derived from 9 longitudinal samples and a combined sample size of 4,361 adults. In contrast, meta-analysis yielded a non-significant effect for adult physical health despite being based on 5 publications derived from 2 longitudinal samples and a combined sample size of 13,404 adults. Importantly, this pattern of results was also in line with our hypotheses.

Likewise, although meta-analysis of intergenerational outcomes was based on 5 studies and a combined sample of 109 mothers and 175 offspring, it yielded a highly significant medium sized effect. Nonetheless, because the data for this intergenerational effect is based on a single longitudinal sample, future replication of results in other samples would increase confidence in intergenerational effects.

The studies included in the meta-analyses assessed social withdrawal across a broad range of ages in childhood and adolescence through 16 years of age. Because the focus of this paper is on long term longitudinal studies, some studies used childhood and adolescent social withdrawal predictors that were based on repeated measurements across childhood and adolescence (Jaffee et al., 2002; Kerr et al., 1996). Also, other studies used social withdrawal predictors that were composites of repeated assessments at ages 6 and younger (Asendorpf et al., 2008; Denissen et al., 2008; Tang et al., 2022), whereas other studies assessed children at one time point at ages ranging from 6 to older (CLP studies, Crum et al., 2006; Ensminger et al., 2002; Fothergill and Ensminger, 2006), so that there was overlap at 6 years of age even between studies that assessed children at mostly younger or older ages. Therefore, it was not possible to enter the age at which social withdrawal was assessed in childhood or adolescence in the analyses.

Moreover, we did not advance hypotheses about the relation between age of social withdrawal assessment in childhood or adolescence and prediction of adult outcomes. We did not advance such hypotheses because previous research had shown that social withdrawal assessed in different developmental periods (both middle childhood and early adolescence) are both predictive of psychosocial adjustment difficulties in early adolescence (Gazelle and Faldowski, 2019). This is partially due to continuity in social withdrawal from middle childhood through early adolescence, but also when social withdrawal emerged in early adolescence it was predictive of concurrent adjustment difficulties (Gazelle and Faldowski, 2019). Similar comparisons of childhood and adolescent developmental periods have not been tested for prediction of adult outcomes (for an exception see Fichter et al., 2009), but based on these findings earlier in development, as well as one study of adult outcomes (Fichter et al., 2009), we would expect that social withdrawal across a broad range of ages in childhood and adolescence would predict adult psychosocial difficulties. Many investigations support this assumption, as many of the investigations included in our meta-analyses successfully utilized assessments across childhood and adolescence as predictors of adult adjustment (Jaffee et al., 2002; Kerr et al., 1996).

The studies included in our meta-analyses also predicted outcomes across a broad range of ages in adulthood (and late adolescence from 17 to 40 years of age). Because the focus of this paper is on long term longitudinal studies, an overarching pattern (e.g., delay in adult developmental milestones) is often assessed across a broad span of adulthood (and late adolescence), but the manifestation of that pattern was assessed in a developmentally appropriate manner (e.g., delay in starting first job in late adolescence and early adulthood vs. delay in achieving a stable career in middle adulthood). Thus, the reviewed studies were able to capture an overarching pattern of psychosocial adjustment despite variation in specific manifestations at different points of adult development. Other outcomes (e.g., internalizing problems) have stable assessments across adulthood, but these assessments flexibly allow for different manifestations across adult development (e.g., life interference at earlier vs. later ages may be manifested at school vs. work or in dating vs. marital relationships). Thus, although the extent of the research available at this point does not allow for a comparison of adult outcomes during different periods of adult development, it does an excellent job of capturing adaptational patterns relevant across adult development.

All studies which met our review criteria were conducted in Western countries. No studies were excluded due to non-English language. In the future prospective long-term longitudinal studies of socially withdrawn children should be conducted in diverse global contexts to achieve a fuller understanding of contextual influences on life-course development. This is important because the implications of childhood social withdrawal for children's peer relationships are culturally specific (Chen and French, 2008) and this may also translate into culturally-specific life-course outcomes. Social withdrawal also has culturally-specific manifestations, such as “hikikomori” in Japan, in which young people physically isolate themselves in their room at home and do not go out into society (Furlong, 2008). Such culturally-specific manifestations of social withdrawal may be especially likely to translate into culturally-specific life-course outcomes.

Most studies of psychosocial adjustment in adults with childhood social withdrawal used designs which compare adults with vs. without childhood social withdrawal. However, designs are needed that compare adults with childhood social withdrawal who demonstrate psychosocial maladjustment to their counterparts who demonstrate adaptive development. This requires testing mediators and moderators of adjustment over time.

This review revealed that few studies have tested interpersonal mediators of the relation between childhood withdrawal and adult psychosocial adjustment (Tables 2–6). Such research is needed to support efforts to interrupt interpersonal processes that lead to psychosocial maladjustment in adults with childhood social withdrawal.

Contemporary research is increasingly investigating biological mechanisms of anxiety development. Although no reviewed studies examined biological mechanisms in anxiety development (despite recent coverage of biological criterions such as BMI), individuals with increasing shyness trajectories demonstrated adult hypervigilance to angry faces (Tang et al., 2017). Such findings are typically interpreted as evidence for a temperamentally based attentional “bias” toward threat which is thought to be a mechanism involved in the development of anxiety disorders. However, attentional patterns can be learned in threatening environments (Shackman et al., 2007) and may be adaptive in such environments, but maladaptive when they transfer to other environments. Therefore, future studies should investigate both environmental and biological origins of sensitivity toward threat and integrate biological assessment (including genetic vulnerability, Hoekstra et al., 2008) into investigations of the life course development of socially withdrawn children.