Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2025 Jul 16.
Published before final editing as: Int J Behav Dev. 2024 Oct 14:10.1177/01650254241287220. doi: 10.1177/01650254241287220

Subtypes of Childhood Social Withdrawal and Adult Relationship and Parenting Outcomes

Emily B Reilly 1, Kenneth A Dodge 1,2, Yu Bai 1, Jennifer E Lansford 1, John E Bates 3, Gregory S Pettit 4
PMCID: PMC12266682  NIHMSID: NIHMS2022716  PMID: 40677633

Abstract

The aims of the current 30-year prospective study were to determine: 1) whether socially withdrawn kindergarten children are less likely than others to enter serious romantic relationships or become parents by age 34, 2) whether socially withdrawn children parent differently than non-withdrawn individuals when they grow up, and 3) whether subtypes of withdrawal are associated with different adult outcomes. Following Harrist et al. (1997), 558 kindergarten children (81% White, 17% Black) were categorized into one of five groups: four clusters of social withdrawal (n = 95 unsociable, 23 passive-anxious, 18 active-isolate, 25 sad/depressed) or non-withdrawal (n = 397), using directly observed school behavior and teacher ratings. About 30 years later (Mage = 34.45 years, SD = 0.62 years), participants self-reported on their romantic relationship and parent status and parenting warmth and harshness. Overall, the group of socially withdrawn children was no more or less likely than the non-withdrawn group to be in a current relationship or a parent, nor did they report any differences in parenting. However, the active-isolate subtype of social withdrawal, characterized by impulsivity and anger, was less likely than the non-withdrawn group to be in a current relationship (B = −1.24, p < 0.05). This study suggests socially withdrawn children in the U.S. fare similarly to non-withdrawn peers in adulthood in their romantic relationships and parenting, but a subgroup of active-isolate children may be at risk of not entering adult relationships.

Keywords: social withdrawal, romantic relationship, parenting, longitudinal

One of the seminal contributions of Dr. Kenneth Rubin was to describe how some children are more socially withdrawn than other children and interact less with their peers. Rubin developed and implemented methods of assessing withdrawal with teacher and parent report as well as direct observations and peer ratings (Rubin et al., 2009). Individual differences in social withdrawal can be seen in preschool (e.g., Nelson et al., 2005) and although social withdrawal is somewhat unstable in elementary ages (Booth-LaForce & Oxford, 2008), it is more stable among the most socially withdrawn children (e.g., Schneider et al., 1998). Rubin also demonstrated that although withdrawn children all engage less with peers, there may be different reasons for this behavior that may be unobservable (Rubin et al., 2021).

The different reasons behind withdrawn behavior can be thought of as subtypes of social withdrawal. Some children may be timid and shy, not interacting with peers because they are anxious or fearful (i.e., passive-anxious; Harrist et al., 1997; Rubin & Mills, 1988). Other children may struggle with anger or impulsivity and be rejected by peers, engaging in more solitary play as a result (i.e., active-isolate; Harrist et al., 1997; Rubin & Mills, 1988). Still others may experience feelings of sadness and perhaps do not seek out social interactions with peers (i.e., sad/depressed; Harrist et al., 1997). Finally, some withdrawn children demonstrate good social skills but may simply be less interested in social interactions (i.e., unsociable; Harrist et al., 1997; Rubin, 1982). An alternative subtyping of social withdrawal using approach-avoidance theory (Asendorpf, 1990) includes shy (i.e., high approach, high avoidance), avoidant (i.e., low approach, high avoidance), and unsocial (i.e., low approach, low avoidance) subtypes. These subtypes align conceptually with the passive-anxious, sad/depressed, and unsociable subtypes (Coplan et al., 2013). Importantly, the differences between these subtypes may be more distinct depending on the cultural context and what social behaviors are valued (Chen, 2015). Even though all these children would be deemed socially withdrawn, these differences could lead to different outcomes. The goal of the current manuscript was to follow subgroups of socially withdrawn kindergarten children into adulthood to examine outcomes in marriage, parenthood, and parenting.

Social Withdrawal in Childhood

Children who are socially withdrawn tend to have different childhood experiences than non-withdrawn children. For example, withdrawn children tend to struggle socially (Coplan et al., 2013) and experience more peer victimization (e.g., Barzeva et al., 2020) and lower quality friendships (e.g., Shin, 2007). However, these experiences may vary by subtype of withdrawal. In multiple societies, shy and avoidant withdrawn children tend to experience more exclusion and loneliness (e.g., Bowker & Raja, 2010; Chen et al., 2015; Valdivia et al., 2005). Shy withdrawn children also tend to struggle with anxiety and depression more than unsociable children (Chen et al., 2015; Coplan et al., 2013). The values attributed to these behaviors in each cultural context are relevant to the experiences of withdrawn children (Chen, 2015). For example, European American and Mainland Chinese mothers provide different reasons for child social withdrawal (Cheah & Rubin, 2004). And for Chinese children, the unsocial subtype is related to problems in school and with peers (Chen et al., 2015; Coplan et al., 2016). Social withdrawal is not particular to a certain gender, but boys are more likely to be of the active-isolate subtype (Harrist et al., 1997) and withdrawn boys may be more likely to experience negative outcomes (Rubin et al., 2009).

Socially withdrawn children may also experience different home lives compared to non-withdrawn children. Specifically, withdrawn children tend to experience more insensitive parenting (e.g., Booth-LaForce & Oxford, 2008; Mills & Rubin, 1998), although very little research has examined differences in parenting experiences among withdrawal subtypes. Parenting experiences may be a dynamic process with children responding to parenting behaviors and parents modifying their behavior based on their child’s withdrawal (Lin et al., 2020). Research has also considered the role of socioeconomic backgrounds in child social withdrawal. Some evidence demonstrates children from lower socioeconomic backgrounds are less likely to engage in large peer groups (Schneider et al., 2000), but some demonstrates no difference (Serbin et al., 2011). Overall, there do appear to be some different childhood experiences among socially withdrawn individuals. The next question is whether differences are present in adulthood experiences.

Social Withdrawal in Adulthood

Few studies have followed withdrawn children over long periods of time, but there is reason to hypothesize that how withdrawn children fare in adulthood may depend on their subtype of withdrawal. Shy and avoidant withdrawn, but not unsocial young adults, report lower wellbeing and self-esteem (Closson et al., 2022; Nelson, 2013). Socially withdrawn children could be delayed in reaching certain life milestones, such as romantic relationships, given their struggles with peers in childhood. Indeed, withdrawn individuals tend to enter romantic relationships later in adolescence and experience shorter relationships (Barzeva et al., 2021). Shy and avoidant adults also report lower relationship satisfaction compared to unsocial adults (Nelson et al., 2021). The studies mentioned all assessed social withdrawal in adulthood and were cross-sectional in design, apart from Barzeva and colleagues (2021) who measured withdrawal in a prospective longitudinal study of young adults. Despite the lack of research on adult relationship status among withdrawal subtypes, there is evidence that shy, rejected, and depressed children experience differences in life course outcomes. For example, although shy children are as likely as others to be married by their early 30s (Schmidt et al., 2017), shy boys tend to get married at a later age than non-shy boys (Caspi et al., 1988). Contrastingly, depressed adolescents are more likely to get married as young adults (Gotlib et al., 1998). Still, it remains unknown whether children with certain subtypes of withdrawal are less likely to be in a serious adult relationship.

Romantic relationship struggles could result in socially withdrawn individuals having fewer children or children later in adulthood. Although childhood withdrawal is not associated with early parenthood, based on a prospective longitudinal study (Serbin et al., 2011), whether withdrawn individuals are delayed in becoming parents is unknown. Studies on shy individuals suggest that this may be true, with shy boys becoming parents later in adulthood (Caspi et al., 1988). More impulsive adolescents tend to engage in more risky sexual behaviors (Dir et al., 2014), but whether these patterns are true for the shy and active-isolate subtypes of withdrawal is unknown. And if these socially withdrawn individuals do become parents, would their parenting behaviors differ? Withdrawn children tend to experience more insensitive parenting, but this may or may not translate into an increased likelihood of becoming insensitive parents. Evidence outside the withdrawal literature suggests parenting differences for certain subtypes: current maternal depression is associated with less sensitive parenting (Bernard et al., 2018), current shyness-anxiety is associated with more overprotective parenting (Root et al., 2016), the expression of anger is associated with child abuse potential (Rodriguez & Green, 1997), and impulsivity is associated with less positive reinforcement (Chen & Johnston, 2007). Nevertheless, withdrawn kindergarten children with these types of behaviors may not display these behaviors in adulthood. Unsociable children with good social skills may be most likely to demonstrate positive parenting, based on evidence that high social competence explains the intergenerational transmission of positive parenting (Shaffer et al., 2009). We are unaware of any study to date that has examined parenting behaviors in parents who were socially withdrawn as children or considered possible differences between withdrawal subtypes in becoming a parent or parenting behavior.

The Present Study

To understand better the adult experiences of socially withdrawn children, the current study aims to test: 1) whether socially withdrawn children grow up to be less likely than non-withdrawn children to be in serious romantic relationships or become parents, 2) whether socially withdrawn children grow up to parent differently than non-withdrawn individuals, and 3) whether these outcomes vary by subtypes of social withdrawal. Based on findings of delays entering romantic relationships in adolescence, it was hypothesized that socially withdrawn individuals would be less likely to be in a current romantic relationship and less likely to be a parent than non-withdrawn individuals. Second, we also expected there would be no differences in parenting warmth or harshness between withdrawn and non-withdrawn individuals, because this broad group includes all subtypes of withdrawal. Third, we expected that the passive-anxious, active-isolate, and sad/depressed subtypes would be less likely to be in a romantic relationship and be a parent than the unsociable subtype, the sad/depressed subtype would report lower parental warmth, and the active-isolate subtype would report both lower warmth and higher harshness than the other subtypes. These hypotheses were based on the research on adult relationship satisfaction and parent depression, anger, and impulsivity described above.

Methods

Participants

Participants were drawn from a larger pool of participants from the prospective Child Development Project study (Dodge et al., 1990). 558 children out of the initial entire sample of 585 children had kindergarten social withdrawal and teacher-reported behavior data and were included in the present study. Participants were recruited during kindergarten preregistration or the first day of school from several schools in three different U.S. cities (n = 170 Bloomington, Indiana; 193 Knoxville, Tennessee; 195 Nashville, Tennessee) to include a variety of socioeconomic backgrounds, across two cohorts (n = 294 from 1987, 264 from 1988). Participant demographics were representative of the geographic regions at the time and mostly white (Table 1). Of the 558 participants with childhood withdrawal data, 427 (76.52%) completed the 30-year follow-up online survey between 2016 and 2019 across the three sites (Mage = 34.45 years, SD = 0.63 years, range = 32.79 – 37.64 years). In-home interviews with participants each summer and school visits each year through adolescence aided longitudinal retention (Lansford et al., 2006). Participants who completed the 30-year follow-up were more likely than the kindergarten sample to have higher family of origin socioeconomic status (SES; t = 4.09, p < 0.001), to be female (χ2 (1, 558) = 17.36, p < 0.001), and to be white (F = 4.92, p < 0.05). Participants provided written consent at each wave of data collection and the current IRB approval was granted from Duke University (protocol 2017-1230) most recently in March 2024.

Table 1.

Descriptive statistics, N = 558

N Percentage
Sex
Female 268 48.03%
Male 290 51.97%
Race
White 451 80.82%
Black 96 17.20%
Other 11 1.97%
Social withdrawal
Withdrawn 161 28.85%
Non-withdrawn 397 71.15%
Relationship status
Single, never married 110 25.82%
Former relationship 38 8.92%
Current relationship 278 65.26%
Parent status
Parent 299 70.02%
Not a parent 128 29.98%
Mean (SD) Range
Family of origin SES 39.41 (14.08) 8.00 – 66.00
Number of children 1.49 (1.36) 0 – 6
Parenting warmth 2.74 (0.62) 0.90 – 4.00
Parenting harshness 0.58 (0.42) 0.00 – 3.00
Open in a new tab

Measures

Kindergarten Measures

Demographics.

Parents reported on demographics in kindergarten, including family SES and child sex (at that time, options of male and female), race (options of Black, White, and other), and ethnicity. SES was measured using the Hollingshead (1975) index of social status based on parent education and occupation.

Observed Social Withdrawal.

Child social interactions in kindergarten were directly observed by trained experimenters who documented the frequency with which children engaged in solitary focused play and solitary unfocused play, among other behaviors, following procedures described by Harrist and colleagues (1997). Observations of child free play in the classroom and on the playground occurred in five-minute intervals over 12 coding sessions over six days. The number of ten-second intervals of solitary play within each five-minute coding session were averaged across the 12 sessions. Interrater reliability for all observations was adequate (kappa = 0.59 to 0.74). Following the procedures by Harrist and colleagues (1997), a social withdrawal score was calculated as the mean of solitary focused and solitary unfocused play for children with school observation data (n = 567). Children who scored greater than 0.5 standard deviation above the sample mean were categorized as socially withdrawn (n = 163), and others were categorized as non-withdrawn (n = 404). As expected, socially withdrawn children engaged in about twice as much solitary play (M = 6.05, SD = 2.92) as non-withdrawn children (M = 3.03, SD = 2.27; t = 11.86, p < 0.001).

Teacher Reports of Behavior.

Teacher reports of child behavior in kindergarten were used to create clustering items to establish subtypes of social withdrawal. Teachers completed questionnaires on the child’s behavior at school. A subset of items from the Teacher Report Form of the Child Behavior Checklist (Achenbach, 1991) and the Teacher Checklist of Peer Relationships (Dodge & Coie, 1987) were included (see Harrist et al., 1997 for list of items). The following seven clustering variables were created by calculating the mean of the relevant items: isolates self (α = 0.73), timid (α = 0.75), anxious (α = 0.73), immature (α = 0.62), sad/depressed (single item), lacks restraint (α = 0.89), and angry/defiant (α = 0.89). Isolates self and lacks restraint were calculated after z-scoring items because these variables included items from two different scales. The remaining non-standardized clustering variables were then z-scored. 558 of the 567 observed participants had teacher-reported behavior data across all clustering variables.

Adulthood Measures

Relationship and Parent Status.

At the 30-year follow-up, participants self-reported their romantic relationship and parent status. Relationship status responses were recoded into 0 = single/never married (n = 110), 1 = formerly married (n = 35 separated or divorced, 3 widowed), and 2 = current relationship (n = 231 married, 47 living with partner). Parent status was coded as 0 = non-parent and 1 = parent (n = 299), including a small number of participants with only non-biological children (n = 17).

Parenting Behavior.

Parenting warmth and harshness were assessed via self-report using the 27-item Parent Practices Scale (Kersteter, 2005; Strayhorn & Weidman, 1988) by participants who were parents. Eight parents who reported they were not involved with or never saw their child were excluded from the parenting behavior analyses. Participants reported on the frequency they engaged in various actions with their first child on a scale of “never” to “many times a day.” Parenting warmth was calculated as the mean of 11 items (α = 0.83) such as “how often do you and this child laugh together” with higher scores indicating more warmth. Parenting harshness was calculated as the mean of four items (α = 0.53) with higher scores indicating more harshness, e.g., “how often do you tell this child they are bad or not as good as others?” This scale has been used in diverse samples of parents, with higher internal consistency in harshness than the current sample (e.g., Garrido & Taussig, 2013; Riany & Ihsana, 2021; Tramonte et al., 2015).

Data Analysis

Cluster analysis was completed on 558 participants with observed social withdrawal and teacher-reported behavior data using Ward’s (1963) method with squared Euclidean distances. A four-cluster solution following that of Harrist and colleagues (1997) was compared to three- and two-cluster solutions using dendrograms and the Dunn (1974) index (Table 2), given the small size of some clusters (n = 18). The two- and four-cluster solutions were used in the analyses, given the better Dunn index of the four-cluster solution and the larger cluster sizes in the two-cluster solution (Table 2). A data-driven approach was used to determine whether to include child sex and/or family SES as covariates by testing their associations with child social withdrawal. Whether childhood social withdrawal was associated with romantic relationship status was tested using an ordered logistic regression model due to the ordinal nature of the adult relationship variable (Harrell & Harrell, 2015). Whether withdrawal was associated with becoming a parent was tested using binary logistic regression, and the associations between withdrawal and parenting warmth and harshness were tested using linear regression models. Parenting warmth and harshness were analyzed as separate outcomes because of their somewhat weak correlation (r = −0.21, p < 0.001). These same models were also tested using the two- and four-cluster categorical variables. All data analyses were completed in R (R Core Team, 2023) using the MASS package for ordered logistic regression (Venables & Ripley, 2002).

Table 2.

Comparison of two-, three-, and four-cluster solutions, N = 161

N (%)
Dunn index Cluster 1 Cluster 2 Cluster 3 Cluster 4
Two-cluster 0.18 95 (59.01%) 66 (40.99%) - -
Three-cluster 0.20 95 (59.01%) 48 (29.81%) 18 (11.18%) -
Four-cluster 0.26 95 (59.01%) 23 (14.29%) 18 (11.18%) 25 (15.53%)
Open in a new tab

Missing Data

Of the 558 participants with kindergarten behavior data, 132 (23.66%) participants were missing romantic relationship status and 131 (23.48%) were missing parent status. Of those who identified as parents with at least some involvement with their child(ren) (n = 282), 11 (3.90%) were missing parenting warmth or harshness data. Using Little’s (1988) test via the naniar package in R (Tierney & Cook, 2023), the data were determined to be missing at random (MAR; χ2 (29, 558) = 941.81, p < 0.001) as opposed to missing completely at random, because missingness was based on cluster assignment with participants in cluster two of the two-cluster solution more likely to be missing adult data (χ2 (2, 558) = 7.90, p < 0.05). Using the mice package (van Buuren & Groothuis-Oudshoorn, 2011), ten imputations were completed, and estimates were pooled across the ten imputed datasets for each model, because imputation produces unbiased estimates for MAR data (Jakobsen et al., 2017). Parenting warmth and harshness were imputed only for participants who identified as involved parents. Sensitivity analyses were also conducted using list-wise deletion and produced similar results. Analytic models with the two-cluster solution had slightly better model fit averaged across imputations (BIC = 979.80, 705.13, 830.40, 636.04) than models with the four-cluster solution (BIC = 980.95, 719.47, 838.03, 640.81).

Results

Whether or not a child was categorized as socially withdrawn did not vary by child sex (B = −0.08, p = 0.66) or family SES (B = 0.01, p = 0.14). Correlations between child measures and adult parenting behavior are displayed in Table 3.

Table 3.

Correlations between child social withdrawal measures and adult parenting behavior, Ns = 265 – 558

1 2 3 4 5 6 7 8 9 10
1. Observed social withdrawal -
2. Isolates self 0.11* -
3. Timid 0.04 0.58*** -
4. Anxious −0.01 0.28*** 0.55*** -
5. Immature 0.04 0.26*** 0.24*** 0.19*** -
6. Sad/depressed 0.04 0.43*** 0.35*** 0.18*** 0.27*** -
7. Lacks restraint 0.03 0.05 −0.07 −0.01 0.48*** 0.14** -
8. Angry 0.02 0.03 0.00 −0.01 0.37*** 0.14** 0.75*** -
9. Childhood SES −0.07 −0.11* −0.15* 0.01 −0.10* −0.15*** −0.21*** −0.17*** -
10. Parenting warmth −0.04 0.00 −0.07 0.03 0.03 −0.11 −0.01 0.00 0.23*** -
11. Parenting Harshness 0.00 −0.01 −0.08 −0.09 0.03 0.07 0.02 −0.02 −0.02 −0.21***
Open in a new tab

Note. *p < 0.05, **p < 0.01, *** p < 0.001

Subtypes of Social Withdrawal

Two-Cluster Solution

The two-cluster solution revealed one cluster of more distressed children characterized by higher scores across all clustering variables (isolates self, timid, anxious, immature, sad/depressed, lacks restraint, and angry) and a second cluster characterized as unsociable (Table 4). Most withdrawn children (n = 95) fell into this unsociable cluster, with a smaller proportion falling into the distressed cluster (n = 66). There were no differences in cluster membership based on family SES (B = 0.001, p = 0.95). But boys were more likely to be in the distressed group (48.8%) than were girls (32.0%; B = 0.71, p < 0.05). Child sex was included as a covariate in the two-cluster models.

Table 4.

Subtypes of social withdrawal in two-cluster solution, N = 558

Clusters ANOVAs
Clustering variable 1 2 Control Withdrawn clusters
F (1, 159)		 All groups
F (4, 553)
Isolates self −.28 .87 −.07 87.19*** 53.94***
Timid −.36 .85 −.05 62.64*** 33.82***
Anxious −.28 .68 −.04 33.57*** 21.01***
Immature −.39 .72 −.03 53.99*** 26.81***
Sad/depressed −.33 .91 −.07 54.12*** 38.33***
Lacks restraint −.27 .39 .00 34.10*** 16.95***
Angry −.33 .52 −.01 35.32*** 14.99***
n 95 66 397 161 558
Open in a new tab

Note. 1 = unsociable, 2 = distressed. *** p < 0.001

Four-Cluster Solution

The four-cluster solution replicated the characteristics of the four clusters identified by Harrist and colleagues (1997; Table 5). Namely, an unsociable cluster characterized by lower scores on all clustering variables (n = 95), a passive/anxious cluster characterized by higher timidity, anxiousness, and self-isolation (n = 23), an active-isolate cluster characterized by higher immaturity and anger and lower restraint (n = 18), and a sad/depressed cluster characterized by higher sadness and self-isolation (n = 25). In the two-cluster solution, the distressed cluster is a combination of the passive/anxious, active-isolate, and sad/depressed clusters. There were no differences in cluster membership based on family SES (B = −0.01, p = 0.33) or child sex (B = 0.24, p = 0.43).

Table 5.

Subtypes of social withdrawal in four-cluster solution, N = 558

Clusters ANOVAs
Clustering variable 1 2 3 4 Control Withdrawn clusters
F (3, 157)		 All groups
F (4, 553)
Isolates self −.28 1.18 .21 1.06 −.07 39.25*** 32.91***
Timid −.36 1.57 −.27 .98 −.05 44.16*** 28.58***
Anxious −.28 1.36 −.02 .57 −.04 19.18*** 16.31***
Immature −.39 −.03 1.71 .71 −.03 37.18*** 23.1***
Sad/depressed −.33 −.33 −.33 2.95 −.07 1245*** 100.7***
Lacks restraint −.27 −.34 1.55 .22 .00 68.31*** 31.3***
Angry −.33 −.35 2.11 .19 −.01 78.28*** 28.07***
n 95 23 18 25 397 161 558
Open in a new tab

Note. 1 = unsociable, 2 = passive/anxious, 3 = active-isolate, 4 = sad/depressed. *** p < 0.001

Adult Outcomes

Withdrawn versus Non-Withdrawn Groups

In terms of mid-30s romantic relationships, socially withdrawn children were no less likely than non-withdrawn children to be in a current relationship (B = −0.12, p = 0.53). There were no differences in whether individuals had become parents by their mid-30s between the withdrawn and non-withdrawn groups (B = −0.12, p = 0.60). Among those who were parents, there were no differences in parenting warmth (B = 0.01, p = 0.97) or harshness (B = −0.21, p = 0.44) between the withdrawn and non-withdrawn groups.

Differences Based on the Two-Cluster Solution

As shown in Table 6, the distressed group was marginally less likely to be in an adult romantic relationship compared to the unsociable group (B = 0.68, p = 0.07) and the non-withdrawn group (B = −0.50, p = 0.09). There were no differences in the likelihood of becoming a parent between the distressed and unsociable group (B = 0.18, p = 0.63) or non-withdrawn group (B = 0.22, p = 0.52). For the distressed group, neither parenting warmth (B = −0.04, p = 0.91; B = −0.05, p = 0.87) nor harshness (B = −0.17, p = 0.62; B = 0.09, p = 0.79) differed compared to the unsociable or non-withdrawn groups.

Table 6.

Descriptive data on adult outcomes for the two-cluster solution

N (%)
Adult Outcome Unsociable Distressed Control
Single 19 (23.17%) 17 (37.78%) 74 (24.75%)
Former relationship 6 (7.32%) 4 (8.89%) 28 (9.36%)
Current relationship 57 (69.51%) 24 (53.33%) 197 (65.89%)
Parent 57 (65.22%) 30 (69.51%) 212 (70.90%)
Mean (SD)
Parenting warmth 2.69 (0.59) 2.77 (0.62) 2.75 (0.64)
Parenting harshness 0.52 (0.40) 0.54 (0.36) 0.60 (0.43)
Open in a new tab

Note. Numbers displayed are raw, unimputed data. N = 426 for relationship status, 427 for parent status, 282 for parenting.

Differences Based on the Four-Cluster Solution

As shown in Table 7, the active-isolate group was less likely to be in a current relationship than both the unsociable group (B = −1.43, p < 0.05) and the non-withdrawn group (B = −1.24, p < 0.05). The groups did not differ in the likelihood of becoming a parent. For those who became parents, there were no differences in parenting warmth or harshness.

Table 7.

Descriptive data on adult outcomes for the four-cluster solution

N (%)
Adult Outcome Unsociable Passive/anxious Active-isolate Sad/depressed Control
Single 19 (23.17%) 4 (23.53%) 6 (54.54%) 7 (41.18%) 74 (24.75%)
Former relationship 6 (7.32%) 0 (0%) 1 (9.09%) 3 (17.65%) 28 (9.36%)
Current relationship 57 (69.51%) 13 (76.47%) 4 (36.36%) 7 (41.18%) 197 (65.89%)
Parent 57 (69.51%) 11 (64.71%) 8 (72.72%) 11 (61.11%) 212 (70.90%)
Mean (SD)
Parenting warmth 2.69 (0.59) 2.95 (0.75) 2.76 (0.57) 2.60 (0.48) 2.75 (0.64)
Parenting harshness 0.52 (0.40) 0.47 (0.34) 0.35 (0.14) 0.70 (0.40) 0.60 (0.43)
Open in a new tab

Note. Numbers displayed are raw, unimputed data. N = 426 for relationship status, 427 for parent status, 282 for parenting.

Discussion

This study extended research on subtypes of childhood withdrawal into adulthood. In this mostly white U.S. sample, social withdrawal in early childhood had little association with variations in several important outcomes in adulthood. Overall, the socially withdrawn group of children did not differ from the non-withdrawn children in their adult romantic relationship status, parent status, or parenting behavior. Despite prior research demonstrating withdrawn adolescents form their first romantic relationships later (Barzeva et al., 2021), they may catch up so that they are just as likely to be in a current relationship by their mid-30s, as in the present study. Indeed, social withdrawal demonstrates low to moderate stability in elementary school (r = 0.20 – 0.60; Booth-LaForce & Oxford, 2008; Rubin & Mills, 1988) and early adulthood (r = 0.44 – 0.62; Barzeva et al., 2021) such that childhood social withdrawal may not be highly predictive of adult outcomes. Nevertheless, the social experiences of withdrawn children (i.e., peer victimization, lower quality friendships) may remain relevant, given the longitudinal associations with adult romantic relationship status and quality (Allen et al., 2022; Takizawa et al., 2014; Wolke et al., 2013). These findings add to Rubin’s work on social withdrawal by demonstrating that withdrawal in childhood may not have long-term implications for common adult milestones like romantic relationships and becoming a parent.

Within the withdrawn group, subtype matters. The distressed sub-group of withdrawn children, and in particular the active-isolate subtype, was less likely to be in a current romantic relationship than the non-withdrawn individuals. Possibly, this difference in relationship experiences could be due to ongoing struggles with anger or impulsivity, and not due to withdrawal alone. This finding supports the premise by Rubin and colleagues (2021) that the motivations for withdrawal (e.g., shyness, unsociability) may not be measurable by the type of withdrawn behaviors displayed by children (e.g., solitary play, watching other children play). Perhaps these motivations are most relevant to developing relationships. Interestingly, the active-isolate subtype was no less likely to be a parent. Across all subtypes, parenting behaviors were similar. Therefore, the higher prevalence of insensitive parenting behaviors experienced by withdrawn children (e.g., Mills & Rubin, 1998) does not clearly carry over to the next generation. Given the importance of cultural values for childhood withdrawal, these findings may not extend to other cultural contexts. For instance, though we did not see any differences in adult outcomes for the unsociable group, this may not replicate in a society such as China, where unsociable children experience school and peer problems (Chen et al., 2015; Coplan et al., 2016). The limited differences in adult outcomes for socially withdrawn children reassure that withdrawn children in the U.S. will likely grow to meet common adult milestones.

Strengths and Limitations

The prospective, longitudinal design of this study is a major strength that afforded testing the association between childhood withdrawal subtype in kindergarten and mid-30s adult outcomes, rather than relying on adult measures of withdrawal or retrospective reports. The direct observations of child solitary play are another strength of this study, in addition to the teacher reports of child behavior that informed the subtyping. The small size of some withdrawal subtypes is a limitation of this study, given we had limited statistical power to test differences in adult outcomes with the four-cluster solution. As such, it is possible that there are differences in these outcomes that we were unable to discover with these small subgroups. To not obscure the developmental cascade from childhood withdrawal to adult outcomes, possible confounding variables assessed after child social withdrawal were not tested (e.g., adult mental health, SES, parenting stress). Although child sex and family of origin SES were tested as possible confounders, these analyses were limited in their ability to test other potential confounders. The reliability of the parent harshness scale was lower than previous studies (e.g., Garrido & Taussig, 2013; Riany & Ihsana, 2021; Tramonte et al., 2015), perhaps since harshness was assessed with only four items. Also, some participants were lost to attrition over the 30-year period. However, attrition was not so great as to limit our ability to calculate unbiased estimates using multiple imputation.

Future Directions

The current study adds to a foundation for future research to further examine the adult lives of those withdrawn in childhood. Future work should explore protective and promotive factors that influence the developmental cascade of childhood withdrawal to different adult outcomes (e.g., Henry et al., 2015; Masten & Cicchetti 2016; Masten et al., 2021). For example, improving peer relationships among more distressed subtypes in adolescence could make it more likely for individuals to get married, become parents, and engage in warm parenting behaviors. Family- and community-level factors could also support interpersonal relationship skills and facilitate these adult outcomes. In addition, research should explore factors that may be especially relevant to the active-isolate subtype. This subtype may benefit most from research on childhood interventions to reduce anger and impulsivity and promote healthy peer interactions. In addition, future research should examine other domains of adult outcomes in those who were withdrawn as children, such as wellbeing and health, peer relationships, work life, and psychopathology. Perhaps children who were withdrawn experience adulthood differently from their non-withdrawn peers, but these differences are not evident in forming romantic partnerships or parenting. Finally, the adult outcomes of socially withdrawn children should be studied in multiple societies.

Conclusion

This study demonstrates that in the United States, individuals who were socially withdrawn as children seem to fare as well as non-withdrawn children in adulthood in terms of romantic relationships and parenting. This is especially true for the unsociable subtype characterized by less interest in social interactions and limited distress. Although the subtype of withdrawal characterized by less restraint and more anger may struggle more with romantic relationships in their thirties, these differences do not extend to their self-reported parenting behaviors. This active-isolate subtype of withdrawal may benefit most from additional support to promote healthy peer and romantic relationships.

Acknowledgements:

This research would not have been possible without Jennifer Godwin and the participants in this study, we thank you. The Child Development Project has been funded by grants MH56961, MH57024, and MH57095 from the National Institute of Mental Health, HD30572 from the Eunice Kennedy Shriver National Institute of Child Health and Human Development, and DA016903 from the National Institute on Drug Abuse.

References

  1. Allen JP, Costello M, Kansky J, & Loeb EL (2022). When friendships surpass parental relationships as predictors of long-term outcomes: Adolescent relationship qualities and adult psychosocial functioning. Child Development, 93(3), 760–777. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Asendorpf JB (1990). Beyond social withdrawal: Shyness, unsociability, and peer avoidance. Human Development, 33, 250–259. 10.1159/000276522 [DOI] [Google Scholar]
  3. Barzeva SA, Richards JS, Meeus WH, & Oldehinkel AJ (2020). The social withdrawal and social anxiety feedback loop and the role of peer victimization and acceptance in the pathways. Development and Psychopathology, 32(4), 1402–1417. [DOI] [PubMed] [Google Scholar]
  4. Barzeva SA, Richards JS, Meeus WH, & Oldehinkel AJ (2021). Social withdrawal and romantic relationships: A longitudinal study in early adulthood. Journal of Youth and Adolescence, 50, 1766–1781. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Bernard K, Nissim G, Vaccaro S, Harris JL, & Lindhiem O (2018). Association between maternal depression and maternal sensitivity from birth to 12 months: A meta-analysis. Attachment & Human Development, 20(6), 578–599. [DOI] [PubMed] [Google Scholar]
  6. Booth-LaForce C, & Oxford ML (2008). Trajectories of social withdrawal from grades 1 to 6: Prediction from early parenting, attachment, and temperament. Developmental Psychology, 44(5), 1298–1313. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Bowker JC, & Raja R (2011). Social withdrawal subtypes during early adolescence in India. Journal of Abnormal Child Psychology, 39, 201–212. [DOI] [PubMed] [Google Scholar]
  8. Caspi A, Elder GH Jr, & Bem DJ (1988). Moving away from the world: Life-course patterns of shy children. Developmental Psychology, 24(6), 824–831. [Google Scholar]
  9. Cheah C, & Rubin K (2004). European American and Mainland Chinese mothers’ responses to aggression and social withdrawal in preschoolers. International Journal of Behavioral Development, 28(1), 83–94. [Google Scholar]
  10. Chen X. (2015). Culture, types of social withdrawal, and children's beliefs: An integrative perspective. British Journal of Developmental Psychology, 33(2), 174–176. [DOI] [PubMed] [Google Scholar]
  11. Chen X, Chen H, Li D, Wang L, & Wang Z (2015). Early childhood reticent and solitary-passive behaviors and adjustment outcomes in Chinese children. Journal of Abnormal Child Psychology, 43, 1467–1473. [DOI] [PubMed] [Google Scholar]
  12. Chen M, & Johnston C (2007). Maternal inattention and impulsivity and parenting behaviors. Journal of Clinical Child and Adolescent Psychology, 36(3), 455–468. [DOI] [PubMed] [Google Scholar]
  13. Closson LM, McVarnock AM, & Cook LE (2022). Is there an upside to social withdrawal? Considering well-being among socially withdrawn emerging adults. Applied Research in Quality of Life, 17(5), 3131–3149. [Google Scholar]
  14. Coplan RJ, Liu J, Ooi LL, Chen X, Li D, & Ding X (2016). A person-oriented analysis of social withdrawal in Chinese children. Social Development, 25(4), 794–811. [Google Scholar]
  15. Coplan RJ, Rose-Krasnor L, Weeks M, Kingsbury A, Kingsbury M, & Bullock A (2013). Alone is a crowd: Social motivations, social withdrawal, and socioemotional functioning in later childhood. Developmental Psychology, 49(5), 861–875. [DOI] [PubMed] [Google Scholar]
  16. Dir AL, Coskunpinar A, & Cyders MA (2014). A meta-analytic review of the relationship between adolescent risky sexual behavior and impulsivity across gender, age, and race. Clinical Psychology Review, 34(7), 551–562. [DOI] [PubMed] [Google Scholar]
  17. Dodge KA, Bates JE, & Pettit GS (1990). Mechanisms in the cycle of violence. Science, 250(4988), 1678–1683. [DOI] [PubMed] [Google Scholar]
  18. Dunn JC (1974). Well separated clusters and fuzzy partitions. Journal on Cybernetics, 4, 95–104. [Google Scholar]
  19. Garrido EF, & Taussig HN (2013). Do parenting practices and prosocial peers moderate the association between intimate partner violence exposure and teen dating violence? Psychology of Violence, 3(4), 354–366. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Gotlib IH, Lewinsohn PM, & Seeley JR (1998). Consequences of depression during adolescence: Marital status and marital functioning in early adulthood. Journal of Abnormal Psychology, 107(4), 686–690. [DOI] [PubMed] [Google Scholar]
  21. Harrell FE Jr, & Harrell FE (2015). Ordinal logistic regression. Regression modeling strategies: With applications to linear models, logistic and ordinal regression, and survival analysis, 311–325. [Google Scholar]
  22. Harrist AW, Zaia AF, Bates JE, Dodge KA, & Pettit GS (1997). Subtypes of social withdrawal in early childhood: Sociometric status and social-cognitive differences across four years. Child Development, 68(2), 278–294. [PubMed] [Google Scholar]
  23. Henry CS, Morris AS, & Harrist AW (2015). Family resilience: Moving into the third wave. Family Relations, 64(1), 22–43. 10.1111/fare.12106 [DOI] [Google Scholar]
  24. Hollingshead AB (1975). Four factor index of social status. Unpublished manuscript. Yale University. [Google Scholar]
  25. Jakobsen JC, Gluud C, Wetterslev J, & Winkel P (2017). When and how should multiple imputation be used for handling missing data in randomised clinical trials–a practical guide with flowcharts. BMC Medical Research Methodology, 17(1), 1–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Kersteter S. (2005). Parent Questionnaire, Kindergarten/Year 1 (Fast Track Project Technical Report) [On-line]. Available from the Fast Track Project web site: http://www.fasttrackproject.org/ [Google Scholar]
  27. Lansford JE, Malone PS, Dodge KA, Crozier JC, Pettit GS, & Bates JE (2006). A 12-year prospective study of patterns of social information processing problems and externalizing behaviors. Journal of abnormal child psychology, 34, 709–718. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Lin H, Harrist AW, Lansford JE, Pettit GS, Bates JE, & Dodge KA (2020). Adolescent social withdrawal, parental psychological control, and parental knowledge across seven years: A developmental cascade model. Journal of Adolescence, 81, 124–134. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Little RJ (1988). A test of missing completely at random for multivariate data with missing values. Journal of the American Statistical Association, 83(404), 1198–1202. [Google Scholar]
  30. Lovejoy MC, Graczyk PA, O'Hare E, & Neuman G (2000). Maternal depression and parenting behavior: A meta-analytic review. Clinical Psychology Review, 20(5), 561–592. [DOI] [PubMed] [Google Scholar]
  31. Masten AS, & Cicchetti D (2016). Resilience in development: Progress and transformation. In Cicchetti D (Ed.). Developmental psychopathology, risk, resilience, and intervention. John Wiley & Sons, Incorporated. [Google Scholar]
  32. Masten AS, Lucke CM, Nelson KM, & Stallworthy IC (2021). Resilience in development and psychopathology: Multisystem perspectives. Annual RSClinical Psychology, 17(1), 521–549. [DOI] [PubMed] [Google Scholar]
  33. Mills RS, & Rubin KH (1998). Are behavioural and psychological control both differentially associated with childhood aggression and social withdrawal? Canadian Journal of Behavioural Science/Revue Canadienne des Sciences du Comportement, 30(2), 132–136. [Google Scholar]
  34. Nelson LJ (2013). Going it alone: Comparing subtypes of withdrawal on indices of adjustment and maladjustment in emerging adulthood. Social Development, 22(3), 522–538. [Google Scholar]
  35. Nelson LJ, Jorgensen NA, & Clifford BN (2021). Shy and still struggling: Examining the relations between subtypes of social withdrawal and well-being in the 30s. Social Development, 30(2), 575–591. [Google Scholar]
  36. Nelson LJ, Rubin KH, & Fox NA (2005). Social withdrawal, observed peer acceptance, and the development of self-perceptions in children ages 4 to 7 years. Early Childhood Research Quarterly, 20(2), 185–200. [Google Scholar]
  37. R Core Team (2023). _R: A Language and Environment for Statistical Computing_. R Foundation for Statistical Computing, Vienna, Austria. <https://www.R-project.org/>. [Google Scholar]
  38. Riany YE, & Ihsana A (2021). Parenting stress, social support, self-compassion, and parenting practices among mothers of children with ASD and ADHD. Psikohumaniora: Jurnal Penelitian Psikologi, 6(1), 47–60. [Google Scholar]
  39. Rodriguez CM, & Green AJ (1997). Parenting stress and anger expression as predictors of child abuse potential. Child Abuse & Neglect, 21(4), 367–377. [DOI] [PubMed] [Google Scholar]
  40. Root AE, Hastings PD, & Rubin KH (2016). The parenting behaviors of shy–anxious mothers: The moderating role of vagal tone. Journal of Child and Family Studies, 25, 1325–1333. [Google Scholar]
  41. Rubin KH (1982). Nonsocial play in preschoolers: Necessarily evil? Child Development, 53(3), 651–657. [Google Scholar]
  42. Rubin KH, & Chronis-Tuscano A (2021). Perspectives on social withdrawal in childhood: Past, present, and prospects. Child Development Perspectives, 15(3), 160–167. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Rubin KH, Coplan RJ, & Bowker JC (2009). Social withdrawal in childhood. Annual Review of Psychology, 60, 141–171. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Rubin KH, & Mills RS (1988). The many faces of social isolation in childhood. Journal of Consulting and Clinical Psychology, 56(6), 916. [DOI] [PubMed] [Google Scholar]
  45. Schacter HL, Lessard LM, & Juvonen J (2019). Peer rejection as a precursor of romantic dysfunction in adolescence: Can friendships protect? Journal of Adolescence, 77, 70–80. [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Schmidt LA, Tang A, Day KL, Lahat A, Boyle MH, Saigal S, & Van Lieshout RJ (2017). Personality development within a generational context: Life course outcomes of shy children. Child Psychiatry & Human Development, 48, 632–641. [DOI] [PubMed] [Google Scholar]
  47. Schneider BH, Richard JF, Younger AJ, & Freeman P (2000). A longitudinal exploration of the continuity of children's social participation and social withdrawal across socioeconomic status levels and social settings. European Journal of Social Psychology, 30(4), 497–519. [Google Scholar]
  48. Schneider BH, Younger AJ, Smith T, & Freeman P (1998). A longitudinal exploration of the cross-contextual stability of social withdrawal in early adolescence. The Journal of Early Adolescence, 18(4), 374–396. [Google Scholar]
  49. Serbin LA, Temcheff CE, Cooperman JM, Stack DM, Ledingham J, & Schwartzman AE (2011). Predicting family poverty and other disadvantaged conditions for child rearing from childhood aggression and social withdrawal: A 30-year longitudinal study. International Journal of Behavioral Development, 35(2), 97–106. [Google Scholar]
  50. Shaffer A, Burt KB, Obradović J, Herbers JE, & Masten AS (2009). Intergenerational continuity in parenting quality: The mediating role of social competence. Developmental Psychology, 45(5), 1227–1240. [DOI] [PubMed] [Google Scholar]
  51. Shin Y. (2007). Social withdrawal and friendships in childhood. Korean Journal of Child Studies, 28(5), 193–207. [Google Scholar]
  52. Strayhorn JM, & Weidman CS (1988). A parent practices scale and its relation to parent and child mental health. Journal of the American Academy of Child & Adolescent Psychiatry, 27(5), 613–618. [DOI] [PubMed] [Google Scholar]
  53. Takizawa R, Maughan B, & Arseneault L (2014). Adult health outcomes of childhood bullying victimization: Evidence from a five-decade longitudinal British birth cohort. The American Journal of Psychiatry, 171, 777–784. 10.1176/appi.ajp.2014.13101401 [DOI] [PubMed] [Google Scholar]
  54. Tierney N, & Cook D (2023). Expanding tidy data principles to facilitate missing data exploration, visualization and assessment of imputations. Journal of Statistical Software, 105(7), 1–31. doi: 10.18637/jss.v105.i07.36798141 [DOI] [Google Scholar]
  55. Tramonte L, Gauthier AH, & Willms JD (2015). Engagement and guidance: The effects of maternal parenting practices on children’s development. Journal of Family Issues, 36(3), 396–420. [Google Scholar]
  56. U.S. Census Bureau. (2023). Estimated median age at first marriage, by sex: 1890 to present. U.S. Department of Commerce. Retrieved February 10, 2024, from https://www.census.gov/data/tables/time-series/demo/families/marital.html [Google Scholar]
  57. van Buuren S, & Groothuis-Oudshoorn K (2011). mice: Multivariate Imputation by Chained Equations in R. Journal of Statistical Software, 45(3), 1–67. doi: 10.18637/jss.v045.i03 [DOI] [Google Scholar]
  58. Venables WN, & Ripley BD (2002). Modern applied statistics with S, Fourth edition. Springer. https://www.stats.ox.ac.uk/pub/MASS4/. [Google Scholar]
  59. Ward JH (1963). Hierarchical grouping to optimize an objective function. Journal of the American Statistical Association, 58, 236–244. [Google Scholar]
  60. Wolke D, Lereya ST, Fisher HL, Lewis G, & Zammit S (2014). Bullying in elementary school and psychotic experiences at 18 years: A longitudinal, population-based cohort study. Psychological Medicine, 44, 2199–2211. [DOI] [PubMed] [Google Scholar]

RESOURCES