Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2022 Apr 1.
Published in final edited form as: Dev Psychol. 2021 Apr;57(4):548–556. doi: 10.1037/dev0001165

Early Child Care Experiences and Attachment Representations at Age 18 Years: Evidence from the NICHD Study of Early Child Care and Youth Development

Theodore E A Waters 1, Sophia W Magro 2, Jude Alhajeri 3, Rui Yang 4, Ashley Groh 5, John D Haltigan 6, Ashley A Holland 7, Ryan D Steele 8, Kelly K Bost 9, Margaret T Owen 10, Brian E Vaughn 11, Cathryn Booth-LaForce 12, Glenn I Roisman 13
PMCID: PMC8478126  NIHMSID: NIHMS1716784  PMID: 34594056

Abstract

The role of early child care experiences on the development of the mother-child attachment relationship has been studied extensively. However, no prospective studies of early child care have addressed how these experiences might be reflected in the content of attachment representations during adolescence and beyond. The goal of this study was to estimate relatively precise associations between child care quality, child care quantity, and type of care in the first 54 months of life and the content of adolescents’ attachment representations around age 18 years (N = 857; 51% female; 78% White, non-Hispanic; M income-to-needs ratio = 4.13), leveraging data from the longitudinal NICHD Study of Early Child Care and Youth Development (SECCYD). We identified a small positive association between the observed quality of early child care (but not quantity or type of care) and secure attachment states of mind as measured by the Adult Attachment Interview (but not the Attachment Script Assessment) at age 18 years that was robust to demographic covariates and observations of maternal and paternal sensitivity during childhood. We observed no significant interactions among child care variables. Associations between early child care experiences and indicators of adolescent attachment were likewise not moderated by maternal sensitivity from infancy to mid-adolescence or by maternal reports of child temperament in early childhood.

Keywords: attachment, child care, caregiving, secure base script, Adult Attachment Interview, dual risk, differential susceptibility

Questions pertaining to the potential benefits and/or risks of early child care experiences for children’s development have been of acute interest to parents, policymakers, and developmental scientists alike for decades. In particular, an enduring, and at times contentious debate regarding the potential impact of non-parental child care on aspects of child development has fueled extensive research and controversy since the late 1970s in the United States as more extensive use of non-parental child care became a reality for a large proportion of families (e.g., National Center for Education Statistics, 2006, Table 3). Despite such scholarly efforts, research on the fundamental question about whether non-parental child care is reliably associated with children’s development has arguably failed to yield clear answers (Dearing & Zachrisson, 2017; Shpancer, 2006). Moreover, even in instances where relatively robust short-term links between child care experiences and aspects of child development have been documented, questions naturally arise regarding the enduring versus transient nature of these associations across development (Belsky et al., 2007).

Table 3.

Results of stepwise regression models estimating relations between child care, parental sensitivity, and adolescent attachment

AAI Coherence β b 95% CI R 2
1. Hours per week in child care .06 0.01 [−0.02, 0.01] .03
  % of time in center care .01 0.04 [−0.38, 0.45]
  ORCE quality composite .17 0.59 [0.33, 0.84]
2. Hours per week in child care .06 0.01 [0.00, 0.01] .12
  % of time in center care −.02 −0.12 [−0.51, 0.28]
  ORCE quality composite .10 0.35 [0.10, 0.60]
  Maternal sensitivity .28 0.59 [0.44, 0.74]
  Paternal sensitivity .06 0.11 [−0.03, 0.24]
3. Hours per week in child care .05 0.01 [−0.02, 0.01] .15
  % of time in center care −.03 −0.15 [−0.54, 0.24]
  ORCE quality composite .08 0.28 [0.02, 0.53]
  Maternal sensitivity .21 0.44 [0.27, 0.62]
  Paternal sensitivity .05 0.08 [−0.05, 0.22]
  Gender .16 0.46 [0.28, 0.64]
  White/non-Hispanic .03 0.09 [−0.15, 0.32]
  Maternal education .09 0.05 [0.01, 0.10]
  Income-to-needs ratio .03 0.01 [−0.02, 0.051
ASA Secure Base Script Knowledge β b 95% CI R 2
1. Hours per week in child care .04 0.00 [0.00, 0.01] .02
  % of time in center care −.06 −0.23 [−0.56, 0.11]
  ORCE quality composite .13 0.35 [0.13, 0.56]
2. Hours per week in child care .05 0.00 [0.00, 0.01] .11
  % of time in center care −.08 −0.33 [−0.65, −0.01]
  ORCE quality composite .06 0.14 [−0.07, 0.35]
  Maternal sensitivity .19 0.29 [0.17, 0.42]
  Paternal sensitivity .19 0.25 [0.14, 0.36]
3. Hours per week in child care .03 0.00 [0.00, 0.01] .15
  % of time in center care −.08 −0.32 [−0.64, 0.00]
  ORCE quality composite .04 0.10 [−0.11, 0.31]
  Maternal sensitivity .13 0.20 [0.05, 0.35]
  Paternal sensitivity .18 0.24 [0.13, 0.35]
  Gender (Female) .17 0.37 [0.22, 0.51]
  Ethnicity (White/non-Hispanic) .03 0.08 [−0.12, 0.27]
  Maternal education .06 0.03 [−0.01, 0.07]
  Income-to-needs ratio .01 0.01 [−0.03, 0.041
Open in a new tab

Note. N = 857 for AAI Coherence estimates: N = 673 ASA Secure Base Script Knowledge estimates. Results were estimated with full information maximum likelihood. Bolded text indicates that the result was statistically significant.

A focal area of investigation reflected in the developmental literature on early child care has been the possibility that frequent and/or low-quality non-parental child care impedes the formation or maintenance of secure parent-child attachment relationships. Bowlby (e.g., 1969/1982) and Ainsworth (e.g., Ainsworth, 1967; Ainsworth et al., 1978/2015) proposed that attachment is a relational construct emergent from the interaction of an instinctive suite of behaviors present in infancy with a complementary set of parental behaviors, attitudes, values, and goals. They argued that attachment promoted the formation of a strong psychological connection with a primary caregiver that extends across space and time and is critical to safety, development, and adaptation. The quality of the attachment relationship is thought to arise from the sensitivity and responsiveness of the primary caregiver to the child’s signals (see De Wolff & Van IJzendoorn, 1997 for a meta-analysis). Bowlby (1969/1982) argued that in cases where consistent, sensitive, and responsive care was provided, infants would most likely co-construct a secure attachment, characterized by consistent and effective use of the primary caregiver as a secure base from which to explore and a haven of safety during distress. In contrast, insecure attachment would more likely result when caregiving was inconsistent, interfering/controlling, or rejecting of infant needs. Given the importance of sensitive caregiving for the establishment of secure parent-child attachments, some developmental scholars proposed that time away from primary caregivers in child care might pose a risk for insecure attachment (e.g., Belsky, 1986).

To date, several studies have examined the link between child care experiences and the quality of mother-infant attachment. This research has targeted three relatively distinct aspects of early child care when studying the significance of these experiences for attachment security: (1) quantity (hours per week spent in non-parental child care), (2) quality (e.g., sensitivity of caregivers in the child care setting), and (3) type (home-based versus center-based care). Several early studies reported that time spent in non-familial child care settings (quantity) increased the likelihood of developing an insecure attachment to mothers during infancy (e.g. Barglow et al., 1987; Belsky & Rovine, 1988). However, in subsequent larger sample studies, these associations have not generally replicated, including in the NICHD Study of Early Child Care and Youth Development (SECCYD)—the largest sample longitudinal investigation in this area to date (NICHD Early Child Care Research Network [ECCRN], 1997; 2001; see also Caldera & Hart, 2004).

In terms of the quality of child care, there is little evidence that the sensitivity of caregivers in the child care setting is directly linked to the quality of mother-child attachment in infancy. For example, the NICHD ECCRN (1997; 2001) found that child care provider sensitivity was not significantly linked to infant or toddler attachment to mothers. At the same time, evidence did emerge for an interactive effect of quality of care in the home and in child care. Specifically, at age 15 months (but not 36 months) infants were more likely to exhibit insecure attachment with their mothers when relatively insensitive maternal care was coupled with relatively insensitive child care (dual risk). Type of care has also been linked to increased risk for insecure attachment in some studies. Sagi and colleagues (2003) examined links between child care and infant attachment security in a large sample study and found that type of care—specifically lower-quality, center-based care—was associated with higher risk for insecure attachment to mothers. In contrast, the NICHD ECCRN (1997; 2001) did not find that type of care was related to infant attachment security. In sum, the literature on the links between child care and attachment security during infancy and early childhood to primary caregivers has been extensive but produced inconsistent evidence of main effects and some indication of more complex interactive effects between child care and maternal caregiving quality.

Despite the numerous studies that aimed to examine the significance of early child care experiences for the development of mother-child attachment in infancy and early childhood, none has examined the longer-term significance of child care characteristics for attachment representations beyond early or even middle childhood. This presents a significant gap in our understanding of the significance of child care experience for attachment, for two reasons.

First, although attachment quality in infancy is specific to a particular caregiver, by late adolescence, attachment-relevant experiences across significant caregivers are thought to be internalized in the form of attachment representations that reflect individuals’ understanding of attachment relationships and interpersonal expectations in times of need (e.g. Bowlby, 1969/1982; 1980). Mental representations theoretically have far broader inputs than the dynamics of a single caregiving relationship early in life and emphasize, at the very least, maternal and paternal caregiving expectations (e.g., E. Waters & Cummings, 2000). For example, attachment representations come to reflect our caregiving experiences across childhood with both mother and father and, potentially, with all those with whom we form secure base relationships (e.g., Howes & Spieker, 2008; Vaughn et al., 2016; T. Waters et al., 2015). Thus, despite inconsistent evidence of links between child care experiences and mother-infant attachment, attachment representations in late adolescence and young adulthood that are reflective of attachment history more broadly may include experiences in the child care setting. Moreover, by exclusively studying associations between child care and the mother-child attachment bond in infancy/early childhood, we overlook an important facet of attachment security and a potential pathway by which early child care can impact security (i.e., through other significant attachment relationships, such as those with fathers).

Second, a focus on infancy and early childhood may foreclose on questions related to the enduring nature and developmental significance of associations between child care experiences and attachment. Numerous studies have found enduring associations between early child care experiences and academic and socioemotional outcomes into adolescence (see Belsky & Pluess, 2012; Belsky et al., 2007; Burchinal et al., 2013; Vandell et al., 2010; Vandell et al., 2017). These results support the notion that early child care experiences have enduring associations with critical developmental outcomes, and highlight a noteworthy gap in the literature on child care and attachment.

Taken together, the more generalized nature of attachment representations and the fact that they can be studied outside the childhood period makes them an excellent candidate to study the prospective associations between early experiences in the context of the child care setting and attachment. As such, we present here a significant departure from prior work on the associations between child care experiences and attachment security—which has traditionally focused on the quality of the early attachment relationship between mother and child—to mental representations of attachment in adolescence that emphasize attachment with both mother and father along with more general expectations around support and care in the context of attachment relationships. To address these goals, we examined data from the age 18-year follow-up assessment of the SECCYD cohort (N = 857) to examine whether the quality, quantity, and type of early child care experiences were reflected in attachment outcomes in late adolescence. Specifically, we examined the predictive significance of non-maternal child care quantity (i.e., hours per week), quality (i.e., observation-based assessments of child care provider sensitivity), and type of child care (center-based or home-based) over the first 54 months of life for security of attachment representations at age 18 years.

The Present Study

Late adolescent attachment representations were assessed in two ways. First, participants completed the Adult Attachment Interview (AAI; Hesse, 2008; Main et al., 1985) either in person or by phone. Second, in cases where AAI assessments were conducted in person, participants were also administered the adolescent version of the Attachment Script Assessment (ASA; Dykas et al., 2006; see T. Waters & Roisman, 2019 and H. Waters & T. Waters, in press, for recent reviews). We have demonstrated in prior reports that the AAI and ASA are robustly predicted by the observed quality of both maternal and paternal caregiving experienced during infancy and childhood in the SECCYD cohort (see Haydon et al., 2014; Steele et al., 2014). These findings lend crucial support to Bowlby’s (e.g., 1969/1982) argument that attachment security transitions from being relationship-specific in infancy to becoming an attribute of the individual later in life through the formation of a mental representation of attachment. As our assessments of attachment security focus on late-adolescents’ mental representations of attachment, rather than relationship-specific security, we also evaluated the robustness of observed associations between child care variables and attachment security in late adolescence to the inclusion of key attachment-related predictors (i.e., maternal and paternal sensitivity) as well as demographic covariates (i.e., biological sex, ethnicity, family income-to-needs ratio, and maternal education).

Next, we explored a series of questions derived from prior published work from the child care and attachment literature. First, we explored the potential interaction among the early child care variables. For example, it is plausible that associations between low quality child care and attachment representations could appear only when hours per week of child care are high. We then examined two distinct conditional hypotheses implied by the literature on early child care experiences. First, we examined whether we could replicate and extend evidence from the SECCYD (NICHD ECCRN, 2001) that the dual risk of insensitive maternal caregiving during early childhood combined with more hours per week in child care might put adolescents at particular risk for poorer attachment-related outcomes. Second, in light of reliable evidence reported by Belsky and his colleagues based on the SECCYD (e.g., Belsky & Pluess, 2012; Pluess & Belsky, 2009) that children described by their mothers as having more difficult temperaments are differentially susceptible to early child care experiences, we examined whether temperament moderated associations between early child care and adolescents’ attachment representations.

Given the inconsistency in the literature on the significance of early child care experiences for the quality of attachment in early childhood, making confident predictions a priori about the direction of main effects of early child care indicators or point predictions for effect sizes is difficult. Nonetheless, the large sample size and quality of measurement available in the SECCYD does lend confidence to the precision of the estimated effect sizes we report. Moreover, the variables used in this analysis were operationalized in a manner precisely consistent with our prior programmatic work on the antecedents of attachment representations in this sample (see Methods) and, in advance, we elected to interpret main effects associations between early child care variables and attachment representations only when robust to four standard demographic covariates and observed sensitivity by primary caregivers. It should also be noted that prior work examining the associations between child care experiences and attachment security relied on data through 36 months in the SECCYD. In contrast, for this study we utilized child care data through 54 months to better capture the full range of early child care experiences and their potential contributions to later attachment security.

Methods

Participants

This study was based on 857 adolescents (51% female; 78% White, non-Hispanic) who completed a follow-up assessment of the SECCYD study at approximately 18 years of age (see Booth-LaForce & Roisman, 2014). The SECCYD is a longitudinal study that began with 1,364 study children and their families at 10 sites across the United States who were then followed from birth through age 15 years. The primary purpose of the SECCYD was to predict children’s developmental outcomes (language, cognitive, social-emotional, physical health) from characteristics of relevant contexts (home, child care, school, neighborhood) and individual characteristics of the family and the child (see http://www.nichd.nih.gov/research/supported/Pages/seccyd.aspx for more details). The follow-up of the NICHD SECCYD and related analyses were approved by the University of Minnesota ethics review board (title: “Follow-up of the NICHD Study of Early Child Care and Youth Development”; IRB ID 1207S16927).

At the follow-up assessment of the active SECCYD cohort at 18 years of age, participants were interviewed in person or via phone about their childhood using the AAI. For a subset of the participants, the ASA (N= 673) was administered directly after the administration of the AAI (generally when the assessment was conducted in person; see Steele et al., 2014). Importantly, the demographic characteristics of the AAI sample and the ASA sample did not differ substantively on gender composition, ethnicity, maternal education, or income-to-needs ratio. There were several significant differences between the age 18 years AAI and ASA samples and the full sample recruited at the first wave, but these differences were trivial to small in magnitude based on Cohen’s (1992) criteria (see Booth-LaForce & Roisman, 2014; Steele et al., 2014 for further details). Below, we describe the specific measures used in the present analyses and the time points of administration. Additional details about the larger SECCYD data collection procedures and instruments at specific time points are publicly available at: https://www.icpsr.umich.edu/icpsrweb/ICPSR/series/00233.

Measures

Early child care

Child care hours per week.

Parents reported children’s hours of routine non-maternal care during phone and personal interviews conducted at 3-month intervals starting at birth through 36 months and subsequently at 4-month intervals until the target child reached 54 months of age. Consistent with prior work by the NICHD ECCRN (e.g., 1997; 2001), non-maternal care included father care, other relative care, in-home non-relative care, center-based care, and child care in a non-familial home. A mean composite variable created from all available child care hours per week data from birth to age 54 months was used for the analyses in the present study.

Child care quality.

Observational assessments were conducted in the primary non-maternal child care location at ages 6, 15, 24, 36, and 54 months. Quality of care was assessed during two visits scheduled within a 2-week interval at 6, 15, 24, and 36 months. A single half-day visit was conducted at 54 months. At 6, 15, 24, and 36 months, four 44-minute observation cycles were collected across the two visits and two 44-minute observation cycles were collected at the 54-month visit using the Observational Record of the Caregiving Environment (ORCE; NICHD Early Child Care Research Network, 1996; 2002). During the observation periods, observers focused on the study child’s behavior, activities, and interaction with the caregiver. At the end of the 44 minutes, the observer made qualitative ratings of the quality of care provided by up to three caregivers. Reliability was greater than .80 for all waves (Vandell et al., 2010). A mean composite of the quality of non-maternal care was computed for each child using all available ORCE data collected in the child care setting.

Proportion of time in center-based child care.

Similar to the hours per week data, mothers reported on all types of non-maternal caregiving arrangements made at 3-month intervals through 36 months and then at 4-month intervals until the target child reached 54 months old. Non-maternal care was classified as either center-based care, home-based care (based in a home other than the target child’s residence), in-home care, father care, or grandparent care. The proportion of epochs the child spent 10 or more hours per week receiving care in a center (as opposed to other forms of child care or maternal care) was computed using all available data on type of care received.

Parental sensitivity.

Observations of maternal sensitivity were collected at 6, 15, 24, 36, and 54 months of age; Grades 1, 3, and 5; and at 15 years of age. Observations of paternal sensitivity were acquired at 54 months; Grades 1, 3, and 5; and age 15 years. Sensitivity was tested as children and their parent (mother or father) were videotaped completing age-appropriate tasks. Identical to earlier studies using SECCYD data (e.g., Steele et al., 2014), sensitivity scores at all ages were based on 15-minute semi-structured observations using the rating items that paralleled the ORCE in infancy and early childhood, and were adjusted for age-appropriateness and task characteristics at each age (e.g., Owen, 1992; Owen et al., 2000). The ratings at age 15 years were made from a coding system based on those used at earlier ages and the Autonomy and Relatedness Coding System of Allen and colleagues (Allen et al., 2000; Owen et al., 2007). Data from all waves were first standardized and then composited to create an index of the observed maternal or paternal caregiving sensitivity composites across the study (for more information on the tasks, coding, and reliability, see Booth-LaForce & Roisman, 2014; NICHD ECCRN 2001; 2004; 2008). Internal consistencies of the maternal and paternal sensitivity composite measures averaged .79 (range .70-.85) and .78 (range .71-.82), respectively (Booth-LaForce et al., 2014). Maternal (α = .83) and paternal (α = .70) sensitivity ratings were z-standardized and then averaged over time to create total summary measures of antecedent maternal and paternal sensitivity (Roisman, Haltigan et al., 2014).

Adolescent attachment measures

Adult Attachment Interview.

The AAI is a semi-structured interview focused on the participant’s relationships with their primary caregivers before the age of 13 years. There are multiple scales applied to AAI transcripts, but in the present study we focused on the coherence of mind score (ranging from 1-9) which serves as the primary indicator of attachment security in the Main, Goldwyn, and Hesse AAI coding system (2003-2008). Coherence of mind scores reflect the extent to which the AAI narrative produced by the participant is detailed, plausible, internally consistent, and not emotionally overwrought (see Crowell, in press; Hesse, 2016, for details regarding the AAI coding system). Highly coherent AAI transcripts are thought to be a reflection of a well-organized and secure attachment representation, whereas lower scores indicate insecurity. For further details on the AAI administration, coding, and reliability procedures in the SECCYD, see Booth-LaForce and Roisman (2014). Coder reliability for the coherence of mind scale in the full AAI sample produced an ICC of .85.

Attachment Script Assessment.

Grounded in Ainsworth’s notion of the “secure base phenomenon” (e.g., Ainsworth, et al., 1978/2015; E. Waters & Cummings, 2000), H. Waters and Rodrigues-Doolabh (2001), designed the ASA to assess individual differences in knowledge and use of the secure base script to organize narratives about ordinary events that involve a parent/child (or adult/adult) dyad faced with a social dilemma. The adolescent version of the ASA is a narrative-based measure of attachment containing a set of four fictional narrative outlines (see Dykas, et al., 2006).

Two stories focus on mother-child relationships and two focus on father-child relationships. The mother-child narrative prompts are titled “The Party” and either “Acne” (completed by girls) or “The Haircut” (completed by boys). The father narrative prompts contain stories entitled “The Tennis Match” and “Studying for an Exam”, which were given to boys and girls alike. Embedded in each story outline is a reference to some sort of difficulty or distress that triggers the adolescent to seek out the attachment figure, who is then expected to resolve or respond to the distress in a sensitive matter (H. Waters & T. Waters, in press). Participants were encouraged to tell the best possible story using the outline implied by the prompt list but were also instructed that they need not use all the words in the prompt, they could change the order, and they could change the words themselves if they chose to do so. To ensure participants understood the task, they first completed a practice story, which was not coded, called “A Trip to the Beach.” Each story is coded from 1-7, with “7” indicating that the story follows a clear secure base script structure and contains multiple elaborations of positive secure base content (e.g., comforting behavior) and “1” indicating that the story does not follow a secure base script structure or directly contradicts the secure base script (e.g. caregiver rejects the child’s bid for support; see Steele et al., 2014 for more details on the administration, coding, and reliability procedures). Coder reliability within and across coding sites produced ICCs between .93-.95 and were considered high. ASA scores for each story were averaged to produce a single score (α = .78; Steele et al., 2014).

Difficult temperament.

In order to evaluate the differential susceptibility hypothesis in the current study, we used the same infant difficult temperament variable utilized in previous reports on the SECCYD dataset (e.g., Pluess & Belsky, 2009; Roisman et al., 2012): At the 6-month visit, infant difficult temperament was assessed with an adapted version of the Infant Temperament Questionnaire (Carey & McDevitt, 1978). Mothers were asked to rate their baby’s behavior from 1-6 on 38 items related to various aspects of temperament (e.g. activity level; adaptability; mood) and an average rating across all the items was computed (α = .81; Roisman et al., 2012). Higher scores reflect more difficult temperament.

Demographic Covariates.

Covariates included child biological sex (1 = male, 2 = female), child race/ethnicity (1 = White/non-Hispanic, 0 = other), maternal years of education, and socioeconomic status. Socioeconomic status was operationalized as an income-to-needs ratio (total family income divided by the year-specific poverty threshold for the appropriate family size) calculated separately at 1, 6, 15, 24, 36, and 54 months; Grades 1, 3, 4, 5 and 6; and age 15 years. Income-to-needs scores were then averaged across waves (α based on standardized items = .97).

Results

Main Effects of Child Care on Attachment Representations

To address missing data, all regression analyses were conducted with full information maximum likelihood (FIML) in R with the package lavaan (Rosseel, 2012). Results using listwise deletion methods were conducted as a robustness check for all regression analyses utilizing FIML and are presented in the online supplement to this manuscript (see Table Sl-S5). Before conducting regression analyses, correlations among early child care variables and attachment representations in late adolescence were examined. Significant associations between child care quality (ORCE composite) and AAI coherence (r = .16, p < .05) and between child care quality and ASA secure base script knowledge (r = .13, p < .05) were observed. However, neither hours per week in child care nor percentage of time in center-based care was associated with either attachment measure (Table 1; Table 2).

Table 1.

Correlations, means, and standard deviations of study variables for AAI analysis sample

1 2 3 4 5 6 7 8 9 10 11
1. AAI Coherence —
2. Hours per week in child care .02 —
3. % of time in center care −.01 .39 —
4. ORCE quality composite .16 −.24 −.18 —
5. Maternal sensitivity .32 −.03 .05 .19 —
6. Paternal sensitivity .18 −.12 −.04 .17 .40 —
7. Temperament −.08 −.15 −.05 −.01 −.16 −.04 —
8. Gender (Female) .19 .02 −.02 .05 .08 .07 .04 —
9. Ethnicity (White/non-Hispanic) .16 −.09 .00 .14 .40 .13 −.12 .04 —
10. Maternal education .25 .09 .14 .16 .51 .24 −.11 .05 .25 —
11. Income-to-needs ratio .20 .11 .17 .21 .42 .21 −.09 .03 .23 .55 —
N 857 857 835 777 857 745 842 857 857 857 856
M 4.99 24.07 0.21 2.94 0.00 0.00 3.16 — — 14.56 4.13
SD 1.44 14.10 0.26 0.40 1.00 1.00 0.40 — — 2.44 3.13
Open in a new tab

Note. All correlations with absolute values greater than or equal to .08 are statistically significant, p < .05.

Table 2.

Correlations, means, and standard deviations of study variables for ASA analysis sample

1 2 3 4 5 6 7 8 9 10 11
1. ASA Secure Base Script Knowledge —
2. Hours per week in child care −.01 —
3. % of time in center care −.06 .36 —
4. ORCE quality composite .13 −.21 −.17 —
5. Maternal sensitivity .27 −.01 .05 .22 —
6. Paternal sensitivity .28 −.10 −.02 .18 .40 —
7. Temperament −.05 −.18 −.05 .01 −.17 −.02 —
8. Gender (Female) .21 .01 −.05 .07 .10 .09 .03 —
9. Ethnicity (White/non-Hispanic) .14 −.05 .01 .13 .41 .12 −.13 .03 —
10. Maternal education .20 .11 .14 .18 .50 .23 −.14 .07 .29 —
11. Income-to-needs ratio .17 .14 .18 .23 .45 .24 −.12 .06 .27 .58 —
N 673 673 661 619 673 581 662 673 673 673 672
M 3.71 24.34 0.20 2.94 −0.02 −0.02 3.16 — — 14.53 4.10
SD 1.05 13.84 0.26 0.39 0.99 1.03 0.39 — — 2.41 2.95
Open in a new tab

Note. All correlations with absolute values greater than or equal to .09 are statistically significant, p < .05.

We next sought to determine whether unique associations with early child care quality could be observed, net of other theoretically relevant correlates of attachment representations in late adolescence (i.e., maternal and paternal sensitivity, demographic factors). Stepwise regression results with FIML are presented in Table 3. For AAI coherence, we observed a unique association with child care quality (β = .17) above and beyond nonsignificant associations with hours per week in child care and percentage of time in center-based care. The association with child care quality was robust after including both maternal and paternal sensitivity (β = .10) and demographics (β = .08). However, this effect was small in magnitude. It should also be noted that the observed association between maternal sensitivity and AAI coherence β = .21 in the final step) remained significant.

Regarding ASA secure base script knowledge, we observed a unique association with child care quality (β = .13) above and beyond hours per week in child care and percentage of time in center-based care. In Step 2, when maternal and paternal sensitivity were included in the model, percentage of time in center-based care appeared significant (β = −.08) and child care quality was no longer significant. However, neither child care quality nor percentage of time in center-based care remained significant after including demographics (Step 3) in the model. It should be noted that the observed associations for maternal (β = .13 in the final step) and paternal (β = .18 in the final step) sensitivity remained significant across all steps of the ASA regression analyses.

As a robustness check, the analyses presented in Table 3 were repeated with the inclusions of the proportion of times secure in early childhood (assessed at 15m, 24m, and 36m using the Strange Situation Procedure; see Groh et al., 2014 for more detail) using both FIML and listwise deletion (Table S6-S7). Although infant attachment security has not demonstrated a robust association to attachment at age 18 years in the SECCYD sample (see Groh et al., 2014; Steele et al., 2014) we thought it important to check for any potential impact of infant attachment security on the results. As expected, the results presented here did not substantively change when infant attachment security was included (i.e., proportion of times secure across three assessment points, see Groh et al., 2014 and Steele et al., 2014 for more detail).

Conditional Associations between Child Care and Attachment Representations

Building on our initial models, we expanded our analyses to examine whether interactions between early child care variables could account for individual differences in AAI coherence or ASA secure base script knowledge. Accordingly, interactions were added in Steps 2 and 3 of the stepwise regression models.

Child care × child care

We first examined the interactive effects of the child care parameters on AAI coherence. No significant interaction emerged between child care quality and the number of hours spent in child care (β = 0, b = 0, 95% CI: [−0.02, 0.02]), between percentage of time spent in center-based care and number of hours spent in child care (β = −.03, b = −0.01, 95% CI: [−0.05, 0.02]), or between percentage of time spent in center-based care and observed child care quality (β = −.02, b = −0.29, 95% CI: [−1.47, 0.88]).

In examining ASA secure base script knowledge, there were again no significant interactions between various aspects of early child care. We did not find statistically significant interactions between child care quality and the number of hours spent in child care (β = 0, b = 0, 95% CI: [−0.01, 0.02]), between percentage of time spent in center-based care and number of hours spent in child care (β = 0, b = 0, 95% CI: [−0.03, 0.03]), or between percentage of time spent in center-based care and observed child care quality (β = .01, b = 0.14, 95% CI: [−0.85, 1.14]).

Child care × maternal sensitivity

We next sought to examine the dual risk hypothesis predicting that individuals would be at heightened risk for insecurity when they received less sensitive care from mothers in combination with less sensitive child care. Although a small, positive association between child care quality and AAI coherence remained after including interaction terms, there was no evidence that the association between child care quality and AAI coherence was larger for those with low maternal sensitivity. Furthermore, no conditional effects of hours per week in child care or percentage of time in center-based care emerged. Regarding secure base script knowledge, we observed a small negative main effect of percentage of time in center-based care (β = −.08, b = −0.34, 95% CI: [−0.66, −0.01]) after including the interaction terms and demographic variables (Step 3). No other main effects nor any conditional effects of child care variables emerged for ASA secure base script knowledge. Full results for analyses examining associations between child care and attachment representations conditioned on maternal sensitivity are presented in the online supplement (Table S2; S4).

Child care × difficult temperament

Lastly, we examined whether associations between early child care and adolescent attachment security were moderated by child temperament. For AAI coherence, net of all child care quality variables (Step 1), there was a small negative unique association of temperament with AAI coherence (β = −.08, b = −0.27, 95% CI: [−0.52, −0.03]). However, no evidence emerged to support the hypothesis that the links of any of the three early child care variables to AAI coherence or ASA secure base script knowledge were conditional upon children’s temperament early in life. Full results for analyses examining associations between child care and attachment representations in late adolescence conditioned on child temperament are presented in the online supplement (Table S3; S5).

Discussion

Although a significant number of studies have focused on links between child care characteristics and early mother-child attachment security, the present study is unique in investigating the relations between child care and adolescent attachment representations. The objective of this study was to address lingering questions related to the association between non-maternal child care experiences and the development of secure attachment by extending prior research into late adolescence using follow-up data from the SECCYD. To do so, we leveraged existing data on early child care quality, quantity, and type, as well as assessments of parental sensitivity to probe potential long-term associations between child care experiences in the first 54 months and attachment security at age 18 years, measured using two well-validated assessments of adolescent attachment representations (i.e., the AAI and ASA).

Results based on the AAI data indicated that higher-quality child care positively predicted security in late adolescence above and beyond demographic covariates and parental sensitivity from infancy through mid-adolescence. This robust association was small in magnitude. Moreover, when examining potential moderators of this association, we found no evidence that the predictive significance of early child care quality (or indeed child care quantity or type) for later attachment status was amplified or attenuated by the other child care variables, the quality of maternal caregiving, or early temperament.

In a subsample of the SECCYD, we were able to test for replication of the observed association between child care quality and AAI coherence using an alternative representational measure of attachment security, the ASA. ASA scores in late adolescence were significantly correlated with child care quality, but this association was not robust to the inclusion of parental sensitivity and demographic covariates in the regression analyses. This may be attributable, in part, to the smaller number of participants who completed the ASA compared with the AAI. The other child care variables were not significantly associated with secure base script knowledge, as assessed by the ASA. Further, none of the associations between child care variables and secure base script knowledge were conditional on the other child care variables, maternal sensitivity, or infant temperament. It should be noted that the ASA was significantly negatively associated with proportion of time in center based care in the regressions examining the dual risk hypothesis [Table S1]. However, we hesitate to draw any conclusion from this result as it was not statistically significant in any of our other analyses and thus should not be considered robust.

The current study demonstrates the possibility of long-term links between the observed sensitivity/responsiveness of child care providers and the content of late adolescents’ attachment states of mind—net of experiences with primary maternal and paternal caregivers. This is consistent with work that suggests children can form secondary attachment relationships with their non-familial child care providers, which may be beneficial (see Howes & Spieker, 2008 for further discussion). However, our results also suggest that these relations are weak in magnitude over the long-term. In line with attachment theory, the more significant caregiving context for the development of secure attachment representations appears to be with primary caregivers in the individual’s family.

In terms of limitations, it is first important to note the normative risk nature of the SECCYD cohort, as well as its modest ethnic diversity, as potential constraining generalizability of the findings reported here. Future research is needed to determine whether the long term association of child care with mental representations of attachment differs in higher-risk contexts or within minority communities where child care quality may be, on average, lower and the amount of time children spend in child care may increase due to a higher prevalence of multiple employment and single-parent households which may need to rely more on non-familial child care. (That said, it should be noted that a wide range of child care provider sensitivity scores was observed in the SECCYD dataset). Second, child care observations were limited to a single child care setting, but in many cases multiple child care arrangements were being utilized by families. Unfortunately, we were unable to examine the potential contribution of these additional (unobserved) care arrangements (Morrissey, 2009; Tran & Weinraub, 2006). Third, causal inferences about the caregiving environments studied in this report are necessarily constrained by the non-experimental research design of the SECCYD. Fourth, we were unable to evaluate a more developmentally nuanced model examining whether the associations reported differed by developmental timing (e.g., child care experiences through 36 months versus through 54 months) as the associations of such child care composites were very high. Fifth and finally, data on the extent to which parents engaged in non-familial child care with their school-aged children before and after school hours, during summers, or over school holidays are available in the SECCYD dataset. However, data on the quality of that care were not collected for those assessments. As such, we elected not to include data on non-familial child care after age 54 months and rather maintained our focus on early child care experiences, which is most consistent with the extant literature. Future research may benefit from examining the role, if any, of later child care experiences in the development of attachment representations.

Supplementary Material

Supplemental Material

Acknowledgments

This work was supported by a grant from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (R01 HD069442) to Glenn I. Roisman.

Contributor Information

Theodore E. A Waters, New York University – Abu Dhabi.

Sophia W. Magro, University of Minnesota

Jude Alhajeri, New York University – Abu Dhabi.

Rui Yang, New York University – Abu Dhabi.

Ashley Groh, University of Missouri.

John D. Haltigan, University of Toronto

Ashley A. Holland, Edgewood College

Ryan D. Steele, University of Minnesota

Kelly K. Bost, University of Illinois – Urbana Champaign

Margaret T. Owen, University of Texas – Dallas

Brian E. Vaughn, Auburn University

Cathryn Booth-LaForce, University of Washington.

Glenn I. Roisman, University of Minnesota

References

  1. Ainsworth MDS (1967). Infancy in Uganda: Infant care and the growth of love. Johns Hopkins University Press. [Google Scholar]
  2. Ainsworth MDS, Blehar MC, Waters E, & Wall S (1978/2015). Patterns of attachment: A psychological study of the Strange Situation. Taylor & Francis (Psychology Press: Classic Editions). [Google Scholar]
  3. Allen JP, Hauser ST, Bell KL, McElhaney KB, Tate DC, Insabella GM, et al. (2000). The autonomy and relatedness coding system. University of Virginia, Unpublished manuscript. [Google Scholar]
  4. Barglow P, Vaughn BE, & Molitor N (1987). Effects of maternal absence due to employment on the quality of infant-mother attachment in a low-risk sample. Child Development, 58(4), 945–954. 10.2307/1130535 [DOI] [PubMed] [Google Scholar]
  5. Belsky J (1986). Infant day care: A cause for concern? Zero to Three, 7(1), 1–7. [Google Scholar]
  6. Belsky J, & Pluess M (2012). Differential susceptibility to long-term effects of quality of child care on externalizing behavior in adolescence? International Journal of Behavioral Development, 36(1), 2–10. 10.1177/0165025411406855 [DOI] [Google Scholar]
  7. Belsky J, & Rovine MJ (1988). Nonmaternal care in the first year of life and the security of infant-parent attachment. Child Development, 59(1), 157–167. 10.2307/1130397 [DOI] [PubMed] [Google Scholar]
  8. Belsky J, Vandell DL, Burchinal M, Clarke-Stewart KA, McCartney K, & Owen MT (2007). Are there long-term effects of early child care? Child Development, 78(2), 681–701. 10.1111/j.1467-8624.2007.01021.x [DOI] [PubMed] [Google Scholar]
  9. Booth-LaForce C, & Roisman GI (Eds). (2014). The Adult Attachment Interview: Psychometrics, stability and change from infancy, and developmental origins. Monographs of the Society for Research in Child Development, 79(3), 1–185. 10.1111/mono.12110 [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Bowlby J (1969/1982). Attachment and loss: Vol. I Attachment. Basic Books. [Google Scholar]
  11. Bowlby J (1980). Attachment and loss: Vol. 3. Loss. New York, NY: Basic Books. [Google Scholar]
  12. Burchinal MR, Vandell DL, & Belsky J (2013). Is the prediction of adolescent outcomes from early child care moderated by later maternal sensitivity? Results from the NICHD Study of Early Child Care and Youth Development. Developmental Psychology, 50(2), 542–553. 10.1037/a0033709 [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Caldera YM, & Hart S (2004). Exposure to child care, parenting style and attachment security. Infant and Child Development, 13(1), 21–33. 10.1002/icd.329 [DOI] [Google Scholar]
  14. Carey WB, & McDevitt SC (1978). Stability and change in individual temperament diagnoses from infancy to early childhood. Journal of the American Academy of Child Psychiatry, 17(2), 331–337. 10.1016/S0002-7138(10)60096-9 [DOI] [PubMed] [Google Scholar]
  15. Cohen J (1992). A power primer. Psychological Bulletin, 112(1), 155–159. 10.1037/0033-2909.112.1.155 [DOI] [PubMed] [Google Scholar]
  16. De Wolff MSD, & van IJzendoorn MH (1997). Sensitivity and attachment: A meta-analysis on parental antecedents of infant attachment. Child Development, 68(4), 571–591. 10.1111/j.1467-8624.1997.tb04218.x [DOI] [PubMed] [Google Scholar]
  17. Dearing E, & Zachrisson HD (2017). Concern over internal, external, and incidence validity in studies of child-care quantity and externalizing behavior problems. Child Development Perspectives, 11(2), 133–138. 10.1111/cdep.12224 [DOI] [Google Scholar]
  18. Dykas MJ, Woodhouse SS, Cassidy J, & Waters HS (2006). Narrative assessment of attachment representations: Links between secure base scripts and adolescent attachment. Attachment & Human Development, 8(3), 221–240. 10.1080/14616730600856099 [DOI] [PubMed] [Google Scholar]
  19. Groh AM, Roisman GI, Booth-LaForce C, Fraley RC, Owen MT, Cox MJ, & Burchinal MR (2014). Stability of attachment security from infancy to late adolescence. Monographs of the Society for Research in Child Development, 79, 51–66. doi: 10.1111/mono.12113 [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Haydon KC, Roisman GI, Owen MT, Booth-LaForce C, & Cox MJ (2014). Shared and distinctive antecedents of Adult Attachment Interview state-of-mind and inferred-experience dimensions. Monographs of the Society for Research in Child Development, 79(3), 108–125. 10.1111/mono.12116 [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Hesse E (2008). The Adult Attachment Interview: Protocol, method of analysis, and empirical studies. In Cassidy J & Shaver PR (Eds.), Handbook of attachment: Theory, research, and clinical applications (2nd ed., pp. 552–598). Guildford Press. [Google Scholar]
  22. Hesse E (2016). The adult attachment interview: Protocol, method of analysis, and empirical studies: 1985–2015. In Cassidy J & Shaver PR (Eds.), Handbook of attachment: Theory, research, and clinical applications (3rd ed., pp. 553–597). Guilford Press. [Google Scholar]
  23. Howes C, & Spieker S (2008). Attachment relationships in the context of multiple caregivers. In Cassidy J & Shaver PR (Eds.), Handbook of attachment: Theory, research, and clinical applications (p. 317–332). Guilford Press. [Google Scholar]
  24. Main M, Goldwyn R, & Hesse E Adult attachment scoring and classification system. University of California at Berkeley; 2003-2008. Unpublished manuscript. [Google Scholar]
  25. Main M, Kaplan N, & Cassidy J (1985). Security in infancy, childhood, and adulthood: A move to the level of representation. Monographs of the Society for Research in Child Development, 50(1/2), 66–104. 10.2307/3333827 [DOI] [Google Scholar]
  26. Morrissey TW (2009). Multiple child-care arrangements and young children’s behavioral outcomes. Child Development, 80(1), 59–76. 10.1111/j.1467-8624.2008.01246.x [DOI] [PubMed] [Google Scholar]
  27. National Center for Education Statistics (2006). Table 3: Percentage distribution of children from birth through age 5 and not yet in kindergarten whose parents reported various combinations of weekly care arrangements, by child and family characteristics: 2005. In U.S. Department of Education, National Center for Education Statistics (Ed.), Initial Results From the 2005 NHES Early Childhood Program Participation Survey. Retrieved from https://nces.ed.gov/pubs2006/earlychild/tables.asp. [Google Scholar]
  28. NICHD Early Child Care Research Network (1996). Characteristics of infant child care: Factors contributing to positive caregiving. Early Childhood Research Quarterly, 11(3), 269–306. 10.1016/S0885-2006(96)90009-5 [DOI] [Google Scholar]
  29. NICHD Early Child Care Research Network (1997). The effects of infant child care on infant-mother attachment security: Results of the NICHD Study of Early Child Care. Child Development, 68(5), 860–879. 10.1111/j.1467-8624.1997.tb01967.x [DOI] [PubMed] [Google Scholar]
  30. NICHD Early Child Care Research Network (2001). Child care and family predictors of preschool attachment and stability from infancy. Developmental Psychology, 37(6), 847–862. 10.1037/0012-1649.37.6.847 [DOI] [PubMed] [Google Scholar]
  31. NICHD Early Child Care Research Network (2002). Early child care and children’s development prior to school entry: Results from the NICHD Study of Early Child Care. American Educational Research Journal, 39(1), 133–164. 10.3102/00028312039001133 [DOI] [Google Scholar]
  32. NICHD Early Child Care Research Network. (2004). Fathers’ and mothers’ parenting behavior and beliefs as predictors of child social adjustment in the transition to school. Journal of Family Psychology, 18(4), 628–638. 10.1037/0893-3200.18.4.628 [DOI] [PubMed] [Google Scholar]
  33. NICHD Early Child Care Research Network. (2008). Mothers’ and fathers’ support for child autonomy and early school achievement. Developmental Psychology, 44(4), 895–907. 10.1037/0012-1649.44.4.895 [DOI] [PubMed] [Google Scholar]
  34. Owen MT (1992). The NICHD study of early child care mother–infant interaction scales. Unpublished manuscript, Timberlawn Psychiatric Research Foundation, Dallas, TX. [Google Scholar]
  35. Owen MT, Klausli JK, & Murrey M (2000). The NICHD Study of Early Child Care Parent-Child Interaction Scales: Middle Childhood. Unpublished Manuscript, The University of Texas at Dallas. [Google Scholar]
  36. Pluess M, & Belsky J (2009). Differential susceptibility to rearing experience: The case of childcare. Journal of Child Psychology and Psychiatry, 50(4), 396–404. 10.1111/j.1469-7610.2008.01992.x [DOI] [PubMed] [Google Scholar]
  37. Roisman GI, Haltigan JD, Haydon KC, & Booth-LaForce C (2014). Earned security in retrospect: Depressive symptoms, family stress, and maternal and paternal sensitivity from early childhood to mid-adolescence. In Booth-LaForce C & Roisman GI (Eds.), The adult attachment interview: Psychometrics, stability and change from infancy, and developmental origins. Monographs of the Society for Research in Child Development, 79(3), 85–107. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Roisman GI, Newman DA, Fraley RC, Haltigan JD, Groh AM, & Haydon KC (2012). Distinguishing differential susceptibility from diathesis–stress: Recommendations for evaluating interaction effects. Development and Psychopathology, 24(2), 389–409. 10.1017/S0954579412000065 [DOI] [PubMed] [Google Scholar]
  39. Rosseel Y (2012). lavaan: An R Package for Structural Equation Modeling. Journal of Statistical Software, 48(2), 1–36. [Google Scholar]
  40. Sagi A, Koren-Karie N, Gini M, Ziv Y, & Joels T (2003). Shedding further light on the effects of various types and quality of early child care on infant–mother attachment relationship: The Haifa Study of Early Child Care. Child Development, 73(4), 1166–1186. 10.1111/1467-8624.00465 [DOI] [PubMed] [Google Scholar]
  41. Shpancer N (2006). The effects of daycare: Persistent questions, elusive answers. Early Childhood Research Quarterly, 21(2), 227–237. 10.1016/j.ecresq.2006.04.006 [DOI] [Google Scholar]
  42. Steele RD, Waters TEA, Bost KK, Vaughn BE, Truitt W, Waters HS, Booth-LaForce C, & Roisman GI (2014). Caregiving antecedents of secure base script knowledge: A comparative analysis of young adult attachment representations. Developmental Psychology, 50, 2526–2538. 10.1037/a0037992 [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Tran H, & Weinraub M (2006). Child care effects in context: Quality, stability, and multiplicity in nonmaternal child care arrangements during the first 15 months of life. Developmental Psychology, 42(3), 566–582. 10.1037/0012-1649.42.3.566 [DOI] [PubMed] [Google Scholar]
  44. Vandell DL, Belsky J, Burchinal M, Steinberg L, Vandergrift N, & NICHD Early Child Care Research Network (2010). Do effects of early child care extend to age 15 years? Results from the NICHD Study of Early Child Care and Youth Development: Age 15 Follow-Up. Child Development, 81(3), 737–756. 10.1111/j.1467-8624.2010.01431.x [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Vandell DL, Burchinal M, & Pierce KM (2016). Early child care and adolescent functioning at the end of high school: Results from the NICHD Study of Early Child Care and Youth Development. Developmental Psychology, 52(10), 1634–1645. 10.1037/dev0000169 [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Vaughn BE, Waters TEA, Steele RD, Roisman GI, Bost KK, Truitt W, Waters HS, & Booth-LaForce C (2016). Multiple domains of parental secure base support during childhood and adolescence contribute to adolescents’ representations of attachment as a secure base script. Attachment and Human Development, 18, 317–336. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Waters E & Cummings EM (2000). A secure base from which to explore close relationships. Child Development, 71(1), 164–172. 10.1111/1467-8624.00130 [DOI] [PubMed] [Google Scholar]
  48. Waters HS, & Rodrigues-Doolabh LM (2001). Are attachment scripts the building blocks of attachment representations? Narrative assessment of representations and the AAI. Paper presented at the biannual meeting of the Society for Research in Child Development, Minneapolis, MN. [Google Scholar]
  49. Waters HS, & Waters TEA (in press). Measuring attachment representations as secure base script knowledge: The prompt-word outline method in adulthood, adolescence, and middle childhood. To appear in Waters E, Vaughn BE, & Waters HS (Eds.), Measuring Attachment. Guilford Press. [Google Scholar]
  50. Waters TEA, Fraley RC, Groh AM, Steele RD, Vaughn BE, Bost KK, Verissimo M, Coppola G & Roisman GI (2015). The latent structure of secure base script knowledge. Developmental Psychology, 51, 823–830. [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Waters TE, & Roisman GI (2019). The secure base script concept: An overview. Current Opinion in Psychology, 25, 162–166. 10.1016/j.copsyc.2018.08.002 [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplemental Material
NIHMS1716784-supplement-Supplemental_Material.pdf (267.1KB, pdf)

RESOURCES