program practices, caregiver stability, and child–caregiver relationships
TRANSCRIPT
Applied Developmental Psychology 24 (2003) 497–516
Program practices, caregiver stability, and
child–caregiver relationships
Sharon Ritchie*, Carollee Howes
Graduate School of Education and Information Studies, University of California,
Los Angeles, CA 90095-1521, USA
Abstract
In this article, we examine program practices, caregiver behaviors, and classroom climates
associated with positive child–caregiver relationships. We used the presence or absence of these
practices and our independent observations of child–caregiver interactions and classroom climates to
predict children’s attachment security. Two hundred and fifty-six children (48% girls) from 22
programs serving underrepresented children and families in Los Angeles and rural North Carolina
participated in this research. Three of the programs served only children from difficult life
circumstances. Over half of the children experienced basic stability and uniform/consistent caregiving,
while primary caregiver assignment and looping were rare. Programs specifically serving only children
from difficult life circumstances were more likely to use relationship practices; 74% of them
experienced all four practices, in programs for the other children, none of them experienced all four
practices. The children from difficult life circumstances were in less acrimonious classrooms, and were
more likely to be with teachers who interacted with them intensely and sensitively. Intense and
sensitive interactions, and spending more time with the primary caregiver were most important in
predicting child–caregiver attachment security. Secondary and negative predictors were membership
in a caregiver direction cluster and being assigned to a caregiver.
D 2003 Elsevier Inc. All rights reserved.
Keywords: Caregiver stability; Caregiver behaviors; Child–caregiver relationships; Program practices
0193-3973/$ – see front matter D 2003 Elsevier Inc. All rights reserved.
doi:10.1016/S0193-3973(03)00028-5
* Corresponding author.
E-mail addresses: [email protected] (S. Ritchie), [email protected] (C. Howes).
S. Ritchie, C. Howes / Applied Developmental Psychology 24 (2003) 497–516498
1. Introduction
Children in early childhood programs spend their days with a variety of caregivers. They
often have different caregivers in the morning as they do in the afternoon, and move through
multiple groupings and activities that are facilitated by different adults. Caregivers however
are not interchangeable. Each child–caregiver relationship is with a particular caregiver. A
child in a classroom may very well have different patterns of relationships with the morning
caregiver, the afternoon caregiver, and the assistant caregivers. Every time the child
experiences a new caregiver, both must then engage in the process of constructing a new
relationship. The nature of the relationships that children construct with their caregivers
influences children’s competence and learning while they are in their current program, and as
they move along in school (Howes, 2000; Howes, Matheson, & Hamilton, 1994; Howes,
Phillipsen, & Peisner-Feinberg, 2000). In this article, we examine program practices that
inhibit or promote the development of secure attachment relationships between caregivers and
children—caregiver stability, caregiver behaviors, and classroom climates associated with
positive child–caregiver relationships.
1.1. Secure attachment with caregivers
In general, child–caregiver attachment security is independent of child–mother attachment
security (Goossen & van IJzendoorn, 1990; Mitchell-Copeland, Denham, & DeMulder, 1997;
Pianta, Nimetz, & Bennett, 1997). Each new caregiver has the opportunity to construct a
different relationship than the child has previously experienced. However, when children
come from difficult life circumstances they may bring to new relationships prior histories of
acrimonious, conflictual, neglectful, or unstable interaction and caregiving. These children
who have experienced difficult life circumstance often tend to act towards new caregivers as
if they too will be untrustworthy partners (Howes & Ritchie, 2002).
Caregivers may have to be particularly sensitive and talented to construct a secure
attachment relationship with a child with prior difficult life circumstances. Children who
have been adopted from Romanian orphanages (Chisholm, 1998), and from the United States
foster care system (Marcus, 1991; Ritchie, 1995) are able to construct secure attachment
relationships with their new caregivers. The likelihood of achieving a secure relationship is
increased when the caregivers are rated as highly sensitive, consistently positive, and
committed to the loving caregiver role. Therefore, in our current work we were particularly
interested in the relationships of the children enrolled in programs only for children with
difficult life experiences.
Ideas drawn from attachment theory and empirical research on child–caregiver relation-
ships suggest that the particular nature of children’s social interactions with their primary
caregivers and the emotional climate of these interactions (Boyce et al., 1998; Cassidy &
Shaver, 1999; Howes, 1999) influence the nature of relationships. Thus, all children who
have warm, responsive, and individualized interactions with caregivers in the context of a
harmonious classroom emotional climate are more likely to form secure attachment relation-
ships with their caregivers (Goossen & van IJzendoorn, 1990; Howes & Hamilton, 1992;
S. Ritchie, C. Howes / Applied Developmental Psychology 24 (2003) 497–516 499
Howes & Smith, 1995; Kontos, Howes, Shin, & Galinsky, 1995). These findings are
consistent with a large and well established body of research on mother–child attachment
relationships, which finds that warm, responsive, and sensitive mothers construct secure
mother–child attachments with their children (Bretherton, 1985). Accordingly, we examined
children’s experiences of caregiving with particular attention to the emotional quality of
child–caregiver interactions.
1.2. Caregiver stability
A body of literature suggests that children who experience instability in caregivers are less
likely to form positive relationships with caregivers. The prototype context for this concern is
foster care and adoption rather than childcare. Children who experience extremes of
instability, particularly as infants, tend to have persistent problems with positive relationship
formation (Marcovitch et al., 1997; Rutter, 1999). Because childcare environments frequently
involve multiple caregivers and high turnover rates there has been some documentation of the
negative impacts of instability of caregivers in childcare. The National Child Care Staffing
Study (Whitebook et al., 1990), the Child Care Employee Project (Whitebook, Phillips, &
Howes, 1993) and the NAEYC Accreditation as a Strategy for Improving Child Care Quality
(Whitebook et al., 1997) have all established as best practice the need to reduce high turnover
rates amongst teaching staff. Children in the National Child Care Staffing Study who were
enrolled in centers with higher rates of caregiver turnover spent less time engaged in social
activities with peers, more time in aimless wandering, and scored lower on assessments of
language development (Whitebook, Howes, Phillips, & Pemberton, 1989). Furthermore,
children who experienced more caregivers between the ages of one and four were more
aggressive with peers than children who had fewer caregiver changes (Howes & Hamilton,
1993). Infants who spent more time with the same caregiver were rated (by the caregiver) as
having a more secure attachment relationship with their caregiver (Raikes, 1993). The
National Child Care Staffing Study, more than any previous research effort, made it apparent
that as long as the early childhood field failed to resolve the staff compensation crisis that
contributes to high turnover, that childcare’s capacity to nurture children and assist families
would continue to be shortchanged (Whitebook et al., 1993). Childcare quality also has been
found to affect turnover. Centers that retained a greater percent of highly skilled teachers are
significantly more likely to receive good or better ratings on overall classroom quality
(Whitebook et al., 1997).
1.3. Relationship-based practices
The response of the early childhood field to the issues of high turnover rates and the
importance of positive child–caregiver relationships has been to develop a number of
practices designed to mitigate the potentially negative influences of caregiver instability
and to enhance positive child–caregiver relationships. Although widely accepted as ‘‘best
practice,’’ and drawn from years of experience, there have been few empirical tests of the
efficacies of these practices. As part of our ethnographic work with programs designated by
S. Ritchie, C. Howes / Applied Developmental Psychology 24 (2003) 497–516500
communities as successful in their work with children and families, we identified a number of
practices the program participants believed to be important in the development of positive
relationships. In this paper, we used the presence or absence of these practices as well as our
independent observations of child–caregiver interactions and classroom climates to predict
children’s attachment security.
To reduce caregiver instability, programs attempted a number of strategies to keep
instability to a minimum. At the most basic level caregivers were assigned to a group of
children, usually a classroom, for 1 year. Caregivers were not rotated between groups and
caregivers did not move to different classrooms or sites during the week. Programs who
articulated a particular concern with relationship formation added at least one of the three
following practices to their repertoire: primary caregiver assignment; uniform/consistent
caregiving, and/or looping. When programs practiced primary caregiver assignment, the
director or head caregiver assigned each child within a group to a particular caregiver. The
assigned caregiver was to provide all emotionally salient caregiving for that child—greeting
the child in the morning, helping the child with meals and toileting or diapering, putting the
child to nap, and monitoring the child’s play throughout the day. Programs that practiced
uniform/consistent caregiving believed that all caregivers in the program should share and
practice a uniform/consistent philosophy. They made purposeful efforts in staff meetings and
in-service training to insure that caregivers maintained shared beliefs around ideas of
positively relating with children, language development, separation from parents, and child
involvement in the program activities. In programs that practiced looping, a group of children
were assigned to a head caregiver for more than 1 year.
1.4. Positive classroom climate
The emotional climate of the classroom is a relatively recent addition to the construct of
childcare quality. It refers to the tone of the classroom and can range from positive and
prosocial to acrimonious and harsh. Early classroom climates in one study, predicted social
competence with peers in second grade (Howes, 2000). Acrimonious interactions disrupt the
learning of the child involved and at times, the entire classroom. Acrimonious interactions
can involve conflicts between teachers and children or between children and children. They
are often marked by verbal or physical aggression, disregard for classroom rules of conduct,
or disputes over materials. We prefer to call them acrimonious rather than conflictual
interactions because in these episodes children introduce disruptive behaviors and teachers
can respond by escalating or de-escalating behaviors. For classroom interaction to be
harmonious, there must be agreement on the rules of conduct, specifically the explicit and
implicit rules for which behaviors are permitted and forbidden in classrooms.
Important to examining the tone of a classroom is an awareness of the nature of the
difficult relationships within it. Attachment theorists have described two attachment
organizations associated with insecure relationships: avoidant and ambivalent/resistant
(Ainsworth, Blehar, Waters, & Wall, 1978). Children with avoidant attachment organiza-
tions turn away rather than seek comfort from adults because they have experienced
rejection and insensitivity from adults (Ainsworth et al., 1978). Since they expect the adult
S. Ritchie, C. Howes / Applied Developmental Psychology 24 (2003) 497–516 501
to reject them, they tend to make ‘‘preemptive strikes,’’ acting in a hostile fashion before
the adult has an opportunity to be rejecting. Alternatively, they may avoid the adult to
avoid being rejected. Children with avoidant maternal attachment histories tend to be rated
by teachers as high in aggression and passive withdrawal (Renken, Egeland, Marvinney,
Mangelsdorf, & Sroufe, 1989).
Children with an ambivalent/resistant attachment organization also do not trust their
attachment figure to provide comfort and emotional security. However, in contrast to children
with avoidant attachments, their experience of the attachment figure has been inconsistent,
and thus confusing (Cassidy & Berlin, 1994). Sometimes, the adult will be there for them and
other times the adult will withdraw from the child. Children with ambivalent/resistant
attachment organization tend to be dependent and hard to comfort. These children may
appear to seek comfort, and then reject the adult’s attempts to provide it. Thus, these children
are both ‘‘clingy’’ and difficult. Children with a history of ambivalent/resistant maternal
attachment are characterized as fearful and inhibited in exploration with both peers and
materials (Cassidy & Berlin, 1994). Children with ambivalent/resistant attachment organ-
izations can also be disruptive within classrooms. Unlike the child with an avoidant
attachment organization, the child with an ambivalent/resistant attachment organization uses
disruptive behavior to draw the teacher into interpersonal conflict.
In a positive classroom climate, the goal of teachers is to maximize harmonious
interactions by keeping acrimonious interactions from becoming conflicts that disrupt
learning. It would be logical to assume that children from difficult life circumstances
would be in more acrimonious classrooms, as children who act out more could prompt
more teacher frustration, more time-outs and reprimands. Consideration, however, of the
teachers who choose to work in classrooms with these children in programs that have a
philosophical base which supports the development of positive relationships can perhaps
reduce the likelihood of acrimony for children who have likely already experienced far too
much of it.
1.5. Summary
As early childhood classrooms become increasingly burdened with federal and state
mandates for child outcomes in literacy and math, it is essential to bear in mind the practices
and program philosophies that may promote children’s success in school which are not
directly linked to academic progress. This study examines the ground work, the basic
principles and practices which allow children to form important relationships with their
caregiver so that they are ready to take the opportunities to learn that are available to them in
the programs they attend. It is essential to remain mindful of this idea as we explore the
pathways that lead children to success. Educational policies that compel an academic agenda
that forsakes the development of the whole child may be both shortsighted and counter-
productive. Children with more positive child–teacher relationships appear more able to
make use of the learning opportunities available in classrooms (Howes & Smith, 1995) and
better adjust to the demands of formal schooling (Birch & Ladd, 1997). Awareness of the
specific practices that may well be the precursors to success for children is important for the
S. Ritchie, C. Howes / Applied Developmental Psychology 24 (2003) 497–516502
lives of the children in the programs as well as for the efficacy of the people who provide
their care.
In summary, we examined the prevalence of practices to mitigate against caregiver
instability and enhance positive relationships in programs serving underrepresented children
and families, associations among these practices, classroom emotional climate, caregiver
behaviors, and the prediction of child–caregiver attachment security from practices, class-
room emotional climate, and caregiver behaviors. Our design included assessing the integrity
of classroom practices, conducting observation of children and caregiver behavior, and
assessing child–caregiver attachment relationship quality.
2. Method
2.1. Participants
2.1.1. Selection of sites
2.1.1.1. Advisory board. Ten respected early childhood professionals identified 23 exem-
plary programs that worked with low income, minority children in Los Angeles and rural
North Carolina. This was not a random sample, but rather one purposefully comprised of
programs that earned the respect of the community they served. Baseline data on environ-
mental quality was gathered at each nominated site (n= 23). Classroom scores on the Early
Childhood Environmental Rating Scale (ECERS) (Harms & Clifford, 1980) averaged 5.67
(SD = .75; range 5.41–6.35). No classroom had an overall score lower than 5. ECERS scores
of 5 and above are considered to be good to excellent in quality. Additionally, researchers
talked with caregivers and directors about who was served by the program and the purpose
and mission of their program. Profiles for each site were developed and presented it to a
community advisory board to structure a discussion on dimensions of quality.
2.1.1.2. Best practices programs. The final selection of 16 participating sites was made by
the project staff based on the following criteria: service to low income children; representa-
tion of ethnic groups similar to that of Los Angeles County and rural North Carolina; a mix of
centers serving children from predominantly one ethnic or home language group, and
programs serving children from diverse backgrounds; representation of children with special
needs both included into programs for typical children and in ‘‘stand alone programs’’; a
variety of distinct practices including practices that appeared to be debatable within the
childcare community (e.g., home language vs. English only; child initiated vs. didactic
approaches to teaching).
In a subsequent phase of the study, the advisory board and the original 16 sites nominated
10 additional partner sites. The criteria for participation included similar demographics to the
original sites, and a willingness to engage in thinking and talking about their practices with
researchers. Partnership sites were chosen to determine if those practices observed in the
community nominated sites could be introduced into similar environments of overall lower
S. Ritchie, C. Howes / Applied Developmental Psychology 24 (2003) 497–516 503
quality as determined at baseline level by the ECERS. Although average environmental
quality scores for these sites were good (mean ECERS score of 5.56), the range of ECERS
scores (3.83–6.06) indicated not all classrooms were in the good quality range. The study
was both a close examination of the process through which change occurs, as well as a look at
the change itself.
2.2. Sample
Two hundred and fifty-six children (48% girls) participated in this research. One hundred
and eighty-three of the total sample of children were randomly selected to have approx-
imately 25% of the children in each program participate. The remaining 77 children were
selected from a pool of children at 12 of the Best Practices Programs whose parents
volunteered to have them participate in a longitudinal study. Ninety-seven percent of the
parents volunteered to participate. We randomly selected two girls and two boys from each of
the 20 classrooms (n= 80) included in the 12 Best Practices Programs used in the longitudinal
study. Three children were persistently absent reducing our n to 77. We used a MANOVA to
examine age, gender, and ethnic differences in the sample of children randomly selected and
those from parental volunteers and found no significant differences.
Thirty-two percent of the children were African American, 39% Latino, 16% White, 7%
Asian, and the remaining biracial. Sixty-two percent of Latino children, 52% of White
children, and 81% of African American children were taught by teachers from their own
ethnic backgrounds. The children averaged 50 months (SD= 10.8) at the time of the first
observation and ranged in age from 15 to 82 months.
A subsample of these children (n = 31) were enrolled in one of three programs that only
included children with particular difficult life circumstances. By difficult life experiences we
mean maltreatment, parental psychopathology or substance abuse, homelessness, and multi-
ple placements. While we assume that there were children with difficult life circumstance in
all programs, these three programs were exclusively for children with difficult life experi-
ences: a therapeutic preschool; a program for emancipated teen mothers who lived with their
children at the facility; and a program for the children of teen age parents who had
experienced domestic violence. The programs we included in this subsample served children
of teen parents who were themselves often struggling with substance abuse and/or domestic
violence, poverty, and low education levels, as well as children who had been diagnosed as
having special needs, (posttraumatic stress disorder, separation anxiety, attachment disorders,
oppositional defiant disorder). Many of the children entered the programs with aggressive,
depressed, or withdrawn behaviors, which suggested that their prior relationships with adults
had been problematic. Many of the children in the programs continued to experience difficult
life circumstances throughout their time in the programs. Only 41% of these children were
girls because the therapeutic preschool enrolled more boys than girls. The ethnic distribution
of this subsample was dissimilar to the larger sample by having somewhat more White (28%)
and somewhat fewer African American children (28%) but similar proportions of Latino
children (45%). The children in this subsample averaged 50.48 months (SD= 15.8) and
ranged in age from 15 to 76 months.
S. Ritchie, C. Howes / Applied Developmental Psychology 24 (2003) 497–516504
2.3. Procedure
The study took place over the course of two school years. In each year, Time 1 data were
collected in November, and Time 2 data in May. Eight trained observers conducted the
observations. Additionally, a research partner acted as a participant observer for at least a year
in the program. The participant observer spent one full day each week getting to know the
programs and the staff, taking extensive field notes, interviewing staff, and in the end, helping
to interpret and make use of the data. Each was comfortable in early childhood classrooms,
and was matched to the ethnicity of caregivers and/or the majority of the children in the
classrooms. In all cases, Spanish-speaking participant observers were in classrooms where
Spanish was the primary language. On the basis of field notes and interview transcripts,
programs were rated as ascribing, or not to practices. Members of the research team who had
not served as participant observers in that particular program conducted the observations and
assessments. Each data collector spent at least 4 h observing each study child and their
primary caregiver at each data collection period.
2.4. Measures
2.4.1. Attachment Q-sort-security of attachment with primary caregiver
To examine the relationships between individual caregivers and children, we used the
Attachment Q-Set (AQS) (Waters, 1990).The AQS has good validity with the Strange
Situation (Howes & Hamilton, 1992; van IJzendoorn, Vereijken, & Riksen-Walrave, in
press). Following their observations, the observers sorted the 90 behavior descriptions that
serve as items in the AQS, i.e., ‘‘this child turns to the caregiver when she is upset,’’ into nine
piles of 10 items each. In this way the observers described the child–caregiver relationship
according to the behavior descriptions. The particular behavior was given a score (1–9)
corresponding to very characteristic (pile no. 9) to very uncharacteristic (pile no. 1). The item
scores for each child–caregiver relationship were correlated with an ideal child–caregiver
relationship (Waters, 1990) to determine the child’s security score. Scores may vary from
� 1.0 to 1.0. A higher score indicates greater security, with a .33 as the determinant for the
lowest score describing a secure relationship. Subscale scores were developed based on
clusters of AQS items (Howes & Ritchie, 1999). As a result each child–caregiver relationship
could be described as high or low on the following subscales: avoidant, resistant, seeking
comfort, using the caregiver as a secure base, and engaging in harmonious interactions. The
avoidant and resistant subscales conceptually correspond to insecure attachment organiza-
tions. The other three subscales correspond to secure attachment organizations. We catego-
rized each child’s child–caregiver attachment relationship based on a cluster analysis of AQS
items (Howes & Ritchie, 1999). The categories of attachment organization were disorganized,
avoidant insecure, resistant insecure, near secure and secure.
Observers were trained to an 85% exact agreement criterion on each item prior to data
collection with the gold standard trainer. Monthly observer–trainer reliability checks were
conducted throughout data collection. Median interobserver reliability for exact item place-
ment was Kappa = .87 (range Kappa = .79 to .96).
S. Ritchie, C. Howes / Applied Developmental Psychology 24 (2003) 497–516 505
2.5. Snapshot observations
We used a ‘‘snapshot’’ (time sampling) procedure to capture aspects of adult–child
interaction including time with caregivers and caregiver involvement and engagement with
the child. We also examined the activities in which the children were engaged as the context
for the adult–child interaction. Over the course of a program morning, 50+ snapshots were
collected for each child. To complete a snapshot, the data collector observed a target child for
20 s, and then had 40 s to record presence of the variables described in the next sections. The
observer then watched the next child and proceeded in the same manner to observe a total of
four study children. This process was repeated throughout the program morning for a total of
4 h (50+ snapshots for each child).
Observers were trained to an 85% exact agreement criterion on each item prior to data
collection. Monthly interobserver reliability checks were conducted throughout data collec-
tion. Median interobserver reliability for exact item placement was Kappa = .84 (range
Kappa = .79 to .93).
Involvement and engagement were coded during the snapshot observations.
2.5.1. Involvement
The adult involvement scale (Howes & Stewart, 1987) rates the intensity of child–
caregiver involvement. Each time the caregiver was within 3 ft of the child, or a child–
caregiver interaction occurred during the 20-s observation period, an aspect of adult
involvement with the child was scored. Therefore, a child could have a maximum of 50
ratings (one for each snapshot scored). Only one of the following scale points could be
recorded in each snapshot: (0) Monitor (caregiver is close to the child but does not engage
in interaction); (1) Routine (caregiver touches the child for routine caregiving, or passes out
materials, but has no verbal interaction with the child); (2) Minimal (caregiver touches the
child only for necessary redirection, answers child requests in a few words, or makes a
brief remark.); (3) Simple (caregiver provides helpful physical support, verbally answers the
child in simple sentences, ‘‘Yes, you need to glue that piece.’’ or initiates simple social
interaction such as ‘‘What book are you reading today?’’); (4) Elaborate (caregiver
maintains close proximity to the child, responds to child’s statements without restating,
or asks and answers complex questions); (5) Intense (caregiver hugs/holds the child,
answers by restating the child’s statement, engages the child in reciprocal conversation, or
plays interactively).
The scale points for the adult involvement codes represent increasing complexity and
reciprocity in adult–child interaction. We created a total score for adult–child involve-
ment by summing the levels of involvement. We also scored the percent of time the
caregiver was within 3 ft of the child during each of the six levels of involvement.
Finally, we created a set of codes representing the proportion of each type of
involvement over the total number of observations and scored ‘‘high’’ caregiver
involvement as the proportion of simple, elaborated and intense involvement scores,
and ‘‘low’’ caregiver involvement as the proportion of monitor, routine and minimal
scores.
S. Ritchie, C. Howes / Applied Developmental Psychology 24 (2003) 497–516506
2.5.2. Engagement
During each observation we coded the type of engagement the caregiver had with the
child. The caregiver was not required to be within 3 ft of the child, and more than one coding
category could be recorded in each snapshot. The following categories comprised the
engagement index: (1) Positive initiation (caregiver initiates positive physical/verbal inter-
action with the child without child’s solicitation); (2) Positive response (caregiver responds in
a positive verbal/physical manner to the child’s social bids); (3) Positive management
(caregiver verbally intervenes, redirects the child or reminds the child of the rules for
behavior); (4) Negative management (caregiver handles problem/misbehavior in a harsh or
negative manner, including time-out, yelling, belittling, spanking, or threatening); (5)
Facilitate peer (caregiver promotes the child’s interaction with other children); (6) Language
play (caregiver engages in language play with the child or group of children, including
rhyming games, reading one-on-one, or a social conversation).
2.5.3. Learning and play activities
This set of codes captures the activity that the teacher provides for the children. These
categories were (1) Creative activities (fantasy play, blocks, and open-ended art); (2)
Language arts (looking at or reading books to self, being read to, listening to a story, music
activities, and circle time); (3) Didactic learning (teacher-modeled art projects or teacher-
directed drill and practice of school skills); (4) Fine motor (use of art materials, puzzles,
stringing beads, pegboards, or Legos); (5) Routine (transitioning to the next activity,
grooming activities, meal, and snack time) (6) Unoccupied (child is not engaged with any
other person, any object or any specific activity). During each snapshot, the observer selected
only one activity that best fit what the child was doing. For analysis, each child received a
score for the proportion of time they were involved in these activities.
The proportion of time for each engagement category over the entire observation period
was scored. In addition, the proportion of time of total engagement for each category was
scored.
2.6. Time with primary caregivers
Because each snapshot observation of a child that included adult involvement or engage-
ment was linked to the identity of the adult, we were able to separately consider the children’s
experiences with their primary caregiver, their other caregivers, or when there was no
engagement with a caregiver. A primary caregiver was defined as the caregiver to whom the
child was assigned, or as the head caregiver/lead teacher.
2.7. Classroom emotional climate
We devised two sets of class scores to capture classroom emotional climate. ‘‘Acrimony’’
represented the degree of child–caregiver conflict, and the ‘‘nature of difficult relationships’’
represented the proportion of children in the room with each type of insecure attachment
relationship (avoidant insecure, resistant insecure, near secure, and secure).
S. Ritchie, C. Howes / Applied Developmental Psychology 24 (2003) 497–516 507
2.7.1. Classroom interaction scale
The classroom interaction scale (Arnett, 1989) is a 26-item 4-point rating scale completed
by the observer of the head caregiver in the classroom at the conclusion of the observation
period. It yields three scores: sensitive (warm, attentive, and engaged); harsh (critical,
threatening, and punitive) and detached (low levels of interaction, interest, and supervision).
Scores from this instrument have been found to predict childcare caregiver’s involvement
with children and the children’s social competence (Howes, Whitebook, & Phillips, 1992)
and attachment security (Howes & Hamilton, 1992). Median interobserver reliability was
Kappa = .85 (range Kappa = .81 to .89).
We created a classroom acrimony score by summing the standard scores for harshness and
detachment with the standard score for proportion of adult engagement in negative manage-
ment, from the adult engagement codes.
2.7.2. Attachment organization
Based on our prior work that found children categorized as disorganized, resistant and
avoidant in AQS attachment organizations to be higher in behavior problems than children
categorized as secure or near secure in AQS attachment organization, we calculated the
proportion of children in each room categorized as disorganized, resistant, and avoidant. This
gave us the number of children who had difficult relationships with their teachers (Howes &
Ritchie, 1999).
3. Results
3.1. Child–caregiver attachment relationships
A security score of .33 and above indicates a secure relationship with the caregiver.
Security scores for children in the study averaged .29 (SD = .24) and ranged from � .52 to
.74. The mean indicates low level secure relationships and the range indicates both some very
difficult relationships as well as some that were very positive. There were no relations
between age and security and no ethnic differences in security scores. Gender and ethnicity
was determined by chi squares and an ANOVAwas used to test age. As expected, children in
the only difficult life circumstances programs had lower child–caregiver attachment security
scores, M(difficult) = .16; SD = .27; M(other) = .31, SD = .23; t(254) = 3.31, p = .001.
3.2. Practices
Sixty percent of the 256 children, (enrolled in eight different programs) experienced basic
stability practices. Eleven percent of the children (enrolled in three different programs)
experienced primary caregiver assignment. Fifty-two percent of the children (enrolled in
seven different programs) experienced uniform/consistent caregiving. Fifteen percent of the
children (enrolled in four different programs) experienced looping or being kept with the
same caregiver and peer group over more than 1 year. Children enrolled in only difficult life
S. Ritchie, C. Howes / Applied Developmental Psychology 24 (2003) 497–516508
circumstances programs were more likely to experience basic stability, 81%; v2(1) = 6.40;p < .01, primary caregiver assignment, 86%; v2(1) = 168.72; p < .001, uniform/consistent
caregiving, 81%; v2(1) = 11.63; p < .001, and looping, 100%; v2(1) = 168.72; p < .001, than
children in other programs (basic stability 57%; primary caregiver assignment 2%; uniform/
consistent caregiving 4%, and looping 4%). Seventy-four percent of the children in the
difficult life circumstances subsample experienced all four practices, while 26% experienced
one. In programs for the other children, 37% experienced one practice, 53% experienced two
and 9% experienced three. None of these children experienced all four practices.
3.3. Classroom climate
3.3.1. Acrimony
Acrimony scores averaged 0 (SD = 2.11) and ranged from � 2.33 to 10.17. There were no
gender, ethnic, or age associations with acrimony. Gender and ethnicity was determined by
chi squares and an ANOVAwas used to test age. Children in only difficult life circumstances
programs had lower acrimony scores than children in other programs, M(difficult) =� 0.68,
SD = .81; M(other) = .09, SD = 2.21; t(98) = 3.64, p < .001.
3.3.2. Proportion children of children in the classroom with insecure attachment
organizations
We categorized approximately half of the children as having secure child–caregiver
attachment behavior organizations. There were no gender, ethnic, or age associations with
attachment behavior organizations. Children in the only difficult life circumstances programs
were less likely to have secure child–caregiver attachment behavior organizations and more
likely to have disorganized or resistant child–caregiver attachment behavior organizations
than children in other programs, v2(4) = 17.15; p < .01 (see Table 1).
The proportion of children in classrooms with avoidant child–caregiver attachment
behavior organizations ranged from none to 50% (M = .10, SD = .11), with resistant
attachment behavior organizations from none to 25% (M = .02, SD = .05), and with
disorganized behavior attachment organizations from none to 25% (M = .07, SD = .07).
Children in the only difficult life circumstances programs were more likely to have higher
proportions of children with avoidant attachment behavior organizations, M(difficult) = .14,
SD = .09; M(other) = .10, SD = .10; t(254) = 2.47, p < .01, resistant attachment behavior
Table 1
Percentage of attachment behavior organizations in different programs
Type of attachment organization All children (%) Only difficult (%) Other programs (%)
Secure 51 29 53
Near secure 29 32 28
Avoidant 10 14 10
Resistant 1 7 1
Disorganized 9 18 8
S. Ritchie, C. Howes / Applied Developmental Psychology 24 (2003) 497–516 509
organizations, M(difficult) = .07, SD = .06; M(other) = .01, SD = .4; t(94) = 5.11, p < .001,
anddisorganizedattachment behaviororganizations,M(difficult) = .18,SD= .11;M(other)= .06,
SD = .08; t(94) = 6.15, p < .001, than children in other programs.
3.4. Adult involvement and engagement
We used K-means cluster analysis to create composite variables. Cluster analysis is a
technique most commonly used to identify taxonomies or data structure. A k-means cluster
analysis is a tool to identify underlying structure in the observed data. In this procedure initial
cluster centers are randomly assigned and then cases are selected to the next closest cluster
based on distance from cluster center. The procedure repeats reassignment and recalculation
of cluster centers until centers cease changing. The cluster analysis using involvement and
engagement codes converged in three iterations. The minimum distance between initial
centers was 1.6465 and the maximum distance by which any center changed was 0.0275
indicating that the structure was stable.
The descriptive statistics for the adult involvement and engagement clusters are in Table 2.
The adult involvement and engagement codes resulted in three clusters that represent different
experiences that children had with their primary caregiver. Children in cluster 1 (31% of the
children) experienced the most simple and minimal involvement and positive management.
We named cluster 1 caregiver direction. Children in cluster 2 (48% of the children)
experienced the most positive initiations and responses and were similar to children in
Table 2
Descriptions and differences among adult involvement and engagement clusters: Proportion of involvement or
engagement
Caregiver direction
cluster (1)
Interactive
cluster (2)
Language play
cluster (3)
F Post hoc
M SD M SD M SD
Adult involvement
Monitor .14 .17 .09 .12 .17 .19 5.68* 3,1 > 2
Routine .05 .08 .03 .06 .06 .11 2.19
Minimal .18 .17 .10 .10 .16 .21 7.98** 1,3>2
Simple .28 .31 .19 .15 .12 .14 9.79*** 1>3,2
Elaborated .26 .21 .32 .19 .28 .21 2.42
Intense .09 .10 .27 .21 .22 .27 19.69 2,3>1
Adult engagement
Positive initiation .14 .13 .33 .18 .16 .15 41.76*** 2>1
Positive response .17 .16 .30 .19 .11 .10 30.64*** 2>3,1
Positive management .47 .24 .11 .10 .08 .08 161.41*** 1>2,3
Facilitate peer .04 .07 .05 .07 .07 .09 2.69
Language play .16 .18 .16 .14 .57 .20 132.66*** 3>2,1
*p < .05. **p < .01. ***p < .001.
S. Ritchie, C. Howes / Applied Developmental Psychology 24 (2003) 497–516510
cluster 3 in intense involvement. Cluster 2 was named interactive. Children in cluster 3 (21%
of the children) experienced the most language play, were similar to children in cluster 1 in
monitored involvement and to children in cluster 2 in intense involvement. Cluster 3 was
named language play.
There were no gender, ethnic, or age associations with adult involvement and engagement
cluster membership. Children in the only difficult life circumstances programs were more
likely to be in the interactive cluster and less likely to be in the caregiver direction and
language play clusters than children in other programs, v2(2) = 19.13; p < .001 (see Table 3).
3.4.1. Time with primary caregivers
We conducted a 2 (type of program)� 3 (type of caregiver) ANOVA with a repeated
measure on the last factor on amount of time. Children spent, on the average, 22% (SD = .16)
of their time with their primary caregiver, 29% (SD = .18) with the other caregivers
combined, and 49% (SD = .21) with no adult. There were no age, ethnic, or gender
differences in time spent with caregivers. We used a repeated measure ANOVA with only
difficult life circumstances programs as the grouping variable to compare time distributions.
There was a significant interaction between program and time, F(2,250) = 2.54, p < .01, and
a significant main effect for time, F(2,250) = 66.41, p < .001. All children spent more time
without an adult than with an adult. Children in only difficult life circumstances programs
spent more time than other children with no adult, M(difficult) = .58, SD = .12; M(other) =
.49, SD = .24, and less time with their primary caregiver, M(difficult) = .13, SD = .07;
M(other) = .23, SD = .15.
3.5. Associations among practices, classroom climate, involvement and engagement with
adults, and child–caregiver attachment security
Descriptive information for the proportion of adult involvement and engagement is in
Table 4. High level adult involvement (M = .70, SD = .27, range: 0–1.00). Adult engagement
for positive initiation, positive response, positive management, and language play were quite
similar (M = .23, SD = .22, range: 0–1:00). Engagement in facilitating peer interaction was
lower (M = .05, SD = .07, range: 0–36).
3.6. Relationship based practices
We used a series of two-way ANOVAwith practices (yes or no) and program for children
with difficult life circumstances (yes or no) as grouping variables to examine differences in
Table 3
Percent differences in adult involvement and engagement clusters
Clusters Only difficult (%) Other (%)
Caregiver direction 10 33
Interactive 86 43
Language play 3 23
Table 4
Proportion of adult involvement and engagement
Mean SD High Low
Involvement
Monitor .12 .15 .67 0
Routine .04 .08 .50 0
Minimal .14 .16 .96 0
Simple .20 .22 1.00 0
Elaborated .29 .20 .88 0
Intense .20 .21 .89 0
Percent high level involvement .70 .27 1.00 0
Engagement
Positive initiation .24 .18 1.00 0
Positive response .22 .19 1.00 0
Positive management .21 .23 1.00 0
Facilitate peer .05 .07 .36 0
Language play .25 .24 1.00 0
S. Ritchie, C. Howes / Applied Developmental Psychology 24 (2003) 497–516 511
child–caregiver attachment security when children did or did not experience the four
identified relationship practices. There were no significant main effects for practice and no
significant interactions.
3.6.1. Classroom climate
We used Pearson product–moment correlations to examine associations between class-
room climate and child–caregiver security scores (see Table 5). Children’s security scores
were negatively associated with the percent of children in their classrooms with avoidant
attachment behavior organizations.
Table 5
Associations between child–caregiver security and classroom climate and time with caregiver
Security scores (Pearson product–moment correlations)
Classroom climate
Acrimony � .08
% avoidant � .21**
% resistant � .19**
% disorganized � .13*
Time with caregiver
Primary .29**
Others .01
No caregiver � .18**
*p < .05. **p < .01.
S. Ritchie, C. Howes / Applied Developmental Psychology 24 (2003) 497–516512
3.6.2. Adult involvement and engagement
We used a two-way ANOVA with involvement and engagement cluster membership and
program for children with difficult life circumstances (yes or no) as grouping variables to
examine differences in child–caregiver attachment security scores. Children in the interactive
cluster (M = .35, SD = .23) had higher scores than children in the language play cluster (M = .29,
SD = .22) and caregiver direction cluster (M = .19, SD = .24), F(2, 251) = 3.52, p < .05. There
were no significant main effects or interactions for program enrollment.
We used Pearson product–moment correlations to examine associations between time with
caregivers and child–caregiver security scores (see Table 5). Children who spent more time
with their primary caregiver and less time away from adults had higher security scores.
Table 6
Predicting child–caregiver attachment scores
Security
R R2 R2 change Final h sri2
Model I: Includes proportion of insecure children as part of classroom climate
Enrollment in program for children with
difficult life circumstances
.20** .04 � .01 � .00
Classroom climate and adult involvement
and engagement
.48** .24 .20
Acrimony � .02 � .03
% avoidant � 1.31 � .08
Caregiver direction � 2.18* � .14*
Interactive 3.00** .20**
Time with primary caregiver 3.58** .25**
Practices .50** .25 .01
Basic stability .04 .03
Uniform .00 .01
Assign � 1.70 � .12*
Looping � .46 � .03
Model II: Does not include proportion of insecure children as part of classroom climate
Enrollment in program for children with
difficult life circumstances
.20** .04 .17 .16
Classroom climate and adult involvement
and engagement
.49** .24 .20
Acrimony � .11 � .10
Caregiver direction � .37** � .24*
Interactive .27** .20**
Time with primary caregiver .56** .45**
Practices .50** .26 .02
Basic stability .03 .03
Uniform .02 .01
Assign � .22 � .20*
*p < .05. **p < .01.
S. Ritchie, C. Howes / Applied Developmental Psychology 24 (2003) 497–516 513
3.7. Predicting child–caregiver attachment security
The final step in our analysis was to test a predictive model using sequential multiple
regression. We completed this analysis twice with and without the proportion of the children
in the classroom with insecure child–teacher attachment relationships as a predictor. In both
equations, we first entered enrollment in a program only for children with difficult life
circumstances because we expected from theory and our descriptive analyses that these
children present special challenges to caregivers. We then entered classroom climate and adult
involvement and engagement together because prior research and our preliminary analysis
suggested relations. In the analysis including attachment as part of classroom climate we used
only one of the three proportion of children with insecure attachment behavior organization
variables, and dummy coded involvement and engagement membership to avoid multi-
colinearity. We entered practices as the last step in the analysis, although there were no simple
relations between practices and security scores, because it was the primary focus of our
analysis. Table 6 displays the standardized regression correlations (h), the semipartial
correlations (sri2), and R, adjusted R2, and change in R2 at each step for each of the
regression models.
In both models classroom climate and teacher involvement and engagement best predicted
security scores; relationship practices added little to the predictive power of the model.
Inspection of the final standardized regression correlations and the semipartial correlations
suggest that membership in an interactive cluster and spending more time with the primary
caregiver were most important in predicting child–caregiver attachment security. Secondary
predictors and negative predictors were membership in a caregiver direction cluster and being
assigned to a caregiver.
4. Discussion
Over half of children experienced basic stability and uniform/consistent caregiving,
while primary caregiver assignment and looping were rare. Traditionally children are
placed in classrooms or groups for 1 year. High turnover rates in childcare, rather than
philosophical approach may account for lack of basic stability for those children who did
not experience the same caregiver over the course of a year. Short of the good luck of
having teachers work in concert with one another without communicating, or teachers who
work alone, uniform/consistent caregiving requires that staff members have the opportunity
to talk with one another to discuss philosophy and approach on a regular basis. Although
many programs do often choose to make regular meetings a priority, some programs
simply do not have the luxury of time or money to insure that these conversations take
place.
Programs enrolling only children who had experienced difficult life circumstances were
more likely to use relationship practices. Three of the programs in this study were
dedicated to helping children specifically through improving their relationships with
adults. These groups were philosophically bent towards both knowing about and deciding
S. Ritchie, C. Howes / Applied Developmental Psychology 24 (2003) 497–516514
upon specific approaches to support the development of positive relationships. It was
within these programs that practices of looping and primary caregiver assignment were
observed.
These same children, those in programs only for children from difficult life circumstances,
had less acrimonious classrooms and were more likely to be in interactive cluster. Here again
were programs that both developed and implemented philosophies that suggested that to
support positive adult–child relationships, it was important that caregivers be intensely
engaged, warm, responsive, and consistent, and that they initiate positive interactions with the
children. Still these classrooms had the highest proportion of children with insecure
relationships. To overcome the relationship damage that some children bring with them
may take longer than a single year. In the therapeutic preschool, it was found in a separate
study that children with more time in the program (often 2 years) had more secure
relationships with their teachers (Howes & Ritchie, 1997). Children in the group of programs
for children with difficult life circumstances spent less time with a primary caregiver than
children in other programs. Children with avoidant, resistant, and disorganized attachment
organizations can be difficult to spend time with. Logically, the children will avoid and/or
resist the caregiver, who in these cases seem to make the most of the time they do spend with
the child, by relating to the child intensely and sensitively.
Membership in an interactive cluster and spending more time with the primary caregiver
were most important in predicting child–caregiver attachment security. Secondary and
negative predictors were membership in a caregiver direction cluster and being assigned to
a caregiver. Children who spend the day in programs where their primary caregiver is
available and accessible to them and who is intensely and sensitively interacting with them
throughout the day are, not surprisingly, more likely to have a more secure relationship
with their caregivers than those who experience minimal interactions, mostly centered
around management. What is important here is using this information to help programs
think about their priorities. It has become a prevalent practice to have lead teachers move
from classroom to classroom, sometimes spending only a few days a week with a specific
group, and sometimes so overwhelmed by administrative duties, that they spend virtually
no time in the classroom. Teachers who interact minimally with children are perhaps a
product of a childcare system that must hire untrained, and possibly unmotivated personnel
to spend whole days and whole years with children. Children who are assigned to
caregivers, even with the best of intentions, may be assigned to those who are themselves
perhaps unable or unwilling to form positive relationships. Currently some programs are
looking at the notion of observing children over the first several months of childcare and
then using the choice of the child to help make more informed decisions about the
assignments of children.
It is important to consider the impact of multiple caregivers on children. It appears evident
that programs must consider their population and become increasingly aware of the specific
needs of the children they serve to devise their philosophies and approaches. It may be that
traditional notions of relationship practices may not be at the heart of what is important.
Further work needs to be done to more closely define those components that genuinely
support the development and maintenance of secure attachment relationships.
S. Ritchie, C. Howes / Applied Developmental Psychology 24 (2003) 497–516 515
Acknowledgements
This work was supported by grants to the National Center for Early Development and
Learning under the Educational Research and Development Centers Program, PR/Award
Number R307A60004, as administered by the Office of Educational Research and
Improvement, US Dept. of Education. We wish to acknowledge the participation and time
of the programs, teachers, and children who allowed us to record their actions, words, and
practices. We would also like to thank the research assistants and project coordinator for their
hard work. Finally, we would like to acknowledge our colleagues’ insights as we developed
this research.
References
Ainsworth, M. D. S., Blehar, M. C., Waters, E., & Wall, S. (1978). Patterns of attachment: A psychological study
of the strange situation. Hillsdale, NJ: Erlbaum.
Arnett, J. (1989). Issues in the training of caregivers. In J. S. Lande, S. Scarr, & N. Gunzenhauser (Eds.), Caring
for children (pp. 241–255). New Jersey: Erlbaum Publishing.
Birch, S., & Ladd, G. (1997). The teacher–child relationship and children’s early school adjustment. Journal of
School Psychology, 35, 61–79.
Boyce, W. T., Frank, E., Jensen, P. S., Kesler, R. C., Nelson, C. A., & Steinberg, L. (1998). Social context in
developmental psychopathology recommendations for future research from the MacArthur network on psy-
chopathology and development. Development and Psychopathology, 10, 143–164.
Bretherton, I. (1985). Attachment theory: Retrospect and prospect. Monographs of the Society for Research in
Child Development, 50, 3–35.
Cassidy, J., & Berlin, L. (1994). The insecure/ambivalent pattern of attachment. Child Development, 65, 971–991.
Cassidy, J., & Shaver, P. (1999). Handbook of attachment. NY: Guilford.
Chisholm, K. (1998). A three year follow-up of attachment and indiscriminate friendliness in children adopted
from Romanian orphanages. Child Development, 69, 1092–1106.
Goossen, F. A., & van IJzendoorn, M. H. (1990). Quality of infant’s attachment to professional teachers: Relation
to infant–parent attachment and daycare characteristics. Child Development, 61, 832–837.
Harms, T., & Clifford, R. M. (1980). Early childhood environmental rating scale. New York: Teacher’s College
Press.
Howes, C. (1999). Attachment relationships in the context of multiple caregivers. In J. Cassidy, & P. R. Shaver
(Eds.), Handbook of attachment theory and research (pp. 671–687). NY: Guilford.
Howes, C. (2000). Social–emotional classroom climate in child care, child–teacher relationships, and children’s
second grade peer relations. Social Development, 9, 191–204.
Howes, C., & Hamilton, C. E. (1992). Children’s relationships with child care teachers: Stability and concordance
with maternal attachments. Child Development, 63, 879–892.
Howes, C., & Hamilton, C. E. (1993). The changing experience of child care: Changes in teachers and in teacher–
child relationships and children’s social competence with peers. Early Childhood Research Quarterly, 8, 15–32.
Howes, C., Matheson, C. C., & Hamilton, C. E. (1994). Maternal teacher and child care history correlates of
children’s relationships with peers. Child Development, 65, 264.
Howes, C., Phillipsen, L., & Peisner-Feinberg, E. (2000). The consistency and predictability of teacher–child
relationships during the transition to kindergarten. Journal of School Psychology, 87–113.
Howes, C., & Ritchie, S. (1997). Changes in child-teacher relationships in a therapeutic preschool program. Early
Education and Development, 4, 411–422.
Howes, C., & Ritchie, S. (1999). Attachment organizations in children with difficult life circumstances. Develop-
ment and Psychopathology, 11, 254–268.
S. Ritchie, C. Howes / Applied Developmental Psychology 24 (2003) 497–516516
Howes, C., & Ritchie, S. (2002). A matter of trust. N.Y., N.Y.: Teachers College Press.
Howes, C. M., & Smith, E. (1995). Children and their child care teachers: Profiles of relationships. Social
Development, 4, 44–61.
Howes, C., & Stewart, P. (1987). Child’s play with adults, toys, and peers: An examination of family and child-
care influences. Developmental Psychology, 23, 423–430.
Howes, C., Whitebook, M., & Phillips, D. A. (1992). Teacher characteristics and effective teaching in child care.
Findings from the National Child Care Staff Study. Child and Youth Care Forum V, 21, 309–414.
Kontos, S., Howes, C., Shin, M., & Galinsky, E. (1995). Quality in family child care and relative care. New York:
Teachers College Press.
Marcovitch, S., Goldberg, S., Gold, A., Washington, J., Wasson, C., Krekewich, K., & Handley-Derry, M. (1997).
Determinants of behavioral problems in Romanian children adopted in Ontario. International Journal of
Behavioural Development, 20, 17–31.
Marcus, R. F. (1991). The attachments of children in foster care. Genetic, Social and General Psychology
Monograph, 17, 365–394.
Mitchell-Copeland, J., Denham, S., & DeMulder, E. (1997). Q-Sort assessment of child– teacher attachment
relationships and social competence in the preschool. Early Education and Development, 8, 27–39.
Pianta, R. C., Nimetz, S. L., & Bennett, E. (1997). Mother–child relationships; teacher–child relationships; and
adjustment in preschool and kindergarten. Early Childhood Research Quarterly, 12, 263–280.
Raikes, H. (1993). Relationship duration in infant care: Time with a high ability teacher and infant– teacher
attachment. Early Childhood Research Quarterly, 8, 309–325.
Renken, B., Egeland, B., Marvinney, D., Mangelsdorf, S., & Sroufe, L. A. (1989). Early childhood antecedents of
aggression and passive-withdrawal in early elementary school. Journal of Personality, 57, 257–281.
Ritchie, S. (1995). Attachment relationships of substance-exposed children with their caregivers and their teach-
ers. Unpublished doctoral dissertation, University of California, Los Angeles.
Rutter, M. (1999). Clinical implications of attachment concepts: Retrospect and prospect. In L. Atkinson, & K. J.
Zucker (Eds.), Attachment and psychopathology (pp. 17–46). NY: Guilford.
Waters, E. (1990). Appendix A: The attachment q-set (version 3.0). Monographs of the Society for Research on
Child Development, 60, 234–246.
Whitebook, M., Howes, C., & Phillips, D. (1990). Taking on turnover. In house report for Center for the Childcare
Workforce, Washington, D.C.
Whitebook, M., Howes, C., Phillips, D., & Pemberton, C. (1989). Who cares? Child care teachers and the quality
of care in America. Young Children, 46 (4), 5–47.
Whitebook, M., Phillips, D., & Howes, C. (1993). Child Care Employee Project. The national child care staffing
study revisited: Four years in the life of center-based child care. Washington, DC: Center for the Early
Childhood Work Force.
Whitebook, M., Sakai, L., et al. (1997). NAEYC accreditation as a strategy for improving child care quality (p. 20).
Washington, DC: National Center for the Early Childhood Work Force.