The Relation of Preschool Child-Care Quality to Children's Cognitive and Social Developmental Trajectories through Second Grade

by Ellen S. Peisner-Feinberg, Margaret R. Burchinal, Richard M. Clifford, Mary L. Culkin, Carollee Howes
The Relation of Preschool Child-Care Quality to Children's Cognitive and Social Developmental Trajectories through Second Grade
Ellen S. Peisner-Feinberg, Margaret R. Burchinal, Richard M. Clifford, Mary L. Culkin, Carollee Howes
Child Development
Start Page: 
End Page: 
Select license: 
Select License

Child Development, September /October 2001, Volume 72, Number 5, Pages 1534-1553

The Relation of Preschool Child-Care Quality to Children's Cognitive and

Social Developmental Trajectories through Second Grade

Ellen S. Peisner-Feinberg, Margaret R. Burchinal, Richard M. Clifford, May L. Culkin, Carollee Howes,
Sharon Lynn Kagan, and Noreen Yazejian

The cognitive and socioemotional development of 733 children was examined longitudinally from ages 4 to 8

years as a function of the quality of their preschool experiences in community child-care centers, after adjusting

for family selection factors related to child-care quality and development. These results provide evidence that

child-care quality has a modest long-term effect on children's patterns of cognitive and socioemotional develop-

ment at least through kindergarten, and in some cases, through second grade. Differential effects on children's

development were found for two aspects of child-care quality. Observed classroom practices were related to

children's language and academic skills, whereas the closeness of the teacher-child relationship was related

to both cognitive and social skills, with the strongest effects for the latter. Moderating influences of family

characteristics were observed for some outcomes, indicating stronger positive effects of child-care quality for

children from more at-risk backgrounds. These findings contribute further evidence of the long-term influ-

ences of the quality of child-care environments on children's cognitive and social skills through the elementary

school years and are consistent with a bioecological model of development that considers the multiple environ-

mental contexts that the child experiences.


As the use of nonparental child care has risen in the United States, the long-term effects of preschool child-care experiences, especially center-based child care, has been a topic of growing interest to the gen- eral public as well as to the research community. A substantial majority of children now regularly experi- ence center-based child care prior to school entry, with even higher rates for preschool-age children than for infants and toddlers. Recent estimates indi- cate that more than half of all 3- to 5-year-old children in the United States attend child-care centers prior to kindergarten, representing three quarters of the pre- school-age children in out-of-home care (West, Wright, &Hausken, 1995). Given these high usage rates, the quality of these early child-care experiences has be- come an important public policy issue in the United States. Of particular interest for research in this field is the extent to which variations in the quality of these preschool child-care experiences influence children's readiness for and success in school.

A key question for parents, professionals, and pol- icy makers involves the long-term impact of child-care experiences on the transition to and success in school. Examination of children's performance in lunder- garten can provide information about the influences of earlier experiences on the transition to school, whereas examination of children's performance in second to third grade provides information about their school success. The mid-elementary school years (second and third grade) seem to be the point in chil- dren's school careers when academic trajectories be- come more stable, and accordingly, more difficult to change (Alexander &Entwisle, 1988). Therefore, ex- amining children's developmental status at this time point provides a good indication of the long-term ef- fects of child-care experiences on children's school success.

Whereas several studies have explored the longitu- dinal effects of early intervention programs, few have examined the effects for children attending commu- nity child-care programs as they make the transition from preschool to school. A number of studies of early intervention programs for low-income children have found some long-term positive effects on chil- dren's cognitive development and academic achieve- ment at least through the third or fourth grade and sometimes longer, especially for indicators of school success such as retention in grade, special education placement, and intellectual functioning (e.g., Burchinal, Campbell, Bryant, Wasik, &Ramey, 1997; Campbell &Ramey, 1994; Lazar, Darlington, Murray, Royce, & Snipper, 1982; Schweinhart, Barnes, &Weikart, 1993).

Research involving community child-care settings provides a more typical picture of the preschool expe- riences of a broad cross-section of children in this country, and provides necessary information for un-

@2001 by the Society for Research in Child Development, Inc. ~11

rights reserved. 0009-3920/2001/7205-0017

derstanding variations in the developmental patterns of children and their success in the early school years. The literature on the long-term effects of child-care center quality contains mixed results from studies mostly conducted outside the United States: some studies found lasting effects through elementary school age (e.g., Broberg, Wessels, Lamb, & Hwang, 1997; Vandell, Henderson, & Wilson, 1988), although others did not (e.g., Chin-Quee & Scarr, 1994; Deater- Deckard, Pinkerton, & Scarr, 1996).

A few studies addressed this issue longitudinally for preschool-age child care, examining the long-term influences on both cognitive and socioemotional de- velopment, after adjusting for child and family fac- tors known to be linked to both child outcomes and child-care quality (i.e., family selection factors such as socioeconomic status, gender, and ethnicity). In a study of Swedish children, child-care quality prior to school entry related positively to children's math abilities at 8 years of age, whereas verbal ability at age 8 related positively to the amount of center-based child care received before age 3% (Broberg et al., 1997). Interestingly, these relations between child- care experiences and cognitive development were not obtained in studies of this same group of children at earlier ages; other studies have similarly found such "sleeper effects" (e.g., Andersson, 1989). In con- trast, a study of children in Bermuda found that nei- ther amount nor quality of preschool child care re- lated to teacher report card ratings of children's academic achievement and social competence through the first 4 years of elementary school, although child- care quality was related to chldren's concurrent devel- opment while in child care (Chin-Quee & Scarr, 1994).

Other studies of socioemotional functioning have shown little or very modest longitudinal effects of child-care quality when family background factors were controlled. In a study of Canadian children, Jacobs and White (1994) found that better child-care center quality was marginally related to greater com- pliance in kindergarten. In a 4-year longitudinal follow-up of children in the United States, child-care center quality was not related to subsequent parent or teacher ratings of relatively low-frequency indicators of behavioral maladjustment, behavior problems, and social withdrawal (Deater-Deckard et al., 1996).

Although there are relatively few longitudinal studies, the relations between children's concurrent development and center-based child-care experiences have been extensively documented in the child-care literature. A number of studies have established that child-care quality has modest positive effects on chil- dren's cognitive and social functioning, after adjust- ing for child and family background factors. Several

Peisner-Feinberg et al. 1535

studies have found that higher quality care is related to better child outcomes in the short term (e.g., Bryant, Burchinal, Lau, & Sparling, 1994; Dunn, 1993; Howes, 1990; McCartney, Scarr, Phillips, & Grajek, 1985; NICHD Early Child Care Research Network [ECCRN], 1998a, 2000; Phillips, McCartney, & Scarr, 1987; Schliecker, White, &Jacobs, 1991; Whitebook, Howes, & Phillips, 1989). Although most studies have found influences of child-care quality on children's outcomes, a few have found little effect (Clarke-Stewart & Gruber, 1984; Goelman & Pence, 1987; Kontos, 1991; Kontos & Fiene, 1987), perhaps because of re- stricted ranges of quality across the centers in these studies, relatively small sample sizes of centers, or both.

In addition to documenting whether child-care quality is related to later performance, it is useful to determine whether family characteristics moderate the influence of child-care experiences. Of particular interest is the issue of whether child-care quality is more strongly related to developmental outcomes among some groups of children, such as those who may already be at greater risk for less optimal devel- opment. Only a few of the studies that related child- care quality to concurrent child outcomes, and none of the studies that followed children to school, exam- ined this question. A Swedish study found that child and family characteristics moderated the influence of child-care quality at 29 months on measures of socio- emotional functioning at 4 years of age (Hagekull & Bohlin, 1995). Better quality child care had a stronger compensatory influence on ratings of aggressive be- havior for children from lower socioeconomic homes and on ratings of internalizing problems and ego strength and effectance for boys. In some cases, simi- lar results were found for younger children, indicat- ing that the attachment security of infants from higher risk home environments was more affected by the quality of child care (NICHD ECCRN, 1997), al- though subsequent studies did not find such moder- ating effects for other socioemotional and cognitive outcomes with the same sample (NICHD ECCRN, 1998a, 2000).

However, none of the existing studies addressed the issue of long-term effects following school entry for a large sample of children in the United States. The present study was designed to examine the relation of child-care center quality in the preschool years to chil- dren's cognitive and social slulls through second grade, using a relatively large sample of children in the United States. We controlled for child and family background characteristics and, further, examined whether they had any moderating effects on the influ- ences of child-care quality. Two different aspects of child-care quality were considered in the design- classroom practices and teacher-child closeness-to determine whether they were differentially associ- ated with children's development. In addition, none of the previous studies had considered the impact of later school environments when examining the longi- tudinal effects of child care. Therefore, we also ex- plored the extent to which the quality of earlier child- care environments and subsequent kindergarten and second-grade school environments contributed to chil- dren's developmental outcomes in second grade. These additional analyses provided information about the extent to which the effects of child care reflected long-term influences of early experiences versus con- tinuity in earlier and later environments.

Our research design was guided by a bioecological theoretical perspective (Bronfenbrenner & Morris, 1998), in which development is viewed as the interac- tions, or proximal processes, that occur between indi- viduals and their environments. The influence of these processes on development varies as a function of person characteristics (measured here as child background characteristics), environmental contexts (measured as family background characteristics and child-care quality), and the time periods in which these processes occur (measured both concurrently in child care and longitudinally from preschool through second grade).

The three major research questions addressed by this study were (1) Is the level of child-care quality in preschool related to children's patterns of cognitive and social development between the ages of 4 and 8 years? (2) How long term are the influences of child- care quality? and (3) Are there differential effects of child-care quality on developmental outcomes for children at greater risk? We anticipated that better quality care in preschool would be longitudinally re- lated to better cognitive and socioemotional out- comes for children, after adjusting for family selec- tion factors and child characteristics. In addition, we believed that these background factors would display moderating influences, so that child-care quality would be more strongly related to cognitive and so- cial skills for children at greater risk for less optimal developmental outcomes.

Further, none of the previous studies had consid- ered the impact of later school environments when examining the longitudinal effects of child care. There- fore, in an additional set of analyses, we explored the extent to which earlier child-care quality predicts chil- dren's cognitive and social development in second grade, after taking into account subsequent (kinder- garten) and concurrent (second grade) school environ- ments. These additional analyses provided information about the extent to which the effects of child care reflect long-term influences of early experiences versus continuity in earlier and later environments.


The present research is part of the Cost, Quality, and Child Outcomes (CQO) in Child Care Centers Study, a study of center-based community child care and children's longitudinal outcomes in four states in the United States. The data included measurement of preschool child-care quality and longitudinal assess- ments of children's language, cognitive, and socio- emotional functioning over a 5-year period from pre- school through second grade. The data were gathered in five phases: (I) measures of the quality of preschool environments during children's next-to-last year in preschool, (2) assessments of children's developmen- tal outcomes during their next-to-last year in pre- school (at about age 4), (3) assessments of children's developmental outcomes during their last year in preschool (at about age 5), (4) assessments of chil- dren's developmental outcomes during kindergarten (at about age 6), and (5) assessments of children's de- velopmental outcomes during second grade (at about age 8). In addition, brief measures of the quality of children's lundergarten and second-grade school en- vironments were gathered during Phases 4 and 5.


The participants included the children and fami- lies who participated in the longitudinal outcomes component of the CQO study, which examined chil- dren's development from preschool through second grade. An initial sample of 401 child-care centers was randomly selected from four regions in the United States: Los Angeles County in California, the Hart- ford corridor in Connecticut, the Frontal range in Col- orado, and the Piedmont region in North Carolina. These regions were selected because they varied widely in both economic climate and the stringency of state regulations regarding child care (e.g., at the time, mandated staff-child ratios for 4-year-olds var- ied from 1:10 in Connecticut to 1:20 in North Carolina).

The outcomes component of the CQO study in- cluded a subsample of 183 of the selected preschool classrooms in 176 of the 401 centers. Only children with complete data on all of the preschool quality mea- sures and with at least some child assessment data and parent survey data were included in the analysis sample, which included 167 classrooms from 160 cen- ters. The mean number of participating children per classroom was 4.4 (SD = 2.4, range = 1-12). Class- rooms were eligible if they served at least one child in the next-to-last year of preschool (i.e., eligible for kin- dergarten in the second subsequent school year). Center refusal rates for the longitudinal outcomes component were modest; for example, of the eligible centers contacted in North Carolina, 18% refused to participate in this phase of the data collection.

Once a center and classroom teacher agreed to par- ticipate, consent forms were sent home to parents of all potentially eligible children, with up to 12 children randomly selected from each classroom. The four cri- teria for inclusion were (1)children were eligible to enter kindergarten in the second subsequent school year, (2) children were enrolled in the classroom dur- ing the quality observations, (3) parents expected to continue at that center the following year, and (4) the primary language spoken at home was English. Re- fusal rates were low; for example, of those eligible to participate in North Carolina, 7% of the parents or children refused (for further description of the sam- pling procedures, see Peisner-Feinberg & Burchinal, 1997).

The present study examined data over a 5-year pe- riod (i.e., from the 3-year-old year in preschool through second grade), which included 826 children in preschool Year 1,579 in preschool Year 2,451 in kin- dergarten, and 418 in second grade. Due to missing classroom or teacher data, the analysis sample in- cluded fewer children: 733 in preschool Year 1,499 in preschool Year 2,399 in kindergarten, and 345 in sec- ond grade. The average age of children each year was

4.3 (SD = .36), 5.1 (SD = .34), 6.0 (SD = .34), and 8.0 (SD = .34) years old, respectively, at the time of the child assessments. The average age at entry into child care was about 8 months (M= .65 years, SD = 35). The sample was approximately evenly divided by gender (51% male), and about 31% of the children were from diverse ethnic groups. Comparisons to re- cent Census data (Bryson & Casper, 1997) suggest that the sample was similar to U.S. families in general, with about 82% of the children living in two-parent families (versus 72% nationally for families with children under 18 years), average maternal education of 14.25 years (compared with 13 years for women in the United States), and average family income of $47,753 (versus median income of $37,500 for all U.S. households).


Data collection took place in two phases during Year 1, with child-care quality data gathered from classrooms first, and child outcomes data gathered from children, teachers, and parents second. Three subsequent phases of child outcomes data collec- tion were conducted during children's last year in

Peisner-Feinberg et al. 1537

preschool, in kindergarten, and in second grade. In addition, a more limited set of information about elementary school quality was gathered from chil- dren's lundergarten and second-grade classrooms and teachers.

Child-care classroom observations. Teams of six to eight assessors from each of the four sites were trained during a week-long session. Interrater reli- ability visits were conducted at the midpoint of data collection, and included both within- and between- state assessments. Two assessors observed each class- room for approximately 3 to 4 hr in a single visit. One assessor gathered the first three measures, and the other collected the last (described below).

Child-care classroom practices measures. Four obser- vational measures of the process quality of classroom practices were used: (1) classroom environment was measured using the Early Childhood Environment Rating Scale, or ECERS (Harms & Clifford, 1980); (2) teacher sensitivity was measured with the Caregiver Interaction Scale, or CIS (Arnett, 1989); (3) child cen- teredness was measured by the UCLA Early Child- hood Observation Form, or ECOF (Stipek, Daniels, Galuzzo, & Milburn, 1992); and (4) teacher respon- siveness was measured with the Adult Involvement Scale, or AIS (Howes & Stewart, 1987). These instru- ments are widely used and have been shown to relate to other measures of child-care quality and to chil- dren's outcomes (Arnett, 1989; Bryant et al., 1994; Dunn, 1993; Howes & Stewart, 1987; Schliecker et al., 1991; Stipek et al., 1992; Whitebook et al., 1989).

The quality of the classroom environment was measured by the ECERS, which examines the devel- opmental appropriateness of classroom practices, in- cluding the activities, materials, equipment, inter- actions, supervision, organization, and scheduling for children. The ECERS contains 37 items in seven subscale areas: personal care routines (greeting/ departing, meals / snacks, nap / rest, diapering / toileting, personal grooming); furnishings and display (furnishings for routine care, learning, and relaxation; room arrangement); language development (under- standing language, using language, reasoning, infor- mal language); fine and gross motor development (fine motor activities and supervision; gross motor space, equipment, time, and supervision); creative ac- tivities (art, music /movement, blocks, sand /water, dramatic play, schedule, and supervision); social de- velopment (space to be alone, free play, group time, cultural awareness, tone, provisions for exceptional children); and adult needs (adult personal area, adult opportunities, adult meeting area, provisions for par- ents). Each item is rated on a 7-point scale from inad- equate (1)to excellent (7). Psychometric analyses in- dicated that a single total score of the child-related items (1-32) most parsimoniously represented our data, Cronbach's a = .96, with scores of 1.0 to 2.9 defined as poor quality, 3.0 to 4.9 defined as medi- ocre quality, and 5.0 to 7.0 defined as good quality (i.e., in the range of developmentally appropriate practices). Interrater reliability (correlations between raters) ranged from .83 to .98 for the total score, with a median of .94.

Teacher sensitivity was rated using the CIS, with 26 items measuring four subscales: teacher sensi- tivity (e.g., speaks warmly to children), harshness (e.g., seems critical of the children), detachment (e.g., doesn't seem interested in children's activities), and permissiveness (e.g., doesn't reprimand children when they misbehave). Items are rated on a 4-point scale indicating how characteristic they are of the teacher, from not at all (1) to very much (4). Psycho- metric analyses suggested a single factor most parsi- moniously represented our data, a = .93. Interrater reliability ranged from .89 to .98 for each subscale, with median subscale scores from .92 to .95.

The extent to which the teaching style was didactic versus child centered was rated using the ECOF, which includes five subscales: child initiation (e.g., children choose peers for activities), academic em- phasis (e.g., teacher uses commercial materials to teach basic skills with little improvisation), discipline (e.g., teacher uses negative disciplinary techniques), performance pressure (e.g., teacher's emphasis is on outcomelperformance, doing it right like others), and negative evaluation (e.g., rewards given for com- pleting work correctly). This measure consists of 24 items scored on 3-, 4-, and 5-point scales, ranging from didactic (low) to child centered (high). A total mean score was computed after converting all items to the metric of a 5-point scale, a = .92. Interrater re- liability ranged from .81 to .97 for each subscale, with median scores ranging from .91 to .95.

Teacher responsiveness to children was measured using the AIS. For this instrument, 2 children (1 male and 1 female) were randomly selected in each class- room and observed for three observations of 5 min each, with teacher-child interactions coded every 20 s. (No additional information was gathered for these children.) This measure codes the level of the teacher's interactions with the target child on a 6-point scale: ignore (ignoreslunaware of child), routine (routine caretaking but no verbal response), minimal (verbally directs, disciplines, or answers request for help), sim- ple (responds to verbal initiations but does not elabo- rate), elaborative (maintains proximity and acknowl- edges and responds to child's statements), and intense (positive physical interaction, engages in conversa- tionlplay with child). For the present analyses, teacher responsiveness was calculated as the percent- age of time the teacher was at least minimally respon- sive to the target child (the four upper points of the scale). The median interrater reliability using Cohen's K was .92 (range = 33-.96).

The four observational child-care practices mea- sures tended to be highly correlated, with correlations from .74 to .91 among the ECERS classroom environ- ment, CIS teacher sensitivity, and ECOF teaching style, and from .26 to .31 between the AIS teacher re- sponsiveness and the other measures. Therefore, a single composite index of the quality of classroom practices was computed. A principal components analysis of the four measures indicated that one factor accounted for 68% of the total variance, and that sub- sequent factors were unnecessary. The classroom practices quality index was calculated for each class- room based on this principal component and was computed as a z score (M = 0, SD = 1).

Elementary school classroom practices measures. Dif- ferent instruments were used to obtain information about practices in children's lundergarten and second- grade classrooms. Briefer observations were conducted in these classrooms than in the preschool classrooms, with an average duration of 1 hr or less. For the lundergarten year, a shortened version of the ECERS was used, based on items that were readily observable, appropriate to the kindergarten setting, and highly correlated with the overall score from the first year's data. This 5-item version of the ECERS was completed while assessors were in the class- rooms for the child assessments. The total mean item score for the lundergarten short ECERS was used for purposes of analysis, a = 33.

In second grade, a modified version of the Instruc- tional Environment Observation Scales, or IEOS (Secada, 1997) was used. This instrument is designed speci- fically to measure the instructional environment of second- and third-grade classrooms. The IEOS rates a number of characteristics of the classroom environ- ments that students experience using a 5-point scale from 1 (low) to 5 (high), and yields information rele- vant to the domains measured in previous years. This measure consists of six subscales: classroom routines (extent to which transitions are well orchestrated ver- sus major disruptions), classroom climate (extent to which the classroom is a place where students feel safe and respected), cross-disciplinary connections (extent to which activities are connected to multiple subject areas), linkages to life beyond the classroom (extent to which activities are connected to competen- cies or concerns beyond the classroom), social sup- port for student learning (extent to which classroom is characterized by atmosphere of support and re- spect for individual learning capabilities), and stu- dent engagement (extent to which students are en- gaged in the activities). lnterrater reliability using Cohen's K ranged from .50 to .79 for each subscale, with a median of .64. A principal components factor analysis performed on our data yielded two factors. The first factor, general climate, included the class- room routines, classroom climate, social support for student learning, and student engagement subscales, a = 34. The second factor, linkages, included the cross-disciplinary connections and linkages to life be- yond the classroom subscales, a = .79.

Child outcomes assessme~t procedures. Information pertaining to children's cognitive and socioemotional functioning was gathered from individual assessments and from teacher ratings each year, and parents provided information on family characteristics. These child assessment measures have been used in a num- ber of studies examining influences on children's de- velopment during the preschool and elementary school years (e.g., Bryant et al., 1994; Burchinal et al., 1997; Dunn, 1993; Osbor~le, Schulte, & McKinney, 1991; Pianta & Steinberg, 1992; Whitebook et al., 1989). Standard test procedures were used for all in- struments, including establishment of basals and ceil- ings for standardized measures.

Site coordinators from each state were trained in the standard procedures for administering the indi- vidual child assessment instruments in a 3-day ses- sion. They then trained four to seven assessors at each site and monitored them throughout the data collec- tion process. Children were assessed individually for about 30 min at their school in the spring of each year, to measure their language ability and math and read- ing skills. Questionnaires were given to teachers after the assessments to provide ratings of children's class- room behavior and the teacher-child relationship. Parent questionnaires were sent to their homes to gather information on family demographic character- istics. Preaddressed, stamped envelopes for returning questionnaires were included. The return rates were high-from 81% to 98% for parents and 81% to 96% for teachers each year.

Individual child assessments. Individual child assess- ments were conducted using two instruments. Recep- tive language ability was measured using the Peabody Picture Vocabulary Test-Revised, or PPVT-R (Dunn & Dunn, 1981); and preacademic slulls were measured using the Woodcock-Johnson Tests of Achievement-Revised, or WJ-R (Woodcock &Johnson, 1990).

The format of the PPVT-R is appropriate for both younger and older children, and involves having

Peisner-Feinberg et al. 1539

them point to the picture that matches the word spo- ken by the examiner. PPVT-R raw scores were con- verted into standard scores on the basis of age (i.e., M = 100, SD = 15 within norming sample). Based on the original test development, this measure has good split-half (Mdn = .80) and test-retest reliability (Mdn = .82), and correlates highly with other measures of vo- cabulary and moderately with intelligence tests and school achievement.

Children's academic achievement in reading and math each year was measured using two subtests of the WJ-R. The letter-word identification subtest mea- sures reading ability, including association of pictures and symbols and recognition of letters and words. The applied problems subtest measures math skills, including understanding of basic numeracy, compar- isons of differing numbers of items, counting, and solving mathematical problems. Test development in- formation showed high internal consistency for these subtests (Mdn = .92, .91), and moderate correlations with other tests of achievement. Rasch scores were used for the analyses, which allowed for calculation of individual growth over time in the longitudinal data.

Teacher surveys. Teachers rated children's social and cognitive skills each of the 4 years, using the Classroom Behavior Inventory, or CBI (Schaefer, Edgerton, & Aaronson, 1978). In addition, each year teachers rated their relationship with each child using the Student-Teacher Relationship Scale, or STRS (Pianta, 1992), thus providing information on one as- pect of children's contemporaneous classroom expe- riences in preschool, kindergarten, and second grade.

For each participating child, the lead teacher was asked to rate the child's social and cognitive skills using the 42-item research version of the CBI. Items represent 10 scales, and are rated for how well they describe the child, using a 5-point scale from not at all

(1) to very much (5). Factor analysis of the 10 CBI scale scores from the first year resulted in three factors accounting for 76% of the variance. These factors rep- licated other factor analyses of the CBI (Osborne et al., 1991), yielding a cognitive / attention factor, a sociabil- ity factor, and a problem behaviors factor. The cogni- tive/ attention factor consisted of the creativity, verbal intelligence, independence, task orientation, depen- dence (reversed), and distractibility (reversed) scales, a = .84. The sociability factor included the extroversion and introversion (reversed) scales, a = .65. The prob- lem behaviors factor consisted of the distractibility, hostility, and consideration (reversed) scales, a = .77.

A second aspect of the quality of children's experi- ences in preschool, kindergarten, and second grade was measured by teachers' ratings of their relation- ship with each participating child, using the STRS.

This measure contains 30 items rated on a 5-point scale indicating how characteristic they are of the par- ticular teacher-child relationship, from definitely does not apply (1)to definitely applies (5).The origi- nal scoring of the measure sums the items into three factors representing different aspects of the teacher- child relationship: closeness (e.g., I share a warm, af- fectionate relationship with this child), conflict (e.g., this child easily becomes angry at me), and overde- pendency (e.g., this child reacts strongly to separation from me). Only the closeness factor was used in these analyses. Based on the first-year data, the internal consistency within this sample was very good (.86) for the closeness factor.

Parent surveys. Parents were asked to complete surveys each year, from which a variety of demo- graphic information was obtained, including family income; parental education; marital status; age of en- try into child care; and child ethnicity, gender, and birth date.

Data Analysis

Both descriptive and inferential analyses were con- ducted to examine the longitudinal influences of child-care quality on children's development. De- scriptive analyses provided information about the quality of child care and school and about family selection factors relating to child-care quality. Two ad- ditional sets of correlational analyses examined the relations between child-care quality and child out- comes over time and between child-care quality and subsequent school quality in kindergarten and sec- ond grade.

The inferential analyses examined longitudinal patterns of development from ages 4 (next-to-last year of preschool) through 8 (second grade) using hi- erarchical longitudinal analyses. A separate analysis was conducted for each of the six developmental out- comes: assessments of children's receptive language ability (PPVT-R standard score), reading ability (WJ-R letter-word identification Rasch score), and math abil- ity (WJ-R applied problems Rasch score); and teacher ratings of children's cognitive and attention slulls (CBI cognitivelattention factor score), problem be- haviors (CBI problem behaviors factor score), and so- ciability (CBI sociability factor score). In these analy- ses, both individual and group growth curves were estimated. A separate slope and intercept were esti- mated for each child and from these the group growth curves were computed as a function of background and child characteristics.

Patterns of development over time on these out- comes were predicted hierarchically from three sets of predictors. State was entered in all analyses to rep- resent the sampling frame including four sites of data collection. The first set of predictors, the background variables, included mother's education (in years), age of entry into child care (in months), child's ethnicity (White = 1, not White = O), and chld's gender (male = 1, female = 0). To describe patterns of change over time, this set of variables also included child age; child age, squared; and the four interactions between child age and mother's education, age of entry into care, child ethnicity, and child gender. The second set of predictors included two different aspects of pre- school child-care quality: observed classroom prac- tices and ratings of teacher-child closeness. Analysis variables included the classroom practices composite index, the STRS teacher-child closeness rating, and the interactions between child age and each of the two quality measures. The third set of predictors, the moderators of child-care quality, initially included all two-way interactions of the four background vari- ables with the two child-care variables and the three- way interactions adding year, to test for changes in these associations over time. Four a priori contrasts were tested. The first contrast tested whether, as a block, mother's education interacted with either classroom practices or teacher-child closeness in two-way interactions or in three-way interactions adding year. The other three contrasts tested the same block of interactions for age of entry into child care, ethnicity, and gender. When any of these four con- trasts were not significant, then that block of interac- tions was dropped from the analysis model. When these contrasts were significant, then the individual interaction terms were examined so that only the sig- nificant interactions were retained along with associ- ated lower order interaction terms. In addition, to protect against the likelihood of Type I error, child- care quality (including both classroom practices and teacher-child closeness) was first tested as a block, and any main effects and interactions were interpreted only if the block test was significant.

An additional set of analyses examined the prior and contemporaneous effects of classroom quality on children's second-grade outcomes. The same six de- velopmental outcomes were examined, using only the child's scores from the second-grade assessment. Hierarchical regression analyses were conducted. The first block of variables included state as a control vari- able and the background variables (mother's educa- tion, age of entry into child care, ethnicity, and gen- der). The second block added the two child-care quality measures from preschool (the observed class- room practices index and STRS teacher-child closeness rating by the preschool teacher) and any inter- action between background and preschool quality identified in the longitudinal analyses. The third block added the two kindergarten quality measures (the shortened ECERS and STRS closeness rating by the kindergarten teacher), and the fourth block added the three second-grade quality measures (the two IEOS observed practices factor scores and STRS close- ness rating by the second-grade teacher).


Descriptive Analyses

Child-care quality. The first set of descriptive analy- ses examined the individual components of the qual- ity of child care (for full details, see Peisner-Feinberg & Burchinal, 1997). Mean scores on each of the obser- vational measures of classroom practices and the teacher-child relationship are included in Table 1. In general, classroom practices tended to be in the me- dium range of quality. The mean ECERS global qual- ity score of 4.38 was well within the medium quality range (i.e., between 3.0 and 5.0), suggesting that chil- dren were attending centers where their routine care needs were likely to be met, but where there were more limited opportunities for learning activities, in- dividual attention, or language stimulation. In gen- eral, the teachers were moderately sensitive, as indi- cated by the CIS mean of 3.01 (range = 1-4). The AIS data indicated that, on average, teachers were ob- served as being at least minimally responsive to the children in their class about 31% of the time. The teaching style observed in the classrooms based on the ECOF was slightly closer to a child-centered than didactic approach. In contrast, teachers reported fairly close relationships to the children, with an aver- age score of 4.17 on the STRS (range = 1-5).

School quality. Second, classroom practices and the closeness of teacher-child relationships in lundergar- ten and second grade were examined descriptively. Mean scores on each measure are presented in Table 1. The same measure of teacher-child closeness was used each year in child care and elementary school, while different measures of classroom practices were used each year. In general, teachers reported fairly close relationships with children each year based on the STRS, with similar average scores each year, al- though the scores were slightly lower in second grade than in preschool or kindergarten. The quality of class- room practices in kindergarten was similar to that in preschool, based on a shortened form of the ECERS used in preschool. Scores on the IEOS indicated that second-grade classrooms tended to be fairly high in general classroom climate, but fairly low in linkages across disciplines and to life beyond the classroom.

Peisner-Feinberg et al. 1541

Table 1 Mean Scores for Quality and Child Outcome Measures over Time


Year 1preschool child care

Teacher-child closeness (STRS)
Classroom practices index

ECERS total
CIS total
ECOF total
AIS % responsive

PPVT-R language
WJ-R reading
WJ-R math
CBI cognitive/ attention
CBI problem behaviors
CBI sociability

Year 2 preschool child care

PPVT-R language
WJ-R reading
WJ-R math
CBI cognitive/ attention
CBI problem behaviors
CBI sociability


Teacher-child closeness (STRS)
Shortened ECERS

PPVT-R language
WJ-R reading
WJ-R math
CBI cognitive / attention
CBI problem behaviors
CBI sociability

Second Grade

Teacher-child closeness (STRS)
IEOS classroom climate
IEOS linkages

PPVT-R language
WJ-R reading
WJ-R math
CBI cognitive/ attention
CBI problem behaviors
CBI sociability

Note: STRS = Student-Teacher Relationship Scale; ECERS = Early Childhood Environment Rating Scale; CIS = Caregiver Interaction Scale; ECOF = Early Childhood Observation Form; AIS = Adult In- volvement Scale; PPVT-R = Peabody Picture Vocabulary Test-Revised; WJ-R = Woodcock-Johnson-Revised; CBI = Classroom Behavior Inventory; IEOS = Instructional Environment Observation Scales.

Family selection factors. Next, the associations be- tween child-care selection and family background characteristics were examined (for more details, see Peisner-Feinberg & Burchinal, 1997). Pearson product- moment correlations were computed among child and family background factors, the observational measures of classroom practices, and the ratings of closeness of the teacher-child relationship. Informa- tion was included about the mother's years of educa- tion; the family's reported monthly income; and the child's ethnicity, gender, and age of entry into child care. As shown in Table 2, children from more advan- taged families were more likely to enter child care at a later age and to experience higher quality child care in the preschool years. Better classroom practices and closer teacher-child relationships were associated with higher maternal education and higher family in- come in child care, but not in elementary school. There was some relation of ethnicity to better quality practices in kindergarten and second grade but not in child care, with White children tending to be in higher quality care than children of color. Teacher-

child relationships were associated with gender each year, with teachers reporting closer relationships with females than with males. From these analyses, mother's education and child ethnicity, gender, and age of entry into child care were selected to represent the family selection factors for subsequent analyses. Family income was not included because it was highly correlated with maternal education and was missing for a portion of the sample.

Correlations among preschool child care, kindergarten, and second-grade quality. Pearson product-moment cor- relations were computed among child-care quality and elementary school (kindergarten and second grade) quality, to examine the relations between earlier and later environments. The two aspects of quality- classroom practices and teacher-child relationships- were examined separately. As seen in Table 2, there was little association between the quality of class- room practices in earlier and later settings. Correlations among child care, lundergarten, and second-grade practices ranged from .06 to .15. There was a slightly stronger association for the closeness of teacher-child relationships over time, although these relations were still modest, ranging from .20 to .30 among preschool, kindergarten, and second-grade ratings.

Correlations of child care and school quality with child outcomes. The associations between children's devel- opmental outcomes and the quality of child care and school experiences over time were examined. Pearson product-moment correlations of children's longitudi- nal outcomes with the teachers' ratings of the close- ness of their relationships with children and the ob- served quality of classroom practices were computed from preschool through second grade. As shown in Table 3, classroom practices tended to relate to chil- dren's language and math skills over time, whereas

Table 2 Correlations among Family Background, Child-Care Quality, and School Quality Measures
Child-Care Kindergarten Second-Grade
Family Background   Quality Quality Quality
2 3 4 5 6 7 8 9 10 11 12

Family background

  1. Maternal education
  2. Ethnicity (White = 1)
  3. Gender (male = 1)
  4. Income
    1. Age of entry Child care quality
    2. 6. Teacher-child closeness (STRS)
  1. Classroom practices index Kindergarten quality

8. Teacher-child closeness (STRS)

9. Shortened ECERS Second-grade quality

  1. Teacher-child closeness (STRS)
  2. IEOS linkages
  3. IEOS classroom climate

Note: STRS = Student-Teacher Relationship Scale; ECERS = Early Childhood Environment Rating Scale; IEOS = Instructional Environ-
ment Observation Scales.
*p < .05; **p < .01; ***p < ,001.

Table 3 Correlations of Child Outcomes with Family Background, and Child-Care Quality, and School Quality Measures

Child-Care Quality Kindergarten Quality Second-Grade Quality
Teacher-Child Classroom Teacher-Child Teacher-Child IEOS
Maternal Closeness Practices Closeness Shortened Closeness Classroom
Outcome Education (STRS) Index (STRS) ECERS (STRS) Climate IEOS Linkages
Year 1preschool        
PPVT-R language        
WJ-R reading        
WJ-R math        
CBI cognitive / attention        
CBI problem behaviors        
CBI sociability        
Year 2 preschool        
PPVT-R language        
WJ-R reading        
WJ-R math        
CBI cognitive / attention        
CBI problem behaviors        
CBI sociability        
Year 3 kindergarten        
PPVT-R language        
WJ-R reading        
WJ-R math        
CBI cognitive / attention        
CBI problem behaviors        
CBI sociability        
Year 5 second grade        
PPVT-R language        
WJ-R reading        
WJ-R math        
CBI cognitive / attention        
CBI problem behaviors        

CBI sociability

Note: STRS = Student-Teacher Relationship Scale; ECERS = Early Childhood Environment Rating Scale; IEOS = Instructional Environment Observation Scales; PPVT-R = Peabody Picture Vocabulary Test-Revised; WJ-R = Woodcock-Johnson-Revised; CBI = Classroom Behavior Inventory.

teacher-child closeness tended to relate to children's social and behavioral skills over time. The composite index of child-care classroom practices during the first year of the study showed modest correlations with children's language and math skills over time. Child-care classroom practices were also related to reading skills at the first assessment, to cognitive and attention skills during the preschool years, and to so- ciability at the first assessment. Child-care teachers' ratings of the closeness of their relationships with children were modestly related to language and math skills at most assessment ages, and were moderately to modestly related to children's social skills over time. Similarly, classroom practices in lundergarten were modestly related to children's kindergarten lan- guage and math skills, and teacher-child closeness in both kindergarten and second grade was moderately to modestly related to children's social skills in ele- mentary school. In contrast, measures of second-grade classroom practices were not related to chil- dren's cognitive or social skills in second grade. In comparison, maternal education was moderately re- lated to children's language and cognitive skills over time, and modestly related to cognitive and attention skills and problem behaviors over time.

Longitudinal Analyses

The inferential analyses examined the association between children's preschool experiences and pat- terns of language, cognitive, and social development from 4 to 8 years of age. State was included as a con- trol variable, and background variables of maternal education, age of entry into child care, ethnicity, and gender were entered to adjust for selection effects and to test for moderating effects on the association between child-care quality and children's developmen- tal patterns. The longitudinal regressions were fit hi- erarchically, entering state, age, and background vari- ables first; the two measures of child-care quality and interactions between child care quality and age sec- ond; and the interactions between background and child-care quality last. Results are shown in Table 4, listing the unstandardized regression coefficients from the final model and group-specific slopes when inter- actions were significant.

Receptive language ability. Analysis of children's language skills showed marked increases over time relative to the PPVT-R norming population. Back- ground characteristics, F(6,655) = 30.74, p < .001, and child-care quality, F(4, 655) = 9.31, p < .001, signifi- cantly contributed to describing patterns of change over time. After adjusting for the background vari- ables, children attending child care with higher qual- ity classroom practices tended to have higher language scores, F(1, 655) = 10.69, p < .002, but the magnitude of this association declined over time, F(l, 655) = 7.99, p < .005. This trend can be seen by examining the co- efficients describing the association for Age X Classroom Practices in Table 4. Figure 1shows this relation for the 25th and 75th percentiles of quality based on the classroom practices index. The association be- tween language scores and quality of classroom prac- tices was strongest when assessed contemporaneously in Year 1, B = 1.85, SE = .38, and declined over time. The quality of child-care practices was a significant predictor of children's language ability through kin- dergarten, but not in second grade. This finding is il- lustrated by examining the coefficients for child-care practices separately for each age. These estimated slopes are listed in Table 4 under the Age X Classroom Practices interaction coefficient, and show a clear decline in magnitude across age.

In addition, children whose preschool teachers rated their relationship as closer tended to have higher language scores over time, F(l, 655) = 7.52, p < .007. Unlike the classroom practices quality index, the association between preschool teachers' ratings of closeness with children and language scores did not significantly decline with age. There were no moder- ating effects for child and family background charac- teristics found for either aspect of child-care quality.

Reading ability. On the basis of the Rasch scores, which describe development using a developmental age score rather than a standard score, children's reading slulls increased between 4 and 8 years of age, as expected. Rasch scores for the letter-word identifi- cation subtest of the WJ-R over time were signifi- cantly related to the background variables, F(6,654) = 11.87, p < .001, but were not significantly related to the child-care measures.

Math ability. Children's math slulls also increased greatly over time, based on the Rasch scores for the applied problems subtest of the WJ-R. Developmen- tal patterns were related to the blocks of background variables, F(6, 635) = 16.19, p < .001; child-care vari- ables, F(4, 635) = 3.56, p < .007; and interactions among background and child-care variables, F(1,635) = 5.53, p < .02. Children tended to have slightly better math skills if they attended a child-care classroom with higher quality practices, F(1,635) = 6.38, p < .02, and if their child-care teachers rated their relationship as closer, F(1, 635) = 4.91, p < .03. There was a mod- erating effect for maternal education on the associa- tion between classroom practices and children's development, but not on the association for teacher- child closeness. The significant interaction between classroom practices and maternal education indi-

Peisner-Feinberg et al. 1545

Table 4 Results from Hierarchical Linear Model Regression Analyses Predicting Children's Longitudinal Outcomes PPVT-R CBI Cognitive1 CBI Problem

  Language WJ-R Reading WJ-R Math Attention Behaviors CBI Sociability
State block            
Age block            
Age Age, squared            
Background block            
Maternal education            
Gender (male = 1)            
Ethnicity (White = 1)            
Age of entry            
Age X Education            
Age X Gender            
Age X Ethnicity            
Age X Entry            
Child-care quality block            
Classroom practices index            
Teacher-child closeness (STRS)            
Age X Practices            
Practices slope, age 4 --            
Practices slope, age 5            
Practices slope, age 6            
Practices slope, age 8            
Age X Closeness            
Closeness slope, age 4            
Closeness slope, age 5            
Closeness slope, age 6            
Closeness slope, age 8            
Quality X Family Interaction block            
Education X Practices            
Practices slope, education = 12            
Practices slope, education = 16            
Education X Closeness            
Education X Age x Closeness            
Closeness slope, age 4, education = 12            
Closeness slope, age 5, education = 12            
Closeness slope, age 6, education = 12 --            
Closeness slope, age 8, education = 12            
Closeness slope, age 4, education = 16            
Closeness slope, age 5, education = 16            
Closeness slope, age 6, education = 16 --            
Closeness slope, age 8, education = 16            

Note: PPVT-R = Peabody Picture Vocabulary Test-Revised; WJ-R = Woodcock-Johnson-Revised; CBI = Classroom Behavior Inventory. *p< .05; ** p < .01; *** p < ,001;ns = nonsignificant.

cated that better quality child care had a stronger as- whose mothers had only a high school degree, and sociation with children's math skills for children nonsignificant for children whose mothers had a col- whose mothers had less education, F(1, 635) = 5.53, lege degree. Figure 2 displays these same develop- p < .02. To illustrate this finding, the coefficients for mental patterns for Math Skills x Maternal Education the classroom practices index were estimated sepa- and Classroom Quality, estimated for children whose rately for children whose mothers had a high school mothers had high school and college degrees for the degree and a college degree as their highest level of 25th and 75th percentiles of quality based on the educational attainment. These estimated slopes are classroom practices index. There were no interaction shown in Table 4 under the coefficient for the Mater- effects with year, indicating that the main effect asso- nal Education X Classroom Practices interaction. The ciations for both aspects of child-care quality re-coefficient is significant and positive for children mained consistent over time.


-High Quality Low Quality

Age (Years)

Figure 1 Children's language skills over time for low-quality (25th percentile) and high-quality (75th percentile) child-care classroom practices. PPVT-R = Peabody Picture Vocabulary Test-Revised.

Cognitive and attention skills. Teacher ratings of children's cognitive and attention skills on the CBI increased and then decreased over time, and individual patterns of change were related to the background, F(6, 496) = 6.50, p < .001, and child-care blocks, F(4,496) = 34.73,p < .001.There was a positive association between child-care teacher-child closeness and cognitive/attention scores, F(1, 496) = 54.82, p <

$ 400 1

+Low Quality, High School Education ..o .-Low Quality, College Education +High Quality, High School Education ..*..High Quality, College Education

4 5 6 7 8

Age (Years)

Figure 2 Children's math skills over time for low-quality (25th percentile) and high-quality (75th percentile) child-care classroom practices by maternal education (high school and college education). WJ-R = Woodcock-Johnson-Revised.

.001, but this associationdeclined over time, F(1,496)= 15.65, p < .001. Table 4 shows the regression coefficients describing the association between teacher-child closeness in child care and cognitive / attention ratings over the 5-year period. Although there was a decline in the influence of teacher-child closeness as children got older, this association was still statistically significant through second grade. There were no moderating effects for child and family background characteristics on this relation.

Problem behaviors. Teacher ratings of children's problem behaviors on the CBI declined slightly over time, and patterns of change were related to the blocks of background variables, F(6,496) = 6.94, p < .001; child-care variables, F(4,496) = 16.43, p < .001; and interactions among child-care and background variables, F(2, 496) = 4.98, p < .008. Children with closer relationships to their teachers in child care were rated lower on problem behaviors through second grade, F(1, 496) = 37.04, p < .001. The magnitude of this association declined over time, F(1,496) = 6.33, p < .02, and maternal education had a moderating effect, so that there was less decline over time for children whose mothers had less education, F(1, 496) = 9.77, p < .002. Table 4 shows the regression coefficients describing the association between childcare teacher-child closeness and problem behaviors ratings for each age, estimated for children whose mothers had high school and college degrees (listed under the Maternal Education X Age x Teacher-Child Closeness interaction).Figure 3 displays these

-=-Low Closeness, High School Education --o -. Low Closeness, College Education +High Closeness, High School Education .. .. High Closeness, College Education

Age (Years)

Figure 3 Children's problem behaviors over time for low-quality (25th percentile) and high-quality (75th percentile) child-care teacher-child relationships by maternal education (high school and college education). CBI = Classroom Behavior Inventory.

same developmental patterns for Problem Behav- iors X Maternal Education for the 25th and 75th per- centiles of quality based on ratings of teacher-child closeness. Whereas child-care teacher-child closeness was a consistent and significant predictor of problem behaviors for children whose mothers had a high school education, it became less predictive over time for children whose mothers had more ed- ucation. In addition, there was a significant interac- tion between classroom practices and age, although no main effect for practices, indicating that higher quality practices were significantly associated with fewer problem behaviors in Year 1, but this associa- tion declined in subsequent years, F(1, 496) = 4.31, p < .04.

Sociability. Teacher ratings of children's sociability on the CBI increased and then decreased slightly over time, and individual patterns of change were re- lated to the child-care block, F(4, 496) = 60.25, p < .001. Ratings of greater teacher-child closeness in preschool were related to higher ratings of sociabil- ity, F(1, 496) = 81.76, p < .001, but the magnitude of this association declined over time, F(1,496) = 33.36, p < .001. Table 4 shows the regression coefficients describing the association between child-care teacher- child closeness and sociability over time. Teacher- child closeness in child care was a significant predic-

Peisner-Feinberg et al. 1547

tor of children's sociability through kindergarten, but was not significant in second grade. There were no moderating effects for child and family background characteristics.

Second-Grade Outcomes in Relation to Background, Child Care, and Second-Grade Classroom Experiences

An additional set of analyses examined the data from the last assessment, second grade, to determine the extent to which child-care measures predicted outcomes in second grade and whether these associa- tions held when adjusting for classroom experiences in lundergarten and contemporaneously in second grade. Hierarchical analyses were conducted, enter- ing state and background variables first; child-care quality variables, including any significant interac- tions between background and child-care measures obtained in the longitudinal analyses, second; kinder- garten quality variables thrd; and second-grade qual- ity variables last. Results are shown in Table 5, which lists the standardized regression coefficients for each variable from the final model. It should be noted that these analyses included fewer children than the pre- vious analyses due to missing classroom observation data in kindergarten or second grade.

Table 5 Results from Hierarchical Regression Analyses Predicting Children's Second-Grade Outcomes

  PPVT-R Language WJ-R Reading WJ-R Math CBI Cognitive/ Attention CBI Problem Behaviors CBI Sociability
Total R2            
Background Maternal education Ethnicity (White = 1) Gender (male = 1) Age of entry            
Preschool child care            
Classroom practices index Teacher-child closeness (STRS) Interactions Education X Practices Education x Closeness            
Kindergarten Shortened ECERS Teacher-child closeness (STRS)            
Second grade IEOS classroom climate IEOS linkages            

Teacher-child closeness (STRS)

Note: Standardized regression coefficients are listed. State was entered as a covariate. PPVT-R = Peabody Picture Vocabulary Test-
Revised; WJ-R = Woodcock-Johnson-Revised; CBI = Classroom Behavior Inventory; STRS = Student-Teacher Relationship Scale; ECERS =
Early Childhood Environment Rating Scale; IEOS = Instructional Environment Observation Scales.
*p < .05; **p < .01; ***p < ,001;ns = nonsignificant.

The results provide some evidence that preschool and attention slulls and sociability and lower ratings child-care quality is related to children's skills in sec- of problem behaviors. Closer kindergarten teacher- ond grade, even after adjusting for background char- child relationships were also associated with higher acteristics and kindergarten and second-grade class- ratings of sociability in second grade. In contrast, room practices and teacher-child relationships. Higher there were no relations between observed kindergar- math scores were associated with better quality child- ten classroom practices and children's cognitive or care classroom practices, F(2, 153) = 3.27, p < .04. In socioemotional functioning. There was one associa- addition, teachers reported fewer problem behaviors tion between observed second-grade classroom prac- in second grade for children who had closer relation- tices and children's outcomes, with higher reading ships with their preschool teachers, F(3, 152) = 3.50, scores for children in classrooms rated lower in class- p < .02. However, this association was moderated by room climate. matemal education, indicating a stronger effect for chil- dren whose mothers had less education.

Effect Sizes

In addition, second-grade teacher-child relationships predicted children's second-grade social skills, Effect sizes were computed using partial correla- after controlling for all other variables, although these tions to provide a context for interpreting the findings associations may be partially explained by common from the longitudinal regression analyses. The data source variance, because the teacher provided both from each year were examined separately for each as- sets of ratings. Closer teacher-child relationships in pect of child-care quality. The partial correlations in- second grade were associated with better ratings of cluded the covariates of state, maternal education, children's social slulls-higher ratings of cognitive child ethnicity, child gender, age of entry into child

Table 6 Effect Sizes of Child-Care Quality and Maternal Education for Children's Longitudinal Outcomes

Child Care Kindergarten Second Grade Outcome (N = 628-633) (N = 318-378) (N = 256-330)

PPVT-R language Classroom practices index Teacher-child closeness (STRS) Maternal education

WJ-R reading Classroom practices index Teacher-child closeness (STRS) Maternal education

WJ-R math Classroom practices index Teacher-child closeness (STRS) Maternal education

CBI cognitive / attention Classroom practices index Teacher-child closeness (STRS) Maternal education

CBI problem behaviors classroom practices index Teacher-child closeness (STRS) Maternal education

CBI sociability Classroom practices index Teacher-child closeness (STRS) Maternal education

Note: Partial correlations are listed. State, child ethnicity, child gender, and age of entry into child care were included as covariates. In addition. for each child-care measure, maternal education and the other child-care quality measures were covaried. For matemal education, both child-care quality mea- sures were covaried. PPVT-R = Peabody Picture Vocabulary Test-Revised; STRS = Student-Teacher Relationship Scale; WJ-R = Woodcock-Johnson-Revised; CBI = Classroom Behavior Inventory.

care, and the other child-care quality measure (i.e., classroom practices index when computing effect sizes for teacher-child closeness and teacher-child close- ness rating when computing effect sizes for classroom practices). In addition, the effect sizes of maternal edu- cation were computed as a contrast for the effect sizes of child-care quality, under the assumption that family characteristics should provide the strongest prediction of child outcomes. In those analyses, both measures of child-care quality were included as covariates, as well as state, ethnicity, gender, and age of entry.

The results from these analyses are shown in Table

6. The first row under each outcome displays the ef- fect sizes for the child-care classroom practices index; the second row, for child-care teacher-child closeness ratings; and the third row, for maternal education. Only effect sizes corresponding to significant results from the longitudinal regression analyses were inter- preted. In addition, effect sizes for teacher-child closeness associated with ratings of social slulls from Year 1 were not interpreted because the resulting large effect sizes were likely influenced by the child- care teacher being the common informant for both types of data. Cohen (1988) recommends that effect sizes around r = .10 be regarded as modest, around r = .30 as moderate, and around r 2 .50 as large. Using those criteria, there were modest effect sizes for child-care practices on preschool and kindergarten language ability, and on preschool through second- grade math skills. Teacher-child closeness showed modest effect sizes for language ability and math skills over time. There were slightly stronger effect sizes for teacher-child closeness for teachers' ratings of children's cognitive and attention skills, problem behaviors, and sociability over time.

Maternal education displayed modest to moderate effect sizes for language and math skills over time, and more modest effect sizes for classroom behaviors of cognitive and attention slulls and problem behaviors. Comparisons of the effect sizes for preschool child-care quality and for maternal education indicate that both were associated with five of the six outcomes, and that the effect sizes for child-care quality were about one fourth to three fourths of those for maternal education. The effects of maternal education, however, reflected both genetic and environmental influences. This con- found was greatly reduced for child-care effects, be- cause genetic influences related to family selection fac- tors were controlled statistically.


As the labor force participation rates for women with young children have increased, so has the proportion

Peisner-Feinberg et al. 1549

of children attending child-care centers during the preschool years. The majority of preschoolers regu- larly receive some form of nonparental care, with more than half of 3- to 5-year-old children in the United States attending child-care centers (West et al., 1995). Whereas many studies have documented both variations in the quality of these child-care en- vironments and associations between child-care quality and children's cognitive and socioemotional functioning, few have examined the long-term im- pact of these early experiences into the school years. Utilizing a longitudinal design, the present study provided evidence for the long-term effects of child- care quality on children's language, cognitive, and social skills through early elementary school. These findings have implications for both the public policy arena and early school practices in terms of enhanc- ing children's readiness for and subsequent success in school.

In most domains of development that were exam- ined, there was evidence for a continued influence of preschool child-care quality on children's skills through elementary school. Longitudinal effects were found for receptive language ability, math ability, cognitive and attention skills, problem behaviors, and sociability, indicating that children who had better quality preschool experiences were more ad- vanced in their development over a 5-year period. As expected, there was some evidence indicating a diminishing influence of child-care quality over time. Although the effects of child-care quality were modest, especially by second grade, they are con- sistent with some of the other longitudinal studies that have documented long-term effects following school entry (e.g., Broberg et al., 1997; Jacobs & White, 1994).

These findings of long-term effects in the present study, after adjusting for child and family characteris- tics, are noteworthy for several reasons. First, these results were obtained on a varied sample of commu- nity child-care programs in four regions of the United States. Whereas long-term effects of preschool experi- ences have been established for a variety of early in- tervention programs (e.g., Campbell & Ramey, 1994; Lazar et al., 1982; Schweinhart et al., 1993), much less is known about the long-term effects across the range of child-care experiences and for children from a range of backgrounds. The children and families in our study were enrolled in typical community child- care programs, which represented a range of quality. Furthermore, the families themselves comprised a va- riety of backgrounds. A generally positive influence of child-care quality was found for children from di- verse backgrounds, although in some cases, stronger influences of child care were found for children at greater risk, who are presumably more similar to the participants in the early intervention programs. Spe- cifically, maternal education showed such a moderat- ing effect for children's math skills and problem be- haviors, indicating that child-care quality had a stronger positive effect on development for children from families with less highly educated mothers. These results are consistent with other studies that, at least within the domain of socioemotional develop- ment, have found stronger effects for child-care qual- ity for children from less advantaged backgrounds (e.g., Hagekull & Bohlin, 1995; NICHD ECCRN, 1997). These findings extend the results of both the early intervention and the child-care research, and suggest that child-care experiences, both concurrently and over the long term, have an even greater influence on some aspects of both cognitive and social development for children at greater risk.

Second, for all but one of the outcomes studied, child-care quality continued to exhibit an influence at least through kindergarten, and in some cases, through second grade, 4 years later. These long-term effects covered a substantial portion of time, representing about half of these children's lives at that point. The finding of consistent kindergarten effects provides evidence that the quality of children's preschool ex- periences is an important predictor of their readiness for school. Children who experienced higher quality preschool child-care settings demonstrated better skills in their first year of school. Other studies sug- gest that these early positive school experiences may enable children to establish a more positive academic trajectory for their future school careers (e.g., Alex- ander & Entwisle, 1988). Further evidence of the last- ing nature of these effects, at least through second grade, was found for some aspects of both cognitive and social development, namely math skills, cogni- tive and attention skills, and problem behaviors.

Third, comparisons with the influence of family characteristics indicate the relative contribution of child-care quality to different aspects of children's de- velopment. As with several other studies (e.g., Chin- Quee & Scarr, 1994; Deater-Deckard et al., 1996; NICHD ECCRN, 1998a, 2000), this study found stron- ger relations with child outcomes for family charac- teristics such as maternal education than for child- care quality. It is to be expected that characteristics of the home would significantly predict children's de- velopment, as the family provides the primary envi- ronment for the child as well as the one consistent environment over time. Other studies have shown that family characteristics, especially indicators of en- vironmental risk such as maternal education and eth- nicity, are consistent predictors of children's cognitive development over time (e.g., Sameroff, Seifer, Bald- win, & Baldwin, 1993). Moreover, recent evidence suggests that the influence of family factors on the prediction of children's outcomes is not diminished by participation in full-time hld care (NICHD ECCRN, 199813). Child-care experiences, in contrast to the fam- ily environment, provide a secondary environment for only part of this time, being replaced by other school environments as well as other preschool or school-age child-care environments as children grow older. Examination of the effect sizes for family and child-care measures in the present study suggests that maternal education was a somewhat stronger predic- tor of children's language and cognitive skills over time, whereas child-care quality exhibited an effect size about one quarter to three quarters as large. In contrast, the closeness of the child's relationship with the preschool teacher tended to be a similar or even stronger predictor of children's behavioral and so- cial skills in the classroom compared with maternal education.

Overall, our findings support a bioecological per- spective, wherein children's development is the prod- uct of proximal processes or interactions that occur within the multiple environments in which they live (Bronfenbrenner & Morris, 1998). Whereas the family is the primary environment, for children in full-time child care, the child-care environment is likely to be the second most frequent environment in which they spend time. Our findings showed that the quality of these secondary environments also plays a role in determining young children's developmental outcomes, not only contemporaneously but also longitudinally through the early school years. The findings of the di- minishing influence of child-care quality over time would be predicted from this model, given that child care becomes an increasingly remote environmental context. However, our findings of the moderating in- fluences of maternal education for some of the out- comes studied suggest that there may be differential pathways that lead to these developmental outcomes. In other words, the long-term influence of the proxi- mal processes occurring within the child-care context may differ on the basis of the characteristics of other environmental contexts existing simultaneously during that point in development.

Fourth, given the nature of longitudinal designs, children have a number of intervening experiences in out-of-home environments over the specific time pe- riod. A more complete rendering of the bioecological approach takes into account the variety of environ- mental contexts experienced by children over this time period, as well as the degree of continuity across these environments over time. From their next-to-last year in preschool through second grade (ages 4 to 8), children have experienced a variety of care and edu- cation settings, including the transitions to and expe- riences in lundergarten, first, and second grade. Even high-quality preschool child-care experiences cannot be expected to serve as an inoculation, protecting children from the potential effects of subsequent neg- ative experiences and superceding the influences of subsequent positive experiences. However, the find- ings of longitudinal child-care effects despite the va- riety of subsequent experiences suggest the long-term importance of early experiences on children's devel- opment. Further evidence for this conclusion was supported by the additional analyses that examined second-grade outcomes as a function of preschool, kin- dergarten, and second-grade environments. Although somewhat reduced, a similar pattern of child-care ef- fects was found, even when controlling for the subse- quent experiences in lundergarten and second grade.

Although the measures of elementary school class- room practices were not as comprehensive as the child-care measures, there is little indication in our data that children who initially experience higher qual- ity environments in child care remain in higher quality environments once they enter school. Correlations among the observed practices measures for child care, kindergarten, and second grade ranged from .06 to .15, and suggest that there was little similarity in the quality of classroom practices over time. There was a somewhat stronger indication of similarity in the na- ture of teacher-child relationships over time, however. The correlations among ratings of teacher-child close- ness over time ranged from .20to .30, with the strongest correlation between kindergarten and second grade.

It may be that children are learning ways of inter- acting and establishing relationships with nonparen- tal caregivers during their preschool years that carry over, at least to some extent, into the relationships they form with their teachers in elementary school. Such an interpretation is consistent with a bioecolog- ical perspective (Bronfenbrenner & Morris, 1998), which acknowledges that person characteristics of the child that develop within a particular environ- mental context (i.e., child care) influence the proximal processes that occur in subsequent environmental contexts over time (i.e., school). For example, it may be easier to establish closer relationships with some children because they engage in more facilitating be- haviors. Similarly, children may learn patterns of behavior that are likely to elicit certain patterns of re- sponses on the part of caregivers, so that continuity in these behaviors over time would enhance the likeli- hood that earlier and later caregivers respond in a

Peisner-Feinberg et al. 1551

similar manner (e.g., Howes, 1999; Pianta & Steinberg, 1992; Pianta, Steinberg, & Rollins, 1995). Al- though there may be adaptations for the school set- ting as opposed to the child-care setting that result in greater consistency from kindergarten to second grade than from child care to kindergarten, there still seems to be a modest effect for such early experiences that is maintained over time.

Fifth, this study examined two different aspects of the child-care experience-the quality of observed classroom practices and the closeness of the teacher- child relationship. The results demonstrated that these two aspects were differentially associated with children's patterns of development. Observational measures of classroom practices tended to relate more strongly to children's cognitive skills, whereas ratings of the teacher-child relationship were related to both cognitive and social slulls, although even more strongly to the latter. These findings suggest that child-care experiences influence cognitive and socioemotional development somewhat differently. Whereas actual classroom practices, including the materials, activities, and interactions, contributed most strongly to chil- dren's language and cognitive development both con- currently and over time, the establishment of closer re- lationships with caregivers provided an additional contribution. Further, these early relationships with caregivers were the strongest longitudinal predictors of children's social skills. It may be that children with positive early experiences with nonparental caregivers learn a pattern of interacting that facilitates their relationshps with future caregivers, as well as their ability to utilize the learning experiences provided in these environments to further their development.

This study's findings have direct relevance to on- going public policy debates regarding the care and education of young children during both the pre- school and early school years. Our results indicate that child-care quality continues to predict children's development during the elementary school years for a variety of cognitive and socioemotional outcomes. Furthermore, our initial examinations of the data showed no evidence of nonlinear relations between child-care quality and children's outcomes, indicat- ing that there is not a specific threshold at which qual- ity begins to have a positive effect. Rather, the linear relations that were found indicate that better quality child care is related to better outcomes for children across the spectrum of quality, so that the more qual- ity is increased, the better off children are. These find- ings provide support for the importance of high-quality preschool experiences as a mechanism for promot- ing school readiness and school success. Because so many children are experiencing center-based child care before they enter school, it is important to exam- ine the longer term effects of variations in the quality of these programs on children's development and subsequent school success. This study provides longi- tudinal evidence that higher quality care is modestly associated with a wide variety of better cognitive and socioemotional outcomes for children from diverse backgrounds, including differences in gender, ethnic background, and level of maternal education. Al- though in some cases the positive effects of higher quality care were even stronger and longer lasting for children at greater risk, higher quality child care was still associated with better outcomes for all groups of children. From a policy perspective, these findings indicate that the need for high-quality child care is of universal importance, and that policies promoting better quality child care have benefits that last into the early school years.


The authors gratefully acknowledge support for this research project by the Carnegie Corporation of New York, the William T. Grant Foundation, the JFM Foundation, the A. L. Mailman Family Foundation, the David and Lucile Packard Foundation, the Pew Charitable Trusts, the Smith Richardson Founda- tion, the USWEST Foundation, and the Educational Research and Development Centers Program as administered by the Office of Educational Research and Improvement, PR / Award Number R307A60004,

U.S. Department of Education. The contents of this article do not necessarily represent the positions or policies of the National Institute on Early Child- hood Development and Education, the Office of Educational Research and Improvement, the U.S. Department of Education, or any other sponsoring organization. The Cost, Quality, and Outcomes Study was conducted by a team of researchers in- cluding Kathleen Bernier, Donna Bryant, Margaret Burchinal, Patricia Byler, Richard Clifford, Debby Cryer, Mary Culkin, Suzanne Helburn, Carollee Howes, Sharon Lynn Kagan, H. Naci Mocan, John Morris, Elleh Peisner-Feinberg, Leslie Phillipsen, Jean Rustici, and Noreen Yazejian. The authors thank the participating children, families, teachers, child-care centers, and schools for their effort and involvement. They also thank the numerous individuals who have assisted with this study over the years.


Corresponding author: Ellen S. Peisner-Feinberg, Frank Porter Graham Child Development Center, University of North Carolina at Chapel Hill, CB# 8180, 105 Smith Level Road, Chapel Hill, NC 27599- 8180; e-mail: Margaret R. Burchinal, Richard M. Clifford, and Noreen Yazejian are also at the University of North Carolina at Chapel Hill; Mary L. Culkin is at the University of Colorado Health Sciences Center, Denver; Carollee Howes is at the University of California at Los Angeles; and Sharon Lynn Kagan is at Yale University, New Haven, CT.


Alexander, K. L., & Entwisle, D. R. (1988). Achievement in the first 2 years of school: Patterns and process. Monographs of the Society for Research in Child Development, 53

(2, Serial No. 218). Andersson, B-E. (1989). Effects of public day care: A longi- tudinal study. Child Development, 60, 857-866.

Arnett, J. (1989). Caregivers in day-care centers: Does train- ing matter? Journal of Applied Developmental Psychology, 10,541 -552.

Broberg, A. G., Wessels, H., Lamb, M. E., & Hwang, C. P. (1997). Effects of day care on the development of cogni- tive abilities in 8-year-olds: A longitudinal study. Developmental Psychology, 33, 62-69. Bronfenbrenner, U., & Morris, P. A. (1998). The ecology of developmental processes. In R. M. Lerner (Ed.), W. Damon (Series Ed.), Handbook of child psychology: Vol. 1. Theoretical models of human development. New York: Wiley. Bryant, D. M., Burchinal, M., Lau, L. B., & Sparling, J. J. (1994). Family and classroom correlates of Head Start children's developmental outcomes. Early Childhood Re- search Quarterly, 9, 289-309. Bryson, K., & Casper, L. M. (1997). Household and family characteristics: March 1997. (US. Bureau of the Census Current Population Reports No. P20-509), Washington, DC: U.S. Bureau of the Census. Burchinal, M. R., Campbell, F. A,, Bryant, D. M., Wasik, B. H., & Ramey, C. T. (1997). Early intervention and mediating processes in cognitive performance of children of low- income African American families. Child Development, 68,935-954. Campbell, F. A., & Ramey, C. T. (1994). Effects of early inter- vention on intellectual and academic achievement: A follow-up study of children from low-income families. Child Development, 65, 684-698. Chin-Quee, D. S., & Scarr, S. (1994). Lack of early child care effects on school-age children's social competence and academic achievement. Early Development and Parenting, 3,103-112. Clarke-Stewart, K. A,, & Gruber, C. P. (1984). Day care forms and features. In R. C. Ainslie (Ed.), The child and the day care setting (pp. 35-62). New York: Praeger. Cohen, J. (1988). Statistical power analysis of the behavioral sciences (2nd ed., pp. 407-424). Hilldale, NJ: Erlbaum. Deater-Deckard, K., Pinkerton, R., & Scarr, S. (1996). Child care quality and children's behavioral adjustment: A

four-year longitudinal study. Journal of Child Psychology

and Psychiatry, 37,937-948.

Dunn, L. (1993). Proximal and distal features of day care quality and children's development. Early Childhood Research Quarterly, 8, 167-192.

Dunn, L. M., & Dunn, L. M. (1981).Peabody Picture Vocabu- lay Test-Revised. Circle Pines, MN: American Guidance Service.

Goelman, H., & Pence, A. R. (1987). Effects of child care, family and individual characteristics on children's lan- guage development: The Victoria day care research project. In D. A. Phillips (Ed.), Quality in child care: What does the research tell us? Washington, DC: National Asso- ciation for the Education of Young Children.

Hagekull, B., & Bohlin, G. (1995). Day care quality, family and child characteristics and socioemotional develop- ment. Early Childhood Research Quarterly, 10, 505-526.

Harms, T., & Clifford, R. M. (1980). The Early Childhood Envi- ronment Rating Scale. New York: Teachers College Press.

Howes, C. (1990). Can the age of entry into child care and the quality of child care predict adjustment in kindergar- ten? Developmental Psychology, 26,292-303.

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). New York: Guilford.

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.

Jacobs, E. V., & White, D. R. (1994). The relationship of child-care quality and play to social behavior in the kin- dergarten. In H. Goelman & E. Jacobs (Eds.), Children's play in child care settings (pp. 85-101). Albany: State Uni- versity of New York Press.

Kontos, S. J. (1991). Child care quality, family background, and children's development. Early Childhood Research Quarterly, 6, 249-262.

Kontos, S., & Fiene, R. (1987). Child care quality, compliance with regulations, and children's development: The Penn- sylvania Study. In D. Phillips (Ed.), Quality in child care: What does the research tell us? (pp. 57-79). Washington, DC: National Association for the Education of Young Chldren.

Lazar, I., Darlington, R., Murray, H., Royce, J., & Snipper, A. (1982). Lasting effects of early education: A report from the Consortium for Longitudinal Studies. Monographs ofthe So- cietyfor Research in Child Development, 47(2-3, SerialNo. 195).

McCartney, K., Scarr, S., Phillips, D., & Grajek, S. (1985). Day care as intervention: Comparisons of varying qual- ity programs. Journal ofApplied Developmental Psychology, 6,247-260.

NICHD Early Child Care Research Network (ECCRN). (1997). The effects of infant child care on infant-mother attachment security: Results of the NICHD Study of Early Child Care. Child Development, 68, 860-879.

NICHD Early Child Care Research Network (ECCRN). (1998a). Early child care and self-control, compliance, and problem behavior at 24 and 36 months. Child Devel- opment, 69,1145-1170.

NICHD Early Child Care Research Network (ECCRN). (1998b). Relations between family predictors and child Peisner-Feinberg et al. 1553

outcomes: Are they weaker for children in child care?

Developmental Psychology, 34,1119-1128.

NICHD Early Child Care Research Network (ECCRN). (2000). The relation of child care to cognitive and lan- guage development. Child Development, 71,823-839.

Osbome, S. S., Schulte, A. C., & McKinney,J. D. (1991). A longi- tudinal study of students with learning disabilities in main- stream and resource programs. Exceptionality,2,81-96.

Peisner-Feinberg, E. S., & Burchinal, M. R. (1997). Relations between preschool children's child care experiences and concurrent development: The Cost, Quality, and Out- comes Study. Merrill-Palmer Quarterly, 43,451-477.

Phillips, D., McCartney, K., & Scarr, S. (1987). Child care quality and children's social development. Developmental Psychology, 23, 537-543.

Pianta, R. C. (1992). The Student-Teacher Relationship Scale. University of Virginia, Charlottesville, VA.

Pianta, R. C., & Steinberg, M. (1992). Teacher-child relation- ships and the process of adjusting to school. In New Di- rections for Child Development: No. 57 (pp. 61-80). San Francisco: Jossey-Bass.

Pianta, R. C., Steinberg, M., & Rollins, K. (1995). The first two years of school: Teacher-child relationships and de- flections in children's classroom adjustment. Development and Psychopathology, 7,295-312.

Sameroff, A. J., Seifer, R., Baldwin, A., & Baldwin, C. (1993). Stability of intelligence from preschool to adolescence: The influence of social and family risk factors. Child Development, 64,80-97.

Schaefer, E. S., Edgerton, M., & Aaronson, M. (1978). Classroom Behavior Inventory. Unpublished manuscript, Uni- versity of North Carolina at Chapel Hill, Chapel Hill, NC.

Schliecker, E., White, D. R., &Jacobs, E. (1991). The role of day care quality in the prediction of children's vocabu- lary. Canadian Journal of Behavioural Science, 23, 12-24.

Schweinhart, L. J., Barnes, H. V., & Weikart, D. P. (1993). Significant benefits: The HighlScope Perry Preschool Study through age 27. Ypsilanti, MI: High/Scope Press.

Secada, W. G. (1997). Instructional Environment Observation Scales. Unpublished manuscript, University of Wiscon- sin, Madison.

Stipek, D., Daniels, D., Galuzzo, D., & Milburn, S. (1992). Characterizing early childhood education programs for poor and middle-class children. Early Childhood Research Quarterly, 7, 1-19.

Vandell, D. L., Henderson, V. K., & Wilson, K. S. (1988). A longitudinal study of children with day-care experiences of varying quality. Child Development, 59,1286-1292.

West, J.,Wright, D., & Hausken, E. G. (1995). Child care and early education program participation ofinfants, toddlers, and preschoolers. Washington, DC: U.S. Department of Edu- cation, National Center for Education Statistics.

Whitebook, M., Howes, C., & Phillips, D. (1989). Who cares? Child care teachers and the quality of care in America (Final report of the National Child Care Staffing Study). Oak- land, CA: Child Care Employee Project.

Woodcock, R. W., &Johnson, M. B. (1990). Woodcock-Johnson Psycho-Educational Battery-Revised. Allen, TX: DLM Teaching Resources.

  • Recommend Us