ACF OPRE: Early Head Start Research and Evaluation Project (EHSRE)

V. EARLY HEAD START PROGRAMS’ OVERALL IMPACTS ON CHILDREN’S DEVELOPMENT, PARENTING, AND FAMILY WELL-BEING

This chapter presents findings from our analysis of the overall impacts of 17 Early Head Start programs on the children and families they served. The chapter begins with a brief discussion of the various ways in which Early Head Start programs work with parents and children and suggests why these programmatic strategies can be expected to have positive influences on children’s development, parenting behaviors and attitudes, and other aspects of child and family well-being. In some cases, the different program approaches implemented by Early Head Start programs, as discussed in Chapter I, are expected to have different patterns of impacts. Those differences, as well as differences in impacts related to patterns of program implementation, are presented in Chapter VI. In Chapter VII, we explore how children and parents who entered the program with different characteristics fared. First, however, this chapter focuses on the overall impacts—the ways in which Early Head Start programs, on average, were found to make a difference in the lives of the families they have served during the first three years of the children’s lives.

In developing hypotheses to guide our analysis and interpretations, we have drawn on research literature, the experiences of other programs, but also, to considerable degree, on site visit discussions with Early Head Start program staff about their theories of change.1 Following the presentation of findings from the national study, we present findings in “boxes” that address impacts of the program on Early Head Start fathers and local research reports that pertain to site-specific findings.

One of the major goals of Early Head Start is to improve the cognitive, social, and emotional development of infants and toddlers in low-income families. Programs seek to accomplish these aims by working directly with the child in center-based care, during home visits, or both, and to support this work through health, nutrition, and parent education services. Programs also support children’s development indirectly by working with parents and providing parent education to support close parent-child relationships, which are expected to enhance the longer-term development of infants and toddlers.

Close relationships provide infants and toddlers with the emotional support necessary for developing trusting relationships with important adults in their lives, learning to regulate their emotional responses, and playing cooperatively with their peers. Trusting relationships also support cognitive development (especially cause-and-effect reasoning) and communication skills. Parent-child interactions that also include talking, reading, teaching, and encouragement of new developmental experiences can promote the cognitive development of infants and toddlers. Parents support their children’s cognitive development by creating a supportive and stimulating learning environment in the home.

A strong parent-child relationship is expected to support and extend the development of infants and toddlers while families participate in the Early Head Start program and well into the future, as parents continue to guide children in the years after Early Head Start services end. In addition, programs focus to some degree on improving parent and family well-being, which can constitute a third, but more indirect, influence on child outcomes. Programs may seek to improve family functioning and in so doing may help parents move toward self-sufficiency; improvements in self-sufficiency, in turn, will offer families more resources to support a more cognitively stimulating home environment and activities for the child.

Parents’ ability to develop a supportive relationship with their children and make progress toward self-sufficiency may depend on their mental health and various aspects of family functioning. For example, parents who are depressed or who live in families with high levels of conflict may have difficulty in nurturing their children and functioning in the workplace. The effects of stress, conflict, and depression on children may be mediated by the parent-child relationship. Programs attempt to address mental health and family functioning in a variety of ways, but it is very challenging for them to overcome these substantial barriers to the development of supportive parent-child relationships and economic self-sufficiency.

Early Head Start eligibility guidelines require that at least 90 percent of enrolled families have incomes below the poverty line. While they have many strengths, families at this income level often struggle for survival, and financial concerns can interfere with parenting. Therefore, to develop support for the children, many programs aim to help families become economically stable and move toward self-sufficiency.

A. HYPOTHESES AND BRIEF SUMMARY OF FINDINGS

Early Head Start was designed as an intervention to support children’s development, promote supportive parent-child relationships, and assist families in their efforts to attain self-sufficiency. As described in Chapter III, Early Head Start programs provided extensive services of many kinds to their families, and the broad range of services families received would be expected to promote such outcomes. Further, in most areas, as reported in Chapter IV, Early Head Start families received substantially more services than their control-group counterparts did. The differences in receipt of parent education, home visits, center-based care, and case management, both overall and at the intensity required by Head Start program performance standards, support hypotheses of both direct and indirect impacts on children, parenting and the home environment, and self-sufficiency outcomes.

The programs’ focus on child development and parenting leads us to expect impacts on child cognitive, language, and social-emotional development and on parenting practices and knowledge. We further expect that the case management support provided by programs has the potential to enhance parents’ mental health, family functioning, and self-sufficiency. In addition, as a consequence of the programs’ focus on family development and enhancements in the quality of child care that programs provide or arrange for, we expect modest impacts on self-sufficiency.

To summarize the 3-year findings briefly, before presenting them in detail, Early Head Start had favorable impacts on a wide range of outcomes for children and parents. For the most part, the impacts found at 2 years were sustained at age 3.

For children, the programs produced positive impacts on cognitive and language development at age 3, sustaining the impacts found when children were 2.
For children, the programs produced favorable impacts on aspects of social-emotional development at age 3, broadening the range of impacts on these behaviors found at age 2. At age 3, Early Head Start children engaged their parents more, were more attentive during play, and showed less negativity toward parents during play compared to control-group children, and levels of aggressive behavior were lower than for control children.
When children were 3, Early Head Start programs continued to have positive impacts on parenting behavior, including emotional support and support for the child’s language development and learning. The programs also led to lower levels of insensitive and hostile parenting behavior and to the use of less-punitive discipline strategies.
At age 3, we found no overall impacts on measures of parent’s health or mental health and family functioning, although some had been seen when children were 2.
Important for parent self-sufficiency, overall results showed continued impacts on training and education activities, with some emerging impacts on employment (but not in average hours worked per week), and fewer subsequent births among Early Head Start mothers.

B. OUTCOME MEASURES USED AT AGE 3

Measures of children’s behavior and development, parenting, and family development were chosen to assess areas that Early Head Start was expected to influence, and that are important indications of favorable early development. We selected measures that had been used in previous evaluations and large-scale studies of children and families. We used multiple methods of measurement, including direct assessments of children; parent report; interviewer observation of the parent and child during the in-home interview; and videotaped, semistructured parent-child interaction tasks that were later coded by trained psychologists following a standard protocol. Use of multiple methods for measuring outcomes within a single domain avoids reliance on any single method that may have particular biases or inaccuracies.

Next, we provide an overview of the domains of child development, parenting, parent mental health, family functioning, and self-sufficiency activities measured at the most recent follow-up point. Descriptions of the particular measures used are provided throughout this chapter in boxes next to each table of impact estimates to help in interpreting the findings in each area. Details on the measures’ psychometric properties are given in Appendix C.

1. Child Development Measures

Cognitive development is a critical area to measure at this early age because of the foundation that knowledge and such skills as problem solving establish for later success in school. Language development is important as a foundation for cognitive and social development. Infants and toddlers are in a particularly sensitive period for language development; language delays during this period can persist, and may inhibit the acquisition of reading skills later on. We conducted direct assessments of children’s cognitive and language development.

Social-emotional development, including persistence and self-control, are developing during infancy and toddlerhood and contribute to children’s ability to learn in a variety of settings. Greater self-control, less-aggressive behavior, and a more positive relationship with the parent are important foundations for relationships with peers and with other adults. We used a combination of parent report and observation to measure children’s social-emotional development.

2. Parenting and Home Environment Measures

To measure the impacts of Early Head Start on parenting behavior and the home environment, we tapped four important areas:

Emotional support, which includes the parent’s warmth and affection toward the child, positive feelings about the child that are conveyed to others, and appropriate responses to needs that the child communicates
Stimulation of language and learning, which includes talking and reading to the child regularly, encouragement for learning basic concepts such as colors, numbers, and the alphabet, and the parent’s approach to assisting the child with a challenging task
Negative aspects of parenting behavior, which include insensitivity, emotional detachment from the child, hostility, anger, and punitiveness
The parent’s knowledge about safety and discipline strategies

Measures of parenting behavior and the home environment were collected using several different methods, including parents’ self-report, observations conducted by in-home interviewers, and coded videotaped interactions with their child, which guarded against biases and inaccuracies that can arise when relying on a single measurement strategy.

3. Measures of Parent Health and Mental Health, Family Functioning, and Self-Sufficiency

Parent health and mental health and family well-being are important, both for supporting parent-child relationships and for parents’ progress toward self-sufficiency. In fact, a number of programs described a theory of change that included such constructs as parent mental health as important expected outcomes. Nevertheless, these outcomes are not the main focus of most program services, and they are particularly challenging for programs to influence. We included brief, parent-report measures of these outcomes that have been widely used in empirical studies and have demonstrated validity. Measures of parent health, mental health, and family functioning include health status, feelings of depression, family conflict, and stress related to parenting. Measures of economic self-sufficiency tap education and training, employment, welfare program participation, family income, and births since enrollment.

4. Data Sources for Child, Parent, and Family Measures

Data come from two major sets of follow-up measures (see Chapter II and Appendix C for details). Assessment of children’s development and some aspects of parenting behavior require standardization or modification as children get older; thus, measures of these constructs were collected at specific age levels (when children were approximately 14, 24, and 36 months old). Outcomes closely related to child development and parenting, including mental health and family functioning, were also collected during the birthday-related interviews. Self-sufficiency activities, like the receipt of program services, are likely to be influenced by the length of the intervention. Therefore, information on these outcomes was collected at intervals after the family enrolled in Early Head Start (on average at 7, 16, and 28 months).

C. OVERALL IMPACTS ON CHILDREN’S DEVELOPMENT

Early Head Start programs had favorable impacts on a broad range of child development outcomes at age 3. This section discusses the programs’ impacts on cognitive and language development and on social-emotional development. Overall, the programs’ impacts on children’s cognitive and language development at age 2 were sustained at age 3, and impacts on social-emotional development at age 3 were greater and broader than they had been at age 2.

1. Overall Impacts on Cognitive and Language Development

Early Head Start enhanced children’s cognitive and language development at age 3, sustaining the positive impacts on cognitive development and language found at age 2 (Box V.1 describes the measures and Table V.1 presents the impacts). Early Head Start children scored higher on the Bayley Mental Development Index (MDI) at age 3 than control children, replicating the findings at age 2. Perhaps even more important, fewer Early Head Start than control children scored below 85 on the MDI (one standard deviation below the standardized mean). Reducing the number of children scoring below this threshold may be indicative of Early Head Start programs potentially reducing the need for special education services. This effect was first seen at age 2 and was sustained through age 3.

At age 3, the Peabody Picture Vocabulary Test, Third Edition (PPVT-III), a test of children’s receptive vocabulary, was administered. Early Head Start children scored higher on the PPVT-III at age 3 than control children. In addition, fewer Early Head Start than control children scored below 85 on the PPVT-III. At age 2, the significant positive vocabulary impacts were based on parent-reported vocabulary; it is noteworthy that this effect was sustained when this widely used, standardized direct assessment of receptive vocabulary was administered when children were 3 years old. Children who spoke Spanish in the home were assessed using the Test de Vocabulario en Imagenes Peabody (TVIP), which measures receptive vocabulary in Spanish. We found no significant impacts on the TVIP standard scores or on the percentage with scores below 100. Fewer than 200 children were assessed using this measure, however.

To investigate supporting evidence for the impacts on receptive vocabulary, we factor analyzed the Bayley and found a “language/reasoning” factor. Early Head Start programs had a significant impact on this outcome.2 While this factor is not an accepted standard measure (and is highly correlated with the MDI at r = .78), this finding suggests an impact of Early Head Start on broader aspects of language development than receptive vocabulary, since the Bayley items include observations of the extent and complexity of the child’s spoken language (language production).

In summary, the positive Early Head Start impacts on cognitive and language development found when children were 2 years old were sustained through age 3. The reduction in the proportion of children scoring below 85 on the Bayley MDI and PPVT-III may be especially important in reducing the likelihood that children would need special services at an early age. It is important to note that although Early Head Start had positive impacts on children’s cognitive and language development, average scores on the cognitive and language assessments for both program- and control-group children remained below the national average.

2. Overall Impacts on Children’s Social-Emotional Development

Because policymakers, parents, and caregivers view positive and negative behaviors differently, and because the evaluation obtained data on both aspects of social-emotional development, we present the results separately here. Although when the children were 2 years old we found no Early Head Start impacts on the positive aspects of children’s social-emotional development, when they were a year older, significant positive impacts were found on some aspects of children’s behavior during play, as assessed by trained observers of videotaped parent-child interactions (see Box V.2 and Table V.2).

Early Head Start children were more engaging of their parents during play; in other words, Early Head Start children, when compared to controls, were more likely to behave in ways that maintained interaction with their parent. They were also rated as more attentive to objects during play at age 3 than were control children, a behavior pattern that, should it persist, could be important for attending to tasks in later preschool programs the children might attend. Early Head Start programs did not have a significant impact on child behavior during the puzzle challenge task, as rated by trained coders of videotaped parent-child interactions, or on child behavior during the Bayley assessment, as rated by trained observers during the in-home interviews.3

BOX V.1

MEASURES OF COGNITIVE AND LANGUAGE DEVELOPMENT

Bayley Mental Development Index (MDI) – measures the cognitive, language, and personal-social development of children under age 3½. Children were directly assessed by the Interviewer/Assessor following a standardized protocol.

The MDI is one of three component scales of the Bayley Scales of Infant Development – Second Edition (Bayley 1993). At 36 months, the child is assessed on his/her ability to follow simple spoken directions that indicate an understanding of prepositions, size comparisons, quantities, colors, and simple numbers; on his or her spoken vocabulary during the assessment; on spatial concepts, memory, and the ability to match shapes and identify patterns.

For example, the child is asked to build a bridge and a wall of cubes; identify the big tree in a picture; count; understand prepositions like in, under, or between; name four colors; sort pegs by color; place shapes into holes of the same size and shape; use the past tense; and repeat short number sequences.

The Bayley MDI was normed on a nationally representative sample of children of various ages so that raw scores can be converted to age-adjusted, standardized scores with a mean of 100 and a standard deviation of 15.

The percentage of children with Bayley MDI below 85 measures the proportion with delayed performance, or scores one standard deviation or more below the mean for their age in the nationally representative, standardization sample.

Peabody Picture Vocabulary Test, Third Edition (PPVT-III) – measures listening comprehension of spoken words in standard English for children and adults from age 2 ½ and over (Dunn and Dunn 1997). The child is presented with four pictures and is asked to point to the picture that matches the word spoken by the interviewer. The PPVT-III was normed on a nationally representative sample of children and adults of various ages so that raw scores can be converted to age-adjusted, standardized scores with a mean of 100 and a standard deviation of 15.

The percentage of children with PPVT-III below 85 measures the proportion with scores one standard deviation or more below the mean for their age in the nationally representative, standardization sample.

Test de Vocabulario en Imagenes Peabody (TVIP) – measures the listening comprehension of spoken words in Spanish for Spanish-speaking and bilingual children from age 2 ½ to 18 (Dunn, Lloyd, Eligio, Padilla, Lugo, and Dunn 1986). The child is presented with four pictures and is asked to point to the picture that matches the Spanish word spoken by the interviewer. The TVIP was normed on a sample of Mexican and Puerto Rican children of various ages so that raw scores can be converted to age-adjusted, standardized scores with a mean of 100 and a standard deviation of 15.

The percentage of children with TVIP below 100 measures the proportion with scores below the mean for their age in the standardization sample. This cutoff was chosen because only 6 percent of the Early Head Start evaluation sample scored below 85. The higher standardized scores on the TVIP compared to the PPVT-III could be attributable to the fact that norms for the TVIP were developed nearly two decades ago.

**TABLE V.1**
**IMPACTS ON CHILD COGNITIVE AND LANGUAGE DEVELOPMENT**
Outcome	Program Group Participantsa	Control Groupb	Estimated Impact per Participantc	Effect Sized
Cognitive Development
Bayley Mental Development Index (MDI) Standard Score	91.4	89.9	1.6**	12
Percentage with Bayley MDI Below 85	27.3	32	-4.7*	-10.1
Receptive Language Development
Peabody Picture Vocabulary Test (PPVT-III) Standard Score	83.3	81.1	2.1**	13.1
Percentage with PPVT-III Below 85	51.1	57.1	-6.0**	-12.1
Test de Vocabulario en Imagenes Peabody (TVIP) Standard Score	97.2	94.9	2.3	27.1
Percentage with TVIP Below 100	36.2	41.2	-5	-9.9
Sample Size
Bayley	879	779	1,658
PPVT	738	665	1,403
TVIP	95	89	184
SOURCE: Parent interview and child assessments conducted when children were approximately 36 months old. NOTE: All impact estimates were calculated using regression models, where each site was weighted equally. aA participant is defined as a program group member who received more than one Early Head Start home visit, met with an Early Head Start case manager more than once, received at least two weeks of Early Head Start center-based care, and/or participated in Early Head Start group parent-child activities.(back) bThe control group mean is the mean for the control group members who would have participated in Early Head Start if they had been assigned to the program group instead. This unobserved mean was estimated as the difference between the program group mean for participants and the impact per participant.(back) cThe estimated impact per participant is measured as the estimated impact per eligible applicant divided by the proportion of program group members who participated in Early Head Start services (which varied by site). The estimated impact per eligible applicant is measured as the difference between the regression-adjusted means for all program and control group members.(back) dThe effect size was calculated by dividing the estimated impact per participant by the standard deviation of the outcome measure for the control group times 100 (that is, it is the impact expressed as a percentage of the standard deviation).(back) Significantly different from zero at the .10 level, two-tailed test. Significantly different from zero at the .05 level, two-tailed test. **Significantly different from zero at the .01 level, two-tailed test.

BOX V.2

MEASURES OF POSITIVE ASPECTS OF CHILD SOCIAL-EMOTIONAL DEVELOPMENT

Child Behavior During Parent-Child Semistructured Play – measures the child’s behavior with the parent during a semistructured play task. The parent and child were given three bags of interesting toys and asked to play with the toys in sequence. The semistructured play task was videotaped, and child and parent behaviors were coded on a 7-point scale by child development researchers according to strict protocols (see Appendix C). This assessment was adapted for this evaluation from the Three Box coding scales used in the NICHD Study of Early Child Care (NICHD Early Child Care Research Network 1999). Two positive aspects of children’s behavior with the parent were rated on a 7-point scale:

Engagement – measures the extent to which the child shows, initiates, or maintains interaction with the parent. This may be expressed by approaching or orienting toward the parent, establishing eye contact with the parent, positively responding to the parent’s initiations, positive affect directed toward the parent and/or engaging the parent in play. Very high engagement receives a 7.

Sustained Attention with Objects – measures the degree to which the child is involved with the toys presented in the three bags. Indicators include degree to which the child “focuses in” when playing with an object and the extent to which the child coordinates activities with several objects and/or explores different aspects of a toy. Very high sustained attention receives a 7.

Child Behavior During Parent-Child Puzzle Challenge Task – measures the child’s behavior with the parent during a puzzle completion task. The child was given a puzzle to play with, and the parent was instructed to give the child any help needed. After 3 minutes, or earlier if the puzzle was completed, the interviewer gave the child a second, harder puzzle and asked the mother not to help the child. If that puzzle was completed or 3 minutes elapsed, another, more challenging puzzle was provided. The puzzle challenge task was videotaped, and child and parent behaviors were coded on a 7-point scale by child development researchers according to strict protocols (see Appendix C). The scales are based on a puzzle task used by Brooks-Gunn et al. (1992) in the Newark Observational Study of the Teenage Parent Demonstration. Two positive aspects of children’s behavior with the parent were rated on a 7-point scale:

Engagement – measures the extent to which the child shows, initiates, or maintains interaction with the parent. This may be expressed by approaching or orienting toward the parent, establishing eye contact with the parent, positively responding to the parent’s suggestions, positive affect directed toward the parent and/or engaging the parent in the puzzle task. Very high engagement receives a 7.

Persistence – measures how goal-oriented, focused, and motivated the child remains toward the puzzle throughout the task. The focus of this measure is on the child’s apparent effort to solve the puzzle, not on how well the child performs. Very high persistence receives a 7.

Bayley Behavioral Rating Scale (BRS) – measures the child’s behavior during the Bayley MDI assessment. The BRS is one of three component scales of the Bayley Scales of Infant Development – Second Edition (Bayley 1993).

Emotional Regulation – measures the child’s ability to change tasks and test materials; negative affect; and frustration with tasks during the assessment.

Orientation/Engagement – measures the child’s cooperation with the interviewer during the assessment; positive affect; and interest in the test materials.

The interviewer assesses the child’s behavior by scoring items on a 5-point scale, with 5 indicating more positive behavior (for example, less frustration and more cooperation). Scores are the average of the items in the subscale.

TABLE V.2

IMPACTS ON POSITIVE ASPECTS OF CHILD SOCIAL-EMOTIONAL DEVELOPMENT
Outcome	Program Group Participantsa	Control Groupb	Estimated Impact per Participantc	Effect Sized
Engagement of Parent During Parent-Child Semistructured Playe	4.8	4.6	0.2***	20.3
Sustained Attention to Objects During Parent-Child Semistructured Playe	5	4.8	0.2***	15.9
Engagement of Parent During Parent-Child Puzzle Challenge Taskf	5	4.9	0.1	8.8
Persistence During Parent-Child Puzzle Challenge Taskf	4.6	4.5	0.1	6.3
Bayley Behavior Rating Scale (BRS): Emotional Regulationg	4	4	0	0.6
Bayley BRS: Orientation/Engagementg	3.9	3.8	0	4
Sample Size
Parent-Child Interactions	875	784	1,659
Bayley BRS	936	833	1,769
SOURCE: Child assessments, interviewer observations, and assessments of semistructured parent-child interactions conducted when children were approximately 36 months old. NOTE: All impact estimates were calculated using regression models, where each site was weighted equally. aA participant is defined as a program group member who received more than one Early Head Start home visit, met with an Early Head Start case manager more than once, received at least two weeks of Early Head Start center-based care, and/or participated in Early Head Start group parent-child activities.(back) bThe control group mean is the mean for the control group members who would have participated in Early Head Start if they had been assigned to the program group instead. This unobserved mean was estimated as the difference between the program group mean for participants and the impact per participant.(back) cThe estimated impact per participant is measured as the estimated impact per eligible applicant divided by the proportion of program group members who participated in Early Head Start services (which varied by site). The estimated impact per eligible applicant is measured as the difference between the regression-adjusted means for all program and control group members.(back) dThe effect size was calculated by dividing the estimated impact per participant by the standard deviation of the outcome measure for the control group times 100 (that is, it is the impact expressed as a percentage of the standard deviation).(back) eBehaviors were observed during the videotaped parent-child semistructured play task and coded on a seven-point scale.(back) fBehaviors were observed during the videotaped parent-child puzzle challenge task and coded on a seven-point scale.(back) gBehaviors were observed during the Bayley assessment and rated on a five-point scale by the interviewer/assessor.(back) Significantly different from zero at the .10 level, two-tailed test. Significantly different from zero at the .05 level, two-tailed test. **Significantly different from zero at the .01 level, two-tailed test.

The positive impacts found in the semistructured play interaction suggest that Early Head Start may improve the parent-child relationship and the child’s ability to focus on objects during play. By enhancing the degree to which young children maintain interest in exploring objects they are playing with and maintain their interaction with their parent, Early Head Start programs may be contributing to behavior patterns that will help children learn in early learning settings.

These impact analyses were supplemented with growth curve analyses for selected outcomes. (The methodology and approach to these analyses are described in Chapter II, with more details in Appendix D.5. Appendix D.5 also includes figures depicting the growth curves that are reported in this chapter.) We undertook these analyses to take advantage of the longitudinal nature of some of the measures. Because growth curves required having the same measures administered at all three ages, these results are limited.4 However, they do show change over time and allow us to conclude not only that Early Head Start programs produced impacts at particular points in time but that, in a few cases, altered the rate of change over time (indicated by a significant impact on the slopes of the curves). For child engagement of the parent and child sustained attention with objects, the group mean differences were significant at 2 and 3 years of age, as found in the overall impact analyses just described. The Early Head Start program experience did not alter the growth trends, however (that is, the program had no significant impact on slopes).

In general, there was a broader pattern of favorable impacts on reducing negative aspects of children’s social-emotional development at age 3 than at age 2. Early Head Start reduced two of the three negative measures of children’s social-emotional development at age 3. The reduction in parent-reported aggressive behavior sustains the findings at age 2 and extends them to behavior in semistructured play with the parent (see Box V.3 and Table V.3). Similar to findings at age 2, Early Head Start children were reported by their parents as being less aggressive than control-group children.

At age 3, Early Head Start children also displayed less negativity toward their parents during semistructured play, an impact that did not appear at age 2 (Table V.3). The growth curve analysis of this outcome similarly showed no program impact on the change in negativity overtime—it declined at the same rate for both program and control children. Early Head Start had no impact on the level of child frustration during the parent-child puzzle challenge task at age 3, as rated by trained observers of videotaped parent-child interactions. This task was not administered at age 2.

As early aggressive behavior is predictive of later conduct problems (Moffitt et al.1996; and Dishion et al. 1995), these findings indicating less negativity toward the parent and less agressive behavior among Early Head Start children. Children may enhance children’s conduct and performance when they enter school.

BOX V.3

MEASURES OF NEGATIVE ASPECTS OF CHILD SOCIAL-EMOTIONAL DEVELOPMENT

Child Behavior During Parent-Child Semistructured Play – measures the child’s behavior with the parent during a semistructured play task. The parent and child were given three bags of interesting toys and asked to play with the toys in sequence. The semistructured play task was videotaped, and child and parent behaviors were coded on a 7-point scale by child development researchers according to strict protocols (see Appendix C). This assessment was adapted for this evaluation from the Three Box coding scales used in the NICHD Study of Early Child Care (NICHD Early Child Care Research Network 1999). Three aspects of children’s behavior with the parent were rated on a 7-point scale, with one reflecting a negative aspect of children’s social-emotional development:

Negativity Toward Parent – measures the degree to which the child shows anger, hostility, or dislike toward the parent. Expressions may be overt (for example, forcefully rejecting a toy offered by the parent or pushing the parent away) or covert (for example, hitting or throwing an object in response to the parent’s behavior). Very high negativity receives a 7.

Child Behavior During Parent-Child Puzzle Challenge Task – measures the child’s behavior with the parent during a puzzle completion task. The child was given a puzzle to play with, and the parent was instructed to give the child any help needed. After 3 minutes, or earlier if the puzzle was completed, the interviewer gave the child a second, harder puzzle and asked the mother not to help the child. If that puzzle was completed or 3 minutes elapsed, another, more challenging puzzle was provided. The puzzle challenge task was videotaped, and child and parent behaviors were coded on a 7-point scale by child development researchers according to strict protocols (see Appendix C). Three aspects of children’s behavior with the parent were rated on a 7-point scale, with one reflecting a negative aspect of children’s social-emotional development:

Frustration with Task – measures the degree to which the child expresses frustration or anger toward the puzzle task, for example, by putting hands in lap, whining, pushing away puzzle pieces, crying about the puzzle, saying it is too hard, or throwing puzzle pieces. Very high frustration receives a 7.

Child Behavior Checklist – Aggressive Behavior – this subscale measures the incidence of 19 child behavior problems that tend to occur together and constitute aggressive behavior problems. Parents completed the Aggressive subscale of the Child Behavior Checklist for Ages 1 ½ to 5 Years (Achenbach and Rescorla 2000). Some behaviors asked about include, “Child has temper tantrums,” “Child hits others,” and “Child is easily frustrated.” For each of the possible behavior problems, the parent was asked whether the child exhibits this behavior often, sometimes, or never. Scores range from 0, if all of the behavior problems are “never” observed by the parent, to 38, if all of the behavior problems are “often” observed.

TABLE V.3

IMPACTS ON NEGATIVE ASPECTS OF CHILD SOCIAL-EMOTIONAL DEVELOPMENT
Outcome	Program Group Participantsa	Control Groupb	Estimated Impact per Participantc	Effect Sized
Negativity Toward Parent During Parent-Child Semistructured Playe	1.2	1.3	-0.1**	-13.8
Frustration with Parent-Child Puzzle Challenge Taskf	2.7	2.7	0	2.2
Child Behavior Checklist: Aggressive Behavior	10.6	11.3	-0.7**	-10.8
Sample Size
Parent Interview	1,107	1,003	2,110
Parent-Child Interactions	875	784	1,659
SOURCE: Parent interviews and assessments of semistructured parent-child interactions conducted when children were approximately 36 months old. NOTE: All impact estimates were calculated using regression models, where each site was weighted equally. aA participant is defined as a program group member who received more than one Early Head Start home visit, met with an Early Head Start case manager more than once, received at least two weeks of Early Head Start center-based care, and/or participated in Early Head Start group parent-child activities.(back) bThe control group mean is the mean for the control group members who would have participated in Early Head Start if they had been assigned to the program group instead. This unobserved mean was estimated as the difference between the program group mean for participants and the impact per participant.(back) cThe estimated impact per participant is measured as the estimated impact per eligible applicant divided by the proportion of program group members who participated in Early Head Start services (which varied by site). The estimated impact per eligible applicant is measured as the difference between the regression-adjusted means for all program and control group members.(back) dThe effect size was calculated by dividing the estimated impact per participant by the standard deviation of the outcome measure for the control group times 100 (that is, it is the impact expressed as a percentage of the standard deviation).(back) eBehaviors were observed during the videotaped parent-child semistructured play task and coded on a seven-point scale.(back) fBehaviors were observed during the videotaped parent-child puzzle challenge task and coded on a seven-point scale.(back) Significantly different from zero at the .10 level, two-tailed test. Significantly different from zero at the .05 level, two-tailed test. **Significantly different from zero at the .01 level, two-tailed test.

D. OVERALL IMPACTS ON PARENTING

Early Head Start programs had favorable impacts on a broad range of parenting behavior, the home environment, and parenting knowledge. Overall, Early Head Start had favorable impacts on several aspects of emotional support for the child and support for language development and learning. Fewer impacts were found on negative aspects of parenting behavior, although there is evidence that the program reduced the use of punitive discipline.

1. Parenting Behavior and the Home Environment

This section discusses Early Head Start impacts on emotionally supportive parenting behavior, on measures of the parent’s support for the child’s language development and learning (including the overall measure of the emotional support and stimulation available in the home environment), and negative aspects of parenting behavior, including insensitivity, hostility, and punitive behavior.

a. Emotional Supportiveness

Early Head Start increased parents’ emotional supportiveness toward their children, as rated by interviewer observations and through coding of behavior during videotaped, semistructured parent-child activities (see Box V.4 and Table V.4). Early Head Start parents exhibited more warmth towards their children during the parent interview session, as rated by the interviewer/assessor using a short subscale of the Home Observation for Measurement of the Environment (HOME) warmth subscale. This finding replicates the positive impact of Early Head Start on emotional responsivity, a similar subscale of the HOME for infants and toddlers, used when children were 2 years old. The positive impact of Early Head Start at age 2 on parent supportiveness observed during parent-child semistructured play was sustained at age 3: Early Head Start parents were rated as more supportive (warmer, more sensitive, and offering more cognitive stimulation) in play than parents in the control group. In the puzzle challenge situation, in which parents were instructed to give needed help as their child tried to complete a series of puzzles (see description in Box V.4), however, there was no significant program effect on emotionally supportive parenting. In other words, Early Head Start parents were no more likely than control parents to show support and enthusiasm for their child’s work, or to display a positive attitude toward the child while the child attempted a complex activity that was challenging to complete (more so than the semistructured play task). This measure was not administered at age 2.

The group differences in parent supportiveness during the semistructured play task also are seen in the results of the growth curve analysis (see Appendix D.5). The growth curves indicate that this measure of supportiveness declined slightly over time, but the decline was the same for both groups of parents (that is, the program did not alter the rate of change). The observed decline for both groups may reflect parent provision of greater autonomy to their maturing, more capable children.

b. Support for Language and Learning

When children were 3 years old, Early Head Start had positive impacts on several aspects of parent support for language and learning and the overall quality of the home environment, continuing the pattern of impacts in this domain originally observed at age 2. These outcomes were measured by a variety of methods—parent report, interviewer observation, and coding by trained observers of videotaped parent-child interactions.

BOX V.4

MEASURES OF EMOTIONALLY SUPPORTIVE PARENTING

Home Observation for Measurement of the Environment (HOME) – measures the quality of stimulation and support available to a child in the home environment (Caldwell and Bradley 1984). At the 36-month assessment, we based our measure on the HOME-Short Form inventory, Preschool version, used in the National Longitudinal Survey of Youth (NLSY). Information needed to score the inventory is obtained through a combination of interview and observation conducted in the home with the child’s parent while the child is present. A total of 37 items were used for the 36-month HOME scale in this study. In addition to a total score, we derived five subscales from this assessment, with one related to emotional support:

Warmth – Measures responsive and supportive parenting behavior observed by the interviewer during the home visit. Items in this subscale are based entirely on interviewer observations of the parent and child during the interview, and include whether the mother kissed or caressed the child during the visit; whether her voice conveyed positive feeling, and whether she praised the child. Scores can range from 0, if none of the positive behaviors were observed, to 3, if all of the behaviors were observed.

Parent Behavior during Parent-Child Semistructured Play – measures the parent’s behavior with the child during a semistructured play task. The parent and child were given three bags of interesting toys and asked to play with the toys in sequence. The semistructured play task was videotaped, and child and parent behaviors were coded by child development researchers according to strict protocols (see Appendix C). This assessment was adapted for this evaluation from the Three Box coding scales used in the NICHD Study of Early Child Care (NICHD Early Child Care Research Network 1999). Four aspects of the parent’s behavior with the child were rated on a seven-point scale, with one aspect related to emotional support:

Supportiveness – this composite measure is an average of parental sensitivity, cognitive stimulation, and positive regard during play with the child. Sensitivity includes such behavior as acknowledgement of the child’s affect, vocalizations, and activity; facilitating the child’s play; changing the pace of play when the child seems under-stimulated or over-excited; and demonstrating developmentally appropriate expectations of behavior. Cognitive stimulation involves taking advantage of the activities and toys to facilitate learning, development, and achievement; for example, by encouraging the child to talk about the materials, by encouraging play in ways that illustrate or teach concepts such as colors or sizes, and by using language to label the child’s experiences or actions, to ask questions about the toys, to present activities in an organized series of steps, and to elaborate on the pictures in books or unique attributes of objects. Positive regard includes praising the child, smiling or laughing with the child, expressing affection, showing empathy for the child’s distress, and showing clear enjoyment of the child.

Parent Behavior during Parent-Child Puzzle Challenge Task – measures the parent’s behavior with the child during a puzzle completion task. The child was given a puzzle to play with, and the parent was instructed to give the child any help needed. After 3 minutes, or earlier if the puzzle was completed, the interviewer gave the child a second, harder puzzle and asked the mother not to help the child. If that puzzle was completed or 3 minutes elapsed, another, more challenging puzzle was provided. The puzzle challenge task was videotaped, and child and parent behaviors were coded on a 7-point scale by child development researchers according to strict protocols (see Appendix C). Four aspects of the parent’s behavior with the child were rated on a 7-point scale, with one aspect related to emotional support:

Supportive Presence – measures the parent’s level of emotional support and enthusiasm toward the child and his or her work on the puzzles; displays of affection and a positive attitude toward the child and his or her abilities.

TABLE V.4

IMPACTS ON EMOTIONALLY SUPPORTIVE PARENTING
Outcome	Program Group Participantsa	Control Groupb	Estimated Impact per Participantc	Effect Sized
Home Observation for Measurement of the Environment (HOME): Warmthe	2.6	2.5	0.1*	9
Supportiveness During Parent- Child Semistructured Playf	4	3.9	0.1***	14.6
Supportive Presence During Parent-Child Puzzle Challenge Taskg	4.5	4.4	0.1	4.2
Sample Size
Parent Interview	1,107	1,003	2,110
Parent-Child Interactions	874	784	1,658
SOURCE: Parent interviews and assessments of semistructured parent-child interactions conducted when children were approximately 36 months old. NOTE: All impact estimates were calculated using regression models, where each site was weighted equally. aA participant is defined as a program group member who received more than one Early Head Start home visit, met with an Early Head Start case manager more than once, received at least two weeks of Early Head Start center-based care, and/or participated in Early Head Start group parent-child activities.(back) bThe control group mean is the mean for the control group members who would have participated in Early Head Start if they had been assigned to the program group instead. This unobserved mean was estimated as the difference between the program group mean for participants and the impact per participant.(back) cThe estimated impact per participant is measured as the estimated impact per eligible applicant divided by the proportion of program group members who participated in Early Head Start services (which varied by site). The estimated impact per eligible applicant is measured as the difference between the regression-adjusted means for all program and control group members.(back) dThe effect size was calculated by dividing the estimated impact per participant by the standard deviation of the outcome measure for the control group times 100 (that is, it is the impact expressed as a percentage of the standard deviation).(back) eBehaviors were observed during the HOME assessment and rated on a yes/no scale by the interviewer/assessor.(back) fBehaviors were observed during the videotaped parent-child structured play task and coded on a seven-point scale. Supportiveness is a combination of Warm Sensitivity and Positive Regard.(back) gBehaviors were observed during the videotaped parent-child puzzle task and coded on a seven-point scale.(back) Significantly different from zero at the .10 level, two-tailed test. Significantly different from zero at the .05 level, two-tailed test. **Significantly different from zero at the .01 level, two-tailed test.

Sustaining the impact found at age 2, the total HOME score was significantly higher for Early Head Start families than for control families, suggesting that overall, Early Head Start children live in home environments that provide more emotional support and cognitive stimulation (see Box V.5 and Table V.5).

Early Head Start programs did not have an impact on the internal physical environment of the home, an index derived from the HOME scale that measures the presence of home furnishings and decorations as well as cleanliness and orderliness. Families in both groups received relatively high scores on this measure (which ranges from 3 to 9) so impacts would likely have been difficult to accomplish.

More importantly, Early Head Start families scored higher on the subscale of the HOME that measures support of language and learning (Table V.5). Thus, Early Head Start improved the amount of cognitively stimulating toys and materials, along with the interactions that children experience in the home. This finding is consistent with impacts found at age 2 on a comparable subscale of the HOME.

In the videotaped parent-child puzzle challenge, Early Head Start parents provided higher quality of assistance to their children as well. This is one of the few puzzle challenge outcomes for which Early Head Start impacts paralleled those in the semistructured play situation. The parent-child puzzle challenge task was not administered at age 2.

Early Head Start parents reported engaging more frequently in a broader range of play activities with their children, a finding that was significant at age 2 and sustained at age 3.

Early Head Start impacts on regular reading to children were mixed at age 3. Similar to the findings at age 2, when children were 3, Early Head Start parents were more likely than control-group parents to report that they read daily to their children (57 percent of program-group parents compared with 52 percent of control-group parents). However, Early Head Start had no impact on the proportion of Early Head Start parents reporting reading to their children regularly at bedtime at age 3, although there had been a favorable program impact at age 2.5 By age 3, 29 percent of control group families reported reading to their children at bedtime, a figure similar to the percentage of Early Head Start families who reported reading at bedtime at age 2, while 32 percent of program parents at age 3 reported reading at bedtime.

At age 3, we found no impact of Early Head Start on parents’ structuring the child’s day by keeping a regular bedtime and following regular bedtime routines. Nearly 60 percent of both program and control groups set a regular bedtime for their 3-year-old children, and nearly 70 percent followed regular bedtime routines. At age 2, the program had an impact on regular bedtimes but not on routines.

In general at age 3, Early Head Start parents provided more support for children’s language development and learning than control parents by making efforts to teach colors, shapes, and numbers, by frequent reading to the child, telling stories, and singing songs, by providing more cognitive stimulation in interaction with the child, and by providing cognitively stimulating books, toys, games, and materials in the home. However, Early Head Start parents were not more likely than control parents to structure the child’s day by setting a regular bedtime or following regular bedtime routines by age 3.

BOX V.5

MEASURES OF THE HOME ENVIRONMENT AND PARENT STIMULATION OF LANGUAGE AND LEARNING

Total Score – measures the cognitive stimulation and emotional support provided by the parent in the home environment. The total includes all 37 items. The maximum potential score is 37.

Support of Language and Learning – measures the breadth and quality of the mother’s speech and verbal responses to the child during the home visit, as rated by the interviewer; whether the parent encourages the child to learn shapes, colors, numbers, and the alphabet; the presence of books, toys, and games accessible to the child; and whether the parent reads to the child several times per week. Items are obtained by a combination of parent report and interviewer observation. The maximum potential score is 13.

Internal Physical Environment – measures the cleanliness, organization, and warmth of the home environment. Items in this subscale are based entirely on interviewer observations during the interview and were each coded on a 3-point scale for this subscale (but on a binary scale for the total HOME). Scores can range from 3 to 9.

Regular Bedtime – measures whether the parent has a regular bedtime for the child. The parent must name the time and report that the child went to bed at that time at least four of the past five weekdays.

Regular Bedtime Routines – measures whether the parent reports having a regular set of routines with the child around bedtime, such as singing lullabies, putting toys away, or telling stories.

Parent-Child Play – measures the frequency with which the parent engages in several activities with the child that can stimulate cognitive and language development, including reading or telling stories, dancing, singing, and playing outside together.

Read Every Day – measures whether the parent reported that she reads to the child “every day” or “more than once a day.”

Read at Bedtime – measures whether the parent reported that the child has a regular bedtime routine and, in response to an open-ended question about activities that are part of that routine, the parent reported that reading is one of the routine activities.

Quality of Assistance – measures the frequency and quality of clear guidance to the child, flexible strategies for providing assistance, and diverse, descriptive verbal instructions and exchanges with the child.

TABLE V.5

IMPACTS ON THE HOME ENVIRONMENT AND PARENT STIMULATION
OF LANGUAGE AND LEARNING
Outcome	Program Group Participantsa	Control Groupb	Estimated Impact per Participantc	Effect Sized
Home Observation for Measurement of the Environment (HOME) - Total Score	27.6	27	0.5**	10.9
Structuring the Child's Day
Percentage of Parents Who Set a Regular Bedtime for Child	59.4	58.2	1.3	2.5
Percentage of Parents and Children Who Have Regular Bedtime Routines	69.3	68.6	0.7	1.4
Parent-Child Activities and Learning Support
HOME: Support of Language and Learning	10.6	10.4	0.2**	9.9
Parent-Child Play	4.4	4.3	0.1*	9.1
Quality of Assistance During Parent- Child Puzzle Challenge Taske	3.6	3.5	0.1*	9
Percentage of Parents Who Read to Child Every Day	56.8	52	4.9**	9.7
Percentage of Parents Who Regularly Read to Child at Bedtime	32.3	29.2	3.1	6.8
Internal Home Environment
HOME: Internal Physical Environment	7.8	7.8	0	-0.3
Sample Size
Parent Interview	1,107	1,003	2,110
Parent-Child Interactions	874	784	1,658
SOURCE: Parent interviews, interviewer observations, and assessments of semistructured parent-child interactions conducted when children were approximately 36 months old. NOTE: All impact estimates were calculated using regression models, where each site was weighted equally. aA participant is defined as a program group member who received more than one Early Head Start home visit, met with an Early Head Start case manager more than once, received at least two weeks of Early Head Start center-based care, and/or participated in Early Head Start group parent-child activities.(back) bThe control group mean is the mean for the control group members who would have participated in Early Head Start if they had been assigned to the program group instead. This unobserved mean was estimated as the difference between the program group mean for participants and the impact per participant.(back) cThe estimated impact per participant is measured as the estimated impact per eligible applicant divided by the proportion of program group members who participated in Early Head Start services (which varied by site). The estimated impact per eligible applicant is measured as the difference between the regression-adjusted means for all program and control group members.(back) dThe effect size was calculated by dividing the estimated impact per participant by the standard deviation of the outcome measure for the control group times 100 (that is, it is the impact expressed as a percentage of the standard deviation).(back) eBehaviors were observed during the videotaped parent-child puzzle task and coded on a seven-point scale.(back) Significantly different from zero at the .10 level, two-tailed test. Significantly different from zero at the .05 level, two-tailed test. **Significantly different from zero at the .01 level, two-tailed test.

c. Negative Aspects of Parenting Behavior

Continuing the pattern observed at age 2, Early Head Start had few impacts on insensitivity, hostility toward the child, and punishment at age 3 (see Box V.6 and Table V.6). Early Head Start parents were less detached in semistructured play than control-group parents, and the proportion of Early Head Start parents who reported spanking the child in the past week was lower than for control-group parents. There were no program effects on ratings of intrusiveness or negative regard toward the child in the semistructured play setting or on detachment or intrusiveness during the parent-child puzzle challenge task. In addition, there was no difference between Early Head Start and control-group families in the amount of harshness expressed toward the child during the parent interview (HOME harshness subscale), consistent with the findings at age 2. Average levels of insensitivity, hostility, and punitive behavior were relatively low among both program- and control-group parents.

The growth curve analyses for detachment, intrusiveness, and negative regard outcomes show similar patterns (see Appendix D.5). All three of these negative behaviors declined as children developed over the two-year period from approximately 1 to 3 years of age, and for the most part, impacts were not significant at any age. Control group parents were higher in detachment than program parents when children were 15 months, and their decrease over time was somewhat greater than was the program parents’ decrease (that is, the programs’ impact on slopes was significant). No impact on change was found for either intrusiveness or negative regard.

Early Head Start parents were less likely to report spanking during the previous week, sustaining a similar finding at age 2. The reported reduction in the use of physical punishment at ages 2 and 3 is consistent with findings discussed in the next section about Early Head Start impacts on parents’ knowledge of discipline strategies.

2. Parenting Knowledge

Our assessment of parenting knowledge at age 3 was more limited than at age 2. Parenting knowledge is not always consistent with behavior. Therefore, in general, we focused the age 3 assessments on a broader range of child development outcomes and parenting behaviors than was true at age 2. It seemed likely that after two or three years of family enrollment, programs would expect behavioral changes to be emerging, and would place greater importance on them than on indicators of knowledge. We limited the measures of parenting knowledge to two important topics: safety practices with respect to child car seats and discipline strategies for common parent-child conflict situations.

Early Head Start had no impact on car seat safety practices, with about 70 percent of both program and control families reporting that they regularly used a car seat for their young children (see Box V.7 and Table V.7). At age 2, we also found no Early Head Start impacts on regular use of car seats.

In response to questions about how they would handle four common parent-child conflict situations (temper tantrums, playing with breakables, refusing to eat, and hitting the parent in anger), Early Head Start parents were less likely to report that they would physically punish their 3-year-old children or threaten physical punishment. At age 2, we found a similar reduction in physical punishment as a discipline strategy. However, in contrast to the findings at age 2, Early Head Start had no impact on the proportion of parents suggesting other discipline strategies, including positive discipline strategies, such as preventing certain situations, distracting the child, and talking to or explaining consequences to the child at age 3. The percentage of parents who suggested only mild discipline strategies (including all discipline strategies except shouting, threatening, or using physical punishment) was significantly higher among Early Head Start parents. Similarly, the most severe discipline strategy mentioned tended to be more severe among control group than program families.

BOX V.6

MEASURES OF NEGATIVE ASPECTS OF PARENTING BEHAVIOR

Parent Behavior during Parent-Child Semistructured Play – measures the parent’s behavior with the child during a semistructured play task. The parent and child were given three bags of interesting toys and asked to play with the toys in sequence. The semistructured play task was videotaped, and child and parent behaviors were coded by child development researchers according to strict protocols (see Appendix C). Four aspects of the parent’s behavior with the child were rated on a seven-point scale, with three related to negative parenting behavior:

Detachment – measures the extent to which the parent is inattentive to the child, inconsistently attentive, or interacts with the child in an indifferent manner. For example, the parent may be inattentive, perfunctory, or cold when interacting with the child, may not respond to the child’s talk or expressions, or may not try to engage the child with the new toys.

Intrusiveness – measures the extent to which the parent exerts control over the child rather than acting in a way that recognizes and respects the validity of the child’s perspective. Higher scores on intrusiveness indicate that the parent controlled the play agenda, not allowing the child to influence the focus or pace of play, grabbing toys away from the child, and not taking turns in play with the child.

Negative Regard – measures the parent’s expression of discontent with, anger toward, disapproval of, or rejection of the child. High scores on negative regard indicate that the parent used a disapproving or negative tone, showed frustration, anger, physical roughness, or harshness toward the child, threatened the child for failing at a task or not playing the way the parent desired, or belittled the child.

Detachment – measures the extent to which the parent is inattentive to the child, or interacts in a perfunctory or indifferent manner. For example, the parent may be inattentive, perfunctory, or cold when interacting with the child, may not respond to the child’s talk or expressions, or may not try to engage the child with the new toys.

Intrusiveness – measures the degree to which the parent controls the child rather than recognizing and respecting the validity of the child’s independent efforts to solve the puzzle. For example, a parent behaving intrusively may complete the puzzle for the child or offer rapid, frequent instructions.

Harshness – measures harsh or punitive parenting behavior observed during the home interview. Items in this subscale are based entirely on interviewer observations of the parent and child during the interview, and include whether the parent scolded the child, physically restrained the child, or slapped or spanked the child. For this subscale (but not for the total HOME score), items were reverse-coded so that higher scores indicate more observed harsh behavior. Scores can range from 0, if no harsh behavior was observed, to 3, if the three types of harsh behavior were observed.

Spanked Child in Previous Week – measures parent’s report that she used physical punishment in the previous week by spanking the child.

TABLE V.6

IMPACTS ON NEGATIVE ASPECTS OF PARENTING BEHAVIOR
Outcome	Program Group Participantsa	Control Groupb	Estimated Impact per Participantc	Effect Sized
Insensitivity
Detachment During Parent-Child Semistructured Playe	1.2	1.3	-0.1*	-9
Intrusiveness During Parent-Child Semistructured Playe	1.6	1.6	0	-5.5
Detachment During Parent-Child Puzzle Challenge Taskf	1.6	1.6	0	-0.2
Intrusiveness During Parent-Child Puzzle Challenge Taskf	2.7	2.7	-0.1	-5.8
Hostility and Punishment
Negative Regard During Parent- Child Semistructured Playe	1.3	1.3	0	-1.6
Home Observation for Measurement of the Environment (HOME): Harshnessg	0.3	0.3	0	2.1
Percentage of Parents Who Spanked the Child During the Previous Week	46.7	53.8	-7.1***	-14.2
Sample Size
Parent Interview	1,107	1,003	2,110
Parent-Child Interactions	874	784	1,658
SOURCE: Parent interviews, interviewer observations, and assessments of semistructured parent-child interactions conducted when children were approximately 36 months old. NOTE: All impact estimates were calculated using regression models, where each site was weighted equally. aA participant is defined as a program group member who received more than one Early Head Start home visit, met with an Early Head Start case manager more than once, received at least two weeks of Early Head Start center-based care, and/or participated in Early Head Start group parent-child activities.(back) bThe control group mean is the mean for the control group members who would have participated in Early Head Start if they had been assigned to the program group instead. This unobserved mean was estimated as the difference between the program group mean for participants and the impact per participant.(back) cThe estimated impact per participant is measured as the estimated impact per eligible applicant divided by the proportion of program group members who participated in Early Head Start services (which varied by site). The estimated impact per eligible applicant is measured as the difference between the regression-adjusted means for all program and control group members.(back) dThe effect size was calculated by dividing the estimated impact per participant by the standard deviation of the outcome measure for the control group times 100 (that is, it is the impact expressed as a percentage of the standard deviation).(back) eBehaviors were observed during the videotaped parent-child semistructured play task and coded on a seven-point scale. (back) fBehaviors were observed during the videotaped parent-child puzzle challenge task and coded on a seven-point scale.(back) gBehaviors were observed during the HOME assessment and rated on a yes/no scale by the interviewer/assessor.(back) Significantly different from zero at the .10 level, two-tailed test. Significantly different from zero at the .05 level, two-tailed test. **Significantly different from zero at the .01 level, two-tailed test.

BOX V.7

MEASURES OF PARENTING KNOWLEDGE ABOUT SAFETY AND DISCIPLINE STRATEGIES

Always Uses Car Seat for Child – measures whether the parent usually uses a car seat or booster seat when taking the child in a car, and whether the child usually sits in the back seat. The small proportion (6 percent) of parents who said they never use a car were coded as using safe practices for this measure.

Discipline Strategies – measures the parent’s strategies for handling four different potential conflict situations with the child: (1) the child keeps playing with breakable things; (2) the child refuses to eat; (3) the child throws a temper tantrum in a public place; and (4) the child hits the parent in anger. Parents provided open-ended answers to how they would respond to each of the four situations, and these responses were classified into the types of discipline strategies, which were coded as binary variables. A parent received a “1” for each strategy that was ever mentioned. In addition, we created the following composite measures:

Mild Discipline – binary variable indicates parents who mentioned only the following types of responses for each situation: prevent the situation; distract the child; remove the child or object; talk to the child or explain the issue; ignore the behavior; put the child in time out; send the child to his or her room; threaten to take away treats or threaten time out; or tell child “No.”

Index of Severity of Discipline Strategies – measures the degree of harshness of discipline strategies suggested. An individual’s score on this index ranges from 1 to 5, and is determined by the harshest strategy that was suggested in response to any of the three conflict situations. Thus, parents who said they would use physical punishment receive a 5; those who did not suggest physical punishment but did say they would shout at the child receive a 4; those whose harshest response was to threaten the child with punishment receive a 3; those who suggest sending the child to his or her room, ignoring the behavior, threatening time out or loss of treats, or saying “No!” receive a 2; and those who suggested only preventing the situation or distracting the child, removing the child or object, talking to the child, or putting the child in time out receive a 1.

TABLE V.7

IMPACTS ON PARENTING KNOWLEDGE: SAFETY AND DISCIPLINE STRATEGIES
Outcome	Program Group Participantsa	Control Groupb	Estimated Impact per Participantc	Effect Sized
Safety Practices
Percentage of Parents Who Always Use Car Seat for Child	69.8	70.8	-0.9	-2
Discipline Strategies
Percentage of Parents Who Suggested Responses to the Hypothetical Situations with Child:
Prevent or distract	70.6	69.3	1.3	2.8
Remove child or object	80.7	81.3	-0.5	-1.4
Talk and explain	70.7	69.1	1.7	3.6
Time out	27	26.9	0.2	0.3
Threaten or command	9.8	13.3	-3.5**	-10.3
Shout	8.7	8.3	0.4	1.4
Physical punishment	46.3	51.1	-4.8**	-9.6
Percentage of Parents Suggesting Only Mild Responses to the Hypothetical Situationse	44.7	40.5	4.2*	8.5
Index of Severity of Discipline Strategies Suggestedf	3.4	3.5	-0.2**	-11
Sample Size	1,107	1,003	2,110
SOURCE: Parent interviews conducted when children were approximately 36 months old. NOTE: All impact estimates were calculated using regression models, where each site was weighted equally. aA participant is defined as a program group member who received more than one Early Head Start home visit, met with an Early Head Start case manager more than once, received at least two weeks of Early Head Start center-based care, and/or participated in Early Head Start group parent-child activities.(back) bThe control group mean is the mean for the control group members who would have participated in Early Head Start if they had been assigned to the program group instead. This unobserved mean was estimated as the difference between the program group mean for participants and the impact per participant.(back) cThe estimated impact per participant is measured as the estimated impact per eligible applicant divided by the proportion of program group members who participated in Early Head Start services (which varied by site). The estimated impact per eligible applicant is measured as the difference between the regression-adjusted means for all program and control group members.(back) dThe effect size was calculated by dividing the estimated impact per participant by the standard deviation of the outcome measure for the control group times 100 (that is, it is the impact expressed as a percentage of the standard deviation).(back) eParents were classified as suggesting only mild discipline if their responses to the four discipline situations included only the following: prevent or distract, remove child or object, talk and explain, time out, ignore child, send the child to his or her room, threaten time out or loss of treats, or tell the child “No.”(back) fThe Index of Severity of Discipline Strategies is based on a hierarchy of discipline practices, from talk and explain, remove child or object, time out, or prevent/distract (1) through physical punishment (5). The most severe approach suggested is used to code this scale.(back) Significantly different from zero at the .10 level, two-tailed test. Significantly different from zero at the .05 level, two-tailed test. **Significantly different from zero at the .01 level, two-tailed test.

Overall, the pattern of findings suggests that Early Head Start parents were less likely to consider physical punishment as an appropriate response to hypothetical discipline situations, but the program did not significantly increase the proportion of parents suggesting more-positive discipline practices, as had been the case at age 2. When children were 3, a sizeable majority of both Early Head Start and control group parents specified more-positive techniques (for example, approximately 70 percent of parents suggested preventing the situation, distracting the child, or talking to the child and explaining consequences).

3. Summary of Impacts on Parenting

Early Head Start had significant impacts on several aspects of emotionally supportive parenting and support for children’s language development and learning when children were 3 years old, measured in a variety of ways (parent self-report, interviewer observation, and structured coding of videotaped parent-child interactions). Early Head Start parents provided more-positive contexts for children’s development through their more-emotionally supportive interactions (observed in the HOME and semistructured play), and cognitively stimulating interactions (including cognitively stimulating assistance in the puzzle challenge, increased incidence of daily parent-child reading and play activities, and the availability of cognitively stimulating toys and materials and interactions as observed in the HOME language and learning scale).

The program had fewer impacts on insensitivity and hostility, but appears to have reduced the use of physical punishment. Early Head Start parents showed less observed detachment during the videotaped semistructured play task, were less likely to report spanking in the previous week, and were less likely to suggest physical punishment as a response to hypothetical discipline situations.

Some domains of parenting were not significantly affected by participation in Early Head Start, including (1) structuring the child’s day by having a consistent bedtime and bedtime routine (over 40 percent of children in both groups did not have a regular bedtime); (2) structuring of the internal physical environment (both groups averaged 7.8 out of 9 possible points, suggesting that both had reasonably organized home environments); (3) regular use of child car safety seats (about 70 percent in both groups); (4) use of positive hypothetical discipline strategies in response to parent-child conflict; and (5) several aspects of insensitivity and hostility toward the child during semi-structured parent-child tasks, and observed harshness during the home interview, which were low on average for parents in both the program and control groups.

Overall, the pattern of impacts on parenting when children were 3 years old was generally very consistent with the pattern of impacts found when children were 2 years old. Moreover, the results suggest that Early Head Start was successful in influencing some of the major categories of parenting that are important for children’s well-being and school readiness. Early Head Start parents were more emotionally supportive, were more likely to read regularly to their children, provided more stimulating home environments, and were less likely to use physical punishment (both actual and hypothetical).

E. OVERALL IMPACTS ON FAMILY WELL-BEING

1. Parents’ Health and Mental Health and Family Functioning

The relatively high level of health services available in communities and the absence of program impacts on receipt of family health services lead us not to expect impacts of Early Head Start on parents’ physical health. We also expected small or no impacts on mental health and family functioning because infant and parent mental health services were often lacking in the communities and because the programs did not have a significant impact on the receipt of parent mental health services. Nevertheless, we examined these outcomes because of their importance to the parent’s ability to function as a parent and provider.

Early Head Start had no impact on parents’ reported levels of health status, mental health, parenting stress, or family conflict at the time children were 3 years old. Favorable impacts on parenting-related stress and negative feelings and on family conflict measured when children were 2 years old did not persist a year later (see Box V.8 and Table V.8). In growth curve analyses, we see that levels of parental distress declined at about the same rates for both program and control parents, although the program group levels were consistently lower than those of the control group (Appendix D.5).

The program had no impact on reported levels of parent-child dysfunctional interaction (Table V.8). At age 3, there were no overall impacts on family conflict, although favorable impacts were found at age 2 (a difference perhaps due to the somewhat different sample included in the growth curve analysis). Growth curve analysis showed an interesting pattern that did not emerge in any of the other outcomes that we could examine over time: Family conflict declined for the program group, while staying about the same across time for the control parents (see Appendix D.5). The difference in the two slopes (rates of change over time) was statistically significant, indicating that family conflict declined at a somewhat faster rate for the program than for the control group.

2. Economic Self-Sufficiency

Early Head Start had favorable impacts on the level of self-sufficiency activities of parents, measured as the proportion ever engaging in the activity in the eight quarters after enrollment in Early Head Start, or in the average hours per week that they engaged in the activity (see Box V.9 and Table V.9). Participation rates in any activity (education, job training, or employment) were higher for parents in the program group than for those in the control group in the third through eighth quarters after enrollment (Figure V.1).

Impacts on education activities were larger than impacts on employment activities. A larger proportion of parents in the program group participated in education or job training activities in the third through eighth quarters after enrollment (Figure V.2). Approximately 20 to 30 percent of Early Head Start parents participated in education or training activities in any quarter, but over the eight quarters, 60 percent of Early Head Start parents participated in an education or training activity (Table V.9).

The overall education impacts generally reflected an increase in high school attendance. No significant impacts were found overall in rates of attendance in other educational programs. Given the persistent impact on high school attendance at 15 months and 26 months after enrollment, we expected to find an impact on the highest grade completed, GED certificate or high school diploma by 26 months after enrollment, but there were no impacts of Early Head Start on attainment of these credentials (Table V.10).

Employment rates were much higher than the percentage in education or training activities. Overall, more than 85 percent of Early Head Start parents were employed at some point during the 26-month follow-up period (see again, Table V.9), while on a quarterly basis, employment rates increased from about 40 percent to 65 percent (Figure V.3). Nevertheless, employment rates seemed to be responding to the strong economy and welfare policies encouraging work rather than the influence of Early Head Start, since employment rates for the program and control groups were not statistically different in seven out of eight quarters after enrollment.

BOX V.8

MEASURES OF THE PARENT’S HEALTH, MENTAL HEALTH, AND FAMILY FUNCTIONING

Health Status – measures the parent’s perception of own health status on a 5-point scale, where 1 indicates poor health and 5 indicates excellent health.

Parenting Stress Index – Short Form (PSI-SF) – measures the degree of stress in parent-child relationships stemming from three possible sources: the child’s challenging temperament, parental depression, and negatively reinforcing parent-child interactions (Abidin 1995). We included two subscales of the PSI-SF:

Parental Distress – measures the level of distress the parent is feeling in his or her role as a parent stemming from personal factors, including a low sense of competence as a parent, stress because of perceived restrictions stemming from parenting, depression, and lack of social support.

The parent answers whether he or she agrees or disagrees with statements such as, “You often have the feeling that you cannot handle things very well,” and “You feel trapped by your responsibilities as a parent,” and “You feel alone and without friends.” Item responses are coded on a 5-point scale, with 5 indicating high levels of parental distress. Scores on the 12-item subscale can range from 12 to 60.

Parent-Child Dysfunctional Interaction – measures the parent’s perception that the child does not meet the parent’s expectations and interactions with the child are not reinforcing the parent. The parent may perceive that the child is abusing or rejecting the parent or that the parent feels disappointed in or alienated from the child.

The parent answers whether he or she agrees or disagrees with statements such as, “Your child rarely does things for you that make you feel good,” and “Most times you feel that your child does not like you and does not want to be close to you,” and “Your child seems to smile less than most children.” Item responses are coded on a 5-point scale, with 5 indicating high levels of parent-child dysfunctional interaction. Scores on the 12-item subscale can range from 12 to 60.

Center for Epidemiological Studies Depression Scale – Short Form (CESD-SF) – measures symptoms of depression (Ross et al. 1983). It does not indicate a diagnosis of clinical depression, but it does discriminate between depressed patients and others. The scale includes 12 items taken from the full, 20-item CESD scale (Radloff 1977). Respondents were asked the number of days in the past week they had a particular symptom. Symptoms include poor appetite, restless sleep, loneliness, sadness, and lack of energy. Items coded on a four-point scale from rarely (0) to most days (3). Scores on the scale range from 0 to 36.

Severe Depressive Symptoms – percentage of parents whose scores on the CESD-SF were 15 or higher. This corresponds to a score of 25 or higher on the full CES-D, which is used to indicate high levels of depressive symptoms (Seligman 1993).

Family Environment Scale – measures the social environments of families along 10 key dimensions, including family relationships (cohesion, expressiveness, and conflict); emphases within the family on aspects of personal development that can be supported by families (for example, achievement orientation; independence); and maintenance of the family system (organization and control) (Moos and Moos 1976). We measured one dimension:

Family Conflict – measures the extent to which the open expression of anger and aggression and generally conflictual interactions are characteristic of the family. Parents respond to items on a 4-point scale, where 4 indicates higher levels of agreement with statements such as, “We fight a lot,” and “We hardly ever lose our tempers.” Items were recoded and averaged so that 4 indicates high levels of conflict.

TABLE V.8

IMPACTS ON PARENT HEALTH, MENTAL HEALTH, AND FAMILY FUNCTIONING
Outcome	Program Group Participantsa	Control Groupb	Estimated Impact per Participantc	Effect Sized
Parent's Physical Health
Parent's Health Status	3.4	3.5	-0.1	-4.9
Parent's Mental Health
Parenting Stress Index (PSI): Parental Distress	24.7	25.5	-0.7	-7.7
PSI: Parent-Child Dysfunctional Interaction	17.8	17.8	0	-0.2
CES-Depression Scale (CES-D; short form)	7.4	7.7	-0.3	-3.7
CES-D: Severe Depressive Symptoms	14.5	14.8	-0.3	-0.8
Family Functioning
Family Environment Scale- Family Conflict (Average Score)	1.7	1.7	0	-4.3
Sample Size	1,107	1,003	2,110
SOURCE: Parent interviews conducted when children were approximately 36 months old. NOTE: All impact estimates were calculated using regression models, where each site was weighted equally. aA participant is defined as a program group member who received more than one Early Head Start home visit, met with an Early Head Start case manager more than once, received at least two weeks of Early Head Start center-based care, and/or participated in Early Head Start group parent-child activities.(back) bThe control group mean is the mean for the control group members who would have participated in Early Head Start if they had been assigned to the program group instead. This unobserved mean was estimated as the difference between the program group mean for participants and the impact per participant.(back) cThe estimated impact per participant is measured as the estimated impact per eligible applicant divided by the proportion of program group members who participated in Early Head Start services (which varied by site). The estimated impact per eligible applicant is measured as the difference between the regression-adjusted means for all program and control group members.(back) dThe effect size was calculated by dividing the estimated impact per participant by the standard deviation of the outcome measure for the control group times 100 (that is, it is the impact expressed as a percentage of the standard deviation).(back) Significantly different from zero at the .10 level, two-tailed test. Significantly different from zero at the .05 level, two-tailed test. **Significantly different from zero at the .01 level, two-tailed test.

BOX V.9

MEASURES OF ECONOMIC SELF-SUFFICIENCY

Education - Parents were asked about education and job training programs that they had participated in during the follow-up period, including the start and end dates for those activities and the typical hours per day and days per week they spent in those activities. From that information we constructed a weekly timeline of education/training activities and indicators of whether parents were in education/training activities during each of the first five quarters following random assignment. We also combined information on hours per day and days per week for all education/training activities to obtain the average hours per week parents spent in education/training activities during the 24-month follow-up period. Averages include zero hours.

Employment - Parents were asked about jobs that they had held during the follow-up period, including the start and end dates for those jobs and the typical hours per week they worked in those jobs. From that information we constructed a weekly timeline of employment activities and indicators of whether parents were employed during the first five quarters following random assignment. We also combined information on hours per day and days per week for all jobs to obtain the average hours per week parents spent in employment during the 24-month follow-up period. Averages include zero hours.

Any Activity - The weekly histories of education/training activities and jobs were combined to create a timeline of participation in any of these self-sufficiency activities and indicators of whether parents participated in any self-sufficiency activities during each of the first five quarters following random assignment. We also added the average number of hours spent in education/training and jobs to get the average number of hours parents spent in any self-sufficiency activities during the first 24 months after random assignment. Averages include zero hours.

Welfare Program Participation - Parents were asked about their receipt of AFDC/TANF cash assistance, food stamps, general assistance, and SSI or SSA benefits, including the amount they received and the months during which they received it. From this information we created a monthly timeline of welfare receipt and a timeline of AFDC/TANF cash assistance receipt, as well as indicators of welfare receipt and AFDC/TANF cash assistance receipt during each of the first five quarters after random assignment. We also added the welfare benefit amounts to obtain the total amount of welfare benefits received, the total amount of food stamps received, and the total amount of AFDC/TANF cash assistance received during the 24-month follow-up period. Averages include zero benefit amounts.

Family Income and Resources - In the Parent Services Follow-Up Interviews, parents were asked about their family income during the last year. We compared information on their annual income and the number of children in their family with federal poverty levels to create an indicator of whether or not the family’s income during the year prior to the third follow-up was above the poverty level or not. Family resources were assessed using the Family Resource Scale (Dunst and Leet 1987) plus items assessing additional resources, in which parents rated the adequacy of 39 specific resources on a scale of 1 (not at all adequate) to 5 (almost always adequate). The item values were summed to obtain a total family resources scale value.

Subsequent Childbearing - In the Parent Services Follow-Up Interviews, parents were asked whether they had borne any children since the previous interview and if so, the birth dates. We used this information to create an indicator of whether the parent had any births and the timing of the births since the enrollment date. For mothers who entered the program in pregnancy, the birth of the focus child is excluded from these counts.

TABLE V.9

IMPACTS ON SELF-SUFFICIENCY ACTIVITIES
Outcome	Program Group Participantsa	Control Groupb	Estimated Impact per Participantc	Effect Sized
Any Self-Sufficiency Activities
Percentage of Parents Ever Employed or in an Education or Job-Training Program in First 26 Months	93.9	90.5	3.4**	11.1
Average Hours per Week Employed at All Jobs and in Any Education or Training in First 26 Months	22.3	20.9	1.5*	9.3
Employment Activities
Percentage of Parents Ever Employed in First 26 Months	86.8	83.4	3.4*	9
Average Hours per Week Employed at All Jobs in First 26 Months	17.1	17.1	0.1	0.5
Education Activities
Percentage of Parents Who Ever Participated in an Education or Training Program in First 26 Months	60	51.4	8.6***	17.2
Average Hours per Week in an Education Program During First 26 Months	4.6	3.4	1.2***	18.4
Sample Size	1,139	1,097	2,236
SOURCE: Parent services follow-up interviews conducted an average of 26 months after random assignment. NOTE: All impact estimates were calculated using regression models, where each site was weighted equally. aA participant is defined as a program group member who received more than one Early Head Start home visit, met with an Early Head Start case manager more than once, received at least two weeks of Early Head Start center-based care, and/or participated in Early Head Start group parent-child activities.(back) bThe control group mean is the mean for the control group members who would have participated in Early Head Start if they had been assigned to the program group instead. This unobserved mean was estimated as the difference between the program group mean for participants and the impact per participant.(back) cThe estimated impact per participant is measured as the estimated impact per eligible applicant divided by the proportion of program group members who participated in Early Head Start services (which varied by site). The estimated impact per eligible applicant is measured as the difference between the regression-adjusted means for all program and control group members.(back) dThe effect size was calculated by dividing the estimated impact per participant by the standard deviation of the outcome measure for the control group times 100 (that is, it is the impact expressed as a percentage of the standard deviation).(back) Significantly different from zero at the .10 level, two-tailed test. Significantly different from zero at the .05 level, two-tailed test. **Significantly different from zero at the .01 level, two-tailed test.

FIGURE V.1

IMPACTS ON ANY SELF-SUFFICIENCY ACTIVITY,
BY QUARTER
[D]
Source: Parent Services Follow-Up Interviews conducted approximately 7, 16, and 26 months after random assignment. Notes: All percentages are regression-adjusted means estimated using models that weight each site equally. The differences between program and control families are estimated impacts per participant. * Program impact is significantly different from zero at the .10 level, two-tailed test. Program impact is significantly different from zero at the .05 level, two-tailed test. *Program impact is significantly different from zero at the .01 level, two-tailed test.

FIGURE V. 2

IMPACTS ON PARTICIPATION IN EDUCATION AND TRAINING PROGRAMS,
BY QUARTER
[D]
Source: Parent Services Follow-Up Interviews conducted approximately 7, 16, and 26 months after random assignment. Notes: All percentages are regression-adjusted means estimated using models that weight each site equally. The differences between program and control families are estimated impacts per participant. * Program impact is significantly different from zero at the .10 level, two-tailed test. Program impact is significantly different from zero at the .05 level, two-tailed test. *Program impact is significantly different from zero at the .01 level, two-tailed test.

TABLE V.10

IMPACTS ON EDUCATION ACTIVITIES AND CREDENTIALS
Outcome	Program Group Participantsa	Control Groupb	Estimated Impact per Participantc	Effect Sized
Types of Education Activities
High School	13.6	9	4.6***	16.2
High School or Alternative	14.3	10.3	4.0***	13.2
Adult Basic Education	4.3	3.7	0.7	3.7
English as a Second Language	3.5	2.5	1	7
GED Preparation	9.8	8.5	1.3	4.6
Any Vocational Education	20	17.3	2.7	7.3
Two-Year College	10.9	10.4	0.5	1.8
Four-Year College	6.4	6.1	0.3	1.5
Degrees and Credentials Received
Highest Grade Completed at Second Followup	11.6	11.6	-0.1	-3
GED Certificate	10	11.1	-1.1	-3.5
High School Diploma	50.3	49.5	0.8	1.6
Vocational, Business, or Secretarial Diploma	17.6	17.4	0.2	0.6
Associate's Degree	3.6	4.8	-1.2	-6
Bachelor's Degree	4.4	5.9	-1.6	-7.1
Sample Size	1,139	1,097	2,236
SOURCE: Parent services follow-up interviews completed an average of 26 months after random assignment. NOTE: All impact estimates were calculated using regression models, where each site was weighted equally. aA participant is defined as a program group member who received more than one Early Head Start home visit, met with an Early Head Start case manager more than once, received at least two weeks of Early Head Start center-based care, and/or participated in Early Head Start group parent-child activities.(back) bThe control group mean is the mean for the control group members who would have participated in Early Head Start if they had been assigned to the program group instead. This unobserved mean was estimated as the difference between the program group mean for participants and the impact per participant.(back) cThe estimated impact per participant is measured as the estimated impact per eligible applicant divided by the proportion of program group members who participated in Early Head Start services (which varied by site). The estimated impact per eligible applicant is measured as the difference between the regression-adjusted means for all program and control group members.(back) dThe effect size was calculated by dividing the estimated impact per participant by the standard deviation of the outcome measure for the control group times 100 (that is, it is the impact expressed as a percentage of the standard deviation).(back) Significantly different from zero at the .10 level, two-tailed test. Significantly different from zero at the .05 level, two-tailed test. **Significantly different from zero at the .01 level, two-tailed test.

FIGURE V.3

IMPACTS ON EMPLOYMENT RATES, BY QUARTER
[D]
Source: Parent Services Follow-Up Interviews conducted approximately 7, 16, and 26 months after random assignment. Notes: All percentages are regression-adjusted means estimated using models that weight each site equally. The differences between program and control families are estimated impacts per participant. * Program impact is significantly different from zero at the .10 level, two-tailed test. Program impact is significantly different from zero at the .05 level, two-tailed test. *Program impact is significantly different from zero at the .01 level, two-tailed test.

Consistent with the employment findings, welfare receipt went down for both program and control group families over time, but Early Head Start had no impact on receipt of welfare or the amount of welfare benefits received over the 26-month period after enrollment (Table V.11, Figure V.4). Early Head Start had no impact on the percentage of families with income above the poverty line at 26 months after enrollment (Table V.12).

Early Head Start mothers were somewhat less likely than control mothers to experience subsequent births during the first two years after they enrolled (Table V.12). In addition, Figure V.5 shows that in quarters 6, 7, and 8 after enrollment, the percentage of program mothers who had given birth to a child other than the focus child since enrollment was significantly lower than the percentage of control group mothers. This delay in subsequent births may have implications for parents’ progress toward self-sufficiency and mental health, as shorter intervals between births can negatively affect parents’ well-being and make it more difficult for them to engage in self-sufficiency activities.

F. HOW IMPACTS ON PARENTING AT AGE 2 MAY HAVE INFLUENCED CHILD OUTCOMES AT AGE 3

Many of the Early Head Start programs believed that an important route to enhancing children’s well-being was to support a strong parent-child relationship. Thus, these programs hoped that impacts on parenting behavior would, over time, yield benefits for children’s cognitive and social-emotional development. To explore how this theory of change might be working during the three years of the evaluation, we conducted analyses that examined the association between child impacts measured at the time of the child’s 36-month birthday and parenting impacts measured a year earlier. These “mediated” analyses controlled for many demographic characteristics that could also affect the size of the impacts, yet must be interpreted with caution because of possible uncontrolled confounding and simultaneity (see Chapter II and Appendix D.9 for details on the design and results of these analyses). Specifically, we developed the following hypotheses as we developed the statistical models for the mediated analyses:

Several aspects of parenting measured when the children were 2 years old are likely to support and stimulate children’s cognitive and language development. Thus, 2-year impacts on parent supportiveness, cognitive stimulation during parent-child play, and support for language and learning, as well as parents’ reading every day may at least partially mediate impacts on the Bayley MDI and the PPVT-III at age 3.
How parents interact with their children is likely to affect how the children respond to them. When parents display greater warm sensitivity, emotional responsivity, and support for language and learning when their children are 2, their children may be more likely to initiate and maintain higher levels of interaction (engagement) with their parents in the play situation when children are a year older. Similarly, if parents act more detached, children may show lower levels of engagement.
When children were 2, if their parents behaved in ways that supported language and learning, had greater knowledge of child development, and felt more confident in their role as parent (that is, had lower levels of parenting distress), these children might be expected to be better able to focus while playing with objects (that is, show higher sustained attention).
Parent discipline styles often are considered to be important mediators of many aspects of children’s behavior and development. The literature suggests an association between physical punishment and aggressive behavior in children. We therefore hypothesized that when the program has an impact on reducing parent spanking at age 2, children will show lower levels of aggressive behavior when they are 3. Other aspects of parenting that are likely to be associated with lower aggressiveness at age 3 include a stable and warm home atmosphere, which could be reflected in parents providing regular bedtimes, being warm and supportive, and having lower levels of parenting distress at age 2.
Finally, we hypothesize that the favorable 2-year impacts on parents’ spanking, parental distress, intrusiveness, and warm sensitivity (that is, lower levels of spanking, distress, and intrusiveness and increased warm sensitivity) will be associated with lower levels of children’s negativity toward their parents when they are 3.

TABLE V.11

IMPACTS ON WELFARE PROGRAM PARTICIPATION
Outcome	Program Group Participantsa	Control Groupb	Estimated Impact per Participantc	Effect Sized
Welfare Program Participation
Percentage of Parents Who Received Any Welfare Benefits During First 26 Months	68.1	66.5	1.6	3.5
Total Welfare Benefits Received During First 26 Months	$5,287	$5,548	-$261	-3.5
Percentage of Parents Who Received AFDC or TANF Benefits During First 26 Months	47	44.7	2.3	4.6
Total AFDC or TANF Benefits Received During First 26 Months	$2,142	$2,160	-$19	-0.5
Average Total Food Stamp Benefits Received During First 26 Months	$2,110	$2,079	$30	1.1
Sample Size	1,139	1,097	2,236
SOURCE: Parent services follow-up interviews conducted an average of 26 months after random assignment. NOTE: All impact estimates were calculated using regression models, where each site was weighted equally. aA participant is defined as a program group member who received more than one Early Head Start home visit, met with an Early Head Start case manager more than once, received at least two weeks of Early Head Start center-based care, and/or participated in Early Head Start group parent-child activities.(back) bThe control group mean is the mean for the control group members who would have participated in Early Head Start if they had been assigned to the program group instead. This unobserved mean was estimated as the difference between the program group mean for participants and the impact per participant.(back) cThe estimated impact per participant is measured as the estimated impact per eligible applicant divided by the proportion of program group members who participated in Early Head Start services (which varied by site). The estimated impact per eligible applicant is measured as the difference between the regression-adjusted means for all program and control group members.(back) dThe effect size was calculated by dividing the estimated impact per participant by the standard deviation of the outcome measure for the control group times 100 (that is, it is the impact expressed as a percentage of the standard deviation).(back) Significantly different from zero at the .10 level, two-tailed test. Significantly different from zero at the .05 level, two-tailed test. **Significantly different from zero at the .01 level, two-tailed test. AFDC = Aid to Families with Dependent Children; TANF = Temporary Assistance for Needy Families.

FIGURE V.4

IMPACTS ON AFDC/TANF RECEIPT, BY QUARTER
[D]
Source: Parent Services Follow-Up Interviews conducted approximately 7, 16, and 26 months after random assignment. Notes: All percentages are regression-adjusted means estimated using models that weight each site equally. The differences between program and control families are estimated impacts per participant. * Program impact is significantly different from zero at the .10 level, two-tailed test. Program impact is significantly different from zero at the .05 level, two-tailed test. *Program impact is significantly different from zero at the .01 level, two-tailed test.

TABLE V.12

IMPACTS ON FAMILY INCOME, RESOURCES, AND SUBSEQUENT CHILDBEARING
Outcome	Program Group Participantsa	Control Groupb	Estimated Impact Per Participantc	Effect Sized
Percentage of Families with Income Above the Poverty Line at Third Followup	42.9	43.3	-0.4	-0.8
Total Family Resources Scale First Followup	149.6	148.7	0.9	4.4
Second Followup	152.9	151.6	1.3	6.7
Third Followup	154.8	153.8	1	5.2
Percentage with Any Births (Not Including Focus Child) Within 24 Months After Random Assignmente	22.9	27.1	-4.2*	-9.2
Average Number of Births (Not Including Focus Child)e	0.3	0.3	0	-6.1
Sample Size	918 - 1,139	857 - 1,097	1,775 - 2,236
SOURCE: Parent services follow-up interviews completed an average of 26 months after random assignment. NOTE: All impact estimates were calculated using regression models, where each site was weighted equally. aA participant is defined as a program group member who received more than one Early Head Start home visit, met with an Early Head Start case manager more than once, received at least two weeks of Early Head Start center-based care, and/or participated in Early Head Start group parent-child activities.(back) bThe control group mean is the mean for the control group members who would have participated in Early Head Start if they had been assigned to the program group instead. This unobserved mean was estimated as the difference between the program group mean for participants and the impact per participant.(back) cThe estimated impact per participant is measured as the estimated impact per eligible applicant divided by the proportion of program group members who participated in Early Head Start services (which varied by site). The estimated impact per eligible applicant is measured as the difference between the regression-adjusted means for all program and control group members.(back) dThe effect size was calculated by dividing the estimated impact per participant by the standard deviation of the outcome measure for the control group times 100 (that is, it is the impact expressed as a percentage of the standard deviation).(back) eLength of followup varies among sample members but is the same for program and control group members.(back) Significantly different from zero at the .10 level, two-tailed test. Significantly different from zero at the .05 level, two-tailed test. **Significantly different from zero at the .01 level, two-tailed test.

FIGURE V.5

IMPACTS ON SUBSEQUENT BIRTHS, BY QUARTER
[D]
Source: Parent Services Follow-Up Interviews conducted approximately 7, 16, and 26 months after random assignment. Notes: All percentages are regression-adjusted means estimated using models that weight each site equally. The differences between program and control families are estimated impacts per participant. * Program impact is significantly different from zero at the .10 level, two-tailed test. Program impact is significantly different from zero at the .05 level, two-tailed test. *Program impact is significantly different from zero at the .01 level, two-tailed test.

Table V.13 summarizes the results of mediated analyses for the full sample.6 The shaded rows indicate the 3-year-old child impacts for which we tested mediated models that included, as mediators, the 2-year parenting outcomes that are listed in the first column.

1. Mediators of Cognitive and Language Development

Our analyses suggest that Early Head Start programs may have produced some of their impacts on children at 3 years of age through the impacts on parenting a year earlier. These analyses indicate that children’s scores on the Bayley MDI at 36 months were related to higher levels of parent supportiveness in semistructured play, greater support for cognitive and language development, and daily reading at 2 years of age. In total, the estimates suggest support for the hypothesis that some of the Early Head Start impact on children’s cognitive development could have occurred because of the program’s impacts on parents’ sensitivity and cognitive stimulation in interactions with the child, and their support in the home for the child’s cognitive and language development.7

Estimates also suggest a positive relationship between 36-month PPVT III scores and parent supportiveness in play and support for cognitive and language development. In total, these estimates suggest that part of the Early Head Start impact on children’s receptive language ability at 3 years of age could have emerged because of earlier impacts on the parent’s sensitivity, cognitive stimulation, and support for the child’s language development across a range of parenting situations (during play, through regular daily reading, and during everyday interactions in the home).

TABLE V.13

ESTIMATED MEDIATING EFFECTS OF PARENTING IMPACTS AT AGE 2 ON EARLY HEAD START IMPACTS ON CHILD DEVELOPMENT AT AGE 3
Parenting Outcome	Estimated Effect of Parenting Outcome on Child Outcomea	Significance Level	Percentage of Impact on Child Outcome Associated With Earlier Impact on Mediatorb
36-Month Bayley MDI
Supportiveness: Semistructured Play	+	***	Moderate
HOME Support of Language and Literacy	+	***	Moderate
Read Daily	+	*	Small
36-Month PPVT-III
Supportiveness: Semistructured Play	+	***	Small
HOME Support of Language and Literacy	+	***	Moderate
Read Daily	+	n.s.	Small
36-Month Engagement of Parent inSemistructured Play
Warm Sensitivity: Semistructured Play	+	***	Small
HOME Emotional Responsivity	+	***	Small
HOME Support of Language and Literacy	+	***	Small
Detachment: Semistructured Play	-	n.s.	Small
36-Month Sustained Attention to Objectsin Semistructured Play
Supportiveness: Semistructured Play	+	***	Small
HOME Support of Language and Literacy	+	**	Small
Knowledge of Infant Development	+	**	Small
Parental Distress	-	**	Small
36-Month Negativity Toward Parentin Semistructured Play
Warm Sensitivity	-	***	Small
Physical Punishment in Last Week	+	n.s.	Small
Parental Distress	+	**	Small
Intrusiveness: Semistructured Play	+	***	Small
36-Month Aggressive Behavior
Warm Sensitivity: Semistructured Play	-	***	Small
Physical Punishment in Last Week	+	***	Moderate
Parental Distress	+	***	Large
Regular Bedtime	-	n.s.	Small
a+ Indicates positive association between the 2-year mediator and the 3-year impact. (back) - Indicates an inverse relationship. bSmall: 0 to under 10 percent (back) Moderate: 10 to under 40 percent Large: 40 percent or more * Significantly different from zero at the .10 level, two-tailed [or one-tailed] test. Significantly different from zero at the .05 level, two-tailed [or one-tailed] test. * Significantly different from zero at the .01 level, two-tailed [or one-tailed] test. n.s. = not significant.

2. Mediators of Child Engagement of Parent and Sustained Attention to Objects

For models of positive aspects of children’s social-emotional behavior during semistructured play, we estimated their association with warm and supportive parenting behavior and cognitive stimulation, which together are expected to influence the child’s positive relationship with the parent. The mediated analysis suggests that the Early Head Start programs’ positive impacts on the children’s engagement of the parent in semistructured play at age 3 are consistent with earlier positive program impacts on the parent’s sensitivity during play, responsiveness to the child, and cognitive stimulation and support for language development in the home.

The child’s attention and focus on play at age 3 (sustained attention) is positively related to parents’ sensitivity and cognitive stimulation during semistructured play a year earlier; support for cognitive development and language stimulation in the home environment in the previous year; and the parent’s knowledge of child development measured at 2 years of age. Sustained attention toward objects during play at 3 years also is inversely related to parental distress measured in the previous year. In total, the mediated analysis estimates suggest that part of the positive impact on children’s sustained attention to objects during semistructured play at age 3 could have come about because of earlier favorable program impacts on parent supportiveness in semistructured play, cognitive stimulation and language support in the home environment and knowledge of child development, and through reductions in parental distress.

3. Mediators of Negative Aspects of Children’s Social-Emotional Development

Our analyses indicate that children’s negativity toward their parents in semistructured play at 3 years of age is inversely related to parents’ warm sensitivity during semistructured play observed in the previous year and positively related to levels of parental distress and intrusive behavior during semistructured play measured in the previous year. The relationship between child negativity at 3 and the parent’s use of physical punishment a year earlier is not significant, however. In total, the estimates suggest that part of the reduction in levels of child negativity toward the parent during semistructured play that came about through Early Head Start participation at age 3 might be associated with the increases in parent warmth and sensitivity during play and reductions in parental distress and intrusiveness during play that Early Head Start produced one year earlier.

The estimates of the mediation model of children’s aggressive behavior at 3 years of age and parenting behavior in the previous year indicate that children’s aggression is inversely related to the parents’ warm sensitivity during semistructured play and positively related to the use of physical punishment and levels of parental distress measured in the previous year. The relationship between aggression and the parents’ setting a regular bedtime for the child is not significant, however. In total, the estimates indicate that part of the Early Head Start impact reducing levels of aggression in 3-year-old children may be attributable to the programs’ positive impact on parents’ warm sensitivity toward the child during play and to the programs’ impact in reducing the incidence of physical punishment in the previous year. The relationship between children’s aggressive behavior and earlier levels of parental distress appears fairly large, but the relationship may be overstated because of shared method variance. Part of the correlation may occur because distressed parents may view their children’s behavior more negatively than an outside observer would. Parental distress and child aggression are both tapping a similar dimension of difficulty with child behaviors, and since they are reported by the same person (although at different points in time), the correlation is likely to be high.

4. Synopsis of Estimates from the Mediated Analyses

In summary, the estimates of models relating children’s behavior at age 3 to parenting behavior measured a year earlier in the full sample suggest some support for the theory of change that at least a portion of the Early Head Start programs’ impacts on children could have come about because of earlier favorable changes the program created in parenting behavior. The estimates of the relationships between parenting behavior and children’s outcomes and the Early Head Start program impacts on these outcomes are consistent with the theory, although the models we have estimated are not structural and therefore cannot establish a causal link between the parenting impacts and impacts on children.

G. FATHERHOOD RESEARCH AND LOCAL RESEARCH CONTRIBUTIONS

The Early Head Start father studies have yielded extensive information about the fathers of Early Head Start children, both from the mothers and from the fathers themselves. In Box V.10, we summarize findings related to father presence, participation in their children’s lives, and impacts of the program on fathers’ interactions with their children.

Following the chapter conclusions, we include a series of local research reports that present site-specific findings related to the themes of this chapter—child development outcomes, parent-child relationships (including both mother-child and father-child), self-sufficiency outcomes, and explorations of factors mediating child outcomes. More details on these brief reports can be found in Volume III.

BOX V.10

FATHERS AND FATHER FIGURES IN THE LIVES OF EARLY HEAD START CHILDREN

Fathers are important in Early Head Start programs and most Early Head Start children are likely to have fathers in their lives. Early Head Start programs have increasingly devoted energies to involving men in program activities, and also to encouraging biological fathers and father figures to be more active participants with their children and families. The Early Head Start father studies began at a time when the majority of the research programs had not implemented specific father involvement components and did not target father outcomes as areas of expected change. Direct assessment of fathers and father outcomes were not included in the original evaluation design, but Father Studies were added to the research to provide descriptive information about the role of fathers or father figures (social fathers) in the lives of their children and to explore how father involvement in children’s lives is related to child outcomes. Here we report features of father presence and participation in the lives of Early Head Start children.

Additionally, the growing prominence of father involvement in programs suggests an exploration of program control differences despite the fact that father involvement was not a part of the original evaluation or strongly emphasized when programs started up. Thus, we examined Early Head Start and control-group differences in father presence, father well-being, and father activities with their children in an exploratory manner. Hypotheses about program effects on fathers and father figures are complex given the early stage of father program development, and the joint possibility that programs could increase father involvement in some families and reduce father participation in others in cases where fathers are abusive to children or mothers or unwilling to seek treatment for substance abuse or mental illness. We reported in Chapter IV that Early Head Start fathers had a fairly low incidence of participation in program services which leads to a hypothesis that fathers would not be affected by the Early Head Start program. On the other hand, Early Head Start fathers were significantly more likely to participate in services than control group fathers which justifies the exploratory examination of program and control differences.

Data about fathers were collected from mothers and from fathers. As described in Chapter 2, our findings about fathers are drawn from mother interviews conducted in all 17 sites (at the time of the 14-, 24-, and 36-month birthday-related parent interviews we asked mothers about their child’s father and any father figures) and from father interviews conducted in the 12 father study sites (when the children were approximately 24 and 36 months old). At 24 and 36 months, 7 of the 12 father study sites conducted a videotaped semistructured play task. At 24 months, the father video sites conducted the teaching task, and at 36 months they conducted the father-child puzzle challenge task. The videotaped tasks were conducted and coded using the same procedures as in the main study parent-child tasks. The father study measures and constructed variables are described in Appendix C.

Based on reports from mothers, most Early Head Start children had some contact with their biological father when the children were 36 months old. If the child did not live with his or her biological father, we asked the mother about the nonresident biological father and how often the father saw the child. If the father saw the child a few times per month or more, we categorized the family as having a father who was present in his child’s life. As reported by the mothers when the children were 36 months old, almost 75 percent of children lived with or had contact a few times per month or more with their biological father.

Mothers reported that almost all Early Head Start children had a father or father figure in their lives. When the mother reported that there was a nonresident biological father, we also asked her about any other men who might be “like a father” to the child. If the mother named a father figure, we categorized the family as having a father figure who was present in the child’s life. Close to 90 percent of the children had either a biological father or a father figure in their lives at 36 months. These rates of father presence were consistent with mother reports of father presence when the children were 14 and 24 months old.

When the children were 36 months old, 40 percent of mothers reported that they were married. Just over one-third were married to the child’s biological father and about 6 percent were married to someone else. At enrollment, 26 percent of mothers interviewed at 36 months reported that they were married (to the child’s biological father or someone else) so more mothers reported being married at 36 months than when they began the program. At 36 months, about half of the mothers reported that they were in a relationship with the child’s biological father in which he was either her spouse, live-in partner, or boyfriend.

Rates of biological father and male presence were similar at 36 months across the Early Head Start and control groups. Biological father presence was 73 and 71 percent and male presence was 90 and 89 percent, respectively, for the Early Head Start and control groups. (See table.) At 36 months, rates of marriage to the biological father in the Early Head Start and control groups were about the same (35 and 36 percent, respectively). Similarly, the proportion of biological fathers who were the mothers’ husband, boyfriend, or live-in partner at 36 months did not differ between the Early Head Start and control groups (49 and 51 percent respectively). These results are not surprising because programs worked individually with families which would have led to some increases and some decreases in father presence.

When the children were 36 months old, Early Head Start fathers and father figures in the 12 father study sites reported that they participated in their children’s lives in a variety of ways. Fathers reported participating in a number of activities with their children, including caregiving, engaging in social activities, cognitive activities, and physical play.

Early Head Start programs had several important impacts on father involvement with children. Based on father interview and videotaped interaction data gathered in the father study sites, fathers whose families participated in the Early Head Start program spanked less, were less punitive in discipline practices and were less intrusive in interacting with their children than fathers in the control group (see table). Early Head Start and control-group fathers did not differ in terms of positive discipline strategies, other parenting behaviors and attitudes, father well-being, and the frequency of caregiving, social, cognitive, and physical play activities with their children.1 Although some programs were working with families to increase father involvement with their children, the majority of the programs were at very early stages in these efforts.

Early Head Start children showed significantly more positive behaviors in interaction with their fathers/father figures than control-group children with theirs. In the semistructured play task, Early Head Start children scored higher on engaging their fathers/father figures in play and demonstrated more sustained attention than control-group children. There was not a significant program effect on father reports of children’s aggressive behavior, negative behavior toward the father, or other measures of the child’s behavior during the puzzle challenge task with the fathers.

In summary, most Early Head Start children are likely to have fathers in their lives and Early Head Start is making a positive difference in some aspects of fathering and father-child interaction. The majority of mothers are not married to the focus child’s father. The program and control groups were similar in father presence and marriage of children’s father to their mothers. On the other hand, even though the program is in early stages in implementing intentional father involvement practices, fathers participated in the program activities considerably more than they would have had they not been involved in Early Head Start and there were some important impacts on their parenting practices as well as on father-child interactions. Some of the impacts on father-child interaction are of the type that would be expected to lead to overall improved outcomes for children.

1Father-child activities, discipline, parenting behavior, and father’s well-being were drawn from father interviews and father-child videotaped interactions when the children were approximately 36 months old. Unlike the mother reported data, the father-reported and father interaction group differences were pooled and not weighted by site because of sample size constraints.(back)

GROUP DIFFERENCES IN FATHER PRESENCE, ACTIVITIES WITH CHILD, FATHER WELLBEING,
DISCIPLINE STRATEGIES, PARENTING BEHAVIORS, AND CHILD BEHAVIOR WITH FATHER
	Program Groupa	Control Groupb	Estimated Impact per Participantc	Effect Sized
Father Presence
Biological Father Present in Child's Life (Percentage)	72.7	70.9	1.8	4
Male Present in Child's Life (Percentage)	89.8	88.5	1.3	4.3
Respondent Married to Biological Father (Percentage)	34.6	35.6	-1	-2
Biological Father is Currently Married to, Lives with, or is Boyfriend of Respondent (Percentage)	48.9	50.5	-1.6	-3.3
Father Activitieswith Child
Frequency of Caregiving Activities Score	48.5	49.3	-0.8	-7.4
Frequency of Social Activities Score	49.2	49.1	0.1	0.7
Frequency of Cognitive Activities Score	49.6	49.1	0.4	3.9
Frequency of Physical Play Score	48.9	49.6	-0.8	-7.5
Father Well-Being
Parenting Stress Index (PSI): Parental Distress	19.4	19.3	0.1	1.4
PSI: Parent-Child Dysfunctional Interaction	14.1	14.3	-0.2	-4.7
CES-D Not at Risk of Depression (Percentage)	61.3	56	5.3	10.7
CES-D: Severe Depressive Symptoms (Percentage)	5.3	7.3	-2	-8
Family Environment Scale - Family Conflict (Average Score)	1.4	1.5	-0.1	-10.3
Discipline Strategies
Index of Severity of Discipline Strategies	3.3	3.4	-0.2	-10.6
Percentage of Fathers Who Spanked the Child in the Past Week	25.4	35.6	-10.2**	-21
Percentage of Fathers Who Would Use Mild Discipline Only	37.9	33.2	4.7	10
Parenting Behavior
Supportiveness During Father-Child Semistructured Play	4.1	4	0.2	17.9
Intrusiveness During Father-Child Semistructured Play	1.4	1.3	0	6.2
Quality of Assistance During Father-Child Puzzle Challenge Task	3.3	3.3	0	-3.6
Intrusiveness During Father-Child Puzzle Challenge Task	2.4	2.8	-0.4**	-30.4
Child Behavior with Father
Child Behavior Checklist-Aggressive Behavior	10.6	10.9	-0.3	-4.5
Engagement of Father During Father- Child Semistructured Play	5.1	4.8	0.3**	29.8
Sustained Attention with Objects During Father-Child Semistructured Play	5.2	4.9	0.3**	32.6
Negativity Toward Father During Father-Child Semistructured Play	1.1	1.1	-0.1	-12.6
Engagement of Father During Father-Child Puzzle Challenge Task	5.2	5.3	-0.1	-8.9
Persistence During Father-Child Puzzle Challenge Task	4.9	4.9	0	-1.9
Frustration During Father-Child Puzzle Challenge Task	2.3	2.3	0	-2.4
Sample Size
Mother Interview	1055	957
Father Interview	356	330
Father-Child Interactions	148	141
SOURCE: Parent interviews in all 17 sites when children were approximately 36 months old. Father interviews and father-child semi-structured interactions in the 12 father study sites conducted when children were approximately 36 months old. NOTE: All mother-reported impact estimates were calculated using regression models, where each site was weighted equally. All father-reported and father-child interaction impact estimates were calculated using regression models that pooled across sites. aA participant is defined as a program group member who received more than one Early Head Start home visit, met with an Early Head Start case manager more than once, received at least two weeks of Early Head Start center-based care, and/or participated in Early Head Start group parent-child activities.(back) bThe control group mean is the mean for the control group members who would have participated in Early Head Start if they had instead been assigned to the program group. This unobserved mean was estimated as the difference between the program group mean for participants and the impact per participant.(back) cThe estimated impact per participant is measured as the estimated impact per eligible applicant divided by the proportion of program group members who participated in Early Head Start services (which varied by site). The estimated impact per eligible applicant is measured as the difference between the regression-adjusted means for all program and control group members.(back) dThe effect size was calculated by dividing the estimated impact per participant by the standard deviation of the outcome measure for the control group times 100 (that is, it is the impact expressed as a percentage of the standard deviation).(back) Significantly different from zero at the .10 level, two-tailed test. Significantly different from zero at the .05 level, two-tailed test. **Significantly different from zero at the .01 level, two-tailed test.

H. CONCLUSIONS

The analysis of the Early Head Start programs’ overall impacts when children were about 3 years old, averaging across all types of programs and all types of families, shows a large number of favorable impacts for children and their parents. In large measure, these impacts sustain the impacts found a year earlier, when the children were 2. For children, the Early Head Start research programs:

Produced sustained, significant positive impacts on cognitive and language development at age 3. Early Head Start children were significantly less likely than control-group children to score in the at-risk range of developmental functioning in these areas.
The programs had favorable impacts on more aspects of social-emotional development at age 3 than at age 2—Early Head Start children engaged their parents more, were less negative to their parents, and were more attentive to objects during play, and Early Head Start children were rated lower in aggressive behavior by their parents than control children.

When children were 3, the Early Head Start programs also continued to have significant favorable impacts on a wide range of parenting outcomes:

Early Head Start parents were observed to be more emotionally supportive and to provide more support for language and learning than control-group parents (for example, they were more likely to read to their children daily).
Early Head Start parents were also less likely than control-group parents to engage in negative parenting behaviors. Early Head Start parents were less likely to report that they spanked their child in the past week, and they reported greater knowledge of mild discipline strategies.
When children were 3, Early Head Start parents did not differ significantly from control parents in any of the mental health outcomes we assessed, although they had significantly less parenting stress and family conflict when children were 2. However, growth curve analyses, while subject to some limitations, suggested that family conflict decreased over time for program but not for control parents.
The Early Head Start programs had some important impacts on parents’ progress toward self-sufficiency. The positive impacts on participation in education and job training activities continued through the 26 months following enrollment, and some impacts on employment began emerging late in that follow-up period in some subgroups. These impacts had not yet resulted in significant improvements in income, however.
Early Head Start families were less likely to experience subsequent births during the first two years after they enrolled and may have been less likely to experience the economic and psychological consequences of rapid repeat births.

Finally, although the programs had less experience in providing services specifically to fathers, they had significant favorable impacts in several areas of fathering and father-child interactions:

Program fathers had significant favorable impacts in several areas of fathering. They spanked less and were less intrusive. In father-child interactions, program children were more engaging of their fathers and showed greater sustained attention than control children did.

Analyses of potential mediators of the impacts on 3-year-old children provide support for programs’ theories of change that indicate program efforts to enhance aspects of parenting and the home environment may contribute to longer-term impacts on children:

Impacts on children’s cognitive and language development at age 3 were associated with parents who were more supportive in their interactions with their children and provided more language and literacy supports in the home at age 2.
Impacts on some of the positive aspects of social-emotional development (engagement of parent and sustained attention) when children were 3 were associated, to a small degree, with such parenting behaviors as warm sensitivity and emotional responsivity, and with parents’ knowledge of infant/toddler development, at age 2.
Impacts showing lower levels of children’s aggressive behavior and negativity toward their parents at age 3 appeared to be mediated by parenting a year earlier that was characterized by less physical punishment, lower parental distress, and greater warmth.

The consistent pattern of statistically significant, favorable impacts across a wide range of outcomes when children were 2 and 3 years old is promising. This pattern suggests that Early Head Start programs, overall, may be improving the balance of risk and protective factors in the lives of the low-income families they serve.

BOX V.11

HOW MUCH BETTER THAN EXPECTED? IMPROVING COGNITIVE OUTCOMES IN UTAH’S BEAR RIVER EARLY HEAD START

L.A. Roggman, L.K. Boyce, G.A. Cook, and A.D. Hart Utah State University

What are the strongest early predictors of later cognitive skills? Do Early Head Start children do better than expected, based on predictions? What aspects of Early Head Start are related to how much better they do? To test whether development is “better” for children in our local Early Head Start group than for the control group, we first examined the strongest early predictors of later Bayley MDI scores (at 36 months). We then developed statistical models using developmental measures at more than one age point, a grouping variable indicating whether or not the child’s family was in Early Head Start, and a set of the strongest early predictors of children’s later cognitive outcomes.

The strongest early predictors of poorer later cognitive skills were earlier measures of cognitive skills. Other early predictors were mothers’ low education, avoidance in close relationships, and poor use of social support. We used these strong correlates as covariates in a statistical test of the interaction between age and intervention. Age changes in the Bayley MDI scores over time showed a significant decline for the control group but not for the Early Head Start group (see Figure 1).

Figure 1

Differences over Time in Cognitive Skills (Bayley MDI Standardized Scores)
[D]

Next, we developed prediction models using the earliest Bayley MDI score and a set of predictors from before families were randomly assigned to Early Head Start or a control group. Compared to expected scores based on early predictors, Early Head Start children were doing better than expected, and the control group children were doing worse. Differences between actual and expected scores from early predictors significantly favored Early Head Start children. The advantage Early Head Start children gained was related to how engaged their mothers were in Early Head Start home visits. Home visit engagement, in turn, was related to more involvement in other Early Head Start activities, more facilitative home visitors, and less maternal avoidance.

In summary, cognitive development was progressing better for children in Bear River Early Head Start than for children in the comparison group. While cognitive skills scores declined for the control group, they did not for the Early Head Start children who maintained age-appropriate progress in developing their cognitive skills. Mothers’ involvement in Early Head Start appeared to buffer early risk factors for poor cognitive development.

BOX V.12

MOTHER-CHILD LANGUAGE AT 14 AND 24 MONTHS: CONCURRENT AND LAGGED ASSOCIATIONS

Elizabeth Spier, Catherine S. Tamis-LeMonda, and Mark Spellmann
New York University

Barbara Alexander Pan and Meredith Rowe
Harvard Graduate School of Education

The quality and quantity of caregivers’ language is one of the most powerful predictors of children’s early language and cognitive development. Thus, a fundamental goal of many Early Head Start practitioners is to encourage parents to talk frequently with their children in ways that are sensitive to children’s emergent language (for example, by asking question like “What is that?” that elicit the child’s own verbal participation). Parenting often mediates the impact of early interventions and parents’ verbal input accounts for much of the variation linking poverty to compromised child outcomes (Hart and Risley 1995). Given the importance of parents’ language input in developing children’s language and cognition, researchers at New York and Harvard Universities have focused on the amount and diversity of language to which young children are exposed during the foundational period of 14 to 24 months. In this study, we merged transcript data from the two local research sites and explored associations between mothers’ language and children’s language and developmental status.

The sample consisted of 146 mother-child dyads participating in the research at the New York and Vermont sites. Forty-eight percent of the participants were white, 25 percent were African American, 17 percent were Hispanic, and 10 percent fell into other groups (for example, West Indian, Asian, mixed ethnicity). All parents spoke English.

We used the semistructured, three-bag task from the national protocol as the basis of mother and child language at both 14 and 24 months. We transcribed play sessions and tabulated the total number of words (tokens) and different words (types) expressed by each mother and child at each age, as well as mothers’ total number of “wh” questions. Children’s Bayley MDI performance and data from the MacArthur CDI were included in analyses.

Findings revealed that mothers’ language predicted most child measures at 14 and 24 months. Maternal word types, tokens, and “wh” questions were consistently associated with children’s comprehension and production on the MacArthur CDI and Bayley MDI scores, as well as on the Bayley Language factor. Maternal word types correlated with children’s types and tokens, albeit weakly. At 24 months, maternal language measures were associated with every measure in children, except for tokens. Lagged correlations indicate that mothers’ earlier language was associated with children’s language and developmental status over time.

Next, simultaneous regressions were conducted in which we examined the joint contributions of child and mother at 14 months to each child and mother outcome. Regressions indicated that mothers and children both contributed unique variance to children’s language and cognitive outcomes, explaining between 11 and 32 percent of the variance in 24-month measures. However, children’s 14-month language did not predict mothers’ later language over and above mothers’ stability. The strongest predictor of 24-month maternal language was the mother’s earlier language, which explained up to 44 percent of the variance in her later language. Children were also stable in language and developmental status.

Together, these findings indicate that mothers’ language at the onset of children’s second year is beginning to make a difference in children’s emergent cognitive and linguistic abilities. Therefore, it is important to encourage mothers to talk to and ask questions of their children well before children speak with regularity.

BOX V.13
FUNCTIONS OF LANGUAGE USE IN MOTHER-TODDLER COMMUNICATION1

Joanne Roberts, Catherine S. Tamis-LeMonda, and Mark Spellmann
New York University

Caregivers who provide children with verbally rich, responsive language environments in the early stages of language acquisition have children who excel in lexical, grammatical, and syntactic abilities and who achieve important language milestones sooner. Understanding links between parenting and children’s emerging language competencies is central to understanding and modeling associations between Early Head Start and developmental achievements in children.

Investigators at New York University have been investigating the language environments to which children are exposed in relation to their early communicative abilities. We wanted to examine associations between mothers’ and children’s language at 14 months, during initial stages of language acquisition. Because children have a limited productive vocabulary at 14 months, we developed a way to assess their communicative intentions that incorporated verbalizations and gesture in determining their communicative intent.

The sample consisted of 75 ethnically diverse mother-child dyads (63.6 percent of the children were male), the first wave of participants at New York University’s local research program. We transcribed maternal speech and actions, as well as all child vocalizations and actions, from the 10-minute, semistructured play task. We coded maternal utterances into 1 of 17 language functions and children’s vocalizations as 1 of 9 functions.

Variation in amount and function of language among mothers and children was dramatic. Mothers expressed between 20 and 331 utterances, children between 0 and 117 utterances. Factor analysis, with varimax rotation, was conducted on mothers’ and children’s language. Three factors of maternal language emerged:

Responsive/Didactic: Language in which the mother is repeating and expanding on the child’s vocalizations, reformulating the child’s behaviors into words, proposing questions to the child, and labeling and describing objects and events
Directive: Language characterized by mothers’ control and direction of children’s actions, as well as by prohibitions and corrections
Uninvolved/Hostile: Language characterized by mothers’ self-directed comments and criticism of the child

For children, two factors of communication emerged:

Communicative: Utterances that are responsive to the social partner or that relate information about objects, events, desires and interactions with others
Distress: Utterances that express discontent, frustration or objection

Analyses showed that the maternal responsive/didactic factor related to children’s communicative factor, as did the maternal directive factor. The maternal directive factor also related to the children’s distress factor. Further breakdown of these associations revealed that mothers’ responsive/didactic language speech predicted children’s imitations, expression of notice, references to actions in play, and declaratives. Mothers’ directive speech related only to children’s objections/refusals, The maternal uninvolved/hostile factor did not relate to children’s language.

These findings show that it is important for programs to support mothers in their use of frequent, didactic-responsive language to encourage children’s verbal fluency. Focusing solely on decreasing uninvolved/hostile communications in mothers, while important to social-emotional aspects of children’s development, is not sufficient for increasing children’s language achievements.

1This research is taken from: Roberts, J. & Tamis-LeMonda, C. S. (2000, June). Functions of language use in mother toddler communication. In J. Atwater (Chair), the social context of early language development for children in poverty. Symposium conducted at Head Start’s National Research Conference, Washington, D.C. (back)

BOX V.14

PARENT RESPONSIVENESS AND CHILDREN’S DEVELOPMENTAL OUTCOMES KANSAS EARLY HEAD S
TART PARTNERSHIP

Jane Atwater, Judith Carta, Jean Ann Summers, and Martha Staker
University of Kansas and Project EAGLE Early Head Start

The Kansas Early Head Start Partnership identified responsive parent-child interaction as an optimal and essential context for promoting children’s development and fostering families’ well-being. In these analyses, we examined parent responsiveness as a predictor of early development for children in multirisk families. In addition, for Early Head Start families, we asked whether their level of engagement in home-based services was related to parents’ responsiveness with their children and to children’s developmental progress. The analysis sample consisted of 74 Early Head Start families and 79 control group families in an ethnically diverse, urban community.

Parent responsiveness was assessed during home-based observations when children were 8, 14, 18, 24, 30 and 36 months old. Responsiveness measures included two composite variables—Parent Talk to the Child and Close Involvement—that provided an index of the parent’s general responsiveness with the child and three specific variables—Prompt/Expansion of Child Communication, Positive/Exuberant Response, and Shared Parent-Child Activities—that described qualitative features of Parent Talk and Close Involvement. Parent engagement in the Early Head Start program was based on Early Head Start staff ratings of the level and consistency of parent participation over time, active interest and involvement during home visits, and parents’ follow-through on individual program goals between visits.

To track children’s developmental progress, we focused on growth over time in children’s cognitive development (performance on the Bayley MDI) and language development (children’s verbal communication during typical activities at home). Child assessments were conducted at 8, 14, 18, 24, 30 and 36 months of age.

The Relationship of Parent Responsiveness to Children’s Development. In analyses of children’s developmental trajectories, every measure of verbal responsiveness (Parent Talk, Prompt/Expansion, and Positive/Exuberant Response) was a significant predictor of Bayley scores. Shared Activity also was positively related to cognitive outcomes and was the only significant predictor of growth in cognitive development from 8 to 36 months. Results for children’s verbal communication were even more striking and consistent. Every measure of responsiveness was a significant predictor of communication outcomes and increases in verbal communication from 8 to 36 months. When parents were more verbally responsive and involved in their children’s activities, children talked more, and their use of words increased more rapidly.

The Relationship of Program Engagement to Parent Responsiveness and Child Development. Parents with the highest level of program engagement had higher rates of verbal responsiveness with their children. That is, the parenting behaviors most clearly related to child outcomes occurred more frequently in families highly engaged in the Early Head Start program. Moreover, engagement in the program was predictive of more positive outcomes in children’s cognitive development and verbal communication and of growth over time in verbal communication. Thus, these results provide evidence of a positive relationship between program engagement and developmental progress and suggest that responsive interactions might be one process that supports that relationship. The results of these analyses provide empirical support for the Early Head Start program’s emphasis on responsive parent-child interactions as a key component of intervention for children and families who experience multiple risks.

BOX V.15

OUTCOMES OF PROGRAM PARTICIPATION AND CORRELATES OF CHILDREN’S COGNITIVE DEVELOPMENT AT THE EDUCATIONAL ALLIANCE’S EARLY HEAD START

Mark Spellmann, Catherine Tamis-LeMonda, Maria Yarolin, Lisa Baumwell, Joanne Roberts, and the NYU
Early Childhood Research Team

In this study, we addressed two research questions:

What child and parent outcomes did participation in Early Head Start affect?
What child and parent characteristics were associated with children’s cognitive development?

We tested two dimensions of program participation for effects on child and parent outcomes: (1) children’s attendance at the Early Head Start child care centers, and (2) the degree of parent involvement with Early Head Start social service staff.

For children, outcomes of program participation included greater cognitive development at 14, 24 and 36 months; greater social development; and greater language development.

Parental domains significantly associated with program participation included the quality of parent-child interaction, the quality of parenting, discipline strategies, parenting stress, psychological well-being, and social support.

We also wanted to explore correlates of children’s cognitive development, as measured by the Bayley Mental Development Index (MDI), which were given when children were 14, 24 and 36 months old.

Observational measures of the quality of parenting and the quality of parent language use showed substantial associations with cognitive development at 24 and 36 months. The quality of parent-child interaction was significantly associated with cognitive development at 24 and 36 months. Self-rated parenting measures were also associated with cognitive development.

Father involvement was associated with children’s’ cognitive development, as was the quality of the home environment.

Emotional social support and advice and guidance social support that mothers received were associated with child cognitive development. Support mothers received from their own mothers, and from their babies’ fathers, was associated with MDI scores.

Program engagement variables were associated with child cognitive development. Four measures of positive program involvement—Social Support from EHS staff, “What I Got from EHS: Growth as a Parent,” “What I Got from EHS: Family-Program Bond,” “What I Got from EHS: Child Development”—were positively associated with children’s cognitive development at 14 and 36 months.

Measures of parents’ emotional well-being were significantly associated with children’s cognitive development. Symptoms of post-traumatic stress disorder and parenting stress were negatively associated with cognitive development. Harsh, rejecting fathering that mothers received when they were growing up was negatively associated with cognitive development of their young children at all three age milestones. The quality of mothering in mothers’ families of origin was associated with MDI scores at 14 and 24 months.

Other aspects of child development also demonstrated significant association with cognitive development. Social development, measured both by parent ratings of children’s social development and by observational measures of child-parent interaction, showed a strong correlation with cognitive development. Mother’s ratings of children’s distractibility, difficult temperament, and difficult behavior were associated with lower MDI scores at 36 months. Children’s health was associated with cognitive development at 36 months.

The wide range of factors associated with cognitive development scores illustrate that children’s cognitive development is embedded in multiple levels of systems, at the child, family, and program levels. The implication of these findings is that early intervention programs are likely to be increasingly effective to the degree that they are able to address every level of the system in which children’s cognitive development is embedded.

BOX V.16

RELATIONS BETWEEN SPECIFIC AND GLOBAL FEATURES OF MOTHER-CHILD INTERACTIONS AND
LANGUAGE

Catherine S. Tamis-LeMonda, Elizabeth Spier, and Mark Spellmann
New York University

Barbara Alexander Pan and Meredith Rowe
Harvard University

The quality of parent-child interactions is one of the most powerful predictors of children’s emerging cognitive competencies, especially language. Many researchers, practitioners, educators, and parents want to know which features of parenting are most relevant to positive outcomes for children, as well as the best ways to capture and evaluate those features in research and practice settings. Many approaches to the coding of parent-child interactions are available, and theoretical orientation and practical constraints guide decisions about which to use.

For example, the national study used measures of caregiver-child interactions during semistructured play based on global ratings of six aspects of behavior in mothers (sensitivity, intrusiveness, stimulation, positive regard, negative regard, and detachment) and three in children (engagement, sustained attention, and negativity). Because global ratings are more efficient to complete than more complex rating systems, large-scale studies frequently rely upon such codings. In contrast, researchers at many local sites, including New York and Harvard Universities, transcribed the full array of verbal and gestural exchanges between mothers and children during the semistructured play tasks to describe and capture specific aspects of parent-child engagements. Both “macro” and “micro” approaches to assessing parenting have merits, and both have limitations. Little is known about whether and how data obtained from the two relate to each another, however. Here, we explore associations between transcriptions of mothers’ and children’s language obtained locally and global ratings of mother-child interactions (obtained at the national level).

Research teams at Harvard Graduate School of Education and New York University Graduate School of Education longitudinally examined mother-child discourse in a total of 146 dyads during the semistructured play task at 14 and 24 months. The sample was ethnically diverse: 47 percent white, 25 percent African American, 17 percent Hispanic, and 11 percent other (for example, mixed ethnicity).

We obtained maternal language samples through transcription of the semistructured play task. We counted the number of different words (word types) each mother and child used; the total number of words (tokens) each mother and child used, and the number of “wh” questions each mother used during the 14- and 24-month sessions. Global ratings of mother-child interactions from this task were those coded nationally by the national evaluation team.

Findings indicated that mothers’ total words, word types, and “wh” questions were positively associated with ratings of sensitivity, stimulation, and positive regard and negatively associated with detachment (rs range from .19 to .66, ps < .05 to .001). We next tested the joint contributions of mothers’ language types, tokens, and “wh” questions to the composite score of “supportiveness” (a composite measure created by the national team by summing mothers’ ratings on the three items). At both ages, maternal language types and “wh” questions (but not tokens) contributed unique variance to the composite measure of supportiveness, together accounting for 40 and 42 percent of the variance at 14 and 24 months, respectively.

In children, associations between language and global ratings of their engagement, attention, and negativity varied with age. At 14 months, children’s word types and tokens were weakly associated with global measures of child engagement and attention (rs range from .17 to .20, ps < .05); by 24 months, however, associations were moderate to strong (rs range from .33 to .51, ps < .001).

In general, results support the validity of national measures of parent-child interactions by demonstrating their strong associations to independently coded, in-depth measures of mother and child language at two local sites. They also indicate that coders are acutely sensitive to mothers’ and children’s language when coding dyadic interactions. Finally, these findings have important implications for program staff. Staff should be sensitized to the importance of mothers’ and children’s language interactions as key expressions and indicators of mutual sensitivity and cognitively rewarding interactions.

BOX V.17

SYNOPSIS OF MOTHERS’ SOCIALIZATION OF TODDLER CONFLICT RESOLUTION

Lisa Baumwell, Tonia Cristofaro, and Mark Spellmann New York University

Young children commonly engage in conflicts with peers. Parents play an important role in transmitting beliefs about how their children should resolve these conflicts. Research suggests that parents’ beliefs, when translated into child-rearing practices, influence children’s social competence. For example, the belief that aggression is a socially acceptable strategy has been found to be associated with children’s aggressive behaviors.

To date, few studies have examined mothers’ beliefs about how their toddlers should resolve peer conflict. Therefore, we sought to characterize mothers’ attitudes about the conflict resolution strategies that their 3-year-olds should employ with intruding peers. We also examined how participation in Early Head Start influences maternal beliefs about conflict resolution strategies.

Sixty ethnically diverse mothers of 27 girls and 33 boys participated in this study. Participants were a subset of the 36-month Early Head Start cohort in New York City. During the 36-month-home visit, mothers completed a self-administered questionnaire on conflict resolution strategies. This is a social problem-solving scale, based on one used by Slaby and Guerra (1988), that required mothers to select strategies that they would want their 3-year-old children to use in four scenarios depicting peer disagreements. Mothers selected one of five strategies appropriate to the scenario. The strategies reflected verbal aggression, physical aggression, walk away, ask an adult for help, and verbal prosocial responses (words with peers).

We calculated frequencies of the five strategies across the four situations. Ninety-two percent of mothers chose ask an adult for help and 75 percent selected verbal prosocial responses at least once. Thirty-eight percent of mothers endorsed walk away, 23 percent chose physical aggression, and only 8 percent supported the use of verbal aggression at least once. In addition, mothers were consistent in the strategies they adopted. Most mothers who selected verbal aggression also selected physical aggression. Mothers who selected prosocial peer responses and ask an adult for help were less likely to select aggression as a strategy to solve peer conflict.

We calculated multiple t tests to examine how participation in Early Head Start influenced mothers’ beliefs about their children’s conflict resolution. Participants whose attendance was rated “fair” to “excellent” at Teen Aid High School and Educational Alliance were compared with control parents. Teen Aid participants chose walking away more. Mothers attending Educational Alliance endorsed physical aggression less and chose asking an adult for help rather than walking away.

In summary, this investigation elucidates mothers’ beliefs about their children’s problem-solving strategies with peers. Our findings suggest that these beliefs can be modified in ways that may help children become more socially competent.

BOX V.18

CHANGE IN PARENT-CHILD INTERACTION IN LOW-INCOME FAMILIES: LINKS TO FATHER STATUS

L.A. Van Egeren, L. McKelvey, H.E. Fitzgerald, R.F. Schiffman, M. Cunningham DeLuca, and M. Hawver
Michigan State University

Contingent responsiveness is a foundation of child socioemotional and cognitive adjustment (Bornstein et al. 1999; and Watson 1985). Among low-income families who experience high rates of single motherhood, inconsistent father involvement, and transitory male figures in children’s lives, how mothers’ and fathers’ interactions with their children mutually develop warrants particular attention. This study examines changes in contingent responsiveness of low-income parent-child dyads over a two-and-a-half-year period.

The sample for this study consisted of 71 families (children, mothers, and men the mother identified as the child’s father or father figure) participating in an ongoing longitudinal study of children eligible for Early Head Start in Jackson, Michigan. At enrollment, 24 months, and 36 months, each parent participated in a teaching task with the child, which was rated using the Nursing Child Assessment Teaching Scale (Sumner and Spietz 1994). To assess the quality of contingent interactions, we used three parental contingency scales—Sensitivity to Cues, Social-Emotional Growth Fostering, and Cognitive Growth Fostering—and two child scales—Clarity of Cues and Contingent Responsiveness to Caregiver. We asked the mother about paternal residency and biological father status.

We used hierarchical linear modeling to derive an overall trajectory for the interaction scores of each parent and child while accounting for interdependencies between parents.

Mean level. At enrollment, mothers were more sensitive than fathers to infant cues but less likely to foster social-emotional or cognitive growth. By 36 months, the pattern had reversed: fathers tended to be more sensitive to cues than mothers but were less likely to foster social-emotional or cognitive growth. Children gave clearer cues to fathers at enrollment but, at 36 months, they showed no difference in behavior toward either parent.

Linear Change. Both parents’ sensitivity to the child’s cues and cognitive-growth fostering increased significantly over time. Although mothers increased in social-growth fostering, fathers decreased substantially. Mothers increased more than fathers in sensitivity to cues and cognitive-growth fostering. Children increased significantly in the clarity of cues and responsiveness, particularly toward mothers.

Father status was consistently related to father sensitivity to cues. At enrollment, residential fathers were less sensitive than nonresidential fathers, and biological fathers were less sensitive than nonbiological fathers. By 36 months, both residential and biological fathers had increased in sensitivity to cues, and residential fathers had also increased in cognitive-growth fostering. Interactions between the two father status variables suggested that the quality of mothers’ and children’s contingent interactions decreased when the father was a nonresidential social father.

The results suggest that fathers and mothers were more similar in their contingent responsiveness toward the child by 36 months than during early infancy. Children’s contingent responsiveness originally favored fathers, then became similar toward both parents. Father status worked in distinct ways for mothers and fathers that were specific to different types of responsiveness.

BOX V.19

FATHER-CHILD INTERACTIONS: MEASURING PAST PATERNAL INFLUENCES

Jacqueline D. Shannon, Catherine S. Tamis-LeMonda, Joanne Joseph, Bonnie Hannibal,
Tracy Poon, Michele Pelnar, and Vanessa Rodriguez
New York University

The Early Head Start Fathers’ group grew out of a need to further understand father involvement in low-income families. In New York City, we examined father-child interactions and whether a father’s interaction style related to paternal relationships in his own childhood.

We examined the interaction styles of 57 ethnically diverse, inner-city fathers with their 24-month-olds (28 boys). Our goals were to:

Describe the nature of fathers’ interaction styles.
Compare the relationship between fathers’ interaction styles and their children’s social, emotional, and cognitive behaviors.
Assess the extent to which fathers’ perceptions of paternal relationships in their own childhoods relate to their own fathering interaction styles.
Explore men’s feelings toward and perceptions of their childhood experiences with their fathers.

Data collection consisted of videotaped father-child interactions during semistructured free play, fathers’ perceptions of paternal childhood experiences measured through the Parental Acceptance-Rejection Questionnaire, and 18 semistructured qualitative interviews. We assessed father-child interactions using the Caregiver-Child Affect, Responsive and Engagement Scale (C-CARES). The C-CARES measures parent-child interactions on 15 parent and 14 child behaviors, which are individually rated on a 5-point Likert scale ranging from 1—“not observed” to 5—“constantly observed.”

We identified three meaningful clusters of father interaction styles:

Responsive/Didactic. These fathers demonstrated great awareness and responsiveness to children’s emotional needs. They were flexible and sensitive to appropriate teaching moments and ways to engage their children in play without being overtly achievement-oriented. This parenting style appeared to be positively associated with children’s social and cognitive abilities.
Overbearing. These fathers were driven to teach their children skills; however, they were highly structured and primarily intrusive. These overly controlling fathers appeared to diminish children’s exploratory and communicative initiatives.
Disengaged. These fathers were uninvolved with their children and unresponsive to them. Their children were also unresponsive to them and only moderately involved with toys, playing with them in a rudimentary, unsophisticated manner.

Because children are not passive recipients of fathering, they might influence their fathers’ interaction styles. Children who exhibit sophisticated language and play might promote sensitive, didactic interactions with their fathers. Similarly, less capable children might be less rewarding social partners, thereby compromising the quality of their fathers’ engagements.

Fathers’ experiences of paternal warmth were not associated with their interaction style. However, overbearing and responsive/didactic fathers were more likely than disengaged ones to perceive lower levels of paternal rejection. All fathers were committed to “being there” physically and emotionally for their children, regardless of the quality of their childhood experiences with their own fathers. Findings support the notion that fathers’ childhood experiences of paternal rejection relate negatively to quality parenting interactions. However, to more fully appreciate how these experiences shape fathers’ interactions and involvement with their children, additional variables should be considered. A deeper understanding of how inner-city fathers’ parenting roles and interaction styles have been shaped could help improve services available to them and their families.

BOX V.20

ANDREYA EARNS HER HIGH SCHOOL DEGREE: THE ROLE OF EARLY HEAD START

Jean M. Ispa and Elizabeth A. Sharp
University of Missouri-Columbia

Andreya1 was 19 and living with her 1-year-old son, her mother, her 16-year-old brother, and her 12-year old cousin in the inner core of a large Midwestern city when we first met her in 1996. She had agreed to participate in our case study research. Like almost all the mothers served by the Early Head Start program in which she was enrolled, she was African American, young, and single. She had been 17 and in the second semester of 11th grade when she’d become pregnant and dropped out of high school.

Still hoping to earn a high school diploma, Andreya had recently begun attending Job Corps classes.2 She had also enrolled in Early Head Start. Looking back five years later, she believes, as we do, that her Early Head Start home visitor played a pivotal role in guiding and supporting her through the challenges that threatened to derail her as she struggled to stay in school. (Volume III contains a case study describing both the barriers and the support Andreya encountered during her Job Corps experience). Here, we summarize the obstacles Andreya faced and the assistance her Early Head Start home visitor provided to her.

A list of the obstacles might begin with Andreya’s poverty and the poor quality of the schools she had attended in her inner-city neighborhood. These conditions help explain the absence of academically successful role models in her family, as well as her quick temper, emotional neediness, and low self-concept. In addition, her partner saw no good reason for her to continue her education. Clearly, it was difficult to study under these circumstances. Other sources of stress were ongoing family conflict, worry over her son’s chronic asthma and aggressiveness, and a second pregnancy and birth before she had completed her Job Corps course work. Because both her children were asthmatic, she missed many days of school to tend to them. The initially unsympathetic attitude of the Job Corps staff toward her absences further undermined her confidence and resolve. Economic hardship exacerbated all of these problems.

On the positive side, her mother and grandmother were unswerving in their messages that she should break with family tradition and be the first in the family to graduate. Moreover, Andreya loved her children and wanted to do whatever was best for them, including completing high school so that she would be better equipped for the job market. Rickie, her Early Head Start home visitor, built on these qualities. He agreed that graduating should be a primary goal and, each time she considered dropping out, warned her of the consequences and shored up her confidence. His contributions went well beyond these discussions, however. During Andreya’s involvement with Early Head Start, Rickie taught her how to manage her temper and her time, encouraged her to set and work toward attainable goals, helped her navigate the social service system, and served as her advocate with the Job Corps staff. He also provided gentle advice regarding her relationships with her children, her mother, her brothers, and her children’s father. This support helped Andreya become the only one in her family to graduate from high school.

1All names are fictitious.(back)

2Job Corps is a federally funded program that provides high school education plus job training. To earn the high school degree, students must complete all high school requirements plus all requirements for their “trade”–the job-specific training.(back)

BOX V.21

VALIDATION OF NATIONAL CHILD LANGUAGE MEASURES AT 14 AND 24 MONTHS1

Barbara Alexander Pan and Meredith Rowe,
Harvard Graduate School of Education

Elizabeth Spier, Catherine Tamis-LeMonda, and Mark Spellman
New York University

At the 14- and 24-month data collection points, the Early Head Start national evaluation relied primarily on the MacArthur Communicative Development Inventory (CDI; Fenson et al. 2000) as a measure of children’s language development. The CDI is a checklist of age-appropriate language skills (for example, vocabulary comprehension and production, use of gestures, sentence types) that parents complete. Studies of middle-class families indicate that mothers in such families are relatively good judges of their children’s language use (Fenson et al. 1994). However, some researchers have questioned the accuracy of reports by low-income mothers or those with lower levels of education (for example, Feldman et al. 2000). Thus, it was important in the current evaluation of Early Head Start to ascertain how accurately mothers in the study assessed their children’s vocabulary.

Research teams at Harvard Graduate School of Education and at New York University Graduate School of Education transcribed and analyzed parent-child discourse observed during the videotaped semistructured play activity (combined sample at two sites: n = 161 at 14 months, n = 158 at 24 months). Approximately 45 percent of the mothers identified themselves as white, 25 percent as African American, and 17 percent as Hispanic. Transcribed spontaneous speech yielded two measures of child language use that we focus on here: the number of different words (word types) produced by the child and the number of total words spoken by the child.

We examined associations between child spontaneous speech measures (word types, total words), parent report measures (CDI scores), and children’s performance on structured cognitive and language assessments (Bayley scores). Note that parents were asked to assess children’s comprehension only at 14 months and that Bayley Language Factor scores were computed only at 24 months.) Results for the combined sample showed that at 14 months, parental report of children’s productive vocabulary correlated moderately well with children’s spontaneous vocabulary use as measured by word types (r = .43, p < .001) and total words (r = .39, p < .001). Bayley MDI scores showed no relationship to spontaneous speech measures and only a weak association with CDI production (r = .17, p < .05). At 24 months, parent report of child language was strongly associated with both spontaneous speech measures (word types: r = .53, p < .001; total words: r = .40, p < .001) and with structured assessments (Bayley MDI: r = .52, p < .001; Bayley Language Factor: r = .61, p < .001 ). These general patterns were found for families in both sites and across ethnic groups, although Hispanic mothers’ report of child productive vocabulary was not associated with child word types at 24 months, possibly due to the small sample size (n = 27).

Regression analyses using maternal report of children’s productive vocabulary to predict children’s spontaneous vocabulary use (word types) and language performance on the Bayley Language Factor confirm that low-income parents accurately report their children’s language development, particularly at 24 months. At age 2, parental report alone accounted for 27.5 percent of variation in child word types and 37.5 percent in Bayley Language Factor scores. Controlling for maternal education, child gender, and birth order, the variation accounted for by maternal report increased to 31.3 percent for word types and to 39.9 percent for Bayley Language Factor.

These results suggest that low-income parents’ reports are congruent with observed measures of children’s language development and that parental report of toddlers’ productive vocabulary at 24 months, as reported in the national evaluation’s interim report is a valid outcome measure of program impacts on child language development (ACYF 2001).

1See full report in Volume III for tables and references.(back)

BOX V.22

ASSOCIATIONS OF MATERNAL AND CHILD ATTACHMENT SECURITY WITH OUTCOMES OF
CHILDREN ELIGIBLE FOR EARLY HEAD START

Susan Spieker, Kathryn Barnard, Michelle DeKlyen, and Dana Nelson
University of Washington

In our Early Head Start study, we used “gold standard” attachment measures for both the mothers and children and related these measures to child outcomes. Immediately after random assignment, mothers participated in the Adult Attachment Interview, in which the coherence of their state of mind with respect to attachment relationships was rated on a 9-point scale. “Coherence” is an indicator of security. It is the adult’s ability to reflect on memories related to attachment while simultaneously having a clear, understandable, and collaborative conversation with an unfamiliar interviewer. When the children were 19 months old, they were assessed in the “Strange Situation,” a separation and reunion paradigm, in which the security of their relationship with the mother was rated on a 9-point scale. Security is the extent to which the infant uses the mother as a source of comfort when distressed and a safe base from which to explore. Both measures are time-intensive and broadly validated.

Based on theory and prior research, we expected that both adult and child attachment security would be protective factors for child outcomes for children eligible for Early Head Start. Thus, we expected that higher security ratings would, in general, predict more positive child language, cognitive, and behavioral outcomes at 24, 30, and 36 months. The outcome measures include Aggressive Behavior (CBCL), Sustained Attention (Semistructured Play), Bayley Mental Development Index, Bayley BRS Orientation Rating, Auditory Comprehension (PLS), Expressive Communication (PLS), and PPVT-III Receptive Vocabulary.

All analyses, which used both the program and comparison groups, consisted of hierarchical regression, in which the mother’s verbal ability, as assessed by the vocabulary subtest of a standard IQ test, was entered on the first step. The mother’s coherence of mind and child’s attachment security were entered on the second and third steps, respectively. Thus, the contribution of coherence of mind was assessed after controlling for the mother's verbal ability (which was correlated .38 with coherence of mind). Unexpectedly, child security was not correlated with the mother’s coherence of mind. Further analyses are planned to discover the reasons for this lack of association. Security was not correlated with mother’s verbal ability, and it was not expected to be.

Not surprisingly, maternal verbal ability was related to all cognitive and language outcomes. Maternal coherence of mind, usually measured before the birth of the child, was associated with child mental ability and orientation at 24 months and child language comprehension at 30 months, even after controlling for the effect of the mother’s own verbal ability. Coherence was also uniquely associated with lower child aggression at 36 months. Finally, child attachment security significantly predicted four cognitive, language, and behavioral scores at 30 and 36 months. These results suggest that the quality of relationships is an important context for child development in the toddler and preschool years. They also suggest that intervention that focuses on relationships, for the mother and, especially, for the mother-child relationship, may have considerable benefit for child behavioral, cognitive, and language development.

BOX V.23

EARLY HEAD START INTERVENTION WITH FAMILIES AND FAMILIES’ INVESTMENT IN CHILDREN

Michaela L.Z. Farber, Elizabeth M. Timberlake, Shavaun M. Wall, and Nancy E. Taylor
The Catholic University of America

United Cerebral Palsy Early Head Start was a federally funded program that promoted child development through a flexible mixture of child and family services. It served young, economically disadvantaged families with children under age 3. The Early Head Start center was in a suburban commercial strip mall in Northern Virginia and served 73 families living in motels, low-rise apartments, and rental houses within a 10-mile radius of the center. The child-focused services included family- or center-based child care and home visiting. The family-focused services included parent mobilization and links to community services to enable parents to fulfill their parenting roles, achieve family well-being, and move toward economic self-sufficiency. To date, however, little is known about how the United Cerebral Palsy Early Head Start services strengthen family functioning, parental investment in children, and children’s social development.

To explore the effect of Early Head Start services, the Catholic University of America research team (1) assessed family needs and aspirations at enrollment; (2) documented the type and amount of Early Head Start services delivered to families; and (3) assessed family functioning and child social development when the enrolled child reached 30 months of age, six months prior to program exit. Next, we explored whether variance in service activities was associated with family status as U.S.-born or immigrant. Finally, we explored whether Early Head Start service activities were congruent with families’ needs and aspirations at enrollment, and, in turn, whether these services helped families achieve greater competency in their pre-exit family functioning when the children were 30 months old. We also explored whether family functioning created a greater family investment in the targeted children and, therefore, improved those children’s social development at 30 months of age.

Findings from multiple quantitative analyses documented Early Head Start services for 32 immigrant and 41 U.S.-born families and identified an Early Head Start service path for all families. Most of the immigrant families received family child care, home visiting, or a combination of the two child care programs. Half of the 41 U.S.-born families received center-based child care with or without home visiting or family child care, and half received a combination of family child care and home visiting. Immigrant families received more parent mobilization services to match their greater need at enrollment. Both immigrant and U.S.-born families received a similar number of links to the community services they needed. Overall, Early Head Start parent mobilization and linking service activities, as mediated by family status, an assessment of family needs and resources, and aspirations at enrollment, created a path that led to increased pre-exit competence in family functioning. The family status at enrollment and pre-exit functioning further affected families’ pre-exit investment in their children. Finally, family pre-exit investment in children and family aspirations at enrollment were reflected in children’s sociobehavioral functioning when they turned 30 months of age. Through meeting the sociocultural needs of Early Head Start families at enrollment, the program activities positively influenced both family functioning and child investment. In addition, when they were combined with families’ aspirations, these activities influenced children’s social development, which was appropriate for their age at 30 months. Further longitudinal study is needed to learn whether these observed effects of Early Head Start services will endure.

1See the discussions in two implementation reports, Leading the Way, Vol. I (ACYF 1999) and Pathways to Quality (ACYF 2002).(back)

2The program-group mean was 5.9; the control-group mean was 5.3, for a positive impact of 0.6, statistically different from zero at the .01 level, with an effect size of 17.8.(back)

3As described in Boxes V.1 and V.2, the measures of child behavior during the Bayley assessment are different kinds of measures than the Bayley MDI, a measure of child cognitive development discussed in the previous section, and on which we did find Early Head Start impacts.(back)

4We were able to conduct these analyses for 3 child and 8 parent measures. The analytic technique limited us to measures that were the same at all three ages, were continuous variables (thus excluding binary and categorical variables), and were not age-normed (since these variables were adjusted for age at each point). Thus, unfortunately, it was not feasible to examine growth for the cognitive and language development outcomes. The sample for these analyses was further limited by the requirement that we could include only sample members who were administered the measures at all three ages. This may account for some differences in outcomes in the growth curve, compared with the point-in-time, impact estimates. For example, parent interview variables (such as parent-child play) were available for 2,110 families at 36 months whereas 1,700 families were interviewed at all three ages, a 19 percent smaller sample. Finally, these analyses produced linear growth curves, which in some cases may not accurately reflect the nature of the changes occurring over time.(back)

5Differences in the way in which these reading outcomes were measured could account for the different percentages reporting regular reading. The daily reading variable was coded based on responses to a direct question about the frequency of reading. The frequency of daily reading could thus reflect both actual behavior and differences in the parent’s knowledge that daily reading is desirable. Reading regularly at bedtime reflects parents’ responses that they follow a regular bedtime routine and that the routine includes reading. While this outcome is not as likely to be influenced by social desirability biases, bedtime is not the only time of the day when reading can occur.(back)

6Appendix D.9 provides greater detail on our hypotheses and the rationale for these analyses, and also includes mediated analyses by program approach, as discussed in Chapter VI.(back)

7To check the robustness of these findings, we also substituted an alternative measure of the frequency of parent reading to the child for reading at bedtime. The alternative variable, Daily Reading, is based on parent report in response to a direct question about how often the parent reads to the child. We found that the proportion of the impact on the Bayley MDI and PPVT-III at 36 months that is associated with daily reading is very similar to the proportion associated with reading at bedtime, and the overall proportion of the impact associated with all of the parenting mediators in each of the models changes by only about 3 percentage points.(back)

Table of Contents | Previous | Next