HighScope Preschool

Study 1

This study was conducted under the former name of the program, Perry Preschool Program.

Summary

Weikart et al. (1978) and five related studies utilized a randomized controlled trial in which 123 preschool-aged children across five successive annual cohorts were randomly assigned to the intervention group or a control group. Data on academic and social outcomes were collected on both groups annually from ages 3 through 11, at ages 14-15, at age 19, at age 27, and at age 40.

Weikart et al. (1978) and the related studies found, compared to children in the control group, children attending the treatment preschool:

• scored significantly and substantially higher on standardized aptitude tests administered in the preschool years;
• demonstrated significantly higher vocabulary skills through the two preschool years and two years beyond preschool;
• evidenced somewhat higher scores on aptitude and achievement measures and teacher ratings of academic potential from kindergarten to fourth grade;
• were less likely to be placed in special education programs (through age 14) or retained in grade (through grade 4);
• showed a significant decrease in self-reported delinquent behavior at age 14 and officially reported crime and delinquency at age 19;
• had significantly higher grade point averages and were more likely to graduate from high school, enroll in further training or education, and be employed at age 19;
• had significantly better educational outcomes, averaged significantly fewer lifetime criminal arrests, and had higher mean monthly earnings at age 27.

A reanalysis of the data from ages 19 to 40 found that the intervention:

• mostly helped education and early employment of women (ages 19 and 27) and later-life income, employment, and criminal activity of men (ages 27 and 40).

Evaluation Methodology

Design: From 1962 to 1967, 123 academically high-risk preschool-age children (ages 3 and 4 years old) were randomly assigned to attend the Perry Preschool Program or to a control group. This group comprised five successive waves of children who entered the program one year apart. Pairs of children matched on initial IQ were randomly split between the groups, then reassigned if necessary to equate the groups' sex ratio and average SES. When the experimental and control groups were compared on the three assignment criteria - initial Stanford-Binet score, SES, and sex - there were no significant differences. The only significant difference between the two groups was in mother's employment: there were more working mothers in the control group than the experimental group. Further analyses showed that children of working mothers scored higher than other children on the study outcome measures, thus this group difference favors the control group. A group of four year olds were selected during the first year for only one year of preschool. However, for all other waves, 3 year olds were selected for 2 years of preschool. Children in the control group had no intervention other than annual testing. Both groups entered the same public schools at age five.

Sample attrition was slight over the long course of the study. Five children from the original sample of 128 children were lost. The median rate of missing data across all measures has been only 5 percent. In no year following preschool through fourth grade were fewer than 90% of all children located and tested.

Children in the experimental group attended preschool for half-days, five days a week from mid-October through May. The staff/child ratio was approximately 1:6 for all waves. In all waves, experimental group children and their mothers received weekly home visits by the preschool teachers.

Sample: The sample of children was selected from African American, low-socioeconomic families located within the attendance area of Perry Elementary School in Ypsilanti, Michigan. In order to be eligible for the study, children from these families had to score between 50 and 85 on the Stanford-Binet Intelligence scale.

Measures: Several instruments were utilized to measure academic achievement including the following: The Stanford-Binet Intelligence Scale, Illinois Test of Psycholinguistic Abilities, Peabody Picture Vocabulary Test, Leiter International Performance Scale, California Achievement Test, Pupil Behavior Inventory (teacher rating scale), and the Ypsilanti Rating Scale (teacher rating scale). The teacher rated Pupil Behavior Inventory and the Ypsilanti Rating Scale were employed to assess the social and emotional development of children. Two instruments were used to gather information about the home environments of all the children in the sample: Cognitive Home Environment Scale and the Maternal Attitude Inventory. For each of the four years following kindergarten children were placed into one of three categories based on their school success: on grade, regular classroom; retained in grade, regular classroom; or placed in special education program. Children who were on grade in regular classrooms were considered "successful."

Analysis: Findings for the preschool period and posttreatment through age 40: ANOVAs and regression analyses, and their nonparametric analogues, were performed to determine whether those who attended preschool did better than those who did not attend preschool.

Outcomes

Preschool Period: Children who attended preschool scored significantly and substantially higher than control-group children on standardized aptitude tests administered in the spring of their first and second preschool years. No evidence was found that boys and girls in the experimental group benefited differentially from the preschool experience. There was also no indication of significant consistent differences among experimental group children belonging to different waves.

Posttreatment (Kindergarten - Fourth Grade): Children from the experimental group consistently evidenced somewhat higher scores than children in the control group on aptitude and achievement measures and teacher ratings of academic potential from kindergarten to fourth grade. While the differences remained significant through third grade on some of the aptitude measures, the magnitude of the measures tended to decrease once treatment ended. On achievement measures, the magnitude of differences between the experimental and control group tended to increase through fourth grade. On teacher ratings of academic potential the experimental children were not significantly rated higher. Although the overall pattern of differences in teacher ratings of social and emotional maturity favored the experimental group after kindergarten, these differences were significant approximately 1/3 of the time from kindergarten to fourth grade and 50% of the time from grade one to grade four. However, those differences that were significant and favored the experimental group tended to become somewhat stronger over time. Children who attended preschool were also less likely than control group children to be retained in grade or placed in special education programs. No evidence was found that boys and girls in the experimental group benefited differentially from the preschool experience. There was also no indication of significant consistent differences among experimental group children belonging to different waves.

Posttreatment (Ages 14-15): Sample attrition was fairly low at this follow-up and varies by measurement instrument. The following is a list of the number of nonmissing cases for each measurement instrument: age 14 IQ, 110 (89%); age 14 achievement 95 (77%); youth interview 99 (80%); parent interview 102 (83%). When background variables such as sex ratio, initial cognitive ability, socioeconomic status, proportion of single parent families, and mothers' years of school were analyzed, groups were equivalent on all variables except mother's schooling for the school achievement test. Further analysis showed the experimental group mothers with the least schooling were not represented among those who took the test. However, the experimental- and control-group members who took the test did not differ significantly in their mothers' years of schooling.

School achievement and attitudes about school: At age 14, there were no significant differences between the experimental and control group on IQ. Preschool education appears to have contributed to school achievement during middle school. At age 14, there was a highly significant difference with 8 percent of items passed in favor of children who attended preschool. Significantly fewer youth in the experimental group received special education services, compared with youth in the control group. There were no significant differences in the number of years retained in grade.

Students in the experimental group rated their school motivation marginally higher than students in the control group. There was no significant difference between the groups on school potential. However, there was a marginally significant difference between the two groups on the values placed on schooling at age 15. Based on a 7-item scale, youth who attended preschool placed a greater value on schooling than those in the control group at age 15. When these items were examined individually, a significantly higher percentage of youth in the experimental group agreed with the statement, "All people should at least have a high school education," compared with youth in the control group. On self-ratings of one's position compared to peers, youths who had attended preschool rated themselves higher on school ability but not on "how smart you are." The groups did not differ in youths' satisfaction with their own school performance.

Parents reported that youth who had attended preschool were more likely to talk about what they were doing in school, than those in the control group. Also, youth who had attended preschool were more likely to do their homework and spent more time doing it, than those in the control group. When their children were 15 years old, a significantly higher percentage of experimental-group parents (over 50%) expressed satisfaction with the school performance of their children, compared with 28% of control-group parents. Parents in the experimental group had higher educational aspirations than the control group but equivalent educational expectations.

Deviant behavior: Preschool education led to a significant decrease in teenagers' self-reported delinquent behavior. The experimental group scored significantly lower than the control group on the serious delinquency scale. There were no significant gender differences on total delinquency. However, boys had significantly higher scores than girls on the serious delinquent behavior scale and three of the categories of delinquent behavior: taking by force from a person, group fight, and damaging personal property.

Teenage employment and general social patterns: There were no statistically significant differences between the two groups on this measure. However, a higher proportion of the experimental group (29%) reported being employed, compared with the control group (16%). There were no significant group differences among the following variables classified as general social patterns: general self-concept, general parent-youth relationship, social patterns of youth, and time use by parents for prosocial activities.

Posttreatment (Age 19): Attrition in the study sample was very low at age 19. The median rate of missing data across all measures has been only 5 percent and the age-19 Young Adult Interview was administered to all but 2 control group study participants. School records were found for all but 11 (4 in the treatment group and 7 in the control group). However, high school grade point averages were obtained for only 64% of the study sample. They were not available for individuals who had dropped out of school in ninth or tenth grade or who had changed schools during high school.

School success: Youths who attended preschool had significantly higher grade point averages than those in the control group. There was no difference between the two groups in the extent to which individuals in the study were retained in grade. Students in the preschool group spent fewer school years in special education classrooms than students in the control group. Individuals who went to preschool spent more time receiving remedial support, whereas those who did not were more frequently classified as mentally retarded. Although the group difference was only marginally significant, preschool attendance led to a more favorable attitude toward high school at age 19. Individuals who attended preschool were significantly more likely than the control group to graduate from high school and to enroll in some form of further education or vocational training after graduation from high school.

Competence in everyday life skills was assessed by the Adult Performance Level Survey (APL). The APL Survey is a multiple-choice test designed to assess skills needed for educational success in modern society. Persons who attended preschool scored significantly higher than the no-preschool group on the APL.

Socioeconomic success: Preschool led to greater socioeconomic success for study participants. An increase in levels of employment of persons who attended preschool is demonstrated by higher proportions of persons working at the time they were interviewed, by more months of employment at ages 18 and 19, and by fewer months of unemployment since leaving high school. As a result of differences in levels of employment, marginally higher earnings are reported at age 19 by the preschool group. Also, a greater degree of economic independence is indicated by a higher proportion of persons in the preschool group supporting themselves on their own (or their spouses') earnings, and by a lower incidence of certain kinds of social services (e.g.; welfare and general assistance). There were marginal differences favoring the treatment group in terms of their reports of saving money and with regard to satisfaction with work. There were no differences between groups in either current-job or aspired-future economic status.

Social responsibility: Based on information obtained from official police and court records and self-report data, the preschool group had lower crime rates and less delinquent behavior than the no-preschool group, as indicated by fewer arrests, fewer cases sent on to juvenile court, fewer months on probation, and fewer persons fined as adults. The proportion of offenders (at least one time) in the preschool group was 31 percent, compared with 51% of individuals in the control group. Those individuals who attended preschool scored marginally lower on the self-report serious delinquency scale. There were no significant differences between groups on the mean seriousness score for arrests across all scored offenses. Fewer pregnancies and births through age 19 were reported by females in the preschool group. Subjects from the preschool group reported undertaking activities for family and friends more frequently than did no-preschool subjects. However, there were no differences between groups on a number of variables dealing with self-esteem, perceptions of health, and use of leisure time.

Posttreatment (Age 27): Data for this study was collected from four sources: an interview with study participants at age 27, school records, crime records, and social services records. At the age-27 follow-up, 117 of the 121 participants who were still living completed interviews and official-records were reviewed for all 123 participants.

Educational performance: On average, the program group completed almost a year more of schooling than the no-program group did. The program group spent significantly fewer school years in educable mental impairment programs, and significantly more school years in compensatory education programs than did the no-program group. In comparison with no-program-group females, program-group females, completed a significantly higher level of schooling and had a significantly higher rate of high school graduation or the equivalent. On a test of general literacy, the program group significantly outscored the no-program group in health information and problem solving, but not in general literacy.

Delinquency and crime: As compared with the control group, the treatment group averaged a significantly lower number of lifetime criminal arrests (2.3 vs. 4.6) and a significantly lower number of adult criminal arrests (1.8 vs. 4.0). Significantly fewer program-group members were frequent offenders - arrested 5 or more times in their lifetimes (7% vs. 35%) or as adults (7% vs. 31%). The treatment group also had significantly fewer arrests for drug-making or drug-dealing crimes (7% vs. 25%). Compared to the control group, treatment group males were arrested significantly fewer times over their lives (3.8 vs. 6.1 arrests, with 12% vs. 49% arrested 5 or more times) and as adults (3.0 vs. 5.4 arrests, with 12% vs. 43% arrested 5 or more times as adults). There were no significant differences between self-reported crime. The only significant difference between the two groups was the amount of time spent on probation, which was significantly lower for program group members compared with no-program group members.

Economic status: Based on age 27 interview data, the treatment group as compared to the control group, had higher mean monthly earnings and had fewer members who had received social services in the previous 10 years. According to social service records and interviews at age 27, a significantly smaller percentage of the program group had received social services sometime in the previous 10 years and significantly fewer program group members reported receiving government assistance, compared with the no-program group members. Significantly more program group members than no-program group members owned their own homes or a second car.

Family formation, health, and social relations: At age 27, significantly more program females than no-program females were married. Program group husbands averaged significantly more years married than did no-program-group husbands. There were two significant differences in childrearing practices between the two groups. Significantly more children of program group parents than no-program group parents regularly used library cards. Also, significantly fewer program parents than no-program parents said their children were turning out better than they had expected them to. The two groups did not differ significantly in their social relations or community involvement. Significantly more program group than no-program group members had been hospitalized in the 12 months prior to their age 27 interview. Members of the program group were significantly more likely to report that they found it very easy to work or study hard all day and they found it very easy to feel close to family and friends, compared with the no-program group. Significantly more of the program group members than members of the no-program group have the goal of travel, adventure, or changing residence in the next five years. Compared to the no-program group, significantly more program group members identified work related problems.

Posttreatment (age 40): With 112 of the original 123 participants interviewed at age 40, Schweinhart et al. (2005) found that the program group had significantly better outcomes, using one-tailed tests of statistical significance, in adulthood on several outcomes: highest level of schooling completed (p < .10), current employment (p < .10), earnings (p < .10), and arrests (p < .10). Statistically significant effects (.05 two-tailed) showed for highest level of schooling completed (females only), having a savings account, arrests for property crimes, and arrests for drug felonies. However, the vast majority of the group comparisons failed to reach standard levels of statistical significance.

Reanalysis (age 19 to 40): A reanalysis of the HighScope data up to age 40 by an independent team of economists adjusted for several methodological and statistical limitations of the original studies. Heckman et al. (2010) noted several problems in the design and previous analyses: 1) deviations from randomization, 2) multiple hypothesis testing of non-independent outcomes, 3) assumptions of normality for small samples, and 4) possible lack of generalizability of the sample. The study made statistical adjustments to deal with the first three limitations and evaluated the representativeness of the sample to deal with the last.

The authors reanalyzed the original data for about 76 outcomes for males and 71outcomes for females. The outcomes included measures of education, health, family, crime, employment, earnings, and economic status. They came from multiple assessments and covered multiple ages (19, 27, and 40).

For each of the outcomes, the study calculated significance probabilities adjusted for deviations from randomization, small sample size, and multiple hypothesis tests. It then compared naïve significance levels that come close to the usual tests of significance with the adjusted significance levels.

Results. The authors first summarized previous results of the program for 715 outcomes: "At the 5% significance level, we obtain a 23% overall rejection rate (13% for males and 22% for females)." The figures suggest about a quarter of the tests showed significant program effects using standard statistical tests.

The reanalysis then demonstrated that the adjusted tests largely confirmed the effectiveness of the program. The authors (p. 42) summarize their findings:

"Males exhibit statistically significant treatment effects for criminal activity, later life income, and employment (ages 27 and 40), whereas female treatment effects are strongest for education and early employment (ages 19 and 27). There is, however, a strong effect of the program on female crime at age 40. The general pattern is one of strong early results for females, with males catching up later in life."

Still, the program influenced relatively few outcomes at adult ages compared to the number of outcomes studied. Using a .05 two-tailed standard, 9 of 71 tests for females reached significance after adjustment, and 4 of 76 tests for males reached significance after adjustment.

Lastly, comparisons of the disadvantaged, largely African American HighScope sample to national samples of disadvantaged children showed similarities. The authors concluded, "The Perry sample is representative of disadvantaged African American populations."

Study 2

Summary

Epstein (1993) conducted a quasi-experimental design study with 200 children.in 26 preschool programs (13 HighScope and 13 comparison). Data were collected at a single time point, presumably posttest, and measured initiative, social relations, creative representation, language and literacy, logic and mathematics, and music and movement.

Epstein (1993) found, compared to control children, treatment children had:

• had more initiative and better social relations but not higher scores on cognitive indicators

Evaluation Methodology

Design: Approximately 8 children were assessed in each of 13 HighScope and 13 comparison programs in the state of Michigan. The programs selected were located in urban and rural settings, operated within a wide range of agencies, and served children across a broad spectrum of socioeconomic backgrounds. Teachers provided the names of children according to random selection procedures. The sample was composed of 200 children, 48.5% from HighScope and 51.5% from the comparison group. The program and comparison groups were closely matched on demographic variables. Child assessments were collected in 1991 over a three-day period. Nearly a third of the children could not be rated on one or more of the Child Observation Record (COR) items. However, there were no significant differences by group between children with complete versus incomplete observational ratings.

Sample: The sample was composed of 200 children, from 15 agencies operating in 21 sites. Children were observed in a total of 26 different program settings. Nearly half came from Head Start programs, a fifth from public schools, and over a third from other non-profit agencies. Males and females were equally represented in the total sample and in each group. The mean age of the children in each group was 4.4 years. Seven of the 15 agencies used to draw the sample served predominantly minority or mixed populations. Forty-three percent of the children were white, 32% Black, and 25% were from another minority group. On average, parents in each group had completed high school and attended 1 to 2 years of college.

Measures: The Child Observation Record (COR) was used to assess child-initiated behaviors in six areas of development: initiative, social relations, creative representation, language and literacy, logic and mathematics, and music and movement. The Developmental Indicators for the Assessment of Learning - Revised (DIAL-R) was used to assess the behavior of children in three areas: motor, concepts, and language.

Analysis: Children's development in the HighScope versus comparison programs were compared using t-. tests. Pearson correlation coefficients explored the relationship between program quality and child outcomes.

Outcomes:
For the DIAL-R items, there was only one significant difference between the HighScope children and the comparison group, while two other items approached significance. HighScope children scored significantly higher on naming verbs and approached significance on naming nouns, suggesting that children in the HighScope group have better language skills than did comparison children. The HighScope group also scored marginally higher than the comparison group on the cognitive task of sorting. Children in the comparison group exhibited significantly more problems during the DIAL-R motor testing and the language test versus children in the HighScope group. The HighScope group rated significantly higher than comparisons on three of the six COR subscales: initiative, social relations, and music and movement. While children in the HighScope program did significantly better than their comparisons on socioemotional indicators, there were no significant differences between the two groups on the cognitive indicators.

Study 3

Summary

Schweinhart et al. (1997, 1986) conducted a randomized controlled trial in which 68 children ages 3 and 4 years old were assigned to conditions that included a HighScope curriculum group, a direct instruction programming group, and a standard nursery school control group.  An additional no-treatment control group (no preschool program experience) was randomly drawn from a subset of the population used to select the first three condition groups. Data measuring intellectual and school performance and social behavior were collected from ages 3 to 15 and at age 23.

Schweinhart et al. (1997, 1986) found, compared to control children, treatment children showed:

• non-significant effects on official records of delinquent behavior at age 15 but significant effects on felony arrests at age 23

Evaluation Methodology

Design: 3- and 4-year-olds between 1967 and 1970, who were at risk of failing in school (according to test scores) and of low socioeconomic status were randomly assigned to one of the three following preschool curriculum programs.

• Direct Instruction programming: This curriculum is based upon behavioral theory and is a programmed learning approach, in which the teacher initiates activities and the child responds to them.
• HighScope Cognitively Oriented Preschool Curriculum: This program is an open-framework approach, based on cognitive-developmental theory, in which teacher and child both plan and initiate activities and actively work together.
• Nursery School Program: This approach is child centered and based upon psychoanalytic theory. Here, the child initiates and the teacher responds and there is an emphasis on free play, similar to the set up of nursery school for younger children.

Assignment was not completely random, however. Each year, children in the incoming wave were randomly assigned to three groups and reassignments were made until the groups were similar on race, gender, and mean IQ. The groups were then randomly assigned to the three curriculum models. Siblings were assigned to the same models, though this had an effect on only 9 of the 68 participants. There were 23 in the Direct Instruction group, 23 in the nursery school group, and 22 in HighScope; 27 were in the 1st wave, 19 in the 2nd, and 22 in the 3rd.

These publications report on program effects at age 15 and age 23. Assessments were conducted yearly through age 8 and then again at ages 10, 15, and 23. This study also utilized a no-treatment control group. These are children from the same community who had no preschool program experience. They had been randomly assigned to the control group of the Perry Preschool study and were selected according to the same criteria used to select the other 3 preschool curriculum participants. This control group, however, was not generated from the entire population of the city, as the other 3 groups were. Instead, it came from one single elementary school and contained only black children. Comparisons are made only on standardized measures of IQ.

Sample: Children in the sample were born in Ypsilanti, Michigan between 1964 and 1966. 65% were black and 54% were female. 26% were from single parent families and the average IQ was 78.3.

Measures: Intellectual and school performance measures include IQ testing given during preschool, achievement scores based on the California Achievement Tests given at the end of first, second, and fourth grades, and the Adult Performance Level Survey (APL) given at age 15. The APL survey measures a person's competence in solving real-world problems and coping with the cognitive demands of adult life.

Measures of social behavior included an 18-item self-reported delinquency scale divided into five subscales: personal violence, property violence, stealing, drug abuse, and status offenses. Other social variables included measures of the family domain, participation in extracurricular activities, and expectations of future education.

Analysis: Percentages or frequencies of categorical variables are presented for each group, with differences evaluated by chi-square analysis. Means of continuous variables are presented, with group differences tested by analysis of variance.

Outcomes:
Attrition and Baseline Group Equivalence: The 3 curriculum groups were equivalent at baseline on race, gender, socioeconomic status, number of single parent families, father's employment, mother's employment, father's years of school, number of persons per household and per room, and child IQ. However, groups were significantly different on mother's years of schooling, with mothers of the nursery school children having the most education (average 10 years). Comparing the overall sample (n = 68) to the no-treatment control group (n = 65), there were no significant differences on mother's years of schooling or on child IQ at entry. The no-treatment control group, though, had significantly more Black participants, as all the sample was Black, and were of a significantly lower socioeconomic status.

Of the 68 participants, 54 (79%) were retained at age 15 and 52 were retained at age 23 (76.5%). Data collections from age 3 through 10 achieved retention rates of 90% or better. All three groups in both the age 15 and the age 23 samples were not significantly different from each other on baseline measures, including group assignment, race, gender, socioeconomic status, family composition, parental employment, father's schooling, persons per household or room, or child IQ. Groups remained significantly different on mother's years of schooling, with the nursery school mothers having the most education.

It should be noted that the first cohort of 8 members of the Direct Instruction group and 8 members of the Nursery School group attended only one year of their preschool programs, while the first cohort of HighScope children attended for two years.

Posttest through Age 10: The mean IQ for the combined Curriculum sample was significantly higher than that of the no-treatment control group at each assessment point through age 10 (though the difference at age 8 was only marginally significant). When the mean IQs of the three preschool curriculum groups are compared, the differences between them are quite small. Only one significant difference emerged at age 5, favoring the Direct Instruction group, followed by the HighScope and then the nursery school groups. For the California Achievement Tests at the end of first, second, and fourth grade, each of the groups gained between 50 and 60 points between testings, but the groups were not significantly different from one another. For the APL Survey given at age 15, the Direct Instruction group scored significantly lower than the other two groups on occupational knowledge and the Direct Instruction group scored lower on writing skills as well (effect was marginally significant, p = 0.06).

Age 15: Overall, the Direct Instruction group reported engaging in significantly more acts of delinquency (13) than the nursery school group (7) and the HighScope group (5). There were no significant group differences on the personal violence or stealing subscales, but there was a marginally significant difference on the drug abuse subscale (p = .06) and significant differences on both the property damage and status offenses subscales, all showing the highest involvement for Direct Instruction youths, compared to both the nursery school and the HighScope groups. Within these subscales, there was a significant difference between groups on arson (marginally significant at p = .06) and running away from home. There were no significant differences between groups on official records of delinquent behaviors or on self-reported mental health.

Direct Instruction youth were also significantly more likely to report that their family members felt they were doing poorly. There were significantly less likely to participate in sports, to have read a book in recent weeks (marginally significant, p = .09), and to have been appointed to an office or job at school.

Age 23: Analysis of school outcomes revealed the HighScope group scored significantly higher than Direct Instruction youth on highest year of schooling planned, but there were no differences in actual schooling completed or in graduation rates. There were no differences in most measures of cognitive functioning (years identified as special ed, educably mentally impaired, or learning disabled) or in truancy, number of failed classes, or repeating grades. However, Direct Instruction youth spent significantly more years identified as emotionally impaired or disturbed than either of the other two groups and the HighScope group spent significantly more years than nursery school youth in compensatory education. There were no differences in literary test scores, or on the majority of measures of household, employment, and income, though HighScope youth were significantly more likely to live with a spouse.

Analysis of measures of prosocial activities, misconduct, and arrests revealed more program effects. Direct Instruction youth were significantly less likely than the other two groups to have ever done volunteer work and reported significantly more sources of irritation than the HighScope group. HighScope youth were significantly more likely than the other two groups to have voted in the last presidential election.

Though program effects on self-reported delinquency were significant at age 15, there were no effects at age 23. While the mean rank on acts of misconduct at age 23 was significantly lower for HighScope than Nursery School youth, the variable did not meet assumptions for analysis of variance and the appropriate, nonparametric analysis was used. Analysis of arrest records revealed significantly more felony arrests for the Direct Instruction group, compared to both HighScope and to Nursery School. Direct Instruction youth also had significantly more felony arrests at ages 22 and above, compared to Nursery School youth, and significantly more arrests for property crimes, compared to HighScope youth. There were no effects on arrests for violent crimes or crimes involving drugs.

Curriculum-group-by-gender analyses were conducted on the 13 variables for which significant group differences were found. The majority of program effects were not impacted by the gender composition of the groups. Effects on years of identified emotional impairment or disturbance, times suspended from work, highest year of schooling planned, sources of irritation, misconduct at age 15, felony arrests, felony arrests at ages 22 and above, property crimes, living with spouse at age 23, and volunteer work remained significant when accounting for gender. Effects on years of compensatory education, misconduct at age 23, and voting in the last presidential election, however, appear to be due to the differences in gender makeup, rather than preschool curriculums. Direct Instruction males were worse off than any other gender-curriculum group in emotional impairment or disturbance, suspensions from work, sources of irritation, and felony and property crime arrests.

Finally, there is some evidence of mediating variables, as felony arrests were correlated significantly, but moderately, with years of emotional impairment/disturbance (r = .54) and self-reported misconduct at age 15 (r = .30).