BARBARA K. IVERSON HERBERT J. WALBERG University of Illinois at Chicago Circle

From a systematic search of educational, psychological, and sociological literature, 18 studies of 5,831 school-aged students on the correlation of home environment and learning in eight countries over a 19-year period were selected. Correlations (the units of analysis) of intelligence, motivation, and achievement with indexes of parent stimulation of the student in the home are considerably higher than those with indexes of socioeconomic status (SES); specifically the medians (and ranges) of 92 simple correlations of home environment and learning are .37 (and .02 to .82) and of 62 multipleregression-weighted composites are .44 (and .23 to .81). Jackknifed regression estimates indicate that gender and SES of the sample affect the sizes of the correlations and suggest priorities for future primary investigations. The analyses suggest that ability and achievement are more closely linked to the sociopsychological environment and intellectual stimulation in the home than they are to parental socioeconomic status indicators such as occupation and amount of education.

THE SOCIAL AND PSYCHOLOGICAL STIMULATION of the child's academic development by parents and other significant persons in the home environment appears to be an important influence on academic ability, achievement, and motivation (3, 9, 19, 29); but, the research has not been quantitatively synthesized to estimate the average correlation between measures of home environment and learning and the dependency of the size of the correlation on characteristics of the samples and research methodology. Drawing on the work of Gage (10), Light and Smith (15), Rosenthal (25), and on quantitative techniques for research synthesis, our purpose is to provide a quantitative summary of the research to consolidate the diverse studies within the research domain of family environment in relation to schoollearning variables.

For historical and theoretical perspective, four approaches to the measurement and study of home environments in relation to academic learning may be distinguished: sociological surveys that include socioeconomic (SES) measures such as parental education, income, and occupation; family-constellation studies that analyze the number, birth order, and spacing of children in the family; the work of the "British school" that emphasizes parental experiences and aspiration for the child, objects and material conditions in the home, and other status variables; and the work of the "Chicago school" that emphasizes specific social-psychological or behavioral processes thought conducive to learning. These four approaches by no means represent opposing views but do constitute fairly distinctive and somewhat separate research traditions. White (29) concluded that measures of home environment account for six times as much variance in achievement scores as traditional SES measures, however.

Research on family constellation also shows low predictability of learning. The typical correlation of the number of children in the family ("sibsize") with academic achievement is .25 (3, 27). With very large samples, birth order and spacing are significant cor

relates of achievement in some work, but their correlations with learning are considerably smaller and more unstable than those involving SES and sibsize.

By the standards of predictive validity and psychological theory, family SES and constellation are less valid, but also less expensive proxies for aspirations, conditions, and processes in the home that are conducive to learning. Walberg and Marjoribanks (31) and Marjoribanks (19) review several studies that show that regression-weighted composites of homeinterview measures of parental characteristics and behavior correlate up to .80 with verbal achievement measures. These reviews also show that SES is only weakly to moderately associated with measures of the home environment. Therefore, this review focuses on studies of the relationship of home environment and school achievement which investigate parents' psychological attitudes and behaviors.

Marjoribanks (19) distinguished the Chicago and British schools of research on home environments. In dissertations directed by Benjamin Bloom (2) at the University of Chicago, Davé (5), and Wolf (34) developed lists of parental behaviors and parent-child interactive behaviors that seemed likely to foster intellectual growth. These "process" variables are specific and changeable; and ratings of them are made by trained home interviewers who ask such questions as: "Do you read to the child?," "Do you discuss his grades with him?," and "Who makes the plans (for family vacations)?" Sets of process variables are summed to provide indexes of "presses" in the home environment; for example, the instrument called the Index of Educational Environment includes academic guidance, achievement, activeness of family, intellectuality of the home, work habits of the family, and language models, all of which were hypothesized as important influences on academic achievement. Similar instruments used in other Chicago school research have focused on the presses for academic guidance, achievement (both for the child and parent), and activeness of the family because these aspects of home environment seem most readily influenced by intervention programs (6). The other presses, language models and intellectuality seem less changeable, involve parent status more than behavior, seem less closely associated with achievement, and are, therfore, not measured in some later Chicago studies (14, 17).

Studies within the British school also attempt to develop valid measures of the home environment (4, 8, 18, 23, 27, 33), but they focus on parental experiences and attitudes, and material conditions in the home rather than on behavioral processes (19). These studies use a variety of home assessment measures such as the Survey of Parents of Primary-School Children (24). Typical questions include: "What do you feel about the ways teachers control the children at (present school)?," and "Has the head teacher, or any other teacher talked to you about the methods they use at (present school)?" Such questions focus on parent attitude and experiences more directly than on parental practices. Fraser (8), who used reading habits of the parents as a home environment measure, and Claeys and DeBoerk (4) and Schaefer (27), who used the Parent Attitude Research Instrument (26) as a home environment measure, classify as studies within the British school (see Table 1). Studies by Kifer (14), Shea (28), and Marjoribanks (17, 18) used modified versions of the Index of Educational Environment (5), and, therefore, they are considered part of the Chicago school. Other Chicago studies include Wolf (34), Dyer (7), Mosychuck (22), Weiss (30), Keeves (12), Kellaghan (13), Marjoribanks (19), and Dolan (6). In addition to the purposes of the present study stated in the opening paragraph--to provide an estimate of the typical correlation of home environment with learning measures across studies and to indicate which sample and study characteriatics are associated with different magnitudes of the correlation--an effort was made to determine if the Chicago and British studies differ in their predictive validity. Such a difference would indicate whether status or process characteristics of the home bear a stronger relation to school achievement.

The 13 references in a recent review by Marjoribanks (19) were the starting points for searching for home environment studies. A search was made of the journal, Child Development for the years 1976 through 1978, the Social Science Citation Index for studies published in 1976 through 1977 that cited earlier work, the Educational Resources Information Center (ERIC) under the descriptors "family environment'' and "family influence," and the references in recent research. All 18 studies that reported simple or multiple correlations of home environment with ability, achievement, or motivation measures for schoolage (first through twelfth grade) samples were selected for analysis. The characteristics of the studies are shown in Table 1.

Eight items of information were recorded for each correlation: the size, age, sex, and socioeconomic status (SES) of the sample; the home-assessment and learning measures employed; Chicago or British school; and type of correlation (Table 2). Several coding procedures require comment. Since some

studies grouped boys and girls in calculating correlations, and others separated them, three coding categories for sex--boys, girls, and mixed--were employed. Where SES was specified, it was coded based on the author's report. Those studies with an unspecified SES were not included in the analysis of the influence of SES on school achievement.

In the analysis of variance, the eight factors were employed as nominal variables as grouped and indicated in Table 2. In the regressions, however, the continuous variables, age and sample size, were left in their full metric precision without grouping in order to take the varying ages and sample sizes into account. The nominal factors were recoded to sets of binary (0, 1) variables to assess the possible effect of the value of each nominal factor on the correlation.

Since the number of correlations varies among studies, studies with greater numbers of correlations are weighted more heavily than others when each correlation is given equal weight as in the correlationweighted analyses. Therefore, analyses were performed where each correlation was given a weight inversely proportional to the total number of correlations in the study from which it was taken; for example, Fraser's (8) six correlations each received a weight of 1/6. This is referred to as study weighted. Although weighted regressions treat each study equally, they do not remove statistical dependencies among the correlations within each study that violate inferential assumptions. The jackknife procedure which tests for independence among the observations (11), was employed, as explained below, to provide

stringent, independent estimates of the regression coefficients and their standard errors. The results of the jackknifed analyses can be compared to other results in terms of the consistency of signs and strengths of relations.

Table 1 shows the chief characteristics of the samples and methodologies of the 18 studies that resulted from the search and selection procedures. The samples range in size from 15 to 3,092; and the grand total across studies is 5,831. There are 92 sample correlations with a median of .37 and a range from .02 to .82 and 62 multiple correlations with a median of .44 and a range from .23 to .81.

Table 2 shows the univariate, study-weighted comparisons for each of the eight factors on correlations of learning with simple indexes and multiple regression-weighted composite indexes of home environment. The sizes of the multiple correlations are significantly related to: SES, type of learning criterion measure, sample size, age, and type of home measure. For simple correlations, learning criterion measure, SES, school, age, and nationality showed significant differences. These results suggest the following interpretation. First of all, that SES should be specified in studies of the home environment, rather than unspecified, insofar as for multiple correlations, lower SES (R = .65) and middle SES (R = .65) means are similar while the mean value in studies in which SES was not specified was lower (R = .45). The substantive meaning of SES in its relationship to home environment and school learning, however, is unclear, and will remain a confounding effect until researchers specify and include it on their analyses.

Secondly, the results suggest that home factors are differentially related to different kinds of achievement (e.g., language, reading, arithmetic are more highly correlated than IQ or intelligence are with home measures). Another differential impact suggested by these univariate statistics is that of age, where multiple correlations are highest (.62) for 9-11 years olds (vs. .30 and .41 for 5-8 year olds and 12-15 year olds), a finding partly reinforced by the mean simple correlations for the three age groups (see Table 2).

Chicago school studies, assessing direct parent-child interactions, appear to be better measures of educationally relevant home factors than British school studies which focus on parental attitudes, habits, and beliefs as measures of home environment, though other methodological differences between studies in each category may account for the higher simple and multiple correlations of the Chicago school. The multiple correlations by sample size may suggest that the studies differ in precision of execution as they get larger, or that studies done on large samples tended to use British school assessment instruments, for example, another possible confounding effect. These univariate results are suggestive, but the conclusions need to be tested more stringently. In order to do this, a series of regressions were computed (Table 3). The first regression was study weighted where each study was given equal weight, despite the number of correlations reported for any given study. From this studyweighted regression of the correlations on the complete set of 25 variables (the two continuous variables and the 23 binary-coded factors), those with t values less than one were first deleted from regressions since they make no independent contribution to the accountable variance. The next regression deleted variables with t values less than two, which are below the approximate .05 significance level (shown in the second double column of figures in Table 3). To control for the unequal number of correlations between studies, a correlation-weighted regression (this weighting process is described earlier in the paper)was run on the reduced set of variables from the best study-weighted model. The results obtained from the correlation and study-weighted regressions are quite similar as the first two double columns of Table 3 show. Alternative weighting, by either correlations or studies makes little difference in the magnitudes of the metric regression weights and their t values, all eight of which are significant at the .05 level on 150 degrees of freedom, although it increases the accountable variance (R²) somewhat to use study weighting. These regression models suggest the following about the correlations between home environment and school achievement: that they are slightly higher for older students, slightly higher where students are identified by gender (rather than leaving gender unspecified), higher for middle SES students, higher in Chicago school studies and that, as would be expected, multiple correlations are slightly higher than simple correlations. Additionally, these results suggest that intelligence measures and motivation measures are related less strongly to home environment conditions than are achievement or reading or math measures.

The problem of interdependence among the variables needs to be noted at this point, however. Even though correlations are calculated on samples of dozens or hundreds of students in each study, the assumption of their independence can be questioned. For example, while a given individual may have a unique home measure or achievement test score, the size of the sample, nationality, or categorization of the study he or she is participating in, will be shared with all the other individuals from the study. The variation of these variables is constrained through this lack of independence. Using the jackknife technique controls for this lack of independence by comparing a

series of regressions, each of which leaves out one of the studies. When the regression with all studies is compared to the regression with a single study removed, the effect of the study will appear through its absence.

The jackknifed equation shows that correlations in studies where the characteristics of the subjects are specified, are significantly different than others. For example, correlations reported separately for boys (probability less than .05) and for girls (.10) are somewhat higher than mixed-gender samples. Also, correlations based on multiple-regression-weighted composites of home environment measures (.05) are higher than zero-order correlations. Because their t values are greater than one, the other four variables contribute uniquely to the accountable variance; but they are not significant; the weights for these variables indicate tendencies for older samples and Chicago (process-oriented) studies to yield higher correlations, and for correlations of home environment with intelligence and motivation to be lower than those with achievement measures. These substantive findings suggest that a process-view of the home can better describe educationally relevant variables than a status or other orientation toward home environment; that an instrument like the Index of Educational Environment (5), captures more information about the educative dimension of the home than does the Parent Attitude Research Instrument (26).

Overall, this quantitative approach to integrating the research on home environments and school learning has pointed to methodological issues as well as the substantive issues already discussed. For example, the jackknifed regression suggests that investigation ought to clearly identify both the gender and SES of students, rather than lumping students altogether in a single group. Also, the small regression weight associated with multiple correlations (.029) suggests that simple indices of home environment may be as efficient and effective as multiple indices in measuring important features of the home. In future research on the educational impact of home environment variables, more carefully specified studies oriented toward parent-child interactions, focusing on educationally modifiable variables (achievement vs. IQ or intelligence) can yield more detailed descriptions leading to specific, effective interventions designed to raise achievement through the involvement and influence of home environments. The correlational studies reviewed here form the basis for development and improvement of quasi-experimental or experimental studies of the relationship of home environment to school and achievement.

The quantitative synthesis suggests that academic ability and achievement are more closely linked to the measures of the sociopsychological environment and intellectual stimulation in the home than they are to parental socioeconomic status indicators such as

occupation and amount of education. More homogeneous samples with respect to age and sex appear to yield somewhat higher correlations. Studies that take into consideration multiple indications of the home environment also yield higher correlations. Although one-way causality cannot be inferred from the correlations, home environment variables, unlike socioeconomic status, are changeable and are worth not only further experimentation but merit contructive efforts to improve them, as well.

Anastasi, A. "Intelligence and family size". Psychological Bulletin, 1956 , 53, 187-209.

Bloom, B. S. Stability and change in human characteristics, New York: Wiley, 1964 .

Circirelli, Victor G. "The relationship of sibling structure to intellectual abilities and achievement." Review of Educational Research, 1978 , 48, (3), 365-379.

Claeys, W., & DeBoerk, P. The influence of some parental characteristics on children's primary abilities and field independence: A study of adopted children. Child Development, 1976 , 47, 842-845.

Dave, R. The identification and measurement of environmental process variables that are related to educational achievement. Unpublished doctoral dissertation, University of Chicago, 1963 .

Dolan, Lawrence. The affective consequences of home support, instructional quality, and achievement. Urban Education, 1978 , 13, (3), 323-343.

Dyer, P. B. A. Home environment and achievement in Trinidad. Unpublished doctoral dissertation, University of Alberta, 1967 .

Fraser, E. Home environment and the school. London: University of London Press, 1959 .

Freeburg, Norman E., & Payne, Donald T. "Dimensions of parental practice concerned with cognitive development in the preschool child." The Journal of Genetic Psychology, 1967 , 111, 245-261.

Gage, N. L. The scientific basis of the art of teaching. New York: Teachers College Press, 1978 .

Glass, G. V. Integrating findings: The meta-analysis of research. In L. S. Schulman (Ed.), Review of research in education, Vol. 5, Itasca: Peacock, 1978 .

Keeves, J. P. Educational environment and student achievement. Stockholm: Almquist and Wiksell, 1972 .

Kellaghan, T. "Relationships between home environment and scholastic behavior in a disadvantaged population". Journal of Educational Psychology, 1977 , 69, 754-760.

Kifer, E. "Relationships between academic achievement and personality characteristics: A quasi-longitudinal study". American Education Research Journal, 1975 , 72, 191-210.

Light, R. V., & Smith, P. V. "Accumulating evidence: Procedures for resolving contradictions among different research studies." Harvard Educational Review, 1971 , 41, 421-471.

Majoribanks, K. "Environment, social class and mental abilities." Journal of Educational Psychology, 1963 , 63, 103-109.

Majoribanks, K. Environments for learning. Slough, England: NFER Publishing Co., 1972 .

Majoribanks, K. "School attitudes, cognitive ability and academic achievement." Journal of Educational Psychology, 1976 , 68, 653-660.

Majoribanks, K. Family environments: Relations with children's cognitive performance and affective performance. In H. J. Walberg's Learning environments and effects. Chicago: National Society for Study of Education, 1979 .

Miller, G. W. Educational opportunity and the home. London: Longman Group Limited, 1971 .

Mosteller, F., & Tukey, J. W. Data analysis and regression. Reading, Mass: Addison-Wesley, 1977 , 133-162.

Mosychuk, J. D. Differential home environments and mental ability patterns. Unpublished doctoral dissertation, University of Alberta, 1969 .

Peaker, G. The regression analyses of the national survey. In B. Plowden (Ed.) et al., Children in their primary schools. London: H. M. Stationery Office, 1967 .

Plowden, B. (Ed.) et al., Children in their primary schools. London: H. M. Stationery Office, 1967 .

Rosenthal, R. Experimenter effects in behavioral research. New York: Irvington Publishers, 1976 .

Schaefer, C. S., & Bell, R. Q. "Development of a parent attitude research instrument." Child Development, 1958 , 29, 339-361.

Schaefer, E. & Clark, P. Parent-professional child interaction & involvement. A paper presented at Teacher Corps Conference at Michigan State University, 1977 .

Shea, J., & Hanes, M. The relationship between measures of home environment and school achievement of Follow-Through Children. Gainsville, Fla. University of Florida. 1977 (ERIC Document Reproduction Service No. ED 013 371).

Walberg, H. J., & Majoribanks, K. "Family environment and cognitive development; Twelve analytic models." Review of Educational Research, 1976 , 46, (4), 527-551.

Weiss, J. The identification and measurement of home environment factors related to achievement motivation and self-esteem. In K. Majoribanks (Ed.), Environments for learning. Slough, England: NFER Publishing Co., 1974 , 141-149.

White, K. R. The relationship between socioeconomic status and academic achievement. Unpublished dissertation. University of Colorado, 1976 .

Williams, Trevor. Abilities and environments. In W. H. Sewell, R. M. Hauser, & D. L. Featherman (Eds.), Schooling and achievement in American society. New York: Academic Press, 1976 .

Wiseman, S. The Manchester Survey, Appendix G. In "Children and their primary schools" (Vol. 2) Research and Surveys. London, England: H. M. Stationery Office, 1967 .

Wolf, R. M. The identification and measurement of home environment process variables that are related to intelligence. Unpublished doctoral dissertation, University of Chicago, 1964 .