School performance and physical growth of underprivileged children : results of the Bogota project at seven years

PHN Tec'hnical Notes RES 8 THE WORLD BANK SCHOOL PERFORMANCE AND PHYSICAL GROWTH OF 'UNDERPRIVILEGED CHILDREN: RESULTS OF THE BOGOTA PROJECT AT SEVEN YEARS June 1983 Population, Health and Nutrition Department * This paper is one of a series issued by the Population,* * Health and Nutrition Department for the information and* * guidance of Bank staff working in these sectors. The* * views and opinions expressed in this paper do not neces-* *sairily reflect those of the Bank.*  RES 8 A B S T R A C T An experiment providing various combinations of nutritional supplementation and "home education" for a cohort of children and their mothers through pregnancy and up to three years of age was conducted in Bogota. This study reports significant effects in children's growth, diet, and school performance four years after the termination of the project's interventions for both supplementation and home education interventions. Prepared by: M. G. Herrera and C. M. Super Harvard School of Public Health June 1983  School Performance and Physical Growth of Underprivileged Children: Results of the Bogota Project at Seven Years It has been repeatedly demonstrated that children malnourished early in life do not achieve normal levels of physical growth and cognitive development. The design and evaluation of interventions have been hampered by inadequate infor- mation concerning factors that covary with early nutrition and later develop- ment. In addition, the causal linkages between early nutritional, health, educa- tional and child care variables and subsequent development are poorly under- stood. Nutrition interventions have yielded variable results, in part because families utilize supplements as income to address needs considered more press- ing. Maternal education may, therefore, influence child growth and development by shaping faLmily attitudes toward greater value in meeting the child's needs. The present report summarizes follow-up results of an experimental interven- tion designed to study effects of food supplementation and maternal education on the growth and development of young children. The data indicate that four years after termination of nutritional and educational interventions significant effects can be found on growth, diet, and school performance.  2 Design of the Bogota Study The Bogota Study of Nutrition and Child Development, a multidisciplinary longi- tudinal study of 443 famil is at risk of malnutrition, was started in Bogota, Colombia in 1973 [1]. Food supplementation and a maternal home education program were utilized to alter selectively nutritional status and home environ- ment during the first 3 years of life. The experimental design of the Bogota Project consisted of a dual intervention study of children at risk of malnutrition extending from the sixth month of ges- tation through 3 years of age. Pregnant women whose families and offspring were study subjects were selected on the basis of two criteria: mothers were in the first or second trimester of pregnancy and families had one or more chil- dren under 5 years of age of whom at least 50 percent were malnourished. Families were randomly assigned to six experimental groups: A) No interven- tion; B) Supplementary feeding from 6 months postpartum to 3 years of age; C) Supplementary feeding from 8 months of gestation to 6 months of age; D) Sup- plexnentary feeding from 6 months of gestation to 3 years of age; Al) Maternal home tutoring designed to increase stimulation of the child-^but no- supplemen- tary feeding; D1) Supplementary feeding from 6 months of gestation to 3 years of age plus maternal education. All groups were given a uniform obstetrical and pediatric health-care program. The outcome data can thus be analyzed utilizing two 2x2 designs to assess the single and interactive effects of a) early and late food supplementation and b) supplementation and maternal education. Food supplements were provided in amouhts sufficient to meet a substantial portion of the recommended dietary allowance (RDA) for each member of the family. Supplements in the form of bread, powdered milk, and cooking oil were  3 given to Groups C, D, and Dl from the time of enrollment, and to group B from the time when the infant reached 6 months of age. Supplements were provided for every family member in amounts sufficient to close the gap between baseline diet and recommended dietary allowances (RDA). (Pregnant women were allot- ted 856 calories and 39 grams of protein, every other person over 1 year of age received 623 calories and 30 grams of protein.) The goal of the maternal tutoring program was to stimulate learning and development of the target child through direct intervention by trained person- nel. This goal was a means of pursuing a second, more important goal: to modify caretaker-child interaction so that cognitive development and positive social interaction would become self-sustaining in the absence of intervention personnel. The infant stimulation program was based on the work of Lambie and collaborators [2]. During the first three years of life, the child's family w-as visited twice weekly by a specially trained paraprofessional (a secondary school graduate with experience in teaching) who interacted with the child and the principal caretaker. Selection bias was avoided by random assignment to experimental and controlI groups. The health-care program permitted surveillance of morbidity and reduced the drop-out rates. Standardized information was obtained on mater- nal anthropometry, prenatal diet and health status of the mother, duration of breastfeeding, nature and timing of additions to the infant diet, family socioeconomic characteristics, quality of the physical and social environment, and maternal attitudes. Family and individual dietary surveys were conducted at the time of enrollment and repeated two months later after supplementation was initiated at the end of the second trimester of pregnancy. Dietary surveys were carried out again when subjects were 18 and 36 months of age. Health and  4 physical developrnent were monitored by longitudinal anthropometry and pros- pective assessment of morbidity. Finally, a battery of age-appropriate psycho- logical tests arnd carefully controlled behavioral observations was administered at several points in time. A number of analyses of intervention effects during the first three years have been completed and published. They indicate that interventions influenced nutritional intake [3]. This increase was reflected in improved physical growth at 3 years of age [4]. Behaviorally, supplementation and maternal education had a variety of effects on overall cognitive development as assessed by the Griffiths scales at age 3 [5]. The maternal tutoring program was found to have specific, anticipated effects on maternal behavior [6]. When the subjects reached age 3, interventions were stopped, although monitor- ing and health services continued. The purpose of the follow-up fou- years alter termination of interventions was to examine the long-term outcome for both physical growth and school competence. The Sample at Seven Years Of the initial 443 cases/families recruited during the second trimester of preg- i.ancy, 7 had dropped out by the time of birth. At termination of the interven- tions, when each subject reached three years, the project had lost contact with 82 other families who moved and could not be traced. An additional 14 families declined to participate and 1 child died, leaving a sample of 339 at aae 3 years. For the present school phase follow-up, project staff in Bogota obtained anthro- pometric measures on 247 children, a figure which includes 57 children not included at age 3 (49 moved and 8 refusals). Preschool test scores were located for 264 children, and at least some information about school enrollment was  obtained for 269. The sample available for the school-phase analysis varies among specific outcome measures, but is at maximum about 70 percent of the size at age 3. The present analysis of school phase outcomes concentrates on the full Supple- mentation and Maternal Tutoring interventions (groups A, Al, D, Dl). Early Sup- plementation and Late Supplementation groups (groups B and C) are included only when they are of particular relevance. There is no statistically significant association between experimental group and availability of data for followup. Dropout is associated with some background variables, especially a summary index of education and literacy use (called 'EDPA'): the greater this index the less likely we are to have followup data (p<. 001). There is a tendancy for the bias to act- ifferently under the two con- ditions of Maternal Tutoring about child development, as indicated by the com- parisons of loss rates between the 3-year evaluation and the present collection of preschool test scores. Table 1 presents the data for groups A, D, A', and D', that is the Supplementation and Maternal Tutoring groups. TABLE 1 Sample Loss from 3-yr Anthropometry to 7-yr Preschool education/literacy percent no data bias to lose level of father n overall Mat-Tut no-Mat-Tut the Mat-Tut group low - 0 19 28% 20% 36% -16% 1 58 55% 45% 53% -7% 2 56 55% 50% 58% - 8% 3 78 63% 55% 64% - 9% 4 52 74% 83% 61% +22% high - 5 16 84% 60% 100% -40% Jn general, more families without Maternal Tutoring have been lost, but this bias  decreases with increasing levels of EDPA (the trend reverses at the highest level, but the number of subjects here is small). One interpretation of these results is that fathers with higher education have both higher goals and better means of achieving them, hence they are more likely to move up economically and out of the slums. In addition, according to this interpretation, Maternal Tutorina altered the family system in a way that differentially motivated or aided families with highly educated fathers to move out, and those with less educated fathers to remain available to the project. The mechanisms involved are not clear, but might involve the mother's support and encouragment for economic and social improvement of the family. One consequence of this bias in dropout rate is that for most of the school phase material EDPA is no longer independent of the Maternal Tutoring experimental condition. (There is a correlation of .16 [p=.O09] but the relationship is curvi- linear, hence the value and direction of the r is arbitrary.) It is therefore difficult to examine outcome data with respect to Maternal Tutorina wrhile hold- ing EDPA constant. Similarly, the Maternal Tutoring intervention has in -effect lost some power of causal interpretation, since it is confounded with EDPA (but see controls below). In summary, about 30 percent of the sample that was available at three years has been lost. Because of the interaction of father's education/literacy (EDPA) with Maternal Tutoring in affecting the loss rate, extra precaution must be taken in interpreting the effects of these variables. Preschool Achievement Test Scores Preschool achievement test scores (precurrentes) were available on 174 chil- dren (groups A, Al, D, and Dl). Age at testing ranged from-5;.6-years to 8.7 yea~rs  7 (Q1=6.2, Q3=7.2, median=8.8). The variation had a significant efiect on children's scores and it is thus necessary to adjust it statistically in the analysis. The Preschool Test consists of 52 items organized around readina readiness, mathematics, and basic knowledge. Items are similar to those found on many preschool examinations, such as tracing a square, identification of left/right body parts, memory for digits, auditory discrimination, and location of missing features in pictures. Three summaries of the data were analysed by several different means. (1) The raw data were examined for major features through (a) principal components (factor) analysis and analysis of variance of the resulting factors; and (b) mul- tivariate analysis of variance. (la) Principal components analysis found a first factor that accounts for 24% of the variance, suggesting the test does not contain a large central core for this population. Verimax rotation did not yield any useful interpretation, with- the first three factors accounting respectively for 14%, 5%, and 5% of the vari- ance. ANOVA of the first (unrotated) factor revealed a Maternal Tutoring x Sex interaction: maternal tutoring slightly increased scores for -male offspring, and slightly decreased scores for the females (p=.03). Covarying out the effect or' age at testing (the range of ages was rather substantial) did not appreciably alter the results. (lb) Multivariate analysis of variance on individual items suggested a marginally signiflcant effect of Supplementation (p=.068), increasing the overall score. There was also the same interaction of Sex with the Maternal Tutoring intervention as found above (la). but this time only of marginal significance (p=.06). Again, using age as a covariate did not alter the results.  (2) A second kind of summary used items scores recoded to achieve more reasonable distributions. Factor analysis on these data was cleaner, yielding a principal component that accounted for 52% of the variance; nevertheless, the results were about the same as above: - There was a signiflcant interaction (p=.03) of Maternal Tutoring and Sex; - Covaring age at testing, did not change the results (noy p=. 05 for the Sex x Maternal Tutoring interaction). - Covarying background variables removed the two-way i-nteraction and did not introduce nlew ones. Background variables used were possessions at 36 months, income at the beginning of the study, EDPA, mother's disinterest/anomie, maternal report of teaching the child at 36 months, matri- monial satisfaction, maternal age, and mothers's education/literacy. The goal of this analysis was to reduce the noise that such variables might have imposed on the Preschool Test results, but the effect, given the correlation of EDPA with Maternal Tutoring intervention, was to reduce the apparent effect. -Covarying both age and background yielded only a very marginal effect of Supplementation (p=.11). (3) The third way of summarizing test results used a priori summary scales: Reading Readiness, Math Summary, Basic Knowledge Summary, and Total Score. Again a similar picture was obtained: -straightforward ANOVA yielded a marginally significant Maternal Tutor- ing by Sex interaction (as before) for Reading Readiness (p=. 06), Math Summary (p=. 07), and Total Score (p=. 05), but not the Basic Knowledge Summary; - Covarying age eliminated all these effects but introdueed a marginal effect of Supplementation, for Reading Readiness only, ,f~C9)r  -- Covaryi.ng background variables removed all effects and interactions; -- Covarying both age and the set of background variables yielded only a maxrginal effect of Supplementation on Reading Readiness (p=.08) and Total Score (p=.10), in both cases the supplemented childxen doing better (e.g. adjusted scores of 28 vs. 23 C diness). To place this effect in some perspective, the 5 point differt ' jading Readiness scores just cited corresponds to an age difference of 33 days. Because of the association between Maternal Tutoring and father's education/literacy, the above analyses were repeated within each level of EDP.A For all levels except 2 and 3, the number of subjects was very small; for all levels except 2, either no reLiable effects were found, or the inferential statistics could not even be computed. Within level 2 (n=243J important results were found for the simple ANOVA model, and more strongly when age at testing was covaried out. In the latter analysis, Supplementation had a reliable effect on Reading Readiness (p=. 04), Math Summary (p=.03), and Total Score (p=.OQ6). The mag- nitude of Supplementation effect on the Total Score was 25-point (52 vs. 7-7); a functionally meaningful difference. There was also a Supplementation by Sex interaction (p=. 0C?) for Math Summary, such t2aat this intervention helped boys but not girls. In stummary, there was a small overall, marginally reliable positive effect of nutritional supplementation on some preschool achievement scores -when appropriate statistical controls are applied. This effect was large and reliable for children with fathers of low to moderate levels of education and literacy. Maternal tutoring appeared marginally to increase Preschool Test scores for sons (but not daughters); however this effect was reduce& or eliminated when age of the child at testing was statistically controlled, ort when the artifactual  10 association between Maternal Tutoring and EDPA is controlled. School Enrollment There is some school attendance information on 269 children. Intervention had no effects on whether or not they were ever enrolled in school. Children who were not enrolled were all in the later cohorts of data collection and therefore not old enough to be in school at the time of the present data collection. There ,as an important effect of Maternal Tutoring intervention on the timing of school entry. Across the whole sample, age at first entry into the educational system was 5.8 years (median=6.0; quartiles 1 and 3 = 5.3 and 6.5 respectively; range = 1.3 to 7.7 years; s.d. = 0.98; n = 150 for this variable). Omittina a few extreme scores (which represent most likely entry into nursery school, rather than the regular system), ANOVA indicates that Maternal Tutoring led to earlier school entry (p= .0008), but primarily in association with absence of Supplemen- tation (interaction of Tutoring and Supplementation, p = .06). The relevant means are: TABLE 2 Mean Age at School Entry Supplementation Maternal Tutoring no yes no 6.0 5.0 yes 5.9 5.6 Given the probLem that Tutoring is no longer a statistically independent inter- vention, thiis analysis was repeated at each level of EDPA. At levels 0 and 1,  there was no test and no effect, respectively. For levels 2 and 3 (i = 18 and 30), the effects given above were replicated; that is, tutoring led to earlier school entry, especially for unsupplemented children. (There was also a significant effect of Sex, p = .01, boys go earlier than girls by 1 year; see below.) There was no effect at level 4 (n = 20), and no test possible at level 5. There was performance data on 156 children in school. Of the 95 in the major four experimental groups, 92 were in first grade at the time of data collection, 3 were in prekindergarten. Table 3 shows that all 3 of the prekindergarten kids were in one of the unsupplemented groups (p of Fisher's exact test < .0001). There was no effect of Maternal Tutoring. TABLE 3 Supplementation effects on Current Grade Su,pplementation Grade prekinder first no 3 44 yes 0 48 (p<.ooo1) Some of the children in first grade were repeating. An analysis of the interven- tions on repeating (in this cross-sectional sample, a noteworthy problem) is given in Table 4. Repeaters were more likely to be unsupplemented than supple- mented, untutored than tutored, boys rather than girls.  12 TABLE 4 Intervention effects on Repeating Fi.rst Grade (Current) Group Repeating no yes p No Supplementation 38 (86%) 6 (14%) Supplementation 46 (96%) 2 ( 4%) .04 No Maternal Tutoring 53 (88%) 7 (12%) Maternal Tutoring 31 (97%) 1 ( 3%) .06 Male 53 (84%) 10 (16%) Female 71 (95) 4 ( 5%) .02 As might be expected from the differences in age at school entry, the average age in first grade differed among experimental groups. Children front the sup- plemented and tutored groups tended to be younger (p = .05 and .001); as did girls, but primarily girls in the Tutored group (p of Sex = .60, p of Sex * Tutored interaction = .0035). Intervention means are shown in Table 5: - TABLE 5 Mean Age in First Grade Supplementation Maternal Tutoring no yes rDo 8.1 7.1 yes 7.5 7.3 School Performance Three performance measures were available: grades awarded in selected sub- jects; passage onto the next grade level in the following year; and award of a  13 prize. Children were rated in first grade for Manual Work, Physical Education, Religious Education, Mathematics, Science, and Reading. Initial analysis indicated that the interventions did not influence grades, except for Physical Education, where there was a complex 3-way interaction (p= 03) and a Maternal Tutoring by Sex interaction (p=. 06). Maternal Tutoring appeared to increase grades for males and decrease them for females, but this effect includes the possible confound of EDPA. A second analysis was restricted to only those first graders who were not repeat- ing. Stil there were no significant effects of Supplementation, Maternal Tutor- ing, or Sex. Removing the effect of age through covariance analysis failed to yield any significant main effects of the three independent variables. A standard set of background variables was also controlled statistically (mater- nal anomie at 30 months, family possessions at 36 months, maternal self-report of teaching at 36, maternal pride during interview at 36, family income at begin- ning of study, and EDPA), in addition to child's age. Under these circumstances, a marginal interaction effect (p = .07, .08, .08) of Supplementation and Sex was found on Total grades during the first year, in the first semester, and for a rotated factor that included Manual Work, Mathematics, and Science, all in the first semester. Supplementation was found to help males on these grades but not females. Family wealth and maternal pride were the backaround factors with major effect. These analyses were repeated by level of EDPA. Controlling for age as well, a number of significant effects were found for the middle level (EDPA = 3): Sup- plementation significantly (p = .007) influenced total grades for first grade,  14 especially in the first semester (mean effect 1.2 standard deviations). Maternal tutoring also increased total grades (p = .02) by .66 standard deviations. There was a marginal interaction (p = .06), such that any one or the combination of interventions was effective (i.e. the control group was low). There were no reli- able effects or interaction at the hiaher levels of EDPA, and subsamples were too small at lower levels for adequate analysis. It should be noted that the dispersion of children among a large number of schools renders analysis of assigned grades difficult, for there are likely to be teacher and school differences in average grades, a variation masking whatever intervention effects might exist. Methods for overcoming this problem (such as matched comparisons with classmates) were not available with the present resources. The meaning of passing first grade may be different for repeaters and non- repeaters. Hence our analysis was restricted to the non-repeaters. In Table 6, only the Sex effect on passing is statistically reliable (girls do better, p = .01). TABLE 6 Intervention Effects on Passing First Grade Group Pass First Grade Fisher's p no yes (one-tailed) No Supplementation 8 (22%) 28 (76%) Supplementation 8 (17%) 38 (84%) ns No Maternal Tutoring 12 (23%) 41 (77%) Maternal Tutoring 4 (14%) 25(86%) ns Male 16 (31%) 36 (69%) Female 9 (13%) 61 (87%) .01  15 It is not possible in a dichotomous analysis to adjust individual scores for age, therefore the above analyses were repeated restricting age to a span of one year. The only important outcome of this restriction was to make the effect of Supplementation on Passing marginally reliable (p = .06 by Fisher's test, one- tailed): all 16 supplemented children passed, while only 24 of the 30 unsupple- mented were promoted. About 17% of the non-repeating sample was given a prize upon passina first grade. Although Supplementation increased the probability of receiving a prize, the results were not statistically reliable (Table 7, rows one and two). Similarly, Maternal Tutoring and Sex had no effects. Reanalysis within levels of EDPA, how- ever, indicated that Supplementation reliably increased the receipt of prizes at level 3 (Fisher's p=.03). (At level 4 the effect was mnarginally reliable: p=.10.) Also at level 3, Maternal Tutoring had a marginal effect or- receiving prizes (P=.1 a). Imposing the 1-year age span yielded no significant results. TABLE 7 Intervention Effects on Receiving a Prize Group Receive Prize Fisher's p no yes (1-tailed) No Supplementation 39 (91%) 4 ( 9%) Supplementation 36 (78%) 10 (22%) ns No Maternal Tutoring 43 (81%) 10 (19%) Maternal Tutoring 24 (86%) 4 (14%) ns Male 45 (87%) 7 (13%) Female 53 (79%) 14 (21%) ns  16 In s-ummary, the Maternal Tutoring intervention was seen to lead to earlier entry into school, especially in the absence of Supplementation. Maternal Tutor- ing also reduced the rate of repeating first grade (as near as can be detected in this cross-sectional analysis), despite the earlier enrollment of these children. Tutoring of the mothers also led to higher school arades for the middle (and presumably lower) levels of EDPA. Supplementation also had effects on school performance. As for Tutoring, Sup- plemented children were less likely to be found repeating first grade, even though they were younger at the time. There was marginal evidence that Sup- plementation led to higher grades (at least for males); this finding is highly reli- able in the middle (and presumably lower) levels of EDPA. Supplemented chil- dren are marginally more likely to pass directly from first grade onto second (without repeating). Within the middle levels of EDPA, supplementation increased the probability of winning a prize at the end of the school ye-ar. Anthropometry A straighforward analysis of variance indicates significant main effects of Sup- plementation and Maternal Tutoring on the major indices-of-growth: height for age (p = .001 and .02 respectively) and weight for age (p =.06 and .04). There were no interactions or Sex effects. The intervention effects were meaningful: Supplementation and Maternal Tutoring each added approximately 2 cm in height and .5 kg in weight. The raw control/experimental means are: Supplementation: 1.06 m vs. 1.08 m and 17.43 kg vs. 17.89 kg Maternal Tutoring: 1.06 m vs. 1.08 m and 17.37 vs. 17.94 kg The results are additive (not interactive), so that children who had received both  17 treatments (group D1) had means of 1.09 m and 18.24 kg, compared to 1.05 m and 17.21 kg in for those in the control group (A). There were no significant main effects on skinfold thickness, or on thorax, arm, or head circumference, except for Sex in the expected direction. Supplement-a- tion had a marginal effect on head circumference, p=. 07, means of 49.0 vs 49.3. There was also a marginal interaction of Maternal Tutoring with Sex on arm cir- cumference (p=.07), such that the intervention was associated with smaller measures for males and larger ones for females. Early Supplementation slightly increased thorax circumference (by about .8 cm) for males but not females (P= 02). The continuing effect of Supplementation four years after termination of the intervention is impressive though predictable because it is consistent with the results seen at age three (85.6 cm vs 88.4 cm). The positive influence of Mater- nal Tutoring on growth was a new finding, for there was no significant effect at age three (86.5 cm vs. 87.5 cm, p=.13). Losses in the sample m-ay have contri- buted to the magnitude of the Maternal Tutoring effect at seven years, but they are not sufficient to artificially produce the effect. The data at 3 years were reanalyzed using only those subjects for whom there was data at 7 years; the original 3-year results were replicated: main effects of Supplementation were seen on height and weight for age, a significant Sex difference on weight, and no effect of Maternal Tutoring. Thus the new-found effects of Maternal Tutoring at 7 years were not an artifact of having only a select subsample for analysis. To examine for an artifact arising from the differential effect of Father's Education/literacy on loss rate, the analysis of variance was repeated at each level of EDPA. The small number. of  subjects made some of the sub-analyses impossible, but in general the results confirmed the initial conclusion. The Maternal Tutoring intervention increased height among children whose fathers had the lowest level of education (p=.03, 1.04 m vs. 1.08 m). The same was true for the second level (p=.02; 1.06 m vs. 1.10 m). It was less reliable at the third level (p=.08, 1.07 m vs. 1.09 m). An effect of the same magnitude (but opposite in direction) was not reliable at the fourth level (1.06 vs 1.04 m, p-.16). At the two highest levels no estimates could be made because of the small number of subjects. A sirnilar pattern held for effects on weight, but with lower levels of reliability. It would appear that the eEfect of Maternal Tutoring is a real one, but it is substantial only at lower pater- nal educational levels. A second approach to removing possible artifacts is to covary out the effect of fathers' education/literacy before performing the analysis of variance. Under this procedure, the effect of Maternal Tutoring remained on height, but was no longer reliable for weight. In addition, effects of Supplementation were then visible. The full results of this last analysis (controlling for both age and EDPA) are as follows: Supplementation, p=.006; mean height Supplemented = 1.084 m vs. 1.062 control. No signiflcant effect on weight. Maternal Tutoring, p=,03; mean height Maternal Tutoring = 1.082 m vs. 1.064 m control. No significant effect on weight. - There were no further effects or interactions for height or weight (con- trolled for sex). It could be argued that the covariance adjustment was inappropriate since the drop-out bias was monotonic up to the highest level of EDPA, at which point it  19 reversed. However, the relationship of EDPA to height was monotonic and that is the covariance for which the results were adjusted. Nevertheless, the above analysis was rerun omitting the highest level of EDPA, with no change in the results. In summary, children in the nutritional Supplementation groups retained at age 7 the 2cm and 5OOg advantage over the control children which they had at age 3 years when interventions ended. In addition, Maternal Tutorina, which had not previously affected physical growth, was now associated with a similar, but independent, gain in height and weight. Diet At age 7, children in the Maternal Tutoring groups had higher levels of protein and calorie intake than other groups. For protein, the averaae level of adequacy in the treated group was 110 percent of standard, compared to 88 percent in the controls (p=.003). For calories, the effect was only marginally reliable (P=. 09), but still noticable in real terms: 70 percent of standard vs. 63 percent. No differences were found for the average per capita level of adequacy in the f-ami- lies as a whole. Maternal Tutoring effects at three years (in the same subsam- ple) were in the same direction as these results, but they were not reliable. Mean energy intake (Maternal Tutoring vs control) was 92 percent of standard vs 90 percent; for protein adequacy, the means were 151 percent vs 130 percent. (p = .72 and .19 respectively). Predictors of school phase variables It is instructive to examine the relative power of the interventions (discussed above) and the naturally ocurring variation in socioeconomic factors. Stepwise  multiple regressions of the Preschool Test acheivemnent scores were performed, using as predictors the Supplementation, Maternal Tutoring, Sex, cumulative gastro-intestinal morbidity from birth to 54 months, weight for age and height for age at both 36 months (termination of the intervention) and school entry (about 7 years), child's age at testing, maternal anomie, environmental sanita- tion, maternal self-report of teaching the child, maternal judgment of the child health, observer ratings of maternal pride in the child, maternal age, maternal and paternal education/literacy, marital satisfaction, family integration into neighborhood groups, family wealth and possessions, income at beginning of the study, direction of change in income during the study, and variability of income. The regression analysis was free to select the most effective combination of 1, 2, 3, etc. predictors. The following subsets of independent variables had the highest explanatory power (maximum R-squared) with each predictor variable making an individual contribution significant at the..10 level or better. The best predictors of Reading Readiness were family base.inc.ome, Supplemen- tation, age at testing, and EDPA. The combination of four predictor variables accounted for 35 percent of the variation in Reading Readiness. For the Math score on the Preschool Test, the most effective predictors -were family'incorre, age at testing, and height for age at three years. The overall R-squared was .38. The Basic Knowledge Score was reliably predicted, with family income beina the major correlate (R-squared = .21, p<.c: 005) The overall Preschool Test score was predicted by the same pattern as the Read- ing Readiness scale: family income, age at testing, EDPA, and Nutritional Supple- mentation (R-squared = .36). Although the interventions did not appear to affect school grades in the overall  21 analyses presented above, multiple regression proved more effective, and the results differed significantly from those obtained for preschool achievement scores. Using a summary of all grades in all 6 subject areas in both semesters, it was found that 66 percent of variation in grades could be attributed to the fol- lowing variables: maternal teaching of the child, maternal age, sex (boys better), maternal education/literacy, and nutritional supplementation. It is evident that school grades are influenced more by maternal factors (especially her 'school- related' behavior) than are achievement scores, which respond more generally to family (i.e. paternal) socioeconomic factors. Nutritional Supplementation is about equally effective in raising both kinds of outcome. Passing the current school grade and winning a prize in class present a related contrast. Passing is responsive to specific maternal factors, namely teachina (self-report) and Maternal Tutoring (R-squared = .44). Obtainina a prize, in con- trast, is a more direct outcome of child variables: height for age, weight for age (indicating smaLI children do better), maternal perception of health, and Nutri- tional Supplementation). The overall R-squared is .39. Repeatina the previous year shows a combination of the two patterns. SUMMARY Four years after termination of the project's interventions significant and mean- ingful effects were found in children's growth, diet, and school performance. Nutritional supplementation was still associated with greater height and weight at age 7 than among controls. Supplemented children showed a marginal advan- tage in preschool achievement scores (large at some levels of father's education/literacy), and in performance in school (as indicated by grades, prizes, and class progress).  22 The Maternal Tutoring program had a marginal influence, at best, on the Preschool Test scores and on passing school, although the children in these con- ditions tended to be younger than control students in the same grade. The Maternal Tutoring effects on passing school were significant for some levels of father's education and literacy. It is especially noteworthy that M1aternal Tutor- ing was associated with increased nutritional intake at age 7 and with an advan- tage in height and weight equal to the effect of Supplementation. These effects were not evident - but perhaps presaged - in the evaluation at three years. One speculative interpretation of this pattern of results is that, at least at some levels of father's education/literacy, Nutritional Supplementation led to a higher level of activity, alertness, and social cooperation, all of which helped on the preschool knowledge test, in classroom performance, in benefitting from class- room instruction, in winning prizes, and in passing on to second grade. Maternal Tutoring, irn contrast, did not have so general an effect, but resulted instead in some specific school-related skills (at least at certain levels of father's - education/literacy). The children's familiarity with the- school-like learning paradigm of interaction with adults, and possibly a larger knowledge base, helped them obtain good grades. The skills did not help iirthe preschool exami- nation scores, in winning prizes, or in passing onto second grade. Tne dietary and growth effects suggest, however, that there may have been a pervasive change in family attitudes toward the child. These summary generalizations understate the magnitude of intervention effects on the children some at risk. Although most of the results cited come close to being significant for the population as a whole, all of them are substan- tially greater among children from poorer and less educated families.  23 It is worth distinguishing three different types of effects. First, some of the effects achieved by the completion of the intervention phase were retained, par- ticularly the increased height and weight of the supplemented children. Second, some effects, most notably the increased growth among the Maternal Tutoring group, became significant after the interventions ceased. These effects may reflect improvements in the child's environment secondary to changes in the mother's attitude towards the child and his or her needs for development. Finally, some of the effects summarized here are likely to be of future conse- quence in their own right. Earlier school enrollment, for example, or receiving higher grades, may have long-term effects on school career regardless of the ini- tial causes of that difference. The data presented here suggest that both Nutritionial Supplementation and Maternal Tutoring altered the child and the child's environment sufficiently to produce results 4 years later. The results are large enough in some cases to be of real significance themselves and to engage recursive change in the child- environment system. The specific patterns that can be seen contribute to-our understanding of the complex human costs of poverty, ignorance, and malnutri- tiorn References [1] Herrera, M.G., Mora, J.O., Christiansen, N., Ortiz, N., Clement, J., Vuori, L., Paredes, B., Wagner, M., and Waber, D. Effects of nutritional supplementation and early education on physical and cognitive development. In: Life-Span Developmental Psychology, Intervention, (R.R. Turner & F. Reese, eds.). New York: Academic Press, Inc., 1980, pp. 149-184. [2] Lambie, D.Z., Bond, J.T., Weikart, D.P. Home teaching with mothers and  24 infcants. Ypsilanti, Michigan: High/Scope Foundation, 1974. [3] Sellers, S.G., Mora, J.O., Super, C.M., & Herrera, M.G. The effects of nutri- tional supplementation and home education on children's diets. Ms. submitted for publication, 1982. [4] Mora, J.O., Sellers, S.G., Suescun, J., and Herrera, M.G. The impact of sup- plementary feeding and home education on physical growth of disadvantaged children. Nutriotin Research, 1981, 1 213-225. [5] Waber, D.P., Vuori, L, Ortiz, N., Clement, J., Christiansen, N., Mora, J.O., Reed, R.B., and Herrera, M.G. Nutritional supplementation, maternal education and cognitive development of infants at risk of malnutrition. American Jour-al of CZinical Nutrition, 1981, 34, 807-813. [6] Super, C.M., Clement, J., Vuori, L., Christiansen, N., Mora, J.0., and Her- rera, M.G. Infant and caretaker behavior as mediators of nutritional and social intervention in the barrios of Bogota. In: Ciulture and Eart-y Interaction, (T. Field, P.H. Leidermen, and P. Vietze, eds.). Hillsdale, N.J.: Earlbaum, 1981.