Contents - Previous - Next

This is the old United Nations University website. Visit the new site at


A number of reports of the effects of nutrition supplementation on mental development scores in preschool children were published in the decade after completion of the INCAP longitudinal study [37-40]. In addition the children's socio-emotional functioning was measured when they were of school age.


Results from the first decade

The only report on the effects of the supplementation in infancy divided children into groups according to the adequacy of their caloric ingestion [37] since energy may have been limiting for growth [25]. Combining atole and fresco villagers, the mothers and infants were categorized into three groups according to the level of supplemental calories ingested (high, medium, or low) up to the age of testing [37]. Statistically significant differences (p < .05) in the expected direction existed between the mental development scores of the three subgroups at the three ages tested (6, 15, and 24 months). Between-group differences for the motor scale were statistically significant at 15 and 24 months, and their magnitude was similar to that of those reported for the mental scale. In all cases the highest and lowest scores corresponded to the groups with the highest and lowest intake of calories respectively.

A number of reports have been published on the preschool data. The original atole-fresco design was used to compare subjects' performance on the battery [38]. The analyses were restricted to 550 children born between March 1969 and February 1973. Since the supplementation was first implemented in January 1969, the sample included both subjects who had and who had not received supplements since conception. Using estimates of both supplement and home diets, children in the atole villages had a higher intake of energy and protein than those in the fresco villages, except for boys in the small fresco village, for whom no differences were found. Analyses were performed within village size and sex groups. Generally, children in the atole villages performed better than those in the fresco villages. Within the large village pair, 6 of 27 comparisons (22%) favoured the atole children. and in the small village pair, 29% favoured the atole children significantly. All but one of the effects observed in the small villages were restricted to girls. Differences were most common for the naming and draw a line slowly tests. Given the number of statistical tests, the overlap of the subjects, and the covariation among tests, it was difficult to determine appropriate significance levels [38].

These analyses had no statistical controls. Subsequent analyses to test the role of potential confounding factors showed that attendance, socio-economic status, and morbidity did not differ significantly by village type.

All the villages were combined and calorie intake was used as a continuous variable in an analysis that included all the subjects ever tested (1,083 children) [39]. Results from three tests (language, memory, perception) and a composite score based on 12 tests were reported. Subjects from the four villages were pooled and stratified by age and sex. Separate correlations were calculated between scores on the psychological tests and the amount of supplement consumed during pregnancy and lactation by the mother and by the child. Most of the correlations with the language, memory, perception, and composite scores for boys and girls 3-7 years old showed a positive association with the supplement. Almost half of the multiple regression models that took supplementation into account were statistically significant. The effects were particularly salient in correlations that involved language. The effects of supplementation on test performance varied inversely with age at the beginning of supplementation. Children exposed to supplementation both prenatally and postnatally were more likely to show effects than those exposed at older age. Another study used a similar analytic strategy and reported essentially the same findings [40]. It also reported a number of tests for confounding variables, such as the effects of repeated testing, the level of cooperation of the residents of the villages, and morbidity. None of these factors were different in atole and fresco villages.

A within-group analytic design divided the sample by age (3-6 years), sex, and treatment [41]. In addition, instead of using the scores on particular tests of the preschool battery, it reported a composite score for "each child at each age by summing all of the cognitive tests in standardized form" (p. 202). The total amount of supplement ingested during the child's lifetime was used as predictor. In this analysis the effect of the supplement on cognitive function within the atole and fresco samples was restricted to boys. The authors specifically reported that the "most striking finding was a lack of association between test performance and supplement ingestion for girls" (p. 206). Both supplements were generally associated in the expected direction with the cognitive composite score. The authors noted: "If the supplement caused increased mental development, a boy who has ingested an extra 200 calories per day since the program began, or about ' cup of Atole a day, should increase his composite score by between one-third and one-half of a standard deviation" (p. 206). Controlling for social and economic variables and other demographic characteristics did not modify substantially the pattern of findings reported for boys and girls.

Design problems

These reports cannot be compared because the data were analysed using differing designs. The supplement variable was handled in four ways: as a village type (atole versus fresco) [38], as a continuous measure (total calories ingested combining atole and fresco children) [39], as a categorized variable reflecting adequacy of ingestion of calories at various times, combining villages [37, 40], and as a continuous measure within village types [41]. The decision to combine fresco and atole villages and examine the effects of calories across the entire sample was a result of analyses of physical growth which suggested that calories, not protein, were associated with improved growth (J-P. Habicht, personal communication, 1990).

The problem with combining the four villages and using calories as the independent variable, particularly for behoviaural data, is that value of the fresco group to control for the social stimulation effects of attendance is lost. Any associations of calorie supplementation and test performance observed could then be a function of the social stimulation received at the centre. Only one of the published studies [38] and a new analysis [1] adhere to the atole-fresco design.

A second problem that limits comparability between the studies is that different tests of cognitive development were used. One selected a few tests [41]; another presented data from 11 separate tests [38]; and a third used a summed, standardized composite measure [41].

Control for confounding variables

Because the design did not permit random assignment of individuals to treatments, the possibility that differences in village characteristics, or in the responses of residents to the two supplements, would differ by treatment is substantial. Each investigator has dealt with confounding variables in different ways. The major variables examined were socioeconomic status (all), morbidity [38, 40], cohort and secular change (all except [39]), repeated testing [40], and attendance and self-selection [38, 40]. None of the first four were associated with treatment; however, the confounding effects of attendance and self-selection were particularly difficult to eliminate.

The fresco-atole design can control for attendance effects only if the attendance levels are similar in the two village types or if the kinds of children who attend are the same in the two village types. Although attendance levels differed for children 2-3 years old, with atole children attending significantly more days than fresco children, the rates of attendance are equivalent to those for 4-5-years-olds. Therefore, the authors of one study [38] argued, since the tests they used to test the hypothesis were administered to 3-5-year-olds, attendance rates should have been similar for the two groups. However, since fresco children had much lower rates of attendance before age 5, the argument is valid only if one assumes that attendance during a prior period in one's life is irrelevant to test performance several years later, contrary to the main hypothesis of the study.

In another study [40] it was maintained that the effects of attendance on test performance were so small that they could be dismissed. An increase of 0.007 standard deviation unit on the congnitive composite is found for 100 days of attendance. Mean differences between atole and fresco children are approximately 0.2-0.3 standard deviation, and at the above rate about eight years would be required to increase a test score by 0.2 standard deviation.

Self-selection also was considered as a a confounding variable [38, 40]. No differences were observed in parental cooperation between the fresco and atole groups [40], and self-selection was not a problem because villages were randomly assigned and house quality did not differ between well and poorly supplemented atole children [38]. Neither of these arguments explores the possibility that the kinds of children who attended the centre were different for the two village types. One could suspect that the motivation for ingesting atole, a warm, sweet, brown, thick drink considered to be nutritious, could differ from motivation for ingesting the Kool-Aid-like fresco.

Social and emotional development

Between March 1979 and August 1979 an investigation of the effects of nutritional supplementation on school-age children's social and emotional functioning was performed using a subsample of 139 children who had received supplementation from birth, ranging in age from 6 years 1 month to 8 years 3 months [42, 43]. The rationale for this investigation was that the major effects of nutritional deprivation would be seen in attention, exploratory behaviours, and emotionality, none of which had been tapped in standardized testing.

A series of structured situations was developed to observe the children's responses to social activities. These included (1) a response to novel materials, (2) behaviour in a relay race in which each runner had to pop a balloon, (3) persistence in attempting to open a clear box with a prize inside, and (4) a game of "Simon says," in which the players have to inhibit themselves to respond only when the proper stimulus appears. The children were assigned to same-sex, same-age, and same-village groups of six. Careful observations of their behaviour in each setting were recorded.

The two reports analysed the same data in somewhat different ways, with essentially the same conclusion: on a number of measures, the children who had ingested more supplement were more active and involved than those who had ingested less. Specifically, they touched more materials in a novel situation, explored more in the novel situation, were more involved in the balloon-popping relay race, tried more strategies to open the box, and were more likely to anticipate a response in "Simon says." The last effect was specific to boys.

In contrast to the significant correlations of supplement ingestion with behavioural variables, there were few relationships of supplementation with cognitive test performance measured at the same time as the behavioural assessment [42]. On the other hand, anthropometric measurements were related to test scores but much less to social and emotional variables.

This investigation presents intriguing findings, consistent with the theories about the effects of malnutrition on behaviour. Unfortunately, it uses the same across-village analysis as three of the reports summarized above. Therefore, it cannot eliminate the hypothesis that attendance at the supplementation centre or self-selection could explain the relationships observed. Given the social nature of the variables measured, an attendance hypothesis is even more probable than with strictly cognitive measures.

The investigators report that relationships between behaviours and supplement were more often found in the calorie-only village than in the protein-calorie village. They speculate that "only at the lowest levels of energy intake will differences in calorie supplementation be sufficient to produce differences in cognitive development . . . malnutrition may have its strongest effects on mental development when environmental stimulation is also limited" [42, p. 554]. However, it is also possible that the effects observed are attendance effects, since from the age of 6 the children in the fresco villages were attending the supplementation centre as much as or more than those in the atole villages.


Results from the second decade

Analytic strategy

In one analysis of these data, additional controls were added to test the hypothesis more rigorously [1]. In brief, these controls were to establish baseline equivalence in both test scores and socio-economic variables, to compare the atole and fresco villages controlling for socio-economic differences, to examine the confounding effects of attendance, and to control for attendance effects in the final analysis. Infancy scores as well as preschool scores were examined. For the preschool battery, factor scores rather than many specific tests were used as the dependent variables to reduce the number of statistical comparisons. The sample comprised children who had been exposed to supplementation from conception to at least 24 months of age, the maximum-exposure cohort. Some of the analytic decisions are discussed in more detail in the following sections.

Baseline assessment

To determine whether initial differences existed between the groups, atole and fresco comparisons were made for children who probably would have been affected relatively little by supplement ingestion: those tested in 1969, 5-7 years old. No systematic differences in test scores were found favouring either village type [1].

Analyses of initial differences in socio-economic variables, which tend to be highly related to cognitive test performance, were also lacking from the initial studies. Atole-fresco comparisons of maternal and paternal education, house quality, and father's occupation prior to the beginning of the study suggest that systematic differences favoured the fresco villages, particularly in parental literacy, before the study began [26]. Therefore, socio-economic differences were controlled statistically in the test of the hypothesis.

Statistical controls

Attendance rates were calculated for each child for each year of age. Patterns of supplementation attendance varied by age and by village. In the atole villages, children attended significantly more in the first four years of life, whereas 7-year-olds attended significantly more in the fresco villages [1]. Differences in attendance patterns by village type by age suggest that motivations to attend varied by supplement. It is possible that mothers were more likely to perceive that atole had nutritional benefit and take younger children, whereas older children, who tended to come on their own, preferred the taste of fresco. A conservative test of the hypothesis is to control for differences in attendance patterns when examining the effect of treatment on test performance, the strategy adopted by Pollitt et al. [1].

Development of measures

The measure developed for the infancy scores was the sum of items passed at each testing period. For the preschool battery, a factor analysis of all tests with adequate test-retest reliability was calculated [1]. Two factors resulted: a general factor consisting of verbal and perceptual-motor tests, and a short-term memory factor loading primarily on memory for sentences and memory for digits.

Results with the infant battery

The model compared the maximum-exposure-cohort atole and fresco children on the dependent measures controlling for attendance and socio-economic covariates [1]. Boys and girls were combined, with sex entered as an independent variable along with treatment. The dependent measures were mental and motor measurements from the infant battery at 6, 15, and 24 months. After controlling for the socioeconomic variables and attendance up to the date of testing, atole boys had significantly higher scores on the mental scale at 15 months, and girls had significantly higher motor scores at 24 months [1]. With the sexes combined, atole children scored higher than fresco children at 24 months on the motor score. There were no interactions with SES.

Results with the preschool battery

Dependent measures for the analysis in the preschool battery were factor 1 and the memory factor at 3-5 years. Too few children were in the maximum-exposure cohort at ages 6 and 7 for analysis. All analyses controlled for socio-economic status and days of attendance up to the date of testing. Atole children scored significantly higher than fresco children on factor 1 at 48 and 60 months. The effects of the supplement were more evident in the lower-SES half of the population than in the higher-SES group [1]. There were no significant differences for boys, whereas for girls two of the six comparisons (33%) were at least marginally significant.


What initially appeared to be a simple quasi-experimental design includes many threats to internal validity [44]. Threats that would affect the interpretation of these findings are (1) initial nonequivalence, and (2) differences in the implementation and response to the treatments. The first problem would emerge if the villages were not initially equal. The second emerges because the treatment was not blinded. Differences in the two treatments may have resulted in different patterns of attendance to the centre. For example, mothers may have been more motivated to bring young children to receive atole, perceived to be of greater nutritional value, than fresco. The INCAP personnel, contrary to stated policies, may have differed in their encouragement of families to attend the supplement centre. Since mothers also ingested the supplement, it is possible that children's intake may have been associated with more active or responsive care-giving behaviours. Attendance was not equal in the two village types, so the effect of attendance on test performance might also have confounded the observed relationships and therefore was controlled for in the analyses.

There was no evidence that the two village types differed systematically in initial test scores [1]. Initial differences were seen in socio-economic status, both in the occupation of the father and in the education of the parents. The education variables, which are frequently associated with cognitive development, favoured the fresco villages, whereas the occupation codes were higher in the atole villages, principally because a higher proportion of men owned their own land [26]. The greater level of education in the fresco villages would tend to mask or hide any effects of supplementation, and the occupation differences might have inflated differences. Thus village differences are not as much of a problem as they would have been if the atole villagers had achieved higher levels of education. However, in these analyses, initial socio-economic status differences were controlled for.

The problem of differences in the implementation of the two treatments is more difficult to eliminate as a confounding factor. Attendance patterns did differ significantly between the atole and fresco villages; younger children were more likely to attend the atole centres and older children the fresco centres. Analyses of the effects of attendance suggest that they were small compared with differences between the two groups [ I, 40]. However, one study added attendance as a covariate in the analyses [1].

For the two studies that used the atole-fresco design, small but significant differences consistently favoured the atole villages [ 1, 48]. Because one analysis controlled for both socio-economic differences and attendance and used factor scores [1], it should be considered the definitive statement of the effects of the nutritional supplementation on cognition during the infant and preschool years. Differences were greater in the lower-SES population, which is consistent with the hypothesis that the supplement is most effective in conditions of deprivation. Differences were also greater for girls than boys.

Other studies also found greater effects of supplementary feeding on girls than boys. These gender-specific effects may operate through parental response to girls' energy or physical size [45], although the data do not allow us to distinguish mechanisms for the sex difference.

These results are consistent with two intervention studies implemented in the late 1960s to examine the cognitive effects of supplementation on infant and preschool cognitive development, in Bogotá, Colombia [46], and Taiwan [47]. Supplementation was associated with small but significant differences in development scores during infancy, particularly in motor development. Later effects are less clear; effects of the Bogotá intervention were reported for preschool childrens' reading readiness [48], but no differences were seen in Stanford-Binet IQ scores of five-year-old Taiwanese as a function of early supplementation [49]. In sum, evidence shows that among low-socio-economic, nutritionally at-risk women, food supplementation alone during pregnancy and lactation has a small, statistically significant effect on the motor and mental development of infant and preschool children, particularly girls.

Two mechanisms were proposed to account for these effects: direct biological or neural effects, and indirect effects related to children's lowered levels of activity or responsiveness to the environment. The data summarized here do not allow one to distinguish clearly between these hypotheses; however, they are consistent with a developmental interpretation.

A developmental model would suggest indirect effects of alterations both in activity and motor ability in interactions between developmental systems within the child, and between the child and its environment [1]. This model rests on two sets of findings: ones showing that among nutritionally at-risk populations supplementary feeding is associated with earlier motor maturation and activity level, and ones showing that these characteristics of the child are associated with more complex and developmentally stimulating interactions.

Supplementary feeding has been linked to gains in motor development not only in the INCAP longitudinal study but in other studies [50-52]. Activity levels also have been found to be lower among mildly to moderately malnourished children [53, 54] and to increase after supplementary feeding [51, 55].

The link between motor maturation and cognitive development has been examined primarily in well-nourished populations. Associations of locomotor skills with perceptual and cognitive development have been reported [56]. Increased locomotion is also associated with social contacts, which broaden social experiences and enhance affective and linguistic interactions with adults [57]. The link between responsive language in adults and growth in linguistic ability in children has been documented frequently, both in the United States [58] and in developing countries [59].

Much less research has been conducted in developing countries on the relations of motor maturation and activity to cognitive functioning. In rural Kenya, children who were more spatially active scored higher on cognitive measures [60]. It is theoretically reasonable to assume that relationships that hold in developed countries will also be found in nutritionally at-risk populations, where low levels of motor maturation and activity are likely to be linked to undernutrition. The association between advanced motor activity and improved cognitive performance provides a plausible explanation of these data.

In a longitudinal follow-up of these children when they reached adolescence and young adulthood, effects of the early supplementary feeding were seen not only in functional performance but also on information-processing tests such as reaction time, which are closely related to brain function [1]. These results suggest direct or neural effects on cognitive development; the possibility of a neural mechanism is not abrogated by these findings.

Contents - Previous - Next