|Early Supplementary Feeding and Cognition (Society for Research in Child Development, 1993, 123 pages)|
In the first section of this chapter, we review the issues that were discussed in Chapters III and IV, with a particular focus on whether the differences in the cognitive functioning of the Atole and the Fresco groups are best explained by the difference in the nutritional history of the subjects. The second section is devoted to a discussion of the results of the analyses of the psychoeducational and information-processing data. We then present a theoretical interpretation of the findings and, finally, consider their programmatic implications.
The assessment of effects with which we are concerned was framed within a quasi-experimental design. As such, the study faces most, if not all, of the well-known risks involved in this type of experimental approach that we listed in our earlier discussion (see Chap. IV). In this case, the villages, but not the individuals, were randomized into one of the two nutritional treatments. One might argue that utilizing this randomization would allow for a more powerful analytic strategy than electing to use the individual as the unit of analysis: theoretically, the four villages could have functioned as the units for analysis. However, in our view, the benefits of such a strategy would be illusory; the analysis would be legitimate only if the design had been that of a true experiment and included randomization as well as blinding of both subject and personnel. The intervention study failed to meet both these criteria. From our own analyses, we now know that randomization did not render the Atole and the Fresco villages comparable on all possible confounding variables, and neither subjects nor field personnel were blinded with regard to treatment.
Our analyses of the follow-up data established long-term developmental effects of nutritional supplementation, particularly among those at the lower end of the social and economic distribution in the villages. To our knowledge, no previous study has tested the effects of supplementary feeding during the first years of life on intellectual functioning assessed 10-15 years later. Our results are the first of this kind to be reported, and most readers of the nutrition-behavior literature will find them unexpected (Pollitt, 1988). However, it is necessary to review alternative explanations before an inference of a nutrition effect can be drawn.
In Chapter IV, we discussed the foremost threats to the validity of the conclusion that we propose; the most important among them were the nonequivalence between villages (both during the intervention and in the 10-year interval between the longitudinal and the follow-up study) with regard to factors capable of affecting the outcomes of concern, differences in the delivery and consumption of Atole and Fresco related to the differing nutritional properties of the two drinks, and the different patterns of attendance at the feeding stations where each of the supplements was administered. Although we lack the necessary information to reject them completely, our analyses of the available data converge to indicate that none of these alternatives can completely account for the differences in the cognitive functioning of subjects from Atole and from Fresco villages: the internal validity of the nutritional explanation is not compromised by these challenges.
In addition to ruling out alternative explanations, we have also established that a nutrition experiment did in fact take place and that the nutritional differences between the Atole and the Fresco villages had demonstrable developmental implications. The differences in the composition and the actual consumption of the two supplements suggest that the individuals who received the Atole were better nourished, particularly since Atole consumption truly supplemented the diet rather than merely substituting for other foods (see, e.g., Chap. IV).
Although the presence of a nutritional treatment is unquestionable, we cannot test the original protein hypothesis, nor can we specify which nutrient(s) determined the observed differences in outcomes. We do know that the two supplements were equivalent in micronutrients per unit of volume; however, the actual consumption of the Atole and Fresco supplements differed. Thus, when actual micronutrient consumption data were analyzed in terms of recommended dietary allowances (National Research Council, 1989), there was no equivalence between the two groups in that regard. Because the subjects in the Atole group consistently consumed more micronutrients than the subjects in the Fresco villages in the postnatal period, we cannot rule out the possibility that factors such as iron contributed to the test differences between groups (Pollitt, in press).
Finally, we must consider whether our nutritional explanation conforms with the current understanding of the nature and determinants of human development. We argued in Chapter III that both theory and data support the possibility of effects of early supplementary feeding on cognitive development in adolescence. Early supplementary feeding of nutritionally at-risk infants and preschool children results in a developmental advantage and is likely to have beneficial long-term effects, just as the provision of educational opportunities to young children living in poverty has been shown to have beneficial long-term effects on their social behavior.
In addition, longitudinal studies of children exposed to stress factors suggest that, while single events are generally not a sufficient condition to affect development, the probability of deviancy increases when multiple stress factors coexist or interact in synergistic fashion. We have found that those children who were at the lowest levels of the SES distribution and received Fresco performed less well on the battery of psychoeducational tests than any of the other groups of children.
In contrast to the strong evidence on the effects of Atole observed among the subjects in the cohort of maximum exposure, the range of supplement effects on the subjects with late exposure (after 2 years of age) was narrower. There were main effects of Atole on the knowledge and numeracy tests and a significant interaction between treatment and grade attained on reading achievement, similar to those found in the cohort of maximum exposure.
This difference in the range of effects observed among the two groups suggests that the behavioral development of children is more sensitive to nutritional factors during the first years of life, particularly during the period of rapid growth in the brain and body. However, the effects on this late exposure cohort show that, in a nutritionally at-risk population, dietary improvements after the second year of life, following the peak period of growth and change, will still have long-term developmental benefits.
In sum, the evaluations of competing explanations suggest that: the dietary experimental intervention is the most likely determinant of the observed differences in test performance.
The findings that we have reported are most prominent in the cohort of maximum exposure. Thus, unless otherwise specified, the discussion that follows focuses on this group.
The coherence of the results summarized at the close of Chapter VI is perhaps their most striking feature. Among them, the most conspicuous is the significant interactive effects of treatment x SES on five of the seven tests (except reading and literacy) included in the psychoeducational test battery. Consistently, the subjects who benefited the most from the Atole were those at the lowest levels of the SES continuum. A second major finding is the interactive effect of treatment x grade: those who had reached the highest grades in school and had received the Atole fared better than all other groups on all three tests tapping reading skills. In the information-processing test battery, there were main effects on memory and paired associates tasks; none of the interactive terms were significant.
The breadth of effects was coupled with their modest size. The maximum R² accounted for by the Atole treatment was 5% (e.g., on the vocabulary test). On some tests, although maximum R² was statistically significant, it represented only about 1% of the variance (e.g., numeracy). However, because these modest effects were generalized over a wide range of mental abilities, they are likely to have resulted in significant differences between the behavioral repertoire of the subjects in the Atole and the Fresco villages. The coherence of these findings extends back, at least in part, to the earlier longitudinal study. At 4 and 5 years of age, the interactive term treatment x SES accounted for significant portions of the variance in the cognitive Factor 1 (derived from the battery of tests administered during the preschool period). As is the case in adolescence, the subjects who benefited the most from the early supplementary feeding were those at the lower end of the SES distribution within the Atole group. Thus, the activation of the processes that resulted in the developmental advantage of the low-SES group must have occurred early in life.
In contrast to what was observed at 4 and 5 years of age, the findings at 3 and 6 years showed neither main nor interactive effects. The absence of effects at 3 years of age agrees with the theoretical model that we propose below, one that incorporates the notion that the probabilities of detecting effects increase as the child grows older. Recall that the effects of treatment during the first 2 years of age on the Composite Infant Scale were restricted to the motor scale at 24 months of age.
The lack of either interactive or main effects at age 6 is not an aesthetically pleasing finding and cannot readily be explained. The personal characteristics and history of the subjects who were included in the statistical tests run for this age group do not differ in any substantive way from those subjects who were assessed at 4 and 5 years of age. The data do not offer any suggestive evidence to explain the finding.
In many respects, the subjects at the lower end of the SES distributions who benefited from the Atole in adolescence were disadvantaged compared to the rest of their community. For instance, on average, the mothers of these children had no more than 1 year of elementary schooling. The fathers had the lowest levels of occupation, and their housing was of the poorest quality across villages (e.g., small, without toilets, thatched roofs). During their preschool years, these subjects performed poorly on the cognitive test battery, and they were also shorter and lighter than other children of the same age.
The rate of physical growth, performance on preschool tests, and SES background of the low-SES subjects who benefited from the Atole suggest that their early development had been at high risk both in absolute terms and also by the standards of their own community. This high risk, moreover, was not derived solely from poor nutrition; it also stemmed from the many other adverse conditions of their environment (e.g., poverty and disease). An assessment of risk during these subjects' early formative years would probably have predicted poor test performance during adolescence. Yet their test performance in the follow-up was comparable to that of subjects at the highest SES level, and it was better than that of a comparable SES group in the Fresco villages.
The protective effect of the Atole conforms with results from longitudinal studies of exposure to early biological (e.g., low birth weight) and SES stress factors in which characteristics of the caregiving environment appear to shield the development of children against the adverse effects of exposure to such risk factors. For example, in a study of children in Kauai (Werner, 1986), the availability of alternate caregivers, the mother's work load (e.g., steady employment outside the household), and the amount of attention given to the children by the primary caregiver proved to be factors that protected the development of children considered to be at high risk in early life.
The significant interactions between treatment and maximum grade attained on outcomes such as the achievement tests suggest that the effects of the Atole were also modified by particular characteristics of the subjects. Recall that the relation between maximum grade attained and the Interamerican reading and vocabulary scores was positive for Atole but not Fresco subjects; Atole children who were in the upper percentiles of grade attainment scored significantly higher than Fresco children.
Intuitively, the interactions between treatment and grade seem to conflict with the interactions between treatment and SES, where those who were worse off were more likely to benefit from the dietary treatment. However, in the context of a rural society living in poverty, the differences in the developmental implications of SES and of maximum grade attained resolve such an apparent discrepancy. SES is a carrier variable of family conditions that were relatively stable over the lifetime of the subjects in the study and that, even among the families who were better off, pointed to a state of unmet basic human needs. On the other hand, maximum grade attained reflects increased exposure to a favorable environment as the children broaden their educational opportunities from one year to the next. This distinction in the developmental meaning of SES and maximum grade attained explains, as discussed below, why those at the lowest end of the SES distribution and those with the highest levels of formal education benefited most from Atole.
Further light on the developmental significance of the two sets of interactions that we have discussed is shed by a consideration of the three-way interaction that emerged in the analysis of the reading achievement test. As noted, the SES X grade interaction was significant for the Fresco but not for the Atole villages. Within the Fresco villages, associations between SES and achievement were positive and significant for both low and high levels of grade attainment. Conversely, these associations were not significant in the case of Atole.
Such a differential pattern in the interactions suggests that the truism that SES affects cognitive test performance is fulfilled in the Fresco group, independent of whether the focus is on opposite sides of the distribution of maximum grade attained. This is not the case with Atole: the truism is challenged since SES has no effect on test scores at either extreme of the school grade distribution - in order words, the high nutrient supplement is acting as a social equalizer.
We now sketch a theoretical model of the linkage that we propose between early nutrition and later development. We recognize that these efforts could be similar to painting broad strokes over a large canvas, representing a reality that may well be far too complex to be tested experimentally. The model is first discussed in general terms, followed by a description of its relevance to the particular findings from the follow-up.
The model builds on well-grounded empirical information that, among nutritionally at-risk children, variations in nutrition status (i.e., protein, energy, and micronutrient deficiencies such as iron and zinc) account for significant variations in physical growth, motor maturation, and physical activity (Husaini et al., 1991; Martorell & Habicht, 1986; Meeks Gardner, Grantham-McGregor, Chang, & Powell, 1990; Rutishauser & Whitehead, 1972; Simonson, Sherwin, Anilane, Yu, & Chow, 1969; Super, Herrera, & Mora, 1990; Vasquez-Velasquez, 1988). In Guatemala, the children in the Atole villages were taller and heavier than those in the Fresco villages, and such differences continued into adolescence (Martorell et al., 1980; Martorell, Rivera, & Kaplowitz, 1990). They were also motorically more mature at 24 months than the children in the Fresco group. Because motor activity was not assessed in the study, we do not know whether the children in the two groups differed in this regard; however, it is likely that this was the case. Both theory and empirical studies suggest that one of the mechanisms available to maintain energy balance among malnourished children is a reduction of energy expenditure in motor activity (Beaton, 1984; Grantham-McGregor, Meeks Gardner, Walker, & Powell, 1990).
We conjecture that the effects of malnutrition on body size, motor maturation, and physical activity are the linkages between malnutrition and delays in behavioral development over time. In particular, we conjecture that small body size, delays in motor maturation, and reduced physical activity contribute to the gradual formation of styles or modalities of social-emotional and behavioral interactions between the malnourished child and the environment that slow cognitive development and educational progress. This proposition is an outgrowth of current theorizing that the associations between particular biological characteristics of children and their environment tend to covary with each other (Plomin, DeFries, & Loehlin, 1977; Wachs & Plomin, 1991). The following is a more detailed analysis of our theoretical proposal.
1. In both industrialized and preindustrial societies, accelerated physical growth and height are associated with comparatively better performance on developmental scales in early childhood and intelligence and school achievement tests in adolescence. While the exact mechanisms behind these associations remain unclear and are likely to differ from one society to another, one possible explanation is that physical growth and body size determine in part ways children are cared for and treated socially. In particular, small children are likely to be treated as younger than their chronological age, and they trigger caretaking behaviors of nurturance and protection at a time when children of a similar age but of average size and maturity are searching for autonomy and independence from their caretakers. The opposite will be true for comparatively larger, more mature-looking children; their physical attributes will induce caretaking behaviors that, on average, are reserved for older children. For example, the mothers of malnourished children maintain more frequent physical contact and closer proximity to their offspring than mothers of well-nourished children of the same age (Graves, 1976, 1978; Lozoff, 1988; Lozoff, Klein, & Prabucki, 1986). Similar patterns of behavioral interactions between caretaker and offspring are also observed in animals; in some species, the frequency of adult-offspring contact correlates negatively with the offspring's age and size (Harper, 1970; Konner, 1976; Moss, 1967; Trivers, 1974).
2. It is now well established that malnutrition delays motor maturation in infants and young children. In keeping with the first proposition, there is also evidence that the timetable of motor maturation influences the nature and range of contacts that the organism has with its physical and social environment. Motor milestones, particularly those of self-locomotion (i.e., crawling, creeping, and walking), transform the child's perception of and physical relations with the physical space and the elements within it (Bertenthal & Campos, 1990; Bremner & Bryant, 1985). These developmental changes lead, in turn, to the acquisition of new perceptual skills (e.g., depth perception). In addition, self-directed locomotion increases social contacts, broadens social experiences, and enhances affective contacts with adults (Gustafson, 1984). Of concern to us here is that, because of delays in motor maturation, malnutrition limits the opportunities that these children have to move about the environment, learn from it, and develop cognitive skills that children who are of a similar age, but more mature, are acquiring.
3. Our last proposition is that physical activity in children is positively related to exploratory behavior - behavior exploring the physical and social environment. We recognize that there is not much empirical evidence to support this proposition; however, the evidence that is available comes mostly from studies of well-nourished groups, where, under ordinary circumstances, the level of motor activity of children is independent of their nutrient intake. In these groups, the critical issues may be the child's capacity to self-regulate and to modulate activity according to the demands of the environment rather than the characteristic level of activity (Wachs, 1990).
We conjecture that there is a positive relation between physical activity and exploratory behavior in those situations where the motor activity of children is partly determined by the intake of energy and micronutrients (e.g., iron). This postulate stems from the notion that one of the mechanisms available to the organism to maintain energy balance is to reduce the expenditure of energy through a decrement of motor activity. Thus, under these circumstances, the average level of activity in children is lower than the level of activity they would display if their nutrient intake would meet their physiological requirements. This low average level of activity is what we believe is associated with the reduced amount of exploration of the environment evinced among malnourished children.
The importance of exploratory behavior has been discussed at length in the developmental psychology literature (Berlyne, 1966; Bruner, 1968; Piaget, 1952). In fact, early exploration of events, people, and objects is seen as the foundation of cognition. With the development of perceptual systems and the maturation of new action systems, the young child discovers the particular attributes of each new physical and social stimulus that are relevant to the context and to his or her developmental stage (Gibson, 1988).
In sum, we propose that the effects of malnutrition on body size, motor maturation, and physical activity mediate the relation between malnutrition and delayed cognitive and behavioral development. In particular, there are three basic propositions: (1) Children who are small because of their nutritional history induce caretaking behaviors and social responses from others that are generally reserved for children of a comparatively younger age. (2) Their slow motor maturation delays the acquisition of particular cognitive abilities and social behaviors. (3) Their low level of motor activity that results from the need to maintain energy balance limits the exploration of the physical and social environment. These three sets of effects are not limited to a particular age; they operate and express themselves throughout the child's early, middle, and late childhood and adolescence. They activate mechanisms that span years in the lives of children who live under social, economic, and physical circumstances that are similar to those in rural Guatemala. The final outcome is the attainment of a level of cognitive competence that is behind the level of other children and adolescents of the same chronological age with a history of having received adequate nutrition.
The definition of effects that we propose is substantively different from that which assumes that malnutrition has direct effects on the central nervous system and, therefore, on cognitive function. However, we do not deny that such direct effects might indeed exist; the two explanatory models are not mutually exclusive. At issue now is to fit the model to the data that show no main effects on the psychoeducational tests but do show significant interactions: in one case, between treatment and SES and, in the second, between treatment and maximum grade attained.
Within the context of a poor rural society in Guatemala, the families with the lowest economic and social resources must face the most severe constraints to meet the health (e.g., medical care, hygiene, and sanitation), nutrition (e.g., dietary quality), and developmental (e.g., educational opportunities) needs of their children. This was also the case among the severely impoverished families who benefited the most from the Atole; in such a context, the Atole compensated in part for the existing constraints and pushed the children in the study toward faster growth and healthier development. In particular, their comparatively larger size, earlier motor maturation, better health, and, possibly, increased exploratory activities must have induced particular caretaking and social responses and led to the development of behaviors and the early acquisition of cognitive skills that compensated somewhat for the severe limitations of their impoverished households, which failed to meet their developmental needs. On the other hand, these effects did not occur at the other end of the SES distribution because, within the limits of a poor, rural community, even in the family and household environment of those who were better off there is not much that can be offered to a healthy child. In other words, there is a ceiling in the provision of potential benefits given to the children in the context of rural family poverty.
A natural extension of this argument is that, if the children who received Atole would have been exposed to an environment without such a ceiling, then the benefit of the nutritional supplement would have been greater. This postulated incremental effect is what we think explains the significant interaction between treatment and maximum grade attained. Despite all their limitations (Gorman & Pollitt, 1992), the schools in the villages made possible the effects of the Atole because they provided opportunities for intellectual growth. Accordingly, the higher the grade attained - that is, the greater the exposure to schooling - the greater the effects of the nutritional supplement.
The explanatory model also accounts for the effects observed among the children included in the late exposure cohort because nutritional supplementation accelerates growth velocity even after the second year of life, albeit at a lower rate than what is observed when it is implemented before the second year. Modest physical and health advantages among those in the late exposure cohort who received Atole are likely to have activated the same process that we have described in connection with the children in the maximum exposure cohort. The difference between the advantages of the two cohorts is a matter of degree of effects.
An empirical test of the model that we propose is not feasible because we lack the necessary data concerning motor development milestones and physical activity measurements; moreover, even if these data were available, the samples would become too small to permit analysis once all the relevant variables were taken into consideration. At this stage, this model remains strictly conjectural.
The provision of health and nutrition services is a means of promoting growth and development among children. In the United States, for example, although the specific objectives of food assistance programs are often not clearly defined, some programs (like the federally funded Special Supplemental Food Program for Women, Infants, and Children [WIC]) target nutritionally at-risk groups in order to prevent growth and developmental delays associated with malnutrition (Pollitt, Garza, & Leibel, 1984; Rush et al., 1988). Do the results of the Guatemala study provide a public health justification for food assistance programs such as WIC? In particular, do the data support a claim that such programs will help prevent cognitive developmental delays? Such preventive effects would have broader social and economic implications if it were to be shown that these kinds of delays affect work capacity and productivity.
The findings that we report support the assertion that, among nutritionally at-risk infants and children, early supplementary feeding contributes to a salutary development of the complex mental abilities required in an industrialized society among those at the lowest socioeconomic levels. However, they do not support a claim that the provision of food alone assures a significant net increase in daily intake. The Guatemala study was a major undertaking, with a solid and efficient administrative infrastructure created for research purposes. There was close access to the target group and continuous prompting to attend a station, where subjects received the supplement and socialized with other people. Without this prompting, the food supplement may not have had the desired effect. In their comprehensive review of supplementary feeding programs in developing countries, Beaton and Ghassemi (1982) concluded that the net increases in food intake were often lower than intended.
A related issue regarding treatment effects and their programmatic implications is the nutritional status of the target population. The prevalence of growth retardation in the population studied in Guatemala points to a high risk of malnutrition among infants and children. For example, a recent analysis of anthropometric data of subjects who had been measured at 3 years of age showed that 25.8% of the sample had severe stunting (3 standard deviations or more below the reference median) and that 42.5% had moderate stunting (2.9-2.0 standard deviations below the median) (Martorell, Rivera, Kaplowitz, & Pollitt, 1991). These estimates are similar to those made prior to the initiation of the longitudinal study in 1969 (Martorell, Habicht, & Klein, 1982).
The external validity of the Guatemala findings must be assessed in context, and generalizations are restricted to populations with a nutritional status similar to that in the rural villages in Guatemala. Moreover, the interactions between the nutritional treatment and SES indicate that, even within these populations, there is a differential response to treatment. Not all those in a nutritionally at-risk population benefited from the intervention.
On the surface, it may appear that findings of SES x treatment interactions speak clearly in favor of targeting treatment to individuals of lower SES. However, a closer look shows that the problem is more complex The interaction between treatment and maximum grade attained indicates, that individuals who may already have some social and educational advantages may be helped to excel even further by nutritional supplementation. Programmatically, it may be easiest to identify and target only those at the lowest end of the SES distribution within a community, but doing so means that other children who could benefit from the supplement will not receive it.
From a policy perspective, it must be recognized that the social or economic significance of the psychological test findings obtained in the Guatemala study is not readily apparent. While the construct validity of the tests was supported on theoretical grounds, no assessment was made of their predictive validity with reference to a behavioral criterion such as social adjustment or work productivity. The wide age range in the sample precluded such an assessment since many subjects were still in school and not earning any income. Moreover, to our knowledge, there is no relevant information from comparable rural populations in Latin America that could be used to establish the ecological significance of the Guatemala findings. Thus, the test score differences cannot be translated into other, more tangible terms, such as work productivity, earnings, or social adjustment. Although it is theoretically justified to assume that variation in scores on tests of reading and numerical ability are associated with variations in social and economic behaviors, it is also conceivable that such covariations may be lower than expected in a society in which basic human needs remain unmet.
The effects of public health programs, particularly behavioral effects, need to be analyzed in the context of a society's explicit and implicit social policy. In the context of rural Guatemala, the benefits of a supplementary program in enhancing development must be contrasted with the consequences of other existing conditions that counter such development. The school system, for example, is terribly inefficient and does not respond to the basic educational needs of the population (Gorman & Pollitt, 1992). Less than half the children enrolled in the first grade finish primary school, and many remain functionally illiterate. It is highly unlikely that the provision of food will prevent or remedy the consequences of not receiving an adequate formal education in a changing society.
Programmatic actions that focus on unmet nutritional needs and that have beneficial effects on human cognitive development are potentially a step forward in social policy. However, in our view, such actions are deceptive if they are framed in the context of a social policy that disregards other basic human needs and does not attend to the overall quality of life. Unmet nutritional needs generally coexist with, among other things, unmet needs in education, housing, sanitation, and health care. Only by meeting all these needs in conjunction with nutritional needs will we have truly moved forward toward a fair, humane society that sustains the rights of children and fosters cognitive and social-emotional development.