Contents - Previous - Next

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

Human nutritional requirements

Nutrition and educational achievement part II. correlations between nutritional and behavioural test indicators within populations where malnutrition is not a major public health problem
A consideration of allowable fibre levels in weaning foods

Nutrition and educational achievement part II. correlations between nutritional and behavioural test indicators within populations where malnutrition is not a major public health problem

Ernesto Pollitt and Nita Lewis
Human Nutrition Center, School of Public Health, University of Texas, Houston, Texas, USA

Two types of data are reviewed here. One refers to feeding modalities in the infancy period, and the other to anthropometric measurements (e.g., height), and they are both related to either cognitive measures or school achievement indicators.


Data from two large epidemiological studies in Great Britain (1) and in the United States (2) have shown the existence of a relationship between the modality of feeding in early infancy and subsequent cognitive ability and school achievement. After controlling for confounding variables, breast-fed infants obtained statistically higher scores than bottle-fed infants in later tests of intelligence, and in reading and mathematical attainment tests. In the British study (1), the sample of 5,362 subjects was selected from all live births during one week in March 1946 and was statistically considered to be representative of all legitimate singletons born in Great Britain during that week. Data were collected from mothers of the infants soon after birth and over the pre-school and school-age periods.

Further limitation to those strictly bottle- or breast fed, availability of complete family background information, and a birth weight of at least six pounds resulted in a sample size of 1,464 and 1,398 children with test scores at 8 and 15 years of age, respectively. Except for word-reading at 8 years of age, there was a statistically significant (p < .01) difference in all test scores in favour of the breast-fed over the bottle-fed infants. In picture intelligence (eight years), the children who had been breast-fed were estimated to score 1.76 points above bottle-fed children when confounding variables were held constant. This difference was identical in the case of the non-verbal ability test (15 years); in mathematics it was 1.55 points (15 years), and in sentence completion, 1.73 points (15 years). Although appropriate statistical corrections were made for the possible influence of social environmental variables, it is still conceivable that there may have been unidentified variables that influenced, in the same direction, the choice of feeding modality and rate of intellectual growth.

The US study was based on a nationally representative sample of 7,119 non-institutionalized children aged 6 to 11 years examined over the 1963 - 1965 period (2). The data set comprises complete medical and developmental histories for each child, school reports, and intelligence testing scores. The present analysis was restricted to 3,599 white children who lived with both of their natural parents and who had complete data sets. It represents, to an extent, a privileged sub-sample from the total pool on which the overall study was based. The measures of cognitive development used were an IQ calculated from two sub-tests of the Wechsler Intelligence Scale for Children, and a school achievement measure derived from the reading and arithmetic sub-test of the Wide Range Achievement Test.

The findings on the modality of early feeding and later performance are in keeping with those from the British National Survey. The IQ and achievement test scores of breast-fed children were one to two points higher, on the average, than those of children who had never been breast-fed. This impact seemed to be totally independent of all other variables that were examined in a regression equation. The authors do not advance any specific explanation for this finding, and simply speculate that it may be related to better health, or possibly to other aspects of child-rearing associated with feeding modality.

Another study was based on a sample of patients from the private office population of a paediatrician (3). The data showed that early feeding history was associated with prevalence of learning disorders. Twenty-nine children with learning disorders were compared with 53 control children referred to the same physician for other neurological conditions. Thirteen per cent of children with learning disorders had been breast-fed, compared with 47.2 per cent of the control children.

These three studies suggest that breast-fed, compared to bottle-fed, children show some small advantage in cognitive tests and some educational attainment indicators. Whether this comparative advantage is associated with the differences in the biochemical composition of breast- and bottle milk, or because both choice of feeding and rate of intellectual growth are causally associated with a third child-rearing variable, is unknown.


In the US Health Survey study already cited (2), a positive and statistically significant correlation was found between height (at 6 to 11 years) and the different cognitive test sources. Children who were one standard deviation above average in height for their age scored more than one point higher on IQ and achievement tests than did children of average height. However, there was essentially no relationship between weight and cognitive development. The authors indicate that this differential finding should not be surprising. They considered height to be a better summary measure of the lifetime nutritional status of the child, while weight conveys information primarily about current nutritional status.

In the Collaborative Perinatal Project of the US National Institutes of Neurological Diseases and Blindness (4), the relationship between anthropometric characteristics and behavioural test scores was also assessed. This longitudinal investigation of pregnant women and their children, conducted in 14 medical centres throughout the US, researched the relationship between factors and conditions affecting the parents, especially the pregnant mother, and the occurrence of central nervous system abnormalities in the offspring During visits to the prenatal clinic, the mother provided interviewers with her medical history, and with socioeconomic and genetic information about herself and her family, and the baby's father and his family. The mother's physical status was re-evaluated prior to delivery, and the events of labour and delivery were recorded. Paediatric, psychological, neurological, and other developmental examinations were conducted up to the time children were close to ten years of age. The sample with which the present review is concerned consisted of 37,945 live-born single births to white or Negro mothers (Puerto Ricans excluded).

The Stanford Binet Intelligence Scale score at four years of age is the outcome variable for the present analysis. The independent or explanatory variables included 29 prenatal variables, 13 neonatal variables, and 23 infancy and childhood variables. Stepwise multiple regression analysis was used gradually to restrict the analysis to those variables with the greatest explanatory power.

For both the white male and female children, head circumference at four years of age and weight at one year of age were potent (statistically significant) explanatory variables of IQ variance. Their contributions to IQ variance, moreover, were independent of the effects of all other variables. In the case of the white males, the effects of weight at one year and head circumference at four years were .61 IQ points per kilogram and .89 points per centimetre, respectively. For white females, it was .77 IQ points per kilogram of weight at one year and .82 per centimetre in head circumference at four years. For the Negro males and females, the stepwise regression analysis also showed important contributions of the anthropometric variables to IQ variance. The most important variables for the Negro males were weight at four years and head circumference at one year. There were .39 IQ points per kilogram of weight, and there was also an increment of .56 IQ points per centimetre of head circumference at one year. In the case of the Negro females, there was an increment of .31 IQ points per kilogram of weight at four years and there was also an increment of .47 IQ points per centimetre in head circumference at one year.

It is not clear why the explanatory anthropometric variables in the regression equations were not the same in whites and blacks.

In conclusion, anthropometric variables both at the time of testing, as well as four years before testing, explained part of the variance of IQ at four years of age. It is likely, although far from conclusive, that this association is due to a nutrition input into the cognitive development of the children.

In summary, the three long-term studies with a very large data base that we have reviewed indicate that there are clear relationships between the growth characteristics of the children and subsequent cognitive performance.


In addition to discussion of nutritional deficiencies, consideration of the effects of hunger and of nutrition intervention in the school on behaviour is included in this section.

Most of the available data on the relationship between protein-energy malnutrition (PEM), iodine, and vitamin-A deficiency on cognition and school performance in developing countries were included in part I of this paper (rood end Nutrition Bulletin, vol. 2, no. 3). The present part is restricted to data on iron deficiency and anaemia and adolescent school behaviour in the United States.

Correlational data collected in the United States suggest that iron deficiency and anaemia in the adolescent period are associated with both behavioural disturbance and low achievement in school (5, 6). In one study of 12- to 14-yearold junior high school students living in an economically deprived area of a black community of Philadelphia, 193 adolescents were selected to participate because of their haematological status. Ninety-two of these students were classified as anaemic (Hb. ranged from 10.1 to 11.4 g/dl), with the remaining 101 students serving as a normal control group (Hb. ranged from 14.0 to 14.9 g/dl). The dependent variable under investigation was scholastic performance, as measured by the composite score on the Iowa Test of Basic Skills.

Results indicated that anaemic students performed less well than the non-anaemic groups. In addition, a sex-by-age-by-haematologic status interaction was found. This triple interaction showed that anaemic girls performed less well than non-anaemic girls at all ages by approximately the same degree (the difference in test score means between non-anaemic and anaemic 12-,13-, and 14-year-old girls was- 0.50, - 0.59, and- 0.54 grade levels, respectively); anaemic boys, however, showed no deficit at 12 years, but greater deficits in performance with increasing age. The difference in test score means between non-anaemic and anaemic 12-, 13-, and 14-year-old boys was 0.34, - 0.25, and- 1.65 grade levels, respectively. In addition, the anaemic group took significantly longer (4.08 seconds) than non-anaemic children (1.81 seconds) in reporting a visual afterimage. Moreover, a teacher evaluation indicated that anaemic males displayed significantly more conduct problems than did non-anaemic males (5 - 7).

It is necessary at this point to distinguish between short-and long-term affects of the nutritional variables in question. Whereas short-term effect refers to the behavioural or cognitive impact of temporary food deprivation or hunger conditions, long-term effect refers to the behavioural impact of a nutritional input variable(s) over months or even years.

In a recently published review of the educational benefits of the US school feeding programme, attention was given to the short-term behavioural effects of morning feeding and hunger (8). Restricted to school children, the review focused on the behavioural impact of eating or not eating breakfast, or of having a mid-morning snack. All studies reviewed were strictly correlational; there were no studies where the dietary variable had been manipulated experimentally. These studies did not yield a uniform set data from which conclusive inferences could be derived. Two researchers (9, 10) examined emotional dimensions of behaviour; two others (11, 12) focused on cognitive components, and two more (13, 14) concentrated on measurements of physical activity. The studies on emotionality suggested that, at least among pre-schoolers and children up to about fifth grade, a morning snack was likely to be beneficial in "very general behavioral terms" to the children. Yet it was not possible to infer, from the information reviewed, what these specific benefits really were. The studies and the data reviewed were too ambiguous and lacked behavioural specificity.

The studies on cognition showed some discrepancies. On the one hand Dwyer et al. (11) found that breakfast did not have a detectable effect on attention. On the other hand, Matheson (12) observed that having breakfast did indeed have some beneficial effects on performance on an arithmetic test and a decoding task. This latter author concluded that students score higher on school-type tests undertaken shortly after food is given than when food is not given.

From the data that Dwyer et al. (11) and Matheson (12) present, it is not possible to classify these contradictions. Conceivably they tapped or tested different mental abilities that had, in turn, different degrees of sensitivity to the dietary variable. It is also conceivable that the home food intake of the populations in the two studies was different. If this were the case, then, given the nature of their experiences, the subjects' response to the dietary school treatment should have been different. These alternative explanations must remain hypothetical. Nonetheless, it should be pointed out that the Matheson study, which is methodologically one of the best in the literature, supports the contention that morning food supplementation in school has beneficial effects on the children's performance on school-type tests.

Finally, in connection with the measurements of physical activity, there again seems to be some contradiction between the data. It must be noted that one evaluation involved a breakfast/no-breakfast condition, whereas the other study compared the effects of various breakfasts having different caloric, carbohydrate, and protein levels. Perhaps the most meaningful finding, in terms of the present concern, is that the omission of breakfast interfered with the children's maximum work output.

In summary, the data give some indication that short-term hunger (due to lack of breakfast) may have some adverse effects on emotional behaviour, arithmetic and reading ability, and physical work output as measured by an ergometer.

Recently a study was completed (15) where there was an experimental manipulation of the morning dietary intake of school children to establish its impact on behavioural testing at noon. The subjects included in this study were 23 girls and 11 boys who ranged from nine to 11 years of age. They were all well-nourished, healthy children. All were admitted twice to a clinical research centre; both admissions took place at about 4:00 p.m., and there generally was a seven-day interval between them. The following morning half of the sample ate breakfast at 8:00 a.m. The other half received no breakfast. On second admission, this order was reversed. The composition of breakfast included a total of 535 kilocalories, 15 grams of fat, and 70 grams of carbohydrate.

On both days tests were administered at 11:00 a.m. Psychological testing included problem-solving tasks that required a fast and accurate response, a vigilance task (i.e., attention to visual stimuli presented rapidly), and a short-term recall test. The mean IQ for the group was 110.4. The data indicated a differential effect of the no-breakfast condition according to the IQ of the child. Among those children whose IQ was above the median for the whole sample, the no-breakfast condition affected the accuracy and speed of their responses in the problem-solving task. They tended to respond faster and made more mistakes when they had not had breakfast. This effect, however, was not seen in the children with an IQ below the median. There was no effect on the vigilance task; and there was a surprising positive effect on the short-term recall. Most children in the no-breakfast condition tended to respond better to one of the six items of the memory task than did those who had eaten breakfast.

The data from this controlled study indicated that 19 hours of overnight and morning fasting affected selected features of mental tempo and information processing, but these effects do not seem to be uni-directional. The data showed that problem-solving competence is adversely affected, especially among those cases with a high IQ. However, it is also apparent that at least one feature of selective attention of short-term memory may be benefited. The behavioural signs detected on the cognitive test need not be taken as an indication of a primary effect on the central nervous system. In fact, variations in arousal level associated with variations in peripheral factors may explain the findings, since the metabolic state determined by the no-breakfast condition may have induced changes in peripheral neuro-physiological receptors. The behavioural response to these changes may, in turn, have affected performance on the tests administered. The findings of this experimental study are in keeping with the types of general conclusions that were derived from the review of the literature, which were that, in terms of the cognitive test performance, it seems likely that the availability of breakfast to school children has a beneficial effect on the children's performance.

In connection with long-term effects of school lunch and breakfast on behaviour, the data show that two investigators (16, 17) found a beneficial effect of school breakfast on school performance. However, other investigators (18 - 22) failed to detect such impact. From the data available from these authors, it is impossible to identify precisely the reasons for such contradictory findings. Most of the reports present only brief descriptions of their samples and methodology. Nonetheless, it should be pointed out that some of the data suggest that there are many important moderating variables (for example, degree of participation in the feeding programmes, teacher expectations of success, food intake on the day achievement tests are administered) that have to be controlled in order to have a reliable assessment of the nature of the correlation between the feeding programme and the measurements of achievement. Conceivably, the differences in the design and the samples in the various studies may also account for some of the different results obtained.

It must be emphasized that probably all the studies that were reviewed on the long-term effects were focused on well-nourished populations. An important question that obviously remains unanswered is whether or not a programme that starts with children who are not well nourished and that brings about a nutritional improvement would benefit the educational status of such children.


In developed countries- especially within populations where malnutrition is not a major public health problem- the picture of nutritional effect on cognitive function is diffuse. This state of affairs is determined by the scarcity of information, and the fact that the information that is available is not conclusive. Unless new data are forthcoming, it will not be possible to determine the importance of nutrition in the estimation of cost-benefit ratios in school intervention programmes. Accordingly, if the criterion used for policy implementation is documentation of the functional consequences of nutritional variability within what is accepted as normal, then this decision must wait. If, on the other hand, the criterion for implementation is the availability of information suggesting the possibility of functional consequences secondary to mild nutrient deficiencies, such as iron deficiency, then there is already a basis for action.

The data reviewed in this paper (parts I and II) have different policy and economic implications for developing and developed countries. Econometricians working in these two types of countries would find different weights for the nutrition factor in economic-production-functions (23). From an educational policy perspective, countries with high rates of malnutrition must attend to these deficiencies if they are concerned with the improvement in the quality of educational results. The data in this respect are conclusive.

In populations where malnutrition is not considered a major public health problem, there may be- although it has not yet been proved conclusively- functional consequences secondary to normal nutritional variability. This is particularly true where the concern is with cognitive effects of short-term fasting. Available experimental data suggest that this condition may act as a "stressor," increasing the chances of error in problem-solving tasks.


1. B. Rodgers, "Feeding in Infancy and Later Ability and Attainment: A Longitudinal Study," Develop. Med. Child Neurol., 20: 421 11978)

2. L.N. Edwards and M. Grossman, "The Relationship between Children's Health and Intellectual Development," in S. Mushkin, ea., Health: What Is It Worth? Measure of Health Benefits, Pergamon Policy Studies (Pergamon Press, New York and Oxford, 1980).

3. J.H. Menkes, "Early Feeding History of Children with Learning Disorders," Develop. Med. Child Neurol., 19: 169 (1977).

4. S.H. Broman, L. Nichols, and W.A. Kennedy, Preschool IQ Prenatal and Early Developmental Correlates (Wiley, New York, 1975).

5. T.E. Webb and F. Oski, "The Effect of Iron Deficiency Anemia on Scholastic Achievement, Behavioral Stability and Perceptual Sensitivity of Adolescents," Pediat Res., 8: 294 (1973).

6. T.E. Webb and F. Oski, "Iron Deficiency Anemia and Scholastic Achievement in Young Adolescents," J. Pediat, 82:827 (1973).

7. T.E. Webb and F. Oski, "Behavioral Status of Young Adolescents with Iron Deficiency Anemia," J. Special Ed., 8: 153 (1974).

8. E. Pollitt, M. Gersovitz, and M. Gargialo, "Educational Benefits of the United States School Feeding Program: A Critical Review of the Literature," Amer. J. Publ. Hlth., 68: 477 (1978).

9. D.A. Laird, M. Levitan, and V.A. Wilson, "Nervousness in School Children as Related to Hunger and Diet," Med. J. Record, 134: 494 (1931).

10. M. Kiester, "Relation of Mid-Morning Feeding to Behavior of Nursery School Children," J. Amer. Dietet Assoc., 26: 25, 11950).

11. J.T. Dwyer, M.F. Elías, and J.H. Warren, "Effects of an Experimental Breakfast Program on Behavior in the Late Morning," (Department of Nutrition, Harvard School of Public Health, Boston 1973, unpublished).

12. N.E. Matheson, "Mid-Morning Nutrition and Its Effects on School-Type Tasks" (Ph.D. dissertation, University of Southern California, Los Angeles, 1970).

13. W.W. Tuttle, K. Daum, R. Larsen,J. Salzano, and L. Roloff, "Effect on School Boys of Omitting Breakfast: Physiologic Responses, Attitudes and Scholastic Attainment," J. Amer. Dietet Assoc., 30:674 (1954).

14. I. Arvedson, G. Sterky, and K. Tjernstrom, "Breakfast Habits of Swedish School Children," J. Amer. Dietet Assoc., 55: 257, (1969).

15. E. Pollitt, D. Greenfield, and R. Leibel, "Effects of Short-term Fasting on Cognitive Test Performance among 9 to 11-YearOld Children," (paper presented at the meeting of the American Academy of Child Psychiatry, Oct. 1978, San Diego, Calif., USA).

16. F. Lininger, "Relation of the Use of Milk to the Physical and Scholastic Performance of Undernourished School Children," Amer. J. Publ. Hlth., 25: 555 (1933).

17. S.W. Krietzman, Evaluation of the Craddock Breakfast Study (Atlanta School of Dentistry, Emory University, Atlanta, Gal, USA, 1973).

18. H.M. Lieberman, l.F. Hunt, A.H. Coulson, V.A. Clark, M.E. Swendseid, and L. Ho, "Evaluation of a Ghetto School Breakfast Program," l Amer. Dietet Assoc., 68: 132 (1976).

19. S.A. Fellers, "A Study of the Effects of Breakfast on Scholastic Attainment, Dropout Rate and Knowledge of Nutrition" (Ph.D. dissertation, Boston University, Boston, 1967)

20. F.F. Tisdall, E.C. Robertson, G.H. Drake, S.H. Jakson, H.M. Fowler, J.A. Long, L. Bronha, R.G. Ellis, A.J. Phillips, and B.S. Rogers, "Canadian Red Cross School Meal Study," Canad. Med Assoc. J., 64: 477 (1951).

21. M.S. Pinkus, "A Study of Pupil Breakfast Habits and Behavioral Patterns in Certain Louisiana Elementary Schools Following Implementation of the National Breakfast Program" (M.A. thesis, Louisiana State University, Baton Rouge, La., USA, 1970).

22. T.M. Koonce, "Does Breakfast Help?" Schl. Food Serv. J., 26: 51 (1972).

23. L. Alexander and J. Simmons, The Determinants of School Achievement in Developing Countries: The Educational Production Function, International Bank for Reconstruction and Development, Staff Working Paper No. 201 (Washington, D.C., 1975).

Contents - Previous - Next