Contents - Previous - Next
This is the old United Nations University website. Visit the new site at http://unu.edu
Ernesto Pollitt and Patricia Kariger
The literature reviewed in this article suggests that breastfeeding positively influences cognitive development. Studies comparing the effects of early feeding methods on cognition consistently show mental test score advantages for breastfed infants over bottlefed subjects. Also, breastfeeding may function as a buffer against adverse developmental outcomes from early traumatic events such as low birthweight or neurologic insults. It is recognized that breastmilk contains long-chain polyunsaturated fatty acids that are necessary for the normal development of the retina and cerebrum. Yet, breastfeeding may not always promote optimal development in children. Infants breastfed by mothers using alcohol or marijuana have lower scores on developmental tests. Thus, while breastfeeding has the potential for contributing to the healthy development of infants, it may also be capable of transferring toxic substances known to interfere with normal development.
Human biology and psychology provide support for the hypothesis that breastfeeding benefits mental development. On the one hand, biochemical components of breastmilk affect particular elements of the neural circuitry that contribute to information processing [1-3]. On the other hand, in addition to the emotional ties that arise between mother and infant from suckling, it is plausible that breastfeeding helps the development of interpersonal communication between infant and caretaker .
Several lines of research bear on the issue of the effects of breastfeeding on mental development.
The most obvious is the search for a direct relationship between the presence or absence of breastfeeding, at one end, and measures of cognition, at the other end. A second line, which has received little attention in the literature, refers to breastfeeding as a primary or secondary preventive agent. Finally, there is research that raises a caveat to the idea that breastfeeding can do no wrong. At issue is whether breastfeeding could operate as a developmental risk factor by the transfer of toxic elements such as alcohol that could limit mental development.
Direct effects of breastfeeding
Several studies have attempted to test the hypothesis that breastfeeding gives infants developmental advantages over those who are bottle-fed. Limitations in research design prevented most of these studies from carrying out a fair test of the hypothesis. By definition, correlational studies do not address issues of causality because of the uncertainty that all potential confounders have been controlled for.
Because developmental scales that are used to test infants and toddlers do not assess the same psychological constructs as those assessed by intelligence tests administered to older children, we have separated studies of infants and toddlers from the remaining studies. Longitudinal studies that include assessments at both ages were placed in both age groups.
Infants and toddlers (table 1)
Five studies [5-9] are included in this group. All used the Bayley Scale of Mental Development for the developmental assessment. Because the study by Rogan and Gladen  also included evaluations with the McCarthy Scales at three, four, and five years of age, we included their later results in the following section.
The definition of the independent variable (feeding method) varies among the studies and prevents pooling the data to draw any generalizations. For example, the number of types of feeding methods ranges from two  to five . Likewise, the nutrient composition of the formulas given to the bottlefed babies was generally not reported, and the nature and number of potential confounders accounted for and the socio-economic status (SES) of the subjects varied among studies.
TABLE 1. Studies of the effect of breastfeeding on development in infants and toddlers
|Study||Study design||Sample||Exclusion criteria||Outcome variables||Significant results|
|Temboury et al., 1994 ||Prospective, longitudinal, correla- tional study Cognitive development measured once between 18 and 29 mo Data on background variables Collected when subject was tested Subjects classified by method of Feeding: bottle, breastfed < I mo (n = 99); breast, breastfed 23 mo (n = 130)||229 newborns from Madrid hospitals; about half middle and half lower SES||Gestational age <36 wk Birthweight <2,400 g Perinatal complica- tions Major birth anoma- lies Poor 5-min Apgar score Breastfed for 2 mo Follow-up testing not completed||Bayley MDI and PDI scores at 18-29 mo Covariates: Maternal education, age, working status SES No. of siblings Psychosocial risk Testers also coded for shyness, temper tantrums, hyperactivity||Low MDI scores at 18-29 mo were predicted by:|
|bottle-fed low maternal||1.86||.044|
|no. of siblings||2.63||.002|
|Rogan & Gladen, 1993 ||Prospective, longitudinal, correla- Tional study Mothers interviewed about feeding Methods from birth to 5 yr Cognitive development measured at 6, 12, 18, and 24 mo Subjects classified by duration of breastfeeding and weaning age:||600+ full-term infants from mostly white, middle SES, well- educated families in North Carolina, USA||Mothers did not intend to breast- feed or lacked in- terest in being in long-term study||Bayley MDI and PDI scores at 6, 12, 18, and 24 mo Confounders: Maternal age, race, occupation, educa- tion, smoking and drinking during pregnancy Child's sex, birth weight, birth order||Mean differences at 24 mo: Very long vs short: +6.7 points, MDI Very long vs bottle-fed: +5.6 points, MDI; +5.9 points, PDI Positive effects for maternal age and birth order were also found|
|Morrow- Tlucak et al., 1988 ||Prospective, longitudinal, correla- tional study Cognitive development measured at 6,12, and 24 mo Household characteristics measured at 12 and 24 mo Subjects classified by duration of breastfeeding: bottle only (n = 157); breast <=4 mo (n = 39); breast >4 mo (n = 23)||229 mostly low SES infants whose mothers were recruited during prenatal care visits at a Cleveland hospital||Gestational age <37 wk Perinatal complica- tions Admission to neonatal care unit Known maternal drug use or psychosis Primary language not English Child placed for adoption Delivery and discharge on weekend Family intends to move from area||Bayley MDI and PDI scores at 6, 12, and 24 mo HOME scores at 12 and 24 mo Confounders: Race, parental edu- cation, maternal intelligence, cigarette use, ma- tennal disciplinary attitude||Mean MDI score advantages for breastfed vs bottle only:|
|Breastfed group||Age tested (mo)||Point advantage|
|*F = 3.24; p = .041 **F = 3.76; p = .025 Regression analyses indicated breastfeeding uration still signif icantly related to MDI score at 12 and 24 mo, even with confounding variables and HOME scores entered in the analyses|
|Young et al., 1982 ||Retrospective, cross-sectional, correlational study Cognitive and motor development measured once at 6, 8, 10, 12, 14, or 16 mo Mothers reported feeding method used Subjects classified by feeding method: artificial, breastfed <= mo; mixed, breastfed, then given cow's milk after 2 mo of age; breastfed breastfed >7 mo||1,000 infants selected from data bank set up from Tunis birth records. Equal numbers of infants from 5 social classes rep- resented for each age group alcohol use||Social class not verifiable||Bayley MDI and PDI scores, body length and weight, disease preva- ance, allergies Confounders: No. of siblings; maternal educa- tion, smoking, alcohol use||MDI mean advantages for breastfed over mixed and artificial com bined, according to sex, SES, and age: +3.86 (t = 3.02; p < .01) for males, high SES, 10 mo +5.85 (t = 2.04; p < .05) for females, high SES, 10 mo +4.97 (t = 2.82; p < .01) for females, low SES, 12 mo +5.93 (t = 2.77; p < .05) for males, high SES, 12 mo +3.50 (t = 1.92; p < .10) for males, high SES, 14 mo None of the confounding variables were found to influence MDI scores|
|Florey et al., 1995 ||Follow-up correlational study examining the influence of feed- ing method on motor and mental development at 18 mo Feeding data obtained ffom hospital forms and records of nurses' home visits in the first 2 mo Bayley MDI and PDI administered to 582 children at 18 mo Subjects classified by feeding method: breasted (n = 228); bottle fed n = 354)||Population-based sample of 846 firstboni single- tons delivered in a hospitalinDundee, Scotland||Bayley MDI and PDI scores at 18 mo||Mean MDI scores: breastfed, 110.2; bottle-fed, 102.5 No group mean differences were found for PDI scores|
TABLE 2. Studies of the effect of breastfeeding on development in children
|Study||Study design||Sample||Exclusion criteria||Outcome variables||Significant results|
|Hoefer and Hardy 1929 ||Prospective, longitudinal, correlational study of growth Subjects classified by dura- tion of exclusive breast- Durabon feeding: artificially fed (mo) (n = 38); S3 mo (n = 78); 4-9 mo (n = 77); 10-20 mo fed (n = 190)||383 children 7-13 yr of age||No data on feeding method||IQ, EQ, PQ scores (tests not specified); age talking||Mean scores for tests according to duration of exclusive breastfeeding:|
|Duration (mo)||n||IQ||EQ||PQ||Age talking (mo)|
|Rogan & Gladen, 1993 ||Follow-up to study described rn table 1 Cognitive development measured at 3, 4, and 5 yr||See table 1||McCarthy Scale scores at 3, 4, and 5 yr 3rd-grade report card grades in English and math||Results on aH scales showed a trend for higher scores in the breastfeeding groups at all ages Mean score advantages ranged from +2.8 to +4.8 pornts for very long breastfed vs short breastfed or bottle-fed No significant differences between groups in 3rd-grade report cards|
|Fergusson et al., 1982 ||Prospective, longitudinal, correlational follow-up study Cognitive development measured at 3, 5, and 7 yr Subjects classified by months of breastfeeding: bottle-fed (0 mo breastfeeding); breastfed <4 mo; breastfed >4 mo||Birth cohort of children partici- patina in a New Zealand devel- opmental study .||Missing data on feeding method||IQ test scores: age 3, Pea- body Picture Vocab- ulary Test; age 5, Stanford-Binet; age 7, Wechsler IQ LC test scores: ages 3 and 5, Reynell Develop- mental Language Scales; age 7, Illinois >4 Test of Psycholinguistic Abilities Confounders: Maternal intelligence, education and child- rearing knowledge, SES; children's gesta- tional age at birth and birthweight||Mean advantages for breastfed vs bottle-fed groups on IQ and LC tests, controlling for confounders:|
|Age (yr)||Breastfeeding duration (mo)||Advantages|
|Rodgers, 1978 ||Longitudinal, correlational study Data from achievement tests obtained at 8 and 15 yr Mothers interviewed every 2 yr from child's birth to age 18 Subjects classified by feeding method: breastfed only (n = I,291); bottle-fed only (n = 1,133). Other subjects classified by age (months) at which they were charged from breast- to bottle- feeding||5,362 infants born in 1 week of March 1946, selected to represent all singletons born that week||Missing data on feeding method||Age 8: performance on school-administered picture intelligence tests and reading tests Age 15: performance on sentence completion, nonverbal ability, and math tests All scores were standard- ized on the survey population to give a mean of 50 and an SD of 10 Confounders: intelligence Parental interest in school, SES, maternal ability and paternal education, family size, birth order||Mean score advantage for breastfed only group vs bottle-fed only group on sentence completion test at age 15 was +3.5 (F = 4.89; p < .001) Regression analyses showed all listed con founders significantly predicted perfor mance on sentence completion test Predicted mean score advantage for breastfed only group vs bottle-fed only group, con founders held constant:|
None of the studies reported significant differences among any of their feeding groups during the first 11 months of life. The earliest differences were observed at 12 months. For example, Morrow-Tlucak et al.  reported that at 12 months the mean Mental Development Index (MDI) of exclusively bottle-fed infants was 10 points lower than that of the babies who were breastfed for at least four months (p < .05) (table 1).
After the first year of life, bottle-fed babies consistently obtained lower MDI scores than the respective comparison groups with the longest periods of breastfeeding. The intergroup differences ranged from 3.5  to 13.9  MDI points. However, there are inconsistencies in the data. For example, on the one hand, Temboury et al.  reported that infants breastfed for more than three months had higher MDI scores than bottle-fed babies (table 1). On the other hand, breastfeeding from 5 to 19 weeks did not give an advantage over bottle-feeding in the Rogan and Gladen  study (table 1).
Developmental scales generally lack the sensitivity to detect subtle delays in mental development during the first year of life (see, for example, ref. 10). It is therefore not surprising that the MDI scores of the bottle-fed and breastfed babies in the studies cited above first appeared after 12 months of age. Of importance here is that the MDI obtained during the second year of life is generally a modest predictor of later intelligence . Pearson correlations between MDI scores and later intelligence quotient (IQ) scores are generally less than 0.50. Consequently, the developmental significance of the observed MDI differences is questionable. A conservative conclusion is that, during the second year, the behavioural repertoire of the breastfed baby is broader than that of the bottle-fed baby.
Children (table 2)
In addition to a study published in 1929 , we found three studies [7,13,14] that reported the comparisons of scores from intelligence and school achievement tests in childhood of breast- and bottlefed subjects. As with the studies of infants and toddlers, these studies do not constitute a homogeneous set. There are differences in the ages at the time of testing, in the tests administered, and in the number and nature of the confounders accounted for. In one study  the last assessment occurred at 5 years of age, and in another  the subjects were tested up to 15 years of age. The tests also differed. For example, Rogan and Gladen  used the McCarthy Scales for the assessment of preschoolers, whereas Fergusson et al.  used the Peabody Picture Vocabulary Test for preschoolers, the Stanford-Binet Scale for five-year-olds, and the Wechsler Scale for seven-year-olds. Rodgers  used a picture intelligence test and tests of specific cognitive abilities (e.g., mathematics).
The studies of children generally included large samples, ranging from 383  to more than 5,000 , and found relatively narrow numerical differences between groups to be statistically significant. An evaluation of the importance of this issue, however, must consider the consistency of the findings across studies. The results form a distinct pattern. Independently of age and the tests used, subjects who were breastfed generally had higher IQs and achievement scores than those who were bottle-fed. For example, a study in Great Britain compared the performance of 1,291 exclusively breastfed (duration not specified) and 1,133 exclusively bottle-fed subjects in tests of sentence completion at 15 years of age  (table 2). The breastfed subjects obtained consistently higher scores than the bottlefed subjects.
Problems of studies of direct effects
There are two salient features in the studies of both infants and toddlers and of children. One is the weakness of the correlational design, which precludes conclusive inferences. The other is the consistency across studies in different populations and social contexts of the cognitive advantage of breastfed over bottle-fed subjects. Although the second feature cannot compensate for the weakness of the first, it strengthens the scientific merit of the original hypothesis.
The search for direct effects of breastfeeding on mental development has closely followed the prescription of the main-effect model in the biomedical literature. This model posits that exposure to a particular influential factor or event during critical early periods of brain growth results in structural changes that, in turn, determine the presence of a particular developmental outcome. At issue here is that this model is rarely indicated today in research on developmental psychobiology, because more often than not it has not worked (see, for example, Sameroff and Chandler ). Theoreticians argue that the influences on development are multiple and interactive, and that one single determinant seldom causes the type of effects attributed to breastfeeding . In the particular area of developmental risk, it has been shown that it is generally the number of risk factors to which an organism is exposed, rather than the nature of the factors, that influences cognition [17, 18].
In conclusion, shortcomings of design have prevented a fair test of the hypothesis that breastfeeding confers developmental advantages to infants. However, the consistency in the results strengthens the scientific merit of the hypothesis and indicates future experimental research, with the admonition that the magnitude of the effects is not likely to be large. Future research should be rooted in basic contemporary premises of human development, which underscore the influence of interactions between developmental determinants.
In addition to providing nourishment for the infant, the act of breastfeeding been associated with the establishment and promotion of the mother-child relationship. It has been proposed that the physiological changes that occur during breastfeeding may be related to the occurrence of emotional changes in the mother . That is, it is suggested that breastfeeding may enhance a mother's feelings of connectedness with her baby, which, in turn, positively influences the developing mother-child bond. However, the data to support this notion are weak, at best. In fact, the concept of bonding has been described as "scientific fiction" .
The potential preventive effects of breastfeeding may be divided into primary effects, by which breastfeeding would prevent a nutritional deficiency known to have adverse effects on mental development, and secondary effects, by which breastfeeding would protect the development of cognitive function from the adverse effects of an early necrologic disorder.
Primary prevention (table 3)
Although there are no published studies addressing this issue head-on, sufficient relevant information exists to document the role of breastfeeding as a primary preventive factor. Iron deficiency is associated with developmental delays among infants and toddlers . In addition, although there are discrepancies in the data, well-controlled clinical trials have shown that the developmental delays of babies with iron-deficiency anaemia are reversed following iron repletion therapy .
The iron in breastmilk is of low concentration but is highly bioavailable, so that breastfed infants are significantly less likely to become iron-deficient during the first six months of life than bottle-fed infants, especially infants fed cow's milk . Thus, if breastmilk prevents anaemia, it also prevents the developmental delays associated with anaemia.
Further support for this argument is found in a randomized, controlled trial recently conducted in Canada on the preventive effect of iron-fortified formula  (table 3). Two groups of infants up to two months of age were randomly assigned to receive Enfalac with 12.8 or 1.1 mg elemental iron per liter. Subsequently, the Bayley Scales of Infant Development tests were administered to the subjects at 6, 9, 12, and 15 months of age. The treatment-by-time interactive term of an analysis of variance accounted for a significant portion of the variance of the Psychomotor Development Index (PDI). In particular, there were significant differences in the expected direction between groups at 9 and 12 months. Fortified formula prevented the delays in psychomotor development observed in those who received low-iron formula. No differences were observed in the MDI. Since the MDI is a poor indicator of mental development during the first 12 months, the absence of differences in MDI is not surprising.
In conclusion, when infants are artificially fed without iron fortification, breastfeeding is likely to operate as a shield against developmental delays associated with irondeficiency anaemia. This is particularly true in lowbirthweight infants, who are at high risk for iron-deficiency anaemia . However, exclusive breastfeeding after six months of age is not likely to have this effect against irondeficiency anaemia .
Long-chain polyunsaturated fatty acids (LCPUFA) are components of neural tissue, and recent research on their effects on selective neural function in laboratory animals and humans provides strongly suggestive evidence of the potential contributions of breastmilk to cerebral function [27, 28]. Briefly, LCPUFA, particularly arachidonic and doeosahexaenoic acids, play important roles in the development of the retina and cerebrum, and deficiencies of these in laboratory animals and low-birthweight infants may cause selective dysfunctions in visual function and learning [1, 27].
In utero, effective transfer from the mother across the placenta satisfies the need for LC-PUFA in the foetus. Later, maternal milk, a rich source of doeosahexaenoic acid, meets the needs of the young infant . Generally, however, the contents of docosahexacnoic acid in formula are insufficient for newborns or young infants, as demonstrated by the low plasma levels of docosahexaenoic acid in formula-fed infants and by visual function differences between breastfed and formula-fed infants . In one study, pre term (four months adjusted age) and full-term infants who were fed human milk performed significantly better than formula-fed infants in tests of visual evoked potentials and of preferential looking acuity. Further, at 36 months the breastfed term infants also performed better than the formulafed term infants in dot stereo acuity and lettermatching ability tests .
TABLE 3. Studies of the primary preventive effects of breastfeeding
|Study||Study design||Sample||Exclusion criteria||Outcome variables||Significant results|
|Moffatt et al., 1994 ||Prospective, longitudinal, randomized clinical trial comparing the effects of iron-fortified and regular formula on the cognitive and motor development of Age (mo) infants 112 subjects received iron fortified and 113 regular formula for 15 mo Cognitive and motor devel opment measured at 6, 9, 12, and 15 mo||225 bottle-fed infants 0-2 mo old born in hospitals in the area of Winnipeg, Canada||Bayley MDI and PDI scores at 6, 9,12, and 15 mo||No significant differences between groups in MDI scores Mean score advantage for iron-fortified formula vs regular formula group in PDI scores:|
|p = .02 (group x time interaction)|
|Agostoni et al., 1995 ||Prospective, quasi-experi- mental study comparing the effect of breastmilk, LC-PUFA-enriched for- mula, and standard for- mula on neurodevelop- mental scores at 4 mo Infants whose mothers chose to breastfeed were ran- domly assigned to receive LC-PUFA-enriched for- mula or standard formula Neurodevelopmental re- sponses measured with the Brunet-Lezine psycho motor development test at 4 mo Blood fatty acid content measured at 4 mo||90 infants born in a clinic in Milan, Italy, during 1 yr Classification of sub- jects: exclusively breastfed (n = 30); LC-PUFA formula (n = 29), standard formula (n = 31)||Gestational age <37 wk Apgar score <7 Any perinatal dis- order or disease||Scores on the Brunet- Lezine psychomotor test, which assesses gross and fine motor function, social reactions, and language||Brunet-Lezine scores significantly higher in the exclusively breasfed and LC-PUFA formula groups than in the standard formula group (p < .05)|
|Blood fatty acid levels significantly higher in the exclusively breastfed and LC-PUFA formula groups than in the standard formula group|
A recent study  compared the developmental test performances of three groups of subjects receiving human milk, conventional formula (with LC-PUFA precursors), or an experimental formula supplemented with pre-formed LC-PUFA (table 3). All eligible subjects were between 37 and 42 weeks gestational age, were free of any type of diagnosable disorder, and had five-minute Apgar scores of 7 or better. Mothers of 30 of the subjects chose to breastfeed their babies. The remaining subjects were randomly assigned to receive either the conventional formula (n = 31) or the experimental formula (n = 29).
A psychomotor developmental test (BrunetLezine) was administered to 86 of the subjects at age four months (two subjects in the conventional formula group and two subjects in the experimental formula group could not be tested). The BrunetLezine Test assesses gross and fine motor function, social responses, and language. A developmental quotient score is obtained from scores on the subscales. The mean score on the Brunet-Lezine Test for all subjects was 101 (SD = 11) and the range was 80-136, similar to the standardized values reported by the authors of the test. Infants receiving the LC-PUFA-enriched formula or human milk scored significantly higher (105.3 and 102.2, respectively) on the Brunet-Lezine Test than the subjects receiving the conventional formula (96.5). There were no differences among the three groups in parental and socio-economic variables. Blood analyses for a subgroup of infants revealed that fatty acid levels were higher for the infants receiving the LC-PUFA-enriched formula or human milk than for the infants receiving conventional formula.
Although these and other similar studies  help our understanding of the relations among breastmilk, LC-PUFA, and cerebral function, they do not speak directly to the question of the contribution of breastmilk to the growth of intelligence. Neither theory nor empirical data suggest that visual acuity, visual evoked potential, or letter discrimination during the first months of life are precursors of the development of central information processing or of abilities that constitute the cornerstones of intelligence. Further, although the Brunet-Lezine Test, the Bayley Scale of Mental Development, and other developmental scales are useful tools to describe the behavioural repertoire of four-month-old infants, in most cases developmental scale scores obtained at that early age do not predict later developmental scale performance even at the end of the first year of life. Developmental scales administered in early life do not measure cognition.
Secondary prevention (table 4)
At issue here is the potential protective effect of breastfeeding against the sequelae of prematurity or an early neurologic disorder. Lucas and collaborators [34-36] have published a series of articles on the short- and long-term effects of breastfeeding and formula-feeding on the performance on developmental and IQ tests by children born prematurely with very low birthweights (1,850 g). The basic design is that of a prospective, randomized trial in five centres to assess the effects of diet on neurodevelopment (table 4). A limitation of these studies is that the definition of the samples is very limited, and the articles do not discriminate between the subjects who participated in the different studies. Further, the direction of the hypothesis tested in each study is not clearly indicated.
A study published in 1989  compared the development of pre-term infants admitted to the neonatal units in Cambridge, Ipswich, or King's Lynn, England, from 1982 to 1984, who were randomly assigned at birth to either donated banked breast milk (REM) or pre-term formula (PTF) (2 g protein, 0.335 MJ; 35 mg phosphorus; 70 mg calcium; 45 mg sodium, 100 ml) as the sole diet. The mothers in this trial (trial I) had chosen not to provide their own milk to their offspring. In trial II, comparisons were also made among the offspring of women who elected to express their own milk to feed their children. These children were also randomly assigned to either BBM or PTF to supplement their intake from maternal milk. The diets were discontinued when the children were discharged from the neonatal unit or when their body weight was at least 2,000 g. Comparisons were made at nine months of age on their performance in the screening inventory of Knobloch et al. , which includes five developmental areas (adaptive, gross motor, fine motor, language, and personalsocial). This assessment tool is a screening inventory rather than a developmental scale, and is not generally recognized as a test of intellectual development, as suggested (p. 320) by the authors. A neurologic examination (Amiel-Tison and Grenier) was also administered to the children, who were classified as normal, equivocal, or impaired.
When the data from both trials were pooled at nine months of age, the subjects who had received PTF, either as the sole diet or as a supplement, had significantly higher scores in the adaptive, fine motor, language, and personal-social areas than those who had received BBM. Further, in trial II those subjects who had received PTF had a higher developmental quotient than those who had received BBM, as well as higher scores in the adaptive, language, and personal-social subscales .A finer analysis showed that among those who had received some milk from their own mothers, those who had received more than 50% of their total intake from BBM were at the highest disadvantage. Moreover, the covariation between diet and development was stronger among those classified as small for gestational age (below the 10th percentile).
Contents - Previous - Next