Contents - Previous - Next

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

Overview of findings from the literature

About 30 studies carried out in developing countries have specifically attempted to compare well-nourished and malnourished children and their families in the same low income communities, while differing in focus and methodology. Most of them still centre on identifying pathology in the malnourished rather than adaptive qualities of the well-nourished, although this distinction is somewhat artificial. Only one study known to the writer, the Burma study cited earlier, has divided children into well, average, and malnourished, so that the top and bottom of the curve can be contrasted with the middle. Dr. Maria Luz Alvarez of Chile has been the most active researcher in the field, with the largest number of publications, mainly in Spanish (see Alvarez in the References). She has compared parental, environmental, and behavioural characteristics of the well-nourished and the malnourished.

Table 2 summarizes in detail the results of 10 studies that characterize the type of research that has been conducted and the range of research findings. The findings will be discussed by topic area in sections that synthesize the broader research literature.

Capsule summaries of a variety of other studies are as follows:

In Thailand, a study by Somchai Durongdej of 365 children aged 0-48 months living in lowincome congested areas of Bangkok found that male sex, high birth weight, low morbidity, good appetite, ante-natal care, well-baby clinic attendance, consumption of colostrum, and current breast-feeding were factors that determined whether babies were well-nourished or malnourished (Durongdej et al., 1987).

In Chile, a study by Maria Luz Alvarez of more than 1,000 families, comparing low-income families with well-nourished versus malnourished infants, disclosed significant differences in the parents' own childhood history and current social-support structure and marital adaptation. Mothers of the well-nourished were more likely to come from intact homes, perceive that they were loved as children, have positive parental models, and a currently stable union with fewer arguments and more demonstration of affection. Fathers had a more stable employment history. Mothers of the malnourished were found to need models of good parenting behaviour, which might be provided by mothers of the well-nourished. Adolescent pregnancy was a risk factor for poor parenting (Zeitlin and Ghassemi, 1986). (Dr. Alvarez has reported these results, taken here from workshop proceedings, in a number of separate papers noted in the References. )

Table 2. Types of relationships reported by existing positive-deviance studies

  Imesi. Western Nigeria (Morley et al., 1968) Solis, Mexico (Chavez et al., 1974)
Study objective Examination of socio-medical factors affecting the lives of children and their eventual use as "indications for special care" of children at risk of malnutrition Determination of the biological and cultural factors that give rise to adequate development among a certain prop ortion of the children in the adverse conditions of a poor environment. This knowledge could possibly be used for the planning of more rational applied programmes
Research design and population characteristics Sample drawn from a longitudinal study of 405 children under five years of age living in the village of Imesi, Western Nigeria. This community from the Yoruba tribe relies mainly on farming for its living Analysis of the differences in food, care, and health ex isting among two groups of children of the same social stratum but with a different nutritional status. Sample drawn from 220 children below four years of age in a rural community
Positive-deviance discriminators Weight-for-age

- Malnourished children MN(52): weight-for-age be- low the 10th percentile (local weight distribution) on one or more occasions at 6, 9 or 12 months of age
- Well-nourished children WN(52): weight between the 50th and 75th percentile (40) or above the 75th percentile (12)

Clinical signs of malnutrition

- MN(37): presented obvious signs of malnutrition
- WN(36): did not present any alteration in growth, nor signs of malnutrition

Mother-infant interaction Not reported Greater mother-child interdependence in MN children; mothers were closely restrictive in the first group while liberal in the case of WN children
Child behavioural characteristics Not examined Not examined. WN children were older than MN ones.
Maternal psychosocial characteristics, nutrition/health practices Not examined Mothers of WN children had modern concepts regarding child illness, while MN children's mothers tended toward "magic concepts" of health and treatment. MN were iso lated in comparison to WN children
Maternal and child physiology, diet, mor- bidity, mortality, etc. Mothers of positive-deviant children were heavier after delivery but mothers' heights. and fathers' heights and weights, were not related to the child's weight Mothers of WN children consumed diets superior in quantity, quality, and variety, currently and in the past, especially during pregnancy
No significant difference in the duration of breast- feeding and time of introducing other foods between the two groups. Fathers of MN children had worse hygiene than those of WN No difference in the duration of breast-feeding, age and type of supplementary feeding, but a difference existed in weaning practices as MN children were weaned quickly and abruptly
Family size and structure, maternal education, and other socio-demographic correlates of malnutrition No difference in family size, fathers' occupation and re- ligion between the two groups child ratio in the WN group No family size difference between the two groups but family composition differed because of a higher adult/
Significant difference in birth order MN children were more common above a birth order of 7 No difference in family income, land property &father's education (mother's education standardized), but food expenditure was higher among families of WN children. A difference was found in the degree of westernization. Mothers of WN children were more inclined towards modernization, had more contact with the external world, were out of the home and town more often, and took their children out more frequently
Community health, social services, family and social support Marital system: low-weight children were associated with broken marriage, father's death, and lack of other family support Not examined
  Mexico (Cravioto and Delicardie. 1976) Kampala, Uganda (Goodall, 1979)
Study objective Examination of the micro-environmental factors that lead to the appearance of severe malnutrition among low-income families, and assessment of the same factors in a matched group without diagnosis of clinical malnutrition Adverse social factors are not the only factors explaining the occurrence of malnutrition since these factors occur in families with no malnutrition. This study seeks to fill the gap in the knowledge in this regard.
Research design and population characteristics From a prospective longitudinal study of 334 children under 5 years of age, 22 malnourished children were matched at birth. for sex, gestational age, birth weight, psychomotor development and socio-economic factors, with well-nourished children Samples were drawn from 107 children with kwashiorkor from inpatients of New Mulago Hospital, Kampala, Uganda ( 1969-1972) and from 111 controls with various diseases not attributable to socio-economic factors, infectious agents; or normal siblings of the malnourished children
Positive-deviance discriminators Clinical signs of severe malnutrition
- MN(22): 15 developed kwashiorkor and 7 marasmus
- MN(22): children selected from the same birth cohort who were never diagnosed as severely malnourished
Clinical signs of severe malnutrition
- MN(50): presented clinical signs of kwashiorkor
- WN(50): did not present any signs of severe malnutrition or infectious disease as they were matched for sex, age, and tribe or origin.
Mother-infant interaction WN children had significantly more favourable environments in terms of home stimulation: the amount of verbal communication, expressions of affection toward the child, mother's co-operation and reaction when the child performed well, and need gratification were significantly higher in the WN group  
Child behavioural characteristics Not examined Not examined
Maternal psychosocial characteristics, nutrition/health practices Significant differences were found in the behavioural responses of the mothers to the test situation. as mothers of WN children were more supportive. communicative, and responsive to their child's performance Mother's contact with the mass media made a significant distinction between the two nutritional groups, as mothers of WN children listened to the radio more frequently Not examined
Maternal and child physiology, diet, morbidity, mortality, etc. Not examined All kwashiorkor children were weaned, even before illness developed, while 13 of the controls were still breast-fed. Statistically significant differences were found between "good" and "bad" reasons for weaning with ' bad" reasons being mostly given in the kwashiorkor group. For most of those children weaning was abrupt. Twenty-eight kwashiorkor children were sent away from home during weaning, as compared to 8 controls
Family size and structure, maternal education, and other socio-demographic correlates of malnutrition No difference in the age, height, or weight of either parent, the mother's number of pregnancies, or the number of live children between the two groups 9 and 18 in the other control group No significant relationships between the presence or absence of malnutrition and variables of personal cleanliness, literacy, and educational level No difference between the two groups on the grounds of socio-economic factors, family size and structure (nuclear or extended) Significantly, more children with kwashiorkor had fathers with a poor income: 23 cultivators and 4 salaried were represented in the kwashiorkor group as opposed to Stability of marriage: only 20 couples of the kwashiorkor group were found to be consistently living together, as compared to 31 in the control group Parent-child relationship: 16 MN children lived with their parents v. 33 in the control group
Community health, social services, family and social support Not examined Child attendant: kwashiorkor children were less likely to have their mothers as attendants in hospital than were control children. The change in attendants in case of child illness was most marked in the kwashiorkor group
  Bangladesh (Graves, 1976) Kathmandu, Nepal (Graves, 1978)
Study objective To explore whether differences in specific components of infant and maternal behaviour could be detected between poorly nourished and adequately nourished young children. The study deals with chronic or moderate malnutrition rather than with severe cases of malnutrition Same objective as the Bengali study, with an additional one to test whether it was possible to reproduce the findings of the West Bengali pilot study in a different cultural setting, using a cross-sectional approach
Research design and population characteristics Samples drawn from children attending the maternity and health clinic in Singur District, West Bengal (1971-1972). The population is mostly agricultural Age of study children ranged from 7 to 18 months Samples drawn from children attending maternal child health clinics around Kathmandu, Nepal Same age of study children covered in the Bengali study
Positive-deviance discriminators Three anthropometric measurements were done: - weight-for-age based on local values - weight/length2 x 100 – mid-arm circumference/head circumference ratio MN(23):
- weight-for-age below 30th percentile
- weight/length2 x 100: at or below 0.14
- mid-arm circumference/head circumference ratio: below 0.280

- weight above 70th percentile
- weight/length2 x 100: at or above 0.15
- mid-arm circumference/head circumference ratio: at or above 0.310

Same 3 anthropometric measurements used in the Bengali study MN(36): Same criteria of selection as in the Bengali study WN(38): Same criteria of selection as in the Bengali study
Mother-infant interaction Mothers of WN children were more responsive and attentive to behavioural signals but did not interfere with the children when they were involved in play. Mothers of MN children were less responsive and demonstrated a lack of mutuality in their interaction. The greater need for close physical contact with the mother may be viewed as a characteristic of insecurely attached children. Maternal distance interaction also differentiated between the two nutritional groups: WN children used this mode of communication significantly more frequently than MN children; a high degree of reciprocity existed between mothers and children in the WN group while it was lacking in the MN group Maternal behaviour toward the child was not different between the two nutritional groups, and the intercorrelations between maternal and child scores showed similar or parallel patterns of reciprocity for the WN and the MN children
Child behavioural characteristics No difference in the intellectual performance levels between the two groups, but the intercorrelations between the development quotient and maternal scores suggested that the determinants of intellectual performance are different: higher performance levels were noticed among the WN children as they were accompanied by low levels of maternal attention, while the MN boys' intellectual performance level was directly dependent on the level of on-going maternal attention and stimulation MN children differed from WN both in exploratory and attachment behaviours: MN showed markedly less vigour when handling the toys; the amount of overall attachment behaviour and of interaction across a distance with the mother was significantly lower than among WN, and they preferred to maintain close physical contact with the mother, staying on her lap rather than moving away from or around her MN children scored significantly lower in their intellectual performance levels. The overall time spent in play was decreased, while time spent breast-feeding increased MN children showed significantly lower levels of exploratory activity and attachment behaviour, especially distance interaction, and a heightened need for physical closeness to the mother
Maternal psychosocial characteristics, nutrition/health practices Mothers of MN children had lower scores in 4 of 6 measures of maternal behaviour; they tended to marry and have their first child at a younger age and expressed a negative, unwelcoming attitude toward pregnancy with the study child No difference was found in maternal behaviour between the two nutritional groups
Maternal and child physiology, diet, morbidity, mortality, etc. Mothers of MN children had a significantly lower average weight All children in both groups were breast-fed; even though there was no difference in the age of introduction of adult foods, there was a significant difference in the inclusion of animal protein source in the children's diet between the two groups Sibling deaths were reported more frequently among MN families. Same observation as in the Bengali study with regard to children's diet was made; the only difference between the two groups was the inclusion of animal protein food
Family size and structure, maternal education and other socio-demographic correlates of malnutrition No significant difference in socio-economic status between the two groups. Parents of MN children had significantly lower education levels Family size and structure: families of MN children tended to have, on average, more children. No statistical difference in family income. Educational levels of parents were lower in the MN group Larger family size among the MN group
Community health, social services, family and social support Families of MN children were more frequently nuclear families (parents, children) rather than extended families (one or more relatives living with the nuclear family) Although the same number of nuclear/extended families was reported in both groups, MN families were more crowded
  Jamaica (Kerr et al., 1978) Kenya (Dixon et al., 1982)
Study objective To examine the role of social functioning in the development of infant malnutrition and to explore the relationship between infant malnutrition and maternal psychosocial behaviour, by comparing mothers of MN children with mothers whose children were matched for age and family income but who were not malnourished Many sub-Saharan East African countries are undergoing rapid social changes: increased population growth, accelerated conversion of traditional agricultural to partial cash-crop economy, and rapid cultural and political changes. It is the purpose of this study to examine the impact of these rapid social changes on the nutritional status of individual children and to see what characteristics and circumstances distinguish the MN children from the rest of the population, and what factors in the family unit are associated with failure to meet a child's nutritional needs
Research design and population characteristics Samples drawn from children admitted to University Hospital (Kingston. Jamaica) for severe malnutrition (malnourished, N = 6; controls, N = 6) Sample drawn from an agricultural Bantu tribe living in the fertile highlands of south-western Kenya (N = 597)
Positive-deviance discriminators Weight-for-age Growth failure after excluding non-nutritional causes and rapid weight gain after dietary therapy (a) Weight-for-age
(b) Major signs of acute malnutrition
(c) a and b combined
Mother-infant interaction Not examined The common and basic risk factor of malnutrition appears to be an alteration in the usual attachment of the child to a primary caretaker. In this study, for large proportions of the MN subsample, societal, familial, individual, and economic events combined to produce the "bonding failure," with nutritional failure as a secondary event. The malnutrition of these children should then be labelled a symptom of an "attachment disorder." Since the Gusii have no tradition of shared child care among adults, care by grandmothers, or adoptive mother, is considered a deviant pattern in their society
Child behavioural characteristics Not examined The MN infant himself may also have contributed to the altered attachment. Interactional difficulties between mothers and MN children were suspected to begin very early, multi-determined and perhaps self-perpetuating. For others, the onset of difficulties may happen later on, linked to a shift in family relationships around the time of weaning and the birth of a sibling Development assessment was impossible to complete in the MN subsample because of the affective characteristics of the MN children, who were sad, apathetic, and difficult to engage in the test situation. This was in contrast to the longitudinal sample children, who readily co-operated after a short warm-up period. The 28 WN children, assessed by Bailey scales, showed scores above average at every age for PDI and MDI
Maternal psychosocial characteristics, nutrition/health practices Both groups followed the same custom of passing on children to grandparents. relatives or other adults. Mothers of MN children (MMC) were often aware of the poor treatment inflicted upon their children but did nothing about it. They showed lack of anxiety about their children's health Early traumatic separation had occurred with similar frequency among MMC and controls, but control mothers thought that this had resulted from major upheavals that were beyond the control of their families, and maintained positive feelings toward their parents, while all MMC perceived that they had been deserted Of MMC who had fathers or father substitutes, 7 had encountered severe beatings and 4 had suffered from sexual abuse By the time of adolescence, most of the controls had left home (independent) while MMC were still living in hostile family situations In spite of an equally deprived environment, the psychosocial functioning of the MMC and controls was significantly different: MMC were apathetic, dependent, passive, and isolated, with poor self-esteem and low energy level, or aggressive, manipulative. and evasive  
Maternal and child physiology, diet, morbidity, mortality, etc.   Three mothers of the subsample were reported chronically ill and were unable to do farm work or care for the child. This was contrasted with the vigour and consistent hard work displayed by Gusii women in general. Three mothers of the subsample were not caring for their children because of alcoholism, as compared to only one in the whole community who was known to drink to excess Unusual incidence of prematurity and small size at birth (55 per cent of cases) and significant morbidity in the pre-natal period in the subsample was compared to a quite low level of perinatal casualty in the whole Gusii village. Weaning: 13 of 20 MN children were weaned between 10 and 36 months. This early weaning pattern for the subsample children was deviant from that of their own community
Family size and structure, maternal education and other socio-demographic correlates of malnutrition Work history of MMC was very poor; 10 wanted to work but were waiting "passively" for someone to find work for them Eight of 10 control mothers who wanted to work were employed for by maternal grandmothers (with 4 accepting Housing: Controls made better use of their homes and gardens. In contrast, many homes of MMC were exceedingly ill-kept The economic situation of families of the subsample seemed to have no consistent relationship with the nutritional status of the child Seven out of 20 MN children were cared for primarily by someone other than the natural mother. Five were cared reluctantly) and 2 by a grown sister while 2 had no specific caretaket. The usual pattern of child care in Gusii population is through the natural mother Seven of the subsample children were illegitimate, hile the whole village had only 3 per cent illegitimacy (about one-tenth the incidence in the subsample)
Community health, social services, family and social support Relationships with men: Control mothers' relationships with men were more stable. In MMC these relationships were not stable, were not based on trust but on money or other compensations, and usually ended in pregnancy. In families where fathers were present, there was often intense parental conflict and physical abuse Relationships with family and friends: Most control mothers had mutually supportive associations with churches, neighbourhoods, and extended families. They considered existing community institutions very supportive. On the other hand, most MMC had close family relationships, characterized by dependency, close supervision, and intense attachment, followed by accusations of having been exploited  
  Santiago, Chile (Alvarez, 1982, 1983) Guatemala (Scrimshaw and Scrimshaw, 1980)
Study objective Maternal cultural and psychological deficiencies have been recognized as important intervening factors in the genesis of infantile malnutrition, but have not been adequately investigated. It is the purpose of this study to compare the non-verbal language (NVL) expressiveness of mothers of malnourished and healthy children To investigate determinants of positive deviance (positive growth status and survival of all children) among plantation wage-earning mothers
Research design and population characteristics Positive-deviance discriminators Samples drawn from children attending health centres of the National Health Service in Santiago. Chile. Weight-for-age using NCHS standards - MN(20): below 10th percentile of weight-for-age - WN(80): above 25th percentile and not hospitalized more than once in the preceding 3 months and for no longer than 10 days Samples drawn from plantation wage-earning mothers (N = 100) Weight-for-age: death was reported among pre-school siblings Positive deviants: all children m the family were well nourished and there were no pre-school deaths Malnourished: low weight-for-age was accompanied by a mortality history
Mother-infant interaction Not examined Not examined
Child behavioural characteristics Not examined Not examined
Maternal psychosocial characteristics, nutrition/health practices No difference in intellectual performance (IQ) between the two groups . . . Mothers with MN children had higher degrees of dissatisfaction or were poorly satisfied with their family life (in 70 per cent of cases), while 85 per cent of mothers of healthy children considered themselves "moderately or completely satisfied'' Mothers of MN children had a low degree of NVL expressiveness 90 v. 45 per cent m healthy children's mothers) Subindex for feeding act showed that all mothers of MN children had low scores while only 45 per cent of mothers of MN children had similar scores Mothers who were more successful (with positive-deviant children) were entrepreneurial and also contributed to family income by petty trading, marketing, animal husbandry, kitchen gardening, etc., whereas those with poorer children were not earning comparable additional income One of the strongest discriminators between mothers in positive-deviant households and those with malnutrition and mortality history was an enterprising versus a fatalistic attitude
Maternal and child physiology, diet, morbidity, mortality, etc. Not examined Not examined
Family size, structure, maternal education and other known socio- demographic correlates of malnutrition All families belong to a low socio-economic level (stratum 5 of Graffai's modified scale) All families were standardized with respect to: number of children; age and birth weight of study children; Job situation of the mother; and absence of other female relative living in the house There are no significant differences between the two groups with respect to number of individuals in family group, number of children, maternal age, education, duration of breast-feeding, and marital stability The only variable in the "past history" that distinguishes significantly the two groups referred to the jobs held by mothers before their infant was born: 65 per cent of mothers with healthy infants had held jobs, compared to only 35 per cent in the other group Positive-deviant mothers contribute to family income from trading, gardening, and other hidden resources, whereas mothers of MN children were not earning comparable additional income
Community health, social services, family, social support Not examined Not available

In Pakistan, a comparison reported by Julian Lambert between 38 households in which all children under five were above 90 per cent of weight-for-age and 134 households having at least one child below 60 per cent, with equal income and food expenditure, found higher consumption of edible oils, dhal, eggs, and rice and longer birth intervals in the well-nourished families (Zeitlin and Ghassemi, 1986).

In Mexico, 25 well-nourished infants, aged 8.5 to 20.5 months, studied by Marian Zeitlin, F. Catherine Johnson, and Robert LeVine, were found to receive more physical assistance from their mothers in eating meals and snacks than 25 age-matched malnourished infants in the same low-income squatter community. The well-nourished also received more powdered milk, which was given in feeding bottles and mixed with atole (sugar water cooked and thickened with corn or rice starch) (Zeitlin and Ghassemi, 1986).

In Bangladesh, a survey of 180 9- to 18-month-old infants, in one upland and one lowland rural site (Zeitlin et al., 1985), found that the mothers of well-nourished infants differed from those of the malnourished in their method of cleaning infants' faeces, mother's own handwashing after defaecation, age of introducing supplementary foods, and educational aspirations for the child. Diarrhoeal rates differed significantly between the well-nourished and the malnourished. These rates were related to the infant's contact with chickens, ducks, goats, cows, and human faecal matter while playing on the ground. and to the dryness of the earth surface where the baby was left to play.

Types of Information Found by Existing Positive-deviance Studies

The few studies that have compared well-nourished and malnourished children living under conditions of socio-ecomomic underdevelopment and poverty have pointed to three types of variables related to good growth:

  1. Socio-economic correlates of malnutrition, such as mother's education, family size and birth spacing, knowledge of modern nutrition and utilization of health services, and other factors associated with growth status in low-income environments.
  2. Physiological determinants of growth, including maternal height, weight and metabolic factors, parity, morbidity, and dietary intake of nutrients.
  3. Psychosocial and behavioural aspects of mother-child interaction, associated characteristics, and the social network surrounding the dyed.

The psychosocial correlates belonging to this third group are the primary focus of this paper, for a number of reasons: variables in the first and second categories are already reasonably well known, and have been addressed by the major child-survival interventions, including PHC activities. The remaining physiological determinants of growth are in the domain of basic biomedical research.

Different Bodies of Literature Relevant to Positive Deviance

Taken in isolation, the findings of the positive-deviance studies are too fragmentary to synthesize into a meaningful whole. However, each finding of these studies supports the conclusions of the broader review of the literature on pages 45-79. This review draws on a wealth of closely related literature that cannot be pigeonholed under the label of positive deviance, but that elucidates the associations that contribute to the positive-deviance concept.

Although psychosocial and behavioural characteristics will be the focus of the literature review, we will first identify the various bodies of literature relevant to positive deviance. While scanning and abstracting information from these literatures, we could not summarize them all in this paper. To do so would be to resynthesize all previous work in the field of MCH nutrition in developing countries.

It is important for the reader to know under which headings information related to positive deviance may be located. The focus on the top (normal) end of the developing-country curves makes it necessary to review literatures from both male nourished and well-nourished populations. Factors that influence child growth have been studied by a wide range of biomedical and social-science disciplines. The number of life-stages affecting growth contributes to the vastness of the topic. Positive deviance in the infant may start in the girlhood of the mother. Therefore, it becomes necessary to consider not only the nutritional status of the infant at the weaning age, but also maternal nutrition before and during pregnancy, birth weight, neo-natal status, and maternal nutrition during lactation. Figures 2 and 3 represent in simplified form three of the groups to be considered and the bodies of literature about these groups that are relevant to positive deviance. These figures have been drawn with reference to weight-forage. However, other nutritional status indicators could have been used. It should be noted that very little attention has been given to positive deviance per se, so that studies with other agendas must be examined for their applicability to positive deviance. Figure 2 represents infants and young children. Curve A shows the distribution of weight-for-age in a typical developing-country population. Bodies of literature relevant to these curves are listed below. Subtopics with particular importance for research in positive deviance are discussed in part 2. Persons interested in any given subquestion should pursue the complete literature pertinent to their subtopic or collaborate with colleagues who are already familiar with this literature. Relevant bodies of literature are:

1. A modest number of clinical, biochemical, and immunological studies of infants at the bottom of curve A, who require inpatient treatment or nutritional rehabilitation.

2. A large number of studies of malnutrition correlates in developing countries, focusing on the bottom or the whole of the growth curve. Few have examined the phenomena at the top of the curve, beyond noting that high socio-economic status characterizes the top. These studies tend to be micro-economic, dietary, and anthropometric in nature with little measurement of behavioural and social variables expected to differentiate positive deviants. Some of these data sets could profitably be reanalysed, controlling for income and looking for descriptors that characterize children with good growth.

3. A large nutrition and infection literature, suggesting that positive deviants are protected from high levels of infection either by parental behaviours or high natural immunity.

4. A growing number of failure-to-thrive studies in industrialized countries that yield richer psychological and social information than analogous studies noted under (1)

5. A moderate number of studies correlating the size of parents and children and adopted versus natural children, comparing racial groups, breast- versus bottle-fed and low versus high socio-economic groups, and linking micro-nutrient deficiencies and maternal smoking to growth.

Fig. 2. Young children. 6 months to 3 years.

6. Infant and childhood obesity studies. This literature is relevant to positive deviance because: (a) stunting combined with obesity is a prevalent form of malnutrition in some locations and (b) some positive deviants in food-scarce environments may have characteristics that would make them obese if food were plentiful.

7. A voluminous literature on child development, mostly unrelated to nutrition but elaborating in great detail upon the psychosocial factors described in this document.

Figures 3a and 3b represent developing versus industrialized country distributions of weight gain in pregnancy and birth weight. Relevant literatures relating to those figures are:

8. More than a dozen supplementation studies in developing and industrialized countries, proving that under certain conditions weight gain during pregnancy and birth weight can be increased by interventions improving maternal food intake. Some of these supplementation projects have gone on to show beneficial effects of maternal supplementation on neo-natal behaviour and child development.

9. A large literature on nutrition and fertility, indicating that birth spacing and total family size are important to positive deviance, in part because of effects on nutritional status during pregnancy and also because of maternal workload and strain on family income.

Fig.. 3a. Pregnancy weight gain.

Fig. 3b. Birth weight

10. A moderate-sized literature, primarily from industrialized countries, relating maternal prepregnancy weight, calorie and protein intake, micro-nutrient status, smoking, alcohol and drug use, blood pressure, glucose tolerance, and infection to weight gain during pregnancy and/or birth weight and/or child development.

11. A growing number of reports on factors associated with prematurity and low gestational weight-for-age, and their sequelae. Most of these are from industrialized countries.

12. A moderate-sized literature of very uneven quality on neo-natal bonding.

13. A growing literature on neo-natal assessment of high-risk newborns, involving cry analysis, interactions between teenage mothers and their newborns, etc. This group of studies does not address the question of what proportion of low-birth-weight newborns in developing countries exhibit behavioural abnormalities. In a country such as Bangladesh, for example, where the average birth weight is very low, the percentage of infants weighing less than 2,500 grams who behave abnormally should be lower than in industrialized countries, where low birth weight is commonly due to causes other than low caloric intake of the mother. Very little or no attention has been given to the high end of the neo-natal weight curve.

14. Lactation studies: relationships involved in determining lactation performance are too complicated to represent in a simple diagram. There is a literature supporting the conclusion that malnourished women produce less milk of slightly lower nutrient density than the wellnourished. Relationships between breast-milk volume and prolatin, diet, frequency and duration of nursing, other breast-feeding behaviours and introduction of supplements still remain indefinite. These topics have obvious relevance to positive deviance.

15. Studies of metabolic efficiency and psychological adaptation to nutrient deficits and to stress have been conducted primarily on animals and obese adults.

Positive Deviance

Factors that characterize infants at the high end of the growth curve in developing countries fall into a number of categories that are worth distinguishing. Typically, the distribution curve of the weights or heights of a malnourished child population is shifted to the left when compared to the distribution of an affluent country. As has been shown in many instances, the distribution does not acquire a major skew but retains a typical symmetrical or almost symmetrical bell-shaped curve. This means that all sections of the population are affected to some degree, including the so-called positive deviants, who are believed to be found at the right hand or upper part of the distribution.

The shape and position of the distribution is influenced by a number of factors. The width of the distribution is probably mostly due to genetically fixed factors affecting body size and shape, and to a minor extent influenced by the errors that occur when taking and recording the measurements. The position of the distribution in relation to that of a well-nourished population is determined by external factors which include nutritional adequacy, social, economic and behavioural variables. It is usually not possible to determine which of the many active factors are responsible for an individual child's precise place in the distribution. But it is possible, and this has been shown many times, to identify the major factors that affect the distribution as a whole. Income, for instance, usually has a positive effect on height but its influence on fatness is inconsistent from one population to another. In some societies and environments lower-income children tend to be thinner while in others they remain the same or fatter than upper-income children unless income is excessively deficient. Theoretically, it should therefore be possible to identify influences that come to bear primarily at the upper end of the distribution, where the "positive deviants" are located, but not over the whole width of the distribution.

Fig. 4.

The main categories of factors characterizing children who grow well under socioeconomic and environmental deprivation are shown on figure 4 by As and Bs numbered from 1 to 4.

A1 on the malnourished-population curve and B1 on the reference-population curve denote in combination A1/B1 the set of factors that distinguish between children falling below and above the internationally defined threshold for malnutrition, regardless of the population to which they belong. Al/B1 interventions would move all malnourished children upwards across the threshold dividing normal and subnormal growth. Total amount of food eaten, nutrient density of the food, and protection from infection are Al/B1 variables. Truly adequate MCH services can be expected to have an Al/B1 effect.

A2/B2 factors tend to characterize infants at the top of both the developing-country and the industrialized-country distributions. Large genetic size potential and certain types of metabolic efficiency should fall into this category. Mother-child interaction variables, child and maternal characteristics, and social-support structures described earlier are expected to fall into this category. These factors may in fact not cluster at the top end of the reference curve.

A3 factors distinguish infants falling at the top of the developing-country distribution but not of the reference curve. Good mothers living with high infant mortality and endemic malnutrition may coax their infants to eat large amounts of food (of low nutrient density) at frequent intervals and may restrict their freedom of movement in a manner that protects them from exposure to infection. These same behaviours manifested in the reference population tend to cause obesity. The Women, Infants and Children's Programs (WIC) nutritionists working with newly immigrated Hispanic mothers in the United States campaign against infant obesity by teaching the mothers not to bottle-feed the baby every time it cries and not to overly restrict its movement.

A4/B4 factors may describe characteristics of the threshold between clinical malnutrition and low average nutritional status in both malnourished and well-nourished populations. Although we picture positive deviance as falling at the top of a continuum, certain factors, such as intimate psychological support for the mother, appear to be necessary for a modestly satisfactory outcome but not sufficient for a superb result. A programme providing services for teenage mothers, for example, could be considered to be an A4/B4 intervention. These factors are important to positive deviance studies for two reasons:

  1. Failure-to-thrive studies provide a starting-point for attempting to identify vari ables that characterize positive deviance.
  2. The percentage of mothers exhibiting stress and depression, child abuse and other non-nurturing behaviours is high in some deprived communities. These variables may be characteristics distinguishing positive deviants from those below them on the growth curve. The Ik tribe of East Africa (Turnbull, 1972), for example, would represent an extreme case of a society in which non-nurturing behaviours were the norm.

Physiological Factors Contributing to Positive Deviance

Some infants and young children are physiologically better able to adjust to the stress of low energy and protein intake and high exposure to infection. Genetically, some have low nutrient requirements and/or enhanced ability to adjust metabolically to reduced food intake. Certain types of inadequate diets facilitate this adjustment more than others. Mind-body links are critical to growth, to the immune response and to most maturational processes.

Most research related to these topics is still in its infancy and is based on animal models, obese adults, cancer patients, and other subject groups rather than human infants and children. Although it is not within the scope of this paper to fully review this research, it is appropriate to orient the reader to the types of physiological mechanisms that are under investigation.

Energy Metabolism

Some degree of adaptation to food scarcity appears to be genetic. Genetic characteristics that reduce the individual's ability to dispose of excess energy and predispose towards obesity and adult-onset diabetes in industrialized, energy-surplus countries serve also to protect their bearers in developing countries from the effects of food scarcity.

When energy intake is restricted, the growing child can reduce energy expenditure by becoming less active, by reducing growth rate, and possibly by reducing the amount of energy that is burned off as heat. Conversely, when energy intake is excessive, the amount disposed of as heat possibly can be adjusted upwards. Reducing the amount of energy wasted as heat should be a harmless form of adaptation unless the mechanisms that produce this reduction also are linked to lower levels of activity, growth, and development. Reducing activity levels and growth rate are adjustments that cannot be classified as satisfactory. Although they often are adaptations consistent with normal health and performance, levels of achievement tend to be below the adapted individual's genetic potential.

Hormones Involved in Energy Conservation

The regulation of food energy as a supply of fuel or body heat will depend on the extent to which the child is nutritionally stressed. Further, the degree of adaptation and its successful implementation in the child will determine the form of PEM the child develops (Gopalan, 1958).

There is a spectrum of what may be called adapted PEM, which is consistent with survival and eventual maturation of the child into a functioning adult. This spectrum ranges from mild growth retardation to marasmus. The marasmic child has adapted well, in the sense that most levels of metabolic indicators are normal in marasmus. Kwashiorkor, on the other hand, displays extreme failure to adapt, as seen by severe biochemical changes in the child (Rag, 1982) and high case-fatality rates. Within the so-called adapted spectrum from mild growth retardation to marasmus, the most efficient adaptations are those that have the fewest detrimental effects on growth and development. At the lower extreme, marasmus may be considered to be an adaptation only in the sense that it is compatible with survival (barring severe infection). Developmentally, the marasmic child is retarded.

A variety of hormones affect the process of adaptation to reduced food intake. The following are some of the main hormones regulating energy utilization.

Thyroid hormones

These hormones regulate basal metabolic rate and thus the heat generated by metabolism (Danforth, 1983).

During nutritional deprivation the levels of protein-bound iodine are decreased, primarily because of the decreased synthesis of thyroxine carrier proteins. The hormone thyroxine, T4, may be selectively deiodinated into either an active energy mobilizing isomer, referred to as T3, or an inactive and energy-sparing isomer, referred to as rT3 (reverse T3). In pregnancy, rT3 increases, permitting the pregnant woman on a limited diet to transfer more nutrients to the foetus. High levels of T3 lead to high loss of energy as heat. Becker (1983) suggests that the coexistence of low T3 and high rT3 levels present in adults suffering from PEM is also present in children. Genetically efficient individuals may have low T3 production and utilization rates or low T3/rT3 ratios and hence be more successfully adapted.

Hormones secreted by neurons of the sympathetic nervous system

Norepinephrine (noradrenaline) or NE, secreted by the sympathetic nerve endings to the organs, regulates basal metabolic rate and the activity rates of many organ systems, including mental activity (Landsberg and Young, 1983). NE also is secreted by the adrenal medulla into the bloodstream, but this source of freely circulating NE is not an important part of the regulatory mechanism we discuss here. Caloric restriction reduces NE, particularly when blood glucose, insulin, and lipid levels fall. NE appears to be a major regulator of dietary thermogenesis, or the burning off of energy as heat that follows a meal (during the period when glucose and insulin levels are elevated). During fasting, reduced NE lowers the amount of energy released for necessary uses into the bloodstream (substrate mobilization). Individuals in whom sympathetic nervous system activity is genetically low and in whom adaptation to calorie restriction occurs with high efficiency would be energy-efficient.


Insulin removes glucose, free fatty acids, and amino acids from the bloodstream into the cells, both for use and for storage. These effects are decreased under conditions of insufficient energy. A high insulin response causes wastage of energy, as glucose is oxidized or cycled about more rapidly than it needs to be (Danforth, 1983). A lower insulin response permits glucose and other substrates for metabolism to remain in the bloodstream at satisfactory levels despite relatively prolonged periods of little or no food intake. In protein-energy malnutrition, less insulin is secreted and the effects of insulin may be reduced by the mechanism described next.

Redistribution of Body Fat

Although not clearly proven, it also appears that undernourished children adjust by redistributing body fat from the limbs to the trunk (Bailey et al., 1985). This is adaptive because the fat cells on the abdomen differ from those on the limbs in having a higher rate of lipolysis in response to epinephrine and lower insulin sensitivity (Kissebah et al., 1982). This implies that fat cells in this location are better able to release energy rapidly into the bloodstream in response to food deficits than cells located elsewhere. In pre-school children in Thailand, a higher ratio of trunk skinfold thickness to triceps skinfold thickness was associated with less stunting and less delay in bone age. Alterations in androgen metabolism may cause a greater predominance of fat to be located in the upper body (Evans et al., 1982).


For the child to maintain a reduced but steady level of growth without overt morbidity and under conditions of protein-energy deficiency requires a series of hormonal adaptations (Rag, 1974), which are as yet imperfectly understood. Protein deficiency leads to a reduction in the plasma amino acids necessary for growth. Several hormones interact to maintain an increased efficiency of amino acid utilization. Under these conditions, when protein intake is insufficient to permit normal growth, increasing the intake of calories alone may not improve the rate of growth but rather destabilizes this adaptive mechanism and leads to overt morbidity in the form of kwashiorkor. Ex perimental restriction of protein in a high-energy diet has been shown in infant Cebus monkeys and baboons to produce a condition resembling kwashiorkor (Samonds and Hegsted, 1978). A genetically slow linear growth rate could be adaptive by reducing nutrient requirement for growth.


This hormone, secreted by the adrenal cortex, withdraws protein from the muscles and returns it to the liver, where it can serve both to maintain the function of the liver and to make extra nutrients available to the body in response to stress. High cortisol levels in marasmus result in wasted muscles but a well-maintained liver function. In kwashiorkor, cortisol is lower, amino acids are not diverted from muscle to liver, and liver damage occurs (Rag, 1974). Cortisol levels therefore play a crucial role in the adaptation to protein-energy malnutrition. Another function of cortisol is to reduce the rate of skeletal growth by inhibiting the stimulatory action of somatomedin in the epiphyseal cartilage. Somatomedin is a serum growth factor that facilitates the anabolic actions of growth hormone. Cortisol also directly inhibits somatomedin synthesis (Smith et al., 1981).

Growth hormone (GH)

Secretion of this pituitary hormone increases as a response to high cortisol, low blood glucose and low blood amino acid levels (Rag, 1974; Brasel, 1980). Growth hormone levels are systematically increased in well-adapted mild to moderately malnourished children. Increased levels of GH also trigger the stimulatory action of somatomedin on epiphyseal cartilage synthesis. Thus, in the adapted child, the effects of increased core tisol and reduced insulin keeping nutrients in the liver and bloodstream ready for use - remain in balance with the effects of increased growth hormone, which channels some of these nutrients into maintaining growth, at a lower than normal rate.

In marasmus, when growth virtually stops, some researchers have found elevated, and some normal levels of GH. Rao (1982) found that GH levels are more likely to be normal in marasmic children and elevated in children with kwashiorkor, and that they can be normalized within a few days of therapy with milk feedings or protein containing diets. GH is also involved in lipolysis, releasing free fatty acids into the bloodstream and thus providing a source of fuel. This lipolysis becomes dysfunctional in kwashiorkor, when amino acid levels are low, and the released free fatty acids become trapped in the hepatic cells causing fatty infiltration of the liver.


The low levels in PEM may be due in part to the general inability of the liver to synthesize protein without sufficient free amino acids, as is illustrated by decreased albumin synthesis during malnutrition. High GH levels may suppress somatomedin synthesis as will high cortisol and low insulin levels. Soliman and co-workers (1986) found that impaired insulin secretion and elevated cortisol levels in malnourished children correlated with percent weight deficit; and that low somatomedin levels were associated with a greater height deficit. Thus it appears that somatomedin reduction is one more regulating mechanism for keeping linear growth proportional to the nutrient intake of the child.

However, Smith and associates (1981) found the same relationship between cortisol and weight deficit and a negative correlation of somatomedin levels with cortisol and growth hormone, without observing the association between decreased somatomedin and decreased stature. Thus, further research needs to focus on the effects of somatomedin on growth before conclusive statements can be made.

Immune Response

All of the immune responses are to some degree genetically regulated. Infants with high response levels are inevitably protected from the severe nutritional disturbances that follow infection. This topic and additional factors, such as malabsorption, require further detailed research.

Psychological Factors

It has long been known that psychological stress causes protein catabolism (Scrimstraw et al., 1966). A stressful caretaker-child interaction, therefore, can be expected to increase protein requirements while tending to decrease the amount of food that the child consumes (Scrimshaw, 1969). Pleasantly stimulating interactions, on the other hand, enhance the child's tendency to exercise its developing organ systems and, hence, to utilize nutrients for growth and development.

Stress and depression also have been shown to decrease primarily cellular but also humoral immune responses (Marx, 1985; see Calabrese et al., 1987, for a review of the role of stress in the neuroendocrine regulation of immunity). Increased cortisol levels in stress and in malnutrition suppress cell-mediated immunity, with the effect of protecting the stressed or malnourished individual from inflammatory responses that may further increase stress. The immune system increasingly appears not to be an independent body system but to be intimately linked to the nervous system, and to play an integral role in the body's overall system of hormonal regulatory responses. There is increasing evidence that psychoactive peptides, such as natural opiates, produced by the human brain in association with psychological mood, attract immune cells to parts of the body where they are needed to fight infection or repair damaged tissues (Ruff et al., 1985). Tumour-fighting immune responses have been shown to be lower in cancer patients who suppressed their anger, lacked good social support, and were apathetic. These last two characteristics tend to typify mothers of malnourished, chronically ill infants.

Implication of Physiological Factors for the Study of Psychosocial Aspects of Positive Deviance

If children who successfully resist malnutrition are genetically different from others, it would be reasonable to ask what was to be gained by studying the psychosocial factors linked to adequate growth amidst poverty. In fact, we have no evidence that the purely physiological components of positive deviance are strong enough to predominate over or to seriously confound the interpretation of the psychosocial components. The fact that some children are genetically more resistant than others adds to the unexplained variance in the results of psychosocial research, just as uncontrolled psychosocial face tors add to the "noise" in physiological research findings. It is desirable and it should be increasingly feasible to address both psychosocial and physiological questions in the same research designs.

On the other hand, it is extremely helpful to discover that psychological stress negatively affects the body's use of nutrients and its resistance to infection, and conversely that psychological well-being may activate the immune system and stimulate the secretion of growth hormone. These mechanisms help to explain why psychosocial factors, such as a smiling happy affect between mother and child, are consistently associated with adequate growth and development.

Overall conclusions and policy recommendations

Three very broad conclusions emerge from the review of the literature which follows. They concern psychological and social wholeness or good health; modernization; and behaviours, technologies, and social structures specifically adaptive for the nutrition and health protection of the infant. These conclusions and their policy implications are presented here in order to provide the reader with a framework for synthesizing and interpreting the subsections in the review of the behavioural, psychological, and social correlates of child growth. Specific programme-design recommendations are given in each subsection.

Psychological and Social Good Health

A major conclusion is that the psychosocial factors associated with adequate growth amidst poverty are not specific to nutrition alone. The same characteristics that predict a good nutritional outcome also predict good cognitive development, health, and longterm development of the individual into a stable, productive member of society.

Not surprisingly, most of these characteristics are common moral virtues and social values: devoted mothering, happy marriages, close families, supportive neighbours and friends, and active communities with employment opportunities and well developed social services. Healthy communities produce healthy families with healthy children.

These indicators of psychological adjustment appear to be necessary for good growth in deprived environments. yet once in place, they also are sufficient to promote favourable child development in almost every area. These characteristics that facilitate favourable child outcomes can be considered to be measures of the psychological and social health of families and communities.

The lack of specificity of these factors implies that action to improve infant nutrition by improving the quality of care and stimulation given to infants and young children, as well as the social services and social support networks available in low-income come munities, can be justified not only on nutritional grounds but much more broadly, in terms of the child's overall development and well-being.

Policy Implications

The policy implications of this conclusion are:

  1. Closer links are requited between health/nutrition services and social and psychological services. New combined service models are needed to protect high-risk mothers and children.
  2. Both health and social services should be increasingly decentralized and should be provided through community self-help programmes such as Alcoholics Anonymous or La Leche League Organization (with external technical and resource assistance). Decentralizing the management of services is desirable because the ability of a community to take responsibility for its own high-risk members appears to yield nutritional, social, and psychological benefits that are equal in value to the actual services provided.
  3. Legislation that binds fathers to support their children financially, and employers to provide maternity and health benefits, is badly needed in rapidly modernizing areas where customary legal and moral obligations have broken down. Social-welfare legislation designed to protect women and young children must be realistic and enforceable.
  4. Development support is needed for the small-scale and informal economic sectors. Community health in most developing regions depends on the productivity and self-esteem of individuals working primarily as small farmers and small-scale food processors, manufacturers, and service providers. Much of what we term "social support" consists of micro-scale material support in the form of help with tasks and child care and timely social interchanges of food and other necessities. Community based micro-interventions are needed to strengthen the social networks that support the nutrition and health of children.


Parents of positive deviants are more likely than others in their communities to be upwardly mobile, to discard fatalistic attitudes, to take the initiative in adopting modern practices for themselves and for their children, and to be more enterprising. They make a more effective use of health services, family planning, and educational facilities, and tend to be further advanced along the demographic transition concerning family size, as they bear fewer children, have higher aspirations for them, and invest more resources in each child.

Policy Implications

The diffusion of education, non-fatalistic enterprising attitudes, and modern practices protects the nutrition and health status of young children in spite of poverty. The shift in attitudes and values towards bearing fewer children, investing more in each and consciously planning for the future of each child yields direct benefits for the feeding, health care, and cognitive stimulation given to the individual child, particularly when resources are scarce. Since attitudinal problems may be at the root of malnutrition and ill health, explicit counseling, given during the post-natal check-up, for example, should discuss the parents, goals for the baby and the type of feeding, care, education, and birth spacing needed to achieve goals. Social marketing should also reinforce these new attitudes in health and nutrition as well as in family-planning programmes.

Behaviours, Technologies, and Social Structures Specifically Adaptive in Protecting Nutrition and Child Health

In addition to global measures of psychological and social wholeness and modernization, there are also positive-deviant behaviour patterns and technologies that are spe cifically adaptive in protecting the nutritional status and health of infants and young children. Actively feeding toddlers instead of expecting them to feed themselves is one such behaviour (Zeitlin and Johnson, in progress). Making sure that the infant was fed by adults, not by children, supplementing the infant's diet with a flour-based porridge, and using a metal hoe to remove faeces from the compound surface were similarly positive behaviours discovered in Bangladesh (Zeitlin, 1989). Very little research has looked specifically for these factors.

The ways in which psychologically competent mothers, receiving high levels of social support, feed and manage the lives of their infants and young children under conditions of resource scarcity should reveal these methods. Once identified they can be reinforced by programmes.

Policy Implications

Further research is needed to identify these behaviours, technologies and structures and then to incorporate them into programmes. The three-stage research model recommended in part 2 incorporates the search for positive-deviance factors into the development of culturally relevant child-survival programmes. This combined approach is considered essential to applied research in positive deviance, since nutritionally adaptive characteristics that are selected for programme reinforcement must be integrated into modernizing child-survival strategies.


Contents - Previous - Next