AMERICAN JOURNAL OF PHYSICAL ANTHROPOLOGY 119:276 –282 (2002) Birth Weight and Environmental Heat Load: A Between-Population Analysis Jonathan C.K. Wells1* and Tim J. Cole2 1 MRC Childhood Nutrition Research Centre, Institute of Child Health, London WC1N 1EH, UK Centre for Paediatric Epidemiology and Biostatistics, Institute of Child Health, London WC1N 1EH, UK 2 KEY WORDS temperature; thermoregulation; fetal growth ABSTRACT Low birth weight, a major cause of infant morbidity and mortality, is caused by different factors in Western and developing-country populations. In addition to differing in terms of ethnicity, maternal size, maternal nutritional status, and disease load, developing-country and Western populations are also characterized by different environmental heat loads. Thermodynamic theory predicts that heat stress is mitigated by reduced size of both mother and offspring, and therefore generates the hypothesis that reduced birth weight may be an adaptation to environmental heat load. The aim of this study was to test the hypothesis that environmental heat load is associated with between-population variability in birth weight. Data on birth weight and thermal environment were obtained from the literature for 140 populations. Low birth weight is a major cause of infant morbidity and mortality, both in developed and developing countries, and is a major contributing factor to the approximately 9 million infant deaths that occur worldwide annually (World Health Organization (WHO), 1992). The incidence of low birth weight, classified by the WHO (1992) as weight less than 2,500 g, is the product of incidences of preterm birth and intrauterine growth retardation. The proportion of low birth weights attributable to these factors differs markedly between the developing and developed worlds, with most low birth weights in developing countries arising from intrauterine growth retardation. A number of factors are associated with low birth weight in developing countries, including ethnicity, low energy intake, low pre-pregnancy weight, short stature, malaria, and general morbidity (Kramer, 1987; WHO, 1992). This epidemiology differs in several ways from that of developed countries, where preterm birth represents a significant proportion of low birth weights, and cigarette smoking is an important additional risk factor. The incidence of low birth weight is highest in Asia, where it exceeds 20% of all births, but it is also high in Africa (15%) and Latin America (10%) (WHO, 1992). In southern Asia, approximately 1 in 3 liveborn infants is underweight (WHO, 1992), whereas in Southeast Asia and Africa, the rate of low birth weights is closer to 1 in © 2002 WILEY-LISS, INC. Further data on several physical, social, and biological confounders were also collated. After adjusting for confounding factors, of which only maternal height and per capita gross domestic product were statistically significant, heat stress showed a significant inverse relationship with birth weight in 108 populations, accounting for an additional 9.6% of the between-population variance. Though based on data collected from more than one source, these results are consistent with the hypothesis that heat stress is inversely associated with birth weight, as previously reported for within-population studies. Further studies are needed to establish to what extent heat stress is a determinant of low birth weight in developing countries. Am J Phys Anthropol 119:276 –282, 2002. © 2002 Wiley-Liss, Inc. 6. It has proved difficult to explain this regional disparity purely in terms of the factors listed above. Following earlier observations of an association between climate and birth weight (Roberts, 1978), thermal stress has been proposed as an additional potential cause of intrauterine growth retardation in developing countries (Wells, 2002). Heat stress is assumed to have played an important role in determining size and shape throughout hominid evolution (Ruff, 1994), due to the fundamental relationship between mammalian physiology and the thermal environment. Thermal stress is the product of internal heat production, capacity for heat loss to the environment, and environmental heat load. Core body temperature must be regulated within narrow limits to avoid significant morbidity or mortality, and various aspects of human morphology and physiology are attributable to thermoregulatory requirements. *Correspondence to: Jonathan C.K. Wells, MRC Childhood Nutrition Research Centre, Institute of Child Health, 30 Guilford St., London WC1N 1EH, UK. E-mail: J.Wells@ich.ucl.ac.uk Received 30 August 2001; accepted 22 February 2002. DOI 10.1002/ajpa.10137 Published online in Wiley InterScience (www.interscience.wiley. com). BIRTH WEIGHT AND ENVIRONMENTAL HEAT LOAD Classical rules describing a relationship between body size and climate were presented in the 19th century (Bergman, 1847; Allen, 1877), and inspired subsequent interest in the role of heat stress in natural selection of body size and physique (Roberts, 1953, 1978; Newman, 1953; Schreider, 1964; Crognier, 1981; Froment and Hiernaux, 1984). Most authors focused on adults, due to the large quantity of data available, and reported that heat stress was positively related to the ratio of surface area to mass, or proxy indices of this ratio. For example, humans conform both to the rule of Bergman (1847) in showing a negative association between body mass and mean annual temperature, and also to the rule of Allen (1877) in having relatively shorter limb lengths in colder climates (Roberts, 1953). Recently, such relationships were reanalyzed in order to assess the effect of secular trends in body size. The slopes of regressions between indices of size and temperature are reported to be more modest in contemporary populations (Katzmarzyk and Leonard, 1998) compared to those studied a half century previously (Roberts, 1953). However, this finding does not negate the role of climate in selecting physique. Ruff (1994) reported that body breadth varies much less than stature, such that the effect of increasing stature is to increase linearity. Thus, although changes in body weight may alter the surface area to mass ratio, adaptive physique may remain preserved in other dimensions that are less sensitive to nutritional influences. Other factors which may confound the temperature/size relationship include altitude (Stinson, 1990), while the strength of associations may also differ between the sexes (Roberts, 1978; Stinson, 1990). In addition to basic physical aspects of size, several other aspects of human physiology contribute to the mitigation of heat stress. These include differential sweat gland distribution (Millington and Wilkinson, 1983), relative blood flow (Hanna and Brown, 1983), and variation in the level of heat production at rest (Roberts, 1978; Hayter and Henry, 1994). Behavior may also influence both heat stress and heat tolerance. Thus the human body may be considered a thermodynamic system, with populations demonstrating adaptation of both physical and physiological properties of the system in relation to the thermal environment. The majority of work on human form and heat stress has been conducted on adults, with infants and children discussed only rarely (Roberts, 1978; Hanna and Brown, 1983). Pregnancy represents a particularly interesting aspect of this general relationship, due to the fact that mother and offspring are characterized by genetic conflict (Trivers, 1974). Both mother and offspring are subject to heat stress, and generate heat in relation to their size and body composition. “Mitigation of heat stress” may be considered a resource subject to competition, similar to the conflict over nutritional resources (Haig, 1993) and indeed influencing its outcome. 277 In hot environments, the physiological changes that accompany pregnancy are predicted to exacerbate maternal heat stress for several reasons. First, fat deposition increases the specific heat of the body, such that the increase in core body temperature due to a given thermal stress is higher in women with greater adiposity (Falk, 1998). Second, the change in body weight and shape decreases the ratio of surface area to body mass, thereby reducing the ability of the body to lose heat to the environment by sweating (Wells, 2002). Third, the combination of weight gain and fetal growth increases basal heat production, although less so in populations inhabiting the tropics than those in temperate climates (Prentice et al., 1989). Fourth, the interaction between increased body mass and physical activity may further exacerbate heat production (Wells, 2002). Finally, the fetus itself also contributes to maternal heat stress through its body composition and metabolic rate (Wells, 2002). For these theoretical reasons, heat stress has been proposed as a possible factor associated with intrauterine growth retardation in hot environments (Wells, 2002). There are two ways in which heat stress might affect birth weight. Chronic heat stress during pregnancy might exert marked effects on fetal growth, and might be a major determinant of intrauterine growth retardation in individuals. Alternatively, heat stress might exert small effects on individual infants, but nevertheless cause a slight shift downwards in the spectrum of birth weight, thus increasing the prevalence of intrauterine growth retardation in the population as a whole. The aim of the present paper is to consider whether the hypothesis that environmental heat load and birth weight are associated is supported by data from a wide range of human populations. METHODS Data on birth weight were obtained from the WHO global database (WHO, 1992). This database summarizes data from the medical literature on birth weight, the prevalence of intrauterine growth retardation and preterm birth, sample size, and year of study. The heterogeneous nature of the literature means that not all data are equivalent. For example, some reports refer to total births, whether live or stillborn, or single or multiple births. Other reports variously list all single births whether live or stillborn, all live births whether single or multiple, or all live singleton births. For a significant proportion of the data, no such information is available. For the present study, the database was initially constructed by selecting the optimum data for each population. The first criterion was that populationspecific data on birth weight and thermal environment must be available. Secondly, the population must comprise at least 200 individuals. Thirdly, data were selected in the order of preference of 1) live singleton births, 2) all live births, 3) single births whether live or stillborn, 4) no data available, 278 J.C.K. WELLS AND T.J. COLE and 5) total births. Where more than one data set was available for a given population, preference was given firstly to larger sample sizes and secondly to more recent reports. Data on thermal environments were obtained from the Hutchinson World Weather Guide (Pearce and Smith, 1998), which lists mean temperature and humidity for various locations in the world’s countries. Heat stress is a combination of environmental temperature and the water content of the air, which influences the ability to lose heat by sweating. Humidity is extremely important in estimating heat stress: at higher temperatures, when the environment is hotter than the body, alternative paths of heat loss such as convection and radiation cease to operate, and sweating is the only possible means of losing heat (Dubois, 1937). Many environments are characterized by seasonality, with heat stress peaking during hot and humid seasons and decreasing during cooler, drier periods. For between-population comparisons, it is therefore necessary to derive an index representing average annual thermal load. Two approaches were considered. Firstly, monthly values for heat stress were calculated using the data on average maximum daily temperature and afternoon humidity. Temperatures were corrected for humidity, using the heat index published by the US Meteorological Bureau (Schneider, 1996). These monthly values were then averaged to give a mean humidity-adjusted temperature per year. The second approach involved a simpler calculation (Pearce and Smith, 1998). For each location, the weather guide provides a monthly rating of discomfort arising from the combined stress of temperature and humidity. The categories were none, moderate, medium, high, extreme, and dangerous. A schematic diagram, showing how data on temperature and humidity were assigned their scale value, is given in Figure 1. The categories were given a numeric score of 0 –5, and a mean monthly score was calculated. To compare the two indices of thermal stress, 10 locations selected randomly from the guide were assessed by both approaches. The two indices were very highly correlated (r ⫽ 0.97), indicating that the simpler method is a valid index of heat load. This simpler method was therefore used in the present study, and was termed “heat index.” Data on latitude and altitude were likewise obtained from the same weather guide (Pearce and Smith, 1998). All latitude data were expressed as positive values, and thus represented distance from the equator, irrespective of hemisphere. Data on maternal height were taken from the tables published by James and Schofield (1985). Data on socioeconomic status (per capita gross domestic product), average daily energy intake, and infant mortality were obtained from the World Bank (1990). This year was selected, as the birth weight data were collected prior to 1992. Data on gross Fig. 1. Schematic diagram indicating how monthly data on temperature and humidity were converted into a simple index of thermal stress. Monthly values were then combined to provide a mean annual index, termed “heat index.” domestic product represented a 3-year average, designed to smooth out fluctuations in prices and exchange rates in each country. It should be borne in mind that the calculated values, in US dollars, do not represent local purchasing power. However, published data on purchasing power parity were not available. Per capita dietary energy intakes were calculated from data on food supplies from the Food and Agriculture Organization, divided by the size of the population. Data on infant mortality, calculated per 1,000 liveborn births up to age 1 year, were used as a general index of environmental disease load. All data for confounding factors necessarily refer to the total population, either national or regional, and in the present analysis it is assumed that they were representative of the sample of mothers providing the birth weight data. All data were transformed where necessary to achieve normality. Birth weight, gross domestic product, and altitude were natural log-transformed, latitude and mortality were transformed to square roots, and heat index and maternal height were unadjusted. In addition, the log-transformed variables were multiplied by 100 to give regression coefficients in percentage units (Cole, 2000). RESULTS Comparable data on birth weight and temperature were available for a total of 140 populations. The median sample size was 5,558 infants, with only 20 samples comprising ⬍1,000 births. However, due to the effects of a small number of very large samples, the average sample size was 97,237. The numbers of populations per birth weight data category BIRTH WEIGHT AND ENVIRONMENTAL HEAT LOAD TABLE 1. Mean birth weight (g) by global region Region No. countries Mean SD Africa Central and South America Western industrialized1 Asia 40 25 35 30 3,076 3,122 3,415 3,024 167 144 87 292 1 Europe, North America, Australia, and New Zealand. TABLE 2. Description of independent variables for 108 populations used in final model Variable Mean Standard deviation Altitude (m) Latitude (degrees)1 Mortality index2 Energy intake (kcal/day) Gross domestic product Maternal height (cm) Heat index (units)3 327 25.1 56.8 2,720 5,025 158.8 1.51 633 16.8 44.2 575 6,850 3.7 1.18 1 Degrees north or south from equator. Deaths per 1,000 live births to 1 year of age. 3 See text for calculation of index. 2 Fig. 2. Relationship between average annual heat index and average birth weight in 140 human populations, using birth weight data from WHO (1992) and heat index values calculated from data in Pearce and Smith (1998). were as follows: total births, 14; no description available, 35; single births, 13; live births, 49; and live single births, 29. Mean birth weight by global region is given in Table 1. Birth weight in Asian and African populations was slightly lower than that in Central and South American populations, but this difference did not reach significance. All three non-Western regions had highly significantly lower mean birth weight than the Western group of populations (P ⬍ 0.0001). A further 10 populations were excluded from Table 1 due to their not being readily included in one of these four groups. Descriptive data for the independent variables are presented in Table 2. The relationship between heat index and population mean birth weight for 140 populations is shown in Figure 2. There was a significant correlation of ⫺0.59 (p ⬍ 0.001) between variables. Correlations between all variables considered in the analysis are given in Table 3. 279 With one exception (altitude and birth weight), every variable was significantly correlated with every other variable, indicating the complexity of associations between geographic factors, size, socioeconomic factors, and indices of health. Heat index was negatively related to maternal size, latitude, gross domestic product, and energy intake, and positively related to infant mortality. There was a strong inverse correlation between energy intake and infant mortality. Stepwise regression analysis was conducted to obtain the best model for predicting birth weight, using the 108 populations with available data as appropriate. The initial regression model included gross domestic product, energy intake, infant mortality, and maternal height. Of these, only gross domestic product and maternal height were significant. Latitude and altitude were then added, with only latitude being significant. Finally, heat index was added, at which point latitude became insignificant. Heat index was more significant than the other two predictors, and its addition to the model increased R2, the proportion of variance explained, from 46.1% to 55.7%. The initial, intermediate, and final regression models for the 108 populations with appropriate data are given in Table 4. The regression statistics indicate that a 1-cm increase in maternal height and a 1-unit decrease in heat index are associated with 0.6% and 2.7% increases, respectively, in birth weight. The same analysis was repeated with the sample restricted to the 65 populations in which birth weight data referred to live births only. The findings were similar to those for the whole sample, except that gross domestic product was no longer significant. The addition of heat index to the model increased R2 from 36.9% to 50.2% (see Table 5). DISCUSSION Mean birth weight is highly variable between human populations, despite the strong relationship between birth weight and subsequent health. The range in the present study was from 2,429 g in Bangladesh to 3,632 g in South Korea. There is little evidence for genetic factors influencing birth weight (Clausson et al., 2000), although within a single population, small but significant differences between ethnic groups were found after adjusting for maternal anthropometric and socioeconomic variables (Shiono et al., 1986). Environmental factors are thought to be responsible for the vast majority of between-population variabilities in birth weight (Kramer, 1987; WHO, 1992), and the purpose of the present study was to investigate whether environmental heat load, previously not considered in the context of health, might be a further factor. In the present study, data were gathered from several different sources. The tabulation of global data on birth weight by WHO (1992) represents an 280 J.C.K. WELLS AND T.J. COLE TABLE 3. Correlation coefficients for relationships between variables included in analysis1 Heat index Latitude Altitude GDP Energy intake Infant mortality Maternal height 1 Birth weight Heat index Latitude Altitude GDP Energy intake Infant mortality ⫺0.59 0.58 ⫺0.14 0.53 0.60 ⫺0.54 0.58 ⫺0.79 ⫺0.28 ⫺0.57 ⫺0.68 0.58 ⫺0.48 ⫺0.17 0.70 0.78 ⫺0.65 0.57 ⫺0.22 ⫺0.24 0.22 ⫺0.33 0.70 ⫺0.69 0.68 ⫺0.77 0.60 ⫺0.50 All correlations above 0.16 significant, p ⬍ 0.05. TABLE 4. Stepwise regression of birth weight (%) in 108 populations Predictor Regression coefficient Standard error t-ratio p-value 684 0.02 1.2 * 10⫺3 4.8 * 10⫺3 0.65 32 0.01 0.00 0.40 0.21 21.5 2.0 0.6 0.0 3.1 ⬍0.0001 0.05 0.55 0.99 0.0029 688 0.01 0.63 0.80 0.42 28 0.01 0.19 0.39 0.31 23.9 3.0 3.3 2.0 1.3 ⬍0.0001 0.0033 0.0015 0.0441 0.1785 713 9.17 * 10⫺3 0.56 ⫺2.71 27 0.00 0.18 0.56 26.4 2.0 3.1 ⫺4.9 ⬍0.0001 0.05 0.002 ⬍0.0001 Initial Model Constant GDP (%) Energy intake Infant mortality Maternal height R2 ⫽ 47.0% Intermediate model Constant GDP (%) Maternal height Latitude Altitude R2 ⫽ 48.0% Final model Constant GDP (%) Maternal height (cm) Heat index R2 ⫽ 55.7%. GDP, gross domestic product per capita. TABLE 5. Regression of birth weight (%) in 65 populations where birth weight data refer to live births only Predictor Regression coefficient Standard error t-ratio p-value Constant Maternal height (cm) Heat index R2 ⫽ 50.2% 691.0 0.74 ⫺2.96 33.0 0.21 0.70 20.8 3.6 ⫺4.2 ⬍0.0001 0.0006 ⬍0.0001 important database for research, but fully comparable data for the world’s nations have never been published. The data utilized were therefore heterogeneous, referring variously to all births, live births, single births, or single live births. Similarly, birth weight data could only be matched against other data relevant to their general population, rather than the sample of mothers giving birth. For some smaller populations, data on maternal height, gross domestic product, energy intake, and infant mortality were unavailable. Despite these shortcomings, a relatively large sample of data points was available for analysis, and the results were similar whether all birth weight data were included, or just those referring to live births. Data on both birth weight and heat stress were reduced to an annual average value, and thus ignored seasonal variation in climate. If births are evenly spaced throughout the year, the use of such averaged data will introduce minimal bias into the model. However, if birth frequency follows a seasonal pattern, then the averaged values will be biased. For example, data from Gambia suggest that births are seasonally regulated, with fewer conceptions occurring in the hungry season, and hence fewer births occur in the harvest season (Prentice and Cole, 1994). Similar birth rate seasonality was reported in Lesotho (Johnston, 1993). In practice, however, the bias due to seasonality of birth frequency is likely to be small. Maternal size, pre-pregnancy nutritional status, and pregnancy weight gain are well-established determinants of birth weight. Energy intake during pregnancy might be assumed to be particularly important, and one possible explanation as to why energy intake was not significant in our model is that our intake data are simply too approximate, being estimated for all adults rather than specifi- BIRTH WEIGHT AND ENVIRONMENTAL HEAT LOAD cally for mothers during pregnancy. However, average per capita intakes are still likely to rank populations with reasonable accuracy, and our results are in agreement with previous research reporting that maternal intake during pregnancy had little impact on birth weight except in famine conditions (Susser, 1991). Likewise, intervention studies have failed to increase birth weight substantially through nutritional supplementation during the third trimester (Ceesay et al., 1997). Maternal height exerted a strong effect on birth weight, as is wellestablished (Kramer, 1987), but data on maternal nutritional status and pregnancy weight gain could not be incorporated in the present study. Maternal height alone accounted for a third of the betweenpopulation variance in birth weight in our analysis, but is itself inversely related to heat stress (r ⫽ ⫺0.48; p ⬍ 0.001). The contribution of gross domestic product to the model was significant, but that of infant mortality index, used as a proxy for maternal disease load during pregnancy, was not. Together, gross domestic product and maternal height explained 46.1% of the between-population variance in birth weight. The addition of latitude increased this only to 47.6%, and it no longer contributed significantly to the model once heat index was added. After taking into account gross domestic product and maternal height, heat index explained an additional 9.6% of the between-population variance in mean birth weight, with the effect highly significant. Almost half the between-population variance in birth weight was not explained by the predictors in our model. Several factors closely associated with low birth weight in developing countries could not be included in the analysis, including ethnic group, maternal nutritional status, pregnancy weight gain, and malaria status. The heat index used in the analysis might act as a proxy for one or more of these variables. However, each of these missing factors is predicted to be influenced by heat stress. For example, theoretical models predict that greater heat load should favor lower body fatness and reduced pregnancy weight gain (Wells, 2002). Malaria may increase heat production through fever. Ethnic differences in body size have also been linked to environmental temperature (Roberts, 1953; Newman, 1953; Schreider, 1964; Hiernaux, 1974; Crognier, 1981; Froment and Hiernaux, 1984), although both sexual selection and adaptation to marginal resources are plausible alternative explanations for such variation in size. Our analysis indicates that both maternal size and, independently of that factor, infant birth weight are inversely associated with thermal stress. CONCLUSIONS Although necessarily preliminary in nature, our analysis fails to refute the hypothesis that heat stress is a significant determinant of population 281 variability in birth weight. Consistent with theoretical models, increased heat stress is associated with reduced birth weight. The issue now requires further investigation using more comprehensive data sets, in particular to determine if the relationship is causal. If supported, the hypothesis could help explain the paradox of particularly low birth rates in the humid environments of Asia. LITERATURE CITED Allen JA. 1877. The influence of physical conditions in the genesis of species. Radical Rev 1:108 –140. Bergman C. 1847. Ueber die Verhaltnisse der Wärmeökonomie des Thieres zu ihrer Grösse. Gottinger Stud 3:595–708. Ceesay SM, Prentice AM, Cole TJ, Foord F, Weaver LT, Poskitt EME, Whitehead RG. 1997. Effects on birth weight and perinatal mortality of maternal dietary supplements in rural Gambia: 5 year randomised controlled trial. Br Med J [Clin Res] 315:786 –790. Clausson B, Lichtenstein P, Cnattingius S. 2000. Genetic influence on birthweight and gestational length determined by studies in offspring of twins. Br J Obstet Gynaecol 107:375– 381. Cole TJ. 2000. Sympercents: symmetric percentage differences on the 100 loge scale simplify the presentation of log transformed data. Stat Med 19:3109 –3125. Crognier E. 1981. Climate and anthropometric variations in Europe and the Mediterranean area. Ann Hum Biol 8:99 –107. Dubois EF. 1937. The mechanism of heat loss and temperature regulation. Stanford, CA: Stanford University Press. Falk B. 1998. Effects of thermal stress during rest and exercise in the paediatric population. Sports Med 25:221–240. Froment A, Hiernaux J. 1984. Climate-associated anthropometric variation between populations of the Niger Bend. Ann Hum Biol 11:189 –200. Haig D. 1993. Genetic conflicts in human pregnancy. Q Rev Biol 68:495–532. Hanna JM, Brown DE. 1983. Human heat tolerance: an anthropological perspective. Annu Rev Anthropol 12:259 –284. Hayter JE, Henry CJK. 1994. A re-examination of basal metabolic rate predictive equations: the importance of geographic origin of subjects in sample selection. Eur J Clin Nutr 48:702– 707. Hiernaux J. 1974. The people of Africa. London: Weidenfield & Nicholson. James WPT, Schofield EC. 1990. Human energy requirements. Oxford: Oxford University Press. Johnston FE. 1993. Seasonality and human biology. In: Ulijaszek SJ, Strickland SS, editors. Seasonality and human ecology. Cambridge: Cambridge University Press. p 5–16. Katzmarzyk PT, Leonard WR. 1998. Climatic influences on human body size and proportions: ecological adaptations and secular trends. Am J Phys Anthropol 106:483–503. Kramer MS. 1987. Determinants of low birth weight: methodological assessment and meta-analysis. Bull WHO 65:663–737. Millington PF, Wilkinson R. 1983. Skin. Cambridge: Cambridge University Press. Newman MT. 1953. The application of ecological rules to racial anthropology of the aboriginal new world. Am Anthropol 55: 311–327. Pearce EA, Smith CG. 1998. The Hutchinson world weather guide. Oxford: Helicon. Prentice AM, Cole TJ. 1994. Seasonal changes in growth and energy status in the Third World. Proc Nutr Soc 53:509 –519. Prentice AM, Goldberg GR, Davies HL, Murgatroyd PR, Scott W. 1989. Energy-sparing adaptations in human pregnancy assessed by whole-body calorimetry. Br J Nutr 62:5–22. Roberts DF. 1953. Body weight, race and climate. Am J Phys Anthropol 11:533–558. Roberts DF. 1978. Climate and human variability. 2nd ed. Menlo Park, CA: Cummings. 282 J.C.K. WELLS AND T.J. COLE Ruff CB. 1994. Morphological adaptation to climate in modern and fossil homonids. Yrbk Phys Anthropol 37:65–107. Schneider SH, editor. 1996. Encyclopedia of climate and weather. New York: Oxford University Press. Schreider E. 1964. Ecological rules, body-heat regulation, and human evolution. Evolution 18:1–9. Shiono PH, Klebanoff MA, Graubard BI, Berendes HW, Rhoads GG. 1986. Birth weight among women of different ethnic groups. JAMA 255:48 –52. Stinson S. 1990. Variation in body size and shape among South American Indians. Am J Hum Biol 2:37–51. Susser M. 1991. Maternal weight gain, infant birth weight, and diet: causal sequences. Am J Clin Nutr 53:1384 –1396. Trivers RL. 1974. Parent-offspring conflict. Am Zool 14:249 –264. Wells JCK. 2002. Thermal environment and human birth weight. J Theor Biol 214:413– 425. World Bank/International Bank for Reconstruction and Development. 1990. World development report. Oxford: Oxford University Press. World Health Organization. 1992. Low birth weight. A tabulation of available information. Geneva: World Health Organization.