RELATIONSHIP BETWEEN SUBSISTENCE A N D AGE AT WEANING IN "PREINDUSTRIAL" SOCIETIES Daniel W. Sellen a n d Diana B. S m a y
Emory University
Cross-cultural studies have revealed broad quantitative associations between subsistence practice and demographic parameters for preindustrial populations. One explanation is that variation in the availability of suitable weaning foods influenced the frequency and duration of breastfeeding and thus the length of interbirth intervals and the probability of child survival (the "weaning food availability" hypothesis). We examine the available data on weaning age variation in preindustrial populations and report results of a cross-cultural test of the predictions that weaning occurred earlier in agricultural and pastoral populations because dairy and cereal production increased the availability of easily digestible, nutrientrich foods appropriate for weaning. We found that, contrary to predictions, supplementation with liquid foods other than breast milk was delayed in agricultural populations relative to less agriculturally dependent ones and complementary feeding with solid foods was delayed in pastoral populations relative to those less dependent on herding. Although the duration of breastfeeding was longer in populations dependent on hunting, there was no qualitative evidence that such populations lacked foods appropriate for weaning. The patterns observed suggest that the relationships between demography and subsistence observed among preindustrial societies cannot be explained by the "weaning food availability" hypothesis. We discuss the implications for understanding the mechanisms underlying prehistoric human demography, subsistence shifts, allocation to parenting and mating effort, and the evolutionary implications of tradeoffs between diet and disease. Received: December 6, 1999; accepted: February 11, 2000.
Address all correspondenceto Daniel W. Sellen, Department of Anthropology, Emory University, Atlanta, GA 30322. E-mail:
[email protected] Copyright 2001by Walter de Gruyter, Inc., New York Human Nature, Vol. 12, No. 1, pp. 47-87.
1045-6767/01/$1.00+.10 47
48
Human Nature, Vol. 12, No. 1, 2001 KEYWORDS: Agriculture; Breastfeeding; Childcare; Comparative method; Cross-cultural studies; Cultural evolution; Horticulture; Hunter-gatherers; Pastoralists; Weanling's dilemma; Women's work.
Cross-cultural studies have revealed broad quantitative associations between subsistence practice and demographic parameters for preindustrial populations (Bentley, Goldberg, and Jasienska 1993; Campbell and Wood 1988; Hewlett 1991; Sellen 1999). Fertility rates are on average higher in populations predominantly dependent on agriculture, and they increase with the degree of dependence on agriculture among populations with mixed subsistence (Bentley, Jasienska, and Goldberg 1993; SeUen and Mace 1997). Childhood mortality rates increase with the degree of foraging activity (Sellen and Mace 1999). Understanding the mechanisms underlying such associations is crucial for generating and testing behavioral models for the Neolithic and modern demographic transitions. It will also be useful for developing a general behavioral ecology of human weaning, and may have implications for the design of health interventions in mode m populations. A number of behavioral mechanisms can be hypothesized to link variation in subsistence mode across human populations to variation in infant mortality, birth spacing, and population fertility rates. However, testable predictions about the association between key measures of subsistence activity and the intervening factors influencing population demography are rarely developed (Sellen 1999). Our current understanding of the mechanisms underlying the patterns reported is poor for several reasons. First, the limitations of aggregate crosssectional data on past populations preclude the explicit linking of demographic outcomes to specific key behaviors that might impact individual fertility rates and child survival. Second, the mechanisms linking subsistence and other behaviors to demographic outcomes in preindustrial populations are rather weakly theorized owing to a tendency for many analysts to adopt group-level arguments. Third, there have been few attempts to draw together the sparse reports of infant feeding practices, maternal work patterns, and disease prevalence to test for their possible associations with demography. Teasing apart the complex inter-relationships between culture, demography, and resource base for a diversity of societies represents a considerable challenge, and the problems compound if we wish to consider past populations for whom records are of limited scope and quality. Nevertheless, such barriers to understanding may be overcome by careful construction of hypotheses and use of existing data. It is plausible that the observed variation in fertility and mortality rates is associated with variation in age at weaning. From a physiological and
Weaning and Subsistence
49
nutritional perspective, weaning is the process of transition of the infant from exclusive breasffeeding to exclusive reliance on foods other than breast milk. This transition is not usually abrupt but rather involves an intermediate phase of partial breasffeeding that varies enormously for individual infants in m o d e m populations (Hull and Simpson 1985; Maher 1992; Raphael and Davis 1985; Winikoff et al. 1988). The associated pattern of decline in the duration and frequency of suckling during this process influences the duration of maternal lactational amenorrhea and the timing of the resumption of ovulation, which, along with other factors such as maternal energy flux, contraceptive use, and sexual activity, influence birth intervals. Thus, any intervening factor that accelerates the process of weaning in preindustrial societies may also increase the fertility of mothers independently of the mortality of infants. It is commonly assumed in the paleodemographic literature that an increased availability of starchy weaning foods and animal milks in ancient populations that adopted pastoral and agricultural production triggered or facilitated early weaning, thus contributing to birth interval reduction, increased fertility, and possibly increased infant mortality (Cohen and Armelagos 1984; Flannery 1973; Larsen 1995). This notion of the deficiency of suitable weaning foods in forager diets has been supported by the !Kung model of human hunting and gathering (Lee 1979b). We refer to this interpretation of the association between preindustrial subsistence and population demography as the "weaning food availability" hypothesis. For comparative purposes milestones such as the introduction of various foods other than breast milk and the cessation of partial breastfeeding can mark the weaning process. (We prefer the term "cessation of breastfeeding" to the term "termination of breastfeeding" because it highlights the fact that we make no a priori assumptions about the agency of the mother and the weanling in determining the timing of this milestone.) Stated more generally, the weaning food availability hypothesis predicts that introduction of foods other than breast milk and cessation of breastfeeding will tend to occur at earlier ages in populations with greater access to regularly available, easily digestible, nutrient-rich foods appropriate for weaning with minimum health risk to the child. Implicit assumptions are that the quality of foods available to young children varies among societies engaged in different subsistence practices, and that parents and caretakers cue on such differences in making choices about feeding young children. One approach to testing this hypothesis is to conduct cross-cultural comparisons on the variables of interest. Given the rapid changes in the relevant parameters occurring in many so-called traditional populations, we must now rely almost entirely on ethnographic records and on the construction of summary measures for whole groups or populations to do this. Although this approach has many limitations, it does allow us to test
50
Human Nature, Vol. 12, No. 1, 2001
a variety of biocultural explanations for variation in fertility and mortality rates across human populations living in different ecological, social, and institutional settings. Thus, in this paper we examine the pattern of variation in markers of the weaning process across preindustrial societies described in the ethnographic record specifically in relation to their mode of subsistence. We ask whether the timing of introduction of foods other than breast milk and cessation of breasffeeding were associated with subsistence activity and whether there were any qualitative differences in the types of weaning foods used or styles of infant feeding among populations differing in their predominant food production activities.
METHODS
We set out to test for associations between dependence on various forms of subsistence and key markers (milestones) of the weaning process using the available data on weaning in preindustrial, natural fertility populations. Our strategy was to collate a large set of estimates of age at weaning, variously defined, for a broad range of so-called preindustrial societies. For the purposes of this analysis, a preindustrial society was operationally defined as one that remained somewhat "marginal" to the emerging industrial capitalist world system at the time of ethnographic or ethnohistorical description. The most reliable indicators of such marginality are a lack of evidence that a modern demographic "transition" (typically an increase in child survival and/or reduction in fertility) and a shift to a predominantly cash-based economy had taken place. Specifically, we aimed to test the following predictions, three quantitative and one qualitative, derived form the general "weaning food availability" hypothesis: Ha: The range and central tendency of age of introduction of supplementary liquid foods tend to be lower in those preindustrial populations with access to the products of agricultural and pastoral production, and higher in those with greater dependence on extracted resources. Fib: The range and central tendency of age of introduction of complementary solid foods tend to be lower in those preindustrial populations with access to the products of agricultural and pastoral production, and higher in those with greater dependence on extracted resources. He: The range and central tendency of age of cessation of breasffeeding tend to be lower in those preindustrial populations with access to the products of agricultural and pastoral production, and higher in those with greater dependence on extracted resources. Ha: The types of foods given to young children during the weaning process were qualitatively different among preindustrial populations who
Weaningand Subsistence
51
were predominantly foragers, pastoralists, or agriculturalists, such that agriculturalists and pastoralists had greater access than foragers to foods appropriate for weaning. To construct a database of weaning indicators we searched the Human Relations Area Files (HRAF) for culture records that included information on weaning practices, as indexed under category 862 ("weaning and food training"), and carried out an additional retrospective "snowball" search of the anthropological literature. We read all of the sources available in the microfiche version of the HRAF and were also able to obtain additional published descriptions. We coded the information into a series of summary variables for use in statistical analysis. These variables included estimates of the timing of key milestones in the weaning process, such as age at first introduction of liquids and solids, age of termination of breastfeeding, and descriptions of typical weaning foods used. There was great variation in the descriptors chosen by various authors, with most reporting a range of ages at which typical weaning events or transitions occurred. For example, statistical estimates of the mean, median and mode of breastfeeding duration were only available for 11, 5, and 3 populations, respectively. We therefore used the midpoint of the reported ranges as a reliable and comparable estimate of the average timing of milestones in the weaning transition for each population. We also compiled lists of foods mentioned as used in weaning in all the ethnographic accounts available. Descriptive data for 133 distinguishable cultural groups, as defined by the original ethnographers, were extracted from 172 sources. We controlled for pseudo-replication of those groups most heavily studied by applying a rigorous re-coding strategy. For some cultural groups, different researchers were found to have collected several estimates of weaning age and practices at roughly the same time on the same populations. If different investigators reported weaning information for the same culture group within 10 years of each other, the estimates and descriptions from the various sources were aggregated or averaged for analysis as a single observation. The midpoint of the total reported range of ages at introduction of non-breast milk liquids and solids and cessation of breastfeeding was calculated as the mean of the midpoint values of the original references. If no midpoint value could be calculated for one or all reports on a single population, the reported mean age at termination of breastfeeding was used to calculate a pooled mean. In several cases,/rfformation on weaning was presented in different formats by the different primary sources; for example, one source would report a range of age at termination of breastfeeding while another would cite a single age value. In these cases age at cessation of breastfeeding as well as maximum earliest and latest ages in all published sources were included in the range in the combined report.
52
Human Nature, Vol. 12, No. 1, 2001
For other cultural groups, data were found pertaining to several populations living in different places or at different times more than 10 years apart, and the data on each distinct population were included as independent cases in the final sample. When available, fieldwork dates were used to make this determination. If no date of fieldwork could be located, the dates of publication were used. Thus, each case in the final sample represented a set of measures taken from populations distinguishable in space, time, and ethnic identity as defined by western-trained observers. After excluding studies on populations engaged in any degree of industrial production, reliable reports of weaning practices were gathered for a total of 113 populations described in the HRAF, and an additional 16 were collected from other anthropological sources. This yielded a final sample of 129 populations from 110 ethnographically distinguished cultural groups for which some information regarding weaning practices was available (Table 1). We collated the weaning data from the HRAF and other sources with estimates of the relative dependence on five basic economies of subsistence available for 1,267 societies coded in Gray's recent revision of Murdock's Ethnographic Atlas (EA) (Murdock 1967, 1981; Gray 1999). Specifically, variables 1-5 in the EA provided coded range estimates of the contribution to the diet and economy of gathering of wild plants and small land fauna; terrestrial hunting, including trapping and fowling; fishing, including shell-fishing and the pursuit of large aquatic animals; animal husbandry; and agriculture. We re-coded the categorical data in the EA into continuous variables as the midpoints of the coded ranges and thus estimated the percentage importance of hunting, gathering, fishing, herding, and agriculture for the whole diet and economy for 98 populations for which some weaning data were available. Matching of populations from the two data sets was achieved by paying special attention to geographic location and ethnic names or synonyms. This allowed us to quantify the multiple dependence of cultural groups on several axes of subsistence and to test for associations with variables that indicate weaning patterns using multiple regression.
RESULTS
Most previous discussions of the relationship between demography and subsistence use a categorical approach to teasing apart populations with different subsistence activities. Such categorizations are of broad conceptual use and may be useful for identifying threshold effects. We therefore first categorized the populations in the present sample according to the predominant mode of subsistence practiced at the time the weaning data were collected. A population was designated as having a "predominantly
Weaning and Subsistence
53
~D
e~
.~.~ "~o~
e~ -~
~
. ~
~ao~-~.~.~
~e
" ~ ' ~ =O=0. ~ ' ~
"O~0= ' ~0=
-~:~;~.~ ~ ~
~
o ~ ~ ~ ~ ~
~ ~
~
g -.~
oam ~
~D
0 0
~N~u:zm~
go~zX
0
e~
} D 0
