Correlates of breastfeeding frequency among nomadic pastoralists of Turkana Kenya A retrospective study.код для вставкиСкачать
AMERICAN JOURNAL OF PHYSICAL ANTHROPOLOGY 98:239-255 (1995) Correlates of Breastfeeding Frequency Among Nomadic Pastoralists of Turkana, Kenya: A Retrospective Study SANDRA J. GRAY Department of Anthropology, University of Kansas, Lawrence, Kansas 66045 KEY WORDS Reproductive ecology, Breastfeeding patterns, Breastfeeding structure, Lactational amenorrhea, Infant growth ABSTRACT Recent research has shown that significant variation in breast-feeding behavior exists among natural fertility populations, all of whom have been characterized as practicing “on-demand breastfeeding. A number of recent prospective studies have contributed to a better understanding of breastfeeding structure and of its consequences for population differences in fertility. Currently, there is a growing interest in quantifying the complex environmental and biocultural interactions which influence that structure; in other words, in establishing an ecology of breastfeeding. In this paper, a carefully structured retrospective study of breastfeeding behavior among nomadic Turkana is used to identify environmental, biobehavioral, and socioeconomicfactors associated with variation in breastfeeding frequency among Turkana women. In agreement with the results of a prospective study conducted as part of the same research, the age (growth) and physical development of nurslings show significant correlations with breastfeeding frequency. Maternal physical status, the depth of the maternal social network, and, t o a lesser degree, rainfall patterns are also significant. All of these factors appear to influence breastfeeding through their effects on maternal participation in herding activities and related absences from camp. Finally, the study also presents new strategies for collecting and utilizing retrospective data, which are notoriously unreliable and difficult to classify accordingto operational definitions recently developed for prospective studies. Results of the present study suggest methods by which the quality and reliability of recall data may be enhanced. o 1995 Wiley-Liss, Inc. In 1989, a study of reproductive biology and behavior of Ngisonyoka pastoralists of South Turkana, northwest Kenya, was initiated to determine causes of the extreme seasonality of births observed in that population (Leslie and Fry, 1989). Among factors hypothesized to affect the timing of births and conceptions were seasonal variation in the nutritional status of Turkana women, in the timing of initiation of weaning and resumption of menses postpartum, and in infant mortality. Since breastfeeding and weaning practices may have a significant effect on any or all of these parameters, a study of 0 1995 WILEY-LISS. INC. infant care and feeding practices among Ngisonyoka was undertaken between May 1989 and late February 1990 (Gray, 1992). In addition to quantifying effects of lactation on birth seasonality, an important objective of the infant feeding study was to develop an ecology of breastfeeding among nomadic Turkana. Our objective was, on the one hand, to quantify the duration of lacta- Received November 12, 1993; accepted May 18, 1995. Address reprint requests to Sandra J. Gray, Dept. ofhthropology, 622 Fraser Hall, KU, Lawrence, K S 66045. 240 S.J. GRAY tion and the structure of breastfeeding, since those two parameters were assumed to be directly linked to fertility patterns. On the other hand, we wished to identify the environmental, cultural, and biological interactions influencing lactation and breastfeeding in this population. I n this paper, the focus is on the second of those two objectives. Specifically, the study examines the effects of selected environmental and biobehavioral factors on breastfeeding structure among nomadic Turkana. In order to contextualize these effects, brief discussions of the model used to design the research and of breastfeeding structure in this population are also included. STUDY AREA AND PEOPLE Turkana District is located in the northwest corner of the Republic of Kenya, between 1'0 and 5'0 north latitude (Fig. 1). The area is semiarid to arid scrub savanna and is subject to marked periodicity in rainfall. In the ideal, rainfall is distributed in a bimodal pattern across the year: the longer rainy season lasts from the end of March through the end of May and is followed by a long dry season, which begins toward the end of July. The dry season may be punctuated by the short rains (erupe) in October or November. In actuality, the timing of the rains and the amount of rainfall at any given locale in Turkana vary dramatically from year to year. Drought is a frequent occurrence and may be anticipated at least once during a 5 year period, although it is impossible to predict precisely when. During drought, some rain may fall, but the pattern is generally one in which there is a violent storm during one isolated month but little or no rain in the preceding or following months. During the last 15 years, there have been three droughts of varying severity: one from 1979-1981, another from 1984-1985, and the most recent one lasting from March 1990 through the beginning of 1992. However, the rains during the year of the breastfeeding study were described by nomadic Turkana as good. Lamphear (1992) includes Turkana people among the Karimojong linguistic cluster, and Vossen ( 1982) classifies Teso-Turkana- speaking peoples as part of the Eastern Nilotic language group (Nilo-Saharan:ChariNile)(see also Greenberg, 1966). The members of the Ngisonyoka subsection of South Turkana, who were the focus of the South Turkana Ecosystem Project (STEP), are primarily nomadic pastoralists, keeping mixed herds of camels, cattle, sheep, goats, and donkeys and moving their animals frequently in response to resource availability, the season of the year, the severity of the dry season, and the prevalence of livestock diseases in a particular area. Vegetation is sparse and is patchily distributed across South Turkana. Acacia ssp. and thorn scrub are the predominant species, but there are also areas in which various species of grasses may also be found, depending on the rainfall. Water for the herds is obtained from wells which are dug in the dry washes that crisscross the region. Herders move their livestock as the need for forage and water demands. During the dry season, vegetation becomes increasingly scarce, while water may be quite distant from forage. To distribute foraging stress more evenly throughout their territory, large herding units are broken down into smaller satellite herds, which are then sent to different parts ofthe range (Ellis and Swift, 1988). Because longer distances must be travelled between wells and forage, and because there are fewer people in each herding unit, labor demands on individuals tend to be more intense during the dry season than during the rainy season. Since the pastoral system is labor-intensive, wives and children contribute substantially to the labor pool. Successful, wealthy herders are those men who possess large herds and many people to tend them. These may include several wives, many children of varying ages, and other members of a n extended social network. Rada DysonHudson (1989) described the subsistence strategy as one in which both herd size and family size are optimized. In such a socioeconomic context, the value placed on children is high, and Ngisonyoka women average seven live births by the end of their reproductive years (Brainard, 1981; Leslie and Fry, 1989; Gray, 1992). Major features of the environment and of 24 1 BREASTFEEDING PRACTICES AMONG TURKANA NOMADS :,Ii I 1 - Hills kilometers 0 20 40 3,6" Fig. 1. Map of South Turkana the pastoral system have been described in detail in several publications (see Coughenour et al., 1985; Dyson-Hudson and McCabe, 1985). Health, biology, and demography of Ngisonyoka were examined in studies undertaken during the 1980s and early 1990s (Mugambi and Little, 1983; Leslie e t al., 1988; Little et al., 1988; Leslie and Fry, 1989; Little and Gray, 1990; Little, in press). Results of recent studies of human reproduc- tion in South "urkana may be found in Leslie e t al. (1993). THE ECOLOGY AND STRUCTURE OF BREASTFEEDING The model used to design the research is shown in Figure 2. Since rainfall patterns and seasonality have been associated with human biology of nomadic Turkana in sev- 242 S.J. GRAY FOOD SCARCITY PATCHY RESOURCES r - - - - - - - - - - - - - - - - - - -1 BIOLOGICAL FACTORS *MATERNAL AGE 8 PARITY .INFANT AGE MORBIDITY MORTALIW NUTRITIONAL STATUS SOCIAL FACTORS STATUS OF HERDING UNIT .c --c oHUMANRESOURCES .SIZE OF HERDS .EXTENT OF SOCIAL NETWORK - -- BEHAVIORAL FACTORS .MATERNAL ACTIVITY SCHEDULES .CHILD CARE PRACTICES SUPPLEMENTATION WEANING PRACTICES Fig. 2. Environmental, biological, socioeconomic, and behavioral factors hypothesized to influenced breastfeeding behavior among nomadic Turkana pastoralists. era1 earlier studies, they were given a primary causative position in the research design. However, rainfall effects were assumed to be mediated by other biobehavioral factors affecting both mothers and their nurslings. These factors, shown as biological, social, and behavioral components in Figure 2, have figured prominently in previous studies of nutrition and dietary practices in Turkana (see Wienpahl, 1984; Galvin, 1985). The two major components of breastfeeding behavior are defined as outcome variables in the model: breastfeeding structure and duration of lactation. Breastfeeding structure is defined by Vitzthum (1994a, p. 309) as “the temporal patterning of suckling duration and frequency.”Although effects on fertility of the duration of lactation have been known for some time (see Bongaarts, 1981),only recently have scientists begun to understand the impact of daily breastfeeding structure (McNeilly et al., 1988). It is now clear that this structure varies markedly among so-called natural fertility populations (Campbell and Wood, 1988),with important consequences for population differences in the timing of the resumption of menses postpartum and of the length of interbirth intervals (Gray, 1994a; Vitzthum, 1994a). The model presented here builds upon the work of several researchers. Popkin et al. (1986) developed a similar research framework which incorporated maternal, infant, and household factors as determinants of infant care and feeding practices. Jones (1988, 19901, applying the results of research by McNeilly et al. (19881, used multiple hazards analysis to demonstrate the effects of both breastfeeding structure and the duration of lactation on the timing of the return to menses postpartum. Vitzthum (1989) and Wood et al. (1985) showed that components of individual breastfeeding bouts (the intervals between bouts and the duration of suckling during bouts) may be determined by quite different sets offactors and must therefore be treated independently in analysis. The work of Panter-Brick (1991) and Vitzthum (1988) successfully integrated cultural variables such as labor demands on mothers and economic change into analysis of factors affecting infant feeding practices. The model also draws heavily on results of studies carried out by the members of the South Turkana Ecosystem Project (STEP), a long-term, multidisciplinary study of human ecology which was initiated in the late 1970s. Independent variables considered here are those that have been demonstrated to play BREASTFEEDING PRACTICES AMONG TURKANA NOMADS key roles in several aspects of human biobehavioral adaptation in South Turkana (Little, in press). From the onset, it was assumed that these same variables would influence decisions relating to the care and feeding of young children. Furthermore, it was hypothesized that care of the youngest and most vulnerable people in Turkana society would be best understood as yet another set of coping strategies in this arid and unpredictable environment. Given the variable ecological context in Turkana, however, a preliminary caveat is in order: infant feeding and caretaking behaviors reported here should be viewed as representative of the social and environmental conditions prevailing at the time of the study. 243 tween episodes suggested two distinct temporal patterns: in one pattern, episodes were separated by less than 15 min, on average, and, in the second, interepisode intervals were significantly longer. It appeared that Turkana children suckled almost constantly during periods of close maternalhursling interaction. During these periods, which might last 1-2 h, children were breastfed whenever they demanded the nipple, regardless of their age (in fact, children older than age 15 months were the most demanding of the breast). It was also the case, however, that several hours of intense breastfeeding activity might be preceded (or followed) by several hours in which no breastfeeding occurred. These nonbreastfeeding intervals among Turkana Breastfeeding patterns among women ranged from 1 or 2 h to the entire nomadic Turkana day, from sunrise to sundown. The breastfeeding study included both Since the length ofintervals between suckprospective and retrospective components. ling activity is a key factor in the return of Results of the prospective component were lactating women to prepregnant hormonal presented in detail in Gray (1994a), in which levels and thus in the duration of postparcomparison of breastfeeding structure tum amenorrhea (McNeilly et al., 1985, among Gainj of New Guinea (Wood et al., 1988; Vitzthum, 1994b), it was hoped that 19851, Quechua of Peru (Vitzthum, 1989), an understanding of the ecology of and Turkana was undertaken. To facilitate breastfeeding among nomadic Turkana comparison with those studies, temporal would explain this bimodal structure and, by patterns of suckling activity were defined as extrapolation, the influence of breastfeeding in Vitzthum (1989) and Wood et al. (1985). practices on fertility and birth seasonality A nursing event was recorded any time the in this population. child took the nipple into its mouth. Events were clustered into episodes: a series of MATERIALS AND METHODS events separated from each other by less than 5 seconds. A breastfeeding session conThe small sample size of the prospective sisted of a series of episodes separated from sample (N = 231, however, limited its useeach other by less than 60 seconds. fulness in elucidating ecological relationThe most meaningful statistical unit of ships, and a decision was made to use the breastfeeding activity among nomadic Tur- larger retrospective sample. The retrospeckana was the nursing episode. Suckling ac- tive component of the study included 101 tivity was not clustered into well-defined motherhursling pairs. Of the 101 women sessions (a series of closely spaced episodes), in the study, 92 were able to reconstruct as had been observed by Vitzthum among their activity schedules (and hence their Quechua women (1989), but occurred in- breastfeeding schedules) for the previous stead as a series of isolated episodes, usually day with some precision. Information prolasting no more than 2 or 3 min but sepa- vided by these women was corroborated by rated by intervals longer than l min (for a other women in the herding camp or by their nice illustration of this pattern, see Fig. 1 in husbands. Nine women were either uncerVitzthum, 1994a). The frequency of suck- tain about the timing of some events during ling increased dramatically as children ap- a 1-3 h period or were judged to be poor proached the second half of their second year. informants, and their responses were exFurthermore, analysis of the intervals be- cluded from analysis. 244 S.J. GRAY TABLE 1. Age structure of the samples of mothers and nurslings in the n r k a n a breastfeeding study: Retrospective component A. Age and panty of mothers in the retrospective study Aee class N Mean age (s.e.1 Mean parity (s.e.1 21-30 years 3 1 4 0 years Over 40 years Total 55 31 6 92 26.1 (0.389) 34.2 (0.455) 44.0 (1.155) 2.2 (0.144) 4.1 (0.311) 5.5 (0.885) 30.0 (0.622) 3.0 (0.183) x B. Age and sex of nurslings in the retrospective study Age class Less than 1 month 1-3 months 4-6 months 7-9 months 10-12 months 13-15 months 1&18 months 19-21 months Over 24 months Total Mean age, months (s.e.1 Females 2 11 8 10 6 4 1 3 1 46 8.3 (0.886) Males 0 14 6 10 6 7 1 2 0 46 7.5 (0.778) Subjects were selected to include women throughout the female reproductive span, and parity ranged from 1-9. Nurslings were selected to include equal numbers of both sexes and to represent a cross-section of ages from birth through the second year. Since Turkana do not keep records of birth dates, ages of women were estimated using a South Turkana event calendar developed by Rada Dyson-Hudson, Paul Leslie, and Eliud Lowoto. More precise birth dates were provided for nurslings, since mothers were often able to recall the exact month of the birth. Estimated age structure of the sample is shown in Table 1. During a detailed preliminary interview, each woman was asked to recall her breastfeeding and activity schedule, her own dietary intake, and intake of nonbreastmilk foods by her nursling for the 24 h period immediately preceding the interview. In addition, each woman and her husband assisted in reconstructing her reproductive history. Husbands also provided information on livestock resources and on size and composition of the herding unit (awi)as a whole, as well as of their wives’ own households (a woman’s household, her ekol, includes those people whom she feeds with the milk from her allotted milch animals). Finally, each of the 101 women in the study was observed at least once when she went to milk her livestock, either in the morning or the early evening of the day of the interview. At that time, the number of her milch animals and the quantity of milk produced by each species were recorded. These data were used to verify information provided during the interview in order to estimate as precisely as possible the socioeconomic ranking of the awi. The use of retrospective data of suckling activity posed a number of methodological problems, which have been discussed recently by Vitzthum (1994b). However, recall data collected during the Turkana study differed somewhat from the type discussed in that paper. Thus, additional-and important-research questions in this study pertain to the usefulness and reliability of recall data of the type collected in Turkana and to the adequacy of those data for identifying sources of variation in breastfeeding structure among nomadic Turkana. In the initial days of the study, mothers were simply asked to report how many times they had breastfed during the preceding 24 h and how long their nursling suckled during each feed. As has been true of other studies (Vitzthum, 1994b), these were completely unrealistic questions and were considered quite impossible by the women in the study (who were always ready to express their attitude concerning the inquiry at hand), since Turkana children were given the breast so frequently. Although they could not provide an actual count of individual suckling episodes, mothers were able to recall when during the day they experienced periods of frequent suckling activity. These periods were punctuated by children’s naps or by mothers’ leaving their children for one or more hours to go to the well, to town, or to visit a neighboring awi. Activities occupying less than an hour included collectingfirewood or milking. Mothers were able to recall the time and duration of their activities with considerable precision by referring to the position of the sun or to some other event for which the time could be reliably estimated. It became clear that the best way to obtain recall data on breastfeeding structure was for mothers to reconstruct their activity schedules for the BREASTFEEDING PRACTICES AMONG TURKANA NOMADS 245 TABLE 2. Actiuitv patterns of nomadic Thrkana women bv time of day Time of day Observation period Early morning (before sunrise) 4:OO AM to 7:OO AM Morning 7:OO AM to 11:OO AM Midday 11:OO AM to 4:OO PM Late afternoon 4:OO PM to 7:OO PM Evening 7:OO Night 9:00PM to 4:OO PM to 9:00PM AM Characteristic activity Rising, building fire, milking animals; breastfeeding between and during tasks Prepare morning meal, make buttermilk and butterfat, look after baby animals, assist men as they move herds from kraals, domestic tasks; depart to herd or water animals, travel to town or to neighboring camps; breastfeeding between and during tasks Domestic tasks, such as beadwork, making clothing, rest or sleep in the cool of the ekol, visit with other women in the awi; water or graze animals; nurslings often nap at this time or visit the ekol of a female relative Women gather firewood, other tasks near camp; animals return to camp (camels and cattle first); women begin milking; small stock return to camp, with herders; herding of animals into kraals; young stock returned to mothers; very busy; nurslings usually carried on their mothers’ backs during these tasks Milk small stock; cook evening meal; talk and storytelling, visit neighbors; dancing and singing (wet season); breastfeeding between and during tasks Sleep with nursling a t breast; in the dry season, women may depart with baby animals to take them to water entire 24 h period. These schedules proved to be consistent with the daily pattern of breastfeeding and work observed during the prospective component of the study (Table 2). In addition to estimates of 24 h breastfeeding frequency and of the number of hours in which no breastfeeding occurred, maternal recall of activities was used to determine labor demands of the pastoral system on mothers, the degree to which mothers relied on alternative caretakers, and the pattern of weaning. In order to determine whether there was diurnal variation in suckling activity, the day was divided into six periods derived from maternal work and sleep patterns observed during the prospective study: very early morning, from approximately 4:OO AM until sunrise, at about 7:OO AM; morning, from sunrise to 11:OO AM; midday, from 11:OO AM to 4:OO PM; late afternoon, from 4:OO PM until sunset a t approximately 7:OO PM; evening, from sunset until bedtime a t about 9:00 PM; and night, from bedtime until early morning, when mothers awoke (Table 2). Other data obtained during interviews were used to quantify biological, socioeconomic, and behavioral factors hypothesized to influence breastfeeding behavior (Fig. 2). Household and livestock data were used in computing three socioeconomic indices: the socioeconomicrank of the herding camp, the depth of the woman’s social network, and the number of her cowives. The ranking was calculated from counts of adults and adolescents in residence, estimates of herd size and composition, and counts of milch animals in the herd. Since livestock resources and human resources in Turkana are linked, camps which had more livestock also had more adults and adolescents in residence (adolescents are responsible for watering and herding the livestock, relieving married women from these tasks); hence, they were of higher socioeconomic ranking. Lower-rank camps might consist of one herder with one wife, one older postreproductive woman, several children under age 5 years, and a few animals, usually goats, sheep, and one or two camels or cows. The ranking of the awi was used as both a n indicator of the overall economic status and self-sufficiency of the herding unit and as a n estimate of its resilience to the stresses ofthe dry season: food scarcity and increased energetic demands on both humans and livestock associated with longer distances travelled between wells, grazing areas, and camp sites. A detailed discussion of socioeconomic differences in this sample may be found in Gray (1994b). The number of close family or friends not in residence in the camp a t the time of the study but with whom a woman reported frequent interactions (several visits over the 246 S.J. GRAY course of a year) was used a s a n index of the depth of her social network. The number of cowives includes only those who resided in the same herding camp as the woman herself; otherwise, they are included as members of the external network. The depth of the social network and the number of a woman’s resident cowives were used as measures of human resources: specifically, the number of people who were available to assist in child care and herding responsibilities and to provide additional food, animals, or money (Wienpahl, 1984; Gray, 1994b). Anthropometric examinations of all mothers and their nurslings were conducted on the same day as the interview. These data were used to assess nutritional status, size, and body composition of mothers and their nurslings. Measurements included recumbent length, head and chest circumferences (children only), height (of mothers), weight, midcalf and upper arm circumferences, and six skinfolds (triceps, suprailiac, subscapular, periumbilical, midaxillary, and midcalf). Maternal stature was measured to the nearest millimeter using a standard anthropometer. Recumbent length was measured to the nearest millimeter using a n infant measuring board. Children were weighed in a basket suspended from a Salter metric scale. If they were too big for the basket, they were weighed on a Health-0-Meter floor scale. In some cases, mothers of children who resisted weighing were asked to step on the scale holding the child, and the measurement was recorded. The child was then taken from the mother, who was weighed again. The difference in the two weights was recorded as the weight of the child. Maternal weight was measured using a Health-0-Meter floor scale and was corrected for the weight of clothing and adornment. All weights were measured to the nearest 0.1 kg. All circumferences were measured to the nearest millimeter using a plastic-coated cloth measuring tape. Skinfold measurements were taken to the nearest mm using a Lange skinfold caliper. Monthly rainfall data were obtained from the meteorological station at the district capital in Lodwar and represented the monthly average of five locations across Turkana (two are in South Turkana). Rainfall influences availability of forage and distances travelled between forage and water; hence, it affects labor and energetic demands on all members of a herding unit. Dependent variables, as used in this paper, refer to reported frequencies of breastfeeding activity. For each motherlnursling pair, these include individual counts for each time of day described in Table 2 above a s well as the summed count for the entire 24 h period. Independent variables refer to factors hypothesized to influence breastfeeding behavior: rainfall, all maternal and infant characteristics, and all socioeconomic indices. However, the designations dependent and independent as used here should not necessarily be assumed to suggest causality; they are used rather to simplify discussion. The term bouts, when used in this paper, refers to extended periods of on-demand breastfeeding as reported by mothers during 24 h recall interviews. It is used with the specific purpose of differentiating recalled breastfeeding data from prospective data; since neither the frequency nor duration of suckling episodes during these bouts was known, it was felt that the use of the suckling unit session (as defined by Vitzthum, 1989) to describe such activity was inappropriate. Univariate analysis was run for all dependent and independent variables. Of these, only anthropometric variables were normally distributed. Factor analysis (Bernstein e t al., 1988) was used to identify intercorrelations and underlying factors in anthropometric data. Most of the data collected for the study were counts (with the exception of anthropometric data), but attempts to normalize them using a square root transformation (Sokal and Rohlf, 1981) were unsuccessful. Failure of the data to satisfy the assumptions of parametric analysis clearly indicated that nonparametric analysis was more appropriate for this study. Two measures of rank-sum association were employed: Spearman’s coefficient of rank-sum correlation and the Kruskal-Wallis X2-approximation to test for differences in location of ranked data. Where only two classes were considered, the Wilcoxon 2-sample rank sum test of differences in central location (normal BREASTFEEDING PRACTICES AMONG TURKANA NOMADS 247 4 5 6 7 8 9101112 1 2 3 4 5 6 7 8 9101112 1 2 3 4 am I Pm I am Time of day Fig. 3. Variation in frequency of breastfeeding bouts reported by nomadic Turkana women by time of day. approximation) was used (Sokal and Rohlf, 1981). Statistical analyses were undertaken using SAS/STAT, Version 6.0, fourth edition. PROC UNIVARIATE was used for univariate analysis; PROC PRINCOMP and PROC FACTOR were used for exploratory analysis of anthropometric data. PROC FREQ and PROC CORR were used to identify associations between variables; PROC NPARlWAY was used for nonparametrie analysis of variance. RESULTS A total of 639 bouts of suckling activity were reported by the 92 women. As is discussed in detail below, there was variation in absolute frequencies of suckling activity among the mothers in the study. Nonetheless, a clear diurnal pattern emerged which was consistent across all women in the sample, regardless of the frequency with which they nursed their infants (Fig. 3). Approximately 34% of all bouts were reported to have occurred in the 7 h from 4:OO AM t o 11:OO AM. Nearly 16%occurred in the early evening hours, between 7:OO PM and 9:00 PM. Twenty-five percent occurred during the night, and 25% of bouts occurred in the 8 midday and late afternoon hours, between 11:OO AM and 7:OO PM. Mothers were more likely to report that they had experienced more than one bout during morning or nighttime hours. The frequency of nighttime bouts was certainly underestimated, since mothers consistently reported that they did not always wake when their children nursed during the night (children slept a t their mother’s breast). Furthermore, mothers were more likely to report that they had not nursed at all during midday or late afternoon hours. The total number of feeds reported for the 24 h period was most strongly correlated with the frequency of breastfeeding activity in morning and nighttime hours (Table 3). Correlation analysis was run to identify associations between bout frequency and the independent variables described above. Statistically significant correlations are shown in Tables 4 and 5. Correlations were gener- 248 S.J. GRAY TABLE 3. Spearman coeficients (r,) of correlations between the total number of breastfeeding bouts reported by Turkana women for 24 h and breastfeeding bouts reported for each time of day Total number of breastfeeding bouts Sunrise Morning Midday Afternoon Evening Night r, Plr,l 0.260 0.0123 0.586 0.0001 0.424 0.0001 0.354 0.0005 0.273 0.0084 0.5648 0.0001 = 0 TABLE 4. Spearman coeficients (rJ of correlations between biological, socioeconomic, and behavioral factors and 24 h breastfeeding frequency as recalled bv Turkana women' Number of bouts (24h) Maternal weight Maternal sum of six skinfolds Age of nurslings Nursling head circumference Nursling midaxillary skinfold Depth of social network Hours mothers and nurslings were apart Rainfall in month of interview Hours with no bouts 0.281' 0.242' -0.218 -0.32P ns 0.256 -0.438 -0.1814 ns -0.27g2 ns ns -0.281~ ns 0.572 ns 'PlrMl = 0 < 0.05 'Association is independent of maternal age or parity or infant age. 3Association is independent of infant age or growth. '0.05 c P < 0.1. TABLE 5. Spearman coefficients (r,) of correlations between socioeconomic and behavioral uariables and frequency of breastfeeding bouts by time of day as reported bv nomadic Thrkana women Night bouts Wife number Number of cowives in residence Depth of social network Number of hours apart Rainfall in the month of interview Morning bouts Midday bouts ns ns ns ns ns -0.370' ns ns 0.442' ns ~0.238~ -0.235l 0.2601 ns -0.204" ' P Ir,l = 0 < 0.05. ' P jr,I = 0 c 0.001. '0.05 < P < 0.1. ally weak but tended to be stronger for the total number of bouts in the 24 h period (Table 4) than for the number of bouts during a particular time of day. Exceptions were the number of bouts during the morning and midday periods with the number of hours mothers spent apart from their infants (r8= -0.37 and -0.44, respectively; P < 0.001) and the number of nighttime bouts with several of the socioeconomic indices (Table 5). The total number of bouts was significantly correlated with several mea- sures of child growth, but these associations were not independent of nursling age, which was itself only weakly associated with 24 h bout frequency. The strongest association of bout frequency with any growth variable was with head circumference, but this association was independent of the age of the child (partial r, = -0.26; P = 0.013). Interestingly, there was no significant relationship between bout frequency and fatness of infants (which had been shown to be independent of age and growth in preliminary factor analysis). Of maternal characteristics, weight and the sum of six skinfolds showed the strongest associations with the reported frequency of breastfeeding bouts. Both associations were independent of any measures of maternal or infant age. Indeed, preliminary principal components analysis had shown that these two variables were the best indicators of maternal size and fatness, respectively. Of the other independent variables, only the depth of the social network and the number of hours mothers spent away from their infants were significantly correlated with the 24 h frequency of breastfeeding bouts. Rainfall showed a very slight association with 24 h bout frequency (Table 41, most probably an artifact of its slightly stronger association with the frequency of nighttime bouts (Table 5). Given the nature of these data, it was impossible to obtain estimates of the duration of suckling during bouts; however, estimates of the number of hours individual mothers went without breastfeeding could be calculated from periods of the day for which they reported no bouts. Correlations between these estimates and the set of independent variables were run; those results are also shown in Table 4. There were no significant associations of the number of hours in which mothers failed to breastfeed with nursling BREASTFEEDING PRACTICES AMONG TURKANA NOMADS age or growth. However, there was a correlation with the fatness of children, as measured by the midaxillary skinfold; as had been suggested by the results of principal components analysis, this association was independent of nursling age. Of the maternal characteristics, maternal fatness (the sum of six skinfolds) was significantly and independently correlated with the number of hours in which no breastfeeding occurred. To summarize, bout frequency was positively correlated with the size and fatness of mothers and the depth of their social network but negatively correlated with age (growth) and head circumference (independent of age) of nurslings. There was a strong negative correlation between bout frequency and the number of hours mothers spent apart from their nurslings and a slight negative association with rainfall in the month of the interview. Failure to breastfeed during part of the day, on the other hand, was negatively correlated with both maternal and nursling fatness but positively correlated with the number of hours mothers and nurslings were apart. All associations, although statistically significant, were relatively weak. In a n effort to clarify associations suggested by the results of correlation analysis, nonparametric analysis of variance in bout frequency was undertaken. On the basis of the results of univariate analysis, women in the sample were assigned to one of three groups: those mothers who recalled nine or more bouts during 24 h (those above the seventy-fifth percentile; N = 20), those who recalled between six and eight bouts (N = 48), and those who reported five or fewer bouts (those below the twenty-fifth percentile; N = 24). Although the three groups exhibited similar patterns of breastfeeding activity over the course of the day (see Fig. 3), women in the low-frequency group were more likely to report that they had failed to breastfeed a t all during a given period of the day (for 36.1% of all periods recalled, they reported no suckling activity). Furthermore, they tended to report only one bout per period in which they did breastfeed (Table 6). Women in the high-frequency group recalled very few periods in which they failed to breastfeed (9.2%of periods recalled) but sev- 249 eral periods in which they experienced two or more bouts (47.5% of periods). Women in the median frequency group generally reported breastfeeding activity similar to that of the high-frequency group but recalled very few periods in which three or more bouts occurred (3.5%of periods recalled). All of the women in the low-frequency group reported that they had failed to breastfeed during a t least one period of the day, and 71% (N = 17) reported failure to breastfeed during two or more periods. Of these, 50%(N = 10)experienced two or more consecutive periods without breastfeeding. Since the majority of nonnursing periods reported by these women were clustered into two or more blocks of time in the morning, daytime, and early evening hours, it is certain that they had experienced interbout intervals in the range of 5-10 h in the preceding 24 h period. Among the other women in the study, the percent not breastfeeding during at least one period of the preceding 24 h was 50% and 63% for high-frequency and mediumfrequency feeders, respectively. Only one woman from the high-frequency group failed to nurse during two or more periods (5%); these were consecutive blocks of time. Of the women in the median-frequency group, 21% (N = 10) failed to nurse during two or more periods; of these, five women experienced two consecutive periods with no suckling activity (10% of median group). Nonparametric analysis of variance (Kruskal-Wallis X2-approximation) was run for each of the variables that were correlated with bout frequency and failure to breastfeed (in Tables 4, 5). The exception is infant weight, which was included because of its high correlation with growth. Results are shown in Table 7. The analysis allows more precise characterization of each of the three groups of motherlnursling pairs. Mothers in the low-frequency group weighed less and tended to be shorter than the mothers in the other two groups. Their social network was more shallow, and they spent significantly more hours away from their infants. The nurslings in this group were slightly larger, as indicated by differences of 2 cm in head size and 1 kg in weight. They were also on average 2.5-3 months older than the 250 S.J.GRAY TABLE 6. Percent of recalled periods' for which zero, one, two, or more than 2 breastfeeding bouts were reported by Turkana women in the retrospectiue study Low frequency Number of bouts reported 0 bouts reported 1 bout 2 bouts 3 or more bouts Total bouts reported Total periods recalled Ratio of boutdperiods Likelihood Ratio Xz, testing for Row Percent of recalled periods, by level of breastfeeding frequency in 24 h Median frequency X 36.1 55.6 8.3 0.0 104 144 0.72 Column IndeDendence: 99.67. P High frequency 13.9 58.7 24.0 3.5 338 288 1.17 = 9.2 43.3 25.8 21.7 203 120 1.70 0 'Periods of the day, as described in Table 2. TABLE 7. Results of analysis of variance between groups of Turkana women in breastfeeding activity and in biological, socioeconomic, and behavioral factors' hypothesized to affect breastfeeding behavior: Retrospectiue study Turkana women grouped by 24 h bout frequency Mean number of feeds recalled (24 h ) Mean number of hours in which no bouts were reported (24 h ) Mean number of hours in which mothers and infants were apart Mean infant age (months) Mean infant weight (kg) Mean infant head circumference (cm) Mean infant midaxillary skinfold (mm) Mean maternal height (cm) Mean maternal weight (kg) Mean infant midaxillary skinfold (mm) Mean depth of mother's social network Mean rainfall, month of interview Low N = 24 Median N = 48 High N = 20 Kruskal-Wallis X2-approximation 4.3 (0.24) 8.0 (0.93) 10.3 (1.09) 10.1 (1.38) 8.1 (0.40) 45.0 (0.63) 4.8 (0.22) 161.4 (1.42) 43.4 (1.24) 28.2 (1.48) 10.3 (1.16) 30.2 (6.40) 7.0 (0.12) 2.8 (0.39) 6.9 (0.53) 6.9 (0.64) 7.1 (0.28) 43.0 (0.50) 5.7 (0.23) 165.0 (0.82) 48.1 (0.75) 31.1 (1.26) 15.2 (1.18) 22.8 (3.36) 10.2 (0.42) 1.9 (0.52) 4.9 (0.69) 7.7 (1.41) 7.1 (0.50) 42.7 (0.77) 5.6 (0.33) 163.8 (2.07) 49.8 (1.66) 35.4 (2.74) 14.6 (1.73) 14.2 (4.40) 77.37 P < 0.0001 31.37 P < 0.0001 13.79 P < 0.001 ns 4.99 P < 0.0825 7.08 P < 0.029 5.82 P < 0.0546 5.06 P < 0.0796 13.81 P < 0.001 ns 8.11 P < 0.0173 4.92 P < 0.0852 'Variables selected for analysis are those t h a t were significant in correlation analysis (Tables 5, 6). Standard errors are shown on the line below the group mean in parentheses. children in the other two groups, although the age difference was not statistically significant. Children in the low-frequency group were also slightly thinner, as indicated by differences in midaxillary skinfold thickness. Differences between the highfrequency and median-frequency groups were negligible and suggested that there were in fact only two distinct groups of mothers in this study: those who reported five or fewer breastfeeding bouts in the preceding 24 h and those who reported more than five. In relation to variation in bout frequency in the latter group, differences in rainfall, although only of slight significance, are of interest. There was an increase of approximately 8 mm rainfall in the month of the interview for each group, moving from high frequency to low. That is, the lowest rainfall occurred during the months in which the women in the high-frequency group were interviewed. BREASTFEEDING PRACTICES AMONG TURKANA NOMADS 251 number of hours spent away from their nurslings suggest that mothers were the princiBoth correlation analysis and nonparapal actors. That is to say, a mother's percepmetric analysis of variance suggested that tion of her child's developmental progress several variables from the set of factors idenappears to have triggered changes in her tified in the original research design were behavior, such that she was more likely to in fact associated with differences in breastfeeding behavior among nomadic Tur- be separated from her child for a prolonged period during the day. Mothers of older nurskana mothers and their nurslings. Clear patlings were observed to spend more time away terns emerged, which were also consistent from the herding camp in the performance with the results of the prospective study of tasks such as herding and watering ani(Gray, 1994a). Physical status of both mothmals or making purchases in town or in visers and children was associated with varia- iting neighbors. Such activities usually retion in bout frequency in this sample. quired that a woman depart from camp in Among children, head circumference and the early morning and return after sunset fatness showed stronger relationships with to avoid travel during the heat of midday, breastfeeding than did age or any other indi- thus explaining the failure of several women ces of growth, such as weight or recumbent to breastfeed during this part of the day. length. The strong relationship between For nurslings, on the other hand, probout frequency and head circumference is of longed maternal absences were stressful. particular interest. The most likely explana- The oldest children in the study (older than tion of this association is that changes in age 18 months) were observed to refuse food breastfeeding frequency among nomadic during the day, preferring to wait until their Turkana in the study were not tied as mothers returned. These children were genstrongly to chronological age or size of chil- erally more demanding of the breast when dren as to behavioral changes which were, their mothers were in camp. In response to in turn, linked to brain growth and cognitive the persistent nursing demands of older childevelopment. Ifthat was the case, then head dren, their mothers spent longer periods circumference may be the best anatomical away. Finally, mothers in the sample recorrelate of key developmental milestones ported that the most effective weaning stratrecognized by mothers. Previous studies of egy was to simply leave their child in the growth of nomadic Turkana children have care of another woman in the awi for several focussed on weight, height, and fatness, with weeks, while the mother herself went to visit particular attention to the falling off of friends or relatives. growth in length and weight during infancy The relationships between nighttime (Little et al., 1993)and to patterns of growth bouts and rainfall, wife number, and the in stature and changes in body composition number of cowives (Table 5) may be related during childhood and adolescence (Little to another preweaning strategy practiced by et al., 1984; Little and Johnson, 1987; Little mothers of older children: that of sending and Gray, 1990). Given the falling-off of nurslings to sleep with another woman in growth during infancy already established the awi rather than with their mother. The in this population, it will now be most intri- mother will, on occasion, go to nurse her guing to examine the pattern of growth in child if it wakes and cries, but it may also head size, particularly if the latter does in- be fed animal's milk in its mother's absence. deed trigger changes in care and feeding of Thus, the rainy season, when milk from the very young children. herds is more plentiful, is an appropriate The relationship between breastfeeding time to initiate such a behavioral change. A bout frequency and age and development of cowife is a logical choice for an alternate Turkana children raises the question of nighttime caretaker, since an important facagency: did mothers or nurslings actively tor in selecting this strategy, in addition to initiate changes in infant feeding behavior? the developmental stage of the child, is the The very strong associations between bout resumption of sexual relations between the frequency, failure to breastfeed, and the mother and her husband. DISCUSSION 252 S.J. GRAY Both maternal size and fatness were associated with recalled bout frequency and/or failure to breastfeed during a part of the day. The logical explanation of these relationships, at least as they concern fatness or weight, is that they are simply a reflection of changes in maternal physical status postpartum: we would expect that mothers of older nurslings would be thinner. However, the relationships identified here were independent of the age of nurslings or of any other measures of nursling growth and development. In other words, after controlling for the age of their children, smaller and thinner mothers did in fact breastfeed less frequently. The peculiar association with maternal size and body composition may be linked to relationships between breastfeeding behavior and socioeconomic status. Both sets of analysis suggested an association between the depth of the social network and breastfeeding frequency: the more shallow a woman’s network was, the fewer the number of breastfeeding bouts she reported. It is the case among nomadic Turkana (and among other groups in the Karimojong cluster) that wealthier and more resilient herding units tend to be members of larger networks (Dyson-Hudson, 1966; Johnson, 1990). Conversely, a shallow social network certainly suggests a less wealthy herding unit. The absence of any direct correlation between recalled bout frequency and the ranking of the awi may imply that the social network was a better indicator of socioeconomic status. Since large herding units may be split into satellite units, livestock and human resources of any given awi at the time of the interview may not have reflected adequately long-term social and economic stability of the unit or its true size and composition. As stated earlier in this paper, the network is one measure of human resources available, as is the overall wealth of a herding unit. A shallow network thus has important implications for the size of the labor pool of the awi. Gray (1992) demonstrated that wives of poorer herders spent more time attending to herding duties outside of the camp, since they had fewer children and cowives who could assist them in such tasks. Thus, it was expected that poorer women would spend more hours away from their infants and hence that they would breastfeed more infrequently. Table 7 shows that women in the low-frequency group spent, on average, 10.1 h out of 24 away from their nurslings; that was 3.5 h and 5.5 h more than women in the median- and highfrequency groups, respectively. The thinness of mothers, coupled with the thinness of their nurslings, may be an index of socioeconomic status. Thinness may reflect poorer nutritional status in general. Size is more problematic, but it is possible that small size of a Turkana woman equates in some way with either lower socioeconomic status of her father’s awi or with lower fertility; if so, smaller, thinner women may be less desirable marriage partners for wealthy men and hence may be more likely to be married to poor herders. This hypothesis cannot be addressed in the present study, but it is certainly of interest for future research. Although of only slight significance, differences between the three groups of women in rainfall during the month of the interview were intriguing, since they ran counter to what might be expected if the onset of the dry season increased labor demands on mothers andlor put them at risk of negative energy balance. If that were indeed the case, lower rainfall should have been associated with longer maternal absences and lower breastfeeding frequency. Instead, rainfall appears to have been lowest in the months in which mothers in the high-frequency group were interviewed. A possible explanation is that enhanced milk production of the herds during periods of higher rainfall increased availability of foods, other than breastmilk, that were considered appropriate for infants. Under such conditions, a mother may have felt less constrained by the nursing demands of her child and may have spent more time away from camp, leaving her nursling with an alternate caretaker and a plentiful supply of animal’s milk (children are in fact introduced to goat’s and camel’s milk by age 4 months). Again, such behavior may have been more characteristic of women from camps of low socioeconomic status, in which mothers contributed significantly in the herding of livestock. The weakness ofthe associations identified here may be an artifact of the data them- BREASTFEEDING PRACTICES AMONG TURKANA NOMADS selves; as discussed earlier, the problems of recalled breastfeeding data are legion. However, the results of this particular retrospective study are encouraging in that, unlike other studies based on maternal recall, they are in agreement with the findings ofthe prospective study (Gray, 1992, 1994a), particularly in relation to the overall frequency of breastfeeding and changes in nursing behavior among older children. In fact, the breastfeeding data appear to be more precise and more reliable than was the case for some independent variables used in the study: the difficulty in determining the socioeconomic rankingoftheawiis acaseinpoint. Nonetheless, the retrospective data were useful in teasing out relationships suggested by the much smaller set of prospective data. That one of the strongest associations identified was between breastfeeding activity and head circumference,which may be the best indicator of the progress ofinfant growth and development in this population, increases the credibility of the results. With regard to the larger study of reproductive ecology of nomadic Turkana of which the breastfeeding study was but one part, the long intervals Turkana mothers spent away from their infants during the day may have important implications for the timing of the return to menses postpartum in this population. It was not unusual, on the one hand, for mothers of children as young as 8-10 months t o spend five or more hours at a time away from their children during the daytime. On the other hand, Turkana infants slept a t their mother’s breast until well into their second year and nursed on demand throughout the night. Thus, in this population, suckling patterns during the daytime hours should prove to have a far more important effect on the duration of lactational amenorrhea and the timing of the next conception than do nighttime patterns (cf.Jones 1988, 1990). It might also be hypothesized that duration of amenorrhea is curtailed among nomadic Turkana, relative to other natural fertility populations practicing socalled on-demand breastfeeding, since marked increases in the length of intervals between breastfeeding bouts were evident by the time infants reached 10 months (see also Gray, 1994a). 253 CONCLUSIONS Of the variables hypothesized to influence breastfeeding structure among nomadic Turkana, nursling development, maternal size and fatness, and the depth the maternal social network were most strongly linked to variation in bout frequency in this sample. All of these effects appear to have been mediated by the number of hours mothers spent away from their children over the course of 24 h, behavior which itself was determined largely by labor demands of the pastoral system. When considered in concert with the results of the prospective study, the results of the retrospective component substantially increase our understanding not only of the structure of breastfeeding in this population but also of the ecological factors which introduce variation into that structure and of their influence on fertility through their effects on breastfeeding structure. Furthermore, the type of recall data employed here suggests a method of increasing the precision and reliability of such data which should be more carefully developed in future research. As Vitzthum (199413) has noted, collection of prospective data is timeconsuming and often invasive, and researchers must find more efficient methods of collecting recall data. It is clear from the results of the Turkana study that future research may be enhanced by using both methods. However, the present study should be seen only as a relatively raw effort; in the future, it is hoped that refinement of data collection methods employed here will increase the range of analytic options available, resulting in stronger associations between breastfeeding behavior and the full array of biological, socioeconomic, and behavioral variables which influence it. At present we have at best faint hints; it is hoped, in the future, that we will have a clearly delineated and more precisely modeled biobehavioral phenomenon. ACKNOWLEDGMENTS Members of the Turkana households with whom I worked were extraordinarily cooperative and hospitable, and I am much in their debt. Turkana women in particular accepted my presence most graciously and my questions with great patience and humor. Many 254 S.J. GRAY _ * to Ravi Duggirala for his assistance in the This research was funded by NSF grant BNS-8718477. LITERATURE CITED Bernstein IH, Garbin CP, and Teng GK (1988) Applied Multivariate Analysis. New York: Springer-Verlag New York Inc. Bongaarts J (1981) The impact on fertility of traditional and changing child-spacing practices. In H J Page and R Lesthaeghe (eds): Child-Spacing in Tropical Africa. London: Academic Press, pp. 11-129. Brainard JM (1981) Herders to farmers: The effects of settlement on the demography of the Turkana population of Kenya. Ph.D. Dissertation, State University of New York, Binghamton. Campbell KL, and Wood JW (1988) Fertility in traditional societies. In P Diggory, M Potts, and S %per (eds.): Natural Human Fertility: Social and Biological Determinants. Houndmills, Basingstoke: MacMillan Press, pp. 39-69. Coughenour MB, Ellis JE, Swift DM, Coppock DL, Galvin K, McCabe JT, and Hart TC (1985) Energy extraction and use in a nomadic pastoral ecosystem. Science 230.519-625. Dyson-Hudson N (1966) Karimojong Politics. Oxford: Clarendon Press. Dyson-Hudson R (1989) Ecological influences on systems of food production and social organization of South Turkana pastoralists. In V Standen and RA Foley (eds): Comparative Socioecology: The Behavioral Ecology ofHumans and Other Mammals. Oxford: Blackwell Scientific Publications, pp. 165-193. Dyson-Hudson R, and McCabe J T (1985) South Turkana Nomadism: Coping with an Unpredictably Varying Environment. HRAFLEX Books Ethnography Series. New Haven: Human Relations Area Files, Inc. Ellis JE, and Swift DM (1988) Stability ofAfrican pastoral ecosystems:Alternate paradigms and implications for development. J . Range Management 4:450-459. Galvin K (1985) Food Procurement, Diet, Activities and Nutrition of Ngisonyoka, Turkana Pastoralists in an Ecological and Social Context. Ph.D. dissertation, State University of New York, Binghamton. Gray SJ (1992) Infant Care and Feeding Practices among Nomadic Turkana Pastoralists of Northwest Kenya: Implications for Fertility and for Child Survival. Ph.D. dissertation, State University of New York, Binghamton. Gray SJ (1994a) Comparison of the effects of breastfeeding practices on birth spacing in three societies: Nomadic Turkana, Gainj and Quechua. J . Biosoc. Sci. 26:69-90. Gray SJ (199413)Correlates of dietary intake of lactating women in South Turkana. Am. J . Hum. Biol. 6:369-383. Jones RE (1988) Return-to menses postpartum in rural Indonesian women. Hum. Biol. 60:853. Jones RE (1990) The effect of initiation of child supplementation on resumption of postpartum menstruation. J . Biosoc. Sci. 22:173-189. Lamphear J (1992) The Scattering Time: Turkana Responses to Colonial Rule. Oxford: Clarendon Press. Leslie PW, and Fry PH (1989) Extreme seasonality of births among nomadic Turkana Pastoralists. Am. J. Phys. Anthropol. 79:103. Leslie PW, Fry PH, Galvin K, and McCabe J T (1988) Biological, behavioral and ecological influences on fertility in Turkana. In EE Whitehead, CF Hutchinson, BN Timmerman and RG Varady (eds): Arid Lands Today and Tomorrow: Proceedings of an International Research and Development Conference. Boulder: Westview Press, pp. 705-712. Leslie PW, Campbell KA, and Little MA (1993) Pregnancy loss in nomadic and settled women in Turkana, Kenya: A prospective study. Hum. Biol. 65:237-254. Little MA (in press) Adaptability ofAfrican pastoralists. In SF Ulijaszek and R Huss-Ashmore (eds): Human Adaptability: Past, Present and Future. Little MA, and Gray SJ (1990) Growth of young nomadic and settled Turkana children. Med. Anthropol. Q. 4:296-314. Little MA, and Johnson BR (1987) Mixed-longitudinal growth of nomadic Turkana pastoralists. Hum. Biol. 59396-707. Little MA, Galvin K, and Mugambi M (1984) Crosssectional growth of nomadic Turkana pastoralists. Hum. Biol. 55:811-830. Little MA, Galvin K, and Leslie PW (1988) Human growth, health and energy requirements in nomadic Turkana pastoralists. In I deGarine and GA Harrison (eds): Coping with Uncertainty in Food Supply. Oxford: Oxford University Press, pp. 288-315. Little MA, Gray SJ, and Leslie PW (1993) Growth of nomadic and settled Turkana infants of northwest Kenya. Am. J . Phys. Anthropol. 92:273-289. McNeilly AS, Glasier A, and Howie PW (1985)Endocrine control of lactational infertility. In J Dobbing (ed): Maternal Nutrition and Lactational Infertility. New York Raven Press, pp. 1-24. McNeilly AS, Howie PW, and Glasier A (1988) Lactation and the return of ovulation. In P Diggory, M Potts, and S Teper (eds.): Natural Human Fertility: Social and Biological Determinants, Houndmills, Basingstoke: MacMillan Press, pp. 102-117. Mugambi M, and Little MA (1983) Blood pressure in nomadic Turkana pastoralists. East Afr. Med. J. 60t863-869. Panter-Brick C (1991) Lactation, birth spacing and maternal workloads among two castes in rural Nepal. J . Biosoc. Sci. 23:137-154. Popkin BM, Laskey T, Litvin J , Spicer D, and Yamamoto ME (1986) The Infant Feeding Triad. New York: Gordon and Breach Science Publishers. BREASTFEEDING PRACTICES AMONG TURKANA NOMADS Sokal RR, and Rohlf FJ (1981) Biometry, 2nd ed. New York: W.H. Freeman and Co. Vitzthum VJ (1988)Variation in infant feeding practices in an Andean community. Michigan Studies in Andean Anthropology, Michigan Discussions in Anthropology 8:137-156. Vitzthum VJ (1989) Nursing behavior and its relation to duration of postpartum amenorrhea in an Andean community. J . Biosoc. Sci. 21:145-160. Vitzthum VJ (1994a) Comparative study of breastfeeding structure and its relation to human reproductive ecology. Yrbk. Phys. Anthropol. 37307-349. 255 Vitzthum VJ (1994b) Suckling patterns: Lack of concordance between maternal recall and observational data. Am J. Hum. Biol. 6:551-562. Vossen R (1982) The Eastern Nilotes: Linguistic and Historical Reconstructions. Berlin: Deitrich Reimer Verlag. Wienpahl J (1984) The Role of Women and Small Stock Among the Turkana. Ph.D. dissertation, University of Arizona, Tucson. Lai D, Johnson PL, Campbell KL, and Maslar Wood JW, IA (1985)Lactation and birth spacing in highland New Guinea. J . Biosoc. Sci. Suppl. 9:159.