Eight-year study of social and ecological correlates of mortality among immature baboons of Mikumi National Park Tanzania.код для вставкиСкачать
American Journal of Primatology 16:199-212 (1988) Eight-Year Study of Social and Ecological Correlates of Mortality Among Immature Baboons of Mikumi National Park, Tanzania RAMON J. RHINE', SAMUEL K. WASSER', AND GUY W. NORTON3 'Department of Psychology, University of California, Riverside, 'Department of Psychology, University of Washington, Seattle, 'Department of Applied Biology, Cambridge University, Cambridge, England The Cox Proportional Hazards Regression Model was used to examine ecological and sociodemographic correlates of mortality among 164 immature yellow baboons born over a n 8-year period at Mikumi National Park, Tanzania. Ecological correlates were derived from seasonal rainfall. Mortality was lowest for immatures born during the late rainy season, when it was likely that nourishment for pregnant and lactating females was greatest. High mortality was associated with above-average rainfall early in the immature's first complete rainy season. This association may result from one or more of the following: exposure, increased ranging, or accelerated vegetation growth, each of which is thought to increase hazards for weanlings. When births for the 8 years of the study were pooled, a birth peak occurred during the early dry season. The larger number of births during a birth peak should intensify competition among mothers and among immatures. Mortality was greatest for immature females born during the season immediately following the birth peak, while lowest for immature females born during the season immediately preceding the peak. Immature female mortality was greatest when the troop size a t birth was large, and was most severe for females born to low ranking mothers. The sociodemographic results are consistent with Wasser's hypothesis that cooperative attacks of female baboons upon other troop females suppress the others' reproduction and the vitality of their infants, thereby improving the relative competitive position of the attackers' own young. Key words: Papio, sociodemographiccorrelates, rainbow patterns, social competition INTRODUCTION The effect of ecological and sociodemographic pressures on survival is a central concern of behavioral ecology [Dittus, 1979; Dunbar, 1980; Malik et al., 1984; Altmann et al., 1985; Hrdy, 19861. This paper is a n investigation of such pressures on the mortality of immature yellow baboons (Pupio cynocephalus) living in Mikumi National Park, Tanzania. Mikumi baboons have been studied almost continuously since 1975 in a habitat described by Norton et al. . These long-term observaReceived for publication February 9,1988; revision accepted July 22, 1988. Address reprint requests to R.J. Rhine, Department of Psychology, University of California, Riverside, CA 92521. 0 1988 Alan R. Liss, Ine. 200 I Rhine et al. tions yielded a large sample of individually identified immatures needed for reliable demographic analyses. Among the many possible ecological influences upon mortality, rainfall in East Africa is a prime candidate for study because of its seasonal effect upon important aspects of the animals’ environment. Rainfall might influence the health of pregnant females, lactating mothers, or immatures either through sickness or through food production, depending upon the amount and timing of rainfall vis a vis immature development [e.g., Dunbar, 1980; Altmann et al., 19851. Also, the amount and timing of rainfall could create hazardous conditions by their impact upon ranging patterns and upon the growth of cover in which immatures could become lost or predators could hide. For example, Rhine et al. [1980, p. 4021 described the plight of a n infant, Magego, the day after her mother was killed by a predator: “The troop was out of sight over 500 m away, and Magego was wandering aimlessly by herself in heavy grass.” The possible impact of social conditions upon the survival of immatures is suggested by the work of Wasser, who found that competition among female baboons at Mikumi was tied to important reproductive events [Wasser, 1983; Wasser & Starling, 1986, 19881. In particular, cooperative female-female aggression (attack coalitions) increased with estrous and pregnancy synchrony. Receipt of attack was in part dependent upon the female’s reproductive state and in part upon the time a female gave birth relative to births of other young in her offspring’s cohort: The first few females to give birth received the highest rates of attack. The effect of such stressful female-female aggression upon pregnancy or mothering may reduce the number and viability of immature competitors the attackers’ own offspring will face during its development. Competition is expected to increase with the number of immature competitors in the infant’s birth cohort. In this study, advantage is taken of troop-size differences and a n &year seasonal birth peak to examine the possible association of birth-cohort size with mortality. An original study troop of 106-130 individuals split into three parts in late 1978 and early 1979. Recent troop sizes averaged 65, 65, and 20. The average birth-cohort size was considerably larger pre-split than post-split. Similarly, average birth-cohort size was largest in the season of a birth peak. Mikumi baboons gave birth throughout the year and did not have a yearly birth season such as that described for Japanese macaques [Lancaster & Lee, 19651; however, when all births were pooled over 8 years, more births occurred in one season than in others. Across those 8 years, the time of birth can be related to the birth peak, and the season of a n immature’s birth in relation to the birth peak can be used in a n examination of the possible association between survival and the immature’s time of birth within its cohort. Although social and ecological conditions are expected to influence immature mortality, their effects are not necessarily separate or additive. Given the complexities of baboon social and ecological life, a network of interactions is more likely. These interactions should cast light upon how and why patterns of reproduction, social stress, and immature survivorship may be related in complex social mammals. METHODS Procedures Mortality patterns were examined from birth through the fourth year of life for 164 immatures born during the 8 years from 1975 to 1982. Four years was chosen as the cut-off because almost no mortality occurred between age four and adulthood. In some cases, separate analyses were also conducted for age cut-offs of 1 or 2 years to determine if ecological and social pressures affecting mortality of younger individu- Mortality of Immature Baboons / 201 T T-I- T -r z 2 a w 5 0 I NOV ' DEC I JAN I FEi3 I MAR APR ' MONTH OF RAINY SEASON MAY JUN Fig. 1. Mean rainfall in inches (+SEM) for November, June, and the six rainy-season months from December to May for the eight-year period from 1975 to 1982. Mean rainfall was negligible during dryseason months not shown [Norton et al., 19871. als differ from those affecting older ones. Two years was chosen as a cut-off because all individuals were fully weaned by this time [Rhine et al., 19851. Mortality was examined for 0-1-year-olds because 53% of immature deaths occurred during that age period. The data are from part of the long-term records kept on Mikumi baboons since 1975. From that year, immatures of known date of birth and all adult animals were individually identified, as were many of the older immatures. From November 1976, all troop members were individually identified. They were well habituated for closeup observation even in conditions of heavy cover [Rhine et al., 1979;Fig. 11. The censusing method used to obtain demographic information was described by Rhine et al. . Briefly, in the morning, at and near the sleeping site, an observer walked through the troop finding and identifying animals. Animals not found in the morning were sought throughout the day with the help of other observers, including those obtaining censuses in the process of recording progressions of moving troops [e.g., Rhine & Westlund, 19811. A date of birth was usually established to within plus or minus 3 days, and the possible error was always less than a month. Immatures that disappeared before the end of their fourth year of life were presumed to be dead. From the time when all animals were individually identified and well habituated, no unknown female or immature male immigrated into any of the Mikumi study troops. All other troops encountered were checked whenever possible to determine if any study animals had transferred into them. Study-troop females and males of 4 years or younger have never been observed in other troops. In contrast, numerous subadult and adult males from the study troops were observed in others. The age at death was calculated as the number of days from the first observation of the animal after its birth until the last day the animal was observed in the troop. In some cases, there was a day or more gap in observation days before the animal 202 I Rhine et al. was first observed, or immediately after the animal was last observed. Under these circumstances, the number of missed observation days in each gap was divided in half and added onto the total number of days the animal was known to be alive in the troop. Maternal dominance, which was quite stable among Amboseli yellow baboons [Hausfater et al., 19821, was estimated from the outcomes of one-on-one supplant data collected in 1975 and 1976 [Rasmussen, 19801,1977 [Klein, 19831, and 1979 and 1982 by Wasser. As occurred also in the Amboseli population [Samuels et al., 19871, the dominance ranks of some adult females changed over time. In such cases, the maternal rank associated with each offspring was that determined from data that were collected closest in time to the offspring’s birth. Dominance ranks were assigned by quartile within troops to allow comparisons of the effects of dominance rank in the pre-split and post-split conditions, irrespective of troop sizes. Cox Model The Cox Proportional Hazards Regression Model [Cox, 1972; Cox & Oakes, 19841 was used for survival analyses. This time-dependent, log-linear model is employed in epidemiological and medical research to predict the effect of a condition on a subject’s time to death, taking into account unique aspects of the subject such as age, sex, and socioeconomic status. The model is also suitable for survivorship analyses of natural populations where the conditions are ecologic and demographic. It utilizes data on all subjects by censoring data from individuals who lived beyond a specified cut-off time, such as survival to 4 years. In the Cox analysis, a regression coefficient reflects the degree to which a covariate increases the probability of mortality before a given time. Specifically, the 6o r W 3 a W [r 12-2 3-5 6-8 9-1 I MONTH OF BIRTH Fig. 2. Frequency of births by month of birth: December to February (early wet season), March to May (late wet season), June to August (early dry season), and September to November (late dry season). Mortality of Immature Baboons / 203 50 40 0 = 20 / 12-2 0 BEFORE 2 Y R S . I I 3-5 6-8 I 9-1 I MONTH OF BIRTH Fig, 3. Percent mortality before two or four years of age by months of birth December to February (early wet season), March to May (late wet season), June to August (early dry season), and September to November (late dry season). regression coefficient is the log difference in the hazard function per unit change in the covariate. The regression coefficient divided by its standard error approximates a z of the normal curve from which the statistical significance of the covariate can be determined. Throughout this paper, statistical significance is at the .05 level or better. Logistic or log-linear regression differs from standard regression by determining the change in probability of an event, rather than the amount of variance accounted for. This probability is reflected by taking e to the exponent of the fi coefficient, expressed as exp@).This expression is interpreted as the multiplicative change in the risk of death caused by one unit increase in a given covariate. In other words, one unit change in covariate i multiplies the probability of death before time x by e to the power of pi. For example, for the number of females in the individual’s birth cohort, an exp(fi)of 1.4 means that an increase of one female in the birth cohort multiplies the individual’s risk of death by 1.4. The covariates entered into one or more of the Cox analyses described below are as follows: rainfall, season of birth, sex of the infant, number of males and females in the infant’s birth cohort, maternal dominance rank, and overall troop size at birth (pre-split-larger; post-split-smaller). RESULTS Ecological Conditions by Season Rainfall. Mikumi has a single rainy season that usually begins in late November and extends into early June. Details of rainfall for 20 years are provided by Norton et al. [1987, Figs. 4 and 51. Figure 1 shows the mean monthly rainfall in 204 I Rhine et al. inches (+SEM) for November through June from 1975 to 1982, the period of the present study; negligible rainfall occurred in the dry-season months (not shown). Based on these data, which are consistent with the 20-year averages, the year was divided into four 3-month seasons as follows: early wet (December-February), late wet (March-May), early dry (June-August), and late dry (September-November). The ecological year was taken to begin with the early wet season. An infant’s birthcohort size was defined as the number of infants born in its ecological year of birth (December-November). Birth peak. When the birth data were pooled across all eight years of this study, a birth peak occurred during the early dry season (Fig. 2). The overall pattern of birth frequencies across the four seasons differs significantly from chance expectancy (x2 = 7.82, df = 3, P < .05), and the largest contribution to this difference is from the peak frequency, which is the only one of the four seasonal frequencies differing significantly from chance (z = 2.71, P < .01). Mortality by season of birth. It might be expected that immature mortality would be lowest for infants born during the birth peak. In fact, mortality was lowest for individuals born in the late wet season, the season just prior to the birth peak, and was highest for those born in the late dry season, the season immediately following the birth peak (Fig. 3). The difference in immature mortality between these two seasons is significant in the Cox analysis for mortality before 1year (z = 2.52, P < .025, exp(0) = 5.0), 2 years (z = 2.47, P < .025, exp(0) = 3.551, and 4 years of age (z = 2.23, P < .05, exp(0) = 2.23). Thus, in comparison with immatures born in the late wet season, those born in the late dry season have a 5.0-fold increase in their probability of mortality before age 1, a 3.55-fold increase before age 2, and a 2.23-fold increase before age 4. Mortality rates of immatures born in the early wet and dry seasons were intermediate and not significantly different from any other season. Food availability. Survivorship is the inverse of the mortality pattern by birth season shown in Figure 3. Survivorship to 4 years corresponds in Figure 4 to two broad measures of food usage-the number of different foods eaten and the number of parts eaten (flower, fruit, pod, etc.) from 185 food species [Norton et al., 19871. The number of foods and parts eaten were highest during the late wet season, the season of birth for which immature survival was also highest. Survival was lowest among immatures born in the late dry season when the number of foods and parts eaten were lowest. Survival to 2 years also closely parallels the feeding curves. Rainfall and mortality patterns. Cumulative rainfall during the ecological year of the infant’s birth was not significantly related to immature mortality, neither was rainfall in the immature’s third ecological year (the 13th through the 24th months after the end of the ecological year of birth). On the other hand, high cumulative rainfall during the immature’s second ecological year, the time when weaning was well along or completed (Rhine et al., 1985), was positively associated with mortality before one (z = 3.08, P < .0005, exp(0) = 1.05), two (z = 2.93, P < .005, exp(P) = 1.09, and 4 (z = 3.86, P < .0005, exp(0) = 1.05) years of age. In a n attempt to determine more precisely the source of these statistically significant findings, the Cox model was applied to rainfall by month during the first full rainy season after the immature’s birth. Except for a brief period in the latter part of November, this is the rainy season of the second ecological year. Only the amount of rainfall in December, the first full month of the rainy season, was significantly associated with immature mortality (z = 2.33, P < .025, exp(0) = 1.2 for all of 1, 2, and 4 years of life). The association between rainfall and mortality is highly significant if December rainfall is entered into the model alone (z = 3.29, P < .0005, exp(0) = 1.24 for mortality before 1 year; z = 3.51, P < .0005, exp(/3) = Mortality of Immature Baboons / 205 TABLE I. Regression Coefficients (R.C.), z's of the Normal Curve, and exp(P) Generated From the Cox Proportional Hazards Regression Model for the Effect of December Rainfall on Mortality as a Function of Immatures' Season of Birth Mortalitv before 1 Yr Seasonofbirth Earlywetseason Late wet season Early dry season Latedry season R.C. z .19 .07 .19 .23 2.21" .61 2.73"" 3.36""" 2 Yr exp(0) R.C. 1.21 1.06 1.21 1.25 .18 .10 .19 .21 z 2.37" 1.06 3.09**" 3.48""" 4 Yr exp(0) R.C. 1.20 1.11 1.21 1.25 .17 .10 .19 .19 z exp((3) 2.69"" 1.35 3.71""" 3.80""" 1.19 1.01 1.21 1.20 * P < .05. **P < .01. ***P < ,001. 1.22 for 2 years; and z = 3.98, P < .0001, exp(/3) = 1.21 for 4 years). In addition, as shown in Table I, December rainfall was significantly associated with mortality of immatures born during all seasons except the late wet season, and the greatest effect was on immatures born during the late dry season. Social Effects and Sex Differences The Cox model was used to explore associations between the immature's probability of mortality and its sex, the number of males and females in its birth cohort, and whether the immature was born before the troop split, when the troop was relatively large, or after the split, when the resulting three troops were smaller. Although none of these variables yielded significant main effects, some key interactions were statistically significant. Interaction terms were calculated by multiplying the independent variables together for continuous variables. When one of the independent variables was discontinuous or categorical (e.g., sex of immature) interaction terms were calculated by creating a new variable for each category of the discontinuous variable (e.g., the interaction terms for sex and birth-cohort size become birth-cohort size on males and birth-cohort size on females). Each of these new variables was set equal to the value of the continuous independent variable for those cases which contain that category; otherwise its value was set to zero. For example, for male immatures, birth-cohort size on males was set equal to birth-cohort size, and birth-cohort size on females was set equal to zero. The opposite would hold for female immatures. These variables were then entered into the regression together. Applying this method via the Cox model, the effects of birth-cohort size on the mortality of each sex were examined. The mortality of a n immature male was unaffected by the number of males or females in its birth cohort; in contrast, female mortality before the fourth birthday significantly increased as the number of females in the birth-cohort increased (z = 2.29, P < .025, exp(/3) = 1.32), and significantly decreased as the number of males in the cohort increased (z = 3.29, P < .001, exp(/3) = .72). Only the latter association was significant for mortality before 2 years of age (z = 2.18, P < .05, exp(/3) = 233). When the same comparisons were made using the Cox model after dividing the data into pre- and post-split conditions, the above effects of birth-cohort size on female mortality remained significant only for animals born in the pre-split condition, when troop size was larger. These results for mortality before 4 years are shown in Table 11. For mortality before 2 years, female mortality again decreased as the number of males in the cohort increased (z = 2.36, P < .025 exp(/3) = 2301, but the effect of female-cohort size on female mortality was no longer significant. 206 I Rhine et al. TABLE 11. Regression Coefficients (R.C.), z's of the Normal Curve, and exp(0) Generated From the Cox Proportional Hazards Regression Model for the Effects of Male and Female Cohort Size on Sex-Specific Mortality Before 4 Years of Age, Pre- and Post-Split Pre-split Variablesa Female immatures No. females in cohort No. males in cohort Male immatures No. males i n cohort No. females i n cohort Post-split R.C. Z exp(0) R.C. Z .28 -.34 2.13* 3.44** 1.32 .71 - .07 - .04 .61 .43 .93 .96 - .05 .71 1.16 .95 1.08 .13 -.13 1.77 1.29 1.44 .07 exp(0) .88 "The regression coefficients for sex and pre- versus post-split as main effects are not significant. *P < .05. **P < .001. Ecological and Social Factors Combined The above association of social variables with mortality of immature females appears to be stronger for mortality before 4 years than before 2 years, This trend became more apparent when ecological and social variables were combined into a single Cox analysis of mortality before 4 years. In this combined 4-year analysis, December rainfall was no longer significantly associated with immature mortality; however, it was significantly associated with mortality before 2 years (z = 2.40, p < .025, exp(0) = 1.24). In contrast, the birth-cohort size variables previously associated with immature mortality (Table 11)remained statistically significant for mortality to 4 years (Table 110,but were not significant for mortality to 2 years. These results suggest that the effects of rainfall have their greatest impact on immature mortality up to the end of weaning, whereas social factors have their greatest impact after weaning. TABLE 111. Regression Coefficients (R.C.), z's of the Normal Curve, and exp@)Generated From the Cox Proportional Hazards Regression Model for Female Immatures Re-Split, Showing the Effect of Maternal Rank by Season of Birth and of Birth-Cohort Size Upon Mortality Before 4 Years of Age Variablesa Maternal rankb Early wet season Late wet season Early dry season Late dry season No. females in cohort No. males in cohort R.C. - .24 1.20 .91 1.40 .34 - .64 2 exp(P) .53 1.28 2.16" 2.73"" 3.46"* * 3.94 * * * .79 3.33 2.48 4.02 1.40 .53 "The birth-season variables for immature females post-split and for immature males pre- and post-split were included in the combined regression analysis (they were not significant). "To account for different troop sizes, highest rank is membership in the first quartile of an animal's own troop and lowest rank is membership in the fourth. * P < .05. + a p < ,01, *:*:rp < ,001. Mortality of Immature Baboons / 207 The association of maternal rank with immature mortality before 4 years was also examined in the combined Cox model. Maternal rank was significantly associated with immature mortality, not as a main effect, but as an interaction with the pre-split season of birth. As shown in Table 111, maternal rank was associated with the mortality of immature females born in the early and late dry seasons. Females born to low-ranking mothers during the season of the birth peak (early dry), and especially those born in the following season (late dry), had a significantly higher probability of mortality than did females born at these times to high-ranking mothers. DISCUSSION Ecological Effects Rainfall and hazards. High mortality was associated with high December rain during the immature’s second ecological year. Brief descriptions will be given of three speculative hypotheses that, individually or in combination, might account for this association of rainfall with mortality. Adequate empirical support of these hypotheses is not available. The first of these hypotheses is that high rainfall early in the rainy season produces scattered rainpools in areas that lack a permanent water supply, making possible increased ranging, which in turn increases the probability of young baboons becoming lost. To test this hypothesis, ranging data by month over several years are needed to determine if mortality and ranging early in the ecological year are greater in years of heavy December rain than in years of light December rain. Immature baboons were seldom known to be sick before they died. With the exception of injured animals, the large majority of mortalities were recorded for individuals present and healthy-appearing one day and gone the next. These observations are consistent with the first hypothesis, but not with the second. The second hypothesis is that high rainfall early in the second ecological year increases the health hazard to immatures through direct exposure [e.g., Dunbar, 19801. Confirming data would show that sickness and mortality are greatest during second ecological years having heavy rain early in the rainy season. Obtaining reliable data on the health of free-ranging immatures without seriously disturbing natural troop behavior is a considerable challenge. If exposure were an important cause of immature mortality, high rainfall during months of the rainy season other than December might also be expected to show some association with immature mortality, which was not the case. The third speculative hypothesis is that high rainfall early in the rainy season accelerates early grass and herbaceous growth, increasing the probability of weanlings becoming lost or being taken by predators, and possibly making food harder to find. To test this hypothesis it is necessary to show that grass growth is accelerated during years of heavy December rain and that increases in the number of lost or preyed-upon immatures accompany such accelerated growth. This is a formidable task. Since data and analyses confirming or disconfirming the three hypotheses are not yet adequately available, and since the three hypotheses are not mutually exclusive, future analyses may show that conditions specified by all three in some degree contribute t o immature mortality. Rainfall and nourishment. It would be surprising if the amount and quality of nutrition available for pregnant or newly lactating females were unrelated to immature mortality. During the dry season at Mikumi, as at Amboseli, there may be a “vitamin or protein deficit in baboons due to reduced availability of fresh green vegetation and perhaps insects as well” [Hausfater, 1976, p. 561. At Mikumi where proteinaceous insects are an important part of the diet [Rhine et al., 19861, 208 I Rhine et al. 70 300 p 200 60 LT 2 LL 0 a r cn az 0 > > LT too 50 3 v) \ 0 0 Z SURVIVORSHIP TO 4YRS. 8 C I 12-2 40 I I 3- 5 6-8 9 MONTHS I75 117 0 0 LL 0 58 0 EARLY WET LATE WET EARLY DRY LATE DRY SEASON Fig, 4. Percent survivorship to four years plotted in relation to the mean number of plant parts and foods eaten per month during three-month seasonal blocks. Food data are from Tables VIII and IX of Norton et al. 119871. insect eating is at its height during the late wet season and is lowest during the late dry season. During the dry season, the browned range is swept by fires, and “prior to the start of the rains a large proportion of the study animals’ range is barren of grass and herbaceous cover and most trees are leafless [see Rhine, et al., 1980, fig. la]” [Norton et al., 19871. Grasses and herbs are among the baboons’ most important food staples [Norton et al., 1987 Table V]. In contrast to the dry season, the range is lush and green by the late wet season during which time, as Mortality of Immature Baboons / 209 Figure 4 indicates, the number of plant-food species and parts available and eaten were greatest. MacArthur and Pianka  predicted that such dietary diversity would be associated with decreased quality and abundance of food. It is difficult to see how their model can explain the close seasonal correspondence for Mikumi baboons between foods eaten and mortality (Fig. 4)unless one makes the dubious argument that food quality and abundance are inversely related to mortality. Given the difference in the character of the vegetation from the dry to the wet season [Norton et al., 19871, it is probable that available food increases in the late wet season, not only in diversity, but also in abundance and quality. MacArthur and Pianka’s predicted relationship makes good sense for animals normally with a limited diet which can be expanded into less salubrious foods during adverse times when their preferred foods are in short supply. But, baboons are eclectic feeders [Norton et al., 19871. Thus, favorable feeding for baboons might occur through selective feeding on only a limited subset of quality foods from among those available, through lesser use of each of many available foods which together provide a quality diet, or more likely, a combination of both [Norton et al., 19871. If the number of food species and parts used per season is associated with the level of nourishment received, which seems likely for Mikumi baboons, then nutrition was highest for mothers of immatures born during the late wet season when survival was also highest, and lowest for mothers of those born in the late dry season when survival was lowest. In the intermediate cases, those born in the remaining two seasons, the early dry and early wet, nutritional conditions were probably less salutary than in the late wet season but less deleterious than in the late dry. Feeding conditions taken together with the possible hazards of rainfall, discussed above, suggest that rainfall patterns produce both benefits and costs, depending upon their timing vis a vis birth and weaning. Sociodemographic Effects In addition to ecological effects, the mortality pattern of Figure 3 may have resulted in part from social pressures associated with the birth peak. Such social pressures were documented by Wasser and Starling [1986,1988],who found rates of aggression among adult females to be highest at the onset of the birth peak. Individuals born during the birth peak in the early dry season had the most comparably-aged competitors also born during that time. Individuals born earlier, during the late wet season when mortality was lowest, did not have this same competitive disadvantage. On the contrary, they had an age advantage in competition with younger cohort mates born during the birth peak a few months later. Individuals born in the season following the birth peak, the late dry season when mortality was highest, seem likely to have had the greatest risk because they had an age disadvantage in competition with cohort mates. In addition, during this period, quality nourishment for pregnant or lactating mothers probably was least available. The sociodemographicresults further suggest that competition was particularly deleterious for immature females born to mothers of relatively low dominance status. The disadvantage of low dominance was most evident for immatures born during the early and late dry seasons (Table III) when birth-cohort competition presumably was highest. Mortality was most marked in females born before the troop split (Table III), when the large troop size is believed to have intensified social competition. Indeed, exacerbation of competitive conditions resulting from large group size may have contributed to the splitting of the troop. Why pre-split competition did not also adversely affect male immature mortality is a mystery remaining to be solved Icf., Dittus, 19791. 210 I Rhine et al. Female Competition The finding of increased mortality with increased cohort size relates to a hypothesis proposed by Wasser  to explain attack coalitions of female baboons. He suggested that attack coalitions benefit an attacker’s own offspring at a cost to competitors. Attack coalitions against females infrequently involved males, and were tied to the number of females simultaneously in estrus or in late pregnancy [Wasser, 1983; Wasser & Starling, 1986, 19881, which are related to the size of the forthcoming birth cohort. Moreover, females with new infants often received high levels of aggression from the remaining pregnant females who were observed on two occasions to commit infanticide [Wasser & Starling, 19881. Wasser hypothesized that females could enhance the survivorship of their own infants by suppressing the reproduction of others, as well as by diminishing the vitality of female infants already present in the troop. Female infants are potential long-term competitors of cohort females who, unlike natal males, do not usually disperse. Complex Relationships and Long-term Studies The analyses in this paper show that many variables can contribute to a population’s mortality patterns, but not necessarily as main effects. For example, the effect of a mother’s dominance rank on the mortality of her offspring was only significant for immature females born during or just after the birth peak, which are thought to be competitively disadvantageous times. The large number of variables and the complexity of the analyses necessary to ascertain interactive effects illustrate the value of long-term studies in providing the large sample sizes necessary to detect such effects in nature. CONCLUSIONS 1. Reproductive success of baboons is the outcome of an intricate web of ecological, social, and demographic conditions. Interactions among the ecological and sociodemographic variables of this study were significantly associated with immature mortality. 2. Rainfall patterns, depending upon their timing in relation to birth and weaning, were both costly and beneficial to immatures. Immatures born during the late wet season, when nutrition for their recently pregnant and newly lactating mothers was probably at its height, received the benefits of rainfall. The costs arose from increased mortality of weaning or recently-weaned young as a function of above-average rainfall early in their first full wet season. It was speculated that heavy rainfall early in the rainy season could adversely affect health or increase ranging and early grass growth, which in turn could increase the weanlings’ chances of becoming lost or preyed upon. 3. In addition t o rainfall patterns, analyses of sociodemographic variables suggest that immature mortality was a negative outcome of female-femalecompetition. The mortality of female immatures increased as a function of troop size at birth and of the number of females in the birth cohort. There was a greater chance of mortality for female infants of low ranking mothers if the infants were born during or especially just after a birth peak, when competition was thought to be high. ACKNOWLEDGMENTS Research at Mikumi National Park evolved from planning started in 1971 at the Gombe Stream Research Centre whose Director, J. Goodall, introduced us to Mikumi and gave invaluable advice and support. The research described herein is part of a long-term effort to which many individuals have contributed [Rhine, 19861, including J. Rogers who assisted in the initial rainfall investigations. Over the years, Mikumi research has been supported by grants to the senior author from the Mortality of Immature Baboons / 211 National Institute of Mental Health, the National Science Foundation, the Harry Frank Guggenheim Foundation, the Leakey Foundation, Biomedical Science Support grants to the University of California (Riverside), Intramural and Intercampus Opportunity grants (Riverside campus), and the Center for Social and Behavioral Sciences Research (Riverside campus). Grants to Wasser from the Harry Frank Guggenheim Foundation have helped to support Mikumi operations. We thank M. Kahn, A. Raftery, and G. Sackett of the University of Washington for their advice in the use and interpretation of the Cox model. For the opportunity to work a t Mikumi, we are grateful to the people and leaders of Tanzania, particularly to the late Hon. D. Bryceson and to the leadership of the Serengeti Wildlife Research Institute, the Tanzanian National Scientific Research Council, and the Tanzanian National Parks. The manuscript benefitted from critical and editorial comments from M. Costello, R. Cox, D. Rhine, G. Sackett, and L. Wasser. In this paper, data from part of one subset of a larger, long-term data base were made available to Samuel K. Wasser for tests of hypotheses about female competition. He is primarily responsible for the following contained herein: Cox analyses, associated definitions, and interpretations of the effects of December rain and female competition on immature mortality. Additional and more complete analyses will be forthcoming from Ramon J. Rhine and Guy W. Norton. REFERENCES Altmann, J.; Hausfater, G.; Altmann, S.A. Demography of Amboseli baboons, 19631983. AMERICAN JOURNAL OF PRIMATOLOGY 8:113-125,1985. Cox, D.R. Regression models and life tables. JOURNAL OF THE ROYAL STATISTICAL SOCIETY 34 (series B):187-220,1972. Cox, D.R.; Oakes, J. ANALYSIS OF SURVIVAL DATA. Green Monograph, Chapman and Hall, 1984. Dittus, W.P.J. The evolution of behaviors regulating density and age-specific sex ratios in a primate population. BEHAVIOUR 49:2651302,197Q. Dunbar. R.I.M. DemoeraDhic and life history variables of a popul&in of gelada baboons (Theropithecus gelada). JOURNAL OF ANIMAL ECOLOGY 49:485-506,1980. Hausfater, G. Predatory behavior of yellow baboons. BEHAVIOUR 56:44-68,1976. Hausfater, G.; Altmann, J.; Altmann, S. Long-term consistency of dominance relations among female baboons (Papio cynocephalus). SCIENCE 217:752-755, 1982. Hrdy, S.B. Sources of variance in the reproductive success of female primates. PROBLEMI ATTUALI DI SCIENZA E DI CULTURA 259:191-203,1986. Klein, H.D. PATERNAL CARE AND KIN SELECTION IN YELLOW BABOONS (PAP I 0 CYNOCEPHALUS). Ph.D. Dissertation, University of Washington, Seattle, 1983. Lancaster, J.B.; Lee, R.B. The annual reproductive cycle in monkeys and apes. Pp.486513 in PRIMATE BEHAVIOR. I. DeVore, ed. New York, Holt, Rinehart and Winston, 1965. Malik, I.; Seth, P.K.; Southwick, C.H. Population growth of free-ranging rhesus monkeys at Tughaqabad. AMERICAN JOURNAL OF PRIMATOLOGY 7~311-321, 1984. Norton, G.W.; Rhine, R.J.; Wynn, G.W.; Wynn, R.D. Baboon diet: A five year study of stability and variability in the plant feeding and habitat of the yellow baboons (Papio cynocephulus) at Mikumi National Park, Tanzania. FOLIA PRIMATOLOGICA 485’8-120, 1987. Rasmussen, K.L.R. CONSORT BEHAVIOUR AND MATE SELECTION IN YELLOW BABOONS (PAPZO CYNOCEPHALUS). Ph.D. Dissertation, Cambridge, University of Cambridge, 1980. Rhine, R.J. Ten years of cooperative research at Mikumi National Park. Pp 13-22 in ONTOGENY, COGNITION AND SOCIAL BEHAVIOR OF PRIMATES. J.G. Else; P.C. Lee eds. Cambridge, Cambridge University Press, 1986. Rhine, R.J.; Westlund, B.J. Adult male positioning in baboon progressions: Order and chaos revisited. FOLIA PRIMATOLOGICA 35:77-116, 1981. Rhine, R.J.; Forthman, D.L.; Stillwell-Barnes, R.; Westlund, B.J.; Westlund, H.D. Movement patterns of yellow baboons (Pupio cynocephalus): The location of subadult males. FOLIA PRIMATOLOGICA 32:241-251, 1979. Rhine, R.J.; Norton, G.W.; Roertgen, W.J.; Klein, H.D. The brief survival of free-ranging baboon infants (Papio cynocephulus) after separation from their mothers. INTERNATIONAL JOURNAL OF PRIMATOLOGY 1:401-409,1980. 212 I Rhine et al. Rhine, R.J.; Norton, G.W.; Westlund, B.J. The Pp. 349-390 in SOCIAL BEHAVIOR OF waning of dependence in infant free-rang- FEMALE VERTEBRATES. S.K. Wasser, ing yellow baboons (Pupio cynocephulus) at ed. New York, Academic Press, 1983. Mikumi National Park, Tanzania. AMERI- Wasser, S.K.; Starling, A.K. Reproductive CAN JOURNAL OF PRIMATOLOGY competitive competition among female yel7~213-228,1984. low baboons. Pp.343-354 i n ONTOGENY, Rhine, R.J.; Norton, G.W.; Wynn, G.M.; COGNITION AND SOCIAL BEHAVIOR Wynn, R.D. Weaning of free-ranging infant OF PRIMATES. J.G. Else; P.C. Lee, eds. baboons (Papio cynocephalus) as indicated Cambridge, Cambridge University Press, by one-zero and instantaneous sampling of 1986. feeding. INTERNATIONAL JOURNAL OF Wasser, S.K.; starling, A.K. Proximate and PRIMATOLOGY 6:491-499,1985. ultimate causes of reproductive suppression Samuels, A.; Silk, J.B.; Altmann, J. Conti- among female yellow baboons a t Mikumi nuity and change in dominance relations National Park, Tanzania. AMERICAN among female baboons. ANIMAL BEHAV- JOURNAL OF PRIMATOLOGY 6:97-121, IOUR 35:785-793,1987. 1988. Wasser, S.K. Reproductive competition and cooperation among female yellow baboons.