Foraging Decisions Among Ach Hunter-Gatherers: New Data and Implications for Optimal Foraging Models Kim Hill, Hillard Kaplan?, Kristen Hawkes, and A. Magdalena Hurtado Department of Anthropology, University of Utah and the "? Department of Anthropology, University of New Mexico This article summarizes 5 years of research on resource choice and foraging strategy among Ache foragers in eastern Paraguay. Successes and failures of simple models from optimal foraging theory (OFT) are discussed and revisions are suggested in order to bring the models in line with empirical evidence from the Ache. The following con- clusions emerge: (1) Energetic returns from various alternative resources and foraging strategies is probably the best single predictor of foraging patterns. (2) Nutrient con- straints should be added only when they significantly improve the predictive power of the model. Importance of meat versus vegetable resources may be one important modi- fication based on nutrients that enhances the ability of OFT models to account for empirical reality in human foragers. (3) Men's and women's abilities and foraging patterns differ enough that they should be treated separately in all OFT analyses. (4) Opportunity costs associated with resources that are processed when foraging is not possible may be sufficiently low to predict that high processing time resources will be included in the optimal diet even when their associated return rates (including pro- cessing) are lower than mean foraging returns. (5) When food sharing is extensive and foraging bands include several adult males and females, foragers may not need to modify foraging strategies in other ways in order to reduce the risk of not eating on some days. KEY WORDS: Hunter-gatherers; Foraging models; South America. INTRODUCTION he study of the determinants of resource choice and diet for hu- T mans is important for many reasons. First, there is good evidence of a direct link between nutrition, health, fertility, and child mor- tality in many different societies (e.g., Butz and Habicht 1976; Received April 12, 1985; revised June 23, 1986. Address reprint requests to: Dr. Kim Hill, Department of Anthropology, University of Utah, Salt Lake City, UT 84112. Ethology and Sociobiology 8:1-36 (1987) © Elsevier Science Publishing Co., Inc., 1987 0162-3095/87/$03.50 52 Vanderbilt Ave., New York, New York 10017 2 K, Hill et ai Mosle'~, and Chef 19~4). Since a good deal of lime is .,penl ~r~ fiw,d acquisition and since the outcome of foraging choices can have hnportan! effects on ~urvi, ,:! ~md reprod~cGon, u~lderst;:nding resource choice shoi1!d be important to ~mderstanding the general adaptive patterns ol humans~ Second, many models of hominid-pongid divergence consider dietat? patterns of principal importance in understanding the morphological and behavioral evolution of the hominid lineage (e.g., Dart 1953; Washburn and Lancaster 1968; Jolly 1970: Winterhalder 1981 : Hill 1982: Kurland and Beck .... erman 1985). We may be able to reach a better tmderstanding of the evu- lutionary trajectorv of hominids if we can reconstruct hominid diets am_{ feeding strategies during different periods of time in the past. Third, ever since Steward ( 1936, 1938) some anthropologists have seen the food quest as a very important determinant of social structure and se~- tlement pattern, especially among band level human societies le.g. Birdseli 1953; Damas 1969; Yellen and Harpending 1972; Wilmsen 1973: Wobst 1974: Smith 1981). Indeed, specific hypotheses proposing that resource explov- ration patterns accotmt ior group size, territoriality, movement patterns, and other aspects of social behavior are common in anthropological literature (e.g., Lee 1972; Jochim 1976: King 1976: Winterhalder 1977: Birdseil 1978: Dyson-Hudson and Smith 1978: Durham 1981; Smith 1981; Cashdan 1983: Hill and Hawkes 1983 ). There is a simitar recognition of the {mportance of understanding for- aging decisions in studies of nonhuman organisms. A direct link between nutrition and fitness is suggested by the data on a wide variety of organisms (Gaulin and Konner 1977). Biologists have hypothesized that the character of exploited resources te.g., low or high quality, large or small packages, low or high processing time required) and their distribution in space and time (evenly distributed or patchy, abundant or scarce, predictable or non- predictable, seasonal or nonseasonal) can be used to predict a wide variety of social behaviors fi-om group size and territoriality to movement patterns and reproductive slrategies among nonhuman organisms. These hypothe.- sized causal relationships have led to a number of predictions that are par- tially supported by data on food resources and social organization of a num,. ber of different animals (e.g., Caraco and Wolf 1975; Clutton-Brock and Harvey 1977; Bekoff and Wells 19811; MacDonald 1983; Terborgh 1984). In order to understand human resource choice, diet breadth models derived from optimal foraging theory (OFT) have been increasingly em- ployed by anthropologists (Smith 1983). Optimal foraging theory is an at- tempt to discover rules that predict decisions about which resources to ex- ploit and how to exploit them. These decisions, via the neural mechanisms (shaped by natural selection) that produce them, are assumed to result in the tendency for living organisms to acquire nutrients as efficiently as pos- sine. This is likely m be whenever more food could increase fitness, tbraging exposes the organisms to greater risks than nonforaging, or more time spent in alternative activities could increase fitness. Models derived fiom ()F'~ Hunter-Gatherer Foraging Decisions 3 make specific behavioral predictions about how an organism can be most efficient. These predictions are compared with behavioral data in an attempt to falsify the models or the assumptions built into them. Although it is im- possible to demonstrate that any organism is not foraging optimally, because many assumptions are built into OFT models, specific models can be falsified (Krebs et al. 1983). Optimal foraging theory diet breadth models are based on the premise that the ratio of costs to benefits resultant from exploiting each potential resource will determine whether it is exploited when it is encountered in the environment. The important independent variables in these models are the return rate defined as the benefits per unit time that can be obtained from resources in question, and the amount of time necessary to search for al- ternative resources. Most models are purposely simplified to consider the benefits in energy alone (calories) and thus ignore the importance of other nutrients. This simplification is based on the probable correlation between energy acquisition rates and fitness for most organisms. Such a simplifica- tion, however, leads to several potential problems (see Pyke et al. 1977, or Smith 1983 for discussion). Nevertheless, models based on maximizing en- ergy return ra'~es have been widely applied and tested because they have been shown to explain a good deal of the observed variance in foraging patterns for different organisms (see Krebs et al. 1983 for review of tests on nonhumans, and Smith 1983 for review of anthropological applications). This article summarizes many of the results from the past 5 years of foraging studies with Ache hunter-gatherers in Paraguay. It emphasizes both the successes and difficulties of using OFT models, and discusses some potentially useful modifications. We collected and analyzed data on Ache foraging patterns, specifically with the intention of assessing the utility of OFT diet breadth models. Although the Ache represent only a single eth- nographic case, we have tried to emphasize the generality of the models used with the hope that other researchers will also attempt to test and modify them, or propose explicit alternative explanations for observed foraging patterns. In previous publications on the Ache, we tested some predictions de- rived from models of OFT (e.g., Hawkes et ai. 1982; Hill and Hawkes 1983). In particular, we concentrated on gathering data to test predictions from models concerning resource choice. We demonstrated, for example, that all 16 resources exploited by Ache foragers during a 4-month field period were characterized by returns after encounter (in calories per person hour) higher than overall Ache foraging returns, including search time (Hawkes et al. 1982). This agrees with the prediction of the optimal diet model (McArthur and Pianka 1966; Emlen 1966; Schoener 1971; Charnov 1973, 1976a; Charnov and Orians 1973) that no resource will be exploited that reduces overall foraging returns. The finding we reported would be improbable if energy returns were not constraining Ache foraging decisions. If we assume that there are at least as many unexploited but potentially edible resources in 4 K. Hill et al. the forest of Paraguay as the number we saw exploited, and assume that the: median return rate for all resources is equivalent to the Ache mean foraging return rate, 1 the probability pf obtaining the observed result is 2 × l(} (0.516 ) if resources are exploited without regard to energy returns. Also in agreement with predictions from the optimal diet model was the finding that low-ranked resources lmonkeys and small birds) were not ex- ploited when men hunted with shotguns rather than with bow and arrows (Hill and Hawkes 1983). Shotguns raised overall foraging returns sufficiently that the pursuit of monkeys and birds would have actually lowered their foraging returns. Despite these encouraging results, many questions re- mained. It has correctly been pointed out that these studies did not dem-- onstrate that some other alternative foraging pattern would not produce higher calories per hour than the one we observed. If this were true, some of the resources exploited by the Ache might actually lower their foraging returns, and thus the same data set might not agree with predictions from the optimal diet model. For example, if some high-return resources were ignored by Ache foragers, this would result in a lower mean foraging return rate. The lower mean tbraging return rate would then lead one to predict the inclusion of low-ranked resources in the optimal diet when they should actually be excluded if the higher ranked resources were not ignored. This means that we cannot unequivocally conclude, based on available data, that the Ache forage in such a way as to maximize their foraging return rate. We also made several observations that did not fit the prediction from the optimal diet model that resources should either always be taken upon encounter, or never taken. Although some sampling of nonoptimal resources is expected (Oaten 1977) we could not easily explain why resources in the optimal set were sometimes passed by. The possibility that resources were patchily distributed and should be abandoned before patches were com- pletely depleted (Charnov 1976b) was considered but did not seem to account for many of the observations. These and other questions led to a ~bllow-up study of Ache resource choice that is presented here. BACKGROUND There are several references to the Ache (also called Guayaki) in historical accounts before the 1960s (see Metraux and Baldus 1963; O'Leary 1963), but the first modern ethnographic reports that became widely available are those of Clastres (1968, 1972), who studied two of the four living Ache groups. The data reported here pertain to the northern Ache, who have only come into unarmed contact with outsiders during the past decade (Hill 1983). The traditional range of the northern Ache is an area about 12,000 km 2 i Both of these assumptions are probably quite conservative. There actually appear to be several hundred edible fruits, birds, and insects the Ache foragers ignore. Most are very small and dispersed and could be expected to give very poor returns per time spent to acquire them. Hunter-Gatherer Foraging Decisions 5 between 54-56 degrees west and 24-25 degrees south. During the 10 years prior to their first peaceful contact with outsiders (1960-70) they probably numbered between 600 and 800 persons (Hill 1983). Tropical broadleaf ev- ergreen forest, which the Ache prefer to open grassland, covers much of the area. Rainfall is quite unpredictable from month to month and year to year, although there is a statistically significant wet and dry season. Annual average precipitation is approximately 1600 mm. Fluctuations in temperature are much more regular, with an annual January maximum around 40°C and a July minimum of about - 3°C. Ecology and climate are more fully described in Hill et al. (1984). The northern Ache, who were full-time hunter-gatherers until the mid- 1970s, currently live primarily at an agricultural settlement (Chupa pou) sponsored by a Catholic mission, but continue to forage frequently in the nearby forest (approximately 25% of all days, Hawkes et al. 1985). Although the shift to part-time residence at an agricultural colony has undoubtedly affected some Ache behaviors, there is no a priori reason why foraging decisions on extended forest trips should not be made using the same criteria that were used by the Ache as full-time foragers (Hill 1983; Hill et al. 1984). The study presented here is designed to determine whether current for- aging patterns can be predicted from current ecological parameters. It is likely to represent past Ache foraging patterns before contact only to the extent that the character and values of relevant variables in resource choice have not changed. On observed foraging trips the Ache take a wide variety of animal spe- cies, among which the most important quantitatively are peccaries (Tajassu tajacu and Tajassu pecari), pacas ( Cuniculus paca), coatis (Nasua nasua), armadillos (Dasypus novemcintus), and capuchin monkeys ( Cebus apella). They also exploit numerous plant products, especially of the palm Arecas- trum romanzolfianurn, from which they take the fruit, the heart, and the starch from the trunk. Fruits and honey are also major resources, with in- sects providing a small but consistent component of the diet. We have previously described Ache foraging trips, reporting on diet choice (Hawkes et al. 1982; Hill and Hawkes 1983), the seasonal pattern of food acquisition (Hill et al. 1984), the sharing of food resources (Kaplan et al. 1984; Kaplan and Hill 1985a), and men's and women's time allocation to activities (Hill 1983; Hill et al. 1985; Hurtado et al. 1985; Hurtado 1985). In addition, Jones (1983, 1984) has published ethnoarcheological analyses of Ache camps and faunal refuse. Recent analyses have concentrated on Ache reproductive strategies (Hill and Kaplan 1986), child development (Ka- plan and Dove 1986), and the relationship between hunting ability and re- productive success (Kaplan and Hill 1985b). Because we are particularly concerned with the mix of resources in the Ache diet, as well as the strategies employed to obtain that mix, two points should be clarified. First, no resources acquired on any foraging trip were brought back to the mission colony for sale or trade. Virtually all resources 6 K. Hill et al, exploited were consumed by men. women, and children of the foraging part3 Second, although the Ache diet at the mission colony is characterized b3 more carbohydrate, less meat, and fewer calories than are consumed in the forest, diet compares favorably with most tribal or peasant peoples in South America, and daily per capital meat consumption among the Ache at the mission settlemen~ appears to be higher than that reported for many po~ ulations of native South Americans (Hawkes et al. 1985). METHODS Data on Ache foraging patterns were collected in two separate field sessions in 1980 and 1981-82. Methods of data collection and analysis for the 1980 field session have been previously reported (Hawkes et al. 1982; Hill and Hawkes 1983). New data on foraging behavior were collected on nine for- aging trips out of the Chupa pou mission between October 1981 and April 1982. Foraging groups included men, women, and children and have been described in detail elsewhere (Hill 1983: Kaplan 1983; Hurtado et al. 1984). All resources acquired were weighed with Homs hanging spring scales and calibration after the field session showed less than 1% error. Caloric equiv- alents were determined from published tables or direct laboratory analysis of food samples (see Hill et al. 1984b for details). Edible portion by weight of each resource was determined from between two and ten measurements in the field. Time spent in food procurement and processing was recorded to the nearest minute with an electronic digital stopwatch. Men's foraging time was determined by clocking them out of camp in the morning and back to camp in the evening. We have previously demonstrated, based on focal individual observations, that 87% of such time is indeed devoted to food getting activities (Hill et al. 1984a). Women's foraging time was measured directly. Definitions of foraging activities (i.e., pursuit time, processing time, search, etc. ~a nd other relevant methodological details have been previously published (Hawkes et at. 1982: Hill and Hawkes 1983; Hill et al. 1984). Return rates for all men and a randomly picked subset of women were calculated for this period. This was done by dividing total calories acquired by each individual by the total number of hours spent in food acquisition and processing activities. The time component of this calculation consists of three parts: (a) time spent searching for resources; (b) time spent pursuing or extracting the resources; and (c) time spent processing, butchering, or cooking the resources. These three components of foraging time are referred to as search time, pursuit time, and processing time. Monitoring the daily behavior of individuals using focal person or instantaneous scan sampling techniques led to approximately 611 man foraging days, and 61 woman for- aging days of return rate data. Although the ethnographer's presence might conceivably alter the return rates of focal subjects, a one-way ANOVA revealed no significant differences between focal men's return rates and Hunter-Gatherer Foraging Decisions 7 those for other men not accompanied by an ethnographer on the same day (F = 0.966, p = 0.41). The procedure for calculating overall foraging return rates along with return rates for specific resources has also previously been described (Hawkes et al. 1982; Hill and Hawkes 1983). Overall mean foraging returns were calculated by dividing all calories acquired by the members of the group in question by the sum of all time they spent in food acquisition and pro- cessing. Return-rates upon encounter with individual resources were cal- culated by dividing the total of all calories acquired of each resource by all time spent pursuing (whether successful or not) and processing it. RESULTS The overall mean foraging return rates for adults during the 1981-82 sample period are presented in Table 1. The rates are slightly higher than those presented earlier (Hawkes et al. 1982; Hill and Hawkes 1983) and should supersede the earlier results, as they are based on a larger sample and more accurate monitoring. We have divided foragers and foraging conditions into several categories in order to examine differences in return rates that may lead to different foraging strategies. Men and women differ in their overall return rates and also show dif- ferent rates depending upon specific conditions during the time they were monitored (Table 1). Men's foraging produces a mean of 1339 calories per hour before processing the food that they acquire. Women's foraging pro- duces a mean of 1221 calories per hour before processing. Adding in pro- cessing time changes calculated return rates only slightly (see discussion below) in the Ache case. The above results agree well with those we cal- culated in previous studies (Hawkes et al. 1982; Hill and Hawkes 1983). Although the data suggest that men's foraging is characterized by a higher Table 1. Mean Foraging Returns in Calories per Hour (Person Days Sampleda ) In "No Forest In Near Move with Oct-Dec Jan-July Forest b settlement" Days ''J Process Men 1619 (247) 1118 (364) 1339 (503) 1018 (108) 1344 (80) 1253 (50) Women 826 (30) 1233 (31) 1221 (50) 302 (ll) 2804 (12) 1087 (50) Daily returns for males as a function of age: Males, 14-20 years 1545 (215) calories per day Males, 21-59 years 9240 (503) calories per day Males, 60-75 years 797 (42) calories per day a All but first and last days of foraging trips. b First and last days of foraging trips. Days on which Ache remain in the same camp site for two or more consecutive nights. d These scores are pooled across subjects but no single individuals contribute inordinately to the data set (see Kaplan and Hill 1985b) 8 K. Hill et aL return rate than women's foraging, one observation suggests that this is n<~t the case. Return rates for women on days that camp does not move are nao~: than twice as high as women's return rates on the days when camp doe~, move. This is important to note because we believe that women, unlike me~"~ do not always search for resources when they are walking to a new campsi{~~ (Hurtado et al. 1984). We have observed women to pass numerous resources without exploiting them when they walk, and estimates of palm densities indicate that they must pass dozens each day and yet only exploit some of them. If women's time spent walking is not actually search time, then won:i- en's foraging return rates have been underestimated. On "no move days ~ all walk time is unambiguously spent searching for resources. The women's return rate on "no move days" is therefore probably a better estimate ol their true foraging return rate. The fact that women's true foraging return rate seems to be higher than that for men presents a problem that is discussed in the next section. In addition to the differences between the sexes in foraging return rate we find differences between individuals and across age classes. For the 25 adult men for whom we have more than 10 sample days of foraging data, the mean returns per man range from 446 to 2124 calories per hour. Although some of this variance is undoubtedly due to hunting luck on the days mon-- itored, we have discovered positive and significant correlations between return rates measured for individual men in 1980 and 1982 (Kaplan 1983, p. 100). These differences also correlate with amount of time spent foraging daily (Hill et al. 1985) and reproductive success (Kaplan and Hill 1985b). This implies that there are long-term stable differences in the hunting abilities of different men. In addition, male adolescents and old men have foraging returns that are almost an order of magnitude lower than adult men Isec Table 1). Such differences are likely to affect individual foraging strategies. Women's return rates also vary from individual to individual. Most im- portantly, Ache data suggest that women who were nursing infants during our sample period are characterized by significantly lower foraging return rates (Hurtado 1985). It is also our impression that there is some variance in foraging returns between individual women based on age, size, and per- sonality characteristics. Several conditional factors can be shown to affect the return rates of Ache foragers. Men's rates appear to be higher during the first half of the warm-wet season when honey is abundant (see Hill et al., 1984 tor descrip- tion of seasonal dietary variance). Women's rates on the other hand appear higher in the late warm-wet season when some important fruits are abundant. These differences, however, are not significant (warm-wet daily mean of hourly return vs. cold-dry, t-test p > 0.05 for both). Both sexes have sig- nificantly higher return rates when they are more than one day distant from Hunter-Gatherer Foraging Decisions 9 the mission, indicating resources depletion2 near the permanent settlement (t-test, p < 0.05 for both). The overall fOraging return rate including all food processing time is presented in Table 1 under the column entitled "with process." Although processing time barely affects men's overall returns, adding it in causes women's returns to decrease by about 11%. This is because vegetable re- sources exploited by the Ache require relatively more processing per unit of food value than do game items. In fact, compared to the amount of time spent in pursuit of animal prey, processing time is almost a negligible com- ponent of their total cost. For example, the largest game animal, white-lipped peccaries, require almost 7.5 man hours of pursuit for one successful kill. Singeing off the hair, gutting the animal, and butchering the animal into suitable for pieces cooking takes just under 15 minutes. Although the meat requires several hours to cook, the actual time spent tending it is only a few more minutes, thus the total processing time is just under 20 minutes, or about 4% of the total time cost of a white-lipped peccary. For palm fiber, on the other hand, approximately 30% of the total time cost is processing time. Now, let us examine how well these foraging patterns conform to pre- dictions of optimal diet models. As mentioned above, the optimal diet model (McArthur and Pianka 1966; Emlen 1966; Schoener 1971; Charnov 1973, 1976a; Charnov and Orians 1973) predicts that none of the resources ex- ploited by the Ache should give returns after encounter (in calories per hour) that are lower than overall foraging returns. This prediction is met for all but one of the 26 resources whose returns we were able to measure. That resource is bamboo larvae, which is taken primarily by women but occa- sionally by men. Since the test reported here is based on an entirely new data set, it represents a partially independent (the same people were mon- itored in both studies) replication of findings reported eariler (Hawkes et al. 1982). The original OFT models were kept as simple as possible in order to ensure their generality. When considering modification of these models one faces a tradeoff between maintaining that generality and increasing the pre- cision of the models with respect to their explanatory power in specific cases. The piecemeal approach described by Krebs (1983) allows for both generality and increasing specificity of models. One begins with the simplest model possible and then adds modifying factors one at a time if they significantly reduce the unexplained residual variance that is of concern to the researcher. Using this approach, the first and most important lesson to be learned from the Ache case is that the energy costs and benefits of exploiting alternative resources allow for reasonably accurate predictions concerning which re- 2 This difference could also be partially due to differences in foraging strategies on the first day of trips vs. other days. Ache walk for more hours and appear to be less involved in active search for resources on the first day of trips. 10 K. Hill e~ ai, sources will be exploited in an area. There appear to be hundreds (i1 ~:, thousands) of edible ~esources in the Paraguayan forest as judged by wha~ is eaten by peccaries, rodents, monkeys, and birds. Many of these item~ (such as insects, small fruits, and small birds) would be expected to gi,~c• very low returns if pursued, and are, in fact, never taken. The finding tha~ all but one of the resources we saw taken by the Ache gave caloric retta~as higher than the mean return rate for the age-sex category of Ache tbragers who exploit it is theretorc quite impressive. Nevertheless, some intriguing difficulties remain. ACHE FORAGING STUDIES: INSIGHTS AND PROBLEMS Results from the previous section, along with the difficulties that we en- countered in attempting to operationalize OFT models and analyze the data that we collected, have led us to some simple insights about studying re- source choice and some modifications of the basic OFT models that may help in predicting human resource use. Most of these modifications were clearly anticipated by the proponents of the original OFT models (e.g., Pyke et al. 1977), all have been discussed in the abstract by recent theoretical works on foraging theory (e.g., Stephens and Krebs 1986), and many have been recognized as general problems with applying OFT to the human case (Smith 1983), At this point it is important to clarify that many of the ideas in the following discussion should be regarded as hypotheses that require further evaluation. Explanations consistent with evolutionary theory and OFT are offered for observations that seem, at least initially, to contradict OFT models. Particularly we were bothered by the observation that the caloric returns from adult male hunting seem to be considerably lower than the caloric return rate that characterizes women's palm collecting activities. This implies that men were choosing a low-return foraging strategy when a higher- return alternative strategy was possible. This observation forced us to deal directly with nutrient constraints in the analysis of data even though we had not anticipated this problem in the data collection phase. Because the ob- servations on Ache foraging patterns were used to derive some of the ex~ planations we propose, the same data cannot be used to support our ideas. Most of the discussion below is an example of the constant interplay between theoretically derived initial expectations, empirical observations, and new interpretations of theory with new expectations. We offer these interpre- tations with the hope of stimulating the collection of new field data among other human foragers. Nutrient Preferences Although the above results are consistent with the basic predictions of the optimal diet model, there is good reason to believe that an alternative for-