1. Introduction

Evolutionary anthropologists propose various adaptive strategies for meat sharing. The payoffs for these strategies vary according to the cost–benefit structures created by different environmental and social contexts (Winterhalder, 1996). The goals of this paper are to (1) contribute a case study on meat sharing from the Ecuadorian Amazon, (2) identify environmental and social variables that influence people's decisions to share meat in this case, and (3) argue that patterns of meat sharing may be best understood as outcomes of multiple adaptive strategies, including strategies to recruit and maintain coalitional support.

Conambo is a community of around 185 Achuar, Quichua, and Zapara speakers who make a living in the Ecuadorian Amazon as horticultural foragers.1 Conambo provides a case study where food resources are relatively abundant, hunters exercise control over the distribution of game, settlement is dispersed, and postmarital residence is matrilocal. Alliances and coalitional membership can be, and often are, volatile, and feuding and male homicide rates are historically high (Bowser, 2000, Bowser, 2002, Bowser & Patton, 2004, Patton, 1996, Patton, 2000, Patton, 2004). In Conambo, the management of political alliances is critical to personal and household well-being, arguably as important as the management of subsistence resources, if not more so.

In this paper, I propose a coalitional support hypothesis to account for meat transfers in Conambo. Additionally, I examine the most commonly cited evolutionary explanations for meat sharing among hunter–gatherers and horticultural–foragers: kin selection, reciprocity, show-off hypothesis, tolerated theft, and costly signaling. I report evidence of coalitional support, kin selection, and reciprocity influences on meat sharing, with equivocal support for the other hypotheses. I argue that meat transfers may be studied within the context of an evolved coalitional psychology, reflecting a strategy for recruiting and maintaining political allies, and that some social and environmental contexts favor the transfer of meat for alliance considerations. I conclude by examining the environmental and social context of meat transfers in Conambo and contrast it with the contexts of previous studies to account for differences in patterns of meat sharing.

2. Life in Conambo

Because evolutionary arguments have focused on hunters' efforts within the contexts of subsistence needs, mating effort, parenting effort, and social norms for cooperation, this section provides relevant information from Conambo on environmental richness, settlement, marriage and fertility patterns, hunting practices, and norms for meat sharing. To assess the relative importance of coalitional support in Conambo, information is provided concerning political life and coalitional interactions.

The community of Conambo is located in the Ecuadorian Amazon along the Conambo River, which flows into the Rio Tigre at the border with Peru. The data in this study were collected during ethnographic fieldwork in 1992 and 1993 and informed by subsequent fieldwork in 1998 and 2002. The region is relatively remote; there are no roads to Conambo, and no markets there. People are economically self-sufficient and acquire food through hunting, fishing, gathering, and slash-and-burn horticulture. Men have the primary responsibility for hunting, and women, for gardening. Traditionally, hunting was done with blowgun and spear. Today, men continue to use blowguns, but muzzle loading shotguns have replaced the spear for hunting ground-dwelling animals such as tapir, deer, peccaries, armadillos, pacas, and agoutis. Monkeys are the most common game consumed; monkeys are taken by both blowgun and shotgun.

Population density is low. Conambo is the largest of eight communities located within the Zaparo territory, where about 400 people hold title to 250,000 ha, a population density of 0.16 people/km2. Settlement in the community is dispersed over 3000 m along both riverbanks. With few exceptions, there is no line-of-sight between houses. Most houses have private trails that connect them to a system of main trails, and a person can travel from one end of Conambo to the other without going through the majority of household compounds. According to Descola (1994, p. 207), individual men in the region have control over hunting territories of over 40 km2. Hunters generally radiate out from their house sites along trails that connect to their wives' gardens and their hunting trails. Individual hunting catchments tend not to overlap with the exception of men whose households are geographically close and typically have consanguine and affinal kinship relationships and strong political loyalties.

Hunters exercise control over the distribution of game, and generalized sharing is not the norm in Conambo. Men go hunting every few days, sometimes with their wives, most often by themselves, and sometimes in pairs or small groups of men. Because houses are widely dispersed and hunting trails emanate from house sites directly to the forest, successful hunters can return home without being seen by members of other households. The decision to share or hoard meat is primarily made by the hunter and his wife, although meat transfers by a young coresident son-in-law may be directed by his father-in-law in this matrilocal society, and meat may be requested by a member of another household who knows the meat has been procured. Meat giving in Conambo is targeted to specific households. Meat is often shared by the hunter or his wife by personally delivering a smoked or uncooked portion of game (often a limb or section of ribs), having a child deliver it, inviting a family to come visit and eat the fully processed meat (usually in a soup) in the hunter's home, or inviting people to help transport portions of a large animal from the kill site.

Subsistence resources are plentiful in Conambo. The hunting is very good, and meat is generally consumed daily. The study of Descola (1994, p. 210) of diet in the region conducted from 1976 to 1978 indicates that adults consume, on average, 3400 cal and 105 g of protein per day. The latter is over three times the amount of protein considered adequate by the World Health Organization (30 g/day). During the time that meat sharing data for this study were collected, I measured the height and weight of all residents in Conambo. Data on body mass index (BMI) for adults in Conambo indicate that people are well nourished. Of the 32 adult women (18 years of age and older) in the sample, the average BMI was 24, which is at the high end of the normal range (BMI standards taken from the Centers for Disease Control website 6/28/04, www.cdc.gov). None were considered underweight (below a BMI of 18.5), and almost a third were overweight (10 had BMIs of 25 or above). The average BMI for 33 adult men was 23, with no men underweight and 10 overweight (see also Orr, Dufour, & Patton, 2001). It appears safe to conclude that the people of Conambo have a more than adequate diet.

Marriages in Conambo are monogamous and relatively stable. Of the 32 men in the community, only five had been married more than once. Four of these men remarried after a single divorce. One man, a shaman, married five times, divorced three times, was widowed once, and fathered 19 children, the most of any man at that time in Conambo. Of the marriages recorded for male heads-of-household living in Conambo in 1992–1993, 82% were still intact. The divorce rate of 18% for men drops to 12% after excluding the shaman. Divorce rates are comparable for women; only four of the wives in this sample of 32 couples had been divorced. Five had been widowed. With monogamy and low divorce rates, men's and women's reproductive success are comparable.

Fertility in Conambo is high, with short birthing intervals and low childhood mortality. On average, women have a child every 2.4 years during their reproductive years (Patton, 1996, pp. 95–97). Of the 10 children born in the 9 months of fieldwork, all but 1 was alive a decade later. Male lifetime fertility in Conambo is comparable with that in the Yanomamö. Among the 11 older men in Conambo who have stopped having children, the average number of children is nine. By comparison, the average lifetime fertility for Yanomamö men is 9.7 (Chagnon, 1979) and about 4 for Ache men (Hill & Hurtado, 1996, p. 411). Excluding the 2 Conambo men who never had children, the average male lifetime fertility is 11. Female lifetime fertility is also high. Among the 12 women in Conambo who have stopped having children, the average number of children is eight. One woman never had children; excluding her from the sample increases women's average lifetime fertility to 9, about twice that of the lifetime fertility of Yanomamö and Ache women [4.8 children for Yanomamö women (Chagnon, 1979), 4.2 to 5 for Ache women (Hill & Hurtado, 1996, p. 411)]. Men and women in all three of these Amazonian populations have much higher lifetime fertility than !Kung men and women do (Hill & Hurtado, 1996).

The current parity in reproductive success between men and women in Conambo is a recent phenomenon. Historically, high rates of male homicide skewed sex ratios in favor of women, and polygamy was common. Harner (1972, p. 80), who conducted fieldwork in 1956–1957 among the nearby Shuar (Jívaro), reports a sex ratio of 2:1 favoring women. He states that “the most common number of wives for a man to have is two, one, or three in that order.” In contrast, in 1992–1993, there was a near 1:1 sex ratio in Conambo, with 95 females and 90 males. These data on marriage and fertility patterns in Conambo suggest that, with the reduction of intensive warfare, there was a corresponding reduction in the variance of men's reproductive success.

Postmarital residence is commonly matrilocal. Men often marry into Conambo from outside the community and therefore have few coresidential kin compared with men in patrilocal societies. Newly married couples typically live within the bride's parent's household, later building a house nearby. This pattern tends to create matrilocal neighborhoods. However, having a brother in another community facilitates opportunities for visiting and finding a marriage partner. Thus, despite matrilocality, there are a number of sets of brothers who have married into Conambo.

Conambo is divided into two ethnic/political groups, which structures many of the social interactions and conflicts in the community. Of the 24 nuclear and extended households represented in this study, 14 were affiliated with the Achuar faction, and 10 with the Quichua faction. This political divide also has spatial correlates; the Achuar-allied households live upriver from the center of the community, and the Quichua-allied households live downriver. These ethnic labels denoting coalitional membership obscure a high degree of interethnic marriage and political realignments after conflicts. Sixty percent (6 of 10) of the Quichua coalition households were headed by an ethnically Achuar man (4) or woman (3); in one case, both household heads were ethnically Achuar. About a third (4 of 14) of the Achuar coalition's households were headed by an ethnically Quichua man (2) or woman (2). Some individuals in Conambo identified themselves as either Achuar or Quichua, depending on the context, the question, or who was asking. The higher proportion of ethnic Achuar on the Quichua side of the political divide can be traced historically to two separate incidences of Achuar–Achuar conflict that ended in homicides, after which the widows and their families relocated across the political divide, and then their children intermarried (for a more extensive discussion of the coalitional and ethnic divisions in Conambo, see Bowser, 2000, Bowser, 2002 and Patton, 1996, Patton, 2000, Patton, 2004).

Genealogical data for the ancestors of people in Conambo indicate that half of the men died in homicides (Patton, 1996, Patton, 2000). This is consistent with male homicide rates reported earlier for this area of the Ecuadorian Amazon (Descola, 1996, Kelekna, 1981, Ross, 1976, Ross, 1984, Ross, 1988) but is high compared with other tribal societies studied during active times of war, which typically have male homicide rates that range from 20% to 30% (Chagnon, 1996). Currently, homicide rates are much lower in Conambo and throughout the region. However, coalitional alliances are inherently unstable, conflicts and open hostilities between households are not infrequent, and tensions between the two main political coalitions are typically high. Residential mobility is a common response to shifting alliances within coalitions and across the main coalitional divide (Bowser & Patton, 2004).

In summary, Conambo is an ethnographic context where settlement is decentralized and widely dispersed, hunters exercise control over meat distribution; subsistence resources are plentiful, marriages are monogamous and stable, male and female fertility is high, postmarital residence is matrilocal, and alliances are unstable. Accordingly, this case study provides a point of contrast to previous studies of meat sharing and an opportunity to broaden the study of how environmental and social constraints differentially influence strategies for the sharing of meat.

3. Methods and results

Testing the six evolutionary explanations for meat transfers in this study requires two qualitatively different data sets. The logic of kin selection, reciprocity, tolerated theft, and coalitional support arguments implicates meat sharing as an adaptive strategy designed to weight the costs and benefits associated with dyadic exchanges. Therefore, meat sharing between households is conditioned on specific relationships and the marginal valuation of meat transfers between the exchanging households. The appropriate tests of these explanations require data on the dyadic relationships between households: degrees of relatedness, meat giving and receiving, spatial distance, and the strength of coalitional loyalties. In contrast, the show-off hypothesis and costly signaling arguments posit that the benefits accrued to better hunters are not tied directly to dyadic exchanges but are derived from the greater access to group resources that higher status provides or from better mating opportunities. The tests for these arguments require data concerning attributes of the hunters (hunting skill, status, and reproductive success). For this reason, this section is divided into two subsections, each dealing with the methods and results of the two data sets and the corresponding theoretical tests.

3.1. Dyadic relationships: kin selection, reciprocity, tolerated theft, and coalitional support

To test for independent and multivariate influences of kinship, reciprocity, tolerated theft, and coalitional support on meat transfers between households, a household-by-household database was constructed with the following variables: meat given, meat received, genetic relationship, spatial distance, and alliance strength.

The meat transfer data were obtained from informants and represent their reports as to which hunters outside of the informant's household had or had not provided meat to the household. One representative from each of the 24 households in Conambo was shown photographs of the male heads of the other 23 households. As each photo was presented, each representative was asked “has this man or anyone from his household ever given meat to your household?” This question minimizes the potential for informants to exaggerate their own generosity. Informants were not asked to identify households to which they had given meat, nor how much they had contributed to other households. Once contributors were identified, informants were asked to rank the men in order of who had given the most meat to their household. Of the 24 informants, 11 were the senior man of house, 9 were the senior woman, and 4 were the next oldest adult in the household (2 sons, 1 daughter, and 1 son-in-law).

Because different households received meat from a different number of donors, these rankings were scaled. The man who gave the most often to a household received a score of 1; those who did not contribute were assigned a score of 0; and the other men were scaled between 1 and 0 according to their ranking. This allowed for comparison of meat-giving rankings across men. These ranking data concern the relative generosity of each man, rather than data on individual transfer events. The data do not permit primary and secondary sharing to be distinguished. However, the majority of meat transfers observed were primary transfers.

The data in this study are based solely on informants' perceptions of meat transferred into their households. Although perceptions may be biased, studies examining informant perceptions of meat transfers and hunting success rankings have reported significant correspondence between perceptions and observed data. Among the Hadza, Hawkes (personal communication) found a strong correspondence between observed meat transfers and informants' perceptions of meat transfers, and hunting skill rankings of Ache men by informants also correspond well with observed data on hunting success (Hill & Hurtado, 1996). These studies find that perceptions are likely to be accurate assessments of actual transfers, as well as meat debt associated with asymmetric transfers.

An important advantage to using perception data over observed data is that informant perceptions are less temporally restricted than the period of fieldwork. People's memories of meat transfers extend across seasons and years, and their perceptions represent the vantage from which people decide their future actions. Therefore, long-term perceptions of meat transfers may provide a better foundation from which to examine the decision-making processes associated with meat transfers than short-term observations of meat transfers do.

To obtain alliance strength data, 38 informants (25 men and 13 women) were asked to make judgments concerning the alliances of all 33 married men in Conambo (see Patton, 1996, Patton, 2000). These data were collected during the 2 months prior to the meat-sharing task. Some adults were not included as informants in the study, 1 due to bad eyesight, 2 were impatient with the task, 1 gave random responses, 16 were absent from the community when the task was conducted, 3 were too old, and 1 refused to do the task.

Informants were asked to judge relative alliance strengths by choosing the two of three men depicted in photos who were more likely to form a coalition against the third man if a conflict involving the three men were to occur. Photos of all the men in Conambo were presented in all possible triad combinations. Each informant was shown 33 randomized triads, the equivalent of making relative alliance strength judgments for 99 dyadic coalitions. In total, informants made 3762 alliance judgments concerning the set of all dyadic relationships among the 33 men. Each man was represented in 228 possible dyadic alliances and appeared, on average, 7.13 times in the same triad with each of the other 32 men in the sample. The alliance strength between men was calculated as the percentage of times the two men were judged to have the strongest alliance in the triads in which they both appeared (i.e., times paired divided by the total number of pairing opportunities).

For this analysis, I use alliance data for the 24 male heads-of-household, the primary persons responsible for hunting and provisioning meat for their households. Data on kin relatedness were compiled from genealogical interviews with all adult men and women in Conambo. Spatial distance between households was measured as straight-line distance in meters, using a GPS instrument. The data on meat giving, alliance strength, kinship, and spatial distance were compiled into a database including variables for 276 dyadic pairs of households in Conambo.

3.1.1. Kin selection

These data indicate that kinship is an important determinant of adaptive motivation for the transfer of meat between households in Conambo. Meat was more likely to be transferred to the households of relatives than to more distant relatives or nonkin. The hunter's genetic relatedness with the heads of receiving households accounted for 11.3% of the variation in meat transfers (Pearson's r=.34, p<.001, N=276). Nonparametric correlations between hunter's relatedness and meat transfers were also highly significant (Spearman's ρ=0.29, p<.001). The hunter's wife's genetic relatedness with the heads of the receiving households accounted for over twice that amount of variation in meat transfers (r=.50, p<.001; ρ=0.50, p<.001; N=276). In combination, summing the genetic relatedness of husbands and wives between households, kinship accounted for 34% of the variation in meat transfers (r=.58, p<.001, N=276), and nonparametric correlations were highly significant (ρ=0.52, p<.001).

There was no meat sharing between 130 of the 276 dyadic pairs of households. The average degree of relatedness between men and the heads of nonsharing households was 0.000, and the average relatedness for their wives was 0.018. Among sharing households, the average relatedness was 0.037 for men and 0.127 for women. Overall, the average relatedness (husbands and wives) between sharing households was 7.3 times greater than the average relatedness between nonsharing households (0.134 to 0.018), a highly significant difference (t=7.26, df=274, p<.001). Among related households, genetic relatedness was highly correlated with degree of meat sharing, according to informant's rankings (r=.52, p<.001; ρ=0.55, p<.001; n=88 dyadic pairs of households for which Hamilton's r was greater than 0).

3.1.2. Reciprocity

There was a significant positive relationship between meat giving and receiving, suggesting that reciprocity-in-kind was also a reason for sharing meat in Conambo. Hunters were more likely to give meat to households from which they received meat (r=.40, p<.001, N=276). Likewise, hunters gave more meat to households that gave more meat to their households, according to informants' rankings (r=.56, p<.001, N=276). Reciprocity-in-kind accounted for 32% of the variation in meat transfers between households, and nonparametric methods yield comparable correlations (ρ=0.52, p<.001).

3.1.3. Coalitional support

Men who share more meat with each other are considered to be most likely to support each other in a conflict within Conambo. There was a highly significant relationship between meat transfers and alliance strength between male heads-of-households, accounting for 22.4% of the variation in meat transfers (r=.47, p<.001, N=276), and nonparametric correlations were also highly significant (ρ=0.46, p<.001; N=276). Additionally, there was a significant but less predictive relationship between meat transfers and alliance strength between the female household heads (r=.16, p=.012; ρ=0.22, p<.001; N=276).

3.1.4. Tolerated theft

As predicted by a tolerated theft model, there was a highly significant relationship between meat transfers and linear spatial distance between houses (r=−.33, p<.001; ρ=−0.33, p<.001; N=276). Households that were geographically closer shared more meat.

3.1.5. Multivariate analyses

A stepwise regression with meat transfers as a dependent variable and the significant bivariate predictions above as independent variables (see Table 1) yielded a highly significant model that predicted over 42% of the variation in the transfer of meat (R=.649, p<.001, N=276). The model retains reciprocity, wife's kinship, men's alliances, and men's kinship as highly significant (β weights of <.001, <.001, .001, and .005, respectively) independent contributors to meat sharing. Women's alliance strength and spatial distance are excluded in the stepwise regression (i.e., they are not independent of the other variables, with β weights of .780 and .464, respectively). Although the meat transfer data are derived from a ranking task, parametric and nonparametric methods using these data produce comparable correlations, indicating that the meat transfer data are behaving like ratio level data. Given this, and the relatively large sample size (N=276), the inclusion of the meat transfer data in a linear regression is justified.

Table 1. Bivariate correlations with meat transfers within the household-by-household data set (N=276)

	Variables	Pearson's r	Spearman's ρ
	Reciprocity	.56*	0.51*
	Wife's kinship	.50*	0.43*
	Husband's alliances	.47*	0.46*
	Husband's kinship	.34*	0.29*
	Spatial distance	−.33*	−0.33*
	Wife's alliances	.156**	0.223*

* p<.001. ** p=.012.

The insignificance of spatial distance in the multivariate regression suggests that tolerated theft has no significant influence on meat transfers in Conambo. Spatial distance becomes an insignificant predictor of meat sharing (p=.200) in a linear regression when minimally paired with men's alliance strength and women's kinship, indicating that the relationship between spatial distance and meat sharing was not related to hoarding costs and was more likely a byproduct of kinship and alliance. However, spatial data may not be a definitive test for tolerated theft in Conambo. The overall distances between households in Conambo were relatively large. In some cases, the nearest household was a 20-minute walk away. Given trail layouts and these larger distances, closer households may have had no greater, or only a marginally greater, opportunity to detect a successful hunter returning home.

4. Discussion and conclusions

In Conambo, transfers of meat are best explained by multiple adaptive strategies, many of which are better understood with reference to the political context of Conambo. People's decisions about transferring meat to other households may be seen as consequential aspects of men's political strategies and the environmental and coalitional contexts in which they are embedded.

Patterns of meat sharing in Conambo are different than those for the !Kung, Hadza, and Ache, from whom the majority of meat sharing data have been reported. In contrast, meat transfers in Conambo are not governed by norms of generalized sharing, under which hunters have little control over the distribution of their efforts and motivations to target meat transfers to kin, reciprocators, and allies are thwarted. Generalized sharing norms are likely to favor showing-off and costly signaling where transfers are not conditioned on the cost and benefit structures of specific dyadic relationships. I suggest that, in Conambo, the need for larger sharing networks is mitigated by environmental richness, which lowers the risk associated with hunting return variance and the need for pooling resources across a larger number of hunters. In societies where hunting return variance is high, cultural norms may be favored that allow for, if not require, generalized rules for sharing. Generalized sharing norms and associated sanctions provide a cultural solution for policing free riders in larger meat distribution networks and eliminate the need for direct accounting of reciprocal exchanges. Generalized sharing norms also preclude a hunter's ability to target meat sharing conditionally. Conambo is in a game-rich environment, allowing for smaller meat-sharing networks and direct accounting and policing of transfers are more manageable. As a result, hunters in Conambo exercise control over meat transfers, can more easily practice conditional giving, and target meat transfers to reciprocating households, kin, and political allies.

Despite kin selection's theoretical clarity, previous studies have not yielded direct evidence of nepotistic influences on transfers of meat. This study shows that, in the absence of generalized sharing norms, people in Conambo act nepotistically when transferring meat to group members. Similarly, the recent study of Wiessner (2002) of !Kung data indicates that kinship may be an important motivation for hunters. Men contribute more to generalized meat sharing in camps where a high percentage of camp members are related to them at the level of cousin or closer (Wiessner, 2002, p. 428). Thus, despite generalized meat sharing, !Kung hunters also may be motivated by nepotism.

Reciprocal altruism also provides a theoretically lucid reason for meat transfers, but data supporting reciprocation in kind as an influence on the sharing of meat are sparse. The data from Conambo join recent data from the Yanomamö (Hames, 2000) and the Hiwi (Gurven et al., 2000, Gurven et al., 2000) in demonstrating reciprocation in meat transfers. I suggest that the lack of generalized sharing norms and relatively small sharing networks allow people in Conambo to track meat debts and to practice scorekeeping, the hallmark of reciprocal altruism (Mesterton-Gibbons & Dugatkin, 1992).

In Conambo, despite a significant correlation between residential propinquity and meat sharing, the multivariate analysis did not support the argument that tolerated theft is an independent influence on meat transfers between households. In Conambo, spatial distance seems to be a proxy measure of alliance strength and kinship (Bowser & Patton, 2004). This may also be the case in other studies that demonstrate correlations between cooperation and residential propinquity (Gurven et al., 2000, Hames, 2000) and may explain why genetic relatedness appears to be secondary to spatial distance in explaining meat transfers among the Hiwi (Gurven, Hill, et al., 2000). In Conambo, spatial distance between households is best modeled by the wife's kinship and the husband's alliances (r=.45, p<.001, N=276; Bowser & Patton, 2004), as might be expected in this matrilocal society. Men's alliances are significantly correlated with men's kinship, but at low levels of prediction (r=.16, p<.001, N=276). A better predictor of alliances between male heads-of-household is their wife's kinship (r=.402, p<.001, N=276). People tend to locate their households closer to those whom they trust. Spatial distance may thus reflect different°s of kinship and alliance, depending on postmarital residence rules, and meat transfers in Conambo may reflect men's alliances more than their kinship to a greater degree than would be expected for hunters living patrilocally.

If spatial distance is an indicator of the strength of alliances among men, tolerated theft still may be an important influence on meat transfers in Conambo. If the cost of hoarding is the threat of withdrawal of political support, then spatial distance may be a proxy for differences in the valuation of hoarding costs. Withdrawal of political support from one's strongest ally, who is likely to be spatially closer, is more costly than lack of support from a weak ally. If fear of losing or weakening a political alliance is a significant influence on a person's decision to transfer meat, then tolerated theft is supported as an adaptive reason for meat sharing in Conambo and is consistent with the coalitional support hypothesis. The data reported here indicate that there are reproductive advantages to being a good hunter, a characteristic of costly signaling arguments for meat transfers. However, better hunters do not contribute more than they receive within their meat-sharing networks. Contexts where meat transfers are targeted within small distribution networks are not generally addressed by costly signaling arguments. While wide distribution networks may better broadcast the signal (Smith & Bliege Bird, 2000), there is no reason why signaling within small groups would not be adaptive. Furthermore, reputations for hunting skill reach beyond the boundary of individual meat distribution networks, such that the benefits of signaling may extend beyond those networks. Thus, Conambo may be a case of “not-so-costly signaling,” where the hunter derives direct and indirect benefits from targeting transfers to allies and family, including an enhanced reputation among those to whom no direct benefits are delivered. The signaling of skills necessary to be a good hunter also provides reliable cues to attributes that are essential in conflict within the historic arena of Conambo politics. Skills in the use of weapons, stealth, and stalking are important in a context where politics are volatile and alliances are often tested. Therefore, hunting skill also signals value as an ally.

Limited individual meat distribution networks do not fit the expectation that meat gifts are used to show off. However, these data clearly indicate that hunters who can target meat transfers may also be motivated by status acquisition within their coalitions, even if those transfers are reciprocated. Given that age-corrected status is associated with hunting skill, it is difficult to completely disentangle showing off from signaling motivations, and both may be influences on meat sharing in Conambo, if not intertwined adaptive behaviors (see Hawkes & Bliege Bird, 2002). If a benefit of signaling hunting skill is social status, generosity in meat giving is an attribute of high status, and men in stable coalitions are more likely to use meat transfers as a strategy for status acquisition (Patton, 2004), then these data offer some support for showing off and costly signaling arguments while indicating coalitional support influences on meat transfers.

The clearest evidence that meat sharing is a strategy to gain and maintain coalitional support in Conambo is that men transfer meat to their political allies. Men transfer more meat to households whose male heads are likely to support them in conflicts against a third party. The strength of a man's alliances predicts over twice as much of the variation in meat transfers to other households as does his genetic relatedness to those households (19% vs. 8%, all percentage of variation figures used in this section are derived from Pearson's r correlation coefficients). It predicts less than reciprocity (32%) and more than spatial distance (11%), and in this matrilocal society, a man's alliance strength is only slightly less predictive than his in-law relationships is (his wife's kinship with other households, 21%). Alliance strength is also sufficiently independent of these other variables to make a highly significant contribution to the multivariate analysis of meat transfers (p=.001, for the men's alliance strength component). When a hunter in Conambo gives game to another household, he does so within a coalitional context, and he conditionally targets his meat transfers with an eye to coalitional support. It should be no surprise that a valuable resource such as meat would be used, both given and withheld, by other political animals as part of their repertoire of coalitional strategies and maneuverings. Evidence of meat transfers for coalitional support has been reported for wild chimpanzees (Mitani & Watts, 2001), implying old roots for human adaptations designed to integrate economic and political cost–benefit calculations.

Whereas other adaptive arguments account for a substantial portion of meat transfers in Conambo, there still are gaps in these explanations and unaccounted for variation in meat sharing. While tolerated theft may explain why meat may be transferred to avoid the loss of coalitional support, it does not explain the giving of meat to gain support, as some high-status men in Conambo appear to be doing. Relationships among hunting skill, status, and reproductive success are evident in Conambo. However, better hunters in Conambo do not contribute to a public good; they conditionally target meat transfers. Hunters give to coalitional allies, indicating that there are political dimensions to showing off and building a reputation as a good hunter. Nepotism and reciprocal exchanges are also political acts. Kin are dependable allies, and meat sharing offers important cues of alliance and trust between households.

The use of meat transfers to manage coalitional support as an adaptive strategy is likely influenced by a number of environmental and social variables: (1) limited line-of-sight between households and a decentralized settlement pattern that reduce the impact of tolerated theft and increase the opportunity for targeted meat transfers; (2) abundant game, such that the cost of acquiring meat above that necessary for household needs is relatively low, encouraging the transfer of meat to people likely to reciprocate in other critical currencies; (3) and a political context with factionalism and shifting alliances that make efforts at gaining and maintaining coalitional support important. Cultural norms that may encourage meat transfers for coalitional support in Conambo include a relatively high value placed on privacy, personal control over one's labor, control over the distribution of one's resources, and matrilocality. Because of matrilocality, men in Conambo have few kin in the community and, hence, a need to acquire and maintain political alliances with nonkin. Marriage and fertility patterns in Conambo may also influence patterns of meat sharing. Overall high fertility, low reproductive variance, monogamy, low divorce rates, and low childhood mortality imply that men are less focused on mating effort, and this may reduce the influences of costly signaling and showing off.

Efforts toward building and maintaining coalitional support are strategies within a suite of possibilities that may be exercised as part of an evolved psychology that influences people's patterns of meat sharing in different contexts. The benefits of coalitional support are likely to influence meat transfers and other forms of cooperation to different degrees in other contexts and in other societies. However, the measurement of coalitional loyalties, or alliance strengths, is not common in ethnographic studies, and confirmation of the differential importance of coalitional support in varying environmental and social contexts awaits the collection of such data. That said, these data on meat transfers in Conambo support the premise that a significant proportion of the variation in meat sharing between households can be explained as political action, an aspect of an evolved coalitional psychology and an important component of people's strategies to gain and maintain coalitional support in a context where allies are critical resources.