Kin on the Wing : patterns in residence , mobility , and alliance for Ache hunter-gatherers

This paper provides a structural and historical overview of the kinship of a group of Tupi-Guarani-speaking hunter-gatherers, the Ache (Guayaki) of eastern Paraguay. I begin by considering the distinguishing features of Ache kin terminology, describing its characteristic tension between the dimensions of generation and crossness, before considering arguments for historical transformations offered for similar cases in lowland South America. The Ache case shows that the “Hawaiianization” of terms in ego’s generation does not necessarily entail an inward-looking endogamy, as some anthropologists (Dole, 1969; Wagley, 1977) have argued. By describing the network of Ache foraging bands as a residence-based form of kin organization, I show that “Hawaiianization” is not only perfectly compatible with the creation of alliances over considerable distances (Asch, 1998; Ives, 1998; Hornborg, 1998), but that “Hawaiianization” and distant marriage actually work together in the production of band alliances. At various points I highlight semantic, ethnographic, and historical data which, despite lying outside the scope of the phylogenetic analysis undertaken in other contributions to this issue, may nonetheless bear on some of its claims.


THE STRUCTURE OF ACHE KIN TERMINOLOGY
This article addresses the question of directional change in kinship terminology through a structural and historical overview of the kinship of a group of Tupi-Guarani-speaking hunter-gatherers, the Ache (Guayaki) of eastern Paraguay, as it existed in the first half of the twentieth century. 1 For heuristic purposes, Ache kin terminology may be roughly described as semi-bifurcate merging at G +1 and generational at G 0 , a pattern anthropologists have variously termed bifurcate generational (Dole, 1969) or Iroquois-generational (Tjon Sie Fat, 1998). 2 In this article, however, I am less concerned about fitting Ache terminology into existing typologies than I am with exploring the tension between the bifurcate and generational tendencies that I believe define it.
I begin by describing how these tensions are expressed in Ache terminology.In exploring these tensions, I consider arguments for the historical "regression" of kinship structures.The Ache case shows that the "Hawaiianization" of terms in ego's generation, i.e. the reclassification of cross and parallel cousins as siblings, does not necessarily entail an inwardlooking endogamy, as some anthropologists (Dole, 1969; Wagley, 1977) have argued.By describing the network of Ache foraging bands as a residence-based form of kin organization, I show that "Hawaiianization" is not only perfectly compatible with the creation of alliances over considerable distances (Asch, 1998; Ives, 1998; Hornborg, 1998), but that "Hawaiianization" and distant marriage actually work together in the production of band alliances.
The nomenclature is presented below in Table 1, 3 and tree diagrams to ego are presented in Figures 1 and 2. All kin terms can be used as both reference terms and terms of address.Affinal terms are generally not used as terms of address, owing to the influence of affinal avoidance. 1I consequently do not discuss changes in Ache relationship terminology since their settlement in the latter half of the twentieth century, such as the introduction of certain Spanish terms and the abandonment of affinal terminology. 2See Wagley and Galvão (1946). 3Terms were elicited during the period 2006-2017 while gathering genealogies in two Ache communities.All the 18 respondents I spoke with had been married before the beginning of the settlement period (1959-1978).I later cross-checked the consistency and accuracy of terms against an extensive archive of transcribed Ache recordings spanning from 1960 to 2017.The terms above illustrate that in its broad features, Ache kinship terminology springs from four salient dimensions: 1) Dimension of generation: a generation may take one of five values in Ache terminology: grandparents' generation, parent's generation, ego's generation, children's generation, or grandchildren's generation.Ego's G +2 and G -2 generations are boundary positions, in the sense that their terms apply for all generations beyond them.Thus, the terms for 'grandfather' (djamo) or 'grandmother' (djary) apply to all persons in the generation above ego's parents, while a genderless term for 'grandchild' (mino) may be applied to those persons younger than ego's children. 4 (Conclusion) Table 1.
Generational dimensions may possess strongly indexical valences, a point long stressed by anthropologists working in lowland South America (Basso, 1975; Kensinger, 1985; Ball, 2015).A particularly important example is the application of male consanguineal terms in G +1 and G +2 beyond the boundaries of Ache society to classify, describe, and establish asymmetrical relations with non-Ache.Since at least the nineteenth century, the Ache have referred to distant enemies (particularly whites) as ãpã and djamo, 'fathers' and 'grandfathers,' independent of any actual relationship established with them. 5I consider these uses indexical and iconic to Ache 'fathers' and 'grandfathers' (and not simply an undifferentiated element of the same category) because ãpã and djamo could not be applied to non-Ache in the same way as they applied to Ache persons.First, the generational salience of ãpã and djamo as Ache kin terms did not transfer to whites; in the latter case, the use of one or the other was completely unrelated to the age or generation of the person in question.Second, these terms had gendered values which differed from their ordinary use as kin terms.In no case did an Ache woman become a daughter (radjy) to a Paraguayan 'father,' and Paraguayan women were not 5 As categorical terms for distant enemies (particularly whites), ãpã and djamo were not possessed.When they referred to actual relationships of 'familiarization' between specific Ache and Paraguayans, these terms were possessed.Thus, after establishing relations with certain whites (ãpã) during the settlement period, Ache men would commonly refer to a Paraguayan ally as 'my father' (cho ãpã) and would refer to themselves as 'their son' (idja ray).categorically referred to as 'mothers' (ei).Finally, this practice was not merely a feature of the generationally-minded terminology of the Ache, but was common among other Tupi-Guarani-speaking groups as well.It seems to have been a well-established practice for the sixteenth-century Tupinamba, who received the Portuguese with the term tamoi ('grandfathers') (Thevet, 1953 [1575]), and the Parakanã, who called them 'fathers' in the twentieth century (Fausto, 2012).
2) Dimension of relative age: male and female egos use the same elder/younger distinction for their same-sex 'sibling' terms, but not for cross-sex 'sibling' terms.Thus, men call their older brothers key'y and their younger brothers tywy, just as women call their older sisters key'y and their younger sisters tywy, with a different term for each crosssex sibling: a man calls his sister ny, while a woman calls her brother kywã.These terms may also be applied to close cousins co-residing with ego.I did come across a few instances in which women referred to their brothers with these 'same-sex sibling' relative age terms, however. 6The only partial hypothesis I can offer to regarding this departure is that Ache sisters very rarely co-resided in adulthood (see below).
3) Dimension of sex: offspring terms are distinguished by both the gender of ego, and for the male ego, the gender of the child, as stated above.An Ache father calls his S, BS and daughters ray and radjy (or in one dialect, tay and tadjy), respectively.A woman's term for her child (memby) does not indicate its gender, although it can be further specified, as in cho memby kybei, 'my male child,' or cho memby kudja, 'my female child.' Specific terms for cross-sex siblings (ny for a man's sister and kywã for a woman's brother) exist alongside pawe, a generic term for sibling that encodes neither the sex nor seniority of the ego or referent. 74) Dimension of crossness: in a terminology with a strong generational tendency where all kin are grandfathers, grandmothers, fathers, mothers, siblings, children, or grandchildren, the crossed term tuty remains a conspicuous outlier. 8he significance of this term has eluded previous ethnographers, and attempts at defining it have not yet yielded any sort of consensus about its meaning.The German anthropologist Mayntzhusen (2009 [1947]) defined tuty as ZS, Cadogan (1968) defined it as ZD, and Clastres (1968) deemed tuty a reciprocal term used between MB and ZS.None of these previous definitions are entirely correct.Tuty is instead a reciprocal term used between MB and Zc of either sex.So in fact tuty designates two relationships: the relationship between MB and ZS and the relationship between MB and ZD.
Among Amazonian groups, this oblique equation is generally thought to express (in vestigial form) the quintessentially Dravidian MB/ZD marriage favored by Tupi-Guarani speaking groups centuries ago (Kirchhoff, 1931; Métraux, 1948;  Watson, 1944; Lévi-Strauss, 1943), which established the repetition of marriage alliances across generations. 9 6 Though these terms find cognates in neighboring Guarani groups, the relations to which they refer appears to vary from case to case.
In the sibling terminology of the Pai-Guarani described by Melià et al. (1976), the Ava Guarani described by Reed (1995), and the Mbya Guarani terminology described by Pissolato (2007), each sex has different relative age terms for same sex siblings (e.g. a male ego's brother is tyke'y or tyvy, while a female ego's sister is tyke or kypy'y) with single terms for cross-sex siblings.By contrast, the Dravidian-type sibling terminology described for the Kaiowa by Watson (1944), makes a relative age distinction for both same and cross-sex siblings. 7Gender may be specified for pawe ('sibling') in the same way as the term memby ('child of woman'), i.e. cho pawe kybei, 'my male sibling,' or cho pawe kudja, 'my female sibling.' 8 Clastres (1968) reports that the term tuty had already been abandoned by the Northern Ache subgroup at the time of his fieldwork in 1963, yet recordings of Northern Ache from around that time suggest otherwise. 9Among the early chroniclers, Léry (1993 [1578]) and Thevet (1953 [1575]) paid particular attention to the Tupinamba avunculate, and in Cardim's (1939 [1584], p. 142) telling of the Tupinamba flood story, humanity was able to rebuild itself after the cataclysm through avuncular marriage.Despite the strong overall similarities between Tupinamba and Guarani, there is no mention of avuncular marriage in the earliest accounts of the Guarani, and the seventeenth century Guarani terminology recorded by Montoya does not encode avuncular marriage.Strangely, Montoya's terminology expresses a lineal tendency for these terms (Montoya, 1876 [1639]), MB = MBS = MBD and FZ = FZS = FZD for both male and female egos.
But avuncular marriage was quite rare for Ache in the twentieth century, comprising less than 3% of the marriages for which I was able to collect data, and was described to me by several informants as 'disgusting.'Those who are affines to each other in G +1 are consanguines to ego.This is reflected in the attitudes and behavior typifying the tuty relationship.For the Ache, the tuty took on a paternal role with respect to his Zc.He was expected to shelter his sister and her children if her husband died or abandoned her, and would be owed bride service by his ZH if his father were absent or dead.He therefore stood as a father-substitute for his Zc and not a husband or father-in-law.The absence of preferential cross-cousin marriage for the Ache in the twentieth century and their use of a developed affinal register further distinguish it from the Dravidian ideal (Dumont, 1983 p. 114). 10The presence of crossness did not entail affinity for the twentieth-century Ache as it evidently did for the Tupinamba and Guarani of the seventeenth century. 11s Lounsbury (1964) has pointed out, many bifurcate merging systems are "semi-bifurcate merging," "distinguishing between 'cross' and 'parallel' kin only among males in the first ascending generation, and only in relation to a male ego for kin types of the first ascending generation" (Lounsbury, 1964, p. 387, note 3).Ache kin terminology, which lacks a specific term for FZ, is an example.Nevertheless, we can propose a case of "vestigial crossness" for FZ in Ache terms for ritual kin or ritual states.A comparison with Guarani is instructive: the Guarani term for FZ (jaiche) is cognate with the Ache term for a woman's birthing helper (waiche).Along similar lines, during brief fieldwork in the 1960s with the Ache Susnik (1974) elicited the term djoare (literally, the "washer" who cleans the polluting blood from the newborn and its parents) for MB.It therefore appears that Guarani bifurcate merging "survives" as terms for male and female birth helpers in Ache. 12aken together, these dimensions exhibit a number of interesting patterns: first, we see that male and female terminologies are not structurally equivalent.Male terminology is bifurcate at levels G +1 and G -1 , but generational at level G 0 , whereas female terminology is bifurcate at level G +1 and generational at G 0 and G -1 .Second, with the exception of terms for a man's S and D, it appears that the generational dimension exists in tension with distinctions in relative age, sex, and the oblique avuncular structure of MB.In cases where generational terms overlap with terms indicating relative age and sex, the Ache generally prefer generational terms.The generational term for G 0 , pawe, for example, has all but eclipsed crossed sibling terminology as well as relative age terms for same sex siblings.Clastres (1968) reported difficulties eliciting tuty and crossed sibling terminology in 1963, and I encountered some inconsistencies in my own attempts to elicit relative age terms for same sex siblings.Nevertheless, the fact that these terms could be elicited, along with the retention of a fuller bifurcate merging terminology at G +1 in the names for birth helpers, raises a host of questions about the history of these terms.

"HAWAIIANIZATION" AS REGRESSION
Bifurcate generational terminologies have typically been explained as the result of historical deviations from more "regular" Dravidian terminologies (Hocart, 1928; Dole, 1957, 1969; Viveiros de Castro, 1993; Tjon Sie Fat, 1998;  Kryukov, 1998; Godelier, 2011).Dravidian terminologies distinguish "same-sex" and "cross-sex" parental siblings, who are treated as consanguines and affines, respectively.The distinguishing feature of these terminologies, according to Dumont (1983), is that affinity is passed from one generation to the next.Not only is WB an affine to a male ego, but he and his children remain affines to ego's child as MB and MBc (Dumont, 1983). 13The problem of bifurcate generation terminologies, as with other "semi-complex systems," involves how they depart from this pattern.In bifurcate generational terminologies crossness is retained in G +1 , but the implicit distinction between cross and parallel is not transmitted from G +1 to ego's generation.Instead, G 0 undergoes a process of "Hawaiianization."The puzzle of bifurcate generational terminologies is why (affinal) cross cousins would be reclassified in (consanguineal) generational terms.
Most discussion on bifurcate generational terminologies in lowland South America has centered on the Upper Xingu, and it was an anthropologist working there, Gertrude Dole, who originally coined the term.Dole  (1969) marshaled comparative evidence to suggest that cross-cousin terms are marked with respect to the crossed terms of the first ascending generation, so that the former terms will be added last or lost first with respect to the latter.In other words, cousin terms were more sensitive indicators of social change. 14As a transitional type midway between a bifurcate merging and a fully generational terminology, Dole (1957, 1969) considered bifurcate generational to have elements of each: it retains a bifurcate pattern in G +1 as long as cross-cousin marriage remains desirable, yet erases the bifurcate pattern at G 0 when exogamy and unilocal residence dissolve and cross cousins co-reside.The ultimate cause of the terminology's Hawaiianizing tendency, according to Dole (1969, p. 107), is therefore acculturation, the "[…] demographic disturbances and disruption […]" that supposedly make exogamy and unilocal residence impossible.Basso (1984) would later dispute the characterization of several Upper Xingu terminologies as bifurcate generational, arguing instead that generational and cross-cousin terms were used in different contexts to index different aspects of sociability.In Basso's view, terminology at G 0 could not be considered generational, strictly speaking.Basso's argument is persuasive, yet the presence of bifurcate general terminologies outside the Upper Xingu, particularly among Tupi-Guarani groups, means Dole's argument could find validation elsewhere, a possibility Dole herself suggested (Dole, 1969, p. 117).
Contrasting the generational tendencies in Ache (Guayaki) terminology with other terminologies in the Tupi-Guarani family led Pierre Clastres to consider similar processes: It is quite surprising to note that, although the Guayaki terms are found in Guarani terminology, it is incomparably richer and more complex than the former; this marked contrast poses the important problem of the historical relations between the two cultures: should we suppose that the Guayaki system is, in its impoverishment, only the archaic foreshadowing of the later Guarani system?Or did the Guayaki and Guarani have the same system of kinship in the past, but the former, as a result of certain circumstances, allowed the deterioration and impoverishment of their terminology?(Clastres, 1968, p. 10-11).Clastres (1998) would eventually favor the latter possibility, thereby including kinship as an element in what would be known as the regression hypothesis, the idea that destructive exogenous forces resulted in the abandonment of moiety structure, horticulture, and a host of other socioeconomic features by a number of Amerindian groups (Lévi-Strauss, 1963 [1958]; Lathrap, 1968; Martin, 1969). 15He argued that the "impoverishment" of Ache socioeconomic organization, presumably captured in the preponderance of generational terminology in its kinship nomenclature, was likely the result of Ache defeat at the hands of their neighboring Guarani enemies (Clastres, 1968, p. 11, 1998, p. 85-86).Other authors (Susnik, 1961, 1983; Noelli; Soares, 1997) have argued that a similar "regression" took place for the Guarani centuries later when waves of Iberian colonization led to a number of dramatic transformations in traditional Guarani political institutions, such as the gradual abandonment of long-house residence and the dissolution of intervillage marriage alliances in favor of village endogamy. 16espite the occasional functionalist undertones of these arguments, such hypotheses have an intuitive plausibility.Kinship terminologies are other sociological classifications, Lévi-Strauss noted, "[…] are not only thought but lived […]" (Lévi-Strauss, 1966, p. 66, emphasis in the original).Classifications of individuals and groups are subject to forces exogenous to inner classificatory logics, particularly epidemics and warfare, that may significantly change the number and quality of persons and relations considered.This is a particularly important point given the profound ways indigenous and colonial expansion have reshaped (and continue to reshape) the Americas.But there is a methodological danger in inferring "functional" changes in social morphology from changes in kinship terminology in cases with little historical documentation (as many explanations invoking "regression" do).The argument that demographical decline "simplifies" a given kin terminology offers little help in determining which specific terms will be lost or altered.Even if exogenous forces determine that a classificatory system will change, it is the principles behind the classification -its inner coherence -that determines how it changes.Consequently the issue is not whether exogenous forces can influence kin terminology, as they clearly can, but how they do so and whether these influences can be demonstrated.Dole (1969) is to be commended for suggesting possible mechanisms explaining why a society would adopt a bifurcate generational terminology.But as it stands, I have some doubts that the complex processes associated with colonialism, such as depopulation caused by epidemics and warfare, can be assumed to have regular effects on kin terminologies.Not only can we find bifurcate generational terminologies for groups with considerably different social morphology and residence patterns, but the Ache case (for which we have some historical data) casts doubt on some of the principal mechanisms Dole provided.

BAND, RESIDENCE, DISTANCE
The Ache trekked year-round in residentially mobile hunting bands.These were small groupings of between 20 and 40 persons, with size and composition varying as persons came from and went to other allied bands.Camps were cleared, occupied, and then abandoned on an almost daily basis, with no semi-permanent base village to which they would return for a portion of each year.They cultivated no crops; to obtain domesticated food like manioc and corn, they raided the gardens of their Guarani and Paraguayan enemies, whom they sometimes called "corn eaters." 17unting bands are not reducible to families, clans, or lineages, but are contingent entities whose size and composition depend upon a number of political and economic factors.Bands often coalesced around a group of brothers, their wives, children, unmarried sisters, and a few sons-in-law.These band leaders were said to "own" their co-residents in the band, an idiom also used to express hierarchical relations at different scales: husbands "owned" their wives and parents "owned" their children.Just as a leader was creator and owner of the band, a man was owner of his wife and parents the owners of their children. 18This notion of ownership did not imply strong authority, however.People came and went freely to and from the band, and individuals had a great deal of autonomy in deciding whom they wished to follow.The bands people moved among were themselves ephemeral assemblages; a band might lose or gain members, disband, or reconstitute itself with a largely different membership over a short span of time, particularly if the band's core relations of siblings and/or affines were affected.
Though each band formed its own mobile community, clusters of neighboring bands made up territorial groups defined by density of kinship ties and frequency of visiting and intermarriage.These extra-band allies were referred to as irõndy, 'habitual companions.' Irõndy referred not to common ancestry but to the habit of living together, so that one's irõndy consisted of those one resided with or could reside with, but not forbears who one had never personally encountered. 19Marriage partners might be drawn from any of the irõndy who hunted and camped in the peripheral reaches of the ekõandy, a territorial descriptor that literally means 'living-space.' It was common to have parents, siblings, and/or children residing in other allied bands throughout the territory, giving the Ache choices about where to live and whom to live with.Of the many interests Ache might consider in choosing residence, the Ache men and women I spoke with expressed the desire to live with their siblings, particularly brothers.Ache men achieved political influence by living alongside their brothers.Women, for their part, preferred to live close to their brothers since men were expected to care for and protect their sisters; like a father, a man was expected to shelter his sister and her children if her husband died or abandoned her.Should a sister's husband mistreat her, he was expected to avenge (djepy) the offense.As mentioned above, he would be owed bride service by his ZH if his father were absent or dead.
Band residence did not differentiate one's affines from one's consanguines.Individuals often had close kin in distant bands, and their resident band was mostly composed of non-relatives.According to a recent study by Hill et al. ( 2011) 17 We lack the historical sources needed to make strong claims about the continuity of pre-Columbian Ache with the Ache of the twentieth century.The Jesuit chronicler Pedro Lozano states that in the seventeenth century, the Ache "[…] roam[ed] like nomads throughout the forest […]" (Lozano, 1874 [1759], p. 416).He also describes them as growing corn (but not manioc) during their seasonal rounds, scattering the seeds in forest clearings and returning months later to harvest it.Though the Ache may have been horticulturalists at some point before the arrival of Spanish colonists, Lozano's report does not support Roosevelt's revisionist claim that the Ache of the mid-twentieth century were typical "Amazonian horticulturalists, who supplement their starchy crops with fish and game" (Roosevelt,  1996, p. 203).Firstly, corn cultivation does not necessarily imply the sedentary residence she attributes to it (Meggers, 1957; Balée,  1992), while secondly and more important, Roosevelt's claim flies in the face of how the Ache have described their livelihood over a century of Ache ethnography.No Ache has ever admitted to cultivating corn or manioc before being taught to do so by non-Ache in the settlement period, a fact the Ache grant particular significance.The inability to grow corn becomes their defining cultural diacritic in one etiological Ache myth explaining their separation from their Guarani enemies.While the twentieth-century Ache occasionally obtained corn and manioc by raiding the gardens of Guarani and Paraguayan enemies, these were clearly not things the Ache grew themselves. 18See Fausto (2012) and Costa (2009, 2017). 19Cognates from other Tupi-Guarani languages may have different referents, (e.g.Tupi-Kawahiva irũ, which refers to a 'same-sex sibling' and Guarani irũ, which Montoya (1876 [1639]) defines as 'friend' or 'fellow'), but all focus on members of ego's generation.These terms refer to a horizontal relation and do not include ideologies of descent.
that reviewed the adult band composition of 58 pre-contact Ache bands, kin made up only a quarter of band residents.Non-relatives made up another quarter, and actual affines and their kin made up roughly half the band.In the Ache case, whole families did not form bands, and bands were not composed of whole families.Life in a band meant living in close proximity with non-relatives.Men and women were forbidden to marry anyone classified as pawe, a 'classificatory sibling' which includes full siblings, half siblings, and both parallel and cross cousins.The Ache did not make use of descriptive terminology or modifiers to specify marriageable collateral kin (e.g.cross cousins) that a classificatory term like pawe would obscure. 20ne should instead seek potential marriage partners among those considered picha.Picha refers to an unrelated peer from an allied band, typically of the same generation with whom one did or could reside, but not kin or an actual affine.Unsurprisingly, the contrast between pawe and picha was made explicit by the Ache: Pawellã-rõ picha, "Those who aren't siblings are picha" (interview, January 5, 2016). 21Like pawe, picha is a relational term, in that someone who is picha to me is someone to whom I am picha.Yet pawe and picha are not symmetrical opposites: not everyone who is picha to my pawe is pawe to each other, so the distinction made between pawe and picha was not an objective designation dividing Ache society into two marriage classes or sections.One's picha represented a heterogeneous field of potential alliances that cross-cut band membership.
While the pursuit of contented relations was a basic virtue for Ache men and women, informants accepted that conflict was inevitable in relations with picha, and behavior that would be unthinkable around a sibling was known to occur (if not tacitly encouraged) among picha.For example, though men were expected to behave with a respectful distance toward their BW, picha and their wives were treated according to a different standard.One informant admitted, "We take the wrists of the wives of non-kin (picha).Those are the ones we typically have affairs with" (interview, January 5, 2016).The typified enmity between pawe and picha was symbolically productive and featured prominently in some of the most important rituals in Ache society: men supported their pawe against picha in the ritual club fights (called tõmumbu, 'head-splitting') that often followed male initiation.Even in cases where relatives relied on picha to accomplish ultimately ameliorative goals, such as helping picha to sever the dangerous connection between themselves and their dead kin after funeral rites, the proximate means of picha were decidedly antagonistic: they do so by insulting the dead and harshly scolding mourners for the grief they expressed.
Given the antagonistic overtones often associated with these relations, picha stood as a model against which kin might measure the satisfaction of their own relations.Likening sibling relations to relations with picha signaled that expectations had eroded in some fundamental way, failing to achieve the kind of care and deference expected in these relationships.One woman described her broken relationship with her brother in these terms: "My brother was often ill-tempered with me.If I brought him palm heart, my brother refused to eat it.He treated me like a picha" (interview, May 28, 2010).In responding to his sister's care with aggression and refusing her generosity, the man did not (normatively) behave like the brother he (categorically) was.He was "like a picha." Nonetheless, some picha had to be made into actual affines.Though there were no fixed rules regarding band endogamy or exogamy, there were clear preferences about where one should marry.Even though an Ache man or woman could usually find marriage partners within one's own band, Ache nevertheless preferred marriage partners who were spatially distant.As one informant recalled to me in 2016, "When we lived in the forest, we looked for wives from the other Ache, from the far-away Ache" (interview, April 2, 2017).Allied bands might be separated from each other by several days' journey, perhaps 25 km or more.The kind and frequency of these distant marriages depended on the relations between one's band and the allied bands nearby, and the ability to visit kin in distant bands served as ready pretext to find a spouse.Though distant picha were preferred, marriage to someone who was not irõndy was extraordinarily rare.While the Jesuit chronicler Lozano (1874 [1759]) 22 reported that in the seventeenth century, Ache warriors frequently raided enemy (irõllã, 'not-companions') groups for women, and Clastres (1998, p. 225) emphasized this point strongly, only slight importance seems to have been given to marriage by capture in the past century. 23There has been only one case of an Ache war party raiding enemies for wives for any Ache subgroup in the twentieth century: the one case Clastres (1998) mentions.Alliances with enemy Ache, Guarani, or Paraguayans were unimaginable (if these groups were even considered "people").The Ache consequently preferred marriages in a middle range, outside one's own band but within a universe of regularly interacting allied bands.
Though men and women were allowed considerable latitude in choosing the partners they 'courted' (gaita), parents exerted their own influences on their children's choices, particularly in early marriages.Parents warned their daughters against mates known for aggression or jealousy and recommended suitable mates based on hunting ability and generosity.Such recommendations could express own interests, since the new husband was expected to provide regular gifts of game to his co-resident WF, WM, and WB during a period of bride service.Ideally, a gift of game to a woman's parents -usually the fatty jowls of a peccary or some other coveted cut of meat -marked the beginning of the man's residence at the hearth of his affines and a promise of gifts to come.
The significance of these gifts and the kind of residence they implied is important, and seems to be historically variable.Seventeenth-century Jesuit missionaries described the Ache living in bands with uxorilocal residence, where political power rested on the alliance between a man and the sons and sons-in-law he could gather around him: He who is fortunate enough to have a daughter born to him is very careful in raising her, because it is through her he will become the head of the others; being the inviolate law of the Guayaki that a son-in-law must follow his father-in-law and become part of the family, because among them they have no chiefs, only that the brothers and sons-in-law get together in a group and recognize their father or father-in-law as the leader; but the power that he enjoys over others is very limited, since each lives according to his own whims (Lozano, 1874 [1759], p. 417-418, emphasis in the original).
Within the outlines of Lozano's description, early twentieth-century patterns of Ache kinship and residence can be glimpsed.As Lozano suggested, there appeared to be some specificity in the relation between an Ache man and his father-in-law, and the reciprocal term used in this relation is semantically marked.Thus, while the term katy is used reciprocally between women and their parents-in-law and between men and their mothers-in-law, men and their WF use a separate reciprocal term: djywẽ.A man's relations with his wife's parents were ideally marked by a respectful distance, and reflecting the restricted interaction between them, neither katy nor djywẽ were used as vocatives.
The relationship between a man and his djywẽ is described in the idiom of exchange, but the reciprocity between affines is achieved through a short bride service and not the repetition of alliances.Fathers reluctant to give away their daughters were considered 'stingy persons' (tãngi) by suitors, just as new sons-in-law or brothers-in-law were considered 'stingy' for failing to provide their real affines with meat.As one informant explained to me, "A man doesn't want to give his sister to a brother-in-law that doesn't give meat to him" (interview, April 18, 2017).
In the twentieth century, however, marriage ties were often too weak to serve as a source of enduring political influence.It was not simply that a father-in-law's authority was limited by norms of autonomy egalitarianism, as Lozano described.Uxorilocality typically lasted for only a few months of bride service; in cases where the new husband was older and established, uxorilocal residence and bride service might be avoided entirely.Most importantly, a father-in-law's authority was structurally limited by the weakness of marriage ties.Partners terminated marriages and affairs at their own behest, and did so with some frequency.Hill and Hurtado (1996, p. 219, 237) report a mean of thirteen husbands for Northern Ache women over their lifetimes.The relative weakness of fathers-in-law meant that residence was less a rule to follow or violate than the outcome of a number of decisions, each carefully weighed among local possibilities. 24hough Lozano described Ache post-marriage residence in the seventeenth century as uxorilocal, it appears that since that time fathers-in-law have increasingly lost the struggle to determine residence and influence through their sons-in-law.Residence was normatively virilocal during the first half of the twentieth century, a point my male informants were keen to emphasize.As one older man explained to me, When those ones lived in the forest, their lover's place was beside them.When a man wanted a woman, he brought her to his fire, he brought her to his fire.The man normally brought the woman.When a woman was loved by a man, he loved her well, then he took her.The man did.The man normally took that one.He was beside his father.He didn't leave his father.It was like that in the forest (interview, April 10, 2017).
Such views express an ideal, and as data collected by Hill et al. (2011, p. 1287) demonstrate, it was an ideal that was unattainable for many men.Nevertheless, as their study shows, men were still more likely than women to co-reside with their close kin, a significant departure from the residential patterns of the seventeenth century described by Lozano. 25 Is it possible that this shift from uxorilocal to bilocal residence is related to the "Hawaiianization" of terms in ego's generation, as Dole (1957, 1969) and others (Wagley, 1977) have suggested?The possibility is intriguing, but the image of an inward-looking endogamy that these authors associate with "Hawaiianization" does not appear to be necessary in this case.For the Ache, the preferences for sibling co-residence and spatially distant marriages offer the strongest structuring principles for the flexible group configuration system.The Ache prohibition on marrying one's pawe is more expansive than most marriage restrictions in Amazonia; it prohibits the cross-cousin marriage typically preferred by many groups in the region.This stands in stark contrast to Dole's (1969) theory that generation terminologies entailed the loosening of marriage restrictions. 26But the greatest departure from Dole's theory (as well as from descriptions of contemporary Guarani groups) involves the Ache preference for distant marriage, which favors distant over co-resident picha.The preference for spatially distant marriage maintains alliances between bands as residential units.Ache society thus appears not merely as a sum of relations but also of locations. 27This then seems less a case of "regressed sedentarism" than an integral feature of Ache mobility and foraging (Costa, 2009).Within a strongly factionalized political context, long-distance marriage produced extensive alliances and scattered kin, 28 and the fragility of many marriage bonds ensured the further dynamism of this system.

Table 1 .
Ache relationship terminology in the twentieth century.