Edward Sapir’s “Terms of Relationship and the Levirate” is a short technical paper in anthropological linguistics, written in the early twentieth century by one of the founders of American structural linguistics. It examines how the custom of levirate marriage—a man marrying his dead brother’s widow, or a widower marrying his late wife’s sister—leaves visible traces in the vocabularies of kinship. Working from Chinook, Yana, Takelma, Tlingit, Shoshone, and Paiute material, Sapir argues that certain kinship terms are not merely labels for biology but fossils of social obligation.
The Case Against Purely Sociological Explanation
Sapir begins by taking stock of a revival of interest in the connection between systems of consanguinity and affinity and particular forms of social organization. The credit for this revival, he says, belongs chiefly to Rivers, who through discussion and example—drawing mainly on Melanesian material—showed convincingly that many groupings of kinship terms are best understood as expressions of particular types of marriage. Many of Rivers’ specific inferences seem far-fetched, Sapir concedes, and no one need follow him in every detail, but he holds the main argument sound.
A footnote complicates the picture. Lowie had tried to show that Rivers’ line of reasoning was too exclusive, explaining by a specific form of marriage what might equally well result from group exogamy. Sapir finds Lowie’s objections partly invalidated: Lowie fails to explain why only certain relationships covered by the same exogamic rule fall under a single term, and by not defining the reasons for that specific delimitation he may prove too much. Sapir’s own view is that the factors governing kinship nomenclature are very complex and only partly explicable on purely sociological grounds. Satisfactory results demand linguistic analysis of the terms themselves, and proper historical perspective demands a feel for the lack of strict accord between linguistic and cultural change—an existing nomenclature may survive sociological developments that ought to have modified it. The paper, he insists, is neither a polemic nor a broad methodological treatise. Its aim is narrower: to call attention to a specific type of marriage as a determinant of part of the kinship nomenclature. Some of the facts he instances are instructive precisely because one might otherwise have inferred actual or former group exogamy—an inference known to be impossible for the Yana and the Chinook.
Two Ways the Levirate Enters the Vocabulary
Sapir defines the levirate as the widespread custom, found among American Indians and many other peoples, by which a man has the privilege or—more often—the duty of marrying the widow of his deceased brother and raising the offspring in his own household. Correlatively, a man has the privilege or duty of marrying the as yet unmarried sister of his deceased wife. He treats these as two forms of the same institution.
How does such a custom find expression in kinship terms? Sapir answers in two distinct ways. First, viewing the levirate as an accomplished fact, the step-relationship comes to be defined in terms of the nepotic relationship: step-father as uncle, step-mother as aunt, step-child as nephew or niece. In tribes with distinct terms for paternal and maternal uncle and aunt, this demands that step-father and step-mother be identified with paternal uncle and maternal aunt respectively; and in tribes distinguishing brother’s from sister’s children, the step-child is identified with the man’s brother’s child and the woman’s sister’s child. Second, viewing the levirate as a potential fact, certain nepotic and ensuing relationships are defined in terms of the filial and fraternal ones: paternal uncle as father, maternal aunt as mother, the man’s brother’s child as son and daughter, the woman’s sister’s child as son and daughter, and the children of father’s brother and mother’s sister as brothers and sisters—as distinguished from “real” cousins, the cross-cousins. One may also expect a man’s sister-in-law and a woman’s brother-in-law to be called wife and husband. Sapir sets out to show that such peculiarities actually follow from the levirate among the Upper Chinook and the Yahi, or Southern Yana.
The Upper Chinook: Step-Kin as Uncle and Aunt
Among the Upper Chinook the identification of step-relationship with nepotic relationship is complete. Sapir lists the terms: i-mut for paternal uncle, i-lsm for maternal uncle, a-lak for paternal aunt, a-gutx for maternal aunt; i-wulx for a man’s brother’s son or a woman’s sister’s son, a-wulx for the corresponding daughters; i-latxan and a-latxan for a man’s sister’s son and daughter; i-tkiu and a-tkiu for a woman’s brother’s son and daughter. The step-relationships fall exactly onto certain of these: i-mut is also “step-father,” a-gutx “step-mother,” i-wulx “step-son,” a-wulx “step-daughter.” These facts, Sapir says, speak for themselves; their dependence on the levirate is too obvious to need extended discussion. He adds only that the levirate is known to have been in force among most or all of the tribes of Washington and Oregon. For the Upper Chinook he infers, with some plausibility, that the custom itself—the fact that a man’s brother’s child and a woman’s sister’s child were alike potentially step-children—was responsible for grouping those two relationships under a single term, in contrast to the distinctive terms for the man’s sister’s child and the woman’s brother’s child. A footnote identifies the Wishram tribe of southern Washington as the source, with data secured at the Yakima reserve in 1905.
Ishi’s Yahi: Father as Uncle, Uncle as Father
The Yahi data, gathered from Sapir’s informant Ishi, who spoke very little English, are, in Sapir’s judgment, fully as instructive. He regards them as thoroughly reliable: the terms were collected slowly and checked repeatedly, and the agreements with Northern Yana cannot be put down to suggestion. Ishi made clear that he saw the facts as simply another way of saying that it was customary for a widow to marry her former husband’s brother and for a widower to marry his former wife’s sister.
The Yahi terms show the levirate at work. Galsi means “father” and also “paternal uncle”; ganna means “mother” and also “maternal aunt.” The paternal uncle, as potential father, is termed father; the maternal aunt, as potential mother, is termed mother; and their children are correspondingly termed son and daughter—H’sipla, literally “little man,” and mari’mipla, literally “little woman.” The maternal uncle (u’dji’yauna) and paternal aunt (mucdi) keep distinctive terms, with their nephews and nieces sharing those same terms. The nepotic vocabulary thus falls into two groups: terms identified with the filial relation through the levirate, which recognize the generation difference, and reciprocal pairs that do not recognize it—the latter type also covering grandparents and grandchildren.
One term repays close analysis. Wa’dat’imauna (plural yeidat’imauna) applies to a man’s brother’s son or daughter. Sapir parses it: the stem -wa’- (plural yei-) is a verb meaning “to sit,” conveying “to have, consider as,” with an incorporated noun for “child” and a participial ending—hence “had, considered as own child,” a potential son or daughter under the levirate. Northern Yana has cognate forms. Cross-cousins have their own terms, while cousins through parents of like sex are simply brothers and sisters. The fraternal terms—for older and younger brother, for a man’s sister, a woman’s brother, older and younger sister—are each also applied to cousins through the paternal uncle or maternal aunt, with “older” and “younger” referring to the parties themselves, not their parents.
The levirate surfaces again in the words for spouse. H’si, “husband” (literally “man, male”), also means “husband’s brother”; mari’mi, “wife” (literally “woman”), also means “wife’s sister.” Sapir lacks data on whether these extend to a woman’s sister’s husband and a man’s brother’s wife, but thinks it likely, noting that Northern Yana has a single term, u’naiyana, for wife’s sister, man’s brother’s wife, husband’s brother, and woman’s sister’s husband—roughly, “potential spouse.” The remaining in-law relations each have distinctive terms in both dialects, differing only phonetically. The paternal uncle’s wife and the maternal aunt’s husband are not treated as typical potential mother and father: the former may become a step-mother, the latter may take the widow to wife and become step-father, but “mother” and “father” are not used for them; they have special terms instead.
The Step-Mother and the Fossilized Uncle
The term p’imo’o, “step-mother,” carries real weight. Ishi said that the maternal aunt’s husband addressed his wife’s children as his own children, implying potential fatherhood—though Sapir cautions that Ishi’s statements left some doubt as to his exact meaning. A woman calls her husband’s brother’s child dat’ipla, the diminutive of “child,” addressing such children as her own, as a potential step-mother or co-mother in a polygamous society. Sapir is unsure whether dat’ipla is an inclusive term for son and daughter or is restricted to this and the step-relation; in Northern and Central Yana, did is the regular word for “child,” which Ishi considered peculiar to those dialects. The Yahi step-terms are wa’nimasi, “step-father” and “man’s step-child”; p’imo’o, “step-mother”; and dat’ipla, “woman’s step-child,” used even of a grown-up step-child. The striking point is the identity of the step-mother–step-child relation with that of the paternal uncle’s wife to the husband’s brother’s child—a clear indication of one form of levirate marriage.
The reciprocal term wa’nimasi has no parallel in Yahi, but its Northern Yana cognate un’lma means “paternal uncle.” That points to the direct, typical levirate—a man marrying his brother’s widow—and implies that wa’nimasi formerly meant “paternal uncle” in Yahi, was displaced by the term for “father” under levirate influence, and survived only in the specialized sense of “step-father.” The identification of paternal uncle with father and maternal aunt with mother is peculiar to Yahi; Northern and Central Yana keep distinct terms for paternal uncle, maternal uncle, paternal aunt, and maternal aunt. Since the Yana terms for paternal uncle and maternal aunt are unanalyzable and therefore ancient, and since the Yahi internal evidence points to a former wa’nimasi meaning “paternal uncle,” Sapir concludes that the Yahi peculiarities are secondary—that levirate influence was increasing, not waning. He regrets that Northern Maidu data, which might show whether this emphasis came from southern influence, were inaccessible to him.
How Far the Argument Reaches
Sapir closes with careful limits. The merging of lineal and collateral lines need not indicate group exogamy; the levirate may be equally or more plausible, and each case must be examined on its merits. He expressly restricts his explanation to the two groups of cases he has dealt with. Classifying cousins through parents of like sex as brothers and sisters is symptomatic of the levirate; classifying all cousins as brothers and sisters, as among the Nootka, is not significant. Among the Takelma of southern Oregon, the terms for younger brother, older brother, younger sister, and older sister were also applied to the children of the father’s younger or older brother and the mother’s younger or older sister, while cross-cousins were classed partly with the paternal uncle and maternal aunt and partly under a distinctive term—significant, since the levirate was obligatory among the Takelma. Identifying a wife’s sister or a man’s brother’s wife with the wife, and a husband’s brother or a woman’s sister’s husband with the husband, is presumptive evidence of the levirate. For the Tlingit, Swanton reports that a woman’s sister’s husband was called husband and a wife’s sister, wife, because on the wife’s death the widower had a right to marry her sister; Livingston F. Jones’s work indicates that this “right” was really a duty, and that both forms of the levirate were customary—when a brother died a surviving brother had to take his widow, and if the wife died a sister or close relative had to be given to the surviving husband. In such marriages no presents pass from the man’s people to the woman’s people, “for this is only making good his loss.” The surviving husband may even select a married sister of his deceased wife, who must leave her husband; or the widow may select a married brother of her deceased husband, who must leave his wife and children—the writer knows more than one such case. In Shoshone, the word for “my husband” or “my wife” also means “my brother’s wife” (male speaking) and “my husband’s brother”; in Southern Paiute, the terms for “my sister’s husband” and “my husband’s brother,” and for “my brother’s wife” and “my wife’s sister,” are derivative forms of the words for husband and wife, probably signifying “my co-husband, my co-wife” or “my possible husband, wife.” Many other American examples, Sapir adds, could doubtless be found.
The paper’s significance lies in its method as much as its findings. Sapir treats vocabulary as a historical record: a term that looks like a mere synonym is often a sediment of obligation, and the linguist who parses a word like wa’dat’imauna can recover a marriage custom that the speakers themselves state as plain fact. His insistence that sociological explanation alone will not do—that kinship terms must be analyzed linguistically and read against the uneven pace of cultural change—set a standard for the anthropological study of kinship, and his Yahi material, taken down from Ishi, remains among the most closely argued demonstrations that grammar and custom can be read in the same breath.
