CAIRN-INT.INFO : International Edition
In memory of Françoise Héritier
“Thought is but a flash in the middle of a long night. But this flash is everything.”
Henri Poincaré (1946 [1905]: 293).

1 The question of the relationship between terminology, alliance and sexuality is among those few precious ethnological aporias that have remained unchanged despite currents and trends. Originally posed by the seminal work on kinship studies, Lewis Henry Morgan’s Systems of Consanguinity and Affinity of the Human Family (1870), it is formulated almost identically in contemporary works. It thus finds an echo even in the recent debate on the nature of parental bonds and their relationship with human sexuality (Barry 2012; Carsten 2000; Sahlins 2012; Shapiro 2012; Strathern 1992; Zonabend & Collard 2013), in that it explicitly calls into question the natural and descriptive or cultural and classificatory nature of the sociolinguistic categories governed by sexual and matrimonial prohibitions.

2 There are many clues that point to a probable connection between terminology, marital preferences and sexual prohibitions. These are generally due to the existence of taxa that merge different genealogical positions under the same term. We interpret these assimilations as testifying to a common matrimonial destiny.

3 These equivalences, which are deemed to be matrimonially significant, may bring together the following groups terminologically:

  1. consanguines and affines;
  2. consanguines with other consanguines;
  3. affines with each other.

4 Debates about the possible link between terminology, marital preferences and sexual prohibitions have often been prompted by the assimilation of affines and consanguines under the same terms. It is this phenomenon that has given rise to the greatest number of interpretations, and sometimes continues to do so, particularly in the analysis of Dravidian systems and certain Australian nomenclatures. These interpretations, if not always the most subtle, at least have the merit of seeming plausible due to the simplicity of the reasoning. This is the case with the classical equations that use the same terms for uncles and aunts as for in-laws, or those that designate, under the same taxon, a father’s sister and a maternal uncle’s wife, on the one hand, and a maternal aunt and a paternal uncle’s wife, on the other. These assimilations are unanimously interpreted as signs of a marriage advantageous to Ego in the first case, and of a practice of sister-exchange between paternal and maternal kin in the second.

5 But this merger of consanguines and affines is not the only one to serve as a basis for ethnological speculation, and the merger of different consanguines under the same terminological taxon raises comparable issues. It concerns, in this case, the supposed reproduction of marital behaviour by close relatives of Ego. Thus, if I refer to FB as F and MZ as M, this must surely indicate that my paternal uncle and my father have a common matrimonial destiny and that both can (or could) marry my mother or her sister indiscriminately. Historically, these assimilations of consanguines with other consanguines served as the starting point for the very first ethnological speculations on hypothetical alliance systems whose terminologies, like the archaeologist’s sedimentary layers, were supposed to preserve linguistic vestiges. It was on the basis of such ‘clues’, which seemed so reliable to 19th-century anthropologists—just as those that bring together consanguine and affine terms seem so convincing today—and whose futility only became apparent much later, that we constructed fearsome anthropological chimeras: those of ‘group marriage’ and the ‘primitive horde’, in particular.

6 Finally, the last series of terminological equivalences to which we attribute an indicative character in matrimonial matters is that which aims to assimilate affines with each other. This is the case, for example, when I call my sister-in-law “wife”, which leaves doubt as to whether or not there is a practice of sororate polygyny or serial sororate marriages, or when we have the equations BW = WZ [1] or W = ZHZ, which seem to suggest a marriage of two brothers with two sisters in the first case, or an “exchange of sisters” in the second.

7 Unfortunately, although all these sets of explanations and easy deductions, of which I have provided here only a brief anthology, have the merit of simplicity, they are as capricious as they are obvious. As soon as one tries to untangle the skein further, the thread breaks, exceptions multiply, and inconsistencies are revealed.

8 Ethnologists have long observed that term-to-term correspondences between terminological classes and marriage or incest classes, while sometimes proven, admit a very large number of exceptions. The same term often covers different or even opposite marriage behaviours and, conversely, similar marriage customs sometimes cover distinct terminological taxa. How many terminologies merge under the same name—“sister” or “brother”, for example—individuals who are marriageable and/or with whom sexual relations are permitted (those whom in our Western terminology we would call “cousins”), and others for whom this is absolutely forbidden (those whom we would consider “real” brothers and sisters)? How many others, on the contrary, divide relatives into multiple terminological taxa when in fact they are all matrimonially and sexually inseparable: all marriageable/sexually available or all unmarriageable/sexually forbidden? Ultimately, the number of exceptions is such that it is objectively unsustainable to support the overall hypothesis of a direct association between these institutions—terminology, marriage and sexuality.

9 Some authors (from Lounsbury and Goodenough to Trautmann and Tjon Sie Fat) have understood this and turned to a strictly linguistic formal study of terminology, leaving it to others to inform on more specifically sociological meanings. Others who retain the idea of providing a sociological and matrimonial interpretation of terminological taxa generally do so with great caution, putting forward timid interpretative hypotheses only at very restricted and localized scales: that of the one society they most often study. In reality, it is only in rare cultural areas (southern India, the Amazonian lowlands, Australia, New Guinea) and in very specific nomenclatures of the bifurcate merging type (Dravidian, Kariera, Yafar, etc., but not Iroquois, for example) that the idea of a terminology as a mirror of alliance continues (sometimes) to be explicitly claimed. But in such cases, and being unable to find reasons for the fact that such an association no longer works outside these specific cultural contexts, ethnologists specializing in these regions, while maintaining this hypothesis, tend to mention it only in the context of exchanges between specialists in the regions concerned, considering it as a kind of ethnological theory for regional use—a culturalist fallback which undoubtedly serves to balkanize kinship studies even more than they already are.

10 For elsewhere (and this elsewhere covers most of the known land and terminologies) an examination of the nomenclatures does not make it possible to establish a close correspondence between these and the local matrimonial categories. Therefore, although elsewhere (in Eskimo, Hawaiian, Sudanese, Iroquois, Buryat, Crow-Omaha, etc. terminologies) the seductive idea that there is an association between terminology and marriage is not entirely abandoned, it is no longer purported as a universal proposition. It is referred to more timidly, on a case-by-case basis, often to account for a specific term, but never for a terminology as a whole [2].

11 While the construction of terminological taxa would still have structural value in matrimonial matters for ethnologists in the Dravidian field, it would apparently only have an indicative value for ethnologists in all other fields.

12 This notwithstanding, it is my opinion, indeed the hypothesis I am defending herein, that there is not a real discontinuity in the overall logic of terminologies that would cause some of these to reflect the rules of marriage or prohibitions on incest while others ostensibly deviate from those.

13 To my mind, the relative instability of the correspondences between terminology, sexuality and marriage that we see in some cultural complexes have less to do with the intrinsic weakness of the relationship than with the fact that it is not as immediate or as naive as assumed. Contrary to long-held belief, from the earliest works to the most recent investigations, this relationship does not establish a term by term correspondence between terminological classes and marriage classes, but exploits the correspondences between the structural variables that make up these classes.

14 In order to understand what I mean by this, I must first make a technical digression regarding the construction of the basic principles common to all terminologies. Only once we have a better grasp of that—with its degrees of freedom, its constraints and its limits—will we be able to fully appreciate the relationship between producing terminologies or term formation and the logic of alliances [3].

Fundamentals in the Production of Terms

15 In order to conduct an analysis of the rules of production or formation for terms that constitute kinship nomenclatures, it is necessary to begin by considering terminologies not from a typological point of view—as belonging to specific “types” (Iroquois, Dravidian, Eskimo, Hawaiian, etc.)—but as possible and finite expressions of a combinatorial system based on a few very simple rules.

16 In the remainder of this text, I will try to demonstrate that all kinship terminologies are built on a common logical ground organized around one postulate or axiom and two very simple rules. Of course, the diversity of terminological variants implies that these rules are themselves subject to minor adaptations and additions of secondary variables (I will provide an example with Pashtun terminology in Appendix II). Nevertheless, whatever their intricacies may be, it is always possible to analyse them ab initio by following this postulate and these two simple rules. I shall therefore begin by explaining them in order to illustrate how they work, using a few simple practical examples. I will then show not only that these principles (postulate and rules) are necessary and sufficient to begin constructing linguistic categories for all real terminologies, but also and perhaps above all that their specificities make it immediately possible to understand, in an elegant and economical way, the reasons justifying the absence of certain terminologies that may be conceived theoretically but do not exist empirically, without resorting to supernumerary factors that construe kinship terminologies as heteronomies.

17 The final part of the text will be devoted to examining a possible reconciliation of these various orientations, which guide the construction of terminologies and organize the matrimonial systems with which they are most often associated.

Conventions and Rules

Postulate and Rules

18 As I have just indicated, in order to account for the main types of terminology in G + 1 (Eskimo, Hawaiian, Sudanese and Iroquoian) [4] and also for the absence of other theoretically possible forms, we need only to posit one postulate and two unique rules, each of which may take two states: true or false (active or inactive) [5]. The assumption is simple: it consists in considering that there are only three relationships involved in terminological constructions: a relationship of ascendancy, a relationship of descent and a relationship of affinity.

19 The relationship of collaterality (of siblings or cousins, for example) is thus thought of as a “complex” relationship built from the “simple” relationships that link Ego and Alter through one or more common apical ancestors. A brother, for example, will be considered a “male descendant of an ascendant (or a pair of ascendants) of Ego”. Of course, this postulate concerns only the strictly logical mechanisms governing the construction of terminologies; it has no indicative or predictive value as to the real sociological importance of such or such a relationship. More precisely, this means that there would not in any way be a sociological subordination of relations of siblinghood to relations of filiation or affinity.

20 Let us now look at these two rules:

  • Rule 1, of relational equivalence: a relationship may or may not be assimilated to a simpler equivalent relationship (the state of the rule will be considered “true” in the first case, “false” in the second); relationships can always be assimilated with each other if they are identical;
  • Rule 2, of gender relevance: the gender of the intermediate link is considered either irrelevant or relevant (again, the state of the rule will be considered “true” in the first case, “false” in the second).

21 Rule 2 corresponds approximately to the criterion that ethnologists call “bifurcation”. However, this concept is only descriptive: it informs us of the effect of applying this rule (the terms “bifurcate” depending on whether they refer to the father’s side or the mother’s side), but it tells us nothing about the reasons behind this “bifurcation”. It therefore seems preferable to replace this purely descriptive concept with a rule that not only describes it, but also makes it explicit, especially since this rule is based on an extremely simple phenomenon: whether or not to take into account the gender of the link through which a relationship is established.

22 Rule 1, based on the notion of “relational equivalence”, has no direct equivalent in our ethnological categories. It covers two phenomena that ethnologists consider separately: the so-called “merging” of bifurcate merging type terminologies (Iroquois, Dravidian, etc.), and also the phenomenon (which does not have a specific name) that leads to the assimilation of some or all collateral and direct lineal descendants in generational (Hawaiian) or lineal (Eskimo) terminologies. The notion of “relational equivalence” thus reflects, in practice, the possibility that different genealogical descriptions may refer to the same terminological object.

Examples of the Application of the Rules

Examples of the application of the rule of relational equivalence (Rule 1)

23 Let us examine, for example, the prevailing description for a paternal uncle of Ego, which following our assumption that avoids the direct use of the relationship of siblinghood, is literally a “male descendant of an ascendant (or of a couple of ascendants) of a male ascendant of Ego” [XH-H] [6].

24 This description is, of course, identical to the prevailing description for each of Ego’s other paternal uncles and there will therefore be only one term to designate them all. By contrast, it is formally different from that of the father: “male ascendant of Ego” [X(H)]. Thus, a priori there should be two different terms for these two descriptions because they are not identical. However, although it is not identical, the description “male descendant of an ascendant (or of a couple of ascendants) of a male ascendant of Ego” is a formally correct, though contrived way (logicians speak in this case of a “well-formed formula”) to describe the position of the father of Ego. Formulated differently, Ego is in fact one of the “sons” of his own ascendants. I will thus say that these two positions (for F and FB), insofar as it is possible to group them under the single description of “male descendant of the ascendants of a male ascendant of Ego” [XH-H], and all propositions that describe a relation to the same individual(s) in a formally correct way, are not identical but equivalent [XH-H ~ X(H)].

25 Therefore, we have two scenarios. In the first, Rule 1 is not applicable (its state is “false”), and the father and his brother, although described by equivalent relationships, will be given two different terms. We will thus obtain a terminology that distinguishes F <> FB. In the second, we apply Rule 1 (its state is “true”) and, therefore, we must reduce two equivalent relations to the simpler of the two. In the latter case, I will therefore reduce the relation “father’s brother” to the simpler equivalent relation “father” and state that a “male descendant of the ascendants of a male ascendant of Ego” can be assimilated to the simpler description of a “male ascendant of Ego” [XH-H ~ XH so XH-H: XH]. This means that in this terminology, a second term is not created, but the one used for the father is reused to refer to the paternal uncle. Of course, the same reasoning can be applied to the description of the mother’s sister who, if the rule of “relational equivalence” is active, will be assimilated to that of the mother. It goes without saying that in such a case, this rule will then cascade to all relationships using the expression “father’s brother” [XH-H], which simply means that the principle of relational equivalence is indeed a transitive principle.

26 Thus, the description for the “children of the paternal uncle” [XH-HX] contains within it the description for FB [XH-H] (they are themselves the “descendants of a male descendant of ascendants of a male ascendant of Ego”; the elements of the expression subject to assimilation are underlined), and thus become “father’s children”, in other words siblings of Ego [XH-H ~ X(H) then XH-HX ~ X(H)X and thus XH-HX : X(H)X]. Obviously, this assimilation will also apply to the children of the maternal aunt who, according to the same procedure, will be assimilated to “mother’s children”, and therefore to siblings of Ego.

27 Let us take another example, that of nepotic relationships. The description for the children of the brother of a male Ego, literally “descendants of a male descendant of the ascendants of a male Ego” [H-HX], is equivalent to one of the formally correct descriptions that can be given for the children of a male Ego (as a male Ego, my own children are indeed “children of a son of my parents”). The first expression may therefore be reduced to the second if Rule 1 is active. In other words, the children of a man’s brother are terminologically assimilated to that man’s children [H-HX: (H)X]. The same will be true for a woman’s uterine nephews and nieces, as long as the state of Rule 1 is true.

28 By contrast, this assimilation will no longer be valid for “cross” nephews (uterine for a male Ego, agnatic for a female Ego), because in this case the descriptions of Ego’s children and her nephews and nieces will not be equivalent. Thus, the description for a woman’s agnatic nephews and nieces (“descendants of a male descendant of the ascendants of a female Ego” [F-HX]) is not a possible (formally correct) alternative description for a woman’s children: “descendants of a female Ego” [(F)X]. Likewise, the description for a man’s uterine nephews and nieces (“descendants of a female descendant of the ascendants of a male Ego” [H-FX]) is not a formally correct description for a man’s children, the “descendants of a male Ego” [(H)X], precisely because of the explicit mention of the different genders involved in these pairs of descriptions.

What about Ego?

29 Let us note in passing that in these last few examples (where BCh (ms) = Ch and ZCh(ws) = Ch), just as in the previous ones involving the assimilation of descendants of a paternal uncle or maternal aunt to siblings of Ego, I have reduced a collateral relationship to a direct relationship. But in that case, why may I not do the same for Ego in relation to Ego’s own brothers or sisters?

30 In our hypothesis, which precludes the use of a relation of siblinghood and only allows the use of relations of ascendancy, descent and affinity, a brother of Ego will be formally described as a male descendant of an ascendant of Ego [or X-H in positional notation]. This description for one of Ego’s brothers of is therefore perfectly identical to that for another brother [X-H]; likewise, the description for a sister as a “female descendant of an ascendant of Ego” [X-F in positional notation] is identical to that for another sister [X-F]. However, the descriptions for a brother and a sister are different from each other, in that they explicitly mention two different genders for Alter. Because of Rule 1, there will be a single term for each set of same-sex siblings whose relationships are identical [X-H is identical to X-H and X-F to X-F], but different terms for opposite-sex siblings [X-H is not identical to X-F] [7]. There will thus be, in the vast majority of cases, an assimilation of same-sex siblings to each other under a single term, rather than a distinction in as many groups as Ego has brothers and sisters, without there necessarily being a merger of siblings of both sexes into a group of “siblings” undifferentiated by gender.

31 But in that case, why does this reasoning no longer apply to Ego itself? Why does Ego not self-designate as a “brother” or “sister” according to sex? The description “male descendant of the ascendants of a male Ego” [H-H] that I use to represent a brother of a male Ego, or the relation “female descendant of the ascendants of a female Ego” [F-F] that I use to refer to the sister of a female Ego are, indeed, also formally correct descriptions to designate Ego itself (Ego is indeed “a child of its own parents”). In this case and thanks to the rule, this contrived but formally correct description could be assimilated to a simpler equivalent description that would allow Ego to be designated.

32 However, in the hypothesis that I am defending here, kinship terms always refer to relationships and not individuals, and although the term Ego does indeed represent an individual, it does not express any relationship. The fact that the Ego has no relation to itself is supported by a fact so obvious that it often goes unnoticed: the universal absence of a kinship term designating the “I”. The Latin “Ego”, which psychoanalysts and ethnologists are fond of, is only an equivalent of the English “I”; in other words, a personal pronoun, not a term of kinship. Consequently, the relationships expressed in the descriptions for the brothers of a male Ego or the sisters of a female Ego [H-H and F-F] cannot, according to Rule 1, be reduced to Ego, because there is simply no terminological description for Ego itself (even though it could appear as an intermediate link in a terminological relationship).

33 Rule 1 therefore allows for the assimilation of a relationship to a simpler relationship, but it does not allow for the transformation of a relationship into a non-relationship. There will therefore always be, in all terminologies, a term for brothers and/or sisters as a whole that will never be reduced to the “non-term” of kinship which is the “I”, the first person singular speaking subject.

Examples of the application of the rule of gender relevance (Rule 2)

34 To understand how Rule 2 works, let us take two relationships of the same level of complexity, those of a paternal uncle and a maternal uncle [XH-H and XF-H]. These two relationships are not identical to each other (the description for a “male descendant of a male ascendant of Ego” is not identical to that of a “male descendant of a female ascendant of Ego”) and there is no formally correct description of the former that is equivalent to the latter. Therefore, these two relationships cannot under any circumstances be assimilated by adhering to Rule 1 alone.

35 However, if I now consider Rule 2 to be active, then the gender of the intermediate links (respectively the father and mother of Ego) is no longer taken into account in the relationship (whereas the gender of Ego or Alter still is).

36 The two relationships may therefore be rewritten as follows:

  • a paternal uncle described as “male descendant of an ascendant of a male ascendant of Ego” becomes “male descendant of an ascendant of an ascendant of Ego” [XH-H: XX-H]. The gender of Alter is still noted here, but that of the father of Ego is no longer noted because, let us recall, Rule 2 acts only on the intermediate links in a terminological relationship, never on Ego and Alter themselves;
  • a maternal uncle whose description was “male descendant of an ascendant of a female ascendant of Ego” becomes, according to the same procedure, a “male descendant of an ascendant of an ascendant of Ego” [XF-H: XX-H].

37 The descriptions for a paternal and maternal uncle are therefore both reduced to one, that of a “son of an ascendant of an ascendant of Ego”. Therefore, if Rule 1 of “relational equivalence” is also active, it may be written:

  • FB = MB: PB [XH-H = XF-H: XX-H].

38 Rather than two terms—one for the paternal uncle and one for the maternal uncle—I now have only one generic term for “brother of an ascendant”; in other words “uncle”. Other assimilations are also possible based on this example. Indeed, the description “male descendant of an ascendant of an ascendant of Ego” [XX-H] that we have just obtained for these “uncles” is also a formally correct description for the father of Ego: he too is “a son of the parents of an ascendant”.

39 In this way, if Rule 1 of “relational equivalence” applies, we will not use a specific term for these “uncles”, but we will instead use the term already used to designate the father, because these two relationships will be assimilated to a simpler equivalent relationship: “male ascendant of Ego” [XX-H: X(H)]. In other words, in the latter case: F = FB = MB. If, on the contrary, Rule 1 is not active (if its state is “false”), then these uncles will be given a specific term common to them that differs from that used for the father of Ego, thus F <> (FB = MB).

40 In the first case (Rules 1 and 2 true), all the men of generation + 1 will be “fathers”, which corresponds to a Hawaiian-type nomenclature, and in the second (Rule 1 false, Rule 2 true), there will be one father and several uncles, the latter not being distinguished from each other, and the terminological equation will be of the Eskimo type.

41 This subdivision of the effects of activating Rule 2 according to the state of Rule 1 works in the opposite direction. If we consider Rule 2 to be inactive, then the gender of the intermediate link is considered relevant and there will be separate terms for FB and MB. This configuration will in turn be subdivided, conditionally, into two distinct configurations depending on whether or not Rule 1 is active. FB can be assimilated to F by distinguishing them from MB in the first case, where they will continue to be distinguished from each other, and from F in the second case. We will therefore obtain either a system where (F = FB) ≠ MB, or another where F ≠ FB ≠ MB. As can be seen, these last two classifications correspond respectively to the classic bifurcate merging terminology (Iroquois/Dravidian, etc.) in the first case, and Sudanese/descriptive terminology in the second.

Configuration of Rules

42 Depending on the permutations of the two states (true or false) of our two rules, we can thus obtain four—and only four—possible taxonomic arrangements for G + 1.

43 As we have just seen, these four configurations correspond precisely to the four major kinship terminology systems existing at this terminological level (G + 1). The following table will allow us to verify this and to understand the categories produced by this combination of our two rules in a concise manner.

Table 1

Effects of the application of terminology rules in G + 1

Rule 1 is “true”:
equivalent relationships can be assimilated
Rule 1 is “false”:
equivalent relationships cannot be assimilated
Rule 2 is “true”:
gender is irrelevant
Rule 2 is “false”:
gender is relevant

Effects of the application of terminology rules in G + 1

44 Of course, the application of these rules can be extended to any other terminological level. In the preceding pages, I have given many examples of such applications of these rules at G 0 (siblings and cousins of Ego) and G-1 (children and nephews and nieces). However, the reader can in fact easily verify that Rules 1 and 2, in all possible states at each genealogical level, are sufficient to produce the terminological assimilations and distinctions that we find in classical terminological arrangements for all generations adjacent to Ego.

Iroquois and Dravidian Terminologies (Bifurcate Merging)

45 In bifurcate merging terminologies (Iroquois, Dravidian, Kariera, etc.) we can assimilate equivalent relationships, so according to Rule 1 of relational equivalence we can therefore assimilate some uncles to fathers and some aunts to mothers. However, because the gender of the intermediate links is relevant (Rule 2 is “false”), only the agnatic “uncles” will take the term “father” and only the uterine aunts will take the term “mother”:

  • [XH-H: X(H) and XF-F: X(F)].

46 Maternal uncles must therefore adopt a specific term, and paternal aunts another one (they cannot use the same term because they are not of the same sex).

47 In the same terminologies and according to the same logic, children of the paternal uncle and those of the maternal aunt will become “father’s or mother’s children” and be assimilated to siblings, whereas those of the maternal uncle and the paternal aunt cannot be assimilated and will be assigned a specific term:

  • FBCh: Sib and MZCh: Sib [XH-HX: X-X and XF-FX: X-X].

48 Again, according to the state of these two rules, children of same-sex siblings will be assimilated to children of Ego, since the relationships “child of a male descendant of a man’s parents” and “child of a female descendant of a woman’s parents” [H-HX and F-FX], which describe the relationship to a man’s agnatic nephews and nieces and a woman’s uterine nephews and nieces, are two formally correct ways of describing the children of a man as opposed to the children of a woman [(H)X and (F)X]. Since Rule 1 is true, agnatic nephews and nieces are assimilated to children for a man and uterine nephews and nieces are assimilated to children for a woman. However, “cross” nephews and nieces (the children of the sister for a man, or of the brother for a woman) may not be assimilated:

  • BCh: Ch(ms) and ZCh: Ch(ws) [H-HX: (H)X and F-FX: (F)X].

49 These assimilations and distinctions that the state of our two rules for nepotic relations allows are precisely those that we find in the real bifurcate merging type terminologies.

Eskimo Terminology

50 In Eskimo terminology, which is the exact opposite of Iroquois terminology in the logic of our rules, equivalent relationships cannot be assimilated. There are “uncles” and a “father”, “aunts” and a “mother”. However, the intermediate gender is no longer relevant and, as a result, the descriptions for “paternal” and “maternal” relationships are now identical and are therefore assimilated (Rule 1). Thus, paternal and maternal uncles become brothers of “parent” (of unmarked gender) and maternal and paternal aunts become sisters of “parent”. FB and MB, as well as FZ and MZ, can thus be grouped together under the same taxon, while distinguishing them from F and M respectively:

  • FB and MB: PB and FZ and MZ: PZ → FB = MB and FZ = MZ [XH-H = XF-H and XF-F = XH-F].

51 Because of this assimilation, the children of these uncles and aunts will be distinguished from siblings and assimilated with each other. Indeed, when an additional step is added to the relationship with uncles and aunts, formerly in the position of Alter, they will henceforth occupy the position of “intermediate links” in the relationship between Ego and Alter. Since the intermediate gender is not taken into account because Rule 2 is active, the descriptions “children of uncles” and “children of aunts” are changed to “children of siblings (of undifferentiated gender) of parents” [XX-HX and XX-FX: XX-XX], and therefore to “cousins”. However, since Rule 1 is not active, these may not be assimilated with the formally equivalent relationship used to designate siblings:

  • FBCh and MBCh: PSibCh and FZCh and MZCh: PsibCh thus FBCh = MBCh = FZCh = MZCh: PSibCh [XH-HX = XH-FX = XF-HX = XF-FX: XX-XX].

52 The same phenomenon occurs with nephews and nieces: since Rule 1 is false, they cannot be assimilated to children as in the case of bifurcate merging terminology. However, since Rule 2 is true, the gender of the intermediate “link” no longer has to be taken into account, and the “descendants of a male descendant of an ascendant of Ego” (“children of brothers”) or the “descendants of a female descendant of an ascendant of Ego” (“children of sisters”) all become indiscriminately “children of siblings”:

  • BCh and ZCh: SibCh [X-HX and X-FX: X-XX].

53 In other words, they are nephews and nieces, which may only be distinguished according to their respective gender and no longer according to that of their parents.

54 Once again, the application of this theoretical model fits perfectly with the empirical data that we know, and this grouping of agnatic and uterine nephews and nieces in the same set, which is itself distinguished from the set of siblings, corresponds precisely to what occurs in the case of real Eskimo terminologies.

Hawaiian Terminology

55 For Hawaiian terminologies, the set of rules is even simpler since both rules are considered “true”. All uncles and aunts will therefore be assimilated to father and mother according to Rules 1 and 2:

  • FB and MB become PB, and FZ and MZ become PZ according to Rule 2 [XH-H: XX-H and XF-H: XX-H];
  • then PB and PZ become F and M according to Rule 1 [XX-H: X(H) and XX-F: X(F)], and so we obtain the following equations: FB = MB = F and MZ = FZ = M.

56 In turn, all their children will be assimilated to siblings of Ego. Rule 1 immediately assimilates parallel cousins, since “siblings” and “parallel cousins” are formally equivalent descriptions here:

  • The FBCH, i.e. “descendants of a male descendant of an ascendant of a male ascendant of Ego” are assimilated to “descendants of a male ascendant of Ego” due to Rule 1 of relational equivalence [XH-HX: X(H)X], i.e. to “father’s children”, therefore to siblings. The same applies, of course, to MZCh who for the same reason become “mother’s children”, therefore siblings [XF-FX: X(F)X];
  • then Rule 2 assimilates cross cousins to children of uncles and aunts because it “erases” the gender of intermediate relationships: FZCh become FSibCh and MBCh become MSibCh [XH-FX: XX-FX and XF-HX: XX-HX];
  • finally, Rule 1 applies once again and assimilates them to siblings, since the previous descriptions obtained, FSibCh and MSibCh, have become formally correct descriptions for FCh and MCh. We thus obtainFSibCh: FCh and MSibCh: MCh [XX-FX: X(F)X and XX-HX: X(H)X].

57 With regard to the sets of nephews and nieces, they will be treated here as “children”, regardless of the gender of Ego. Rule 2 will first make them “children of siblings”:

  • the “descendants of a male descendant of an ascendant of Ego” and the “descendants of a female descendant of an ascendant of Ego” are assimilated to “descendants of a descendant of an ascendant of Ego”, therefore to “children of siblings” [H-HX: H-XX and H-FX: H-XX, F-FX: F-XX and F-HX: F-XX].
  • then Rule 1 will transform them into “children”: the “descendants of a descendant of an ascendant of Ego” are assimilated to “descendants of Ego” [H-XX: (H)X and F-XX: (F)X].

Sudanese Terminology

58 Finally, Sudanese terminology is the exact opposite of Hawaiian terminology, since neither Rule 1 nor Rule 2 is “active”. Thus, following the model outlined here, no assimilation will occur and each of the types of uncles and aunts, cousins, nephews and nieces will be referred to by separate terms.

59 This is what we find in real Sudanese terminologies, which in principle combine a descriptive classification in G 0 and G -1 with a specifically Sudanese terminology in G + 1.

60 By considering the application of the two states (true or false) that these two rules can take, as I have just done for some significant positions, I think we understand better what justifies the congruence of the types of assimilation most often found at different genealogical levels in the same terminology. For example, there is the fact that the partial assimilation of cousins depending on the gender of the intermediate parent in a bifurcate merging type terminology is accompanied, in the vast majority of cases, by a partial assimilation according to the gender of the intermediate parent. This congruence of equivalences, which we often find at various levels of a nomenclature, reflects the simple fact that it is the same set of rules, the same logic that guides and directs the production and consistency of the set of terminological taxa.

61 Indeed, by using the principles outlined here, we can also unravel a number of aporias that haunt the classical anthropological analysis of kinship terminologies. One of the most important of these is the non-existence of certain terminologies that are conceivable in theory. For example, Françoise Héritier (1981) rightly noted the empirical absence of the equation where (MB = F) ≠ FB, to which should, of course, be added its female counterpart FZ = M ≠ MZ, which is also absent. However, this absence is not a terminological hapax and we can in fact enumerate several others. There are, indeed, several terminological equations that are theoretically possible but never realized in practice. A few examples involving similar terminological categories include the following:

  • (MBCh = FZCh = Sib) ≠ FBCh = MZCh
  • (FZCh = Sib) ≠ (MBCh = FBCh = MZCh)
  • (MBCh = Sib) ≠ (FBCh = MZCh = FZCh)
  • (Ch(ms) = ZCh(ms)) ≠ BCh(ms)
  • (Ch(ws) = BCh(ws)) ≠ ZCh(ws))
  • etc.

62 All these examples of imaginable but absent equations, including the one noted by Françoise Héritier and those I have just listed, are in reality only conceivable in a framework where terminological taxa proceed from a purely stochastic arrangement and where all possible permutations would therefore be producible and indeed produced.

63 However, these same equations are no longer viable nor conceivable, since it is assumed that the terminologies derive not from chance but from logic, specifically if we believe that they follow the rules of production of terms as I have just mentioned.

64 It is not really necessary to resort to ad hoc sociological considerations to account for the absence of these particular equations, since these absences are only due to the various possible states of a specific combinatorial system. In fact, it is sufficient to list all the possible states of our two rules for each of the terminological levels considered here (G + 1, G 0, G -1; a relatively simple analysis to perform) to see that not only do we obtain all the terminological equations that we find in actual nomenclatures, but more importantly that no combination of these two rules, regardless of their state (true or false), will ever lead to one of the terminological configurations listed above; in other words, to one of those equations that we can of course imagine, but which ultimately do not occur in reality, and have no tangible correspondence to empirical ethnographic data.

Principles of Convergence

65 Of course, what has just been presented only touches on the subject of terminology construction. Many more explanations would still be needed to account for terminologies that include “terminological slices” (“tranches terminologiques”) which borrow from several of the logical categories identified here. Similarly, it would be necessary to continue this analysis further in order to understand the more complex mechanisms at work in certain terminological sub-categories, particularly with respect to denotations used for distant terms (for example, for the generation of grandparents or classificatory cousins in Dravidian equations, for skewing rule equations in Crow-Omaha terminologies, etc.) or, quite simply, to understand the link between consanguine and affine terminology [8].

66 However, although the logical mechanisms that have just been outlined require even more detail for an exhaustive exploration of the field of study of kinship nomenclatures, they are nevertheless sufficient at first glance to understand the logic of terminological fundamentals, the criteria that are always present in all nomenclatures. This will therefore allow us to answer our initial question: what relationships can be maintained between terminology and alliance? In order to do this, we need to provide some sociological content and flesh out the two rules that we examined earlier.

67 It is obvious that Rule 1 of relational equivalence refers to the form of a genealogical space and is built on the notions of distance and proximity, while Rule 2 of gender relevance refers more to its content and sets out the differentiating factor which, by specifying the nature of the relations operating in this space, will allow its form to be worked out.

68 In my opinion, these operations on content and form are not aimed at a precise matrimonial order, as is implied by the systematic correspondence that ethnologists assume between terminological and matrimonial classes. They do not seek to assimilate potential affines with each other, nor to distinguish them from unmarriageable consanguines, but rather to distinguish radically opposed categories of kinship, taxa that are perfectly antithetical.

69 This distinction of opposing categories of kinship will therefore be carried out on the basis of the notion of “kinship”, as I have considered it in other works (Barry 2008 and 2012), and not on the basis of the concepts of “descent” or “alliance”. Thus, within the limits of the expressions made possible by the interplay of the two rules mentioned, the terminological categories adopt a configuration which, in each case, will best distinguish kin (who one cannot marry) from those who are not kin (and who are therefore possible spouses). It does not matter, then, whether in order to achieve this opposition they assimilate one of these two extremes with individuals who do not fall under either of these extremes, or, on the contrary, distinguish individuals who are in roughly equivalent situations in terms of kinship and therefore of marriageability. The following examples will make these remarks more readily understandable.

Terminology and Unisexual Kinship Principles

70 In the theoretical exposition developed in the book La parenté (Barry 2008), I detailed four main forms that can be adopted by “kinship groups”—individuals who consider themselves “relatives”—which, in my opinion, are at the root of the incest prohibitions and alliance systems identified by ethnologists.

71 I have referred to these four principles as the “principles of uterine, agnatic, parallel and cognatic kinship”, depending on the mode(s) of affiliation with these groups. Here I simply wish to recall the main implications of each of these principles for the form of kinship and alliance systems, and to harmonize them with the types of terminology associated with them. Indeed, it seems clear to me that the logic at work in these principles of kinship, whose primary function is to regulate marriage prohibitions and possibilities, is very close to that governing the production of terms in nomenclatures.

72 In my 2008 essay, I argued that a system based on a principle of uterine kinship, which underlies practices related to what is called “Arab marriage”, for example, assumes that the “true kin” are primarily relatives in the female line, whom one cannot marry (and who are, indeed, generally forbidden in societies practising “Arab marriage”). The category diametrically opposed to this one, which groups together individuals who are no longer truly “kin”, was therefore made up of relatives in the strictly agnatic line, who were therefore the preferred spouses. Meanwhile, cognates were in an intermediate grey area, where the difference between maternal versus paternal cross kin was no longer truly relevant, and, without ever being preferred, they could in principle be married. How could terminology therefore 1) mark that in this system the primary opposition is based on the unisexual character (uterine versus agnatic) of the relationship between Ego and Alter, and 2) emphasize the fact that one of the two sexes “carries kinship” whereas the other is no longer really a carrier of a shared kinship?

73 If we stay within the “classical” terminological forms identified above and apply our two rules, of gender relevance and relational equivalence, there is only one form that can fully satisfy both of these conditions. This form is the one proposed by Sudanese/descriptive terminology. Iroquois terminology, for example, would of course emphasize the importance of the gender criterion (by distinguishing “unisexual relationships”—i.e. “parallel kin”—from others—i.e. “cross kin”), but at the same time it would combine the two unisexual lines—those of patrilateral and matrilateral parallel kin—within the same category. As we have just seen, these two unisexual lines are situated at opposite ends of the spectrum of kinship and marriageability: the uterine line is that of matrimonially prohibited or avoided “kin”; the agnatic line that of matrimonially preferred quasi “non-kin”.

74 In order to radically distinguish agnates from uterines based solely on the interplay of the terminological rules that I stated above, there is only one solution: to distinguish all consanguines at the level of uncles and aunts, and, primarily, at the level of cousins. In doing so, we will achieve the desired differentiation between the two categories that are perfectly opposed from the perspective of kinship, namely those of patrilateral and matrilateral parallel kin. But, to do this, we must also distinguish cross patrilateral and matrilateral cousins without this distinction being in any way significant (or at least necessarily significant). This latter distinction is, in fact, merely a correlation of growth[9], rendered indispensable by the desire to distinguish parallel kin from one another, while resorting only to the expressions made possible by the different states of the rules (1 and 2) for the production of terms.

75 As can be seen, the distinction into four classes of cousins in Sudanese terminology therefore neither reflects the existence of four distinct matrimonial categories, nor does it reflect a marked opposition between siblings and cousins from the perspective of marriage (uterines are distinguished here from siblings, for example, whereas they are in principle not marriageable either). The distinction does not even distinguish between marriageable versus unmarriageable cousins (patri- and matrilateral cross cousins and agnatic parallel cousins are all distinguished here even though they are in principle all possible spouses). We understand that the distinction into four classes specific to Sudanese terminology is simply the only possible terminological expression that can be used to oppose two unisexual categories—those of patri- and matrilateral parallel kin—which must not be combined in the same matrimonial group.

Sudanese Terminology and Uterine Kinship

76 If we accept this reading, we are also admitting the strength of the link between these two perfectly congruent forms from the perspective of the opposition of kinship categories: that of a Sudanese/descriptive terminology and that of the practice of “Arab marriage”.

77 Indeed, this association—which, however, is never discussed by anthropologists working on “Arab marriage”, precisely because it does not obviously associate terminological categories with matrimonial classes—simply corresponds to the strongest positive correlation between a nomenclature and a marriage practice. Even if this fact is obscured in the ethnological literature on kinship, this interdependence is, in the final analysis, much more pronounced than the much more well-known and discussed association between Dravidian terminology and cross-cousin marriage. Thus, if we look at the sample of societies [Table 2] that Georges P. Murdock (1967) coded as practising “Arab marriage” (code Qa), we find that there are 24 cases of a total of 27 for which we also have information on terminology. Of these 24 “useful” cases, 19 associate “Arab marriage” with Sudanese/descriptive terminology (codes D and S), i.e. 79% of the total.

Table 2

Association between “Arab marriage” and terminology in Murdock’s data

SocietyType of marriageNomenclature
BajunQa (= “Arab marriage”)D
Afghans (Pushtun)QaE

Association between “Arab marriage” and terminology in Murdock’s data

78 Let us examine in more detail the five “exceptions” to this association between “Arab marriage” and Sudanese/descriptive terminology:

79 — The first of these exceptions, the Kurds coded Z, in fact also distinguishes parallel cousins from each other as Murdock himself points out, but does so using a specific formula whereby one term assimilates FZCh to MZCh, another designates FBCh and a third designates MBCh.

80 — Balinese terminology, which Murdock codes here as Hawaiian (H), has actually been “standardized” by this author. In fact, Balinese nomenclature does allow the two sets of parallel cousins to be distinguished, but it places agnatic parallel cousins in one category and in a different category, all other so-called cousins “by women” (which is also symptomatic of the effect of the uterine bond in “Arab marriage”). For while the same term, misan, applies to all first cousins (second cousins being called midon), a distinction can be made between PPC (misan kapurusa, “cousin in the direct male line”) and other cousins (MPC, MCC and PCC known as misan ulian luh, “cousins through females”) (cf. Geertz & Geertz 1975: 200, note 40). It is only by omitting to mention this particular taxon reserved for FBCh that Murdock was able to classify Balinese terminology as “Hawaiian”. Let us recall that, in the hypothesis that I have just outlined here, what is supposed to be sought in the terminologies associated with “Arab marriage”, making them congruent with it, is less a particular “type” of nomenclature than, to be precise, the existence of this differentiation between the two categories of parallel cousins that Balinese terminology also makes, in its own unique way.

81 — We find exactly the same classification for a third case, that of the Pathans (Pashtuns), which Murdock has also “standardized” in “Hawaiian” terminology. However, as Fredrik Barth tells us:


“Pathans distinguish between Father, Mother, Father’s Brother, and Mother’s Brother, but classify Father’s Sister and Mother’s Sister together. The terms for the two kinds of uncle and for aunt are extended to first cousins and second cousins of the parents as well. Sibling terms are extended to the children of all these persons, except to the children of Father’s Brother, real or classificatory (tre), for whom there is a special term (tarbur). A differentiation of types of ‘siblings’ may of course be expressed, but only by constructions such as ‘aunt’s son’ (da tror zoe) or ‘mother’s brother’s daughter’ (da mama lur). Patrilateral parallel cousin is uniquely separated from all other cousins and siblings by a separate term. Furthermore, this term carries the subsidiary connotation of ‘enemy’”.
(1959: 11; original italics)

83 As with Balinese terminology, the Pathans therefore separate the terms for patrilateral and matrilateral parallel cousins. The fact that they merge the latter with the cross cousins while Sudanese terminology distinguishes them (and distinguishes between cross cousins) has no real impact on the link established between terminological and matrimonial logics. For in the model that I support here, whether in this particular terminology or within the framework of classical Sudanese terminology, this link is based solely on expressing the necessary opposition of categories that are extremes from the perspective of kinship (and therefore of marriage): the two types of parallel consanguine cousins in the case of those societies that have a “principle of uterine kinship” and therefore a matrimonial system based on the practice of “Arab marriage” [10].

84 — The fourth case, that of the “Afghans”, actually corresponds in Murdock’s case to the Pushtun and, more specifically, to the Ghilzai tribe (also called Ghaljis or Khiljis; one of the two great Pushtun confederations of Afghanistan along with that of the Durrani). However, the term “Pushtun” (or “Pukhtun”) is actually only a self-ethnonym used in Afghanistan by those who speak “Pashto” and who are also referred to as “Pathans” (Pashtuns). My earlier remark about Pathan terminology, which distinguishes between children of the father’s brother on the one hand and other cousins on the other, is of course also valid here.

85 — Of the five “exceptions” identified in this table, perhaps only the last one, the “Kanawa” (the term does not refer to an “ethnic group”, but to the Hausas, the inhabitants of the emirate of Kano in Nigeria), is based on an “Iroquois” classification in G 0 and does not seem to make the same terminological distinction between parallel cousins. However, here again, it is noted that the terminology is Sudanese in G + 1 (MB: kawu; FB: abbani; FZ: babani, gwaggo; MZ: inna, iya; M: uwa, inna, iya; F: uba, baba) and that, while the term for “cousins” (aboki or taubashi) refers to FZCh and MBCh, parallel cousins may be named descriptively (‘yan maza zar, for the children of two brothers and ‘yan ma/ta zar, for those of two sisters, the term zar therefore indicating something made up of one and the same gender, unmixed) rather than by the terms used for siblings [11].

86 Thus, even if one were to accept the Hausa case as an exception, the fact remains that 23 of the 24 societies listed in this corpus, or 96% of cases, associate “Arab marriage” with a terminology based on 1) taking gender into account and 2) distinguishing categories for agnatic and uterine parallel kin [12]—in other words, on the fundamental elements that make it possible to characterize a “principle of uterine kinship”. Such a perfect correlation seems to me to be particularly meaningful and amply supports the thesis on the relationship that links terminology and marriage in that this relationship is based on the opposition of kinship categories (kin versus non-kin), rather than on a simple relationship of equivalence of terminological and matrimonial taxa as is generally assumed.

87 It is also very significant to note that, in these systems based on the principle of uterine kinship, those practising “Arab marriage”, the category of strict agnates—quasi non-kin, therefore—tends not to be described as consanguines. This is the case with the Merinas of Madagascar who, before French colonization, practiced Arab marriage and had a Sudanese/descriptive terminology. However, in this terminology, there were only two terms used to describe siblings: mpiray tam-po (“full” siblings) and mpiray kibo (uterine half-siblings); none for agnatic half-siblings, who were, in short, expelled “out of kinship” (Vogel 1982). The same is true in the terminology of the Beri of Sudan studied by Marie-José Tubiana (1985) which, as with Arab or Fulani nomenclature, is Sudanese in G + 1. In G 0, two systems are present: either FBCh and FZCh as “children of my father” and MBCh and MZCh as “children of my mother” are distinguished, or else the terminology is partly Iroquois but only the matrilateral parallel cousins are assimilated to siblings, because patrilaterals “leave the system” and are designated by a specific term, as are cross cousins who are assimilated with each other and placed in a separate class. In such a system, the author noted, only “parallel matrilateral cousins with whom marriage is impossible remain strongly marked as ‘my mother’s children’” (Ibid: 233).

Sudanese Terminology and Agnatic Kinship

88 It is possible to go further in this first reading. Indeed, this particular division proposed by Sudanese terminology should not, logically, be of interest only to systems based on the principle of uterine kinship, those favouring “Arab marriage”. It should also apply to all those who favour one gender over the other and thus contrast the two categories of parallel kin with each other. In other words, if it is congruent with the categories of kinship based on what I have referred to as a “principle of uterine kinship”, it should be equally congruent with those categories of kinship based on a “principle of agnatic kinship”.

89 Systems based on a principle of agnatic kinship (not filiation), where only individuals in the agnatic line are considered as “true” kin and where individuals in the strictly uterine line are no longer considered as “true” kin (while cognates range between these two extremes), are very rare. In my 2008 essay, I cited two examples of this, and detailed one of them: that of the Han. So is it just a coincidence that Han terminology is also based on a “Sudanese/descriptive” logic, which one would hardly expect to find in this Far Eastern cultural area more often associated with bifurcate merging type terminologies [13]?

90 Indeed the logic of the opposition of the categories of parallel consanguines may be even more obvious here. Thus, in the Chinese village of Hsin Hsing in Taiwan studied by Bernard Gallin (1966), rather than distinguishing all the categories of cousins as in mainland China, MBCh and FZCh are grouped together under the same term ku piao (instead of using chiu piao and ku piao for them), without however adopting a bifurcate merging type terminology, but instead a particular classification where FBCh ≠ MZCh ≠ (FZCh = MBCh).

91 As we saw above with Merina terminology, it will be recalled that the category of agnatic parallel kin tends to be marginalized in societies based on a principle of uterine kinship and practising “Arab marriage”. Conversely, in the Han system, and more generally in those based on the principle of agnatic kinship, it is logically the category of uterine parallel kin who tend to “leave the system”, with members being demoted to the status of “almost foreign”. This is quite understandable in those systems where uterine relations are no longer, as I have said and as Gallin writes, truly “kin”:


“Therefore, in order to categorise the Chinese system of kinship terminology for first cousins, we must consider the following: (1) the initial and most significant bifurcation between patrilateral parallel cousins and the other three types of cousins (the patrilateral parallel cousin is also the only one with whom marriage is, of course, strictly forbidden), and (2) the secondary bifurcation between the patrilateral and matrilateral cross-cousins and the matrilateral parallel cousins (a distinction apparently not found universally throughout China or even throughout Taiwan). This secondary bifurcation, which, terminologically, actually tends to separate matrilateral parallel cousins from both types of cross-cousins as well as from the patrilateral parallel cousins, is also indicated in the fact […]that while the villagers have these kinship terms for matrilateral parallel cousins, in actuality many of them do not apply the terms to their matrilateral parallel cousins. Rather, they think of them merely as ch’in ch’i (relatives), so that Hsin Hsing villagers say that marriage to a matrilateral parallel cousin is the closest one can come to marrying a stranger. Such is clearly not the case for any of the other three types of cousins”.
(1966: Appendix II, 288-289)

93 As we can see, while Sudanese/descriptive terminology is particularly congruent with the principles of unisexual kinship (whether agnatic or uterine), these principles can always accommodate other, rarer modes of terminological classification in G 0. These could be those we have just examined where FCh ≠ MZCh ≠ (MBCh = FZCh), or those where FBCh ≠ (MZCH = MBCh = FZCH), but also Buryat terminology, which distinguishes “paternal” from “maternal” kin and where (FBCh = FZCh) ≠ (MBCh = MZCh), etc.

94 The only imperative that these terminologies must always respect is not so much a specific form as an idea: that the two categories of parallel kin, which are extremes from the perspective of kinship and marriage, cannot and must never be merged under the same term.

Terminology and Principles of Parallel Kinship

95 We have, I hope, begun to identify the logic behind the relationship between terminology and marriage with these first examples of matrimonial systems based on a unisexual kinship principle (whether agnatic or uterine). It will now be easier for us to understand how other expressions of kinship principles relate to various types of nomenclature.

Dravidian Terminologies and Cross-Cousin Marriage

96 Let us consider the classic case of bifurcate merging terminologies (Dravidian, Kariera, etc.). Their relationship to what I have called a “principle of parallel kinship” (Barry 2008) is particularly obvious. According to this principle, the two unisexual lines (those of the parallel cousins and therefore, ad minima, those of the siblings) are the lines of “kin”, and are therefore unmarriageable. Only the genealogical chains comprising individuals of both sexes (the lines of consanguine cross cousins) will, in these systems, “leave” kinship and therefore become marriageable again.

97 So, if we continue to follow the hypothesis that terminologies seek to mark the maximum difference between categories of kinship that are opposite to each other, and do so only by exploiting the state of the two rules of term production that we have examined above, then of course the two sets of parallel cousins and the siblings, on the one hand, and the two categories of cross cousins, on the other, should be opposed terminologically as a priority. It does not matter whether, in order to do so, we need to distinguish between them or to assimilate the respective elements making up these two opposing categories with each other, because, once again, the aim is not to find a direct correspondence with the matrimonial classes, but to distinguish opposites from the perspective of kinship, always within the constraints imposed by the rules of terminology construction.

98 The nomenclatures that best lend themselves to the expression of this principle of parallel kinship are of course, first and foremost, those based on a principle of bifurcate merging and, more specifically, where this principle transcends generations, rather than those where it relates only to generations immediately adjacent to Ego (as is the case, for example, with Iroquois nomenclature). Dravidian (or Kariera) terminologies, which oppose the two sets of parallel cousins (assimilated to siblings) and the two sets of cross cousins (assimilated to affines) regardless of the generation considered, is therefore the most appropriate.

Dravidian Terminology and “Oblique Marriage”

99 Moreover, the terminological equations of the Dravidian models certainly reflect much more closely, at least at some levels, the transmission of unisexual principles in kinship (a double unisexual principle in this case) than most other nomenclatures, and the rules of relational equivalence and gender relevance that organize the terminologies are applied more rigorously here than elsewhere.

100 Sudanese terminology thus reflects the fact that the gender counts in the definition of kinship, but it also tells us that the idea of the “closeness” of kinship is tempered by the idea of collaterality and genealogical distance: equivalent relationships cannot be assimilated and same-sex collateral kin are distinguished. This amounts to saying that a mother’s sister, for example, is not “truly” considered to be a mother, even if, in the case of a principle of uterine kinship, she will indeed be the “closest” relative after the mother. On the contrary, Dravidian terminology, both by taking account of the gender (Rule 2) and by allowing for the assimilation of equivalent relationships (Rule 1), seems to take for granted that two siblings of the same sex are indeed considered to be one and the same being, that a mother’s sister is a mother and a father’s brother is a father. This (quasi-)complete identification of same-sex siblings, which refuses the idea of collaterality, genealogical distance and, of course, any desire for a “biologizing” reading of kinship, allows us to better understand certain forms of marriage commonly associated with Dravidian systems.

101 This will be the case, in particular, with “oblique marriage” with the sister’s daughter [14]. Indeed, if the genealogical space were “uniform” here, the ZD would be perceived (in Western genealogical logic, even from the point of view of our fellow geneticists who study consanguine marriages) as a “closer” relative in direct uterine line than the MZD for example, and much closer than a classificatory MZD (MMZDD, MMMZDD, etc.). She should therefore not be marriageable, since parallel unisexual lines are “kin” lines and matrimonially prohibited. But, to be precise, the genealogical space of Dravidian systems is distorted, as it were, by the interplay of relational equivalences, which apply here both to the terminological field and to the field of kinship calculation. Thus, the terminological logic which holds that MZD = Z [15] also applies to her position in Ego’s kinship group: she really is a sister if we consider her from the perspective of identity and marriageability.

102 On the other hand, a man’s uterine niece, ZD ms, who might appear “genealogically” closer if one were to adopt the perspective of a Western observer, for whom the genealogical space is uniformly extended [16], will ultimately be considered by societies that use these forms of Dravidian classifications as a more “distant” relative than the uterine parallel cousins, to the point of sometimes becoming marriageable.

103 This distancing in kinship of relationships that are close genealogically (which is merely the inversion of the phenomenon that we have just analysed in the case of parallel cousins, where an already distant genealogical relationship is nevertheless considered to be close from the perspective of kinship) can also be explained very simply based on our calculation rules for the production of kinship terms. In this case, in fact, as we have already seen in the examples of the application of our rules, the rule of relational equivalence (Rule 1) cannot be observed, since the gender is considered relevant (Rule 2) and the male Ego and his sister are not of the same sex.

104 In this way, an agnatic niece will be assimilated to a “daughter” for a man [H-HF: (H)F], whereas a uterine niece of the same man cannot, and will remain a “sister’s daughter” [H-FF: ≠ (H)F]. However, with the application of our two rules, a man’s matri- or patrilateral parallel cousins have indeed become “sisters”; in other words, the closest uterine relatives of Ego after his own mother and daughter.

105 Only a man’s uterine niece thus retains a status of “2nd degree” collateral kin (child of a sister), whereas his agnatic niece and parallel cousins will be perceived as direct (“1st degree”) kin. In other words, in Ego’s perception of the specific “curvature” of the Dravidian genealogical space that surrounds him, his uterine niece is a more distant relative than his agnatic niece and patri- or matrilateral parallel cousins; she is almost as distant, in fact, as cross cousins, who are also marriageable.

106 It is therefore understandable why a large number of Dravidian systems prohibit marriage with all parallel cousins (who are, terminologically, “sisters”) and with nieces in the agnatic line (assimilated to “daughters” for a man), but allow it both with cross cousins (who remain, terminologically, “cousins”) and with uterine nieces (who are, terminologically, his only “nieces”). The association, in these particular configurations, between the possibility of marrying cross cousins and “oblique marriage” only with the sister’s daughter, but not with the brother’s daughter is thus in perfect agreement at all levels, not only with the idea of a “principle of parallel kinship”, but also, in the context of the analysis of terminologies that concerns us here, with the rules of term production as just explained [17].

107 However, it is not yet sufficient to state that the expression of a principle of parallel kinship fits perfectly with Dravidian-type terminology. It must also be shown that this reading accounts for more facts than the classical interpretation. The exercise is not as arduous as the overabundant literature on Dravidian systems might suggest. In fact, it is sufficient to state that just as we cannot explain the prohibition on the two categories of parallel kin in a Dravidian marriage model on the basis of the idea of exchange of women between generally unilinear [18] lineages, neither are there any satisfactory interpretations of the fact that Dravidian terminologies assimilate all parallel kin and distinguish them from all cross kin [19].

108 This double assimilation of the patri- and matrilateral parallel kin cannot, in fact, be justified by notions of descent or consanguinity. If Dravidian terminologies were based on these notions, then in these systems’ predominantly unilinear framework (particularly in South India, where unilinear systems are omnipresent) at least one of the categories of parallel kin would not fall de facto within the category of consanguines, nor within the descent group, and should therefore not be assimilated to the category of “siblings”. We should have, in a patrilineal situation, an equation of the type (FBCh = Sibs) ≠ MBCH ≠ FZCh ≠ MZCH and, in a matrilineal situation, something like (MZCh = Sibs) ≠ MBCH ≠ FZCh ≠ FBCH. The Dravidian terminological assimilations therefore cannot in any way be derived simply and logically from a unilinear mode of descent; moreover, this is what Emeneau already remarked in 1937 and, before him, Rivers as early as 1914 [20].

109 Contemporary interpretations of these Dravidian terminologies most often follow neo-Dumontian readings, where terminologically “consanguinized” individuals and equations assimilating parallel (patri- and matrilineal) cousins to siblings are seen as reflecting these individuals’ belonging to Ego’s sociological group. According to this theoretical perspective, this is what justifies the fact that they are all unmarriageable. In doing so, these interpretations closely follow the syllogistic reasoning of those very old anthropological chimeras that we mentioned at the beginning of this text, according to which the delimitations of terminological classes serve for the immediate identification of consanguines or affines. But as we have just seen, in the absence of “moieties” or bilinear filiation—conditions that are, in the vast majority of cases, not realized in this Dravidian framework—the reasons for these assimilations of all “parallel kin” to “members of Ego’s group” are clearly sociologically null and void.

110 Now, if we agree to change the paradigm somewhat by considering that Dravidian terminology, as I have proposed above, follows the terminology construction rules that we have described here, and if we allow these rules to express a “principle of parallel kinship” that contrasts “kin” (and not consanguines, nor the members of the descent group) with “non-kin”, then the terminological assimilation of all parallel (matri- versus patrilateral) cousins to siblings and their differentiation from the class of cross cousins and other affines will no longer raise any problems, whether sociological or logical.

111 In a “principle of parallel kinship”, kinship comes from the transmission of both “uterine kinship” by women and “agnatic kinship” by men, without these gender-related “principles” being confused within a kinship principle that is also transmitted by both sexes, but in an undifferentiated (gender-neutral) manner, as is the case with a “principle of cognatic kinship”, for example. In a “principle of parallel kinship”, all uterine and agnatic parallel cousins are therefore “kin”, regardless of the descent group to which they belong, since they are all directly linked to Ego by a unisexual principle.

112 This is what the terminology reflects very accurately by activating both Rule 1 and Rule 2 (whose state is held to be “true”). This activation allows the terminological assimilation of all kin in unisexual lines (“parallel cousins”) to siblings, and contrasts this category with “cross cousins”, who are non-kin (or “quasi-non-kin”). These two opposing categories—unmarriageable kin and possible spouses—as described in the terminology, thus correspond exactly, in the context of these Dravidian societies, to the matrimonial opposition of “kin” and “non-kin” (and not of consanguines and affines) implied by a system based on what I have called a “principle of parallel kinship”.

Crow-Omaha Terminology and Unilateral Marriage

113 Before leaving the framework of this principle of parallel kinship and related terminologies, let me add a few brief thoughts about a more specific system, that of Crow-Omaha terminologies.

114 Here again, if we follow our idea that the link between terminology and marriage is based on the need to terminologically dissociate positions that are “extreme” from the perspective of kinship, we find that it helps to explain the frequent association between Crow-Omaha nomenclatures and a very specific form of marriage. Indeed, as Claude Lévi-Strauss (1967 [1949]) or Françoise Héritier (1981) have pointed out, although these terminologies can sometimes be associated with global prohibitions concerning entire lineages (parents, grandparents, etc.) according to the classic definition of “semi-complex systems”, we nevertheless find them mostly linked to societies that have adopted forms of marriage in which the two categories of parallel cousins, but also one (and only one) of the categories of cross cousins, are prohibited, and in which one may therefore only marry a unilateral cross cousin, whether matrilateral or patrilateral [21].

115 What do these terminologies ultimately tell us in their own language, if we look at them from the perspective of this logic of opposing extremes that we have applied to the nomenclatures examined so far? They tell us that parallel cousins are different from cross cousins (this is the principle of bifurcate merging), but also that the cross cousins are not similar to each other. This is the effect of the skewing rule, which radically distinguishes the two categories of cross cousins (often assimilating one of them to nephews and nieces or grandchildren, and the other to uncles and aunts or grandparents) [22].

116 The fact that these nomenclatures are then associated with the possibility for a male Ego of marrying one of his cross cousins, but not the other, nor any of his parallel cousins, is thus once again in total harmony with such a distribution of terminological categories.

117 Crow-Omaha nomenclatures, however, do not assimilate one category of cross-cousins to “affines»”—as may be the case in Dravidian terminology to mark the “marriageability” of the latter—and the other with “consanguines” to underline the prohibition of marriage [23]. What counts, as we will now understand, is neither to assimilate those who are not marriageable, nor to group together those who are: the essential goal, here as elsewhere, is to avoid classifying with the same taxon individuals who, from the perspective of kinship (and therefore, secondarily, of marriageability), occupy perfectly irreconcilable positions.

Terminology and Principle of Cognatic Kinship

118 The final possible expression of kinship principles, the “principle of cognatic kinship”, also recognizes the role of both parents in the identity of offspring, as did the principle of parallel kinship, but unlike the latter it does not differentiate these relationships by gender. The essential characteristic of these cognatic systems is thus their desire for symmetry, for uniformity of the genealogical space surrounding Ego, which requires any genealogical relationship to be perfectly equivalent to another of identical form and length, regardless of the gendered nature of the links that constitute them.

119 Distinguishing between kin and non-kin presupposes attention to the form and complexity of the relationship, rather than to its content. It does not matter whether the area covered by kinship is very restricted (for example, it does not exceed, as in pagan Rome or in contemporary Western societies, the level of uncles and aunts, and cousins, who are already no longer really “close kin”, are all equally marriageable), or whether on the contrary it is—officially, at least—considerable (extended to 7th canonical degree kin, as was the case in medieval Christian Europe). What is characteristic of the conception of kinship and the matrimonial prohibitions it generates is that individuals of the same genealogical level are all treated the same from the perspective of identity and marriage: they are all “kin” (and therefore all prohibited) or else they are all “non-kin” (and therefore all matrimonially authorized). As such, the gendered nature of their relationship with Ego does not at any time enter the equation.

120 This essential criterion of the irrelevance of gender is precisely what characterizes two terminologies strongly associated with these forms of cognatic prohibitions: those known as Eskimo and Hawaiian. Although the correlation between these terminologies and specific marriage systems is a little less absolute than that between the Sudanese terminology and a unisexual kinship principle, it is nonetheless very strong.

121 Referring again to Murdock’s (1967) data for which we have information on both forms of marriage and nomenclatures (i.e. 730 societies), we see that of 73 Eskimo terminologies, 55 are associated with prohibitions (or possibilities of marriage) that are cognatic and perfectly symmetrical with respect to Ego (N, O or Q codes) [24], representing 75% of Eskimo terminologies. With regard to Hawaiian terminology, no fewer than 199 of 254 occurrences, i.e. 78% of these nomenclatures, involve these same codes linked to cognatic marriage prohibitions and possibilities [25].

122 It is, I readily admit, less easy to establish on what sociological criterion the last two terminological forms, Eskimo and Hawaiian, are truly distinguished. I will therefore venture the following hypothesis with some caution. One possible interpretation would be to suppose that, in addition to the notion of symmetry, Hawaiian nomenclature adds the idea that “genealogically” close siblings and collaterals are truly “sociologically” similar, whereas Eskimo terminology is based more on the idea of a systematic and constant “distance” from kinship, in successive steps based on a direct genealogical count [26]. This is consistent with the two strictly opposed views: that cousins are “close” relatives like siblings, and that siblings are “distant” relatives in the same way as cousins [27].

123 But whatever may distinguish them, what unites the Hawaiian and Eskimo terminologies is, conversely, particularly obvious. These nomenclatures are by far the most suitable for “complex alliance systems”, insofar as their logic of term formation can be related to the salient features inherent in the logic of constructing a “principle of cognatic kinship”. By distinguishing, as they do, relatives in symmetrical and equivalent concentric circles (either according to generation for Hawaiian terminology, or generation and level of collaterality for Eskimo terminology), they give the genealogical space a smooth, symmetrical and egocentric form in which the criterion of gender is absent. It is precisely this effect of cognatic and egocentric symmetry, which does not take into account the gender of relatives, that characterizes the adoption of a “principle of cognatic kinship”. This principle will thus define concentric circles of matrimonial prohibitions around Ego and only be open to possible alliances according to genealogical distance.

124 Hawaiian or Eskimo terminologies therefore express, within the limits of their own constraints and complying only with possible states of the rules for constructing terminologies that we have seen here, not only the original idea of a principle of cognatic kinship—namely the irrelevance of gender—but also the idea that only genealogical distance ultimately makes marriage possible, once kinship has become sufficiently diffused that it is no longer clearly an obstacle.

125 *

126 It seems to me that anthropologists analysing alliance and kinship systems have far too long neglected Eskimo, Hawaiian and especially Sudanese terminologies, which are all nomenclatures that make no apparent correspondence between matrimonial classes and terminological classes. Admittedly, no useful a priori clue emerged from these linguistic constructs that would have made a direct and obvious reading possible in terms of marriageability. Thus, growing weary, anthropologists fell back on Dravidian terminologies, for which the correspondence between matrimonial classes and linguistic categories seemed much more obvious.

127 We have seen, however, that such a correspondence, when all too evident, is sometimes misleading. And although many ethnologists have engaged in analysing Dravidian terminological classes based on the much-discussed opposition between consanguinity and affinity without batting an eyelid, these terminological classes are in fact purely linguistic, ethereal and without any real sociological content. They do not, in most cases, refer to any empirical reality that can be expressed in terms of descent, nor in terms of those who are classified as “consanguines” as belonging to any “group” of which Ego is a member.

128 Rather than following this superficial lure, and by instead choosing to look at the logic of the rules of production of kinship terms that structure all terminologies rather than at the classes generated by some of them, it seems to me that in this text we have been able to begin reconciling nomenclatures and alliance systems. This reconciliation, though less immediate and obvious to grasp than what is usually proposed, has at least the distinct benefit of linking all the main terminologies to all the major alliance systems and not, as has been the case up to now, a handful of the former to a minority of the latter.

129 What ultimately brings these terminologies closer to the forms of alliances associated with them is not, according to the perspective we have outlined herein, that they directly and explicitly tell us who is or is not marriageable. Indeed it would be naively intentional to believe that linguistic facts provide clues to make it easier for us to grasp them—a bit like with the “signature theory”, when it was believed that nature itself revealed to man some of its functions (Foucault 1966). Rather, what terminologies describe to us—in their language and in their way, rather than ours—is how they use their own logic to bring together the salient features that are particular to the modalities in which kinship is constructed. And it is only in a second stage that the logic proper to governing kinship subsequently defines matrimonial prohibitions and possibilities, that is to say, the alliance system.

130 Therefore, the possibility of approximating nomenclature and alliance refers to a tangible reality, as ethnologists have often assumed, and probably a universal one, as they have sometimes hoped. However this rapprochement is always indirect, for it is above all a translation—and sometimes, to translate is to betray, Traduttore, traditore—in aiming to reflect in terminological designations, those categories and representations that allow us to express the idea of kinship itself.

Appendix I

Positional writing

131 In order to explain the rules that organize the construction of nomenclatures, it is possible to use a so-called “positional” writing convention that I devised some years ago [28] and that has since been used by some of our colleagues (see, in particular: Hamberger & Daillant 2008; Hamberger, Houseman & Grange 2009; Hamberger, Houseman & White 2011: 537-538).

132 The traditional writing system used in works on kinship in anthropology is, despite appearing very “technical” on the surface, only a simple shorthand, a simple paraphrase of the syntax of natural languages. Describing an expression such as “my mother’s brother’s daughter” is, after all, a matter of replacing each term in the expression with its abbreviation in English: M (= Mother), F (= Father), B (= Brother), Z (= Sister), etc. Far from being a “scientific” language that is objective and independent of the culture in which it was developed, it is intrinsically linked to this culture. This complete reducibility to common language makes it both incomplete and imprecise.

133 Conversely, positional writing is completely independent and thereby allows for both a simplified, exhaustive and intuitive expression of the objects it describes, and in particular, reveals some attributes that otherwise remain indiscernible in common language or its abbreviated version in traditional notation. This positional notation thus makes it possible to describe all possible and imaginable genealogical relationships in a simple manner, while traditional notation can only transcribe a small number of them. Finally, as we shall see, its formal nature allows propositions to be calculated logically, which is not possible within the framework of traditional notation.

134 But rather than continuing to extol its merits ab abstracto, let us instead examine the conventions of this positional writing and some concrete examples of notation that will verify its advantages over the traditional notation system.

135 In positional writing, a genealogical relationship (a “string”) is read from left to right, so Ego and Alter are placed at the beginning and end of the string respectively. Three unique symbols are used: X for individual, H for man and F for woman. Three operators are used: dash “–” [29], parentheses “()” and full stop “.”. Symbols are always read ascending when they are to the left of a dash “–” and descending when they are to the right. A dash indicates the position of the apical ancestors when specifying them is unnecessary. If it is necessary to specify them, then parentheses are used instead of the dash, within which the symbol(s) of the ancestors will be indicated: for example, (H) will mean that Ego and Alter have only one male ancestor in common, (F) that they have only one common ancestor, and (HF) that they are from a common pair of ancestors [30]. The use of parentheses thus makes it possible to specify all the cases of genealogical relations involving the notion of half-siblinghood (which is absent from the classical notation). Finally, a full stop indicates a marriage between individuals on either side of it and at the same time recommences an ascending reading to its right.


A few examples
(X) designates an Ego, (H) a male Ego and (F) a female Ego.
(X)H refers to an individual’s son, (X)F a daughter and (X)X a child.
X(H) refers instead to an individual’s father and X(F) to the mother [31].
XH(H) will therefore refer to an individual’s paternal grandfather, XF(H) to the maternal grandfather, XH(F) to the paternal grandmother, XF(F) to the maternal grandmother, and so on.

137 As far as collaterals are concerned, H-H will thus designate the brother of a male Ego since we start from a man, we “go up” to his immediate ancestors (the parentheses are replaced here by a dash if the identity of these common ancestors does not concern us), in order to subsequently “go back down” to a man who is the descendant of these ancestors. H-F will thus be a man’s sister according to the same logic.

138 X-H will refer to a brother in general (the brother of an “individual” of unspecified sex) and X-F will refer to a sister.

139 Instead of the previous proposals, we could of course have written H(HF)H, H(HF)F, X(HF)H and X(HF)F if we wished to specify that they were full siblings (from the same father and mother).

140 If, on the contrary, we wanted to speak of a man’s agnatic half-brother, then we would have written H(H)H, and F(F)F for a woman’s uterine half-sister.

141 Let us note immediately that these last two examples, although extremely simple and obvious to write in positional notation, already surpass the capacity of traditional notation to transcribe relationships, unless periphrases are introduced: respectively “B (same father only) ms” and “Z (same mother only) ws” [32].

142 For slightly more distant relationships, F-HH will designate a woman’s agnatic nephew and F-FH her uterine nephew, while F-FX will designate a woman’s uterine nephews and nieces, F-XX a woman’s agnatic and uterine nephews and nieces, and X-XX agnatic and uterine nephews in general (of a man or a woman).

143 At a slightly more complex level, HH-HF will describe a man’s patrilateral female cousin, FH(F)HH the son of the uterine half-brother of a woman’s father (a matrilateral cross “half uterine male cousin”, a relationship which, again, cannot be translated in the traditional notation system, except by using an obscure periphrasis: “M’s half B (same father) S ws”).

144 For affines, HH-HF.HH-HF will designate for example the female patrilateral parallel cousin of the husband of my female patrilateral parallel cousin (FBDHFBD ms in ordinary notation), and F.H.F [33] will designate a co-wife of a woman (HW ws in traditional notation).

Appendix II

An Example of the Application of Gender Rules and a Nomenclature with “terminologic Slices”

145 Pathan terminology is one of those unusual and rare systematics which, though still based on the same two rules (1 and 2), add as an additional condition that their state varies according to the sex. For example, they allow relational equivalence for men but not for women, or vice versa, which again leads to a distinction between the two sets of parallel cousins. These terminologies lead to assimilations for 4th degree kin (cousins) that are not possible for 3rd or 2nd degree kin.

146 Let us look at this Pathan example to understand the logic behind it and verify that it does not deviate from the rules we have examined so far. This is a terminology in G + 1 where FB ≠ MB ≠ F but where M ≠ (MZCh = FZCH). As can be seen, this terminology therefore applies an Eskimo logic (Rule 1 inactive, Rule 2 active) only to female referents, with the sisters of a father or mother becoming “sisters of a parent” [XF-F and XH-F = XX-F] without being assimilated to the mother since Rule 1 is inactive [X(F) ≠ XX-F], while applying a Sudanese logic (Rules 1 and 2 inactive) to male terms. We therefore have two “uncles” and a single category of “aunts”:

  • FB [XH-H]
  • MB [XF-H]
  • MZ = FZ = PZ [XX-F].

147 It is this logic, which continues unchanged at the level of the children of siblings of parents, which gives rise to the Kurdish nomenclature mentioned above where FBCh ≠ MBCh ≠ (MZCh = FZCh).

148 But to complicate things further, the Pathan case involves a nomenclature comprising “terminological sections”, thereby yielding different states for our two rules depending on the generation under consideration. By adding a generation (that of the cousins), it starts from the previous results and activates Rule 2, but only on this new intermediate link (that of the uncles and the aunts), which will thus give us (starting from the three previous results):

  • FBCh = FsibCh [XH-HX = XH-XX]
  • MBCh = MsibCh [XF-XX = XF-XX]
  • PZCh = PsibCh [XX-FX = XX-XX].

149 And this point, Rule 1 of relational equivalence will be activated only for the female links at this genealogical level. This will therefore give us (starting from the previous results):

  • FsibCh [XH-HX]
  • MsibCh = PSibCh and MsibCh = MCh so (MsibCh = PSibCh) = MCh due to the transitivity of the relationships.

150 We will therefore have “children of father’s brothers” on one hand, and a category of “siblings” on the other, which includes all the other categories in G 0. Here, therefore, where the equation FB ≠ (MB = F) is not possible starting from the application of our rules in G + 1, we can however obtain, even in extreme and very rare cases, the equation FBCh ≠ (MBCh = Sib) in G 0 without deviating in any way from the logic of our rules, simply by accepting the idea that they may be treated differently according to gender (that of Alter and/or that of the intermediate links that make up the relationship between Ego and Alter).


  • [1]
    In the notation system, the “=” sign will be used here as a formal symbol for the identicalness of two relationships, the “≠” sign for non-identicalness. The sign “~” will be used to designate equivalence and the sign “:” will symbolize the assimilation of two relations, the one to the left of the symbol being assimilated to the one to the right of the symbol.
  • [2]
    The treatment of Chinese terminology (which is Sudanese/descriptive in nature) by Claude Lévi-Strauss (1967 [1949]), Francis Hsu (1945) and T. S. Chen and J. K. Shryock (1932), for example, offers a textbook case for the use of such an indexing procedure on a case-by-case basis: the few terminological equations that find a correspondence with alliance practices are cited as “obvious” arguments in support of such a logic of equivalence of classes, but at the same time all these authors are silent on a whole series of other terminological equations that are not echoed in the logic of alliances that they anticipate, or even contradict.
  • [3]
    This article considers only the construction of “classical” terminologies (Iroquois, Hawaiian, Eskimo, Sudanese, etc.). I will devote another text to the construction of more complex but, ultimately, equally frequent types of terminology. I am thinking in particular of “mixed terminologies”, commonly referred to in French as “tranches terminologiques” (“terminological slices)” (whose importance is generally underestimated, even though these combinatorial systems alone account for about half of all terminologies), but also of the variants that exist in many of the basic types identified here (Buryat, Iroquois, Kariera, Dravidian, Yafar, Kuma, etc., to name but a few of the variants derived from a single type based on bifurcate merging).
  • [4]
    The Dravidian, Kariera, Yafar, Kuma, Crow-Omaha, etc. formulas at this generational level of G + 1 all fall within the assimilation or class distinction operations proposed by one of these types, even if they differ from each other in some more distant positions. The rules we shall examine are, of course, applicable at all these levels and for all terminological positions. However, for these more distant levels, the application of the rules discussed here may sometimes be much more complex (I give an example of this with regard to Pathan terminology in Appendix II, below). For obvious reasons of space, therefore, in this paper I will only consider, as a first step, the terminological taxa of the levels that are generally held to be discriminating, i.e. the classification of generations into G+1 and/or G 0, which served as a basis for Robert Lowie and later George P. Murdock in establishing the major classical types.
  • [5]
    The criterion of generation is not to be set as a “rule” (and therefore liable to adopt different states) as it is always held to be “true”: all terminologies distinguish at least between direct ascendants and descendants (although Crow-Omaha nomenclatures may subsequently assimilate some collaterals of different generations).
  • [6]
    I will systematically provide here [in square brackets] a transcription of the simple descriptions of the terminological positions given in ordinary common language (or by using common English abbreviations, such as MBD, FZD, etc.) in a new notation system, called “positional notation”. A short vade mecum describing the rules and uses of positional notation is given in Appendix I of this article. This notation system is now used by a number of our colleagues in the field of kinship studies, however I choose herein to keep the common language descriptions to avoid confusing some readers who might be put off by the idea of having to learn a notation system in order to read this text. However, while the common language descriptions are sufficient for the overall understanding of the hypotheses set out in this text, they are nonetheless much more contrived and ultimately counter-intuitive than those in positional notation. Therefore, those willing make the little extra effort of looking at the transcriptions that accompany the common language version will soon realize, I think, not only the simplicity but also the immediate and obvious qualitative advantages of this more logical, concise and formal notation for immediately understanding the mechanisms guiding the terminological logics described herein.
  • [7]
    Note that these examples of unconditional assimilation of identical relations in the case of siblings, however trivial, account for and justify the existence of the universal rule laid down by Alfred R. Radcliffe-Brown, known as the “unity of the [same-sex] sibling group”, a rule which describes the results of this basic equation, but without providing the keys to it. Of course, the scenario in which the group of same-sex siblings bears a single term illustrates only the most general and common case. But we regularly find specific forms resulting from the effects of adding secondary criteria: age, for example. Thus, by adding this simple variable we obtain two different descriptions: elder male descendant of an ascendant of Ego vs. younger male descendant of an ascendant of Ego (X-H- and X-H+ in positional notation). The rule of assimilation of identical relationships still remains valid, even in cases where these secondary criteria (age, marital status, etc.) must be taken into account, but the relationships thus amended and supplemented are then simply no longer identical with each other.
  • [8]
    I do not mention the latter in this first text because, as all ethnologists who have worked on the subject of nomenclatures know, when affine terminologies exist (and sometimes there are simply no terms for affines, or only a very limited number of them), they are often organized with a different logic from that for terms concerning consanguines.
  • [9]
    I borrow this expression by analogy with the phenomenon of “correlations of growth” in the field of biology that were studied by Charles Darwin (1859; see also Stephen Jay Gould 1982 [1980]). In the latter context, the expression designates those elements which, starting from a given state of an organism, will be modified simply because others are modified, without the modifications of the former being relevant in terms of evolutionary adaptation.
  • [10]
    On these assimilations, see Appendix II.
  • [11]
    A fairly comprehensive English-Hausa dictionary can be found at
  • [12]
    If we leave Murdock’s data aside temporarily and turn to other examples of societies practising “Arab marriage” (cf. Barry 2008), we will again see that the Sudanese form in G + 1 and descriptive in G 0 is indeed the one almost always associated with this system (including in the case of societies, such as the Merinas or the Tswana, which do not, however, fall within the Sudanese or Arab-Persian linguistic area).
  • [13]
    Even in its earliest form, Han terminology already presented an equation of the type FBCh ≠ (MBCh = FZCh = MZCh), i.e. it distinguished the two sets of parallel cousins from each other.
  • [14]
    On the relationship of “oblique marriage” with cross-cousin marriage, see also Laurent Barry & Jean-Pierre Goulard (1998).
  • [15]
    A “female descendant of the ascendants of a female ascendant of Ego” is assimilated to a “female ascendant of Ego” [XF-F: X(F)] because of Rule 1, and thus her daughter becomes herself a “daughter of a mother” and thus a sister [XF-FF: X(F)F].
  • [16]
    In our Eskimo terminology, the rule of relational equivalence is held to be “false” and collaterals are therefore distinguished from direct ascendants and thought to be more “distant”, a classification that reflects the Western conception of kinship in which genealogical distance is equivalent to identity distance.
  • [17]
    One will of course notice that Dravidian terminology, because of its particular terminological equations, will add to the prohibitions on “genealogical” parallel kin (i.e. relationships composed exclusively of unisexual links) other prohibitions for genealogically “cross” but terminologically parallel individuals (for example, FMFZSDCh are considered in a Dravidian system as “parallel kin” and are matrimonially prohibited).
  • [18]
    To the extent that a single ban on the unilinear lineage of Ego is sufficient to proscribe the women of his group and thus forces him to marry those of other groups; cf. Laurent Barry (2008), for a more thorough critique of the aporias of the “theory of exchange”.
  • [19]
    Except by multiplying, as Kenneth David (1973) does for example, the ad hoc rules. In particular, by postulating the implementation in Dravidian terminologies of an imperative of linguistic symmetry of terminological categories—an imperative that is purely logical (or even purely aesthetic)—that would in turn (we do not really know why or how) have very real sociological repercussions, simply by doubling the prohibited matrimonial classes.
  • [20]
    Yet it is precisely on this implicit postulate that the post-Dumontian analyses of these Dravidian nomenclatures are based, insofar as they continue to see in them an opposition between “affinity” and “consanguinity” and, following Louis Dumont, to assume that “alliance” and “allies” are inherited from generation to generation among a group of “consanguines” who, in order to remain permanent and always have the same “allies” available, can therefore only be unilinear (or bilinear). The imperative of unilinearity is evident in Louis Dumont’s very formulation of the marriage alliance: “the rule defines the marriage of an individual in relation to the marriage of one of his ascendants. For example, in a patrilineal and patrilocal society, marrying one’s matrilateral cousin means reproducing one’s father’s marriage; it is one’s grandfather’s marriage that is reproduced in the patrilateral formula, etc. In general, the rule determines a cycle of repetition of a certain kind of marriage […]. In other words, the result of the rule is that the marriage is passed on from one generation to the next in much the same way as membership in a kinship group is passed on. Thanks to the rule, marriage acquires a diachronic dimension; it becomes an institution that transcends generations and which I shall call a ‘marriage alliance’, or simply an ‘alliance’” (1975: 48, quoted in Zimmermann 1993: 88).
  • [21]
    On the prevalence of the association of Crow-Omaha terminology with these forms of unilateral cross-cousin marriages, see in particular the comparative data reported by Alf Hornborg (1988: 244) and by George P. Murdock (1949). See also Alfred R. Radcliffe-Brown (1953: 22 ff.) and Jean-Claude Muller (2000).
  • [22]
    This, in matrimonial language, would result in a “principle of parallel kinship”, under which one of the categories of cross cousins would be considered too “close” for marriage to be possible.
  • [23]
    In the “strong” forms per Floyd Lounsbury’s (1964) classification, Crow-Omaha nomenclatures will, in fact, merge one of these categories with direct ascendants and the other with direct descendants (for example, in some Omaha variants, MBCh will be assimilated to individuals of the maternal grandparents’ generation while FZCh will be “daughter’s children”).
  • [24]
    With the exception of Qa for “Arab marriage” or Qc for the possibility of marrying all cousins, but with a marked preference for cross cousins.
  • [25]
    These same codes are associated in only 170 cases with any of the 403 societies that use other types of terminology, representing only 42% of the total.
  • [26]
    This hypothesis may be partly supported by the fact that only 4% of the 254 cases of Hawaiian terminology are associated with a matrimonial system that allows marriage with all cousins (Code Q), while 16% of the 73 examples of Eskimo terminology allow this form of marriage.
  • [27]
    This would correspond, for example, to certain possibilities sometimes associated with this nomenclature, such as the possibility of marrying all one’s cousins, but also one’s own siblings, in Hawaii itself.
  • [28]
    The basic principles are recalled here; a fuller description can be found in Laurent Barry, “History and Technical Specificities of the Genos Programme”, Llaccan site, 2004, “Collection and Processing of Field Data” School [].
  • [29]
    Some authors who have used positional writing have used the empty brackets “()” instead of the dash; there is of course no disadvantage to this alternative writing if it is well specified before use.
  • [30]
    For more particular cases, we can write (X) for Ego and Alter who are from a single common ancestor whose sex is unknown, (XX) when it is not known whether they descend from a single common ancestor or a couple, (HX) when it is known that they have at least one male ancestor in common but it is not known whether they also have a common female ancestor, and (FX) in the opposite case. It should also be pointed out that all these cases, which are described precisely with this positional notation system, cannot be transcribed in the traditional notation system.
  • [31]
    By convention, Ego is put in parentheses because he is the individual in the apical position of his genealogical chain (even though it would comprise only him in this case). As can be seen, parentheses are also used for Ego and Alter when they are in the position of apical ancestors in a genealogical chain, in order to indicate the reading direction (ascending or descending). In general, parentheses are always used when a description is ambiguous.
  • [32]
    Relationships that are somewhat more complex, but still very easy to describe in positional writing—for example: H(H).F.HF(F).H which designates a second spouse (other than the maternal grandfather) of the maternal grandmother of a co-wife (other than the mother) of the father of a male Ego—become almost impossible to describe unambiguously in traditional writing. I invite the reader to try it with this last example, limiting himself/herself to the use of classical abbreviations.
  • [33]
    The notation F.H.F is unambiguous and therefore it is not necessary to use parentheses—(F).(H).(F)—to indicate the apical ancestors and reading direction.

In this article, the author investigates the modalities of constructing the main terminologies of kinship and the relationship that these logics maintain with those governing alliance systems. The proposed analysis will highlight the central role played by two main factors in the development of nomenclatures, namely, the consideration or non-consideration of gender and the relational equivalence of certain terminological strings rendered possible under certain conditions. These two criteria appear necessary and sufficient to provide the basic logical framework for the main terminologies identified by anthropologists. Based on these results, it appears that there is not a mere identity of the logics at work in the fields of nomenclatures and alliance systems, but rather a real congruence of the distinct principles that organize them.

  • kinship
  • kinship terminologies
  • kinship systems
  • alliance
  • filiation
  • matrimonial exchange
  • sexual prohibition
  • positional writing

Logiques terminologiques

Les taxinomies de parenté et leur relation aux systèmes d’alliance

Dans cet article, l’auteur s’interroge sur les modalités de construction des principales terminologies de parenté et sur le rapport que ces logiques entretiennent avec celles régissant les systèmes d’alliance. L’analyse proposée met en exergue le rôle crucial joué par deux principaux facteurs dans l’élaboration des nomenclatures, à savoir celui de la prise en considération ou non du genre et celui de l’équivalence relationnelle de certaines chaînes terminologiques rendues possibles sous certaines conditions. Ces deux critères apparaissent nécessaires et suffisants pour fournir l’armature logique de base aux principales terminologies recensées par les anthropologues. En partant de ces résultats, l’on voit émerger non pas une identité des logiques à l’œuvre dans les domaines des nomenclatures et des systèmes d’alliance, mais bien une réelle congruence des principes distincts qui les organisent.

  • parenté
  • terminologies de parenté
  • systèmes de parenté
  • alliance
  • filiation
  • échange matrimonial
  • interdits sexuels
  • écriture positionnelle

References cited

  • Barry, Laurent, 2008 La Parenté. Paris, Gallimard (“Folio. Essais” 498).
  • 2012
    “La parenté au singulier”, in Emmanuel Désveaux & Michel de Fornel, eds, Faire des sciences sociales, 3. Généraliser. Paris, Éd. de l’Ehess (“Cas de figure” 23): 121-149.
  • Barry, Laurent & Jean-Pierre Goulard, 1998 “Un mode de composition de l’alliance : le “mariage oblique” ticuna”, Journal de la Société des américanistes 84 (1): 219-236 [].
  • Barth, Fredrik, 1959 “Segmentary Opposition and the Theory of Games: A Study of Pathan Organization”, The Journal of the Royal Anthropological Institute of Great Britain and Ireland 89 (1): 5-21.
  • Carsten, Janet, ed., 2000 Cultures of Relatedness. New Approaches in the Study of Kinship. Cambridge-New York, Cambridge University Press.
  • Chen, T. S. & J. K. Shryock, 1932 “Chinese Relationship Terms”, American Anthropologist (new ser.) 34 (4): 623-669.
  • Darwin, Charles, 1859 On the Origin of Species by Means of Natural Selection, or the Preservation of Favoured Races in the Struggle for Life. London, John Murray.
  • David, Kenneth, 1973 “Until Marriage Do Us Part: A Cultural Account of Jaffna Tamil Categories for Kinsman”, Man (new ser.) 8 (4): 521-535.
  • Dumont, Louis, 1975 Dravidien et Kariera. L’alliance de mariage en Inde du Sud et en Australie. La Haye-Paris, Mouton / Paris, Éd de l’Ehess (“Textes de sciences sociales” 14).
  • Emeneau, Murray B., 1937 “Toda Marriage Regulations and Taboos”, American Anthropologist (new ser.) 39 (1): 103-112.
  • Foucault, Michel, 1966 Les Mots et les choses. Une archéologie des sciences humaines. Paris, Gallimard (“Bibliothèque des sciences humaines”).
  • Gallin, Bernard, 1966 Hsin Hsing, Taiwan. A Chinese Village in Change. Berkeley, University of California Press.
  • Geertz, Hildred & Clifford Geertz, 1975 Kinship in Bali. Chicago-London, University of Chicago Press.
  • Gould, Stephen Jay, 1982 [1980] Le Pouce du panda. Les grandes énigmes de l’évolution. Translated from English by Jacques Chabert. Paris, Grasset.
  • Hamberger, Klaus & Isabelle Daillant, 2008 “L’analyse de réseaux de parenté : concepts et outils”, Annales de démographie historique 116 (2): 13-52 [].
  • Hamberger, Klaus, Michael Houseman & Cyril Grange, 2009 “La parenté radiographiée : un nouveau logiciel pour l’analyse des réseaux matrimoniaux”, L’Homme 191: 107-138.
  • Hamberger, Klaus, Michael Houseman & Douglas R. White, 2011 “Kinship Network Analysis”, in Peter J. Carrington & John Scott, eds, The Sage Handbook of Social Network Analysis. London-Thousand Oaks, Sage: 533-549.
  • Héritier, Françoise, 1981 ’Exercice de la parenté. Paris, Gallimard-Le Seuil (“Hautes Études”).
  • Hornborg, Alf, 1988 Dualism and Hierarchy in Lowland South America. Trajectories of Indigenous Social Organization. Uppsala, Academiae Upsaliensis / Stockholm, Almqvist & Wiksell International.
  • Hsu, Francis L. K., 1945 “Observations on Cross-Cousin Marriage in China”, American Anthropologist (new ser.) 47 (1): 83-103.
  • Lévi-Strauss, Claude, 1967 [1949] Les Structures élémentaires de la parenté. Paris-La Haye, Mouton.
  • Lounsbury, Floyd, G., 1964 “A Formal Analysis of Crow- and Omaha-Type Kinship Terminologies”, in Ward H. Goodenough, ed., Explorations in Cultural Anthropology. Essays in Honor of George Peter Murdock. New York, McGraw-Hill: 351-394.
  • Morgan, Lewis Henry, 1870 Systems of Consanguinity and Affinity of the Human Family. Washington, Smithsonian Institution.
  • Muller, Jean-Claude, 2000 “Des “chiffres et des lettres” : discours locaux et ordinateurs”, L’Homme 154-155: Question de parenté: 489-504.
  • Murdock, George Peter, 1949 Social Structure. New York, Macmillan.
  • Murdock, George Peter, 1957 “World Ethnographic Sample”, American Anthropologist 59 (4): 664-687.
  • Murdock, George Peter, 1967 “Ethnographic Atlas: A Summary”, Ethnology 6 (2): 109-236.
  • Murdock, George Peter, 1970 “Kin Term Patterns and their Distribution”, Ethnology 9 (2): 165-208.
  • Poincaré, Henri, 1946 [1905] La Valeur de la science. Geneva, Constant Bourquin (“Classiques français du XXe siècle” 5).
  • Radcliffe-Brown, Alfred Reginald, 1953 [1950] “Introduction”, in A. R. Radcliffe-Brown & Daryll Forde, eds, Systèmes familiaux et matrimoniaux en Afrique. Revised translation by Marcel Griaule. Paris, Presses universitaires de France (“Bibliothèque de sociologie contemporaine”): 1-107.
  • Rivers, William Halse Rivers, 1914 Kinship and Social Organisation. Cambridge-London, Constable & Co.
  • Sahlins, Marshall, 2012 What Kinship Is—And Is Not. Chicago, University of Chicago Press.
  • Shapiro, Warren, 2012 “Extensionism and the Nature of Kinship”, Journal of The Royal Anthropological Institute 18 (1): 191-193.
  • Strathern, Marilyn, 1992 After Nature. English Kinship in the Late Twentieth Century. Cambridge-New York, Cambridge University Press.
  • Tubiana, Marie-José, 1985 Des troupeaux et des femmes. Mariage et transferts de biens chez les Beri (Zaghawa et Bideyat) du Tchad et du Soudan. Paris, L’Harmattan (“Bibliothèque Peiresc” 4).
  • Vogel, Claude, 1982 Les Quatre-mères d’Ambohibaho. Étude d’une population régionale d’Imerina (Madagascar). Paris, Selaf (“Langues et civilisations de l’Asie du Sud-Est et du monde insulindien. Langues, cultures et sociétés de l’océan Indien” 13).
  • Zimmermann, Francis, 1993 Enquête sur la parenté. Paris, Presses universitaires de France (“Ethnologies”).
  • Zonabend, Françoise & Chantal Collard, 2013 “Parenté sans sexualité : le paradigme occidental en question”, L’Homme 206: 29-58.
Laurent Barry
School of Advanced Studies in the Social Sciences (EHESS) Social Anthropology Laboratory (Las)—Université Psl, Paris
Translated by
Shaun Murdock
This is the latest publication of the author on cairn.
Uploaded on on 20/05/2021
Distribution électronique pour Editions de l’E.H.E.S.S. © Editions de l’E.H.E.S.S.. Tous droits réservés pour tous pays. Il est interdit, sauf accord préalable et écrit de l’éditeur, de reproduire (notamment par photocopie) partiellement ou totalement le présent article, de le stocker dans une banque de données ou de le communiquer au public sous quelque forme et de quelque manière que ce soit.
Loading... Please wait