Chapter 3: The Principle of Exogamy and its Applications


Only the custom of marriage between cross-cousins, which in its application means that the same person is my father’s father and the brother of my mother’s mother, enables us to understand that Latin avunculus, derived from avus ‘paternal grandfather’ signifies ‘maternal uncle’.
Correlatively, nepos ‘nephew’ (indulged by his maternal uncle, but subjected to the strict patria potestas), beside this original sense (confirmed inter alia by Gr. anepsíos ‘cousin’, literally co-nephew), takes on the meaning “grandson” wherever the Indo-European patrilinear system was imposed with increasing rigor.
In contrast to nepos, the designation of “son”—generally as “offspring”— presents a considerable diversity in the Indo-European languages: we can glimpse in this traces of a structure of kinship where the relationship between father and son was eclipsed by that of maternal uncle to nephew.


A common term designates “grandfather” in most Indo-European languages: it is represented by Latin avus and the corresponding forms. But in certain languages the sense offers a noteworthy variant: it is no longer “grandfather” but “uncle,” and in particular “maternal uncle.”
We shall now enumerate these forms proceeding in the order of increasing complexity.
To Latin avus corresponds the Hittite term of the same sense, huhhaš. The relationship seems surprising considering how different the forms are. It finds its explanation in an archaic stage of Indo-European phonology. Hittite preserves an ancient laryngeal phoneme (written h), which has disappeared in other languages, but which is there indirectly manifested by the modifications of timbre or vocalic quantity. We shall use the notation *H. The common prototype can be reconstructed as *HeuHos.
Like Latin avus and Hittite huhhaš, Armenian haw ‘grandfather’ presents the word without a suffix. The initial h of Armenian has nothing to do with that of Hittite; it is a secondary aspiration due to a recent phenomenon: etymologically, the Armenian form pre-supposes an ancient initial vowel. The same recent aspirate has developed in the parallel Armenian word for “grandmother,” han, which is compared with Hittite hannaš ‘grandmother’, Latin anus ‘old woman’, Greek annís, glossed “mother of the mother or of the father,” Old High German ana ‘grandmother,’ etc.
As against Hittite huhhaš, Lat. avus, Armenian haw ‘grandfather’, the forms in the other languages fall into special groups. We have first the group of Slavic and Baltic: Old Slav. ujǐ, originally *auios; in Baltic the Old Prussian awis, Lithuanian avýnas. As for the sense, we observe that the Balto-Slavic *auios signifies “uncle.” The Lithuanian avýnas, a secondary derivation, designates especially the mother’s brother, the maternal uncle.
The Celtic forms represent two distinct developments. On the one hand there is Old Irish aue, Middle Irish ōa, which also come from *auios but designate the “grandson.” On the other hand, Welsh ewythr, Breton eontr, presuppose a derivative *awen-tro- and signify “uncle.”
In Germanic, we have a series of derivatives with a suffix in -n forming a new root *awen-: in Gothic this *awen- is by chance represented only by the feminine awo ‘grandmother’ (dative sing. awon); the masculine form is attested in the Icelandic afe ‘grandfather’. This stem *awen- is presumably represented in Old High German in the word ōheim, German Oheim ‘uncle’, which is reconstructed hypothetically as a compound *awun-haimaz. We do not know how to interpret the second element: it may be a derivative of the name of the residence (Heim, cf. home) “he who has the residence of the grandfather” (?), or as a nominal form from the root *k w ei-(Gr. timḗ, τιμή) “he who has the esteem (?) of the grandfather”; but this root does not appear elsewhere in Germanic. Everything in this reconstruction remains uncertain and this is detrimental to the analysis of the sense. In any case, Old High German ōheim and the corresponding forms of Old Engl. ēam, Old Fris. ēm, likewise signify “uncle” and not “grandfather.”
Such are the facts arranged according to their forms. It will, however, be noticed that not all languages figure here: Greek and Indo-Iranian are missing. These two dialect groups have new terms. In Greek, the “grandfather” is called páppos (πάππος), a form of address belonging to the language of children; it is not found in Homer, but it is the only one known in prose both of literature and inscriptions. In Sanskrit, the “grandfather” is called pitāmaha-, a descriptive compound in which the two elements are in an unusual order. It has been explained as an imitation of the compound with an intensive reduplication mahāmaha ‘very great, all powerful’; this reveals the recent date of this designation. Moreover, Indic does not here agree with Iranian, which has a distinct word, found both in Avestan and Old Persian, nyāka ‘grandfather’, Persian niyā, a term with no etymological connections.
We can now see the great problem posed by the evolution of sense between Indo-European *awos and its derivatives and compounds. The fact that these derivatives are formed with the help of suffixes in -yo, -en, explains nothing. What we have to find out is how, starting from the word for “grandfather,” the same word came to be used for the “maternal uncle.” The question does not arise only in the different dialect groups, but within Latin itself, since, along with avus ‘grandfather’, we have the diminutive avunculus as the term for ‘uncle’. The problem has been recognized since ancient times and it has often been discussed. It is already found in Festus : “avunculus, matris meae frater (brother of my mother and not of my father) traxit appellationem ab eo quodtertius a me, ut avus… est” (because he occupies the third degree in relation to me, like the grandfather)— or, another explanation, “quod avi locum obtineat et proximitate tueatur sororis filiam” (because he takes the place of the grandfather and is responsible for the supervision of his sister’s daughter). It never designates anything else than the maternal uncle.
An idea presents itself immediately: if avunculus is attached to avus, is it not because avus designated the maternal grandfather? Avunculus could thus be explained as the son of the real avus. This was supposed by Delbrück, and Eduard Hermann has insisted on this explanation. [1] This idea is not acceptable either in fact or in theory. Let us take the examples of avus collected in the Thesaurus; none has the sense of “maternal grandfather.” All the definitions of the ancients connect avus with paternal lineage. In the Origines of Isidore of Seville we read: “avus pater patris est; patris mei pater avus meus est” (“Avus is the father’s father; my father’s father is my avus”). If the ancestors are enumerated, a beginning is always made with the father, pater, and then avus, proavus, etc. are listed. For the maternal grandfather, the specific expression avus maternus is used. Similarly in Hittite, huhhaš is exclusively the paternal grandfather; we have an additional proof in the plural huhhanteš which designates the fathers, i.e. the ancestors, the forebears; it is in the paternal line that the ancestors are to be found.
This is a question of fact; let us now consider the theoretical reason. In a system of classificatory kinship, no special importance is attributed to the mother’s father. In agnatic filiation, account is taken of the father and the father’s father; on the other hand, in uterine filiation, the mother’s brother is considered. But the mother’s father has no special position. It follows that one could never have designated as avunculus such an important person as the maternal uncle with a term derived from avus, if avus indicated the mother’s father, a relationship which is of no particular importance.
The difficulty which philology cannot solve unaided finds its solution in the structure of exogamic kinship. We have to envisage the situation of EGO with reference to his avus and his avunculus. We can represent the situation figuratively by a schema indicating the relationships after the lapse of two generations. We have to remember that following the principle of exogamy, the two different sexes always belong to opposed moieties: therefore marriage must always take place between members of opposed moieties.
Smith I is the avus, the father of EGO’s father. At the same time, Smith I is the brother of the mother of EGO’s mother: avus designates therefore, in one and the same person, the father’s father and the brother of the mother’s mother, that is the maternal great-uncle. The double relationship to EGO of this single person follows automatically from the marriage of cross-cousins. Starting with Jones II, the same scheme begins anew: the son of Jones I marries the daughter of his father’s sister, his cross-cousin; the avus is always the paternal grandfather and maternal great-uncle. To sum up: Smith I is the father’s father (or avus) of Smith III, who is EGO. But Smith I is at the same time the brother of the mother of Jones II, who in his turn is the brother of the mother of Smith III (EGO). For EGO, Smith I will be avus and Jones II avunculus.
Starting with EGO, his mother’s brother, his avunculus, is the son of the sister of his father’s father, of his avus. This is always the case. In this system, a relationship is established between maternal uncle and nephew, while in agnatic filiation, it is established between father and son.
Accordingly, if avus refers in reality to the maternal great-uncle, the maternal uncle could be called “little avus” or avunculus. This solution is a simple consequence of the necessities of the system. This suggests that we should ascribe the sense of “maternal great-uncle” to avus rather than “grandfather”: one and the same person, the brother of the mother’s mother, is at the same time the father’s father. In his authoritative work on ancient Chinese society, Granet [2] draws attention to the same correspondence: the agnatic grandfather is always the maternal great-uncle. This rule applies also in other societies: it has the typical character of a necessary rule.
Latin, thus re-interpreted, offers some important evidence: but in historical times the sole meaning attested is the agnatic signification of avus as “grandfather,” “father’s father.” The etymological relationship with avunculus implies and reveals another type of filiation, given that avunculus is the mother’s brother.
This general structure conditions the diverse elements which it comprises. The way is open to a structural conception of Indo-European kinship and of the vocabulary of this kinship, because it contains classes and relationships between classes. This makes intelligible the variety of terms and the dissymmetry of the designations for uncles and aunts in Latin: patruus for the “father’s brother” but avunculus for the “maternal uncle”; in the feminine matertera, the mother’s sister, the “quasi-mother”, but amita for the father’s sister. The relationship of fraternity between members of the same sex puts them in the same class. As the father’s brother or the mother’s sister are of the same sex as the personage in virtue of whom they are defined, the terms which designate them are derivatives from the primary term. But the mother’s brother, or father’s sister, being of the opposite sex, have different names: this is an illustration of the principle of exogamy (see figures below).

Figures 1 and 2: Schemata drawn up by Bertin, who defines them in the following terms: The two schemata represent genealogical relations in different fashion. In both cases the information is the same, both for the individuals and their relationship.


Figure 1. In this schema (a traditional genealogical tree) the individuals are represented by points (of different shape according to sex, and black or white according to family) and the relationship by lines (of different design, according to the kind of relationship: filiation or alliance).


Figure 2. In this schema, less orthodox, certainly requires some effort of adaptation: the individuals are represented here by lines (different according to sex and family), their relationships by a point (representing by itself alliance and filiation). But the figure thus obtained brings out better the special relationships of cross-cousins here studied. This second system of representation has the added advantage of facilitating the recording of genealogical information that is infinitely more complex and ramified, and presenting it in easily read form (which the first type of representation does not permit).

In general, in our modern languages, this distinction has been lost. However, it is not necessary to go back very far to discover various pointers to the privileged position which the maternal uncle occupied.
For the ancient Germanic world, we refer to Tacitus, Germania XX, 4:
Sororum filiis idem apud auvnculum qui ad patrem honor; quidam sanctiorem artioremque hunc nexum sanguinis arbitrantur et in accipiendis obsidibus magis exigunt tanquam et animum firmius et domum latius teneant.
The sons of the sister are just as dear to their avunculus as to their father; there are even some who believe that this blood tie (that of the avunculus) is more sacred and close (than that of paternity). They insist on it by preference when taking hostages, because thus they think they have a better hold on their minds and a wider hold on the family.
With the Celts, too, we find concordant testimony. The great heroes of the epic call themselves after their mothers. The relationship between Cuchulainn and his mother’s brother Conchobar is a good illustration of this type of relationship. In Homer, this structure remains recognizable, although the designation of the maternal uncle has been remodeled as mḗtrōs (μήτρως), a secondary derivative made on the model of pátrōs (πάτρως), which is the equivalent of Latin patruus. The ancient noun has disappeared, but the old idea has survived. In the Iliad, the only two examples of the term mḗtrōs are particularly significant:
1) Apollo appears in disguise to Hector to encourage him in his moment of weakness; he takes on the appearance of his maternal uncle (mḗtrōs) in order to have more authority (Il. 16, 717).
2) Tlepolemus, the bastard son of Heracles, has killed the maternal uncle of Heracles; he has to flee, followed by the “sons and grandsons” of Heracles; by this murder he has provoked the hostilities of the whole of his kindred (Il. 2, 661 ff.).
It would certainly be possible to find other examples of this kind which often pass unnoticed. Here we cite only one from Herodotus (IV, 80). At the moment when Octamasadas, the king of the Scythians, is getting ready to fight Sitalkes, the king of the Thracians, the latter makes him say: “Why should we fight since you are the son of my sister?”
Much the same is testified by a fact of the Armenian vocabulary: k c eri ‘maternal uncle’ is a derivative of k c oyr ‘sister’. This morphological relationship appears clearly if we substitute the respective prototypes: k c oyr goes back to *swesōr and k c eri to *swesriyos. The maternal uncle is therefore literally designated as “he of the sister,” after his sister, who is the mother of EGO. This is an explicit expression, probably a substitute for another more ancient one, which underlines the specific nature of the maternal uncle in the system of Armenian kinship. All this brings out, in a way that is all the more convincing because the facts come from languages and societies which have long become separate entities in the Indo-European world, the special position of the “maternal uncle,” and it makes the formal relationship between avus and avunculus more probable.
Correlatively, the word for “nephew,” a term represented in nearly all the languages, shows a parallel variation of sense: it means both “grandson” and “nephew.”
First we list the forms in their etymological relationships: Skt. napāt, naptr̥, fem. naptī; Av. napāt, fem. napti; Old Persian napā (nominative); Lat. nepōs, feminine neptis; Old Lithuanian nepuotis, feminine nepte; Old Engl. nefa; Old High German nefo; Old Slavic netĭjĭ < *neptios; in Celtic, Old Irl. nia, Welsh nei. We must also cite Gr. anepsiós (ἀνεψιός), but separately: it does not signify “nephew,” but “cousin.”
According to the language, *nepōt- is sometimes “grandson,” sometimes “nephew” and sometimes both.
In Vedic, napāt is the “grandson” or, more vaguely, the “descendant”; it is “grandson” in Iranian, too, especially in Old Persian, where it is clearly defined in the genealogy of the Achaemenid kings. The modern Iranian forms like Persian nave always refer to “grandson”; for “nephew” Persian employs descriptive compounds, “brother’s son” and “sister’s son.”
In contrast to Indo-Iranian, the languages of the West, except for Latin, have *nepōt in the sense of “nephew.” If in Latin, nepos seems to apply at will to “nephew,” to “grandson” or to “descendant,” in Germanic, Slavic and Celtic, the corresponding term denotes the nephew, in fact always the son of the sister. This special expression for the descendant by reference to the mother’s brother emerges even in Latin in certain uses of nepos.
A study by Joseph Loth [3] of the sense of nepos in the Latin inscriptions in Brittany has shown that it always refers to the sister’s son; nepos therefore has the same sense as in the corresponding Celtic word nia in Irish and nei in Welsh, which designate the sister’s son, while the brother’s son in Irish is called mac brathar, a descriptive term. Apart from this, there are in Celtic legends traces of a uterine kinship; in the Ogham inscriptions, filiation is established through the mother. In Latin authors, too, we can collect important testimony. Thus in Livy (V, 34) the Gaulish king Ambigatus, wanting to rid his kingdom of surplus population, asked the two sons of his sister (sororis filios) to lead a portion of the tribe to new territories. This is not only a feature of the Celtic societies. According to a Lacedaemonian tradition, reported by Herodotus (IV, 147), the royal power in Sparta had been assumed by Theras, the maternal uncle of the heirs who were still too young to reign and whose guardian he was.
What are we to make of the classical use of nepos? Certain etymologists, confronted with the double sense of nepos, “nephew” and “grandson,” which are distinguished in other languages, have thought that it was a vague term with no well-defined meaning.
It is nothing of the kind. What we find in all languages is that when we retrace the history of their words we meet precise meanings which later usage may have extended. This is particularly the case with terms for kinship, where words must have an exact sense, because they determine each other mutually. Insofar as it designates “nephew,” nepos often has an emotional overtone: the nephew is a spoilt child, dissipated, spendthrift. This connotation implies a certain type of relationship between the nephew and his mother’s brother. In effect, ethnographers have observed that in societies where the relationship between the maternal uncle and nephew prevails, it has a sentimental value, inverse to that which unites father and son. Where relations between father and son are strict and rigorous, the other is indulgent, full of tenderness. Inversely, where the father is indulgent to his son, the relationship between nephew and uncle is more rigid; he educates the child, inculcates rules of conduct and initiates him into religious rites. The two relationships of kinship are in correlation: they are never established on the same sentimental footing.
Now we know that in Latin the relationship between father and son was characterized by its severity: the father was invested with the right over life and death over his son, and he sometimes exercised this right. In ancient Roman society, the patria potestas was not subject to appeal. It had to be tempered by another relationship, precisely that between uncle and nephew, in the type of filiation which this supposes.
As for the duality of the sense “nephew” and “grandson,” the explanation of this is given by the homologous relationship between the name of “uncle” and that of “grandfather.” Just as avus, in the paternal line “brother of the mother’s mother,” produces the diminutive avunculus for the “brother of the mother,” similarly and correlatively, the name of the grandson may designate at the same time the nephew of the mother’s brother. The two changes are symmetrical; the son of the sister’s daughter receives the same name as the son of the sister. However, the increasingly rigorous patrilineal tendency of Indo-European kinship often secured the predominance of the agnatic signification: “son’s son.”
The related Greek term anepsiόs (from *a-nept-iyo-) signifies “cousin” in the sense in which we understand the term. The form itself furnishes important testimony: the literal sense is “those who are co-nephews,” which supposes as the point of departure for the element -nept- not the sense of “grandson” but that of “nephew.” Thus the “nephews” of brothers and sisters called each other by this term, which is an indirect proof of the priority of the sense “nephew.” However, the sense of “grandson” was not completely abolished in proto-historic times, to judge by the gloss of Hesychius which must come from literary sources: νεόπτραι· υἱῶν θυγατέρες ‘neόptrai: daughters of the sons’. This feminine could be restored as *νεπότραι (*nepótrai), feminine of *νεποτήρ (*nepotér), which would have designated the son of the son.
In its historic nomenclature, Greek has a new term for “grandson” which is huiōnós (ὑιωνός), derived from huiόs ‘son’, and, correlatively, for “nephew” a descriptive term adelphidoûs (ἀδελφιδοῦς) ‘descendant of the brother’. It may seem natural that the term for “grandson” should be related to that of “son” by way of derivation, as in Gr. huiōnós, or by a composition, as in Engl. grandson, Fr. petit-fils. For this reason, the cases where the “grandson” is called “little grandfather” will seem more curious and noteworthy. Such is the Irish aue ‘grandson’, which goes back to *auyos, a derivative of *auos ‘grandfather’. Similarly, OHG enencheli (German Enkel) ‘grandson’ is etymologically a diminutive of ano ‘grandfather’. Old Slav. vŭnukŭ, Russ. vnuk ‘grandson’ has been connected with it, and this is close to Lith. anukas, unless the Lithuanian word is itself a loanword from Slavic. Closer to us, in Old French, the grandson was called avelet, a diminutive of ave, ève ‘grandfather’. It is the term which has been replaced by the analytical expression petit-fils. Thus, at least in three languages, the “grandson” is called “little grandfather.” There must be a reason why such an expression has been created independently in several different societies, In fact, it is an instance of a shift for which there are parallels. Numerous systems of kinship contain reciprocal terms employed between the two members of what may be called a pair: the mother’s father and the daughter’s son address each other by the same term. In this peculiarity of the vocabulary there is once again a classificatory reason. In many societies we find the belief that a newly born child is always the reincarnation of an ancestor, going back a certain number of generations. They even believe that, strictly speaking, there is no birth, because the ancestor has not disappeared, he has only been hidden away. In general, the process of reappearance is from grandfather to grandson. When a son is born to somebody, it is the grandfather of the child who “reappears,” and this is why they have the same name. The young child is, as it were, a diminutive representation of the ancestor which it incarnates: it is a “little grandfather,” who is born again after an interval of a generation.
With the word for “son” we encounter an unexpected problem. For such a close relationship Indo-European languages present a large variety of designations. The most common one is *sūnu-, attested in Skt. sunu-, Avestan hunu-; Got. sunus; Lithuanian sunus; Slavic synŭ; and, with a different suffix, Gr. huiús (ὑιύς) ; Tokh. A soyä, Tokh. Β . Hittite is isolated with its uwa (nominative uwas). Also isolated is Luvian, which has titaimi, Lycian tideimi ‘son’ (really “nursling”). The Latin fīlius has no immediate correspondent in this sense, and Celtic mace (< *makkos) is again different. The Armenian term for “son” ustr has been adapted to the word for “daughter” dustr, which corresponds to Gr. thugátēr (θυγάτηρ). The form *sūnu- seems to be derived from *su- ‘give birth’; the word thus designates the son as being the “offspring.”
The discordance between the terms for “son” have been highlighted in an article by Meillet [4] who, if he did not solve the problem, has at least made it manifest.
Starting with Latin filius, we can try to understand the nature of the process. Filius is linked in Latin itself to an etymological family represented by felo, fecundus, etc., which imply the notion of “sucking” (Umbrian feliuf, acc. plural, ‘lactentes, sucklings’). The real significance of the word is clear: to explain how it entered into the nomenclature of kinship, we shall have to consider filius as an adjective which has taken on the function of a noun. Here we have the same phenomenon as appears in consobrinus, patruelis, where the adjective, at first joined to a substantive, finally supplanted it: patruelis, consobrinus represent frater patruelis, frater consobrinus. It may be conjectured that filius has evolved from a group which we may hypothetically posit as *sunus filius; the true term was eliminated from the analytical expression, the more expressive term alone survived. How is this to be explained? We observe that this instability of the term for “son” contrasts with the constancy of the word for “nephew.” The fate of filius must be correlated with that of nepos: the important descendant, in a certain type of kinship, is the nephew rather than the son, because it is always from uncle to nephew that inheritance or power is transmitted. The descendant is for his father simply his offspring, which is expressed by the term *sunus. We know, further, that the brothers of the father are regarded as fathers; the sons of brothers are brothers to one another and not cousins, cf. frater consobrinus distinguished from frater germanus. Consequently, the sons of two brothers are in their eyes equally “sons”; hence a man will also call the offspring of his brother “sons.” But how can the proper son be distinguished from the son of the brother? The introduction of filius ‘nursling’ fills this gap. Then, when the relation of maternal uncle to nephew ceased to be important, and when the “Grossfamilie” broke up, it was filius alone which came to designate specifically the descendant of EGO.
Through the vicissitudes of *nepot- and *sunu- we discern the difficulties which societies experienced, when passing from one system to the other, in normalizing the system of agnatic kinship, which had become established, and the only one recognized in law, and in adapting or replacing the inherited terms of previous structures. Their meaning wavers between archaic relationships and the more modern ways of regarding kinship, and it is not always easy to puzzle out the manner in which these nomenclatures have been organized or transformed in each language. [5]


[ back ] 1. Göttinger Nachrichten, 1918, pp. 214f.
[ back ] 2. Civilisation chinoise, 1929, p. 247.
[ back ] 3. Comptes rendus de l’Acad. des Inscr, 1932. 269ff.
[ back ] 4. Mémoires de la Société de linguistique de Paris, 21, 1920, p. 45.
[ back ] 5. We have not touched here on two particularly complex problems: the degrees of ancestry (“grandfather,” “great-grandfather” etc.), and the relations of cousinhood (Lat. sobrinus, consobrinus). We have treated this in detail in an article in L’Homme, vol. V, 1965, pp. 5-10.