Use the following persistent identifier: http://nrs.harvard.edu/urn-3:hul.ebook:CHS_Benveniste.Indo-European_Language_and_Society.1973.
Chapter 6. Formation and Suffixation of the Terms for Kinship
From the morphological point of view, the great unity of the Indo-European vocabulary of kinship emerges from the existence of the class suffix *-ter (or *-er), which not only characterizes a great number of the most ancient terms (*pǝter, etc.), but still continues to figure in the most recent creations or remodeled expressions.
Even when they differ from one language to another, the terms which designate social units—clan, phratry, tribe—are often formed from roots expressing a community of birth, Greek génos, phrátra, phulḗ, Latin gens, tribus.
Less specific than *-ter, and also less studied, is the suffix *-w(o)-/-wyo-which seems to have indicated homostathmic (= ‘at the same level’) proximity: *pəter father”— Greek patrō(u)s, Sanskrit pitrvya– “father’s brother.” The anomalies presented, for instance, by Gr. patruíos “stepfather,” Skt. bhrā́tr̥vya– “brother’s son, later “cousin” > “enemy” may lead us not to question the ancient values of the suffix, but to interpret the deviation which it has undergone by reference in each case to the particular system in which these forms occur.
After this review of the terms which permit us to reconstitute the general organization of kinship, it may be useful to examine a number of questions concerning the form of these terms together with their function. There are, in fact, special features of morphology peculiar to this group which give great unity to it. Particularly notable are certain suffixes characteristic of the kinship words, whether it is because they are found only here or because they assume a special function.
Among the suffixes we cite in the first place –ter or –er, which is the suffix of kinship par excellence. Not only does it serve to constitute some of the most ancient terms of this series, but it kept its proper value after the parent language split into dialects and it remained productive. The initial state of this class suffix is furnished by the common ending of four fundamental designations which cannot be further analyzed: *pəter, *māter, *dhugh(ə)ter, *bhrāter; and, further, in kinship by alliance: *yen(ə)-ter ‘wife of the husband’s brother’.
These are primary words, which are unanalyzable, where this ending is constant and from which it has been extracted with its proper value. Later it was extended to new designations in at least some of the languages: *nepōt– ‘nephew’, or “son-in-law,” has a secondary formation *nepter, which was introduced into the declension of napāt– in Indo-Iranian; e.g. the Sanskrit accusative naptāram and the stem of the oblique cases in Avestan, nafəδr-, which goes back to *naptr-.
The “son-in-law” is in Skt. jāmātar-, in Avestan zāmātar-. The corresponding form in the other languages also exhibits final –r, although the stem has suffered various alterations: Lat. gener, Gr. gambrós. Whatever the particular history of these forms may be, they all come from the same root, extended by a suffix in –er or –ter, and we can see that the –r– is secondary, from the fact that the Avestan terminology, alongside zāmātar– ‘son-in-law’, also has zāmaoya (= *zāmavya), probably “brother of the son-in-law,” which is today continued by Pashto zūm ‘son-in-law’.
There are terms connected with Latin avus, avunculus which in Celtic designate the maternal uncle: Welsh ewythr, Breton eontr, go back to *awontro-; we recognize here, in a thematic form, the same suffix –ter.
Let us recall finally *daiwer ‘husband’s brother’, Lat. lēvir, etc., everywhere with –er.
We see that the formation in –ter or in –er is from its origin attached to many terms of kinship. It remained productive of new terms in this lexical class in the subsequent history of the languages. One of the clearest examples of this extension is observed in Middle and Modern Persian, where this suffix, eliminated by the loss of endings, has been secondarily restored. After the ancient series pitar– ‘father’, mātar– ‘mother’, brātar– ‘brother’, duxtar– ‘daughter’ developed according to phonetic laws into pit, māt, brāt, duxt, the characteristic ending –or was restored, resulting in the present-day Persian forms: pidar, mādar, brādar, duxtar, and, by analogy, pusar ‘son’ (for pus). This process of morphological repair began in Middle Persian. Few suffixes have preserved such great vitality.
There is another proof of the antiquity of this formation; it is given by one of the most ancient terms characterized by the suffix, the word for “daughter,” and in a language the Indo-European character of which is now assured as a member of the Luvio-Hittite group. This is Lycian, where the word for daughter is cbatru (accus. sing.). The phonetic detail of the reconstruction is not completely certain. However, we may suppose that the initial Lycian group cb– goes back to an early *dw-; we have a parallel development in the word or compositional element signifying “two”: Lycian cbi < *dwi. We can thus reconstruct a proto- Lycian *dwatr, which corresponds to Gr. thugátēr, with modification of the dorsal plosive between vowels: *duga– > *duwa-. In any case, we can identify here the same final –er or –ter as in the other languages.
Those terms of kinship which have the suffix –ter are further characterized by the nature and importance of certain of their derivatives.
Above, the question of the phratry has already been discussed and particularly the connection which this term shows between “consanguineous brother” and “classificatory brother.” The phratry is a grouping which is inserted in its proper place in the series of Greek terms which mark social divisions. We have three groups, in order of increasing size, génos (γένος), phrátra (φράτρα), phulḗ (φυλή), these being three concentric divisions of ancient Greek society.
Roman society similarly exhibited three divisions, but they are not exactly the same: first the gens, then the cūria, and lastly the tribus. In this triple organization the terms of the first rank are comparable, the others diverge; but the actual organization is much the same. These are the units which we express by the series clan, phratry, tribe.
In fact, Gr. génos and Latin gens correspond without completely coinciding. There is a difference in suffixal formation: the morphologically Lat. genus = Gr. génos, but gens is a feminine in –ti. Thus between Greek and Latin the formal connection is established as *genes-/*genti-. By its formation Latin genu– corresponds to Skt. jāti– ‘birth’. The abstract noun in –ti– denotes the “birth” and, at the same time, the class of persons united by the tie of their “birth”; this fact serves us a sufficient definition of a certain social group. To the same lexical family belongs the Avestan term zantu-, which differs from it only by the suffix –tu and likewise designates as “birth” an important segment of Iranian society. If we disregard these variations of suffixes, the principal ancient languages agree in making membership of a “birth” the foundation of a social group. 
As for the second division, the Latin term cūria, equivalent to Gr. phratría, is quite different: cūria has no correspondent either in Greek or elsewhere. It is, however, possible to explain the form cūria in Italic itself as *co–viria “collection of viri” on the evidence of Volscian covehriu, which has the same sense. It is at the same time both a place of reunion and an important division of the Roman people. In contrast to phratría in Greek, the expression cūria does not bring out any tie of kinship between the members of this unit. By this it reveals its more recent origin, which is also confirmed by its limitation to Italic.
It is more difficult still to establish a connection between Greek phulḗ and Latin tribus. The problem is the etymological formation of tribus. It is to be presumed that the two terms underwent analogous processes of development. The ancients already saw in tribus a unit consisting of three groups. It would thus be a compound having tri– as its first term. In fact, in the historical Indo-European traditions, especially among the Greeks, we know of such triple groups. We have the testimony for three Dorian tribes in a Homeric epithet: Δωριέες τριχαί(w)ικες ‘the Dorians (divided) into three wik-’ (cf. Gr. (w)oîkos (w)οῖκος). In the Greek territory which was occupied by Dorians in ancient times, a district of Elis is called Triphulía (Τριφυλία), clearly attesting the division into “three tribes” of the first inhabitants. We would have here the rough correspondent of the Latin tribus, if it signifies “a third (of the territory).” It is in fact not impossible that tribus, like Umbrian trifu, its only correspondent, contains a nominal form *bhu-, which coincides exactly with Gr. phu– (in phulḗ). However, we do not find any historical testimony which would support this primary meaning of the term. At an early date, tribus gave rise to important derivatives, such as tribunus, then tribunal, and the verb tribuo, but they give no evidence for a connection with “three.”
Among the types of formation peculiar to words of kinship other than –ter or –er, we must mention a number of secondary formations in *-w– and *-wyo-; they merit special attention, all the more because they have been less well studied. This type is represented in Latin by patruus ‘father’s brother’, ‘paternal uncle’, cf. Gr. pátrōs (πάτρως) ‘father’s brother’, from *patrōw-, and the symmetrical feminine mḗtrōs (μήτρως) ‘mother’s brother’. We must compare with patruus the words of the same sense, Sanskrit pitr̥vya-, and Avestan tūirya < *(p)tr̥wya-; cf. Persian afdar and Pashto trǝ ‘father’s brother’, and further in Old High German fatureo (German Vetter) < *faðurwyo, and probably Old Slavic stryj ‘uncle’.
This type of derivation exists in Greek with a rather different sense: patruiós (πατρυιός) signifies “stepfather,” mētruiá (μητρυιά) ‘stepmother’; also in Armenian yawray ‘stepfather’ and mawru < *mātruvyā ‘stepmother’.
On the basis of the word for brother and by the same morphological device, Skt. bhrā́tr̥vya-, Av. brātūirya– was constituted. But the sense of these terms has provoked much discussion. The examples are few and not decisive. Is the meaning “the brother’s son” or “the son of the father’s brother”? Is he the nephew or the cousin? For the sense of Skt. bhrā́tr̥vya– we have a formal indication in Pāṇini, who gives this brief definition: bhrātur vyac ca, that is to say from bhrātr̥ ‘brother’, the derivative indicating descent is also formed by –vya-. Thus, apart from the normal derivative in –iya– for “descending from,” there is another formation in –vya– with the same sense: the upshot is that bhrātrvya– signifies “brother’s son,” and not “the son of the father’s brother,” the translation given generally by scholars. There is no doubt that Av. brātūirya– (variant brātruya-, i.e. brātr̥vya-, fem. brātruyā-) should also be interpreted as “brother’s son,” since for “son of the father’s brother,” there exists a clear analytical designation, tūirya–puθ ra ‘son of the tūirya’, that is, of the paternal uncle. A confirmation is also given in modern Iranian by the Pashto language of Afghanistan, where wrārə (from *brāθr(v)ya-) means “nephew,” that is “the brother’s son.”
Up till now the facts do not seem open to dispute. But for Skt. bhrā́tr̥vya-, apart from the sense of “nephew,” we have also that of “rival, enemy,” which is well attested. This fact has made certain etymologists hesitate, following Wackernagel, to admit that “brother’s son” was the initial sense of bhrā́tr̥vya-, in spite of the Iranian correspondent forms. In their view, bhrā́tr̥vya– would rather have signified “cousin” (= son of the father’s brother), because it is difficult to imagine the “nephew” acting as “rival,” whereas among cousins rivalry is easier to understand. In Arab society, “cousin” takes on the sense of “rival,” “enemy.” But the truth is that this notion appears to be alien to the Indo-European world; between the anepsioí of the Homeric society, the relationship of cousinhood, far from occasioning rivalries, is an amicable relation. Wackernagel thinks, however, that in the case of bhrā́tr̥vya– a prehistorical change took place from “cousin” to “nephew,” a transition which would find a parallel in Spanish sobrino, etymologically “cousin,” which today has become the word for “nephew.”
All this seems to us disputable, both for the reconstruction of the ancient state and for the chronology of the senses. If we keep to the given facts, we have to admit that Indo-Iranian bhrātr̥vya– designates the “brother’s son” and nothing else. As for the sense of “rival, enemy,” we observe that it is limited to Sanskrit. Iranian for its part explains the connection between the two notions. We find in Pashto (Afghanistan) the kinship term tǝrbur ‘cousin’, to be analyzed as tǝr ‘paternal uncle’ and *pūr ‘son’, going back to *ptǝrvya–putra– ‘son of the father’s brother’. Now this word does not only designate the “cousin” but also the “rival,” “the enemy.” Hence the sense of “enemy” is attached to an analytical expression “son of the paternal uncle,” while the “nephew” is called wrārǝ (< *brāθr(v)ya-) a term which does not imply rivalry, any more than the old Avestan brāturya– does. This is clear confirmation of the testimony of Pāṇini on the sense of Sanskrit bhrā́tr̥vya– as “brother’s son, nephew,” not “cousin.” The initial relationship between pitr̥vyà– and bhrā́tr̥vya– in Sanskrit was as follows: pitr̥vyà– signified “father’s brother” and bhrā́tr̥vya– ‘brother’s son’. This is also the situation in Iranian of the correspondent terms. The forms and their senses must therefore be attributed to Indo-Iranian. It is from this finding that we must start to reconstruct, as far as possible, the connection of these terms in the Indo-European period. This formation is certainly of Indo-European date; in fact, outside Indo-Iranian there are ancient representatives, as we have seen, in Greek, Latin and Germanic. We are here confronted with a lexical category which may be presumed to be homogeneous, but with local aberrations.
To give an explanation of it we must introduce here two theoretical considerations, one bearing on the terminology of kinship, the other on the morphology of the terms.
We believe it is necessary, in particular, in defining the changes which have come about in the course of history in the application of the words to the degree of relationship, to distinguish the relationship between members of the same generation, which we will call homostathmic (= at the same level), and the relationship between members of different generations, which we will call heterostathmic (= of different levels).  The relationship of fraternity is homostathmic, the relationship of ancestrality is heterostathmic.
In the formation of the terms for kinship themselves we must pay attention to the nature of the suffix when this seems to have, as in the present case, a distinctive value. The Indo-European morpheme *-wo, *-wyo-, which forms the secondary derivatives in question, should indicate some kind of connection with the basic term. We can give precision to the nature of this connection by considering the function of this suffix in a class of primary nominal derivatives; these are the adjectives indicating spatial position, like Ved. pū́rva-, Ir. parva– ‘previous, first’; Gr. deksi–wós, Got. taihswa ‘right(side)’; Gr. lai(w)ós, Lat. laevus, Old Slav. levŭ ‘left(side)’; Ved. viśva– ‘all’, sarva– ‘entire, intact’, Lat. salvus ‘id.’; Ved. r̥ṣvá– ‘elevated, high’, Av. ǝrǝšva– ‘id.’, etc. By analogy, we conjecture that the derivative in –w– from a term of kinship will have indicated a situation of proximity to the person indicated by the basic term, a particularly close relationship which in some way is homogeneous with the basic term.
This class of derivatives for kinship in *-w– is represented in Indic by pitr̥vya– and bhrā́tr̥vya-. But if they occupy almost the same lexical position in Indic, these two terms differ greatly in their Indo-European distribution: the first is widely attested over an extensive area, the second is limited to Indo-Iranian. There are reasons for thinking that the first term is the original one and that the other has been adapted to it by secondary assimilation, and only in a part of the territory.
Other indications confirm this relative chronology. The forms which in western Indo-European correspond to Skt. pitr̥vya– show, so to speak, the foundation of the function and even of the form of the suffix. This is seen especially in ancient Greek, where several derivatives are thus made with –w-. There is, first, pátrōs (attested first in Herodotus and Pindar) “father’s brother” and mḗtros (Homer, Herodotus, Pindar) “mother’s brother,” both derived by means of *-ōu– from patḗr and mḗtēr. This formation thus indicates in general the nearest relations of the same generation (hence excluding filiation). We have here a homostathmic relation to the basic term. Consequently, “father’s (or mother’s) brother” is the degree of kinship to which this suffixal function is appropriate, and it is sometimes, particularly in the plural, found extended to the whole group of the nearest relations of the father or the mother. This suffix, in the thematic form *-wo-, is that which recurs in a similar function in the Latin patruus ‘father’s brother’. Latin has no correspondent of Greek mḗtrōs ‘mother’s brother’, any more than any other language has. For this relationship Latin says avunculus, Sanskrit mātula-. The variety of these designations shows that they are of different date. Whereas Lat. avunculus is connected with avus by an ancient relationship which is repeated in other languages (see above, Book Two, Chapter Three), the Greek and Indic expressions are secondary: Greek mḗtrōs is evidently coined after pátrōs and Skt. mātula– (for *mātura-) is a purely Indic formation. They are recent substitutes for an Indo-European designation which disappeared when the mother’s brother ceased to have a privileged position with respect to the father.
Another reason may also have contributed to its elimination. We get some idea of it in a very complex process of competition between two suffixal formations of ancient Greek which considerably modify our ideas of Indo-European. Apart from pátrōs ‘father’s brother’, which corresponds exactly to the sense, but not exactly to the form of Skt. pitr̥vyà-, Greek has the term patruiós, which corresponds to the form οf pitr̥vyà-, but does not have the same meaning: patruiós designates the “stepfather.” Now, while pitr̥vyà– ‘father’s brother’ has no Sanskrit homologue in the feminine (*mātr̥vyā does not exist and doubtless could not have existed), Greek patruiós ‘stepfather’ was accompanied by the feminine mētruiá ‘stepmother, second wife of the father’. In fact, in the lexical history of Greek, the primary term is mētruiá (attested from Homer on and in all the dialects), which is strongly characterized by its affective connotation and its metaphorical usage (the stepmother, bad mother) as compared with the patruiós, which is both late and rare and purely descriptive, and obviously an analogical formation based on mētruiá. We must conclude that the formal concordance between Skt. pitr̥vyà– and Gr. patruiós is deceptive: it is due to a simple convergence of forms created independently and at different dates. The only terms to be taken into consideration are in Indic the masculine pitr̥vyà– ‘father’s brother’, and in Greek the feminine mētruiá ‘stepmother’. The formation in *-w(i)yo– has been utilized in comparable, but not identical, ways in Indic and Greek: in Indic pitr̥vyà– denotes the nearest relation to the father, in fact his brother; in Greek, where pátrōs has the same sense, use was made of the suffix to form from mḗtēr a derivative mētruiá, which designates the “mother by substitution,” the “stepmother.”
Owing to the lack of ancient data, we know less about the fate of this formation in *-w(o)- or *-wyo– in other languages. It is highly probable that Old Slavic stryjǐ ‘father’s brother’ (a Panslavic term, with the exception of Russian) continues, with a phonetic treatment that is somewhat obscure in detail, the same original as Skt. pitr̥vyà-. This type is represented in Germanic by Old High German fetiro ‘father’s brother’, which is distinguished from ōheim ‘mother’s brother’, just as Lat. patruus is from avunculus. In the history of High German, fetiro has passed from meaning “father’s brother” to “son of the father’s brother,” hence modern German Vetter ‘cousin’. But this is an exceptional evolution. Everywhere else, this term, or its feminine equivalent, has kept its homostathmic value.
Let us consider now the second term of kinship exhibiting the same suffix, that is Skt. bhrā́tr̥vya-, Av. brātūirya– (cf. above). It is, as we have seen, limited to Indo-Iranian. This alone is a reason for thinking that it is less ancient than pitr̥vyà-. Besides this we may note that the two expressions are not homologous: bhrā́tr̥vya– ‘brother’s son’ indicates a heterostathmic relationship, different from pitr̥vyà– ‘father’s brother’, which is homostathmic. Thus there is morphological conformity but disparity of sense; these two features must be kept in mind; they must be explained together. We shall find the reason for it in the general structure of this terminology.
If Indo-Iranian bhrātr̥vya– is not applied to the same level of kinship as pitr̥vya-, this is because the position of the basic term required it. Given the value of the suffix, if the derivative pitr̥vyà– from pitr ̥– ‘father’ was applied to “the father’s brother,” then bhrā́tr̥vya-, with the same formation, strictly speaking ought to designate only “the brother’s brother,” which is nonsense, at least in Indo-European, where all brothers have the same relationship to each other. It was, therefore, applied to another degree of proximity, “son of the brother” which, by a shift of one generation, answers a double purpose: in the first place it served to differentiate the “brother’s son” from the “sister’s son,” who had a different designation (*nepōt-, Indo-Iranian napāt-); in the second place it specified the notion more clearly than another derivative, bhrātrīya-, which also meant “brother’s son” according to Pāṇini and which, being duplicated, was eliminated. But when napāt– was applied indiscriminately to the son of the brother or the son of the sister, Skt. bhrā́tr̥vya-, now becoming available, was reinterpreted either as “son of the father’s brother,” or as “quasi-brother,” which practically comes to the same thing and designated the “cousin.” The connection with EGO again became homostathmic; then, in the social conditions which seem to have been peculiar only to India, the kinship of cousins was associated with the behavior of rivals. This was the origin of the double meaning of bhrā́tr̥vya– in classical Sanskrit as “cousin” and “rival.”
This whole evolution took place only on Indian territory. No trace is to be found in Iranian, where brātr̥vya– (Av. brātūirya-, etc.) seems never to have deviated from the initial sense of “brother’s son.” But this conflict between the terms for “nephew” and “cousin” was renewed in the modern phase of the Romance languages, in Ibero-Romance, where the representatives of Latin nepos, sobrinus, consobrinus ended up by forming a new system. 
Thus, in every case, it is not one single term which must be considered in isolation, but rather the whole group of relationships: it is by this that the history of each of the terms is conditioned. Apart from the general structure of Indo-European kinship, we must recognize in each language at a given period a particular structure which must be interpreted in its own right. It is by starting from bhrā́tr̥vya– with the sense of “brother’s son,” given in the Indian tradition, that we are able to reconstruct the conditions for the passage to the sense “cousin” and later “rival,” which was effected in classical Sanskrit. More than with any other lexical group the terms of kinship oblige us to keep to, and to combine, the two aspects of one and the same methodological requirement, the structural study of the terminology as a whole and the consideration of the levels in each language and each society.
[ back ] 1. The precise sense of the terms génos, gens, zantu– will be studied below in Book Three, Chapter Two.
[ back ] 2. These terms have been proposed and employed in an article in Ľ Homme, V, 1965, p. 15.
[ back ] 3. See the article already cited (Book Two, Chapter Three, n. 5) in L’Homme.