Use the following persistent identifier: http://nrs.harvard.edu/urn-3:hul.ebook:CHS_Benveniste.Indo-European_Language_and_Society.1973.
Chapter 1. The Importance of the Concept of Paternity
Father and mother, brother and sister do not constitute symmetrical couples in Indo-European. Unlike *māter ‘mother’, *pəter does not denote the physical parent, as is evidenced, for instance, by the ancient juxtaposition preserved in Latin Iupiter. Nor is *bhrāter ‘brother’ a term of consanguinity: Greek, in phrá̄tēr, preserves better than any other language the sense of “a member of a phratry,” a classificatory term of kinship. As for *swesor (Lat. soror), this word designates literally a feminine being (*sor) of the group (*swe)—another classificatory term of kinship, but not symmetrical with *bhrāter.
Of all the terms of kinship the most securely established is the name for father: *pəter, Skt. pitar-, Arm. hayr, Gr. patḗr, Lat. pater, Old Irl. athir, Gothic fadar, Tokharian A pācar, Tokharian Β pācar. Only two of the forms diverge from the common model: in Irish and in Armenian, an alteration of the initial p took place. In Tokharian the ā of pācar does not represent an ancient long vowel; and the c (=ts) is a development of the Indo-European palatalized t.
The testimony of a certain number of languages reveals another term. In Hittite we find atta, a form corresponding to Latin atta, Gr. átta (ἄττα), Gothic atta, Old Slav. otǐcǐ (a form derived from atta, coming from *at(t)ikos).
It is a piece of good fortune that we know in Hittite the form atta because the ideographic writing masks the phonetic form of most of the terms of kinship: only “father,” “mother,” “grandfather” are written out; we do not know the words for “son,” “daughter,” “wife” or “brother” because they are written solely by means of ideograms.
Gothic has two nouns, atta and fadar. It is customary to quote these on one and the same plane. In reality the name for father is always atta. We have a single mention of fadar, Gal. IV, 6, where a vocative ἀββᾶ ὁ πατήρ ‘Abba! Father!’ (ἀββᾶ, a traditional form of invocation in Aramaic, taken up by the Greek nominative-vocative) is translated abba fadar. The translator, seemingly wanting to avoid *abba atta, has recourse to the old word common in other Germanic dialects, which has given in Gothic itself the derivative fadrein ‘lineage, parents’.
Everywhere else, Greek patḗr is rendered as atta, including the formula atta unsar ‘Our Father’. Why is it that *pəter does not appear either in Hittite or in Old Slavic? We do not answer the question if we are content to say that *atta is a familiar expression for *pəter. The real problem is much more important: does *pəter designate properly and exclusively physical paternity?
The term *pəter has a pregnant use in mythology. It is a permanent qualification of the supreme God of the Indo-Europeans. It figures in the vocative in the god name Jupiter; the Latin term Jupiter is taken from a formula of invocation: *dyeu pəter ‘father Heaven’, which corresponds exactly with the Greek vocative Zeû páter (Ζεῦ πάτερ). Besides Jupiter, the nominative Diēspiter has also been preserved, which corresponds in Vedic to dyauḥ pitā. To the testimony of Latin, Greek and Vedic we must add that of Umbrian Iupater and, finally, a form less well-known, but interesting, Deipáturos (Δειπάτυρος), glossed in Hesychius: θεός παρὰ Στυμφαίοις ‘God of the Stymphians’, the inhabitants of Stymphaea, a town in Epirus. In this region occupied by an ancient Illyrian population some part of the Illyrian heritage has survived in the Dorian dialect: the form Deipáturos may be a vocative of Illyrian origin. The area of this divine invocation is so vast that we may be right in assigning it to the common Indo-European period as a mythological use of the name for “father.”
Now, in this original usage, the relationship of physical parentage is excluded. We are outside kinship in the strict sense, and *pəter cannot designate “father” in a personal sense. The passage from one sphere to the other is no easy matter. These are two separate ideas, and in some languages they can be mutually exclusive. To make this difference clear, we may refer to the observation of a missionary, W. G. Ivens, who has given an account of his experience in the Western Pacific. When he tried to translate the Gospels into Melanesian, the most difficult part was to express the Pater noster, since no Melanesian term corresponded to the collective notion of Father. “Paternity in these languages is only a personal and individual relationship”;  a universal “father” is inconceivable.
The Indo-European distribution corresponds on the whole to the same principle. The personal “father” is atta, which alone survives in Hittite, Gothic and Slavic. If in these languages the ancient term *pəter has been replaced by atta, this is because *pəter was originally a classificatory term, a fact of which we shall find confirmation when we come to study the name for “brother.” As for the word atta itself, a number of features serve to define it. Its phonetic form classes it among “colloquial” terms, and it is not an accident that names similar or identical with atta for “father” are found in very different languages which are not related, e.g. Sumerian, Basque, Turkish, etc. Furthermore, atta cannot be separated from tata which in Vedic, Greek, Latin, Rumanian, is a traditional childish way of addressing the father affectionately. Finally, as we shall see apropos of the Germanic adjective “noble”: *atalos > edel, adel,  this appellative has produced a number of derivatives which have their place in the vocabulary of institutions.
It follows that atta must be the “foster father” who brings up the child. This brings out the difference between atta and pater. The two terms have been able to coexist, and do in fact coexist, very widely. If atta has prevailed in part of the territory, this is probably due to profound changes in religious ideas and in social structure. In fact, where atta alone is in use, there is no longer any trace of the ancient mythology in which a “father” god reigned supreme.
For the name of the “mother” almost the same distribution of forms is to be observed: the IE term *māter is represented in Sanskrit by mātar-, Av. mātar, Arm. mayr, Gr. mḗter (μήτηρ), Lat. mater, Old Irl. mathir, Old Slav. mati, Old High German muotar. But Hittite has anna-, which makes a pair with atta ‘father’, cf. Lat. anna, Gr. annís (ἀννίς) ‘mother of the mother, or of the father’. The names of father and mother are of parallel formation: they have the same ending in –ter, which had become the characteristic suffix of kinship names, and which later was extended in a number of languages to the whole group of names designating members of the family. 
We can no longer analyze *pəter or *māter, so that it is impossible to say whether from the beginning the ending was a suffix. In any case, this –ter is neither the morpheme of agent nouns, nor that of comparatives. We can only state that, originating in *pǝter and *māter, it became the indicator of a lexical class, that of kinship names. This is why it has become generalized in other terms of this class.
It is probable that the two names for “mother,” *māter and *anna, correspond to the same distinction as that between *pəter and *atta for “father.” “Father” and “mother,” under their “noble” names, express symmetrical ideas in ancient mythology: “Father Heaven” and “Mother Earth” form a couple in the Rig Veda.
Further, only the Hittite group has made anna– (Luvian anni-) into the term for “mother,” like atta (Luvian tati-) for “father.” Elsewhere, the sense of *anna is rather vague; Lat. anna, poorly attested, seems to designate the “foster mother” and this does not accord with Gr. annίs, given in a gloss of Hesychius as “the mother of the mother or of the father.” Terms of this nature do not convey any precise placing in the system of kinship.
The name of “brother” is IE *bhrāter, as emerged from the equation of Skt. bhrātar, Av. brātar, Arm. ełbayr, Gr. phrā́tēr (φράτηρ), Lat. frāter, Old Ir. brathir, Goth. broþar, Old Slav. bratrŭ, bratŭ, Old Pruss. brati, Tokharian prācer. The Hittite name is still unknown. The Armenian form can be explained phonetically by an initial metathesis: bhr– > (a)rb-, which has provoked a dissimilation of the two consecutive r into l–r.
One important fact does not appear in this picture: while Greek has, it is true, the form phrā́tēr, the correspondent of *bhrāter, in the vocabulary of kinship *bhrāter is replaced by adelphós (ἀδελφός) (from which comes adelphḗ, αδελφή ‘sister’). A substitution like this could not be an accident of vocabulary; it is a response to a need which concerns the whole of the designations for kinship.
According to P. Kretschmer  the replacement of phrā́ter by adelphós may be due to a new way of regarding the relationship of “brother” which made phrá̄tēr into the name for a member of a phratry. In fact, phrá̄tēr does not mean the consanguineous brother; it is applied to those who are bound by a mystical relationship and consider themselves as descendants of the same father. But does this necessarily imply that this is an innovation of Greek? In reality Greek preserves here the “broad” meaning of Indo-European *bhrāter which is still reflected in certain religious institutions of the Italic world. The “Arval Brothers” (fratres arvales) at Rome, the Atiedian Brothers (fratres Atiedii) of the Umbrians, are members of confraternities. Where these associations remained alive and their members had a special status, it was necessary to specify by an explicit term the “consanguineous brother”: in Latin, for the blood brother, the expression used was frater germanus, or simply germanus (Spanish hermano, Portuguese irmão), a brother of the same stock. Similarly, in Old Persian, when Darius in his royal proclamations wanted to talk of his consanguineous brother, he adds hamapitā, hamātā ‘of the same father, of the same mother’, cf. in Greek homo–pátrios, homo–mḗtrios. In fact, the “brother” is defined with reference to the “father,” which does not necessarily mean the “progenitor.”
In the light of these facts, *bhrāter denoted a fraternity which was not necessarily consanguineous. The two meanings are distinguished in Greek. Phrá̄tēr was kept for the member of a phratry, and a new term adelphós (literally “born of the same womb”) was coined for “blood brother.” The difference is also reflected in a fact which has often escaped attention: phrá̄tēr does not exist in the singular; only the plural is used. On the other hand, adelphós, which refers to an individual kinship, is frequently used in the singular.
Henceforward, the two kinds of relationship were not merely distinguished but actually polarized by their implicit reference: phrá̄tēr is defined by connection with the same father, adelphós by connection with the same mother. Henceforth only the common maternal descent is given as a criterion of fraternity. At the same time this new designation also applies to individuals of different sex: adelphós ‘brother’ produced the feminine adelphḗ ‘sister’, a fact which completely overturned the old terminology.
There is a specific term for “sister”: Indo-European *swesor is represented in Sanskrit by svasar, Av. xvanhar, Arm. kcoyr (the phonetic result of *swesor) Lat. soror, Got. swistar, Old Slavic sestra, Tokharian šar.
Greek is apparently missing from this picture although the Greek correspondent of *swesor is preserved in the form éor (ἔορ). But this is only a survival preserved by the glossographers. Just as phrá̄tēr conveys a special sense, so the word éor, phonetically corresponding to *swesor, is given with divergent meanings. It is glossed as θυγάτηρ ‘daughter’, ἀνεψιός ‘cousin’, and ἔορες· προσήκοντες ‘relatives’. The term, which is very vague, was applied to a degree of kinship which the commentators were unaware of. This obliteration was due to the creation of adelphḗ ‘sister’, and this in its turn was produced by the transformation of the term for “brother.”
What is the proper sense of *swesor? This form is of exceptional interest because it seems open to analysis as a compound *swe–sor, formed from *swe, well known as a term of social relationship,  and an element *-sor, which appears in archaic compounds where it denotes the female: the ordinal numbers for “third” and “fourth” have, alongside the masculine forms, feminines characterized by the element *-sor: Celtic cetheoir, Vedic catasra, Av. čataṅrō, all deriving from *kwete–sor.
It is probable that *-sor is an archaic name for “woman.” It can be recognized in Iranian in the guise har– in the root of Av. hāiriši ‘woman, female’, where it has a suffix in –iš–i, the morpheme which we find again in the feminine mahiṣi ‘queen’. It is also possible that Skt. strī (< *srī) ‘woman’, is a secondary feminization of the ancient *sor. Thus we can identify the two elements of the compound *swe–sor, etymologically “the feminine person of a social group swe.” Such a designation puts “sister” on a quite different plane from “brother”: there is no symmetry between the two terms. The position of the sister is defined by reference to a social unit, the swe, in the bosom of the “Grossfamilie,” where the masculine members have their place. Later on, at the appropriate time, we shall study more closely the sense of swe.
Unlike the word for “sister” we have no means of analyzing the name for “brother,” apart from isolating the final –ter itself, as in the case of “mother” and “father.” But we can offer no explanation for the root *bhrā-. It is useless to connect it with the root *bher– of Lat. ferō because we know of no use of the forms of this root which would lead to the sense of “brother.” We are not in a position to interpret *bhrāter any more than we can *pəter and *māter. All three are inherited from the most ancient stock of Indo-European.
[ back ] 1. W. G. Ivens, Dictionary and Grammar of the Language of Saea and Ulawa, Solomon Islands, Washington, 1918, p. 166.
[ back ] 2. Book Four, Chapter Eight.
[ back ] 3. Cf. Book Two, Chapter Six.
[ back ] 4. Glotta, vol. II, 1910, pp. 201ff.
[ back ] 5. See Book Three, Chapter Three.