Inka Conical Clan

Download as pdf or txt
Download as pdf or txt
You are on page 1of 37

University of Pittsburgh- Of the Commonwealth System of Higher Education

Proto-Indo-European Kinship
Author(s): Paul Friedrich
Source: Ethnology, Vol. 5, No. 1 (Jan., 1966), pp. 1-36
Published by: University of Pittsburgh- Of the Commonwealth System of Higher Education
Stable URL: http://www.jstor.org/stable/3772899 .
Accessed: 22/06/2014 19:14

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

University of Pittsburgh- Of the Commonwealth System of Higher Education is collaborating with JSTOR to
digitize, preserve and extend access to Ethnology.

http://www.jstor.org

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions
Proto-Indo-European Kinship1

Paul Friedrich

University of Chicago

". . . alle Geschichte ist eben ein Fluss der Entwicklung und
sie Hsst sich demnach nicht in feste Gestaltungen, sondern nur
in Stromungen zerlegen"?Paul Vinogradoff

This paper is part of a more comprehensive study of the evolution of Russian


kinship from the reconstructed stages of Proto-Indo-European (PIE) and
Common Slavic, through the historically attested Old Russian and nineteenth
century Russian, to the trends in contemporary Soviet Russia. By evolution
I refer to how a system of interdependent variables shifts and realigns over a
long time span. Kinship terminology is thought of as a distinctive, Janus-
faced phenomenon which shares many of the formal properties of the
tightly-knit paradigms of a grammar, while also symbolizing a particular
social aspect of changing cultural organization. In the inter-disciplinary
essay that follows I have attempted to show how several different kinds of
phenomena have been related through time as parts of one culture-historical
system.
Two distinct scientific traditions have been brought together. The first
is the vast body of Indo-European scholarship, notably that of Meillet and
Vasmer, and in particular the analyses of early kinship terms by philologists
such as Delbrikk, Schrader, Brugmann, and Trubachev. Second, I have drawn
on the ethnological and social anthropological study of kinship semantics as
exemplified by Kroeber, Radcliffe-Brown, Lounsbury, and Leach, and more
particularly on those who have improved the historical-comparative method
in this area: Rivers, Spoehr, Eggan, and Murdock. In accordance with Indo-
Europeanist standards a considerable amount of data is cited in evidence,
while, consonant with the standards of social anthropology, the relevant
kinship theory has been made as explicit as possible. I do not pretend to
contribute much to the etymological or phonological understanding of PIE
kinship terms, aside from treating them within one general theoretical
framework. Nor do I pretend to add much to our understanding of Omaha
kinship, aside from fully elucidating one instance of the type. In other words,
my use of Indo-European reconstructions and of kinship theory has been
conservative. Yet the present working integration of linguistics and social
anthropology has a certain epistemological interest because it would appear
that kinship semantics is to over-all semantic structure more or less what

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions
2 Ethnology

phonology has been to linguistic structure; the tightness of the two sorts
of paradigms admits a relatively great degree of formalization and of appli?
cation of the comparative method. It is probably no accident that Saussure's
reconstruction in the 1870s of the laryngeals of PIE phonology was followed
shortly by a comparable tour de force in the form of Delbmck's reconstruction
of an essentially Omaha system of kinship.
The stages of the argument below deal successively with the following
topics in Proto-Indo-European kinship: the immediate blood relationships,
then the extended patriarchal family and the relationships through marriage,
and finally the Omaha hypothesis and the refutation of cross-cousin marriage.
Let us begin with a review of the areal and chronological background.

Area and Time Horizon

On the one hand, the identification of a single proto-language with a


single archeological culture is precarious. On the other hand, kinship termi?
nologies do not evolve in a vacuum, and a purely formal analysis of them is
at least implicitly arid. In what follows I therefore assume that kinship
terminologies are part of natural languages and that a prehistoric reconstruc?
tion, to be convincing, must be realistic in the sense of more or less con-
forming to what we have come to know on the basis of field work in
live cultures. In the same spirit I have assumed that terminologies are signifi-
cantly related to the ecology and economy of real people localized in space
(Eggan 1955: 519-548); one burden of the present paper is to indicate,
insofar as possible, the approximate associations of material surroundings
and social organization, even though unambiguous causal relations cannot
be established in the case of speech communities such as the Proto-Indo-
European one where a mixed economy was practiced for at least a millennium
of time over a considerable geographical area.
During the first half of the fourth millennium B.C. a community speaking
the heterogeneous dialects of a Proto-Indo-Hittite language seems to have
been located somewhere in or between the Altai Mountains of Central Asia
and the swampy forests of northeastern Europe. But as Piggott (1950:248)
concludes, "responsible linguists and archeologists have agreed in regarding
the area of origin as relatively limited, and lying somewhere between the
Danube and the Oxus." About the beginning of the third millennium the
Proto-Hittites broke off and eventually reached Anatolia; on philological
grounds, their entry into the latter country can be dated as far back as 1900
B.C. Another section of the community, possibly the Proto-Slavs, may have
produced the Tripolye culture of the western Ukraine, from about 3000 to
2100 B.C. Other PIE peoples may have been scattered through what is today
Poland and the northern Ukraine. In other words, a broad belt of dialects
running from proto-Baltic-Slavic to proto-Indo-Iranian may have been spread
across a continuous ecological zone of fairly open country that allowed rapid
communication.
On several independent grounds I would agree with Schrader and Gim-
butas, who think the most probable homeland to have been the somewhat

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions
Proto-Indo-European Kinship 3

more specific area of the foothills and steppe country from northwest of the
Caucasus to north of the Caspian. In the first place, a constellation of typo?
logical linguistic rarities such as the single "general vowel," taken together
with the basic grammatical design, would link PIE in a Sprachbund with
Proto-Caucasian or General Caucasian as exemplified by languages such as
Kabardian (Kuipers 1960; Trubetskoy 1939), although adequate comparative
work remains to be carried out.2 Second, all the ingenious arguments about
PIE words for fauna and flora (Thieme 1953) that have been used in favor
of northeastern Europe can be applied just as well to Cis-Caucasia, where,
for example, the honey bee, the wolf, and the "shining one" (the beech)
are found in abundance; and in the tributaries to the Black and Caspian Seas
there swim salmon that attain the length of 92 centimeters. In the third
place, several dozen sound etyma refer to (presumably) domesticated animals
such as sheep, oxen, pig, and horse, to domesticated grains such as barley,
and to a variety of cultural artifacts such as the boat, rudder, and oar and
the axle, yoke, and four other specific wagon parts; one form may have
denoted a metal, but there is dispute as to whether it was copper or bronze.
In any case, one can reconstruct a fairly mobile, late Neolithic or Early
Chalcolithic culture of animal-breeding agriculturalists.3
The way of life inferred on purely linguistic grounds jibes with the
archeological "Kurgan" culture of the Cis-Caucasian-Caspian area, where
wagons, livestock, and a mixed economy are fully attested in the excava-
tions. In addition, Kurgan archeology specifically indicates small houses,
small, fortified settlements (Greek polis, Sanskrit pur-as), and the burial of
leaders beneath mounds (Russian \urgdny). The speakers of PIE dialects
probably spent the winter months in the foothills or other sheltered areas and
hunted and pastured on the Cossak steppe during the warmer season, just
as did the Bashkir, Kalmyk, Kazak, and other "South Russian" semi-pas-
toralists of the last century.
During the middle of the third millennium the western sections of this
Kurgan culture, speaking dialects of PIE that were already strongly dif?
ferentiated, appear to have moved toward the west and south, and to have
participated in the so-called Transcaucasian Copper Cultures. Shortly there?
after, having acquired and probably improved on the battle ax and the light,
spoked chariot, the members of the same community spread into southern
and eastern Europe, disrupting Tripolye in the Ukraine about 2100 B.C.
The proto-Greeks probably entered Greece and western Anatolia about 1950;
their advent is associated with the diffusion of "Grey Minyan Ware" pottery.
Other branches had reached the Rhine and the Baltic by 1900; their spread
seems to be associated with the battle axes and corded ware of the Kurgan
culture and its derivatives. Yet other branches moved southwards into Meso?
potamia, establishing the Mitanni confederacy, and also into Egypt as the
Hyksos invaders. An eastern wing, presumably speaking early Indo-Iranian
dialects, entered Afghanistan and had conquered the Indus Valley by about
1600.
From India to the Rhine the migrations of these intruders caused a re-
arrangement of the pre-existing cultural patterns. Everywhere the early

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions
4 Ethnology

Indo-Europeans appear to have established themselves as a dominant class


with local kings or petty chieftains ruling over two or three subordinate
levels. In fact, a form for "ruler" has one of the sounder PIE etymologies,
with phonologically perfect correspondences between widely removed etyma
?Sanskrit ra:j-, Latin rex, and Keltic ri:g (Gothic rei\s represents a borrow-
ing from the latter). Several non-cognate but semantically similar forms,
such as Gothic \uni and Greek basileus, increase the probability that some
fairly definite notion of local, petty king was nearly ubiquitous in early Indo-
European culture.
The relation of the linguists' conclusions to the archeological evidence has
been well summarized by Gimbutas (1963: 827), although I must add the
qualification that a superposed "aristocratic class" with an Omaha social
organization could have co-existed for centuries or even millennia over
plebian strata of artisans and agriculturalists with a different kinship
system and speaking non-Indo-European as well as Indo-European languages.
. . . the Indo-European speakers would have had to be ruled by a powerful aristocracy
who managed to impose the same language and the same social organization over a
vast area of Europe and Asia. Moreover, the written records of the second millennium
B. C. have proven it by vividly disclosing the military prowess and political authority
of the Hittites, Kassites, and Mitanni in Anatolia and Mesopotamia and of the Hyksos
in Palestine and Egypt. [To which might be added the Achaeans in Greece and the
Vedic Aryans in the Indus Valley: PF]. That the Kurgan people succeeded in con-
quering almost two-thirds of the European continent was probably largely due to
their social organization and to the possession of wheeled vehicles. The stock-keeping
and the small patriarchal communities living in small villages on fortified high river
banks [or hilltops: PF] as well as the presence of vehicles is evidenced by Kurgan
finds dating from the last centuries of the third millennium and from successive centuries.
The Kurgan elements fully correspond with the early stratum of Indo-European words
concerning social structure, pattern of habitation, architecture (small rectangular
timbered houses), economy (predominantly stock breeding, farming on a small scale)
and religion (horse sacrifice, sun symbolism, etc).

During the first quarter of the second millennium the material culture of
eastern Europe was already differentiated in a manner corresponding roughly
to the probable location of the major linguistic stocks such as Teutonic and
Slavic. In sum, the dispersal of the Indo-European stocks in all directions but
due north (into the Ural Mountains) is correlated with the expansion over
the same space and time of a distinctive archeological complex; the rationale
of the age-area model solidly supports the hypothetical North-Caucasian-
Caspian origin of the Indo-Europeans. "The only plausible explanation of
the currency of languages so similar over so large an area at the beginning
of historical periods is that they derive from dialects of a fairly homogeneous
prehistoric language which had been disseminated by migrations out of a
smaller region" (Crossland 1957: 16).

The Kinship System

Let us now turn to the linguistic and philological evidence on the kinship
system. In the present paper I have considered ten stocks (Tocharian and
Albanian being slighted for several reasons), although only three to five

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions
Proto-Indo-European Kinship 5

reflexes are normally cited in the text. Special weight has been attached to
the oldest documents in Indic and Greek; both the Rigveda and the Homeric
epics depict cultures of the second millennium B.C. and present a full panoply
of terms enmeshed in an ethnographically realistic context (Piggott 1950: 244-
289). As regards the linguistic scholarship, many of the etymologies and
interpretations cannot be accepted today because they were worked out by
Indo-Europeanists indifferent to or ignorant of social anthropology. On the
other hand, I propose that Delbriick and several others have been essentially
sound in concluding that PIE kinship was patriarchal, patrilocal, and patri?
lineal, and with a system of terms and statuses that would now be classed as
"Omaha." To some extent, their ideas have been substantiated by the Hittite
texts.4
External, ethnological factors also support Delbruck's complex theory.
The tribes of the Caucasus and the South Russian steppes have always been
notably patriarchal, and most of them are patrilocal and have either patri?
lineal or double descent (Kovalevsky 1893: 272-273); by descent I refer to
rules for assigning status on the basis of birth and blood relationship. Patri?
lineal descent and inheritance are comparatively well adapted to economies
based on the herding of livestock by men, on hunting, and on the periodic
division of movable estate (Goodenough 1963: 927); these were very prob?
ably important economic components of large sections of the Proto-Indo-
European speech community. Finally, the Kazak and other semi-pastoral
groups in Southwest Asia have retained Omaha terminologies and patrilineal
avunculates with extraordinary conservatism for hundreds of years (Krader
1955)-

The Kinship Terms

Proto-kinship terms are formulas or diagrams that tell us which systematic


correspondences of phonemes or morphophonemes appear in a set of related
languages; they are also an approximation to the overt forms which formerly
labeled the nodes in a prehistoric network of meanings. What are the recon-
structed terms in this case, and what are their implications ?
It seems probable on the basis of correspondences in meaning alone that
the Proto-Indo-Europeans recognized at least two levels of ascending and
descending generations. But reconstruction beyond parents and children
cannot be made in some daughter stocks, nor beyond grandrelatives (i.e.,
grandparents and grandchildren) in any stock. In most daughter stocks the
grandparents, great-grandparents, etc. are designated by the parent term plus
various prefixes such as pra- (Slavic).
The putative term for grandrelative raises interesting problems. PIE
(h)an/en is reflected in seven stocks, including Common Slavic "grand?
child," vUngJ(U, based in turn on the forms in ten Slavic languages. Stankie-
wicz (1962) has argued that this term is most plausibly interpreted as vUn-
( descending) plus -g/u^U (link), but the majority would tie it to PIE (h)an.
Kurylowicz (1935: 74) has argued that the root began with a Pre-Proto-Indo-
European laryngeal that, as one would expect, was reflected only in Hittite

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions
6 Ethnology

{hannas) and lost elsewhere without affecting the vowel quality. A second
point, of considerable semantic interest, is that, aside from Slavic and Ger-
manic "little grandson" (e.g., Old High German eninchili), the reflexes in
the other four stocks refer to an elder female relative, such as Latin anus
(old woman), Lithuanian anytas (husband's mother), or Hittite hannas
(grandmother). (h)an may also be linked to another well-attested PIE root
meaning "old, old person," (s)en, that began with the "s-movable" and is
reflected in Latin senex and Old Irish sen; the Armenian reflexes hana and
hin further increase the possibility of a relation to (h)an/en. I would con-
clude that the term was probably used to refer to persons in alternate genera?
tions, that it may have denoted any kind of grandrelative, but that it was
possibly limited to the grandchildren and the father's mother or some more
generic category of elder woman in the speaker's family group or household.
All six terms for immediate blood relatives are based on correspondences
of both form and of meaning in seven to eleven of the daughter stocks, and
all are reflected in both the Rigveda and Homer. They and all the other
reliable PIE reconstructions imply that the difference between generations
was discriminated. The line between blood relationships and relationships
by marriage was also sharply drawn in the sense that the reconstructed affinal
terms do not appear with consanguineal referents, and vice versa. The excep?
tions to this generalization have a unique or random character and represent
special histories within one of the stocks or individual languages. Finally,
the sex (as well as the grammatical gender) of the relative referred to was
implied by all the PIE terms except (h)an/en.
In Chart I the symbol /H/ stands for the first laryngeal, which is realized
as a mid-central vowel (shwa) between consonants, and a component of
length after vowels.5 The capital G stands for "found in Homeric Greek,"
the S for "found in the Rigvedas." Here and elsewhere the abbreviations
have the following meanings: Fa = father, Mo = mother, Pa = parent,
So = son, Da = daughter, Br = brother, Si = sister, Sb = sibling, Ch =
child, Hu = husband, Wi = wife. The numbers refer to the daughter stocks
in which are found phonologically more or less regular reflexes with the
specified referent cited.

The term for son, swHnws, is attested in seven stocks, including Vedic
Sanskrit sw.nuh, Avestan hunus, and Common Slavic synl. Following Brug-
mann (1905), many linguists would agree that swHnws, perhaps alone of
the six primary terms, derives from a verbal root, specifically swH-/sew-/sw-,
meaning "to give birth;" Sanskrit shows su:te:, "she gives birth," and sutd-h,
"son." Also, Hittite has the verb has(s)- "to give birth," the participle of
which (hassant-) means "own son" (J. Friedrich 1952: 62). These semantic
patterns are paralleled in several daughter languages. The Latin word for son
is fi:lius (feminine fi:lia). The Slavic word for child is deti, which is related
to the word for "to suckle, milk" (doiti). The roots of both the Latin and
Slavic words, and for several allied notions, go back to PIE dheH-, meaning
"to suckle, to be capable of bearing children" (Benveniste 1933: 15). In
short, the term-concepts for son seem to be related to those for female

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions
Proto-Indo-European Kinship 7

CHART 1

The Proto-Indo-European Consanguines

(maternal) physiological processes. This has been advanced as an argument


for matriliny (Isachenko 1953: 55-56). Yet these and other, more question-
able etymologies, even if valid, in no wise demonstrate matriliny or matri-
archy. In the first place, recognition of maternity is a cultural universal.
Second, the tie between mother and son is often the most dominant emo?
tionally in patrilineal and patriarchal systems, including those of Omaha
type such as of the Kalmyk and Kazak (Krader 1955).
A second term for son is reflected in a proposed set of cognates that in?
cludes Proto-Italic putyo, Sanskrit putram, Greek paid-, and Old Slavic
pUta. However, the phonological correspondences present serious difficulties,
and the term is semantically marginal, since it more probably denoted "boy"
or "lad." It definitely postdates swHnws and probably emerged very late in
PIE times. Within some of the daughter stocks its reflexes give terms for
"daughter" or "girl," as in Indic putri: and Latin puella (from puer).
Trubachev (1959: 51) and other Marxists, by rather fanciful etymologizing,
have related the second and very late "son" term to the PIE root for "penis"
and explained its emergence as part of "the victory of father right."
The term for daughter, dhwgHteir, is probably one of the oldest in the
entire PIE set and is reflected with considerable phonetic regularity in nine
stocks, including Homeric thugdteir and Armenian dustr, although the re?
flexes, like those for "son," have been lost in the Italic and Keltic languages,
Oscan fuitir constituting the lone exception. Both the Balto-Slavic forms
(e.g., Lithuanian du\te:) and the Germanic ones (e.g., Gothic dauhtar)
show the expected loss of the PIE shwa in the middle syllable of a word.
Since Sanskrit /h/ normally reflects a PIE voiced velar stop (always from
gh, and from gh and gwh before front vowel), some scholars have mistakenly
suggested that Sanskrit duhitd: reflected a PIE dhwghHteir, with dissimila-

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions
8 Ethnology

tion of the initial stop following Grassmann's Law. But this runs counter to
strong arguments; above all, the transition from dhwgHteir to Sanskrit
duhitd: is paralleled by several other forms, whereas the alternative transi?
tion from dhwghHteir to Greek thugdteir would be unique.
Other philologists have derived dhwgHteir from the verb for "to milk";
for example, Sanskrit duhitd:/duh- (daughter/to milk) is paralleled by
Common Slavic dU\ti, dllsti/doiti. This hypothesis is weak on linguistic
grounds but may be paralleled by a genuine folk etymology; daughters
usually do function as milkmaids in the cattle-breeding cultures of Eastern
Europe and Central Asia (e.g., the Don and Terek Cossaks).
The most widely attested of all the PIE terms is that for "mother," as in
Old Irish maithir, Primitive Germanic mo:Ber, and Common Slavic mdt(i).
And the constituent phonological elements of the PIE maHteir are reflected
so regularly as to invalidate attempts to derive the term from "childish
babble," or at least to push the originating babble far back of the prehistoric
time horizon considered here. A far less frequent term of secondary status
was nan-/nana-, reflected in several languages including Tocharian and
Hittite. I know of no positive evidence for attributing distinct denotata to
maHteir and nana; judging from the daughter stocks, the latter was prob?
ably a contextual variant for use on less formal occasions.
The term for father (pHte:r) is reflected in nine stocks, including Vedic
pitdr and Gathic (Persian) pta:/ta: (the latter from compounds where the
shwa is lost internally). Other cognates include Primitive Germanic fdder
and even Tocharian paitar. Regular also is the hayr of Armenian, where
initial /p/ goes to /h/, and Old Irish aithir reflects the regular loss of PIE
/p/ initially before vowels in Keltic. An alternative and definitely secondary
term for father (atta/tata) is reflected in a number of stocks and has aroused
a good deal of inconclusive philological discussion; it may be the source of
Common Slavic otltsl. The peculiarities of the Slavic father term, together
with the two uniquely Slavic innovations for the wife's parents, probably
symbolize important social differentiations within the Slavic community very
early after the breakup of the PIE speakers, or perhaps even during the many
centuries or millennia that Slavic constituted one of the PIE dialects.
The analysis of the sibling terms particularly merits attention. BraHte:r,
while denoting "brother" in most of the ten stocks where it is reflected, as in
Primitive Germanic bro:&er, encompasses a broader range of agnates in some
cases. For example, many Slavic languages use derivations from the fraternal
term to designate a wide set of secondary male relatives, perhaps most strik-
ingly in the case of Russian bratdn, meaning (with some dialectal variation)
a brother, elder brother, cousin, brother's son, or nephew. The Homeric
phra:te:r actually meant "member of the same phratry," and the phratry,
within the original framework of the archaic Greek charter, was a patrilineal
group with several important ritual and political functions.
The evidence for the classificatory extension of the brother term is paralleled
by that for the "sister," reflected, in the first place, in Homeric heores, Russian
sestrd, and an otherwise almost complete set that even includes Tocharian
sesdr. Second, Brugmann, Benveniste (1934), and others have analyzed PIE

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions
Proto-Indo-European Kinship 9

swesoir as composed of the roots for "own" (swe-)y which occurs elsewhere
in the kinship terminology, and for "woman," as reflected, for example, in
the first syllable of Avestan hair-is-i. The PIE swesoir was probably of con?
siderable antiquity by the end of the third millennium: "II semble bien en
effet que par ailleurs les deux elements de swesoir n'etaient plus per^us
distinctement des la language commune . . ." (Benveniste 1934: 106). Illus-
trating a somewhat similar conceptualization, the Pre-Irish form for "daugh?
ter" can be analyzed as "the one born inside" (some sort of kinship group).
These two etymologies, as well as structural considerations, tempt one to
postulate for swesoir a more inclusive meaning of "own woman" that would
override generations. But this cannot be done because there is no positive
evidence that swesoir was extended to the father's sister or to the daughter.
Another scholar has argued that the soir element derives from the PIE root
for "blood" (esor/esr), reflected in Sanskrit dsr\, Greek ear, and Hittite
eshar (Mayrhofer 1952: 32-33), but this hypothesis seems untenable because
the /-r/ of "blood" (alternating morphophonemically with /-n/ is actually
the suffix of a heteroclite with the shape VsH-.)
In Ossetian, an Iranian language of the Caucasus, the descendent of the
PIE swesoir is today xo/xwoeroe and means "the woman of my clan," just as
Ossetian oervad, the reflex of PIE braHteir, means "clan brother" (Abaev
1949: 61-63). This type of evidence in "relic areas" of the daughter stocks
reinforces the hypothesis of a classificatory terminology in PIE; specifically,
the two sibling terms may have included certain kinds of cousins, e.g.,
father's brother's children, as is still the case in many Balkan and Caucasian
cultures. Analysis of the two sibling terms as classificatory, as well as other
etymologies of basic kinship terms, is given additional validity through a
second areal consideration: in many Caucasian languages such as Kabardian
the term for example, for brother (qf?a-s), consists of two segments, "son"
(q?) plus "sibling" (i), the sister term consists of daughter plus sibling, and
so forth (Kuipers 1960: 91); the areal pattern seems to be that some basic
terms are analyzable.
Several other PIE morphemes have more special implications. One solid
root is orbho-, reflected in five stocks with the socially related meanings of
either "orphan" or "inheritance" (Ernout-Meillet 1951: ii, 828); in the archaic
peasant system of nineteenth century Russian, for example, orphans adopted
into a patrilocal household usually acquired full rights of inheritance,
and similar patterns may have prevailed in PIE. The term for "widow"
(wydh(e)waH-) is well attested by words with precisely this meaning in
eight stocks, with languages ranging from Old Irish (fedb), to Common
Slavic (vldova), and even to Hittite (sal) udati (J. Friedrich 1952: 237).
One widely accepted and inherently reasonable etymology derives the term
from the root wydh-, "to be empty, inadequate." The strength of the cor-
respondences for "widow" as against those for "widower" may indicate that
the former status was differentiated, perhaps through some rule for the remar-
riage of women; the remarriage of widows was forbidden, for example,
among some early Teutonic and Greek tribes. Or widows may have been
marked for some mortuary ritual; sacrifice and interment with the husband

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions
io Ethnology

have been postulated on purely archeological evidence from the chiefly tombs
of the Cis-Caspian-Caucasian homeland area. In Roman law a widow could
not remarry out of the clan of her deceased husband (Rose 1925: 121), and
concubines were still being sacrificed by the Varangian conquerors of Russia
as late as the ninth and tenth centuries A.D. All such patterns are but extreme
expressions of viripotestality.

The Affinal Set

The category of in-laws is explicit in many stocks; for example, definite


words in Sanskrit (sambandhi) and Russian (svojstvo) function as labels
for distinct classes of affines. In the case of the PIE set, eight secondary affinal
terms invariably implied a distinction of the sex of the linking relative. Five
of the affinal terms were primarily or exclusively employed by a woman
speaking of her husband's close blood relatives and constitute what I call the
virilateral set. All of the five had precise and mutually exclusive denotations,
which will be considered in turn.
First, the term for husband's sister was gHlows, reflected in Armenian tal,
Homeric galois, Common Slavic zlJly or zely, and in two other stocks. In
yet other stocks the same semantic space was perceived but was attached to
phonetically unrelated labels, such as Vedic nanandr. Incidentally, both the
Sanskrit and Slavic forms have been independently folk-etymologized as "the
spiteful, malevolent girl." Second, the term for husband's brother, daHyweir,
was reflected in seven stocks, e.g., Vedic devdr, Old English tacor, Old
Russian deverl, and Latin levir.
Third, as has long been recognized, a special term for husband's brother's
wife is particularly diagnostic of patrilocal aggregates or of some jural
analogue to them. The cognates of ynHteir include Sanscrit yaitar, Armenian
ner, and Latin ianitriiceis. Several of the reflexes, such as Homeric eindteres
and Common Slavic jeNtry, usually occur as morphological plurals or at
least semantic collectives to refer to the collectivity of the wives of brothers.
Unlike any other PIE term, ynHteir was self-reciprocal; a woman was
ynHteir to her ynHteir. As a logical analogue, there is a correspondence
between phonetically dissimilar terms in Sanscrit, Greek, Old Norse, and Old
Russian (svojdf(U), all meaning wife's sister's husband. I would conclude
that neither the term nor the concept of wife's sister's husband existed in the
PIE community, and that the formations in question reflect the development
of extended households, coupled with the bilateral descent of later times
(Friedrich 1964:156-162).
All three of the terms for a woman's affinal age-mates are reflected in six
to seven stocks, some of them separated by thousands of miles; the corre-
spondences could, at the latest, be due to interdialectal borrowings during the
second half of the third millennium B.C. The a priori possibility must be
entertained that some of these terms were extended to the cousins of the
husband who were classed as his siblings because they were part of his
lineage or of his extended family, as was sporadically true even in the
bilateral system of nineteenth century Russian. Here, as in all prehistoric

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions
Proto-Indo-European Kinship ii

reconstruction, one cannot establish absolute verities but only degrees of


probability.
Two terms have been traditionally reconstructed as meaning husband's
father (swe\wros) and husband's mother (swe^rwHs), reflected respectively
in nine and seven Indo-European stocks. Cognates for the latter include
Armenian s\esur and Latin socrus, -u:s. In fact, all stocks but Hittite and
Tocharian contribute to one or both sets. The two terms for parents-in-law
provide one of the neatest demonstrations of Verner's Law, whereby PIE
voiceless stops correspond to voiced spirants in Germanic when not preceded
by an original accent as reflected in Sanskrit and Greek (in this case the
Greek evidence on accent has been eliminated by analogy):

PIE Vedic Sanskrit Russian Germanic Old High German

swelftrwHs svasru sve\rov' svegra: swigar

swe\wros svdsura svfo\or svehra swehur

The term for husband's mother is probably older than that for husband's
father. This is partly because all the reflexes for the former reconstruct to a
base of a single type whereas those for the latter go back to two alternates
(Grammont 1924: 111-118). In the second place, (swe-)f(rwHs is a perfectly
canonical member of a small and probably ancient lexical set that also includes
words for "mouse, sow, fish" and that lacks an e-alternation in the oblique
cases; derivation of (swe-)\wros (with metathesis of /r/ and /w/) would
simply be another instance of a regular and enormously productive process
in PIE morphology?from a base with addition of the thematic vowel?
whereas derivation in the opposite direction would be anomalous and
unique. Temporal priority for the husband's mother term?established
on purely linguistic grounds?fits with the hypothesis of an extended
patrilocal family where the primary relation of the in-marrying woman is
to her mother-in-law.
Both of the PIE parent-in-law terms appear to be compounded of a
reflexive, pronominal morpheme for "own" {swe-), which reappears in
what may be later secondary formations for other affines, such as the Old
High German swaiger, "wife's brother." Swe- (discussed above in connection
with swesoir) appears to have been conjoined with a morpheme for "chief,
power" (l(rwH-), plus affixes denoting sex gender (Schrader 1901: 753). Such
etymological transparency suggests that at some time during PIE unity
(2900-2300 B.C.) the women shifted from a previous usage to these special
descriptive compounds for the husband's parents. Paralleling this hypothetical
but likely development, we find that as late as Homeric and Vedic times
men were still using generic terms with the notion of "own, attached" for
the wife's father (Homeric pentheros and Vedic sam bandhin, both deriving
from PIE bhendh-, "bind, attach"). On the other hand, key exceptions in
the ancient texts make it seem quite possible that derivations from swe\wros
and swe^rwHs could also be used for the wife's parents (Delbriick 1889:

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions
12 Ethnology

516, 528). The regularity and relative functional importance of patrilocal


extension appears to have obscured the exact status (frequency, normality,
and so forth) of this alternative; for example, in the Iliad heXuros appears in
key passages involving the daughters-in-law of Priam, but in the plays of
Plautus, because of a different dramatic emphasis, the term socer occurs only
with the meaning of wife's father.6
Complementing the virilateral set is a very sound proto-term, snwsos,
specifically denoting son's wife, with reflexes in six stocks, e.g., Vedic snusdi,
Latin nurus, Greek nuos, and Proto-Germanic snuzd-. The last form illus?
trates Verner's Law, intervocalic PIE /s/ going to Germanic /z/ when not
immediately preceded by stress. Various interesting etymologies have been
advanced, such as one deriving the term from "son" (swHnws) and another
relating it to "suckle" as "the suckler of children" (e.g., Latin nuitri). It
may be derived from the root for "new" (new-)y i.e., "the new or recent
member of the family," as in Trubetskoy's parallel etymology of newisthai
discussed below. But many linguists would probably follow Brugmann (1907:
315-322) in deriving the term from the verbal root for "tie, bind" (snew-);
the son's wife may have been conceptualized in terms of one of her major
functions: to link two families by marriage (in modern German dialects
Schnur still means both daughter-in-law and knot). The same line of
etymologizing can carry us back to even remoter time depths since both the
kin term and the putative verbal root may derive metaphorically from a more
basic root for sinew or tendon, the son's wife thus holding together two
limbs or parts of a larger kinship body.
An apparently spurious phonological problem has led to a semantically
interesting solution. The problem is that PIE snwsos, if it were feminine,
would allegedly be the unique feminine o/e-stem. Brugmann (1907) argued
cogently that snwsos was originally masculine and meant "link, coupling,
connection." But because it was used so often for the son's wife?the link to
another kin group?it came to be associated with feminine articles and
modifiers and gradually was itself changed to feminine gender. The transitions
would have been from early PIE twos snwsos (masculine modifier and
term), to late PIE twai snwsos (feminine modifier with masculine form
being used with a feminine referent), to twai snwsai (feminine modifier and
term), as illustrated by Slavic ta snU\hd. This semantic process has many
parallels, as in the German dialects where one finds die Fraulein. Meillet, on
the other hand, has mitigated the signiflcance of this ingenious analysis by
showing that the PIE o/e-stems were not sexually marked although there
was a very high correlation with masculine gender; in fact, snwsos is re?
flected by a-stems in most of the daughter stocks such as Slavic, the Greek
nuos constituting the prominent exception.7
A point of considerable importance is that the wide attestation of
snwsos includes no evidence of reference to brother's wife. In a society with
extended patrilocal families one would expect a term for brother's wife on
the basis of contrast and complementation within the system. The absence of
any such term is a good example of what Goodenough (1956: 213) has called
a "zero lexeme," i.e., a situation where, in an otherwise fairly regular para-

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions
Proto-Indo-European Kinship 13

digm, a hypothetical and expectable concatenation of components may be


covered by a circumlocution but is not actually labelled by a specific term.
A possible answer is suggested by the Common Slavic nevest(\)a, with
the meanings of son's wife, brother's wife, bride, virgin, etc; in Russian and
some other languages, the denotations fall into two phonetically marked
subgroups?nevesta meaning bride or virgin, and nevest\a meaning brother's
or son's wife (the latter denotation overlapping with that of sno\hd). Most
of the prolific etymologizing by the Slavists falls into two camps. One relates
the term to the verb for "to lead" (ved-), in which case Common Slavic
nevest\a connoted "the woman led or brought into the household;" this
line of reasoning is paralleled by much philological evidence from Slavic,
Vedic, Greek, and Latin that the bride was thought of as "led in" (uxorem
ducere). By the second etymology, Common Slavic nevest\a is derived from
the verb for "know" (vest-) and means "the unknown one;" this is sup?
ported by a great deal of external, ethnological evidence on Slavic taboos
designed to shield the receiving household from the alien spirit forces of the
in-marrying bride and even her name. Despite their differences, both these
schools assume that nevest\a was purely Slavic. In a radical departure,
Trubetskoy (cf. Trubachev 1959: 92) rejected both "the unknown one" and
"the led one" as Slavic folk etymologies and proposed that the term was
originally "an independent, non-complex word" related to Gothic hauhists
and Sanscrit navisthah, and then proceeded to derive it by an elaborate and
barely tenable series of steps from a PIE newystha:, a feminine superlative
from newos, meaning "the newest, youngest female." Trubetskoy's theory
deserves attention because of its ingenuity and because it fills a semantic
slot, the "zero lexeme" predicted on structural grounds.
A final problem regarding secondary affines is the PIE ?enHr, reflected
in a number of alleged correspondences. One set of such correspondences
occurs in eight stocks with the meaning of daughter's husband. Another
occurs in six stocks with the meaning of sister's husband. Phonetically similar
shapes that may be etymologically related to genHr are associated in six other
stocks with seven kinds of denotata, of which the most significant are the
Keltic forms meaning "daughter" (e.g., Welsh geneth). The Avestan form
zaimaoya means the daughter's husband's brother, and implies a more inclu-
sive component of "younger men in the family linked through the daughter's
marriage."
Only a few of the many phonological and semantic questions provoked
by this putative kinship term can be entertained here. Most important is the
fact that the Greek and Indo-Iranian forms for the daughter's and sister's
husband contain an /-m-/, as in gambros, and that these forms simply
cannot be reconciled with the /-n-/ in all the other forms, such as Latin
gener. Nowhere else in Indo-European does a Greek and Indo-Iranian
/-m-/ correspond to an /-n-/ in the other stocks. The /-m-/ in question may
derive from the PIE root for "bind, marry" (?m-, gem-); the Greek form
probably comes from gamos, "marriage," with the -ros as a secondary suf-
fixation and the /-b-/ as a phonologically conditioned intrusion. In short,
the semantic slot may have been signalled by two proto-allomorphs.

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions
14 Ethnology

Let us turn to the meaning. If we take the scatter of unique denotata to


represent post-PIE developments, the striking fact is that within five stocks
the same terms are used for both daughter's and sister's husband, as in
Homeric Greek (Iliad v: 474), Common Slavic zeNtl (Schrader 1904: 20),
and the gener in the plays of Plautus (ca. 200 B.C). The meaning of
daughter's husband is unquestionably both earlier and primary, but the
reference of sister's husband can be postulated for late PIE times. Thus,
from the point of view of the male speaker, genHr signified the class of
men who, by marrying a woman in the speaker's family, linked it to
another. A well-reasoned etymology by Kiparsky derives the term from the
verbal root meaning "to know" (gn-, gen-), the term thus meaning "the
known man in an affinally linked family." This lumping of the two primary
references of sister's and daughter's husband illustrates within the affinal set
the so-called "skewing rule" of Lounsbury's Omaha III; by this rule, a
man's sister's and daughter's relatives are designated by the same term. In
some languages the reflexes of genHr also have a secondary, generic meaning
of "affine," and two stocks show "groom." On the other hand, genHr is
nowhere reflected as the father's sister's husband, as often happens in Omaha
systems. One cannot postulate this denotatum for the term in PIE.8 Nor is
there evidence that genHr?as one might expect in this Omaha type?was
ever extended reciprocally from the daughter's (and sister's) husband to the
wife's brother, or to the wife's father (the one exception is Classical Greek).
No matter what particular conclusion one draws, the PIE genHr term
presents a fascinating problem of inference. On the one hand, ten stocks
contain a phonetically similar form, eight of which have nearly identical
denotations, and two others, denotations that are semantically related. On
the other hand, the forms appear to have different origins, some derived
from PIE gm-, gem- ("to marry"), others related to PIE genos (patrilineal
kinship group) and genH- ("to procreate"), whereas yet others have been
related to PIE gn-, gnoi- ("to know"). The situation resembles that where
phonetic similarity in false cognates is due to grammatical resemblances of
forms in the mother tongue (Bloomfield 1933: 352); in this case the phonetic
resemblance may be due to a semantic resemblance of distinct forms in PIE.
In particular, the Greek and Indo-Iranian forms with /-m-/ may reflect a
late folk etymology by which a term resembling genHr was derived from
the verb for "to marry." I emphasize that the Indo-Europeanists have not
been able to resolve the various discrepancies in the phonetic and semantic
evidence.
As a hypothesis that might shed some light, I would suggest that the
status of daughter's and sister's husband is of great importance to mobile
peoples allied to each other by distant patrilocal marriage. In very late PIE
times (ca. 2300), the dialects were barely intelligible and in some cases
already differentiated into languages. But families, especially upper class
ones, probably continued to intermarry across the linguistic boundaries. The
kinship terms for brother- and son-in-law, different in derivation but similar
in sound, evolved as a consequence of and in order to facilitate interdia-
lectal and interlingual communication. The same causes may underlie the

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions
Proto-Indo-European Kinship 15

remarkable similarity, mentioned above, between PIE snwso-s and the South
Caucasian nosa, "son's wife."
The affinal terms discussed so far are presented in Chart 2. In this chart
the triangles denote male sex, the circles denote female sex, the equal signs
denote marriage, the horizontal, linking bars denote siblingship, and the
vertical bars denote parenthood.

The Patrilocal Unit

The affinal terms discussed so far can now be conjoined with the evidence
from legal texts and archeological sites in order to give a fuller idea of norms
and behavior.
In the first place, the five virilateral affinal terms plus snwsos would
normally imply patrilocality (or viripatrilocality), namely, that upon mar?
riage the girl removed to the familial group of her husband's father. By way
of contrast, the emotional and legal ties of the husband to his wife's relatives
appear to have been comparatively neutral or unimportant. The conclusion
of Meillet (1937: 391) deserves to be cited, although it requires serious
qualification as to the status of the wife's brother (syVHr-), and the sec?
ondary usage of the parent-in-law terms, not to mention the ambiguity of
his phrase "parente masculine":
Tout ceci indique un etat social ou la femme entrait dans la familie du mari, mais ou
le mari n'avait pas avec la familie de sa femme une parente. II s'agit de ces "grandes
families*' a parente masculine, telle qu'on les observe encore chez les Serbes (zadruga)
et chez les Armeniens.
Put more generally, the manner in which the sex of the linking relative was
regularly discriminated in all the affinal terms except those for spouse is as
diagnostic of unilocality as the overriding of generation in Crow-Omaha
systems is diagnostic of unilineal descent.
In the second place, all the evidence from the early texts, including the
Vedas, Homer, the Hittite and Roman laws, and Tacitus, indicates strong
patripotestality, that is, a concentration of rights to the father as against

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions
16 Ethnology

obligations to the son. The relations between Telemachus and Odysseus are
typical. The father often had the right of life and death over his own
children, as in the Roman patriapotestas. In short, the extended family was
patriarchal.
Third, there is external evidence on an extended family group and its
chief. A wide and long-recognized set of correspondences is at once phono-
logically regular and widely divergent in meaning, ranging from a patri?
archal household (the Homeric F6i\os), to a village quarter (Gothic weihs
and Latin uixus), to a family plot of some kind (the Common Slavic
vlsl), to a local district (the Indian vis), to a clan or lineage of some
sort in yet another Homeric context (Meillet 1937: 392). The associated
meanings of this set of cognates indicates some sort of extended familial
group, possibly linked to a plot or domicile, possibly connected to a patri?
lineage, although there is admittedly no way of deciding unequivocally
between these semantic antecedents. In addition, there is a second, excellent
set of correspondences, ranging from Greek despoteis to Old Prussian
waispattin, that generally serve to denote the head or chief of a family; they
contain the same root for power (pot-) that shows up in the term for
husband (potys) and, possibly, for nephew (nepot-); some component of
authority must be postulated here to cover the clearly politico-governmental
meanings. For such words which symbolize the intersection of the kinship
and some other semantic domain I would suggest the expression "quasi-
kinship terms."
Last, we must consider the PIE domicile. The proto-language actually
shows two words for house: (1) domos, as in Sanskrit ddmah, Greek domos,
and Russian dom, and (2) a second, West Indo-European set with /b/ that
is reflected in Germanic, Old Welsh treb, Lithuanian troba, and Latin trabs.
Some physical features of this domicile can be determined on philological
grounds?for example, the walls were probably of wattle. But for the dimen?
sions and the number of rooms we must look to the archeological evidence.
As a matter of fact, the third millennium sites of the postulated PIE home-
land, whether northeastern Europe or the Cis-Caspian-Caucasian area, reveal
relatively small houses or huts of one to three rooms, each room with a floor
space of about nine to twenty square yards. It has been shown above that
the affinal terminology makes patrilocality probable. Putting the two patterns
together, we may conclude that the dwellings in question accommodated
either immediate families or small patrilocal or joint fraternal aggregates
consisting of two to four immediate families, and that in many cases the
patrilocally grouped individuals may have lived together in separate but
adjoining or neighboring huts within the same village. This type of patri?
local group "on the ground," to use Leach's catchy idiom, is eminently
compatible with an Omaha system. But more cross-cultural research needs
to be done on the probabilities of association between various kinds of groups
on the ground, patrilocal and otherwise, and the various types of Omaha
terminology. After their dispersal, the speakers of Indo-European languages
constructed many large houses and "palaces," notably the Mycenaean megara
and the great halls of northern Europe.

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions
Proto-Indo-European Kinship 17

The house types of the PIE homeland fall sharply into two halves: the
small huts just described and the multichambered long houses of Tripolye
and Danubian I. Archeologists generally assume that the Tripolye culture
was matrilocal and matrilineal, but this is actually a hypothesis that remains
to be tested by the new techniques currently being developed by Dietz,
Binford, and others for inferring social structure from artifacts. The particu?
lar material traits could perfectly well have meshed with patrilocal extended
families; the "female figurines"?even one per room as in Tripolye?no
more prove matriliny than do ikons of the Virgin Mary. European archeolo?
gists generally feel that the remarkable Tripolye culture was not Indo-
European, but others, including the present writer, think that they might
have been occupied by one stock such as the Slavs; despite the repeated
invasion and intrusive migration of alien peoples, the Slavs have since the
dawn of history repeatedly reoccupied and repopulated the Ukrainian and
Cossak steppe. In addition, the Common Slavic kinship terminology includes
certain unique innovations involving affinal and consanguineal affiliation
through women, that may reflect culture contact with a non-Indo-European
people.
This brings us to the single affinal term that implies a linking wife rather
than a linking husband: syVHr, the wife's brother. This term, like those for
the avunucular-nepotic dyad and the husband's brother's wife, is of singular
signiflcance for the historical theory here being developed. In the type of
Omaha system which PIE appears to have had, a single powerful term?
powerful in the sense of referring to many kin types?may comprise a wide
spectrum of referents ranging from the wife's eldest brother, to any true
brother, to some larger category that includes all her lineage brothers
(especially parallel cousins) or clan brothers, to a maximal classification that
covers all her male clansmen in her own or descending generations. Strong
solidarity and ritual obligations often link the woman and such a "brother,"
the latter "sewing together" or binding the new couple at the wedding. The
referential range of such a term can expand and contract. In later stages,
and consonant with altered social conditions, what was originally a classifica?
tory, Omaha type of "wife's brother" may be reduced to denoting only the
wife's (eldest) natural brother; or various formal derivations from the
original term may separately denote the various kin types. With these con?
siderations in mind, let us turn to the PIE syVHr.
First, the distribution is peculiar in that it is reflected in only two stocks :
Slavic and Indic. One cannot preclude borrowing between dialects during
the first half of the third millennium B.C, when the speakers were still
contiguous. Nor can one rule out the possibility that during PIE times the
adjacent Slavic and Indic "kinship idioms" constituted?with respect to this
feature at least?a special variety of the overall PIE Omaha. In the second
place, the original term is very widely reflected in the languages of the two
stocks. The Common Slavic surl has yielded forms in Serbian, Polish, Rus?
sian, and many other languages and dialects of the three divisions of Slavic.
There are two extensions in meaning: in Old Russian suric denoted the
wife's brother's son, and in Czech two distinct derivations refer to the wife's

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions
18 Ethnology

brother's son and to his wife. Within Indic, the Vedic cognate is sya:la-h.
The /!/ in this case is the East Indic reflex of PIE /!/ or /r/ that appears
in the written form of many Sanskrit words. The u/a: correspondence
is so inconclusive?leads in such diametrically opposed directions?that I
have only postulated an unspecified nuclear vocalism.
The modern languages of India provide evidence especially interesting and
relevant to semantic analysis. Eight of these languages have a reflex that
denotes wife's brother's wife (e.g., Hindi salhaj); ten show a denotation of
wife's sister (e.g., Hindi sa:li); ten show the wife's sister's husband (e.g.,
Hindi sairhu); and almost all show additional reflexes with the presumably
basic or primary denotatum of wife's brother (Ghurye 1955: 243). These
post-Vedic extensions are exactly the residue one might expect from a
classificatory term for "wife's brother" in PIE; members of the original class
have been subdivided and assigned to formal derivations of what was once
a single kinship term. But there is no evidence that syVHr was used for
wife's father, as one would expect on the basis of Lounsbury's rule for
Omaha III whereby a linking woman's brother is lumped with her father.
Etymological and legal evidence buttresses the distributional arguments
just advanced. Thus the soundest of the various etymologies of syVHr
derives it from the PIE verbal root meaning "to sew" (as reflected in Slavic
siti), by somewhat the same reasoning that the son's wife (snwsos) was
interpreted as the linking sinew. The already noted analyzability of kinship
terms in the Caucasian languages lends credibility to these explanations,
although on the basis of our experience with living systems it seems unlikely
that even a large minority of kinship terms should be reducible to verbal
roots, and even less likely that the same root (e.g., pot- or sew-) should
underlie two or more kinship terms. More immediately germane is the fact
that in many early legal systems, such as those of Scandinavia, the brother
did play a major role: in arranging the dowry, at the wedding,
regarding
his sister's inheritance rights, protecting his sister in case of mistreatment,
and in repaying the bride-wealth on separation. In Roman law the status of
guardian over the mother and sister passed within the family from father
to son. In peasant Russia these fraternal functions were very prominent in
early times and endured in modified form well into the present century.
This type of ethnological evidence suggests that the PIE syVHr denoted
several non-elder male relatives of the wife and that
they probably had
considerable ritual and legal significance.
Chart 3 presents some of the semantic dimensions that have been discussed
above in connection with the affinal terminology. The meaning of the
symbols in the chart is as follows: A = affine(s); C = consanguine(s);
AC = consanguine of (Ego's) affine; CA = affine of (Ego's) consanguine;
Link = linking (with reference to the sex of the linking M =
relative);
male; F = female; Gen = generation (of the relative in relation to ego's
? =
generation); +, 0, senior, same, or junior generation.
Let us conclude the discussion of the affinal set by
turning to the spouse
terms. Proto-Indo-European lacks a word referring specifically or exclusively
to a spouse. Scholars have therefore thought that
gwena:, "woman," might

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions
Proto-Indo-European Kinship 19

CHART 3
Semantic Discriminations for the Proto-Indo-European Secondary Affines

have denoted the wife, or at least various classes of women preferred or


prescribed for secondary marriages, such as the brother's wife. In a parallel
manner, men- denotes "man," "husband," or both in many stocks. Thus
there is some basis for believing that both these terms were originally used
with generic meanings, and that they were combined with the possessive pro-
noun to express the notion of a spouse. In the daughter stocks the two roots
have been narrowed down to the specific denotatum, or expanded to the ge?
neric one, or have evolved into some combination of both, as in Russian
muzh, meaning husband, and muzhchina, meaning man. In addition, several
stocks show a correspondence for "husband, master" (e.g., Sanskrit pdtih,
Greek posts, Latin potis), with such strong connotations of "master, power?
ful one" that we are justified in taking it as evidence of viripotestality. Here,
as elsewhere in comparative semantics, the connotations of the proto-term are
inferred from the polysemous, associational meanings of what are otherwise
phonologically regular correspondences.
An intriguing solution for the uxorial category rests in part on an ety?
mology of the Latin word for wife, uxor. This word is probably of great
antiquity and may, in the opinion of Ernout and Meillet, derive from the
root ew\-, "to be habituated, to learn"; by this interpretation, w\soir would
have referred to a category of originally alien women to whom one became
accustomed through marriage. Similar patterning is found in contemporary
Armenian, where amusin means spouse, whereas the closely related verbal
form, usanim, means "I learn." The meaning of "accustomed woman" is not
only congruous with community exogamy and distant patrilocal residence,

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions
20 Ethnology

but falls into a pretty paradigm when taken together with the sounder
analyses of swe^rwHs and swesoiri

swe?l{rwHs mother-in-law, "own mistress"

swe?soir sister, "own woman"

w\?soir spouse, "accustomed woman"

Before turning from the affinal to the consanguineal system, a few words
are in order regarding an important and functionally interrelated set of
norms and values. In the first place, there is some evidence on bride-capture.
For example, the warrior caste in ancient India had a special term for the
practice; symbolic bride-capture was an integral part of many early or
peasant wedding rites, notably among the Slavs; and it used to be a hallmark
of life in the Caucasus. One may realistically postulate that bride-capture
figured in the PIE marriage ritualism and that it may have been an occa?
sionally practiced alternative among the small and probably exogamous com?
munities of the putative PIE homeland. In other words, bride-capture is
not entirely the romantic fantasy of anthropologically nai've, nineteenth
century philologists. In the second place, there is widespread evidence of
bride-wealth, be it in frozen, archaic phrases such as the Greek term for
virgin (alphesiboia, "a girl who gets her parent a good price"), or in the
statements of ancient texts, as when Strabo reports a span of oxen as the
normal bride-price among the Indians. Such bride-wealth would have served
to compensate the girl's kin. Third, there is considerable evidence for
polygyny, reported sporadically for the Vedic royalty, the Old Persians
(Herodotus i: 115), the Achaean Greeks, the early Germans, and the Old
Norse and Old Russians. In most instances, polygyny appears to have been
an alternative that was limited to the upper strata of the society. Fourth,
most of the evidence from the epic songs, the legal texts, and the ethnological
sources indicates that the husband had strong rights over the wife; the
psychological subordination and jural obligations of the in-marrying girl
were particularly clear in the extended households of nineteenth century
peasant Armenia, Russia, Serbia, and North India. On the other hand,
certain economic rights of the woman appear to have been carefully pro?
tected and there was probably considerable individual and regional variation
in the domestic power of the wife. Fifth and last, the ethnological and
philological evidence on concubines is supported by an interesting set of
apparent cognates dredged up by Dumezil (1940: 68-69): Avestan pairi\ai
(enchantress, prostitute) and, with the meaning of concubine, Old Irish
airech-, Greek para\oitos, and Armenian hare (from pargyai, although it
may be a Sassanian loan). To conclude, all five of the kinship patterns just
sketched require further research, but the brief review has been suggestive.
Bride-capture, bride-wealth, polygyny, a strong manus mariti, and con-
cubinage, all articulate functionally with the patrilocal family and the
Omaha terminology that are being argued on other grounds.

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions
Proto-Indo-European Kinship 21

The Omaha Problem

Patriliny and the Omaha complex, having already been referred to at


several points, may now be tackled directly. The case for patriliny is
enormously buttressed by external evidence, i.e., by statements in ancient
texts referring to descent and descent groups. Despite taxonomic confusions
in the earlier literature, the external evidence on patriliny is generally strong.
It is summarized in part below.
Northern and western Europe appears to have been an area of endemic
bilaterality combined with agnatic structures of various kinds, whereas
ancient Greece and India yield the best evidence on patriliny. I have not
found these distributions to be convincingly explicable in terms of ecology
or culture contact. Various Keltic and early Germanic categories, such as
the Gothic sibja, referred to bilateral descent groups; Irish cland, to take
a second example, referred to the children of an eponymous ancestor. But
even in these stocks personal naming was patrilateral, and agnatic ties were
comparatively emphasized. In the early Norwegian laws regarding wergeld
and blood vengeance, for instance, the father's brother and brother's son
generally ranked just after members of the immediate family (Vinogradoff
1900: 27). Many Germanic tribes had additional, named agnatic categories.
For the Slavs the evidence on early bilateral organization is overshadowed
by that for patriliny, ranging from the deep genealogies of the Serbian
peasants and Polish aristocrats to the shorter lineages of the Great Russian
Cossaks. The descent system of the ancient Indians was explicitly patrilineal.
The Vedas testify to agnatic groups with depths of four to thirteen genera?
tions, and the Vedic gotra was an exogamous, non-localized patrilineal sib.
Old Latin, like Irish, had a bilateral descent group with certain functions in
ritual and marriage, but the agnatio was an Ego-oriented patrilineal kindred,
and the gens was a non-localized patrilineal group associated with an altar,
ancestral spirits, vengeance and marriage rules, and descent reckoned from
an ancestor who might be either historical or fictive (Ghurye 1955: 152-188).
The Greek genos, though not mentioned in Homer, nevertheless figured
as part of the archaic charter, and membership definitely passed from father
to son (Dittenberger 1885). In ancient Greece the genos was "a great patri?
archal family" of one or more constituent households that retained certain
traditional lands (Asheri 1963: 7). By the fifth century B.C. the genos was
a group of families tracing descent back as far as 30 generations, regulating
marriage to some extent, and governing certain rituals (Hignett 1952).
Thirty gene: made up the phratry, a confused territorial unit after the migra?
tions into Greece but apparently a non-localized patrilineal group before that
time.
In addition to these social structural parallels between the various Indo-
European tribes, there is a set of regularly corresponding terms sharing the
component meaning of (patrilineal) kinship group: Sanskrit janas, Greek
genos, Latin gens, genus, and Gothic fyini. Philologists often confuse the
members of this set with those that underlie the previously mentioned
proto-term for "daughter's husband." In fact, comparison of the two sets

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions
22 Ethnology

leads one to reconstruct two forms with identical roots, genH-os and genH-r,
both of which involve the e-vocalism and which belong together in terms of
laryngeal theory. I take this to be an example of a fairly frequent coinci-
dence whereby two similar terms may denote two semantically related
ideas. In the present case, genH-os may have denoted not merely "patri?
lineal group" but more specifically the patrilineal group or groups that were
linked through marriage to that of the speaker. Consonant with such an
interpretation, genH-r may have denoted that member of the group or
groups who was functionally the most salient?the group member mar?
ried to the daughter of the speaker. But it must be remembered that there is
no direct evidence in any of the daughter stocks for the hypothesis that
genH-os denoted something as specific as a uxorilaterally linked patrilineage.
Some suggestion of how PIE speakers conceived their patrilineal group?
ings comes from two etymologies that closely parallel one another. First, the
genH- root is related in sound to a second set that means two discrete but
polysemously related things: (i) "to give birth, generate, geminate," and
(2) "knee, link," etc. (both having the form gon-, gen-). In the second
place, PIE (s)f(we/ol appears to have denoted both "to grow" and, by a
second set, "member, knee," as reflected in Greek \01lon, "member," and
\01lem, "backbone," and s\elos, "shank." In Common Slavic \oleno meant
"knee, stem, offspring." In short, the PIE peoples, like many Central Asiatic
tribesmen, may have conceptualized their families, lineages, and clans as
metaphorical parallels to the human body or a tree, in which kinship statuses
and subgroups functioned as knees, elbows, branches, members, etc.
I would conclude that the evidence indicates patrilineal descent. First, the
Proto-Indo-Europeans probably behaved in a manner that produced aggre-
gates of people related through men, and, second, they seem also to have
had a patrilineal ideology, an explicit, consciously articulated system for
reckoning descent. The culture probably had bilateral kindreds as well, but
matriliny only is attested among the Lycians of Asia Minor.
The terms for secondary and tertiary consanguines?blood relatives linked
to the speaker by one or two intervening relatives?lend indirect support to
the hypothesis of patriliny in PIE times. In some instances, e.g., certain
Vedic passages, the father term was extended to include the father's brother.
And on purely linguistic grounds the PIE term for the latter (pHtrwos)
should be interpreted as a derivation from the former (pHteir), or both
are derived from one common antecedent. In other words, the avuncular
terminology was probably "merging" in pre-PIE times. But the bulk of the
evidence makes it yet more probable that during much of the PIE period
the father was differentiated from the father's brother. In three stocks there
is a distinct if obviously derivative term for the latter: father's brother is
pitrvya in Vedic, patrois in (Pindaric) Greek, and patruus in Latin. In yet
two other stocks the same derivation is less simple but etymologically rather
sound: Zend tuiirya and Common Slavic stryjl (Vey 1931; Mikkola 1908-
09: 65). The very complexity of these etymologies increases the probability
that within PIE itself there was a terminological-conceptual distinction
between the father and his brother. In those stocks which lack a reflex of

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions
Proto-Indo-European Kinship 23

pHtrwos, the father's brother is symbolized by a variety of patterns, such as


descriptive compounding. Finally, the Germanic classifications and some
Greek evidence (Delbruck 1889: 500, 502) suggest that pHtrwos was actually
used for a larger set of secondary and tertiary agnates, for some definable
set of older men in the speaker's patrigroup. In any case, the "father's
brother" was distinguished from the "mother's brother" (awyos); the two
kinds of uncle were "bifurcated." Since collateral distance was recognized
between father and his brother, one may class the PIE avuncular terminology
as bifurcate collateral. Such bifurcate collaterality is a good index of unilocal
residence and of unilineal descent (Murdock 1941: 148) or of some other
juridical or spatial differentiation of the relatives in question.
Less positive evidence is provided by the amital nomenclature. The father's
sister was denoted in several ways in the daughter stocks. Descriptive com-
pounds appear in Greek, Armenian, and Sanskrit (pituhsvasar). Baltic
shows a special form (teta, from a Balto-Slavic root for aunt), whereas
many Slavic languages show special terms derived from the root for father's
brother (e.g., Old Church Slavic stryj\a). Latin had a special form, amita.
The Old High German special form, Base, was later extended to mother's
sister, father's brother's wife, niece, etc. In these various stocks we therefore
note a widespread tendency to differentiate the father's sister by linguistically
distinct patterns, possibly indicating some sort of convergence after the
breakup of the PIE community. On the other hand, the absence of a PIE
term for father's sister does not strongly imply that the speakers lacked a
way of referring to such a relative (e.g., by means of circumlocutions), or
did not refer to her; it does accord with the theory that distant or at least inter-
communal patrilocal residence removed such a relative from the speaker's
immediate ken. Classical Mongol and a few other Central Asiatic Omaha
systems also lack a term for father's sister.
The other aunt, the mother's sister, may have been named by the pattern
just described for the father's brother, i.e., by a derivation from the term for
"mother," as in Old English moidrie and Latin matertera (the latter also
meaning "a kind of mother"). But the mother's sister is no where actually
identified with the mother, and even the derivations follow distinct morpho-
logical patterns. In a parallel manner the term for the stepmother is derived
from that for mother in a number of stocks, as in Armenian mauru and
Common Slavic matjexa or mdshte\ha, but these terms appear to represent
convergence, and the PIE term cannot be reconstructed. Incidentally, the
evidence is generally weak for the sororate, or marriage with the deceased
wife's sister. I would conclude that the amital nomenclature was probably
collateral and possibly bifurcate. Perhaps the terms were derived from the
distinct terms for uncles, as is still true in several of the daughter stocks such
as Slavic. In any case, the amital nomenclature must play a secondary role
in a theory based on positive evidence.
Throughout this paper the attempt is to reconstruct conceptual categories
as labelled by proto-morphemes and kinship terms. The Omaha theory rests
in part on one such category, a true structural keystone: awyos. (The over?
all evidence appears on Chart 4, where parentheses refer to lesser frequency

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions
24 Ethnology

CHART 4
Evidence on Proto-Indo-European nepoit- and awyos

or salience). Awyos is attested in the daughter stocks with meanings that


range from just the mother's brother (e.g., Lithuanian avynas), to the
parent's father (Hittite huhhas), to the mother's father (Lycian xuga).
Even this distribution would be highly significant, since it groups the
mother's brother and her father. But within two stocks the mother's brother
and the parent's father are lumped together, as in the Old English eiam of
Alfred's Orosius. The largest and tightest network comes in Latin, where
avus meant the parent's (usually the mother's) father, while avunculus or
"little grandfather" was used to denote the mother's brother; ava meant
mother's sister and avia the mother's mother. In sum, the single proto-term
awyos is reflected with one type of denotation (grandparent, mother's father)
in one group of stocks, with a second type of denotation (mother's brother)
in a second group, and with the combination of both denotations in yet two

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions
Proto-Indo-European Kinship 25

other stocks; the third grouping is highly congruous with a developed system
of patrilineal descent. I would conclude that awyos in PIE, or at least in
several dialects of PIE, denoted a set of older men in the mother's patrigroup.
This conclusion is supported by the fact that the postulated awyos category
was linked to its logical reciprocal; in Old Irish the form (h)aue denoted
not an avuncular but a nepotic or grandchild category. In some medieval
Germanic dialects the same term for uncle (e.g., oiheim) or for nephew
(e.g., neffe) occurs in texts with the reference of either mother's brother or
sister's son; that is, either term could be used by either relative for his alter.
But further inquiry is called for by the argument of Schrader (1904: 15)
that this reciprocal usage actually postdated Old High German; oiheim
itself is a secondary formation possibly based on a folk etymology, since
-heim is a distinct root.
On the basis of the Hittite huhhavs, Kuryowicz (1935: 74) has argued
rather convincingly that the Pre-PIE term had an initial laryngeal; this
squares phonologically, since /h/ or zero are the Hittite reflexes of the second
laryngeal, which yields PIE /a/, and /hh/ corresponds to the third laryngeal,
which yields PIE /o/. Austin (1942: 22) has derived both the PIE and
Hittite forms from Proto-Indo-Hittite xauxos, but this is contested. These
and other factors argue for the extreme antiquity of the term and concept in
question. Interestingly enough, however, awyos is not directly reflected in
the kinship terminologies of either of our major sources of evidence: Homer
and the Rigveda.
The PIE awyos category was completely paralleled by the nepotic cate?
gory; nephews and nieces were apparently classed down wards with grand?
sons and granddaughters as nepot- and neptyH-. Both the avuncular and
nepotic categories overrode the distinctions of generation and degree of
collaterality. In Germanic, Latin, Slavic, and Keltic there is evidence that
the downward classification of the nepotic kin specifically involved the
sister's (as against the brother's) son and daughter. This is illustrated by Old
High German nevo, by the nefa and nift of Old English, and finally by Old
Irish niae and necht. Such a grouping of a man's sister's children with his
grandchildren is the logical reciprocal of grouping the mother's brother with
the grandfather. As Delbruck (1889: 504) put it, "Ich nehme also an, dass
die Bezeichnung nepotes von dem dvos ausging. Ist dieser der mutterlicher
Grossvater, so sind die nepotes ihm gegeniiber Enkel, ist er Oheim, so sind
sie ihm gegeniiber Neffen."
This kind of equation is diagnostic of Omaha systems found in many parts
of the world. More particularly, it illustrates Lounsbury's Omaha III skew-
ing rule, which equates relatives linked through a man's sister and his
daughter; it also illustrates the corollary to this rule, which equates a
linking woman's brother with her father. However, the evidence, though
encouraging, is not conclusive that PIE exemplified the yet more fully
developed type of Omaha system wherein only the mother's (but not the
father's) father and only the daughter's (but not the son's) children are
lumped, respectively, with the maternal uncle and sister's children. In such
a case, the avuncular category refers clearly to older men in the mother's

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions
26 Ethnology

patrigroup, whereas the nepotic category refers to the children of non-elder


females of the speaker's patrigroup.
There is no evidence for the downward classification of the brother's
children, who may have been lumped with a man's own children; it is also
possible, though less probable, that a sister's children were classed with the
children of a female speaker. Were such the case, it would illustrate what
Lounsbury has called the "merging rule," whereby offspring of siblings of
the same sex are seen as belonging to the same category. On the other hand,
since there is strong positive evidence that the father's brother (pHtrwos)
was distinguished from the father, it seems unwise to assume, on the basis
of negative evidence, that the children of a man were lumped with those of
his brother. In all these discussions it must be remembered that "brother"
and "sister" probably referred not only to the true siblings but to the mem?
bers of the classificatory units analyzed above as "own woman" or "clan
sister" and as "clan brother."
Aside from the logical implications of the reconstructed nomenclatures,
certain etymological and ethnological evidence suggests inferences about
the structure of sentiment. Brugmann (1903) related awyos to Homeric aia
(earth) and the Greek conceptualization of the earth as the primary source
of life. On the other hand, there is some evidence from Latin and early
Keltic literature of the mother's brother playing a stereotypically fond and
protective role vis-a-vis his sister's son. Tacitus in his exemplary ethnography
of the primitive Germans depicts a beneficent avunculate, and other Ger-
manic and Old English oral literature indicates a special degree of fondness
and loyalty between these two relatives which in matters of vengeance
and even inheritance seems at times to have taken precedence over
the father-son tie; the relations of Beowulf to Hygelac and of Gawain to
Arthur are illustrative (Gummere 1901: 137). Similar patterns have been
consistently reported for many Central Asiatic tribes, and from the Caucasus
since Herodotus. Partly for these reasons philologists have argued that awyos
originally connoted "beneficent, friendly" and may have been derived from
the PIE root for "to rejoice" (aw-), as in Sanskrit dvati, "he is rejoicing." In
addition, the PIE nepotic terms have been etymologized (Trubachev 1959:
77-78) as "non-powerful" or "non-older" (ne- plus pot-/pt-). All these etymol?
ogies are precarious on phonological and grammatical grounds but deserve
mention here because they point to the sort of affective relationship between
classificatory uncles and classificatory nephews that is often associated with
patriarchy and patrilineal descent; the awyos may have acted as the permis-
sive "male mother" to the demanding sons of his clan sisters, particularly of
his own sisters and daughters (Radcliffe-Brown 1952: 15-32). As Delbruck
(1889: 505) noted long ago, the uncle would "den Kindern nicht mit dem
Ernste des vaterlichen Grossvaters, oder als patruus, sonder mit der Freund-
lichkeit des Gonners gegeniibertreten." Such an avunculate meshes neatly
with an Omaha system.9
The case for an Omaha system developed above provides indirect or in?
ternal arguments for PIE patriliny, arguments that are strongly supported
by the results of cross-cultural, statistical analysis. Cottrell (1965) points out

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions
Proto-Indo-European Kinship 27

that the 45 Omaha terminologies in Murdock's "World Ethnographic Sam?


ple" show the following correlations: 42 with patrilineal descent, 0 with
matrilineal, 2 with bilateral, and 1 with double descent. The proportion of
42 out of 45 is most significant, even after we allow for variations in the
meaning of "patrilineal," for the fact that cross-cousin terminology is only
one of the diagnostic traits of an Omaha terminology, and, finally, for the
obvious fact that the probabilities are very different in the opposite direction,
only about one in four patrilineal systems being "Omaha" by any definition.

The Cousin Terms and Cross-Cousin Marriage

We have now discussed the immediate family, the affines, the avuncular-
nepotic set, and patriliny. The fifth and final problem is the PIE cousins.
First, the cousin term is comparatively unimportant in many ancient texts,
such as Homer, where anepsios is mentioned only five times. Second and
more significant, a morpheme for cousin cannot be reconstructed for PIE,
nor even for some of the daughter languages such as Common Slavic.10 This
alone would suggest that cousin types were classed with other relatives, in
various possible ways, some of which should be considered. For example,
all the cousins to a given degree may have been lumped with the siblings
(with maybe an adjective attached to the sibling term, as in modern Rus?
sian) ; the PIE roots for brother and sister are in fact based on nearly perfect
correspondences of form and meaning in nearly all stocks, as would be
consonant with, although not necessarily generated by, a classificatory sys?
tem in which the terms for primary consanguines were extended to col?
laterals.
A second possibility, suggested by Balkan data, is that the father's brother's
children were classed with siblings, while the cross-cousins were classed with
spouses, and the mother's sister's children were excluded altogether from the
kinship domain. In Latin, too, the father's brother's children were types of
siblings (frater or soror patruelis), as against the mother's siblings children
(consobrinus, -a) and the father's sister's (amitinus, -a). A particular affinity
or classificatory grouping of the father's brother's children with one's own
siblings cannot simply be assumed, however. In many patrilineal societies,
conflicts over inheritance and the succession to authority combine to make
the connection with such parallel cousins either ambivalent or marked by
hostility, as has in fact been argued for the Vedic Aryans, where bhratrvya
may have denoted both "parent's (primarily father's) brother's son" and also
"rival, enemy."
I would conclude that the most likely grouping of cousins was of the
Omaha type. That is, consonant with the Omaha avuncular and nepotic
groupings, all parallel cousins may have been classed with siblings, while
cross-cousins were ranked asymmetrically upward and downward in the
Omaha fashion. The maternal cross-cousins would be ranked with the
mother's brother and the mother (or mother's sister), just as the mother's
brother was grouped with the mother's father as awyos. By the same token,
the father's sister's children would be ranked downward with the sister's

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions
28 Ethnology

children, just as the sister's children appear to have been classed downward
with the grandchildren. Despite the negative quality of the evidence, and
the stress on pattern congruity, this third hypothesis has the highest prob?
ability, in part because it fits most logically with the avuncular-nepotic evi?
dence, in part because it appears that Omaha avuncular-nepotic sets are
associated with Omaha cousin terminology about half the time, although
the exact statistics await further research. Moreover, in the putative home?
land area many tribes such as the Kalmyk Mongols have asymmetrical
cousin sets of the Omaha variety. The hypothesis of Omaha cousin termi?
nology is also strengthened by the external evidence on patriliny. But all
these alternatives must be entertained with caution; the PIE problem raises
once again the obvious point that Omaha systems are more fruitfully and
realistically conceptualized in terms of a network of variables rather than of
any one trait such as the cross-cousin terminology.
The foregoing considerations bring up a subject of considerable interest in
light of contemporary theory (Levi-Strauss 1949; Homans and Schneider
1955), namely, the possibility of cross-cousin marriage in the culture of the
PIE speakers. I will not here discuss marriage with the father's "sister's"
daughter because it is so unlikely as the dominant or prescribed form in
conjunction with patriliny (Needham 1962: 101-121). On the other hand,
cross-cousin marriage with the mother's "brother's" daughter is often articu?
lated with a system of patrilineal descent and must be given serious attention.
Such cross-cousin marriage, whether preferred or prescribed, tends to be
associated with certain terminological equations; for example the mother's
brother's daughter may be equated with the wife, the mother's brother with
the wife's father, and so forth. Yet there is no positive evidence that awyos
was used for the wife's father. Nor is there evidence that snwsos was used for
a (woman's) niece, nor that the term for sister's son (nepoit-) was
used for son-in-law, nor that the term for son-in-law (genHr) was used for a
nephew. In fact, there is no positive evidence for any of the groupings that
are typically associated with matrilateral cross-cousin marriage. The negative
picture is summarized below (from the point of view of a male speaker).

i. MoSi^FaBrWi 5. SoWi^WiBrDa
2. MoSiHu^FaBr 6. SoWi ?- BrDa
3. MoBr ?= WiFa 7. SoWiFa ?? WiBr
4. MoBrWi ?> WiMo 8. DaHu ?- SiSo

The results are not entirely surprising because matrilateral cross-cousin


marriage is unlikely in an Omaha system, where the cross-cousins are char?
acteristically ranked out of the speaker's generation and identified with lineal
relatives; for example, the mother's brother's daughter may be referred to as
"mother." To obviate precisely such logical contradiction and status conflict,
most societies with both Omaha cousin terminology and cross-cousin mar?
riage also have special purificatory rituals or terminological subclassifications
of the cross-cousins. In addition to the unilateral alternative just considered,
one might expect preferred or prescribed cross-cousin marriage of the bi-

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions
Proto-Indo-European Kinship 29

lateral type?with either the fathers sister's daughter or the mother's brother's
daughter. But the case for such "simple exchange" is equally poor in terms of
the anticipated groupings. Furthermore, the strength of the set for "in-laws"
and the absence of specific terms for the grandparents argue strongly against
a cross-cousin system (Hocart 1925: 201). The Proto-Indo-European termi?
nology, though classificatory, is not of the cross-cousin type. I am led to con-
clude that preferred or prescribed cross-cousin marriage was absent, and that
recent Marxist theorizing on the subject is untenable.11

Conclusions

Our conclusions refer, not to two-by-two correlations, or to a set of traits


atomistically conceived, but to a highly complex, multidimensional system
of interdependent variables. In brief, PIE culture had patriarchal, patrilocal
families that probably lived in small houses or adjacent huts. Villages were
small, distant, and presumably exogamous. In addition to a large virilateral
affinal set (consonant with patrilocal groups), the terminology at the avun-
cular-nepotic and "brother-in-law" levels was of Lounsbury's Omaha III
type. Both internal and external evidence for patriliny is excellent. The cross-
cousin terminology was most probably asymmetrical Omaha, but cross-
cousin marriage appears most unlikely. The entire mosaic provided by the
linguistic, legal, ethnological, and archeological evidence indicates a system
of the Omaha type, which White (1939: 240) has rightly called "typical of
the patrilineate in its most highly developed form."
As much as postulating a network of terms, concepts, and behavior, this
paper raises or at least sharpens a whole set of questions which could serve
as guidelines to future research. Any one of the major categories below could
lead to an article as substantial as the present one.
1. Homeland Archeology and Kinship Categories. In terms of house type
and settlement patterns the putative homeland falls into two divisions: (1)
from the Central Ukraine to the Danube, and (2) the Cis-Caspian-Caucasian
area. For the third millennium B.C. one needs to know with far greater
qualitative and quantitative precision about the average size of dwellings,
the number and size of hearths, stoves, and religious altars per dwelling, the
evidence for clustering of dwellings around compounds or some other focus,
the number of dwellings per community, the average distance between com?
munities, the relation of communities to hilltops, navigable streams, and
swamps, and the relative importance of hunting, stockraising, and agricul?
ture. Areal archeology in terms of more quantitative information on the
distribution of features and complexes would greatly increase the validity
and reach of certain kinds of inference, and would facilitate probabilistic
statements about patrilocal units, exogamy, and the like. Here is one case
where anthropology must involve archeology to a significant extent.
2. Family and Descent Groups. First, the linguistic, legal, and ethnological
evidence on bilateral groups and bilateral descent needs to be systematically
tabulated and compared, with the objective of specifying the nature and
extent of these institutions in PIE culture and their functional relation to

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions
30 Ethnology

patrilineal structures. Second, even more lexicographic and textual digging


is necessary for certain key nodes in the terminological system that include,
for example: the brother's children of a male speaker; the son's children (to
what extent were they separated from the daughter's children?); the wife's
father and wife's brother's son (can a case be made for their having been
grouped with the wife's brother?); the father's sister and her husband and
children (is there evidence for their having been classed with the sister and
her husband and children?). Provisional answers to some of these questions
have been given above, but more evidence is needed. Third, one must
assemble and evaluate all the evidence and arguments in favor of an alliance
system (with patrilineal descent and matrilateral cross-cousin marriage).
Even more, the British Marxist school of PIE matriliny and matriarchy, and
the Slavist school of PIE as an Australian type system, need to be thoroughly
explored and criticized.
3. Legal Patterns and Archaic Customs. Some student with the requisite
legal, linguistic, and ethnographic training should review the early codes,
particularly the archaic Greek, early Latin, Old Norse, and Hittite, and also
the customary law of the peasantry, particularly of medieval Europe, the
nineteenth century Balkans and Russia, and the semi-nomads of Afghanistan
and India. The legal codes first referred to are often brief and could be
collected in terms of large tables of traits that are functionally most important
within or related to the kinship system, such as wergeld and vengeance obli?
gations, dowry and bride-wealth, bride-capture and wedding rites, patri-
potestality and viripotestality, polygyny and concubinage, and the economic
and political rights of women within the patrilocal family. Such tables & an
analysis would significantly tighten up many of the probabilities regarding
the range of the kindred, the role of bride-wealth, etc.
4. Intensive Study of Other Proto-Systems and of Key Texts. In addition
to the comparative method, reaching back into the prehistory of the phylum,
additional philological and ethnographic analysis needs to be done on certain
of the more-or-less self-sufflcient microcosms represented in ancient law codes
and literary classics, particularly the Celtic, Roman, and Hittite laws and
above all the Rigveda, which has never been thoroughly analyzed in and
for itself by a modern anthropologist. The Homeric epics have profited from
excellent analyses by Glotz, Miller, Finley, and others, and we have today
what amounts to an ideal control over all the relevant evidence in the form of
lexical concordances and the precise records of the classical archeologist.
But now the entire material needs to be reviewed and interpreted in the
light of contemporary kinship theory. The results of such analyses could be
related to subsequent stages of the same culture and back wards to the PIE
system. New insights and probabilities for the proto-system, reconstructed by
the comparative method of linguistics and anthropology, may emerge when
we reconsider the basic evidence in terms of sociolinguistic and humanistic
analyses of the social contexts, patterns of sentiment, and the contrasts
between the terms and the restrictions upon their co-occurrence in various
svntactic frames. Here as everywhere in the study of PIE kinship, a rigorous
division of labor along methodological lines may lead only distorted and

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions
Proto-Indo-European Kinship 31

devious interpretations. Paraphrasing Jakobson, one might conclude, "I am a


student of early Indo-European kinship; all relevant evidence will be enter-
tained."

NOTES

i. Gratitude is expressed to Murray Emeneau and Henry Hoenigswald for introducing


me to Indo-European studies, and to Floyd Lounsbury for suggesting in a personal
communication the Omaha character of PIE. I am particularly indebted to Robbins
Burling, Eric Hamp, and Ward Goodenough for their trenchant and informed review
of this or earlier versions. I wish to thank Wallace Chafe, Calvert Cottrell, Fred Eggan,
David Schneider, Roulon Wells, and Nicholas and Gertrude Vakar for their helpful
comments. I also gratefully acknowledge partial support from the National Defense
Education Act, Title VI, Section 602.
2. PIE phonology was worked out by persons ignorant of Caucasian systems, although
the reverse is not true. For the entire world only PIE, Kabardian, and Abaza (another
Northwest Caucasian language) have been phonemicized with only one vowel (or no
vowel if one chooses to call this "a feature of openness"). The 48 consonants of
Kabardian, and the 67 of Abaza are extremely rich in glottal and velar sounds, thus
paralleling the 12 velar stops and the three (or four) "laryngeals" of PIE (Alien 1956).
Although I have reviewed much of the linguistic and some of the archeological evidence
on PIE origins, the present context only permits me to point at what I feel are the
decisive arguments.
3. Danubian I and Tripolye lacked wagons, as did the "North German Plain" till ca.
1700 B.C. The wheeled vehicle is a peculiarly valuable type of evidence because, like
a kinship system, it consists of a limited number of named, interdependent variables.
4. Otherwise, fifteenth century Hittite evidence includes the following on women:
residence with the wife's father as a frequent alternate (Gurney 1961: 101); the implied
presence of matrilocal households as an alternate (Gurney 1961: 162); equal rights of
the mother over her daughter and the right of a mother to (independently?) expel and
reincorporate her son (J. Friedrich 1959: 77); the independent position of the Hittite
queen; and the dominant position of the Sun Goddess "directing the government of
the king and queen of Hatti" (Gurney 1961: 139). A second class of feminalateral
traits includes a special kinship term for sisters by the same mother (J. Friedrich
1952: 62). The Lycians of southwestern Asia Minor (linguistically close cousins to
the Hittites) were still matrilineal by 600 B.C, according to Herodotus (i: 173); this,
like the Hittite evidence, probably reflects a Pre-PIE substratum.
5. A proto-phoneme is a hypothetical class of sounds reconstructed by comparing forms
in daughter languages that correspond in sound and in meaning. A proto-morph is a
proto-phonemic sequence with a definite meaning. A protomorpheme is a class of one
or more such morphs that contrasts with other classes of morphs. A proto-lexeme is a
word-like unit containing one or more morphemes. The proto-phonemic transcription
of PIE includes: (1) the aspirates are unit proto-phonemes symbolized here by a
digraph, such as /bh/; (2) all the PIE semi-vowels /m, n, 1, r, w, y/ have both syllabic
and non-syllabic allophones, the latter in any position bounded by a vowel, except after
heavy syllables; (3) the consonants with superscripts /K, g/ are palato-velars, and those
with a raised /w/ are labio-velars; (4) /H/ stands for the first laryngeal proto-phoneme,
with two allophones: length after vowels or semi-vowels and "shwa" (a mid-central
vowel) between consonants; (5) /e/ stands for the first laryngeal plus the general
vowel, /a/ for the second laryngeal plus the vowel, and /o/ for the third laryngeal
plus the vowell; (6) /sh/ has been used for /s/ in Sanskrit, /s/ in Lithuanian, etc,
whereas /th/ has been used for the theta of Greek.
I have not followed the convention of starred forms since it should be clear from
the context when I am referring to reconstructions. The base stem or the nominative
singular of the morphemes or terms has generally been cited (the base stem symbolized
by a following hyphen). The cited forms have not, in general, been footnoted, since the

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions
32 Ethnology

understanding is that all the PIE terms have been cross-checked in Delbriick, Pokorny,
Vasmer, Trubachev, and other sources.
6. Goody (1959) has spaded through Morgan (1871) and found that of 137 kinship
terminologies only 13 were reported without giving a term for the wife's father. But,
as he partly suggests, the reportage on affinal terms is so poor in Morgan?and indeed
many other sources?that we often cannot rely on such tests." Goody has performed
a valuable service in at least raising the question of the "sociological reality" of the
so-called "patriarchal family," a question the answer to which is frequently obscured
by the widespread confusion, especially by philologists, of succession, descent, and
inheritance.
7. This is illustrated by the remarkable resemblance of snwsos to South Caucasian (e.g.,
Mingrelian nosa and Laz nusa), as reported in Trubachev (1959: 32); such formal
similarities may reflect early borrowing from Slavic and Iranian through the sort of
bilingualism and intermarriage that still marked relations between the Great Russian
Terek Cossaks and the Northwest Caucasian Chechen in the last century (Tolstoy
1863: 160).
8. Lumping the sister's and daughter's husband is typical of Omaha systems where
the unilocal residential group is dominant, as against those where the unilineal descent
group is dominant and the sister's and father's sister's husband are lumped; I regard
this as a major dimension in the typology of Omaha systems. As White (1939: 567-568)
has cogently written, the family and clan are "rivals, competitors, so to speak, in the
game of relative classification." In this gameful rivalry the family and clan were almost
equal in PIE kinship.
9. Isachenko (1953) argues that nepois originally meant cross-grandnephew, i.e., a
man's sister's grandchildren or a woman's brother's grandchildren. This follows from
his general approach, which is to postulate an Australian type of system with cross-
cousin marriage, and then to explain the evidence accordingly. He leans heavily on
Morgan and Thompson (and on Stalin, Engels, and other non-obvious authorities),
while totally ignoring the relevant social anthropology that has been appearing since
the turn of the century (aside from off-hand allusions to "ill-tempered" attacks by
"bourgeois ethnographers" such as A. L. Kroeber). Isachenko's use of Indo-European
scholarship also leaves much to be desired. Nonetheless, he does raise some interesting
points. The most important (Isachenko 1953: 60-61) concerns the suffix -ter, -tor,
analytically distinguishable from the agentive suffix. Isachenko claims that -ter/-tor
appears in the kinship terms for those relatives who would be grouped together under
matrilocal residence and cross-cousin marriage classes: mother, father, brother, daughter,
and husband's brother's wife. Aside from certain linguistic problems, the argument
advanced does not explain why brother is included in the matrilocal group, while the
sister, sister's husband, and daughter's husband (genHr), are not. The husband's
brother's wife (ynHteir) could equally well belong in the -ter "in-group" under a
system of patrilocal residence and patrilineal descent. Isachenko was more or less
followed by Trubachev, with many specific reservations and, of course, a more massive
and valuable documentation.
10. About the PIE cousins, the opinion of Delbriick (1889: 506) was: "Hierbei hort
die Gemeinsamkeit der Bezeichnungen vollig auf. Ich kann daher nur einen Uberblick
iiber die Tatsachen der Einzelsprachen geben . . . darf man wohl schliessen, dass
Vettern und Cousinen in der Urzeit sich als Briider und Schwester bezeichnet haben."
11. Notwithstanding Thompson (1949: 80-83), who illustrates the aprioristic use of
negative evidence in conformity with Marxist dogma perhaps nowhere more revealingly
than in his arguments about the Greek sibling terms, adelphos, -e:, which, because they
are clearly analyzable as "from the same womb," are cited as strong evidence for
matriliny. Actually, they are late formations, and clearly former modifiers which have
assumed nominal functions. In earlier times an adjective was used with the sibling
term in order to set off monomatric siblings because, for example, the incest taboos
were different for them, as we know from Greek legal history. Homeric Greek had
additional terms for the children of sisters. With the decline of extended patrilineal
categories and the increasing importance of immediate relatives and the bilateral

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions
Proto-Indo-European Kinship 33

family, the former adjectives came to be used for siblings, and the original sibling
terms survived only in compounds. In sum, the transition was from phra:te:r (clan
brother), to adelphos phra:te:r (monomatric brother as against other kinds of
"brother"), to adelphos (true brother).
Miller (1952-53), whom I take to be another member of the British school of Greek
matriliny, has constructed an ingenious argument that (Pre-) Homeric Greek kinship
operated with three matrilines with the men circulating and the women marrying their
father's sister's son. Aside from matriliny, Miller's key assumption is that recurring
proper names indicate a kinship group, that is, marriage categories. He is able to
account very well for the otherwise anomalous ten-generational genealogy of Odysseus.
However, the data he cites?and particularly the Autolykos category in Odysseus'
genealogy?can be accounted for just as elegantly by assuming matrilateral cross-cousin
marriage and three intermarrying patrilines. The full implications of this exciting
internal evidence for a patrilineal alliance system in Pre-Homeric Greece certainly
requires further study. It may be possible to demonstrate that the proto-Greeks and
Homeric Greeks combined definite feminalateral groups with patrilineal descent and
descent groups. Rose (1925: 121-124), writing of the early Romans, makes a weak case
for restricted exchange between two or at least a small number of patriclans.
On the other hand, a possibly matrilineal or even "matriarchal" system for Pre-PIE
cannot be dogmatically excluded, if only because of the time lapse between the second
half of the third millennium B.C, when Greek and Italo-Celtic split off and began
moving south and west while Indo-Iranian moved eastwards, and the first evidence of
Homer and the Vedas about a thousand years later. If we go back to Proto-Indo-
Hittite the time span reaches almost two millennia, during which a matri-system could
have passed away. Lounsbury has recently elucidated the close logical affinity between
Omaha and Crow systems, and Eggan has reported very rapid transitions from one to
the other.
Without the external evidence for various patterns, the PIE nomenclature could be
accounted for by the minimal assumptions of bilateral descent, extended patrilocal
families, and marriage between distant, exogamous communities such that the couple
was sharply separated from the wife's family.
In the same connection, be it noted that we do not need to assume uniformity in the
Omaha kinship of the dialects and regions of the PIE people. In the kinship as in the
phonology of Indo-European, the comparative method strongly suggests that the
parent language was composed of various dialects. Sousberghe and Robles Uribe (1962),
doing synchronic work on contemporary Tzeltal, have shown (implicitly) how different
parts of an Omaha network can be differently realized in different communities of the
same tribe.
For an attempt to debunk evolutionary interpretations of Omaha systems, on the
basis of a statistically oriented and almost entirely synchronic analysis of a single system,
see Tax (1955).

BIBLIOGRAPHY

Abaev, V. I. 1949. Osetinskij jazyk i folklor. Marx-Lenin Institut.


Alien, W. S. 1956. Structure and System in the Abaza Verbal Complex. Transactions
of the Philological Society 127-176. Oxford.
Asheri, D. 1963. Laws of Inheritance, Distribution of Land, and Political Institutes in
Ancient Greece. Historia 12: 1-21.
Austin, W. 1942. Is Armenian an Anatolian language? 18: 22-25.
Benveniste, E. 1933. Le participe indo-europeen en -mno-. Bulletin de la Societe de
Linguistique de Paris 34: 5-21.
- 1934. Un nom indoeuropeen de la "femme." Bulletin de la Societe de Lin?
guistique de Paris 35: 104-106.
Bloomfield, L. 1934. Language. Chicago.
Brugmann, K. 1903. Beitrage zur griechischen, germanischen und slavischen Wort-
forschung. Indogermanische Forschungen 15: 93-97.

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions
34 Ethnology

1904. Griech. vivc, vioc, vlvoc, und ai. sunus got. sunus. Indogermanische
Forschungen 17: 483-491
- 1907. Nuos, nurus, snusa und die griechischen und italienischen Femina auf -os.
Indogermanische Forschungen 21: 315-322.
Buck, C. D. 1949. A Dictionary of Selected Synonyms in the Principal Indo-European
Languages: A Contribution to the History of Ideas. Chicago.
Childe, V. G. 1926. The Aryans. London.
- 1958. The Dawn of European Civilization. New York.
Cottrcll, C. 1965. Changes in Crow Kinship. M.A. thesis, University of Chicago.
Crossland, R. A. 1957. Indo-European Origins: The Linguistic Evidence. Past and
Present 12: 16-46.
Delbriick, B. 1889. Die Indogermanischen Verwandtschaftsnamen: Ein Beitrag zur
vergleichenden Altertumskunde. Leipzig: Abhandlungen der Koniglichen Sachs-
ischen Gesellschaft der Wissenschaften 11: v, 380-606.
Dittenberger, V. 1885. Die cleusinischen Keryken. Hermes XX: 1-40.
Dumezil, G. 1940. Series etymologiques armeniens. Bulletin de la Societe de Lin-
guistique de Paris 41: 68-69.
Eggan, F. 1934. The Maya Kinship System and Cross-cousin Marriage. American
Anthropologist 36: 188-203.
- 1937. Historical Changes in the Choctaw Kinship System. American Anthro?
pologist 39: 34-53.
- 1955. The Social Anthropology of North American Tribes. Chicago.
Ernout, A., and A. Meillet 1951. Dictionnaire etymologique de la langue latine. 3d. edit
3V. Paris.
Friedrich, J. 1952. Hethitisches Worterbuch. Heidelberg.
- 1959. Die hethitischen Gesetze. Leiden.
Friedrich, P. 1963. An Evolutionary Sketch of Russian Kinship. Symposium on Lan?
guage and Culture: Proceedings of the 1962 Annual Spring Meeting of the
American Ethnological Society, ed. W. L. Chafe and V. E. Garfield, pp. 1-26.
Seattle.
- 1964. Semantic Structure and Social Structure: An Instance from Russian.
Explorations in Cultural Anthropology: Essays in Honor of George Peter
Murdock, ed. W. H. Goodenough, pp. 131-166. New York.
Ghurye, G. S. 1955. Family and Kin in Indo-European Culture. Bombay.
Gimbutas, M. 1956. The Prehistory of Eastern Europe, pt. 1. Cambridge, Mass.
- 1960. From the Neolithic to the Iron Age in the Region between the Vistula
and the Middle Dnieper Rivers: A Survey. International Journal of Slavic Lin?
guistics and Poetics 3: 1-12.
- 1963. The Indo-Europeans: Archeological Problems. American Anthropologist
65: 815-837.
Goodenough, W. H. 1956. Componential Analysis and the Analysis of Meaning. Lan?
guage 32: 195-216.
- 1963. Review of Matrilineal Kinship, ed. D. M. Schneider and K. Gough
American Anthropologist 65: 923-928.
Goody, J. 1959. Indo-European society. Past and Present 16: 88-92.
Grammont, M. 1924. L'interversion. Streitberg-Festgabe, pp. 111-118. Leipzig.
Griffith, R. T. H. 1897. The Hymns of the Rigveda. Benares.
Gummere, F. B. 1901. The Sister's Son. An English Miscellany (Presented to F. J.
Furnivall), pp. 133-150. Oxford.
Gurney, O. R. 1961. The Hittites (revised). Baltimore.
Hignett, C. 1952. A History of the Athenian Constitution to the End of the Fifth Cen?
tury B.C. Oxford.
Hocart, A. M. 1925. The Cousin in Vedic Ritual. Indian Antiquary 54: 16-18.
- 1928. The Indo-European Kinship System. Ceylon Journal of Science 1: 179-204.
Homans, G. C, and D. M. Schneider. 1955. Marriage, Authority, and Final Causes: A
Study of Unilateral Cross-Cousin Marriage. Glencoe.
Homer. 1960-63. The Iliad (2V.); The Odyssey (2V.). A. J. Murray. Cambridge, Mass.

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions
Proto-Indo-European Kinship 35

Isachenko, A. B. 1953. Indoevropejskaja i slavjanskaja tcrminologija rodstva v svete


Marksistskogo jazykoznanija. Slavia 22: 44-80.
Jakobson, R. 1936. Sur la theorie des affinites phonologiques entre les langues. Ap-
pendixed to Principes de phonologie, by N. Trubetskoj. Paris.
Kiparsky, V. 1942. Der Schwiegersohn als "Bekannte." Neuphilologische Mitteilungen
^43:113-120.
Korosec, V. 1932. Ehe in Hatti. Reallexikon der Assyriologie 2: 293-296.
Kovalevsky, M. 1891. Modern Customs and Ancient Laws of Russia. London.
Kuipers, A. 1960. Phoneme and Morpheme in Kabardian. The Hague.
Krader, L. 1955. Principles and Structures in the Organization of the Asiatic Steppe-
Pastoralists. Southwestern Journal of Anthropology 11: 67-92.
Kurylowicz, J. 1935. fitudes indoeuropeens. Cracow.
Lavrovskij, P. 1869. Korennoje znachenije v nazvanijakh rodstva u Slavjan. Zapiski
Imperatorskogo Akademii Nauk 12: 1-120. St. Petersburg.
Levi-Strauss, C. 1949. Les structures elementaires de la parente. Paris.
Lounsbury, F. G. 1964. A Formal Account of the Crow- and Omaha-type Kinship
Terminologies. Explorations in Cultural Anthropology: Essays in Honor of
George Peter Murdock, ed. W. H. Goodenough, pp. 351-393. New York.
Mayrhofer, M. 1952. Gibt es ein idg. sor "Frau"? Studien zur Indogermanischen
Grundsprache 4: 32-33.
Meillet, A. 1937. Introduction a l'etude comparative des langues indo-europeenes. Paris.
Mikkola, J. J. 1908-09. Zur slavischen Etymologie. Indogermanische Forschungen 23:
120-127.
Miklosich, F. 1886. Etymologisches Worterbuch der slavischen Sprachen. Wien.
Miller, M. 1952. Greek Kinship Terminology. Journal of Hellenic Studies 73: 46-52.
Mongait, L. 1961. Archeology of the U.S.S.R. Trans. M. W. Thompson. London.
Murdock, G. P. 1949. Social Structure. New York.
- 1957. World Ethnographic Sample. American Anthropologist 59: 664-687.
Needham, R. 1962. Structure and Sentiment: A Test Case in Social Anthropology
Chicago.
Piggott, S. 1950. Prehistoric India. London.
Pokorny, J. 1959. Indogermanisches etymologisches Worterbuch, v.i. Bern and
Miinchen.
Radcliffe-Brown, A. R. 1952. Structure and Function in Primitive Society. Glencoe.
Rivers, W. H. R. 1914. Kinship and Social Organization. London.
Rose, H. J. 1925. Patricians and Plebians of Rome. Journal of Roman Studies 12: 120-
124.
Schrader, O. 1904. Ueber Bezeichnungen fur die Heiratsverwandtschaft bei den Indo?
germanischen Volkern. Indogermanische Forschungen 17: 11-36.
Sousberghe, L. de, and C Robles Uribe. 1962. La parente chez les indiens tzeltal.
L'Homme 2: iii, 102-120.
Stankicwicz, E. 1962. The Etymology of Common Slavic vUnokU/vUnukU. Slavic and
East European Journal 6: 28-33.
Tax, S. 1955. The Social Organization of the Fox Indians. Social Organization of
North American Tribes, ed. F. Eggan, pp. 243-282. Chicago.
Thieme, P. 1953. Die Heimat der indogermanischen Gemeinsprache. Abhandlungen der
Geistes- und Sozialwissenschaftlichen Klasse, Akademie der Wissenschaften und
der Literatur 535-610. Wiesbaden.
Thompson, G. 1949. Studies in Ancient Greek Society: The Prehistoric Aegean. London.
Tolstoy, L. 1863. Kazaki. Sobranije Sochinenij 3 (1951 edition). Moscow.
Trubachev, O. 1959. Istorija slavjanskikh terminov rodstva i nekotorykh terminov
obshchestvennogo stroja. Moscow.
Trubetskoj, N. 1939. Gedanken uber das Indogermanenproblem. Acta Linguistica 1: 81-

Vasmer, M. 1950-58. Russisches etymologisches Worterbuch. 3V. Heidelberg.


Verner, K. 1877. E^ne Ausnahme der ersten Lautverschiebung. Zeitschrift fur Ver-
gleichende Sprachforschung 3: 97-130.

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions
36 Ethnology

Vey, M. 1931. Slave st- provenant de i.-e. *pt-. Bulletin de la Societe de Linguistique de
Paris 32: 65-67.
Vinogradoff, P. 1900. Geschlecht und Verwandschaft im altnorwegischen Rechte.
Zeistchrift fiir Sozial- und Wirthschaftsgeschichte 7: 1-43.
White, L. A. 1939. A Problem in Kinship Terminology. American Anthropologist 16:
569-570.

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 19:14:21 PM


All use subject to JSTOR Terms and Conditions

You might also like