Using queries on the AustKin database to find kinship patterns and

advertisement
Using queries on the AustKin database to
find kinship patterns and systems1
Patrick McConvell and Rachel Hendery
Society for the Athropological Sciences/Society for Cross-cultural
research conference, 19th February 2010
1. Introduction
1.1 Aim and historical background
The aim of this paper is to show how queries on a large database of Australian
indigenous kinship terminology can capture sets of languages representing types
of pattern and system within typologies proposed for kinship both on a world
scale and in Australia. It has often been noted that compared to conceivable
variation in such patterns and systems, actual variation found is quite narrow.
From the start of systematic kinship comparison with Morgan, ethnologists have
proposed limited numbers of types defined primarily by equations (polysemies)
of kinterms. One approach was that of Murdock, later systematized in the
Standard Cross-Cultural Codes which proposed patterns of polysemy in various
sub-sets of kin (generations, in-laws). Other approaches attempted to combine
patterns in different subsets into system types, like the Australianist scheme of
Radcliffe-Brown. This paper tests how both these kinds of atpproaches may be
captured with database queries., including employing parsimony to reduce the
queries needed to circumscribe each to a minimum. A number of language
groups are shown to represent ‘mixture’ or ‘overlap’ of different types. This does
not necessarily indicate that the search criteria are flawed but can be a
significant result in itself.
1.2 The AustKin Project
The project "Tracing change in family and social organisation in Indigenous
Australia, using evidence from language"2 (henceforth referred to as "AustKin")
1 This research was supported under the Australian Research Council’s Discovery Projects
funding scheme (project number DP0878556); the Australian National University (ANU), and
the Centre National de la Recherche Scientifique (CNRS) through the Centre de Recherche et
de Documentation sur l’Océanie (CREDO). The software for this project was developed by
Laurent Dousset of CREDO and uses a geo-spatial-interface developed by the Research School
of Humanities (RSH) at ANU using the AUSTLANG
(http://austlang.aiatsis.gov.au/disclaimer.php) coordinates and language list developed by
Kazuko Obata of the Australian Institute of Aboriginal and Torres Strait Islander Studies
(AIATSIS).
2 The official name of the project is “Tracing change in family and social organisation in
Indigenous Australia, using evidence from language”. The chief investigators are Harold Koch
and Ian Keen at the Australian National University. Laurent Dousset, of CREDO in France, is a
partner investigator. Patrick McConvell is a Senior Research Fellow working on the project,
is a large collaborative database project, collating and comparing kinship
terminology from languages across Australia. As of January 2010 the database
contained around 68,000 words in 487 word lists spanning 257 language
varieties (as defined by AUSTLANG3). Each word list in the database consists of
kinship terms from a single source (grammar, dictionary, etc) for a single
language or dialect, along with the original definitions from that source, and
standardised glosses in the form used by anthropologists, e.g. MFZ for 'mother's
father's sister'. The database can be searched using simple and complex queries,
and the results of these queries can be displayed on a map of Australia. The
project is described in more detail in Dousset et al (unpublished).
2. Capturing patterns of polysemy
2.1 Single equation queries
One of the types of queries that can be run on the AustKin database yields
languages in which certain types of equations are found. For example, entering
the query "MM=FFZ" yields the map shown in figure 1, where the blue arrows
represent all languages in the database in which the same term is listed for the
gloss "mother's mother" as for the gloss "father's father's sister".This,
incidentally, is one of the equations which we use to define ‘Kariera’ systems as
proposed by Radcliffe-Brown (see further later).
and Rachel Hendery is its Research Assistant. Jeanie Bell (Batchelor Institute of Indigenous
Tertiary Education), Claire Bowern (Yale University), and Barry Alpher (of Washington D.C.)
are further participants. The project is scheduled for completion at the end of 2010. The
project’s website is at the following address: http://austkin.pacific-credo.fr. Note however
that the access to files and databases created under this project are for now restricted. Wider
public access is planned by the end of 2010.
3 http://austlang.aiatsis.gov.au/disclaimer.php
Figure 1: MM=FFZ
2.2 Combining queries
The results of two such queries can be combined and overlaid on the same map
for comparison (e.g. see figure 2). As well as mapping polysemies in this way, we
can map inequalities, such as languages in which mother's mother is not the
same term as father's mother. The latter query would be a useful addition to a
polysemy search such as the one in figure 1, since a simple search for MM=FFZ
will also return languages which only have one term for the grandparental
generation, and languages which have only one term for any female member of
the grandparental generation. In order to rule these cases out, combining the
polysemy search with one or more inequalities is necessary. Excluding
languages which have MM=FM (a characteristic polysemy in the Aluridja
system) is an additional requirement for defining Kariera, to be discussed later.
The map in figure 2 shows on the same map the searches MM=FFZ (blue arrows)
and MM≠FM (red arrows). Therefore the only languages in the database for
which the polysemy MM=FFZ is significant (i.e. not simply an artifact of a term
with much wider semantics) are the languages marked with both a red and a
blue arrow, i.e. Kariyarra, Martuthunira, Nyamal, Arabana, Wilyakali, Buandig,
Yuwaalayaay, Wangaaypuwan, Kok Kaper, Umpila, Ayabadhu, Oykangand, Ogh
Undjan
Figure 2: MM=FFZ≠FM
3. Standard Cross-Cultural Codes
3.1 The Omaha cousin pattern
AustKin equation queries can be used to capture the kinship patterns recognized
in the Standard Cross-Cultural Codes (Murdock 1970, White et al. 2009) and the
Ethnographic Atlas. This can be done by combining a series of polysemies and
inequalities. We will illustrate this here for the SCCC definition of the Omaha
cousin pattern. The definition of the Omaha pattern in Murdock (1970:165) is as
follows:
The children of a mother's brother and of a father's sister are terminologically
distinguished from siblings, parallel cousins, and each other, but are not
designated by special terms. Instead, a mother's brother's children are
terminologically equated with relatives of an ascending generation, normally with
mother's brother and mother and a father's sister's children are equated with
relatives of adescending generation, normally with a man's sisters' children and a
woman's own children.
This can be captured by the following series of equations:
1. MBS≠B ("The children of a mother's brother and of a father's sister are
terminologically distinguished from siblings [...]")
2. MBS≠FBS ("[…] parallel cousins [...]")
3. MBS≠FZS ("[…] and each other […]")
4. MBS=MB ("[…] a mother's brother's children are terminologically equated
with relatives of an ascending generation, normally with mother's brother [...]"
5. MBD = M ("[…] and mother" [...])
6. FZD = ZD, 7. FZS= ZS ("[…] normally with a man's sister's children […]")
8. FZD=D, 9. FZS=S ("[...]and a woman's own children.")
For completeness, one would need to repeat the inequalities (1—3 above) for
MBD, FZS and FZD. It may be unlikely that a kinship system would have e.g. a
cross-cousin/parallel-cousin distinction for MBS but not MBD, but in practice the
database will not contain terms for all cross cousins and nephews/nieces for all
the languages, so building some redundancy in by including as many equations
as possible in the search will ensure that languages with incomplete word lists
are not missed.
Searching for all the above logically possible polysemies and inequalities and
displaying them on the same map is, however, rather impractical. Fortunately it
is not strictly necessary, as some elements of Murdock's definition are almost
ubiquitous in Australia: a distinction between parallel cousins and cross-cousins;
and the distinction between cross-cousins and siblings.
3.2 Cross/parallel cousin/sibling neutralization
The few languages currently in the Austkin database which have a sibling—
cross-cousin polysemy are Kija, Arrernte4, Ngaatjatjara, Ngaanyatjarra, Kokatha,
Gubbi Gubbi, Dunghutti and Kattang (Dousset 2010). Those with cross—parallel
cousin neutralization are Arrernte, Buandig, Wadi Wadi, Dhauwurd Wurrung
and Gunnai. The results of the queries used to generate these two lists are
shown (on the one map) in figure 3. The languages which have sibling=crosscousin apparently have parallel cousin= cross-cousin also (although this
information may not be currently recorded in our sources) but it seem as if the
opposite implication does not hold, since it seems parallel cousins in the
Victorian languages mentioned do not have the same term as siiblings unlike in
most of Australia..
4 The word with sibling—cross-cousin polysemy in Arrernte is arrwempe, a relatively recent
loan-word from Western Desert. According to Henderson & Dobson (1994), each of these two
senses (cross-cousin, sibling) is only used by "some speakers". (There is no indication of
whether or not any one group of speakers uses the term for both senses.) Arrernte also has
other terms for sibling and cross-cousin that do not have this polysemy.
Figure 3: languages in the database with cross-cousin—sibling or cross—
parallel polysemies
3.3 Narrowing down to core Omaha
This leaves us with the relatively simple search for the equation of MBCh with M
and MB; and the equation of FZCh with mZCh and fCh. The results are shown in
figure 4.
Figure 4: MBCh = M, MB; FZCh = mZCh, fCh (Omaha)
Based on the results from figure 3, we exclude Gureng Gureng, Dunghutti,
Kattang, and Buandig. The remaining languages: Wirangu, Guugu Yimidhirr,
Warumungu, Marra, Nunggubuyu, Ayabadhu and Anindilyakwa are the
languages in the AustKin database that then fit Murdock's definition of an Omaha
cousin pattern.
This result captures some of the key areas which are known to have Omaha
skewing in Australia: South –east Arnhem Land,, Cape York Peninsula and
Central Australia (represented by Warumungu). Another well-know Omaha
group in the North Kimberleys is not yet entered in the database. Other Central
Australian groups which do have Omaha (Gurindji, Arrernte) are not here
because this feature is not entered in the source sets. This is because Omaha
skewing is a contextual overlay rather than part of the basic kinship system
(McConvell and Alpher 2003, McConvell 2010) so not consistently reported in
sources.
The Standard Cross Cultural Codes (SCCC) simplifies Murdocks’s definition of
Omaha by omitting the explicit condition that cross-cousins are distinguished
from parallel cousins and siblings. This means that Gureng Gureng, Dunghutti,
Kattang and Buandig are Omaha languages according to the SCCC definition, but
not according to Murdock's.
Only some of the languages captured as having Omaha patterns have the full set
that one might expect, that is for both genders of cousin and both sides
(patrilateral and matrilateral cousin). This may be because of missing data in
some cases but in other the Omaha pattern is only partial. Where we can
establish that this is not a question of missing data, then this is important and
could lead to examination of just what partial varieties of this pattern (and other
patterns we will be discussing) exist.
The fact that we don't have a single case of both MB-raising and FZ-lowering in
the same language in our search results is due in part to missing data, but the
phenomenon of different patterns of reciprocals of skewed cousin terms is also a
factor. Often more than one reciprocal pattern in a single language (eg in
Arrernte, Denham and White 2005, McConvell 2010). In Kattang and Dunghutti
in central coastal New South Wales, MBD is not called M 5
It is possible that there are implicational relations within the partial patterns.
From current limited data the implicational hierarchy: if MBD=M, then MBS=MB
(but not the other way around!)seems to work. If this is a valid or plausible
generalization it could be used to estimate values of gaps in the data. However at
our present stage of knowledge, the putative generalization might simply be a
matter of what sort of data we are likely to lack: if a set has a term for MBD it
always gives the term for MBS as well, but not vice versa..
This example discussed above of Omaha skewing demonstrates how SCCC
5
According to our sources there is an equation MBD=FZ in Dunghutti. This seems to be
connected to Omaha patterns but is unusual. In Kattang, all cross-cousins are kamiin. This is a
widespread Pama-Nyungan term originally meaning ‘mother’s mother’ (McConvell 2008) but
in a number ofareas including this one switching to ‘father’s mother’. FM is often a spouse
term and could be used for ‘cross-cousin’ in the context of cross-cousin marriage.
patterns can be captured using a combination of polysemy and inequality
queries. For some patterns, a search of the AustKin database returns many
results; others are relatively rare. Most importantly, there are some patterns of
kinship term polysemy common in Australia that are not included in the SCCC
list of patterns.
3.4 Australian patterns not found in SCCC
The following patterns (among others) occur in Australia but are not included in
the SCCC list.
1. Cross-cousin-=spouse equations are found fairly commonly in Australia
where there is cross-cousin marriage., in the ‘Kariera’ areas of the Pilbara
and Cape York Peninaula, Western Desert etc. This is a classic Dravidian
pattern but not explicit in Murdock’s typology, where ‘spouse’ is not
specifically dealt with. There is something similar however in the Siblingin-law set, Pattern E, where siblings-in-law are equated with crosscousins, which also occurs in Australia.
2. A distinction between mother's younger and elder brother, but no
distinction between father's younger and elder brother. (found in e.g.
Arrernte, Ngayawung, Yaraldi). The SCCC codes allow only for the
opposite situation (a distinction between young and elder on the father's
side, but not the mother's).
3. One term for spouse's sibling (i.e. HB=HZ=WB=WZ), but distinct terms for
each of man's sibling's spouse, woman's brother's wife and woman's
sister's husband (e.g. Arabana, among others).
4. Equivalence between cross-cousin's child and own child or (man's)
sister's child. This is simply not a type of relationship covered in the
Murdock typology, but it became important to determine whether
opposite sex or same sex cross-cousin’s child was classified as own, as a
distinction betweem Dravidian and Iroquois OK was refined in later work
(Lounsbury 1964, Scheffler 1971, Kronenfeld 2004). Dravidian and
Iroquois in this sense ar both found in Australia, in different regions, and
it is a key distinction but often overlooked so the relevant data is not in a
lot of our sources.
5. Equivalence of grandparent and sibling (or sometimes specifically elder
sibling) eg in Ngatjumaya, Wangaaypuwan. This is a common type of
‘alternate generation’ equivalence in Australia. This is touched on later as
it overlaps with Kariera patterns in some areas.
6. The Kariera grandparent configuration: MM=FFZ, FF=MMB, MF=FMB,
FM=MFZ. (to be further discussed below).
Such gaps can be remedied by adding further categories to the relevant SCCC
lists. However the fact that SCCC patterns are predicated on predefined sets
makes it necessary to put the equation into one or other of the sets where the
equation crosses between them. For instance, equations between grandparents
and siblings (5 above)can be defined in relation to either siblings or
grandparents or, redundantly, both. There should be some non-arbitrary way of
making such a choice.
One difficulty with the SCCC codes that is harder to address is the co-existence of
multiple patterns within the same language. Fundamentally this is a conceptual
problem: we tend to think of a language as fitting one particular pattern for any
kinship relationship, e.g. we might describe a language as having a "bifurcate
pattern for grandparental terms". In fact, however, many languages in Australia
(and elsewhere too) have partially synonymous alternative terms for some
kinship relationships, each of which can follow a different pattern This issue is
further discussed in the section on ‘Overlaps’ below
4. Kinship typology in Australianist Anthropology
4.1 Radcliffe-Brown
Murdock’s typological scheme and those arising from it were built on ‘patterns’
each referring exclusively to a particular predeifined subset of kin types e.g
siblings, cousins, granparents. These modules were not put together into bigger
‘systems’ nor were the correlations between patterns within the different
modules explored to any extent [6
The Australianist approach on the other hand stemmed from an earlier
tradition of a typology of ‘systems’, going back to Morgan, who constructed
broad types and in his later work, arranged into a linear evolutionary scheme.
Radcliffe-Brown abandoned the evolutionist and diffusionist diachronic
frameworks early in his career and when working in Australia produced a
general typology of Australian systems which is still largely in use today. This
combined several features of the kinship terminology into ideal types, to which
typical marriage patterns were also added. Obviously there were local variations
which did not exactly fit these types,.The marriage systems did fit reasonably
well with the main kinship types but there were many groups in which this was
not the case. As argued further below, a preferable approach would be to classify
kinship systems and marriage type separately and then study the correlations
and mismatches.7
6
There was a large amount of research based on statistical correlations beween different
aspects of the SCCC including between kinship patterns and other socio-cultural patterns,
including in the journal Cross-Cultural Resarch.This continues today to a limited extent eg one
can now run cross-tabulations on the Ethnographic Atlas website hosted by University of
Kent Canterbury.
7
In an older tradition of anthropology such mismatches were also explained in terms of a
time-lag between kinship and marriage, where one system could be ‘survival’ reflecting
earlier patterns. This type of diachronic theorizing was attached trenchantly by RadcliffeBrowne and Malinowski and virtually disappeared form Australianist anthropology –
unfortunately as while some of the hypotheses proposed were absurd, some were not and
historical linguistics can sometimes add independent evidence for earlier states of kiship and
marriage.
His major work (1930-31) provided a template for much thinking about overall
variation in Australian kinship. The two main types proposed , Kariera and
Aranda, were said (quite justifiably in most cases) to be associated with two
types of marriage: bilateral cross-cousin marriage (Kariera) and bilateral second
cousin marrage (man with a MMBDD/FFZSD). These two marriage systems
were coupled with two distinctive kinship systems: one of the key distinctive
features was the patterning of grandparental terms.
Kariera
FM=MFZ
FF=MMB
MM=FFZ
MF=FMB
Some systems also neutralise gender so that FM=MF etc. Thie is (correctly we
think) not thought of as the basis of an important typological distinction between
systems.
Aranda:
FM≠MFZ
FF≠MMB
MM≠FFZ
MF≠FMB
So in an Aranda system all the grandparental equations stated for Kariera above
are absent. Instead of two grandparent terms (or four if there is a gender
distinction as In Kariera, in the Aranda system there are four terms if gender is
not differentiated (or eight if gender is differentiated in all).
Radcliffe-Brown also had several other types which will not be discussed here.
4.2 Contributions after Radcliffe-Brown
Elkin, who succeeded Radcliffe-Brown in the chair of anthropology at Sydney
refined the system. He added in particular more detail of the asymmetrical
variety of marriage and kinship terminologies : MBD marriage for a man, with
MBD and FZD having different terms, and Elkin predicted, three grandparent
terms with MF=FMB but FF and MMB distinct.
Scheffler returned to the task of an Australia-wide kinship typology after having
worked with Lounsbury’s extensionist ‘reduction rules’ formalism. This
approacjhprovides the ability to make formal generalizations over a wider set of
kinship terms in single languages, and comparatively across languages, and is
further compared to what is being proposed in this paper in a later section,
Scheffler’s book on Australian kinship is a work of insight and careful
scholarship which amends Radcliffe-Brown’s and Elkins schemes and
reinterprets them in terms of reduction rules, and another concept of superclasses.
There are certainly ways in which the Australianist typology can be further
integrated with the more general typologies that have been proposed.
Kariera , for instance, which we are about to discuss in detail can alternatively
be seen as a variety of Dravidian, with Dravidian types of equation extended to
the grandparental generation (McConvell and Keen in press)8 In Dravidian
systems, mother’s brother is equivalent to father’s sister’s husband and spouse’s
father. If these equations are extended to linking relatives in the grandparent
generation, it follows that :
• MMB = MFZH = MHF = FF
• FMB = FFZH = FWF = MF
And if father’s sister is equivalent to mother’s brother’s wife and spouse’s
mother then:
• MFZ = MMBW = MHM = FM
• FFZ = FMBW = FHM =MM
4.3 Capturing Kariera in AustKin
As noted above the major division of kinship systems in Australia had as its
primary division Kariera and Aranda systems.
Queries on AustKincan be combined in sequences to capture types of kinship
systems recognized in the Australianist tradition (e.g. ‘Kariera’, ‘Aranda’). We
look at the Kariera system as an example.
Radcliffe-Brown proposed quite a long list of criteria for ‘Kariera (reviewed
below) , not in such a rigorous way as Murdock and others later set out kinship
patterns. It is not completely clear which of these criteria need to be satisfied in
order for a system to be deemed ‘Kariera’. One option would be to talk of ‘full
Kariera’ if all or most of the criteria are satisfied, and ‘partial Kariera’ (or X%
Kariera) if only some are. Such an approach does not single out which of the
criteria are most important in defining the class.
8
The term ‘Kariera’ has been used more generally in ethnological work in recent years by
Allen [refs] and Hage [refs]. Hage’s usage does not conform to how the term is defined by
Radcliffe- Brown nor how it is used by Australian ists. The typical pattern of Kariera
grandparent terms which is discussed below in section ?? is not included in the criteria but
alternate generation equivalence is. Alternate generation equivalenc is also stressed by
Radcliffe-Brown, Scheffler and others for Australia but is quite distinct from the criteria for
AAKariera. Some Kariera systems have AGE, some do not. This deviation may be traceable fto
Radcliffe-Borwn having included the fact that grandpatents and grandchildren use terms
reciprocally between them as part of the description of Kariera. However as argued below
this is not really one of the key criteria of Kariera and many Kariera systems do not have it.
In practical terms for AustKIn, such an approach raises problems. It magnifies
the issues mentioned above (large number of equations, co-existence of multiple
patterns in a single language, incomplete data). In order to overcome these
problems, we need to pare down the number of equations used to capture a
given type
More useful would be a method in which certain features were considered
‘indicators’ or ‘diagnostic equations’, that is their presence usually or always
implies the presence of other Kariera features. It is then an empirical matter to
find out, using AustKin searches for instance, whether chosen diagnostic
features actually do predict other important features of Kariera.
Radcliffe-Brown's definition of the Kariera kinship type is as follows:9
In the second ascending generation the grandparents and their brothers and sisters,
and all other relatives are divided into the following four kinds: 1. Father's father; with his brothers, husbands of the father's mother's sister's and
the brothers of the mother's mother.
2. Father's mother; with her sisters, wives of the father's father's brothers, and
sisters of the mother's father.
3. Mother's father; with his brothers, husbands of the mother's mother's sisters,
and brothers of the father's mother.
4. Mother's mother, with her sisters, wives of the mother's father's brothers, and
sisters of the father's father.
Each of these groups of relatives is denoted by one term of relationship.
[…]
The terms for grandparents are used reciprocally for grandchildren. [...…]
In the first ascending (parents') generation a man distinguishes four kinds of relatives.
'Father' including own father, father's brother, mother's sister's husband, father's
father's brother's son, mother's mother's brother's son, etc.
'Mother' including own mother, mother's sister, father's brother's wife, mother's
mother's sister's daughter, etc.
'Mother's brother' including the brother of any woman called 'mother' and the
husband of the sister of any man called 'father'.
'Father's sister' including the sister of any man called 'father' and the wife of any man
called 'mother's brother'.
In his own generation a man has distinct terms for older and younger brothers and for
older and younger sisters, the actual relation in age to himself being the determining
factor in the use of the terms. He has names for male cross-cousins and for female
cross-cousins.
9 The elided material in the following quotation is only further clarification or exemplification
of the relevant principles.
In the first descending (children's) generation a man again distinguishes only four
kinds of relatives 'son', 'daughter' 'sister's son' and 'sister's daughter'.
(Radcliffe-Brown 1930-31:47-48)
The number of equations that would have to be combined to capture all of the
above features of the system would be enormous, and is impractical even to
illustrate here. There are separate equations for four different generations, as
well as a set of equations to check for reciprocality between every second
generation. Even just to capture the grandparent generation (not including
reciprocals) the following equations would be necessary:
FF=FFB,
FF=FMZH
FF=MMB
MF=MFB
MF=MMZH
MF=FMB
FM=FMZ
FM=FFBW
FM=MFZ
MM=MMZ
MM=MFBW
MM=FFZ
MM≠FM
FF≠MF
While it would be easy enough to automate a query that searched for all of these
equations, as well as all of those required for the other three generations and the
reciprocals in Radcliffe-Brown's definition, it is unlikely that many word lists
contain all of the necessary terms, and it is also possible that there are systems
that we would want to call 'Kariera' that nevertheless fall short of some of the
above definition. Assuming that the total number of equations and inequations
required for a full "Radcliffe-Brown-Kariera" search is around 70, the best we
could hope for is that such a search would return the "Kariera-score" of each
language in the database: i.e. fulfils 62/70 of the searched polysemies, or 45/70,
or 11/70. We could then set an arbitrary cut-off and call e.g. all languages that
have more than 50 of the relevant characteristics 'Kariera'. Or we could rank
languages on a continuum of more—less likely to be Kariera, or even more—less
prototypically Kariera. All of these are possible solutions.
However, some of the above equations are likely to be redundant. For some
polysemies, it turns out to be the case that if a language has polysemy A, it
always has polysemy B. Some of the equations do more "work" than others: if all
except three languages in the database have a given polysemy, it might be the
case that it is a feature of Kariera, but it is not a particularly useful feature for
distinguishing Kariera from other systems. And some of the elements of
Radcliffe-Brown's definition are simply not considered essential defining
features of Kariera systems by modern anthropologists. For two hypothetical
languages that each might score 66/70 on the "Radcliffe-Brown-Kariera scale",
one might be lacking the four core grandparental polysemies— FF=MMB,
MF=FMB, MM=FFZ, FM=MFZ—while the other lacks the polysemies FF=FMZH,
MM=MFBW and does not distinguish between older and younger brother, or
older and younger sister.
We can test what equations are good diagnostics of a Kariera system in RadcliffeBrown's sense by comparing the results from fuller searches (using equations
from each generation level) to subsets of these equations and seeing which
subsets most closely approximate the fuller results. We carried out a number of
searches of that kind with up to 27 equations and found that the grandparental
equations are most helpful. The details of these searches are not included in this
paper.
4.4 Diagnostic subsets of the Kariera equations :Grandparent equations
We focus here on the grandparental equations. While Radcliffe-Brown does not
privilege the grandparent part of his Kariera definition about the other features
of the system, subsequent anthropologists frequently do. Scheffler's Kariera
definition, for example, can be represented as the following equations:
MMB = FF
FMB=MF
FFZ=MM
MFZ=FM
MMBDD =MBD
FMBSD = FZD (p 53)
WM=FZ (p 58)
WMB=MMBS=MFZS=F (p 79)
These crucially involve sibling terms as in FF-MMB, but FF=MM etc also captures
a Kariera feature where gender distinctions are not found between siblings as in
may grandparental patterns in Australia. Unfortunately, terms for grandparents'
siblings are not present in the database for most of the relevant languages. Out of
the above list of languages, a set of queries describing Radcliffe-Brown's
grandparent generation definitions for Kariera therefore returns only Kariyarra
and Nyamal as "perfect" matches. It also returns strong matches (all of the
equations, or all but one of the equations are found: the missing equation being
due to missing data) for Kokatha, Antakirinya, Wailpi, Wilyakali, Wangapuwan,
Yuwaalayaay, and Umpila. As we will see below, some of these are compromised
by having ‘overlaps’ of different equations in their patterns.
In an attempt to winnow down the grandparental equations further, we can
attempt to capture the Kariera type by only two of the grandparental equations,
the cross-grandparents, rather than four
FM=MFZ and MF=FMB
The languages which have both FM=MFZ and MF=FMB are Kariyarra,
Yuwaalayaay, Kokatha, Wangaaypuwan, and Umpila. Although these are mostly
genuinely Kariera (except Kokatha which is an overlap system, to be discussed
below) this search does not return a large enough number of the Kariera
languages in the database.
FF=MMB and MM=FFZ
This search for the parallel grandparent equations is the best diagnostic we have
found. 10This yields the groups marked on the map ? below with two arrows, red
for FF=MMB and blue for MM=FFZ. The languages with only one arrow are not
considered to be Kariera in this test but may be partially affiliated to this type.
The languages in the Pilbara satifying both criteria of the diagnostic are
Kariyarra itself (which gave its name to the type) and Nyamal. In the Daly River
region is Marringarr . In Cape York Peninsula there are several Kariera groups in
the database: Kok Kaperr. Oykangand, Ogh Undjan, Umpila, Ayabadhu, and
Guugu Yimidhirr. In Central New South Wales there is Yuwalayaay and
Wangaaybuwan.In South Australia, there is Kokatha, Antakirinya, Arabana,
Wailpi and Wilyakali.
A number of these are not neat and tidy Kariera systems but involve mixtures or
‘overlaps’ of systems which we will be discussing in the next section. There are
several terms for FF and MM in Wangaaybuwan and all of these are used for the
sibling of the same sex as well: FF=B and MM=Z. Wilyakali has only a partial
overlap of this parallel grandparent= sibling pattern in FF=MMB=eB.
There are three systems on the map which should be excluded but instead of
using a inequality search in this case we have simply drawn a line around those
which do not belong in the Kariera category.
Wailpi has a term anyani which in addition to MM and FFZ also has the meanings
FM and MFZ. This kind of crossing of the parallel-cross divide means that this
should be excluded from the list of Kariera. This can be achieved by adding a
negative inequality in the search to exclude this type of equation from a Kariera
search. FM=MM , the typical ‘grandmother’ category of most European languages
is also found as a feature of the Aluridja system of the Western Desert and some
neighbouring areas. Kokatha and Antakarinya are Western Desert dialects and
also of this Aluridja type with a split by gender between grandmother (MM/FM)
and grandfather (FF/MF) but with some extensions to siblings of the other
parallel or cross grandparent which gives false positive results for Kariera (eg
MM=FFZ). It may be that this should be considered an overlap between a Kariera
and Alurdija type of patterning and significantly occurs at the eastern edge of
the Western Desert abutting on to languages which have more Kariera features.
10
It is possible also to add the equation MM=FF which does provide some more genuine Kariera
cases where siblings have not been mentioned in word lists, but this has not been done here.
GRANDPARENT
=SIBLING
OVERLAP
ALURIDJA PATTERN OVERLAP
Figure 5: Results of a ‘Kariera’ search using two equations
5. Overlapping Kinship Term Patterns And Systems
5.1 Reports of pattern overlaps in the ethnographic liternature
Tyler (1969:488) criticizes previous ethnographers for failing to properly report
or explain alternate kinship terms or patterns. In his description of the Koya
terminology he reports terms from two languages and several subvarieties in
regular use in the community, forming a single pool of terminology rather than
being selected solely on the basis of a speaker’s or addressee’s language identity.
Some of the terms have semantic differences from each other. For instance in
Telugu the same term taata is used used for MF and FF; and the same term
vaava for MM and FM In contrast in Koya, there are distinct terms for MF and FF
and for MM and FM..That is there are different kinship polysemy patterns in the
same speech community.
Heider (1974) refers to a situation which he calls ‘multi-ply’ variation in kin
term usage operates among the Grand Valley Dani of New Guinea. One example
concerns five sibling terms (Heider 1974:243 ff) : one has the meaning of
opposite-sex sibling, one elder and one younger sibling, and the two others
simply sibling. While people were questioned about the usage of these different
terms, no clear reasons for choice were apparent, and while the idea that some of
the terms may come from different dialects is mentioned, nothing definitive is
stated.
Both these examples involve mainly hypernymy/hyponymy (category inclusion)
– the Telugu grandparent terms are hypernyms of the more differentiated Koya a
cross and parallel terms; in Dani the general sibling terms include the other
terms more differentiated by age or relative gender.
The Omaha equations in Australia which are often ‘overlays’ in Kronenfeld’s
terms (McConvell 2010) are also hypernymic with regard to the underlying
system: eg the MB also covers MBS in the overlay, but the underlying system has
two distinct terms.
5.2 Overlap of patterns in Australia
The examples of ‘overlap’ in Australia that we have already encountered are in
part of this hypernymic type but also of a different type in which both terms have
in addition to the overlapping another non-overlapping area. I suggest reserving
the name ‘overlap’ for the latter type , and calling the former type hypernymic.
Overlap in the sense we are using it takes two forms: one is where the overlap
involves one term which extends in two directions (one-term overlap) and the
other is where there are two terms which overlap in meaning (two-term
overlap).
Take as an example of what we are calling one-termoverlap the term pingay in
Kattang and Dunghutti, .This is a combination of Omaha skewing MBS=MB and
cross-parallel neutralization MBS=B in a single term. This is illustrated in Figure
6.
In Wangaaypuwan the terms kaaka, and kaakampa are used for B, FF and MMB
This is a combination of the ‘Kariera’ component FF=MMB with the alternate
generation merger of sibling and parallel grandparent (see Figure 7)
An example of a two-term overlap in the database is in Wirangu. There are three
grandparent terms that follow the "bisexual" pattern: kaparli (MM, FM), tyamu
(FF, MF) and pakarli (FF, MF), but alongside these there are also two terms that
are used only for the paternal grandparents: wiya (FM) and muma (FF) and these
latter are also used (among several others) as parental terms (wiya = M, muma =
F). The latter is an unusual type of polysemy in Australia.
The question then arises: is it legitimate to call these sets of patterns
‘combinations’ or ‘overlaps’ of other more basic patterns or do they constitute
patternings or separate systems in their own right.? Pushed further, this kind of
example could undermine the notion of a limited number of systems.
We will not carry this argument too much further in this paper as it goes beyond
what we have set out to do here. However it seems to us that this an empirical
question and one which can be answered by using the very same tools that we
have been developing here. If, for instance, it can be shown that these ‘combined’
or ‘overlap’ systems actually occur in zones which are on the borders of two
zones each of which has the component patterns in a ‘purer’ form, then the
notion of a ‘combination’ or ‘overlap’ of patterns seems quite justified.
Figure 6
MB
MBS
B
Figure 7
B
FF
MMB
Figure 8
M
MM
FM
Mixed or overlap systems like this are only a problem if one thinks of kinship
patterns as properties of a language or a cultural group rather than properties of
a word or set of words. They can also be excluded from database searches easily
enough if one does wish to only return languages with more typical examples of
a pattern.
We can, in fact, deliberately search for such mixed cases easily by combining the
same polysemies and inequalities in a different way. For example, the above
case of mixed grandparent systems would be returned by a search for languages
with MM=FM, FF=MF, FM≠FF and MM≠FM or FF≠MF. This search would return
all languages which have both a bisexual pattern for grandparents and also
another grandparental term or set of terms that does not follow the bisexual
pattern. (This search returns not only Wirangu, but also Ngaatjatjara,
Ngaanyatjarra, Pitjantjatjara, Yugambeh and Bundjalung).
6. Other Approaches To Kinship Typology
6.1 Diagnostic and analytical approaches
While analytical approaches such as reduction rules (Lounsbury, Scheffler) and
kinship maps/kinship algebra (Leaf, Read, Fischer) are valuable in discovering
overall principles in kinship systems, the approach outlined here for discovering
recognized patterns and systems in kinship databases fulfils a different and also
valuable role in cross-cultural kinship research.
While the approach of looking for diagnostic features and equations of
recognised types has been somewhat left behind in the building of other
analytical frameworks in kinshio (Kronenfeld 2004:250;253) we believe it still
has heuristic value .The exposition above argues for the use of the AustKin
database and its search and mapping capabilities to recognise kinship ‘patterns’
and ‘systems’ of types described in the world and Australianist kinship
literature.
There are a number of contending approaches to the formal analysis of kinship
sytems, some of them associated with computer applications. The question then
arises: do the method and tools advocated here have a usefulness beyond, or
different from these other approaches?
The answer we give here is that there is a distinct role for the method we are
proposing here. It is not intended to serve the same aims as the other analytical
methods, some of which will be reviewed below.. Therefore the method
proposed here is not, we believe, in competition with the previous analytical
approaches.
The other approaches, which we term ‘analytical’ are seeking a small set of
principles which explain the patterning of kinship terminology in at least one
language. For such principles, to be highly valued, they should have explanatory
power over a wide range of languages/societies. This is an important goal in
semantics and kinship studies. However necessarily and inevitably this
endeavour involves (and has involved) major disagreements about what the
appropriate or best schemes are .
The proposal we are discussing here has a less lofty goal, to produce a heuristic
for identification of recognised and often named patterns and systems in
databases of kinship terminology data. It is thus more in line with the work of
the typologies of the Murdock and the SCCC, and in its geographical aspect, the
tradition of Driver and colleagues working on North American cultures.
One of the important characteristics of the AustKin search proposal is that it is
the input and processing routines of the tool is completely transparent and does
not require high-level theoretical decisions about principles underlying systems.
Analytical schemes that do require such theoretical decisions or choice of
theoretical framework are inappropriate to such a typologising tool, which is
intended to be user friendly, not requiring users to become involved in reflection
or debate about whether the framework or analysis is correct or the best
possible.
This is not to say that this tool is of no use to those who wish to construct higherlevel theoretical schemes. As well as detecting patterns and systems described in
the literature from new data sets, the tool also offers an array of possibilities of
detecting other patterns which occur to an analyst. For instance if an analyst
proposes that pattern X predicts or implies pattern Y in a language, that can
easily be tested using the search mechanism. The AustKin search tool with its
mapping capability can also factor geographical relationships into the search. For
instance in the last section we proposed the category of ‘overlap patterns’ which
to our knowledge has not been discussed much in the literature.11 We further
suggested a hypothesis that such overlap patterns occur primarily in the
geographical interface of two regions, in each of which one of the component
patterns of the overlap pattern is dominant. Given adequate accurate data in the
database such hypotheses can be tested.
6.2 Analytical approaches
Below we briefly review some of the main previous and current approaches to
kinship analysis. Each of these has its own merits but none are appropriate to
use directly for kind of basic typological search tool we are proposing. We make
some preliminary remarks about how the AustKin search tool or similar could
contribute to research in these frameworks.
Componential analysis
In the 1960’s-70’s there was a wave of publications in componential
ethnosemantics, many of which dealt with kinship. We will not discuss these
approaches in detail here. They were analytical in the sense we are using the
term, and grew out of structuralism, and specifically notions of ‘distinctive
features’ pioneered by Jakobson first in phonology then in other fields. In
kinship, terminological systems were analysed as terms which had sets of
positive and negative values for a set of features. While some features were
recognised as needed in a universal scheme, there was continuing disagreement
over what were the appropriate features. This kind of impasse leads to the
inability to carry out comparison in a way that is generally accepted. Another
failing of componetial analysis of kinship was the lack of a concept of polysemy
in most work (Scheffler 1972: 324): practitioners sought unitary feature bundles
to describe the meaning of each term.
This leads to problems wth many kinship systems. Crow and Omaha systems
were particularly intractable without formal recognition of polysemy or
extension. Such systems merged adjacent generation kin types but only in
certain contexts, eg. for Omaha between +1 and –0 generations only on the
matrilateral side.The development of reduction rules was in part a way to get
11
This is distinct from the notion of ‘overlay’ which is applied by Kronenfled (2004) to Crow and
Omaha equations since they are used only in certain contexts and a more basic system used in
others. However the two concepts are related: ‘overlays’ are a kind of ‘overlap’’ if both the
overlay pattern and the basic pattern invlve different directions of extension.
around this problem (Lounsbury 1969).
Reduction rules
Lounsbury introduced this formalism and Scheffler further developed it. In
order to analyse systematic polysemy as extension from a basic kintype to
others in various ways. The rules rewrite one kin-type (represented as a string of
links F, Z etc) as another (more basic) string often in certain environments of kin
type elements before or after the string to be rewritten. The rules are ordered
and may br recursively repeated..
This type of framework provides a useful way of describing extension patterns,
and has been used to good effect in Australian kinship by Scheffler (2008/1978).
There are a limited set of rules, some of which are candidates for being
universals. A number of the rules apply not just to one kinship pattern but a
number of other complex strings where the description of the input and
environment of the rule is met. This considerably simplifies a description of a
system, whereas in contrast, in the AustKin search set-up, each instance of the
rule has to be separately entered.
For instance, in some Australian Omaha skewing patterns not only is the kintype
MBS realised asMB but the kintype MMBS is realised as MMB. If a reduction rule
is of a form which converts MBS to MB, in the context of no linking relative or
linking relative M_ then both the reductions (and potentially other correct ones)
will be covered. As the AustKin search is presently set up, there is no option but
to spell out all the equations/polysemies rather than use a generalised rule.
This kind of requirement to spell out all the equations is not a disadvantage
when one considers the purpose of the tool we are recommending. In that way a
user can find out exactly what is going on in different languages through the
search rather that have to construct a more complex general rule first. One
approach would be to adopt the ‘indicator’ or diagnostic’ strategy : to find out via
AustKin searches which equations reliably predict which others in actual
empirical cases, then investigate whether these implicational patterns reflect
what have been suggested in thhe reduction rules studies.
It would be possible to generate the main equations predicted by a reduction
rule and then interrogate the database with the result. It might be possible
alternatively to change the search in such a way that it finds partial string
equations rather than full strings, for instance where XMBS and XMBare equated.
This has not been done so far in AustKin..
Kinship algebra/kinship mapping.
This approach (Leaf, Read, Fischer) is often presented as a way to carry out
initial collection of a kinship system and thus contrasts with other analytical
approaches which assume a set of equations of kin type strings then try to
explain those patterns. It also stresses (following Schneider) that it should not be
assumed that ‘kinship’ is solely a matter of genealogy but that the investigator
should be guided by what members of the society say about the meaning and
relationship between kinterms.
It is clear then that this approach differs in some fundamental ways not only
from the AustKin search tool proposed here, but also other analytical approaches
such as reduction rules. It is possible to add non-genelogical terms to AustKin
but the highly constrained semantics provided by kin type strings would be
absent. On the other hand much of what is done under the kinship
algebra/mapping rubric is in fact beuilt on fairly standard genealogical concepts.
Apart from the initial stage of ‘mapping’ kinship there is a second analytical stage
in kinship algebra where differential principals are isolated which drive a
number of features of each type of kinship system. Arriving at a justifiable
generalisation about such key principles is not a simple process and does not
necessarily yield an immediate or single solution. Regardless of the value or
otherwise of this procedure, it is very far from the kind of transparency and
user-friendliness which should characterise the AustKin search tool. It seems
likely,as with reduction rules above, that AustKin type searchers ccould be used
as a heuristic in establishing kinship algebra solutions, but we have not
attempted this here.
7. Conclusions
This paper has proposed a way of using the search capacity of the AustKin
database to find regular patterns and systems in the kinship terminology data
for Australia (and potentially for any other kinsip data base similarly organised
for other continents).
The inspiration for this approach grows from the work of Murdock and the
Standard Cross-cultural codes approach to kinship patterns, and the first task set
for the AustKin search tool was to capture some of the SCCC patterns. We do not
share a general opinion that this ‘diagnositc equation approach to typology,
flawed though it is, is completely superseded and to be discarded. One reason is
that those interested in kinship still use such descriptors of linship patterns and
systems and find them comprehsnible and of practical use. Rendering them into
search menisms in a database will have practical value. Secondly, though, it is not
conclusively shown that diagnostic approaches also have no theoretical value,
and we suggest some ways in which they may be assessed and refined using the
same AusKIn search tools.
In testing the application of AustKin searches to SCCC kinship patterns, ome of
the cousin and grandparent patterns were successfully retrieved from the
database and mapped.
Not all the significant patterns of kin equations are represented in the SCCC
sample, and in particular some of the most significant patterns in Australia are
absent. We therefore looked at what Australianist anthropologists from
Radcliffe-Brown on have used in kinship typologies, focussing on the notion of a
‘Kariera’ system. ‘System’ in this tradition is a broader concept than ‘pattern’
consisting of several kin equation patterns which tend to cluster together, and
forms of prescriptive marriage.
This notion of system is too broad, with too many features, but even in RadcliffeBrown there is an incipient notion of a core set of features within these, which in
the case of Kariera is the set of grandparent equations. We suggest that these or
even a smaller subset within them are a ‘diagnostic’ or ‘indicator’ set which
predicts other typical feature of ‘Kariera’ – and this empirical validation can be
carried out by the AustKin search tool.
Another issue is the status of ‘mixed’ or ‘overlap’ systems where a language has
either single terms or sets of two terms which extend polysemy in different
directions, so that the language cannot be uniquely assigned as having solely one
or other pattern. This phenomenon is not widely discussed but is probably
widespread - and not only in Australia. The mapping interface of AustKin
enables us to see where such overlaps exist and this may help to wards their
explanation.
Finally we dal with some of the higher-level analytical schemes that have been
proposed to explain the different types of systems that exist. This task is
different from simply capturing the patterns as we are proposing for the AustKin
search tool, but the AustKIn search tool can be useful in providing an initial
survey of pattern and system distribution which may then be subject to analysis
and esplanation.
References
Denham, Woodrow W. and Douglas R. White 2005. Multiple Measures of
Alyawarra Kinship Field Methods 17: 70-101. DOI:
10.1177/1525822X04271610
Dousset, Laurent. 2010. Horizontal” and “vertical” skewing: similar objectives,
two solutions? Paper to Crow-Omaha Seminar, Amerind Foundation,
Arizona.
Dousset, Laurent, Rachel Hendery, Claire Bowern, Harold Koch, and Patrick
McConvell. Unpublished MS. Developing a database for Australian
Indigenous kinship terminology: The AustKin project.
Heider, Karl G. 1978. Accounting for variation: a non-formal anlaysis of Grand
Valley Dani Kinship Terms. Journal of Anthropological Research. 34.2:219262
Keen, Ian 2004 Aboriginal Economy and Society. Oxford: Oxford University Press.
Kronenfeld, David 2004 Definitions of Cross versus Paralle: Implications for a
new typology(An appreciation of A. Kimball Romney) Cross-cultural
research .38:249-269.
Lounsbury, Floyd G. (1964), "A Formal Account of the Crow- and Omaha-Type
Kinship Terminologies", in Ward H. Goodenough (ed.), Explorations in
Cultural Anthropology: Essays in Honor of George Peter Murdock, New
York: McGraw-Hill,351–393
McConvell, Patrick.2010. ‘Omaha skewing in Australia and New Guinea: overlays,
dynamism and change’. Paper to Crow-Omaha Seminar, Amerind
Foundation, Arizona.
McConvell, Patrick and Barry Alpher.2003. On the Omaha trail in Australia:
tracking skewing from east to west. Anthropological Forum. 12.
McConvell, Patrick and Ian Keen. In Press.
Murdock, George Peter 1970 Kin term patterns and their distributions.
Ethnology. 9(2): 165-208.
Radcliffe-Brown, A. R 1930-31 The social organization of Australian tribes.
Oceania 1:34-63.
Scheffler, Harold 1972 Kinship Semantics. Annual Review of Anthropology 1. 309328.
Scheffler, H. W. 1971. "Dravidian-Iroquois: The Melanesian evidence",
Anthropology in Oceania. Edited by L. R. Hiatt and E. Jayawardena, 23154. Sydney: Angus and Robertson.
Scheffler, Harold W 2008 [1978] Australian Kin classification. Cambridge Studies
in Social Anthropology. Cambridge: Cambridge University Press. .
Tyler, S. 1 969. Context and variation in Koya kinship terminology. In Cognitive
Anthropology, ed. S . A. Tyler, pp. 487-503. New York: Holt, Rinehart and
Winston
Radcliffe-Brown, A. R 1930-31 ‘The social organization of Australian tribes’.
Oceania 1:34-63.
White, Douglas R., Michael Burton, William Divale, Patrick Gray, Andrey
Korotayev, Daria Khalturina 2009 Standard Cross-Cultural Codes.
http://eclectic.ss.uci.edu/~drwhite/courses/SCCCodes.htm, accessed 30
November 2009.
Download