Dr Janet Bagg, University of Kent, Canterbury, UK, 1996
This paper describes work being done at the University of Kent, UK in connection with two Economic and Social Research Council funded projects in Social Anthropology. One seeks to develop methods for computer use appropriate to the discipline, in particular the use of database management systems and interfaces, exemplifying these in a study involving both fieldwork and archival research in Southern Corsica. The other project examines marriage practices in the town of Ascoli Satriano, Puglia, southern Italy, focusing on their relationship with kinship, neighbourhoods, heirship strategies and patterns of political and ecclesiastical succession. The Corsican research, although more limited in scope, has similar aims. The original paper was given at the Association for History and Computing conference, at the University of Hull, UK in April 1994.
Kinship studies have taken somewhat of a back seat in anglophone social anthropology during the last fifteen years. This has partly been in reaction to the way in which the importance of kinship relations was overemphasised in structural-functionalist studies, and also due to a shift in emphasis towards symbolic analysis and later still the rise of interpretative concerns. Meanwhile, kinship continued to be an important area of research in French anthropology. Here, although similar themes to its English-speaking counterparts were important, a parallel tradition with its origins in Claude Lèvi-Strauss work on 'elementary structures of kinship' (Lévi-Strauss 1949) ensured that the area did not suffer neglect.
Recent work by French and Italian social anthropologists and historians, not necessarily following a structuralist line, has produced a new and active area of research focusing on kinship, family and marriage in 'complex' societies. Their studies are either wholly founded upon archival evidence, or upon a combination of such sources with fieldwork in the contemporary town or village. The methodological link between historical research and ethnography fits well with a tradition established in mediterranean anthropology at the University of Kent by Professor John Davis (Davis 1976).
Central pieces of empirical research in this area have been those of Martine Segalen in Brittany (Segalen 1985) and Nanterre (Segalen 1990), and those of Gerard Delille in the Campania (Delille 1985) and Manduria (southern Puglia). Through large scale genealogical reconstruction, they have identified a significant amount of repeat marriage of different types. Incidence and type of kinship relation are found to vary through time, by region and by socio-economic group. Delille has also noted regional variation in the relationship of kinship to residential patterns. Campania was characterised by groups of houses containing kin related through males (quartiers lignagers) while Puglia lacked such concentrations but showing a tendency towards what he terms matrilineal inflexion. This shows itself in particular in neighbours being related through women.
Delille's model is based upon his own archival research in Campania and Manduria, but little work has been done to test these ideas in other regions. Ascoli lies in northern Puglia, an area dominated by the finely balanced sheep/grain economy of the Neapolitan royal customs house up to the late 18th century and later by large scale grain producing estates and hired labour. This has produced an ecological character quite different to those areas examined in detail by Delille and others. Such local studies are of more than parochial interest. It is far too easy for a large and diverse geographical area, even a whole modern nation state to be characterised by broad sociological features, with little attention paid to variation which may indicate important historical divergences. Corsica too illustrates this point. Most ethnographic studies (e.g. (Ravis-Giordani 1983)), plus related archival research, have been concentrated in the north-west of the island which is characterised by the importance of the pastoral economy and forms of common rights in property. Less work has been done in the south, where individual ownership has long been the norm. Because of such regional bias, the whole island tends to be characterised by the social relations salient in the north.
The projects share the aims of examining kinship, marriage and residence. Both have common requirements for information management. Other areas are also vital to one or both projects, such as record linkage, dealing with mappable data and using textual information linked to more structured data. Therefore we wish to take a common approach in computing implementation. This has the benefit of allowing me to use the Ascoli data for testing programs before enough Corsican data has been entered.
One of the fundamental strengths of social anthropology has been, and still is, to show the importance of links between different areas of social life. This power, although initially developed under the functionalisms of Malinowski, Radcliffe-Brown and others, has not only continued through more recent theoretical shifts but been enhanced by them. Given such concerns, when considering how to use computer methods in ethnographic research ( whether on contemporary or past populations ), it is very important that information is organised in a way that makes it as easy as possible to investigate possible links. Because of this, and given our current real-world ( i.e. supported ) software possibilities, we have chosen to use a relational database management system to organise our information. This permits us to keep track of different parts of the information sources used by the projects ( parochial and civil registers, census type documents, fieldwork notes and interviews, notarial documents, land surveys ) and to make links between the different aspects of social life recorded. Access is by three means:
1. ad-hoc SQL queries - the simplest method. Very easy to write. Suitable for the type of one-off questions often raised in research.
2. designed interfaces. These are suitable for frequently used tasks. Useful for input, editing and presentation. The overheads vary with the type of interface. Tasks requiring little transformation of data can easily be constructed with product supplied interface builders. More complex tasks require programs in conventional languages (C/++) some of which may operate in conjunction with the products own interface systems.
3. links to other programs. Using interfacing methods to link the database system to existing analysis programs. Such software usually requires particular forms of input so that interface programming must be used to transform the data accordingly. Recent real-world advances in connectivity make it easier to link some database systems to other programs such as spreadsheets and word processors.
In examining kinship, we have identified two initial requirements:
1. The identification of kinship links between two individuals. This is particularly relevant for marrying couples, in the context of the work on marriage and kinship mentioned above but may also be used in looking at other social relationships such as godparents or credit relations.
2. Calculating 'lineages' in the genealogical sense and using these classifications to map other information. For instance, residence data - testing for kin clusters similar to those identified by Delille as quartiers lignagers ( though not necessarily expecting them to be directly comparable).
Over the past 25 years, several approaches have been tried in using computers to examine genealogical data. They fall into 4 main groups.
1. Programs for professional and amateur genealogists. These have some interest in demonstrating methods of information presentation. Most, however, use specific file formats and can not be attached to database systems holding related information. Where database systems are used they tend to be of the more limited and archaic type such as the XBase series (.DBF). None of them provide any analytical routines. Output is usually in a form useful to genealogists, such as pedigrees, but not to social anthropologists.
2. Analytical programs in anthropology. A classic early study examined the empirical incidence of cousin-marriage among pastoral nomads with patrilineally based kinship and compared this with the perceived incidence (Randolph and Coult 1968). Programs at this period were very much specific to the problem and relied upon specially formulated data. Output was severely constrained by the existing technology. Other approaches, mostly from the USA and France, are based upon graph theory (White and Jorion 1992) and algebraic description (Read and Behrens 1992). Segalen's work uses programs written by P. Richard to analyse a total population to find the incidence of repeat marriage (Segalen and Richard 1986). The programs use special data files and require knowledge of the genealogical generation of each person. A similar approach has been taken by Marion Selz in examining marriage patterns in her work with Prof. Hérétier in the area of complex systems of kinship (Selz-Laurière 1987). The programs used by Delille and the researchers at the Ecole Française de Rome, derive from the latter and use XBase (.DBF) format files. A large part of this work has been concerned with questions relating to marriage, and has been influential in decisions that I have made.
3. Graph Browsers. These are programs designed to traverse and display information recorded in the form of a mathematical graph. One example is Ryan's Gnet, the use of and problems with which I have written about elsewhere (Bagg and Ryan 1991). Graph-browsers are too generalised to be very useful, even in visualising genealogical data, which has certain very specific requirements involved in laying out the graphs. As the are intended to be general, they do not contain any application specific analytical tools. However, in terms of data structure and presentation possibilities they have been useful guides.
4. Prolog. This is a programming language based on the idea of information and rules. The simple ancestor search is a classic example used to introduce the basics of this programming language (for an anthropological example see (Fischer 1994) chapter 7 Prolog would appear to be a useful and simple way to write genealogical programs. This is however a much more complex task than at first appears. A rule-based approach such as this has much to recommend it but the language is not widely available in what might seem its most practical use, as a database system query system, or as a language for writing interfaces.
As stated above, our information is very complex and derived from many sources. Because of the importance of making links between pieces of information in social anthropology, it is essential that data is stored in a way that not only makes it easily accessible but that allows great flexibility in retrieval paths. At present, the most practical solution which is actually supported, at the University of Kent as at most academic institutions, is a database system from a major supplier. The projects use Ingres 6.4, a choice made on three grounds: local support, interface building tools and the development of knowledge extensions. For software development purposes, data has been exported to MS Access on 486 based machines but this product, as with all micro-based database systems, is less easy to use for work requiring bulk processing (e.g. record linkage) and complex ad-hoc queries than its workstation-based siblings. It is however the preferred option for entering information and can be used in fieldwork situations on a portable.
The data model is designed to be incremental, with a set of core tables (Bagg and Colclough 1992), and other groups of tables linked to the core by one or more key fields. So far data for the Ascoli project has been drawn from 4 types of source:
1. State of Soul registers: five registers have been entered covering the period 1709-1765. All except the 1709 register give, in the vast majority of cases, names, ages, residence, communicant status and place of origin, plus the names and places of origin of both parents and the former spouses of widowed persons. Such detailed information has made linkage fairly easy and has allowed the construction of a table consisting of a unique set of individuals and another giving the relationships between them. Only three types of relationship are specified in this table: mother, father and spouse. Parental relationships also specify a family number which is given to each couple and their children. Other tables record information specific to the registers.
2. Parochial and civil life-event registers. These are entered as discrete sets of tables linked by event number and, afterwards, linked to the core tables through person ids and family numbers. Data from these sources, through the linkage process, is being used to modify and expand the core tables derived from the State of Souls registers.
3. Cadasters. These are surveys of property which may also include census type listings which are not directly comparable with the State of Souls registers. The Catasto Onciario of 1753 has been entered using a set of appropriate tables with entry numbers and person ids as keys and linked to the core tables by pid;
4. Notaries records. These are complex textual documents. At present there are tables of persons involved in each contract, information about each contract, key words and notes. Each contract is given a unique number which is used to link the tables.
As the Corsican project is at a much earlier stage, much less information has been entered but it will follow a similar model where possible. All of the development carried out so far in interface software has used the Ascoli data for testing but is intended to be of wider application.
Two interface modules are presented here, both run as options from the same program, which uses ODBC links to MS Access in the PC version and Embedded SQL in the Unix/Ingres version, within C++ code.
1. Finding any common kin for two people.
This is being used to examine marriages recorded in the State of Souls registers. The program finds separately for the husband and the wife:
1. The direct ancestors in both male and female lines.
2. The spouses of the siblings of each ancestor.
3. Other spouses for each direct ancestor.
4. The direct ancestors of such spouses.
These searches result in a temporary table for each of the couple giving id of relative, relationship string (e.g. FFF), id of originating person, line of ascent (see below) and generation of each related person. The tables are then compared to find persons common to both sets of kin. From these, the related person closest in generation to the marriage is selected for each pair of descent lines. After initial experiments, it was decided to restrict results a comparison of the direct line of one member of the couple and all lines for the other (and vice versa).
Links to a common ancestor are presented from each person in the married couple, for example: person 2404 is the brother of 2408 and the husband of 4277's sister. The generation upward from both husband and wife is also given. In the example (Table 1) the link is at the same generation.
Marriage 2408 to 4277
2404 B gen: 0 line: 1 || ZH gen: 0 line: 3 RCA
4282 BW gen: 0 line: 12 || Z gen: 0 line: 1 RCA
Table 1: An example of the program output showing in this case a sibling exchange.
A further piece of information is the line of ascent, given again for both husband and wife. Line 1 is the direct line up through fathers and mothers, including siblings. Line 2 denotes the affinal line ( ascent of previous spouses ). Further lines trace the ascent of the spouses of people in the direct line. Marriages in the direct line are consanguineous (with 'blood' kin). Those between the direct line and the affinal line are affinal (with in-laws). Other marriages where common kin are found within the above constraints are classified as affinal relinkage - a second marriage between two sets of consanguineous kin. These are denoted by the codes CON, AFF and RCA.
The third type of marriage is by far the most common within the Ascoli data set. This is not surprising - Delille thinks that it is unlikely that many consanguineous or affinal marriages will be found within less than 6/7 generations. At present we have no more than 5 generations in any ascent line but will of course have more as further data is entered and linked.
What is more interesting is the high number of sibling exchanges ( two brothers marrying two sisters or two brother and sister pairs getting married). This is confirmed by material in the notarial contracts which sometimes treat the dowry exchanges in such cases as one, undivided transaction. This type of marriage will have to be investigated in more detail, examining its frequency over time, whether it was practised by many or a restricted number of families, property implications in terms of dowry and other transmission and its association with residence patterns. It is hoped that the notaries contracts and perhaps ethnographic enquiry might prove helpful.
At present either all marriages or an individual marriage can be examined. When more marriage data has been entered, with dates and linkage to other demographic information, marriages should be examined by cohort. This will allow us to assess whether practices of kin marriage are changing through time.
A further requirement will be to specify the population from which marriage choices might be made. To do so, the age ranges within which spouses are likely to be chosen (e.g. +n years and -n years relative to the age of ego) must be defined, to determine how many of the eligible men or women for each marriage decision are kin or not. This will help us to understand whether kin marriage is purely a result of limited choice. Because of difficulties with allowing for migration only those persons "in observation" at any one time can be examined as candidates.
Another area for development is the establishment of some form of descriptive classification of affinal relinkage, similar to the degrees of consanguinity and affinity familiar from canon and civil law.
2. Descent and residence.
Delille has identified areas of towns in Campania which show marked concentrations of people from the same patriline (related through males). He has also noted that in Puglia, relationships through the matriline (related through females) tend to be more important.
An interface module has been built to plot residence classified into nine districts upon diagrams of matrilines, patrilines and bilineal descent. Each of the districts is given a unique colour and the districts are grouped into broader divisions through the use of similar colours. Reddish for the north-west of the town, blues for the south-west area around the castle, gold for the commercial district of the piazza and green for the 18th century development to the south-east. The three types of descent can be examined and residence in each register plotted (those not resident at each date are shown in grey) for each diagram.
Example diagrams: The descendants of Potito Bortoni and Angiola Ferrachia of Ascoli in 1731 and in 1754
In this case, those related through females can be seen to group in the north-west of the town although their brother Niccolo also lives there, in the house of one or other of his sisters. They appear to more stable in residence. Lorenzo's children lived in this area when they were young but moved later, probably after marriage. Apart from Antonio's house in the castle district in 1731, all of the residences were rented. It is probable that the large number of rented properties in Ascoli had a considerable effect upon residential patterns. Only a few families occupy their own house through the generations. Others may own a house but rent it out.
A single case is merely an example and conclusions cannot be drawn from it. Further work is needed to establish a method of determining levels of residential clustering which takes into account both number of kin and size of district.
Several necessary improvements have been mentioned above. Some await further data entry and linkage, others can be implemented almost immediately. The marriage analyses need to be more tightly controlled so that we may postulate whether choice of a relative as a spouse was a matter of chance, lack of alternatives or deliberate selection. Comparative analysis over time will also be important. For residence patterns, some form of measurement is clearly required to discover the presence of and frequency of clusters of kin. To understand patterns over time several factors must be taken into account as well as descent line such as sex, marital status, age and tenancy.
This paper has described some experimental interface software for examining large amounts of data covering the whole urban population. Such general analyses are only one part of the projects. Much will also be gained from close reading of notarial documents and from interviews conducted as part of the contemporary ethnographic study. The latter will not be a case of reading the past from the present, but will allow a commentary upon such practices as kin marriage in a contemporary context and also in reflecting upon the past. Other tools are being developed to help organise and navigate such textual material and to link it to the general sources.
A good database system is currently the only practical way to organise and retrieve large and complicated collections of information. Now that such systems can include large blocks of text and binary data, the complexity is likely to be greater than ever. At the same time, ever more elaborate questions are being asked of the material and new methods of presenting results (both summary and detailed need to be found). Fortunately, because of interface building kits and ODBC links to existing products (e.g. spreadsheet programs) simpler presentation methods have become easier to achieve leaving more time to concentrate on more complex areas.
(1) ESRC Grants: Dr J. Bagg R000234538 "Anthropological information and computer representation: modelling time-related data" and Dr. N. Colclough R000234749 "Marriage strategy in North Puglia, 1730-1990: a computer aided study".
Bagg, J. and N. Ryan 1991 Interacting with Diagrams: Using a Graph Browser with a Research Database. In H. Best, E. Mochman and M. Thaller (eds) Computers in the Humanities and Social Sciences: Achievements of the 1980s, Prospects for the 1990s, pp. 280-285. K. G. Saur, Munich.
Bagg, J. and N. Colclough 1992 The Ascoli Database, BICA 8:16 - 22..
Davis, J.H.R. 1976 An account of changes in the rules for the transmission of property in Pisticci, 1814-1961. In J.G. Peristiany (ed) Mediterranean Family Structures, pp. 287-304. Cambridge University Press, Cambridge.
Delille, G. 1985 Famille et propriété dans le Royaume de Naples (XVe-XIXe siècles). Ecole Française de Rome, Rome-Paris.
Fischer, M. D. 1994 Applications in Computing for Social Anthropologists. Routledge, London.
Lévi-Strauss, C. 1949 Les structures élémentaires de la parenté. Presses Universitaires de France, Paris.
Ravis-Giordani, G. 1983 Bergers Corses: les communautés villageoises du Niolu. Edisud, Aix.
Randolph, R. R. and A.D. Coult 1968 A computer analysis of Bedouin marriage. South-western Journal of Anthropology 24: 83 - 99.
Read, D. W. and C. A. Behrens 1992 Computer representation of cultural constructs: new research tools for the study of kinship terminologies. In M.S. Boone and J.J. Wood (eds) Computer Applications for Anthropologists, pp. 228-50. Wadsworth, Belmont.
Segalen, M. 1985 Quinze générations de bas Bretons: parenté et société dans le pays bigouden Sud (1720-1980). Presses Universitaires de France, Paris.
Segalen, M. 1990 Nanterriens: les familles dans la ville. Presses Universitaires du Mirail, Toulouse.
Segalen, M. and P. Richard 1986 Marrying kinsmen in the pays bigouden sud. Journal of Family History 11:109-130.
Selz-Laurière, M. 1987 Parenté et informatique. Mathématiques et Sciences Humaines 97:57 - 66.
White, D. R. and P. Jorion 1992 Representing and computing kinship: a new approach. Current Anthropology 33(4):454 - 463.