Le Médiéviste et l’ordinateur
Le Médiéviste et l’ordinateurHistoire médiévale, informatique et nouvelles technologies
n° 41 (Hiver 2002) : L’apport cognitif

THE DEEDS PROJECT : Towards the dating and analysis of english private charters of the twelfth and thirteenth centuries

Michael Gervers, PhD
Director, DEEDS Project,
Professor of History, University of Toronto
e-mail : 102063.2152@compuserve.com
website : http://www.utoronto.ca/deeds

The Norman contribution to the administrative history of medieval England is a commonplace of academic inquiry into the period. Among the many novelties which followed the Conquest of 1066, there is one in particular whose consequences have blurred our understanding of social, political and economic change for the nearly two and a half centuries separating the Conquest from the end of the reign of Edward I in 1307. That particularity is the custom of not including a date of issue in records recording property transfer. Such records, known as deeds or charters, were the most ubiquitous records of the time and it is estimated that for the twelfth and thirteenth centuries alone over a million have survived to this day as originals or cartulary copies.

There was considerable diversity on the Continent at the time of the first millennium in the tendency to include, or exclude, a date reference when property conveyances were recorded in writing, depending upon where and when they were issued. There can be no doubt that the closer the tie between the issuer and the Roman tradition, the greater the likelihood that a charter would bear a date. There is increasing evidence to suggest that in Burgundy, as in Normandy, the custom of including dates declined steadily from the late tenth century until the middle of the eleventh century, when they became quite rare. Although situated within the Frankish realm, these territories lay well outside the formal control of the French king. It is above all this detachment from royal administrative supervision which may explain why, when the charters of the French monarchy continued to be dated, those of some, at least, of the more independent outlying provinces did not. In this regard it was England’s fate that the Norman Conquest took place when it did, for with William I came the then-current Norman custom of not dating charters at any level of society. The Norman dukes, with an administrative system developed when ties with the French Crown were weak, were clearly unconcerned about the formal letter-writing conventions which European monarchs had adopted from the papal chancery. They bothered little with traditional formality, dispensed with all that was not absolutely essential to the message they wished to convey, and only occasionally included a date. They sought brevity and conciseness in their charters and carried that administrative principle with them to England where it became a long-standing tradition in its own right. It was undoubtedly this tradition, rather than a conscious desire to dispense with the concept of time, that led to the enduring English phenomenon of the undated charter. In Normandy, dating returned to the charter text as royal administrative authority spread in the region after 1204, while in England it was with the accession in 1189 of Richard I, whose administrative experience was rather Continental than English, that dating was regularly introduced to records emanating from his royal chancery. Attachment to tradition being what it is, it still took well over a century for that custom to be adopted consistently by those who drafted charters elsewhere in the realm. It was not until the early years of the reign of Edward II (1307-27) that it became customary to include dates in private conveyances.

The development of a methodology to date undated medieval charters has been the primary occupation of the DEEDS Project over the past eight years. The traditional method of ascertaining at least a relative chronology for undated records has been through the association of personal names in those records with their counterparts in dated sources. This approach works reasonably well when one, but preferably two or more office-holders whose dates of tenure are known appear in the same record. It becomes distinctly hazardous, however, in the case of individuals to whom a clearly defined chronology cannot be attached, since namesakes, even when occurring in groups (as they do from one generation to the next in rural environments), can often be shown to be different people. And even when the identification is certain there is usually no way of telling at what point in the life-span of that individual a given record was issued. For lack of sufficient contextual evidence inaccurate attributions can easily be made in the case of unknown individuals and dubious dates consequently assigned to the records in which they appear. Furthermore, once a circa date has been attributed to such a record and the work goes into print, that date tends thereafter to take on a historical value of its own. It is given the benefit of the doubt by subsequent scholars as though it actually belonged to the record in question. In this manner, dates are assigned to undated charters which in turn are used as evidence in the dating of other undated charters. The potential for error in this process is rather greater than one would like it to be, and probably far greater than one could imagine. In a recently published posthumous article by John Horace Round, the author points out how the respected editor of the Colchester Cartulary misdated a dated royal charter by more than 200 years because he did not know that the king’s style as registered in the dating clause, « Henricus Rex Anglie et Francie et Dominus Hibernie », designated Henry VI rather than Henry III [1]. When errors of this sort can be made with dated royal charters, how much more vulnerable might be dates assigned through the association of personal names to undated private ones.

Problems arising from the misidentification of different individuals bearing the same name, or of the same individual bearing the same name, not to mention the same individual bearing different names, are compounded in those records whose witness lists were omitted, as often happened when they were copied into a cartulary. In the case of records in which no more than two or three otherwise unknown names appeared, other means were obviously needed to identify chronological change. Palaeography was not an option as the great majority of the sources from the period have survived only in later copies. For the same reason, sigillography was also insufficient, regardless of its importance in those relatively rare cases where originals exist together with their seals.

Seeking an alternative solution, DEEDS turned to a route suggested by the English medieval historian, Sir Frank Stenton, eighty years ago. Faced with the reality that only five percent of the records he was editing bore dates, or could be assigned accurate dates based upon recognizable names or events cited in them, he became convinced after making a meticulous examination of his sources that charter chronology was inextricably tied to the growth and development of the formulae which appeared in them [2]. The term « formula » is somewhat ambiguous as it suggests a set of words whose order does not change regardless of context. Since nothing could be further from the truth in terms of the medieval legal phraseology of the twelfth and thirteenth centuries, DEEDS prefers the expression « word-pattern » to « formula » and uses it to designate any group of two or more words which occurs in more than one record.

In 1983, using what was then considered « new » methodology, the DEEDS Project produced a microfiche concordance of the content of the 957 charters in the secunda camera of the English Hospitaller cartulary of 1442 [3]. Careful perusal of the concordance made it unquestionably clear that these charters alone contained tens of thousands of identifiable word-patterns, that is groups of two or more words in one record which were exactly similar to a corresponding group in another record. The wide variety of dates associated with the different patterns led to the supposition that rather than being predominantly static, with the occasional change resulting from a historical event or made in response to a royal statute, the language of the medieval charter was in constant flux. Taken as a whole, it seemed reasonable that if the chronology of change were clear, it should be possible to use the rates of appearance of word-patterns in the record, and their disappearance from it, to determine a fairly accurate date for any given record of this sort. In other words, chronological groupings reflected in the word-patterns from the concordance suggested word-pattern matching to be a potentially viable means for dating the undated charter.

To make the proposed methodology viable DEEDS researchers began by identifying the dated private charters (that is, those records which were issued neither by the royal family, nor by the papacy) which were available in the printed record, and converting them into electronic format. The corpus of charters, derived from some 115 published cartularies and collections, presently contains over 6,200 such records, each of which is represented as a plain (ASCII) text. These records date from the late eleventh to the early fourteenth century. Identifying that number of dated acts meant searching through some 75,000 charters, since dated records represent, on average, only eight percent of the extant corpus. The text, converted from the published copy through scanning and Optical Character Recognition (OCR), is separated from its printed context (page numbers, English summary, editor’s notes, rubric), and verified manually for accuracy against the printed source, itself retained in the system as an image file for subsequent checking and comparison. The records are then stored in a database (ORACLE) and managed through a custom built programme developed by the DEEDS Project called the Document Manager (DM). The DM serves as an interface which allows processed records to be loaded and managed as documents of predefined format, where the records themselves are linked to the attributed data.

One of the major challenges faced by the researcher working from published records rather than from manuscript sources is dealing with a myriad of editorial methods and principles devised by individual editors. Wherever possible, the computer is used to eliminate or circumvent such differences. Letter « v » is invariably read as « u », and « j » as « I »; letters « c » and « t » are standardised in certain combinations. No distinction is made between upper and lower case letters, and punctuation is removed. More problematic are cases where editors (and sometimes medieval copyists as well) have abbreviated texts or replaced what they interpret as purely formulaic phrases with « etc. ». We mark up obvious lacunae and retain abbreviations, but disregard them in research queries, and always retain the « etc. ».

In addition to the raw text of which it is comprised, each record is associated with a specific set of attributes : 1) date, 2) date type (assigned by an editor or internal ; if internal, by AD, regnal year, feastday, or event) ; 3) record type (grant, agreement, quitclaim and so on) ; 4) whether the source is an original record or a copy, and if a copy, the nature and date of the source ; 5) place of issue ; 6) parish and county concerned ; 7) the general nature (religious or lay) of issuer and recipient ; 8) the specific nature of the issuer and recipient (religious order or institution, or an individual’s name, title and occupation) ; 9) in the case of religious houses, name of the house and dates of foundation and dissolution ; 10) length of the text; and 11) total number of records published in the source collection. The raw text and attributes are integrated into one document by means of Extensible Mark up Language (XML), which is subsequently stored in the database.

Editors do not concur in their assessment of record types. As a consequence, DEEDS attaches its own designation in a standardised approach based on words of disposition (GRANTS invariably include forms of dedi and/or concessi, sometimes together with liveravi, tradidi, dimisi, assignavi, contuli, legavi, and donavi ; CONFIRMATIONS incorporate forms of confirmavi, and concessi et ratam habui ; QUITCLAIMS, forms of quietum clamavi, sometimes in conjunction with remisi, concessi, recognovi, resignavi, relaxavi, confirmavi, and dimisi ; LEASES, forms of dedi, dimisi, concessi, and tradidi ; SALES, forms of vendidi, quietum clamavi, concessi, forisaffidavi, and confirmavi ; AGREEMENTS, forms of convenit, or conventio and compositio). In order to confirm the totality of dispositive words and phrases concomitant with record types, those words and phrases are marked up and lists of patterns established for each type.

In addition to the words of disposition, Markup is used to distinguish the other diplomatic divisions and subdivisions of a records. Thus, the standard charter composition of PROTOCOL, CORPUS, and ESCHATOCOL is further broken down such that the parts of the PROTOCOL may be identified as the invocatio, intitulatio, and salutatio ; the CORPUS according to the divisions of specific record types (generally the promulgatio or notification, the narratio or background, the dispositio or words of action, and the conditional clausulae) ; and the sections of the ESCHATOCOL as the sealing clause, the witness list, the place of issue and the date. With the record number as the central point of reference, these segments are used in subsequent analysis to confirm and distinguish record types, and within each type to identify chronological changes in terminology which in turn point to developments in the spheres of legal, social and economic activity.

Before being stored in the database, the printed charter texts are collected using a scanner, processed by photo-imaging software, converted to the Vector format with an OCR programme, held temporarily and manipulated in Word Processing programmes, and finally processed by the DM. The DM creates a structured document that contains all the available information related to a specific record in such a way that it appears as a single entity in which all related information is wrapped around the text core. The text core is in turn broken down into a content model allowing the text to be treated as a hierarchy of content parts and subparts. Each record is described in XML terms. What is eventually stored in the database is nothing more than an XML document composed of a structured text core and attributed data.

Once edited, marked up and cast into a structural model, the data is ready for exploratory analysis. The DEEDS Project has developed specialized programmes for generating data files compatible with industrial Statistical Software packages (SPSS and S-Plus). These are used to respond to queries issued against the corpus of records stored in the database. Statistical software permits the application of advanced statistical algorithms to the output of the queries and obtains results in the form of standardized charts and reports. Research results are published initially in Intranet. Plans are under way to make them available to our colleagues around the world on the Internet. The DEEDS Project has also developed programmes for evaluating the probable date of a given charter based on analyses of word-pattern frequency.

The research project employs an array of software programmes, programming and mark up languages, and data management technologies. For the development of specialized software it uses Java, C, XML, SQL, PL/SQL, HTML, and JavaScript. Data is stored and managed on an ORACLE database. Research results are published over the Internet using a Microsoft Internet Information Server.

Summary of research intent and results

The primary objective of the DEEDS research project is to develop a methodology for the analysis of the medieval « letter » or charter using a Structural Content Model (SCM). Once established, that model can be used : to make chronological evaluations, to identify forgeries, to distinguish between or identify authors ; and to undertake a wide range of historical research based upon analyses of content.

Preliminary analyses confirm that chronological boundaries can be assigned to undated charters by comparing the frequency distribution of terms and expressions generated from a given undated charter against an existing Control Set (CS) of records whose attributes have known values. While presently consisting of 6,200 records, the CS established by DEEDS grows constantly as additional dated charters are identified and processed.

Hand in hand with the establishment of correct chronology is the identification of forgeries, since the purpose of forgery is to change past intent. DEEDS proposes to approach the identification of forgeries through the analysis of content, particularly in terms of words in context, their placement in the text, and the frequency of their usage. Sections containing words, phrases or concepts which appear to be outside their normal chronological context will be singled out as having possibly been entered by an « improver » or forger. Few forgeries entail complete records ; it is the component parts which are likely to have been tampered with.

By adding chronological, spatial, lexical and structural components to inquiries of this sort, researchers at DEEDS anticipate that it will be possible to improve existing methods of authorship evaluation. The result in the case of charters will be to identify the work of individual scribes, the date range associated with their activity, and their patron or patrons.

When the record attributes have been defined or verified, the SCM may be used for a wide range of subsequent legal, financial, military and social studies, as well as for matters concerning land use and ownership. Analysing undated charters is particularly significant as such records are the most important source for interpreting the social, political, economic and administrative history of the eleventh through thirteenth centuries. That history reflects the individual circumstances which determined changes in the construction of legal texts and the constant adoption, formulation and adaptation of words and word patterns. Put quite simply, changes in word usage and expression are an immediate reflection of social change. Only when the chronological and spatial context for a large number of charters has been established will it be possible to use them to determine when, and why, these changes took place.

A typical charter is a legal record, which has a logical structure and specific lexical, spatial, temporal and derivative attributes. The methodology we propose will enable us to verify or assign attributes based on a comparison with similarities appearing in the Control Set of charters. The SCM uses mark-up language which follows the Backus-Nour Form. It is based on a set of mathematical rules that describe the overall structure of the content of the whole, and all parts, of a typical record. Using SCM, a record can be conveniently parsed into a virtual tree-like structure for further computer-aided analyses.


SELECT BIBLIOGRAPHY OF THE DEEDS PROJECT

1978

M. Gervers, « Medieval Charters and the Computer: An Analysis Using Mark IV », in Computers and the Humanities, 12, p. 127-136.

1981

M. Gervers, The Hospitaller Cartulary in the British Library (Cotton MS Nero E VI). A Study of the Manuscript and its Composition, with a Critical Edition of Two Fragments of Earlier Cartularies for Essex, Toronto, Pontifical Institute of Mediaeval Studies (xx, 386 p., 6 pl., 12 diagr., 36 tabl., 9 maps).

1982

M. Gervers (ed.), The Cartulary of the Knights of St. John of Jerusalem in England, Secunda Camera : Essex, London, Oxford University Press for the British Academy, 1982 (lxxxiv, 735 p., 2 pl., 1 tabl., 2 maps).

1983

M. Gervers, with the technical assistance of L. C. Kordecki, A Microfiche Concordance of the Cartulary of the Order of St. John of Jerusalem, secunda camera (British Library, Cotton ms. Nero E VI, fols. 289-467), Toronto, Deeds Project (Scarborough Campus, University of Toronto).

1984

 L. C. Kordecki, «Computer techniques and medieval land transfers : the DEEDS project», in Journal of the Society of Archivists, 7/5, p. 299-311.

1986 

N. W. Alcock, Old Title Deeds: a guide for local and family historians, Chichester, p. 13-20.

1987 

M. Gervers, «The DEEDS Project and a Survey of the Essex Textile Industry in the Twelfth and Thirteenth Centuries », in History and Computing, ed. Peter Denley and Deian Hopkin, Manchester University Press, p. 81-89.

1990

M. Gervers, G. Long & M. McCulloch, « The DEEDS Database of Mediaeval Charters: design and coding for the RDBMS ORACLE 5 », in History and Computing, 1 no. 3, p. 1-12.

1992

M. Gervers, « Donations to the Hospitallers in England in the Wake of the Second Crusade », in The Second Crusade and the Cistercians, ed. M. Gervers, New York, St. Martin’s Press, p. 155-61.

1996

M. Gervers (ed.), The Cartulary of the Knights of St. John of Jerusalem in England, part 2, Prima Camera, Essex, London, Oxford University Press for the British Academy (cxii, 324 p., 1 pl., 1 map).

1997 

M. Gervers & R. Fiallos, « The Dating of Medieval English Private Charters of the Twelfth and Thirteenth Centuries », in A Distinct Voice : Medieval Studies in Honor of Leonard E. Boyle, O.P., ed. Jacqueline Brown & William P. Stoneman, Notre Dame (Indiana), p. 455-504.

2000 

M. Gervers (ed.), Dating Undated Medieval Charters, Woodbridge, Sf. – Rochester, NY, Boydell & Brewer, and Budapest, Collegium Budapest / Institute for Advanced Study (237 p.).

M. Gervers, «Introduction » to Dating Undated Medieval Charters, ed. M. Gervers, Woodbridge, Sf – Rochester, NY – Budapest, p. 1-12.

M. Gervers, «The DEEDS Project and the Development of a Computerized Methodology for Dating Undated English Private Charters of the Twelfth and Thirteenth Centuries », in Dating Undated Medieval Charters, ed. M. Gervers, Woodbridge, Sf – Rochester, NY – Budapest, p. 13-36.

2001 

M. Gervers, « Changing Forms of Hospitaller Address in English Private Charters of the 12th and 13th Centuries », in Expanding the Frontiers of Medieval Latin Christianity. The Crusades and the Military Orders, ed. József Laszlovszky and Zsolt Hunyadi, Budapest. In press.

M. Gervers, «The Hospitaller and Templar Commandery as an Economic Unit in England », in La Commanderie, institution des ordres militaires dans l’Occident médiéval, Paris, CNRS. In press.



The author is indebted to Michael Margolin (Programmer) and Gillian Long (Project Manager) for their assistance in the preparation of this report.

[1]. J. Horace Round (revised and completed by W. R. Powell), « St. Thomas Becket’s sisters and other studies », Essex Archaeology and History, 31, 2000, p. 154-160 spec. p. 156.

[2]. F. M. Stenton (ed.), Transcripts of Charters Relating to the Gilbertine Houses of Sixle, Ormsby, Catley, Bullington, and Alvingham, Publications of the Lincoln Record Society for 1920, 18 (Horncastle, 1922), p. x-xi, xxxiii-xxxiv.

[3]. See Bibliography, 1983


 

© CNRS - IRHT 2004 - Contacts - Infos légales - Rédaction