Scribes as Agents of Language Change 9781614510543, 9781614510505

Citation preview

Esther-Miriam Wagner, Ben Outhwaite, Bettina Beinhoff (Eds.) Scribes as Agents of Language Change

Studies in Language Change

Edited by Cynthia Allen Harold Koch Malcolm Ross

Volume 10

Scribes as Agents of Language Change

Edited by Esther-Miriam Wagner Ben Outhwaite Bettina Beinhoff

ISBN 978-1-61451-050-5 e-ISBN 978-1-61451-054-3 ISSN 2163-0992 Library of Congress Cataloging-in-Publication Data A CIP catalog record for this book has been applied for at the Library of Congress. Bibliografische Information der Deutschen Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutschen Nationalbibliografie; detailed bibliographic data are available in the internet http://dnb.dnb.de. © 2013 Walter de Gruyter, Inc., Boston/Berlin Cover image: iStockphoto/thinkstock Typesetting: PTP-Berlin Protago-TEX-Production GmbH, Berlin Printing: Hubert & Co. GmbH & Co. KG, Göttingen ♾ Printed on acid-free paper Printed in Germany www.degruyter.com

Acknowledgements The “Scribes as Agents of Language Change” conference that initially brought together the scholars who produced this book was born at another sociolinguistics conference in Oxford, organised by the coptologists Jennifer Cromwell and Eitan Grossman. In subsequent discussions, Eitan was instrumental in conceptualising the aims and objectives of the Scribes conference, and this book – and the success of the conference – owes much to his initial vision. We should like to extend a big thank you to all the delegates and participants at the Scribes conference for contributing to this intellectual endeavour and for making it such a great success. In addition, Laura Wright and the participants in the Sociolinguistics Seminar Series at Cambridge, in particular Lesley Milroy, Clive Holes and Peter Trudgill, have also provided insightful observations. We are grateful for the financial and administrative support, in particular from Helga Brandt, that we received from the Centre for Research in the Arts, Social Sciences and Humanities (CRASSH) at the University of Cambridge. Thanks are due to the Master and Fellows of St John’s College, Cambridge, not only for hosting the conference, but also for their generous financial support. We should like to thank Peter Matthews, who was instrumental in the search for suitable speakers and in gaining financial assistance from the college; Mark Nicholls, who supported us with money from the presidential chest and helped with the general organisation; Richard Beadle and Robert Tombs. The St John’s Catering and Conferences Department created a fantastic atmosphere for us to share ideas in. Financial support was also received from the Genizah Research Unit, and the staff of the Unit must be collectively thanked for their assistance to the organisation of the conference and the creation of this book. The work of the Unit is supported by the Arts and Humanities Research Council, UK, and the Friedberg Genizah Project. Mouton de Gruyter have been inspirational publishers to work with, and our thanks go to Emily Farrell, Jennifer Carlson, Lara Wysong, and our production editor Wolfgang Konwitschny.

Contents Acknowledgements 

 v

Part I: Introduction 

 1

1

Esther-Miriam Wagner, Ben Outhwaite and Bettina Beinhoff Scribes and Language Change   3

Part II: From spoken vernacular to written form 

2

3

 19

Mark A. Williams Biblical Register and a Counsel of Despair: two Late Cornish versions of Genesis 1   21 Markus Schiegg Medieval Glossators as Agents of Language Change 

 39

4

Roger Wright How scribes wrote Ibero-Romance before written Romance was invented   71

5

Anja Busse Hittite scribal habits: Sumerograms and phonetic complements in Hittite cuneiform   85

Part III: Standardisation versus regionalisation and de-standardisation 

6

7

8

 97

Terttu Nevalainen Words of kings and counsellors: register variation and language change in early English courtly correspondence   99 Florian Dolberg Quantifying gender change in Medieval English 

 121

Merja Stenroos Identity and intelligibility in Late Middle English scribal transmission: local dialect as an active choice in fifteenth-century texts   159

viii   

9

   Contents

Ben Outhwaite Lines of communication: Medieval Hebrew letters of the eleventh century   183

Geoffrey Khan 10 The historical development of early Arabic documentary formulae  Nicholas Zair 11 Individualism in “Osco-Greek” orthography 

 199

 217

Stefan Reif 12 How a Jewish scribe in early modern Poland attempted to alter a Hebrew linguistic register   227 Part IV: Idiosyncracy, scribal standards and registers 

 239

Alexander Bergs 13 Writing, reading, language change – a sociohistorical perspective on scribes, readers, and networks in medieval Britain   241 Esther-Miriam Wagner 14 Challenges of multiglossia: scribes and the emergence of substandard Judaeo-Arabic registers   261 Ivar Berg 15 Variation in a Norwegian sixteenth-century scribal community 

 277

Dmitry Bondarev 16 Language change induced by written codes: a case of Old Kanembu and Kanuri dialects   291 Index 

 325

Part I: Introduction

Esther-Miriam Wagner, Ben Outhwaite and Bettina Beinhoff

1 Scribes and Language Change 1 Scribes as agents of language change The present volume gathers scholars working on the scribal cultures of a wide range of languages, from the earliest written to later medieval languages, and covering most of the major language families. What brings them together is a shared interest in the potential for sociolinguistic investigation of historical textual sources, to improve our knowledge of how language change can be driven through writing – specifically through the engine of the scribal pen (or reed or quill or stylus) – rather than speech alone. In particular, the contributors investigate how different communities of scribes have promulgated standard and substandard registers of languages, and examined how sociolinguistic methods can be used to determine the spread of linguistic innovations within such scribal networks. Far from simply making “the best use of bad data” (Labov 1994: 11), this volume seeks to draw attention to the importance of written sources for the investigation of historical sociolinguistics, and to calibrate appropriate tools to discern how language change occurs within the written register. Scribal evidence is of great value for the study of historical sociolinguistics as it is often the sole reliable source, if not the only source, of evidence for diachronic language change. Observing language change over time, we can see that scribes are both active agents of change but also have a more passive role when – perhaps unknowingly and unintentionally – documenting variation or developmental tendencies and patterns in language. Scribal evidence can also shed light on mechanisms of language change that would otherwise remain obscure, as for example when revealing the mutual influence of co-existing codes (see Berg¹ and Schiegg), or when showing how rapidly language change can take place (Dolberg). Thus, studying scribal evidence is an attempt at making visible the “invisible hand” (Keller 1994) of language change. At this point we should recognise that in using the word “scribe”, we risk conjuring up the image of ancient Egyptian or medieval monastic artisans, carefully copying texts on stone, or with a reed or quill pen. And while our definition of scribe does encompass these early masters of the scribal arts, we are using

1 All italicised names refer to articles in this volume.

4   

   Esther-Miriam Wagner, Ben Outhwaite and Bettina Beinhoff

the term to denote all producers of texts, those who have given us the written evidence of their languages. In the earliest cultures where writing emerged, it was indeed a craft and we know of the existence of scribal castes and families, for instance in ancient Babylonia (see the lemma “Schreiber” in Reallexikon der Assyriologie 12: 251). But we should be careful, when assessing the history of written texts, not to generalise too much from these precious early extant sources, which are usually model texts or ceremonial inscriptions produced by a relative elite. For example, most of our knowledge about Sabaic, a Semitic languages spoken and written in Yemen, until recently had come from royal inscriptions on stone and dedication plaques produced by professionals, before a surprise discovery of hundreds of letters and other documents written on wood completely changed the size and nature of our corpus, and thus utterly altered what was previously assumed about the written culture of Ancient South Arabia (see Stein 2010). Similarly, the oral culture of the Arabian desert has in fact left behind a sizeable written record in the form of scattered graffiti on rock outcrops, showing that a number of people in this overwhelmingly oral culture were willing and able to produce crude texts (Macdonald 2005). Thus we should not discount the value of all written sources for the textual history of a variety of cultures. In exclusively focussing on the producers of high-status texts, we are in danger of ignoring a substantial body of work by the everyday writers, the scriveners of utilitarian prose, or the scribblers of graffiti: indeed, the majority of texts in most societies are probably “utilitarian prose”. All of these individuals worked within their own literary contexts, as a part of the wider written culture of their societies. Thus,“scribes” in the context of historical linguistics should denote the literate, that is, the actively literate, those who write, and not just those who write for a living. The contributions in this volume touch upon a great number of reasons for language change, especially in situations of diglossia, bilingualism and wherever codes co-exist. Strictly speaking, “every language is a multiplicity of codes” (Croft 2000: 92) given that every single language has a variety of registers which interact to varying degrees. This interaction can lead to mutual influence of these registers, for example in communities where individuals are fluent in both the high and low variety of a language and use those in different social situations (see also Bondarev). By its very nature, scribal evidence is set within the written register; however, because registers tend to interact, the written language can provide some insight into the spoken register of the time. Socially relevant variation may not be as easily accessible in writing as it is in speech (see Nevalainen), but some of the contributions in this volume show how the spoken language can be reflected in writing (see Wright and Wagner).

Scribes and Language Change   

   5

It is reasonable to assume that scribal evidence can show attempts at standardising or inflicting norms on a language. In the Jewish community of Fustat, Egypt, for example, the knowledge of reading was fairly common, but as in the older Middle Eastern civilisations of Assyria, Babylonia and Ancient Egypt, writing was an art acquired by persons who had special reason to do so (Goitein 1971: 228). Scribes were, therefore, at times a small, literate minority in the wider society and could thus be assumed to act within a relatively restricted network through which traditions and norms were passed on in daily practice or during the process of learning and teaching how to read and write. This was especially the case in the Middle Ages and before. Thus, scribal tradition seemed to operate more with spontaneously evolving norms rather than relatively fixed standards (see Goossens 1994 and Berg). Depending on the size and density of these networks, certain written norms may have been preserved for longer than the spoken forms were in use (following the network approach of Milroy 1987; for a discussion of written vs spoken forms see Wright). At the same time, scribes may well have been torn between norms of writing and reflecting actual speech. For example, Bergs argues that in the dictation of letters, for instance, scribes may have influenced the outcome so that the resulting document may be more representative of the scribe rather than of its ostensible author. The latter point indicates that individual scribes may have had considerable influence on language change; either by adhering to traditions or by introducing innovations. These decisions were then, as they are today, often driven by prevailing attitudes and linguistic ideology (see for example Williams, Reif and Steenroos). The scribal evidence presented in this volume reveals how scribal groups or even individual scribes have in the past contributed to language change. Historical sociolinguistic research has thus far neglected the role of the individual in language change (for a detailed discussion see Thomason 2000 and 2007), though many of the contributions in this volume show that it is an area well worth further investigation. Although the papers presented here draw upon a wide array of languages and cultures, there are clear parallels in the linguistic investigations, and three major points emerge as the main foci of interest. Firstly, the various processes are explored through which scribes became involved in the emergence of the written form of languages, i.e., the mechanisms of Verschriftung/Verschriftlichung, whereby a vernacular is put into writing for the first time. This scripting of a spoken language may be purposeful and ideologically driven, a more or less accidental process, out of practical need, or can arise from employing older or even archaic forms of a written language that then can come to represent in writing a newly emerging vernacular.

6   

   Esther-Miriam Wagner, Ben Outhwaite and Bettina Beinhoff

The second main area of discussion concerns the manner in which linguistic standardisation took place and the means by which accepted forms of the language were created, adopted or maintained by scribal groups. This raised further issues such as the deliberate introduction of regional forms, and de-standardisation of written languages in general. Lastly, the activities of individual scribes and their schools are clearly of great interest, and the contributors from across the whole spectrum of languages and time-periods discuss how the idiosyncrasies of individual scribes are discernible in medieval manuscripts, how particular scribal schools set language standards, and how specific scribal registers developed. These often highly-artificial language registers are created in the enforced mixing of a spoken vernacular and existing older, literary writing conventions, and appear inevitably to result in the use or creation of orthographic, morphological or syntactic forms that do not naturally belong to either of the two parent registers.

2 From spoken vernacular to written form Several avenues for the emergence of written forms of a language in the process of the Verschriftung (or Verschriftlichung) – the transcoding of phonetic information into graphic symbols – are examined through a number of examples in specific languages. Whereas purposefully-designed written forms of a language can be created by an individual or a group of people to enable speakers to translate works from another language into their own (for instance Cyrillic), in other cases, as shown by Williams’s description of the codification of the Brythonic language Cornish, a writing system is solely – and belatedly – developed to preserve an endangered language on the brink of extinction. In other languages, there is a quasi-accidental, possibly slightly more organic, process than the design of a language form on the drawing board, whereby it emerges from an immediate and practical need. Notes taken in a scribe’s own native tongue have led to the eventual development of a literary language, as Schiegg shows through the example of Old High German, the written form of which originated in marginal glosses. Schiegg describes how 8th-century Frankish monks used their styli to scratch words in vernacular German in the margins or between the lines of Latin manuscripts, either to annotate demanding passages of the parent text or, at times, it appears simply to amuse themselves. He examines two aspects of how these glossators served as promoters of language change. Firstly, such “dry-point glosses” are the first texts to be written in Old High German. The glossators’ use of their vernacular began a process that resulted in a substantial

Scribes and Language Change   

   7

change in the written language usage in medieval northern Europe: the eventual replacement of Latin with Old High German. Secondly, glossators – writing either for themselves or for a restricted and local audience of fellow scribes – are promoters of progressive language forms, employing an idiom of conceptional orality that cannot be encountered in other text types until centuries later. Glosses fulfill an important role in the emergence of language forms; the annotation of an older writing system is a phenomenon that has significantly contributed to the codification of various languages. In Korean, for example, a writing system was developed to aid the reading of Chinese texts, and the earliest written form of Basque glossed Latin texts. For the glossator, there is an immediate and practical need to put their spoken language into writing, and at the same time, the private context inherent to the act of glossing provides the reader/writer with a setting intimate enough to mount the threshold and commit their language to parchment or paper for the first time. Another interesting phenomenon associated with Verschriftung can be discerned in the use of older language forms to write the vernacular. The gradual emergence of the Ibero-Romance languages from Vulgar Latin is obscured by the peninsular scribes’ continued use of Latin, even after it ceased to be a living, spoken language, but, as Wright demonstrates in his study of a document from the monastery of Vairão, Portugal, the disguise is a thin one, and the underlying vernacular shows through. In the Vairão text, the morphology, the semantics, the syntax and the vocabulary are those of the speech of the 10th-century Kingdom of Galicia. The only linguistic aspect of the text that is not Ibero-Romance is the spelling, the written form of the words, because the scribe, bound by traditional practice, was not trying to reproduce the phonetics of his speech in a direct transcription but using the scribally-correct forms of orthography that he had been taught. In contrast to the view of many Latinists, who tend to dismiss these texts as “bad” or “corrupt” Latin, written by ill-trained scribes with astounding ignorance, incompetence and stupidity, Wright shows that the scribes in fact worked extremely intelligently in a difficult and evolving context where there was an everwidening gap between the written language that they had been taught and their everyday spoken tongue. He points to marked parallels in English orthography, where spellings have greatly diverged from spoken English pronunciation over the last five centuries, and it is easy to find other languages where the orthography mirrors much older strata of the language, such as, for instance another Romance language, French, or even Chinese, whose script is “indifferent to the phonemic constitution of [its] phonemes”.² Wright’s example of scribal practice

2 Haas (1982: 21).

8   

   Esther-Miriam Wagner, Ben Outhwaite and Bettina Beinhoff

in Galicia exemplifies how changes in the spoken language may not initially be represented by matching evolutions in the written mode, and how a traditional written form may be a vehicle for an underlying vernacular. A similar process is evident in the Jewish version of Middle Arabic, JudaeoArabic (see Wagner). Normative Arabic grammars of the 8th/9th century compiled and propagated “acceptable” forms of the written language, a prescribed register based upon the idiom of the Qur’an. From a religious imperative, Muslim Arabic speakers aspired to write a standard based on this Classical Arabic register, although from the very first centuries of Islam the vernacular had been markedly diverging from this written norm, through the rapid language change brought about by the migrations and integrations of the Islamic conquests. Jews, however, had no such religious duty and were not bound to the standard Muslim Arabic; instead they developed their own written language forms. Judaeo-Arabic, therefore, employs elements from the standard written language almost as logograms to represent vernacular forms, in a similar fashion to Wright’s Galician scribe. To give just one example, an uninflected version of the normally inflected Classical Arabic relative pronoun allaḏī is used to represent the uninflected vernacular illi, thus combining conservative orthography with vernacular morphology and syntax. Alongside the example of this Arabic relative pronoun it also becomes evident that the contrast written vs spoken does not suffice to describe the complex communication situation of languages. The Tiberian vowel signs in the few rare vocalised letters suggest a pronunciation of this relative pronoun allaḏī/illi as äldi. This äldi is a mixed form found in medieval Judaeo-Arabic that belongs to the inventory of neither the Classical base language nor the spoken vernacular but that is part of the reading tradition of documents. As Fleischer (2009: 162–164) has stated, reading a text in ancient times and in the early Middle Ages meant reading it aloud, whereas silent reading was only a later development.³ The reading tradition was therefore a third component of language beside spoken and written varieties. Wright has shown this phenomenon in his Galician texts; the writer wrote “Latin” but when that text was read aloud it was recited in Ibero-Romance. Similar reading traditions can be found in a large number of Semitic languages: Syriac texts are read out in modern Neo-Aramaic dialects in today’s Christian Aramaic communities.⁴ Or, ancient Hebrew is read out with the help of vocalisa-

3 Indeed, the necessity of reading a text aloud gave impetus to the increased use of Hebrew in the medieval Near East (Outhwaite). 4 Compare Khan (2007: 103).

Scribes and Language Change   

   9

tion systems developed between the 7th and the 11th centuries.⁵ In these cases, writing systems belonging to the older stages of a language are used to codify later forms of the same language. In other cases, though, we can find an even more extreme phenomenon, in that one language and its corresponding writing system are used to write another language, perhaps from an altogether different language family.⁶ For example, Classical Chinese texts can be read in Japanese with the help of extralinear symbols.⁷ Texts written in administrative Aramaic were read out in Persian at the Persian courts.⁸ An intriguing parallel is described by Busse concerning the Verschriftung of the ancient Indo-European language Hittite, where an older writing system, Sumerian-Akkadian (itself a hybrid system), is adapted by official scribes to the needs of the contemporary chancery language. In this case, Sumerian logograms are used to write Hittite words, and would, in the reading tradition, have been read and understood as their corresponding Hittite forms. One of the more astonishing results of Busse’s research shows that writing verbs with sumerograms becomes even more common in the younger (New Script) stages of Hittite script. We thus find a conservative written feature growing more dominant as time goes on. Busse presumes that these spellings may have been introduced into other registers from long-term records, i.e. writings that were kept and recopied for a long time, such as ritual and omen texts, which may have influenced scribal habits due to their deeply entrenched prestige.

3 Standardisation versus regionalisation and de-standardisation According to Vachek (1982: 45), “in absence of a written norm, a language falls short of its optimal development”. Although the term “optimal” may prompt discussion and require clarification, it is certain that the Verschriftung of a language induces a process of further language development. One of the most noticeable impacts is that written language normally serves as a vehicle of standardisation. Standardisation is defined by Milroy and Milroy (1999: 26) as “intolerance of optional variability in language”, and is normally “motivated in the first place by

5 See Khan (2012), in particular Chapter 6. 6 This phenomenon was termed alloglottography by Gershevitch (1979). 7 See Lurie (2012). 8 See Gershevitch (1979), in particular p. 143.

10   

   Esther-Miriam Wagner, Ben Outhwaite and Bettina Beinhoff

various social, political and commercial needs” (1999: 22). In the case of written language, it is probably foremost driven by a desire to disambiguate language. Standard language should be regarded more as an idea than a reality, a prescribed set of norms, driven by “prestige”, selected and accepted by a certain set of individuals, before being subject to geographical and social diffusion (Milroy and Milroy 1999). While a central authority may play an important role in the process of codification, the role of scribes in the creation of standardised languages should be emphasised, Bible translators being a particular case in point. Translations and commentaries of the Holy Bible have been credited with the creation of a number of standard written languages, for example Wulfila’s Gothic Bible, Saadiah ben Joseph’s Tafsīr (translation/commentary on the Hebrew Bible), which drove the emergence of a standardised Judaeo-Arabic, the Luther Bible’s shaping of New High German, and the powerful and longlasting influence of the King James Bible on Modern English. Beyond the Bible, and in the absence of a need for translation of basic religious texts, other major literary works may serve as role models for standardisation. One may think of the centrality of Confucius’s works to the imperial Chinese educational system. Prestigious literary works can shape and change written language forms for generations to come, simply by setting a dominant precedent. The English language provided much of the focus for the question of standardisation processes, such as Nevalainen’s investigation of the underlying causes of the conservatism found in 16th-century correspondence. That the speed of standardisation processes can be much faster than has been anticipated has been shown by Dolberg, who examined the rapidity and directed nature of language change evidenced at the seam of Old and Middle English (12th century). He demonstrates that a language feature, in this case referential gender agreement, develops from a minor variant to the dominating form within one generation only. While this is known to happen in situations of language contact-induced language shift (compare the abrupt creolisation described in Thomason & Kaufman 1988), this has not yet been observed for language standardisation. Stenroos’s study of several 15th-century sources raises a number of important questions with regard to the whole question of “standardisation”: was it simply a lack of regional variance or something with more ideological underpinnings? She discusses general ideas about the standardisation of English, and points out critically that those who focus solely on geographical variation may consider written English to be largely “standardised” by 1500, while in fact the written language continues to be highly variable well into the 17th century. There is a tendency among scholars to relate variation in this period to an ongoing process of standardisation, often interpreting forms that are identical to those that survive in later standard English as “progressive” and other forms as “conservative” or “reces-

Scribes and Language Change   

   11

sive”. Stenroos argues that such a perspective may result in a skewed view of the actual variation, and that terms such as “standard” and “standardisation” with their social and educational connotations, may be misleading when applied to late medieval materials. She shows that regional usages were selected by highly competent writers, and actively chosen and preferred throughout the 15th century. Modern conceptions of “standard” usages, with their social and educational connotations, may therefore be misleading when applied to late medieval materials. Regionalisation is often discussed in connection with de-standardisation (Schmidlin 2011: 64).⁹ According to Ferguson (1988: 120–123), de-standardisation processes should not be seen as separate and necessarily diametrical to language standardisation but rather as embedded within general standardisation processes. Mattheier (1997: 1), following the lead of Daneš (1982), has pointed out that in the modern era two developmental tendencies may lead to de-standardisation, which can easily be applied to the situation obtaining with Stenroos’s Middle English scribes. Firstly, we find that a loss of prestige in the codified standard language along with decreased loyalty to the authority that issues language standards leads to the emergence of new registers. This is the case in 11th-century Palestine where the gradual disappearance of the traditional centres of Jewish religious governance led to a concomitant decline in the use of Hebrew in favour of Arabic by Jews (Outhwaite). Secondly, there is an increase in the prestige of a substandard register, in the case of Mattheier and Daneš the spoken varieties, in the case of Stenroos’s Middle English scribes the regional forms. The prestige of a given variety appears to be the key element shaping the use of registers, both in spoken and written varieties. The standards set by particular scribal schools may be influential for the language as a whole. For Arabic, Khan argues that a number of developments that occurred in the 9th century CE were due to the impact of the appointment in Egypt of administrators from the eastern Islamic provinces. This core of eastern scribes led the change towards cursive writing, possibly influenced by Pahlavi writing style, and a stylistic shift in the Islamic chancery from the original Arabic legal formulae (simple, objective) to a more Byzantine style of document composition (complex, subjective) in the 3rd Islamic century. Together with an influx of Aramaic loanwords, this points toward an origin for the newly emerging chancery standard in the scribal schools of the East. Khan’s analyses show that it is not necessarily the seat of power that influences standards. Instead, the origin of writers, in this case the importation of an assemblage of scribes from the eastern periphery of the empire, trained in centres with a long-established scribal tradition,

9 A good discussion of the term de-standardisation can be found in Auer (1997: 135–140).

12   

   Esther-Miriam Wagner, Ben Outhwaite and Bettina Beinhoff

may be a decisive factor. Stenroos’s research on English regionalisation points to similar conclusions; scribes from regions at the periphery of an empire may impose their own regional standards, and if historical circumstances support the prestige of these forms, they can easily spread. In the absence of normative literary works or a leading administrative centre that, through its prestige, enforces specific scribal practices, considerable individual variation can be found, as Zair has shown with the example of Oscan, a member of the Italic family. Oscan writers are not following any kind of externally mandated orthographic rules, but are instead simply relying on their intuitions, based on orthographical conventions of Greek, as the best way to spell a given sound in a given context. Decisions regarding how to write seem therefore to be to the largest part an individual choice. Although widespread changes are usually the result of a concerted group or an extended network, it is not beyond the means of one, ideologically-driven, individual to effect language change. One man’s drive to put his personal stamp on the written language is chronicled in Reif’s contribution, which demonstrates how Shabbethai ben Isaac Sofer tried to mould the language of contemporary liturgy, a particularly entrenched and conservative idiom. Despite this, Shabbethai, a 16th/17th-century Polish Jewish scribe, succeeded in leaving his impact on the language of the traditional prayers recited by many communities of Jews all over the world.

4 Idiosyncracy, scribal standards and registers The issue of regionalisation, as demonstrated by Stenroos, also shows how individuals or groups of writers may aspire to leave their own stamp on written norms, by introducing forms that mark a register as regional. Other writers set precedents and introduce new forms that may eventually gain prestige and become standard forms at a later point. But interestingly, idiosyncracies themselves cannot often be labelled as either straightforwardly progressive or conservative. Nevalainen, for example, has pointed out that individual English letter-writers may adopt some progressive forms, but shun others. So the language often displays an interesting mix of conservative and progressive forms, and her results indicate that “individuals vary in the degree to which they participate in ongoing linguistic changes”. She remarks on a sociolinguistic phenomenon that is not entirely unexpected, that upwardly mobile administrators are less likely to display innovative linguistic behaviour than their contemporaries, including even the royalty in whose service they were working. Her most obvious example

Scribes and Language Change   

   13

is Thomas Cromwell (d. 1540), who rose from humble beginnings to become an influential courtier and who writes in a most conservative style. This behaviour is very typical for upwardly mobile people and occurs in contemporary society, as shown, most famously, by Labov (1972) and Trudgill (1974). Nevalainen confirms that the language of rulers, government officials and secretaries is more conservative and resistant to change than those of individuals in a less official capacity, in this case merchants. A clear parallel may be seen in the Jewish communities of the East in the period leading up to the mid-11th century, where those in official positions at the Jewish courts and academies of Palestine and Egypt chose to write the great majority of their considerable correspondence in the traditional language of the Jews, Hebrew, while the wide-ranging mercantile class adopted wholeheartedly the newer idiom, the written form of their vernacular, Arabic, for all their correspondence (Wagner and Outhwaite 2012). Nevalainen also describes how English royal secretaries and government officials both maintained and created register variation, and transmitted professional repertoires, all of which was reflected in their participation in the linguistic changes occurring over time. Secretaries and officials typically do not follow their monarch’s (in this case Henry VIII’s or Elizabeth I’s) private usage, which can be found in the monarch’s own holographs, from which we can infer that professional scribes were very much aware of internal standards. Writing styles, however, may not only vary from individual to individual, but within one individual’s oeuvre. Among authors of written texts in medieval Fustat, variation was often more the norm than an exception. Even the most prominent of scribes might spell the same word in two adjacent lines differently (Goitein 1971: 237). It is further interesting to see how the individual writer might demonstrate variation over time, as a result of the influence of his peers, for example. Nissim b. Ḥalfon, an Egyptian Jewish merchant writing in the 11th century, uses the less common spelling of ׄ‫ ט‬for writing Arabic ẓ in his earlier Judaeo-Arabic letters, but, as time goes on, he shows a preference for the spelling of ẓ with the Hebrew letter ׄ‫צ‬, a spelling that is used by the overwhelming majority of his contemporaries, and, indeed, those who correspond with him (Wagner 2010: 30). Bergs emphasises how scribes can show strikingly divergent linguistic behaviour when writing for different authors. A young man writing for an older female relative may use a markedly conservative form of language that he deems appropriate for her letters, whereas in a letter written for himself he will employ more progressive forms. The status of the addressee is also crucial: the level of formality in correspondence is dictated by the social distance between writer and reader. We can perhaps extrapolate from this that any set of two interlocutors assumes a slightly distinct register in discourse.

14   

   Esther-Miriam Wagner, Ben Outhwaite and Bettina Beinhoff

Reflections on linguistic behaviour being subject to the relationship between writer and reader emphasise the means employed to reduce the distance that the act of letter writing inevitably engenders as opposed to oral conversation. For example, in Arabic correspondence an author may choose to switch to a more colloquial style of language to create an intimacy between writer and reader. This is one of the factors that drives the occurrence of socially less acceptable registers of so-called “Mixed” or “Middle Arabic”.¹⁰ Wagner calls this “language of intimacy”, following Davies (1981), and shows that colloquial expressions in JudaeoArabic letters are designed to appeal to the addressee on a distinct level in order to establish a closer relationship. Schiegg refers to the same phenomenon when scratched by Frankish monks as a “language of immediacy”, following Ernst and Elspass (2011), and with the example of early medieval manuscripts from Germany he demonstrates how Latin remained the language of distance while the newly emerging vernacular became the “language of immediacy”. Wagner also suggests that merchants fulfill an important role in the setting of linguistically innovative precedents and in the development of registers. The matter of specific scribal registers is worthy of investigation. Written registers ordinarily emerge from spoken languages but over time tend to preserve a more conservative state of the language. It has also been suggested that spoken varieties become more complex (Chambers 2009), intensive and extensive (Oesterreicher 1993) in the course of their Verschriftung. Chambers (2009: 19) states that “the continuum from vernacular to standard is marked by definable layers of increasing structure-dependent grammatical devices and articulated phonological contrasts”. A frequently-quoted example of this phenomenon is the use of certain subordinate structures in writing, which may be much rarer or even entirely absent in the spoken language. Oesterreicher (1993: 18–21) explains the necessity of linguistic development for written languages to meet the requirements of Distanzsprachlichkeit, such as the widening of the communicative functions of idioms, lexical differentiation, and the development of syntacticalintegrative, complex means of expression. In addition to this complexification of written language, further developments can be observed that have little or nothing to do with spoken language forms. Thus within the written register, artificial forms that were never part of the parent vernacular may arise, spread and become productive within specific text genres.

10 The most comprehensive introduction to the intricacies of Middle/Mixed Arabic is the conference volume “Moyen arabe et variétés mixtes de l’arabe à travers l’histoire”, edited by Lentin and Grand’Henry (2008).

Scribes and Language Change   

   15

An example of such an artificial form is the Middle English article θeo referred to by Stenroos, which is neither part of the vernacular repertoire nor of literary traditions but a scribal creation to mark the regional register of the Longleat manuscript. This resembles phenomena occurring in certain Late JudaeoArabic texts (Wagner), where specific lexemes are used to add a literary sheen to otherwise utility prose. For example, the negation lam is used in standard Arabic to negate the past and can only be combined with a particular form of the verb, the apocopate. It is a literary form, not found in the vernaculars, but, owing to this prestige, it becomes the epistolary negator of choice in Judaeo-Arabic correspondence of the 18th and 19th centuries, having considerably widened its function to include negation of the past, present and future, and, among other roles, as a nominal negation particle. Its widespread use probably grew from its markedness as a literary form, a “classical” lexeme which was an easy element to add to a letter to lend a level of prestige to the writing. Ultimately it became so widespread, that it eclipsed all negators, classical and vernacular, in Judaeo-Arabic letters. Because epistolary Late Judaeo-Arabic is grammatically and lexically quite removed from standard Arabic writing traditions, the employment of this peculiar mix of vernacular and literary elements creates a special register found exclusively in letters: prestige elements rubbing shoulders with colloquial forms, which preserve both a certain level of formality while also giving the correspondence a feeling of authenticity and intimacy. A similar phenomenon can be found in Norwegian, where, according to Berg, there is a systematic difference in the stylistic levels between “pure” Danish and a scribal language with Norwegian traits. Local Norwegian scribes developed a register located between the spoken Norwegian vernacular, on the one hand, and the familiar written variety, Danish, on the other. The scribes started using a mixture of both, thereby creating a new distinct hybrid idiom for less formal administrative writings. A similarly artificial register, Tarjumo (a variety of the West African language Old Kanembu), has been investigated by Bondarev. Used for recitation of Qur’anic interpretation, it consists of a complex combination of written and oral features. Bondarev also demonstrates how this written register in turn influences the spoken language.

5 Output The contributors of this volume repeatedly refer to the important act of a particular individual form or of a whole register reaching a “threshold”: once it has been committed to paper, parchment, papyrus or clay tablet, an innovative

16   

   Esther-Miriam Wagner, Ben Outhwaite and Bettina Beinhoff

written form acts as a precedent, acquiring a level of acceptability and respectability. This holds true not only for spoken languages put into writing for the first time (Verschriftlichung), such as the Old High German found scratched into the margins of monastic books (Schiegg). It applies similarly to specific phonological, morphological or syntactic forms that are entered into a pre-existing written mode. Judaeo-Arabic has many examples of this, whereby originally vernacular or regional forms have entered and remained a part of the standard idiom, for instance the dialectal bi- prefix of the imperfect verb, which makes an entrance into texts in passages of reported speech (marking the vernacular imperfect) but which subsequently may be found in much wider use, until it eventually enters the standard morphological inventory of epistolary Judaeo-Arabic (Wagner). Scribal novelties or dialectal peculiarities can thus, through their committal to writing, successfully enter the standard language, emerging as a form untainted by its previous use. On a more general level, the contributions have shown that scribes, through their output and their traditions, have provided copious reliable material on which to base further investigations into theories of language change. For example, Dolberg’s investigation proves how quickly referential gender agreement spread in 12th-century English. This finding provides significant evidence of how abrupt and sudden changes can occur in language standardisation processes. Other studies (and especially the contribution by Nevalainen) suggest that specific language behaviour, such as the use of conservative features by the upwardly mobile, has not changed much throughout the centuries. While written language does not allow for direct inferences about the everyday spoken language of the period, this is not the point: the great range of studies presented in this volume provide valuable insight into the history and development of written languages, for their processes of standardisation, and for the paramount position of scribes – whether working independently, together, ideologically or more casually motivated – as drivers of language change.

6 References Auer, Peter. 1997. Führt Dialektabbau zur Stärkung oder Schwächung der Standardvarietät? Zwei phonologische Fallstudien. In Klaus J. Mattheier & Edgar Rathke (eds.), Standardisierung und Destandardisierung europäischer Nationalsprachen, 129–161. Frankfurt/ Main: Peter Lang. Chambers, J.K. 2009. Cognition and the linguistic continuum from vernacular to standard. In Markku Filppula, Juhani Klemola & Heli Paulasto, Vernacular Universals and Language Contacts, 19–32. London: Routledge.

Scribes and Language Change   

   17

Croft, William. 2000. Explaining language change: an evolutionary approach. Harlow & New York: Longman. Daneš, František. 1982. Dialektische Tendenzen in der Entwicklung der Literatursprache. In Jürgen Scharnhorst & Erika Ising (eds.), Grundlagen der Sprachkultur. Beiträge der Prager Linguistik zur Sprachtheorie und Sprachpflege, vol. 2, 92–113. Berlin: Akademie Verlag. Davies, Humphrey Taman. 1981. Seventeenth Century Egyptian Arabic: A Profile of the Colloquial Material in Yūsuf al-Širbīnī’s “Hazz al-Quḥūf fī šarḥ Qaṣīd Abī Šādūf”. Berkeley: University of California at Berkeley PhD thesis. Ernst, Oliver & Stephan Elspaß. 2011. Althochdeutsche Glossen als Quellen einer Sprachgeschichte “von unten”. NOWELE 62/63. 249–283. Ferguson, Charles A. 1988. Standardization as a form of language spread. In Peter H. Lowenberg (ed.), Language Spread and Language Policy: Issues, Implications, and Case Studies, 119–132. Georgetown: GUP. Fleischer, Jürg. 2009. Paleographic clues to prosody? – Accents, word separation, and other phenomena in Old High German manuscripts. In Roland Hinterhölzl & Svetlana Petrova (eds.), Information Structure and Language Change. New Approaches to Word Order Variation in Germanic, 161–189. Berlin/New York: de Gruyter. Gershevitch, Ilya. 1979. The Alloglottography of Old Persian. Transactions of the Philological Society 77. 114–190. Goitein, Shelomo D. 1971. Mediterranean Society. The Jewish Communities of the Arab World as Portrayed in the Documents of the Cairo Genizah. Vol. II: The community. Berkeley: University of California Press. Goossens, Jan. 1994. Normierung in spätmittelalterlichen Schreibsprachen. Niederdeutsches Wort 34. 77–99. Haas, William. 1982. Introduction. On the normative character of language. In William Haas (ed.), Standard languages. Spoken and written, 1–36. Manchester: Manchester University Press. Keller, Rudi. 1994. On language change: the invisible hand in language. London: Routledge. Khan, Geoffrey. 2007. Aramaic in the medieval and modern periods. In J. Nicholas Postgate (ed.), Languages of Iraq, ancient and modern, 95–113. Cambridge: The British School of Archeology in Iraq. Koch, Peter & Wulf Oesterreicher. 1985. Sprache der Nähe – Sprache der Distanz. Mündlichkeit und Schriftlichkeit im Spannungsfeld von Sprachtheorie und Sprachgeschichte. Romanistisches Jahrbuch 36. 15–43. Khan, Geoffrey. 2012. A short introduction to the Tiberian Masoretic Bible and its reading tradition. Piscataway: Gorgias. Labov, William. 1972. Sociolinguistic patterns. Philadelphia: University of Pennsylvania Press. Labov, William. 1994. Principles of linguistic change. Vol. 1 Internal factors. Oxford: Blackwell. Lentin, Jérôme & Jacques Grand’Henry (eds.). 2008. Moyen Arabe et variétés mixtes de l’arabe à travers l’histoire. Actes du Premier Colloque International (Louvain-la- Neuve, 10–14 mai 2004). Louvain-la-Neuve: Peeters Press. Lurie, David. 2012. Realms of literacy: Early Japan and the history of writing. Cambridge: Harvard University Press. Macdonald, M.C.A. 2005. Literacy in an oral environment. In Piotr Bienkowski, Christopher Mee & Elizabeth Slater (eds.), Writing and Ancient Near Eastern Society: Papers in Honour of Alan R. Millard, 49–118. New York/London: T&T Clark International.

18   

   Esther-Miriam Wagner, Ben Outhwaite and Bettina Beinhoff

Mattheier, Klaus J. 1997. Über Destandardisierung, Umstandardisierung und Standardisierung in modernen europäischen Standardsprachen. In Klaus J. Mattheier & Edgar Rathke (eds.), Standardisierung und Destandardisierung europäischer Nationalsprachen, 1–9. Frankfurt/Main: Peter Lang. Milroy, James & Lesley Milroy. 1999. Authority in language. Investigating Standard English. London: Routledge. Oesterreicher, Wulf. 1998. Textzentrierung und Rekontextualisierung. In Christine Ehler & Ursula Schaefer (eds.), Verschriftung und Verschriftlichung, 10–39. Tübingen: Gunter Narr. Schmidlin, Regula. 2011. Die Vielfalt des Deutschen: Standard und Variation. Berlin: de Gruyter. Schreiber (‘Scribes’). Reallexikon der Assyriologie und der vorderasiatischen Archäologie 12. 273–280. Stein, Peter. 2010. Die altsüdarabischen Minuskelinschriften auf Holzstäbchen aus der Bayerischen Staatsbibliothek in München. Band 1. Die Inschriften der mittel- und spätsabäischen Periode. Tübingen/Berlin: Ernst Wasmuth Verlag. Thomason, Sarah Grey. 2000. On the unpredictability of contact effects. Estudios de Sociolingüística 1.1. 173–182. Thomason, Sarah Grey. 2007. Language contact and deliberate change. Journal of Language Contact 22. 41–62. Thomason, Sarah Grey & Terrence Kaufman. 1988. Language contact, creolization and genetic linguistics. Berkeley: University of California Press. Trudgill, Peter. 1974. The social differentiation of English in Norwich. Cambridge: Cambridge University Press. Vachek, Josef. 1982. English orthography. A functionalist approach. In William Haas (ed.), Standard Languages. Spoken and Written, 37–56. Manchester: Manchester University Press. Wagner, Esther-Miriam. 2010. Linguistic variety of Judaeo-Arabic in letters from the Cairo Genizah. Leiden: Brill. Wagner, Esther-Miriam and Ben Outhwaite. 2012. “These two lines…”: Hebrew and JudaeoArabic letter-writing in the Classical Genizah Period. In Jennifer Cromwell and Eitan Grossman (eds.), Beyond Free Variation: Scribal Repertoires from the Old Kingdom to Early Islamic Egypt (forthcoming). Oxford: Oxford University Press.

Part II: From spoken vernacular to written form

Mark A. Williams

2 Biblical Register and a Counsel of Despair: two Late Cornish versions of Genesis 1 Sources: The two late Cornish translations of Genesis 1 discussed in this paper are both found in the Gwavas Collection in the British Library. The first, almost certainly by John Keigwin (d. 1716), is found there exclusively; the second, likely by Nicholas Boson (d.?1703), is found in addition in the Borlase MS, copied by Dr William Borlase and now in Cornwall County Records Office. Neither are autographs and both have been transcribed at some point by non-Cornish speakers. A reconstruction of the original forms of these text is to be found in Padel (1975). Abstract: Cornish is the only Celtic language spoken in Britain that did not benefit from the production of an Early Modern vernacular Bible. Nevertheless, in the first quarter of the 18th century two learned Cornishmen, John Keigwin and John Boson, produced apparently independent translations of the first chapter of Genesis into Late Cornish. Cornish at this point was already obsolescent (likely with fewer than 4000 speakers), and the two translations coincided with a decisive and disastrous language-shift in favour of English, which led eventually to language death some decades later. This article aims to contextualise Keigwin and Boson’s activity as scribes and translators in the context of language obsolescence, exploring their reasons for attempting an ambitious translation into an increasingly moribund peasant vernacular. The issue is pursued in two ways. Firstly, a close analysis and comparison of the Cornish linguistic forms they deploy is undertaken, looking for evidence of the ways in which terminal linguistic decay was both negotiated (and betrayed) by the last learned speakers of the language in the absence of stable orthography. Secondly, that linguistic information is used to build up a sociolinguistic picture of Keigwin and Boson’s perhaps differing ambitions and intentions for their translations, and, poignantly perhaps, of their hopes for the language.

1 Introduction The subject of this piece is two translations of the first chapter of the Book of Genesis into Late Cornish, which are among a number of fragments of biblical

22   

   Mark A. Williams

translation into the language produced in the first quarter of the 18th century.¹ Taking verses 1–5 as a sample, my concern here is to press the orthographical and grammatical detail of the two versions in order to explore two areas. The first consists of the possibilities of register which existed in this terminal stage of the traditional language; the second encompasses the sociolinguistic implications of these acts of translation in the context of a particularly extreme kind of language change, namely moribundity and terminal language shift. Probing the details of scribal practice in these texts can, I argue, illumine the ways in which Cornish speakers were negotiating the accelerating and clearly inevitable obsolescence of their tongue.

2 The History of Cornish I begin with some essential linguistic and historical background. Cornish is a member of the Brythonic or Brittonic sub-group of the Celtic branch of the IndoEuropean language family, in contrast with the Goidelic sub-group, which consists of Irish, Scottish Gaelic and Manx. A traditional view of the relationship between the various members of the sub-group is set out in Figure 1.²

PROTO-CELTIC

BRITISH CUMBRIC

WELSH

BRETON

CORNISH

Figure 1: The traditional view of the relationship of the Brittonic languages.

1 For previous discussion see Tschirschky (2003) and Spriggs (2006). 2 Other models of the relationship between the various branches are possible and have advocates; see Russell (1995: 127–129).

Biblical Register and a Counsel of Despair: two Late Cornish versions of Genesis 1   

   23

Traditional Cornish shares about 80 % of its lexicon with its nearest relative, Breton, and the two are considerably closer to each other than either are to Welsh. This similarity is best explained by positing the existence of a distinct SouthWestern group in Common Neo-Brittonic, as Figure 1 sets out: a number of the sound changes characteristic of Breton and Cornish appear to date back as far as the 5th and 6th centuries CE (see Russell 1995: 128). The timeline of the development of Cornish is therefore as follows: 450–550 CE: 700–1100: 1100–1600: 1600–ca.1800: 1900–present:

British → “Common Neo-Brittonic”, but likely with a South-Western group already separating in the 6th century Old Cornish Middle Cornish Late Cornish Revived (or Neo-)Cornish, in several varieties

Figure 2: Cornish Timeline.

It may prove useful to begin by offering a brief sketch of the sociolinguistic and literary situation of Late Cornish. The 1700s were the final century of the traditional language: by the beginning of the 18th century, the language had been under constant pressure from English for roughly eight centuries, and had become confined to a last holdout in the far west of the peninsula, essentially the western parts of the hundreds of Kerrier and Penwith. But the final crisis came in the first decade of the century, which marked a decisive language shift in favour of English. At this point, it seems that speakers even in this final far-western redoubt ceased to pass the language on to their children. Cornish had been in retreat as a community language for several hundred years, but at this point its terminal decline suddenly seems to have accelerated: Iwan Wmffre’s generous estimate is that there were probably less than 5,000 speakers of Cornish alive at the turn of the 18th century (see Wmffre 1998: 1). We are looking therefore at the endgame of a longstanding state of language obsolescence. The actual point of language death, however, is harder to specify, as a number of different markers are available. The last monoglot seems to have died around 1675, but the last fluent native speaker, the famous Dolly Pentreath of Mousehole, expired over a century later in 1777. A few persons after this point retained some knowledge of the language: there are records of younger contemporaries of Dolly in Mousehole who could understand but not speak the language well, and a fisherman named William Bodinar wrote a letter in 1776, when he was 65, describing how he had acquired Cornish as a second language from older men whilst fishing at sea in his boyhood. Though he never spoke it at home, Bodinar

24   

   Mark A. Williams

has a good claim to have been the last speaker of traditional Cornish as he clearly outlived Dolly Pentreath (see Jones and Singh 2005: 137–138). As Mari Jones and Ishtla Singh have noted, there are two sets of reasons for the obsolescence of Cornish. The first is economic: the development of mining industries in West Cornwall during the early 17th century, with the accompanying influx of English speakers, is likely to have precipitated language-shift and hastened the demise of the language (see Jones and Singh 2005: 137, on which this discussion draws). Again, this represents the end-result of a long-standing tendency: from the Middle Ages, due to the strategic importance of the Cornish coast, the Cornish economy became increasingly outward-looking; as trade links with England increased in importance, Cornwall’s ports became crucial centres of commercial activity. As such, they inevitably became the foci of an accelerating process of linguistic anglicisation with the result that knowledge of English rapidly became essential for any Cornish person hoping to advance themselves. Psychologically – and in a manner reminiscent of the decline of the other Celtic languages – it seemed that knowledge of English was linked to success, whereas Cornish became indelibly associated with economic disenfranchisement and rural backwardness (see Durkacz 1983: 214–216). This involved a striking degree of social stratification with regard to the language: writing around 1680, William Scawen reported that “the poorest sort at this day [that speak Cornish] … are laughed at by the rich that understand it not.” (Scawen 2000: 292). The second set of reasons for the decline of Cornish are bound up with religion. After the uniting of the bishoprics of Devon and Cornwall, the Bishops of Exeter recognised the importance of having trained clergy who could communicate with their flocks in their native language. As Jones and Singh have noted, in the third quarter of the 13th century they founded the Collegiate Church of St Thomas of Canterbury at Glasney (Penryn), apparently with the specific aim of setting up a focus for cultural production in the Cornish vernacular (see Jones and Singh 2005: 136–137). Unfortunately, Glasney was dissolved in 1545 as part of the Reformation, with the result that the there was a rapid falling off in the cultural prestige of Cornish, as little new literature was being produced in the language. In addition, the Reformation led to the weakening if not the severance of Cornwall’s traditional ties with Brittany and the closely-related Breton language – ties which had already been abraded by the political unification of Brittany with France in 1532. In the late 16th century, Cornish also conspicuously failed to acquire two things which might have delayed or prevented language death: a translation of the Prayerbook, and even more importantly, a vernacular Bible. This is in striking contrast to Welsh, which thanks to an Act of Parliament and the labours of Bishop Morgan had a complete vernacular Bible from 1588. The history of the translation

Biblical Register and a Counsel of Despair: two Late Cornish versions of Genesis 1   

   25

of the Bible into Welsh is complex and my account here is highly streamlined, but the crucial point to grasp is that from 1588 Welsh speakers were in possession of a rich and transdialectical model of magisterial, matchlessly dignified, and largely consistent prose, influenced in a natural way by both the heritage of the medieval bardic literature and the Humanist scholarship of its translators. This extraordinary text almost certainly acted as a bulwark against the decline of Welsh and as a brake upon dialectical divergence.³ Cornish was not so fortunate. The Cornish themselves had blotted their copybook in 1549 by rebelling against the Act of Uniformity and the imposition of the English prayerbook, but while it was declared lawful in 1560 for those without English to learn the premises of the reformed Church in Welsh or Cornish, the fate of the two languages diverged at this point. The survival of Cornish was left up to individual efforts, no doubt because of the obvious difference in the relative sizes and strengths of the two speech communities in last quarter of 16th century. Biblical translations are therefore an obvious place for issues of linguistic prestige to be rehearsed, but it is striking that the two versions of Genesis 1 under consideration in this study are, in a sense, sadly belated: they were produced 130–140 years after the Welsh equivalents, and of course it is far from certain that a complete vernacular Bible would have slowed the decline of Cornish even had one been produced in the late 1500s. The two versions of Genesis 1 we have were produced by a pair of men implicated in a group that seems to have consisted of the last learned speakers of what we might term an increasingly moribund peasant vernacular. The term “peasant vernacular” is apt because, despite Glasney College, Cornish (unlike Welsh and Irish) had never been the vehicle of learned or courtly discourse. This group of individuals, some related by blood, others by looser ties, have left behind evidence for the kind of self-reflexive thinking about the status and fate of Cornish undertaken by its speakers during the first half of the 18th century, in which Cornish was still, barely, the language of a homogenous community, albeit one under extreme pressure. An examination of the linguistic and orthographic detail of the fragments of late Cornish they produced is therefore of great value for identifying the ways in which these writers were reacting to, conceptualising, and coming to terms (or not) with the manifestly terminal state of their language. A simplified diagram of the relationships between this last group of learned Cornish speakers can be set out as follows. Straight lines indicate blood-relations; a wavy one represents demonstrable friendship or contact. At the centre we have the Boson family: the precise details of the Boson family tree remain unclear but the current scholarly consensus is that Nicholas Boson

3 For the history and sociolinguistic effect of Welsh biblical translation, see Gruffudd (1988).

26   

   Mark A. Williams

John Keigwin of Mousehole (1642−1716) Nicholas Boson (fl. 1660−?1703)

John Boson (fl. 1700−?1730)

Thomas Boson (fl. ?1670−?1700)

William Gwavas (1676−1741)

Elizabeth Veale (née Gwavas) (d. 1791) Figure 3: The relationships between the last learned writers of Late Cornish

was the father of John, and that Thomas was a cousin.⁴ They appear to have been a well-off family in Paul, two miles south of Penzance. John Keigwin, the probable author of one of our two versions, is remarked upon by several contemporaries for the completeness of his knowledge of Cornish; John Boson, the likely author of the other version, taught Cornish to the antiquary William Gwavas, a barrister in Penzance, and his translation is preserved in a volume that belonged to Gwavas, now in the British Library, discussed below. It should be noted that the identification of Keigwin and Boson as translators of the two fragments is agreed to be highly probable, but cannot be certainly proven from the evidence: I will refer to the two translators conventionally by these names, though the reader must bear the caveat in mind. At this point we must turn to the texts themselves, and it may be of use here to have both Keigwin and Boson’s versions printed in full before a detailed discussion of their nature is undertaken. Genesis 1:1–5, probably translated by John Keigwin: Verse Ye 1 En delathuaz deewe aveth wraz neave ha no are  Ye 2d Hathera an noare Hebb Roath: ha veth ha tewlder wor bedgeth an downder: ha sperez deow reog gwayath wor bedgeeth an dowrow Ye 3d Ha Deewe Lauaraz Grenz boaz goollo: ha enna thera Gollo

4 Padel (1975: 5): The Cornish writings of the Bosun family (Nicholas, Thomas and John Bosun, of Newlyn, circa 1660 to 1730). Despite the title, ‘Bosun’ is spelled ‘Boson’ throughout, and the title is corrected in the Corrigenda.

Biblical Register and a Counsel of Despair: two Late Cornish versions of Genesis 1   

   27

Ye 4th Ha Deewe welhaz an Gollo troua Daa: ha Deewe roeg Deberrhee an an [sic] gollo thurt an teulder: Ye 5th Ha Dewe a grihaz an Gollo deeth ha an teulder e a Grihaz noaaze ha gothewhar ha metten o an Kessa Jornah. (British Library Add. MS 28554, Gwavas Collection, ff. 117r–119v.) Genesis 1:1–5, probably translated by John Boson: 1. En dalleth Deu grwes an neue ha an aor. 2. Ha thera an bez heb composter, ha heb kanifer tra; ha tuylder rag mêr a dounder, ha spiriz Deu reeg guaya var budgeth an dour. 3. Ha Deu laveraz, gwrens boz golou, ha thera golou. 4. Ha Deu gwellas an golou, tho vo da, ha Deu debarris an golou vrt an tuylder. 5. Ha Deu gwres an golou deth, ha an tuylder e gwres nôz, ha an gothihuar ha metten vo an kenza jorna. (In Padel 1975: 51) It should be noted that these are merely extracts: both Keigwin and Boson translated the whole of Genesis 1. These fragments do not, however, represent the first attempts to translate parts of the Bible into Cornish. Keigwin’s younger contemporary William Rowe (or Carew), b. 1660, of St Buryan, had previously rendered Genesis 3:1–24 and two sections of the Gospel of Matthew; these seem likely to have predated Keigwin and Boson’s attempts, but the point is at present unprovable. A word too is due on the nature of the textual transmission, as in neither case do autograph copies of the translations exist. The two versions of Genesis 1 we have are both found in the Gwavas Collection in the British Library, as was mentioned above. Keigwin’s is found exclusively there, but Boson’s version is found both in Gwavas and in another collection known as the Borlase MS, copied by Dr William Borlase and now located in Cornwall County Records Office. (The text of Boson’s version given above is Oliver Padel’s composite of readings drawn from both, both being of equal value as attestations of Boson’s lost original; see Padel 1995: 54.) The Gwavas version has a note recording that it was written by “Mrs Veale, the Eldest daughter of W. Gwavas Esquire”: importantly this means that both versions of Genesis 1 we possess have been copied by individuals who were not native speakers of Cornish, and as we will see this may be the reason for the errors which can be seen in them.

28   

   Mark A. Williams

3 Linguistic observations The first thing to notice is that the two versions are essentially rather similar – perhaps unsurprisingly, since, as we saw, both emerged from the intellectual milieu of the last educated speakers. To give an example, both use identical phrasing for the phrase in 1:3, ‘Let there be light’, using an unexpected form of the verb gul, ‘to do’ and the verbal noun ‘to be’: ‘Let there be light’ Keigwin: Grenz boaz goollo Boson: gwrens boz golou 3s pres. subj. of gul, ‘doing, making’ + verbal noun boaz, ‘being’ + gollo, ‘light’ = lit. ‘Let light do being’ Strikingly, in neither case does this correspond to the Welsh version(s) of the phrase, all of which use a 3s imperative of bod, ‘to be’ + goleuni, ‘light’, cognate of course with the Cornish words. Cornish did have such a form of the verb ‘to be’, bethens/bythens, but both translators seem to want to reserve boaz for predicative and substantive uses, avoiding deploying it as a term for ‘exist, come to be’. A second shared feature is that in both translations the orthography is in flux in a way that is normal for Late Cornish. (Again, this is in contrast with Welsh, the orthography of which had been largely stabilised by the 1588 Bible.) By “in flux” I mean two things. First, neither writer appears to be looking back towards the somewhat more consistent orthography of the Middle Cornish drama – much of which was on biblical subjects, and therefore might have suggested itself (in part) as an obvious model for dignified writing. Secondly, there is no attempt at imposing consistency of forms even in Late Cornish terms: the same words are frequently spelled differently within a single sentence. We can certainly speak in terms of tendencies: long vowels for example are sometimes marked by writing the vowel double, but this is far from universal (e.g. budgeeth ‘face’ = /ˈbudʒe:th/). Thus what we have orthographically is an anarchic mixture of a traditional spelling reminiscent of Middle Welsh, and the borrowed, and admittedly flexible, conventions of English. To give one example, a feature that is already common in Middle Cornish but greatly predominant in both passages is the use of a final –e to indicate that the vowel of the previous syllable is long. Contrast two uses of Keigwin’s: deeth noaaze

/de:δ/ ‘day’, Welsh dydd, /di:δ/ [with vowel doubled to indicate length] /no:z/, ‘night’, W. nos, /no:s/

Biblical Register and a Counsel of Despair: two Late Cornish versions of Genesis 1   

   29

Here we have the presence of two different systems for marking vowel length side-by-side. Both may even be found within same word if we understand the double aa of noaaze as an attempt to convey a long English diphthong oa (as in, e.g., toad) as a way of spelling Cornish /o:/; but bolstered with the final –e borrowed from post-Great Vowel Shift English orthography, as in abode. We can also contrast Keigwin’s noaaze with Boson’s form nôz for the same word: Boson’s spelling is far more reminiscent of 18th-century Welsh orthography and evokes the Cornish notation of the antiquarian and linguist Edward Lhuyd (1660–1709). (Indeed, it is striking that the use of the circumflex to mark vowel length is not present in Keigwin’s version at all.) And yet Boson also uses neue for /ne:v/, i.e. the English-derived system with a final vowel. So though the orthography is variable in both cases, it is clear that Boson is, broadly speaking, rather more consistent than Keigwin, and he is also experimenting with diacritics (the circumflex), which ultimately offered the possibility of a fully coherent and consistent system for marking long vowels. One characteristic and general feature of Late Cornish texts that appears in both our versions is a very erratic marking of initial consonant mutation. Consonant mutation is a notorious feature of all six modern Celtic languages, consisting of a form of (largely) grammaticalised sandhi whereby the initial consonant of a word undergoes regular phonetic changes dependent on the preceding word, and often also on the grammatical role the word is playing in the phrase (see Russell 1975: 231–257). In Middle Cornish (as in Breton) there were no less than five different types of mutation, of which overwhelmingly the most common is termed soft mutation or lenition. This denotes a slackening in articulation of the initial consonant according to a fixed pattern: speaking very crudely, in the Brittonic system which Cornish reflects, voiceless stops when lenited become voiced, and voiced stops become their corresponding fricatives. One very common environment in which lenition occurs is when a noun follows most simple prepositions – this was absolutely standard in Middle Cornish, and remains so in spoken and written Modern Welsh, and in Breton (and indeed in revived forms of Cornish, most of which look back to Middle rather than Late forms of the language.) And yet neither Keigwin nor Boson use it: Absence of expected lenition after simple prepositions: Keigwin: wor bedgeth [for expected wor vegdeth] ‘upon [the] face’ Boson: var budgeth NB budgeth/bedgeth < Fr. visage, either via Breton bisaig or from Anglo-Norman. Boson:

heb composter, ha heb kanifer tra [for expected heb gomposter, ha heb ganifer tra] ‘without equilibrium, without anything’

30   

   Mark A. Williams

However there is a striking difference between Keigwin and Boson’s versions in the matter of lenition. Boson simply does not mark lenition at all, whereas Keigwin marks only the lenition of verbal forms and not of nominal ones. This is striking because one very common syntactical structure in Cornish (as in Modern Breton and Middle Welsh) involved the lenition of verbs. In this structure, the so-called “Abnormal Order”, a positive statement could be expressed as follows: (pro)nominal SUBJ or OBJ + meaningless preverbal particle a, triggering lenition + inflected VERB form As can be seen, here it is the initial consonant of the verbal form which invariably undergoes lenition. Keigwin applies this rule consistently: Deewe a wraz [unlenited gwraz], ‘God created’ Dewe a grihaz an Gollo deeth ha an teulder e a Grihaz noaaze [unlenited crihaz is 3s preterite of cria, ‘cry, call’: the –h– seems to be a hiatus marker] ‘God called the light “Day” and the darkness He called “Night”’ Now a, our meaningless preverb, seems to have been commonly subject to omission in speech, like a number of other particles in Late Cornish; other writers of the last phase of the traditional language intermittently drop it, whilst retaining the lenition of the verb which it triggers (see Wmffre 1998: 66–67). Keigwin’s translation of Genesis 1:4 provides two examples of Abnormal Order phrases in which the preverbal particle has been elided, but the expected lenition has been retained: Ha Deewe welhaz an Gollo troua Daa: ha Deewe roeg Deberrhee an an [sic] gollo thurt an teulder, ‘And God saw the light, that it was good; and God did divide the light from the darkness’ for expected: Deewe a welhaz [unlenited gwelhaz] ‘God saw’ Deewe a roeg [unlenited groeg/grug] ‘God did’ Note that in both these examples /g/ disappears when lenited, as it does in all three Neo-Brittonic languages: originally /g/ had changed to the corresponding fricative, /ɣ/, which was inherently unstable. If Keigwin drops the preverbal particle but retains the lenition, Boson on the other hand never writes the particle, nor does he lenite verbal forms in this characteristic subject-fronted order. He gives us, for example: Deu gwellas an golou ‘God saw the light’ [for expected Deu a wellas an golou]

Biblical Register and a Counsel of Despair: two Late Cornish versions of Genesis 1   

   31

Subject-Verb-Object (SVO) order without a particle or lenition is highly unusual in terms of Brittonic syntax more generally: it is not available as an option in either Middle Welsh or Breton.⁵ (SVO order with the particle a is only used in very formal written Modern Welsh, but remains standard in spoken Breton). This phenomenon admits of two explanations. It is possible that in the last stages of the language the mutation system itself was breaking down into anarchy, with expected mutation occurring or not ad lib; something like this seems to have occurred in Late Manx (see Thomson 1969: 189–190). Alternatively, the absence of expected lenition could merely be an orthographic quirk. It should be noted that neither Middle Cornish nor Middle Breton had marked mutation in their orthography as consistently as had Welsh, and even there marking was not total (Middle Welsh orthography tends not to distinguish between /d/ and lenited /δ/, for instance). This tendency, native to the historically written forms of the language, could well simply be being extended here – for both Keigwin and Boson writing the spoken language is clearly an unusual act, and therefore liable to idiosyncrasy. The latter seems more likely: in both Breton and Welsh lenition in particular has continued to be very obvious and fairly consistently applied in the spoken, informal forms of the language, despite the fact that the functional load borne by mutation is actually rather small. Indeed, in Welsh in particular lenition has tended to spread at the expense of other mutations, so it seems unlikely that it could have evaporated to so great an extent in Late Cornish (see Wmffre 1998: 18). The main point of significance here, however, is the contrast of convention with Welsh and Irish. From the Early Modern Period scrupulous orthographical marking of consonant mutation in these languages had tended to be a sign of a dignified, formal register, and yet it is striking that neither of our translators seem to feel that such precision is incumbent upon them in Cornish. Paradoxically, it is possible that Boson’s consistent avoidance of such marking should be taken as an idiosyncratic and historically-innovative attempt at formal clarity. Once again we have the sense that both writers are making their own rules, in an uncertain spirit of some experiment. I am tempted also to note that Boson’s dropping of both the preverbal particle and the expected lenition of verbal forms means that his phrases often calque English word order – indeed the very order of the Authorized Version: En dalleth In the beginning

Deu God

grwes created

an the

neue heaven

ha and

an aor the earth

5 In Middle Welsh, the particle a can be dropped before oed, the 3s imperfect of the verb bot, ‘to be’ (later bod), but this is the only exception; see Evans (1964: 61).

32   

   Mark A. Williams

It might seem as though there is an absence of the definite article (‘the’) before dalleth, ‘beginning’, but in fact the preposition en, ‘in’, merged with the article, an, to give e n, ‘in the’, always written en. In other words, Boson’s Cornish is a precise calque in every aspect to the English.

4 Stylistic observations What, therefore, can we conclude about the intentions of Keigwin and Boson in making these translations, and the role of their scribal practices in the history of the final phase of traditional Cornish? First, with regard to register both versions are clearly in a written form of spoken Late Cornish: neither aim at any kind of “literary” archaism nor at a harking back to the forms of the Middle Cornish drama. Both use the word teulder for ‘darkness’, for example, a typically late form which had replaced Middle Cornish tolgo (/ˈtulgo/). Keigwin’s version is completely typical of Late Cornish in the instability of /r/ in unstressed syllables (a feature which space does not permit me to discuss here), the sporadic omission of the preverbal particle in the Abnormal Order, and the variety of orthographic strategies in relation to vowels and diphthongs. He also admits very late forms such as assimilated kessa for kensa, ‘first’. But Keigwin also seems to have been a very careful translator, with an instinctive sense both of faithfulness to the text and to the natural range of the language: this bears out the high esteem in which he was held as a scholar and Cornish-speaker by contemporaries. An extant Cornish epigram by John Boson, with his accompanying (rather loose) translation, survives in the Gwavas Collection as a testament to Keigwin’s linguistic proficiency: En Tavaz Greka, Lathen ha’n Hebra, En Frenkock ha Carnoack deskes dha, Gen ol a Gormola Brez ve dotha Garres ew ni, ha Neidges Ewartha. ‘In Greek & Latine & in Hebrew Tongue In Ffrench & Cornish mighty strong With all the Glory of his mind― Is chang’d & left us here behind.’⁶

6 ‘On the death of Mr. John Keigwin written the 20th of April 1716 by Mr. Boson’ (Padel 1975: 59).

Biblical Register and a Counsel of Despair: two Late Cornish versions of Genesis 1   

   33

The mention of Hebrew is striking, and it is possible that Keigwin’s translation shows signs that he may also have been looking at the Hebrew text as he translated, rather than working solely from the English. Let us take one sentence from Keigwin’s version, with the corresponding verse in the Authorized Version: Hathera an noare Hebb Roath: ha veth ha tewlder wor bedgeth an downder: ha sperez deow reog gwayath wor bedgeeth an dowrow ‘And the earth was without form, and void; and darkness was upon the face of the deep. And the Spirit of God moved upon the face of the waters.’ There are several possible Hebraisms here. In the phrase Hebb Roath: ha veth (/heb ro:th ha veth/) is it conceivable that the alliteration of h and the repetition of monosyllables ending in –th is intended to evoke the Hebrew tohu vavohu? Keigwin’s phrase is not a direct translation of ‘without form, and void’ – which would have been perfectly straightforward to express in Cornish. Rather it means literally ‘without shape and a(ny) shaped thing’, with two nouns dependent on the preposition heb, ‘without, lacking’; veth is from Latin factum, ‘a thing made’, older feth, with characteristically Late Cornish voicing of an initial f- (Wmffre 1998: 19–20). Further, the verb ‘to be’ in ha tewlder wor bedgeth, ‘and darkness [was] upon the face’ is unexpressed in the Hebrew (but not Cornish) manner, though of course the Authorized Version highlighted the insertion of ‘was’ typographically. In short, it is not impossible that Keigwin’s punctiliousness involved keeping one eye on the Hebrew as he translated, given that there is evidence that he had at least some knowledge of that language. The language of Boson’s translation is also essentially spoken Late Cornish, and indeed is often close to Keigwin’s, but he does show signs of certain ambitions for his text. The first of these is his move in the direction of a more stable orthography, and the peculiar suppression, almost eradication, of consonant mutation – a departure, as far as we can tell, from the spoken language, and perhaps to be interpreted as an idiosyncratic attempt at formal clarity: it is of course also an anglicising turn, eliminating one of the most distinctively Celtic features of the language. The absence of mutation in Abnormal Order phrases means that the Cornish seems to calque the English word-order of the Authorized Version: a tacit admission that his Cornish version of Genesis is dependent on, and subservient to, the English. But Boson also seems to be aiming for a certain amount of balance and stylistic variation, ostentatiously avoiding the repetition of words and phrases even where the original text demands it. Let us turn to a pair of examples of this phenomenon from his translation – the relevant elements in the Cornish and the translation have been formatted identically to highlight their correspondence.

34   

   Mark A. Williams

Boson’s unwarranted variations: 1. 2.

En dalleth Deu grwes an neue ha an aor. Ha thera an bez heb composter, ha heb kanifer tra; ha tuylder rag mêr a dounder, ha spiriz Deu reeg guaya var budgeth an dour.

Literal translation: 1. 2.

In the beginning God made the heaven and the earth. And was the world without evenness, and without every thing; and darkness before the face of the deep, and the spirit of God did move upon the surface of the water.

We see here the deliberate avoidance of repetition: an aor, ‘the earth’ in 1:1 becomes an bez, ‘the world’ in 1:2; the prepositional phrase given twice as ‘upon the face of’ in the Authorized Version (Hebrew al-pəne … al-pəne) is rendered with two different Cornish expressions. The intention may have been to bring over a sense of stylistic variatio from English prose-writing. In that sense, therefore, an agenda for the language – the shaping of literary prose – seems more evident in Boson’s version than Keigwin’s, but Boson is also the less instinctive and careful translator of the two. Keigwin inhabits Cornish, where Boson affects it. Nevertheless, neither man uses English borrowings, for example, which were very widespread in Cornish; this is probably the clearest sign of the writing of consciously elevated prose, harking back to the Middle Cornish drama, in which pure Cornish seems to have been the sign of dignified language. (The Devil is the first character to use English in the medieval play-cycle known as the Ordinalia, and English phrases belong to the meaner sort in general.) Nevertheless, even excellent Middle Cornish involved much thoroughly-naturalised English vocabulary, which both Keigwin and Boson seem to be carefully limiting, conscious, no doubt, of the question of English infiltration. Further, neither version is without error. The very first line of Keigwin’s text includes an example of eyeskip dittography, highlighted below in bold. As our sole text is not Keigwin’s autograph, this may not be his own slip but could easily have entered the text when it was copied: Verse Ye 1 En delathuaz deewe aveth wraz neave ha no are  Ye 2d Hathera an noare Hebb Roath: ha veth ha tewlder wor bedgeth an downder There was a Late Cornish form of the conjunction yn weth, ‘also’, spelled a weth, but Keigwin always writes w for /w/, not v for /w/ as does Boson; also there is no

Biblical Register and a Counsel of Despair: two Late Cornish versions of Genesis 1   

   35

conceivable semantic reason for an ‘also’ to be present in Keigwin’s careful rendering. Boson introduces the wrong verb in one line of his rendering; as we have two manuscript versions of his translation, we can say with greater confidence that this error goes back to Boson’s original: 5: Ha Deu gwres an golou deth, ha an tuylder e gwres nôz, ha an gothihuar ha metten vo an kenza jorna Literal translation: ‘And God made [3s preterite < gul, ‘do, make’] the light Day, and the darkness he made Night. And the evening and morning were the first day.’ In the rest of Boson’s translation of Genesis 1 he consistently uses the expected verb form, kries (= Keigwin’s crihaz), ‘called’, the 3s preterite of the verb kria, ‘call, cry’. The confusion of verb may be aural: both incorrect gwres and correct (and, strikingly, correctly lenited) gries might well have been sounded homophonously as /gre:z/.

5 Cultural observations and conclusions Turning from the linguistic to more broadly cultural questions, I think it is clear that one thing both translators shared was the ability, via education, to look over the metaphorical parapet at the imminent death of Cornish and to mount some kind of response. I suspect the impulse to begin the translations may have come from the effect of the visit of Edward Lhuyd to Cornwall at the turn of the 18th century. The famous Lhuyd, an indefatigable Welsh scholar and linguist to whom we owe the insight that the Celtic languages are genetically related, had been contacted by a group of Cornish scholars, led by John Keigwin himself, and he accepted the invitation to travel to Cornwall to study the language. As a result, Early Modern Cornish was the subject of a crucial study published by Lhuyd in 1702 (Williams 1993). It is clear from this anecdote that the learned Keigwin was clearly more than capable of making analogies between Cornish and Welsh, and of aiming to further interest in Cornish amongst Welsh scholars. There is also no doubt that Keigwin must have been aware of the massive importance and prestige of the Morgan Bible (and its 1620 revision by Bishop Richard Parry), and it is possible that his translation of Genesis 1 was undertaken in a spirit of imitative experiment – perhaps simply to see if it were possible.

36   

   Mark A. Williams

Boson is in some ways a more interesting case, in that the Boson family seem to have been what we would now recognise as “language activists”. His father, Nicholas, has left a fascinating and rather difficult document called Nebbaz Gerriau dro tho Carnoack, ‘A Few Words about Cornish’, perhaps from c.1675 (Padel 1975: 24–37). This awkward, hobbled text is an attempt to compose in Late Cornish something between a pamphlet and a learned essay, and the effect is poignant when one sets it beside the formal clarity and Ciceronian elegance of similar pieces by the Welsh Humanists of the preceding century. In Nebbaz Gerriau, Boson père records that, Cornish-speaker though he is, Cornish was not his mother-tongue: his mother forbade the servants and neighbours to speak to him in Cornish in a deliberate attempt to prevent him acquiring the language (a mark of provincialism and poverty). He went on to school and then away to France, and only learned Cornish upon “coming to get business in the world”, in late adolescence or young adulthood. Boson clearly conceived a passion for the language, and is aware of its sister relationship with Welsh and the pressure all three Brittonic languages found themselves under in his day, and of the more secure status of Welsh. Nicholas Boson’s mention of the book written for his children – and his son’s evident mastery of the language – shows how strong the elder Boson’s partisanship of Cornish was, and presumably tells us something about the pro-Cornish nature of John Boson’s upbringing. But in sociolinguistic terms, it also means that John’s Cornish is likely to have been acquired primarily from a second language learner. This is in itself highly significant as an example of the phenomenon whereby endangered languages tend to decline in social status until they are spoken by the rural poor, but language revivalists tend to be disproportionately well-off and middle class. In other words, we can surmise that the Boson family were engaged in a consciously revivalist strategy; indeed, one could make the case that they were not only among the last speakers of traditional Cornish, but also, by virtue of the fact that the childhood transmission of the language to Nicholas Boson had been broken, among the first revivalists. John Boson’s linguistic trajectory therefore anticipates to a degree that of the small number of neo-Cornish native speakers who now exist who have been brought up as fluent speakers by language-activist parents who have learned Cornish themselves. If both Keigwin and Boson’s translations of Genesis 1 are rooted to different degrees in a consciously revivalist strategy, it is unclear what difference they hoped their translations could make in such a dwindling community of speakers. As we cannot date their translations with any degree of precision, it is difficult to say whether there is mutual influence between them. Boson’s translation could have been undertaken after Keigwin’s death; equally each man could have been well-aware of the other’s efforts. There is a certain bold energy to be commended

Biblical Register and a Counsel of Despair: two Late Cornish versions of Genesis 1   

   37

in both that they, so to speak, began at the beginning; but their experience as translators must have been a melancholy one. Was the Bible chosen – by their predecessor Carew as much as by Keigwin and Boson – because it was the most centrally and powerfully culturally-authoritative text, or was the prospect of a full Cornish Bible in their minds as they translated? On a simply human level, the thought can hardly have eluded either man; as a result, when both Keigwin and Boson broke off at the end of Genesis 1:1, one wonders whether it was in the expectation of sharpening the quill to begin again the next day – or in something approaching despair.

6 References Durkacz, Victor Edward. 1983. The decline of the Celtic languages. Edinburgh: John Donald. Evans, D. Simon 1964. A grammar of Middle Welsh. Dublin: Dublin Institute for Advanced Studies. Gruffudd, R. Geraint. 1988. Y Beibl a droes i’w Bobl draw: The translation of the Bible into the Welsh tongue. Cardiff: The BBC. Jones, Mari C. & Singh, Ishtla. 2005. Exploring language change. London: Routledge. Padel, O. J. 1975. The Cornish writings of the Bosun family (Nicholas, Thomas and John Bosun, of Newlyn, circa 1660 to 1730). Redruth: Institute of Cornish Studies. Russell, Paul. 1995. An introduction to the Celtic languages. London and New York: Longman. Scawen, William. 2000. Antiquities Cornuontanic: The causes of Cornish speech’s decay. In Alan M. Kent & Tim Saunders (eds.), Looking at the mermaid: A reader in Cornish literature 900–1900, 281–293. London: Francis Boutle. Spriggs, Matthew. 2006. Additional thoughts on the medieval “Cornish Bible”. Cornish Studies 14. 44–55. Thomson, R. L. 1969. The study of Manx Gaelic [Sir John Rhŷs Memorial Lecture British Academy 1969]. Proceedings of the British Academy 55 [Special issue]. Tschirschky, M. W. 2003. The medieval “Cornish Bible”. Cornish Studies 11. 308–316. Williams, Derek R. 1993. Prying into every hole and corner: Edward Lhuyd in Cornwall in 1700. Redruth: Dyllansow Truran. Wmffre, Iwan. 1998. Late Cornish. Munich: LINCOM Europa.

Markus Schiegg

3 Medieval Glossators as Agents of Language Change Sources: The glosses analysed in this paper were found in two Early Medieval manuscripts, written by monks in cloisters in East Francia. The Maihingen Gospels were glossed in the 8th century in Echternach (now a city in eastern Luxembourg). Currently this codex is stored in the University Library of Augsburg (Ms. I, 2, 4°, 2). The second manuscript, Clm 6263, was glossed around the 10th century in Freising (now in southern Germany). As a result of secularisation, it was brought to the Bavarian State Library in Munich. Abstract: In the 8th century CE, Frankish monks used styli to carve words in vernacular German on the margins and between the lines of Latin manuscripts to annotate demanding passages of the Latin text. This paper examines two aspects of how these glossators served as promoters of language change. First, dry-point glosses are the first documents to be written in Old High German; the glossators’ use of the vernacular resulted in a substantial change in written language usage: the gradual replacement of Latin by Old High German. Second, glossators are promoters of progressive language forms, as they often employ language of conceptional orality that cannot be encountered in other text types until centuries later. This paper’s functional and sociolinguistic approach on progressive spellings of vernacular glosses proves valuable for the reconstruction of the early medieval language of orality.

1 Introduction 1.1 Approach The aim of this article is to explore how and to what extent medieval glossators served as agents of language change. The introductory section summarises the relevance of glossing both for internal and external linguistics (1.2.). I pursue the current functional paradigm in glossography, which stands in contrast to the older lexicographical paradigm (1.3.). The corpus used for this article consists of the Old High German (OHG) glosses in two Latin manuscripts from early medieval times. The dry-point glosses from Echternach are among the oldest in OHG that have been transmitted to us (2.1.). Here, glossators are agents of language change

40   

   Markus Schiegg

because they initiate vernacular writing (2.1.1.). Some of the glosses in this manuscript are striking because – although they are the oldest sources of OHG – they deliver linguistic forms that often appear only centuries later in other documents (2.1.2.). To broaden the database, I present similar phenomena in another glossed manuscript from Freising (2.2.2.) and explain such peculiarities with the specific communicative conditions of glossing with dry-point (2.2.3.). Section 3 systematises these observations from a variationist perspective. Here, I apply Koch and Oesterreicher’s Model of the Conceptual Continuum (3.1.) to the East Francia communicational space (3.2.). A diachronic comparison of the linguistic situation in three periods (the 7th, 8th and 10th centuries) will trace medial and conceptual shifts within this space that result from glossing and other vernacular texts. In the concluding section, I discuss whether or not such changes were intended by the glossators.

1.2 The Relevance of Glossing for Linguistics When you read a demanding text in a foreign language, you probably have a pencil in your hand and annotate difficult words in the margins or between the lines. This act of glossing is a universal phenomenon that can be encountered in any age and any culture that produces and studies written texts (Henkel 2009: 468). Glosses provide important data for linguistics. Some fragmentary IndoEuropean languages like Thracian, Macedonian and Illyrian are mainly transmitted through the glosses collected by the Greek scholar Hesychius of Alexandria (5th/6th  century CE). Here, glosses are essential sources for gaining access to the lexis of these lost languages.¹ In the debate on whether Macedonian can be considered a separate language and not only a Greek dialect, a Macedonian sound change is brought into discussion that may only be derived from glosses and some personal names (Fortson 2010: 459–464). Glosses can thus be useful to reconstruct the phonology of past languages. Yoshizawa and Kobayashi (1999) have analysed several Japanese, Chinese and Tibetan paper manuscripts that are annotated with stylus-impressed glosses without any ink. In 1993, they discovered

1 For our knowledge of OHG, the mere quantitative number of glosses should attest their high relevance: Jochen Splett estimates that about two thirds of the OHG lexis has been transmitted only through glosses (2000: 1196). At the moment, Splett’s numbers are shifting in favor of the glosses. The number of known OHG glosses and especially dry-point glosses increased tremendously in recent years. As a result, Elvira Glaser’s list of 70 known manuscripts with OHG dry-point glosses (1996: 55–63) has recently been doubled (Nievergelt 2012).

Medieval Glossators as Agents of Language Change   

   41

glosses from the 5th to the 10th centuries in manuscripts from Dunhuang in China. One of these glosses’ purposes is to aid the pronunciation of the complex characters of Classical Chinese. Besides lexicography and phonology, glosses can also shed light on the morphology and even the syntax of languages (Schmid 2009). However, glosses are not only relevant for reconstructing past languages and their internal history, but also for establishing the external history of languages and sociolinguistic contexts. The glosses in the Dunhuang manuscripts not only give phonological information but may allow us to glance into the Buddhist classroom and “the method of Buddhist teaching” (Yoshizawa and Kobayashi 1999: 5). The classroom is only one of several contexts where glossing occured (Lapidge 1982; Wieland 1985). Regularly, the sophistication of the glosses as well as the glossator’s scope of interests exceeded school-level. For example, the glosses in an 8th-century manuscript from Freising in Southern Germany (Clm 6305) give detailed comments on complex theological concepts as well as text-critical remarks (Glaser 1996: 559; 643). They stem from a highly educated scholarly monk who annotated the text, supposedly in his private cell and for his own purposes. Michael Baldzuhn (2001: 496) analysed a 1000-years old, densely glossed codex from Trier, Germany, which is striking as a result of its huge size, its marvellous decoration and its luxurious décor. This codex was glossed in the library where educated monks inserted their expertise. Over centuries and after different glossing enterprises such manuscripts developed into accumulators of the knowledge of medieval times. With the exception of a few weighty codices like the one from Trier, the use of a manuscript was not restricted to only one place. A teacher, for example, could have carried a codex to the library to copy information from commentaries, glossaries and other glossed manuscripts into his codex; then he brought it into his cell to study for himself or prepare for class; finally, in class he used these annotations and inserted new glosses again.

1.3 Glossographic Paradigms and Seminal Studies Despite their enormous variability and manifold importance for linguistics, analyses of glosses remained extremely narrow until recently. Since the first editions from the 19th century, like the important five-volume edition by Steinmeyer and Sievers (1879–1922), the focus of research was for a long time exclusively on lexicography. The reason for this probably derives from the attribution of glossing to school contexts, where uneducated pupils relied on their mother tongue to annotate difficult Latin words. As a result, glosses were generally judged as primitive and deficient (Ernst and Glaser 2009: 1002).

42   

   Markus Schiegg

In the 1980s, research interests began to change; Christoph März (1996) even notices a paradigm shift in glossography. Instead of collecting lexical items, current research rather focuses on the individual manuscript and aims to analyse “all manifestations of the glossographer’s activity” (Derolez 1992: 12) – including Latin glosses and even non-linguistic signs like accent marks. This functional approach is suitable for reconstructing the specific conditions under which a monk studied a Latin text. Glossing at school was thus identified as only one of several different contexts in which this technique was used (1.2.). In general, the departure from mere lexicographic interests allows deeper insights into the cultural technique of glossing and opens up research for questions like the ones pursued in this article. Being only visible under certain lighting with a flashlight and a magnifying glass, dry-point glosses had for a long time been excluded from glossographic research. Elvira Glaser (1996) established new foundations for systematic editions and linguistic analyses of such glosses. New paleographic micro analyses combined with linguistic and contextual investigations (e.g. Ernst 2007) ensued from Glaser’s study. For instance, new editions thoroughly differentiate between the layers of glosses by different glossators in one manuscript. Unfortunately, even editions from the 1990s are often not precise enough and still require corrections.² Two recent studies are seminal for this article. “Graphematik und Phonematik” was published by Ernst and Glaser in the profound two-volume handbook on OHG glossography (2009; abbreviated BStH). The authors assert the high relevance of phonological analyses of glosses – and especially dry-point glosses – for our knowledge of OHG grammar. Due to their original and spontaneous character, drypoint glosses regularly deliver linguistic forms that may give hints for the reconstruction of the medieval spoken language (2.2.3.). In “Althochdeutsche Glossen

2 See for example my analysis of the glosses in the manuscript Augsburg, Archiv des Bistums, HS 10 (Schiegg 2012). In the most recent edition, Schulte assumes that all the dry-point glosses in the manuscript were written by the same glossator (1993: 58). However, I could clearly differentiate at least two glossators with dry-point glosses. One glossator – or probably even more than one – scratched his glosses with a sharp instrument into the manuscript; another glossator imprinted them. In addition, Schulte differentiated between only 4 glossators who used a feather quill. The glosses that he subsumes under one glossator, however, are different both in the color of ink and the style of the characters, which made me argue for three further glossators. This had significant consequences for the dating of the glosses because the three feather quill glosses of my “glossator 4” show – besides paleographic peculiarities – very progressive spellings (syncope of prefix vowel and for Germ. /t/; compare 2.2.2.). These findings made me date these three glosses from the 11th century instead of Schulte’s general dating of these glosses from the 10th century (1993: 56).

Medieval Glossators as Agents of Language Change   

   43

als Quellen einer Sprachgeschichte ‘von unten’” (2011), Ernst and Elspaß follow a perspective ‘from below’ that focuses on data that had tended to be neglected in language histories, such as private texts with less formality that show affinity to the oral language and the language of immediacy (3.1.2.). The authors conclude that OHG texts in general and glosses in particular deliver linguistic forms from the language of immediacy, while Latin occupies the pole of the distance (3.2.).

2 Vernacular Glossing and Language Change in Two Early Medieval Manuscripts 2.1 The Glossators of Echternach 2.1.1 Writing the Vernacular In the western world, vernacular glossing was first developed in Ireland and Anglo-Saxon England, where Christianity and education flourished in medieval times. From the 7th century onwards, Irish and Anglo-Saxon missionaries brought Christianity to the continent and, along with their religion, they introduced their elaborate techniques of studying and learning (Moulin and Glaser 2009: 1262). One of the earliest centres of these missionary activities was the cloister in Echternach (in today’s Luxembourg), which was founded in 697/698 by the Anglo-Saxon monk Willibrord (Moulin and Glaser 2009: 1257). Other early missionary centers were Fulda, Würzburg and Mainz, founded by Boniface (Angenendt 2001: 270). Being “one of the most important continental monasteries of the Early Middle Ages” (Ó Cróinín 1999: 86), Echternach can be described as the “cradle of Old English and Old High German writing activity on the continent, and as the centre from which we have the oldest surviving original glosses in Old Irish” (Ó Cróinín 1999: 95). This resulted in a multilingual community of missionary monks from Ireland and England as well as local recruits. Evidence both of the craftsmanship and language contact in Echternach is provided by the Maihingen Gospels (Augsburg UB Ms. I, 2, 4°, 2). Extensive insular influences can be observed in the style of the characters as well as in its rich and superb illumination.³ Together with another codex from the same monastery,

3 High quality reproductions of initials, its insular script, illumination as well as two dry-point glosses can be found in the catalogue of the manuscript’s exhibition in Augsburg (Hägele and Wurst 2010: 50–54).

44   

   Markus Schiegg

the Echternach Gospels (Paris, BN. lat. 9389), this manuscript bears the oldest testimonies of OHG: dry-point glosses. Such glosses are not written with feather quill and ink but carved into the parchment with styli, small instruments made of iron or wood. Styli had originally been used to write on wax tablets, where script can easily be erased. Such tablets were already known in ancient times and were often used at school to practice writing skills. Besides glossing, styli can also be used on parchment to mark the borders of the text and the lines to write on, to indicate the beginning of a chapter, as well as for drawings and sketches for initials. As we have early testimonies of Irish dry-point glosses from the 7th century (Ó Cróinín 1999: 86), recent research assumes that this technique was transferred to the continent by insular monks (Glaser and Nievergelt 2009). Vernacular glossing first appeared in the missionary centers and soon spread from cloister to cloister (Glaser 1996: 67). The Maihingen Gospels carry several dry-point glosses. Glaser (1997) offers the most recent edition of 23 OHG glosses that were written in Echternach, probably in the first half of the 8th century. The insular paleography of the glosses, for example the characteristic (Glaser 1997: 5), shows influences of the missionary activities. Testimonies for the multilingual community in Echternach are the 13 Old English glosses, listed in Ó Cróinín (1999: 87). Other entries with styli like crosses, lines and punctuation marks hint at the intensive use of the manuscript in general (Glaser 1997: 5). In addition, Glaser lists five glosses that have not yet been identified linguistically (1997: 15–16). In 2006, Andreas Nievergelt analysed the manuscript again and found some new vernacular dry-point glosses (Nievergelt 2009: 27). To come to a first conclusion, glossators from Echternach adopted the technique of vernacular glossing in the first half of the 8th century from the Irish and the Anglo-Saxons. They were beginning to write down a language that had only been used in oral contexts before. This can certainly be classified as a significant medial change.

2.1.2 Peculiarities in the Glosses of the Maihingen Gospels A closer analysis of the OHG dry-point glosses in the Maihingen Gospels reveals some striking graphematic features.⁴ On the one hand, we encounter several

4 The dating of the glosses was conducted by analysing paleographic features, not through a phonetic analysis of the glosses (Glaser 1997: 16). The following spellings refer to Glaser’s readings in this edition, where she corrects some mistakes in older editions.

Medieval Glossators as Agents of Language Change   

   45

archaic spellings (see Bergmann 1977: 91; Glaser 1997: 16) that support the paleographic dating of the glosses to the first half of the 8th century: – in the gloss naic for Germ. /ai/ shifts into at the end of the 8th century (Braune 2004: §44) (naic means ‘he flexed his knees’ and glosses the Latin flexo). – in gauma for Germ. /au/ shifts before labial consonants into in the 9th century (Braune 2004: §46) (‘feast’; glosses conuiuis).⁵ – in droc for Germ. /ō/ is diphthongised into during the 9th century (Braune 2004: §38) (‘he carried’; glosses conferens). – in ueo is a an intermediate form between Germ. /eu/ and the spelling that first occurs in the first half of the 9th century (Braune 2004: §47; §48) (‘how much’; glosses quanto). – in gacaman⁶ shifts into at the beginning of the 9th century (Braune 2004: §71) (‘to tame’; glosses domare). Besides these archaic forms, however, there are two kinds of spellings that do not fit into the scheme above and require further explanation: – in the glosses hase and ginos for Germ. /t/, which usually changes into (or ) with the Second Consonant Shift (Braune 2004: §160) (hase ‘hate’ and glosses odio; ginos means ‘comrade’ and glosses socii). Braune also refers to the hase gloss and evaluates this peculiar spelling as a Vertauschung (‘confusion’). – in lan is usually spelt as because it represents the velar nasal [ŋ] (Braune 2004: §128) (‘long’; glosses prolixae). There have been three different interpretations of the spelling. Hofmann (1963: 38), Schützeichel (1964: 22) and Bergmann (1977: 91) postulate Old English influence. Here, they see an argument against the traditional thesis of the spread of the Second Consonant Shift from the south to the north. This would have led to a spelling that can usually be encountered in the south. Moulin and Glaser (2009: 1263) adopt this argumentation. Indeed, the manuscript shows other influences in paleography as well such as the characteristic , mentioned above.

5 In contrast, the gloss nodi (‘necessary’) already shows the monophthongisation of Germ. /au/ before dental consonants into . 6 Glaser is not sure whether the in this gloss could also be a . A would imply that there had been no Second Consonant Shift to an affricate, which could only be explained as an influence from an Anglo-Saxon scribe. Glaser’s second interpretation as a seems more convincing because we sometimes encounter such spellings for an affricate (1997: 9–10; also see Schmid 2007: 52).

46   

   Markus Schiegg

Annina Seiler (2011) recently analysed the use of letters in Old English, Old High German and Old Saxon texts that had been labelled superfluous by Classical Latin grammarians (k, q, x, y, z). For our case, she concludes that some local scribes were struggling to find a suitable letter for the fricative, so they decided on the . In OHG texts, this letter was used to write down the Germ. /s/ that probably sounded similar to the fricative.⁷ Ernst and Glaser (2009) go one step further and assume that the two sounds may already have collapsed much earlier than the grammars state. In the recent edition of Paul’s grammar on Middle High German, we read that these two spellings had not collapsed in writing until the end of the 12th century (Paul 2007: 172). Ernst and Glaser give several further examples and show that for can especially be observed in glosses (2009: 1011). Sonderegger has also noticed such spellings in his analysis of the St. Galler Vorakte (1961), short drafts of charters. These documents are only accidentally transmitted to us as fragments on the reverse of the respective charter. In comparison to the charters, these aides de memoire show strikingly progressive forms.⁸ Sonderegger concludes that such spellings in personal notes are less conservative than the much more archaic and formulaic charters (1961: 283). In analogy, this in the Maihingen Gospels may also be interpreted as a progressive form resulting from spontaneous note taking. The second spelling, in lan for the velar nasal [ŋ] may be explained in the same way. Glaser had at first guessed at an abbreviated gloss (1997: 13). In their later article, Ernst and Glaser interpret this again as a very progressive spelling and show several further examples of dry-point glosses with such a spelling (2009: 1015). These forms, however, regularly occur only centuries later, even after Middle High German (Paul 2007: 150). According to Penzl, they are “extremely rare before the fourteenth century” (1968: 343). Perhaps, the combination of archaic and progressive spellings in the Maihingen Gospels results from different glossators (Hofmann 1963: 37). However, Glaser and Moulin state that it is hardly possible to identify layers of glosses by different glossators from different times here (2009: 1264). With the current state of research, the assumption that progressive spellings exist in the Maihingen Gospels needs to remain speculative. Further analyses of the manuscript, an

7 However, Seiler’s assumption that scribes who were familiar with the early medieval Latin grammarians’ advice to write a for the affricate (Donat) and for the fricative (Isidor of Seville) actually followed this advice and wrote a (2011: 177) does not seem convincing to me. This thesis would require further analyses of the training of the glossators who used the . In general, they do not seem to be uneducated at all. 8 E.g. the consequent writing of the a-Umlaut and the reduction of unstressed vowels (Sonderegger 1961: 265–268).

Medieval Glossators as Agents of Language Change   

   47

edition of the recently discovered glosses, as well as a differentiation of layers of glosses (if possible) are necessary to reach a secure interpretation of the spellings. Another prerequisite for further linguistic and contextual interpretations would be a functional analysis of the OHG glosses in comparison with the Old English glosses (Glaser 1997: 5). However, such striking spellings occur in other manuscripts with dry-point glosses as well (Ernst and Glaser 2009: 1008–1018). The following section (2.2.) focuses on the glosses in a manuscript from Freising that has been edited and analysed both in grammatical and contextual respects by Oliver Ernst (2007: 137–265). An analysis of the most striking glosses (2.2.2.) will reveal that they all have progressive spellings in common. I will argue that such forms result from the specific communicative conditions of the glossators (2.2.3.).

2.2 The Glossators of a Manuscript from Freising (Clm 6263) 2.2.1 Overview of the Manuscript and the Glosses The Clm 6263 was written at the beginning of the 9th century in Freising, in today’s southern Germany.⁹ It includes homilies by Gregory the Great on the gospels, texts that were used extensively in the time of the Carolingians. Parish priests of the 8th and 9th centuries were obliged to own homiliaries for the liturgy. In cloisters, Gregory’s homilies were used in liturgy as well as for private studies and for learning Latin at school (Fiedrowicz 1997: 24). Several people worked with the Clm 6263 in early medieval times. In his edition, Ernst is able to differentiate four monks who inserted dry-point glosses and three monks with feather quill glosses, all of them probably from Bavaria around the 10th century (2007: 254). Altogether, there are 45 OHG dry-point glosses, 10 OHG feather quill glosses and some glosses in Latin, both in dry-point and feather quill. As usual, several entries with styli cannot be identified linguistically (Ernst 2007: 202). Ernst’s detailed analysis of parallel glosses (2007: 217–230), i.e. glosses in other manuscripts with the same text that are identical to the glosses in the Clm 6263, is useful to ascertain if the glosses are original or copied. More than 400  manuscripts with Gregory’s homilies have been transmitted to us (Ernst 2007: 144), of which 27 have OHG glosses (Bergmann 2009: 542). Five of the

9 For references to descriptions of the manuscript and older editions of the glosses see Ernst (2007: 137; 144).

48   

   Markus Schiegg

ten feather quill glosses in the Clm 6263 show parallels in other manuscripts, whereas the dry-point glosses seem to be original entries, inserted independently of other manuscripts.

2.2.2 Peculiarities in the Glosses of the Clm 6263 Some of the dry-point glosses in the Clm 6263 show striking spellings. The following glosses are written by one individual (“glossator 1”) who scratched his glosses into the manuscript with a sharp stylus. Most of the OHG dry-point glosses in this manuscript (28 of 45) are by this glossator. Interestingly, these glosses share their striking phenomena with those from the Maihingen Gospels: – for in glassannem (‘conceded’; glosses Latin concessa) (Ernst 2007: 170–171) as well as in das (conjunction ‘that’; glosses Latin ne and is part of a longer syntagma; OHG das upi si ueliho¹⁰ glosses ne si quem) (Ernst 2007: 165–166). – for [ŋ] in biduan (‘urged’; glosses urguetur) (Ernst 2007: 152–153) as well as in rafsun (‘reproach’; glosses inuectionis) (Ernst 2007: 175–176). These glosses are from the 10th century, about two centuries later than the glosses in the Maihingen Gospels. Nevertheless, these two phenomena are still stunningly progressive because they do not appear in grammars as “regular” spellings until the 12th and 14th centuries respectively (2.1.2.). Other progressive forms like the reduction of unstressed vowels in alse (instead of also; ‘so’) (Ernst 2007: 184) are quite usual from the end of the 9th century on (Braune 2004: §76) so that it is not surprising to find them in these glosses from the 10th century. Two glosses, however, even lose their unstressed vowels. In glassannem, we would expect a gi-prefix (or if reduced: a ge-prefix) as well as an -emu or -emo ending (Braune 2004: §248; Ernst 2007: 171). The second gloss, flio (for f[ir]lio[san]; ‘to forfeit’; glosses intereat), shows both syncope and an abbreviated ending (Ernst 2007: 181). Ernst and Glaser (2009: 1017) assume that syncope particularly occurred in spoken language and spontaneous speech. This can again be validated with examples from Sonderegger’s analysis of the St. Galler Vorakte (1961: 273) as well as with text types that resemble the spoken language like the Paternoster from Freising and even Otfrid’s verses (Ernst and Glaser 2009: 1017).¹¹

10 The reading of the last letter is not secure. 11 See footnote 31.

Medieval Glossators as Agents of Language Change   

   49

Abbreviations are characteristic for this glossator. Sometimes he indicates abbreviated words, for example with a short horizontal line through the letter d in the gloss uerd (abbreviated for uerdan ‘to become’) (Ernst 2007: 169). Some abbreviations are more extensive. In the gloss pi, we can only see the beginning of a noun (probably piscof, ‘bishop’; glosses apostolus) (Ernst 2007: 174). Another example is the gloss ig, which is perhaps the ending of the adjective garawig (‘equipped’) and glosses reparari (Ernst 2007: 168). One gloss in this manuscript by another glossator (“glossator 4”) is worth mentioning because its peculiar spelling is also classified as progressive in Ernst and Glaser (2009: 1013). This glossator’s 13 glosses are not scratched but imprinted into the manuscript and can thus be differentiated from the previous glosses (Ernst 2007: 237). The gloss reads taz (‘that’) and glosses the Latin quod (Ernst 2007: 187).¹² The interesting letter in this gloss is the ; in earlier texts Germ. /þ/ had shifted into a Bavarian/Alemannic or a Frankish . Braune (2004: §167, Anm. 9) judges the spelling as a writing error for . In her edition of the Clm 6300, a manuscript from Freising that has several dry-point glosses from the 8th/9th century, Glaser (1996: 410) found seven spellings for Germ. /þ/. She notices that most of them are pronouns (e.g. teru, ‘the’) as parts of a syntagma after the preposition mit (‘with’) and concludes that such spellings cannot be occasional mistakes. Instead, this could be a type of assimilation known as Notkers Anlautgesetz from documents of the 10th century (Glaser 1996: 459), where lenis plosives become fortis, (a) if the word before ends with a fortis plosive or (b) generally at the beginning of a sentence (see Braune 2004: §103). Hence, we again encounter quite a progressive spelling in a dry-point gloss. To come to a preliminary conclusion, the glosses in the Clm 6263 show peculiarities that deviate from the picture of OHG that grammars present. The drypoint glosses by “glossator 1” have strikingly progressive forms in common. The progressivity seems to be quite systematic, as we encounter certain categories that appear regularly in dry-point glosses. The following section will reconstruct the communicative conditions of the glossator and explain the emergence of such forms.

12 In contrast to “glossator 1”, this glossator obviously writes the more common for Germ. /t/.

50   

   Markus Schiegg

2.2.3 Communicative Conditions of the Glossators To reconstruct the communicative conditions of a glossator is a demanding task and any conclusions drawn often remain speculative. However, the glosses in the Clm 6263 give some hints about the monks who glossed, about their audience and their intentions. As Gregory’s homilies were used in different contexts in medieval times (2.1.1.), it is quite probable that the seven glossators of this manuscript also glossed in diverse contexts. An analysis of the places in the manuscript that are glossed can provide initial clues about the glossators’ intentions. “Glossator 1” focused almost exclusively on the beginning of the text. 25 of his glosses are between fol. 2v and 36v; only three glosses appear on later pages. Ernst interprets this arrangement of glosses as results from studies that had been conducted independently from the chronology of the liturgical year (2007: 255).¹³ Thus, this glossator’s reason for glossing was probably not for the regular preparation of sermons, but rather for an individual study of the text. The other glossators, however, each glossed selected homilies, which could indicate concrete preparations for the presentation of the respective homilies. Other hints for the contexts of glossing are the non-linguistic entries with dry-point. On pages 183v–184r and 190r, there are vertical dashes that indicate speech pauses.¹⁴ On fol. 188r, Ernst discovered horizontal, ascending lines that he interprets as signs to raise one’s voice while presenting the text (2007: 257). Such phonetic aids indicate that the manuscript was also used in presentations,

13 The arrangement of the homilies in the Clm 6263 obviously does not conform to the liturgical year (for further references see Ernst 2007: 255). 14 Such entries have hardly been analysed in glossographic research yet. One exception is the study by Robinson (1973) on the purpose of Anglo-Saxon syntactical glossing in a Latin manuscript. The reason for the neglect of this phenomenon in glossography is probably the non-linguistic character of these entries. However, they could provide essential hints for the reconstruction of the contexts of glossing and should be included in analyses. Dashes such as the ones in the Clm 6263 could be classified as phonetic glosses (see Wieland 1983) for the most frequently cited functional classification of glosses). In my analysis of the manuscript Augsburg, Archiv des Bistums, Handschrift 10 (Schiegg 2012), I encountered several vertical dashes in dry-point as well. For example, on fol. 76v they occur in two functions. A segmental use of dashes marks the accent of a word and thus aids pronunciation (e.g. ignóras; prouincíae; feruóre). The suprasegmental use is similar to the Clm 6263. Here, the dashes mark speech pauses. They support the existing punctuation marks or offer additional ones (e.g. before the conjunction quia or atque). Sometimes, they are used as a hint for the lector to stress the contrast between two opposing ideas (e.g. […] quorum unus euticius | alter vero florentus dicebatur.).

Medieval Glossators as Agents of Language Change   

   51

probably in sermons that had often been read directly from a manuscript in medieval times (2007: 258). In the Clm 6263, such dashes especially appear in a chapter where “glossator 1” did not gloss anything, so they probably stem from “glossator 3” or “glossator 4” who glossed with dry-point in those passages. This correlates to our assumptions about the glossators: “glossator 1” supposedly did not gloss to prepare a sermon and the other glossators used the manuscript for public presentations. The different types of glosses by “glossator 1” reveal more details about his person and intentions. He employs several syntactic glosses that ease the connection of sentences (e.g. the glossing of cui with demo sin; ‘whom’ [Ernst 2007: 160]) or explain difficult constructions (e.g. the glossing of the Latin ablative absolute concessa potestate praedicationis with the past participle glassannem [Ernst 2007: 170]). Glosses in which inflectional endings and parts of the stem are abbreviated (e.g. pi) only give lexical information. Other abbreviations without the stem (e.g. ig) are reduced to grammatical information. Thus, the glosses by “glossator 1” show a high variety of functions, dependent on the specific difficulties of the respective passages. His glosses are not formulaic but show sophistication and a high degree of understanding so that he was probably not a beginner in Latin, and the glosses were probably not produced at school (Ernst 2007: 256). The abbreviated glosses by “glossator 1” can shed further light on his communicative conditions. The two examples above are abbreviated so intensively that it is hardly possible for a stranger to decipher their meaning spontaneously. That is why Ernst (2007: 257) interprets them as private aides de memoire that may stem from a monk’s individual work with the text. Another interpretation of heavily abbreviated glosses is that they could be used as a hint by a teacher for himself to give an explanation or an excursus on a certain passage in class (Ernst 2007: 568). What both possible contexts have in common is that the glossator himself was the intended audience. In general, abbreviated glosses often occur together with the technique of dry-point glossing (Ernst 2009: 300).¹⁵ Such glosses do not need the preparation of ink, and styli were available in each monk’s cell. So this technique of glossing seems appropriate for spontaneous note-making. Moreover, dry-point

15 Analyses of abbreviations in glosses have been sparse until recently due to the lack of quantitative studies, Ernst’s assumption of a correlation between dry-point glossing and abbreviations is based on his observations. However, he hints at the fact that abbreviations seem to be quite common in Old English scratched glosses as well. A study that he does not cite is Scott Gwara’s analysis of the manuscript Corpus Christi College, Cambridge MS 326, in which Gwara again found “hundreds” (1997: 202) of abbreviated Old English scratched glosses (also see fn. 23).

52   

   Markus Schiegg

glosses are often barely visible or legible for anybody except the glossator himself and are often intended for private use (Glaser and Nievergelt 2009: 225). To conclude, several factors indicate that “glossator 1” glossed in the course of his private studies. The position of his glosses speaks against glossing for liturgy; the sophistication of the glosses against the use at school. In addition, his dry-point technique as well as the several abbreviations hint at spontaneity. Such private and informal situations seem to evoke progressive spellings, while other texts that are written in planned situations and get copied several times remain more conservative. Through dry-point glosses we gain access to a progressive phoneme system that does not appear in most other texts until the Middle High German period or even later. In the following section, I contextualise these observations with a diachronic analysis of the communicational space of East Francia and illustrate systematically to what extend glossators can be considered as agents of language change.

3 Glossators as Agents of Language Change 3.1 Language Change 3.1.1 Language Change and Dry-point Glosses The findings in the preceding section have documented two different facets of language change that can be encountered in glosses. In manuscripts like those from Echternach, glossators for the first time write down their vernacular language, which results in a mixture of Latin and OHG in medieval manuscripts. In addition, the progressive spellings both in the Maihingen Gospels and the Clm 6263 indicate language changes that will occur much later in other texts. How can these two points be combined and how do these changes affect the communicational space in East Francia? Following Eugenio Coseriu (1958: 78–80), Wulf Oesterreicher differentiates between three aspects of language change (2001: 224): (1) innovation, (2) adoption and (3) shifts in markedness. While step (1), innovation, can be encountered in individual texts, (2), adoption, appears in historic languages. This step is necessary to call a deviation language change (2001: 223). Step (3) is a type of change where an already adopted innovation in one variant moves into another place inside the variational space and changes its diatopic, diastratic or diaphasic markedness. The technique of vernacular glossing is an innovation that was brought to Echternach and other Frankish cloisters through the missionary work of English

Medieval Glossators as Agents of Language Change   

   53

and Irish monks. This technique was not only used by these foreign monks, but was soon adopted by local recruits. The vernacular glosses in the Maihingen Gospels, both in Old English and OHG, are suitable examples of this process. Vernacular glossing quickly spread from cloister to cloister. The striking spellings that have been presented in the preceding sections are quite similar in the two manuscripts that were glossed by different people in different places and times. The same spellings can be found in other dry-point glosses (Ernst and Glaser 2009) so that they are certainly not idiosyncrasies of the respective glossator. Thus, both aspects, written vernacular and progressive forms, can be characterised as language change. Aspect (3), shifts in markedness in the communicational space of East Francia in general, will be investigated in the following sections. Koch and Oesterreicher’s Model of the Conceptual Continuum offers a useful framework to trace the medial and conceptual changes.

3.1.2 The Model of the Conceptual Continuum The Model of the Conceptual Continuum illustrates the central principles of organisation in variational spaces (in multilingual societies: communicational spaces) and was developed by Peter Koch and Wulf Oesterreicher (1985; see recently 2007 and Koch 2010). This model is valid for any speech community and is derived from universal patterns of human communication.

C

language of immediacy

graphic code phonic code

B

language of distance

A

Figure 1: The Model of the Conceptual Continuum (adapted from Koch and Oesterreicher 1985: 23).

54   

   Markus Schiegg

The authors differentiate between two dimensions: medium and conception. Discourse can either be represented in the phonic or graphic medium. There is a clear dichotomy between this medial opposition – written text versus spoken utterance. The second dimension, conception, characterises the style or mode of an expression and is stretched as a continuum between two poles. On the left, there is language of immediacy (informal and unplanned utterances). On the right, language of distance (a literal, planned and elaborate kind of language). The two triangles visualise the fact that there is a great affinity between spoken utterances and immediacy and between script and the language of distance. However, medium transferability allows the phonic code to be written down (“Verschriftung”) as well as a written text to be read aloud. The communicative conditions of a discourse determine its conceptual profile and the strategies of verbalisation. So a spontaneous conversation in a private setting with familiar partners about a free topic usually leads to the use of a language of immediacy (A), while a planned, elaborate presentation for an anonymous audience – despite the phonic code – resembles the language of distance (B). Language of immediacy (“orality”) can also occur in the graphic code, for example in a private letter to a friend (C).¹⁶ Although such a letter is planned to some extent and usually has fixed topics, it is quite informal and private. In a recent article, Peter Koch applied the model to the history of languages and showed that such histories can actually only be conceptualised as histories of variational or communicational spaces (2010: 159). For the purposes of this article, the Model of the Communicational Continuum offers a useful framework to localise the changes inside the early medieval communicational space of East Francia. In the following sections, I will compare the situations in the 7th century (before glossing; 3.2.1.), at the beginning of the 8th century (after the advent of the first dry-point glosses; 3.2.2.) and finally in the 10th century (the existence of different OHG text types; 3.2.3.).¹⁷

16 For an overview of different “types of orality in text” with examples from Latin documents see Oesterreicher (1997). 17 I am aware of the massive political changes in the Frankish Empire between the 7th and 10th centuries that forbid the assumption of a continuous empire as well as a continuous communicational space. However, a thorough consideration of the changing circumstances of the Frankish Empire would go beyond the scope of this article and needs to be systematised in further work.

Medieval Glossators as Agents of Language Change   

   55

3.2 Changes in the Communicational Space of East Francia 3.2.1 The Communicational Space before Vernacular Glossing The early medieval Frankish Empire can be divided into West Francia with Romance languages and East Francia with Germanic languages. After the missionary activity, when monks like Boniface had transferred Christianity as well as the Latin language to East Francia, the communicational space in this area was shared by at least two languages – a local variety of OHG¹⁸ and Latin. So we can call this a diglossic situation¹⁹ inside a communicational space.

graphic code G3

G4

G2 language of immediacy

G1

P1

elaborate orality

P2

P3

P4

language of distance

phonic code

Figure 2: The Communicational Space before Vernacular Glossing.

18 The different varieties that show the changes of the Second Consonant Shift – each to a different degree – are subsumed under the term Old High German (e.g. Old Bavarian, Old Alemannic and Old Frankish). This does not imply one language, but, as Albrecht Greule stresses, describes an ensemble of Germanic tribal languages in the eastern part of Francia (2004: 142). Earlier research had assumed that the vernacular language from court (Karolingische Hofsprache) had been in supra-regional use (for the research history and further references on this topic see Ernst and Elspaß 2011: 255). However, this conception cannot be verified empirically and implicates a vernacular language of distance, which will be contested in this article. 19 I use the term diglossia in Fishman’s sense (1967), who extends Ferguson’s (1959) concept of diglossia to genetically not related (or historically distant) languages. The use of Latin, however, was restricted to cloisters and courts and the number of people who could read and write in the 7th century probably did not exceed a few hundred in this whole area. Thus, the majority of the population of East Francia was monolingual and only had control over the language of immediacy in the phonic medium (P1 ).

56   

   Markus Schiegg

Figure 2 gives a schematic depiction of this space, based on the Model of the Conceptional Continuum (3.1.2.). Areas in darker grey represent Latin, areas in light grey a variety of OHG; white areas are not verbalised. Sections in the phonic code are tagged with a P (in the graphic code: G). The indices Pn and Gn represent the conceptual degree of distance with ascending numbers. There is a clear split through this communicational space. At the pole of the language of immediacy, OHG is used in the phonic code (P1). Near the other pole, Latin is the written and spoken language in formal and supra-regional communication (P3 and G3). The communicative conditions of the separate sections of this figure will be discussed in the following. Novices had to learn Latin in cloister schools; their mother tongue was a variety of OHG, which they used for private and spontaneous conversations (P1).²⁰ As a conceptual continuum is an anthropological constant (3.2.1.), the area of P1 is not restricted to the utmost pole of immediacy, but also stretches to more formal and distant areas. The right margin of P1 is called elaborate orality and is manifested for example in oral poetry or ritual prayers (Koch and Oesterreicher 2007: 357). The pole of distance, however, was exclusively reserved for Latin as the language of the church, literacy and science (Ernst and Elspaß 2011: 263). Throughout medieval times, Latin was both the written (G3) and spoken (P3) Lingua Franca (Stotz 2002: 150). The reason for this was its supra-regional comprehensibility as well as its – in contrast to the vernacular – fully developed concepts for theology, science and administration. Stotz concedes, however, that the oral use of Latin was restricted. Spoken Latin must either have been based on written drafts (2002: 152) or was delimitated to situations with fixed linguistic requirements and little need for spontaneity, e.g. recitations of prayers, liturgies or schematic conversation at school²¹ (2002: 151). Thus, area P3 can be categorised as rather distant. The

20 We have evidence from the 10th century that some teachers forced their pupils to speak Latin at school, for example Ekkehart II. (d. 990) in St. Gall and Ohtrich (d. 981) in Magdeburg. Gottfried von Viterbo from Bamberg even notices that Latin was the only language in the whole area of the cloister. The situation was probably similar in the 7th century. Such testimonies, however, may be utilised as negative evidence for the fact that Latin was not used spontaneously and by choice, but that monks in private contexts still preferred to speak OHG (Stotz 2002: 152). Moreover, OHG scribbles of notes, short pagan verses and spells in Latin manuscripts support the thesis that the vernacular was constantly in use as a language of immediacy. 21 Herbert Penzl (1985: 240) gives several examples of the text type “Conversational education in a foreign language”. He mentions texts of exemplary Latin conversations that had been written in the 10th century for Anglo-Saxon novices.

Medieval Glossators as Agents of Language Change   

   57

blank area between spoken OHG and Latin (P2) visualises the different conceptual character of these two languages. Latin was the only language of writing in East Francia of that time. Even Germanic tribal laws like the Lex Baiuvaiorum were translated into Latin.²² Written Latin (G3) may be located in a similar horizontal area to spoken Latin (P3). Most texts used in cloisters are Christian, be it the Bible, texts by the church father or Christian interpretations of Vergil. The earlier style of Christian Latin can be characterised as sermo humilis (e.g. due to its paratactic syntax) and had been developed separately from the more sophisticated language of Latin literature (Stotz 2002: 35–40). In later centuries, the lucid style of the Bible was sometimes disregarded (Stotz 2002: 25). Gregory’s plain prose even fostered such critique that Melanchthon denied him his epithet ‘the Great’ (Berschin 1986: 307). However, some other Christian writers such as the late-antique Christian poet Prudentius, who wrote artistic hymns (Stotz 2002: 61), employed a highly sophisticated, literary style. He is considered to have been a widely read author in medieval times, which can be proved by the extraordinarily densely glossed texts (Stricker 2009: 497). To conclude, typical texts from the Bible and Gregory the Great’s writings can be located in the left part of G3, Prudentius in the right part. A systematic analysis of the conceptual character of the diverse texts used in cloisters does not seem to have been conducted yet, however. The written areas near the pole of immediacy (G1/G2) and the pole of the utmost distance (G4) are still empty. Being a foreign language of distance, written Latin of immediacy is not encountered in East Francia.²³ On the other hand, there had been a tremendous deterioration of writing abilities in the times of the Merovingians, which even led to problems with church law, e.g. depraved baptismal formulae (Angenendt 2001: 108). Besides regional pronunciations (Stotz

22 Some words in these tribal laws, however, remained in the vernacular. These are usually technical terms that were required to identify the delict and to determine its penalty. For example, touching a virgin or a married woman (‘horcrift’) resulted in a penalty of 6 solidi; pulling such a woman’s skirt above her knees (‘himilzorun’) cost 12 solidi, as did snatching off her headdress (‘walcwurf’) (von Schwind 1926: 355). 23 In West Francia this area could be filled with Latin dry-point glosses. However, linguistic analyses of Latin dry-point glosses have hardly been conducted yet. One exceptional study is a monograph by Michael McCormick (1992), who provides a detailed edition and analysis of hundreds of Latin dry-point glosses. Some of the glosses are probably from West Francia from the 9th century (1992: 31) and – again – show several abbreviations (1992: 9–10). In his linguistic analysis, McCormick notices progressive linguistic forms of Late Latin or Protoromance origin as well as glosses that replace Latin words that were no longer common in Romance languages (1992: 14).

58   

   Markus Schiegg

2002: 77), the degenerated abilities in Latin led to a reduction in its character as a language of distance so that both G4 and P4 must be left empty.

3.2.2 Changes with the Advent of Vernacular Dry-point Glosses When vernacular dry-point glossing appears in East Francia in the first half of the 8th century, the vernacular for the first time shifts from the phonic to the graphic medium.²⁴

graphic code G3

G4

G2 language of immediacy

G1 ʻVerschriftungʼ P1

elaborate orality

P2

P3

P4

language of distance

phonic code

Figure 3: Changes with the Advent of Vernacular Dry-Point Glosses.

24 There had already been written Germanic in runes before. However, the use of runes had been restricted to short inscriptions (e.g. on jewellery, weapons and stones) and, according to Derolez, runes were not a script for everyday use (1959: 3). Michelle Waldispühl recently developed a new method for a functional analysis of runes. Her minutely detailed catalogue of epigraphic parameters allows a systematic description of runes and – as a second step – new insights into their functionality. In contrast to Derolez, Waldispühl concludes that the high diversity of elaborateness hints at a multitude of intentions and functions in the use of Germanic runes (2011: 39). The only known runes in OHG are some dry-point glosses from St Gall that have recently been discovered by Nievergelt (2009). Such manuscript runes, however, have no direct connections to the Germanic runes, but were used in Christian contexts inter alia as experiments with foreign alphabets or for cryptography (2009: 73; see also Ernst 2007: 382). For now, I follow Koch, who does not consider runes as the beginning of the Verschriftung of the German language, but as rudimentary first steps in the approach to writing the vernacular (2010: 168).

Medieval Glossators as Agents of Language Change   

   59

This vertical shift from P1 to G1 is termed Verschriftung (Oesterreicher 1993: 272). In dry-point glosses, this shift tends to take place near the left pole of immediacy because such glosses had often been inserted in moments of privacy and spontaneity (also see for example the communicative conditions of the glosses in the Clm 6263; 2.2.3.). Hereby, no elaborate orality gets written down, which explains the broad white area to the right of G1. The tiny area left of G1 remains blank as well because direct transcriptions of spontaneous speech do not appear in everyday life (Koch 2010: 161).²⁵ A horizontal transfer alongside the axis of the conception is termed Verschriftlichung (Oesterreicher 1993: 272). Prerequisite for a language that can be used for any communicational need in the written medium are two processes: during the intensive extension (intensiver Ausbau), a language develops the structural needs for the functions of a language of distance (e.g. precise subordinate conjunctions). The extensive extension (extensiver Ausbau) describes the gradual move to the pole of the distance in a communicative community and the adoption of any possible communicational function (Oesterreicher 1993: 276). Effects of this process of the Verschriftlichung of German, however, had not yet appeared at the beginning of the 8th century. The progressive spellings of some dry-point glosses are not part of the Verschriftlichung, although they indicate changes that also appear in the spellings of more conservative vernacular texts centuries later. The following section provides a brief look at the communicational space in the 10th century and discusses to what extent the Verschriftlichung of German had taken place by then.

3.2.3 The Communicational Space in the 10th Century Charlemagne’s educational reform at the end of the 8th century had the central message errata corrigere, superflua abscindere, recta cohortare studemus.²⁶ This led to the correction as well as the standardisation of corrupted ecclesiastical texts (3.2.1.). As an aid for supra-regional communication, Charlemagne introduced a new typeface, the Carolingian Minuscle (Angenendt 2001: 312). All this led to an extension as well as a petrifaction (Koch and Oesterreicher 2007: 365)

25 Using technical aids, today’s linguists are able to make narrow phonetic transcriptions of spontaneous speech. 26 Cited from Haubrichs (1995: 72); translation: “We strive to correct the mistakes, to cut the outgrowths and to strengthen the right things”.

60   

   Markus Schiegg

of Latin as the language of distance.²⁷ By this point written Latin stretched to the utmost distance (G4).

graphic code

G4 G3 G2

language of immediacy

G1 P3

P2 P1

elaborate orality

P4

language of distance

phonic code

Figure 4: The Communicational Space in the 10th Century.

By the 10th century, several different OHG texts had come into existence (G1 and G2).²⁸ However, the use of the vernacular in the graphic code was still an exception and a marginal phenomenon. OHG texts were always embedded in Latin codices. The vernacular was usually inserted in empty spaces, between the lines, in the margins (e.g. glosses), at the bottom of pages (e.g. Muspilli) or on empty pages at the beginning and the end of a codex (e.g. Hildebrandslied). Most OHG texts have only been passed down to us because the Latin texts inside the same codex were worth storing.²⁹ In addition, the quantity of OHG is much lower than Latin. By far the largest amount of glosses is written in Latin (Bergmann 2003: 39), which often makes the glossators’ use of the vernacular surprising.³⁰ In addition, the

27 For West Francia this meant a final break of the continuous Latin variational space into the dichotomy between local idioms and a standardised Latin language (Koch and Oesterreicher 2007: 365). 28 Stephan Müller (2007) provides a new, well-organised collection of several prominent OHG texts. He suggests eleven categories of communicational areas, in which certain text types regularly appear in the vernacular (e.g. law and administration, missionary activities, everyday life, magic and medicine). 29 For the similar situation of the earliest Romance texts see Frank (1994: 100). 30 Obviously, some vernacular glosses provided aids for pupils who had low competence in Latin. By trend, advanced glossators chose the vernacular as well, when Latin paraphrases would have been too cumbersome, e.g. for technical terms: animals, plants, professions and weapons (Bergmann 2003: 46).

Medieval Glossators as Agents of Language Change   

   61

scope of vernacular text types was quite reduced. In several fields like religion or science, Latin remained the only common language. As a consequence, OHG texts do not occupy any position in the language of distance (G3/G4). Even literary texts like Otfrid’s harmony of the gospels show some characteristics of the language of immediacy (Ernst and Elspaß 2011: 271).³¹ The extension of German to more distant areas required several further centuries.³² Although OHG was not a language of distance, its texts are not limited to the area of utmost immediacy (G1). The scribes of OHG were monks in cloisters who regularly wrote religious texts. Their audience was clerical as well, so we can assume a certain degree of stylistic sophistication (Masser 1997: 50). In addition, most texts have a direct relation to a Latin text, be it glosses, translations of the Bible or formulae for confession (Ernst and Elspaß 2011: 255). These texts were copied again and again so that conservative linguistic forms were passed on for centuries. Some OHG feather quill glosses were still copied in the 13th–15th centuries (Stricker 2009). To conclude, we can expect a spread of OHG texts to the area G2, which can be classified as the beginning of the Verschriftlichung of German. The area G2 itself has a certain horizontal extension and consists of texts with different conceptual characteristics. Thus far very little attention has been paid to this so that there is no method to classify the conceptual characteristics of an OHG text. A model for such an enterprise may be Àgel and Hennig’s (2007) technique for counting concrete linguistic features of immediacy in a text (e.g. deictic expressions) and calculating its horizontal positioning in the conceptual continuum. However, I have strong doubts about whether this method can be applied to OHG texts. Due to their quantitative marginality, most OHG texts are singular testimonies and have an individual conceptual profile. Thus, the com-

31 Interestingly, OHG grammars regularly list peculiarities in Otfrid’s literary text together with striking spellings of glosses (e.g. reduction of unstressed vowels); see Braune (2004: §71, Anm. 4 or §77, Anm. 3) and Ernst and Glaser (2009: 1017). Perhaps such spellings are much more common than assumed. The language of immediacy was constitutive for OHG. The reason why this has not been noticed until now is probably the common practice of normalisation in text editions of medieval vernacular texts. 32 Mattheier notices that before the 14th century, there had not been any text genres that were not written in Latin first and then translated into German (2003: 212). A characteristic of the middle ages as well as early modern times was the “gradual displacement of Latin from the written domain” (2003: 231). However, throughout most of the history of the German language, there has been a foreign language at the pole of the distance, be it Latin, French or English. Mattheier even argues that today, German has “lost […] its position as a fully developed standard language” (2003: 239) because its relevance is diminishing in business, political life and science. Moreover, “[t]echnical terminologies, for example, are no longer developed in many scientific areas” (2003: 239).

62   

   Markus Schiegg

pilation of objective and comparable features of immediacy seems futile. Moreover, the vernacular transmission from the early medieval age significantly differs from our conception of a text. Such texts often provide only fragments of complex communicational situations. The vernacular was written down according to the specific pragmatic needs of a communicational situation (e.g. fragmentary drafts for charters in the St. Galler Vorakte). To identify the conceptual profile of such a text, its individual communicative context needs to be reconstructed minutely. To assess its conceptual profile, it would even be necessary to analyse each version of a copied text individually. On the surface, copied glosses appear to be identical with the original ones. However, they are inserted under completely different communicational conditions. When the two versions of the Paternoster from Freising (A and B) are compared, different kinds of gaps both below and above the word boundaries can be identified (Ernst and Elspaß 2011: 277). The configuration of the gaps can provide hints which can help us to reconstruct the contexts and intentions of a scribe (Masser 1997: 58).³³ In longer texts, the conceptual profile can even shift within a text itself. In the OHG Tatian, one of the scribes (“scribe ζ” in research literature) uses more modern forms than the others and even executes a final review to modernise several forms of the earlier scribes (Schmid 2007).³⁴ All this leads to the conclusion that due to the specifics and the complexities of the communicational conditions, the degree of immediacy of OHG texts can hardly be assessed from the mere collection of surface features. Reading a text in ancient as well as in medieval times meant reading it aloud (Fleischer 2009: 162). As a consequence, we encounter different traces of such reading processes like accents (2.2.2.), gaps between words or neumes in these texts. In our model, texts from the area G2 are transferred to the phonic medium P2. The same applies for the shift from G4 to P4. These two changes are again medial and have no conceptual implications. However, they fill spaces of the communicational space that had been empty before. To sum up, throughout early medieval time, Latin remained the language of distance and the vernacular the language of immediacy. Due to the standardi-

33 Based on Masser’s studies, Jürg Fleischer (2009) recently analysed the relevance of accents, word separation, punctuation and other phenomena for the reconstruction of prosodic information of OHG texts. 34 Hans Ulrich Schmid (2007) analysed the different kinds of corrections in the OHG Tatian. Besides corrections concerning the layout, “scribe ζ” corrected (and in most cases modernised) the language of the earlier scribes in several linguistic areas – graphematics, phonematics, morphology, syntax and lexis. Schmid’s observations show parallels with the findings on the progressive phenomena in dry-point glosses and could provide further hints for glossography about other types of progressive linguistic forms that were already existent in OHG times.

Medieval Glossators as Agents of Language Change   

   63

sation through the Carolingian Reform and the increase of different OHG texts, there was a shift to the direction of the distance, both in Latin and in the vernacular. This can be considered as the beginning of the Verschriftlichung of German. However, the horizontal extension of the OHG conceptual continuum remained rather narrow (Ernst and Elspaß 2011: 263). As a marginal phenomenon the German language did not replace Latin anywhere that early.³⁵ All in all, it is perhaps justified to ask, whether glossators can still be understood as agents of language change at all, if their changes affected only an area near the pole of immediacy. The concluding section will ponder this question.

4 Conclusion: Glossators as Agents of Language Change? Are glossators agents of language change? This question can be divided into two: (A) Does glossing result in language change? (B) Do glossators act as promoters of these changes; do they intend to change language? This article has mainly dealt with question (A). Glossing results in language change in two respects. Dry-point glosses are the first written documents in the vernacular, which leads to a considerable change inside the communicational space of East Francia. Now, for the first time the vernacular moves from the phonic to the graphic code (Verschriftung) and thus forms the beginning of written German. Later documents in OHG that had been located further away from the pole of immediacy, however, do not directly evolve from glosses. Earlier research tended to construct ascending stages from glossing, to interlinear translations, to translations and finally to a German language of literacy (Sonderegger 2003). Such a teleological perspective, however, implicitly devaluates glosses as the first, clumsy steps of a long process of the evolution of German. It does not take account of the fact that glossing is a fully developed and functionally adequate technique of studying. The marginality of OHG in general as well as the intentions of the glossators to study Latin forbids such a perspective.³⁶ A moderate conciliation could treat glossing as an unprecedented technique of writing the

35 One exception may be confessions that commonly appear in the vernacular in sacramentaries and according to Bernhard Bischoff have gained the Hausrecht (‘domiciliary rights’) there (1971: 103). 36 For further critique of such teleological models see März (1996) and Ridder and Wolf (2000).

64   

   Markus Schiegg

vernacular and as a prerequisite for, but not as a direct basis of, the conceptual extension (Verschriftlichung) of German. The progressive spellings in dry-point glosses, however, indicate linguistic changes that also take place in the language of further distance later. In variationist terms and from the perspective of the OHG language, progressive dry-point glosses indicate a later shift from marked to unmarked in the registers of the language of distance. This shift is not restricted to phonology and morphology, as the examples in this article have suggested. Glosses even provide hints for syntactical changes, e.g. the grammaticalisation of the demonstrative pronoun as an article with the function of a morphological marker (Schmid 2009: 1086), which again does not appear in text documents until centuries later. Thus, such private and spontaneous dry-point glosses can indeed narrow the “temporal gap between the introduction of new forms in speech and their first recordings in written texts” (Raumolin-Brunberg 1996: 17). To put everything in a nutshell, glossators have tremendous influences on language change in medial respects. Regarding the conceptual factor, they do not cause direct changes, but indicate later changes in the language of distance. Hence there are movements inside the communicational space. Glosses appear step by step in Frankish cloisters so that we can conclude that glossing results in language change and glossators can indeed be considered as agents of language change. To answer question (B), whether glossators intended to change their language or not, it is necessary to refer to their individual contexts and actual intentions. For example, “glossator 1” in the Clm 6263 intended to study a Latin text intensely and inserted notes in OHG that only he himself could use in later situations. Some of the other glossators of this manuscript prepared themselves for the presentation of a sermon. Hence the intention of all these glossators was not to create a written German language or to promote progressive linguistic forms.³⁷ They used the vernacular because it was practical for their needs. As a consequence, however, language change ensued (see question A). Thus, the individual intentions of the glossators deviate from the actual results. This effect can be characterised as an invisible hand process in Rudi Keller’s sense (2003: 113): The cumulation of several individual acts resulted in structures beyond the scope of the individuals, the Verschriftung and later the Verschriftlichung of the German language.³⁸

37 The situation is different for “scribe ζ” in the OHG Tatian. Apparently, he intentionally modernised the language of the earlier scribes (see footnote 34). 38 For a similar reasoning, see a recent article by Eckard Meineke. He argues that the cultivation of German, which he describes as the emancipation of the German language from its status

Medieval Glossators as Agents of Language Change   

   65

5 References Ágel, Vilmos & Hennig, Mathilde. 2007. Überlegungen zur Theorie und Praxis des Nähe- und Distanzsprechens. In Vilmos Ágel & Mathilde Hennig (eds.), Zugänge zur Grammatik der gesprochenen Sprache, 179–214. Tübingen: Niemeyer. Angenendt, Arnold. 2001. Das Frühmittelalter. Die abendländische Christenheit von 400 bis 900. 3rd ed. Stuttgart: Kohlhammer. Baldzuhn, Michael. 2001. Schriftliche Glosse und mündlicher Unterricht. Das Beispiel der älteren lateinisch-volkssprachlich glossierten Aviane (9.–11. Jahrhundert). In Rolf Bergmann, Elvira Glaser & Claudine Moulin-Fankhänel (eds.), Mittelalterliche volkssprachige Glossen, 485–512. Heidelberg: Winter. Bergmann, Rolf. 1977. Mittelfränkische Glossen. Studien zu ihrer Ermittlung und sprachgeographischen Einordnung. 2nd ed. Bonn: Ludwig Röhrscheid. Bergmann, Rolf. 2003. Volkssprachige Glossen für lateinkundige Leser? Sprachwissenschaft 28. 29–55. Bergmann, Rolf. 2009. Überblick über die Gregor-Glossierung. BStH. 525–548. Berschin, Walter. 1986. Biographie und Epochenstil im lateinischen Mittelalter. Band 1. 5 volumes. Stuttgart: Hiersemann. Bischoff, Bernhard. 1971. Paläographische Fragen deutscher Denkmäler der Karolingerzeit. Mittelalterliche Studien 5. 101–134. Braune, Wilhelm & Ingo Reiffenstein. 2004. Althochdeutsche Grammatik. Band 1. Laut- und Formenlehre. 15th ed. Tübingen: Niemeyer. BStH = Bergmann, Rolf & Stefanie Stricker (eds.). 2009. Die althochdeutsche und altsächsische Glossographie. Ein Handbuch. 2 volumes. Berlin/New York: de Gruyter. Coseriu, Eugenio. 1958. Sincronía, diacronía e historia. El problema del cambio lingístico. Montevideo: Universidad de Montevideo. Derolez, René. 1959. Die Hrabanischen Runen. Zeitschrift für deutsche Philologie 78. 1–19. Derolez, René. 1992. Anglo-Saxon Glossography: A Brief Introduction. In René Derolez (ed.), Anglo-Saxon Glossography, 11–43. Brussels: WLSK. Ernst, Oliver. 2007. Die Griffelglossierung in Freisinger Handschriften des frühen 9. Jahrhunderts. Heidelberg: Winter. Ernst, Oliver. 2009. Kürzung in volkssprachigen Glossen. BStH. 282–315. Ernst, Oliver & Stephan Elspaß. 2011. Althochdeutsche Glossen als Quellen einer Sprachgeschichte ‘von unten’. NOWELE 62/63. 249–283. Ernst, Oliver & Elvira Glaser. 2009. Graphematik und Phonematik. BStH. 995–1019. Ferguson, Charles A. 1959. Diglossia. Word 15. 325–340.

as a language of the lower classes, was not a final or causal process, but must be considered as an invisible hand process because it came about as a by-product of other projects like the Christianisation of the Germanic tribes (2011: 109). See also Beatrix Weber’s study of the late medieval ‘Elaboration’ (‘Ausbau’ ) of English. She demonstrates that singular and disparate acts of writing texts in English gained momentum and cumulated in the Verschriftlichung of the English language (2009: 3).

66   

   Markus Schiegg

Fiedrowicz, Michael. 1997. Gregor der Grosse. Homiliae in Evangelia. Evangelienhomilien. Erster Teilband. In Norbert Brox et al. (eds.), Fontes Christiani. Zweisprachige Neuausgabe christlicher Quellentexte aus Altertum und Mittelalter. Freiburg/Basel: Herder. Fishman, Joshua A. 1967. Bilingualism with and without diglossia; Diglossia with and without bilingualism. Journal for Social Issues 23,2. 29–38. Fleischer, Jürg. 2009. Paleographic clues to prosody? – Accents, word separation, and other phenomena in Old High German manuscripts. In Roland Hinterhölzl & Svetlana Petrova (eds.), Information Structure and Language Change. New Approaches to Word Order Variation in Germanic, 161–189. Berlin/New York: de Gruyter. Fortson, Benjamin W. 2010. Indo-European language and culture. An Introduction. Chichester: Wiley-Blackwell. Frank, Barbara. 1994. Die Textgestalt als Zeichen. Lateinische Handschriftentradition und die Verschriftlichung der romanischen Sprachen (ScriptOralia 67). Tübingen: Narr. Glaser, Elvira. 1996. Frühe Griffelglossierung aus Freising. Göttingen: Vandenhoeck & Ruprecht. Glaser, Elvira. 1997. Addenda und Corrigenda zu den althochdeutschen Griffelglossen aus Echternach. In Elvira Glaser & Michael Schlaefer (eds.), Grammatica ianua artium. Festschrift für Rolf Bergmann zum 60. Geburtstag, 3–20. Heidelberg: Winter. Glaser, Elvira & Andreas Nievergelt. 2009. Griffelglossen. BStH. 202–229. Greule, Albrecht. 2004. Über die Anfänge deutscher Sprachkultur und Sprachkultivierung. In Albrecht Greule, Eckart Meineke & Christiane Thim-Mabrey (eds.), Entstehung des Deutschen. Althochdeutsch und Altniederdeutsch – Wörter und Namen – Texte und Glossen, 133–142. Heidelberg: Winter. Gwara, Scott. 1997. Further Old English Scratched Glosses and Merographs from Corpus Christi College, Cambridge MS 326 (Aldhelm’s prosa de virginitate). English Studies 78,3. 201–236. Hägele, Günter & Gregor Wurst. 2010. novum opus ex veteri. Vom Judas-Evangelium zur Furtmeyr-Bibel. Biblische und apokryphe Handschriften aus Spätantike und Mittelalter. Exhibition catalogue. Augsburg: University Library. Haubrichs, Wolfgang. 1995. Geschichte der deutschen Literatur von den Anfängen bis zum Beginn der Neuzeit. Bd. I/1: Die Anfänge: Versuche volkssprachiger Schriftlichkeit im Mittelalter. 2nd ed. Tübingen: Niemeyer. Henkel, Nikolaus. 2009. Glossierung und Texterschließung. Zur Funktion lateinischer und volkssprachiger Glossen im Schulunterricht. BStH. 468–496. Hofmann, Josef. 1963. Altenglische und althochdeutsche Glossen aus Würzburg und dem weiteren angelsächsischen Missionsgebiet. Beiträge zur Geschichte der deutschen Sprache und Literatur (PBB Halle) 85. 27–131. Keller, Rudi. 2003. Sprachwandel. Von der unsichtbaren Hand in der Sprache. 3rd ed. Franke: Tübingen. Koch, Peter. 2010. Sprachgeschichte zwischen Nähe und Distanz: Latein – Französisch – Deutsch. In Vilmos Ágel & Mathilde Hennig (eds.), Nähe und Distanz im Kontext variationslinguistischer Forschung, 155–206. Berlin/New York: de Gruyter. Koch, Peter & Wulf Oesterreicher. 1985. Sprache der Nähe – Sprache der Distanz. Mündlichkeit und Schriftlichkeit im Spannungsfeld von Sprachtheorie und Sprachgeschichte. Romanistisches Jahrbuch 36. 15–43. Koch, Peter & Wulf Oesterreicher. 2007. Schriftlichkeit und kommunikative Distanz. Zeitschrift für germanistische Linguistik (ZGL) 35. 346–375.

Medieval Glossators as Agents of Language Change   

   67

Lapidge, Michael. 1982. The Evidence of Latin Glosses. In Nicholas Brooks (ed.), Latin and the Vernacular Languages in Early Medieval Britain, 99–140. Leicester: Leicester University Press. März, Christoph. 1996. Von der Interlinea zur Linea. Überlegungen zur Teleologie althochdeutschen Übersetzens. In Joachim Heinzle (ed.), Übersetzen im Mittelalter. Cambridger Kolloquium 1994, 73–86. Schmidt: Berlin. Masser, Achim. 1997. Wege zu gesprochenem Althochdeutsch. In Elvira Glaser, Michael Schlaefer & Ludwig Rübekeil (eds.), Grammatica ianua artium. Festschrift für Rolf Bergmann zum 60. Geburtstag, 49–70. Heidelberg: Winter. Mattheier, Klaus J. 2003. German. In Ana Deumert & Wim Vandenbussche (eds.), Germanic Standardizations. Past to Present. (Impact: Studies in language and society, 18), 211–244. Amsterdam/Philadelphia: Benjamins. McCormick, Michael. 1992. Five Hundred Unknown Glosses from the Palatine Virgil. Città del Vaticano: Biblioteca Apostolica Vaticana. Meineke, Eckard. 2011. Textgebundene Formen der lateinisch-deutschen Zweisprachigkeit im frühen Mittelalter. In Michael Baldzuhn & Christine Putzo (eds.), Mehrsprachigkeit im Mittelalter. Kulturelle, literarische, sprachliche und didaktische Konstellationen in europäischer Perspektive, 109–145. Berlin/New York: de Gruyter. Moulin, Claudine & Elvira Glaser. 2009. Echternacher Glossenhandschriften. BStH. 1257–1278. Müller, Stephan. 2007. Althochdeutsche Literatur. Eine kommentierte Anthologie. Stuttgart: Reclam. Nievergelt, Andreas. 2009. Althochdeutsch in Runenschrift. Geheimschriftliche volkssprachige Griffelglossen. In Joachim Heinzle (ed.), Zeitschrift für deutsches Altertum (ZfdA) (Beiheft 11). Stuttgart: Hirzel. Nievergelt, Andreas. 2011. Nachträge zu den althochdeutschen Glossen. Sprachwissenschaft 37,4. 375–421. Ó Cróinín, Dáíbhí. 1999. The Old Irish and Old English Glosses in Echternach Manuscripts. In Michele Camillo Ferrari (ed.), Die Abtei Echternach 698 – 1998, 85–101. Luxembourg: Cludem. Oesterreicher, Wulf. 1993. Verschriftung und Verschriftlichung im Kontext medialer und konzeptioneller Schriftlichkeit. In Ursula Schäfer (ed.), Schriftlichkeit im frühen Mittelalter (ScriptOralia, 53), 267–292. Tübingen: Narr. Oesterreicher, Wulf. 1997. Types of Orality in Text. In Egbert J. Bakker & Ahuvia Kahane (eds.), Written Voices, Spoken Signs: Tradition, Performance, and the Epic Text, 190–214. Cambridge (Mass.)/London: Harvard University Press. Oesterreicher, Wulf. 2001. Sprachwandel, Varietätenwandel, Sprachgeschichte. Zu einem verdrängten Theoriezusammenhang. In Ursula Schaefer (ed.), Varieties and Consequences of Literacy and Orality / Formen und Folgen von Schriftlichkeit und Mündlichkeit. Franz H. Bäuml zum 75. Geburtstag, 217–248. Tübingen: Narr. Paul, Hermann. 2007. Mittelhochdeutsche Grammatik. (Re-edited by Thomas Klein, Hans-Joachim Solms & Klaus-Peter Wegera. 25th edition). Tübingen: Niemeyer. Penzl, Herbert. 1968. The History of the Third Nasal Phoneme of Modern German. PMLA 83,2. 340–346. Penzl, Herbert. 1985. “Stulti sunt Romani.” Zum Unterricht im Bairischen des 9. Jahrhunderts. Wirkendes Wort 3. 240–248.

68   

   Markus Schiegg

Raumolin-Brunberg, Helena. 1996. Historical sociolinguistics. In Terttu Nevalainen & Helena Brunberg-Raumolin (eds.), Sociolinguistics and Language History. Studies Based on the Corpus of Early English Correspondence, 11–38. Amsterdam/Atlanta: Rodopi. Ridder, Klaus & Norbert Richard Wolf. 2000. Übersetzen im Althochdeutschen: Positionen und Perspektiven. In Wolfgang Haubrichs al. (eds.), Theodisca. Beiträge zur althochdeutschen und altniederdeutschen Sprache und Literatur in der Kultur des frühen Mittelalters. Eine internationale Fachtagung in Schönmühl bei Penzberg vom 13. bis zum 16. März 1997, 414–447. Berlin/New York: de Gruyter. Robinson, Fred C. 1973. Syntactical Glosses in Latin Manuscripts of Anglo-Saxon Provenance. Speculum 48,3. 443–475. Schiegg, Markus. 2012. Althochdeutsche Griffelglossen am Beispiel der Handschrift 10 des Archivs des Bistums Augsburg. Archive in Bayern 7. Schmid, Hans Ulrich. 2007. Zu den Korrekturen im althochdeutschen “Tatian” (Cod. 56 der Stiftsbibliothek St. Gallen). In Claudia Wich-Reif (ed.), Strukturen und Funktionen in Gegenwart und Geschichte. Festschrift für Franz Simmler zum 65. Geburtstag , 43–70. Berlin: Weidler. Schmid, Hans-Ulrich. 2009. Syntax. BStH. 1077–1088. Schulte, Wolfgang. 1993. Die althochdeutsche Glossierung der Dialoge Gregors des Großen. Göttingen: Vandenhoeck & Ruprecht. Schützeichel, Rudolf. 1964. Neue Funde zur Lautverschiebung im Mittelfränkischen. Zeitschrift für deutsches Altertum und deutsche Literatur 93. 19–30. Seiler, Annina. 2011. Litteras superfluas – Zum Gebrauch “überflüssiger” Buchstaben im Althochdeutschen, Altsächsischen, Altenglischen. In Elvira Glaser, Annina Seiler & Michelle Waldispühl (eds.), LautSchriftSprache. Beiträge zur vergleichenden historischen Graphematik, 167–183. Chronos: Zürich. Sonderegger, Stefan. 1961. Das Althochdeutsche der Vorakte der älteren St. Galler Urkunden. Ein Beitrag zum Problem der Urkundensprache in althochdeutscher Zeit. Zeitschrift für Mundartforschung 28. 251–286. Sonderegger, Stefan. 2003. Althochdeutsche Sprache und Literatur. 3rd ed. Berlin/New York: de Gruyter. Splett, Jochen. 2000. Lexikologie und Lexikographie des Althochdeutschen. In Werner Besch, Anne Betten, Oskar Reichmann & Stefan Sonderegger (eds.), Sprachgeschichte. Ein Handbuch zur Geschichte der deutschen Sprache und ihrer Erforschung. Vol. 2, 1196–1206. 4 volumes. Berlin/New York: de Gruyter. Steinmeyer, Elias v. & Eduard Sievers. 1879–1922. Die althochdeutschen Glossen. 5 volumes. Berlin: Weidmannsche Buchhandlung. Stotz, Peter. 2002. Handbuch zur lateinischen Sprache des Mittelalters. Band 1. Einleitung. Lexikologische Praxis. Wörter und Sachen. Lehnwortgut. 5 volumes. München: Beck. Stricker, Stefanie. 2009. Entwicklungen der Glossographie im 13. Jahrhundert und später. BStH. 1553–1573. Stricker, Stefanie. 2009. Überblick über die Prudentius-Glossierung. BStH. 497–510. von Schwind, Ernst. 1926. Lex Baiwariorum (Monumenta Germaniae Historia. Leges. 02/5.2) , 180–491. Hannover: Hahn. Waldispühl, Michelle. 2011. Methodische Überlegungen zur Rekonstruktion der Schriftverwendung bei den südgermanischen Runeninschriften. In Elvira Glaser, Annina Seiler & Michelle Waldispühl (eds.), LautSchriftSprache. Beiträge zur vergleichenden historischen Graphematik, 31–46. Chronos: Zürich.

Medieval Glossators as Agents of Language Change   

   69

Weber, Beatrix. 2009. Sprachlicher Ausbau. Konzeptionelle Studien zur spätmittelenglischen Schriftsprache (Arbeiten zur Sprachanalyse 52). Frankfurt am Main: Peter Lang. Wieland, Gernot R. 1983. The Latin glosses on Arator and Prudentius in Cambridge University Library, ms GG.5.35. Toronto (Ont.): Pontifical Institute of Medieval Studies. Wieland, Gernot R. 1985. The Glossed Manuscript: Classbook or Library Book? Anglo-Saxon England 14. 153–173. Yoshizawa, Yasukazu & Yoshinori Kobayashi. 1999. Stylus-Impressed Writing on the Dunhuang Manuscripts. IDP News (Newsletter of The International Dunhuang Project) 13. 4–5.

Roger Wright

4 How scribes wrote Ibero-Romance before written Romance was invented Sources: The document investigated in this paper was written in the monastery of Vairão in the diocese of Porto, and is dated 18th August 991 CE. The document is now in Lisbon (Archivo Nacional, Ordem de S. Bento, Mosteiro do Salvador de Vairão [PT-TT-MSV/3/1/3], maço 1, no.3, recto), and was printed in the Portugaliae Monumenta Historica as no. 163. Abstract: In much of Western Europe Latin continued to be a living, spoken language after the Roman Empire, and the research of generations of Romanists has helped show that many phonetic changes occurred in the first few centuries CE. Spelling rules, however, do not change until somebody authoritative changes them. Thus in the same way as written English spellings have gradually diverged markedly from spoken English pronunciations over the last five centuries, so Latin speech and writing gradually diverged from each other during the first millennium CE. So in the tenth century, with the exception of the intellectuals in Catalonia who were in touch with French culture, the Peninsular scribes therefore learnt to write words in their ancient traditional form while speaking the incipient Ibero-Romance of the age. The analysis of a document from the monastery of Vairão in the diocese of Porto, dated 18th August 991 CE, shows that the morphology, the semantics, the syntax and the vocabulary of this document are those of the speech of tenthcentury Galicia. That speech is usually referred to by modern specialists as IberoRomance. The only aspect of the text that is not Ibero-Romance is the spelling, the written form of the words, because the scribe was not trying to reproduce the phonetics of his speech in a direct transcription but to achieve the traditionally correct forms that he had been taught. In contrast to the view of many Latinists, who tend to judge these texts as “bad” or “corrupt Latin”, and from whose perspective these texts have been written with astounding ignorance, incompetence and stupidity, the author would like to show that the scribes worked intelligently in a difficult context where the written language as taught was ever diverging from the one they spoke.

Specialists in Historical Linguistics have usually preferred to investigate languages that have never been written or the prehistoric stages of literate languages.

72   

   Roger Wright

Trask’s Dictionary of Historical and Comparative Linguistics (2000), for example, has surprisingly little to say about the way we should use written data in an analysis of speech. But this reluctance has not arisen because written documentation is irrelevant to the study of the history of spoken languages; indeed, the direct data are all written, and there are general points that can be made about their nature as well as individual cases which can be explored. Over the last 35 years (e.g. Wright 1982, 1995, 2003) I have argued that the development of Latin into the Romance Languages in the Early Middle Ages is not a special case, but broadly followed the general pattern; these arguments are summarised here. When people are learning to write for the first time, including modern primary school children and early medieval apprentice scribes, their main difficulty concerns the spelling. There need be no problem about the syntax, because the word order and the constructions can be exactly the same in writing as they would be in speech and do not have to be altered. It is, of course, true that speech tends to operate in less formal registers than writing does, but there is no problem about reproducing the word order and constructions of any register of speech on the page if we wish. Nor need there be any great problem about the vocabulary, because there are very few words, if any, that can only be used in speech and never at all in writing, or vice versa. The great difficulty for the primary pupil and the early medieval apprentice scribe comes in the spelling of the words. The written forms of words are standardised, and even those developed within an alphabetical system are not phonetic script except, perhaps, at the moment in which they are initially invented. That is why it is, or was, normal for schoolchildren to come home from school with lists of words to learn how to spell. Although there are correspondences between sounds and letters that are also essential to learn in the initial stages, these correspondences, in living languages, are not enough in themselves to enable us to read and write; if we try reading an English text aloud as if it were in a phonetic script, with a specific sound for each letter, we are unlikely to be intelligible. When alphabetical systems are first elaborated, these sound-letter correspondences usually inspire the initial written forms of each word. In real life it is not always as simple as that in practice, but even so, when studying the past, to some extent we have to assume that the correspondences applied then. In the case of my own surname, for example, in which fifty per cent of the letters are silent, philologists of the future would be right to deduce that once upon a time those letters had not been so silent, and that there was a phonetic motivation for their existence in the fifteenth century even if there is not now. A large number of English (and French) words have a standard written form which is not identical to any kind of phonetic or even phonemic transcription of their normal pronunciation. In the case of Latin, it is similarly assumed, probably rightly, that

How scribes wrote Ibero-Romance before written Romance was invented   

   73

the way in which words were usually written had corresponded to the sounds that were pronounced in these words at the time of the standardisation of their written form. In most cases, that form was fixed more than a century BCE. But sound changes were in progress all the time, even then. For example, we can tell from the metrical conventions used by Virgil and his contemporaries that words beginning with a letter h- and those ending with a letter -m counted for practical versification purposes as if they respectively began and ended with the adjacent vowel. The evidence of the later Romance languages shows that all such words did at some point lose the final [-m] and the initial [h-] sounds (except for the final nasal consonant of a few monosyllables; e.g. Latin CUM >Spanish and Italian con, ‘with’), and it is likely that these developments were already under way before the turn of the millennium from BCE to CE. Once the standard written form of a word has been fixed, schoolchildren and apprentice scribes naturally continue to be taught to use that form regardless of phonetic developments. In the case of Modern English and Modern French, this continues to be true; even though many English and French people make spelling mistakes when writing, the forms they are taught at school have only occasionally been updated to reflect the general phonetic changes that have affected speakers of any level of education. Latin was not immune from this. In much of Western Europe Latin continued to be a living spoken language after the Roman Empire, and the research of generations of Romanists, working with the knowledge of what the Romance languages were going to attest later, has helped show that many phonetic changes occurred in the first few centuries CE. But spelling rules do not change until somebody authoritative changes them. Thus in the same way as written English spellings have gradually diverged markedly from spoken English pronunciations over the last five centuries, so Latin speech and writing gradually diverged from each other during the first millennium CE. A normal written English text of the present day would seriously mislead a historical linguist of a thousand years in the future if he or she were to take it to be a phonetic transcription of what we actually say now; similarly, a Latin text of the tenth century CE from the northern Iberian Peninsula, written at a time when most of the standard forms taught to scribes had not changed but the pronunciations had developed considerably (to be more like what we now think of as being Old Spanish and Portuguese), is misleading if we take it to be a direct phonetic transcription of the writer’s speech. I refer here to the tenth-century Iberian Peninsula rather than to the Romance-speaking world more generally, because in the far North of that world, in Carolingian France, there had by that time already been implemented the important reforms in the nature of official writing and reading aloud which led to the creation of the separate ecclesiastical register that we now call “Medieval Latin”; these reforms only came to the Iberian Peninsula rather later (compare Wright 1982, 2003; Van

74   

   Roger Wright

Acker, Van Deyck and Van Uytfanghe 2008). And I refer to the northern half of the Iberian Peninsula because the southern half was under Muslim domination, where many people of all religions were bilingual in Arabic and Romance but by then, if literate at all, nearly always only in Arabic (Wright 2003: 158–174). So in the tenth century, with the exception of the intellectuals in Catalonia who were in touch with French culture, the Peninsular scribes who wrote the many texts which have survived to be studied a millennium later (by e.g. Davies 2007) had necessarily learnt to write words in their ancient traditional form while speaking the incipient Ibero-Romance of the age. They did not call it that, of course; it is, I think, legitimate for modern linguists to refer to the language spoken then in the Iberian Peninsula as “Romance”, or by the tenth century, “Ibero-Romance”, but the speakers did not do so themselves, and we need to remain aware that there can be awkward consequences if historical linguists refer to a language with a different name from that used by that language’s speakers. Sometimes this is necessary; we do not suppose that the speakers of Proto-IndoEuropean called their own language “Proto-Indo-European”, but we do not know what they did call it and we need to refer to it in some way. In this case, we do know that everybody still called their language “Latin”, whether spoken or written. The fact that we (the philologists of a millennium later) usually give their spoken language the name of “Romance” should not mislead us into thinking that the language they spoke and the language they wrote should be analysed as if they were two different languages. They still thought of it all as one language called lingua latina. This continued to be the case well into the second millennium; the word “Romance”, all over the Romance-speaking world, still had to wait another couple of centuries before being used in a linguistic sense, and then when it was, it was first used to refer specifically to the new written Romance modes adopted in the late twelfth and early thirteenth century rather than to speech (Wright 2003: 36–48). The document printed below, dated 18th August 991 CE, comes from the monastery of Vairão in the diocese of Porto. Vairão is north of the River Minho, so is now in Galicia in Spain, but at that time Porto and the region around it were also in Galicia (and thus part of the Asturian-Leonese kingdom), since Portugal had not yet taken its own independence. The document is now in Lisbon, and was printed in the Portugaliae Monumenta Historica as no. 163 (there are 184 documents from Portugal written before the year 1000 printed there). (chrismon) dumuium quidem non est set multis mane pleuis acque nodisimu eo quod ego abuit Intenzio sagulfu presbiter cum gontigio presbiter pro eglesia uogauolum scõ martinũ q est fundata In uila uillaredi quia dicia sagulfus presbiter quia era sua ereditate et sacauit gontigio frater suas escribturas et suas firmidades In concilio et diuindigauit Ipsa eglesia de

How scribes wrote Ibero-Romance before written Romance was invented   

   75

sagulfu et de alios suos eredes quia conparara gontigio ipsa eredidate de et suos parentes de cencerigu: et de alios todos suos eredes et fundauit asperigu ipsa eglesia et auidauit Ila p plures et post relinq Ila In manu de suos filios nominib~ gontigio et amarelo presbiter ingenua de alios suos eredes et otinueru Ila In faciem de suos eredes p plures anos sine aligua Inquiedatione oituanta anos sine aligua Inquiedatione et postea coliuit gontigio presbiter ipse sagulfu In sua casa pro li facere seruizio bono et placcitum rouorado In concilio que non abuise de Ilo aligua soposida mala In ipsa eglesia et postea Inrubit Ipse sagulfus presbiter ipso placitum et facia se eredario In ipsa eglesia et fuit gontigio presbiter cum isto placito ad concilio ante aluitum aluitizi gomeze benegas et gudinu beneegas ederonio aluitizi truitesindu nantildizi et aliorum multorum filio bonorum et anouit se sagulfu In ueridate et cedauit se sagulfu a pedes de dona domitria et de gontigo presbiter et feceron se Ipsos Iudices rogadores et rogauerunt multum pro ipsum placitum que rouorara et que eisisse sagulfu de ipsa eglesia et anplius non fecese Inbidem alia soposida et pro suo seruizio dederunt Inde de Ilo decimu inter pane et a beuere lx modios p rogo per ipsas oras que disera In ipsa eglesia et una sagia uizione et sup ipso placito adrouorabit alio ad domna domitria In ipso conzilio et ista anuzione ante ipsos Iudices sagulfus presbiter in ac escribtura manu sua +s In ipse era xxviiiia cod erit xv kalendas setenbres ‘There is no doubt, but it is clear and well-known to many people, that I, priest Sagufe, had a dispute with priest Gontigio concerning the church called St Martin’s, which was founded in the town of Guilhabreu, because priest Sagufe said that it was his own inheritance; then brother Gontigio took his documents and guarantees to the tribunal and claimed that church back from Sagufe and his other heirs, on the grounds that he, Gontigio, had bought that inheritance from his relatives at Cencerigo; indeed, from all his other heirs. And Asperigo founded that church, and lived there for many years, and afterwards left it to his sons Gontigio and priest Amarelo, with no claim on it from his other heirs; and they held it, instead of the other heirs, for many years without any problem – eighty years in all, without any problem – and after that, priest Gontigio brought up that Sagufe in his own house, to treat him well, and had come to an agreement in the tribunal so that there would be no bad faith concerning it; and afterwards that priest Sagufe broke that agreement and made himself the heir in that church; and priest Gontigio went along with this agreement to the tribunal, before Alvito Alvitez, Gomez Venegas, Gudino Venegas, Eteronio Alvitez, Truitesindo Nantildez and many other good and noble people. And Sagufe acknowledged the truth, and knelt at the feet of Doña Domitria and the priest Gontigius. And those judges acted as investigators, asking him at length about that agreement which he had accepted. And they decreed that Sagufe should leave that church and not act any more in bad faith there. And for the work he had done there they gave him, out of the tithe, sixty modios worth of food and drink, and for the offices he had officiated in there, a sash with ribbons on; and in addition to that agreement, he agreed to another one for Doña Domitria in that tribunal; and priest Sagufe has made this admission of his error before those judges in this very document written with his own hand (X). In the year era 1029, 18th August.’

The witnesses are presented in three columns at the end; their signatures are not the originals, and all have been reproduced by the scribe of the surviving clean document.

76   

   Roger Wright

gomeze benegas confirmans aluitus aluitizi confirmans gudinu benegas confirmans conf ederonio aluitizi confirmans truitesindo nantildizi confirmans menendus conf arcerigu manuel presbiter maurgado presbiter mido guandilizi ts rizila ts aluitus gondesindizi ermemiru ts cagidu gundulfizi frogia arualdizi froila ts censoi ts coram testes it sunt amarelo mestalizi ts gundesindu mestalizi ts astruedu ts gontado didazi ts gondesindu cepa fafila uizoi astrulfizi fredenando uizoizi ts

This concerns a lawsuit between a certain Gontigio (who was a priest, or perhaps a monk – the distinction was none too clear in Galicia at the time) and his adopted son Sagufe, over which of them owned the church of San Martín in the town which is now called Guilhabreu; here Sagufe accepts that Gontigio was in the right. There is no space here to comment on all the interesting details of the language, because there are too many (more are explained in Wright 2010); I just wish at the moment to indicate that it is worth investigating the language from the point of view of the scribe, who was probably part of the clerical team of an aristocratic estate. His handwriting suggests that he had a strong and lively personality. We have quite a good idea, from much other evidence, about how people spoke in Galicia at that time, as regards their phonetics, morphology, syntax, semantics and vocabulary, and we can reconstruct something, at least, as regards the sociolinguistic variations of the age. It is instructive to compare these reconstructions, particularly the phonetic ones, with texts such as these which were written by those who spoke that way, and who had been taught to write according to the normal old-fashioned standards. For their teaching does seem to explain a number of the textual peculiarities. I dedicated an article some twenty years ago to a reconstruction of how scribes appear to have been taught to write in tenth-century Galicia (Wright 1991; in English, Wright 1995). One facet of that study is particularly relevant; although many non-standard spellings can be attributed to features of contemporary phonetics, others are found rather surprisingly and commonly in words that had undergone no phonetic change at all in the relevant feature. One example is the word now written in Galicia as sobrinho, meaning ‘nephew’; this derived from Latin SOBRINUM, meaning ‘cousin’, which in tenth-century Galicia was more often written with an initial sup- than with sob- (or than with sop- or sub-). In the

How scribes wrote Ibero-Romance before written Romance was invented   

   77

1991 article it was pointed out that the unrelated word sobre, meaning ‘on’ and already pronounced at that time as [só-bre], developed from Latin SUPER, and the scribes were naturally told that the word should be written with the letters sup-; as it usually was, indeed (and as it is in the above document, in the phrase sup ipso placito). So it looks as if the scribes deduced that the word for nephew should be written with the letters sup in the same way, as if it were related to the word [sóbre]. In fact these two words were not related at all, but it was not a stupid guess to think that they were. It has occurred to me since that it is likely that some of the teachers who taught the scribes were under that misapprehension themselves and had actually told their pupils to write the word that way. For all scribes had to have been taught by somebody, and scribes in monastic and aristocratic houses would usually have been taught in a moderately formal context. In any event, it would be wrong for us to deduce from the written form sup- that anybody ever pronounced the word as [su-prí-no], with [u] and [p], at any time. Another instructive example concerns the Latin word ipse in its various morphological forms; in the above document from Vairão there are ipse, ipsa, ipso, ipsos, ipsum and ipsas, and in all the lexeme turns up seventeen times. In earlier times it had meant ‘the same’, but in Ibero-Romance its semantics had changed to mean ‘that’; it had also come to be pronounced without the consonant [p] (it is now Spanish ese, Galician ise and Portuguese isso). But the scribe had no hesitation in writing the letter p every time. From this we can deduce that he been taught to write this word that way in his training; his teacher had simply said “this is how you write the word [í-se]”. The letter p represented no sound in that written word then, any more than it does now in the written form of the English words psychology and pneumonia, which we are also taught to write with a p in our training. Whenever we find a word in these documents whose written form hardly ever changes (except in the inflections), we can assume that the scribe has been taught to write it that way. And if the document was read aloud later, as most were, the reader would have been trained the same way and would have no problem in reading the word written as ipse aloud as [í-se] in Galicia (or [é-se] in Castilla). The main point to be made in connection with reading aloud is that, in all literate societies, readers are interested in identifying the words and their meanings rather than their sounds (compare Wright 2005); thus the fact that the written form was not a direct transcription of the spoken form was of no more significance to the reader-aloud than the fact that if you read my surname aloud you say [rájt] rather than [wríght]. Indeed it would be completely wrong to read wright as [wríght], and it would have been wrong to read ipse as [ip-se]. In documents of this kind, this technique is most obvious in the cases of abbreviations and numerals. Standard abbreviations had necessarily to be taught as whole units, so we con-

78   

   Roger Wright

sistently get the word ecclesia (meaning ‘church’) written as egle~, sancto (‘holy’) as scõ and the word presbiter (‘priest’) as pr~ (whatever their syntactic function); if they recognised the forms, readers would as usual be led straight by them to their own lexical entry, where the phonological information is stored for every word, and read these words aloud in unabbreviated form, in their usual Romance pronunciation, as (here in Portugal) [i-ɡré-Ʒa], [sã-to] and [prés-te]; reading the words as [e-gle], [scõ] and [pr] could not even have occurred to them. Similarly, the number sixty is written in the document as lx, but could never have been read as [lks]; in the Vairão area it would have been read aloud as [se-sa-ě-ta]. This is obvious, but it is only an exaggerated example of what happens in reading with every word; if the reader recognises the written form, then he or she pronounces the word in the ordinary vernacular way. It is also worth stressing that a reading aloud would include all the lexical items in the text without changing them. The performance was not a question of translating Latin into Romance, as has sometimes been suggested, but simply of reading aloud in the ordinary way, as we would read an English text giving the words of the text their normal contemporary pronunciation. There are interesting semantic features here too. We can be sure, for example, that the written form ipse in these documents corresponded to the spoken [í-se] or [é-se] of the time because it has the semantics of that age; it means ‘that’ and not ‘the same’. In some uses in the Vairão document the old meaning would be all but impossible. Such cases are instructive; when a word in these documents combines the traditional written form with the more recently evolved semantics, that shows without any doubt that the text is in the written mode of the scribe’s speech, rather than any kind of attempt at reproducing the language of a thousand years earlier. An example, which I once studied at length (Wright 2004), is the form sedeat. Sedere was regularly used in the Iberian Peninsula as a suppletive verb for esse, ‘to be’, including providing the forms of the present subjunctive, where it developed phonetically to Ibero-Romance [sé-ja] (and subsequently Modern Spanish [sé-a]). It also developed semantically, from the original meaning ‘be seated’ to simply ‘be’, and was often used as the passive auxiliary. In the Ibero-Romance documents of this period the word is usually spelt in the old way, having been taught as such, but used with the developed meaning and grammar, as an auxiliary with no trace of sedentary semantics (Wright 2004: 284– 290). In the same way, the parentes of our document are ‘relations’ (the meaning which the word had in the tenth century) rather than ‘parents’ (the meaning which it had had a thousand years before). For semantic changes happen all the time, regardless of the word’s spelling. Scribes and writers are not often likely to be aware of these developments, and rarely want to achieve the antiquated semantics of such words; there is not a single word in this document from 991 that

How scribes wrote Ibero-Romance before written Romance was invented   

   79

seems to have an anachronistic meaning that did not exist in tenth-century IberoRomance. The technical legal terms are certainly those of the tenth century: intenzio ‘dispute’, concilio ‘meeting’ or ‘tribunal’, firmidades ‘guarantees’, placitum ‘lawsuit’, soposida ‘bad faith’, rogadores ‘investigators’, decimu ‘tithe’, anuzione ‘admission of error’ (as in Portuguese carta de agnição), etc. It would have been absurd, from every point of view, legal, linguistic and practical, for the scribe to try to latinise the terminology semantically. The vocabulary of the document is that of the age, not of a thousand years before. Almost all the personal names are Germanic, as is the second part of the toponym villaredi (< VILLA FRITHS); so is the verb sacar, which had a precise legal significance in Germanic. There are derived forms, composed of free morphemes and affixes that existed but had not been combined in the Roman Empire, such as diuindigauit, inquiedatione, adrouorabit; the root and the affix are not new but their combination is. The ecclesiastical terms are of Greek origin. That is, the scribe used the vocabulary of the time, even though it was written in unreformed spelling. Inflectional morphology is of interest, too. Except in the pronouns, there was no nominal morphology in Ibero-Romance other than markers of number and of gender. But the main traditional inflections were taught to scribes in their training, so they knew what they were; yet, in most of the texts of this time, it becomes clear that the scribe had only a hazy idea of when to use which ending. That is, they knew their morphology but not their syntax. This may be connected with the fact that the main grammar book, if they had one at all, was the fourthcentury Ars Minor of Aelius Donatus, which concentrated on morphology and had nothing to say about syntax. So some of the morphological choices can look chaotic, as if the scribe had chosen the endings more or less at random – in this document, for example, (…) de suos filios nominib~ gontigio et amarelo pr~ (…). But he does know what the actual endings were; he does not, for example, invent endings that did not exist (such as an *-erum or *-irum on the analogy of -orum y -arum). The verbal morphology, however, looks far more like the older Latin; this is because Ibero-Romance had not changed the verbal inflectional system in the same wholesale way as the nominal. Thus the spoken ending [-ów] which had developed in Galicia from the original -AVIT was represented by the letters -auit, which had been taught to the scribe as a whole unit, in the same way as the form ipse had been taught him for [í-se]. The original final [-t] of these verb forms was no longer pronounced there, but in the third person past verb forms the letter -t is written here every time: abuit, sacauit, diuindigauit, fundauit, auidauit, coliuit (in error for coluit), fuit, anouit, cedauit (although the original form had been the irregular cessit), and adrouorabit (with the letter -b- more usual in the future,

80   

   Roger Wright

although here the word has past meaning); there is also the form inrubit (‘broke’) with -bit; while the eleven other cases of the third-person singular (that is, those that did not end in -uit) all lack the normative but silent final letter -t: mane, dicia, era, conparara, relinqui, abuise, facia, rouorara, eisisse, fecese, disera. Among these there are three third-person singular forms of the past subjunctive, in which the inflection then pronounced in Galicia as [-í-se] is written in three different ways, all wrong: abuise with -ise, eisisse [ej-ςé-se] with -isse, and fecese with -ese. We can deduce from all this evidence that the scribe had been taught to use the written form -uit for the past indicative, but had received no such instruction concerning the past subjunctive (or if he had, he had not taken it in: the correct form of the inflection would have been -isset). As regards the syntax of this document, we can see that the preferred word order is VSO, Verb-Subject-Object, particularly, but not only, in subordinate clauses: e.g. … quia dicia Sagulfus presbiter quia era sua eredidate (‘[VSO] which said Sagufe that [VS] was his inheritance’); et sacauit gontigio frater suas escrituras (‘and [VSO] drew out Gontigio his documents’); … quia conparara Gontigio ipsa eredidate (‘that [VSO] had bought Gontigio that inheritance’); … et postea coliuit gontigio presbiter ipse sagulfu (‘and then [VSO] raised Gontigio that Sagufe’) … et postea inrubit ipse sagulfus presbiter ipso placitum (‘and [VSO] then broke that Sagufe that agreement’), etc. Verb-Subject-Object was the unmarked order of speech at that time, but it had not been the unmarked order of the language of a thousand years earlier. Word order is not mentioned in the grammars: nobody has said anything to the scribe about that in his training, so he naturally writes in his normal word order (Blake 1991/96). So we see that the morphology, the semantics, the syntax and the vocabulary of this document are those of the speech of tenth-century Galicia. That speech is usually referred to by modern specialists as being Ibero-Romance, so it would seem to make sense to say that this document was written in Ibero-Romance. The only aspect of the text which is not Ibero-Romance is the spelling, the written form of the words, because the scribe was not trying to reproduce the phonetics of his speech in a direct transcription but to achieve the traditionally correct forms that he had been taught. That is merely normal. When we write a modern English text now, we represent on the paper the modern English morphology, semantics, syntax and vocabulary, using an old spelling system which is far from being a transcription. But that mismatch does not stop our text from being in modern English. That just is the way we write our own language. In the same way, even though the scribe would have referred to his own language as being Latin (pronounced [la-dí-no]), I see no good reason for modern analysts not to call the language of a text like this “Ibero-Romance”. It is true that a new systematic spelling system was not going be invented for the language for another couple of cen-

How scribes wrote Ibero-Romance before written Romance was invented   

   81

turies, but that seems a comparatively irrelevant consideration. Those semantic, lexical, morphological and syntactic features of this kind of text that look like Ibero-Romance need no further explanation for the way they appear; such as, for example, the meaning of the word written as ipse, which has the meaning that it has in these texts because the scribe had such semantics in his lexical entry for, and in his use of, the word. Those aspects of these documents which do not look like reconstructable Ibero-Romance can usually be explained through the scribe’s training; for example, ipse has a letter p here because he was told to write it that way. Unsurprisingly, the same general truths applied in 991 CE as they do today. Latinists, however, do not see it this way. They tend to see the language of such a text as being Latin, arguing solely from the orthography. Yes, they call it “bad Latin”, “corrupt Latin”, “decadent Latin”, even “barbarous Latin”, but always “Latin”, even though from this perspective these texts have been written with astounding ignorance, incompetence and stupidity. But historians would prefer not to think that the royal, ecclesiastical and aristocratic houses regularly employed scribes who were incompetent, stupid, ignorant, barbarous, corrupt, decadent and bad. Emiliano (e.g. 2002, 2003, 2005, 2008) has demonstrated without any doubt that the scribes of this time and place were not barbarous, but working intelligently in a difficult context where the written language as taught was ever diverging from the one they spoke. And this is a theme worth stressing: most of the scribes of the tenth-century Asturian-Leonese kingdom were intelligent. They were not capricious barbarians making daft errors at random. If we do our best to work out from their texts how they were taught and what they were trying to do, we can use their evidence as crucial aids to understanding the language of the time (this is the essence of sociophilology). And thus, as long as we remember that they were trying to achieve the spellings standardised over a millennium earlier, we can take the semantics, the vocabulary, the syntax and even to some extent the morphology of documents of this type at almost face value.¹

1  Wright (2010) is a related study in Spanish of these and several further aspects of the same document; we are grateful to the volume’s editor, Ana María Cano, for permission to develop some of those arguments here. The present chapter is itself expanded, including a suggested phonetic transcription of a reading of the document, in Wright (2012).

82   

   Roger Wright

References Blake, Robert. 1991/96. Syntactic aspects of Latinate texts of the early Middle Ages. In Roger Wright (ed.), Latin and the Romance Languages in the Early Middle Ages, 219–232. London, Routledge. Reprint Pennsylvania: Penn State Press. Davies, Wendy. 2007. Acts of Giving: individual, community and church in Tenth-Century Christian Spain. Oxford: Oxford University Press. Emiliano, António. 2000. O mais antigo documento latino-português (882 a.D.) – edição e estudo grafémico. Verba 26. 7–42. Emiliano, António. 2003. O estudo dos documentos notariais latino-portugueses e a história da lingua portuguesa. Signo 11. 77–126. Emiliano, António. 2005. Representational models vs. operational models of literacy in LatinRomance legal documents (with special reference to Latin-Portuguese texts). In Roger Wright and Peter Ricketts (eds), Studies on Ibero-Romance Linguistics dedicated to Ralph Penny, 17–58. Newark [Delaware]: Juan de la Cuesta. Emiliano, António. 2008. O conceito de latim bárbaro na tradição filológica portuguesa: algumas observações gerais sobre pressupostos e factos (scripto-)linguísticos. In Javier Elvira et al. (eds.), Lenguas, reinos y dialectos en la Edad Media ibérica, 191–231. Madrid: Iberoamericana /Frankfurt: Vervuert. Finbow, Tom. 2008. Limiting logographic Latin: (non-)separation of orthographic words in Medieval Iberian writing. In Roger Wright (ed.), Latin tardif ~ latin vulgaire VIII: actes du VIIIe colloque international sur le latin vulgaire et tardif, 521–531. Hildesheim: Olms. Herculano de Carvalho e Araujo, Alexandre & Joaquim Jose da Silva Mendes Leal. 1867. Portugaliae Monumenta Historica, Diplomata et Chartae. Lisbon: Academia Real das Ciências. Herman, József. 2000. Vulgar Latin. Pennsylvania: Penn State Press. Menéndez Pidal, Ramón. 1972 (7th edition). Orígenes del español. Madrid: Espasa-Calpe. Monteagudo, Henrique. 2008. Letras primeiras: o Foral de Burgo de Caldelas, os primordios da lírica trobadoresca e a emerxencia do galego escrito. A Coruña: Barrié de la Maza. Pensado, Carmen. 1991/96. How was Leonese Vulgar Latin read? In Roger Wright (ed.), Latin and the Romance Languages in the Early Middle Ages, 190–204. London, Routledge. Reprint Pennsylvania: Penn State Press. Pérez González, Maurilio. 2008. El latín medieval diplomático. Archivum Latinitatis Medii Aevi [Bulletin du Cange] 66. 47–101. Stengaard, Birte. 1991/96. The combination of glosses in the Códice Emilianense 60 (Glosas Emilianenses). In Roger Wright (ed.), Latin and the Romance Languages in the Early Middle Ages, 177–189. London, Routledge. Reprint Pennsylvania: Penn State Press. Trask, R. Larry. 2000. The Dictionary of Historical and Comparative Linguistics. Edinburgh: University Press. Van Acker, Marieke, Rika Van Deyck & Marc Van Uytfanghe (eds). 2008. Latin écrit – roman oral? De la dichotomisation à la continuité. Turnhout: Brepols. Wright, Roger. 1982. Late Latin and Early Romance, in Spain and Carolingian France. Liverpool: Francis Cairns. Wright, Roger. 1991. La enseñanza de la ortografía en la Galicia de hace mil años. Verba 18. 5–25.

How scribes wrote Ibero-Romance before written Romance was invented   

   83

Wright, Roger. 1995. The teaching of orthography in tenth-century Galicia. In Roger Wright (ed.), Early Ibero-Romance, 181–208. Newark [Delaware]: Juan de la Cuesta. Wright, Roger. 2003. A Sociophilological Study of Late Latin. Turnhout: Brepols. Wright, Roger. 2004. La representación escrita del romance en el Reino de León entre 1157 y 1230. In José María Fernández Catón (ed.), Orígenes de las lenguas romances en el reino de León, siglos IX–XII, vol. I, 273–293. León: Archivo Histórico Diocesano. Wright, Roger. 2005. El léxico y la lectura oral. Revista de Filología Española 85. 133–149. Wright, Roger. 2010a. Gontigius, Sagulfus, Domitria y el hijo de muchos otros buenos. In Ana María Cano González (ed.), Homenaxe al Profesor Xosé Lluis García Arias, 407–420. Oviedo: Academia de la Llingua Asturiana. Wright, Roger. 2012. Writing and Speaking. In Scrivere e leggere nell’alto Medioevo (Spoleto, CISAM, Settimane di Studio LIX, 2012), 273–292. Spoleto: Fondazione Centro Italiano di Studi sull’alto Medioevo.

Anja Busse

5 Hittite scribal habits: Sumerograms and phonetic complements in Hittite cuneiform Sources: This paper deals with some general orthographic conventions concerning the use of Sumerograms in Hittite cuneiform. There is a great variety of Hittite manuscripts of different text types (e.g. rituals, annals, omen texts, letters, instructions and laws). More than 33,000 clay tablets with Hittite cuneiform dating from the 2nd millennium BCE have been excavated in Anatolia and northern Syria since the early 20th century. Most of them are now kept in Turkish (e.g. Ankara, Istanbul, Çorum) or international museums (e.g. British Museum, Louvre, Pergamon Museum). The greatest number of texts comes from archives in the Hittite capital Hattusa (Boğazköy). We still do not know much about the scribes. They were professionals and some are known to have been state officials or members of the royal family, as well. Only rarely female scribes can be identified. For more information see van den Hout (2009b). Abstract: When the Hittites adopted the cuneiform writing system from a northern Syrian variant in the middle of the 2nd millennium BCE, it had already gone through more than 1000 years of development. As a result, Hittite cuneiform is a complex writing system with graphic components of three different non-related languages (Sumerian, Akkadian and Hittite). Hittite scribal circles¹ were less focused on the Sumerian language component of cuneiform script than scribes in the southern Mesopotamian tradition where the cuneiform writing system had been invented (Schretter 2004: 464–466). Instead, the scholarly language of these scribes was based on Hittite and Akkadian (Weeden 2011: 354). Sumerian may only have been understood by a proportion of the Hittite scribes. However, it remained a part of the cuneiform writing system in logograms and determinatives. In this paper, we will focus on the use of Sumerian logograms and Hittite phonetic complements as orthographic patterns in different chronological stages. At the same time, we will point out some domains that, if comprehensively studied, offer more information about how Hittite scribes used this complex writing system to render their own language.

1 See Gordin (2010) for more detailed information on Hittite scribal circles and Scriptoria.

86   

   Anja Busse

1 Patterns on the graphic level Hittite cuneiform is a logo-syllabic writing system. Cuneiform signs can render syllables, words and determinatives, and all of these features may appear in one word. This paper will focus on the use of Sumerian logograms and Hittite phonetic complements as orthographic patterns in different chronological stages. Two patterns can be observed: in the first, the scribe had a choice between rendering a Hittite word with syllabic signs or logographically² as a Sumerogram³, whereas in the second, if he chose a Sumerogram it could be written with or without a Hittite phonetic complement.

Pattern 1:

Pattern 2:

variant I syllabic

variant II logographic (Sumerogram)

variant II a with phonetic complement⁴

variant II b without phonetic complement

The Hittite language can be subdivided into three main chronological stages⁵: Old Hittite (OH, ca. 1650–1450 BCE), Middle Hittite (MH, ca. 1450–1350 BCE) and New Hittite (NH, ca. 1350–1190 BCE). This classification refers to the language used in the texts. More important for research on scribal habits is the palaeographic classification with its three main stages that refer to the script in order to date the

2 Marquardt (2011: 118–123) sees logographic writing in Hittite cuneiform as “Ausweichschreibung” when the scribe had different possibilities of syllabic rendering for one logogram and a way of avoiding the wrong spelling of a Hittite word. However, he does not specifically analyse the distribution of logographic vs. syllabic writing for different text types or genres. 3 Hittite scribes also used “Akkadograms” for lexemes but these were not actually logograms. It is more likely that they were recognised and read by the scribes as Akkadian words. For more information see Weeden (2011: 10). An exception is the use of Akkadographic prepositions to render syntactic relations that would have been expressed with case morphemes in Hittite. 4 We only refer to Hittite phonetic complements here. There are some Akkadian complements on Sumerograms in Hittite cuneiform as well but these might be only graphic since they often show the wrong case ending (like e.g. dingir-lim, Akkadian g.sg. ilim, which stands in Hittite sometimes for the nominative of siunas ‘god’). 5 Classification according to Hoffner and Melchert (2008: xvii).

Hittite scribal habits: Sumerograms and phonetic complements in Hittite cuneiform   

   87

individual copies of a text⁶: Old Script⁷ (OS refers to the oldest Hittite texts until the reign of Telipinu, ca. 1650–1500 BCE), Middle Script (MS refers to texts from the reign of Telipinu to Suppiluliuma I., ca. 1500–1350 BCE) and New Script (NS refers to texts from from the end of Suppiluliuma I. to the end of the Hittite Empire, ca. 1350–1180 BCE).⁸ In Hittite Studies syllabic writing is often regarded as a dating criterion for older texts.⁹ Likewise, the omission of phonetic complements with Sumerograms is said to be an indication for very late texts. So far, these orthographic phenomena have not yet been sufficiently analysed and cannot be used as general criteria for dating. Even palaeographic criteria like ductus and sign forms have to be used carefully since scribes could have been influenced by the texts they copied or certain scribal habits of their own tradition.¹⁰

2 Variants of pattern 1: logographic or syllabic writing Texts classified as Old Script do not show many innovations. We find a tendency to prefer syllabic writing. However, Hittite cuneiform has been used as a complex system with logograms, determinatives and phonetic complements from the earliest texts on. In younger texts the general amount of Sumerograms grows but we have to bear in mind that the corpus of Old Script is considerably smaller than the New Script corpus. Between the two stages there is a transitional period (Middle Script) in which we can already trace some changes. At the graphic level, the chronological differences are not the same for every word. While some words have been written syllabically and/or logographically from the earliest attestations onwards, others do not appear with Sumerograms until Middle or even New Script. Moreover, we can find Sumerograms with a stable pattern once it has been established and Sumerograms with relatively

6 For the dating we refer to the Würzburger online Konkordanz [http://www.hethport.uniwuerzburg.de/hetkonk/] (30th of September, 2011). 7 Palaeographic criteria to determine OS have been established on sign forms, width of column dividers, point where the writing begins on the tablet, spacing of signs and words (Archi 2003: 6). 8 See van den Hout (2009a: 22) for the dates. A comprehensive discussion of the dating problem is given in van den Hout’s article and in Archi (2010). There are additional stages such as, for example, Late New Script (LNS). They will have to be taken into account for a more detailed analysis. 9 See e.g. Neu (1970: 53) and Hoffner (1997: 239–245). 10 For a conclusion on the dating problem see Archi (2010) and van den Hout (2009a).

88   

   Anja Busse

unstable patterns, which would change according to the grammatical function and the text types. Thus, we could start our analysis of Sumerographic writing in Hittite cuneiform with words that have been attested from the earliest Hittite records on to determine their spelling in different stages and genres.

2.1 Constant use of logographic writing We can observe a rather conservative pattern concerning some Sumerograms that were adopted from Syro-Akkadian cuneiform and infrequently appear syllabically in Hittite. Some of these words are not yet identified in the Hittite language and keep appearing only as Sumerograms until the end of their attestation. A few examples of words¹¹ that appear as Sumerograms¹² from OS on:

word

Sumerogram

Hittite syllabic example

water house god son gold mountain king

a é dingir dumu kù.gi Ḫur.sag lugal

wa-a-tar (nom./acc.sg.) pár-na-aš (gen.sg.) ši-ú-uš (nom.sg.) – – – ḫa-aš-šu-u-ú-i (dat.-loc.sg.)

Table 1: Sumerograms that appear from OS onwards.

2.2 Innovative use of logographic writing a) Verbs We can observe that text types and genre seem to play important roles regarding the use of Sumerograms in both Akkadian and Hittite cuneiform, maybe even more than the chronological stage of a text.¹³ This has already been suggested

11 Examples: Weeden (2011) and Kloekhorst (2008). 12 Some of these words never appear syllabically. 13 Another interesting point that excludes use of logograms for shorthand is the comparison of two versions of “Muwatallis Prayer” (NS) by Singer (1996: 160) [cited in Weeden 2011: 52–53] that shows syllabic writing in the “hastily written probably dictated” version and more logograms in the “neat version”.

Hittite scribal habits: Sumerograms and phonetic complements in Hittite cuneiform   

   89

by von Soden for Akkadian: “Die Verwendung der Wortzeichen ist nicht nur je nach der Zeit, in der ein Text geschrieben wurde, sehr verschieden, sondern hat vor allem auch in den einzelnen Schrifttumsgattungen einen sehr ungleichen Umfang” [‘The use and frequency of logograms is different not only in texts from different chronological stages but most notably in different types of texts.’] (von Soden 1967: 7). He points out that there are almost no logograms in hymns from the Old Babylonian period (ca. 1950–1530 BCE) but in the first millennium BCE omens and rituals were written predominantly with logograms and these were frequently complemented. Concerning Akkadian verbs E. Robson observes that “… logograms and phonetic complements for verbs are not a common feature of Akkadian in cuneiform until the first millennium BCE. Even then, their use was mostly restricted to royal inscriptions, literary texts, and other scholarly works”.¹⁴ However, verbs like šâmum ‘to buy’ and nadānum ‘to give’ appear as Sumerograms (in.ši.in.šam2 and sum) in contracts from Terqa/Hana in the second millennium BCE.¹⁵ “In the Old Babylonian version (forerunners of the omen series) of šumma izbu, only twenty logograms are used, all substantives. There are no verbs written logographically in this text. In the limited (Akkadian) source material from Boğazköy, we find logograms for twenty different substantives and one verb” (Leichty 1970: 30). In contrast, Sumerograms for verbs are attested in the Middle Babylonian (ca. 1530– 1000 BCE) omen texts. In Hittite cuneiform, verbs are almost exclusively written syllabically in Old Script. Exceptions are found in the Old Hittite ritual¹⁶ and omen texts where some Sumerographic verbs can already be found. An example would be tuš-aš for the participle asandas ‘sitting’.¹⁷ According to Friedrich and Kammenhuber (1975–1984: 194) the earliest logographic attestation for the verb ar- ‘to stand’ is gub-aš for the participle arandas ‘standing’. It is interesting that Gelb (1961: 20–22) points out that in Old Akkadian (ca. 2500–1950 BCE) only logograms for participles appear and only rarely logograms for other forms of the verb.¹⁸

14 Eleanor Robson: “Categories of signs”. Knowledge and Power. [http://knp.prs.heacademy. ac.uk/cuneiformrevealed/aboutcuneiform/categoriesofsigns/] (retrieved 29th September 2011). 15 See Podany (2002: 161–162). Podany (2002: 17) points out that “Sumerograms were increasingly used to replace syllabic Akkadian spellings in many different types of words, including personal names.” 16 Several examples in StBoT 26: 275. 17 In the Old Hittite “Gewitterritual” we can find sír-ru for ishamianzi (‘they sing’). This pattern (Sumerogram with Akkadian phonetic complement) points to an underlying Akkadian word. 18 But Gelb suggests that the structure of these logographic writings is very uncommon at this stage.

90   

   Anja Busse

One could thus hypothesise that in both scribal traditions logograms for verbs were at first introduced for nominal forms. The following table shows some forms¹⁹ of verbs that are attested from OS on and also appear as Sumerograms with phonetic complements later:

verb

OS

NS

sipand- (to sacrifice) war- (to burn) iya- (to do/make)²⁰ mēmai- (to speak) walh- (to hit) wemia- (to find) epp-/app- (to take/seize)²¹ aniya- (to work)

ši-pa-an-ti (3.sg.pres.act.) ú-ra-a-ni (3.sg.pres.med.) i-ya-mi (1.sg.pres.act.) me-e-ma-aḫ-ḫi (1.sg.pres.act.) wa-la-aḫ-zi (3.sg.pres.act.) ú-e-mi-ez-zi (3.sg.pres.act.) e-ep-mi (1.sg.pres.act.) a-ni-e-mi (1.sg.pres.act.)²²

bal-an-ti bil-ni dù-mi du11-aḫ-ḫi gul-aḫ-zi kar-zi dib-mi kin-mi

Table 2: Verbs in different graphic variants.

As we have seen, there are similarities in the development of Sumerographic verbs in Akkadian and Hittite. The use of the logographic pattern for this word class seems to be restricted in the early stages of attestation. One possible influence for a change of this and other scribal habits in cuneiform, might have come along with the Kassites (Weeden 2011: 365), a people with an isolated language who gained power over Babylonia in the second half of the second millennium and learned cuneiform writing from their own perspective. In Hittite cuneiform the habit of writing verbs with Sumerograms is enhanced in NS and could probably have spread from one genre to others through scribes involved in the process of copying long-term records from different genres.²³ These texts were kept and recopied for a long time, i.e. they were important and perhaps also prestigious on a graphic level. Thus, texts with special orthographic

19 Syllabic examples: Kloekhorst (2008), Sumerographic examples: Weeden (2011) and Kloekhorst (2008). 20 According to Marquardt (2011: 35), we find predominant logographic attestation in NS. 21 According to Marquardt (2011: 32), there is no Sumerographic attestation until 13th century. 22 Examples from Friedrich and Kammenhuber (1975–1984: 86–87). 23 Van den Hout (2011: 4) distinguishes between long term records with duplicates (historiography, treaties, edicts, instructions, loyalty oaths, laws, oracle theory, hymns and prayers, festival and ritual scenarios, mythology, foreign language compositions, hippological texts, lexical lists) and short-term records with just one copy (letters, land deeds, administrative texts, court depositions, oracle practice, vows).

Hittite scribal habits: Sumerograms and phonetic complements in Hittite cuneiform   

   91

styles like ritual or omen (an imported genre from Mesopotamia) texts could have been the source for the spread of certain scribal habits.

b) Local adverbs A similar orthographic change is also attested with another word class, local adverbs. They appear more often Sumerographically in later texts – a pattern that was rare in Old Script. Some of the local adverbs (like egir-pa) are already attested in earlier stages but not common for OS, see Neu (1970: 53). Examples of local adverbs²⁴ that are attested from OS onwards:

local adverb

OS

NS

āppa (back) āppan (behind) katta (downwards) kattan (under) sarā (up) sēr (above) istarna (in the middle) mēnahhanda (in front of, before)

a-ap-pa a-ap-pa-an kat-ta kat-ta-an ša-ra-a še-e-er iš-tar-na me-e-na-aḫ-ḫa-an-da

egir-pa egir-an gam gam-an ugu šà igi-anda

Table 3: Local adverbs in different graphic variants.

3 Phonetic complements Another orthographic pattern that is connected with logographic writing is the use or omission of phonetic complements.

3.1 Function of phonetic complements Phonetic complements provide additional information for the reading of a sign in logo-syllabic writing systems. They disambiguate the reading of logograms by repeating the first or, as in cuneiform, the last syllables of a word that are

24 Syllabic examples from Kloekhorst (2008), Sumerographic examples from Christiansen (2006), Weeden (2011) and Kloekhorst (2008).

92   

   Anja Busse

actually part of the logogram. In contemporary alphabetic writing systems there are far fewer logograms and accordingly fewer phonetic complements, but one can still find a few examples, such as the numbers with their varying readings: 20  (twenty)  – 20ies (twenties) or 2 (=two) – 2nd (=second). Or, in abbreviated words, which are popular now in mobile or online communication, but also found in normal written communication, e.g. Xmas (Christmas) or Xing (Crossing).

3.2 Phonetic complements in cuneiform The Akkadian scribal habit of writing Sumerograms with phonetic complements²⁵ was adopted by the Hittites, as well. Akkadian phonetic complements were used to disambiguate the reading of the Sumerograms in Akkadian. They are “… rare in Old Babylonian literary texts, but become more frequent in the first millennium, when knowledge of Sumerian had died out” (Weeden, 2011: 6).²⁶ A complete analysis of all phonetic complements in Hittite cuneiform is yet to be done, but there are certain groups of logograms that show peculiarities in the use of phonetic complements. Some Sumerograms seem to have more stable patterns in their use while others switch between use and omission of phonetic complements. Word class seems to play an important role for the stability of this pattern, but again different text types show different characteristics. Furthermore, the forms have to be analysed in the context of frozen writings in stereotyped contexts.²⁷

3.2.1 Stable variants of pattern 2 a) Complementing Sumerograms of Hittite local adverbs As we have seen, the use of logograms for local adverbs is more common in younger texts. Yet, if Hittite local adverbs appear as logograms, some of them are

25 This scribal habit goes back to a development that had already started in the Early Dynastic period in Sumerian cuneiform as intially syllable signs were used to render grammatical morphemes in Sumerian. 26 M. Worthington (personal communication in 2011) points out that in first-millennium Akkadian there is a tendency towards graphic fossilisation. Logograms keep their phonetic complements regardless of their grammatical function. This is especially noticeable with -tim as a phonetic complement to Sumerograms representing feminine nouns. 27 Frozen forms of logograms with phonetic complements appear for example in the Akkadian omen series šumma izbu.

Hittite scribal habits: Sumerograms and phonetic complements in Hittite cuneiform   

   93

almost always complemented while others are always uncomplemented. Hittite local adverbs appear in diverse semantic functions (with a directival or locatival sense) that start to change and get mixed up in the Middle Hittite language period. The effects of this semantic change on the graphic level and the use of logograms and phonetic complements have not yet been explored.

with complement

without complement

gam-an (kattan) egir-pa (āppa ) egir-an²⁸ (āppan) egir-an-na (appanda) igi-anda (mēnahhanda)

gam (katta) ugu (sarā/sēr) šà (istarna)

Table 4: Local adverbs with and without phonetic complements.

b) Complementing Sumerograms of Hittite verbs As described in 2.2.(a), frequent use of Sumerographic verbs begins in New Hittite Script. They appear always with phonetic complements. They also disambiguate Sumerograms for verbs like dù, which can not only represent Hittite iya- (to do) but also Hittite kis- (to become). In Hittite cuneiform these complements tend to be more complex than those used for example with nouns. However, in the Akkadian version of the omen series šumma izbu Sumerographic verbs usually only need one syllabic sign as phonetic complement (see Leichty 1970: 27–30). This could be due to the different structure of the Semitic languages but further investigation is needed.

3.2.2 Unstable variants of pattern 2: complementing Sumerograms of nouns The use of phonetic complements with nouns can be described as less stable than with verbs or local adverbs. Most Sumerograms for nouns appear with or without phonetic complements according to the need for disambiguating grammatical relations. There are also graphic phenomena that exhibit interaction with the syntactic level. Thus, we can find a tendency to write a phonetic complement when the scribe followed the Hittite word-order and no complement when he followed the Sumerian word-order. This happens for example with

28 There are a few cases of egir without complement: KUB 19.55 rev. 4.

94   

   Anja Busse

genitive constructions:²⁹ Sumerian word-order and logographic rendering: kur lugal (‘land of the king’) vs. Hittite word-order and use of phonetic complement: lugal-aš ìr-iš (‘servant of the king’). Another subject for analysis could be case endings. For example from Middle Hittite onwards, formal case syncretism is attested. This contradicts the general Hittite tendency to avoid ambiguity, but, perhaps, there is a disambiguation on the graphic level. Certain groups of words are hardly ever complemented. These include terms for metals and professions.³⁰ Both groups appear already early within Hittite texts and almost always as Sumerograms.³¹ In the case of metals one reason for this could be that they are mass nouns and often appear with units of measurement. When they are complemented they usually stand in the genitive to indicate the material of objects. Moreover, words that frequently appear in lists could be less often written with syllabic signs or phonetic complements. Consequently, for a comprehensive analysis of Sumerographic writings not only syntactic constructions but also differences between the types and genres of texts have to be considered.

4 Conclusion and perspective As we have seen, first, there have been only few changes in the orthography of Sumerographic writings in the Hittite scribal tradition. Hence, the scribes used the patterns they had adapted and at the same time preferred syllabic writing. Some of these old patterns, for example the rendering of certain Sumerographic nouns, remained a part of the Hittite scribal tradition until the end of the written records. In New Script, the Sumerographic rendering of verbs that appeared only in a few texts and only with participles at an earlier stage is enhanced. Similar processes are involved in the orthography of local adverbs. However, we still do not know much about the presumably small scribal circles in the earlier stages of Hittite cuneiform and the changes that occurred in the Middle and New Hittite period, but foreign influence cannot be completely excluded. Beckman suggested the presence of scribes from Babylonia in Hattusa already in the 15th century BCE. In the time of the Hittite Empire, there was a greater need for scribes to carry out all the administrative tasks. The king of this period may also have aspired to meet

29 Compare Weeden (2011: 33) for a similar phenomenon but on another syntactic level. 30 See Weeden (2011: 36 and 366–368). 31 See e.g. Neu (1970: 530).

Hittite scribal habits: Sumerograms and phonetic complements in Hittite cuneiform   

   95

international standards in his correspondence with foreign rulers and officials. This could have facilitated the use of scribal habits with international prestige. The way Hittite scribes rendered their language with graphic elements of Sumerian origin depends on the individual education of the scribe as well as on text types, word class and syntactic constructions. Some scribal habits might have spread from texts with special orthographic patterns to other text types, and others might have resulted from a greater interest in foreign scribal habits and reflections on cultural, linguistic and graphic aspects of the cuneiform writing system.

5 References Archi, Alfonso. 2003. Middle Hittite – ‘Middle Kingdom̕. In Gary Beckman, Richard Beal and Gregory McMahon (eds.), Hittite Studies in Honor of Harry A. Hoffner Jr. on the Occasion of his 65th Birthday, 1–12. Winona Lake: Eisenbrauns. Archi, Alfonso. 2010. When Did the Hittites Begin to Write in Hittite? In Yoram Cohen, Amir Gilan and Jared L. Miller (eds.), Pax Hethitica – Studies on the Hittites and their Neighbours in Honour of Itamar Singer, 37–46. Wiesbaden: Harrassowitz. Beckman, Gary. 1983. Mesopotamians and Mesopotamian Learning at Hattuša. JCS 35. 97–114. Bryce, Trevor. 2002. Life and Society in the Hittite World. Oxford: Oxford University Press. Charpin, Dominique. 2010. Reading and writing in Babylon. Cambridge (MA)/London: Harvard University Press. Christiansen, Birgit. 2006. Die Ritualtradition der Ambazzi. Wiesbaden: Harrassowitz. Friedrich Johannes & Annelies Kammenhuber. 1975–1984. Hethitisches Wörterbuch. Zweite, völlig neu bearbeitete Auflage auf der Grundlage der edierten hethitischen Texte. Band I: A. Heidelberg: Winter. Gelb, Ignace J. 1961. Old Akkadian Writing and Grammar. Chicago: University of Chicago Press. Gordin, Shai. 2010. Scriptoria in Late Empire Period Hattusa: The Case of the é giš.kin.ti. In Yoram Cohen, Amir Gilan and Jared L. Miller (eds.), Pax Hethitica – Studies on the Hittites and their Neighbours in Honour of Itamar Singer, 158–177. Wiesbaden: Harrassowitz. Güterbock, Hans G., Harry A. Hoffner Jr & Theo P. van den Hout. 1989–. The Hittite Dictionary of the Oriental Institute of the University of Chicago (CHD). Chicago: Oriental Institute. Hoffner, Harry A. Jr. 1997. The laws of the Hittites. Leiden/New York/Köln: Brill. Hoffner, Harry A. Jr & H. Craig Melchert. 2008. A Grammar of the Hittite Language. Winona Lake: Eisenbrauns. Kloekhorst, Alwin. 2008. Etymological Dictionary of the Hittite Inherited Lexicon. Leiden/ Boston: Brill. Košak, Silvin et al. Konkordanz der hethitischen Keilschrifttafeln. Online-Databasis [http://www.hethport.uni-wuerzburg.de/hetkonk/]. Leichty, Erle. 1970. The omen series Šumma izbu. Locust Valley, N.Y: J.J. Augustin. Marquardt, Henning. 2011. Hethitische Logogramme – Funktion und Verwendung. Wiesbaden: Harrassowitz. Neu, Erich. 1970. Ein althethitisches Gewitterritual. Wiesbaden: Harrassowitz.

96   

   Anja Busse

Neu, Erich. 1983. Glossar zu den althethitischen Ritualtexten. Wiesbaden: Harrassowitz. Podany, Amanda H. 2002. The land of Hana. Kings, chronology, and scribal tradition. Bethesda: Capital Decisions Ltd. Riemschneider, Kaspar K. 2004. Die akkadischen und hethitischen Omentexte aus Boghazköy. Dresden: Technische Universität. Robson, Eleanor. 2010. Categories of signs. Online contribution to: Knowledge and Power, Higher Education Academy. [http://knp.prs.heacademy.ac.uk/cuneiformrevealed/ aboutcuneiform/categoriesofsigns/]. Rüster, Christel & Erich Neu. 1989. Hethitisches Zeichenlexikon. Inventar und Interpretation der Keilschriftzeichen aus den Boğazköy-Texten. Wiesbaden: Harrassowitz. Rüster, Christel & Erich Neu. 1991. Deutsch-sumerographisches Wörterverzeichnis, Materialien zum hethitischen Lexikon. Wiesbaden: Harrassowitz. Schretter, Manfred. 2004. Zur Rolle der altmesopotamischen Schule bei der Gestaltung und überregionalen Verbreitung kultureller Orientierungsmuster. In Robert Rollinger & Christoph Ulf (eds.), Commerce and Monetary Systems in the Ancient World. Means of Transmission and Cultural Interaction. Melammu Symposia 5. Oriens et Occidens 6, 461–469. Stuttgart: Franz Steiner Verlag. Singer, Itamar. 1996. Muwatalli’s Prayer to the Assembly of Gods through the Storm-God of Lightning. Atlanta: American Schools of Oriental Research. van den Hout, Theo. 2009a. A Century of Hittite Text Dating and the Origins of the Hittite Cuneiform Script. Incontri Linguistici 32. 11–35. van den Hout, Theo. 2009b. Schreiber D. Bei den Hethitern. Reallexikon der Assyriologie und der vorderasiatischen Archäologie 12. 273–280. van den Hout, Theo. 2010. The rise and fall of cuneiform script in Hittite Anatolia. In Christopher Woods (ed.), Visible language – inventions of writing in the Middle East and beyond, 99–106. Chicago: Oriental Institute of the University of Chicago. van den Hout, Theo. 2011. The elements of Hittite. Cambridge: Cambridge University Press. Weeden, Mark. 2011. Hittite Logograms and Hittite scholarship. Wiesbaden: Harrassowitz.

Part III: Standardisation versus regionalisation and de-standardisation

Terttu Nevalainen

6 Words of kings and counsellors: register variation and language change in early English courtly correspondence Sources: The 16th-century materials discussed in this article come from the electronic Corpus of Early English Correspondence (CEEC) and are edited from holograph and secretarial originals. The editions are listed in Nevalainen and Raumolin-Brunberg (2003: 223–234). King Henry VIII’s holograph letters to Anne Boleyn are housed in the Vatican Library in Rome; his secretarial letters and their copies are found in various archives and libraries in Britain, including the British Library, as are the letters of the King’s chief ministers. Other letters, such as those by Thomas Cromwell, and the merchant correspondence discussed, are held in The National Archives in Kew. Abstract: Personal letters are a valuable source of authentic language use for language historians. For the purposes of historical sociolinguistic studies, perhaps the single most important characteristic of a personal letter is that it is holograph, that is, written entirely by the person in whose name it appears. As the quantity of the material is often at issue in the empirical study of language history, royal letters can serve as a relevant source of primary evidence for language historians. One reason for the high survival rate of royal letters is that they consist of important government business. In the Middle Ages a special bureaucracy was set up in England for the running of it, with a central role assigned to the clerks of the Chancery. However, from the late medieval period onwards, holograph letters written by royalty in English have also been preserved for posterity. They provide valuable material for assessing the linguistic repertoire of the royalty as individuals rather than as part of the administrative organisation of their time. In this article I will discuss the range of linguistic variation in royal letters, both holograph and secretarial, in the Tudor period. I will contrast the holograph letters of Henry VIII with those written by his secretaries, and analyse them from the viewpoint of ongoing language change. In the second half of the article I will move on to consider the role of the King’s secretaries and administrators as agents of language change, contrasting them with their contemporaries in other walks of life. My material comes from the Corpus of Early English Correspondence, which incorporates the King’s correspondence, ranging from his love letters to Anne Boleyn and more official holographs written in the 1520s and 30s to secretarial letters under the sign manual from 1517 to 1542. My results indicate that individu-

100   

   Terttu Nevalainen

als vary in the degree to which they participate in ongoing linguistic changes. The monarch and his secretaries display partly similar, partly divergent profiles of linguistic variation. Overall, upwardly mobile administrators are less likely to display innovative linguistic behaviour than their contemporaries, including the monarchs in whose service they were.

1 Background assumptions The question of scribal influence on language change contrasts the spoken and written media and, within the latter, the registers found in the context of private and official writing. My focus will be on the latter distinction, but I am not suggesting that the processes of language change found in writing are restricted to the written medium. Many, if not most, changes may be assumed to have a spoken language basis and as such represent what sociolinguists call a “change from below” in terms of social awareness. While acknowledging the need to study the written language in its own right and compare the two media of communication, I subscribe to the view proposed by Douglas Biber (e.g. 1988, 1995) that linguistic variation often has more to do with register variation than with a simple written–spoken dichotomy. Biber’s comparative work shows that even a language like Somali, with a relatively short history of literacy, displays register variation in writing. Consequently, I would maintain that socially relevant variation can be recorded in writing as well as in speech, and that writings from the past need not be an exception to this rule. There is, of course, no denying that the textual evidence preserved from earlier periods varies greatly depending on the language and period of time investigated, and access to more varied data sets indeed remains one of the major challenges in historical sociolinguistics focusing on earlier periods. The field can therefore be described in William Labov’s famous dictum “the art of making the best use of bad data” (1994: 11). These efforts are supported by the uniformitarian principle, which suggests that the present can be used to explain the past. The hypothesis that the “processes, events and causes” of language variation and change are today “the same that operated to produce the historical record” informs Labov’s own work (2010: 375). He nevertheless has reservations concerning the extent to which the principle applies to the earliest prehistory of a language, since little can be known about how different it was from the present. It must also be added that historical “sameness” cannot be assumed at the substantive level of reconstruction: although there are many continuities, medieval and modern western societies, for example, have

Register variation and language change   

   101

different social structures, regardless of the theories used to describe them (see further Burke 1992: 61–63). At the same time, similarities of social organization can have a comparable impact on language use despite diachronic and diatopic distance. Labov (1994: 23) argues that, in contrast to rural communities, cities have always been centres of linguistic innovations. There is plenty of historical evidence to suggest that, characterised by migration and social mobility, interaction in cities could give rise to language change, typically the levelling of regional variation (for a survey, see Nevalainen 2011: 282–286). Looking for comparable circumstances of this kind is, of course, normal practice in history, and serves as the foundation for tracing diachronic correspondences. My contribution to this volume explores register variation in official and personal correspondence.¹ The period is the first half of the 16th century and the writers under discussion are King Henry VIII and his secretaries and government officials. The data come from the Corpus of Early English Correspondence (CEEC), which is introduced in Section 2. The analysis divides into two parts, beginning in Section 3 with a comparison of the linguistic usage of the King in his personal correspondence and that found in letters composed by his secretaries in his name. Section 3 weighs the expectation that secretarial usage, which had its roots in medieval curial style, would be more conservative than the monarch’s own, private usage. Section 4 takes a more general view of the Tudor language community and explores the extent to which royal administrators and government officials participated in language change in progress, compared to their contemporaries. Section 5 discusses the results and returns to the argument that, as centres of migration and socio-economic activity, cities promote linguistic innovation and language change.

2 The Corpus of Early English Correspondence 2.1 Composition The primary data for this study come from a corpus of personal correspondence, which covers the period from the early 15th century to 1800. The compilers have aimed to provide material for historical sociolinguistic research by focus-

1 My research for this study was supported by the Academy of Finland (Academy professorship 2010–2014).

102   

   Terttu Nevalainen

ing on personal letters, i.e., holograph letters sent by one individual to another. However, where no such material was available in sufficient quantities, secretarial letters, drafts, and copies were included in the interest of social, regional, and temporal coverage (Nevalainen and Raumolin-Brunberg 2003: 43–52). The rich metadata recorded for each letter allows the researcher to study the letter writers both as individuals and as members of various communities and social groups, as well as in relation to the recipients of the letters. It also provides each letter with an authenticity code, making a distinction between holographs as opposed to secretarial letters, and authorial or secretarial drafts and copies. Two versions of the corpus have so far been published, namely the Corpus of Early English Correspondence Sampler (CEECS, 1999) and the Parsed Corpus of Early English Correspondence (PCEEC, 2006). The data used in this study comes from the original, unannotated version of the Corpus of Early English Correspondence (CEEC, 1998), which provides regionally and socially stratified material from 778 people over a period of 270 years. The letter writers are categorized according to social status, domicile and gender in Table 1.

Social status

Domicile

Gender

Royalty 3 % Nobility 15 % Gentry 39 % Clergy 14 % Professionals 11 % Merchants 8 % Other 10 %

Court 8 % London 14 % East Anglia 17 % North 12 % Other 49 %

Female 26 % Male 74 %

Table 1: Writer variables in the Corpus of Early English Correspondence, 1410–1681.

Table 1 shows the uneven social distribution of the corpus, which reflects partly the rate of literacy in late medieval and early modern England, and partly the survival of the written material now available in edited form. Apart from professional men, it was the nobility and the gentry who contributed most to the written record of this period. There was an additional bias in literacy skills: although most adult members of the highest social ranks could read, considerably fewer high-ranking women than men of the same social status knew how to write prior to the 17th century (Nevalainen and Raumolin-Brunberg 2003: 40–42). This gender bias is clearly in evidence when the corpus is divided into subcorpora: holograph letters produced in their entirety by female and male writers, compared to letters produced by scribes and secretaries for their female and male employers. Figures 1 and 2 show the absolute frequencies of words in these four

Register variation and language change   

Figure 1a: Holograph letters by female writers in the CEEC (raw data).

Figure 1b: Secretarial letters and copies of women’s letters in the CEEC (raw data).

   103

104   

   Terttu Nevalainen

Figure 2a: Holograph lettaers by male writers in the CEEC (raw data).

Figure 2b: Secretarial letters and copies of men’s letters in the CEEC (raw data).

Register variation and language change   

   105

categories with periods adjusted to the chronological distribution of the data. They confirm that the amount of holograph data produced by both women and men grows over the centuries (1a, 2a). The development is particularly striking in the case of women, who contributed hardly any holograph letters in the 15th century, but whose letters were nearly all holograph from the late 16th century on. With a maximum of about 480,000 words of holograph (Figure 1a) and 58,000 words of secretarial letters per period (Figure 1b), the amount of femaleauthored data in the CEEC is much more limited than male-authored material. But the number of authentic women’s letters rises steadily over time, and there was little need to include female correspondents’ secretarial letters or copies from the 17th century onwards.² Because of this “complementary distribution” of female holograph and scribal letters, an extensive comparison between the two in any one subperiod would not be practicable. Figures 2a and 2b, in turn, show that male correspondents contributed significantly more holograph letters (a maximum of 1,050,000 words per period in 2a) than secretarial letters or copies throughout the study period, and that no secretarial letters were sampled for the last two decades of the 18th century (2b). What looks like a remarkable increase in the number of secretarial letters in the decades around 1600 (260,000 words) is largely due to a number of individual writers’ letter-book copies and drafts of their out-letters. These include men like Nathaniel Bacon (1546–1622) and Thomas Wentworth (1593–1641), an East Anglian and a Northerner, respectively, who were sampled extensively in order to have a more balanced regional coverage in those particular decades. Figures 2a and 2b suggest that the 16th century is a good period to compare and contrast male holograph and scribal or secretarial letters in the CEEC. They furnish the material for the present study, where the focus is on royal and administrative correspondence from the first half of the 16th century.

2.2 Royal letters Akrigg (1984: 24–30) provides a general classification of letters by Tudor and Stuart monarchs, focusing on James VI of Scotland and I of England (1566–1625), but the classification is also applicable to earlier and later royal correspondence. The vast majority of royal letters were drafted and composed by clerks and secretaries and signed by the monarch, who did not necessarily take more than a

2 With one exception, the 18th-century copies of female-authored letters in Figure 2b are Mary Wollstonecraft’s private correspondence.

106   

   Terttu Nevalainen

glance at the text itself. However, unlike James I, there is evidence that Henry VIII (1491–1547) was in the habit of correcting the drafts presented to him by his clerks (Akrigg 1984: 24; compare Lerer 1997: 87). Akrigg classifies royal letters into three major categories: (1) those “under the sign manual”, signed at the head of the letter, (2) those signed at the close, and (3)  holograph letters. The first category consists of routine government paperwork, authenticated by the monarch with his signature, but not drafted or dictated by him. The second category contains secretarial letters that the sovereign signed at the close. These were considered more important than letters under the sign manual. The third category consists of holograph letters written entirely by the sovereign in whose name they appear. Akrigg adds that a number of royal holograph letters only survive in manuscript copies. The process of compiling the CEEC revealed the extent to which editions of royal correspondence were modernised. Stemmler’s 1988 edition of King Henry VIII’s love letters to Anne Boleyn (?1501–1536) is an exception in that it retains the authors’ original spelling and wording throughout, although it follows the common practice of capitalising the initial letters of proper names, adding modern punctuation, and expanding abbreviations. The edition includes the facsimiles of the originals in the king’s own hand, now the property of the Vatican Library, Codex Vaticanus 3731 A. Stemmler’s edition contains all the surviving seventeen holograph letters by King Henry to Anne Boleyn, eight in English and the rest in French. They are all undated but were written in 1527 and 1528, when the King was negotiating the annulment of his marriage to Catherine of Aragon. By way of illustration, one of the shorter letters is reproduced in (1). (1) The resonable request off your last lettres, with the pleasur also that I take to know them trw, causyth me to send yow now thes news: The legate whyche we most desyre aryvyd att Parys on Sunday or Munday last past, so that I trust by the next Munday to here off hys aryvall att Cales. And then I trust within a wyle after to enyoy that whyche I have so long longyd for to god pleasur and oure bothe comfort. No more to yow at thys present, myne awne darlyng, for lake off tyme, but that I wolde yow were in myne armes or I in yours, for I thynk it long, syns I kyst yow. Writtyn affter the kyllyng off an hart att a XI off the kloke, myndyng with god grace tomorow mytely tymely to kyll Another. By the hand off hym, whyche I trust shortly shall be yours – Henry Rx (King Henry VIII to Anne Boleyn, 1528; Stemmler, ed., 1988: 144; CEEC)

Besides these private love and news letters to Anne Boleyn, the CEEC includes the King’s correspondence over three decades, ranging from his holograph letters written in the 1510s and 20s to Thomas Wolsey (1471–1530), Cardinal and Lord Chancellor, to secretarial dispatches under the sign manual addressed to six

Register variation and language change   

   107

other recipients, dated between 1517 and 1542. They are all drawn from standard original-spelling editions, although the names of the royal secretaries are not usually recorded. The letter in (2) illustrates a secretarial letter under the sign manual addressed to Sir Henry Clifford, the first Earl of Cumberland (1493–1542), who was knighted at Henry’s coronation in 1509 and created Earl of Cumberland in 1525. (2) To our right trustie and right welbeloved cousin therle of Cumberlande. Henricus Rex By the Kinge Right trustie and right welbeloved cousin wee greete you well. And albeit for certaine matters of great importance touchinge us and the wealth of this our realme, wee were minded to send for you to repaire hither unto us. Yet tendringe your labors and eschewinge your charges in the same, wee have thought good to send unto you our trustie servantes William Brererton, one of the Gentlemen of our Privie Chamber and Robert Leighton our Chaplin and Thomas Writhesley one of the clerkes of our Signet, to open and declare our mind and pleasure unto you, desireinge you to give firme credence unto the same. And straightly charge you to keepe secret such matter as they shall shew unto you without disclosinge or communicatinge it to any personne or then those whome ye understand by your said servantes to be made privie thereunto without failing as wee singulerly trust you. Yeven under our signett at our castell of Windesor the xiijth day of June. (King Henry VIII to the First Earl of Cumberland, 13 June, 1530; Dickens, ed., 1962: 38; CEEC)

3 Holograph vs. secretarial letters 3.1 Formality As can be seen from the above illustrations, holograph and secretarial letters differ in style. The use of the royal we in the secretarial letter (2) has an obvious distancing effect, already created by the heavily formulaic address phrase To our right trusty and right welbeloved cousin. The power difference between the sovereign and the addressee is made explicit by the signature line, Henricus Rex By the Kinge. In several respects, royal secretarial letters follow the late medieval conventions of what Burnley (1986) calls curial prose, i.e., legal and diplomatic correspondence based on earlier Latin and Anglo-Norman models. It is characterised by features such as frequent use of present participles (touchinge, tendringe, eschewinge, desireinge), lexical doublets (open and declare, mind and pleasure, disclosinge or communicatinge) as well as various anaphoric devices of cohesion, particularly constructions of the form ‘the same (+ Noun)’ and ‘the said + Noun’

108   

   Terttu Nevalainen

(in the same, unto the same; your said servantes). Burnley (1986: 611) traces such practices to the emergence of English into administrative and documentary texts at the end of the 14th century. By contrast, the King’s letter to Anne Boleyn in (1) dispenses with such formulaic practices, even omitting the address formula at the beginning of the letter (but a closing formula testifying to the holograph status of the letter is included). Instead, the letter puts the correspondents’ relationship on a closely intimate footing towards the end by addressing the recipient as myne awne [own] darlyng. No royal we appears in this letter, which is written in the first-person singular. Combined with the frequent use of I, the letter displays features typical of spoken interaction, including as many as three instances of the pragmatic marker I trust and one of I think (see Brinton 1996: 212–215). However, regardless of the intimate nature of the letter, Anne Boleyn is consistently addressed as you, not as thou. Thou was an option in 16th-century English, but its use was socially and regionally circumscribed, and the use of you was the norm even between husband and wife in letters written by the nobility and the gentry (Nevala 2004, Walker 2007). The King’s letters to Anne Boleyn were all signed Henry Rx or simply H R. The distance created by this subscription was nevertheless minimal in comparison with secretarial letters, where the King’s name was given its full Latin form, Henricus Rex.³

3.2 Modernity and conservatism From the modern speaker’s perspective, it is not immediately obvious which one of the two sample letters would be linguistically more “modern”, and whether scribes and secretaries as a group can be viewed as agents of linguistic change. If anything, the stylistic variation discussed above would suggest that secretarial business letters involved more conservative linguistic features and less individual choice than royal holograph letters addressed to an intimate. However, we find that both (1) and (2) are “conservative”, for example, with regard to their Noun Phrase structures, which can have two definite determiners preceding the noun (oure bothe comfort, this our realme). The royal holograph letter also makes use of the long possessive form mine in determiner function (myne awne darlyng, myne

3 That the holograph and secretarial letters differ markedly with regard to their personal pronoun use becomes evident if they are fed into a Principal Component Analysis. I did this using the Text Variation Explorer program, which showed no overlap between the two kinds of letter category when two principal components based on pronoun use were extracted.

Register variation and language change   

   109

armes) and the relative pronoun which with reference to human antecedents in restrictive as well as non-restrictive relative clauses (the legate whyche and hym, whyche, respectively). It also shows the -th ending in the third-person present tense singular (causyth). Because of the nature of the communication, features such as the use of mine as a determiner do not occur in secretarial letters, but other forms and constructions undergoing change at the time do. To answer the question of how conservative or progressive the King’s secretarial letters were as opposed to his holograph messages, quantitative comparisons are in order.

3.3 Language change in royal letters Although all the surviving English letters of Henry VIII to Anne Boleyn have been included in the CEEC, his holograph letters in the corpus only amount to 2,120 words, including the two early letters that he wrote to Thomas Cromwell. The time span they cover is 1516–1528. By contrast, the royal secretarial letters have a total of 10,550 words, and cover the period 1517–1542. The 16th century belongs to the prenormative era in the history of English, which means that not even spelling practices were strictly regulated at the time. Variant forms are common in both holograph and secretarial letters. However, in some cases the King’s private spellings could be quite consistent. Table 2 contrasts his spelling of the second-person pronoun you with that of his secretaries. With one exception, the King used , whereas his secretaries overwhelmingly preferred . With the benefit of hindsight, one can say that, in this instance, secretarial usage prevailed in the emerging standard language.⁴

Holograph letters Secretarial letters

  1 237

 (2 %) (97 %)

59  8

Total (98 %)  (3 %)

 60 245

Table 2: Variant spellings of you.

4 Although mixed spelling practices were common in the late 15th and early 16th century, both kinds of royal usage had their earlier models and contemporary followers. For further discussion, see Nevalainen (2002).

110   

   Terttu Nevalainen

But in some other cases this conclusion is not warranted. If we look at the use of the second-person pronoun at the time, we find that the traditional subject form ye was in the process of being replaced by the object form you in the course of the 16th century (Nevalainen and Raumolin-Brunberg 2003: 60–61). A closer comparison of holograph and secretarial usage shows clear signs of the latter undergoing change during the period studied. Ye (spelled or ) was the subject form found in the secretarial letters until 1535, but it was increasingly being replaced by you between 1536 and 1542. Here the monarch’s own usage was more “modern” as the King had adopted the incoming form in his private communication well before his secretaries started to use it in his official correspondence.

Holograph letters 1516–1528 Secretarial letters 1517–1535 Secretarial letters 1536–1542

Total

–

  (0 %)

12

(100 %)

12

41

(100 %)

–

  (0 %)

41

20

 (26 %)

57

 (74 %)

77

Table 3: Forms of the second-person subject pronoun.

Other pronominal changes underway in this period involved the wh-relativizer paradigm. The complex form the which, often attributed to French influence, was giving way to the simple which (Nevalainen and Raumolin-Brunberg 2003: 73–75). As shown by Table 4, mixed usage is found in both holograph and secretarial letters, but, in aggregate, the secretarial letters prove somewhat more conservative than the holograph ones. which Holograph letters Secretarial letters

20 28

the which (95 %) (85 %)

1 5

Total  (5 %) (15 %)

21 33

Table 4: The relative pronouns which and the which.

An innovation in the relative pronoun paradigm was the introduction of who as the subject form in the 15th century. This development is a reflection of the encoding of humanness in the animacy hierarchy (Nevalainen 2012, Nevalainen and Raumolin-Brunberg 2002). Here the royal letters subcorpus would suggest that the secretarial letters are on the whole more innovative than the King’s holo-

Register variation and language change   

   111

graphs. But since the numbers in Table 5 are small, one would not want to read too much into them. All one can say is that the King favours which, and that the two instances of who in the secretarial letters come from two different letters.

Holograph letters Secretarial letters

which

who

that

6 3

– 2

5 9

Total (45 %) (64 %)

11 14

Table 5: Subject relative pronouns with human antecedents.

Our provisional conclusion concerning the leadership of linguistic change is a tie: the King’s holograph letters are more advanced as far as the incoming subject pronoun you is concerned, but less so when it comes to the spelling of this form. On the other hand, his secretarial letters appear more innovative if we consider the introduction of the subject relative who, but less so when variation between the forms the which and which is considered. It should be added that there are ongoing changes with regard to which both kinds of royal letter proved equally conservative. The third-person singular present tense indicative suffix is a case in point. Both secretarial and holograph letters only used the southern form -th (appeareth, causyth, longyth, etc.), which was in the process of being replaced by the originally northern form -s in the course of the 16th century. The Royal Court was not in the vanguard of this process of change (Nevalainen and Raumolin-Brunberg 2003: 177–180).

4 Agents of linguistic change 4.1 Framework Another more comprehensive approach to the issue of linguistic leadership is to analyse the language use of royal secretaries and others engaged in administrative correspondence, and compare it with their contemporaries. The statistical analysis proposed in Nevalainen, Raumolin-Brunberg and Mannila (2011) takes into account the phase of the change in progress and relates individual writers to the section of the contemporary language community that is available in the CEEC. The evidence provided by each individual is compared to that produced by the reference group. The bootstrap method (1,000 samples/period/change) is used to generate information about the variation in the use of the incoming form.

112   

   Terttu Nevalainen

Altogether six 16th-century processes of change were analysed, including those discussed above: the second-person subject form (ye vs. you), the thirdperson singular present indicative suffix (-th vs. -s), the form of the wh-relative (the which vs. which), and the form of the first- and second-person possessive determiners (mine vs. my). In addition, the disappearance of multiple negation (we cannot see nothing → we cannot see anything) and the loss of the preposition of in the object of the gerund (writing of the letter → writing the letter) were traced in the corpus. Focusing on Henry VIII’s contemporaries and near contemporaries, the corpus allows us to analyse their progressiveness or conservatism with respect to these ongoing processes of change in slices of 20-year periods. It is thus possible to test the idea suggested by the discussion of curial style that professional administrators and secretaries were linguistically more conservative than their contemporaries in other walks of life. As noted above, apart from secretarial letters and copies, the material included in the CEEC consists of holographs; this is also the case with the letters written by high government officials and administrators in their own name. The writers studied here were active in the first half of the 16th century, and most of their letters included in the corpus date to the period from 1520 to 1550.

4.2 Results Table 6 lists individuals who were found to be conservative with respect to at least two ongoing changes in any 20-year period in the first half of the 16th century. The longest process (verbalisation of the gerund) took over two and a half centuries to be (nearly) completed, and the shortest (ye → you) ran its course in about eighty years. The results of the analysis are particularly interesting in two respects. First, conservative writers come from two different backgrounds: (a) government offices, being typically holders of high administrative positions during King Henry VIII’s reign, and (b) the wool trade, i.e. merchants active in the City and on the Continent, writing in the 1540s. Secondly, the assumption that individuals are consistently either progressive or conservative is not borne out by this analysis (Nevalainen, Raumolin-Brunberg and Mannila 2011: 27–32). The second observation is interesting in that linguistic conservatism does not appear to be simply a matter of register variation. Table 6 records the relationship between the letter writers and the recipients of their letters. “Family” includes both nuclear and non-nuclear family members, and “other” records more distant recipients. The government officials mostly communicated in their official capac-

Register variation and language change   

   113

ity with non-family recipients, whereas the merchants corresponded with their family on both business and family matters.

Writer

Lifespan

Relation to letter recipient

Conservative Intermediate Progressive N of changes N of changes N of changes

Administrators Cromwell, Thomas

1485–1540

family; other 4

1

1

Fox, Richard

1447–1528

other

3

2

1

Elyot, Thomas

1490–1546

other

3

1

2

Gardiner, Stephen

1495–1555

other

3

–

3

More, Thomas

1478–1535

family; other 2

1

3

Wyatt, Thomas

1503–1542

family; other 2

3

1

Saunders, Ambrose

a. 1528–?

family

3

1

2

Cave, Anthony

1493–1558

other

2

2

2

Johnson, Otwell

1518–1551

family

2

2

2

Johnson, Richard

1521–?

family

2

1

3

Johnson, Sabine

a. 1528–?

family

2

1

3

Preston, Richard

not known

other

2

2

1

Merchants

Table 6: Relative conservativeness of individuals with respect to six changes in progress in the first half of the 16th century (modified from Nevalainen, Raumolin-Brunberg and Mannila 2011: 31).

The single most conservative writer in Table 6 is Thomas Cromwell, who was less inclined than his contemporaries to embrace the six incoming features: he was conservative in four changes and progressive in only one, i.e. the replacement of multiple negation with single negation and negative polarity items. As the origins of the decline of multiple negation can be traced back to professional usage in the late 15th century, this is not surprising if we think of Cromwell’s career as the Secretary to Cardinal Wolsey and King Henry VIII’s chief minister (Nevalainen and Raumolin-Brunberg 2003: 149–150, Nevalainen 2006). Richard Fox, Thomas Elyot and Stephen Gardiner were all conservative with respect to three ongoing changes, Thomas More and Thomas Wyatt with respect to two. However, Sir Thomas Wyatt, poet, courtier and ambassador to Henry VIII,

114   

   Terttu Nevalainen

was a middle-of-the-roader in as many as three other changes and Richard Fox, secretary and councillor to Henry VII and Bishop of Winchester, in two changes. An opposite pattern is shown by Stephen Gardiner, Henry VIII’s s principal secretary, ambassador and Bishop of Winchester, and Sir Thomas More, lawyer and secretary to Henry VIII, Lord Chancellor and humanist scholar, in that they were progressive with respect to three ongoing changes, as well as by Sir Thomas Elyot, senior clerk of the King’s council, diplomat and author, who was progressive with respect to two.⁵ Only one member of the wool merchant community, Ambrose Saunders, was conservative in three changes. He was John Johnson’s brother-in-law and business partner in Johnson and Co. The others listed in Table 6, John’s two brothers, Otwell and Richard Johnson, and Anthony Cave, the Merchant of the Staple to whom John was apprenticed, were conservative with respect to two changes, but progressive with respect to at least two others, as was Ambrose Saunders. John’s wife Sabine Johnson is also listed in Table 6, but her balance is tipped in favour of three ongoing changes. Richard Preston, who was in the service of the Johnson family, only provided enough data for five out of the six ongoing changes and proved less progressive than the rest. John Johnson, the head of the family enterprise, does not appear in either list because he fell between the two extremes in four changes out of six. It is noteworthy that Henry VIII himself is not included in Table 6, which is based on the minimum of two processes of change in the conservative category. The output of his holograph and secretarial letters places him in the intermediate category with respect to two changes, progressive in two others and conservative in only one. These scores tally with the observation discussed above that in some cases the King’s private usage was ahead of that of his secretaries, whereas in others the situation was the reverse. The quantitative method introduced in this section has put the question of agents of linguistic change on a broader empirical footing, making the results more representative of contemporary 16th-century usage within the limits of the available material.

5 William Paget, clerk of the Privy Council and Secretary of State, created Lord Paget of Beaudesert, might also have been added to the list of conservatives with his minimum of two changes but was excluded because his letters were all letterbook copies and written in 1547 or later.

Register variation and language change   

   115

5 Discussion 5.1 Linguistic divergence? Due to lack of space, it has not been possible to evaluate the relevance of the duration and stage of a change in progress to the degree in which language users participated in it. If a change is nearing completion, fewer people can obviously be conservative with respect to it, whereas the opposite is the case with incipient changes, where only a few people can qualify as progressive. However, the period studied proved a good choice in that none of the six processes was completed and only one (which) was nearing completion in the first half of the 16th century. Unlike the other recessive forms, the which had never been a majority variant and was clearly on its way out at the time. It was also shown that, overall, linguistic conservativeness is a relative notion. The great majority of the individuals in Table 6 were not only conservative with respect to some ongoing processes but at the same time were also progressive with respect to some others. As the processes advanced, they did not change sides from conservative to progressive or vice versa from one period to the next. Rather, where there was variation it occurred between the extremes and the inbetween position. What Table 6 does not show is the extent to which the government officials and royal administrators agreed on the individual changes they resisted. A closer look reveals that all six were unanimous in rejecting the northern suffix -s, and five in disfavouring the subject form you. Four were conservative with respect to the short determiner my. Only one was progressive in the latter two cases. These findings suggest a fair amount of consistency in their conservative professional use. The merchant circle was more divided with respect to these processes: four of its six members listed in Table 6 proved conservative with respect to you, but two were progressive; three resented my, but three promoted it; two disfavoured -s but three were among the leaders of that change. Sabine Johnson, the only woman among them, proved progressive in each of these cases. These three ongoing changes were thus treated differently by administrators and merchants. The processes that the officials did not promote were all related to the spoken medium, and two of them spread from the North of the country to the South (-s and my; Nevalainen and Raumolin-Brunberg 2003: 60–62, 67–68). We might think that these findings could also reflect age differences between the two groups, the merchants being generally younger than the administrators. This may indeed be a factor with some of the processes but not all: it was

116   

   Terttu Nevalainen

shown in Section 3 that King Henry VIII systematically used you in both object and subject functions (Table 3). There is also evidence that the third-person -s was found in the City of London earlier than at Court (Nevalainen and RaumolinBrunberg 2003: 178). Hence both register and locality played a role in the diffusion of these features. However, these differences were significantly levelled when three members in both communities proved progressive in abandoning double negation and promoting single negation accompanied by nonassertive indefinites. Four merchant community members were also progressive in the use of the relative pronoun which, as were three of the royal administrators. On the other hand, the two promoters of the verbalisation of the gerund were both government officials. In none of these three cases was Sabine Johnson in the progressive camp. The reason for that may be that these processes were typically associated with professional written usage. It is, however, noteworthy that they were not restricted to the administrators but two of them were progressively promoted in the merchant community as well.

5.2 Separate spheres? A capital city and major port, London was by far the most important city in early modern England. Political and economic centralisation in the Tudor and Stuart periods was reflected in the massive increase in the population of London. Estimates for 1550 vary between 70,000 and 120,000 inhabitants, but converge on 200,000 for 1600 and c. 400,000 for 1650 (Boulton 1987: 3, Finlay and Shearer 1986: 39). Relevant to the present study, Beier and Finlay (1986: 11–13) note that the number of high government officials probably stayed roughly the same during the period of intense growth. Officials of gentry rank and above have been estimated at about 400 during the reign of Queen Elizabeth I (1558–1603). But the figures rise if the households and secretaries of the major officials are taken into account. The officials included in the CEEC were all among these major government officials and administrators active at Westminster and formed a powerful coalition. The studies based on the CEEC indicate that those resident at Court exerted considerable linguistic influence on language changes in progress in the Tudor period (Nevalainen and Raumolin-Brunberg 2003: 157–184). Language change diffuses both horizontally and vertically, in regional and social terms. The capital city’s apparently separate spheres were interconnected in various ways. Economic activity created what Archer (2008) describes as a complex web of interdependency. He finds that suppliers of luxury materials, for example, served as brokers between City and Court. Merchants constituted

Register variation and language change   

   117

a particularly mobile group with extensive networks and loose network ties, which made them linguistic brokers par excellence. Johnson and Co. was active not only in London and Calais, which formed part of England at the time, but in France and in the Low Countries. They were particularly well connected in London. Winchester (1955: 25–31, 212–213) suggests that Otwell Johnson, who had been apprenticed to Sir John Cage, the Comptroller of the Royal Household, and lived in the City, was acquainted with many government officials. London news travelled fast: in a letter to his brother John, Otwell provides an eyewitness account of the public execution of Queen Katherine Howard, Henry VIII’s fifth wife, in 1542. The corpus of correspondence shows that language change typically diffused from the capital region to the rest of the country (Nevalainen and Raumolin-Brunberg 2003: 199). As we have seen, there were notable differences between City and Court but also evidence to the effect that this short distance was easily crossed in linguistic terms as well.

6 Conclusion In this paper I have studied two potentially conflicting hypotheses: first, that the language of rulers, government officials and, by extension, their secretaries is conservative and resistant to change, and second, that cities promote linguistic innovation and language change. Both are shown to hold true, the latter more generally than the former. The analysis of linguistic conservatism picked out two groups of letter writers who occupied different worlds in Tudor London, namely government officials and merchants. The incoming features that did not find favour among the administrators were all morphophonological (you, -s, and my), and two of them had northern origins. In this respect, government officials continued to use traditional southern forms, current dialectally and in legal and administrative documents and in religious use at the time. These three changes were also relatively slow to find unanimous acceptance within the rather more heterogeneous merchant circle. There was no consistently conservative individual – although Thomas Cromwell came close to being one. What is more, the leading conservative officials also proved progressive with respect to some ongoing changes. They were among the leaders of three linguistic changes connected with formal written registers, such as the disappearance of multiple negation. These processes were not confined to government quarters but were promoted in the merchant community as well.

118   

   Terttu Nevalainen

A similarly varied picture emerges in miniature from the case study on the usage of Henry VIII and his secretaries. Despite their agreement on issues like the third-person verbal suffix -th, the secretaries did not follow the King’s private usage, for example, in either the spelling or the form of the second-person pronoun you, and could be as divided as the government officials with regard to relative pronoun usage. We therefore find that royal secretaries and government officials both maintained and created register variation, and transmitted professional repertoires, all of which was reflected in their participation in linguistic changes in progress over time. In conclusion, when change is under way, variability is the norm. Incoming features were found in both official and private letters but more wide-ranging variability was witnessed in the mobile merchant circle than among government officials. Future work will no doubt throw more light on how these processes were patterned in registers other than correspondence.

7 References Akrigg, G.P.V. (ed.). 1984. Letters of King James VI & I. Berkeley, Los Angeles & London: University of California Press. Archer, Ian W. 2008. City and Court connected: The material dimensions of royal ceremonial, ca. 1480–1625. Huntington Library Quarterly 71(1). 157–179. Beier, A.L. & Roger Finlay 1986. The significance of the metropolis. In A.L. Beier & Roger Finlay (eds.), London 1500–1700: The Making of the Metropolis, 1–33. London & New York: Longman. Biber, Douglas 1988. Variation across Speech and Writing. Cambridge: Cambridge University Press. Biber, Douglas 1995. Dimensions of Register Variation. Cambridge: Cambridge University Press. Boulton, Jeremy 1987. Neighbourhood and Society: A London Suburb in the Seventeenth Century. Cambridge: Cambridge University Press. Brinton, Laurel. 1996. Pragmatic Markers in English: Grammaticalization and Discourse Functions (Topics in English Linguistics 19). Berlin and New York: Mouton de Gruyter. Burke, Peter. 1992. History and Social Theory. Cambridge: Polity Press. Burnley, J.D. 1986. Curial prose in England. Speculum 61(3). 593–614. CEEC = The Corpus of Early English Correspondence, 1998, comp. by Terttu Nevalainen, Helena Raumolin-Brunberg, Jukka Keränen, Minna Nevala, Arja Nurmi & Minna Palander-Collin. Helsinki: University of Helsinki. http://www.helsinki.fi/varieng/CoRD/corpora/CEEC/ index.html Dickens, A.G. 1962. Clifford Letters of the Sixteenth Century (Publications of the Surtees Society 172). Durham and London: Surtees Society. Finlay, Roger & Beatrice Shearer 1986. Population growth and suburban expansion. In A. L. Beier & Roger Finlay (eds.), London 1500–1700: The Making of the Metropolis, 37–59. London & New York: Longman.

Register variation and language change   

   119

Labov, William. 1994. Principles of Linguistic Change. Vol. 1. Internal Factors. Oxford: Blackwell. Labov, William. 2010. Principles of Linguistic Change. Vol. 3. Cognitive and Cultural Factors. Oxford: Blackwell. Lerer, Seth. 1997. Courtly Letters in the Age of Henry VIII: Literary Culture and the Arts of Deceit. Cambridge: Cambridge University Press. Nevala, Minna. 2004. Address in Early English Correspondence. Its Forms and Socio-Pragmatic Functions (Mémoires de la Société Néophilologique de Helsinki 64). Helsinki: Société Néophilologique. Nevalainen, Terttu. 2002. What’s in a royal letter? Linguistic variation in the correspondence of King Henry VIII. In Katja Lentz & Ruth Möhlig (eds.), Of Dyuersitie & Chaunge of Langage: Essays Presented to Manfred Görlach on the Occasion of his 65th Birthday (Anglistische Forschungen 308), 169–179. Heidelberg: Universitätsverlag C. Winter. Nevalainen, Terttu. 2006. Negative concord as an English “vernacular universal”: Social history and linguistic typology. Journal of English Linguistics 34(3). 257–278. Nevalainen, Terttu. 2011. Historical sociolinguistics. In Ruth Wodak, Barbara Johnstone & Paul Kerswill (eds.), The Sage Handbook of Sociolinguistics, 279–295. London: Sage Publications. Nevalainen, Terttu. 2012. Reconstructing syntactic continuity and change in Early Modern English regional dialects: the case of who. In David Denison, Ricardo Bermúdez-Otero, Christopher McCully & Emma Moore (eds.), with the assistance of Ayumi Miura, Analysing Older English, 159–184. Cambridge: Cambridge University Press. Nevalainen, Terttu & Helena Raumolin-Brunberg. 2002. The rise of the relative who in Early Modern English. In Patricia Poussa (ed.), Relativisation on the North Sea Littoral (LINCOM Studies in Language Typology 7), 109–121. München: LINCOM Europa. Nevalainen, Terttu and Helena Raumolin-Brunberg. 2003. Historical Sociolinguistics: Language Change in Tudor and Stuart England. Longman Linguistics Library. London: Pearson Education. Nevalainen, Terttu, Helena Raumolin-Brunberg & Heikki Mannila. 2011. The diffusion of language change in real time: Progressive and conservative individuals and the time-depth of change. Language Variation and Change 23. 1–43. Stemmler, Theo (ed.). 1988. Die Liebesbriefe Heinrichs VIII. an Anna Boleyn. Zürich: Belser Verlag. Walker, Terry. 2007. Thou and You in Early Modern English Dialogues: Trials, Depositions, and Drama Comedy (Pragmatics and Beyond N.S. 158). Amsterdam: Benjamins. Winchester, Barbara 1955. Tudor Family Portrait. London: Jonathan Cape.

Florian Dolberg

7 Quantifying gender change in Medieval English Sources: This study analyses the two latest, extant components of the AngloSaxon Chronicles, to wit the first and second Peterborough Continuations. These two conclude the Peterborough Chronicle, which, as far as is known, is the Chronicle version that was maintained longest, for almost a century after the Norman Conquest. This Chronicle is contained in a manuscript designated Bodleian Library, Laud misc. 636, and is kept at Oxford. This manuscript is quite beyond doubt a Peterborough document, and both Continuations are unanimously understood to have been written by one local scribe each. The first Continuation comprises near-contemporary piecemeal updates from 1122 to 1131 CE, the second Continuation, written about 1155 CE or a little later, covers the intervening time and events by topic rather than by chronology, indicating non-contemporary production en bloc. For more information on this manuscript please consult Section 2.1.5 and references cited there. Abstract: This study investigates changing gender assignment and exponence at the interface of Old and Middle English by quantitatively analysing the Peterborough Chronicle Continuations. Multivariate analyses of the results reveal significant facilitatory and inhibitory effects of formal, semantic and extralinguistic variables, supporting and expanding upon previous research. Pronominal exponents have adopted the new, referential ME gender-system in virtually all instances by mid-12th-century, while adnominal exponents still display some variation. Among the latter, gender reassignments effecting compliance with the new, referential ME gender-system drastically increase from around 40 % in the First Continuation (≈ 1120–1130) to a little over 80 % in the Second Continuation (≈ 1155). This indicates an instance of very fast-paced language change, in which a language feature (here: referential gender agreement) develops from a minor variant to the dominating form from just one generation to the next.

1 Introduction For a number of reasons, English is a highly interesting object of study for historical linguists. To start with, documents in Old English (OE), the language of Germanic tribes after they invaded Britain in the 5th century CE (see, for example,

122   

   Florian Dolberg

Stenton 1943[1989]: 1–3; Pyles 1964: 100), survive in comparably large numbers. Most texts are coherent works, as opposed to short inscriptions or brief notes. This fortunate state of documentation is mainly brought about by two factors: first, England’s early Christianisation in the sixth and seventh century by missionaries sent from both Hibernia (today’s Ireland) and Rome, bringing Latin scripture and hence a writing culture to the hitherto largely illiterate Anglo-Saxons and their chiefly oral culture (see e.g. Hunter Blair 1963[1965]: 8–9; Denison 1993: 9). Second, efforts ascribed to King Alfred elevated the vernacular to a language for record keeping, ecclesiastical matters and scholarly purposes in the late ninth century. These exertions brought forth a remarkably stable and homogeneous Schriftsprache used all over England (see Wrenn 1933: 85), often referred to as standard West Saxon (sWS).¹ This standard moderately evolved over the next century, so that documents from Alfred’s time (849–899) are classified as early West Saxon, while those of Ælfric of Eynsham (955–1010) and his contemporaries are categorised as late West Saxon (see Pyles 1964: 107). Another fascinating characteristic of English is that it changed radically and rather quickly in comparison to other Germanic languages, such as German. Old English is a typical Germanic language in that it is largely synthetic. Features include elaborate inflectional morphology, a relatively free word order, and agreement in gender, number and case.² By contrast, Middle English (ME), typically dated from the Norman Conquest to the end of the fifteenth century, shows a drastically impoverished system of inflections and agreement, while Modern English (ModE) exhibits a rather analytic system, with rigid word order and little to no inflections and agreement. One category that underwent a particularly profound upheaval in medieval English is gender.³ Gender is arguably “the most puzzling of the grammatical categories” (Corbett 1991: 1), possibly because it appears somewhat unnecessary, in the sense that it does not seem to mark relevant cognitive distinctions, unlike for instance, number (uniqueness vs. multitude) or case (semantic roles, e.g. agent, patient, goal, etc.). This is reflected in the fact that some languages, German for instance, display widespread use of gender, while others, such as Finnish, do not

1 It should be born in mind that this standard does allow and indeed frequently displays variation, rendering it rather dissimilar to the modern notion of orthography. 2 To prevent misunderstandings, gender, as used throughout this article, refers exclusively to the grammatical category and its realisation, not to the socio-psychological construct. 3 Taking the Norman Conquest in 1066 as the dividing line between Old and Middle English is convenient and traditional, but misleading: It implies a dichotomy which is in fact a continuum. Therefore, and because this study focuses on changes that occurred from Old to Middle English, this paper refers to the period in question as Medieval English.

Quantifying gender change in Medieval English   

   123

feature this category at all. Hockett (1958: 231) gives a concise definition of what gender is, though not what it does: “Genders are classes of nouns reflected in the behaviour of associated words.” Hence, a language’s gender-system needs to be described in two dimensions. On the one hand, there is the notion of gender assignment, defining how many gender classes a language distinguishes and on what basis a given noun is assigned to one of these classes. On the other hand, there is gender exponence, delineating which parts of speech agree with the gender of the noun they are associated with. The following section briefly sketches the systems of gender assignment and exponence in Old and Middle English, which we take as the beginning and endpoint of the development this paper aims to elucidate. The present study’s aims and objectives then derive from this basis.

1.1 Gender-systems of Old and Middle English Both endpoints of the Medieval English continuum utilise a tripartite gendersystem with the values masculine, feminine and neuter. Regarding Old English agreement, adnominal elements (determiners, adjectives, and some numerals) agree with the noun they are associated with in gender, number and case. Germanic languages generally code these categories with fusional affixes or forms, so that, syncretisms aside, one form or affix simultaneously encodes gender, case and number: for instance, the OE distal demonstrative þone encodes masculine accusative singular. Outside the noun-phrase, anaphoric third-person singular pronouns agree with their antecedents in gender and number. The nouns themselves do not carry gender markers. Dahl (2000) uses the term lexical to refer to the system of OE gender assignment, meaning a noun’s gender needs to be stored as part of the lexical entry as it is not necessarily indicative of sex or any other attribute of the referent. However, this does not mean that OE gender assignment is completely arbitrary: nouns denoting males are generally masculine and nouns describing females tend to be feminine, if the referent is human or an animal whose sex is salient (e.g. fearr.masc ‘bull’ vs. cuu.fem ‘cow’). Thus, gender, at least for human and some animal referents, does correlate with sex. By contrast, neutral referents, i.e. objects, substances, abstract notions, and life forms of unknown sex, may be of any gender.⁴ The left-hand column in Table 1 below provides an overview of OE gender assignment and exponence.

4 Neutral, in analogy to neuter for neither masculine nor feminine, is intended to signify neither male nor female.

124   

   Florian Dolberg

Old English (lexical):

(late) Middle English (referential):

Exponents:

– Determiners – Adjectives – 3rd sg. pronouns

– 3rd sg. pronouns only

System of assignment:

– Males → mostly masculine – Females → mostly feminine – Neutrals → any gender (m, f, n)

– Males → ‘he’ – Females → ‘she’ – Neutrals → ‘it’

Table 1: Old and Middle English gender-systems in comparison.

By (late) Middle English, this system had changed radically, as shown in the righthand column of Table 1 in English from the late Middle Ages until the present day, only third-person singular pronouns expound gender, and there is no agreement within the noun phrase. More importantly, the mechanism of gender assignment also changed: the pronounced tendency in OE for nouns denoting human and some animal referents to reflect sex in gender became a hard and fast rule for all nouns.⁵ This system of assignment is termed referential, as the referent determines the noun’s gender (see Dahl 2000).⁶ The examples below illustrate the systems of gender assignment and exponence of Old and Middle English as well as demonstrate that in OE not all nouns describing humans reflect their referent’s sex in their gender. (1) Ðá ðæt wíf geseah, ðæt hit him When that.neut woman.neut saw that it.neut him næs dyrn not-was hidden ‘When the woman saw that she was not hidden from him’ (Ælfric’s Gospel of St Luke, ed. Bright 1893: 38) (2) Wedde nat a wiffe for hir inheritaunce Wed not a.uni woman.fem for her.fem inheritance ‘Don’t wed a woman for her inheritance’ (Benedict Burgh’s Cato Major, ed. Förster 1905: 320)

5 Exempt from this are poetic devices, such as metaphor and personification, e.g. referring to the moon, nature or a ship with she. 6 Dahl’s (2000) terminology is used forthwith as it is of greater descriptive accuracy than the traditional terms “natural” and “grammatical” gender, and because the latter pair implies a “natural” gender-system to be less grammatical than a “grammatical” one and vice versa.

Quantifying gender change in Medieval English   

   125

Example (1) represents lexical gender assignment in Old English from around 1000 CE (see Bright 1893: xii).⁷ Example (2) shows the loss of adnominal gender exponence and referential gender assignment of anaphors that characterise Middle English from the mid-fifteenth century (see Förster 1905: 298). This brief characterisation of OE and ME gender-systems enables us to identify this study’s aims and objectives in the following section.

1.2 Organisation, aims and objectives The present paper has a twofold purpose and hence consists of two principal parts. To prepare the ground for the two empirical analyses, Section 2 documents the methodology and operationalisation employed, beginning with the introduction of some desiderata for a database and on this basis identifying suitable documents (see 2.1.). In Section 3, this study assesses the overall nature of this instance of language change. This part investigates whether the change in gender assignment (i.e. from lexical to referential gender) proceeded as a smooth, directed development, rather than by the Old English lexical system collapsing noticeably before the purely referential one of ME emerged – effecting an intermittent period of confusion or chaos as some studies report (e.g. Ausbüttel 1904; Markus 1988) (hypothesis 1). In other words, how much interim confusion or entropy did this paradigm shift effect? These sections also investigate whether adnominal gender exponence ceased before, rather than after, the system of gender assignment changed from lexical to referential (hypothesis 2). Stenroos summarises her findings for southwest Midlands ME gender in favour of the former: “Anaphoric pronouns thus seem to retain the Old English gender-system at least for some time after the loss of gender within the noun phrase” (2008: 459). From the analyses in Section 3, we also obtain an idea about the timescale in which the changes in gender assignment and exponence occurred. The text cohorts selected for analysis are about one generation apart (see 2.1.5.); if they noticeably differ in gender assignment and/or exponence, this would indicate fast-paced language change (hypothesis 3). Finally, it is by no means established that adnominal and pronominal gender exponents behave alike. In fact, Corbett’s (1991: 226) agreement hierarchy, given in (3), predicts the likelihood of referential agreement to increase monotonically from left to right. Restated for diachrony, anaphoric pronouns should adopt the referential system before adnominal

7 Please note that wif and hit are co-referential in example (1).

126   

   Florian Dolberg

gender exponents (hypothesis 4). In order to gain insight into these issues, a quantitative analysis of gender exponents ensues. (3) attributive < predicate < relative pronoun < personal pronoun The second part of this study (Section 4) is more explorative and scrutinises which predictors facilitate or inhibit a given noun-phrase to undergo gender reassignment. These predictors may be language-internal – pertaining to the form or the meaning side of the linguistic sign – as well as extralinguistic. After identifying and motivating putative predictors, their impact on the likelihood of a gender reassignment to occur is quantified by means of multivariate statistical analyses. Following a discussion of the findings in Section 5, Section 6 rounds off this article with a conclusion and outlook.

2 Methodology and operationalisation After identifying Anglo-Saxon writing that is likely to be representative and well suited to illuminate changing gender assignment and exponence in 2.1., Section 2.2. presents the method of data classification employed and explains its underlying rationale.

2.1 Textual basis Ideally, this study would analyse spoken language. As this is not possible, the next best option are texts that resemble spoken language. Four criteria have been used to identify sources most likely resembling speech, which are also otherwise well suited to retracing the development of gender assignment and exponence in Medieval English. However, the number of extant Medieval English texts is limited, their survival coincidental and hence possibly not representative of the overall linguistic picture, so that there is probably not a single perfect candidate. Necessity hence dictates compromise.

2.1.1 Prose rather than poetry or rhetoric Poetry consists of language carefully constructed with regard to formal aspects, such as rhyme, alliteration and meter. Following Jakobson’s (1960) classic model of language’s communicative functions, any instance of language is governed by

Quantifying gender change in Medieval English   

   127

a different hierarchical ordering of these six functions, to the effect that “[t]he verbal structure of a message depends primarily on the predominant function” (Jakobson 1960: 353). The hallmark of poetry then appears to be that form takes precedent over function, to the effect that its function is rendered to a large degree aesthetic. Hence, a poem’s structural make-up arguably influences word choice and syntax, possibly even grammar, to a non-trivial extent. Moreover, poetic language is unlike that used in every-day life: another part of poetry’s appeal lies in aesthetically pleasing ways of expression, for instance by means of animisms, anthropomorphisms, personifications, similes, and obviously metaphor and metonymy, none of which is typical of spontaneous speech.⁸ A similar line of argument rules out texts written for oral delivery. Although seemingly closer to speech than texts intended for silent reading, texts of the former type have a decidedly didactic purpose and make use of rhetorical and poetic devices to increase persuasiveness. Thus, poetry, along with hagiographies and sermons, represents highly planned manifestations of language, quite unlike spontaneous, natural speech. Non-fictional prose, by contrast, arguably focuses more on the information relayed, and is hence likely to feature less planned and more natural language.

2.1.2 Original Anglo-Saxon rather than translations Translations harbour the risk of interference. Close translations, glosses in particular, are prone to interference as they closely mirror the source language and hence are neither idiomatic nor resembling speech: Jones (1988: 22) judges them to “have no more usefulness for the student of morphological and especially syntactic change than vocabulary lists”.⁹ Hogg (1992: 15) arrives at an equivalent assessment, and Wrenn (1933: 83) pithily states: “A gloss is not language in the full sense.” However, the Anglo-Saxon vernacular did not replace Latin as the language for scholarly, administrative and ecclesiastical purposes, as documents in or containing Latin continued to be produced throughout the Old English period. More-

8 Of course, it is generally accepted by now that everyday speech and even human language in general heavily rely on metaphor and metonymy to facilitate communication and conceptualisation (see e.g. Lakoff and Johnson 1980). Nonetheless, the function, role and make-up of poetic metaphor is quite unlike the “ordinary” type. See for instance Lakoff and Turner (1989: 67–85) for an accessible overview. 9 Interestingly and in spite of this assessment, Jones (1988) analyses the OE glosses to the Lindisfarne Gospels and the Durham Ritual for changes in gender.

128   

   Florian Dolberg

over, Latin remained a learned lingua franca in Europe for centuries. Potential Latin influence hence cannot be ruled out for any surviving vernacular document. The issue of interference is thus rather one of probability and degree: risk and magnitude of interference is highest among glosses, and arguably the lowest likelihood and extent of Latin influence is exhibited by those texts which recount roughly contemporary events of local, regional or national character, as these are comparatively unlikely to be modelled directly on a Latin template. Prime examples of such texts are those parts of the Anglo-Saxon Chronicle that were constructed in relatively close temporal and spatial proximity to the events they relate.

2.1.3 Frequent reference to non-human entities Whereas the previous criteria pertain to idiomaticity and representativity of the language as such, this one is geared more towards achieving a balanced database: as explicated in 1.1., Old English nouns frequently reflect the referent’s sex, if the referent is human or an animal whose sex is salient. Hence, nouns whose gender corresponds to their referent’s sex already comply with the emergent, purely referential gender-system still in use today. Should these nouns exhibit variation or change in gender assignment, this would measure the amount of confusion or entropy triggered by a system in flux rather than the progress of change from a (partly) lexical to a purely referential system of gender assignment (see 2.2.). By contrast, nouns that denote neutral entities are typically not of neuter gender (see 2.2.) and hence must have been gender-reassigned at some point during the emergence of the referential gender-system. In short, it is the realm of neutral, sexless referents in which the change in gender assignment happens. Conversely, accidental, erroneous reassignments of gender ending previous congruence of sex and gender are indications of concomitant confusion or entropy. They would ensue, if at all, almost exclusively in masculine/feminine nouns describing humans and, to a lesser degree due to their relative scarcity, in neuter ones referring to inanimates. Ceteris paribus, a text solely about people is hence far less likely to exhibit variation in gender assignment than one exclusively about non-human entities. Consequently, if we want to quantify both the process and progress of change from partly lexical to purely referential gender assignment as well as the amount of confusion or entropy this involves, we are best advised to look for texts in which both types of referents are similarly prominent.

Quantifying gender change in Medieval English   

   129

2.1.4 Reliably dated Unlike the other criteria discussed here, this one leaves comparatively little room for compromise, if a realistic portrayal of the development of medieval English gender assignment and exponence is to be achieved. The textual basis for analysis needs to stem from a time when the change was under way, but was not yet complete: Curzan (2003) analysed pronominal gender assignment in texts from the Helsinki corpora of Old and Middle English and reported “no significant decrease in grammatical [i.e. lexical] gender agreement from 750 to 1150 CE” (Curzan 2003: 91), but “[t]he written texts in the Helsinki Corpus dated between 1150 and 1250 CE seem to provide the critical snapshot of the shift from grammatical [i.e. lexical] to natural [i.e. referential] gender” (Curzan 2003: 106). Such texts will display variation in gender assignment and exponence, which allows identification of predictors that facilitate or inhibit gender reassignment (see 4.). Unfortunately, medieval English documents rarely state their time or place of production, so that we are reliant on indirect evidence to localise a text in time. Obviously, the more independent lines of evidence indicate that a text originates from a certain timeframe, the more reliable this dating is. These types of evidence include palaeographic aspects, such as letter forms and conventions of manuscript production that changed over time. Also, identification of individual scribes on the basis of their handwriting can help to date a manuscript, if the same handwriting is contained in another manuscript, whose provenance is known. Thirdly, the text’s contents may include information from which the text’s time of production can be deduced: reference to an event or person thus indicates that the text in question must have been written after that event, respectively in or after the lifetime of the person mentioned. This type of evidence is obviously most likely to obtain from sources dealing with issues of society, politics, and history. Another type of evidence potentially conducive for dating historical texts and manuscripts are linguistic parameters. However, using language to date and localise a text can be methodologically inauspicious, as it bears the risk of circular reasoning.¹⁰ Hence, we should concentrate on non-linguistic evidence to establish the history and provenance of texts.

10 The circular reasoning would look like this: previous scholars date a document to a certain time on the basis of its language, and we then would use this dating to retrace the development of that language, thus effecting identity of explanans and explanandum.

130   

   Florian Dolberg

2.1.5 The Peterborough Chronicle Continuations The Peterborough Chronicle Continuations fulfil the above criteria rather well. The Peterborough Chronicle, contained in Bodleian Library, Laud misc. 636, is one of nine surviving manuscripts of the Anglo-Saxon Chronicle, and was maintained the longest, until 1154 CE. A chronicle is clearly non-fictional prose; the portions analysed here are not too stilted in terms of register and hence probably resemble speech (see 2.1.1.). In addition, as a document relating history, both human and non-human entities feature prominently (see 2.1.3.). The First Continuation spans the annals 1122–1131, of which 1125–1131 find entry into the present analysis, the Second Continuation comprises the entries 1132–1154 and is analysed in its entirety. The language of both continuations is described as “distinctively East Midland in dialect” (Irvine 2004: ciii) and as deviating from the standard West Saxon of the copied part extending to 1121 (see Home 2007: 20). Knowles (1949: 424) reports abbeys drafting their monastic personnel from the local populace in the thirteenth and late twelfth century, which lets Clark (1970: xxxvii) surmise that this practice was also followed in the 1120s and 1150s. Regarding the First Continuation’s time of composition (see 2.1.4.), the text shows tell-tale signs of contemporary production. Pertaining to its contents, Clark (1970: xxv) notes that the text professes ignorance about events to come: annal 1127 relates the accession of a new Peterborough abbot named Henry and gives an account of his rather reprehensible background, ending with a summarising statement betraying no knowledge of the detrimental influence that man was going to have on the abbey, related in later annals: Þis was his ingang. of his utgang ne cunne we iett noht seggon. (Peterborough Chronicle, Annal 1127, line 72, ed. di Paolo Healy 2000).¹¹ Besides this explicit reference to the monks’ own ignorance about future events, contemporary production is indicated by characterisations that are hardly justified with the benefit of hindsight: annal 1123 introduces Henry of Angély, apparently the same who would obtain the abbacy in 1127, “as a defender of monastic interests, without a hint of what he was later to mean for Peterborough” (Clark 1970: xxv). Palaeographic evidence also supports this interpretation: the first continuation features several changes of ink as well as the appearance of the writing, chiefly coinciding with the end of annals. These changes suggest the first continuation as having been entered in six blocks: 1122; 1123; 1125 to 1126, line 11; 1126, line 12 to 1127; and 1128 to 1131 (see e.g. Whitelock 1954: 14, Ker 1957: 425; Clark

11 “This was his entry. Of his exit we cannot yet say naught” (my translation).

Quantifying gender change in Medieval English   

   131

1970: xvi, xxv; and Irvine 2004: xix). Plummer (1899[1952]: xxxv) interprets these variations as changes of hands rather than variations within one hand, but he also deems the First Continuation contemporary (see Plummer 1899[1952]: xlvii). However, the assumption that there were only two scribes for the entire text, with a changeover of scribe at annal 1132, appears to be the scholarly consensus (see e.g. Bergs and Skaffari 2007: 7; Home 2007: 20–22). Contemporary production and a strong local focus make the existence of a Latin template and hence Latin interference unlikely (see 2.1.2.). The Second Continuation covers the years 1132–1154, but contains only six “annals”: 1132, 1135, 1137, 1138, 1140, and 1154. The material is ordered by topic rather than chronology: [U]nder 1137 the story reaches forward to 1144 (St William of Norwich) and even to 1155 (the end of Martin’s abbacy); then it reverts to 1138; then under 1140 is unfolded without reference to date, and not always in true order, an epitome of the military and political history of nearly fifteen years. (Clark 1970: xxv)

It appears that the Second Continuation subsumes matters of church and society under 1137 and those pertaining to military and politics under 1140 (see Clark 1970: xxv–xxvi). This very strongly suggests an entry en bloc rather than piecemeal updating. For instance, the consecration of the new minster at Peterborough – the previous one was destroyed by a fire in 1116 – is reported under 1137, but nevertheless gives the year as ðæt was anno ab incarnatione Domini Mcxl. a combustione loci xxiii, i.e. 1140 CE and 23 years after the place burned down. Moreover, again under 1137, the reign of King Stephen, also known as the Anarchy (1135–1154), is reported twice to have lasted 19 winters. This correct assessment must come from the benefit of hindsight. Furthermore, the text’s appearance is uniform, displaying no variation of ink or writing characteristic of the first continuation (see Clark 1970: xxv). An entry en bloc after 1154 is thus well evidenced and generally accepted. The date of production typically given is 1155 or a few years later (see e.g. Whitelock 1954: 31; Clark 1970: xii, x; lrvine 2004: xix). This gives us two text cohorts of similar length, written by one local scribe each. The one taken from the First Continuation dates to the 1120s and very early thirties, and the one comprising the Second Continuation dates to the late 1150s. Hence, the texts were composed roughly 30 years (i.e. one generation) apart, which may be enough to observe language change.

132   

   Florian Dolberg

2.2 Classification of noun phrases In the referential gender-system of Middle English and later stages, the referent’s sex is the sole and infallible predictor for gender in all nouns. By contrast, it is only partly effective in the lexical system of Old English and allows for exceptions (e.g. wif.neut, see examples 1 and 2). Hence, a sensible approach is to divide the noun-phrases obtained from the database according to whether an Old English noun reflects its referent’s sex in its gender, and thus already complies with the emerging referential system of Middle English. In order to sort the noun-phrases manually extracted from the texts into these two classes, each noun’s gender, its referent’s sex, and the gender of associated modifiers/anaphors is ascertained on the basis of Old English reference works (Bosworth and Toller 1889/1921, Bright 1923, Quirk and Wrenn 1957, Campbell 1959, Hall 1960, Mitchell 1985) These reference works typically reflect standard West Saxon (sWS). Therefore, we take sWS as a yardstick against which we measure changes in gender assignment. This is a purely methodological decision and does not imply that sWS grammar adequately describes the language of the Peterborough Chronicle Continuations or that sWS grammar was in any way psychologically real to the scribes. Rather, we quantify the progress of change in gender assignment and exponence by measuring to what extent the Peterborough Chronicle Continuations deviate from this standard. The examples (4) to (7) below illustrate the distinction between lexical class and referential class noun-phrases. (4) se ceose the.masc cheese.masc ‘the cheese’

(annal 1131, l. 9) (neutral referent)

(5) al ðæt all.neut that.neut ‘all that iron’

iren iron.neut

(annal 1137, l. 31) (neutral referent)

(6) Þat ilce the.neut same.neut ‘the same day’

dæi day.masc

(annal 1154, l. 9) (neutral referent)

(7) twa papes two.fem/neut popes.masc ‘two popes’

(annal 1129, l. 31) (male referent)

Quantifying gender change in Medieval English   

   133

As the nouns themselves did not change form in the event of a gender reassignment, it is necessary to utilise other evidence, provided by agreement: an instance of gender reassignment is manifest if a noun’s gender as listed in the reference works is different to the gender marking of its adnominal modifiers and/or anaphoric pronouns. Example (4) features a noun whose gender does not reflect its referent’s sex. Nouns of this type make up the lexical class, and we expect them to change their gender to the effect of achieving congruence of gender and sex (see hypothesis 1 in 1.2.). This change occurs if gender exponents modifying this noun (determiners, adjectives) or referring back to it (third-person singular anaphoric pronouns) reflect the referent’s sex rather than the gender the noun had in Old English, as is the case in example (6). The frequency of cases exemplified by (6) – i.e. originally lexically gendered nouns having acquired a new gender effecting congruence of gender and sex – thus measure the progress of change from lexical to referential gender assignment. The change is complete when all nouns in the lexical class reflect the referent’s sex in their gender. By contrast, all those nouns whose gender reflects their referent’s sex already in Old English belong to the referential class, as example (5) illustrates. These nouns are compatible with the emerging referential gender-system of Middle English, so that a gender change as shown in example (7) renders the noun in question incompatible with either gender-system. Hence, the frequency of cases parallel to example (7) measures how much entropy the paradigm shift from lexical to referential gender generates. The smaller this frequency, the more directed the development is. Table 2 below summarises the classification explicated above.

Definition

Example

Prognosis

gender ≉ sex ceose.masc ‘cheese’ need to change gender to comply with the new referential system Referential Class gender ≈ sex man.masc ‘man’ already comply with the new referential system and should not change gender

Lexical Class

Table 2: Lexical and referential class nouns.

While incongruence of sex and gender among nouns with male or female referents is rare, nouns that denote neutral entities are mostly not of neuter gender:¹²

12 Neutral entities are those that are perceived as sexless: objects, substances, abstract notions and life forms other than some game and domesticated animals.

134   

   Florian Dolberg

they typically display no correspondence between the referent’s sex and the noun’s gender. If we hypothetically assumed an equal distribution of masculine, feminine and neuter among Old English nouns for neutral referents, then two-thirds of them would need to be gender-reassigned, since only neuter ones are compatible with the new, referential gender-system. In fact, this proportion of neuters is considerably lower: Platzer (2001: 39) reports that only 21.94 % of OE nouns denoting neutral entities are actually of neuter gender.¹³ Bluntly put, neutral, i.e. sexless referents shun neuter gender in Old English. This means that the large majority of nouns describing male or female referents are found in the referential class, in addition to those neuter nouns that denote neutral referents (e.g. mynster.neut ‘minster/monastery’). Conversely, the majority of nouns describing neutral referents populate the lexical class. These are joined by very few non-feminine nouns referring to female entities, such as wif.neut ‘wife/woman’ or wifman.masc ‘woman’. Males appear to be always referred to with masculine nouns, the only exceptions in this data being two instances in which spelling variants of heafod.neut ‘head’ are used metaphorically to refer to leaders, parallel to modern expressions such as head of state (see example 10 below). As these two cases display an incongruence of sex and gender, they belong to the lexical class. Having divided the database according to gender assignment (lexical vs. referential), we proceed analogously with gender exponence: following Corbett’s (1991: 226) agreement hierarchy, pronominal gender exponents should exhibit a higher propensity to display a correspondence of gender and sex of the referent than adnominal exponents (see 1.2., hypothesis 4).

gender assignment gender exponence

gender ≉ sex

gender ≈ sex

adnominal pronominal

lexical class adnominals lexical class pronouns

referential class adnominals referential class pronouns

Table 3: Subdivisions of the database.

Furthermore, different (sets of) predictors may influence adnominal gender exponence as compared to the pronominal gender exponence investigated in

13 This compares to a congruence of sex and gender among human referents in 96.67 % of cases (see Platzer 2001: 39).

Quantifying gender change in Medieval English   

   135

Section 4 below. Consequently, this divides the database into four subsets, as shown in Table 3. Finally, a few noun-phrases feature gender exponents that reflect neither the noun’s gender in Old English, nor the referent’s sex – i.e. its Middle English gender. These cases, exemplified in (8) and (9), are instances of random fluctuation rather than directed development towards congruence of sex and gender and are consequently included in the referential class. (8) se the.masc ‘the town’

burch town.fem

(9) ðone hehmesse the.masc high.mass.fem ‘the high mass’

(neutral referent) (annal 1130 line 9)

(neutral referent) (annal 1125 line 17)

3 Development of gender assignment and exponence In total, the Peterborough Chronicle Continuations contain 1646 noun-phrases from annal 1125 to the final entry for 1154. These were manually extracted from the text and are distributed fairly evenly between the two samples: the analysed part of the First Continuation yields 870 noun-phrases and the Second Continuation contains 776 noun-phrases. Of these 1646 noun-phrases, all those must be discarded that do not exhibit gender agreement. A noun-phrase may lack agreement for one of three reasons: first, the noun may simply be bare, i.e. not modified with adnominal elements or referred back to with an anaphoric pronoun. Secondly, even in sWS, the inflectional paradigms display manifest syncretisms in all three dimensions of agreement (i.e. case, gender, and number).¹⁴ These syncretisms precipitate that even though a given part of speech normally agrees with its associated noun, it does not do so consistently. Consider Table 4 for illustration: the distal demonstrative in principle agrees with the noun it modifies. However, in the plural, it does not mark gender; in the feminine accusative it does not in the differentiate number;

14 Some of these syncretisms very probably pre-date Old English, such as one form coding both masculine and neuter in the dative singular, which is also found in modern German.

136   

   Florian Dolberg

and neuter singular it uses the same form for both nominative and accusative case. Hence, a gender exponent may not display gender agreement in a given instance, to the effect that the noun-phrase in question drops out of the analysis.

 

masculine

singular

plural

nom

acc

se

þone

gen

dat

inst

þæm,

þon,

þam

þy

nom/acc

þæs neuter

þæt

feminine

seo

þa

gen

dat

þara,

þæm,

þæra

þam

þa

þære

Table 4: Declension table for the distal demonstrative pronoun (‘that’, ‘the’, ‘those’), adapted from Campbell (1959: 290).

Finally, we analyse a document from the border zone between Old and Middle English, where the inflectional paradigms for case, number and gender are in the process of disintegrating. Hence, the text contains aberrant forms, which cannot be assigned to a particular category without engaging in arbitrary judgement calls. Thus, to minimise subjectivity, these aberrant forms are ignored. The resulting database contains 722 gender-marked noun-phrases and is summarised in Table 5 below:

total NPs

gendered NPs

% gendered

First Cont. (annals 1125–1131)

 870

457

52.53 %

Second Cont. (annals 1132–1154)

 776

265

34.15 %

Total

1646

722

43.86 %

Table 5: Noun-phrases from the Peterborough Chronicle Continuations.

It is conspicuous that the Second Continuation contains substantially fewer gender-marked elements, suggesting that the loss of adnominal gender exponence was well under way around 1150 (see 5.1.). The gender-marked noun-phrases are sorted according to lexical/referential class and gender exponent for each Continuation separately, shown in Table 6. The figures in Table 6 require some comment: what stands out most are the comparably low numbers of lexical class items in both Continuations. The reason for this is a combination of the Old English gender assignment system as such and a consequence of text type: generally speaking, as explicated in 2.2.,

Quantifying gender change in Medieval English   

Gender exponence

adnominal pronominal

First Cont. (annals 1125–1131)

   137

Second Cont. (annals 1132–1154)

Lexical class

Referential class Lexical class

Referential class

52  7

212 186

 64 152

39 10

Table 6: Distribution of gendered noun-phrases in the database.

almost all of the nouns denoting sexed entities reflect the referent’s sex in their gender and hence belong to the referential class. These are joined by about 20 % of nouns denoting unsexed entities, which happen to be of neuter gender (see Platzer 2001: 39). Conversely, the lexical class almost entirely consists of neutral entities referred to with masculine or feminine nouns. Ceteris paribus, the latter group is arguably smaller than the former. Additionally, reference to individual people tends to be recurrent, yet not so for neutral entities mentioned, which would explain the rather extreme difference in numbers between lexical and referential class pronouns (see Table 6 above).

3.1 Pronominal gender exponents Due to the very dissimilar ratios of lexically versus referentially gendered nounphrases among pronominal gender exponents as compared to adnominal ones, it makes sense to analyse the former separately from the latter. We begin with the smallest group, lexically gendered pronouns: with one exception aside (see example 10 below), all lexical class pronouns ceased to agree with their antecedent’s Old English gender and rather reflect the referent’s sex. (10) eall ða heaued læred & læuued þæt wæs on Engleland all the heads.neut learned and lay that.neut were on England (male referents) ‘All the head men, learned and lay, that were in England’ (annal 1125 line 32) The sentence in example (10) is a tricky case: heaued.neut ‘head’ is used metaphorically, meaning ‘leader’. In light of this, it is even more surprising that the anaphor reflects the gender of the source domain noun (‘head’, i.e. neuter) rather than the sex of the target domain (‘leader’, i.e. male). Moreover, both anaphor and antecedent are singular, as is the verb, while the context and other parts of speech (eall ‘all’, ða ‘those’) clearly establish that a group of people is meant, not an individual – or a body part, for that matter. It is methodologically unsound to generalise from one instance, so we forgo assigning a motivation to this case. It

138   

   Florian Dolberg

shall suffice to demonstrate that the language of the Peterborough Chronicle Continuations displays a large amount of variation and inconsistencies in aspects of grammar so basic that we normally take them for granted. For a bird’s eye view of pronominal gender exponence, compare Figures 1 and 2 below. The former contrasts lexical class pronouns of the contemporary First Continuation with those of the Second Continuation written about 30 years later. The latter does the same for referential class pronouns. In both figures, dark grey signifies the number of gender-reassigned instances, while light grey symbolises those that abide by their Old English gender. The vastly different intervals on the Y-axes are a result of a skewed distribution, already discussed above. As an almost perfect mirror image to the lexical class pronouns, those of the referential class never change their gender, again save one exception, shown in Figure 2 as a narrow dark grey sliver atop the left column. Both exceptions occur in the First Continuation, to the effect that pronominal gender in the Second Continuation is employed exactly as it is today; third-person singular pronouns invariably reflect their referents’ sex in 100 % of cases. The First Continuation, with one out of six lexical class pronouns abiding by the Old English system and one out of 186 referential class ones having a gender incompatible with either system (see example 11 below) shows the change from lexical to referential gender assignment among pronouns to be virtually, but not quite entirely complete.

Figure 1: Lexical class pronouns.

Figure 2: Referential class pronouns.

Quantifying gender change in Medieval English   

(11) his his

   139

dohter daughter.fem

þæt he æror hafde giuen þone kasere that.neut he earlier had given that emperor (female referent) ‘his daughter, whom he earlier had given to the emperor’ (annal 1126, l. 3)

In this case, the expected form is þa, the feminine accusative singular distal demonstrative.¹⁵ Instead, the neuter form occurs, even though the noun is feminine and its referent clearly female. While this might herald the future use of þat as an indeclinable relative pronoun, we should view this case as random fluctuation for the time being, as parallel instances are absent so far. Moreover, the sentence in question exhibits another agreement “error”: in Old English, recipient roles in ditransitive constructions are coded in the dative, while example (11) codes it in the accusative, i.e. þone.acc rather than þæm.dat (see Table 4). This again underlines the state of flux English grammar was in at that time.

3.2 Adnominal gender exponents Turning to adnominal gender exponence, Figures 3 and 4 below illustrate the ratio of reassigned (dark grey) to non-reassigned nouns (light grey). Figure 3 shows that in the First Continuation, about two out of five lexical class nounphrases abandon their Old English gender in favour of achieving a correspondence between gender and their referent’s sex, as illustrated in example (6). Only one generation later, in the Second Continuation, this proportion increases to a little more than four reassignments out of five lexical class noun-phrases. Conversely, Figure 4 depicts gender reassignments in the referential class, which end the previous correspondence of gender and sex and “erroneously” effect a gender assignment incompatible with either system, as shown in example (7). These occur in almost every fifth instance among the gendered noun-phrases taken from the First Continuation, about half as frequently as reassignments in the corresponding lexical class data (see Figure 3). The trend of referential class reassignments from the First to the Second Continuation is reversed with respect to the one observed in the lexical class: three decades later, the ratio of “erroneously” gendered noun-phrases dropped from under one in five to under one in six. In comparison, the development of gender assignment in adnominal exponents (see Figures 3 and 4) is similar to that of gender-marked pronouns as

15 Old English did not have relative pronouns, but used demonstratives and/or the uniform relative particle þe for this purpose (see e.g. Campbell 1959: 291).

140   

   Florian Dolberg

Figure 3: Lexical class adnominals.

Figure 4: Referential class adnominals.

shown in Figures 1 and 2 above: both among pronominal and adnominal gender exponents, reassignments in the lexical class increase in frequency from the late 1120s (First Continuation) to the late 1150 (Second Continuation), while this frequency decreases in the referential class. However, adnominal exponents are less advanced than pronominal ones: in terms of deviance from the developing referential gender-system (i.e. non-correspondence of gender and sex), two out of 193 items (1.04 %) of the pronominal gender exponents from the First Continuations do not comply with this coming system, while out of 262 adnominal gender exponents 30 lexical class cases remain unchanged and 39 referential class ones are “mis-assigned”, resulting in 26.34 % deviation. From the First to the Second Continuation, the 1.04 % deviance from purely referential gender assignment among pronouns dwindles to zero, since the Second Continuation pronouns are in total compliance with the emerging referential gender-system. Deviation among all adnominal gender exponents decreases from 26.34 % in the First Continuation to 16.5 % of cases (i.e. seven nonreassigned lexical class cases and ten “erroneously” reassigned referential class items out of 103 gender-marked noun-phrases, shown in the left-hand columns of Figures 3 and 4). Having characterised the overall development of gender assignment and exponence in the Peterborough Chronicle Continuations, we can now move on to

Quantifying gender change in Medieval English   

   141

ascertain predictors inhibiting or facilitating a gender reassignment to occur in a given noun-phrase.

4 Predictors for gender reassignment A number of previous studies scrutinised gender assignment in Medieval English; their results provide a starting point for the ensuing analysis of predictors for gender reassignment. The following briefly summarises some of their most important findings and hypotheses. Jones (1967a, 1967b, 1988) proposes that gender markers are re-classified as case markers. Given that Germanic languages code case, number and gender together in fusional affixes or forms, it stands to reason that one category may influence the others. However, most of the evidence Jones brings in support of the gender-to-case-hypothesis, pertains to one particular form, the masculine accusative singular distal demonstrative þone. This does not suffice to support Jones’ claim that the Medieval English gender-system in total turned into a case system, rather than a weaker hypothesis that this particular form came to mark accusative/object case only and lost its gender specificity (i.e. masculine). Moreover, the studies’ methodology is somewhat opaque. It appears to count hits (þone in nonmasculine object-case positions) but to ignore misses (e.g. se in non-masculine subject-case positions and other permutations). Also, as mentioned above (see 2.1.2.), Jones’ studies use glosses as a textual basis, which is a less than ideal choice. These concerns aside, said studies warrant including case as a possible predictor for gender reassignment: Jones’ gender-to-case-hypothesis predicts that gender reassignments should be especially frequent in the accusative. Curzan’s (2003) study focuses on pronominal gender exponence in Middle English (1150–1250), finding considerable variation. Specifically, she reports that anaphors display an increasing likelihood of innovative referential agreement with increasing distance from their antecedents (Curzan 2003: 99, 119). This result also obtains from Köpcke and Zubin (2009: 141–142) for Modern German, which has a gender-system parallel to that of Old English. While apparently a robust predictor, it appears to be only effective for pronominal gender exponents. However, distance from antecedent, or more generally, from the associated or controlling noun, can be implemented into the agreement hierarchy (Corbett 1991: 226), reproduced again as (12) below: (12) attributive < predicate < relative pronoun < personal pronoun

142   

   Florian Dolberg

We can understand the agreement hierarchy as a measure of distance, increasing from left to right: attributives occur in the same phrase as their controlling noun, and thus are on average closest to their controller. Predicative exponents are in the same clause, which translates to a somewhat larger mean distance. Relative pronouns appear in the same (matrix) sentence as the controlling noun, hence being further removed still. Anaphoric pronouns, finally, display the largest average distance to their controller, as they may occur several sentences later. On this premise, a diachronic application of agreement hierarchy predicts anaphoric pronouns to adopt referential gender first, while adnominal gender markers should abide by the Old English lexical system for the longest time. Curzan (2003) also proposes dividing the body of nouns into semantic categories under the term of lexical diffusion, based on Harris and Campbell (1995). She reports these categories as differing in their readiness to adopt the innovative referential agreement: landmarks (e.g. rivers, dwellings), heavenly bodies (e.g. sun, moon) and notions of time (e.g. day, month) appear to be especially resilient to the influx of referential agreement (Curzan 2003: 100, 114). Stenroos (2008) analyses 20 southwest Midland texts from 1175 to 1300 for pronominal gender assignment, contrasting animate and inanimate referents. Among animates, human referents are consistently referred to with pronouns reflecting the referent’s sex, regardless of the noun’s original gender.¹⁶ Animals maintain their Old English gender regardless of their sex, while human and animal infants are consistently neuter. Among inanimates, half of the originally feminine or masculine nouns are referred back to in accordance with the OE lexical system, while a third of them take only the neuter (h)it as an anaphor. In sum, pronouns for human(oid) entities are invariably of referential gender, those referring to inanimates are fairly evenly divided between lexical and referential gender assignment, and all those referring to animals abide by lexical gender assignment in the period covered (see Stenroos 2008: 459). She derives from these findings an emerging “Southwest Midland Gender-system” in which acquisition of neuter gender in formerly masculine or feminine nouns is largely facilitated by factors included towards the right-hand end of the hierarchy of individuation (see Figure 5 below).

16 Nouns denoting imaginary humanoids (demons, elves, gods, etc.) behave likewise, although their referents, in default of existence, have no body and hence no sex. Instead, their culturally conventionalised gender (here referring to the socio-psychological construct) is construed as sex and determines the gender (referring to the grammatical category again) of the pronoun (see also Zaenen et al. 2004: 121).

Quantifying gender change in Medieval English   

   143

This notion is promising for predicting changes in gender assignment as it is known to be a pervasive typological parameter for case, gender and number as well as a number of other grammatical systems (see e.g. Silverstein 1976; Timberlake 1977; Dixon 1979; Comrie 1981; Croft 2001, 2004). In short, the hierarchy of individuation collapses the dimensions of humanness/animacy, concreteness and countability into one cline. On this cline, proper names are most individuated, followed by humans, animals, physical objects and abstracts. Least individuated are mass nouns. Each language or variety employing a mixture of referential and lexical agreement may place the cut-off point anywhere on this cline. Siemund (2008) finds West Somerset English to refer only to abstracts and mass nouns with the neuter pronoun, while Stenroos (2008) finds reference with neuter pronouns to include also physical objects in southwest Midland Middle English. Both varieties of English yield evidence that non-neuter mass nouns and abstracts switch to neuter before count nouns and concretes.¹⁷ Figure 5 summarises the hierarchy along with Siemund’s (2008) and Stenroos’ (2008) findings.

Figure 5: Hierarchy of Individuation with cut-off-points for Standard Modern English (dashed line), South-West Midland Middle English (black line), and West Somerset English (grey line). Adapted from Sasse (1993: 659), Siemund (2008: 140), and Stenroos (2008: 468).

This brief overview of previous research endows the ensuing multivariate analysis with a number of putative predictors likely to influence gender reassignment.

17 It is interesting to note that these two similar varieties are separated by only about 100 miles (ca. 160 km) in space, but by several hundred years in time.

144   

   Florian Dolberg

4.1 Variables The following sections reformulate the predictors reported in the previous section as six variables, which find entry into the binary logistic regression models (see 4.2.). The first two variables (4.1.1. and 4.1.2.) are located on the form side of the linguistic sign, the following three (4.1.3. to 4.1.5.) pertain to the meaning side, while the last one (4.1.6.) is of extralinguistic nature.

4.1.1 Case According to the gender-to-case hypothesis as put forth by Jones (1967a, 1967b, 1988), case should influence gender assignment. Old English distinguished four cases: nominative, accusative, genitive and dative.¹⁸ Because the case system underwent changes at least as fundamental as the gender-system, the case that the context requires – i.e. by verbal rection or preceding prepositions – frequently differs from the case that is morphologically marked on the noun-phrase. Hence, it is necessary to code the two manifestations of case as two separate variables, termed morphological and structural case.¹⁹ Moreover, a given form may code for more than one morphological case due to (pre-) OE syncretisms in the case paradigm: strong feminine nouns, for instance, inflect in –e for genitive, dative and accusative singular as well as (in later or non-sWS texts) nominative and accusative plural (see e.g. Campbell 1959: 234–241). Conversely, many verbs and prepositions license more than one structural case: the preposition to ‘to’ for instance triggers dative, genitive, or accusative (see Mitchell 1985, vol. I, 498, 513–514). This two-sided ambiguity requires coding structural and morphological case as four binary variables each. In addition, we code clashes of structural and morphological case as another, ninth binary variable.

18 Additionally, it features remnants of an instrumental case. Instrumental case forms play no role in the present data, having been supplanted by the dative. 19 The notion of structural case is not universally accepted (see Huddleston and Pullum 2002: 456–457). This study remains neutral on this issue and uses structural case as a descriptive tool only.

Quantifying gender change in Medieval English   

   145

4.1.2 Number As mentioned before (see 1.1. and 4.), gender, case, and number are coded together in fusional affixes and forms. On the basis that what is coded together may well mutually influence each other, this study includes number. Old English distinguishes singular and plural,²⁰ giving rise to a binary independent variable.

4.1.3 Abstractness This variable, as well as the following one, obtains from the hierarchy of individuation. For ease of implementation and for higher precision, the two are encoded separately. Abstractness enters the analysis as a binary variable, with all those entities being counted as abstract that cannot be perceived via the five senses and/or need references to other concepts and their interrelation to make sense. For example, sunrise is concrete, as it can be seen and conceptualised on the basis of that visual input alone. By contrast, sin is abstract, since it has no colour, shape, smell, texture, taste or sound. To conceptualise sin, we need concepts of religion, gods, morality, authority and numerous others in addition to knowing how these concepts are related.

4.1.4 Countability The second variable taken from the hierarchy of individuation encodes whether the referent is conceptualised as a discrete object or as an amorphous, unbounded substance. This difference in conceptualisation is reflected linguistically in the mass/count noun distinction, with the former not occurring with the indefinite article or in the plural.²¹ Stenroos (2008: 466) found countability facilitating pronominal gender assignment, which is reason enough to investigate whether this predictor also influences agreement within the noun-phrase.

20 Old English also had a dual, but only for first and second person personal pronouns, which are not gender marked. Hence, we can ignore it. 21 Except in type readings (e.g. wines of Australia).

146   

   Florian Dolberg

4.1.5 Sex The sex of a noun-phrase’s referent may influence the likelihood of gender reassignment. As female referents occur too infrequently in the database to yield meaningful statistical results, sex is coded as a binary variable with the values sexed (i.e. male/female) versus unsexed (i.e. neutral).

4.1.6 Time of composition Time of composition of a noun-phrase arguably influences the likelihood of gender reassignments to occur (see Figures 1–4). To ascertain whether this influence is statistically significant, and in order to quantify by how much a later noun-phrase is more likely to sport gender reassignment, the regression models below include time as a binary variable with the values First Continuation (i.e. late 1120s) and Second Continuation (i.e. late 1150s).

4.2 Binary logistic regression models A binary logistic regression analysis quantifies simultaneously the impact of several independent variables (predictors) on one binary dependent variable (i.e. whether or not a gender reassignment occurs), which makes it an ideal tool for the present purpose.²² Since pronominal gender exponents exhibit next to no variation in this database (see Figures 1 and 2), the intended extension backwards in time of Curzan’s (2003) and Stenroos’ (2008) analyses of pronominal gender assignment is not possible. Instead, we must contend ourselves with quantifying predictors for variation in adnominal gender assignment. As before, we distinguish between lexical and referential class noun-phrases and construct one model for each class.

4.2.1 Baseline models The first step is to establish a baseline likelihood for a gender reassignment to occur: in sum, 54 out of 91 lexical class adnominals display reassigned gender

22 For an in-depth characterisation of this statistic, the interested reader is referred to the relevant literature (e.g. Pampel 2000, Backhaus et al. 2008, Gries 2009).

Quantifying gender change in Medieval English   

   147

(see Figure 3), so a model would be correct in 59.34 % of cases just by predicting a reassignment to occur in every instance. The referential class contains a total of 49 reassignments out of 274 items (see Figure 4), so the baseline accuracy is 82.12 % by predicting no reassignments to occur at all. These baseline models are given in Tables 7 and 8 below, where “1” signifies reassigned gender and “0” absence thereof. predicted

 

observed

0

1

correct

0 1

0 0

37 54

  0 % 100 %

total

 

 

 59.34 %

Table 7: Predictive accuracy of the baseline model for lexical class adnominals.

predicted

 

observed

0

1

correct

0 1

217  47

0 0

100 %   0 %

total

 

 

 82.12 %

Table 8: Predictive accuracy of the baseline model for referential class adnominals.

The predictive accuracy of these baseline models (59.34 % for the lexical class and 82.12 % for the referential class) should increase by factoring in all those (and only those) variables that significantly influence gender reassignment, resulting in the final, minimal adequate model.²³

23 The minimal adequate model refers to that model which achieves the highest predictive accuracy with the fewest predictors, in accordance with the principle of parsimony (Occam’s razor).

148   

   Florian Dolberg

4.2.2 Model fitting and relevant variables The easiest way to arrive at a minimal adequate model is to begin with a saturated model, i.e. one that includes all 13 variables (nine for case and one each for number, abstractness, countability, sex, and time, see 4.1.), and to exclude the least significant variable (i.e. the one with the highest p-value).²⁴ A new model is calculated, from which again the least significant of the remaining variables is dropped, and so forth, until the model contains only significant variables.

Variable

2nd Cont. plural noun structural GEN structural DAT abstract referent constant

Regression

Standard

coefficient B

error

 2.720 −3.219 −2.012  1.807  1.571 −4.192

 .785 1.012  .922  .662  .702 1.306

Wald

df

p

Odds ratio

12.003 10.123  4.758  7.440  5.010 10.311

1 1 1 1 1 1

.001 .001 .029 .006 .025 .001

15.175   .040   .134  6.091  4.809   .015

Table 9: Variables included in the minimal adequate model for lexical class adnominals.

Variable

2nd Cont. unsexed referent plural noun structural NOM morphological DAT struct./morphol. clash constant

Regression

Standard

coefficient B

error

 −.819  2.890  2.291  1.206  1.694  1.529 −4.006

.489 .609 .538 .587 .479 .511 .900

Wald

df

p

Odds ratio

 2.811 22.517 18.122  4.216 12.530  8.940 19.827

1 1 1 1 1 1 1

.094 .000 .000 .040 .000 .003 .000

  .441 18.002  9.888  3.339  5.443  4.614   .018

Table 10: Variables included in the minimal adequate model for referential class adnominals.

After a couple of iterations, we arrive at a minimal adequate model each for the lexical and referential class adnominals shown in Table 9 and Table 10 above. A

24 Significance is measured as probability of error (p), which ranges from 0 to 1. A p-value of 0.05 or below counts as significant, meaning that, assuming the null hypothesis to be true (i.e. no influence of the variable in question), the chances of obtaining this (or a more extreme) result is at or below 5 %. A p-value of 0.1 or below is rated marginally significant.

Quantifying gender change in Medieval English   

   149

brief description of the most important measures reported in these tables may be helpful: a positive regression coefficient indicates facilitation, a negative one inhibition. The standard error measures the average difference between observation and prediction, i.e. the variable’s predictive accuracy. The odds ratio is simply the frequency of an event’s occurrence divided by its frequency of nonoccurrence under the condition specified by the variable. An odds ratio of 1 means no effect of the variable, one close to zero signifies a large inhibitory effect, and one approaching infinity a large facilitatory one. Two variables significantly influence the likelihood of gender assignments in both classes: in the Second Continuation, a lexical class noun-phrase is 14.175 times more likely to be gender-reassigned,²⁵ while a referential class noun-phrase is 0.441 times (i.e. 55.9 %) less likely to sport a reassigned gender in comparison to those of the First Continuation, though only marginally significant (p < 0.1). The effect of plurality is likewise reversed in the two models; plural lexical class noun-phrases are 96 % less likely to undergo gender change than singulars, while referential class plurals display an 8.888-fold increased probability of doing so. Among the lexical class, morphological case has no significant influence, but structural genitive and dative have opposite effects: a noun-phrase in genitive position is 0.134 times (86.6 %) less likely to switch gender than one that is not, whereas noun-phrases in dative position display a 5.091 times higher probability for undergoing gender change than noun-phrases in other positions. Of the semantic predictors, only abstractness of the referent has a significant impact, increasing a noun-phrase’s likelihood for gender change by a factor of 3.809. Turning to the referential class, most significant variables reported in Table 12 code formal aspects, all of which are facilitatory: structural nominative (i.e. subject position) increases a referentially gendered noun-phrase’s probability to change gender 2.339 times, noun phrases marked for dative display a 4.443-fold increased propensity for switching gender, and if structural and morphological case do not match the noun phrase in question, likelihood of gender reassignment is 3.614 times higher. Finally, noun-phrases describing unsexed referents are 17.002 times more likely to adopt a new gender than those referring to sexed entities.

25 The formula to transform odds ratios (rightmost columns in Tables 9 and 10) into percentages of difference is quite simple: % Δ = (eb – 1) * 100 (see Pampel 2000: 23)

150   

   Florian Dolberg

4.2.3 Final models Inclusion of the variables discussed in the previous section increases the predictive accuracy of the lexical class model from the baseline of 59.35 % (see Table 7) to 83.5 % (see Table 11 below), which translates to an increase of correct predictions by 59.4 %, brought about by including the variables listed in Table 9 above. This is reflected in Nagelkerke’s r² of 0.613, indicating a very good fit.  

predicted

 

observed

0

1

correct

0 1

25  3

12 51

67.6 % 94.4 %

total

 

 

83.5 %

Table 11: Predictive accuracy of the final model for lexical class adnominals.

 

predicted

 

observed

0

1

correct

0 1

213  34

 4 13

98.2 % 27.7 %

total

 

 

85.6 %

Table 12: Predictive accuracy of the final model for referential class adnominals.

For referential class adnominals, the fitted model in Table 12 predicts 85.6 % of cases correctly, thus being 19.46 % more accurate than the baseline model with 82.12 % correct predictions. This more moderate improvement is reflected in Nagelkerke’s r² of 0.386, which translates to a satisfactory model fit.

5 Discussion 5.1 The nature of change Hypothesis 1 predicts a directed development from the OE lexical gender-system to the entirely referential one of (l)ME. This prediction is supported by the lexical

Quantifying gender change in Medieval English   

   151

class gender exponents showing a very much larger contingent of reassignments that effect congruence of sex and gender as required by the new system than those in the referential class. Referential class reassignments, which end the previous correspondence of gender and sex and thus effect a gender assignment incompatible with either system, do occur, but are much less frequent. Understanding the latter as “errors” or “slips”, brought about by a system in flux, we can see that this amount of entropy decreases moderately, but noticeably and marginally significantly (see Table 12) as the change progresses towards completion: the First Continuation has 42.31 % of its lexically gendered noun phrases reassigned to effect referential gender and is thus near the half-way mark. Also, it features a fairly high contingent of “accidental” reassignments ending the prior congruence of gender and sex (18.57 %) among adnominal exponents. About thirty years later, the change nears completion, as 82.05 % of originally lexical class nouns adopted referential gender, and the amount of entropy, measured in referential class reassignments, drops to 15.63 %. Pronouns display a more advanced state of affairs: in the First Continuation, only two gendered pronouns deviate from referential gender of ME, while one generation later, all gendered pronouns reflect the sex of the referent. Hence, the development of medieval English gender is mostly a directed development with only some concomitant confusion. The widespread confusion reported by e.g. Ausbüttel (1904) or Markus (1988) indeed seems to be an artefact brought into existence by analysing texts from too wide a geographical and/or chronological range (see Stenroos 2008: 445). The present results also provide support for hypothesis 3, which predicts a fast-paced change: from just one generation to the next, reassignments of lexical class adnominal exponents almost double from roughly two out of five to about four out of five. Gendered pronouns show less of a development, simply because almost all that occur in the First Continuation already comply with the ME gender rule, so that there is very little room for development. This little room, however, is used to effect total compliance with the new referential system in the Second Continuation. Pronouns, then, changed earlier and/or faster than adnominal exponents, as predicted by hypothesis 4, derived from the agreement hierarchy. At the same time the present results falsify hypothesis 2 (loss of adnominal gender exponence before change to referential assignment), which is obtained from Stenroos’ (2008: 459) findings in later material from the southwest Midlands, where pronouns cling to the OE referential system even after adnominal gender exponence ceased. Stenroos’ results only seemingly contradict Corbett’s (1991: 226) agreement hierarchy, as the agreement hierarchy predicts what kind of agreement occurs, not if agreement occurs: once a category ceases to expound gender, the hierarchy no longer applies, as is the case with the uniform OE relative particle þe, which often occurs without an accompanying demonstrative. In

152   

   Florian Dolberg

the present data, the contingent of gender-marked noun-phrases decreases from the First to the Second Continuation from about one in two to just over one in three (see Table 5). For comparison, on average 50 % of noun phrases expressed gender in standard West Saxon,²⁶ suggesting that loss of adnominal gender exponence did not commence before the 1130s, but then took off quickly, effecting a reduction of gender exponence by 33 % in less than four decades. What still remains to be made sense of is the stark contrast between this study’s and Stenroos’ (2008) findings for pronominal gender: while the Peterborough Continuations display next to no pronominal gender variation, the later texts Stenroos analysed appear still to be in transition from lexical to referential assignment. The most obvious explanation may be dialectal variation: it is conceivable that the change in gender assignment moved sedately across England, in which case it just had not yet taken full effect in the southwest Midlands. More convincing, though, are the parallels between Stenroos’ (2008) proposed “Southwest Midland Gender-system” and the assignment system reported by Siemund (2008) for West Somerset English (see Figure 5). The relatively close geographical proximity of these two varieties suggests that in terms of gender, England’s South West underwent a development somewhat different from what became Standard English.

5.2 Predictors for gender reassignment in adnominals Gender assignment in adnominal exponents is influenced most by formal predictors. By contrast, both Stenroos (2008) and Curzan (2003) identified semantic predictors for pronominal gender assignment in ME. At this point we cannot ascertain whether the reason for this difference is due to a later stage in the development of English gender – in which case formal factors influencing language change would precede semantic ones – or whether it is effected by a difference in gender exponence – adnominal versus pronominal. Given that the amount of cohesion is greater between a noun and its adnominal modifiers than between antecedent and anaphor – noun-phrases agree internally in case, number and gender, whereas anaphors need only agree in gender and number – the latter interpretation appears more plausible. The following discussion refers to Tables 9 and 10.

26 Based on a 180 NP-sample taken from the Old English Orosius (ed. Bately 1980), which is “usually described as ‘typically eWS’ [early West Saxon]” (Bately 1980: xxxix), and dates to the late 9th/early 10th century (Bately 1980: xxiii).

Quantifying gender change in Medieval English   

   153

To begin with formal predictors, plural has the largest effect on gender reassignment likelihood in both models. It inhibits reassignment in the lexical class by a factor of 0.04 (−96 %) but facilitates it in the referential class by a factor of 8.888. Plural noun-phrases mark gender only on adjectives in nominative/accusative case, and do not distinguish masculine and feminine. Overgeneralisation could explain why the direction of influence is opposite in the two models: masculine/feminine plural adjectives in nominative/accusative case end in –e, while neuter ones are typically morphologically unmarked (sometimes they end in –u). As most members of the referential class refer to males (which are of masculine gender), while most lexical class items denote neutrals (which increasingly often acquire neuter gender), a regularisation towards the neuter ending would produce this effect. Despite the (formerly) neuter form being arguably less frequent than the masculine/feminine variant, the former then ultimately prevails, because it is morphologically unmarked and thus more “natural”, according to the theory of natural morphology (see e.g. Mayerthaler 1981). Among lexical class nouns (see Table 9), the inhibitory effect of structural genitive (86.6 % less) can be motivated by referring to the genitive as the lone survivor of the Germanic case paradigm in Modern English, theorising that historically stable grammatical categories inhibit language change in other paradigms. This logic also explains the facilitatory effect (5.091 times higher) of structural dative in the same model: the case system disappeared almost entirely, leaving nothing but the genitive and a few pronominal remnants of a generalised object case, suggesting that unstable categories promote change in other paradigms. For the referentially gendered noun-phrases, a clash of structural and morphological case promotes gender reassignments by a factor of 3.614. This also invites the interpretation that instability in one paradigm fosters instability in another, in analogy to the argument given above. Eluding a straightforward explanation for the moment are the effects of both structural nominative and morphological dative increasing gender change probability by a factor of 2.339 and 4.443 respectively (see Table 10). One, but probably not the only reason, is that, especially in the Second Continuation, dative nouns at times occur in subject position, effecting a case clash.²⁷ The extralinguistic variable passage of time has the second largest effect (14.17 times higher for the later dataset) in the lexical class model, but a rather smaller, marginally significant one (55.9 % less in the later cohort) in the refer-

27 A witness to this seems to be the ME/ModE place name Canterbury, (OE Cantwaraburh), which fossilised a dative form (byrig, dative singular of burh, compare Bosworth and Toller 1889/1921: 145).

154   

   Florian Dolberg

ential class model. This provides a quantification of what has been discussed in 5.1. above. Turning to semantic predictors, abstractness facilitates gender reassignment in lexical class adnominals (3.809 times higher), but not in referential ones. Abstractness thus influences both adnominal and pronominal (see Stenroos 2008) gender assignment, while countability, the second variable extracted from the hierarchy of individuation, has no significant effect on adnominal gender, but appears to influence pronouns (see Stenroos 2008: 464). Understanding referential class reassignments as manifestations of confusion or entropy (or, bluntly put, “errors”) would help explain why they seem relatively impervious to semantic predictors such as abstractness, but are susceptible to formal ones. The 17.002-fold higher reassignment likelihood of referential class nouns describing neutral referents being reassigned is best explained in terms of the corresponding decreased reassignment probability of nouns with male referents (female referents are scarce): as the change from lexical to referential gender assignment creates the greatest variance among neutrals, it is plausible that neutrals of neuter gender (hence in the referential class) are consequently more likely to be confused in terms of gender. In other words, the influence of this variable may ultimately not derive from attributes of the referent (i.e. sex), but rather from the circumstance that the transition from lexical to referential gender required the largest amount of restructuring among neuter referents. As a concomitant effect, then, some neuters denoting neutral referents were “accidentally” also reassigned (see example 7). Additionally (or alternatively), conceptual salience may be a driving factor: the other principal group in the referential class are males, who are human and thus more resistant to losing congruence of sex and gender, as humans are conceptually more salient than things or ideas. This study cannot confirm a higher resilience of certain semantic categories against adopting referential gender as found by Curzan (2003: 100, 114): 49 out of all 120 gender reassignments (40.83 %) and 7 out of 17 reassigned pronouns (41.18 %) in the Chronicle describe temporal, celestial or geographic entities. For example, ‘day’, the most frequent temporal concept in the database, occurs six times with its original masculine gender, all of them in the First Continuation, also six times with the innovative neuter gender, and 15 times without gender marking. One possible explanation could be that Curzan’s observation only holds for pronouns, and that the number of lexical class pronouns in this study is just too small to make the effect visible. In a similar vein, the present data provides no indication for gender markers becoming case markers, which Jones (1967a, 1967b, 1988) largely derives from an assessment of the masculine accusative singular distal demonstrative þone used with non-masculine nouns. To support Jones’ notion, we would need to find

Quantifying gender change in Medieval English   

   155

a positive correlation between accusative case and gender reassignments, i.e. a significant influence of the former on the latter. But neither minimal adequate model features accusative case. Instead, the models show structural or rather morphological dative to promote gender changes.

6 Conclusion and outlook In sum, the Peterborough Continuations portray the change from lexical to referential gender assignment as a remarkably quick and decidedly directed development. The latter characteristic ties in nicely with Stenroos’ (2008) assessment of gender development in southwest Midlands Middle English, even though the endpoints of the respective developments are distinct (see Somerset English as described by Siemund (2008)). Also in line with Stenroos (2008), this study finds no intermittent stage of confusion in between Old and Middle English gender as reported in earlier studies (e.g. Ausbüttel 1904; Markus 1988), although the rather smooth transition from lexical to referential gender assignment does effect some concomitant, and apparently transient entropy, as evidenced by “accidental” reassignment of referential class nouns. In contrast to the studies by Curzan (2003) and Stenroos (2008), this paper shows semantic predictors for gender reassignment to be secondary to formal ones, suggesting that formal characteristics chiefly influence noun-phrase-internal agreement, and semantic ones mainly affect anaphoric agreement. Finally, this study leaves some room for improvement: first, its database is not large enough to meaningfully address the issue of types and tokens. Also, it focuses on a fairly narrow section of medieval English, to wit early to mid-12thcentury Peterborough, while the amount of variation across time and space is extensive. Hence, the present results can only tentatively generalise beyond this period and area. These shortcomings notwithstanding, this paper hopefully constitutes a small, but pertinent, contribution to development of gender-systems in particular and language change in general.

7 References Ausbüttel, Erich. 1904. Das persönliche Geschlecht unpersönlicher Substantiva einschließlich der Tiernamen im Mittelenglischen seit dem Aussterben des grammatikalischen Geschlechts. In Lorenz Morsbach (ed.), Studien zur englischen Philologie XIX. Halle: Niemeyer.

156   

   Florian Dolberg

Backhaus, Klaus, Bernd Erichson, Wulff Plinke & Rolf Weiber. 2008. Multivariate Analysemethoden (12th ed.). Berlin: Springer. Bately, Janet M. (ed.). 1980. The Old English Orosius. Oxford: Oxford University Press. Bergs, Alexander & Janne Skaffari. 2007. Invitation to the Peterborough Chronicle and its Language. In Alexander Bergs & Janne Skaffari, (eds.), The Language of the Peterborough Chronicle. Frankfurt/Main: Peter Lang. Bosworth, Joseph & Thomas Northcote Toller. 1898/1921. An Anglo‐Saxon dictionary, based on the manuscript collections of the late Joseph Bosworth. Oxford: Clarendon Press. Bright, James W. (ed.). 1893. The Gospel of St Luke in Anglo-Saxon. Oxford: Clarendon Press. Bright, James W. 1923. An Anglo-Saxon reader and grammar. London: Allen and Unwin. Campbell, Alistair. 1959. Old English grammar. Oxford: Clarendon Press. Clark, Cecily. 1970. The Peterborough Chronicle 1070–1154. Oxford: Oxford University Press. Comrie, Bernard 1981. Language Universals and Linguistic Typology. Blackwell: Oxford. Corbett, Greville G. 1991. Gender. Cambridge: Cambridge University Press. Curzan, Anne. 2003. Gender shifts in the history of English. Cambridge: Cambridge University Press. Croft, William. 2001. Radical Construction Grammar. Oxford: Oxford University Press. Croft, William. 2004. Typology and Universals. Cambridge: Cambridge University Press. Dahl, Östen. 2000. Animacy and the notion of semantic gender. In Barbara Unterbeck & Matti Rissanen (eds.), Gender in grammar and cognition, 89–116. Berlin: Mouton de Gruyter. Denison, David. 1993. English Historical Syntax: Verbal Constructions. London/New York: Longman. diPaolo Healey, Antonette (ed.). 2000. Dictionary of Old English Corpus in Electronic Form. Toronto: University of Toronto. Dixon, Richard M.W. 1979. Ergativity. Language 55. 59–138. Förster, Max (ed.). 1905. Die Burghsche Cato-Paraphrase. Archiv für das Studium der neueren Sprachen und Literaturen 115 (15). 298–323. Gries, Stefan Th. 2009. Statistics for Linguistics with R: A Practical Introduction. Berlin: DeGruyter Mouton. Hall, J.R. Clark. 1960. A concise Anglo-Saxon dictionary. (4th ed.). Toronto, Buffalo, London: Toronto University Press. Harris, Alice C. & Lyle Campbell. 1995. Historical syntax in cross‐linguistic perspective. Cambridge: Cambridge University Press. Hockett, Charles Francis. 1958. A Course in Modern Linguistics. New York: Macmillan. Hogg, Richard M. 1992. Introduction. In Richard M. Hogg (ed.), The Cambridge History of the English Language, Volume I: The Beginnings to 1066. Cambridge: Cambridge University Press. Home, Malasree. 2007. The Peterborough Chronicle in Context. In Alexander Bergs & Janne Skaffari (eds.), The Language of the Peterborough Chronicle. Frankfurt/Main: Peter Lang. Huddleston, Rodney & Geoffrey Pullum. 2002. The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press. Hunter Blair, Peter. 1963[1965]. Roman Britain and Early England, 55B.C. – A.D. 871. London: Nelson. Irvine, Susan (ed.). 2004. The Anglo-Saxon Chronicle: MS E. Cambridge: Brewer. Jakobson, Roman. 1960. Closing Statement: Linguistics and Poetics. In Thomas A. Sebeok (ed.), Style in Language, 339–377. New York and London: The Technology Press of Massachusetts Institute of Technology and John Wiley & Sons.

Quantifying gender change in Medieval English   

   157

Jones, Charles. 1967a. The functional motivation of linguistic change. English Studies XLVIII. 97–111. Jones, Charles. 1967b. The grammatical category of gender in early Middle English. English Studies XLVIII. 298–305. Jones, Charles. 1988. Grammatical gender in English: 950 to 1250. London: Croom Helm. Ker, Neil R. 1957. Catalogue of Manuscripts containing Anglo-Saxon. Oxford: Clarendon Press. Knowles, David. 1949. The monastic order in England: a history of its development from the times of St Dunstan to the fourth Lateran Council 943 – 1216. Cambridge: Cambridge University Press. Köpcke, Klaus-Michael & David Zubin. 2009. Genus. In Elke Hentschel & Petra M. Vogel (eds.), Deutsche Morphologie, 132–154. Berlin: de Gruyter. Lakoff, George & Mark Johnson. 1980. Metaphors we live by. Chicago: University of Chicago Press. Lakoff, George & Mark Turner. 1989. More than Cool Reason: a Field Guide to Poetic Metaphor. Chicago: University of Chicago Press. Mayerthaler, Willi. 1981. Morphologische Natürlichkeit. Wiesbaden: Athenaion. Markus, Manfred. 1988. Reasons for the loss of gender in English. In Dieter Kastovsky & Gero Bauer (eds.), Luick revisited: Papers read at the Luick Symposium at Schloss Liechtenstein. 15–18.9.1985, 241–58. Tübingen: Narr. Mitchell, Bruce. 1985. Old English Syntax. Volume I: Concord, the Parts of Speech, and the Sentence; Volume II: Subordination, Independent Elements and Element Order. Oxford: Clarendon Press. Pampel, Fred C. 2000. Logistic Regression: A Primer. Thousand Oaks, CA: Sage. Platzer, Hans. 2001. ‘No sex, please, we’re Anglo-Saxon?’ On grammatical gender in Old English. Vienna English Working Papers 10. 34–47. Plummer, Charles. 1899[1952]. Two of the Saxon Chronicles Parallel. Vol. II – Introduction, Notes and Index. Oxford: Clarendon Press. Pyles, Thomas. 1964. The Origins and Development of the English Language. New York: Harcourt, Brace & World, Inc. Quirk, Randolph & C. L. Wrenn. 1957. An Old English Grammar. New York: Holt, Rinehart & Winston. Sasse, Hans-Jürgen. 1993. Syntactic categories and subcategories. In Joachim Jacobs, Arnim von Stechow, Wolfgang Sternefeld & Theo Vennemann (eds.), Syntax. Ein internationales Handbuch zeitgenössischer Forschung / An International Handbook of Contemporary Research (Handbücher zur Sprach- und Kommunikationswissenschaft / Handbooks of Linguistics and Communication, 9), 646–686. Berlin: de Gruyter. Siemund, Peter. 2008. Pronominal Gender in English: A Study of English Varieties from a CrossLinguistic Perspective. London: Routledge. Silverstein, Michael. 1976. Hierarchy of features and ergativity. In R.M.W. Dixon (ed.), Grammatical Categories in Australian Languages (Linguistic Series 22), 112–171. Canberra: Australian Institute of Aboriginal Studies. Stenton, Frank Merry. 1943[1989]. Anglo-Saxon England. Oxford: Oxford University Press. Timberlake, Alan. 1977. Reanalysis and actualization in syntactic change. In Charles N. Li (ed.), Mechanisms of Syntactic Change, 169–183. Austin: University of Texas Press. Whitelock, Dorothy (ed.). 1954. The Peterborough Chronicle (The Bodleian Manuscript Laud Misc. 636.) Copenhagen: Rosenkilde & Bagger.

158   

   Florian Dolberg

Wrenn, Charles Leslie. 1933. “Standard” Old English. Transactions of the Philological Society. 65–88. Zaenen, A., J. Carlette, G. Garretson, J. Bresnan, A. Koontz-Garboden, T. Nikitina, M. C. O’Connor & T. Wasow. 2004. Animacy encoding in English: Why and how. In D. Byron & B. Webber (eds.), Proceedings of the 2004 ACL workshop on discourse annotation Barcelona, July 2004, 118–125. East Stroudburg: Association for Computational Linguistics.

Merja Stenroos

8 Identity and intelligibility in Late Middle English scribal transmission: local dialect as an active choice in fifteenth-century texts¹ Sources: This article discusses two examples of the use of regional dialect in the late Middle English period. The first example is a field survey from Barmston, Yorkshire, dating from 1473, contained in a manuscript held in the Humberside County Record Office at Beverley (shelf mark DDCC/19/I). It was copied by at least four different scribes. Their identities are not known, nor are the circumstances in which the survey was carried out; however, the manuscript was probably in the use or possession of the rector of the local church. The text has been edited and described by Sandvold (2010) and was included in the Linguistic Atlas of Late Medieval English (McIntosh, Samuels and Benskin, 1986) as Linguistic Profile 1130. A transcription is available in the Middle English Grammar Corpus (MEG-C) version 2011.1. at http://www.uis.no/research/humanities/the-middle-englishscribal-texts-programme/meg-c/. The second example consists of a group of texts with West Midland connections, in particular two fifteenth-century manuscripts containing very similar scribal usage: Longleat, Marquess of Bath’s MS 5 (folios 1–35) and Oxford, St John’s College 6. The former contains a lectionary prefixed to a Wycliffite New Testament, while the other contains Lydgate’s Troy Book. Again, nothing is known about the scribes of these texts; however, the St John’s College manuscript is connected to the Leinthall family in North Herefordshire. Both texts were included in the Linguistic Atlas of Late Medieval English, as Linguistic Profiles 7520 and 7510 respectively; their dialectal characteristics were described by the present writer in Black (1997: 215–230). Abstract: The standardisation of English is often taken to begin in the fifteenth century; however, the term is used differently by scholars depending on their focus and interests. Scholars who focus on geographical variation may consider written English to be largely “standardised” by 1500; at the same time, written language continues to be highly variable well into the seventeenth century. While there seems to have been little active pressure towards uniformity, there is a ten-

1 The discussion in Section 3 of this paper builds upon research carried out by Silje N. Sandvold in her Masters thesis (Sandvold 2010). I am very grateful for her permission to include this material in the present paper.

160   

   Merja Stenroos

dency among scholars to relate variation in this period to an ongoing process of standardisation, often interpreting forms that are identical to those which survive in later standard English as “progressive” and other forms as “conservative” or “recessive”. This paper argues that such a perspective may result in a skewed view of the actual variation, and that terms such as “standard” and “standardisation” may not be very useful when applied to fifteenth-century materials. It presents two case studies of texts that do not fit into a unidirectional view of the standardisation process: a field survey from the East Riding of Yorkshire and a lectionary from the West Midlands, prefixed to a Wycliffite New Testament. These texts show that strongly regional usages could be actively chosen and preferred throughout the fifteenth century, and that modern conceptions of “standard” usages, with their social and educational connotations, may be misleading when applied to late medieval materials.

1 Introduction The concept of “standardisation” has played a central role in the study of linguistic variation in late Middle English. The entire Middle English period (ca. 1100–1500) is traditionally defined in terms of standardisation, beginning with the decline of the Late West Saxon Schriftsprache after the Norman Conquest (a process of “destandardisation”) and ending with “standardisation” in the fifteenth century. The intervening period is characterised by extensive variation in the written mode, which to a considerable extent seems to reflect geography. It is the geographical character of the variation, rather than the variability itself, which is seen as the crucial characteristic of Middle English: “while dialects have been spoken at all periods, it is only during the late medieval times that local usage is regularly reflected in writing” (McIntosh, Samuels and Benskin 1986, I: 3). Consequently, even though written English continued to be highly variable throughout the sixteenth century and well into the seventeenth, “standardisation” is commonly seen to begin in the fifteenth century, as the variation gradually seems to lose its geographical salience at that time. This paper argues that the traditional dichotomy between “standardised” and “local” written usages in the fifteenth century, often with the latter viewed as “conservative” or “uneducated”, may not be a very useful approach. It presents two case studies of fifteenth-century English texts, both of which show an active selection of regionally highly marked usages.

Identity and intelligibility in Late Middle English scribal transmission   

   161

The first text is a Yorkshire field survey, held in the Humberside County Record Office at Beverley (shelf mark DDCC/19/I). According to its incipit, it dates from 1473. It was copied by at least four different scribes, one of whom differs markedly from the others both in linguistic usage and script. The identities of the scribes are not known, nor are the circumstances in which the survey was carried out; however, markings in the manuscript suggest that it, at least for a while, was in the use or possession of the rector of the church in Barmston. The text has been edited and described by Sandvold (2010) and was included in the Linguistic Atlas of Late Mediaeval English (McIntosh, Samuels and Benskin, 1986) as Linguistic Profile 1130.² The second example consists of two manuscripts of the mid-fifteenth century: Longleat, Marquess of Bath’s MS 5 (folios 1–35) and Oxford, St John’s College 6. The former contains a lectionary prefixed to a Wycliffite New Testament, while the other contains Lydgate’s Troy Book; both texts display a related and markedly West-Midland usage. Again, nothing is known about the scribes of these texts; however, the St  John’s College manuscript is connected to the Leinthall family in North Herefordshire. Both texts were included in the Linguistic Atlas of Late Mediaeval English, as Linguistic Profiles 7520 and 7510 respectively; their dialectal characteristics were described in Black (1997: 215–230). These examples show that regional usages were selected by highly competent writers during this period, and may have been expected by the users of particular texts. It is suggested that a more useful way of studying linguistic variation in the fifteenth century might be to avoid the term “standardisation”, with all its ideological and political baggage, altogether, and simply focus on the distribution of linguistic forms.

2 The question of “standardisation” in the fifteenth century During the fifteenth century, there was a general trend of written English becoming less regionally variable. Benskin (1992: 71–72) has described this as follows:

2 A transcription of the entire text, produced by Silje Sandvold, is available in the Middle English Grammar Corpus (MEG-C) version 2011.1. at http://www.uis.no/research/humanities/ the-middle-english-scribal-texts-programme/meg-c/.

162   

   Merja Stenroos

At the close of the fourteenth century, the written language was local or regional dialect as a matter of course (…) by the beginning of the sixteenth century (…) English was still far from uniform (…) but in the written language of ca. 1500, local dialect had become very nearly a thing of the past.

If standardisation is taken to mean simply a reduction in regional variation, what Samuels (1963 [1989]: 75) memorably referred to as a “purging (…) of (…) ‘grosser provincialisms’”, then we can certainly talk about standardisation from an early stage; indeed, a volume edited by Schaefer (2006) titled The beginnings of standardization deals with the fourteenth century. On the other hand, it is highly debatable whether such a broad use of the term is useful in this context. In the same volume, Nevalainen (2006: 117) suggests that the term “supralocalisation” might be more appropriate for the Middle English situation, “especially if standardisation is understood (…) as the selection of a reference dialect to be imposed on the language community from above”. Such a reference dialect certainly did not exist in the fourteenth century. It is still controversial to what extent one should be postulated for the fifteenth. In the last decades, there has been a tradition of identifying such a dialect, usually known as the “Chancery Standard”, following an influential paper by Samuels (1963, reprinted with some revision in Smith 1988; see Benskin 2004 for a critical discussion of the name). Samuels identified four “incipient standards”, three of which (Types 2, 3 and 4) represent “London usage” at different points of time; of these, the administrative usage termed Type 4, or Chancery Standard, becomes the “successful” standard, eventually adopted by writers all over the country. Samuels’ types have often been seen to represent successive local standards, each representing the London usage until superseded by the next type. However, such a model is difficult to reconcile either with the empirical evidence or with current views on linguistic variation and change. As Hanna (2005: 26–28) has shown, there is considerable chronological overlap between texts showing “Type 2” and “Type 3” features (see also Roberts 2011). As with any large centre of population and migration, the capital would have been likely to show complex linguistic variation; while only a small proportion of the population may at this stage have been involved in text production, there is no reason to assume that they would have formed a linguistically homogeneous group. A chronological sequence of “types” may therefore not be the most useful way of viewing the development of the written language in London. The idea of a single-source origin of “Standard English” has also increasingly been thrown into doubt. In a seminal paper, Benskin (1992) showed that “standardisation” was not a process that took place in London, but rather all over the country, and that its study thus needs to be moved away from its London focus. Wright (2000: 6) has stressed the sociolinguistic complexity of the process:

Identity and intelligibility in Late Middle English scribal transmission   

   163

“Standardisation is shown not to be a linear, unidirectional or ‘natural’ development, but a set of processes which occur in a set of social spaces, developing at different rates in different registers in different idiolects” (compare also Hope 2000: 53). The early processes, which took place in the fifteenth century, were not yet producing uniformity, but rather blurring earlier geographical distinctions. This was described in an important paper by Samuels (1981[1988]: 90) as follows: spellings which have hitherto been members of regional systems become like the coins when two currencies are combined; they have the same functional value as before, but they pass from writer to writer, or from writer to printer and back again, and their regional significance is lost.

Samuels made a distinction between “colourless” and “standardised” usage, the latter implying a scribe’s acceptance of forms external to his regional repertoire. Such a distinction would, however, seem to stop making sense at the stage where “currencies are combined” and external forms can no longer be defined in terms of geography. The only criterion for “standardised” as opposed to “colourless” usage would then be one of hindsight: any usage involving forms that came to be part of the later standard would be defined as “standardised”. The term “supralocalization”, introduced by Nevalainen and RaumolinBrunberg (2003: 13, 157) avoids this distinction, and instead focusses on the central development at this stage: the loss of geography as a variable. The term also avoids the ideological baggage connected to the term “standardisation” (see Milroy 2000). From a present-day perspective, a command of “standard language” is easy to equate with education; the use of regionally marked forms in whatever period may then come to mark a lack of education. Such a view would seem to be implicit in the following formulation in a paper by Cuesta and Ledesma, dealing with fifteenth and sixteenth-century northern documents: [W]e have found that texts from about the same date present different percentages of Northern features, which indicates that the literacy of the scribes at that time varied greatly depending on their level of education (Cuesta and Ledesma 2004: 304).

Here, education and “literacy” are equated with “standardisation”. However, there seems to be little evidence that would justify such an equation, at least in a fifteenth-century context. On the contrary, there is ample evidence that highly educated writers could use regionally strongly marked registers; in a well-known study of the Paston Family, Norman Davis (1955: 124–125) showed that men with a university background were using markedly Norfolk forms in the 1460s. Those Paston brothers (John II and John III) who worked in London (as a courtier and

164   

   Merja Stenroos

a servant to a noble family respectively) adopted, unsurprisingly, features of London usage; however, as Davis (1983: 28) noted in a later paper, “a generally observed written standard was still far from attainment in the fairly reputable society represented by these brothers”. Even the inherent prestige of the written usage of London as “a model worthy of imitation” (Davis 1955: 130) should perhaps not be taken for granted for all domains and contexts; as the present paper seeks to show, regionally strongly marked registers continued to be actively selected throughout the fifteenth century by perfectly competent, and presumably educated, scribes.

3 The first case study: the Barmston survey The first example is a field survey or “bounding” of the village of Barmston in the East Riding of Yorkshire. The survey is contained in a booklet of six large folios, Beverley, Humberside County Record Office, DDCC/19/I. It lists the open field strips and “tofts” with their holders, moving clockwise around the village centre; it is organised in six sections, the first five of which deal with the main fields in turn and one that lists the tofts throughout the village, again, it seems, in clockwise order. The village lies on the eroding Yorkshire coast, and parts of the surveyed areas have now disappeared into the North Sea. According to the dating clause, the survey was carried out in 1473. The text was produced by at least four different scribes.³ Three of them (A, B and D) use a northern variety of written Middle English. Their repertoires vary but each one produces several clearly northern forms; these include aboun ‘above’, brade ‘broad’, bryge ‘bridge’, dale ‘portion of a common field’, gyne ‘given’, gyrs ‘grass’, halding ‘holding’, ilkon ‘each one’, kirk ‘church’, os ‘as’, qwylke ‘which’, ryg ‘ridge’, thai ‘they’ and tham ‘them’. All three scribes produce identical letter forms for and , with spellings that may be transcribed ye ‘the’, yis ‘this’. The fourth scribe, Scribe C, produced a fairly short passage, the lower half of fol. 4r. The text consists of fifteen lines and is written in a script that differs considerably from the neat anglicana produced by the others. Sandvold (2010: 55) defines it as a secretary hand with anglicana forms, and describes it as lacking “something of the fluency of the other scribes”. On the verso side, scribe D takes over and copies the same passage before proceeding with the remaining text. The

3 A fifth one may have contributed a few lines on fol. 2r; see Sandvold (2010: 52–53). Scribes A, B, C and D in the present discussion correspond to Sandvold’s scribes 1, 3, 4 and 5.

Identity and intelligibility in Late Middle English scribal transmission   

   165

same passage is thus copied twice. This could not have been an accident, as the texts are on two sides of the same sheet, and both are provided with a clearly visible title, identifying the passage as dealing with the North Field of Barmston. Even though the shared passage is a short one (256 words in Scribe C’s version and 290 in Scribe D’s), there is a clear dialectal difference between the two scribes. Scribe C does not produce any markedly northern forms. Those items for which northern forms are used by the other scribes show dialectally unmarked, nonnorthern forms such as þey ‘they’, holdynge ‘holding’, fre-hold ‘freehold’; the only exception is the single word kyrke ‘church’. Of all four scribes, C is the only one to distinguish between and . In contrast, Scribe D has markedly northern forms, including thai ‘they’, haldyng ‘holding’ and frehald ‘freehold’; he does not distinguish between and . The rewriting suggests that scribe C’s work was for some reason not considered acceptable. The lack of northern spellings is unlikely to have been a problem in itself, as scribe C’s language seems to be truly “supralocal”, with no regional peculiarities of any kind. It may be noted that, considering the highly formulaic and monotonous character of the text, the two passages differ a good deal in phrasing, and seem to be fairly free renderings of the same content, perhaps made from compressed notes.⁴ Scribe D’s version is longer and seems to aim at more precision; this is particularly notable in the headings: C. The northe Felde of Barnston ‘the North Field of Barmston’ D. Memorandum that heer begynnes the bowndes of ye North Feild of barnston to begyn at the Estermar ley Close of ye west syde of the forsayd north Feyld of the said Town ‘Memorandum, that here begin the bounds of the North Field of Barmston, starting with the easternmost lea close of the western part of the foresaid North Field of the said town’ Scribe D makes numerous small adjustments that seem to add to the precision of the text; thus, where scribe C has a brode-land of croses next þe meir ‘a broadland of Cross next to the boundary’, D writes a brode-land next the meer of Crosse haldyng, making the meaning (that the strip of land belonged to the holding of Cross) unambiguous. Similarly, scribe C’s longynge to kellynge ‘belonging to Killing’ is written out more fully by D: longyng to ye hows of kelyng ‘belonging

4 For a full comparison of the two passages, see Sandvold (2010: 83–84, 100–101).

166   

   Merja Stenroos

to the house of Killing’. As Sandvold (2010: 85) shows, the C text also seems to contain a misunderstanding: C. Jtem ij gaires lyes fro Wynkton’ mer’ to þe hede grippes logynge to þe maner D. Jtem ij gyrs leys fro Wynktoun mer’ to the | hede gryppes longyng’ to ye maner ‘Item, two grass leas from Wynkton Mere to the main dikes, belonging to the manor’ The reading gaires lyes is cryptic on its own, while gyrs leys makes good sense as a northern form of ‘grass leas’ or pastures. It is possible that scribe C did not understand the northern form gyrs for ‘grass’, if he confronted that in an exemplar or in dictation. On the whole, his text comes across as rather less detailed and accurate than that of the D scribe, and at least part of this seems to have to do with dialectal competence. As the document deals with local geography, and with something as important as field rights, knowledge of the dialect would have been an important asset; apart from dialectal words for features of the landscape, such as carr ‘bog’, scarth ‘?rock, cliff’ and slak ‘valley’, the scribe should preferably also master the system of land units and measures, which varied considerably between different parts of the country. In addition, a confident knowledge of the local dialect, and its written expression, would contribute to the precision and unambiguity of this type of text. This would seem to be in complete contrast to the common conception that the uniformity imposed by standardisation was especially important in legal documents, as it would lead to less ambiguity. However, there is probably no contradiction here: rather, “legal documents” is a vast category, containing texts written for very different purposes and for the benefit of different users (compare Rissanen 2000: 119). They were also produced by scribes with different backgrounds: some trained at Westminster, others in local writing or business schools; still others were more or less self-taught. The text produced by scribes A, B and D of the Barmston survey certainly does not represent the uneducated scribblings that might be associated with the term “non-standard”, but rather a register that is acceptable and expected. As for scribe C, his usage may be termed “supralocal”, as it does not show any northern, or otherwise localisable, features. However, there is no sense in which it may be considered to represent an institutionalised, accepted usage (which “standardised” would imply) in this context; on the contrary, it seems as though the scribe’s non-northern linguistic repertoire might even have rendered him unsuitable for the copying of the document.

Identity and intelligibility in Late Middle English scribal transmission   

   167

4 The second case study: a Marcher orthography? 4.1 The lectionary in Longleat, Marquess of Bath’s MS 5 The second example involves a group of fifteenth-century texts that seem to represent a specifically West Midland writing tradition. The first text is contained in the Marquess of Baths’ MS 5 at Longleat Hall.⁵ It is a lectionary, consisting of a list of the New Testament passages used in the church liturgy, as well as the actual readings for the Old Testament passages. The text is bound together with the later Wycliffite translation of the New Testament, which also contains some glosses to Matthew and Luke. Lectionaries seem to have been included with the Wycliffite Bible from the beginning, and they are present in almost a hundred of the ca. 250 surviving manuscripts (Peikola 2005–2006: 343, 347). Approximately thirty of the lectionaries are accompanied by the full text of the Old Testament readings. This particular text belongs to a small group of four extant manuscripts that contain such fulltext readings added directly into the list of New Testament incipits and explicits, rather than as a separate section.⁶ The lectionary is contained in five quires, consisting of 36 folios in total; the text is incomplete and the last page contains a catchprase suggesting that at least one quire has been lost. The lectionary may have been produced separately from the particular Bible text to which it is attached; however, it was clearly designed for use with a New Testament. The entries for New Testament readings follow the usual design described by Peikola (2005–2006: 350), indicating the liturgical occasion (in the left margin), the biblical book and chapter, a referential letter, and the incipits and explicits of the reading (separated by the word ‘ende’). However, they are laid out as a list rather than in table format; this choice of layout would have been necessary in order to incorporate the Old Testament readings, which are written out as single columns of prose. As with most Wycliffite manuscripts, the Longleat manuscript is not dated. The hands are not very helpful in this regard; they represent the kind of plain, working textura hand that is found in biblical manuscripts throughout the Wycliffite period. The New Testament copy to which the lectionary is attached was dated to ca. 1430 by Madden in his private notes (Hargreaves 1961: 194); however, the lectionary may not be of the same date.

5 This manuscript is discussed by Hargreaves (1961: 294–298) 6 See Dove (2006: 61).

168   

   Merja Stenroos

According to Peikola (2005–2006: 347), most of the lectionaries were produced in the first three decades of the fifteenth century; Dove (2007: 58–59) notes that “one or two” are from the late fourteenth. A comparison of the present text with Peikola’s (2005–2006, 2011) studies suggests neither an early nor a very late textual stage.⁷ Peikola associates the “list” format with an early stage; however, the format is here dictated by the choice of including the full Old Testament texts. The Longleat text gives, overall, a somewhat less polished and professional impression than most of the London-produced lectionaries, with uneven lines and right margins. At the same time, the manuscript is carefully decorated with ornamented initials and decorative strokes along the left-margin ruling. The most interesting aspect of the text is, however, its language: the lectionary is written in a markedly western usage, quite unlike the language of the New Testament and of Wycliffite writings in general. The Wycliffite movement produced vast amounts of texts, most notably the Bible translation but also a large number of sermons and treatises. The majority of these are written in a remarkably homogeneous variety of Middle English, known as “Type 1” or the “Central Midland Standard”, following Samuels (1963). The latter term refers to the dialectal characteristics of the texts, rather than to any historical evidence of provenance. Both the Bible translation and the sermon cycle originated in a centre of learning with a large library, almost certainly Oxford (compare Hudson 1988: 109; 2008: 335); however, much of the production of Bible manuscripts is likely to have taken place in London, and Hudson (2008: 336) suggests that “it may be that much more improbable and remote places harboured copyists”. As there seem to be no local documents in this variety of Middle English, and manuscripts containing it were certainly produced in different parts of the country, “Type 1” may be taken to represent the written register of a text community without a single geographical focus. The New Testament in the Longleat manuscript is written in this variety, which may be illustrated as follows: Augustin in þe iiij book of þe trinite. temptacioun is don in þre manerys, bi suggestioun eþer tisinge, be deliting and consent; and whan we ben temptid we fallin sum tyme, eþer bi liking eþer bi consent; but God, which made man in þe wombe of þe virgyn wiþowt syn in þe worlde, suffrid noon aȝenseying in himsilf (Longleat 5, f. 38v, the Gloss to Matthew).

7 For example, the incipits and explicits are of medium length compared to Peikola’s examples (2011: 117, 119). A more detailed comparison of the Longleat text (which was not available to Peikola) to the other versions would be of interest.

Identity and intelligibility in Late Middle English scribal transmission   

   169

‘Augustin in the fourth book of the trinity. Temptation is done in three ways, by encitement or enticing, by pleasure and consent; and when we are tempted we sometimes fall, either because of pleasure or by consent; but God, who (was) made man in the womb of the virgin without any worldly sin, allowed no contradiction within himself’.

The lectionary (henceforth L), however, is written by a different scribe in a completely different dialect, very clearly of a West Midland character. It was included in the Linguistic Atlas of Late Mediaeval English (McIntosh, Samuels & Benskin 1986) as Linguistic Profile 7520, and it was localised in the far North of Herefordshire, near the Shropshire border. In Black (1997: 223), the following written forms, all of which occur regularly in the manuscript, were identified as suggestive of this localisation: uche ‘each’, whuche ‘which’, muche meche moche ‘much’, lesse ‘less’, þorgh ‘through’, bote ‘but’, þagh ‘though’, ȝit ȝut ‘yet’, aȝayn(-) ‘again(st)’, sulf ‘self’, þanne ‘then’, whanne ‘when’, dyede ‘died’, dredde ‘dreaded’ (sg), fuyr ‘fire’, luytul ‘little’, kuynde ‘kind’, kun ‘kin’, furste ‘first’, dud- ‘did’, stud- stid- ‘stead’, beo ‘be’, heold ‘held’ (sg), þeose ‘these’, scheo ‘she’, schuld(-) ‘should’, han ‘have’ The most distinctive feature of the scribal usage of L is the frequent retention of and spellings for the traditional West-Midland front rounded vowels. Apart from þeose and scheo in the above list, the spellings occur in words such as beo ‘be’, seo ‘see’, þeof ‘thief’, cheose ‘choose’, heold ‘held’ and leornynge ‘learning’. The spellings and distinguish fairly regularly between the reflexes of Old English long and short y: thus fuyr, luytul, kuynde but kun, furste, dud, stud. The appearance of these spellings is remarkable, considering that most of them are exceptional even in West Midland texts after the mid-fourteenth century: only the spelling heo ‘she’ survives as a regularly used form into the fifteenth century, presumably because of its functionality in making an orthographic distinction between he ‘he’ and heo ‘she’ (see Black 1999: 72, 76). That the spelling does not reflect any local survival of unmerged /ø(:)/, going back to the Old English sound spelled , is suggested by the regular unhistorical spelling þeo of the definite article. At the same time, with the exception of þeo, unhistorical use of the , and spellings does not seem to occur in the text. This suggests that the text belongs to a continuous tradition of Western writing, which retains orthographic features that were characteristic of this area in the thirteenth and early fourteenth centuries. In a fifteenth-century context,the text has a decidedly archaic look, made even more exotic by the frequent þeo spellings:

170   

   Merja Stenroos

In þo days sieknede ezechye tyl to þeo deþ / & jsay þeo profyt þeo sone of amos entrede to hym / & seyde to hym / þeo lord sayþ þeose þynges / dispose þyn hous for þow schalt dye & þu schalt not lyue / & ezechye turnde hys face to þeo wal : & prayede þeo lord & seyde / lord y byseche haue þu mynde ; j byseche . how j ȝeode byfor’ þe in truwþe / & in parfyt herte / & j dude þat þat was good by-fore þyn yen . Ezechye wepte wt gret wepynge ; & þeo word of þeo lord was mad to ysaye & seyde / go þu & say to ezechye / þeo lord god of dauid þy fader sayþ þes þynges / j haue j-herd þy preyer ; & j say þy teeres / & lo j schal adde on þy dayes fiftene ȝier (Longleat 5, fol. 6r–6v). ‘In those days Ezechiel fell mortally ill and the prophet Isaiah, son of Amos, came to him and said to him: the Lord says these things: dispose of your house because you will die and not live. And Ezechiel turned his face to the wall and prayed to the Lord and said: “Lord, I beseech you to remember how I went before you in truth and in perfect heart and I did what was good before your eyes”. Ezechiel wept greatly, and the word of the Lord came to Isaiah and said: go and say to Ezechiel: “The Lord, God of your father David, says these things: I have heard your prayer and I saw your tears, and Lo, I shall add fifteen years to your days”’.

The text also contains some morphological features that would have been obsolete in many varieties of Middle English by the mid-fifteenth century, but that seem to have been retained longest in the southern and western margins (as well as, to some extent, Norfolk). The plural forms of modals and of strong verbs in the preterite are retained almost without exception: schullen ‘shall’, mowen ‘may’, risen ‘rose’, drunken ‘drank’. The prefix j- (Old English ge-) is retained in ca. 7 % of otherwise prefixless past participles, including j-herd ‘heard’, jrise ‘risen’, jwryten ‘written’. The personal pronouns show the system current in most fifteenth-century texts from the southern half of England: the third-person plural paradigm is þay/þey – her – hem, and the third-person feminine singular shows the “she” type; however, the latter is most commonly spelt scheo. At the same time, the Longleat text shows a few regularly occurring forms that cannot be reconciled with a Western localisation within the Atlas framework; these are ony ‘any’, eyþur eþur ‘or’ and ȝoue(n) ‘given’. All these forms belong to the list of typical “Central Midland Standard” forms as defined by Samuels (1963 [1989]: 67; 1969 [1989]: 141). Other forms in the text, all occurring as minor variants, also seem to indicate a Central or East Midland element: ich ‘each’, siche ‘such’, neyþer neyþur neþur ‘nor’, silf sylf ‘self’, stid- styd- ‘stead’, þerk ‘dark’, goith ‘goes’ All except þerk and goith belong to the “Central Midland Standard” list; the form þerk is one of the distinctive forms of Samuels’ “Type 2”, a variety (or an assemblage of forms) typical of fourteenth-century London texts (Samuels 1963[1989]: 70, 79 n.7; see also Hanna 2005: 5–7), and of East Anglian usage. These minor

Identity and intelligibility in Late Middle English scribal transmission   

   171

variants suggest strongly that the text was copied by a West Midland scribe from an exemplar in a non-Western dialect; the forms would agree very well with the London-based production of copies in varieties of the “Central Midland Standard”. The opposite scenario – that the text was copied by a London scribe from a West Midland exemplar – would be possible in theory but unlikely (see discussion in Section 4.4.).

4.2 Oxford, St John’s College MS 6: Lydgate’s Troy Book The Atlas localisation of the Longleat lectionary would have been based on the similarity of its language to that of other texts associated with this area, both earlier and approximately contemporary. In particular, the text shows remarkable similarities with another text included in the Atlas material: a copy of Lydgate’s Troy Book surviving in Oxford, St John’s College MS 6. The St John’s College manuscript (henceforth referred to as J) was dated to 1450–75 in the Manual of Middle English Writings (Renoir and Benson 1980: 2168) and to the mid-fifteenth century by Hanna (2002: 7). The manuscript is of 134 folios and is written by a single scribe in a compact Anglicana Formata hand, in two columns. The large initial on the first folio contains a coat of arms and the text “Jaffie bien Leinthale”; Hanna (2002: 8) associates these with the arms of Sir Rowland Leynthall of Leinthall and Hampton Court in North Herefordshire, who fought at Agincourt. The manuscript thus seems to be connected to the Leinthall family of Herefordshire, and was probably produced for a member of the family. The scribal usages contained in J and L are extremely similar, although not identical. The shared usage includes the following regularly occurring forms:⁸ suche ‘such’, whuche ‘which’, muche ‘much’, many mony manye monye ‘many’, þilke þylke etc ‘the same’, lesse ‘less’, fro from ‘from’, þorgh ‘through’, bote ‘but’, ȝyf ‘if’, þan ‘than’, ny ‘nor’, aȝayn/- ‘again(st)’, þanne ‘then’, whom wham ‘whom’, whanne whan ‘when’, yen ‘eyes’, morwe ‘morning’, syluer seluer ‘silver’, fleschss ‘flesh’, hye hiegh ‘high’, oune ‘own’, trewe truw- ‘true’, cam ‘came’, ȝoue(n ‘given’, seyde ‘said’, pruyde ‘pride’, fuyr ‘fire’, luytel ‘little’, kyn kun ‘kin’, furst(e ‘first’, yuel ‘evil’, worch- ‘work’, worschup ‘worship’, swerd ‘sword’, sche scheo heo ‘she’, hit it ‘it’, present plural indicative -en, past participle prefix J-, wol ‘will’, woot ‘wot’

8 The comparison between the two usages is to a large extent based on a study carried out by the present author in her PhD thesis (Black 1997: 227–230), using a 300-item dialect questionnaire.

172   

   Merja Stenroos

Where the two texts differ, they mainly show the same variants in different proportions, J tending to have more “colourless” or supralocal forms as majority variants (see Table 1).

Item

L

J

EACH THOUGH OR NOR YET SELF SAW MAN THANK FATHER DID BE THESE US THEY SHALL pl MAY pl

uche ((eche)) þagh 5 þogh 1 þaugȝ 1 þouȝ 1 or 8 oþur 5 eyþer 5 eþur 4 neþur (ny) (((neo))) ȝit 4 ȝut 3 ȝyt 3 ȝitte 1 sulf (self) ((silf sylf)) say ((siegh seygh sieȝ)) Man þank- 2 fader fadur dud- ((did- dyd-)) beo be þeose þese ous ((us)) þay ((þey)) schullen ((schul)) mowen 2 mown 1 moun 1 may 1

eche þogh ((al-be)) (((þagh))) or nor ny (ne) ((neyther)) ȝut(te; ȝet(te self ((sylf)) saw(e 4 ((saugh sey)) man ((mon)) þank- 4 þonk- 3 fader dyd ((dude dede)) be þese (these) vs (((ous))) þey (þay) ((they)) schal (((schullen))) may

Table 1: Differences in linguistic usage between L (Longleat 5, lectionary) and J (St John’s College 6, Lydgate’s Troy Book).

The use of spellings that is so striking in L is restricted to scheo, heo ‘she’ in J. However, and spellings occur commonly, although their proportions vary greatly lexically: apart from the words cited above, they appear e.g. in burthe ‘birth’, buyldyng ‘build’, pruyde ‘pride’, schut ‘shut’ and stude ‘stead, place’; as in L, these spellings distinguish quite regularly between OE long and short [y]. While the usage is, on the whole, less strongly dialectal, many of the regular Western forms in L occur as minor forms in J: so þagh ‘though’, ouþer ‘other’, dude ‘did’, schullen ‘shall (pl.)’. J shows only occasional examples of the retention of plural forms in the modals and strong form preterites, but has more or less the same proportion of the j-prefix in past participles (7 %) as L has. Like L, the J usage also contains a number of forms that do not fit into a Western localisation within the LALME framework. Just like L, J shows the form (J)- ȝoue(n) ‘given’ as a majority form, while others occur as minor variants:

Identity and intelligibility in Late Middle English scribal transmission   

   173

ychon ‘each’, swyche ‘such’, ony ‘any’, al-be ‘though’, neyther ‘nor’, i-fere ‘together’, sylf ‘self’, eyes ‘eyes’, work ‘work’ (vb), schat ‘shalt’ While most of these forms would have been common in fifteenth-century London, a localisation of this specific assemblage in the LALME framework would point to a small area including southwest Norfolk and southern Cambridgeshire, an area consistent with Lydgate’s geographical background (compare Black 1997: 222). Purely orthographical differences between L and J are few, and all involve relative frequency rather than absolute selection. Thus the L text uses the letters and more frequently than J does, and unstressed vowels are usually spelled in J but in L. Most interestingly, however, the two texts seem to share several distinctive features of orthography, none of which are common in late Middle English; these include the following: – A tendency to double vowel symbols, e.g. maad ‘made’, chees ‘chose’, goold ‘gold’, hoote ‘hot’ and rood ‘rode’ (J); maad ‘made’, chees ‘chose’, heere ‘hear’, hoolde ‘hold’ and roos ‘rose’ (L). The doubling is occasionally extended to short vowels, as in saat ‘sat’ (J), and even unstressed ones, as in fredoom ‘freedom’ (L). – The use of the digraph for /e:/, e.g. fiet ‘feet’, ȝier ‘year’, hier ‘here’, tieth ‘teeth’ (J); chiere ‘cheer’, fiede ‘feed’, fiet ‘feet’, ȝier ‘year’, hiere ‘hear’, sied ‘seed’, schiep ‘sheep’, tieþ ‘teeth’ (L). – Doubling of , e.g. powwer ‘power’, sywwe ‘sue’, towward ‘toward’, trowwe ‘think’ (J); bowwyd ‘bowed’ (L). – Doubling of , and , e.g. cachche ‘catch’, fechche ‘fetch’, gruchchynge ‘grudging’, rechchelesched ‘carelessness’, wrechchede ‘wretched’; neghgheþ ‘approaches’; syththe seththe ‘since’ (J); gruchchynge ‘grudging’, strechcheþ ‘stretches’ (L). All except the first feature may be characterised as rare. The doubling of vowel symbols is not uncommon in Wycliffite texts, but it was considered untypical of the West Midlands (“Herefords or the surrounding counties”) by Samuels (1969 [1989]: 139). The doubling of , , and , on the other hand, is restricted to a very small group in the LALME material; apart from L and J, this includes the collection of romances and other longer texts in Lincoln’s Inn Hale 150, from the first part or middle of the fifteenth century, and the fourteenth-century miscellany in BL Harley 2253. All four texts are localised in LALME within a small area covering Northwest Herefordshire and Southwest Shropshire. Finally, the use of for /e:/ is extremely rare, but seems to occur occasionally in West

174   

   Merja Stenroos

Midland texts: the text of the C version of Piers Plowman contained in BL Add. 34779, localised in LALME in Shropshire as Linguistic Profile 4218, has hier ‘here’.⁹ The scribal usages of L and J are clearly closely related. They share a large number of distinctive forms, both as majority variants and as regulary occurring minority forms. While the proportions of variants differ in some cases, the cooccurrence of a remarkable number of uncommon spellings such as aȝayn(-), fleschss, hiegh, as well as the spellings and double consonants, suggest either a shared, distinctive set of spelling conventions or the usage of a single scribe. L and J could represent the usage of a single scribe, faced with different exemplars or copying for different purposes. As the scripts are different, it is impossible to be certain whether or not they might in fact have been copied by a single scribe; both hands share a certain unevenness, but the hand of J is much more vertically condensed than the L hand (this may, of course, have to do with copying space and layout). The similarities may reflect textual history: J may well have been copied from an exemplar written by L or vice versa. However, they could also reflect a shared writing tradition; this possibility is made particularly interesting in view of the other surviving texts that show a related usage.

4.3 The dialectal context of L and J Some of the distinctive orthographic characteristics of L and J, as well as the geographical dialect markers present in them, are shared by other texts localised in the same area in LALME, most notably Lincoln’s Inn Hale 150 and Harley 2253. Lincoln’s Inn 150 (henceforth I) was localised in the extreme southwest of Shropshire (the Clun area) in LALME, as LP 4037. It has manuscript connections with the Fitzalan family, whose Shropshire holdings were in Clun and Oswestry (Barnicle 1927: x–xiv). It is dated slightly earlier than J: according to Hanna (2005: 16) it was produced “shortly after 1400”, while Horobin and Wiggins (2008) place it in the first quarter of fifteenth century and Macrae-Gibson (1979) at “ca. 1450”. It contains five works: Libaeus Desconus, Arthur and Merlin, Kyng Alisaunder, The Sege of Troye and the A-version of Piers Plowman. It is a long and thin manuscript, written by one scribe in single columns. It used to be thought of as a “minstrel manuscript”, used by an ambulant teller of tales; this assumption is, however, now discredited, and it is considered more likely that the book was produced

9 These spellings would presumably reflect the operation of the Great Vowel Shift, shown by other studies to have been in progress in the West Midlands by the late Middle English period (Black 1998; Steinbrenden 2011).

Identity and intelligibility in Late Middle English scribal transmission   

   175

for, and used within, an aristocratic household (for references, see Horobin and Wiggins 2008). Hanna (2005: 16) has described the contents of the Lincoln’s Inn manuscript as “with one exception (…) composed of well-known London-circulating texts”: The exemplars underlying Hale 150 presumably ended up where they did as the result of forays in and out of the capital by a servant of prominent local lords as he went about the business of his masters, the Fitzalan earls of Arundel, in their guise as barons of Clun and Oswestry.

The linguistic usage of I has many similarities with L and J; most interestingly, it also shows a retention of and spellings, with the same unhistorical extension of to the definite article; in I, the unhistorical appears also in neo ‘not’/ ‘nor’. At the same time, there are many differences of detail; this, and the fluent Anglicana script produced by the I scribe, clearly different from the others, suggests strongly that we are dealing with more than just one idiosyncratic Western scribe. The Harley 2253 manuscript is much earlier than the others, from the first half of the fourteenth century. While its dialect was localised by Samuels (1983[1989]: 260–261) in the Leominster area, its scribe is known to have worked in Ludlow (Revard 1979: 2000). It consists of a large miscellany of verse and prose texts, both religious and secular, in English, French and Latin, and it is likely to have been produced for a wealthy household in the West Midlands, possibly an episcopal one (Boffey and Edwards 2008: 386–387). Apart from the use of doubled consonants and the and spellings (which are, of course, less remarkable in a Western text of this earlier date), it agrees with L (and most often J) in a large number of distinctive forms (see Black 1997: 230, 271). These four manuscripts seem to form a geographically focussed group. They are dialectally closely related, and three of them have known external connections with a very limited area in the Marcher region of the West Midlands. In addition, they share some unusual orthographic features which may point to a shared scribal tradition. Most of the contents of the fifteenth-century manuscripts consist of texts that were commonly reproduced in London, and their linguistic usage suggests that they have been copied from exemplars in an eastern or London dialect. This means that the scribes were actively translating the texts into a West Midland usage. While the dates of all three texts are uncertain in detail, they are all placed in the fifteenth century, and at least one of them (J) is likely to date from the middle, or even the third quarter, of the century. By this time, we might expect the copying of London exemplars to result in the spread of London language; this is, however, far from the case with the present texts.

176   

   Merja Stenroos

4.4 A Marcher variety? In the first half and middle of the fifteenth century, regional variation in written usage was still common, even though a process of supralocalisation was under way in the southern part of the country (McIntosh, Samuels and Benskin 1986, I: 3). It was shown in Black (1999: 75) with regard to the texts localised in Herefordshire in LALME, that regional “colouring” in this period could be to a large extent related to textual history: texts originating elsewhere would show less dialectally marked usage than texts with “local” origins. The interesting question here is why several texts that clearly go back to exemplars in other dialects appear in a markedly Western language. In this period, the mobility of scribes and texts was making written English less local; varieties such as the “Central Midland Standard” seem to have been well-known and understood everywhere, and were routinely reproduced in Bible manuscripts. Accordingly, it may be asked why several scribes would choose to translate sizeable texts into a markedly regional, even local, language. One possibility is that the use of local forms was intended to make reading aloud easier. All three fifteenth-century manuscripts contain texts that were almost certainly habitually read aloud: romances, secular and religious verse and liturgical readings. While opinions are divided on whether the English lectionaries were used in church, and whether they might even have been read aloud at church services (Deanesly 1920: 176; Hudson 1988: 199; De Hamel 2001: 182; Peikola 2005–2006: 365–367), they would at the very least, like any Wycliffite readings, have been read aloud in a home setting. Such an explanation is possible, but cannot be more than partial. Firstly, it would not explain the presence of purely orthographic shared features, such as the consonant spellings in fleschss or wrechchede. Secondly, even if rounded vowels in words like fuyr, dude, beo might have survived in the speech of a small area in the western borderlands, the extensions of the spelling to scheo ‘she’, þeo ‘the’ and neo ‘not’/‘nor’ suggest that the convention was an orthographic one (even though the last two forms might possibly reflect a centralised, even rounded, unstressed vowel). The use of and spellings was a highly salient marker of Western texts in the thirteenth and early fourteenth centuries. Forms such as uche, whuche, mony and þa(g)h, shared by all the four texts, were also highly traditional in the north Herefordshire – south Shropshire area, even if they were by no means limited to it. Black (1997: 271, 377–378) traced a continuous tradition of dialectal writing in this area, linguistically quite different from that of the other parts of Herefordshire, going back at least as far as the written language of Ancrene Wisse (early thirteenth century). This tradition was tentatively related to the powerful

Identity and intelligibility in Late Middle English scribal transmission   

   177

Marcher lordship of Montgomery, held by the Mortimers, as Earls of March, from the thirteenth century until 1425, and in the mid-fifteenth century by Richard of York, father of Edward IV and Richard III. The context of a Marcher lordship could account for the active preservation of a local writing tradition. The Marcher area had its own system of jurisdiction and was ruled by very powerful noblemen, many with high ambitions; it was not until 1536 that the Marcher area was made fully a part of England in legal terms. At the same time, there would have been a continuous traffic between London and the Marches, as both the lords themselves and their servants would travel between their main provincial strongholds and the capital, and it has been shown that this included an important book trade (see e.g. Doyle 1983). Two of the fifteenth-century texts discussed here are directly connected with the great aristocratic families in and near the Marcher area. The St John’s College manuscript contains the Leinthall family arms, while the Lincoln’s Inn manuscript is connected with the Fitzalans. The Longleat manuscript does not show any such direct connections; however, it would fit well into the context of an aristocratic household in this area. As De Hamel (2001: 180–181, 183) points out, from the early fifteenth century, when the lectionaries start appearing in large numbers, the Lollard Bibles seem to become increasingly distanced from the heretic movement and find their ways into numerous households, including aristocratic ones, for devotional use. While such households might have been perfectly orthodox, it may be noted that the Marcher lords had shown interest in the Lollards, even supported them, in the earlier period before persecution began. The households of the Marcher lords would have been in the position to support a local book production; it would perhaps not be surprising if local conventions, even “house styles”, developed, based on the written tradition. Some of the conventions, such as the retention of and its transfer to new contexts, may seem both “unnatural” in relation to phonology and dysfunctional in terms of national usage. However, this is also true of many of the spellings that became part of standard written English; these may have been selected as high-status forms partly because of their difficulty, even though they also came to have a useful function in distinguishing between homonyms ( vs vs ). In the same way, it is conceivable that Western features, such as the spellings, had an identity-reinforcing function in the community for which they were produced (and also distinguished between some homonyms). As far as the lectionary is concerned, one might envisage a more specific reason for producing a usage that was not only local but also, to some extent, archaising. The De heretico comburendo act of 1409 made it illegal to own English Bible copies produced since the time of Wycliffe. At least in some documented cases, attempts were made to forge early Bibles (see e.g. De Hamel 2001: 177; Dove

178   

   Merja Stenroos

2007: 39); it has even been suggested that the unchanging textura hand characteristic of so many Wycliffite Bibles was a conscious attempt to produce archaiclooking, undatable texts. In an area where old writing traditions were well known and at least to some extent maintained, it could have made sense to produce a lectionary, intended to be prefixed to a bought copy of the New Testament, in a writing system made to look as archaic as possible. This would not, however, explain the very similar usage of the Lincoln’s Inn text, which also uses “pseudo-archaic” features such as þeo; it is also unlikely that a simple attempt at archaising for a specific purpose could produce a usage that was, on the whole, as historically consistent as that found in the Longleat lectionary. It seems more likely that the spellings represent some kind of conscious upholding of a local identity: a text community with long traditions, connected to the households of noblemen almost as powerful as the king.

5 Conclusion It was noted in Section 2 that recent research has cast doubtboth on the idea of a single source for Standard English and on the existence of a coherent “standard” or model variety in the fifteenth century. The aim here has been to show that, whether or not such a variety existed, it was very far from becoming a default norm for “educated” writers or important texts. The examples of geographically marked usage shown here are by no means exceptional, and it is well known that strongly “dialectal” writing was still produced much later (see e.g. Benskin 1992: 93 and Britton 2000). However, the present examples show that geographically marked usage was not only produced in the fifteenth century but that texts already available in a supralocal usage, or a London/East Midland dialect, were actively translated into a geographically strongly marked written usage. The examples discussed here are clearly not the work of scribes who produced local dialect because they did not know better. Scribe D of the Barmston document, who reproduced the passage already written by scribe C, turning it into northern usage, is shown to be the more competent legal scribe in all details, including the script and the phrasing of headings. The scribes of the Marcher texts were translating London texts to produce copies with a local identity, using a set of features from an ancient orthographic tradition to achieve this. In both cases, we have highly competent scribes, who select geographically marked registers although supralocal ones are available, and even produce (or reproduce) geographically marked orthographic innovations, such as the use of þeo.

Identity and intelligibility in Late Middle English scribal transmission   

   179

It is not suggested here that the choices necessarily reflected the scribes’ own preferences. The expectations of the intended readers or users will have been crucial, and whoever commissioned the texts may have played an important role. All the texts were clearly intended for use: as with present-day spoken language, their registers were part of living variation, selected for the marking of identity or for the sake of intelligibility. Our understanding of the text communities in which they were produced can only be fragmentary, and very much work remains to be done on the language, manuscript context and historical background of these and other texts. It is, however, crucial that the choices made by the scribes should be studied on their own terms, and without recourse to hindsight, rather than as a marginal part of the history of standardisation.

6 Sources 6.1 Primary sources Beverley, Humberside County Record Office, DDCC/19/I London, Lincoln’s Inn 150 London, Harley 2253 Longleat, Marquess of Bath’s MSS 5, folios 1–35 Oxford, St John’s College 6

6.2 Electronic resource Stenroos, Merja, Martti Mäkinen, Simon Horobin and Jeremy Smith (compilers), The Middle English Grammar Corpus (MEG-C), version 2011.1., University of Stavanger, http://www.uis. no/research/humanities/the-middle-english-scribal-texts-programme/meg-c/.

7 References Barnicle, Mary Elizabeth (ed.). 1927. The Seege or Batayle of Troye: A Middle English Metrical Romance. London: Oxford University Press. Benskin, Michael. 1992. Some new perspectives on the origins of standard written English. In J.A. van Leuvensteijn & J.B. Berns (eds), Dialect and Standard Language in the English, Dutch, German and Norwegian Language Areas, 71–105. Amsterdam: Koninklijke Nederlandse Akademie van Wetenschappen.

180   

   Merja Stenroos

Benskin, Michael. 2004. Chancery Standard. In C. Kay, C. Hough and I. Wotherspoon (eds), New Perspectives on English Historical Linguistics: Selected Papers from 12 ICEHL, 1–40. Amsterdam: John Benjamins. Black, Merja. 1997. Studies in the dialect materials of medieval Herefordshire. Glasgow: University of Glasgow PhD thesis. Black, Merja. 1998. Lollardy, language contact and the Great Vowel Shift: spellings in the defence papers of William Swinderby. Neuphilologische Mitteilungen 99. 53–69. Black, Merja. 1999. Parallel lines through time: speech, writing and the confusing case of she. Leeds Studies of English 30. 59–81. Britton, Derek. 2000. Henry Machyn, Axel Wijk and the case of the wrong Riding: the south-west Yorkshire character of the language of Henry Machyn’s diary. Neuphilologische Mitteilungen 101. 571–596. Cuesta, Julia Fernandez & Maria Nieves Rodriguez Ledesma, 2004. Northern features in 15th–16th century legal documents from Yorkshire. In Marina Dossena & Roger Lass (eds), Methods and data in English historical dialectology, 287–308. Bern: Peter Lang. Davis, Norman. 1955. The language of the Pastons Proceedings of the British Academy 40. 119–144. Davis, Norman. 1983. The language of two brothers in the fifteenth century. In E.G. Stanley & D. Gray (eds), Five Hundred Years of Words and Sounds: A Festschrift for Eric Dobson, 23–28. Cambridge: D.S. Brewer. De Hamel, Christopher. 2001. The Book: A History of the Bible. London: Phaidon. Dove, Mary. 2007. The First English Bible. Cambridge: Cambridge University Press. Doyle, A.I. 1983. English books in and out of court from Edward III to Henry VII. In V.J. Scattergood & J.W. Sherborne (eds), English Court Culture in the Later Middle Ages, 163–181. London: Palgrave Macmillan. Hanna, R. 2002. A Descriptive Catalogue of the Western Medieval Manuscripts of St. John’s College, Oxford. Oxford: Oxford University Press. Hanna, Ralph. 2005. London Literature, 1300–1380. Cambridge: Cambridge University Press. Hargreaves, Henry. 1961. The marginal glosses to the Wycliffite New Testament. Studia Neophilologica 33. 285–300. Hope, Jonathan. 2000. Rats, bats, sparrows and dogs: linguistics and the nature of Standard English. In Laura Wright (ed.), The Development of Standard English 1300–1800: Theories, Descriptions, Conflicts, 49–56. Cambridge: Cambridge University Press. Horobin, Simon & Alison Wiggins. 2008. Reconsidering Lincoln’s Inn MS 150. Medium Ævum 77. 30–53. Hudson, Anne. 1988. The Premature Reformation: Wycliffite Texts and Lollard History. Oxford: Clarendon Press. Hudson, Anne. 2008. Lollard literature. In N.J. Morgan and R.M. Thomson (eds), The Book in Britain. Vol. II: 1100–1400, 329–339. Cambridge: Cambridge University Press. McIntosh, Angus, Michael L. Samuels & Michael Benskin. 1986. A Linguistic Atlas of Late Mediaeval English. 4 vols. Aberdeen: Aberdeen University Press. Macrae-Gibson, O. D (ed.). 1973, 1979. Of Arthour and of Merlin. 2 vols. EETS o.s. 268, 279. Oxford: Oxford University Press. Milroy, James. 2000. Historical description and the ideology of the standard language. In Laura Wright (ed.), The Development of Standard English 1300–1800: Theories, Descriptions, Conflicts, 11–28. Cambridge: Cambridge University Press.

Identity and intelligibility in Late Middle English scribal transmission   

   181

Nevalainen, Terttu. 2006. Fourteenth-century English in a diachronic perspective. In Ursula Schaefer (ed.), The Beginnings of Standardization, 117–132. Frankfurt am Main: Peter Lang. Nevalainen, Terttu & Helena Raumolin-Brunberg. 2003. Historical Sociolinguistics: Language Change in Tudor and Stuart England. London: Longman. Peikola, Matti. 2005–2006. “First is writen a clause of the bigynnynge thereof”: The table of lections in manuscripts of the Wycliffite Bible. Boletin Millares Carlo 24–25. 343–378. Peikola, Matti. 2011. Copying space, length of entries, and textual transmission in Middle English tables of lessons. In Jacob Thaisen & Hanna Rutkowska (eds), Scribes, Printers and the Accidentals of their Texts, 107–124. Frankfurt am Main: Peter Lang. Renoir, Alain & C. David Benson. 1980. John Lydgate. In A. E. Hartung (ed.), A Manual of the Writings in Middle English, 1050–1500, 1809–1920, 2071–2175. New Haven, Connecticut: The Connecticut Academy of Arts and Sciences. Revard, Carter. 1979. Richard Hurd and MS Harley 2253. Notes and Queries 26. 199–202. Revard, Carter. 2000. Scribe and Provenance. In Susanna Fein (ed.), Studies in the Harley Manuscript: The Scribes, Contents, and Social Contexts of British Library MS Harley 2253, 21–109. Kalamazoo, MI: Western Michigan University Medieval. Rissanen, Matti. 2000. Standardisation and the language of early statutes. In Laura Wright (ed.), The Development of Standard English, 1300–1800, 117–130. Cambridge: Cambridge University Press. Roberts, Jane. 2011. On giving Scribe B a name and a clutch of London manuscripts written c. 1400. Medium Ævum 80.2. 247–270. Samuels, Michael L. 1963. Some applications of Middle English dialectology. English Studies 44. 81–94. Reprinted in 1989 in Margaret Laing (ed.) Middle English Dialectology: Essays on some Principles and Problems, 64–80. Aberdeen: Aberdeen University Press. Samuels, Michael L. 1969. The dialects of MS Bodley 959. In Conrad Lindberg (ed.), MS Bodley 959, Genesis – Baruch 3.20 in the Earlier Version of the Wycliffite Bible, 329–339. Stockholm: Almqvist & Wiksell. Reprinted in 1989 in Margaret Laing (ed.) Middle English Dialectology: Essays on some Principles and Problems, 136–149. Aberdeen: Aberdeen University Press. Samuels, Michael L. 1981. Spelling and dialect in the late and post-Middle English periods. In Michael Benskin & M.L. Samuels (eds), So Meny People Longages and Tonges: Philological Essays in Scots and Mediaeval English Presented to Angus McIntosh, 43–54. Edinburgh: Middle English Dialect Project. Reprinted in 1988 in J.J. Smith (ed.), The English of Chaucer and his Contemporaries. Aberdeen: Aberdeen University Press. 86–95. Samuels, Michael L. 1984. The dialect of the Harley Lyrics. Poetica (Tokyo). 39–47. Reprinted in 1989 in Margaret Laing (ed.) Middle English Dialectology: Essays on some Principles and Problems, 64–80. Aberdeen: Aberdeen University Press. Sandvold, Silje Nising. 2010. Scribal variation in a legal document: a study of the bounding of Barmston (1473). Stavanger: University of Stavanger MA thesis. http://brage.bibsys.no/ uis/handle/URN:NBN:no-bibsys_brage_12694?locale=en Schaefer, Ursula (ed.). 2006. The Beginnings of Standardization. Frankfurt am Main: Peter Lang. Stenbrenden, Gjertrud F. 2011. The Chronology and Regional Spread of Long-Vowel Changes in English, c. 1150–1500. Oslo: University of Oslo PhD thesis. Wright, Laura. 2000. Introduction. In Laura Wright (ed.), The Development of Standard English 1300–1800: Theories, Descriptions, Conflicts, 1–8. Cambridge: Cambridge University Press.

Ben Outhwaite

9 Lines of communication: Medieval Hebrew letters of the eleventh century Sources: The Hebrew letters and other documents discussed in this article come from the Cairo Genizah, a celebrated collection of medieval Jewish manuscripts found in the “sacred storeroom” (genizah) of the Ben Ezra Synagogue, Fustat. The contents of the Genizah are now dispersed worldwide, but the greatest part forms the Taylor-Schechter Collection in Cambridge University Library. Other large collections of Genizah material can be found in the Jewish Theological Seminary, the John Rylands Library, Manchester, and the British Library. Many Hebrew letters of the 11th–12th centuries are edited in Gil (1983 and 1997). Most manuscripts from the Cairo Genizah are now freely available to view online (with registration), through the portal of the Friedberg Genizah Project (www.genizah.org). Abstract: Medieval Hebrew has suffered neglect from linguists when compared to other varieties of Hebrew, both more ancient and more modern. Viewed primarily as a literary-liturgical language, without a living native-speaker tradition, it had been regarded as an artificial idiom, not worthy of much scholarly interest. The discovery of a vast amount of documentary material in Medieval Hebrew from the Cairo Genizah, however, points to a vibrant Hebrew language tradition in Palestine and Egypt in the Fāṭimid period, in which Hebrew enjoyed a major role as a language of communication between the traditional centres of Jewish religious governance, the talmudic academies, through to the scattered communities of the Jewish diaspora. The 11th century in particular is a flourishing period for Hebrew letters, and through an examination of the writings of a prodigious letter-writer, the Palestinian Gaʾon Solomon ben Judah, and some of his contemporaries we can discover the motivation and effect that these champions had on the history of the Hebrew language.

Medieval Hebrew is the Cinderella of the Hebrew language family. Long ignored and rarely given the linguistic status that it deserves, it is overshadowed by its Biblical, Rabbinic and Modern sisters. The lack of a continuous native-speaker tradition for Hebrew through the Middle Ages has resulted in neglect from lin-

184   

   Ben Outhwaite

guists (Rabin 1970: 325).¹ When interest has been shown, the language has often been dismissed as somehow falling short: lacking the underpinning of a regularly spoken idiom, it is less than a real language, nothing but an artificial assemblage put together from the building-blocks of the biblical and rabbinic languages (see Outhwaite 2012: 3–8). The most comprehensive recent study of the Hebrew language in all its periods described Medieval Hebrew as “not, properly speaking, a ‘language’ comparable to [Biblical Hebrew] or [Rabbinic Hebrew]” (Sáenz-Badillos 1993: 204). There is no doubt that this is a view shared by the authors of other studies of the Hebrew language. Joel Hoffman’s work, In the beginning: a short history of the Hebrew language (2004), is a particularly cruel example. It devotes the first one hundred and eighty pages to Biblical Hebrew and the Rabbinic language. Pages 187 onward concentrate on Modern Hebrew from the time of Ben Yehuda to today. Medieval Hebrew is consigned to pages 180–183 as “non-spoken Hebrew”. Such short shrift has its origins in the scholarship of previous generations. The Jewish Encyclopedia, a masterpiece of the Wissenschaft des Judentums movement published in New York in 1901–1906, is still sometimes a useful work to dip into today, but in particular it acts as a valuable touchstone for late 19th-century European and New World scholarship on Judaism. It fell to Caspar Levias, who is now thought of mainly as an accomplished aramaist, to provide the entry on the Hebrew language in volume twelve of the Encyclopedia. Since he was writing before the creation of the state of Israel, and thus before the birth of Modern Hebrew as a distinct linguistic entity, Levias divided Hebrew into three, Biblical Hebrew, Rabbinic Hebrew and “Neo-Hebrew”, the latter covering the period of the Amoraʾim (the scholars cited in the gemara [commentary] of the Talmuds, c. 200 CE onwards) up to the present (as was). To sum up the linguistic value – as opposed to the literary value – of one thousand seven-hundred years of Jewish culture takes Levias but one dismissive moment: “This period is of no interest to the student of Hebrew philology” (Jewish Encyclopedia 1901–1906: xii 308). We should probably excuse Caspar Levias, though perhaps not those who repeat his mistakes many decades later, since he was writing only a few years

1 And as Horvath and Wexler (1994: 244–245) point out, Chaim Rabin is himself guilty of just this: “Thus, in a 43-page state-of-the-art survey of Hebrew linguistics (1970), Rabin devoted no more than two pages to unspoken ‘Medieval Hebrew’.” This is slightly unfair, given that Rabin is one of the few scholars to have made a linguistic study of Medieval Hebrew, which was belatedly published as The development of the syntax of post-biblical Hebrew (Leiden, Brill, 2000); however, despite its broad title, it examines only the syntax of European prose. It is invaluable, however, for its introductory survey of Medieval Hebrew, which, though dated, surpasses anything in the standard reference books.

Lines of communication: Medieval Hebrew letters of the eleventh century   

   185

after the most important discovery of modern Jewish scholarship had been made, a seismic event in the history of Jewish letters and in the philology of the Hebrew language, the discovery of the Cairo Genizah. When the lecturer in rabbinics Solomon Schechter brought back to Cambridge University Library nearly two hundred thousand manuscript fragments from the storeroom (“genizah”, a storeroom for sacred texts that can no longer be used) of the Ben Ezra synagogue in Fusṭāṭ, he changed at a single stroke the entire landscape for the academic study of medieval Jewish culture. The scholarly world now had a wealth of material with which to examine minutely the literary and historical “archive” of an influential and rich Jewish community placed at the very centre of a huge network of cultural, commercial and social links.² It was too soon, however, for Levias and his fellow scholars, and there are only brief references to early discoveries from Schechter’s collection in the Jewish Encyclopedia, but all subsequent scholarship on the medieval Jewish world must take into account the fragments of the Cairo Genizah. Prior to the Genizah, work like Levias’s on the medieval language had relied on comparatively few texts, and these were mostly the writings of European Jewry from the High Middle Ages, chiefly poems from the golden age of Spain and difficult halakhic commentaries from Germany; the former were triumphs of style over intelligibility, the latter Levias declared aesthetically unappealing (Jewish Encyclopedia, 1901–1906: xii 309). The impression given was that Hebrew functioned either as a vehicle for poetry or as a medium for the legalistic wranglings of talmudic scholars. The Genizah, however, provides an enormous slice through the written culture of the Jews of Islamic lands for a period in excess of one thousand years. Given that the vast majority of the world’s Jewish population lived under Islam in the early Middle Ages (Brody 1998: xx), it became clear that previous analyses were utterly unrepresentative of the medieval Jewish world in its entirety, and that the study of the Hebrew language, in particular, had been greatly hampered by a paucity of data. That Hebrew was the favoured language of medieval Jewish poets, particularly the payṭanim [liturgical poets], was well known before the discovery of the Genizah, but the great extent to which Hebrew functioned in other literary and

2 The Cairo Genizah in origin is not an archive, but a vast heap of discarded written material, most, but not all, of a sacred character (Bibles, prayer books and the like). What makes the Genizah particularly important for historical linguistics is the amount of non-sacred and non-literary texts that were deposited, by accident or design, such as personal letters, legal documents, commercial deeds etc. For the full story of the discovery of the Genizah and its significance for the history of Judaism and for the medieval world in general, see the recent books by Glickman (2011), Hoffman & Cole (2011), and Reif (2000).

186   

   Ben Outhwaite

non-literary genres among the Jews of the Mediterranean and Middle East must have come as a surprise to those who, like Levias, had accepted the commonlyheld belief (still prevalent in some quarters) that since Hebrew ceased to be the native spoken language of any Jews, it also ceased to function as a productive written idiom. This belief holds that from 200 CE (with the codification of the Mišna) until its revival in the late 19th century by Ben Yehuda and others as the culmination of the European haskala [Jewish enlightenment] movement, Hebrew was a moribund idiom relegated to a relatively few, discrete, written literary-liturgical genres and used for the recitation of prayer. Even recent surveys of Hebrew such as Sáenz-Badillos’s History of the Hebrew Language dwell upon Hebrew as above all a poetic medium, and as the arabicised pseudo-Mischsprache of the language of translations from Arabic, yet fail to mention its far more straightforward use as a non-literary, communicative idiom. Yet the Cairo Genizah has preserved hundreds of examples of Hebrew being used for everyday written communication between Jews from as early as the 9th century up to the 17th–18th centuries.³ In choosing to examine the linguistic situation in the 11th century, we are choosing a period rich in primary sources, the Genizah having preserved examples of all the genres of writing produced by the communities of Egypt, Palestine and other communities with whom the Jews of Fusṭāṭ kept in communication. The period is particularly well represented in the written record, since at that time the Jewish community of Fusṭāṭ was flourishing under the relatively benign rule of the Fāṭimids, paper had become easily available, and a considerable proportion of the Jewish community was able to write. Fusṭāṭ maintained substantive cultural, religio-legal and commercial links with the Jewish communities in other parts of the world, most regularly with the talmudic academies (the Yešivot, which provided religious training, guidance and leadership to the diaspora communities, in return for the payment of tithes) in Iraq and Palestine, but also with Europe (Spain, France, the Byzantine Empire) and as far afield as Yemen or central Asia. Communication relied upon the sending of letters along established networks between the academies, through primary communicative hubs (of which Fusṭāṭ was the most important, but there were also Jewish centres like Damascus, Palermo, Qayrawān, and many more around the Mediterranean)

3 Gil (1997: ii 13–17) has identified Cambridge University Library T-S NS 308.122 as a Hebrew letter from the academy of Pumbedita, Iraq, and dates it to c. 850 CE. On the whole, though, the Genizah does not provide us with much documentary material datable to before the mid10th century. The earliest dated manuscript in the Cairo Genizah is a piece of Bible written in Iran, T-S NS 246.26.2, which contains a colophon dating it to 903–904 CE. The fact that the Genizah is poor in material earlier than this does not mean, of course, that such material never existed.

Lines of communication: Medieval Hebrew letters of the eleventh century   

   187

where letters might be copied to be served up to multiple destinations, before eventually reaching their addressees in the scattered Jewish communities of the Diaspora. All this was serviced by the travels of merchants, mostly Jewish but not always, who carried letters as they plied their trade. Hundreds of such letters from the 11th century have been preserved alongside the more sacred texts deposited in the Genizah. When Solomon ben Judah, the embattled, and occasionally embittered, Gaʾon [Head] of the Jerusalem Yešiva in the second quarter of the 11th century, sent letters to his fellow Jews in the old city of Fusṭāṭ, Egypt, to provide news of the Holy City, rule on religious matters, settle political disputes or to solicit funds for his academy, he addressed the official heads of the community there (the heads of the Palestinian or Babylonian factions) or important individuals among its leading citizens. Taking his lead from the geʾonim who served before him, both, it seems in Jerusalem and in the academies of Iraq, Solomon communicated with them mainly in the Hebrew language of his forefathers, the holy language of Judaism. Although today we would classify the product of Solomon’s pen as Medieval Hebrew (and perhaps further sub-classify it as “Medieval Documentary Hebrew”), in his eyes and in the terminology of the time he was writing in the Lešon ha-Qodeš [The Holy Language], the same language in which the twentyfour books of the Jewish Tanaḵ (Bible) were composed, and which, it increasingly appears, enjoyed an unbroken (though at times considerably attenuated) existence as a medium of written communication from antiquity throughout the Jewish Middle Ages. A resident but probably not a native of Palestine (contemporary detractors referred to him as al-Fāsī, “the man from Fez”), Gaʾon Solomon b. Judah’s first language was Arabic and, as was the usual practice of the time among the Jews of Islamic lands, when he wrote Arabic he usually used Hebrew characters, i.e. Judaeo-Arabic, a written vernacular that functioned for a wide variety of text types. The great scholar of the documentary texts of the Genizah, S. D. Goitein, describes this idiom as “a semiliterary language of rather regular usage and considerable expressiveness” (Goitein 1973: 5). This expressiveness enabled it to be turned to all manner of genres, and Judaeo-Arabic serves both for the composition of literary and religio-legal texts as well as functioning as a language for everyday correspondence in this and subsequent periods. The fact that JudaeoArabic was written in Hebrew characters (rather than in Arabic script) appears to be less a statement of religious or ethnic identity and more a by-product of an educational system that centered around the teaching of the Hebrew language. Copious evidence from the Genizah suggests almost universal education among Jewish males (at least among middle-class town-dwelling Jewish males, the constituency the Genizah best represents), and its intention was to prepare them for

188   

   Ben Outhwaite

their adult duty of reading the Torah (Goitein 1967–1993: ii 175). Advanced students might go on to study the Jewish law codes of the Mišna and Talmud, but all children were trained in the recitation of the Pentateuch so that they could stand up in the synagogue and read out the weekly paraša [Torah chapter], as required. Beginning with the writing out of the Hebrew characters and the Tiberian vowels that accompanied them, the children eventually moved on to the copying of whole biblical books, not infrequently the pedagogically-important book of Leviticus (Olszowy-Schlanger 2003: 47–56). Once sufficient fluency had been reached in reading, the art of writing was not pursued any further (Goitein 1967–1993: v 177–178). While Arabic was the language of the rulers of Palestine and Egypt, and knowledge of it was essential even in the Jewish community, it was regarded in this period as a foreign, and not a Jewish language, even when written in Hebrew characters. When Solomon b. Judah and his contemporaries refer to it, they often call it “the language of the Hagri”,⁴ a gentilic found in Psalms 83:7 where it originally referred to an unidentified tribe, probably of Aramaean origin. The word is commonly found in the Hebrew of eleventh-century letters to refer to Arabs, and specifically in reference to their language. That Solomon even regards Arabic written in the Hebrew script as a foreign language is made clear by examples such as his responsum [a legally-binding answer to a question of religious law] preserved in the Genizah, where he replies to his correspondent in Hebrew (T-S G2.10, recto, ll. 17–18): “This question arrived before us to the gate of the Academy […] and since it is in the language of the Hagri, we have ordered that the reply be written in the language of the Hagri”, before switching from Hebrew into JudaeoArabic to make the ruling on the original query. Other geonic (the period between the early 7th century and the middle of the 11th century, named after the ge’onim, Heads of the Talmudic Academies) contemporaries attest the same practice of marked code-switching, and it is found too in Babylonian responsa of the 10th– 11th centuries (Brody 1998: 189; Outhwaite 2004: 63).⁵ Although to our eyes, the use of Hebrew graphemes for writing down Arabic might imply an “ownership” or “judaisation” of the written form of the language, this does not appear to be the

4 For instance, Solomon b. Judah refers to the sending of letters, “I have sent three letters: the first, for you, is in the language of the Hagri” (Cambridge University Library T-S 20.181 recto l. 16). 5 While code-switching is always marked and conscious in Hebrew letters, it is probably less of a conscious act in Judaeo-Arabic contexts, where it is quite usual to find a large number of Hebrew calques, loans and even whole expressions. The reverse is not true for most Hebrew correspondence of this period, which avoids any significant Arabic intrusion.

Lines of communication: Medieval Hebrew letters of the eleventh century   

   189

case, with Solomon b. Judah and others not noting a distinction between Arabic and Judaeo-Arabic in their replies. The clear implication of Solomon’s code-switching is that, in his opinion (shared by the other writers of responsa), the fitting language for the written communication of halakhic subjects was not Arabic, but Hebrew, else they would have not felt a great need to comment on it. In an earlier generation it would have been Aramaic, while this was still a spoken language of the Jews in Iraq and Palestine, and indeed the language of earlier responsa is predominantly in this language. But with the decline in knowledge of Aramaic, following the Islamic conquests, and the adoption by the Jews of Arabic as their vernacular, Hebrew reasserted itself as the language of halakha [religious law] and, probably as a result, of much correspondence in general. It is likely, however, that Hebrew never ceased being used as a written medium, halakhic or otherwise, in Palestine and perhaps Egypt even during the pre-Islamic period, since a number of Palestinian literary works are written in Hebrew rather than Aramaic, and that number increases greatly in the geonic era, and although Jewish manuscript finds from this period are extremely rare, there are tantalising glimpses of Hebrew being used in correspondence among the Oxyrhynchus papyri from Egypt, from perhaps the 4th–6th century CE (Cowley 1915: 209–213). The Palestinian Gaʾon Solomon b. Judah, however, is remarkable in the number of his Hebrew letters recovered from the Genizah and the range of uses to which he put them. He has left behind ninety-nine letters that he either wrote or signed, and of those only about fifteen are in Judaeo-Arabic.⁶ Thus four-fifths of all his extant correspondence is in Hebrew, a non-spoken idiom, rather than in his native language, Arabic. Moreover, it appears that he is following in the footsteps of his geonic predecessors, who have also left behind (though far fewer in number) letters in the holy tongue. His immediate predecessor as Gaʾon, Solomon ha-Kohen, served for less than a year as head of the Jerusalem Academy, and as a consequence leaves little trace in the Genizah, but of his four preserved letters all are in Hebrew (Gil 1983: ii 81–91). Before him, Josiah Gaʾon (died in 1025), leaves twelve Hebrew letters and two in Judaeo-Arabic (Gil 1983: ii 27–81). The Babylonian geʾonim, beginning with Nehemiah ha-Kohen (died 968 CE), leave less of a first-hand written record in the Genizah, but their extant correspondence (particularly of the earlier geʾonim) is predominantly in Hebrew.⁷ Solomon b. Judah chose to write in the Hebrew language, when the language of the Hagri would have served as (or perhaps more) efficiently as a communi-

6 Most of them are edited in Gil (1983: ii 91–299). 7 Much of it edited by Gil (1997), vol. ii.

190   

   Ben Outhwaite

cative idiom, given that it was the spoken tongue of both sender and recipient. There is certainly plenty of evidence for the use of Judaeo-Arabic as a medium of epistolary communication: there is a much greater proportion of Judaeo-Arabic correspondence in the Genizah, from the Classical Period (the late 10th to the mid-13th centuries CE, the periods of Fāṭimid and Ayyūbid rule in Egypt) up to the arrival of Spanish Jewish exiles in the 16th century.⁸ Throughout the Classical Period, Judaeo-Arabic functions in all literary modes and is used for correspondence of all types. Commercial correspondence, however, probably accounts for the greatest proportion of Judaeo-Arabic letter-writing. For the Jewish mercantile class of Egypt and North Africa, Judaeo-Arabic was the most effective written medium to do business. The international language of trade was Arabic, and Hebrew, in many cases, was deficient in the necessary vocabulary for the expanding world of commerce. Moreover, commercial correspondence is a more transparent form of writing, and seeks to convey information with the minimum of artistry. While Judaeo-Arabic letter-writing is not devoid of aesthetic qualities, it is noticeable that commercial letters tend to avoid the lengthy openings so commonly found in Hebrew correspondence and dare to broach the business of the letter after only a few standard, formulaic expressions of politeness. Indeed, one of the commoner openings for a Judaeo-Arabic letter of the 11th–12th centuries is the straightforward waṣala kitābuka [your letter arrived], when it is a reply to a previous letter, or just kitābī [my letter]. The communicative function is therefore predominant in the Judaeo-Arabic letter, and where a greater degree of formality or flattery is required, it is sometimes achieved through the use of a Hebrew opening to the letter, before an unmarked switch to Judaeo-Arabic once the formalities are dealt with and the main business is to be discussed.⁹ With the numerous examples of successful Judaeo-Arabic correspondence in circulation around him, it is remarkable that Solomon b. Judah maintained an

8 The Genizah does not provide a large amount of material that can be reliably dated to the period of Mamlūk rule in Egypt (1250–1517). Texts begin again to be deposited in significant numbers in the Genizah from the Ottoman period onwards, when the Jewish community of Cairo was swollen by the arrival of immigrants from Spain. In this latter period, Hebrew reasserted itself in the written record due to the new arrivals’ ignorance of Arabic. 9 “Unmarked” in the sense that it goes unremarked by the writer, nor is the change delineated by a pragmatic marker of any type. This switching between Hebrew for (essentially empty, formulaic) praise and Arabic for the content of the letter is very frequent in the 11th century. For example, T-S 13J15.23, a letter written by Eli ha-Kohen b. Ezekiel the cantor in 1071 CE, opens with eight and a half lines of rhymed Hebrew praises before switching midway through line 9 to Judaeo-Arabic when the real purpose of the letter is reached, “I have already sent your honour several letters”. Edited in Gil (1983: iii 85–87).

Lines of communication: Medieval Hebrew letters of the eleventh century   

   191

adherence to Hebrew in his correspondence, given that his letters were not simply formulaic expressions of a pastoral nature, and indeed often involved matters of intercommunal importance or of the movement of considerable sums of money. There must have been a greater significance to the act of writing in Hebrew, which outweighed any potential communicative benefits that Judaeo-Arabic could provide. There are probably two interlinked factors at play in the choice of Hebrew over Judaeo-Arabic in the 11th century. While Hebrew did clearly bear a greater aesthetic load than Arabic, as evidenced by its use for liturgical and secular poetry of the period, and its appearance, when called for, in the openings of Judaeo-Arabic letters, it was not the case, as Rina Drory (1992: 60–61) has put it, that in the Middle Ages Arabic was a transparent medium of communication whereas Hebrew was an opaque medium of a predominantly literary-aesthetic character.¹⁰ Instead we should be aware that many of the Hebrew letters that we now possess were originally intended for public performance, that is to say, they were intended to be read out before the community, and the usual place for this was the synagogue, after one of the regular services. In a letter to his chief supporter and head of the Palestinian faction in Egypt, Efraim b. Šemarya, Solomon b. Judah writes the instruction “and the long sealed scroll is to be read in the hearing of [lit. “ears of”] all the people to inform them about the man who rejected the title of Ḥaver and chose instead the title of Alluf”.¹¹ This accounts for why we find notices of excommunication, requests for tithes, acknowledgements of donations, and the sundry pronouncements of the geʾonim all written in Hebrew – public notices all, delivered in the form of a letter – but also why personal letters between family members, or business arrangements between merchants are principally written in Judaeo-Arabic. In this, and previous periods, the language

10 There are also frequent examples from the Genizah of Hebrew being used primarily for communicative purposes, simply because the writer clearly did not know Arabic. T-S NS 142.126 is a good example of this, perhaps one of the first letters that the 13th-century scholar Yeḥiʾel b. Elyaqim wrote on his unwilling arrival in Egypt. He was from Byzantium and, although he subsequently settled in Egypt and achieved a certain status, he was not an Arabic speaker or writer when he arrived, and was thus forced to write in Hebrew (of a particularly biblical character) in order to have his plight understood. See Outhwaite (2009: 207–214). 11 T-S 20.181 recto ll. 22–23 (edited in Gil 1983: ii 178–180). Solomon b. Judah was insulted by the leader of the Babylonian faction in Egypt who went against tradition, and stirred up factional politics, by disdaining the title of ḥaver [member], which was an honorary title given by the Palestinian Yešiva of which Solomon was head, preferring to use the Babylonian title of alluf [chief] and thus clearly nailing his colours to the wall at a time when inter-factional solidarity was, in Solomon’s eyes, a more important consideration.

192   

   Ben Outhwaite

of the synagogue service was Hebrew (with some Aramaic embellishment in the form of the reading of the Aramaic targum [translation of the Bible]). Although in much earlier times, other languages had been used for Jewish prayer, for instance Greek (Elbogen 1993: 199), by the time of the geʾonim, Hebrew had become the only acceptable language for public worship and Arabic was not to be used in the synagogue (Goitein 1967–1993: ii 220–221). Thus for the traditional elite of the community, the religious leaders and their local appointees, it was necessary to use the acceptable liturgical language, if their words were to reach their flock. Moreover, in attaching their pronouncements to the formal occasion of the synagogue service, these communications acquired a solemnity and authority that derived from their apparent inclusion in the public display of religious worship. That only Hebrew was acceptable for reading in public in the synagogue also accounts for the reason why we find the language in use at the opposite end of the social scale. It would otherwise be a strange dichotomy to note that the high variety, non-spoken Hebrew, is the preferred written medium of both the religious elite and the indigent underclass. It is the middle-classes, especially the mercantile class, who adhere to the more intimate mode of Arabic. The very lowest in society, at least in socio-economic terms, emulate their superiors by showing a preponderance of Hebrew in their correspondence. It is the commonest language of begging letters, a popular documentary genre in the Genizah, either written by the poor unfortunates themselves or at their behest by a local scribe. While in some cases these are the foreign poor, often hailing from Byzantium (Cohen 2005: 72–108), where they would be unlikely to have knowledge of Arabic, they are supplemented by many locals, or at least by Jews from other Arabic-speaking lands, down on their luck. The immediate purpose of begging letters is often stated explicitly in them: the writers request a pesiqa [collection], a term denoting a public appeal among the congregation for funds to help the author, which would have taken place after the service in the synagogue (Cohen 2005: 220–224).¹² The use of Hebrew reaches across social divisions, therefore, with both the religious elite and the indigent employing Hebrew in their communications in order to most effectively reach the largest audience and maximise the potential benefits. Not all Hebrew correspondence can be accounted for in this way, through the necessity for it to be read aloud in the synagogue. Drory asserts that Saʿadya, the Gaʾon of Sura in Iraq in the second quarter of the 10th century, wrote his pastoral

12 Cohen calls it a “pledge-drive”. A good example is T-S 13J13.16, an 11th-century begging letter by a woman who has lost her nose through some disfiguring disease: “May my lord order a charitable collection (pesiqa) in any place that our lord desires” (recto, ll. 20–21; edited in Outhwaite 2009: 203–206).

Lines of communication: Medieval Hebrew letters of the eleventh century   

   193

correspondence in Hebrew to give “an official, solemn, almost regal complexion” (Drory, 1992: 58). However, I find it more likely that as his pastoral correspondence was intended for communal consumption it had to be in the language suitable for public recitation. But there are clear cases of Hebrew letters written by Jewish leaders whose contents were probably only intended for a few discreet ears, rather than those of “all the people”. Solomon b. Judah was a prodigious letter-writer and quite a few of his letters appear to lack the tone of solemnity, replacing it with an intimate familiarity, particularly when addressing his regular correspondents such as Efraim b. Šemarya or Sahlān b. Abraham.¹³ Moreover they sometimes share details that were probably not for public consumption.¹⁴ It is also not unusual for Solomon or other leading members of the Jewish community to be the recipients of letters in Hebrew, rather than Judaeo-Arabic.¹⁵ In these cases, the choice of Hebrew must have been less a practical matter of public performance, and more a symbolic step that bound the writer and addressee through a common, but exclusive, code in the wider Islamic society. The Palestinian geʾonim were seated at the head of the scholars who made up the Academy of the “Pride of Jacob”, the Yešiva of Jerusalem. They were the inheritors of an ancient tradition of Jewish religious governance that stretched back to the Great Sanhedrin, and were the final earthly court capable of sitting in judgement on Jewish religious matters. For the Gaʾon Solomon b. Judah to have eschewed the sacred, eternal language of Jews in favour of the language of the

13 While this can simply be a matter of tone, it can be objectively discerned in the use of the first person singular, instead of the plural of majesty (customary for geonic correspondence), the curtailing of the poetic praises in the opening of the letter, the use of nicknames (some of an amusing or insulting nature), as well as the writer passing on gossip or referring to family matters. 14 For instance, many of the Hebrew letters written to Efraim b. Šemarya during the dispute over the gaonate with Nathan b. Abraham (1038–1042 CE) contain virulent accusations against third parties that the Gaʾon would probably not have wanted to become public knowledge. In other letters the Gaʾon relates sorry details of his health or describes melancholy family matters that would be out of place in an open missive. In one fragmentary Hebrew letter to Sahlān b. Abraham in Fusṭāṭ c. 1029 he moans that as he is writing he is surrounded by “five women without a man”, the female members of his household having met with different misfortunes, Cambridge University Library Or.1080 J105, recto l. 25 (Gil, 1983: ii 147–149). Solomon b. Judah often sheds his regal complexion in favour of pathos. 15 When these were formal requests for halakhic rulings (šeʾelot requesting tešuvot), Hebrew was the preferred medium, as noted above, and were in theory to be read aloud by the members of the Yešiva so that a discussion of the assembled scholars could take place, followed by the Gaʾon giving the authoritative ruling (Brody 1998: 46–47). However, this represents only a proportion of Hebrew correspondence.

194   

   Ben Outhwaite

temporal rulers of the land would have been a step that acknowledged the secondary status of the Yešiva and of his own subservient position within Islamic society. The Palestinian Academy had won the right from the Fāṭimid government to administer many of the legal affairs of the Jewish communities in the Fāṭimid Empire, and it retained much of its original status as the arbiter of Jewish law and custom. To not administer and arbitrate in the Jewish language would have been a lost opportunity, and a recognition of temporal, if not also spiritual, defeat. Moreover, on a personal level, schooling centered upon proficiency in the Hebrew language, and a man could be measured by this ability.¹⁶ It was important therefore for the prime religious authority in Palestine to assert his position as the leading scholar through his mastery of the language. There is a nationalistic character to the use of Hebrew in the 11th century, when, sanctioned by the Fāṭimids, the Jerusalem Academy exercised considerable power in the Jewish communities that owed it allegiance. In choosing not to deploy the language of the conquerors when fulfilling his public role, Solomon b. Judah asserted a Jewish identity independent from the Islamic state, and provided a written idiom that bound the members of the religio-ethnic group through a shared linguistic identity. Some of his flock reacted appropriately, and joined in, adorning themselves with two thousand years of shared Jewish culture, and cementing their religious and ethnic bonds through wielding the language of the Torah. There are others, however, who did not apparently feel a need to mark their membership of this group through the Hebrew language. Solomon b. Judah, for instance, never writes to Abraham ibn Furāt, a physician to the Muslim governor of Ramla, in Hebrew, only Judaeo-Arabic (Outhwaite 2004: 63–64). Ramla was the Fāṭimid capital of Syria-Palestine and thus Abraham was probably the most important Jewish official in Palestine. But as a Fāṭimid courtier, he inhabited a very different world from Solomon, and his acknowledgement to a shared religious identity with Solomon and the Yešiva was probably not something to be stressed, as Solomon must have recognised: they inhabited different social spheres. He was a member of the temporal elite, not the religious, for whom facility in Arabic was a social advantage, and acknowledgement of membership of the Jewish minority was probably a hindrance. There are other courtiers, who tended to inhabit Cairo rather than the Jewish quarter of Fusṭāṭ, and overwhelmingly

16 Drory quotes the Karaite philosopher Sahl b. Maṣliaḥ as he accused Jacob b. Samuel of linguistic ignorance: “So far I have found nearly sixty errors in your letters, some of spelling, some of meaning; and you do not qualify as one worthy of arguing with Bǝne Miqra [followers of the Bible = Karaites]” (Drory 1992: 58–59).

Lines of communication: Medieval Hebrew letters of the eleventh century   

   195

they communicate and are communicated with in the Arabic language which, for them, both denoted and facilitated the exalted social status they had achieved. Looking at a contemporary of Solomon b. Judah, we can see the switching of codes, between Hebrew and Arabic, in action and how it represents the social spheres in which each code was preferred. The Genizah preserves correspondence relating to the geonic schism of 1038–1042 CE, when a young pretender, Nathan b. Abraham, tried to usurp the Palestinian gaonate from the incumbent, Solomon b. Judah. We possess a number of letters written by Nathan both before and during the event, and even though it remains a relatively small sample, it is extremely instructive to examine his use of different codes in the multilingual Jewish society. When we first come across Nathan, around 1035, he is in business. The one commercial letter we have from him is, as expected, in Judaeo-Arabic, and it lacks any kind of linguistic adornment, either in the form of a Hebrew preface or any extended opening blessings in Judaeo-Arabic (indeed, it begins abruptly with “your letter arrived”).¹⁷ It is even addressed in Arabic script, something in which Nathan was proficient. The next time we hear from Nathan, he is writing to one of his patrons, a leading Jew in Qayrawān, and so Nathan switches to the register suitable for addressing such a prominent person, and writes his letter in Hebrew, prefaced with about five lines of elegant blessings, which not only praise the recipient but also establish Nathan’s credentials as a scholar, proficient in the sacred language.¹⁸ It is already clear by this stage that Nathan has failed in business and has begun to think about a different career, and for that he must exercise his Hebrew. Further letters follow, some in Hebrew, others in JudaeoArabic (but with ornate Hebrew openings), as Nathan corresponds with a number of different prominent men. Like his rival, when he writes to Abraham ha-Kohen ibn Furāt in Ramla, he too writes in Judaeo-Arabic, switching code to best suit this government official. Once Nathan had proclaimed himself Gaʾon at a solemn ceremony in a synagogue in Ramla (while Solomon excommunicated him in a different synagogue across town), he writes almost exclusively in Hebrew. Moreover, he places his name and title prominently at the head of the paper, in imitation, it seems, of Babylonian geonic practice (his rival, Solomon b. Judah, does not do this), and to further enrich his Hebrew he adorns his prose with entirely superfluous vowel signs, underlining his religious and scholarly credentials and lending an air of biblical authority to his pronouncements.¹⁹ One can see, therefore, that

17 T-S Ar.54.93, edited by Gil (1983: ii 300–302). 18 T-S 8.3, edited by Gil (1983: ii 305–307). 19 And indeed, these vowel signs are not only superfluous, since they are used to mark common words the consonantal spelling of which any educated reader would be familiar with,

196   

   Ben Outhwaite

as Nathan rose (in a particularly ruthless manner) through a society that gave a great weight to the mastery of the written word, he employed the different codes and varieties available to him to suit his changing social status and assist his rise. The 11th century marks the high point for the use of Hebrew in communicative discourse in the High Middle Ages. It is attested in hundreds of letters from the period, and is found as the regular written language of both the religious elite and the lower classes (as well as being the international language of Jews who shared no common tongue). It is employed to be read aloud in the synagogue and asserts and reinforces membership of a defined religio-ethnic group at a time of competing powers and external pressure. Hebrew enjoys a revival as a written idiom, pushed by the traditional centres of Jewish governance, the academies, and the custodians of tradition, the geʾonim, who champion the use of Hebrew in the face of the victory of the Islamic state, and do something to stem the influx of JudaeoArabic written culture. Thanks to the centrality of the academies in Jewish life of this period, it is possible that the strength of Hebrew discourse in the 11th century excels anything since the language ceased to be spoken. Certainly prior to the 11th century we have less evidence of a wide use of Hebrew, although this could just be a symptom of the comparatively smaller number of sources available to us. However, the copious evidence from the Genizah attesting to the linguistic situation in the following centuries clearly shows that there is a sharp decline in the use of Hebrew for correspondence and other non-literary modes. JudaeoArabic, and Arabic proper, usurped the roles enjoyed by the Hebrew language at its medieval high-water mark, to the point that Maimonides (died 1205) ruled in his Mišne Tora that vernacular languages could be used in place of Hebrew for prayer in the synagogue. It is no coincidence that this decline in the prestige of Hebrew mirrors the decline, and ultimate disappearance, of the traditional centres, the academies in Iraq and Palestine – the authorities that were pushing Hebrew as their idiom of choice. Solomon b. Judah is perhaps the last great gaʾon of the Jerusalem Yešiva, as following him the gaonate is mired in disputes and collapses from the powerful pressure of external events in Palestine, as first the Saljūqs and then the Crusaders force the academy to leave the Holy City, before it eventually takes refuge, a shadow of its former greatness, in Egypt. Robbed of its

but are also unusual. They are not the Tiberian signs, the dominant system for denoting the vocalisation of biblical and other texts, but the supralinear signs of the Babylonian system, rarely found in Palestine in this, or any other, period. Moreover, they are what appears to be a personal modification of the Babylonian system, used, I can only suggest, to give an air of scholarly authority (and specifically the more ancient Babylonian geonic authority) to Nathan’s mischievous missives. I intend to write further on this interesting matter elsewhere.

Lines of communication: Medieval Hebrew letters of the eleventh century   

   197

sacred foundations, severed from the heart of the Jewish religion, the Palestinian Academy loses its authority within the vibrant and ambitious Egyptian Jewish community, and leadership passes to the local elite. This elite was employed in caliphal offices and assimilated to Fāṭimid Egyptian culture; they had remained mostly separate from the traditional networks nurtured by the geʾonim. For them, Hebrew was a necessary skill – a language for study or prayer – and a pleasurable diversion – a language of high literature –, but it was not the language they were accustomed to doing business in.

References Brody, Robert. 1998. The Geonim of Babylonia and the Shaping of Medieval Jewish Culture. New Haven and London: Yale University Press. Cohen, Mark. 2005. Poverty and Charity in the Jewish Community of Medieval Egypt. Princeton: Princeton University Press. Cowley, A. E. 1915. Notes on Hebrew Papyrus Fragments from Oxyrhynchus. The Journal of Egyptian Archaeology 2:4. 209–213. Drory, Rina. 1992. “Words beautifully put”: Hebrew versus Arabic in tenth-century Jewish literature. In Joshua Blau & Stefan C. Reif, Genizah Research after ninety years. The case of Judaeo-Arabic, 53–63. Cambridge: Cambridge University Press. Elbogen, Ismar 1993. Jewish Liturgy: a comprehensive history (translated by R. P. Scheindlin). Philadelphia–New York–Jerusalem: The Jewish Publication Society and the Jewish Theological Seminary of America. Gil, Moshe. 1983. Palestine during the first Muslim period. Tel Aviv: University of Tel Aviv. 3 volumes. [Hebrew] Gil, Moshe. 1997. Studies in Jewish history in Islamic lands in the early Middle Ages. Tel Aviv: University of Tel Aviv. 4 volumes. [Hebrew] Glickman, Mark. 2011. Sacred treasure – the Cairo genizah: the amazing discoveries of forgotten Jewish history in an Egyptian synagogue attic. Woodstock, Vermont: Jewish Lights Publishing. Goitein, S. D. 1967–1993. A Mediterranean Society: the Jewish communities of the Arab world as portrayed in the documents of the Cairo Geniza. Berkeley: University of California Press. 5 volumes with an index volume by P. Sanders. Goitein, S. D. 1973. Letters of medieval Jewish traders. Princeton: Princeton University Press. Hoffman, Adina & Peter Cole. 2011. Sacred Trash: the lost and found world of the Cairo Geniza. New York: Nextbook. Hoffman, Joel. 2004. In the beginning: a short history of the Hebrew language. New York: New York University Press. Horvath, Julia & Paul Wexler. 1994. Unspoken “languages” and the issue of genetic classification: the case of Hebrew. Linguistics 32. 241–269. Olszowy-Schlanger, Judith. 2003. Learning to read and write in medieval Egypt: children’s exercise books from the Cairo Geniza. Journal of Semitic Studies 48:1. 47–69.

198   

   Ben Outhwaite

Outhwaite, Ben. 2004. “In the language of the Hagri”: the Judaeo-Arabic letters of Solomon ben Judah. In Shulie Reif (ed.), The written word remains. The archive and the achievement – Articles in honour of Professor Stefan C. Reif, 52–69. Cambridge: Taylor-Schechter Genizah Research Unit at Cambridge University Library. Outhwaite, Ben. 2009. Byzantium and Byzantines in the Cairo Genizah: New and old sources. In Nicholas de Lange, Julia G. Krivoruchko and Cameron Boyd Taylor (eds.), Jewish Reception of Greek Bible Versions, 182–220. Tübingen: Mohr Siebeck. Outhwaite, Ben. 2012. “Clothed in glory and decked in splendor”: Medieval Hebrew since the discovery of the Cairo Genizah. British Association for Jewish Studies Bulletin. 3–8. Rabin, Chaim. 1970. Hebrew. In T.A. Sebeok (ed.), Current Trends in Linguistics. Volume 6: Linguistics in South West Asia and North Africa, 304–346. The Hague: Mouton. Rabin, Chaim. 2000. The development of the syntax of post-biblical Hebrew. Leiden: Brill. Reif, Stefan C. 2000. A Jewish archive from Old Cairo: the history of Cambridge University’s Genizah Collection. Richmond: Curzon. Sáenz-Badillos, Angel. 1993. A history of the Hebrew language (translated by John Elwolde). Cambridge: Cambridge University Press.

Geoffrey Khan

10 The historical development of early Arabic documentary formulae Sources: The documentary papyri analysed in this paper date from the first three centuries of the Islamic era. Most are from Egypt, although a small number are from sites in the Levant and Iraq, such as Damascus, Nessana (ʿAwjāʾ al-Ḥafīr) near Beʾershevaʿ, Khirbet al-Mird in the Judaean desert, and al-Sāmarrāʾ. To this number is added 33 documents from a private family archive in Khurasan (modern Afghanistan) in the early Abbasid period, and documents from a Pahlavi collection at Berkeley. Abstract: The formulaic structure of both Arabic legal documents and Arabic letters in the Umayyad period was brought by the Arabs to the conquered territories. Some features of the formulae can be shown to have parallels in pre-Islamic Semitic formula traditions. In the Abbasid period new formulae were introduced. The new legal formulae were developed by jurists, who sometimes drew on other legal traditions that existed in the Abbasid, formerly Sassanian, heartlands. The innovations in letter formulae appear to have been stimulated by the enhanced importance of court ceremonial protocol in the Abbasid period. Whereas in the Umayyad period the writer of a letter presented himself as remote from the recipient, in the new style introduced in the Abbasid period the writer presents himself as being in the virtual presence of the recipient. Some of the innovations of documentary formula that arose in the Abbasid heartlands appear earlier in documents from Khurasan (modern Afghanistan) than in those from Egypt.

1 The provenance of Arabic documents from the early Islamic period The vast majority of extant original documents from the early Islamic period have been found in Egypt. These date from the very beginning of the Arab settlement in Egypt in the first century of the Hijra (hereafter AH, the Muslim calendar beginning in 622 CE) and continue to be attested through the following centuries. In the first three Islamic centuries the documents are written on papyrus, the ancient writing material of Egypt. From the fourth century onwards papyrus was replaced as the common writing material in Egypt by paper, which had been originally introduced into the Islamic world in the eastern provinces. Although thousands

200   

   Geoffrey Khan

of Arabic papyrus documents have been preserved from the first three centuries AH, they are not evenly distributed across this period. By far the largest proportion of the extant papyri from Egypt are datable to the third century AH. A small number of Arabic documents on papyrus have been discovered at sites outside Egypt in the Levant and Iraq. These include papyri from Damascus,¹ Nessana (ʿAwjāʾ al-Ḥafīr) near Beʾershevaʿ,² Khirbet al-Mird in the Judaean desert,³ and al-Sāmarrāʾ.⁴ Some papyrus documents that have been discovered at sites in Egypt may, indeed, have originally been written elsewhere. This is the case, for example, with P.Khalili I 6, which is an account of expenditure of a Christian monastic community in Northern Syria or Iraq. Until recently, very little early Arabic documentary material had been discovered in the eastern Islamic world comparable to the Arabic papyri from Egypt. The only document available was an Arabic letter written on parchment from Central Asia datable to 100 AH/718–179 CE. This was discovered in 1933 in the ruins of a fortress on Mount Mūgh, situated in the valley of Zarafšān in Tajikistan (ancient Sogdiana).⁵ The early Arabic documentary material from the eastern extremities of the Islamic world has now been increased by the discovery a few years ago of Arabic documents from a private archive of a family resident in Khurasan in the early Abbasid period.⁶ They consist of thirty-two legal and administrative documents datable from 138/755 to 160/777 and a private letter from the same period. Place names mentioned in the documents indicate that they were written in a region between Bāmyān and Samangān in present day north-eastern Afghanistan. Like the document from Mount Mūgh, these newly discovered documents are on parchment. They were discovered together with a corpus of documents written in Bactrian, an Iranian language, which appear to have belonged to the same family archive.⁷ A few fragments of Arabic documents from the eastern provinces datable to the third century AH have recently come to light amongst a collection of Pahlavi documents at Berkeley.⁸ These are likely to have originated in Iran, as is the case with the Pahlavi documents. One of their notable features is that they are written on paper, whereas documents from the same period written in Egypt are

1 Abbott (1938). 2 Kraemer (1938). 3 Grohmann (1963). 4 Herzfeld (1912: pl. xxxvib). 5 This was published by Krachkovski and Krachkovskaya (1934). 6 The documents are published in Khan (2007a). 7 The Bactrian documents are published in Sims-Williams (2000) and (2007). 8 Khan (2007b).

The historical development of early Arabic documentary formulae   

   201

on papyrus, indicating that paper was in use as a writing material for documents in the eastern provinces earlier than in Egypt. This is in conformity with the statement of al-Jāhiẓ, writing in the 3rd century AH/9th century CE, that “the papyri of Egypt are for the West what the papers of Samarqand are for the East”.⁹

2 Legal documents The formulaic structure of early Arabic legal documents differs from that of the local traditions. The Arabic legal papyri from early Islamic Egypt have a different structure from that of the contemporary Byzantine Greek or Coptic documents. Distinctive elements of the Arabic documents include the predominate use of third person objective style in a “monumental” type of introductory formula consisting of a demonstrative pronoun referring to the document (‘This is a release …’, ‘This is what so-and-so bought …’) and the lack of autograph witness clauses. The Greek and Coptic documents, by contrast, generally use subject style and have autograph witness clauses. The early Arabic formularies are overall much simpler than the Greek and Coptic and lack many of the clauses that make the Greek and Coptic more legally watertight, such as warranty clauses or validity clauses. It is clear that the Arabic formulary is not based on the Greek or Coptic but rather was an independent tradition that was brought by the Arabs to the lands that they conquered. This is shown clearly in the case of a few bilingual ArabicGreek documents from the first century AH that are of a legal nature. One such document, dated 67 AH,¹⁰ is found at Nessana, and the Arabic version differs in structure from the Greek, but conforms to the patterns of other Arabic legal documents from the early Islamic period. The Arabic document, for example, closes with a list of names of witnesses without signatures whereas the Greek has an autograph witness clause.¹¹ Likewise the Arabic legal documents in the newly discovered corpus from Khurasan differ radically in structure from that of the contemporary legal documents written in Bactrian that emanate from the same family archive. It should be noted, however, that the Arabic legal documents from Khurasan resemble closely the structure of Arabic legal documents from Egypt. The only possible explana-

9 Quoted by al-Thaʿlābī (1867: 97) and al-Suyūtī (1881: 28). 10 Kraemer (1958: 156–160). 11 For further details see Khan (1994a).

202   

   Geoffrey Khan

tion for this is that the Arabs brought to the eastern provinces their own formulary tradition, which was ultimately of the same origin as that used in Egypt. The Arabic legal formulary tradition that was brought by the Arabs at the time of the conquests and used in the early Islamic period contains a number of elements that correspond to legal traditions that are known to have existed in the Near East in the pre-Islamic period. One example of this is the witness formula šahida fulān ʿalā nafsihi ‘he witnessed for himself’, which has parallels in Aramaic and Hebrew legal documents dating from the first half the first millennium CE from the Judaean desert and Dura Europos.¹² The Arabic term barāʾa ‘clearance (from legal claims)’ and the associated verbal forms, which are found in the earliest Arabic legal papyri from Egypt, correspond to cognate forms found in the Nabataean Aramaic documents from the Judaean desert datable to the first two centuries CE, e.g. wʾbryt ytky ‘and I have cleared you’, mn klʾ ktbt mbrʾ ytky ‘I have written clearing you from everything’, kḥlyqt mtntʾ wbrʾwnyʾ ‘according to the norm of gifts and quittances’ (Khan 1994a: 364). Parallels to certain components of the early Arabic formulary can be found also in Epigraphic South Arabian legal texts, such as the use of an initial demonstrative pronoun.¹³ The Arabic formulary of legal papyri in Egypt exhibits a watershed in their development between the second and third centuries AH. In the third century new formulaic patterns appear in the Arabic documents that have parallels in Byzantine Greek and Coptic documents. These include elements such as warranty clauses, validity clauses, formulae referring to legal ‘acknowledgements’ (ʾiqrārāt) and autograph witness clauses. These newly emerging elements in the Arabic documents were not, however, direct continuations of the local Egyptian tradition but rather were introduced into the legal formularies by Islamic jurists active in Iraq in the second century AH. These jurists derived certain elements from local legal traditions in Iraq that were of pre-Islamic origin. Although there is a similarity between these elements and the Byzantine legal formularies in use in Egypt, it does not necessarily follow that the jurists were drawing on a Greek tradition. Certain linguistic features of the newly introduced Arabic formulae indicate that their immediate origin was in an Aramaic legal tradition that was related to the Byzantine Greek tradition rather than directly in the Greek. An example of this is the use of the Arabic noun darak ‘overtaking by a claim’ and the associated verb ʾadraka ‘to overtake (a claim)’ to refer to claims of a third party in the formula of warranty clauses. This is likely to be an Arabic imitation of the Aramaic legal term ʾaḏraḵtā, which occurs in Jewish Talmudic law. The legal act

12 Khan (1994a: 364). 13 See, for example, the texts published by Stein (2010).

The historical development of early Arabic documentary formulae   

   203

of ʾaḏraḵtā granted a creditor the right to take possession of property pledged to him as surety by a debtor, if the debtor sold this property.¹⁴ Subsequent developments in Arabic legal formularies that appear in documents in later centuries can likewise be attributed to the influence of the formularies of the jurists. The formularies of the jurist al-Ṭaḥāwī (d. 321/933), for example, had a clear influence on those found in Arabic legal documents from Fāṭimid Fusṭāṭ that are preserved in the Cairo Genizah.¹⁵ There was a certain delay between the introduction of new formulaic elements by the jurists and their appearance in the legal documents. One important insight that we have from the Khurasan corpus of Arabic documents is that some features of the formularies that were developed by the Iraqi jurists in the second century appear in Khurasan earlier than in Egypt. This is seen, for example, in the case of the new warranty formula. In the extant texts of the formularies (šurūṭ) of the Iraqi jurists the term darak and the verb ʾadraka is first found in the formulary attributed to ʾAbū Ḥanīfa (d. 150/767) and ʾAbū Yūsuf Yaʿqūb (d. 182/798): fa-mā ʾadraka fulān ibn fulān fī ḏālika min darak fa-ʿalā fulān ibn fulān ḵalāṣ ḏālika ʾaw radd al-ṯaman ‘Whatever claim is made against so-and-so son of so-and-so, it is the duty of so-and-so son of so-and-so to clear that or return the price’.¹⁶ Warranty clauses containing such terminology are not found in the Arabic documents from Egypt before the third century AH. The earliest case I am aware of is P.Berl.Arab. 11 (276 AH, Fayyūm): fa-mā ʾadraka fulān ibn fulān fī hāḏā al-širāʾ min darak ʿulqa ʾaw tibāʿa li-ʾaḥad min al-nās … ‘Should any claim be made against so-and-so the son of so-and-so with respect to this purchase by way of attachment or right due to any person ….’ The terminology is, however, now attested over a century earlier in a legal document from the Khurasan corpus dated 145 AH: fa-mā ʾadrakaka min subul ʾIbrāhīm ʾaw ġayrihi fa-ʿalayya ḵalāṣuhu ‘Whatever liability overtakes you with regard to ʾIbrāhīm or anybody else – it is incumbent upon me to clear it’ (P.Khurasan 25). We may summarise the development of Arabic legal formularies as follows. The Arabs had a legal formulary tradition in the pre-Islamic period which they brought with them to various regions of the Middle East at the time of the Islamic conquests. The formularies underwent radical changes in the Abbasid period. These changes can be attributed in large measure to the activity of jurists based in Iraq. The changes first emerge in documents in the eastern provinces and only

14 Gulak (1939: 118) and (1926: 314–333). 15 Khan (1993: 51–55). 16 This has been preserved in al-Ṭaḥāwī’s Kitāb al-Šurūṭ al-Kabīr, edited by Wakin (1972: text 21).

204   

   Geoffrey Khan

later find their way into the documents further West in Egypt. It is relevant to add here that some innovations in legal formularies made in the East never reached Spain, in the far West of the Islamic world, where the legal formularies of the later Middle Ages (6th–9th centuries AH) continued to include archaic elements.¹⁷

3 Letters As with legal documents, there is a radical shift in the development of epistolary formulae written in Egypt between the second and third centuries AH. In the first two centuries, the opening formula of letters had the following structure: i. Basmala: bism illāh al-raḥmān al-raḥīm ‘In the name of God the Merciful and Compassionate’ ii. Address: min fulān li-fulān ‘From so-and-so to so-and-so’ or li-fulān min fulān ‘To so-and-so from so-and-so’, the person with the highest rank being placed in initial position. iii. Greeting: al-salām ʿalayka ‘Peace be upon you’ iv. Blessing: ʾaḥmadu ʾilayka allāh allaḏī lā ʾilāh ʾillā huwa ‘I praise for your sake God. There is no other god than He’. v. Formula marking transition to body of letter: ʾammā baʿd fa- … ‘As for after (it), then …’ This structure with the distinctive blessing formula ʾaḥmadu ʾilayka allāh allaḏī lā ʾilāh ʾillā huwa is attested in extant letters written in Egypt from the 90s of the first century AH through to the last quarter of the second century.¹⁸ Occasionally the greeting and blessing formulae are omitted, but the early letters are still distinguished from those of the third century by placing the address after the basmala.¹⁹ This epistolary formula is found also in letters that were written in the eastern provinces in the early Islamic period. These include the letter discovered in Central Asia published by Krachkovski and Krachkovskaya, which is datable to ca. 100 AH,²⁰ and a letter written in Khurasan datable to the middle of the second

17 See Khan (1993: 43–44). 18 The latest datable document I am aware of is PERF 624, a decree issued by the finance director of the governor ʿAbdallāh ibn al-Musayyab, who was in office from 176 to 177 AH. 19 See Khan (1992: 126–127) for details. 20 Krachkovski and Krachkovskaya (1934).

The historical development of early Arabic documentary formulae   

   205

century.²¹ The use of the same epistolary formula in Egypt and in the eastern peripheries of the Islamic world in the Umayyad and early Abbasid periods suggests that, as is the case with the legal formularies, the Arabs brought with them to the conquered territories their own letter formulae and did not take them from the practices local to the various places where they settled. This is demonstrated by the fact that the early Arabic epistolary formula differs from that of contemporary letters written according to the pre-Islamic local practice in other languages. In Egypt, for example, we have several letters written in Greek that are contemporary with the Arabic documents from the early Islamic period but exhibit several points of difference in their formula. Of particular importance in this respect is the correspondence of the governor Qurra ibn Šarīk who issued correspondence of a similar nature in both Arabic and Greek. The opening formulae of the two types of letter are as follows: (i) Arabic letters from Qurra ibn Šarīk: ‘In the name of God the Merciful and Compassionate. From Qurra ibn Šarīk to so-and-so. I praise God. There is no other god than He. As for after (it), then + body of letter’ (ii) Greek letters from Qurra ibn Šarīk: ‘In the name of God. Qurra ibn Šarīk, Governor, to so-and-so + body of letter’ As can be seen the Greek letters have the address after the initial invocation (equivalent to the basmala), but they lack anything that corresponds to the blessing or transition formulae. It has been shown above that some features of the early Arabic legal formularies correspond to elements that are found in other Semitic pre-Islamic traditions, e.g. the formula šahida fulān ʿalā nafsihi ‘he witnessed for himself’ and the term barāʾa ‘quittance’. One may expect, therefore, to find the same background to elements of the early Arabic letter formulae. Of particular significance in this respect is the transition formula that introduces the body of the letter. As has been shown, this is lacking in the Greek opening formula. Such transitional markers are, however, found in the formulae of Aramaic and Hebrew letters of the pre-Islamic period. These letters also have a greeting formula containing the word šlām and šālōm respectively, which corresponds to the formula containing the word salām occurring in some of the early Arabic letters.²²

21 This letter, which is in the Khalili collection, came to light after the administrative and legal documents from Khurasan had been prepared for publication. 22 For these transition markers see Alexander (1978) and Pardee et al. (1982).

206   

   Geoffrey Khan

Official documents from Egypt and Khurasan datable to the second century AH that relate to tax and are issued by the financial administrators (ʿummāl) of a governor (ʾamīr) open with the address and have the operative clauses in subjective style. In these respects, therefore, they resemble letters. They are distinguished from letters, however, by the lack of the characteristic blessing formula and, crucially, by the “monumental” style of opening with a demonstrative pronoun referring to the document (hāḏā kitāb min … ‘This is a document from …’), which reflects the fact that the document had the status of an legal instrument of proof. This type of documentary structure is found, for example, in agricultural leases, which are extant from the early Abbasid period.²³ In these documents the government official refers to himself in the first person and the lessee is referred to in the second person. A similar format is found in official documents from the second century, some from the late Umayyad period, that grant permission to travel for the purpose of finding work and paying tax in a different region.²⁴ These open hāḏā kitāb min and the financial administrator who issues the document refers to himself in the first person in the operative clauses. Of a similar structure are tax receipts issued in the early Abbasid period.²⁵ The early Arabic epistolary formula continues several decades into the Abbasid period in the second half of the second century. By the third century, however, letters written in Egypt have a completely different formula.²⁶ The distinctive features of this are the removal of the address from the text of the letter, the introduction of a new style of blessing formula and the lack of any systematic formula marking the transition to the main body of the letter. A typical opening structure is as follows: i. Basmala: bism illāh al-raḥmān al-raḥīm ‘In the name of God the Merciful and Compassionate’ ii. Blessing: ʾaṭāla allāh baqāʾaka wa-ʾadāma ʿizzaka wa-karāmataka wa-taʾyīdaka wa-saʿādataka wa-salāmataka … ‘May God prolong your life and cause to endure your strength, your honour, your support, your happiness and your health …’

23 Frantz-Murphy (2001: 22–23). 24 P.Cair.Arab. 174–175, Rāġib (1997). 25 This is the case with tax receipts in the Khurasan corpus written in the middle of the second century AH and “official” tax receipts from Egypt of the same period; cf. Khan (2007a: 27–28) and Frantz-Murphy (2001: 64–65). 26 For a detailed study of the formulaic structure of such letters see Grob (2010). For a discussion of the historical and social setting of the letters see Reinfandt (2010).

The historical development of early Arabic documentary formulae   

   207

In most private letters the address was removed altogether from the recto of the letter and only written on the verso, which was visible to the person delivering the letter when it was folded. In some cases it was not completely removed from the recto but written above the basmala outside the text of the letter.²⁷ The older practice of including the address in the text after the basmala is found only in a few documents from the third centuries, which are items of high-level official correspondence or petitions to high dignitaries. These include, for example, PERF 763 (dated 242 AH, ed. Grohmann 1952: 149), an official letter of appointment from the heir apparent al-Muntaṣir billāh, and P.Cair.Arab. 172 (edited by Grohmann 1952: 121), a petition to the caliph al-Muʿtazz billāh (252–255/866–869). The new type of epistolary formula occurs in the fragments of Arabic letters on paper from Iran that have been preserved in the Pahlavi archive at Berkeley. Radio-carbon dating of these Arabic fragments indicates that they were not written later than the third century AH. The use of paper makes it unlikely that they were written before the third century.²⁸ In the letter from Khurasan datable to the middle of the second century, on the other hand, the old epistolary formula is used. The blessing formula characteristic of the epistolary style that emerges in extant letters in the third century was used at that period to address recipients of all social ranks. It is found in letters addressed to ʾamīrs, e.g. P.Khalili I 16, and to rulers, e.g. the aforementioned petition to the caliph al-Muʿtazz billāh (252–255/866–869, P.Cair.Arab. 172, edited by Grohmann 1952: 121) and a letter addressed to Aḥmad ibn Ṭūlūn (254–270/868–884) in Egypt (Princeton Firestone Library No. 2002–136). The only distinctive structural feature of the blessings in such letters is that the high-ranking addressee is referred to in the third person (ʾaṭāla allāh baqāʾahu ‘may God prolong his life’) rather than the second person (ʾaṭāla allāh baqāʾaka), which is characteristic of letters written to people of lower rank. By the Fāṭimid period (10th–12th centuries), however, the blessing formula ʾaṭāla allāh baqāʾahu was not used in documents addressing rulers, but only to people of lower rank.²⁹ A new blessing formula was introduced for correspondence addressed to the Fāṭimid caliphs, which had the form: ṣalawāt allāh wa-barakātuhu wa-nawāmī zakawātihi wa-ʾafḍalu salāmihi wa-taḥiyyātihi ʿalā mawlānā wa-sayyidinā … ‘The benedictions of God and his blessings, his increasing benefactions and most excellent peace and greetings upon our master and

27 Khan (1992: 127, 141). 28 Khan (2007b); Azarpay et al. (2007). 29 Khan (1993: 310).

208   

   Geoffrey Khan

our lord ….’. This blessing and variants of it are found in many letters, including petitions and reports, that were addressed to Fāṭimid caliphs.³⁰ It was also used outside the written medium of documents as a prayer for the Fāṭimid ruler by Jews in their communal worship (Goitein 1982: 57). Another distinctive feature of letters written to Fāṭimid rulers is the obeisance formula al-mamlūk yuqabbil al-ʾarḍ ‘the slave kisses the ground’. This formula was introduced into petitions during the reign of al-ʾĀmir (1101–1130 CE). It is not found in petitions to earlier Fāṭimid caliphs.³¹ The introduction of the formula at the time of al-ʾĀmir appears to reflect a development in the protocol of the court ceremonial during his reign, whereby the custom of kissing the ground in the presence of the caliph was reintroduced after having been discontinued for some time. This is alluded to in the chronicle of the vizier Ibn al-Maʾmūn al-Baṭāʾiḥī: ʾijtamaʿ ʾumarāʾ al-dawla li-taqbīl al-ʾarḍ bayn yaday al-ḵalīfa al-ʾĀmir ʿalā al-ʿāda allatī qarrarahā mustajaddatan ‘The ʾamīrs of the state gathered to kiss the ground before the caliph al-ʾĀmir according to the custom that he had reestablished’.³² The entire set of opening formulae in letters to Fāṭimid caliphs, including the blessings and the obeisance formula, is likely to be a verbal expression of the court ceremonial that was performed during the audience with the ruler. The lack of address (‘to X from Y’) in the text of the letters is consistent with such an interpretation. Such an address would express separation between the sender and the recipient. Rather, the sender presents himself as being in the virtual presence of the caliph, offering blessings and obeisance as he would at an audience at court. The name of the sender is placed outside the text of the letter at the top of the sheet, which was referred to as the tarjama ‘heading’ in the medieval handbooks for secretaries.³³ The structure of the opening of letters to Fāṭimid caliphs, therefore, was as follows: (i) Name of sender (tarjama) (ii) basmala (iii) Blessing on caliph (iv) Obeisance formula (al-mamlūk yuqabbil al-ʾarḍ) Petitions addressed to viziers, who took de facto control in Egypt in the late Fāṭimid period (12th century CE), had a different blessing. This typically had

30 Khan (1993: 307–310). 31 Khan (1990: 24–26). 32 Fuʾād Sayyid (1983: 21). 33 Stern (1962: 190–191).

The historical development of early Arabic documentary formulae   

   209

the form ḵallada allāh taʿāla mulk al-majlis al-sāmī al-sayyidī al-ʾajallī al-juyūšī al-sayfī al-nāṣirī al-kāfilī al-hādī wa-ʾaḍada bihi al-dīn wa-ʾamtaʿa bi-ṭūli baqāʾihi ʾamīr al-muʾminīn wa-ʾadāma qudratahu wa-ʾaʿlā kalimatahu ‘May God, exalted is he, perpetuate the dominion of the lofty seat, the most excellent lord, commander of the armies, sword of Islam, the defender, the protector, the guide and support of the faith through him and allow the Commander of the Faithful to enjoy his long life and cause his power to endure and exalt his word’.³⁴ Such petitions addressed to viziers in the late Fāṭimid period otherwise have the same structural elements as are found in letters addressed to Fāṭimid caliphs, including the verbal obeisance formula al-mamlūk yuqabbil al-ʾarḍ. In the early Ayyūbid period (late 12th–early 13th centuries CE) this structure was maintained and is attested in extant petitions addressed to Saladin (1171–1192 CE) and al-ʿĀdil (1200–1218 CE). After al-ʿĀdil, however, petitions to rulers opened directly with the obeisance formula rather than an initial blessing on the ruler, a practice which continued into Mamluk times (1250–1517 CE).³⁵ A further development in the Ayyubid and Mamluk periods was the distribution of the usage of the obeisance formula. As remarked above, it was originally introduced to recreate court ceremonial in the presence of the ruler. After the Fāṭimid period, however, it begins to be used more widely in letters addressed to recipients of lower rank and is found in numerous private letters written in the Ayyubid and Mamluk periods. Its status was “downgraded” and it lost its original association with court ceremonial. It is possible that the type of epistolary formula that was introduced in the third century AH and replaced the early Arabic epistolary formula likewise had its background in court ceremonial. As we have seen, the type of letter formula that appears in the third century lacks the address in the text of the document, whereby the expression of physical remoteness between the sender and the addressee is removed. This differs from the early Arabic formula which, as in most Near Eastern parallels, makes this remoteness explicit by having the address in the text after the basmala. Of particular significance is that there are references in the historical sources to the use of the formula ʾaṭāla allāh baqāʾahu ‘May God prolong his life’ as a blessing on the Abbasid caliphs at court audiences. This is attested already for the caliph al-Manṣūr (136–158/754–775), who, according to Ibn al-ʾAṯīr (al-Kāmil fī al-Taʾrīḵ, ed. Cairo, 1886, part 6, p. 9), was addressed at court as ʾamīr al-muʾminīn ʾaṭāla allāh baqāʾahu. The letter formula with this blessing, therefore, may have originally been an imitation of court protocol, in which the

34 Richards (1973: 141). 35 Khan (1990: 26–30).

210   

   Geoffrey Khan

sender presents himself as being in the virtual presence of the addressee rather than being in a remote location. As we have seen, in the third century AH this blessing formula is used in letters addressed to rulers and also in those addressed to people of lower rank. By the Fāṭimid period, however, it had lost its association with court protocol, a new style of blessing being used at court, and was only used in letters addressed to recipients below the rank of ruler.

4 Script We have seen that a major shift occurs in the formulaic structure of both legal documents and letters written in Egypt between the second and third centuries AH. In the case of legal and administrative documents the Khurasan corpus now shows us that many of the innovations began in the eastern provinces and were transferred subsequently to Egypt. Their introduction into Egypt is likely to have been facilitated by the appointment of numerous high level Iranian administrators in the country in the Abbasid period (9th century CE). Members of some Iranian administrative families served for many generations in Egypt. In the third century AH Iranians were increasingly appointed as district and higher level governors in Egypt, offices that previously had been given generally to members of the caliph’s family. In 242 AH, in fact, a policy was introduced to exclude Arabs systematically from governorships.³⁶ We do not have documentary evidence for a similar delay between the introduction of the new epistolary formulae in the East and in Egypt. As remarked above, however, it is likely that the formula originated in Abbasid court ceremonial, which, if Ibn al-ʾAṯīr can be trusted, existed already at the time of al-Manṣūr in the middle of the second century AH. Concomitant with the shift in formulae in the third century in Egypt there is a shift in the style of script used in all types of documents. In the third century most documents begin to be written in a hand that is far more cursive than that of the first two centuries AH. The radical nature of this shift makes it unlikely that it developed in Egypt by natural evolution, but rather was introduced from outside. Evidence for this is now provided by the corpus of Khurasan documents from the second century. Many of the documents in this corpus, especially those of an administrative nature, exhibit a script that is more cursive than that of Arabic papyri from Egypt of the same period. It corresponds more closely to the cursive

36 Al-Kindī (1912: 202); Frantz-Murphy (2001: 81–83).

The historical development of early Arabic documentary formulae   

   211

type of script characteristic of the papyri from the third century AH onwards. It is probable, therefore, that the appearance of a more cursive script style in the papyri from the third century AH was another aspect of eastern administrative practice that was introduced into Egypt by officials trained in the eastern provinces. This, therefore, would explain the shift in documentary script type in the Arabic papyri. Many of the officials who drew up the administrative documents in the Khurasan corpus were of Iranian background. Terms used in the nomenclature of these officials such as al-’Iṣbahbaḏ (< Sasanian spāhbed) suggest that some of them were members of Iranian administrative families who could have been in state service over several generations. What may be of crucial significance is that these circles of Iranian administrators who produced the highly cursive Arabic documents in the early Abbasid period would have at a slightly earlier period been writing administrative documents in Pahlavi. The use of Pahlavi was in use in the eastern Islamic administration at least until 78/697, when, according to the historical sources, the official language changed to Arabic. Various Pahlavi administrative documents have been discovered in recent decades that were written in the early Islamic period. The dates proposed for these range from the 1st/7th to the 2nd/8th century,³⁷ suggesting that Pahlavi survived in administrative documents after 78/697. The most conspicuous feature of the script of the Pahlavi administrative documents is its advanced degree of cursiveness, which resulted in many of the Pahlavi letter shapes becoming similar in appearance. Angles are transformed into curves and curves into straight strokes.³⁸ One possible explanation for the development of similar cursive tendencies in the Arabic documentary script of the eastern Islamic empire, therefore, could be that they were introduced through the influence of a Pahlavi “substrate” by administrators who were trained in the Pahlavi administrative tradition. It should be noted that Pahlavi was not the local Iranian language of Khurasan in the early Abbasid period. The extant corpus of Bactrian documents indicates that the local population had a tradition of writing documents in the local Bactrian language. Administrators trained in Pahlavi are likely, therefore, to have come from Iran, nearer the Abbasid, and formerly Sasanian, administrative centre. This is, indeed, indicated by the form of some of the Iranian elements in the names of the administrators in the Arabic documents from Khurasan. The term al-ʾIṣbahbaḏ, for example, is derived from Middle Persian spāhbed ‘army-commander’ and differs from the local Bactrian

37 Gignoux (1991, 1996); Azarpay (2003). 38 Hansen (1938); de Menasce (1953, 1957); Weber (1973, 1983, 1984, 1992); Gignoux (1999).

212   

   Geoffrey Khan

form spālbid.³⁹ The seals with astral imagery used by these administrators in the Arabic documents are also not a local Bactrian tradition but rather are characteristic of iconography originating in the Sasanian heartlands.⁴⁰ The “eastern” innovation in Arabic script which had a radical impact on the documentary hand in Egypt did not have such a thorough-going influence on the Arabic script used in the Maġrib in the far West of the Islamic world, which retained many of the features of the early script down to modern times.

5 Concluding remarks The formulaic structure of both Arabic legal documents and Arabic letters in the Umayyad period were brought by the Arabs to the conquered territories. Certain features of the formulae can be shown to have parallels in pre-Islamic Semitic formula traditions, such as those in Aramaic, Hebrew and Epigraphic South Arabian. In the Abbasid period new formulae were introduced. The new legal formulae were developed by jurists, who sometimes drew on other legal traditions that existed in the Abbasid, formerly Sassanian, heartlands, in particular Aramaic traditions. The innovations in letter formulae appear to have been stimulated by the enhanced importance of the protocol of court ceremonial in the Abbasid period. Whereas in the Umayyad period the writer of a letter presented himself as remote from the recipient, in the new style introduced in the Abbasid period the writer presents himself as being in the virtual presence of the recipient. Some of the innovations of documentary formula that arose in the Abbasid heartlands appear earlier in Khurasan than in Egypt, indicating the close relationship of Khurasan with the administrative centre in the early Abbasid period.

6 References Abbott, Nabia. 1938. Arabic papyri of the reign of Ǧaʿfar al-Mutawakkil ʿalā allāh (AH 232–4/AD 847–61). Zeitschrift der Deutschen Morgenländischen Gesellschaft 92. 88–135. Alexander, Philip S. 1978. Remarks on Aramaic epistolography in the Persian period. Journal of Semitic Studies 23. 155–170. Al-Kindī. 1912. Kitāb al-Wulāt wa-Kitāb al-Quḍāt (ed. by Rhuvon Guest). Leiden: Brill.

39 Sims-William (1997: 5). 40 For the seals with astral images in the documents see Khan (2007a: 86–88). I am grateful to Judith Lerner for drawing my attention to the background of these seals.

The historical development of early Arabic documentary formulae   

   213

al-Suyūtī. 1881. Kitāb Ḥusn al-Muḥāḍara fī ʾAkhbār Miṣr w-al-Qāhira. Vol. 2. Bulaq. al-Thaʿlābī. 1867. Laṭāʾif al-Maʿārif (ed. by Pieter de Jong). Leiden: Brill. Azarpay, Guitty. 2003. Rare Pahlavi texts now at Bancroft. Bancroftiana, Newsletter of the Friends of the Bancroft Library 123. Azarpay, Guitty, Kathleen Martin, Martin Schawrtz & Dieter Weber. 2007. New information on the date and function of the MP Berkeley Archive. Bulletin of the Asia Institute 10. 17–29. De Menasce, Jean. 1953. Recherches de papyrologie pehlevie. Journal Asiatique 241. 185–196. De Menasce, Jean. 1957. Ostraca and papyri. Corpus Inscriptionum Iranicarum Part III. Pahlavi Inscriptions. Volume IV–V. Ostraca and Papyri. London: Percy Lund, Humphries & Co. Frantz-Murphy, Gladys. 2001. Arabic agricultural leases and tax receipts from Egypt 148–427/756–1035. Corpus Papyrorum Raineri XVII. Vienna: Hollinek. Fuʾād Sayyid, Ayman. 1983. Passages de la Chronique d’Égypte d’Ibn al-Maʾmūn. Cairo: Institut français d’archéologie orientale. Gignoux, Philippe. 1991. Une nouvelle collection de documents en pehlevi cursif du début de septième siècle de notre ère. Comptes Rendus de l’Académie des Inscriptions et BellesLettres 135 Nov–Dec. 683–700. Gignoux, Philippe. 1996. Six documents Pehlevis sur cuir du California Museum of Ancient Art. Bulletin of the Asia Institute 10. 63–72. Gignoux, Philippe. 1999. Ostraca, papyri et parchemins. Empires perses, d’alexandre aux sassanides, Dossiers d’archeologie 243. 24–25. Goitein, Shelomo Dov. 1982. Prayers from the Geniza for Fatimid caliphs, the head of the Jerusalem Yeshiva, the Jewish community and the local congregation. In S.R. Brunswick (ed.), Studies in Judaica, Karaitica and Islamica presented to Leon Nemoy on his Eightieth Birthday, 47–57. Ramat Gan: Bar Ilan University Press. Grob, Eva. 2010. Documentary Arabic private and business letters on papyrus: Form and function, content and context. Berlin/New York: de Gruyter. Grohmann, Adolf. 1952. From the world of Arabic papyri. Cairo: al-Maaref Press. Grohmann, Adolf. 1963. Arabic papyri from Ḫirbet el-Mird. Louvain: Publications Universitaires. Gulak, Asher. 1926. ʾOṣar ha-Šṭarot ha-Nehugim b-Yisraʾel [Collection of legal documents customary in Israel]. Jerusalem [in Hebrew]. Gulak, Asher. 1939. History of Jewish law. Talmudic period. Vol. 1: Law of Obligation and its guarantees. Jerusalem [in Hebrew]. Hansen, Olav. 1938. Die mittelpersischen Papyri der Papyrussammlung der Staatlichen Museen zu Berlin. Berlin: de Gruyter. Herzfeld, Ernst. 1912 . Erster vorläufiger Bericht über die Ausgrabungen von Samarra. Berlin: Dietrich Reimer/Ernst Vohsen. Khan, Geoffrey. 1990. The historical development of the structure of medieval Arabic petitions. Bulletin of the School of Oriental and African Studies 53. 8–30. Khan, Geoffrey. 1992. Arabic papyri. Selected Material from the Khalili Collection. Oxford: Oxford University Press. Khan, Geoffrey. 1993. Arabic legal and administrative documents in the Cambridge Genizah Collections. Cambridge: Cambridge University Press. Khan, Geoffrey. 1994a. The pre-Islamic background of Muslim legal formularies. ARAM 6. 193–224. Khan, Geoffrey. 1994b. An Arabic legal document from the Umayyad period. The Journal of the Royal Asiatic Society 3rd series 4. 357–368.

214   

   Geoffrey Khan

Khan, Geoffrey. 2007a. Arabic documents from Early Islamic Khurasan. London: Nour Foundation in association with Azimuth editions. Khan, Geoffrey. 2007b. The Arabic paper fragments from Berkeley. Bulletin of the Asia Institute 17. 31–34. Krachkovski, I.Y. & V.A. Krachkovskaya. 1934. Le plus ancien document arabe de l’Asie Centrale. In Sogdiĭskĭ Sbornik. 52–90, Leningrad. Reprinted in I. Y. Krachkovski. 1955. Izbrannye Socineniya 1. 182–212. Moscow-Leningrad. Kraemer, C.J. 1938. The Colt papyri from Palestine. In Actes du Ve Congrès International de Papyrologie, 238–244. Brussels. Kraemer, C.J. 1958. Excavations from Nessana. Volume 3. Non-literary papyri. Princeton: Princeton University Press. P.Berl.Arab. = Abel, Ludwig (ed.). 1896–1900. Ägyptische Urkunden aus den Königlichen Museen zu Berlin. Arabische Urkunden. Berlin: Weidmann. P.Cair.Arab. = Grohmann, Adolf (ed.). 1934–1962. Arabic Papyri in the Egyptian Library. Cairo: Egyptian Library Press. P.Khalili I = Khan (1992) Pardee, Dennis. 1982. Handbook of ancient Hebrew letters: A study edition. With a chapter on Tannaitic letter fragments by S.D. Sperling, with the collaboration of J. D. Whitehead and P. E. Dion. Chico: Scholars Press. PERF = Karabaček, Josef von, Papyrus Erzherzog Rainer. Führer durch die Ausstellung. 1896–1900. Vienna: collection’s own publication. Rāġib, Yūsuf. 1997. Sauf-conduits d’Égypte Omeyyade et Abbasside. Annales Islamologiques 31. 143–168. Reinfandt, Lucian. 2010. Arabisches Briefwesen. In Claudia Kreuzsaler, Bernhard Palmer and Angelika Zdiarsky (eds.), Stimmen aus dem Wüstensand. Briefkultur im griechischrömischen Ägypten, 115–179. Wien: Phoibos Verlag. Richards, D.S. 1973. A Fāṭimid petition and “small decree” from Sinai. Israel Oriental Studies 3. 140–158. Sims-William, Nicholas. 1997. Four Bactrian economic documents. Bulletin of the Asia Institute. New Series 11. 3–15. Sims-William, Nicholas. 2000. Bactrian documents from Northern Afghanistan I: Legal and economic documents. Oxford: Nour Foundation in association with Azimuth Editions and Oxford University Press. Sims-William, Nicholas. 2007. Bactrian documents from Northern Afghanistan II: Letters and Buddhist texts. London: Nour Foundation in association with Azimuth Editions. Stein, Peter. 2010. Die altsüdarabischen Minuskelinschriften auf Holzstäbchen aus der Bayerischen Staatsbibliothek in München. Bd. 1. Die Inschriften der mittel- und spätsabäischen Periode. (Epigraphische Forschungen auf der Arabischen Halbinsel 5). Tübingen: Wasmuth. Stern, S.M. 1962. Three petitions of the Fāṭimid period. Oriens 15. 172–209. Wakin, Jeanette A. 1972. The function of documents in Islamic law: The chapters on aale from Ṭaḥāwī’s Kitāb al-Shurūṭ al-Kabīr. Albany, New York: State University of New York Press. Weber, Dieter. 1973. Einige Beobachtungen an Pahlavi Papyri. Acta Orientalia 35. 83–88. Weber, Dieter. 1983. Die Pehlevifragmente der Papyrussammlung der Österreichischen Nationalbibliothek in Wien. In Papyrus Erzherzog Rainer (P. Rainer Cent.). Festschrift zum 100-jährigen Bestehen der Papyrussammlung der österreichischen Nationalbibliothek, 215–228. Vienna: Hollinek.

The historical development of early Arabic documentary formulae   

   215

Weber, Dieter. 1984. Pahlavi, Papyri und Ostraca: Stand der Forschung. In Woijciech Skalmowski & Alois Van Tongerloo (eds.), Middle Iranian Studies, Proceedings of the International Symposium organized by the Katholieke Universiteit Leuven 1982, 25–43. Leuven: Peeters. Weber, Dieter. 1992. Ostraca, Papyri und Pergamente. Corpus Inscriptionum Iranicarum. Part III. Pahlavi Inscriptions. Volume IV–V. Ostraca and Papyri. London: on behalf of Corpus Inscriptionum Iranicarum by School of Oriental and African Studies.

Nicholas Zair

11 Individualism in “Osco-Greek” orthography¹ Sources: Oscan inscriptions written in the Greek alphabet are datable from the 4th to the 1st centuries BCE. Counting generously, we have around a hundred, of which the vast majority come from ancient Lucania and Bruttium in Italy (modern day Basilicata and Calabria), with perhaps half-a-dozen others from a little further north, and about the same number from Messina in Sicily. Most were found by excavation, and are in museums in Italy. The inscriptions are found on a range of materials: the majority are engraved in stone, but we also find them on lead and bronze tablets, bronze helmets, terracotta objects (especially pots), coins, and stamped on tiles. Although the names of those responsible for dedications or public works are sometimes included in the text, this is seldom very informative without further historical information. Furthermore, it is often unclear to what extent the instigator of the inscription was responsible for the spelling of the text, since texts on stone were presumably written by specialist stonemasons, and professional scribes may have written other texts. Abstract: One of the alphabets used to write Oscan, a language of ancient Italy, was a slightly modified version of the Greek alphabet. It has usually been assumed that the alphabet and orthography used in the inscriptions in the Greek alphabet from Southern Italy can be attributed to one or more specifically “OscoGreek” scribal schools or orthographic traditions, with relatively little variation in orthographic conventions. In this article I use evidence from our corpus of Oscan inscriptions to argue that a) in fact there is more variation than has commonly been acknowledged, and b) we should be open to the idea that individual writers of Oscan, at least some of whom are likely to have been fluent and literate in Greek, may have taken different approaches to writing Oscan using the Greek alphabet. This might include being literate in Greek, but not Oscan; not recognising that “Osco-Greek” had a separate alphabet/orthography; or sometimes choosing to follow “Greek” rather than “Osco-Greek” conventions (if these in fact existed).

1 I am grateful to James Clackson and Katherine McDonald for their comments on an earlier version of this article. Mistakes and flaws are naturally my responsibility.

218   

   Nicholas Zair

Oscan is one of a group of related languages, known as the Sabellic (or Sabellian) languages, which were spoken in the Italian peninsula in the 1st millennium BCE. The Sabellic languages are part of the Indo-European language family, and probably formed an intermediate family (Proto-Italic) with Latin and Faliscan. Apart from a very small number of words glossed in Latin and Greek works, all our evidence for the Sabellic languages comes from inscriptional evidence of one type or another. Apart from Oscan, the best attested Sabellic languages are Umbrian, whose major remains are seven bronze tablets, written between the 3rd and 1st centuries BCE, recording the rituals and regulation of a brotherhood of priests, and South Picene, of which we have about 25 inscriptions dating from the 6th to the 4th century BCE. A handful of other Sabellic languages are attested in small numbers of inscriptions over Central and Southern Italy. In terms of numbers of inscriptions, Oscan is by far the best attested of the Sabellic languages, with around 400; more continue to be discovered. Our evidence for Oscan dates from the 4th century to the 1st century BCE; some doubtful examples may come from the 1st century CE. It was spoken from Campania to Calabria, with a few inscriptions from Messina in Sicily (see footnote 7).² Italy in the first millennium BCE was a centre of transmission and adaptation for alphabets,³ all of which can be traced back ultimately to the Greek alphabet brought by colonists as early as the 8th century. Oscan in particular is found written in at least four different alphabets: the Etruscan alphabet,⁴ the Latin alphabet,⁵ the Oscan (or “National” or “Native”) alphabet,⁶ and the Greek alphabet, which is the focus of my discussion here.

2 For introductions to the Italic group and the Sabellic languages, with further references, see Fortson (2010: 274–308); Clackson and Horrocks (2007: 37–76); Wallace (2007). 3 In this article I will use “alphabet” to mean a collection of signs (“letters”) which are used to represent sounds or groups of sounds (thus e.g. is a letter which reflects the sounds /ks/). Graphemes, which represent a letter without necessarily reflecting the actual shape of the attested sign, are enclosed within (thus e.g. is used to represent the sign which usually appears as either H or Ͱ). By “orthography” is meant the actual use of these letters in spelling words. Alphabets and orthography can be used to write phonemically and phonetically (subphonemically); phonemes are encased within / /, phones between [ ]. 4 Principally in a small group of 4th-century inscriptions found around ancient Saticula (near Sant’ Agata dei Goti in the province of Benevento). 5 Most importantly a long legal inscription from Bantia in the early 1st century (the Tabula Bantina); a few other late inscriptions in the Latin alphabet are found scattered over the Oscanspeaking area. The so-called “North-Oscan” languages are largely found written in the Latin alphabet, from as early as the 3rd century. 6 The Oscan alphabet was primarily derived from the Etruscan alphabet, as is demonstrated by its letter-forms, its use of for /f/, and its lack of , but with a Greek alphabet as an acces-

Individualism in “Osco-Greek” orthography   

   219

The Oscan alphabet in particular has long been seen as the model of an alphabet that has developed strict orthographic norms, and in which it is surprisingly difficult to find much in the way of variation over a fairly large area of central Italy. The same approach has tended to be taken to Oscan written further south, in the regions of Lucania and Bruttium (roughly modern day Basilicata and Calabria).⁷ Here the Greek alphabet is used, whose orthographic conventions have been seen as fairly fixed across our corpus of around 85 inscriptions; differences in the systems used by the various inscriptions are viewed as due to changes over time, or the existence of separate scribal schools. This is owed to the pioneering work of Lejeune (1970 and 1972) on the orthography of the “OscoGreek” inscriptions, who explicitly talks in terms of an “Osco-Greek” alphabet, “traditions graphiques” and “écoles orthographiques” (e.g. 1970: 272, 276; 1972: 10). According to Lejeune, apparent exceptions to the rules he discovered (which will be discussed further below) are because the writers of these inscriptions belonged to other scribal schools, which themselves presumably also had fixed, albeit different, orthographical rules. A problem with this approach is that over time, with new discoveries, the number of exceptions to Lejeune’s system has increased as a proportion of our relatively small corpus of inscriptions, and there seems to be a limit to the number of scribal schools that could reasonably be posited. I would like to consider the alternative possibility that at least some of our inscriptions are written by people who are not following any kind of externally mandated orthographic rules, but are instead simply using their intuitions as to the best way to spell a given sound in a given context. In the case of the “OscoGreek” inscriptions, the most plausible influence on the writers’ intuitions about how best to spell Oscan is likely to have been the orthography of the Greek alphabet used to write Greek. As a demonstration of the sort of process I envisage, I would like first to consider an inscription written outside Lucania and Bruttium, from Teano in

sory, which led to the retention of a value /g/ for the letter , which has the Etruscan shape, but not its value /k/, and the reintroduction of the letters and (Lejeune 1957: 100; Crawford et al. 2011: 19–22). In its original form, the alphabet had only the four vowel symbols , , and . In the 3rd century BCE, two new vowel signs were created by adding diacritical marks to and to give and . 7 A handful of inscriptions written in the Greek alphabet are known from Messina in Sicily; these are often included in discussions of the use of “Osco-Greek” orthography, but I do not include them here because I suspect that they are heavily influenced by the norms of the Oscan alphabet (Messina was conquered by Oscans from Campania early in the 3rd century BCE). I intend to discuss this at greater length elsewhere.

220   

   Nicholas Zair

Campania (Teanum Sidicinum 25).⁸ This inscription is dated to around 300 BCE, and consists of the words πλατωρ ουψε ‘Plator made this’ scratched around the rim of a plate. It seems likely that this is an example of Oscan being written by someone who was primarily used to writing Greek. We know that Plator spoke Greek because we have another signature of his which says πλατωρ εποι[η]σε, Greek for ‘Plator made this’. Plator’s name is in fact Messapic, so it is possible that Greek may also not have been his native language; but he was probably much more fluent in Greek than in Oscan, since the Oscan 3rd sg. perfect verbal ending /-εd/, which we would expect to be written as -εδ, has been replaced with the Greek 3rd sg. aorist ending /-e/ in ουψε. Plator’s Oscan inscription is exceptional in that the usual way of writing Oscan in Campania was to use the Oscan alphabet rather than the Greek alphabet. It is possible that Plator had learnt to write Oscan according to “OscoGreek” norms (insofar as these existed; see below) in Lucania or Bruttium prior to moving to Campania; but a perhaps more plausible hypothesis, especially given his imperfect Oscan, is that Plator, not being literate in the Oscan alphabet, was simply applying Greek letters and Greek orthography to the writing of Oscan, rather than using an orthography specially developed for writing Oscan using the Greek alphabet. Thus, for the Oscan sound /u:/ in the first syllable of /u:psεd/, Plator used the grapheme for the sound nearest to /u:/ in Greek, which was written .⁹ For the sequence /ps/ he naturally used the symbol /ps/. As already noted, this sort of haphazard use of the Greek alphabet to write Oscan is quite different from the traditional view of the use of the Greek alphabet further south. However, any attempt to describe the use of the Greek alphabet to write Oscan must take into account the possibility that at least some writers of Oscan in the south were doing so in a manner similar to that perhaps represented by Plator, in other words adapting the orthography familiar to them from Greek in a more-or-less ad hoc manner in order to write Oscan. We are able to identify this possibility in the inscription by Plator because he is a non-native speaker of Oscan making language-learner errors, and because he writes in the Greek alphabet in an area of Italy where Oscan is usually written in the Oscan alphabet. In the south of Italy, where almost all Oscan inscriptions are written in the Greek

8 Inscriptions in this article are referred to first according to the classification of Crawford et al. (2011), secondarily according to that of Rix (2002): e.g., Potentia 1/Lu 5. Some inscriptions that are included in Crawford et al. (2011) are not in Rix (2002). 9 It is not clear exactly whether in Greek represented [o:] or [u:] (Threatte 1980: 239); it is also possible that Oscan /u(:)/ may have been phonetically closer to [o(:)] than to [u(:)]. Greek reflected /y(:)/ rather than /u(:)/.

Individualism in “Osco-Greek” orthography   

   221

alphabet, mostly by first- or extremely competent second-language speakers, our ability to recognise Plator-style orthography and to distinguish it from a formal “Osco-Greek” orthography will be extremely limited. There is considerable evidence for Greek-Oscan bilingualism in our corpus, as demonstrated by inscriptions including: Buxentum 3/Lu 45, a lead tablet with a Greek sentence followed by a list of names using Oscan morphology; Crimisa 1/ Lu 23, a text inscribed on stone beginning σακαρακιδιμαι ‘in the priesthood’, with, below it and upside down, in Greek επι ιερ(εως) ‘in the priesthood’; Petelia 2, a lead curse tablet consisting of a list of Oscan names with (largely) Oscan morphology, followed by a clause in Oscan and then two lines of Greek;¹⁰ Laos 2/Lu 46, a lead curse tablet, in which the Oscan names with Oscan morphology in the accusative have final -ν as in Greek, rather than -μ as in Oscan; Potentia 39, where, in an otherwise totally Greek inscription, we find an Oscan dating formula. As a result, it seems quite possible that some of our inscriptions in the Greek alphabet from the South of Italy may have been written by people who had learned the Greek alphabet and Greek orthography primarily for the purposes of writing Greek, and applied this knowledge on an individual – and hence perhaps idiosyncratic – basis to writing Oscan. To some extent, this possibility has already been noted by Poccetti (1988: 141–158), who observes that “official/public” inscriptions in the area of the Brettii are written in Greek rather than Oscan, and argues that consequently there was less impetus for the creation of orthographic norms as found in the Lucanian area, and greater influence from Greek orthography on the writing of Oscan in Bruttium. However, it does not seem to me that we should be so quick to contrast the less rigid and more “Hellenised” practice further south with that of Lucania, not least because, contrary to Poccetti’s and Lejeune’s assumptions, we find considerable variation in the use of the Greek alphabet to write Oscan in Lucania too. In the light of Poccetti’s distinction between “public/official” = consistent and “private” = inconsistent orthography, we might attempt to explain such variation in Lucania in the same way, so that e.g. in the lead curse tablets mentioned above we might find more variation/influence of Greek norms, while a fixed “OscoGreek” orthography would have been developed for “official” inscriptions. Apart from the fact that this risks reducing our corpus into groups of a size so small as to be meaningless, it quickly becomes clear that this is not the case. For example, in the evidently “official” law code from Roccagloriosa (Buxentum 1/Lu 62) /e/ is spelled , and (see below), and uses both the signs H and Ͱ; and

10 Arbabzadah (2007) provides a thorough discussion of questions regarding Greek-Oscan bilingualism in this inscription.

222   

   Nicholas Zair

Potentia 1/Lu 5, which commemorates the erection of bronze statues by decree of the senate by Heirens Pomponis, shows the same variation in the spelling of /e/, spells /o/ with both and , uses both and for /f/, and writes final /-d/ with both and . This does not particularly point us in the direction of a settled “official” orthography for Oscan in the Greek alphabet.¹¹ I would like to raise the possibility that there may have been much more individual variation – even in the “official/public” sphere – than has hitherto been recognised. I do not wish to claim that there were never areas, groups or individuals which developed their own consistent orthography, but I do want to emphasise the large amount of variation in the small corpus available to us, and that Greek orthography may have influenced how individuals came to write Oscan in the Greek alphabet. The bedrock of the idea that there was such a thing as a systematic Oscan orthography in the Greek alphabet is Lejeune’s (1970: 288–305, 1972: 9–10) picture of the representation of Oscan vowels. A brief sketch of Lejeune’s idea is necessary here. Oscan had seven vowels /y/, /i/, /e/, /ε/, /a/, /o/, /u/; according to Lejeune, with only a couple of exceptions, inscriptions in the Greek alphabet originally represented /i/ by , while had to do service for both /e/ and /ε/. Around 300 BCE, under the influence of the Ionic version of the Greek alphabet, the use of for /e/ was introduced, resulting in the system /i/ , /e/ , /ε/ . Oscan also had a diphthong /ej/, which had been written before 300 BCE. After the adoption of for /e/, a digraph was introduced to write /ej/, thus avoiding confusion with /e/. A similar situation pertained in the back vowels, where /u/ was originally written , but after 300 BCE was written , while /o/ was written . In order to avoid confusion, the diphthong /ow/, originally written , adopted the spelling or .¹² Such a rigid system, with its identifiable turning point around 300 BCE, would be an extremely strong argument for the existence of an orthography specifically for writing Oscan, as opposed to Greek, in the Greek alphabet. However, as will already be obvious from the examples given above, there are many problems involved in positing such a system. Some of these were evident at the time and were explained away by Lejeune as reflecting an intermediate stage of uncertainty around 300 BCE, or reflecting late changes in the vocalism of Oscan, or by the existence of different scribal traditions; others have arisen with the discovery of new inscriptions

11 Buxentum 1 is dated to 300–200 BCE, while Potentia 1/Lu 5 is from 125–100 BCE, so inconsistency of spelling in “official” inscriptions cannot be deemed to be an “early” or “late” feature. 12 In certain circumstances was also used for /o/, but the details do not concern us here.

Individualism in “Osco-Greek” orthography   

   223

over the last forty years. Thus, there are several examples of for /e/ and for /u/,¹³ and several inscriptions which use more than one way of representing /e/ or /u/ inconsistently.¹⁴ This variation in spelling within and between inscriptions is found throughout the time-span of our “Osco-Greek” inscriptional evidence. Consequently, the evidence that such a system as proposed by Lejeune ever existed seems extremely shaky. A particularly interesting case study is the bronze law tablet from Roccagloriosa (Buxentum 1/Lu 62), which is dated to between 300–200 BCE. In terms of its orthography for representing vowels, it fits Lejeune’s system particularly badly, since, in addition to using , and for /e/, it consistently uses for both /u/ and /ow/ (Gualtieri and Poccetti 2001: 200–205). But another aspect of its orthography is also worthy of attention in the context of the present discussion, which is its spelling of the sequence [ŋg]. This sound occurs in our corpus primarily in the common word τανγινοδ/τανγινοτ (abl. sg.) ‘decree’ (Atina Lucana 1/ Lu 2, Cosilinum 1/Lu 3, Potentia 1/Lu 5, Potentia 9/Lu 6, Potentia 10/Lu 7). The spelling with may suggest that some people at least had developed a specifically Oscan orthographic practice, since in Greek the standard spelling of this sequence was (although the spelling was always possible)¹⁵. In the Roccagloriosa tablet, however, we find the word spelled (τ)αναγγινουδ (side B, line 1) – apparently because the scribe started to write τανγινουδ, before “correcting” himself and using the spelling (Gualtieri and Poccetti 2001: 233). Consequently, it looks as though, at least in this instance, the scribe of Buxentum 1 was actively choosing to use a “Greek” spelling in an Oscan inscription. It is not clear how to interpret the first mistaken attempt to write . It is possible that the scribe started to write using an “Osco-Greek” convention of spelling [ŋg] as and then preferred to use “Greek” . Or the scribe could have been using “Greek” orthography in both instances, merely replacing substandard with standard . If the former applies, this may be an instance not of Plator-style use of “Greek” orthography to write Oscan, but an active preference in the habits of one scribe for “Greek” orthography in place of “Osco-Greek” orthography. It is striking that almost the reverse pattern of usage of “Greek” vs “Osco-Greek” spelling habits is

13 For example, μεδικιαι /mεddekkjaj/ (Metapontum 1/Lu 37), νανονις /nanunis/ (Potentia 10/ Lu 7). 14 In addition to Buxentum 1 and Potentia 1, e.g. Potentia 40/Lu 13, which has μεδδικεν /mεddekkjen/. I intend to write about the representation of Oscan vowels in the Greek alphabet at greater length elsewhere. 15 It is found from the 5th to the 1st century BCE in inscriptions from Attica (Threatte 1980: 597–601), and in Italy in the 4th century Tabulae Heracleenses from Lucania (see below).

224   

   Nicholas Zair

found for the sequence /ps/, which is almost always spelled with in Oscan inscriptions, but is found once spelled in Potentia 1/Lu 5 (ω)πσανω.¹⁶ This suggests that for some sounds or sequences of sounds “Greek” orthography may have been preferred by most writers, while for others “Osco-Greek” orthography was preferred. Scare-quotes have been liberally bestowed on the terms “Osco-Greek” and “Greek” in the preceding paragraphs, because I would like to suggest that for at least some writers there may not have been a clear difference between the two. On the face of it, it is difficult to claim that the Oscan inscriptions were written just using the Greek alphabet of the time, not least because the “Osco-Greek” alphabet used to write Oscan seems to be distinguished from the “Greek” alphabet used to write Greek by the presence of the two letters /w/ (digamma) and /f/.¹⁷ But, I would argue, this is not necessarily how the situation was seen at the time, at least by everyone. Firstly, although digamma is not a part of the Ionic Greek alphabet per se, it is found in Doric Greek inscriptions from the South of Italy which otherwise use the Ionic alphabet, most notably in the late 4th century Tabulae Heracleenses from Herakleia in Lucania (IG XIV, no. 165). So at least in the 4th century use of digamma would not have been a distinguishing feature of Oscan rather than Greek in the Greek alphabet. Even after ceased to be used to write Greek in Italy, it may have remained in the local Ionic abecedary as a “dead letter”. It is well known that “dead” letters can remain in abecedaries for a long time after they have stopped being used (or were never used) to write a given language (e.g. in Etruscan; Wallace 2008: 17–19). It therefore seems to me quite possible that (some) literate Oscan-Greek bilinguals considered there to be a single alphabet containing the letters and , in which the latter happened usually only to be used when writing Oscan (and were usually “dead” from the point of view of writing Greek); however, sometimes “Oscan” seems to have “leaked” into “Greek” spelling. An early example of this is found on a drachm from the Ionic colony of Velia in Lucania from between 425 and 420 which has on it ϜΕΛH (either ‘Velia’ or short for ‘Υελήτων ‘of the Velians’), with “Oscan” (or at least non-Ionic) but “Greek” for Ionic /ε̄/ at the end of the word (Dubois 1995: 138; Wil-

16 According to Poccetti (1988: 156–7), the use of is Brettian, whereas is used in Lucania. However, (ω)πσανω in Potentia 1/Lu 5 is actually the only example of in Lucania, and is found also in the following Lucanian inscriptions: Laos 1, Laos 2/Lu 46 (admittedly in an inscription with Hellenised morphology), Heraclea 2/Lu 61, Potentia 20/Lu 28, Potentia 21/ Lu 29, and Potentia 26/Lu 20 (in a very broken context). is also found in the inscriptions from Messina, but in this case it may be down to influence from the Oscan alphabet. 17 Apparently was originally borrowed from the Oscan alphabet.

Individualism in “Osco-Greek” orthography   

   225

liams 1992: 42, plate X no. 20).¹⁸ Something similar has happened in the spelling of Greek δύο ‘two’ as δυϝο in Buxentum 3/Lu 45 with the “Oscan” spelling of the subphonemic glide in δύο [duwo]. I did not intend to discuss here all possible examples of specifically Oscan orthographic conventions, or spellings that may instead be due to influence from Greek orthographical practices; but I hope to have shown that there are several possible ways in which individuals may have treated the writing of Oscan in the Greek alphabet, which may want to make us re-think our ideas about “OscoGreek” orthography. Thus, some writers of Oscan may only have been literate in Greek, and, perhaps like Plator, have written Oscan on the basis purely of their knowledge of Greek orthography; others, perhaps like the scribe of the Roccagloriosa tablet, may have been aware of the existence of an “Osco-Greek” orthography, but preferred, at least some of the time, to adopt Greek spelling conventions. Yet others, like the writer of Buxentum 3, may not have distinguished between Oscan and Greek orthography at all, using the same rules (which may reflect originally “Osco-Greek” or “Greek” conventions) to write both. Given the sociolinguistic situation in southern Italy in the second half of the 1st millennium BCE, and our relatively small corpus of inscriptions, it seems to me that the default position of those studying “Osco-Greek” orthography should be to be more open to the existence of considerable variation at the micro level; the existence of “scribal schools” in any strict sense cannot be simply assumed, but must be proved. My own view is that individual decisions regarding how to write Oscan are more important in this context than has hitherto been emphasised.

References Arbabzadah, Moreed. 2007. Linguistic studies on curse tablets from the ancient world. An examination of the Petelia Tablet and its implications for the study of Greek-Oscan bilingualism. Cambridge: University of Cambridge MPhil thesis. Clackson, James & Geoffrey Horrocks. 2007. The Blackwell history of the Latin Language. Oxford: Blackwell.

18 Another drachm has ϜΕΛΙΕ ‘Velia’ (Williams 1992: 42, plate X no. 22), which is much more “Oscan”, insofar as is used instead of for Ionic /ε̄/; the inscription is also written rightto-left, the usual direction of the Oscan alphabet (this is also found in some early “Osco-Greek” inscriptions, e.g. the 4th-century BCE Lucania or Brettii or Sicilia 3/Lu 18, Lucania 1/Lu 19. Furthermore, final -IE instead of contracted Ionic -H may come from the Oscan name of the city. But note that the Ionic final /ε̄/ is still used – in Oscan the name would be expected to be ϜΕΛΙO (or ϜΕΛΙΩ).

226   

   Nicholas Zair

Crawford, Michael, W. M. Broadhead, J. P. T. Clackson, F. Santangelo, S. Thompson, & M. Whatmough. 2011. Imagines italicae. A corpus of Italic inscriptions. London: Institute of Classical Studies. Dubois, Laurent. 1995. Inscriptions grecques dialectales de Grande Grèce I. Colonies eubéennes. Colonies ioniennes. Emporia. Genève: Librairie Droz. Fortson, Benjamin W. 2010. Indo-European language and culture: An introduction. Malden, MA: Blackwell. Gualtieri, Maurizio & Paolo Poccetti. 2001. Frammento di tabula bronzea con iscrizione osca dal pianora centrale. In Maurizio Gualtieri & Helena Fracchia (eds.), Roccagloriosa II. L’oppidum lucano e il territorio, 187–295. Naples: centre Jean Bérard. IG XIV= Kaibel, Georgius & Albertus Lebegue. 1890. Inscriptiones Graecae Siciliae et Italiae Additis Graecis Galliae Hispaniae Britanniae Germaniae Inscriptionibus. Berlin: Georgium Reimerum Lejeune, Michel. 1957. Notes de linguistique italique: XIII. Sur les adaptations de l’alphabet étrusque aux langues indo-européennes d’italie. Revue des Études Latines 35. 88–105. Lejeune, Michel. 1970. Phonologie osque et graphie grecque. Revue des Études Anciennes 72. 271–315. Lejeune, Michel. 1972. Phonologie osque et graphie grecque II. Revue des Études Anciennes 74. 5–13. Poccetti, Paolo. 1988. Lingua e cultura dei Brettii. In Paolo Poccetti (ed.), Per un’identità culturale dei Brettii, 11–158. Napoli: Istituto Universitario Orientale. Dipartimento di studi del mondo classico e del mediterraneo antico. Rix, Helmut. 2002. Sabellische Texte. Die Texte des Oskischen, Umbrischen und Südpikenischen. Heidelberg: Universitätsverlag C. Winter. Threatte, Leslie. 1980. The Grammar of Attic Inscriptions. Volume One: Phonology. Berlin & New York: Walter de Gruyter. Wallace, Rex E. 2007. The Sabellic languages of Ancient Italy. München: Lincom. Wallace, Rex E. 2008. Zikh Rasna. A manual of the Etruscan language and inscriptions. Ann Arbor & New York: Beech Stave Press. Williams, Roderick T. 1992. The silver coinage of Velia. London: Royal Numismatic Society.

Stefan Reif

12 How a Jewish scribe in early modern Poland attempted to alter a Hebrew linguistic register Sources: This article deals with the life and work of Shabbethai Sofer of Przemysl (c. 1535 – c. 1640), whose extant works exist only in various autograph manuscripts around the world, such as MS Opp.782 in the Bodleian Library of the University of Oxford, MS 51 in the Rosenthaliana Library of the University of Amsterdam, and MS 2973.3 in the Library of the Jewish Theological Seminary of America in New York. Abstract: This paper demonstrates the ideology-driven efforts of a 16th/17thc. Polish Jewish scribe, Shabbethai ben Isaac Sofer, to alter the existing Hebrew writing standards, and reflects on the various registers of written Hebrew throughout the ages. Although Shabbethai’s expertise as a Hebrew grammarian is undoubted, aspects of his linguistic philosophy may now be questioned in the light of more modern notions relaying to the definition and development of Biblical and Rabbinic Hebrew.

The German-Jewish (i.e. central European) scholarship of the late 18th, and much of the 19th century, has much to answer for, especially in the matter of our understanding of the history of the Hebrew language. As one of the results of the emancipation and renaissance that they experienced in central Europe, learned Jews built an image of a “Golden Age of Spanish Jewry” during the Islamic hegemony of the Iberian peninsula in the 10th, 11th and 12th centuries and idealised its rationalism, its poetry, its studies of biblical exegesis, its grammatical expertise and its accurate Hebrew language. This “enlightened” Sefardi tradition, as they saw it, was contrasted in their minds with the religious outlook of the contemporaneous Jewish communities in Ashkenaz, that is, in Franco-Germany, during the same period. They identified that latter Ashkenazi outlook as pietistic, convoluted and centred on the casuistic study of the Talmud and the minute details of Jewish religious law, and some of them characterised its Hebrew as inaccurate and second class, containing Aramaisms borrowed from talmudic sources and departing from the pristine language of the Hebrew Bible. Their task was consequently to promote, in as many ways, as possible, the more ancient and authentic language of the Hebrew Bible, rather than the parvenu brogue that had replaced

228   

   Stefan Reif

it in numerous Jewish contexts (Schorsch 1983, Feiner 2002, Ben-Sasson 2004; see also Myers 1995). In the matter of their linguistic ideology, it may also be added that in the German-speaking countries of the Austro-Hungarian and Prussian Empires, Yiddish was regarded as bastardised from pure German. The dominant variety of German at that time was considered to be the true heir to the classical Mittelhochdeutsch and was regarded as vastly superior to the “corrupt” language spoken by the Jews of Eastern Europe. If the truth be told, when the medieval Sefardi culture was being favourably compared with its Ashkenazi counterpart, the same people who were making such a comparison were certainly drawing similar parallels with regard to modern German culture being distinct from that of Poland and Russia in their own day. As a clever pun has it, the Jews had become Germans “by the grace of Goethe”. What they were consequently celebrating was, to a large extent, the “Golden Age of German Jewry”. The cultural superiority of one linguistic register to another was therefore a concept that was not only socially familiar to them but also one of their scholarly presuppositions.¹ We shall in due course pass judgement on these and similar views in the light of more recent historical and linguistic assessments. First, however, we have to refute the notion that the proponents of enlightened scholarship in central Europe of some two centuries ago were totally innovative in the assessments they offered, as well as the preferences they expressed, in the matter of variant Hebrew language forms. In fact, there was something of a Polish-Jewish enlightenment some two or three centuries earlier and, given what has been said about the scholars of Jewish Germany, it is hardly surprising that they should either have eschewed a close study of that earlier place and period, or perhaps even chosen to ignore or distort any evidence that emerged from there. What may be deduced from recent research is that various Ashkenazi centres of Jewish life, especially in and around the extensive Polish kingdom of the 16th century, championed new patterns of thought and fresh ideologies. Such spiritual and cultural developments may even have occurred under the impact of their growing connections with the wider Jewish diaspora and with the Jewish homeland in Ottoman Palestine. Also influential were the spread of printing, with the consequent dissemination of much information in new text editions, and the expanding significance of kabbalistic notions within various areas of Jewish thought and practice.

1 Max Weinreich (1980: 281) offers a summary of such attitudes: “Yiddish was no autonomous language, not even one of the many variants of German, but an uncouth corruption of German, a result of the degenerative process that set in among the Jews after they had been excluded from society”.

Altering a Hebrew linguistic register   

   229

But in the present context what should occupy our particular attention is the fact that a fresh and ultimately more intense interest was kindled in an improved understanding, and more impressive mastery, of Hebrew language and grammar. Those enthused by such an interest looked to the Hebrew Bible, in the form standardised and promoted by the Ben Asher family of 10th-century Tiberias, on Lake Galilee in the Holy Land, as the yardstick by which to measure all other Hebrew. Those enthusiasts also admired and wished to emulate the expert Hebrew linguists of Spain, Provence and Italy in the medieval and late medieval periods. A popular, if not the most popular, figure was the biblical exegete and Hebrew grammarian, David Qimḥi (1160–1235), whose works were widely regarded as of central significance. His theories and text-books therefore attracted considerable attention and became the focus of specialised linguistic analysis, mainly favourable but also at time adversarial (see Elbaum 1990, especially pp. 33–64 and 260– 273, and Reif 1973). When two rabbinic luminaries of 16th-century Poland challenged each other’s approaches to learning, the matter of linguistic competence was employed as one of the weapons. Moshe Isserles (1525–1572) was a leading talmudist and codifier of Jewish religious law with a penchant for reading philosophical texts and an apparent willingness to encourage his students to do likewise. Shelomo Luria (1510–1574) was no less expert in Talmud but decried the contemporary obsession with pilpul – the creation of artificial problems by comparing numerous talmudic texts with others with which they were not originally connected – and the use of extreme forms of casuistry to solve these. His criticism, however, extended to the current interest in philosophy. Having noted that some of the students had copied an Aristotelian text into their prayer-books, he censured Isserles for having permitted the adulteration of religious texts with such material. “I have no enthusiasm whatsoever for such and similar texts”, he wrote, “and the only suitable place for reading them is in the toilet”. While offering such criticism, he also alludes to the failure of his contemporary Isserles to demonstrate competence in the Hebrew language. “If you, sir, will forgive me for saying so”, he continued, “it would be more valuable for you to study Hebrew grammar, since the quality of your writing is generally disastrous, with errors in person, gender and number” (included in Isserles’s responsum no. 6 published in his collected responsa (as in ed. prin., Cracow, 1640); see also Elbaum 1990: 156–59). Although such an exchange between these distinguished rabbis reflects the various cultural developments that were taking place around them, it should also be acknowledged that Luria was not the first to find linguistic fault with the kind of Hebrew that had been widely used by talmudists and experts in halakha (Jewish religious law) for numerous centuries and had its own register – that is, a vocabulary, syntax, and vocalisation that did not match the standard form of

230   

   Stefan Reif

the Masoretic Hebrew used for the Hebrew Bible from the tenth century onwards. Rabbinic figures of the 14th and 15th centuries, such as Ya‘aqov ben Moshe Moellin and Avraham Klausner (as well as a less well-known Ya‘aqov Naqdan), had expressed themselves ill-at-ease with the rabbinic register and pointed to the Hebrew Bible as the authoritative form of the language. What had happened was that the growing commitment to the vocalised text of the Hebrew Bible that had been declared authoritative by the Masoretes had brought with it a realisation that the Hebrew used within the rabbinic spheres of talmudic study, halakhic codification, midrashic compilation, textual commentary, and, indeed, correspondence between the learned Jews of various centres, was of quite a different order. With its own linguistic characteristics, it constituted a more popular, even vernacular, form of the language used within rabbinic circles, with more concern for practical application and less commitment to any notion of linguistic orthodoxy or purity (Reif 1979). Despite his absence from all editions of the Encyclopaedia Judaica and the lack of an entry for him in Wikipedia, the great expert in, and champion of, Hebrew language and grammar in late 16th- and early 17th-century Poland was undoubtedly Shabbethai ben Isaac Sofer who flourished for most of his life in the town of Przemyśl, in the south of the country. Having been born there or possibly in Lublin, around 1565, he studied in Lublin with the outstanding codifier of Jewish law, Mordechai Jaffe, himself a student of Shelomo Luria, and appears to have adopted some of the iconoclastic views of the latter, possibly by way of Jaffe. Jaffe, in his code, certainly displays a concern for Hebrew that is reminiscent of the linguistic philosophy of his teacher. More importantly for the present discussion, Shabbethai spent some of his youth perfecting his knowledge of Hebrew language with a renowned grammarian and poet of the day, Yiṣḥaq ben Yequti’el ha-Kohen Katz Koska(s), or Kuskes. Shabbethai was apparently resident for a number of years in the latter’s school in the Polish province known then as “Russia”, an area approximately between Tarnopol and Kamenetz that later became Austrian Galicia and is now western Ukraine. Both Lublin and “Russia” were some hundreds of kilometres from Przemśyl, indicating that Shabbethai was enthusiastic enough about his educational ambitions to travel long distances to achieve them. It was apparently from Yiṣḥaq Katz that Shabbethai inherited his extensive knowledge of, and enthusiasm for, Hebrew grammar. It should, however, be stressed that Yiṣḥaq Katz’s keenness for accurate language was, according to Jacob Elbaum, as much associated with kabbalistic notions as with rational ones, akin to the approach of the Ashkenazi mystics of the 12th century (Reif 1979: 7–8, Elbaum 1990 196–197 and 269–270). Although he does sometimes mention the kabbalistic significance of the linguistic point he is making, Shabbethai is interested in grammar per se, in the

Altering a Hebrew linguistic register   

   231

literal meaning of words, and in the accuracy of Hebrew. He subscribes to the view that knowledge of Hebrew grammar is a sine qua non for any true understanding of kabbala. In a long and highly active career, he composed numerous grammatical, liturgical and halakhic treatises, as well as some essays and a variety of Hebrew poems, and prepared for the press editions of some works. He annotated existing grammar-books, compiled a grammatical text-book of his own, and wrote an essay demonstrating from authoritative Jewish sources that a Jew had an obligation to study grammar. Incidentally – and this will still strike a chord with many a scholar – an Oxford manuscript records his personal oath, as a fairly young man, that he will never lend books without surety, nor act as security for others, since neglect of such matters has caused him considerable loss on a number of occasions! His edited text and lengthy commentary on the Ashkenazi liturgy – his most extensive work, and recorded by an amanuensis of his in a beautifully preserved manuscript – was (and is) of major importance. It was commissioned by the Council of Three Lands, the representative body of Polish and Lithuanian Jewry at the spring fair in Lublin in 1610. It was approved as an authorised version early in 1617 and printed soon afterwards in that same year. Shabbethai was determined to promote the study and practice of accurate Hebrew language and grammar, to follow the latest trends in novel Jewish education, and to demonstrate high standards of rational, historical and bibliographical learning. Fearlessly polemicising against the works of those (both living and dead) who, to his mind, did not meet such standards, he defended David Qimḥi against the strictures of the Hebrew linguist Elijah Levita (1469–1549), who taught and wrote in Renaissance Italy. Shabbethai also strongly criticised his contemporary, Rabbi Meir of Lublin, for expressing the view that, in the matter of the pronunciation of the divine name (adonay), grammar was not of the essence but had been championed by a separate band of specialists in that field. Other luminaries who were targets for Shabbethai’s barbs included Rabbis Ya‘aqov Moellin, David Abudraham (in 14th-century Spain) and Moshe Isserles, as well as Rabbis Mordechai Jaffe and Shelomo Luria, both of whom had (as earlier noted) inspired his work in different ways. The grammatical ignorance of a rabbinic leader in his own city of Przemyśl, Rabbi Moses Math, would have been forgiven by Shabbethai had Rabbi Math not chosen to argue strongly in a halakhic context in favour of what Shabbethai categorised as the incorrect vocalisation of Hebrew. What is perhaps even more remarkable than Shabbethai’s productivity is the fact that he achieved it after becoming totally blind at some point between the years 1602 and 1608, when he was no more than about forty years old (Reif 1979: 9–27, Satz and Yitschaki 1987–2002, Reif forthcoming).

232   

   Stefan Reif

Shabbethai appears to have commenced his professional career as a scribe and hence (as early as 1591) attracted the epithet “sofer” or “ha-sofer”. This activity, however, was certainly no longer possible after he had been struck down by what could well have been glaucoma in the following decade of his life. We must conclude either that the nomenclature, having once been used, stayed with him, or that even from the start it was being used to describe someone with a passion for grammatical and linguistic accuracy and a great expertise in these areas. Certainly his other sobriquet ha-medaqdeq ha-gadol (‘the outstanding grammarian’) makes it clear that he enjoyed such a reputation. Be that as it may, he set about establishing that the only true register for the Hebrew of the traditional prayerbook was not Rabbinic Hebrew, as recorded in the talmudic and midrashic works and later literary works deriving from such sources. It was, rather, the language of the Hebrew Bible which had to be followed. There were a number of obstacles that stood in the way of his achievement of such an ambition. Rabbinic Hebrew was more familiar and more extensively used than its biblical counterpart by most of the learned members of the Jewish community at large and, unsurprisingly, it dominated the linguistic form of most of the literature inspired by the rabbinic sources. There were also linguistic components that had become a traditional part of Ashkenazi Jewry’s form of Hebrew for some five hundred years and included matters of vocalisation, vocabulary and morphology. In addition, there were what we might call more popular, or vulgar factors at work. Congregants adopted forms that sounded right to them, cantors intoned what they had heard from their predecessors, and printers gave authority to what was familiar to them, regardless of what precise form of language it might constitute, by recording it in their editions and disseminating these widely. It thus came about that, to a substantial degree, demotic rather than scholastic forms became the norm. What contributed further to the complicated linguistic situation was the issue of censorship. Either demanded by theological opponents outside the Jewish community, or self-imposed as a pre-emptive defence from within it, the censorship of texts led to unnecessary variations and strange compromises, especially in liturgical formulations (Reif 1979: 28–41). Shabbethai had no truck with such tendencies and classified all, or at least most of them, as linguistically erroneous, meaningless and even blasphemous. In his view, prayer should be recited in the language of the Hebrew Bible and with the same degree of linguistic stringency that is applied to the formal synagogal reading of Scripture. Non-biblical forms should not be retained in the liturgy and, whenever any doubt arises about the correct reading of a text, the biblical and not the rabbinic precedent should be followed. Citing manuscripts and early editions, as well as major halakhic and philological authorities, in support of his theories about Hebrew language, Shabbethai set about attacking his foes with

Altering a Hebrew linguistic register   

   233

great ferocity. The synagogal cantors come in for particularly harsh censure. He claims that they “simply do not know the meaning of what is coming out of their mouths” (Shabbethai manuscript in Berliner 1909: 61).² It is not the inalienable right of those intoning a Hebrew text to lengthen a vowel for as long as they wish, as they do so merely “in order to show off their voices”.³ They extend such sounds for unconscionable lengths of time “to suck up to congregants and to find favour with them but this will not please God nor will their prayers be heard by Him”. Such a performance on their part, argues Shabbethai, is not surprising, since “their behaviour in other matters is equally unsatisfactory”. ⁴ Some cantors tried to justify their reading of Hebrew by reference to what Shabbethai considered to be linguistic ignorance; for him, their incorrect reading was nothing more or less than “an incurable disease”.⁵ He went so far as to hold them responsible for the hardships and misfortune suffered by the Jewish community, for blasphemy, and for Israel’s failure to achieve ultimate consolation from on high.⁶ Shabbethai also drew attention to common linguistic errors perpetrated and perpetuated by teachers of children, as well as by members of the community. Some vowels were never correctly read and some sounds were erroneously altered in vulgar pronunciation. Interestingly, some such pronunciations have survived to this day in spoken Yiddish, as with the word for ‘worry’, de’aga which is pronounced dayga (Reif 1979: 37 and 94). Even when there were clear biblical precedents for Jews to follow, they preferred to maintain the Rabbinic Hebrew traditions. Changes were not to be made for the sake of change but only if justified by his definition of correct Hebrew. He was especially intolerant of so-called experts who set about making changes in the Hebrew of the liturgy in accordance with their notions of Hebrew grammar. He argued that there was no basis for what he regarded as “their pretensions at wisdom and at grammatical expertise”.⁷ One such “expert” who was visiting Przemyśl gave instructions for a change in the public reading of a Hebrew liturgical expression which, for Shabbethai, demonstrated his total theological ignorance in matters concerning the kingship of God. Had he realised what he was saying, opined the learned gram-

2 The original manuscript (henceforth MS), once owned by the Beth Din of the United Synagogue in London, was sold at auction to a private bidder at the Christie’s auction of June 23, 1999, in New York. MS, General Introduction, f. 16c; Satz, 1:49; in Berliner 1909: 61. 3 MS, General Introduction, f. 4d; Satz, 1:1.2; in Berliner 1909: 16. 4 MS, f. 32d; Satz, 1:98. 5 MS, General Introduction, f. 16d; MS, f. 243b; Satz, 1:50–51 and Satz 3:969; in Berliner 1909: 63. 6 MS, Special Introduction, f. 27c; MS, f. 52a; Satz, 1:83–84 and Satz 2:223. 7 MS, General Introduction, f. 6ab; Satz, 1:16; in Berliner 1909: 21.

234   

   Stefan Reif

marian from Przemyśl, that “expert” would have slit his own throat rather than allowing himself to mouth such a view. Shabbethai categorically rejected such attempts at superior linguistic knowledge based on references to expert opinion, by explaining that those making them had misunderstood a context, and by citing texts that constituted a refutation of their views.⁸ He was even not averse to employing somewhat crude terminology to describe what he viewed as their misguided efforts. When they raised questions about texts that were to his mind wholly unjustified he described them as ‘making things hard for themselves’ (ha-maqšim ‘aṣmam).⁹ All these errors had been canonised by their publication in the printed prayer-books. Printers were therefore another target for his sharpened arrows. Together with a number of contemporary rabbinic leaders, he criticised them for inconsistency, unnecessary abbreviation, and the inclusion of inveterate errors, leading to popular ignorance of correct Hebrew and playing havoc with its pronunciation. As a result of their behaviour, prayer was rendered ineffective, Jewish exile was being prolonged and the messianic redemption had been prevented. Even where standardisation had taken place, the wrong versions had been given permanence by the printers.¹⁰ Although he continued with his literary efforts and his powerful linguistic campaign for many years, it seems that Shabbethai ultimately despaired of changing the situation in Poland since the latest texts that he dictated in matters grammatical, which are to be found on the final folio of a manuscript preserved in New York, reveal that he settled in the Holy Land. These brief notes were written in Jerusalem and date from 1637 and 1639. In an Oxford manuscript there is also a note about his recitation of a memorial prayer at the Western Wall in Jerusalem and his composition of some stanzas of poetry on the occasion of his presence there (Reif forthcoming). But his work was not forgotten in Europe. Although the Khmelnitzki massacres of 1648 brought darkness to a hitherto enlightened period of Polish-Jewish history, Shabbethai’s linguistic ideas were taken up and promoted in a modest fashion by various Hebrew linguists in the two centuries following his death in about 1640. The argument between Biblical and Rabbinic Hebrew continued among specialists, leading up to the 19th century when his-

8 MS, General Introduction, ff. 4d; 10b and 17d; Satz, 1:12; 1.30 and 1.53; MS, f. 56b; Satz, 2:286; in Berliner 1909: 16, 36, 66. 9 MS, General Introduction, ff. 11a, 14c and MS, f. 8a; Satz, 1:33, 1.44, and 2.32; in Berliner 1909: 40 and 54. 10 MS, General Introduction, ff. 8b, 10cd and 13bc; Satz, 1:24, 1.32, and 1.40; in Berliner 1909: 19, 29 and 39.

Altering a Hebrew linguistic register   

   235

torical scholarship once again took on board the ideas that Shabbethai had once championed (Reif 1979: 53–62, Cohen 1999, Cohen 2006). But how valid are these ideas, when assessed in the light of the improved knowledge that we now enjoy early in the 21st century? That knowledge has expanded in a number of ways that are of considerable relevance for this whole discussion. But before we consider how current notions may differ from those of Shabbethai, as well as of his predecessors and successors in the field, we must recognise that he was truly an expert in the grammar based on the Hebrew closely defined by the 10th-century Masoretes and used in the versions of the Hebrew Bible that became standard, in manuscript and in print, as a result of their intensive efforts and those of many scribes who followed their lead. Only very rarely did Shabbethai depart from the principles and details that are to be found in the standard grammar books of the medieval and modern periods. All credit must be given to him for having mastered this area of learning and having enthusiastically promoted it throughout his career. He was, however, committed to the supposition that such a form of Biblical Hebrew represented a pristine language, the ideal medium not only for the Hebrew Bible but also for use with the traditional prayers. He was of course aware that the language of the Bible differed from that of the rabbinic literature but by expressing, with only some exceptions, such a strong preference for the former over the latter in the case of liturgical use, he indicated his view that there was a qualitative difference between the two linguistic registers. What is more, he appears to have made no distinction, at least as far as the Hebrew of the prayers was concerned, between errors, inaccuracies and inconsistencies on the one hand, and the general use of Rabbinic Hebrew on the other.¹¹ Current views of the history of the Hebrew language call into question the validity of Shabbethai’s assessment of the different forms of the Hebrew language. With regard to the language that the Masoretes standardised for application to the reading and transmission of the Hebrew Bible, this may well have preserved elements of the ancient characteristics of the language used by the Hebrews, Israelites and Jews in the biblical period, and they certainly had a live tradition on which to base a substantial part of their reconstruction. At the same time, there can be little doubt that by attempting to impose on the whole Hebrew Bible a degree of consistency that could not have been present throughout all the many centuries of its development, they were to a degree opting for a scholastic rather than an historical form. This is not to say that Masoretic Hebrew is simply

11 All this becomes obvious from even cursory glances at any parts of his liturgical work in the editions of Berliner (1909), Reif (1979) and Satz and Yitschaki (1987–2002), also cited above.

236   

   Stefan Reif

an artificial exercise; it did undoubtedly transmit what it had inherited, but it has to be acknowledged that it also smoothed out the linguistic wrinkles. The evidence from the Dead Sea Scrolls and from the Cairo Genizah manuscripts testifies to the remarkable degree of authenticity that was maintained by the Masoretes but, at the same time, indicates an earlier and greater degree of variation than they were prepared to acknowledge and tolerate. In that case, any supposition that the register being used for the synagogal reading of the Scriptures is a purer, more pristine and altogether superior form of the Hebrew language must be open to serious challenge (Tov 1992: 21–38, 100–117, 155–197 and 201–230; van der Kooij 2002). What follows from these remarks is that post-biblical, or perhaps more accurately, non-Masoretic, Hebrew was one of a number of forms of the language that once existed, perhaps side by side, and without any categorical assumption about their relative qualities. It happened to be the one that gradually developed an identity as the register used for the transmission, much of it originally oral, of the total corpus of rabbinic traditions. It is conceivable that some linguistic elements in such a register may even antedate their equivalents in Masoretic Hebrew and it should not be forgotten that rabbinic texts are vastly more extensive and diverse than those of the Hebrew Bible. As far as the linguistic definition of postbiblical Hebrew is concerned, there is today a much greater degree of sophistication about its classification and chronological evolution. This has been made possible not only by a closer study of the linguistic features of Rabbinic Hebrew, as it is currently known, and by a comparison of the shape it took at different periods, but by the availability and analysis of many more manuscripts, a large proportion of them emanating from the Cairo Genizah. Contemporary scholars are aware that Rabbinic Hebrew cannot be characterised and assessed on the basis of early modern, and even later, printed versions of texts that had their origins between one and two thousand years ago. With the more reliable evidence of earlier manuscripts, experts can now appreciate that the Rabbinic Hebrew of the early Christian centuries reflected the language that was used in the Mishnah and early midrashic works. This language was based on a vernacular Hebrew that is reflected in epigraphic sources such as the Bar Kokhba letters of the second century and has similarities with the Palestinian Aramaic that the Jews used in their homeland and which is recorded in the Yerushalmi Talmud. What occurred in the later talmudic and post-talmudic periods was a steady erosion of these linguistic traditions in favour of Biblical Hebrew and Babylonian Aramaic. This left only a minority of the manuscripts with the more authentic linguistic characteristics of Rabbinic Hebrew but it is these latter texts on which current scholarship bases its historical conclusions (Sáenz-Badillos 1993, Pérez Fernández 1997: 1–15; see also Birnbaum 2008).

Altering a Hebrew linguistic register   

   237

And so to our conclusions about Shabbethai Sofer’s scholarly efforts: His grammar of Biblical Hebrew was of admirable standard, his concern for linguistic consistency was at least two centuries ahead of its time, and his awareness of the differences between Hebrew and Aramaic on the one hand, and Biblical Hebrew and Rabbinic Hebrew on other, were those of a meticulous scholar. There were indeed mistaken forms used by rabbis, cantors and printers that did require rectification and Shabbethai inspired later hebraists to take due account of Hebrew grammar in their studies. Where he must, however, surely come in for criticism is in his approach to Rabbinic Hebrew. It is not a corrupt form with unjustifiable Aramaic accretions but a valid register used more commonly than Biblical Hebrew for a wide variety of Jewish literature. It is not in all its aspects a later and less sound version of Biblical Hebrew but, rather, a form of the language with ancient origins and its own authentic elements. Some of what Shabbethai regarded as egregious errors were very old forms that had survived the attempt by many scribes and authorities to “biblicise” them, and that are documented in some of the most original manuscript traditions. But if the Hebrew of the early liturgy did indeed once reflect the popular rabbinic register, and not the scholastic one employed for the Masoretic Bible, there is no doubt that the efforts of generations of Jewish linguists ensured that it ultimately altered its linguistic allegiance from the former to the latter. In that respect, whether linguistically justified or not, the efforts of Shabbethai, and those of his scholarly ilk, succeeded in leaving their impact on the language of the traditional prayers today recited by many communities of Jews, especially of modern European background, who may never even have heard of Shabbethai Sofer.

References Ben-Sasson, Menahem. 2004. Al-Andalus: The so-called “Golden Age” of Spanish Jewry – A critical view. In Christophe Cluse (ed.), The Jews of Europe in the Middle Ages (tenth to fifteenth centuries), 123–135. Turnhout: Brepols. Berliner, Abraham. 1909. Abhandlung über den Siddur des Schabtai ha-Sofer aus Przemyśl. Frankfurt a. M.: Kauffmann. Birnbaum, Gabriel. 2008. The language of the Mishna in the Cairo Geniza. Academy of the Hebrew Language: Jerusalem. Cohen, Chaim E. 1999. Ashkenazic mishnaic reading traditions in eighteenth-century grammatical treatises. Lĕšonénu 62. 257–283. Cohen, Chaim E. 2006. Mishnaic Hebrew as reflected in Ashkenazic prayer books. Lĕšonénu 68. 73–82. Elbaum, Jacob. 1990. Openness and insularity: Late sixteenth century Jewish literature in Poland and Ashkenaz. Jerusalem: Magnes Press.

238   

   Stefan Reif

Feiner, Shmuel. 2002. Haskalah and history: The emergence of a modern Jewish historical consciousness. Oxford: Littman. Myers, David N. 1995. Re-inventing the Jewish past: European Jewish intellectuals and the Zionist return to history. New York and Oxford: Oxford University Press. Pérez Fernández, Miguel. 1997. An introductory grammar of Rabbinic Hebrew. Leiden, New York and Köln: Brill. Reif, Stefan C. 1973. A defense of David Qimḥi. HUCA 44. 211–213. Reif, Stefan C. 1979. Shabbethai Sofer and his prayer-book. Cambridge: Cambridge University Press. Reif, Stefan C. Forthcoming. Another glance at a gifted grammarian: More on Shabbethai Sofer of Przemysl. In Judith Olszowy-Schlanger & Irene Zwiep (eds.), Hebrew linguistic thought and its transmission in the Middle Ages and Early Modern Times. Sáenz-Badillos, Angel. 1993. A history of the Hebrew Language. Cambridge: Cambridge University Press. Satz, Yitzhak & David Yitzchaki (eds.). 1987–2002. Siddur Ha-Medaqdeq Ha-Gadol. Baltimore: Ner Israel Rabbinical College. Schorsch, Ismar. 1983. The emergence of historical consciousness in Modern Judaism. Leo Baeck Institute Year Book 28. 413–437. Tov, Emanuel. 1992. Textual criticism of the Hebrew Bible. Minneapolis and Assen/Maastricht: Fortress Press and Van Gorcum. van der Kooij, Arie. 2002. The textual criticism of the Hebrew Bible before and after the Qumran discoveries. In Edward D. Herbert & Emanuel Tov (eds.), The Bible as book: the Hebrew Bible and the Judean desert discoveries. London: British Library and Oak Knoll. Weinreich, Max. 1980. History of the Yiddish language. Chicago and London: University of Chicago Press.

Part IV: Idiosyncracy, scribal standards and registers

Alexander Bergs

13 Writing, reading, language change – a sociohistorical perspective on scribes, readers, and networks in medieval Britain Sources: The material analysed in the first part of this paper belongs to the group of documents known as the Paston Letters. These are letters, memoranda, indentures, testaments, and other minor writings composed by members of the Paston family from Norfolk between 1421 and 1503. While many of the originals were lost, we still have a number of contemporary copies, which were retained by their authors and subsequently archived, e.g. for legal purposes. After a long and troublesome history, the 421 texts analysed in this paper are now kept at the British Library, the Bodleian Library, Magdalen College, Oxford, and Pembroke College, Cambridge. There are a number of editions available, dating back to 1787. The most authoritative edition is Davis (1972). The second part of the paper deals with the famous final continuation of the Anglo-Saxon Chronicle, the so-called Peterborough Chronicle (MS Laud 636, Bodleian Library). The Peterborough Chronicle is divided into two parts. The first part is written by a number of unknown hands and deals with historical events between 1122 and 1131; the second part was written in about 1155 by a single (unknown) scribe in Peterborough and deals with the years 1132 to 1154. Both texts are of particular importance not only because of their enormous historical value but also their linguistic significance, as they are usually described as containing some of the first specimens of Middle English. Several important editions are available, such as Clarke (1970) and Irvine (2004). Abstract: This paper investigates the relationship of authors and scribes in medieval Britain. It begins with a more general discussion of the various models of linguistic change that explicitly include written language as a source and medium for linguistic change. It is argued that both Samuels (1972) and Angus McIntosh (1956) provide valuable input here since Samuels’s model does not prioritise in any way written or spoken language as possible sources of linguistic change, and McIntosh’s model explicitly puts written language and its systematicity at its centre. Both these models are important for studies on the influence of scribes, as scribes first and foremost are actors in the written, not the spoken medium. As a next step, this paper discusses the production of (late) medieval texts. It is shown that many texts were in fact not autographed, but dictated in some way by their authors and actually produced by scribes. There are, for example, virtually no writings by female scribes left since there was an extremely high

242   

   Alexander Bergs

degree of illiteracy among female speakers before about 1600. Nevertheless, we do have texts by female authors. This naturally leads to the question what role scribes actually played. How influential were they with regard to the linguistic structure of the text they produced? The present paper argues that their influence may have been quite extensive when it comes to orthographical conventions, but that their influence on syntax and morphosyntax may have been much smaller than expected. In order to discuss this claim, this paper presents one case study on the Late Middle English Paston letters, i.e. 421 family documents (about 245,000 words) produced by members of the Paston family between 1421 and 1503. Some family members wrote their own texts and worked as scribes for others. Similarly, some identifiable scribes worked for more than one family member. A detailed study of two morphosyntactic variables (the new th-pronouns and relative clause structures) shows that in both cases there is no distinct scribal style to be identified. On the contrary, scribes show very different linguistic behaviour when writing for different authors, including themselves. This leads us to suspect that their influence at least in the domain of morphosyntax may have been fairly limited. A second case study looks at a scribal innovation in written language features in McIntosh’s sense. The final continuation of the Anglo-Saxon Chronicle, the so-called Peterborough Chronicle from 1155, is characterised by a very unusual use of the new digraph for the former runic signs “thorn” and “eth” . The spelling was extremely rare before the Peterborough Chronicle and for about 250 years afterwards. This paper discusses the possible motivation of the scribe and the patterns of innovation to found here. It also offers some speculation about why these new written language features did not catch on in the linguistic community. In its summary and conclusion this paper points out that current studies on linguistic change need to focus on both successful and unsuccessful innovations. The role of scribes as agents of linguistic change is an extremely complex one, and their actual part in the game can only be accurately identified, described, and analysed if we look at both actual changes and failed innovations.

1 Models of language change Most of current language change theory is based on speakers and hearers, rather than writers and readers. Investigations from that particular angle pose few problems in contemporary linguistics, but can be a big problem for historical linguistics, where we usually do not have speakers and hearers as such, but

Writing, reading, language change   

   243

rather writers and readers. Since this paper is explicitly concerned with language change as it happens in (or is at least influenced by) the written medium, it is important to begin with some thoughts on how language change can happen, and how spoken and written language can be related in this respect. There are at least two language change theories that have explicitly included written language in their models. These were developed by Chafe (in 1984), and before him by Samuels (in 1972). Chafe discusses linguistic innovations that can start either in “speaking” or in “writing”, and which then enter into a dialectal relationship in which this innovation is present or absent in one of the two, or both. In Model 1 below we have a feature that first only surfaces in spoken language. It takes a while before the same feature also appears in written language. This means that in the meantime we have some sort of layering, i.e. a situation where a new linguistic feature is present in spoken language, but absent in writing. This is surely one of the most common patterns that we find in linguistic change (which often, but not always, begins in the spoken mode). Model 2 assumes a different path. Here, a feature is present both in speaking and in writing, but, for some reason, drops out of use in the written mode, only to resurface (possibly) at a later time. Reasons for this disappearance could be, for example, prescriptivist ideas about style and register.

Figure 1: Model 1.

Figure 2: Model 2.

The third model shows a development where a certain form is present in writing, but absent from speaking. This can be the case in “changes from above”, for example, where new forms are introduced in highly formal, literate styles. Then, for some reason such as stylistic stigmatisation or sociocultural constraints, the

244   

   Alexander Bergs

form may disappear from writing, to be lost completely (with a chance of resurfacing again at some later point in time).

Figure 3: Model 3.

These three models are very valuable in so far as they explicitly discuss the relationship of written and spoken language in linguistic change processes. However, they also seem to miss one important point: innovations do not necessarily have to start in one mode or the other and then surface in the other mode after some delay. Rather, they can also appear and disappear simultaneously, as Samuels’s (1972) model demonstrates (see below).

Figure 4: Model 4.

Here, the arrow marked “B” shows the traditional view of an innovation starting in spoken language which then shows up after a delay in writing. Arrow “C” shows the exact opposite, while the two arrows marked “A” are cases of simultaneity. Since there is no reason to assume that simultaneity in spoken and written language change is in principle impossible (and in fact we can imagine a number of scenarios where this could happen, particularly in present-day languages with their pervasiveness of writing!), I believe that the model developed by Samuels is actually more flexible and accurate in its descriptions of possible pathways of linguistic change.

Writing, reading, language change   

   245

Unfortunately, however, the whole issue is further complicated by the fact that we also need to distinguish between Spoken Language and Written Language as quite different ontological beasts. Angus McIntosh, one of the founding fathers of the groundbreaking Linguistic Atlas of late Medieval England (McIntosh et al. 1986), pointed out that there are linguistic features that belong only to the spoken mode (so-called S-features). One example of these S-features would be sound changes that are not reflected in writing (e.g. the raising of long /i:/ in the Great Vowel Shift to /ai/, as in OE fife [fi:f] to ModE [faif] – both are spelled with the same vowel symbol. Changes in the consonants are a different matter). There are also features that belong only to the written mode (so-called W-features). These, obviously, would include some changes in orthography that have no resonance in pronunciation, e.g. the medieval practice of spelling [u]  and [w]  as and , which gradually fell out of use in the early Modern period). Some features, of course, pertain to both (e.g. the phonetic fusion of going and to into gonna in modern English, which is also reflected in the informal spelling ). For McIntosh it was very important to point out that certain changes only happen in the written medium, i.e. in W-features. So, when we study scribal factors in linguistic change, or even scribes as agents of linguistic change, these W-features are, of course, of special interest. But we also need to keep in mind that, in theory at least, scribes can have influence on other linguistic levels, from morphology through syntax to lexicon. And, as Samuels has shown, writing may influence speaking. Only think of the history of the word waistcoat in English (Smith 1996: 15–17), which has the traditional pronunciation [ˈwɛskɪt], and a more modern one, based on its orthography: [ˈweɪst(t)ˌkəʊt].

2 Reading and writing in the Middle Ages We need to consider the fact that in earlier language stages, particularly before 1500, most language users were illiterate and had to employ scribes in order to write. Obviously, precise figures for the number of illiterate language users are hard to develop, and there is considerable debate about reliable measurements and guestimates, even for the 17th century. And the further we go back in time, the less reliable our figures become. In one influential study, Cressy (1980) suggests that, on average, male illiteracy in the middle of the 16th century could have been around 80 %, female illiteracy as high as 95 %. However, there is also considerable variation, influenced not in the least by sociological factors such as education, place of living, and occupation. Also, we should not forget that, especially in the Middle Ages, reading and writing were not regarded as the same skills and

246   

   Alexander Bergs

were often taught independently (see Graff 1987; Clanchy 1993: 125; Nevalainen and Raumolin-Brunberg 1996: 25). Whereas writing was a form of art or a craft, reading was an intellectual endeavour (Chartier 1996). Also, the process of writing itself was highly complex in the Middle Ages and before. We can find direct and clear dictation, next to drafts on which the scribes elaborated independently, next to scribes actually composing letters according to certain key words (see Davis 1971). What this means is that in these processes scribes may have influenced the linguistic product they were working on. Documents may not be representative of their authors, but of their scribes. Thus, scribes may have had serious influence on linguistic change. Finally, we also should not forget that scribes in the original sense (i.e. monastic scribes working in scriptoria copying manuscripts) were working under conditions that are not the same as we find today. Contemporary culture wants to see copies as close to the original as possible; for the Middle Ages this was not so. Despite the fact that there are many accurate copies, scribes also enjoyed the freedom to change and correct their originals as they deemed fit – sometimes quite radically (see Cerquiglini 1989 for the importance of variance in literary works; Hanna 1996 and 2010). And again, this can be an important source of linguistic change. If scribes enjoyed this kind of freedom (maybe even in dictation or when copying from a draft) the scribes are sometimes to be seen as linguistic innovators, not the authors themselves. Similarly, we also need to keep in mind that writing in the Middle Ages was a time-consuming and very expensive process. The cost of parchment or vellum (still in use in the 15th century!) fostered the optimal use of space in a manuscript. But even after the widespread introduction of (much cheaper) paper in the 14th and 15th centuries, we might still assume that the use of paper was nowhere near as prevalent as in modern times. This leads me to believe that writing usually was not an ad hoc process, but rather something planned, and maybe even sometimes supervised in scriptoria, for example. At the same time, we may also assume that writing became more and more natural and ad hoc the further we progress in time. And from the 15th century onwards indeed we find an increasing number of small personal notes and other miscellaneous pieces of written language.

3 Case Studies In the following, I will present two brief case studies from two different periods and cultural contexts that deal with scribes in medieval England. First, I will discuss the role of authors and scribes in the late medieval (i.e. 15th-century)

Writing, reading, language change   

   247

Paston letters. The aim is to show that scribes may indeed have been less influential than one might think, especially in the domain of morphosyntax. In other words, Samuels’s C arrow is of course an option, but only when it comes to the language of authors, not necessarily that of scribes. The second case study comes from the early Middle English (12th-century) Peterborough Chronicle, i.e. the final continuation of the Anglo-Saxon Chronicle. Here we see very interesting scribal innovations, which I would like to discuss, particularly with regard to the issue of innovation and diffusion in linguistic change. So this is a study that looks at scribes and McIntosh’s W-features, and as we will see, they are an innovation that in Chafe’s sense originated in writing, dropped out again, and then resurfaced in writing.

3.1 Authors and Scribes in the Paston Letters Let us begin with the Late Middle English Paston Letters and the influence of authors and scribes on the linguistic form of those letters. “The Paston Letters” comprise 421 documents (i.e. personal and official letters (349), indentures (16), memoranda (15), testaments and draft wills (12), petitions (11), inventories (9), declarations (2), and several other text types, including even one recipe and one schedule) with a total of about 245,000 words. These were written (i.e. composed) by members of the Norfolk Paston family (11 male, 4 female authors), between 1421 and 1503. The authors belong to three generations of the family (see family tree with biographical data on page 259); the three most productive authors (John II, John III and Margaret) produced between 78 and 107 documents, with more than 40,000 words per author. This study is based on Davis’s authoritative edition published in 1971 (Davis 1971). We can identify about 22 different hands in the letters. The male family members sometimes wrote themselves, sometimes dictated; the female family members were largely illiterate and dictated all their documents. And this is of particular interest here. Some of the letters by female authors were written by male family members. So, for example, Margaret Paston dictated some letters to her three sons John III, John II and Edmond II. Most of them also wrote their own documents as authors and as scribes. Now, if they had been influential as scribes when they took down their mother’s dictation we must assume that their own writings and their writings as scribes should be very similar. Or, in reverse, if they were not influential, the letters written for their mother should be different from their own, unless mother and sons used the same kind of language – which clearly is not the case (see Bergs 2005 for a fuller documentation of the particular differences). We will test this hypothesis on the basis of two independent mor-

248   

   Alexander Bergs

phosyntactic variables: the development of the third person plural pronouns, and the use of different relativizers. Table 1 below shows the use of third-person plural pronoun forms by Margaret and her sons. They had at their disposal the traditional Anglo-Saxon forms that begin with simple , i.e. ¹ ‘them’, and ‘their’, or alternatively the new forms, borrowed from Old Norse, beginning with a dental fricative: and . Note that the nominative form ‘they’ (former Anglo-Saxon ) had already changed by the 15th century, so that we find no variation for this form in the letters (for a full story on the language internal factors involved here, see Bergs 2005, Chapter 4). Table 1 below differentiates between the three scribes Edmond II, John II, and John III, and the respective authors (Edmond II, John II, John III, and Margaret). Let us begin with Edmond II and Margaret. Edmond wrote both for himself and for his mother. The traditional pronoun forms, however, only appear when Edmond II acts as scribe for his mother. He never uses them when he is writing for himself. John II also has a comparatively high rate of hem when writing for his mother (50 %).

Scribe

Author

Hem

Them

Edmond II

Edmond II Margaret John II Margaret John III Margaret

 5  3  4 60 21

 8  9 98  4 77 11

John II John III

Here

Their

Total

 1

  9  15 116  11 196  40

1

2 4

15  3 57  4

Table 1: The use of third person pronoun forms by Edmond II, John II and John III when acting as authors and scribes.

Only about 2 % of his own object pronouns have this traditional form ; the vast majority is modern . The figures for John III are superficially different, but essentially show the same pattern. When he is writing for himself, he uses hem about 40 % of the time. When he is writing for his mother, hem is almost twice as frequent as modern them. The possessive forms ( versus

1 Needless to say, these are simplified representations. The actual forms differed considerably: , , , are just a few of the variants. These were all simplified to , and for the purposes of the present paper. The same applies to the modern forms, which are simplified to , , .

Writing, reading, language change   

   249

) are even more illuminating. Only about 4 % are traditional when John III is the author, but when he acts as scribe for his mother Margaret traditional and modern forms are on a par. In sum, all three speakers appear to have been much more conservative when they wrote for their mother. As a Pearson’s chi-square test with Yates’ continuity correction of the variable role (author vs. scribe; e.g. John II writing for himself vs. writing for his mother) and pronoun form (“old/h-form” vs. “new/th-form”) indicates, this difference is statistically significant (Χ2 = 29.012, 1df, p < 0.001). While the authors themselves strongly preferred the th-forms, the statistical analysis shows that they significantly exhibited more h-forms and fewer th-forms when acting as a scribe for their mother.² Consequently, it seems as if the actual author of the letters played a much greater role in the selection of the pronoun forms than the scribe. How else could we explain the reasonably clear preferences that the scribes show in their own writings and which differ from the choices that we see in the dictated letters? A second group of letters that is particularly interesting consists of letters by different dictating authors (e.g., John I and his Margaret) that were actually written down by one particular scribe (e.g., James Gloys, a servant and chaplain for the family). The hypothesis is the same as the one outlined above: if the language in the letters were essentially that of James Gloys, the letters should be linguistically very similar, despite being dictated by different authors. If they are different, this can only be due to the authors, since the scribe was the same. And, indeed, the findings are quite illuminating. Gloys, when writing for John I, used the possessive pronoun her(e) nine times in 1449 (letter no. 36, 1,062 words), and modern their only once. When he was writing for John I’s wife Margaret in 1469, he used their nine times and her(e) not at all (letter no. 200, 874 words). These figures need to be taken with a pinch of salt, however. The temporal distance of twenty years between the letters may also have played a role. As Bergs (2005, Chapter 7) shows, the changes in the family happen very rapidly and people change their language drastically within ten years and less. It is particularly noteworthy in this case that the letter from Margaret from 1469 shows so many innovative th-forms. Table 1 above shows that she seems to

2 When looking at the residuals of the chi-square test, it turns out that authors used significantly more h-forms and fewer th-forms when acting as scribes only: the chi-square-per-cell values for these two effects exceed the threshold of 6.24 (with Χ2 = 18.88 and Χ2 = 6.58, respectively), which corresponds to a Bonferroni-corrected p-value for multiple tests of 0.05. I am extremely grateful to Thomas Hoffmann for providing me with his expertise in statistical analysis. Without his additions, this whole bit would have been so much more impressionistic.

250   

   Alexander Bergs

have been fairly conservative. Nevertheless, the findings are still not implausible, particularly when compared with the letters that John II wrote for Margaret, where we also find some innovative forms, particularly in the possessive. Let us compare the findings for Gloys with those for another scribe, John Wykes. Wykes wrote both for Margaret Paston and for John Paston II, Margaret’s son. He used the objective pronoun form hem twice, but them only once when writing for Margaret in 1465 (letter no. 190, 1,069 words, to her husband John I), in 1469 when writing for John II he used hem only once, them twice (letter no. 242, 1,308 words, to Walter Writtle). Unfortunately, the figures are really very low here and certainly do not allow for any definite conclusions. Still, they point in the same direction as those presented above, and they suggest the same tendencies, namely that scribes did not (always) compose letters in their own style, but actually paid attention to the forms that the authors used or at least those they believed to be appropriate for certain authors, e.g. male or female, young or old, conservative or modern, etc. The advantage that we have when we compare the two letters written by Wykes is that the letters are of comparable length and written roughly at the same time, so that even very low figures can give us some interesting insights into the relationship of authors and scribes in this case. A question that we cannot answer on the basis of this data, however, is how far the language represented here is actually that of the authors, or rather that of the scribes representing that of the authors. Based on current studies in style, style-shifting and accommodation theory (see, e.g., Eckert and Rickford 2002) it is not unlikely that experienced scribes were very good at anticipating and representing the language that was typical for a particular author. This in turn may have reflected sociolinguistic variables such as age, gender, class, but also stylistic factors such as the addressee or the type of letter that was composed (see Bergs 2007). Unfortunately, these questions will have to remain unanswered at this point, not the least because of the scarcity of detailed data before 1500. Nevertheless, in sum we can say that the scribes in the Paston Letters on the whole seem to have been much less influential at least in the domain of morphosyntax than we might believe. They certainly had some influence on orthography, and perhaps even on the lexicon, but the morphosyntax of the letters (at least with regard to the two variables studied here) rather reflects the language of the authors, and not that of the scribes.

3.2 The scribe as agent in the Peterborough Chronicle Let us now turn to the second case study, the Peterborough Chronicle, and scribal influences in W-features. Here we are dealing with the third part (the entries 1132–

Writing, reading, language change   

   251

1154 CE) of Text E of the Anglo-Saxon Chronicle, preserved as Bodleian MS Laud Misc 636 (and edited by Clark in 1970 and Irvine in 2004). It can be shown that this third part was written in a hand completely different from those of the first two sections and that the entries were made in Peterborough, by a scribe from that town, as a single block in or around 1155 (Clark 1970; Curzan 1996). In all likelihood, the author/scribe was English, and this is particularly important, as it has often been suggested that seemingly “irregular spellings” in Early Middle English are due to Anglo-Norman scribes who were not familiar with English orthographic conventions. This lack of competence arguably led to a certain degree of randomness in the spelling in these early manuscripts. This is partly reflected in dismissive verdicts such as the one by Bennett & Smithers (1974: 374, my emphasis): “the philological value [of the Peterborough Chronicle] is reduced by a slight admixture, not of scribal divergences from an ‘author’s’ form of ME (since they are virtually of autograph status), but of forms from the standard written language of Late OE, which was W[est] [S]axon […], and by a disordered system of spelling”. However, I think it can be argued in line with Milroy (1992a: 193–194) and Clark (1992: 121–125) that this is in fact only the myth of the Anglo-Norman scribe, often originating in a modern prescriptivist attitude to spelling. As a matter of fact, if spellings really had been random and chaotic they would have caused severe communication problems. Since we have no evidence whatsoever of these problems, we can assume that most spellings somehow made sense to contemporary authors and readers. Thus, as linguists dealing with linguistic variation, we should try to investigate in how far the variation that we find in these documents is structured in the sense of Labov, Weinreich and Herzog’s structured heterogeneity (1968). It appears to be the case that structured variation particularly in spelling has so far often been neglected in Middle English studies, so that, in other words, beginning with the hypothesis that spelling variation was not random but rule governed, we need to take spelling variation seriously and to describe and analyze the discernible patterns in our manuscripts. In the present case study, I am interested in the variation of the spelling of the interdental fricatives. Scribes could choose between the traditional runic forms “thorn” , “eth” , and the new digraph .³ All three forms could be used

3 Hogg (1992: 76–77) suggests that the spelling may have originated from Irish, where it was used in the 8th century, but later replaced by the runic graphs “thorn” and “eth”. Similarly, can also be found in Old Saxon for the voiceless interdental fricative, though there is no evidence before the 9th century. It seems fairly obvious that both accounts do not fit in with what we know about the Peterborough Chronicle.

252   

   Alexander Bergs

for both the voiced and the voiceless interdental fricatives [θ] and [ð]. Old English and Early Middle English had a simple voicing rule which said that within the foot, the interdental fricative was voiced when it appeared between voiced segments, and voiceless initially or finally. This rule, however, seems to have been purely phonological, so that we do not find any clear and straightforward correlation with “thorn” or “eth” as spelling variants. In how far this is also true for the Peterborough Chronicle will have to be seen in the following. It is indeed easy to show that all three forms could be used for both the voiced and the voiceless allophone in the Peterborough Chronicle⁴: [θ]: (1137/53), (1135/23), (1140/52) (1137/40), (1135/11), (PetC-2 1131/2) (1137/6), (1135/6), (Havelock, line 88) [ð]: (1137/50), (1135/11), (1137/22) Since there is no straightforward correlation with phonetic values, we are dealing with McIntosh’s “Written language” features. As has been pointed out above, this does not mean that variation here would be random or insignificant: “[…] orthographic variation rarely attracts much interest unless it offers at least the possibility of clues about phonetic variation or unless […] some variant form is at once striking and rare. It is important here to stress the fact that there is a great deal to be learnt from a thorough examination of those numerous cases of orthographic variation which have no phonetic implications, such as the examples (Þ/th, i/y, ʒ/gh, u/v […]” (McIntosh 1963: 5).

Since this paper explicitly investigates the role of scribes as agents of linguistic change, the investigation of orthographic (W-) features as genuinely scribal features becomes even more important and interesting. In the edition of the Peterborough Chronicle produced by Cecily Clark in 1970 we find about 2,700 words all in all, and 384 occurrences of , , and the new digraph . The new digraph accounts for about 60 (16 %) of all spellings, “thorn” is the most popular form (69 %), “eth” is the least common form (15 %). The following tables describe their frequency in relation to various well-established factors, such as phonetic value, position in the word, and word type.

4 All quotations from the Peterborough Chronicle are based on Clark’s (1970) edition and follow the (year/line) format.

Writing, reading, language change   

Total





384 (100 %)

266/270 (69.3 %)

58 (15.1 %)

60 (15.6 %)

   253

Table 2: Total occurrences of , and .

Table 3 below differentiates between what must have been the representation of the voiced and voiceless interdental fricatives. It was assumed that voicing essentially occurred between voiced segments in the foot. Word initial and final occurrences were assumed to be (still) voiceless, as initial voicing only developed during the Middle English period and was not completed before the end of the 15th century, when function words such as the showed initial voicing (Lass 1992: 59). While all forms were possible in all environments, seems to have been particularly frequent in a voiced environment (70 % of all cases, in contrast to “thorn” and “eth”, both of which account for less than 10 % each). This means that more than 65 % of all voiced fricatives are represented by , which only makes up about 16 % of the overall occurrences. , however, was also current in the system for the same function, so that we can find some unexplained variation at this point (e.g. with the word other, which could be spelled with either or ).





[θ] [ð]

251 (94.4 %)  15 (5.6 %)

53 (91.4 %)  5 (8.6 %)

18 (30.0 %) 42 (70.0 %)

Total

266 (100 %)

58 (100 %)

60 (100 %)

Table 3: Occurrences of , , representing [θ] and [ð].

Tables 4 and 5 below show the occurrences of the three spelling variants in different environments. Table 4 looks at the position in the word, i.e. initially or finally versus medially. Table 5 distinguishes between occurrences lexical versus grammatical words. The latter may appear to be a rather unusual conditioning factor.





I/F M

247 (92.9 %)  19 (7.1 %)

52 (89.7 %)  6 (11.5 %)

18 (30.0 %) 42 (70.0 %)

Total

266 (100 %)

58 (100 %)

60 (100 %)

Table 4: Occurrences of , , in I(nitial)/F(inal) or M(edial) position.

254   

   Alexander Bergs





Gram Lex

250 (92.6 %)  20 (7.4 %)

56 (96.6 %)  2 (3.4 %)

41 (68.3 %) 19 (31.7 %)

Total

270 (100 %)

58 (100 %)

60 (100 %)

Table 5: occurrences of , or in grammatical and lexical words.

However, in a series of influential studies, Betty Phillips was able to show that word class may indeed have had an effect on scribal choices (see, e.g., Phillips 1995). With regard to position in the word we find essentially the same picture, of course, i.e. was mostly used in medial position. Interestingly, however, ten out of the 18 occurrences in initial or final position are in the word . The other eight occurrences are single instances of mostly grammatical words. In general, however, it seems that there was a slight preference for in grammatical words, while occurred more often in lexical items. Moreover, was more common in initial/final position (as in ), probably as a result of unpacking assimilated sound sequences. A very frequent combination in the manuscript is (as a shorthand form for “and”) followed by (a phonetically reduced version of the determiner ). This reduced form obviously could not be represented by the runic letters or , since these represented interdental fricatives, not plosives. Following however, we would expect some phonetic assimilation of the fricative to a plosive. This plosive could be represented graphically with , hence after or : [ante]. It seems not unlikely that this was one of the starting points for the loss of the traditional forms and the introduction of . If speakers realised that the spelling was a representation of the former determiner spelling or , they might want to use a similar representation even without the assimilation environment. Otherwise, they would have a very different representation in the context and elsewhere. One way of expanding the form was to take the representative of the plosive and to add some “fricative flavour”. This could simply be done by adding an . A similar process can be observed in the Latin “translation” of Greek φ, where we see a combination of plosive

and . We could describe this process as some kind of “graphemic unpacking” of the assimilated form into . meanwhile was used in only very few lexically conditioned cases and culturally prescribed spellings (such as < ðat>). By culturally prescribed we mean that the scribe was somewhat influenced by his predecessor⁵, at least through

5 “The word has a culturally prescribed spelling which the scribe has learned from others” (Hockett 1959, cited in Stockwell and Barritt 1961: 77).

Writing, reading, language change   

   255

reading the first continuation of the Chronicle. The scribe(s) of the first continuation wrote their text in about 1131, the scribe we are analysing here wrote his part in 1155 so there are only 24 years between them. It is also not unlikely that they came from the same area and were educated at the same monastery. Clark says in her edition that “as Peterborough monks, its writers were probably natives of the district; for in the late 12th century and in the 13th century, established abbeys recruited their monks locally, and this was probably the practice in the earlier twelfth century also” (Clark 1970: xxxvii; see Blake 1992: 12–13; Bauer 1986: 206–207 on the development and abandoning of scribal conventions in monastic circles). Lass comments: “Perhaps some significant part of linguistic structure has nothing at all to do with mind or semiosis, but derives simply from a kind of recipient’s inertia. Speakers have to accept what’s there already” (Lass 1997: 12). In general, however, it seems safe to say that “eth” was mostly restricted to a few fossilised spellings, and that it was no longer productive with our scribe. The interesting point now is that this scribe produced something like a onehit-wonder. as an innovative spelling remained extremely rare until about 200 years later and even then its use was extremely variable. In Patience (British Library MS Cotton Nero A X, ca. 1375 CE), for example, we even find the use of spelling for purely aesthetic (one might almost say calligraphic) purposes: it contains alternating lines that show and . In sum, what does that mean? James Milroy (1992b) developed a model of language change that comprises speaker innovation as the initial act of linguistic change, followed by diffusion and finally actuation. Only when an innovation is picked up by other speakers do we see it spread and diffuse and eventually become a new part of the linguistic system (i.e. actuation proper in the sense of Weinreich, Labov and Herzog 1968). What we have seen in the Peterborough Chronicle is a most interesting story of a scribe as an agent of linguistic change. Here we have an innovator of a new W-feature – but from all we can tell, the community resisted this incipient change. Why did the change not catch on then and there, but only three hundred years later? First, we may assume (but this is only an informed guess) that this document might have been written by High Candidus, a highly controversial person in Peterborough, or one of his pupils. This might explain why the change did not spread (simply because the scribe was not in high regard, especially since the contents of the final continuation were also politically problematic at the time). On the other hand, it is not quite clear why such an important task (after all, this is the Anglo-Saxon Chronicle!) was left with somebody so controversial, and why this person was not subject to stricter invigilation in the scriptorium. This, however, will have to remain a mystery. Another feature also needs to be considered here: there was no great need for English documents around 1200 so that this blurs our picture on the one hand but also leads

256   

   Alexander Bergs

to the idea that because of the probably low production rate there was no chance for a critical mass to develop. Eventually, all we can say is that though we have been able to identify a genuinely innovative scribe, the socio-cultural context of his text production probably reduced the chances of this innovation spreading. Nevertheless, what we see here is a very clear case of a scribe as an agent of (albeit unsuccessful) linguistic change.

4 Summary and conclusion This paper presented some general thoughts on reading, writing and language change, in order to highlight the importance of identifying scribes and authors and what roles they play in linguistic change. In particular, it was pointed out that scribes, and not only authors, may actually influence the linguistic system not only on the purely written level (in its W-features), but also on other levels such as lexicon and morphosyntax. This means, in turn, that it is important to carefully distinguish between authors on the one hand, and scribes on the other when conducting investigations in socio-historical linguistics. This task seems to be especially important (and difficult) before 1500. These ideas were exemplified on the basis of two case studies: the role of authors and scribes in the use of two morphosyntactic variables in the Late Middle English Paston Letters, and the role of the individual scribe in the development of modern English orthography in the Early Middle English Peterborough Chronicle. In the Paston Letters it was shown that scribes were indeed less influential than one might think, given the fact that many of the documents were not autographed, but dictated. While we may assume some influence in the domain of lexicon and orthography, authors seemed to have had an influence on the morphosyntax of their dictated texts. In the Peterborough Chronicle we saw an important scribal innovation in so-called W-features, i.e. the development of the new graphic symbol . Despite the fact that the scribe used this new form in a mostly structured way, it did not catch on in the linguistic community of the time, and only resurfaced about two hundred years later. So what we can observe here (in more or less detail) is one scribe as one actual agent of linguistic change. Unfortunately, this agent is not successful, but as Milroy (1992b) claims: the study of linguistic change should not only be concerned with successful innovations, i.e. actual change, but also with those innovations that do not catch on. In any case, it should have become clear that scribes are indeed important agents in linguistic change, and that they deserve more attention in future research.

Writing, reading, language change   

   257

5 References Bauer, Gero. 1986. Medieval English scribal practice. Some questions and some assumptions. In Dieter Kastovsky & Aleksander Szwedek (eds.), Linguistics across historical and geographical boundaries. In honour of Jacek Fisiak on the occasion of his fiftieth birthday, 199–210. Berlin & New York: Mouton de Gruyter. Bennett, J.A.W. & G.V. Smithers (eds.). 1974. Early Middle English verse and prose. Oxford: Oxford University Press. Bergs, Alexander. 2005. Social networks and historical sociolinguistics. Studies in morphosyntactic variation in the Paston letters, 1421–1503. Berlin & New York: Mouton de Gruyter. Bergs, Alexander. 2007. Letters: A new approach to text typology. In Terttu Nevalainen & S.–K. Tanskanen (eds.), Letter writing, 27–46. Amsterdam: John Benjamins. Blake, Norman. 1992. Introduction. In Norman Blake (ed.), The Cambridge history of the English language, Vol. II. 1066–1476, 1–23. Cambridge: Cambridge University Press. Cerquiglini, Bernard. 1989. Éloge de la variante. Paris: Cerf. Chafe, Wallace. 1984. Speaking, writing, and prescriptivism. In Deborah Schiffrin (ed.), Meaning, form, and use in context, 95–103. Linguistic applications. Washington, D.C.: University of Georgetown Press. Clanchy, Michael. 1993. From memory to written record. Cambridge: Cambridge University Press. Clark, Cecily. 1970. The Peterborough Chronicle. 2nd ed. Oxford: Clarendon. Clark, Cecily. 1992. The myth of the ‘Anglo-Norman scribe’. In M. Rissannen, Ossi Ihalainen, Terttu Nevalainen & Irma Taavitsainen (eds.), History of Englishes. New methods and interpretations in historical linguistics, 117–130. Berlin: Mouton de Gruyter. Cressy, David. 1980. Literacy and social order. Cambridge: Cambridge University Press. Curzan, Anne. 1996. Third person pronouns in the Peterborough Chronicle. Neuphilologische Mitteilungen 3. 301–314. Davis, Norman. 1971. Paston letters and papers of the fifteenth century. Oxford: Clarendon. Eckert, Penelope & John Rickford (eds.). 2002. Style and sociolinguistic variation. Cambridge: Cambridge University Press. Graff, Harvey J. 1987. Labyrinths of literacy. Bloomington: Indiana State University Press. Hanna, Ralph III. 2010. Medieval manuscripts and readers. In Corinne Saunders (ed.), The Wiley companion to medieval poetry, 196–216. Oxford: Blackwell-Wiley. Hanna, Ralph III. 1996. Pursuing history. Middle English manuscripts and their texts. Stanford: Stanford University Press. Hogg, Richard 1992. A grammar of Old English. Oxford: Blackwell. Irvine, Susan. 1994. The Anglo-Saxon Chronicle. MS E. Cambridge: D.S. Brewer. Lass, Roger. 1992. Phonology and morphology. In Norman Blake (ed.), The Cambridge history of the English language, Vol. II. 1066–1476, 23–156. Cambridge: Cambridge University Press. Lass, Roger. 1997. Historical linguistics and linguistic change. Cambridge: Cambridge University Press. McIntosh, Angus, Michael Louis Samuels & Michael Benskin, with the assistance of Margaret Laing and Keith Williamson. 1986. The linguistic atlas of late medieval English. Aberdeen: Aberdeen University Press.

258   

   Alexander Bergs

McIntosh, Angus. 1989 [1963]. A new approach to Middle English dialectology. In Angus McIntosh, Michael Louis Samuels & Margaret Laing (eds.), Middle English dialectology: essays on some principles and problems, 22–31. Aberdeen: Aberdeen University Press. McIntosh, Angus. 1989 [1956]. The analysis of written Middle English. In Angus McIntosh, Michael Louis Samuels & Margaret Laing (eds.), Middle English dialectology: essays on some principles and problems, 1–22. Aberdeen: Aberdeen University Press. Milroy, James. 1992a. Middle English dialectology. In Norman Blake (ed.), The Cambridge history of the English language, Vol. II. 1066–1476, 156–204. Cambridge: Cambridge University Press. Milroy, James. 1992b. Linguistic variation and change. Oxford: Blackwell. Nevalainen, Terttu & Helena Raumolin-Brunberg (eds.). 1996. Sociolinguistics and language history: Studies based on the Corpus of Early English Correspondence. Amsterdam: Rodopi. Phillips, Betty. 1995. Lexical diffusion as a guide to scribal intent. A comparison of ME and spellings in the Peterborough Chronicle and the Ormulum. In Henning Andersen (ed.), Historical Linguistics 1993, 379–386. Amsterdam: John Benjamins. Samuels, M.L. 1972. Linguistic evolution – with special reference to English. Cambridge: Cambridge University Press. Smith, Jeremy. 1996. An historical study of English. Oxford: Blackwell. Stockwell, Robert & Westbrook Barritt. 1961. Scribal practice: Some assumptions. Language 37(1). 75–82. Weinrich, Uriel, William Labov & Marvin Herzog. 1968. Empirical foundations for a theory of language change. In Winfred P. Lehmann & Yakov Malkiel (eds.), Directions for historical linguistics, 95–195. Austin: University of Texas press.

Generation 4

William IV ’ Bridget Heydon (?1479-1554)

’ Agnes Morley

Edmond II ’ Katherine Spellman (?1443-?1504) ’ Margaret Monceaux

Margery ’ Richard Calle

Elizabeth ’ Robert Poyning (?1429-1488)

William Paston I ’ Agnes Berry (1378-1444) (?1400-1479)

Edmond I (1425-1449)

John III ’ Margery Brews (1444-1504) (?1455-1495)

Christopher

John II (1442-1479)

Generation 3

(1421-1466) (?1420-1484)

John I ’ Margaret Mautby

Generation 2

Generation 1

Walter William III (?1456-1479) (?1459-after 1504)

Clement II (1442-?1479)

Family members of whom letters have survived are printed in boldface.

Anne ’ William Yelverton

William II ’ Lady Anne Beaufort (1436-1496)

Geneological tree and biodata of the Pastons family

Writing, reading, language change       259

Esther-Miriam Wagner

14 Challenges of multiglossia: scribes and the emergence of substandard JudaeoArabic registers Sources: The letters analysed in this paper were found amongst thousands of other letters in a storeroom in a synagogue in Old Cairo at the end of the 19th century and were brought from Egypt to Cambridge University Library, where they now form part of the Cambridge Genizah Collections, which contain 200,000 fragments of manuscripts and early printed works. The letters were written from the 11th to the 19th centuries by members of Jewish communities of the Middle East, North Africa and Spain, mostly merchants or professional scribes. All letters are described in greater detail in Wagner (2010, 13–24). Abstract: The dichotomy between written and spoken Arabic, which arose out of historical circumstances during the Islamic conquests, is one of the most striking features of Arabic, resulting in a multitude of literary and spoken registers. This dichotomy can also be found within the Jewish sociolect of Arabic, Judaeo-Arabic, which is the name given to a variety of Arabic language forms used by Jews that differ from the sociolects employed by their Muslim and Christian neighbours. Written Judaeo-Arabic differs markedly from Classical Arabic, the prescribed standard written variety. In particular epistolary Late Judaeo-Arabic exhibits a large number of vernacular forms, which are typically only used in spoken registers and ordinarily never enter the realm of written Arabic. Not only vernacular forms occur; there are also a large number of literary elements, often used hypercorrectly when judged by the rules of Classical Arabic, which have become standardised. The employment of this peculiar mix of typical spoken and written elements creates an exclusive epistolary register: the deliberate use of distinctly literary forms marks the register as literary, whereas the use of colloquial forms gives it the necessary authenticity and intimacy. Particular agents of these artificial epistolary registers appear to be merchants, who had the need to rely on the cordial relations with the addressees of their letters in their business dealings and thus utilised vernacular forms to convey a kind of intimacy between them and their trading partners, but who were also familiar with inherited Classical Arabic literary conventions and forms, which they incorporated into their language to elevate their epistolary register and give it a literary tinge.

262   

   Esther-Miriam Wagner

1 Written, spoken and so-called “mixed Arabic” The dichotomy between written and spoken Arabic is one of the most striking features of Arabic, and was termed “diglossia” by Ferguson (1959), which is still the most widely used term among Arabists and linguists. Hary (1992 and 2003), however, has suggested nomenclature such as multiglossia and continuglossia to describe the linguistic situation more accurately, as there are other poles within the Arabic spectrum beside the dichotomy between written standard and vernacular. As a result of the Islamic conquests in the 7th century, Islam, and Arabic with it, spread over a vast area in relatively short time. Whilst various spoken varieties of Arabic developed in different geographical regions, the prescribed written variety remained more or less the same for all Arabic speakers. This literary standard, Classical Arabic, probably emerged from a register associated with poetry and divination and is the language variety in which the Qur’an and the religious literature were codified in the first centuries of Islam. The spoken Arabic at the time just before the conquests is by some scholars thought to be grammatically similar to Classical Arabic, with subsequent changes due to “pidginisation and creolisation” (Versteegh) of the language leading to the emergence of Neo-Arabic dialects.¹ It has, however, become clear in recent years, for example through works by Troupeau (2008), that this assumption is untenable.² Although there probably was a high-level spoken register associated with poetry and divination featuring many forms of the so-called Classical Arabic linguistic repertory, a lower register of spoken Arabic has to be assumed for the ordinary vernacular of Arabic speakers. As already mentioned, we are not only dealing with a spectrum of two poles, standard written and spoken Arabic; a multitude of distinct registers exist between those two extremes, both spoken and written varieties. Spoken registers may be heavily influenced by written standards, for example intellectuals may choose to converse with each other in a register mostly based on the written variety, and educated people from different Arabic-speaking countries may converse in a form closely related to Standard Arabic for the purpose of mutual understanding. At the same time, forms from the spoken language may enter the literary varieties: authors may employ these colloquial forms to connect with their readership, to

1 See for example Versteegh (1984), in particular pp. 2–3. 2 Troupeau (2008: 470) argues that the Classical Arabic grammatical case endings (iʿrāb), although a feature of the register used for poetry and divination in pre-Islamic times, were probably never part of the dialects spoken by the Arab bedouins.

Scribes and the emergence of substandard Judaeo-Arabic   

   263

portray their protagonists in a specific light. Therefore the language used in any context is greatly dependent on the interaction between any two interlocutors, or between writer and reader, which leads to a large number of distinct spoken and literary varieties.³ Due to segregation of religious groups, we also find particular sociolects or “religiolects”⁴ being spoken, most famously in the diverging dialects of Muslims, Jews and Christians in Baghdad. Here the Christian and Jewish varieties were based on an older, urban spoken Arabic, and the Muslim sociolect on a newer, rural spoken variety. This also applies to written varieties. Jews and Christians in particular were not religiously obliged to use Standard Arabic in the same way that Muslims were. An additional factor can be found in the desire of Christian and Jewish speakers to segregate themselves linguistically from the Muslim population and to create their own sociolect, thus providing the speaker with a social identity and enforcing ties within the social network.⁵ And in the case of the Jews, there was another very significant aspect: they wrote Arabic not in Arabic script, but with Hebrew characters.

2 Epistolary Judaeo-Arabic The present article is concerned with this written Jewish variety of Arabic, in particular with the type of language used in letters. Although all written Jewish registers display differences, to various degrees, to the Muslim written varieties, letters make for particularly rich research material because utility prose is closer to the spoken vernacular than literary texts. In addition, the more literary writing may find itself – consciously or otherwise – following existing traditions and conventions popular in contemporary Muslim Arabic literature. Although epistolary writing often reflects the vernacular more closely than any other written texts, Judaeo-Arabic letters also exhibit features that are distinctively Classical Arabic in origin. Furthermore, some phenomena are neither vernacular nor Classical Arabic features but represent varieties somewhere along the Arabic continuum. Some of these can be categorised as part of substandard writing, i.e. belonging to a standardised variety that does not conform with the

3 A good introduction to the topic can be found in Diem (2006). 4 This term was introduced by Hary (2009: 12–13). 5 For this phenomenon, compare Milroy and Milroy’s (1992) study on Irish sociolects.

264   

   Esther-Miriam Wagner

generally accepted, prescriptive and prestigious Classical Arabic standard, in general, or as part of the epistolary register in particular.⁶ The Judaeo-Arabic letters investigated in this article all come from the Cairo Genizah collections, and were part of a corpus investigated in my PhD dissertation.⁷ The letters range from the 11th century to the beginning of the 19th century, with considerable differences in style, orthography, morphology and syntax. In the following, the use of three typical epistolary phenomena, two examples of distinct literary elements and one of a typical colloquial form, will be investigated, to show the sociolinguistic intricacies of Judaeo-Arabic epistolary writing as a mixed register between the spoken language and literary forms.

2.1 The relative pronoun allaḏī / əldi The Classical and Modern Standard Arabic relative pronoun is allaḏī, which also has separate feminine, dual and plural forms, and is inflected in the dual and plural. The vernacular, on the other hand, has the relative pronoun illi, which is uninflected and serves for all genders and numbers. The relative pronoun used in late epistolary Judaeo-Arabic is a mischform of these two mentioned relativisers. On the one hand, the pronoun is found with Classical Arabic orthography, ‫ אלדי‬in Hebrew script based on Arabic script . On the other hand, it adopts the morpho-syntactical properties of the vernacular pronoun illi and remains uninflected for all genders and numbers. In short, vernacular illi is basically replaced by a classicised, more literary, relative pronoun. In the society that provides the context for Judaeo-Arabic writing, there is however not only a contrast between spoken and written language, but there is a third layer of discourse, the so-called reading tradition of texts, an integral part of multilingual communities and/or societies with long traditions of oral recitation.⁸ From vocalised Judaeo-Arabic texts, such as T-S Ar.18(1).113, we learn that the reading of this classically spelled ‫ אלדי‬was in fact something like əldi and not allaḏī, so even the pronunciation is a hybrid of classical and vernacular Arabic. Therefore, whereas the following example resembles Classical Arabic orthogra-

6 For a more detailed account, see Wagner (2010), in particular pp. 3–8. 7 See Wagner (2010) for Judaeo-Arabic language history, methodology, historical background and further information about the scribes themselves. 8 For the long Middle Eastern history of these reading traditions, see Khan (2012), in particular Chapter 6, for Hebrew, and Gershevitch (1979) for Aramaic and Persian. For a European equivalent, compare Roger Wright’s article in this volume.

Scribes and the emergence of substandard Judaeo-Arabic   

   265

phy, we can assume that the pronoun was probably pronounced əldi when the letter was read aloud. (1) ‫ וכיף גמילכם אלדי אשמלני וסתרני‬CUL Or 1080 J35/28 wa-kayfa jamīl-kum allaḏī/əldi ašmala-nī and-like good.deed-your which wrap.3sg.past-me wa-satara-nī⁹ and-protect.3sg.past-me ‘like your good deed that wraps me and protects me’ In examples where the antecedent is feminine, dual or plural, it becomes more apparent that we are not actually dealing with a Classical Arabic construction: (2) ‫ וכדלך אלצרה אלדי וצלת צחבתי‬Bod MS Heb d. 47.62/17 wa-ka-ḏālika al-ṣurra allaḏī/əldi waṣalat and-like-this the-purse which arrive.3sg.f.past ‘likewise the purse which arrived with me’

ṣuḥbat-ī with-me

(3) ‫ אלסבאג בן ענד אלדי קבטייתין ואל‬T-S 10J7.2/14f wa-l qibṭiyyatayn allaḏī/əldi ʿinda Bni l-Ṣabbāġ and-the Egyptian.linen.dual which with Ibn as-Sabbāġ ‘the two Egyptian linens that are with Ibn as-Sabbāġ (the son of the dyer)’ These examples demonstrate that while the letters employ Classical Arabic orthography in spelling out the relative pronoun, the morphology/syntax is influenced by the vernacular pronoun. Quite astonishingly, this is a very stable feature of epistolary Judaeo-Arabic. Even letters from the Ottoman period, which are generally written in an extremely colloquial way, employ almost exclusively the Classical Arabic spelling well into the 19th century, giving the letters a literary air. In addition, the reading tradition employs a mischform of vernacular and Classical pronoun in the pronunciation, thereby creating an artificial register of

9 As no standard description system exists for mixed Arabic registers, the transcriptions in this article are composed of a relatively artificial mix of vernacular and Standard Arabic forms to reflect the mix of forms employed in them. Whereas the examples from the Classical Genizah period still largely follow Classical Arabic writing convention, the phrases from Late JudaeoArabic letters show many vernacular forms, and the transcription therefore follows mostly the conventions of the modern Egyptian vernacular.

266   

   Esther-Miriam Wagner

letter recitation, ensuring that letter writing despite its many colloquial elements remains a quasi-literary register.

2.2 The role of the negation particle lam in substandard writing Research into the use of negation particles in any Arabic text reveals remarkable sociolinguistic phenomena. This is caused by the wide array of negation particles that Classical Arabic deploys, regulated by a strict set of rules regarding their morphological, syntactic and semantic context.¹⁰ For example, the negation lam can only negate past actions and must be followed by a certain mood, the apocopate; lan with the subjunctive is reserved for the future; whereas lā can express the prohibitive, a negative wish or the future, depending on the mood of the following verb. The strategies to embed distinctly literary or vernacular particles vary greatly from text to text, and we can usually draw sociolinguistic conclusions from their usage. Mejdell (2008) has stated that in mixed Arabic texts vernacular elements may enter intentionally or as a result of interference, whereas the elements introduced from Standard Arabic are always an intentional choice. Therefore the deliberate use of the vernacular negation particles as opposed to the literary can give us a lot of information about the intentions and aspirations of the writers. For example, in the language of modern newspapers the negator mā, although a Classical Arabic form, is rarely used because mā is the common negation particle used in vernacular speech, and therefore its usage is somewhat tainted and considered “common” and low-prestige.¹¹ The particle lam, on the other hand, does not occur in the dialect as a regular negation particle¹², and is therefore marked as highprestige and a part of educated literary style. In mixed Arabic registers, the use of the negations varies considerably. In Christian Arabic texts analysed by Blau (1966–67), or in the 14th–16th-century Muslim works analysed by Palva (2008), lam occurs only very infrequently and mā is common. In the Christian material, it is possible that mā was the preferred particle because lam was seen as a distinct part of Muslim literary traditions, and

10 For a summary see Fischer (2002: 171–175), Wright (1967: II 299–306) and Brockelmann (1960: 176–180). 11 See Wehr (1953: 38), and similarly, Holes (1995: 195). 12 In Egyptian, lam can be found in the strong sense of ‘never’, but it is disputable whether this is an artificial archaising or poetic usage, see the discussion in Wagner (2010: 142–144).

Scribes and the emergence of substandard Judaeo-Arabic   

   267

the Christians wanted to distance their written idiom from the Muslim by avoiding overtly Classical Arabic usages. In the case of Palva’s material, the lack of lam is probably owing to the narrative prose style employed in the folk stories. On the other hand, in some of the earliest Judaeo-Arabic texts, such as the 9thcentury papyri published by Blau and Hopkins (1987), the predominant negation particle is lam, probably an attempt to mark the language as literary. Whereas lam is less frequent in letters from the classical Genizah period (10th–15th century), it resurfaces as the main negation particle in Late Judaeo-Arabic letters.¹³ This Late Judaeo-Arabic of the Ottoman period is of particular interest for historical linguistics, as by that time many of the older writing traditions, which upheld the standards and practices of Classical Arabic, had largely disappeared. Instead, a number of vernacular forms occur in letters from the 18th–19th century, and the language is far removed from general literary conventions of the time. On the other hand, as was the case with the relative pronouns described in the previous section, we find a number of Classical Arabic forms, apparently used to mark the register as “literary” despite its colloquial forms. The negation particle lam is one of these Classical Arabic forms, and it is used in most negative phrases. It is not only found in places where one would expect it according to Classical Arabic grammar, but it has considerably enlarged its function. For example, it occurs as a nominal negation particle, a function it does not hold in Classical Arabic at all. (4) ‫ ולם הווה עלה חדנה‬AIU VIIE 92/8 wa-lam huwa ʿala ḥad-na and-not he on side-our ‘he is not on our side’ (5) ‫ ואן כאן לם לכום קאביל תערפונה‬T-S NS 99.23/15f wa-in kān lam lu-kum qābil tiʿarrif-na and-if be.3rd.sg.past not for-you inclination you.should.let.know-us ‘and if you do not have the inclination, let us know’ Restrictions imposed by Classical Arabic prescriptive grammar as to the imperfective mood of the accompanying verb are also abolished, and the particle lam may appear in an analytical construction together with the perfect to negate the past:

13 This is a phenomenon it shares with a number of other linguistic features, including orthographical peculiarities, which point to the possible existence of an undercurrent sub-substandard writing tradition, see Wagner (2013).

268   

   Esther-Miriam Wagner

(6) ‫ ולם כאן ילזם נקול לכם‬AIU VIIE139/28 wa-lam kān yilzam niqūl lu-kum and-not be.3rd.sg.past be.necessary.3rd.sg.prs speak.1st.pl.prs to-you ‘it was not necessary that we tell you’ (7) ‫ פי לם חבינא נכתב להום‬AIU VIIE 93/9f fa lam ḥabbēna niktib lu-hum and not love.1pl.past write.1pl.prs to-you ‘we did not want to write (it) for them’ (8) ‫ לם פכרתו פינה‬T-S NS 99.23/38 lam fakkirtu fī-na not think.2pl.past of-us ‘you (pl.) did not think of us’ Whereas lam can only negate past actions in Classical Arabic, the Ottoman letters attest to it also negating the present: (9) ‫ ולאכין אחנה לם נחוב אלא גיר אל טייב‬AIU VIIE 139/28f wa-lākin iḥna lam nuḥubb illa ġayr and-but we not want.1pl.prs only except ‘we do not want anything but the best’

al-ṭayyib the-best

(10) ‫ ואל מווגוד פיל בנצ֗ר לם יגי עשר תכיאס‬T-S 13J25.24/31f wa-l mawjūd fil banḍar lam yīgi ʿašar t-akyās and-the existant in-the town not go.3m.sg.prs ten particle-bags ‘what is available in town does not exceed 10 bags’ (11) ‫ אחנה לם ינפענה דיל שרכייה‬T-S NS 99.23/9 iḥna lam yinfaʿ-na di-l we not benefit.3m.sg.prs-us this-the ‘this partnership does not benefit us’

širkiyya partnership

The particle lam can even be used for the prohibitive and to express a negative wish: (12) ֗‫ וכדה קולנה לכום לם תקרבוה בדיל ע‬T-S NS 99.23/12 wa-kida qulna lu-kum lam tiqarrabu-h bi-di-l ʿereḵ and-thus say.1pl.past to-you not you.should.approach-it with-this-the value ‘thus we told you that you should not approach it at this value’

Scribes and the emergence of substandard Judaeo-Arabic   

   269

(13) ‫ ורבונה לם יקטע יסראל‬AIU VIIE 93/m.7 wa-rabbu-na lam yiqṭaʿ Yisrael and-lord-our not cut.3sg.prs Yisrael ‘our lord shall not cut off (his love towards) Israel’ As we can see, lam in these Late Judaeo-Arabic texts has considerably enlarged the functions it may hold as opposed to its usage in Classical Arabic, where it is restricted to the negation of past events. In various grammatical descriptions of Middle Arabic, such as Blau (1966–1967: 305), the appearance of lam in these non-classical functions has been judged as a hypercorrection, i.e. as a form used by aspiring but ignorant writers trying to employ Classical Arabic forms in an attempt to give themselves and their texts an educated air but failing as they use the forms grammatically incorrectly. Judging by the breadth in which lam is used in the Late Judaeo-Arabic letters, and by the large number of writers employing it almost as the sole negation, we can, however, hardly talk of hypercorrections when it comes to the 18th/19th-century sources. These forms have, rather, become a regular and established part of epistolary writing; they are a conscious choice within the epistolary register. The reason why lam becomes the most commonly used negation particle – even used in nominal negations – may be due to its prestige: because it does not normally occur in the spoken language and its usage in Classical Arabic is so heavily regulated, it is the most prominent literary Arabic negation particle. Because of this marking, the writers of Judaeo-Arabic letters seem to employ lam as a means to “elevate” their language. The most remarkable feature of the above examples, however, is the astonishing mixing of forms: very colloquial forms appear next to the negation particle lam, distinctly marked as high style and literary. In (9) and (11), for instance, we can find the vernacular pronoun iḥna ‘we’ (naḥnu in Classical Arabic) used side by side with lam, whereas in (11) and (12) the vernacular demonstrative pronoun di ‘this’ (in Classical Arabic hāḏā) occurs next to lam. The parallel employment of both distinctly literary and vernacular forms seems to mirror communication strategies generally observable in epistolary writing. According to Koch and Oesterreicher (1986: 18), the language used in private letters is located in the middle ground on the scale between spoken and written forms of language. Writers and readers of letters therefore have to bridge a certain distance imposed on them as opposed to oral communication between two interlocutors. To achieve this, colloquial expressions are used to bring a certain intimacy to the writing; these forms also provide a certain flair of authenticity because they mirror expressions that the writer would actually employ in speech. On the other hand, epistolary language is to a degree a literary medium, too, so the use of the

270   

   Esther-Miriam Wagner

negation particle lam distinctly marked as “high style” provides the necessary credentials for a written form of language. In Clive Holes’ (2008) investigation of Middle Arabic elements in 19th- and 20th-century correspondence between Gulf rulers, one of the most interesting findings was that the combination of dialectal morphemes with Classical Arabic lexical items produces acceptable hybrids, while the mixing of Classical Arabic grammatical items with a colloquial lexicon is apparently unacceptable. A nice parallel in the Late Judaeo-Arabic material is the fact that lam may occur in connection with verbs that have the colloquial prefix bi-, the so-called bi-imperfect. (14) ‫ רייאל‬۲۰ ֗‫ ולם ביירצ֗ו יביעו ע‬T-S 13J25.24/81 wa-lam bi-yirḍu yibīʿu ʿereḵ and-not particle-want.3pl.prs sell-3pl.prs value ‘they do not want to sell at the rate of 20 riyyāl’ (15) ‫ לאן לם בייעגבו אל חאל‬T-S 13J25.24/64f liʾanna lam bi-yiʿjabu(-u) because not particle-like.3pl.prs-it ‘since they do not like the situation’

al the

20 20

riyyāl riyyāl

ḥāl situation

To summarise, since lam is so specialised and does not normally occur in the spoken language, it is marked as distinctly Classical Arabic and therefore literary. Because of this marking, the writers of Judaeo-Arabic letters tend to employ it, probably as a means to “elevate” the language of their letters, to the degree that in Late Judaeo-Arabic it becomes by far the most commonly used negation particle, and can be found in both verbal and nominal negations.

2.3 bi-imperfect While the above mentioned Classical Arabic forms lend a certain air of literariness to the epistolary register, there are also many colloquial forms that find their way into letter writing earlier than into literary works. One of these vernacular forms is the so-called bi-imperfect, which starts to appear in the letters from the 12th century onwards. This bi-imperfect is a common verbal form in the spoken language; in some dialects, such as Lebanese and Syrian Arabic, it has superseded the unmarked imperfect and become the regular present tense form. Because colloquial forms such as this bi-imperfect are marked as such, they are initially barred from use, and they do not appear simply in their vernacular function in the texts. The writers of the earliest examples, rather, seem to be quite

Scribes and the emergence of substandard Judaeo-Arabic   

   271

aware that this is colloquial form, and they shy away from using it. It is not surprising therefore that, in the earliest example we do find (16), the bi-imperfect is employed in a pejorative way, in the reported direct speech of an angry man. The use of the bi-imperfect here may have served to portray the young man in a bad light, as a brute and uneducated man. On the other hand, the use of a colloquial form can serve to convey a sort of intimacy, to show one’s familiarity with the addressee or one’s (pretended) humbleness or plainness. Many of these early forms in the corpus are in the first person, which is probably not coincidental. (16) ‫אד לם אגוזהא ואלא באקתל אחדהם‬ iḏ lam ujawwiz-hā wa-illā if not marry.1.sg.prs-her and-except aḥad-hum¹⁴ one.of-them ‘if I don’t marry her, I will kill one of them!’

b-aqtul particle–kill.1.sg.pres

(17) ‫ ומא באעלם כיף תבין אלאמור‬TS 13 J 20.24/v.3 wa-mā b-aʿlam kayfa tubayyin al-umūr and-not particle-know.1.sg.prs how explain.2m.sg.prs the-matters ‘I do not know how you will explain the matters’ (18) ‫ ולא באטאלבה בשי‬GW VIII/24 wa-lā b-uṭālibu-h and-not particle-claim.1.sg.prs-him ‘I am not claiming anything from him’

bi-šay with-anything

Judging by these examples, we can see a clear pattern: the bi-imperfect is used in certain situations, where it either reports direct speech, perhaps in a pejorative way, adds a spoken flavour to the text, or serves in a communicative function in which the writer uses the bi-imperfect in the first person to express his humbleness or close connection to the reader. These forms seem to set a precedent and “open the gates” for the form to be used regularly in the language of correspondence, to the point that it becomes widely used in Late Judaeo-Arabic letters with the very same functions as in the spoken language.

14 From the Genizah collection in the National library in Jerusalem. Published in Qiryat Sefer 1 (1924: 55–61), but no siglum is given.

272   

   Esther-Miriam Wagner

3 The role of merchants in the setting of precedents Most of the letters analysed in the corpus for my PhD dissertation, which also forms the basis for my investigations in this article, were written by Jewish traders. Compared to contemporary material written by other individuals, both in private and official letters, those composed by merchants seem to form very homogeneous corpora within themselves. They also appear to show a number of typically registerial forms lacking from other correspondence, i.e., forms which only appear at certain times, in certain geographical areas, and are apparently limited to business correspondence. One of these registerial forms is, for example, the relative particle an, which is originally derived from an indefinite accusative ending. This an first occurs in sentences such as wa-naḥnu taḥta ḥāl an ʿaẓīm ‘while we (were) in a terrible state’, to form a attribution in ḥāl an ʿaẓīm between noun (ḥāl ‘state’) and adjective (ʿaẓīm ‘terrible’), in a place where the original accusative ending could have theoretically occurred (although not in this concrete example, as the preceding preposition requires a genitive ending). From there it appears to have spread in construction such as katabtu ilayka kutub an ʿidda ‘I wrote you a number of letters’, in which it forms an attributival link between a noun (kutub ‘letters’) and another (ʿidda ‘number’). It finally appears as relativiser in attributival clauses, a function it may have acquired because of the homophonous complementiser an ‘that’, e.g. mā lī ṯawb an nalbas ‘I don’t have any garment to wear’ (ṯawb ‘garment’ an nalbas ‘I could wear’); it appears to be particularly favoured in clauses with a temporal referent, e.g. min yawm an anfaḏtuhā ‘from the time I send them’ (yawm ‘day’ an anfaḏtuhā ‘I sent it’). These constructions were most likely never part of the spoken language; rather, they appear exclusively in the written language, and are largely limited to the 11th century and to epistolary writing of Egyptian traders, in particular. It is probably no coincidence that many innovative forms are encountered in letters written by the Jewish merchants. Business correspondence differs from other documentary sources insofar as it is usually composed in an immediate act of writing, outside of the “corset” of formulae that restricted, for example, chancery writers and trained scribes. Traders were required to react and write immediately, to pen copious amounts of correspondence, and they had to write expressively and very precisely to be understood in their instructions. In addition, merchants often adopt a distinctive argot, having a need to develop coded or secret language to disguise the specialised knowledge employed in their trade, and, indeed, can pride themselves in working with forms of language that mark them out as a separate social group. Many good examples of such

Scribes and the emergence of substandard Judaeo-Arabic   

   273

spoken mercantile languages can be found; from the relatively limited market argots used among Yemeni businessmen (Walter 2003 and the goldsmiths in Cairo (Khan 1995–1997), to a horse and cattle traders’ sociolect, related to other Yiddish-influenced argots such as Rotwelsh, Yenish and masematte, that emerged in Germany (see Guggenheim-Grünberg 1954) and even spread to the US (see http:// www.jochnowitz.net/Essays/WesternYiddish.html). Particular written forms of mercantile language developed too: Yiddish-speaking merchants in 17th-century Prague resorted to encoding parts of their business letters in a secret alphabet (Landau 1911: xxx, 63–64), and Italian traders had developed their own style of handwriting, mercantesca, distinct from the contemporary chancery hands cancelleresca, by the 14th century (Ceccherini 2009). Traders from all periods encounter particular circumstances in the act of writing, as they typically have to react quickly to business matters and compose a large number of letters. This leads to very short-term acts of writing and usually to texts that are not as formulaically restricted as documents. In addition, traders have the desire, and the self-confidence as a distinct and often important social group, to develop their own registers, in speech and in writing. Thus it is among the Jewish merchants of the Judaeo-Arabic letters investigated in this article. They chose to compose their correspondence in language varieties that did not conform with the prescriptive “high” writing standards of the time, but which formed a homogenous and relatively standardised written register in itself.

4 Summary and conclusions The diglossia of Arabic is responsible for the continual emergence of a multitude of grammatically quite diverse spoken and written registers, and also facilitated the extreme divergence of some of the sociolects employed by the members of different religions. Epistolary Judaeo-Arabic, in particular in Late Judaeo-Arabic, shows an extremely mixed register made up of features borrowed from the spoken language and literary forms. Three typical epistolary Judaeo-Arabic phenomena were investigated in this article, which serve to demonstrate the mixed nature of the register used in letters as situated on a scale of “literarity” between the spoken language and the most literary forms. The first concerned the relative pronoun, which retains its Classical Arabic orthography, but takes over the morphosyntactical properties of the vernacular pronoun illi and remains uninflected for all genders and numbers. Vernacular illi is, therefore, basically represented by a classicised, more literary relative pronoun, while being pronounced əldi in the reading tradition. The

274   

   Esther-Miriam Wagner

second feature, the negation lam, is marked as distinctly Classical Arabic and employed as a means of “elevating” the epistolary register, to the degree that it becomes the main negation particle in Late Judaeo-Arabic. On the other hand, colloquial forms, such as the bi-imperfect, appear to first occur in letter writing to express direct speech, and to convey familiarity between writer and addressee. The employment of this peculiar mix of literary and vernacular forms side by side in epistolary Judaeo-Arabic creates a special register used exclusively in letters. Whereas the deliberate use of distinctly literary forms, such as the Classical Arabic relative pronoun allaḏī (ǝldi) and the negation lam, marks the register as literary, the use of colloquial forms, such as the bi-imperfect, gives it the authenticity and intimacy necessary to reduce the distance that the act of letter writing inevitably engenders. Merchants play an important role as agents of language change as the creation process involved in writing business correspondence facilitates the occurrence of precedents and the introduction of innovative forms. These traders also needed to rely on the cordial relations with the addressees of their letters in their business dealings and thus utilised vernacular forms to convey a sense of intimacy between them and their trading partners, in a business world whose economy was largely build on trust, familiar connections and cordial relationships. On the other hand, they were also familiar with inherited Classical Arabic literary conventions and forms, which they incorporated into their language to elevate their epistolary register and mark it as distinctly literary.

5 References Blau, Joshua. 1966–67. A Grammar of Christian Arabic Based Mainly on South-Palestinian Texts from the First Millennium. Louvain: Secrétariat du CorpusSCO. Blau, Joshua & Hopkins, Simon. 1987. Judaeo-Arabic Papyri. Jerusalem Studies in Arabic and Islam 9. 87–160. Brockelmann, Carl. 1992. Arabische Grammatik. 24th edition. Leipzig: Langenscheidt. Ceccherini, Irene. 2009. Merchants and Notaries: Stylistic Movements in Italian Cursive Scripts. Manuscripta 53. 239–83. Diem, Werner. 2006. Hochsprache und Dialekt im Arabischen. Untersuchungen zur heutigen arabischen Zweisprachigkeit. Abhandlungen für die Kunde des Morgenlandes Band XLI, 1 (2nd edition). Wiesbaden: Harrassowitz. Fischer, Wolfdietrich. 2002. Grammar of Classical Arabic. New Haven: Yale University Press. Gershevitch, Ilya. 1979. The Alloglottography of Old Persian. Transactions of the Philological Society 77. 114–190.

Scribes and the emergence of substandard Judaeo-Arabic   

   275

Guggenheim-Grünberg, Florence. The horse dealers’ language of the Swiss Jews in Endingen and Lengnau. In Uriel Weinreich (ed.), The field of Yiddish, 48–62. New York: Linguistic Circle of New York. Hary, Benjamin. 2009. Translating Religion. Leiden: Brill. Holes, Clive. 1995. Modern Arabic. Structure, Function and Varieties. London: Longman. Holes, Clive. 2008. The “mixed” Arabic of the letters of 19th and early 20th century Gulf rulers. In Jérôme Lentin & Jacques Grand’Henry (eds.), Moyen arabe et variétés mixtes de l’arabe à travers l’histoire, 193–229. Louvain-la-Neuve: Peeters Press. Khan, Geoffrey. 1995–1997. A note on the trade argot of the Karaite goldsmiths of Cairo. Mediterranean Language Review IX. 74–76. Khan, Geoffrey. 2012. A short introduction to the Tiberian Masoretic Bible and its reading tradition. Piscataway: Gorgias. Landau, Alfred. 1911. Jüdische Privatbriefe aus dem Jahre 1619. Wien/Leipzig: Wilhelm Braunmüller. Lentin, Jérôme. 2008. Unité et diversité du moyen arabe au Machreq et au Maghreb: quelques données d’après des textes d’époche tardive (16ème–19ème siècles). In Jérôme Lentin & Jacques Grand’Henry (eds.), Moyen arabe et variétés mixtes de l’arabe à travers l’histoire, 305–319. Louvain-la-Neuve: Peeters Press. Mejdell, Gunvor. 2008. “Middle Arabic” across time and medium/mode – Some reflections and suggestions. In Jérôme Lentin & Jacques Grand’Henry (eds.), Moyen arabe et variétés mixtes de l’arabe à travers l’histoire, 355–372. Louvain-la-Neuve: Peeters Press. Milroy, Lesley & James Milroy. 1992. Social Network and Social Class: Toward an Integrated Sociolinguistic Model. Language in Society 21/1. 1–26. Palva, Heikki. 2008. Notes on the language form of some 14th–16th century Arabic manuscripts, written in Hebrew characters. In Jérôme Lentin & Jacques Grand’Henry (eds.), Moyen arabe et variétés mixtes de l’arabe à travers l’histoire, 373–389. Louvain-la-Neuve: Peeters Press. Troupeau, Gérard. 2008. Réflexions sur la nature de l’ ʾiʿrāb. In Jérôme Lentin & Jacques Grand’Henry (eds.), Moyen arabe et variétés mixtes de l’arabe à travers l’histoire, 457–471. Louvain-la-Neuve: Peeters Press. Versteegh, Kees. 1984. Pidginization and Creolization: The Case of Arabic. Amsterdam: John Benjamins. Wagner, Esther-Miriam. 2010. Linguistic variety of Judaeo-Arabic in letters from the Cairo Genizah. Leiden: Brill. Wagner, Esther-Miriam. 2013. The crude, the late and the marginal: Diversity in Judaeo-Arabic letters. In Maria Angeles Gallego and Juan Pedro Monferrer (eds.), The Semitic Languages of Jewish Intellectual Languages. Proceedings of the Memorial Conference for Dr Friedrich Niessen, forthcoming. Leiden: Brill. Walter, Mary Ann. 2003. Kalaam, Kalaarbaam: An Arabic Language Game in Hadramaut, Yemen. Paper given at the 10th Annual Symposium about Language and Society – Austin (SALSA). Downloadable at: http://faculty.wcas.northwestern.edu/~maw962/docs/salsapaper.pdf Wehr, Hans. 1953. Zur Funktion arabischer Negationen. Zeitschrift der Deutschen Morgenländischen Gesellschaft 103. 21–39. Wright, William. 1967³. A Grammar of the Arabic Language, Translated from the Grammar of Caspari and Edited with Numerous Additions and Corrections by W. Wright. Paperback edition. London: Cambridge University Press.

Ivar Berg

15 Variation in a Norwegian sixteenthcentury scribal community Sources: The texts to be discussed were written by scribes of Scandinavian origin employed by the archdiocese during the period 1510–1537. Official letters and charters have sometimes been preserved by the recipients, whereas many drafts and various administrative notes have survived in the diocesal archive, which the last Catholic archbishop brought with him when he fled to the Netherlands following the Reformation in 1537. This archive is today kept in the Norwegian National Archive in Oslo, together with most other documents written in Norway in the Middle Ages. Abstract: Since the political union with Denmark left Norway with no central administration in the late Middle Ages, the chancery of the archbishop in Trondheim was the only scribal milieu of any significance in the country. From 1510 onwards Danish replaced Norwegian as the language used by the archdiocese, and the scribes had to find their way between Norwegian tradition, Danish models and their own spoken language (lists of scribes show that they were of diverse geographic origin). In a time of no regularised orthography, we will expect to find individual variation and interference from the spoken language of the scribes. However, mutual influence between the scribes should eventually result in a more or less heterogeneous orthographic norm. Of the texts in Scandinavian (many charters were written in Latin, especially those addressed to the Holy See), the more formal ones appear to conform to contemporary Danish, whereas less formal writings, especially bookkeeping only meant for administrative purposes, show far more Norwegian interference. The correlation between stylistic level and dialectal variation is striking, and seems more important than individual variation. I claim that such register variation is a way to gain indirect access to language attitudes in written corpora. The systematic difference between stylistic levels, between “pure” Danish and a language with Norwegian traits, demonstrates that there existed within the chancery an idea about what language was suitable for official use; a meta-linguistic awareness of “good” and “bad” language. There also seems to be a development in time, as many of the first writings after the language shift appear more normal Danish. Thus the scribes apparently found a middle road between Norwegian and Danish as they got used to the language.

278   

   Ivar Berg

1 Introduction This contribution will present some of the findings of a PhD project in which I have studied the chancery of the archbishop in Nidaros/Trondheim¹ in late medieval times as a scribal community. Here, I shall address the shift to Danish as the written language in Norway and examine especially the orthographic variation in the early 16th century in relation to dialectal and register variation. I hope to demonstrate how a comparison between orthographic variation and register variation may give us indirect access to language attitudes in written corpora. First I shall have to give a historical backdrop of the late Middle Ages in Norway and discuss some principles regarding the study of orthographic variation, before I move on to patterns of evidence observed and try to draw some conclusions, both on my specific case study and more generally on the use of written sources in language history. During the late Middle Ages Norway lost its independence and came to be dominated by its more powerful neighbours, Sweden and Denmark, through various political unions; after 1380 no king resided permanently in Norway. Power, as well as the writings of a growing administration, was increasingly centralised in Copenhagen. The decline was cultural as well as political. After about 1350 hardly any literary works were written within the Norwegian borders, writing being restricted to administrative and legal functions, mostly documenting economic transactions. The Norwegian borders corresponded roughly to those of the Nidaros archdiocese, and thus the Catholic Church became the strongest national power, as shown by the archbishop’s political role as leader of the Council of the Realm (Riksrådet). As the political unions left Norway with no central administration in the late Middle Ages, the chancery in Trondheim came to be the only significant scribal milieu in the country, and by far the largest such institution in Norway until the Reformation in 1537. It dealt not only with religious activities, but also with the administration of the archdiocese’s vast properties throughout Norway and the fief administered on behalf of the king.

1 The archdiocese was called Nidaros after the city’s original name, however, in the late Middle Ages the city itself was usually called Trondheim.

Variation in a Norwegian sixteenth-century scribal community   

   279

1.1 The language shift The political development and the prestige of the language associated with political power were important elements in the language shift,² as Danish replaced Norwegian as the written language used in Norway during the late 15th and early 16th  century (Mæhlum 2005; Haugen 1976: 245–248 and 329–332; the classic account in Norwegian is found in Indrebø 2001.) The Danish kings put their own men into important administrative posts in Norway, and together with a great deal of intermarriage within the Scandinavian nobility, this led to an increasing number of the higher classes – those associated with writing – being Danes using Danish. But Norwegians too started to write Danish. This happened either as a sudden shift, which was mostly the case in institutions (such as the dioceses), or by introducing Danish elements into basically Norwegian texts in a gradually increasing amount, so that at some point the language must in fact be considered Danish. In the Trondheim chancery the shift appears to have been a sudden one, as the new archbishop in 1510, the Dane Erik Valkendorf, ostensibly enforced Danish as the official language of the see. The scribes then had to find their way between the local tradition and dialect and Danish written models. The language shift shows a dimension of successive domain loss, starting in the highest social layers (see Mæhlum 2005: 1912–1913). Public letters from the king are written in Danish from 1450; the highest level of Civil Service (where we find many of the immigrated Danes) shifted during the following decades and the church followed suit around 1500 (the various sees at slightly different times). Among local civil servants and judges Norwegian was in use well into the 16th century, nevertheless, with only a few rural exceptions, Danish was the only available written code in Norway by around 1530.

1.2 The scribes and their spoken language The scribes of the ecclesiastical chancery were a mixed bunch. Salary lists preserved for the archdiocese during 1532–1537 (published in Seip 1936) show a staff of about twenty employees mentioned with the title “scribe” (scriffwer). Most of them worked at the archbishop’s palace in Trondheim or at his nearby castle, a few at his residence in Bergen. Some of them had nicknames showing that they were of foreign origin, like Jens jwte, perhaps from Jutland, although jute was

2 I use the term “language shift” here, although Norwegian and Danish were certainly dialects from a strictly linguistic point of view (see below).

280   

   Ivar Berg

used about other Danes as well; Michil helsing, Helsingland is a historical district in Sweden; and we even find one Henrick tysk, ‘German’. Nevertheless, the majority were probably Norwegians recruited from the cathedral school in Trondheim. The difference between Norwegian and Danish was dialectal in a linguistic sense and the two varieties were mutually intelligible, however, there were some isoglosses that distinguished most Norwegian dialects in the early 16th century from most Danish. Some important points on similarities and differences can be summarised as follows (see Brøndum-Nielsen 1950–1973; Hansen 1962, 1971 for details on Danish): Diphthongs: Danish early on monophthongised the proto-Scandinavian diphthongs, whereas they were retained in Norwegian. Unstressed vowels: The unstressed vowels were all reduced to schwa in Danish, usually written ‹e›. Lenition: The unvoiced plosives /p, t, k/ changed to /b, d, g/ in inter-vocalic or post-vocalic word-final environments in Danish. Flexional morphology: Norwegian retained more of the old Scandinavian inflections than did Danish. Syntax: Syntactically, the differences were small and negligible. Lexicon: There were notable differences in the lexicon, both in inventory and in certain word forms. Some of these phonological changes affected Norwegian dialects, but only later and marginally (for instance lenition on the south-western coast, monophthongs in the south-east), and not the dialects in the Trondheim area. Although there is no way of establishing with certainty the scribes’ geographic and dialectal background, it seems safe to assume that they spoke a dialect which differed from Danish in the aforementioned features. Thus, they could not produce written Danish based on their own dialect, but had to learn how to spell.

2 Orthography, norm, and variation Medieval orthography was not regular and standardised in the modern sense, and we find individual variation and dialectal spellings. However, “the scribes did not spell ‘at will’, but according to variable (and historically mixed) conventions”, as Milroy (1992: 134) remarks on the orthography of Middle English. Their point of departure was a set of sound–letter correspondences, yet the range of variation was much wider than we are used to. As a counter-force to this variation

Variation in a Norwegian sixteenth-century scribal community   

   281

we expect that mutual influence between the scribes of a chancery would result in some kind of orthographic norm, more or less heterogeneous. In this way our modern standard languages emerged out of the central chanceries of Europe in the early modern period (see Vandenbussche and Deumert 2003). “Norm” in a medieval context cannot be understood in the same manner as we use the term today: “Das Mittelalter kennt noch keine Normierung durch Kodifizierung, sondern nur spontane Normierung” (Goossens 1994: 84). The codification of standard languages by grammarians and academies belongs to modern times, whereas a common understanding of how a written text should look like naturally developed within scribal milieus, both based on education and cooperation in daily practice (“spontane Normierung” in Goossens’ terms). This distinction can also be given terminological expression by “standard” and “norm” respectively. The relationship between the two notions may be expressed as follows, from a description of the development of Modern English by Nevalainen and van Ostade (2006: 288): “[S]tandardisation is often facilitated by the prior development of suitable supralocal norms, being, as it were, superimposed upon them.” A local norm in use in a chancery (or similar institution) may gain more widespread use and become established as a supralocal norm, which later may be codified as a standard. An important point for the present context is that the language shift in Norway cannot be viewed as a continuous development of the written language, but rather as a break in tradition. This situation is parallel to those described by Goossens (1994) for Occitan in southern France and Low German in northern Germany; neither evolved into, but was rather replaced by, French and High German respectively. The language written in early 16th-century Norway was much closer to Danish than to the Norwegian tradition going back to Old Norse, which to a certain degree prevailed in other text types and regions than those discussed here. Thus I believe it to be a sound methodological assumption that influence from older Norwegian writings is negligible, and that orthographic features which agree with spoken Norwegian and differ from Danish should be treated as dialectal interference and not as written loans from older Norwegian writings. This assumption cannot be made for genres that follow older writings more closely, as certain juridical matters.³

3 King Magnus VI’s national law from 1274 in Old Norse was in force and continuous use until 1604, and legal language used many fixed phrases from the law code. However, such texts are not treated here.

282   

   Ivar Berg

2.1 Style and dialect Generally, non-standard features are more likely to surface in informal genres than formal ones (see for example Finegan and Biber 2001; Sebba 2007). Even if the term “standard” is problematic in my context, the scribes did attempt to write some sort of Danish, and norwegianisms were deviations from this. If we accept this assumption, then studying texts on different levels of formality may give us indirect access to language attitudes, as the most formal texts should be in what the scribes considered to be the “best” language. Differences between a draft and the final version of a document may also be useful when both are preserved, as the final version presumably was more elaborate and carefully written. The two works cited above both deal with present-day language and society, and we must address the question of how far we may press the old axiom of “using the present to explain the past” (as formulated e.g. in Labov 1994). The basis for socio-historical linguistics⁴ must be the uniformitarian principle “that the linguistic forces which operate today and are observable around us are not unlike those which have operated in the past” (Romaine 1982: 122; compare the discussion in Labov 1994: 21–25). The problem is of course that the social forces governing modern society are available to us in a totally different manner than those of the past, known only from fragmentary records. The relationship and interaction between language and social factors, which is the sociolinguistic object of study, is only available to us as far as we can reconstruct the social setting of language. Thus, the uniformitarian principle is a working hypothesis, not because it might offer the ultimate answer, but simply because we have no other point of departure. And its application depends “on locating points of contact and similarity between the present and the past” (Labov 1994: 20). I believe that some kind of stylistic or register variation is such a point of contact and similarity. Roberge (2006: 2311) claims that “social and stylistic variation may be inferred from text type (private versus public), hypercorrection, and authorial intent.” Following this, I have tried to group texts according to their function (“authorial intent”) or genre (“text type”), e.g. internal administrative writings, official declarations, letters to various addressees, and drafts or copies of such texts. I have then searched for orthographic patterns within and between groups, to see if the hypothesis of a correlation between dialectal interference and level of formality is reflected in the corpus (compare the method in Romaine 1982, which also looked at stylistic levels).

4 Or historical sociolinguistics, both terms are used (see for example Roberge 2006).

Variation in a Norwegian sixteenth-century scribal community   

   283

The stated connection between language and level of formality also implies that variation between several versions of the same text, or between texts with the same function, presumably was regarded as unmarked. When two texts were supposed to carry the same meaning, orthographic variation between them must have been considered neutral, whereas the occurrence of a feature only in informal texts indicates that it was stigmatised, or at least considered unsuitable for more formal text types.

2.2 Orthographic and “real” variation Variation in a written corpus may be of two distinct types: a) purely orthographic, or b) reflecting phonological reality. The first type points to the allowed variation within the norm and is interesting as a sign of the stage in the standardisation process, whereas b) is the traditional interest of linguists. In language change two variants of a variable will usually fluctuate for some time, before one of them gains the upper hand. This applies regardless of whether the change happens in the spoken language or only in writing. I shall briefly consider type a), nevertheless, my main focus is variation that points to the aforementioned linguistic differences between Norwegian and Danish. Even though Norwegian preserved more of the inflectional morphology (for instance, some dialects have retained the dative case to the present day), this left no impact on the written language. The variation appears to be restricted to the spelling of individual lexemes and/or the aforementioned phonological features, of which the use of diphthongs is an especially prominent Norwegian feature; /ei/ is by far the most frequent one, for instance in the indefinite articles ein and eit. Lenition or lack thereof is a bit harder to judge from the sources, as there was also a chronological development in Danish orthography, where e.g. ‹th› must be considered the spelling of /d/ in many words, before ‹d› became the standard spelling (‹þ› > ‹th› > ‹d› in many pronouns and demonstratives, where the change in spelling occured much later than the phonological change of initial /θ/ > /d/).

3 Findings and patterns The sources I use consist mainly of charters, as no literary works in the traditional meaning were written in Norway during this time. There are very few charters from the archdiocese during the first period after Danish came into use in 1510.

284   

   Ivar Berg

The tendency, however, seems to be that writings for which the archbishop was responsible used an orthography within the normal scope of variation in Danish. In contrast to this, a couple of charters issued by the cathedral chapter show far more Norwegian interference. In the Latin sermon book Breviarium Nidrosiense, printed in Paris⁵ for the diocese in 1519 and edited by members of the chapter, there are two pages in Norwegian, despite the language used in official letters from the see at that time. This clearly demonstrates how the language shift was initiated from above by the archbishop, and it is probable that he brought with him new scribes from Denmark; else it would be hard to explain the smooth shift. This was probably the case for a scribe who followed archbishop Erik at least during 1515–1522, as several of the letters issued by Erik are written in the same hand, a hand which shows hardly any Norwegian features. The anonymity of most works from this period is another complicating factor. Only careful palaeographic studies can identify scribes, making it difficult to judge whether we are dealing with individual or group variation. No matter how consistent a scribe was in his orthography, if other scribes had different practices the orthography of a larger corpus will vary. Thus, even variation between different genres, which, according to my previous line of thought should be treated as the result of language attitudes (good/bad language in different settings), may also be explained as different scribes being assigned to different tasks. Nevertheless, I will argue that this would be just as clear an indication of language attitudes as an individual scribe varying his orthography, as the assignment of scribes must then have been governed by their perceived linguistic competence. The sources flow more richly in the 1530s, largely due to the random preservation of documents. From this period more informal writings (drafts, transcripts, administrative notes) are preserved, precisely the kind of texts most frequently showing Norwegian features, and which are not known from the years immediately following 1510. It is necessary to restrict the comparison to texts with similar functions. In some respects the writings from this period, after twenty years of Danish, show more Norwegian features than the earlier ones. This may indicate that the scribes had found a middle road between their Danish models and Norwegian tradition and spoken language. The Norwegian interference also seems to correlate – to a certain extent – with the degree of formality of the document. Normally, official letters to prominent Danes are written in “well-formed” Danish, although not without Norwegian forms; the occasional diphthong betrays their origin. However, in more informal notes, one might find different spellings marked by heavy Norwegian interference. One example is a note on the reasons

5 There was no printing press in Norway until 1643.

Variation in a Norwegian sixteenth-century scribal community   

   285

for arresting the nobleman Nils Lykke in 1535.⁶ The reason for its existence is obscure; it may have been a draft for some otherwise unknown letter, but in any case was not itself meant to be sent anywhere. It has several of the mentioned features (diphthongs like weit ‘know.pres.’, eyn/eit ‘one/a’, unstressed a in wttan, innan ‘without, within’, and typical Norwegian forms like leet ‘let.pret.’, honom ‘him’ where Danish would have different vowels (lot, hannem).

3.1 Free orthographic variation Medieval orthography allowed some variation in the orthography with no relation to the correspondence between a sound and a sign/letter, either in the individual orthography of a single scribe or as equal variants in larger corpora. As mentioned above, this range of variation can be identified either by variation within a text, or by comparing several versions of the same text. I shall first look at a letter of instruction from the archbishop and a few others to their envoys at a council in Copenhagen in 1534. The text is known in the original (called A here) and a transcript (B) which the archbishop ostensibly kept for his archive.⁷ There were clearly two different scribes at work, as A shows a clear preference for word-initial ‹f-› where B writes ‹ff-›; both have mostly intervocalic ‹ffu› for /v/, but B mixes this with ‹ffw›; A often has ‹sz› for /s/ (e.g. the relative particle szom), which is hardly found in B. Apparently each scribe chose whether to write single or double f word-initially, so this variation is not free on an individual level, however, the difference was not marked and ‹f-/ff-› must be regarded as equal variants on the group level. A second example may be the five almost identical letters sent to notables in Denmark regarding the election of King Christian III in 1535.⁸ 578 and 582 seem to be written by the same hand, and display some of the free variation. 578 twice has ‹y› where 582 has ‹i/ii› (almyndelige ‘common, usual’, sydenn ‘later’ vs. almindeligt, siidenn). Both texts use ‹u/w› interchangeably; 578 has kunde ‘could’, skwlle ‘should’, kwnde, 582 the opposite distribution.⁹ Also, there seems to be variation between word-final ‹-d/-dt›, as in budt, bekend and the other way round.

6 Printed as Diplomatarium Norvegicum XI no. 621. 7 The original is printed as Diplomatarium Norvegicum XVI no. 566, the transcript as XII no. 548. 8 Printed as Diplomatarium Norvegicum XVI no. 578–582. 9 Including the two instances of kunde, which occur at different points in the text.

286   

   Ivar Berg

This kind of orthographic variation was probably unmarked and simply free allographs of the same grapheme; the letters i and j are often used interchangeably, similarly w, v, and u. Other variables seem to be more or less consistent for the writer, but vary on the group level, e.g. ffw/ffu intervocally, as in the frequent verb hava ‘have’. This is a consequence of the sources, as briefly mentioned above: Even if each scribe stuck to one variant in his writings, different individual practices would show up as variation in a larger corpus. I think these two cases, individual and group variation, as far as they are possible to identify, should be treated as different principles. One reason for this is that the different individual norms may be signs of a norm change in progress, which will only become evident over a longer time span. Individual variation, on the other hand, should point to a flexible norm.

3.2 Cadastres One text type where Norwegian traits surface in abundance is cadastres, registers of property and revenue. The cadastre book known as Olav Engelbrektssons jordebog (henceforth: OEJ) (published by Brinchmann and Agerholt 1926), made for the archdiocese in the early 1530s, is a late example of clearly Norwegian place names, as even names were adapted (more or less) to Danish spelling during the 16th century (Indrebø 1927). There are many names ending in -wick(en) ‘wick, bay’ with ‹k› for Danish ‹g›, forms with diphthongs like Geiitesøin (p. 11) and many more. There are danified names and forms as well; most prominent is the almost consistent use of ‹e› for stressless vowels. One might find the preservation of Norwegian in such books obvious, as they consist of lists of local place names, which were not easily transferred to Danish. However, as this was sometimes done even in OEJ and danification of place names became common later in the century, part of the explanation must also be the purely administrative function of the cadastre in a time when Danish was not as established as it later came to be. Another notable feature of OEJ is that where there are differences between Old and Modern Norwegian, this book mostly has a modern form. I shall give an example witnessing the change from Old Norse initial /hv/ > Modern Norwegian /kv/, which has taken place in all Norwegian dialects except in the south-east. There are many names in OEJ with this consonant cluster, as Quam, written Hwam in similar cadastres from the late 15th century. It has often been claimed that such lists were mostly copied from older, similar works. However, the Modern Norwegian name forms cannot have any other origin than speech. If we compare OEJ with the earlier known cadastres for the archdiocese, this becomes quite clear:

Variation in a Norwegian sixteenth-century scribal community   

   287

As shown with Quam–Hwam example, the books from the 15th century all show traditional spelling which can easily be explained as following the Norwegian tradition or copied from earlier writings, whereas OEJ clearly breaks the tradition; this shows that the older Norwegian writing tradition did not influence the spelling of place names as much as has previously been claimed, at least not in OEJ. One of the scribes (as usual anonymous, known in the edition as “hand E”) in the cadastre book shows an interesting pattern: He wrote part of the land register itself, and also a short list of land that was purchased by the see under archbishop Erik, 1510–1522. This was some sort of preliminary work for the cadastre, and may have been written considerably earlier; the language is clearly more Danish than in the cadastre itself. This scribe was evidently capable of varying his orthography at will, which shows a very high scribal competence. The difference can be explained in two ways: a) Either the list of purchases was written earlier (perhaps as early as under the Danish archbishop), at a time when it was more important to conform to “normal” Danish; or b) the Norwegian forms were only allowed in the specific genre “land list” where they were mostly copied from older registers, whereas the list of purchases was based on new letters of transactions with more “modern” forms, in the sense that they were adapted to Danish spelling. As the name forms in the register appear to be independent of the older Norwegian tradition and not copied from older similar works, I believe the former explanation is the correct one, and that the relationship to the new written code was more important than older registers. This is contrary to the traditional claim by Indrebø (1927, 147) that the list of purchases was written from scratch based on fresh information, whereas the cadastre itself was mainly based on older models (op. cit. 150). Furthermore, it may be possible to infer from this that some time after the introduction of Danish, the norm loosened up and allowed certain Norwegian features; it was not mandatory to write Danish exactly like the Danes did. I shall pursue this further in the next section.

3.3 A local norm Although I have mentioned some variation between different text types the general impression is that the scope for variation was fairly limited. And in the material from the 1530s it was evidently acceptable to introduce certain nonDanish forms in the written language, even though, for instance, the spelling of the very frequent indefinite article must have been easy to change, had they cared. This Danish with a Norwegian touch to it became a relatively stable norm of the scribes in Trondheim, and later texts show many of the same features. One example is a note on the priests’ observation of mass, issued by the chapter in

288   

   Ivar Berg

1554.¹⁰ The text shows the same traits we are becoming used to: diphthongs (only in the article ein), lack of lenition with ‹t› for Danish ‹d› (maathe ‘way’, ytthermere ‘furthermore’) and the phrase honnom totthe ’seemed to him’, which was impossible in Danish. The use of such a local norm came to an end as the force of standardisation grew stronger and education improved, and even Norwegians learnt to spell correctly and write proper Danish. The development, I believe, took the following route: 1. Archbishop Erik brought a new written code with him, probably together with Danish scribes. The cathedral chapter and probably the previously employed scribes took some time to accomodate to the new code. 2. During the following years, a loose norm evolved within the chancery in Trondheim, a norm which accepted certain Norwegian features and words, although it was basically Danish. 3. With improved education and the growing force of standardisation throughout early modern times, this gave way to a uniform written code with minimal geographical differences, based on the language in Copenhagen. Although Danish was not formally codified for many centuries the variation became gradually more restricted, and norwegianisms are rather scarce after 1600, as shown by Iversen (1921).

4 Conclusions I have here mostly disregarded the purely allographic variation and focused on the variation which is due to differences between Norwegian and Danish. Although the Norwegians learned to write Danish quite well, the use of diphthongs in particular mark their language as quite different from that written by Danes. This is relatively scarce in official letters – although always present, except in some of the first writings after the language shift – but very prominent in less formal administrative writings. It also seems that the initial difference between writings of the archbishop in “normal” Danish and the more Norwegian language of the cathedral chapter gave way to a local, loose norm. It incorporated some Norwegian elements, but these were not highly regarded, as they are most frequent in informal writings. A couple of conclusions may be drawn from this:

10 AM 332 fol. 7v; printed as Diplomatarium Norvegicum XII no. 654.

Variation in a Norwegian sixteenth-century scribal community   

–

–

   289

The 16th-century scribes had a metalinguistic awareness of the differences between Norwegian and Danish, and Danish was the prestige variety used in the most formal contexts; Norwegian was more “suitable” in administrative matters. The example from the cadastre book shows that at least one scribe was able to adjust his spelling at will. This proves a very high scribal competence and a conscious willingness to adhere to orthographic norms.

The examples provided here also show how important it is to take the sociolinguistic circumstances of text production into account when we are dealing with written sources in language history, and I hope the outlined methodology may provide a way to access the language attitudes of past societies.

5 References Brinchmann, Christopher & Johan Agerholt (eds.). 1926. Olav Engelbrektssøns jordebog. Register på St. Olavs jorder, forfattet under erkebiskop Olav Engelbrektsson. Anhang: Erkebiskop Gautes jordebøger. Norges Rigsarkiv. Oslo: Nationaltrykkeriet. Brøndum-Nielsen, Johannes. 1950–1973. Gammeldansk Grammatik i sproghistorisk Fremstilling I–VIII. Copenhagen: Schultz/Akademisk. Finegan, Edward & Douglas Biber. 2001. Register variation and social dialect variation: The register axiom. In P. Eckert & J. R. Rickford (eds.), Style and Sociolinguistic Variation, 235–267. Cambridge: Cambridge University Press. Goossens, Jan. 1994. Normierung in spätmittelalterlichen Schreibsprachen. Niederdeutsches Wort 34. 77–99. Hansen, Aage. 1962. Den lydlige udvikling i dansk. Fra ca. 1300 til nutiden. 1: Vokalismen. Copenhagen: G.E.C. Gad. Hansen, Aage. 1971. Den lydlige udvikling i dansk. Fra ca. 1300 til nutiden. 2: Konsonantismen. Copenhagen: G.E.C. Gad. Haugen, Einar. 1976. The Scandinavian Languages: An introduction to their history. London: Faber and Faber. Indrebø, Gustav. 1927. Maalet i norske jordebøker i det 16. hundradaaret. In Festskrift til Hjalmar Falk 30. desember 1927, 142–155. Oslo: Aschehoug. Indrebø, Gustav. 2001. Norsk målsoga (2. ed.). Bergen: Norsk Bokreidingslag. [1. ed. 1951]. Iversen, Ragnvald. 1921. Bokmål og talemål i Norge 1560–1630. 1: Utsyn over lydverket. Oslo: Dybwad. Labov, William. 1994. Principles of Linguistic Change I: Internal Factors. Oxford: Blackwell. Mæhlum, Brit. 2005. Sociolinguistic structures chronologically III: Norwegian. In Oskar Bandle et al. (eds.), The Nordic languages: an international handbook of the history of the North Germanic languages, HSK 22.2, 1907–1922. Berlin: de Gruyter. Milroy, James. 1992. Linguistic Variation and Change. Oxford: Blackwell.

290   

   Ivar Berg

Nevalainen, Terttu & I. T.-B. van Ostade. 2006. Standardisation. In Richard Hogg & David Denison (eds.), A History of the English Language, 271–311. Cambridge: Cambridge University Press. Roberge, Paul. 2006. Language history and historical sociolinguistics. In Ulrich Ammon, Norbert Dittmar & Klaus Mattheier (eds.), Sociolinguistics/Soziolinguistik III (2. ed.)., HSK 3, 2307–2315. Berlin: de Gruyter. Romaine, Suzanne. 1982. Socio-historical linguistics: its status and methodology. Cambridge: Cambridge University Press. Sebba, Mark. 2007. Spelling and society. Cambridge: Cambridge University Press. Seip, Jens Arup (ed.). 1936. Olav Engelbriktssons rekneskapsbøker 1532–1538. Noregs Riksarkiv. Oslo: Dybwad. Vandenbussche, Wim & Ana Deumert (eds.). 2003. Germanic Standardizations. Past to Present. Amsterdam/Philadelphia: John Benjamins.

Dmitry Bondarev

16 Language change induced by written codes: a case of Old Kanembu and Kanuri dialects Sources: Kanuri is a major language of West Africa. Together with its related variety Kanembu, it is spoken by over four million people in Nigeria, Niger, Chad, and Cameroon. Extensive glosses of its historical stage, Old Kanembu, can be found in the Borno manuscripts (SOAS Library, MS. 380808), written in the old and modern-day Borno by teachers and advance students of Qur’anic exegesis. The Borno corpus originates from four manuscripts photographed by David Bivar in 1950s (a total of 230 pages in photographic and microfilm form). In 2005–2007, the corpus of digitised manuscripts was substantially increased to more than 4,600 digital pages, including five more copies of the Qur’an and fifteen other bilingual (Arabic and Old Kanembu) manuscripts. The collection now spans a period of several hundred years, from the late 17th to late 20th centuries, and derives from a number of different places in Nigeria. In addition, there are 20 hours of audio material in Tarjumo – a recitational variety of Old Kanembu – part of which has been transcribed. Abstract: With the development of Qur’anic studies in the ancient Islamic Kanem kingdom situated on the north-east of Lake Chad (the 13th to 14th centuries), Kanembu (Kanem-bu: ‘the Kanem people’) became the accepted language of Qur’anic interpretation, evolving into a codified technical language known as Old Kanembu (OK). It has survived in written attestations in the commentaries to the Qur’an (manuscripts of the 17th to early 19th century) and, as a modernised variety, in the network of Islamic scholars (northeast Nigeria) who speak Kanuri – a descendant of OK. As a result of an accumulation of archaic features and invented translational techniques, the meaning of the exegetical text in OK is only accessible to the learned circles of Islamic scholars (ʿulamāʾ) and students. This has created an unusual quasi-diglossic situation: exclusively for translation from Arabic, OK is never used for independent composition and is always linked to, and recited from, a particular Arabic source. OK has no communicative function even for the most educated (ʿulamāʾ), nor is it used in the mosque in ʿiḍa (‘sermon’). The paper starts off with a question as to whether OK can be considered one of the codes or “repertoires” (Fishman 1972:  47, Croft 2000:  92) or “channels/instances” (Romaine 1982:  14) of the same language and demonstrates how various palaeographic, philological, and linguistic techniques allow us to

292   

   Dmitry Bondarev

establish linguistically valid layers of the OK corpora. The paper then shows that however artificial and detached from the spoken form the linguistic medium may be, as OK is, its users act in the same way as users of High or literary written varieties of the language in typical diglossia. That is, the scribes and reciters of OK interact with other Kanuri-speaking communities and play a role in shaping the development of the language at large. Finally, drawing on selected isoglosses and evidence from modern-day conversations between the traditional scholars and laymen, it will be shown that some dialectal variation in Kanuri can be explained by ongoing communicative interaction between Kanuri-speaking communities and Old Kanembu scholarly networks.

1 Introduction¹ The speaker of a single language is in a way multilingual because of the multichannelled nature of the language and because “every language is a multiplicity of codes” (Gumperz 1968/1972; Romaine 1982: 14; Croft 2000: 55, 91–92). Coexistence of codes, or registers, in the language resembles a contact situation where language X interacts with language Y with the effect of mutual or one-way influence of one of them on the grammar and lexicon of the other. The codes of the language are not acquired by the speaker’s “critical age” (by around adolescence) but rather accrued through the speaker’s life (Kerswill 1996, Aitchinson 2000, Croft 2000: 57). Not all codes available in a language are accessible or required by the speakers, and hence a language community speaks several codes some of which are largely understandable by the whole community but others are not. The degree of understanding of the code and the number of its users vary. Situations of extreme gaps between the most common codes understandable by the majority and the restricted codes used by fewer speakers are familiarly described in terms of diglossia or secret languages. It has been noted for a long time that in diglossia situations the two varieties of the language – High

1 Acknowledgments: I am grateful to Doris Löhr for helpful discussions on Kanembu dialects, to Jonathan Owens for his valuable comments on Arabic data, and to Abba Tijani and Imam Habib for their insights into Kanuri and Tarjumo. The research on Old Kanembu and Kanuri and Kanembu dialects was carried out as part of two projects supported by the Arts and Humanities Research Council (AHRC), UK, “Early Nigerian Qur’anic manuscripts” (2005–2008) and by Deutsche Forschungsgemeinschaft (German Research Foundation, DFG)/AHRC “A study of Old Kanembu in early West African Qur’anic manuscripts and Islamic recitation (Tarjumo) in the light of Kanuri-Kanembu dialects spoken around Lake Chad” (2009–2011).

Language change in Old Kanembu and Kanuri   

   293

and Low – influence each other (Ferguson 1959/1972, cf. also the Sanskrit terms tatsama and tadbhava for loanwords from H variety into L used in Indo-Aryan studies [Grierson 1920]) and that this kind of “internal” inter-influence, even in linguistically homogenous societies, is in many ways similar to language contact/ multilingual situations whereby two or more different (and mutually unintelligible) languages contribute to either reciprocal or “one-way” change (compare Gumperz 1968/1972: 220, 225, Samuels 1972: 92, Kerswill 1996: 179). Literary and written varieties of the language constitute one of the multiple codes and as such are in contact with other variants of the same language through communicative activity of interacting networks of speakers. Despite the assumption that the literary/written form gives little to understanding how the language works and changes (see Schneider 2008), it has variously been maintained in the literature that written and spoken media are nothing less than different channels of the same language (Ferguson 1959/1972; Gumperz 1968/1972; Fishman 1972; Samuels 1972; Romaine 1982; Chafe 1982, 1994; Croft 2000). Since codes affect each other (and so induce variation and change of the language as a whole), understanding the role of the literary code in a larger linguistic landscape requires measuring the code’s sociolinguistic “weight”, i.e. its prominence in the everyday life of the speech community. In diglossic societies with rich written and/or literary traditions, such codes are easily identifiable through a corpus of written texts and high register communication (ritual, praise songs, mass media, scholarly discussions, lectures, etc.). However, some languages have developed a very specific kind of literary code which critically differs from familiar diglossic situations because of the absence of an independent body of texts and/or communicative functions. Old Kanembu as surviving in its written and oral form is one such language or rather a code within the Kanuri language.² This article addresses how the “Old Kanembu code” interacts with Kanuri dialects. Section 2 starts with the question as to whether Old Kanembu can be considered one of the codes or “repertoires” (Gumperz 1968/1972: 230; Fishman 1972: 47; Croft 2000: 92) or “channels/instances” (Romaine 1982: 14) of the Kanuri language. Section 3 shows that however artificial and detached from the spoken form the linguistic medium may be, as Old Kanembu is, its users act in the same way as users of High or literary written varieties of the language in typical diglos-

2 The Saharan (of the Nilo-Saharan linguistic macro-phylum) languages Kanuri and Kanembu form a continuum of dialects spoken around Lake Chad by a population of an estimated four million people. Kanuri dialects, mainly spoken west of the lake, by far outweigh Kanembu dialects, spoken north-east/east of the lake, by approximate ratio of 90 % to 10 %. Other Saharan languages are Teda/Daza (collectively known as Tubu) and Beria (also known as Zagawa).

294   

   Dmitry Bondarev

sia. That is, the scribes and reciters of OK interact with other Kanuri-speaking communities and play a role in shaping the development of the language at large. Section 4, drawing on selected isoglosses, demonstrates that some dialectal variation in Kanuri can be explained by the past interaction between Kanuri-speaking communities and Old Kanembu scholarly networks. In Section 5, I give evidence of the ongoing interaction between these two media, as it happens in modern-day conversations between the traditional scholars and laymen. A summary is given in the Conclusions.

2 Old Kanembu: from language to metalanguage With the development of Qur’anic studies in the ancient Islamic Kanem kingdom situated on the north-east of Lake Chad (in the 13th to 14th centuries), Kanembu (Kanem-bu: ‘the Kanem people’) became the accepted language of Qur’anic interpretation, evolving into a codified technical language known as Old Kanembu (OK). It has survived in written attestations in the commentaries to the Qur’an (manuscripts of the 17th to early 19th century) and, as a modernised variety known as Tarjumo (T), in the network of Islamic scholars (northeast Nigeria) who speak Kanuri – a descendant of OK.³ As a result of an accumulation of archaic features and, possibly, invented translational techniques, the meaning of the exegetical text in OK/T is only accessible to the learned circles of Islamic scholars (ᶜulamāᵓ) and their students. In this chapter, I refer to OK/T as a written/literary medium as opposed to spoken dialects. However, OK/T is a complex combination of written/oral features (Bondarev & Tijani in press). For the purposes of the present study we can neglect this multilayered characteristic of OK and consider it as a single literary entity. Nonetheless, it is solely its oral component (i.e. recitation) that makes OK recognisable by an other-than-scholarly audience as a literary medium (see next section). At the same time, the modern-day scholars and their students use OK/T as a written medium in advanced Qur’anic and Islamic studies, and it was likewise used in the past.

3 Throughout the chapter, I use OK/T as a cover term for both manuscript attestations and its modern-day recitation variety, Tarjumo, when both forms are discussed as a single entity. For specific reference to the language of the manuscripts and the language of recitation OK and T are used respectively.

Language change in Old Kanembu and Kanuri   

   295

2.1 OK/T and dialect variation OK/T has inherited archaic features from different periods, dialects and geographic regions. This is evident from the OK/T heterogenic phonology, morphosyntax and lexicon reflecting both western (Kanuri) and eastern (Kanembu) dialects. Verbal morphology is one of the most salient areas of mixed grammatical structures in OK/T. Modern Kanuri and Kanembu dialects have a diversity of dialect-specific verbal paradigms. For example, the verb da ‘to stop’ in past tense 3rdplural form has different exponents in three Kanuri dialects (Yerwa, Mowar and Manga) and one Kanembu (Tumari) as shown in (1). In the OK manuscripts, there are two different forms for the same past tense category: one, which only rarely occurs () corresponds with the Yerwa and Manga, the other, the most frequent one, () corresponds with Mowar and Tumari. The modern-day T has only retained morphological form I, i.e. dajada. However, the T form is not entirely identical to form I in OK, the first vowel of the portmanteau suffix -ada (3p.past) having been adjusted to non-fronted [a] Yerwa pronunciation. (1) Yerwa: Manga: Mowar: Tumari: OK(I): OK(II): T:

da-y-áda⁴ da-y-éra da-y-éino da-g-áino

da-j-áda

stop-past-3p.past

‘they stopped’

There are hundreds of similar examples with multiple irregular morphological correspondences between the OK/T and the modern-day Kanuri and Kanembu dialects. The same eclecticism is found in OK/T syntax. As an example, consider NP coordination. Modern Kanuri dialects have a coordinator clitic -a attached to the final element of coordinated NPs (2a). Some Modern Kanembu dialects employ an -n clitic for coordination (2b). The same clitic -n is used in OK/T coordinated NPs (2c). On the other hand, there is an alternative coordinating structure

4 Angled brackets are used for grapheme representations. In OK data, a macron (¯) above the vowel represents the so-called “weak” letters used in Classical Arabic for representation of long vowels, but used for high and falling tone in OK. High tone in Kanuri is indicated with an acute accent (e.g. á), low tone is unmarked, falling with a circumflex (â), and the rising with a hacek (ă).

296   

   Dmitry Bondarev

in OK/T where the NPs are juxtaposed without a coordinator (represented as zero in 2d): (2) a. Kanuri Yerwa: Áli-a Músa-a Ali-and Musa-and b. Kanembu Kuwuri: Áli-n Músa-n Ali-and Musa-and

‘Ali and Musa’

‘Ali and Musa’

c. OK (1YM/Q2:140):

prophet Ibrahim-and prophet Ismail-and ‘the prophets Ibrahim and Ismail’ d. OK (1YM/Q2:62):

Ø

At the lexical level, we also find a mix of dialectal features: (3) ‘they’ ‘hand’ ‘to fear’ ‘to call’

Kanembu

Kanuri

OK/T

tandí ndúko dow ku

sandí músko ri bowo

/tandí /músko /dap /ku & bogo

Table 1: Mixed lexical correspondences.

2.2 Archaic features in OK/T Many lexical and grammatical features of OK/T are unique in that they have no parallel in any modern dialects. These features represent much earlier stages in the history of Kanuri and Kanembu. Some of them are reflexes of proto-Saharan, as being cognate with the Saharan languages Teda/Daza, such as the existential verb ‘to be’ (4a) and a system of locative postpositional clitics still present in Teda and Daza but absent in Kn and Kb (Bondarev 2010: 245). Others show pre-Kanuri/

Language change in Old Kanembu and Kanuri   

   297

Kanembu morphosyntactic patterns no longer productive in the language, e.g. singular forms (4b): (4) a. OK/T archaic verbs *do(g) ‘know’ (only retained in earlier manuscripts) * ‘to be(come)’ ~ Daza cek (Chonai 1999: 95), cǝk (Lukas 1953: 113); Tubu tugo (Le Coeur 1956: 312) b. OK/T: the alternation k- / y- (k- = a non-productive singulative prefix) humans: ‘a person’ → ‘people’ (cf. MK: kâm > âm) non-humans (only in the manuscripts): ‘a stone’ → ‘stones’ (cf. MK: kâu ‘a stone’> kau-wá ‘stones’) ‘a day’ → ‘days’

2.3 Arabic influence on OK/T Apart from retentions of earlier linguistic stages, OK demonstrates a number of features resulting from adjustment of its structures to Arabic. Vocabulary is the most obvious area of Arabic influence. Lexical items with specific socially important religious connotations and those frequently used in the Qur’an, once borrowed, become an integrated part of the host language. This explains their occurrence across all OK manuscripts, modern-day T recitations and modern Kb and Kn dialects: (5) OK T/Kn / kitáwu < Ar. kitāb ‘book’ / náwi < Ar. nabīy ‘prophet’ / ráma < Ar. raḥma ‘compassion’ / lárdǝ ‘land’ < al-arḍ ‘the-land’ Arabic loan words frequently coexist with genuine Kanembu words. For example, MS.2ShK has the Arabic loan for ‘going ahead’ in Q79:4, whereas MS.1YM uses a native verb for the same meaning in the same place in the Qur’an: (6) a. Q79:4 MS. 2ShK ‘going ahead’ VN ‘going ahead’ < Ar. sabq ‘going ahead’

298   

   Dmitry Bondarev

b. Q79:4 MS.1YM ‘going ahead’ ‘going ahead’ < bǝrfo ‘before, ahead’ Not all loans attested in the manuscripts found their way into the modern-day language. The term ‘fast’ as used in the manuscripts was formed from Arabic ṣaum ‘fast’ (ka+ṣaum) with the once productive nominal prefix ka-. The prefixed form has apparently been substituted with the one derived from the same Arabic word with the definite article al-: Kn ashâm < Ar. al-ṣaum [aṣ-ṣaum]) In contrast to lexicon, morphosyntactic influence of Arabic on OK is not so easily recognisable. There are only a few discernable features that may be accounted for by Arabic syntax, such as (a) the use of some prepositions possibly modelled on PP’s in Arabic, (b) surfacing of independent pronouns (contra to typically pro-drop Kn/Kb) and (c) use of the core case markers (subject and direct object) on the (pro)nouns (conditioned by pragmatics in Kn/Kb), the last two features being most likely “designed” to clearly illustrate grammatical relations in the Arabic clause (Bondarev 2010 and forthcoming/b; Bondarev and Tijani in press). Other than that, the OK/T syntax is independent from Arabic and even if there were calques or inventions to convey the Arabic syntax in OK, they must have been created at a very early stage of the language since they cannot be identified on the basis of the syntactic structures known in modern Kanuri and Kanembu dialects. One of the numerous examples of Arabic-independent syntax is the OK genitive construction which does not copy the Arabic word order (H NDET) but rather uses NDET-GEN-H word order as shown in (7): (7) 1YM/Q7:54, Arabic (H NDET): rabb al-ᶜālamīn lord def-creatures.gen ‘Lord of all creatures’ OK (NDET-GEN-H):

creatures-gen lord-sj ‘Lord of all creatures’

2.4 Tarjumo: reflexes of OK and dialect features Synchronically, Tarjumo – the direct descendant of OK – is characterised by syncretism resulting from OK structures and from the influence of the modern-day western Kanuri dialects. (8) illustrates the point:

Language change in Old Kanembu and Kanuri   

   299

(8) (Q2:38 Ar. wa-lā hum yaḥzunūn ‘they will not grieve’) OK: grieve-prog-vn-comp fut.3p-be-neg ‘they will not be grieving’ Tarjumo (mix of OK, Kb and Kn forms): var. 1. roro-tǝ́-wa-ro sandí-ye ta-tadé-nyí grieve-vn-assoc-comp 3p-sj fut.3p-become-neg ‘they will not be with grieve’ var. 2. roro-tú-wu-ro tandí-ye ta-tadí-bú grieve-vn-ag-comp 3p-sj fut.3p-become-neg ‘they will not become grievers’ In (8 var.1) the Tarjumo phrase is partly based on (western) Kanuri dialects and partly on (eastern) Kanembu dialects (underlined). Thus, the associative marker -wa is non-existent in the OK manuscripts (Bondarev 2010) but it exists in all contemporary dialects as does the negative verbal morpheme -nyí. On the other hand, the pronoun sandí ‘they’ has a form typical of western Kanuri dialects. The variant Tarjumo phrase in (8 var.2) is slightly more aligned with OK in that it uses the predicate negator -bú unattested in modern dialects. The agentive noun morpheme wu is typical western Kanuri whereas the pronoun tandí is only found in the eastern Kanembu dialects. This particular example also shows that some OK grammatical and lexical structures have been reanalysed in Tarjumo in accordance with modern Kanuri grammar. Since all Kanuri dialects lost the full inflected existential verb * ‘to be(come)’ (4a), with its subsequent substitution with a defective existential verb mbéjí/mbézái ‘there is/are’, some of the morphological exponents of the verb * have been reinterpreted, including the one in (8 OK), i.e. ta-dīgi-bū (FUT.3p-be-NEG) ‘they will not be(come)’. This word-form was reinterpreted in Tarjumo as a form of a non-existent Old Kanembu verb *tadi with the same meaning ‘to be’ with the resultant form ta-tadí-bú (FUT.3p-be-NEG) ‘they will not become’ (8 var.2) (see Bondarev in press). The only form of the verb * which did survive in Tarjumo with the original form–meaning relation is the 3rd-singular imperfective cigí ‘there is’ from OK ‘there is’.

2.5 Old Kanembu/Tarjumo: a metalanguage and a code As a result of the accumulation of archaic, dialect and Arabic features OK/T is unintelligible to Kanuri and Kanembu speakers who have not received special training. Since OK/T can only be acquired through formal education, it resembles

300   

   Dmitry Bondarev

High codes in classic diglossia. On a par with H varieties in diglossia, OK/T exclusively serves in a special domain for which other (L) varieties cannot be used, that is the domain of the Qur’anic interpretations and translation of other Arabic texts. However, unlike typical diglossic H codes, OK/T is effectively inseparable from the source language because the function of OK/T is to translate or rather provide grammatical understanding of Arabic. Hence it functions as a linguistic metalanguage used for grammatical analysis of Arabic (Bondarev and Tijani in press). This has brought about an unusual quasi-diglossic situation: used exclusively for translation from Arabic, OK/T is never used for independent composition and is always linked to, and recited from, a particular Arabic source. There is no conceivable way for traditional Kanuri scholars to use OK/T separately from the Arabic source in teaching, (silent and aloud) reading and reciting. Whenever OK/T is used, it is an Arabic source which goes first and then every Arabic verb phrase, noun phrase or postpositional phrase is invariably followed by the corresponding structures in OK/T. With such specialised use, OK/T has no communicative function even for the most educated ᶜulamāᵓ, let alone for anyone outside their scholarly network. The sermon (ᶜiḍa) in the mosque, for example, may be delivered in two local languages – Hausa and Kanuri – but never in OK/T. Absence of two crucial features – speech function and written literature – makes OK/T a very special code of the language, unparalleled in diglossia. Ferguson, for example, (1959/1972: 236) lists four functions of classical diglossia related to communicative function (sermon in church or mosque, speech in parliament, political speech, university lecture, and news broadcast) but none of the like is even remotely present in OK/T. Similarly, the diglossia criterion of a body of written literature (Ferguson 1959/1975: 236) is not relevant for OK/T, which has no independent written texts, being only an accompaniment to the Arabic. Importantly, the interlinear glosses in Old Kanembu are normally not regarded by the Tarjumo users as a standard legitimate model for modern-day writing in Tarjumo. Rather, it is a complex concept of professional, religious and historical identity that links modern-day ᶜulamāᵓ to the early annotated Qur’an manuscripts (Bondarev and Tijani in press). It is the lack of these two criteria strongly associated with classical diglossia that makes it problematic for OK/T to be considered as a code in sociolinguistic terms. The ᶜulamāᵓ community that uses this language does not speak it and has no continuous independent texts to read in it. If we maintain the assumption (see literature in Introduction) that codes mutually interact through their speakers, then OK/T with its covert metalinguistic function has no speakers’ support that could “lift it up” to the level of “visibility” by other codes. However, there is a highly relevant criterion of prestige and power (Ferguson 1959/1972: 245; Ward-

Language change in Old Kanembu and Kanuri   

   301

haugh 2006: 89–90) which allows us to discuss OK/T in terms of diglossia and interaction of codes.

3 Prestige lifts up the OK/T metalanguage to a typical diglossic code The unintelligibility of Tarjumo to ordinary Kanuri Muslims and its role as the language of religious explanation makes Tarjumo a sacred code comparable with Arabic (the first sacred language is Arabic, the second is Tarjumo). For the ᶜulamāᵓ community (including students and their teachers proficient in T) the exegetic language has become an element of cultural identity in a complex identity formula “Kanuri = Muslim/Islam = Tarjumo”, where Tarjumo represents an extra dimension of Kanuri and Muslim identity. On the one hand, Tarjumo is associated with Qur’anic commentary and thus with the sacred knowledge possessed by Kanuri Muslim society. On the other hand, Tarjumo links learned Kanuri to the KanemBorno past famous for its early adherence to Islam (in the 12th or 13th century) and to the outstanding religious and cultural figures of Kanuri/Kanembu history (Bondarev in press). Thirdly, the notion of Tarjumo is linked to the educational and intellectual hierarchy within the ᶜulamāᵓ – the possessors of esoteric knowledge. Thus, proficiency in Tarjumo is associated with one’s status and prestige in traditional Kanuri-speaking Muslim society (Bondarev and Tijani in press). Paradoxically, even though T is entirely incomprehensible to the ordinary public, all the ᶜulamāᵓ practicing the language agree that the purpose of T is to explicate the meaning of the Qur’an in language understandable to Kanuri speakers (Bondarev and Tijani in press). This misconception generated by T users themselves has even reached scholarly publications. Thus, Bobboyi (1992: 58) says that “[Tarjumo] allows easy comprehension of the Qur’an for the ᶜilm student as well as for the interested Muslim public” (emphasis mine) and Mustapha (1987: 176) also writes that the duty of such a student is “to render the meaning of the exegete’s work into the vernacular so that people can understand the message of the Qur’an” (emphasis mine). Moreover, the high status and sacredness of T makes the ᶜulamāᵓ be seen as speakers of OK/T. In literature we find references to a literary language in which the scholars communicate, including my own earlier publication (Bondarev 2006: 138; see also Bulakarima 1996; Dobronravin 1999). The combination of insiders’ construal and outsiders’ reception of OK/T changes the whole picture of OK/T as an inaccessible metalanguage: on the one hand, the users of OK/T do identify OK as a linguistic code (“a language to convey the meaning of the Qur’an”). On the other hand, this “insider’s” uplifting of OK/T

302   

   Dmitry Bondarev

from metalanguage to a proper-language status makes non-ᶜulamāᵓ communities think that OK/T is a conversational linguistic medium. This purely extralinguistic interplay of higher and lower statuses reverses the very nature of OK/T characterised by “zero-conversational” and “zero-literature” features. As a result, OK/T is lifted up as a proper High status code typical of diglossia. Therefore, since OK/T does function as a code in the language it is possible to analyse its role in the development (maintenance and change) of the Kanuri language as a whole. This can be done through understanding whether the ᶜulamāᵓ may propagate variation in the language based on the structures of OK/T. The ᶜulamāᵓ form a high prestige group and their network is characterised by strong ties, i.e. high density of communicative interaction within their own network (Milroy 1992; Croft 2000: 169). The imams proficient in T share the same educational background and are regularly summoned by the Shehu (traditional ruler) of Borno. According to Labov’s “high prestige” theory of introduction of new linguistic variants into a larger community of speakers (Labov 1980), members of close-tie communities act as channels for variation and so we would expect that the Kanuri ᶜulamāᵓ’s role in propagating a variant should be that of Labov’s “high prestige” groups. Milroy (1992) argues against Labov’s view and shows that it is the speakers with weak social ties who are responsible for introduction of a new variant into the language. Nonetheless, the weak-tie theory does not undermine the propagation role of the Kanuri ᶜulamāᵓ. On the one hand, they do see themselves as a distinct and prestigious social group and they are duly recognised as such by the Kanuri society. But despite clear boundaries outlining their network, the traditional ᶜulamāᵓ (proficient in OK/T) are among the most socially engaged groups because – being imams and religious authorities – they are public figures who regularly interact with almost all layers of the Kanuri society – at the level of their families and neighbourhoods, engaging in help and advice. Their network is characterised by the highest “multiplexity” (Croft 2000: 169), i.e., they have the highest proportions of contacts outside their own network and local neighbourhood. This particular strong-tie and multiplexity status of their network allows for bringing both Labov and Milroy’s approaches together as valid corroborating hypotheses. Both Labov and Milroy discuss propagators of novel variations. Conversely, the users of OK/T rather act as prescriptive force preventing changes in the language (see next sections). What is important for this discussion however is the fact that the ᶜulamāᵓ community can act as channels of interaction between OK/T and other codes of the language.

Language change in Old Kanembu and Kanuri   

   303

4 Two phonological features in Kanuri and Kanembu: influence of the written medium Although the interaction between written and spoken media is usually reciprocal (see Samuels 1972: 6), the influence of the written form on the spoken is not easily identifiable.

4.1 Direction of influence As shown in Section 2, Old Kanembu has inherited archaic features from different periods, dialects and geographic regions. This is evident from heterogenic morphosyntax and lexicon reflecting both western (Kanuri) and eastern (Kanembu) dialects. Influence of the spoken language on the literary medium is visible in the manuscripts and audible in recitation. It is something of a commonplace that the written language may retain earlier linguistic structures, as Old Kanembu does in (4), see also for English (Samuels 1972: 109–110), Harari (Wagner 1983), Geᶜez (Tosco 2000) and Russian (Uspensky 2002). It is not unusual that these structures may undergo attrition under the pressure of the language spoken by the users of the written medium, see example (8) for Old Kanembu; compare the influence of Russian on Church Slavonic (Uspensky 2002: 80–85) ), Low variety of Tamil on High Tamil (Britto 1986: 274) and Amharic on Geᶜez (Bausi 1990; Sima 2010). If the direction of change in certain areas of phonology and grammar is predictable and confirmed by internal and external reconstructions (for example a change from intervocalic voiceless velar stops > voiced velar stop > approximants > zero, as is the case in Kanuri), and if these stages are represented across the manuscripts of different periods – provided a philological analysis has established the correct chronological relationship between the artefacts and linguistic material they represent – then the changes in the written language are readily available to demonstrate: we may say that manuscript A represents a variant that precedes the one represented by manuscript B. This kind of written-medium variation only indirectly contributes to our understanding of how the language changes as a result of interaction of its codes because the spoken counterpart(s) of the past written code(s) are not usually available for investigation and can only be reconstructed with varying degrees of approximation.⁵

5 For some well-preserved manuscript cultures, there might be a rare chance of establishing with a certain precision how the spoken medium conditioned variation in the written one, see

304   

   Dmitry Bondarev

Capturing the converse direction of influence – written > spoken – is no less problematic unless the study concerns a contemporaneous speech community with an uninterrupted manuscript tradition. In this case, the analysis of the written > spoken influence may prove fruitful for not only pinpointing the result of change but also for observing an ongoing change and hence for understanding why the language develops at the written/spoken interface. In the case of OK/T, even though dialect variation reflected in the written medium, as shown in (1–3) and (8), does demonstrate that the spoken medium has left traces on the written it is impossible to ascertain how and why this influence took place, for there are no historical attestations of most of the spoken dialects. Conversely, the opposite relation between the written records and spoken dialects may prove more instructive for understanding language change because we do have diachronic attestations of the written form. Since the written data can relevantly represent earlier stages of the language, it can be contrasted with the resultant stage(s) in the contemporaneous dialects. A separate problem related to the influence of the written form is the register at which the written/spoken interaction operates. The received wisdom says that “the literary form may influence only the formal register of the spoken language” (Samuels 1972: 6). However, the influence of the written form may transcend the formal register and encompass many others (as was the case for Russian, with its written officialese influencing the spoken colloquial during the early 20th century after the October Revolution). The potential mutual transparency of registers makes the task of capturing written > spoken interaction productive for understanding language change from the perspective of code interaction and underlines the importance of the written medium for overall development of the language. This written > spoken direction of influence is the focus of the remaining sections.

4.2 Geographic and dialectal mapping of OK/T scholarly networks The highest activity of the ᶜulamāᵓ who practice OK/T is observed in the southwest of Kanuri speaking population, particularly in Maiduguri, the capital of Borno state, where the majority of Kanuri speakers speak a Yerwa dialect. Further north-east, the importance of OK/T diminishes to the extent that even imams of

Bergs’ (2005) work on the “Paston letters”, spanning 80 years of continuous correspondence in 15th-century England.

Language change in Old Kanembu and Kanuri   

   305

the older generation hardly use it. The modern-day distribution of OK/T practice is a reverse picture of the past: the area to the north-east of Lake Chad (Kanem) was once the centre of scholarly activity before it gradually moved to the west (Garumele and Gazargamu) and finally to the south, where Maiduguri is situated (Figure 1). With the shift of the OK/T scholarly centre the networks of scholars have also been reshaped, resulting in the weakening of the OK users’ network in the north-east and a strengthening in the south-west.

Figure 1: Kanem–Borno and Kanembu–Kanuri dialects.

We will now look at two features of Yerwa Kanuri compared with its archaic literary form and see if OK/T may account for dialect variation in spoken Kanuri. These features concern the phonological changes that had already taken place in the past, and therefore can only hint at some kind of interactions between the written/literary and spoken varieties, without a clear picture of how exactly these interactions took place.

306   

   Dmitry Bondarev

4.3 zǝma in Yerwa vs jǝma in other dialects The modern-day Yerwa Kanuri spoken in and around the state capital Maiduguri has two important Islamic terms that seem phonologically odd. These are zǝ́ma ‘Friday’ and ‘Friday prayer’ borrowed from Arabic jumaᶜa and zánna ‘paradise’ from Arabic janna. The two words are unusual in that they have the initial /z-/ (phonetically realised as a free variation [z] ~ [d͡z]) unlike other j-initial Arabic loans which have /j/ (realised as [j] (= IPA [d͡ʑ]) only) both in Yerwa Kanuri and the source Arabic (cf. jaráwa ‘test’ < Ar. jarraba ‘to test’, jaáwu < jawāb ‘answer’). The question arises as to why these two terms borrowed from Arabic j-initial words are found among z-initial Yerwa words whereas all other Arabic j-initial words are borrowed with the initial /j/. The exceptional phonology of the two terms becomes even more salient when compared with their phonological representation in Kanuri and Kanembu dialects other than Yerwa. As shown on the isogloss Figure 2, both words have initial /j-/ in all other dialects. A possible answer to the question of this particular distribution of /z/ and /j/ in Yerwa lies in the way the words were read by the users of Old Kanembu in the past. I will now suggest the arguments for this assumption. Previous scholarship in Kanuri and Kanembu dialects have shown that palatal affricates were non-existent at earlier stages of the language (Cyffer 1997; Chonai 1999; Bondarev 2005) and that Kanuri and Kanembu dialects only recently (within the last 200 years) underwent palatalisation of affricates – together with sibilants and (dialect specific) alveolar and velar plosives (Lukas 1931, 1937; Cyffer 1997; Hutchison 1981; Löhr 1997). In the Gazar variety of Kanuri spoken in and around the old capital Gazargamu and recorded (in its spoken form at the turn of the 19th century) by Sigismund Koelle in the 1830’s (Koelle 1854a and b), the two words in question have initial non-palatal affricate /dz/. I have argued elsewhere (Bondarev 2005, forthcoming/a, in preparation) that Old Kanembu as attested in manuscripts had a pair of non-palatal affricates /ts/ and /dz/, graphically represented as ‫< ﺙ‬th> and ‫< ﺝ‬j> respectively. The former underwent palatalisation /ts/ > /c/ in some words and sibilation /ts/ > /s/ in others, and the /dz/ palatalised to /j/ in all dialects except for Yerwa where etymons in /dz-/ were retained, albeit with a free variation [dz] ~ [z]. Since Islam was established in Kanem at least 600 years prior to Koelle’s record of Kanuri, the most important Islamic terms starting with /j/ (janna and jumaᶜa) must have been borrowed into Old Kanembu around that time and their phonology should have been that of Old Kanembu in its pre-palatalised stage. Hence, these words must have been rendered with closest homorganic initial consonants available at the time, i.e. /dz/: OK *dzana < Ar. janna and OK *dzǝma < Ar.

Language change in Old Kanembu and Kanuri   

   307

jumaᶜa. The majority of j-initial Arabic words were borrowed much later when /j/ had already been in use: jarawa ‘test’ < Ar. jarraba. The above presumptions give us a relative timeline for mutation of /dz/ in ‘Friday’ and ‘paradise’, in Old Kanembu, Kanembu and Kanuri. Table 2 represents a possible scenario for this change:

timeline /j/ in Arabic jumaᶜa and janna

1200s–1500s

1600s–1800s →

→

*/dz/ OK

→ → →

/dz/ OK in Borno MSS /dz/ Kn of Borno /dz/ (~ [j]?) Kb

1900s–present →

/z/ T

→ → →

/z/ Kn Yerwa /j/ non-Yerwa Kn: /j/ Kb:

Table 2: Ar. /j/ > OK /dz/ > KY&T /dz~z/; other MK and MKb /j/.

The graphic encoding of the two terms in the manuscripts may seem problematic because these words are always written with the grapheme . However, the grapheme has both [j] and [dz ~ z] readings in modern day Tarjumo and its homography (one grapheme for more than one sound) was attested as early as in Douisburg 1913. The underspecification of the grapheme is not helpful for reconstruction of the /dz/ sound in ‘Friday’ and ‘paradise’ but it is understandable why other graphemic devices were not used (like or a combination of + ). The OK ‘paradise’ and ‘Friday’ are only written above the corresponding Arabic original word and hence visually modelled on the Arabic. This is exactly what happens in modern Tarjumo: the scholars see the Arabic , write above it (sometimes though) but read it [dzánna].⁶ The retention of the word-initial non-palatal /z/ ([z ~ dz]) in Yerwa is by no means uniquely restricted to these two terms. In Cyffer and Hutchison’s Kanuri dictionary (1990) there are 165 z-initial tokens, including loans from Arabic. These loans (28 items) differ from ‘Friday’ and ‘paradise’ in that their initial consonant originates not from Arabic [j] but from other Arabic consonants, such as Ar. [z] (‫< ﺯ‬z>), e.g. KY zâdǝ < Ar. zād ‘sustenance’, [ð] (‫< ﺫ‬dh>), e.g. KY zákar < Ar dhakr ‘penis’, and three “emphatic” pharyngealised consonants [ðˤ] (‫< ﻅ‬ẓ>), e.g. KY záhir < Ar ẓāhir ‘clear’, [dˤ] (‫< ﺽ‬ḍ>), e.g. zárap ‘bullet’ < ḍaraba ‘to strike’ and [sˤ] (‫< ﺹ‬ṣ>), e.g. KY zǝ́m < ṣāma ‘to fast’. If we exclude the 28 loans of this kind

6 Also see (Bondarev forthcoming/b) for Q.79: 7 in three different Borno Qur’an manuscripts where a genuine Old Kanembu word-form is written as ‘the one who follows’ and pronounced [zuzurtəma] in Tarjumo.

308   

   Dmitry Bondarev

and the two odd ones from Hausa, we are left with about 135 lexical “genuine” Kanuri items with initial /z/. This is almost twice the number of the genuine /j/ initial Yerwa Kanuri words, i.e. 135 to 69.⁷ Out of the 69 items, there are 34 where /j/ occurs (in predicted environments) before the front vowels /i/ and /e/, e.g. jíli ‘kind’ and jê ‘rope’.⁸ Since there is a phonotactic constraint on the /z/ + /i, e/ sequence, the 34 /j/ initial items are in complementary distribution with /z/ and as such should be considered as (former?) allophonic realisations of /z/ in this specific environment. Hence, we only have 35 items in /j-/ that can occur in phonological environments permissible for initial /z/, i. e. before non-front vowels (cf. jána ‘knife’ vs. zána ‘paradise’). This is almost one fourth of initial /z/ in Yerwa Kanuri: 135 vs 35. It is clear then that the items ‘Friday’ and ‘paradise’ are among a large set of /z/ initial words not affected by otherwise typical crossdialect palatalisation and the question about retention of /z/ in these two items can be reformulated as follows: what prevented total palatalisation of the initial /z/ in Yerwa Kanuri? Since /z/ > /j/ occurred in all other dialects including those neighbouring Yerwa and entirely intelligible to Yerwa speakers (such as Mowar, for example), there can hardly be any plausible purely phonetic factors accounting for the lexically restricted palatalisation of /z/ in Yerwa Kanuri, or, in diachronic terms, for slowing down the process of palatalisation.⁹ The features of the substratum Chadic language(s) and contact linguistic situations could have been possible factors that caused resistance to palatalisation. However, the data from contact Chadic languages (Schuh 2002, 2003; Löhr to appear) more readily provide support for palatalisation as an areal phenomenon, and so retention of non-palatal features in Yerwa Kanuri should be considered unique and opposite not only to other Kanuri and Kanembu dialects but also to neighbouring Chadic languages. Given the impossibility of explaining the retention of initial /z/ in Yerwa by intralinguistic phonological mechanisms or by contact-induced processes, the

7 In this /j/ initial count, loans from Arabic (35), Hausa (6), English (3), and French (1) are excluded. 8 /z/ can only marginally occur before front vowels in free variation, e.g., in a traditional title Zérma ~ Jérma ‘equerry’. 9 Why the 35 /j/ initial Yerwa words have palatal /j/ before front vowels at all is a different question and each word would probably require an individual explanation (e.g., the distinction jána ‘knife’ ~ zána ‘paradise’ could have been explained by dissimilation, but these words are complete homophones in the “j-dialect”). Possibly, as discussed below, lexical diffusion of [j] as a variant of (z) started off at some point in the pre-Yerwa dialect but did not go further under pressure of OK/T.

Language change in Old Kanembu and Kanuri   

   309

extralinguistic social factors seem to provide a plausible account for retention of /z/. It is likely that a prescriptive force superimposed by the Borno Islamic scholars upon the speakers was a key factor for “conservation” of /z/ initial words. If we project the timeline of */dz/ > /dz/, /z/ and /j/ on the diachronic and geographic activity of the Kanem-Borno ᶜulamāᵓ we will see that the feature [-palatal] (in /dz/ and /z/) is retained in the areas where OK and Tarjumo have been more intensively practiced. timeline

1200s–1500s

1600s–1800s →

*/dz/ > /dz/ /z/

Geography of OK practice

*/dz/ OK

Kanem

→

/dz/ OK in Borno MSS /dz/ Kn of Borno Borno: Gazargamu, Kukawa

1900s–present →

/z/ T

→

/z/ Kn Yerwa Borno: Maiduguri

Table 3: Distribution of */dz/ > /dz/, /z/ on the time scale of practising OK/T by the ᶜulamāᵓ.

Figure 2: Isogloss ‘Friday’.

310   

   Dmitry Bondarev

Remarkably, this distribution coincides with the bundles of isoglosses in Bulakarima (2001: 114–120) for palatal/non-palatal consonants /z, j, t͡s, t͡ʃ/. The combined isoglosses delineate the non-palatal area which starts from the region of the old Borno capital Gazargamu and runs south-east to Maiduguri (Figure 2). This non-palatal “passage” from Gazargamu via Kukawa to Maiduguri fits into a recent (over the last 200 years) migration of the Borno scholars, caused by religious wars and dynastic shift in the 19th century. As stated in Section 3, the Borno ᶜulamāᵓ have always been an influential social group within Kanuri society. In addition to being a distinct prestigious group, their networks are essentially multiplexible, i.e., the ᶜulamāᵓ intensively and extensively interact with other people outside their own network because of their public status as imams of the mosques and religious authorities within their neighbourhoods. The two words in question (zǝ́ma ‘Friday’/‘Friday prayer’ and zánna ‘paradise’) are among the most important concepts of Islam and it is their religious semantic force and prestigious status of OK and T that kept these words intact (i.e. [z~dz] initial) in OK and T (coexisting alongside /j/ initial Arabic originals). Since the [z~dz]-initial pronunciation of these words has been reinforced by the high-status OK/T and since both words are very frequent in religious and everyday discourse in Kanuri, the ᶜulamāᵓ, in the areas where their presence was most prominent as it was in Gazargamu and is in Maiduguri, must have acted as prescriptive maintainers of the “correct” pronunciation. Not only did this prescriptive force preserve /z/ initial pronunciation in the two important religious terms but it was apparently successful enough to withstand the pressure induced by cross-dialect palatalisation. This may well account for the high proportion of initial /z/ in YK compared to the genuine Kanuri initial /j/. In other words, OK hindered the process of lexical diffusion which could have brought about the unconditioned /z/ > /j/ change. This process however only resulted in a lexical split with a bare 25 % of unconditioned j-initial words. To conclude the Section I therefore advance the hypothesis that the non-palatal feature of voiced fricative /z/ in zǝ́ma ‘Friday’ and zánna ‘paradise’ is only preserved in Yerwa Kanuri because it is spoken in the centre of the traditional scholarly network. This retention likely affected most of the other /z/ initial words. In all other dialects, where such networks disappeared, there was no restriction on total palatalisation *dz > j and so *dzǝ́ma, *dzánna and most of the other */dz/ initial words have had the initial consonant palatalised.

Language change in Old Kanembu and Kanuri   

   311

4.4 h-initial words Another area of Kanuri phonology where the influence of scholarly circles has possibly left a trace in Yerwa Kanuri is /h/-initial words. The initial /h/ had two cycles in the history of Kanuri and Kanembu. The Old Kanembu manuscripts retain an old *h (possibly a glottal fricative), which was a marginal phoneme only occurring word-initially in a few words. This phoneme was subsequently lost in most of the dialects only to reappear later as modernday /h/ under influence of Arabic. The reflexes of *h with corresponding graphemic representation in the manuscripts are shown in Table 4 (based on Bondarev forthcoming/a):

Reconstructed for Old Kanembu

Manuscripts (grapheme ‫< ﺡ‬ḥ>)

Kanuri and Kanembu

‘I’

*hú

‘we’

*handí

‘five’

*húku

(Kn) wú (Kb) ú (Kn) andí (Kb) yandí (Kn) úwu (Kb) ûu (Kn) yen (Kn) hap/ap (Kb) ap

‘to reach out’ *hin ‘to lift up’ *hap



*h > w/ Ø *h > y/ Ø *h > Ø *h > y *h > h/ Ø

Table 4: Reflexes of *h with corresponding graphemic representation.

From the above correspondences we can see that native word-initial /h/ has weakened to an approximant or zero,¹⁰ and only in some words in one dialect (i.e. Yerwa, see below) was the /h/ retained. Disappearance of the phoneme *h at some stage in the history of Kanuri is also evidenced by a few /x/-initial Arabic loanwords which were borrowed without word-initial fricative: alak ‘to create’ (< Ar. xalaqa), alífa ‘caliph’ (Ar. xalīfa).¹¹ At some later stage, the influx of Arabic loanwords with initial and medial /ħ/ and /x/ became greater, which resulted in the introduction of a new phoneme

10 It is not clear if the glottal stop in the vowel-initial words has a phonemic value. 11 A somewhat more complex history of borrowing may lurk here: the source initial consonant in these two words is velar /x/ but not pharyngeal /ħ/ as in other h-initial Arabic words in Yerwa and so there is a possibility that the output of borrowing from the source phonemes /x/ and /ħ/ was different.

312   

   Dmitry Bondarev

/h/ (see also a similar phonemicisation of /ʔ/ and /h/ in Hausa (Newman 2000: 319, 395): word-initial: Kn/Kb hak(k)a ‘correct’ (< Ar. ħaqq), habâr ‘news’ (< Ar. xabar); word-medial: Kn/Kb wahar ‘to be late’ < Ar. ᵓaxar ‘to delay’. Some of the Arabic loans possibly show that they were introduced to the language twice, first when no /h/ was present, and then, when /h/ became phonemically available, the initial segment was either restored/reinforced or just borrowed anew. For example, there is a free variation lawâr ~ habâr ‘news’ (< Ar. xabar).¹² In most of the dialects old */h/-initial words did not restore their onset value. In Yerwa however, there are a few native verbs where the word-initial /h/ occurs before /a/ (some of the verbs can probably be old Arabic loans): hap ‘to lift up’, ha ‘to stretch out’, hak ‘cough up’, har ‘dry up’ (possibly from Ar. ħarr ‘hot’); ham ‘cool down’ (< Ar. [antonymic ?] ħamm ‘heat’). If har ‘dry up’ and ham ‘cool down’ are of Arabic origin at all (alternatively, they could have been reflexes of Saharan roots)¹³ they must have been borrowed at the time when there was word-initial */h/ in Old Kanembu. The retention of /h/ before /a/ in these Yerwa words may well not be diachronically continuous. There is historical evidence that in the Gazar dialect spoken 200 years ago /h/ in these words was in free variation with Ø, e.g. har ~ Øar ‘to dry’ (Koelle 1854b: 299). As mentioned earlier, Gazar was spoken in and around Birini Gazargamu which was the centre of the scholarly activity at the time and where most of the 17th and 18th century manuscripts were written. Unlike Gazar however, the manuscripts are consistent in representing /h/-initial, e.g., ‘lift up, raise’. The same consistency has been maintained in modern-day Old Kanembu manuscripts and recitation. The Gazar variation h ~ Ø then suggests coexistence of a stable “literary” prestige phoneme /h/ (encoded as in the manuscripts) and free variation [h] ~ [Ø] both pointing to the earlier phoneme *h. It is impossible to say whether variation [h] ~ [Ø] was a feature of all dia-

12 Another interpretation would be borrowing from Ar. pl. al-axbār > *lahbar which would result in lenition of /h/ in coda position, cf. MK allo < Ar. lawħ ‘writing tablet’. 13 See possible cognates of Yerwa Kanuri har ‘to dry’ and ham ‘to cool’ in the Saharan languages Teda and Daza: cor ‘to dry’ (Lukas 1953: 202; Le Coeur 1956: 376) and wop/wou/wab ‘to cool’ (Lukas 1953: 183; Le Coeur 1956: 318–319). There is a typical o ~ a correlation in Saran languages which allows for internal Saharan placement of Yerwa Kanuri har ‘to dry’ and ham ‘to cool’, but the issue needs further investigation.

Language change in Old Kanembu and Kanuri   

   313

lects as closely associated with scholarly activity as Gazar was. In modern Yerwa, initial /h/ in these words is a full-fledged phoneme and it is not clear whether /h/ was stabilised recently or it was never in free variation. What the current phonemic status of /h/ in this word-set does show is consistent correlation with the /h/ in the manuscripts and Old Kanembu/Tarjumo recitation. Of all tokens already mentioned, the verb hap ‘lift up, raise’ is likely to be a pivotal item in scholarly influence on the spoken medium. The verb is used for translation of the Qur’anic Arabic rafaᶜa ‘to raise’ (in reference to raising (by God) the sky, Mount Sinai, or raising in rank, etc.). This verb occurs 28 times in the Qur’an and so does the Old Kanembu/Kanuri verb hap ‘to raise’ (in reality it is much more frequent because of specific recitational techniques). Therefore, hap is the most frequent of any h-initial native words used in exegetic context. In the dialects distant from OK/T recitational practices, neither ᶜulamāᵓ nor the ordinary speakers maintain the h-initial feature in hap (Table 5 and Figure 3), let alone other liturgically less frequent (originally) h-initial words.¹⁴

Old Kanembu

Tarjumo

Yerwa

Mowar

Kuwuri

Tumari

Bol

hap

hap

hap

ʔap

yap

ʔap

ʔap

Table 5: hap ‘to lift up’ in OK, T, MK (Yerwa and Mowar) and MKb (Kuwuri, Tumari, and Bol).

A summarising picture of the development of /h/ in Kanuri can be shown in four stages as follows: 1) Native word initial *h was lost and was only surviving in manuscripts. 2) Without any h-like fricatives in Kanuri/Kanembu, Arabic words with velar and pharyngeal fricatives (/x/ and /ħ/) were borrowed with no initial fricative. 3) The mass introduction of Arabic words with initial and medial /x/ and /ħ/ brought about a new phoneme h. This phoneme did not affect native words with the former *h-initial sound in the dialects where the sound was lost. 4) However, in the dialects closest to the centres of literary activity (which maintained written ) the sound *h was not lost entirely, being reflected in free variation h ~ Ø before /a/ in native *h-initial words. In these dialects the fricative value [h] of the syllabic onset was reinforced by the new phoneme /h/ of Arabic origin and thus [h] in these *h-initial words was rephonemicised.

14 There are contradictory data from the Kanembu dialect Kowono (my field notes) where initial [h] occurs in the same words as in Kanuri Yerwa. It is however unclear if it is a retention of initial h or a late reintroduction due to contact influence of Arabic [ħ] and Tubu [h].

314   

   Dmitry Bondarev

Figure 3: Isogloss ‘lift up’: Yerwa [hap] ~ other dialects [ʔap] &[yap].

5 Old Kanembu > Kanuri: ongoing interaction While the two previous cases were meant to demonstrate the written-mediuminduced changes that had already taken place at earlier stages of the language and, as such, are complete, the function of a case clitic ye may provide evidence for an ongoing process of interaction between the speakers of MK and the users of OK. The postpositional clitic ye is the subject marker in OK, T and MK. There is however a crucial functional difference between the marker ye in OK and T on the one hand and MK on the other. In OK and T, ye is syntactically conditioned and is obligatory on subject noun phrases (NPs), whereas in MK the subject-marking choices are more complex. In modern-day Yerwa Kanuri ye is essentially a differential subject marker conditioned by: a) agential semantics of the referent; b) verbal transitivity; c) cognitive status of the referent (shared knowledge) and d) TAM. In the remainder of this Section I will discuss how prescriptive judge-

Language change in Old Kanembu and Kanuri   

   315

ments of the Tarjumo experts make speakers simplify conditions for the choice of ye-marked subject NPs in modern Kanuri – the situation similar to what Samuels (1972: 100, 106) describes as the “conservative influence of the written language” resulting in speakers’ decisions in favour of one variant, which “may run counter to functional factors in the spoken language” and may produce “an ‘affected’ form”.

5.1 Subject marking in Old Kanembu and Tarjumo The role of ye marker as a purely syntactic device operating on subject NPs in OK and T is illustrated in (9a–c). (9) In Old Kanembu ye marks the syntactic subject (pro)noun. a. ye-marked subject of transitive VP

Iblis-ye 3p-do paradise.from lure.3s.appl.prf ‘Iblis lured them from Paradise’ b. ye-marked subject, intransitive VP

3p-ye stay.3p ‘they stay’ c. non-agential passive subject with ye

3s-io 2p-ye fut-appl-pass-return-appl-2p.fut ‘you will be returned to him’, In (9a), ye marks the subject (Iblis) of the transitive verb ‘to lure’ and in (9b) it marks the pronominal subject (‘they’) of the intransitive verb ‘to stay’. Importantly, the clitic ye also appears on passive subjects as demonstrated in (9c) where the pronoun ‘they’ is ye-marked as the syntactic subject but its semantic role is patient. An equivalent passive construction in MK is ungrammatical: (10) *shí-ro nandí-ye yir-wál-tǝ-w-uwin 3s-io 2p-ye appl-return-pass-appl-2p.prf. ‘you will be returned to him’ When Tarjumo is used for grammatical analysis of Arabic, ye is invariably described as a marker of fāᶜil ‘subject NP’:

316   

   Dmitry Bondarev

(11) From an interview with Imam Goni Bashir, Fezzan Ward, Maiduguri, 27 September 2009 Tarjumodə nyiro ye laa gulzaiya misalro, jaa Zaidu Zaidu-ye ishi gulzaiya ye adə fāᶜil-lo nonǝma ‘When they tell you to analyse ye in Tarjumo, for example when they say [in Arabic] jaa Zaidu ‘Zaidu-ye came’ then you know that ye marks fāᶜil [subject]’ (Bondarev and Tijani in press)

5.2 Differential subject marker ye in Modern Kanuri and prescriptive simplification of its choice Counter to the purely syntactic function of ye in OK and T, in Modern Kanuri yemarking on the subject arguments is conditioned by overlapping morphosyntactic, semantic, and pragmatic factors and manifests (conditioned) split-intransitive alignment (Bondarev, Jaggar, Löhr & Tijani 2011). This complex requirement can be schematised as follows: If [+familiar/inferable], [+agential/volitional], [+transitive], [–FOC.TAM] then YE. Speakers usually agree on the use of ye in [+familiar/inferable], [+transitive], [+agential/volitional] environment (12) and zero ye in [+familiar/inferable], [–transitive], [–agential/volitional] environments (13): (12) situation familiar to addressee, agential subject daré-ye wúa kapkóno police-ye 1s.do stop.3s.past ‘the police stopped me’ (13) familiar situation, non-human/non-volitional subject kâu Ø motandéro sukkurû gǝlasndé kálzǝ daáiye stone Ø car.2p.io fall.down.3s.cnv glass.2p break.3s.cnv stop.1p.past ‘a stone fell on our car, the windscreen broke, and we stopped’ In borderline cases with non-human subjects of active intransitive verbs (e.g. ‘come’) speakers’ judgements vary; cf. (14) non-human “letter” and (15) human “Ali” referents:

Language change in Old Kanembu and Kanuri   

   317

(14) ngô, wotíya Ø isǝ́na look, letter Ø arrive.3s.prf ‘look, the letter has arrived’ (15) nonǝ́ma, bíska Áli-ye isǝ́na. know.2s.prf yesterday Ali-ye come.3s.prf ‘you know, Ali came (back) yesterday’ [with the expected gifts] As mentioned above, the grammatical analysis of Arabic in Tarjumo describes ye as a marker of fāᶜil ‘subject NP’ applied to both transitive and intransitive (as in (11) ‘Zaidu-ye came’) subjects. However, the influence of Arabic grammatical terms does not stop here and has a further effect on grammatical concepts expressed in T which leads to inconsistency. According to the traditional Arabic grammar, intransitive and transitive verbs are clearly distinguished under terms fiᶜl lāzim (intransitive) and fiᶜl mutaᶜd (transitive). The Tarjumo grammatical tradition follows the Arabic model but in a peculiar way: when the Tarjumo scholars clarify why in Modern Kanuri subjects of transitive clauses like (12) take ye, they explain it by the use of the transitive verb that has a surfaced direct object (Ar.: mafᶜūl bihi ‘direct object’). This explanation shows a preference to use ye on transitive subjects regardless of cross-syntactic (both transitive and intransitive) occurrence of ye in the manuscripts and recitation. Note, that this preference has nothing to do with Arabic where the syntactic subject is always nominative irrespective of verbal transitivity. On the one hand, this kind of Tarjumo-based grammatical explanation shows a clash between MK and Arabic grammar (i.e. differential subject marking in Kanuri vs non-differential in Arabic) but on the other hand, the Tarjumo-based approach applied by the Tarjumo experts to Kanuri demonstrates an effort to tidy up Kanuri grammar in the complex area of syntactic, semantic and pragmatic variation. Hence the prescriptive solution: “put ye on any transitive subject”. This prescriptivity does not stay within the scholarly network. During my various research trips in Maiduguri in 2005–2009, I conducted interviews on several grammatical features in Kanuri, including differential subject marking. If the speakers were not certain about their choices, they often resorted to the advice of one of the ᶜulamāᵓ who lived in the neighbourhood. These consultations usually brought about interpretations different from original elicitations. Thus, after having talked to someone of authority in Tarjumo about the use of ye, speaker A. reanalysed all instances of discourse-triggered (= familiarity/inferability) ye on subjects of transitive verbs as conditioned by transitivity only:

318   

   Dmitry Bondarev

(16) (= 12) [on the speakers “improved” analysis, familiarity is irrelevant] daré-ye wúa kapkóno police-ye 1s.do stop.3s.past ‘the police stopped me’ This evidence is still inconclusive because the test was only run with a few Kanuri speakers. However, all those who had based their final judgement on prescriptive advice, independently reanalysed ye in (12/16) as conditioned by transitivity. If this prescriptive reanalysis had a potential for language change, it would have simplified pragmatic factors and reinforced the syntactic and semantic ones – a hypothetical situation with obligatory ye-marking on transitive and agential subject. Since this hypothetical change has not taken place, the illustrated pragmatic – non-pragmatic variation may only show a stabilised situation of coexistence of two grammatical microsystems – non-differential subject marking in OK/T and differential one in MK – with no apparent potential for change in the spoken language. What this example clearly shows however, is how influential the traditional scholars may be when it comes to their opinion on language use. Even though OK/T is entirely detached from the spoken medium, its users are potential propagators of variation.

6 Conclusions OK/T, as it is attested in the early manuscripts and modern-day exegetical practices, is a complex specialised language that reflects various grammatical and lexical structures from past stages of the Kanembu and Kanuri dialects. OK/T is a non-communicative medium better understood as a metalanguage exclusively used for codified translation from Arabic. It has been shown that though OK/T is a technical language non-transparent to laymen, it can nonetheless be identified as a proper written/literary code, or register, of the Kanuri language due to the social status of the traditional Islamic scholars versed in OK/T. The multiplex nature of networking ties, characteristic of the traditional scholarly milieu, positions the OK/T code as one of the vehicles of language variation and change. The potential of the OK/T written/literary code to induce change in spoken dialects has been illustrated by two selected dialect variables (z) and (h). In the case of (z), the major southwest dialect Yerwa spoken in the centre of the OK scholarly activity has non-palatal /z/ retained in the important Islamic terms zǝ́ma ‘Friday’/‘Friday prayer’ and zánna ‘paradise’ and also in many other z-initial words, whereas other Kanuri and Kanembu dialects have the /z/ palatalised

Language change in Old Kanembu and Kanuri   

   319

to /j/. Hence, the process of lexical diffusion of [j] as a variant of (z) was stopped by OK/T bringing about a lexical split. In the case of (h), the native word initial *h was lost but subsequently was rephonemicised in the dialects spoken by the users of OK/T. The mapping of OK/T on dialect variation has proved to be valuable for demonstrating changes induced by the written medium, and I hope that this study of the influence directed by the written and literary media on the spoken codes of the language adds to deconstructing the dominating hierarchy in linguistics that privileges the spoken medium over the written/literary. The ongoing process of interaction between OK/T and spoken Kanuri has been illustrated by the example of the function of the (subject) case clitic ye. The prescriptive force imposed by experts in OK/T on Kanuri speakers creates variation in analysis of the ye: the speakers simplify semantic and pragmatic statuses of the subject referents marked by ye in transitive constructions. The examples discussed above are far from providing an integral picture of interaction between written and spoken media in Kanuri. Nonetheless, I hope to have demonstrated at least three possible avenues for further investigation: (a)  accounting for dialect variation induced by the written/literary medium, (b) looking for the potential emergence of a change caused by prescriptive advice on the part of the users of the written/literary medium, and (c) considering literary codes/registers valid for changes induced by code contact in general and not only for formal registers of the spoken language. With regard to the characteristics of social networks involved in the literary/ spoken interaction in modern-day Kanuri society, it was impossible to carry out a detailed analysis of every single place where OK/T is practised and where it has become obsolete. However the tendencies observed are these: wherever OK/T is practised it becomes one of the factors for language maintenance and change (even though the literary form is unintelligible to the ordinary speakers). Where OK/T is not practised, the written/literary form of the language is irrelevant and the dialects change due to other internal and external factors. With these tendencies in mind I finish with a broad-strokes picture of OK/T “framed” by Kanuri and Kanembu dialects. Starting from about the 14th century Kanem, Old Kanembu has been moving further southwest to the province of Borno with the migration of the royal court and the scholarly communities whose many privileges depended on the links with the ruling dynasty. On its way, this literary language accumulated various linguistic features from the dialects spoken in the centres of scholarly activity. Once Old Kanembu moved forward, those dialects were left behind and the bond between them and the written/literary medium relaxed. This geographic shift of literary activity is variously reflected in dialect variation within Kanuri and Kanembu;

320   

   Dmitry Bondarev

and many features of Yerwa Kanuri spoken in Maiduguri can be explained by the ongoing communicative interaction between the Kanuri-speaking communities and Old Kanembu/Tarjumo scholarly networks.

7 Abbreviations and diacritics 1YM – the “Yerima Mustafa Qur’an” 2ShK – the “Shetima Kawo Qur’an” 1(#) s/p – (personal) singular/plural pronouns ACC – accusative AG – agentive/agential APPL – applicative Ar. – Arabic ASSOC – associative CAUS– causative CNV– converb COMP – complementiser DEF – definite DO – direct object FOC.TAM – focus TAM FUT – future GEN – genitive H – head IO – indirect object Kb – Kanembu Kn – Kanuri KY – Kanuri Yerwa MK – Modern Kanuri MKb – Modern Kanembu NEG – negative NDET – determined noun (in genitive construction) OK – Old Kanembu OK/T – Old Kanembu and Tarjumo as a continuous linguistic entity PAST – past tense morpheme PASS – passive PRF – perfective PP – pre/postpositions Q – Qur’an

Language change in Old Kanembu and Kanuri   

   321

SJ – subject marker T – Tarjumo (an offspring of OK)

8 References Aitchison, Jean. 2000. Psycholinguistic perspectives on language change. In Richard D. Janda & Brian D. Joseph (eds.), Handbook of historical linguistics, 736–743. Oxford: Blackwell. Bausi, Alessandro. 1990. La frase relativa nelle lingue semitiche d’Etiopia. Quaderni del Dipartimento di Linguistica 1. 105–123. Bergs, Alexander. 2005. Social networks and historical sociolinguistics: Studies in morphosyntactic variation in the Paston letters (1421–1503). Berlin & New York: Mouton de Gruyter. Bobboyi, Hamidu. 1992. The ᶜulamāᵓ of Borno: a study of the relations between scholars and state under the Sayfuwa, 1470–1808. Evanston: Northwestern University PhD thesis. Bondarev, Dmitry. 2005. Fonologicheskiy contrast ili orfograficheskaya vol’nost’? (o dvuh grafemah v jazyke bornuanskih gloss) [Phonological contrast or orthographic freedom? (Two graphemes in the language of the Bornu MSS glosses)]. In Valentin Vydrin & Sevir Chernetsov (eds.), Ad Hominem, Nikolai Girenko Festschrift, 7–31. St Petersburg: Kunstkamera. Bondarev, Dmitry. 2006. The language of the glosses in the Bornu Qur’anic manuscripts. Bulletin of the School of Oriental and African Studies 69/1. 113–140. Bondarev, Dmitry. 2010. Complex clauses in Old Kanembu/LG. In Georg Ziegelmeyer & Norbert Cyffer (eds.), Aspects of co- and subordination: case studies from African, Slavonic, and Turkic languages, 213–250. Cologne: Rüdiger Köppe. Bondarev, Dmitry. In press. Multiglossia in West African manuscripts: a case of Borno, Nigera. In Jörg Quenzer, Jan-Ulrich Sobisch & Dmitry Bondarev (eds.), Manuscript Cultures: Mapping the Field. Berlin & New York: Mouton de Gruyter. Bondarev, Dmitry. Forthcoming/a. Old Kanembu and Kanuri in Arabic script: phonology through the graphic system. Bondarev, Dmitry. Forthcoming/b. Qur’anic exegesis in Old Kanembu: linguistic precision for better interpretation. In Dmitry Bondarev & Tal Tamari (eds.), Transmission and interpretation of the Qur’an in African languages and styles. [Special issue] Journal of Qur’anic Studies. Bondarev, Dmitry. In preparation. Old Kanembu: a descriptive grammar of the language of the Qur’an glosses. Bondarev, Dmitry, Philip Jaggar, Doris Löhr & Abba Tijani. 2011. Differential subject marking in Kanuri: agentivity, pragmatics, and split-intransitive. In Doris Löhr, Eva Rothmaler & Georg Ziegelmeyer (eds.), Kanuri, Borno and Beyond – Current Studies on the Lake Chad Region, 27–57. Köppe: Köln. Bondarev, Dmitry & Abba Tijani. In press. Performance of multilayered literacy: Tarjumo of the Kanuri Muslim scholars. Britto, Francis. 1986. Diglossia: a study of the theory with application to Tamil. Washington, DC: Georgetown University Press.

322   

   Dmitry Bondarev

Bulakarima, Shettima Umara. 1996. Kanembu-Kanuri relationship: a proposal. Berichte des Sonderforschungsbereichs 268, 8. 37–47. Bulakarima, Shettima Umara. 2001. A study in Kanuri dialectology: phonology and dialectal distribution in Mowar. Maiduguri: Awwal Printing & Publications. Chafe, Wallace. 1982. Integration and involvement in speaking, writing and oral literature. In Debora Tannen (ed.), Spoken and written language: exploring orality and literacy, 35–53. Norwood, NJ: Ablex. Chafe, Wallace. 1994. Discourse, consciousness, and time: the flow and displacement of conscious experience in speaking and writing. Chicago & London: The University of Chicago Press. Chonai, Hasan. 1999. Genealogical relationships in Teda-Daza language group (Central Saharan linguistic family): phonology and etymology. Moscow: Russian State Humanitarian University PhD thesis. Croft, William. 2000. Explaining language change: an evolutionary approach. Harlow & New York: Longman. Cyffer, Norbert & John Hutchison. 1990. Dictionary of the Kanuri language. Dordrecht: Foris Publications/Maiduguri: The University of Maiduguri. Cyffer, Norbert. 1997. A survey of the Kanuri language. In Norbert Cyffer & Thomas Geider (eds.), Advances in Kanuri scholarship. Westafrikanische Studien 17, 17–66. Köln: Köppe. Dobronravin, Nikolay. 1999. Arabograficheskaya pis’mennaya traditsiya Zapadnoy Afriki. [Arabic-script written tradition of West Africa]. St. Petersburg: SPb University. Duisburg, Adolf von. 1913. Grundriss der Kanuri-Sprache in Bornu (Archiv für das Studium deutscher Kolonialsprachen, Band XV). Berlin: Georg Reimer. Ferguson, Charles Albert. 1959/1972. Diglossia. In Pier Paolo Giglioli (ed.), Language and social context: selected readings, 232–251. Harmondsworth: Penguin. (Originally published in Word. Journal of the Linguistic Circle of New York 15/2. 325–340). Fishman, Joshua. 1972. The sociology of language. Berlin: Walter de Gruyter. Grierson, George. 1920. Indo-Aryan vernaculars (continued). Bulletin of the School of Oriental Studies 3/1. 51–85. Gumperz, John. 1968/1972. The speech community. In Pier Paolo Giglioli (ed.), Language and social context: selected readings, 219–231. Harmondsworth: Penguin. (Originally published in International encyclopaedia of the social sciences, New York: Macmillan, 381–386.). Hutchison, John P. 1981. The Kanuri language: a reference grammar. African Studies Programme: University of Wisconsin, Madison. Kerswill, Paul. 1996. Children, adolescents, and language change. Language Variation and Change 8. 177–202. Koelle, Sigismund Wilhelm. 1854a. Grammar of the Bornu or Kanuri language. London: Church Missionary House. Koelle, Sigismund Wilhelm. 1854b. African native literature, or proverbs, tales, fables, and historical fragments in the Kanuri or Bornu language. London: Church Missionary House. Labov, William. 1980. The social origins of sound change. In William Labov (ed.), Locating language in time and space, 251–266. New York: Academic Press. Le Coeur, Charles & Marguerite Le Cœur. 1956. Grammaire et textes Teda-Daza. Dakar: Memoires IFAN (Limoges, impr. de A. Bontemps).

Language change in Old Kanembu and Kanuri   

   323

Löhr, Doris. 1997. The development of Kanuri orthography from 1854 until the present. In Norbert Cyffer & Thomas Geider (eds.), Advances in Kanuri scholarship. Westafrikanische Studien 17, 77–113. Köln: Köppe. Löhr, Doris. Forthcoming. Malgwa-English dictionary. Lukas, Johannes. 1931. Die Sprache der Káidi-Kanembú in Kanem. Berlin:Verlag Dietrich Reimer. Reimer (Ernst Vohsen) A.-G. Lukas, Johannes. 1937. A study of the Kanuri language, grammar and vocabulary. London & New York: Published for the IAI by the Oxford University Press. Lukas, Johannes. 1953. Die Sprache der Tubu in der Zentralen Sahara. Berlin: Akademie-Verlag. Milroy, James. 1992. Linguistic variation and change. Oxford: Blackwell. Mustapha, Abubakar. 1987. The Contribution of Sayfawa ‘Ulama’ to the Study of Islam: c. 1086–1846 AD. Kano: Bayero University PhD thesis. Newman, Paul. 2000. The Hausa language: and encyclopedic reference grammar. New Haven & London: Yale University Press. Romaine, Suzanne. 1982. Socio-historical linguistics: its status and methodology. Cambridge: Cambridge University Press. Samuels, Michael Louis. 1972. Linguistic evolution: with special reference to English. Cambridge: Cambridge University Press. Schuh, Russell G. 2002. Palatalization in West Chadic. Studies in African Linguistics 31. 97–128. Schuh, Russell G. 2003. The linguistic influence of Kanuri on Bade and Ngizim. Maiduguri Journal of Linguistic and Literary studies (MAJOLIS) 5. 55–89. Schneider, Edgar W. 2008. Variation and change in written documents. In Justin K. Chambers, Peter Trudgill & N. Schilling-Estes (eds.), The handbook of language variation and change, 67–96. Oxford: Blackwell. Sima, Alexander. 2010. Amharische Syntax in späten Originalwerken des Gǝʿǝz. Aethiopica 13. 89–103. Tosco, Mauro. 2000. Is there an “Ethiopian Language Area”? Anthropological Linguistics 42/3. 325–365. Uspensky, Boris Andreyevich. 2002. Istoriya russkogo literaturnogo yazyka (XI–XVII) [History of the Russian literary language (11th–17th centuries)] 3rd edition. Moscow: Aspect Press. Wagner, Ewald. 1983. Harari-Texte in arabischer Schrift: mit Übersetzung und Kommentar. Wiesbaden: Franz Steiner. Wardhaugh, Ronald. 2006. An introduction to sociolinguistics. 5th edition. Oxford: Blackwell.

Index Abbasid 199–212 acceptability 6, 8, 10, 14, 16, 163–166, 192, 264, 270, 287, 291 “accuracy” (of Hebrew) 231–232 administrators 11–12, 99–101, 112–117, 206, 210–212 adnominal gender exponents 139–141 adoption (of language features) 52 Akkadian 9, 85–96 alloglottography 9 alphabet 58, 72, 92, 217–225 Ancient South Arabia 4 – see also Epigraphic South Arabian Anglo-Saxon 44, 50, 121–122, 127–128, 248 – see also West Saxon Arabic 199–215 – Classical 261–275 – Judaeo-Arabic 8, 10, 13, 15–16, 187–196, 261–275 – Middle 8, 14 – “mixed” 260–261 – Neo-Arabic 260 Arabic script 210–212 Aramaic 8, 9, 11, 189, 192, 202, 205, 212, 236–237, 264 authenticity 15, 99, 269, 274 Barmston 164–166 Basque 7 Ben Ezra synagogue, Fusṭāṭ 183, 185 Bible 10, 21–37, 61, 167–168, 176–178, 187, 191–196, 227, 230, 232, 237 bi-imperfect 268–269 bilingualism 4, 74, 201, 221, 224 Boleyn, Anne 99, 106–109 Boson, Nicholas 21–37 Brythonic 6, 21–37 Byzantium 186, 191–192 – see also Greek, Byzantine cadastres 284–285 Cairo Genizah 183–198, 259 cancelleresca 271 chancery 9, 11, 99, 272–273, 277–281, 288

Chancery Standard 162 charters 46, 277, 283 Celtic 21–37 Chinese 7–8, 10, 40–41 code switching 188–190, 195 colloquial 14–15, 261–274, 304 communicational space 39–69 complexification 14 Conceptual Continuum 40, 53–54 Confucius 10 conservatism 9–16, 46, 52, 59, 61, 88, 101, 108–117, 160, 249–250, 315 consonant mutation 29–31 contact see language contact continuglossia 262 copying 3, 41, 90, 166, 174, 188, 246 Cornish 6, 21–37 Corpus of Early English Correspondence (CEEC) 99–119 creolisation 10, 262 Cromwell, Thomas 13, 99, 109, 112–113, 117 cuneiform 85–96 curial prose 101, 107, 112 cursivity 11, 210–212 Cyrillic 6 Danish 15, 277–290 de-standardisation 6, 9–12, 160 deviation 52 dictation 5, 166, 241, 246–249, 256 diglossia 4, 55, 262, 273, 291, 300–302 distance 13–14, 53–64, 107–108, 269, 274 Donatus 79 dry-point glosses 6, 39–69 Echternach 43–44 educational systems 10, 187 Egypt 5, 11, 13, 186, 190–191, 196, 199–215 elite 4, 192–197 endangered 6, 36 English 10, 13, 71, 80, 121–158 – Old 10, 121–158 – Middle 10, 15, 121–158 – Modern 10

326   

   Index

Epigraphic South Arabian 202 Etruscan 218

“invisible hand” 3, 64–65 Iraq 186, 189, 199–203

formality 13, 15 , 31, 43, 56, 77, 107, 117, 190, 243, 282–284 Freising 47–48 Fusṭāṭ 5, 186–187, 193

Japanese 9 Jerusalem 187, 194 Judaeo-Arabic see Arabic jurists 199–203, 212

Galicia 8, 71–83 gender 10, 69, 102, 121–158 gender change 121–158 gender reassignment 141–143, 150–155 Genesis 21–22, 26–37 genre 88, 90, 282, 284 geʾonim 187–189, 196–197 German – Old High 6, 16, 39–69 – New High 10 glosses 6–7, 40–41, 127–128, 167, 291 – dry point 6, 39–69 glossators 6–7, 39–69 Gothic 10 Greek 12, 192 – Byzantine 201–202, 205 Greek alphabet 218

Kanembu see Old Kanembu Kanuri 291–323 Keigwin, John 21–37 Korean 7

Hebrew 10, 12–13, 32–34, 183–198, 227–238 – Biblical 227, 234–237 – Medieval 183–198 – Rabbinic 227, 232–237 Hebrew script 187–188 Hebrew vocalisation 195–196 Henry VIII 99–119 Hittite 9, 85–96 holograph 99–119 Ibero-Romance 7, 71–83 identity 159–181, 194, 263, 300–301 ideology see linguistic ideology idiosyncrasy 6, 12, 31, 53, 175, 221 immediacy 14, 53–63 innovation 5, 12–15, 52, 87, 100–101, 110–111, 117, 141–142, 178, 199, 204, 210, 212, 228, 242, 244–250, 255–256, 272, 274 interlocutors 13, 263, 269 intimacy 7, 14–15, 108, 192, 261, 269, 271, 274

language attitudes 5, 228, 251, 277–278, 282, 284, 289 language contact 10, 43, 292–293, 308, 313 language death 21–24 language-learner error 220 language shift 10, 21–24, 204, 210, 277–288 Latin 7–8, 14, 39–69, 71–83, 127–128 Latin alphabet 218 lenition 29–31 level of formality see formality Leviticus 188 lingua franca 56, 128 linguistic ideology 5, 228 literacy 4–5, 63, 74, 77, 100, 102, 122, 163, 217, 224–225, 241–247 logograms 8, 85–96 logo-syllabic writing system 85–96 Maihingen gospels 43–47 March 167–178 mercantesca 271 medium transferability 54 merchants 13–14, 99, 102, 112–118, 190–191, 195, 261, 272–274 metalanguage 294, 299–302, 318 middle class 36, 187, 192, 272–273 migrations 8, 101, 162, 279, 310 monks 6, 14, 39–56, 61, 130, 255 morphology 153 – Hebrew 232 – Hellenised 224 – Ibero-Romance 7–8, 71, 76–81 – Judaeo-Arabic 265 – Kanuri 295

Index   

– Norwegian 280, 283 – Old English 122 – Old Kanembu 295 – Oscan 221 multiglossia 259–262 multiplexity 302, 310, 318 Nathan b. Abraham 193–196 negation 15, 112–117, 264–270, 274, 299 norms 5, 221, 278–279 Norwegian 15, 277–290 Notkers Anlautgesetz 49 Old Kanembu 15, 291–323 optimal language development 9 oral culture 4, 15, 56, 122 oral delivery 127 orality 14, 54, 127, 236, 264, 269, 294 – conceptional 7, 39 – elaborate 56–60 orthography – Arabic 8, 264–265, 273 – Cornish 21, 28–33 – English 7, 167–171, 173, 245, 250, 256 – Hittite 94 – Ibero-Romance 7, 81 – Norwegian 277, 280–287 – Osco-Greek 217–226 – West Saxon 122 Oscan 12, 217–226 Oxyrhynchus 189 Pahlavi 11, 199–200, 207, 211 Palestine 11, 13, 186, 188–189, 194 papyri 189, 199–202, 210–211, 267 Paston letters 163, 241–242, 246–250, 256, 259 Pentateuch 188 Persian 9, 264 Peterborough Chronicle 121–158, 241–258 Peterborough Continuations 121–158 phonetics 6–7, 29, 44, 50, 71–82, 218, 220, 252 phonetic complement 85–96 pidginisation 262 poets 185 poetry 56, 126–127, 186, 191, 262 Portugal 71–83

   327

precedent 10–16, 271–274 prescription 243, 251, 264, 267, 273, 302, 309–310, 314–319 prestige 9–12, 15, 24–25, 35, 90, 95, 164, 187, 196, 266, 269, 279, 289, 300–302, 312 progressiveness 7, 10, 12–13, 39, 42, 46–64, 109, 112–118, 160 pronouns 49, 64, 79, 108–111, 116, 118, 121–158, 170, 201–202, 206, 242, 248–250, 283, 298–299, 315 – see also relative pronouns rapid language change 3, 8, 10, 249 reading 245–246 reading aloud 8, 54, 62, 72–73, 77–78, 176, 192–193, 196, 263, 298 reading silent 8, 127, 298 recessiveness 10–11, 160 referential gender agreement 10 registers 3–18, 21–22, 31, 72–73, 99–101, 112, 116–118, 130, 163–168, 178–179, 195, 227–238, 243, 261–275, 277–278, 282, 292–293, 304, 318–319 regionalisation 6, 9–12, 159–181 relative pronouns 109–112, 116, 118, 126, 129, 141–142, 285 – Arabic 8, 264–267, 272–274 religiolect 261 respectability 16 Romance 7, 55, 57, 60, 71–83 royalty 4, 12–13, 81, 85, 89, 99, 101–118, 319 runes 58 Sabaic 4 Sabellic languages 217–218 Second Consonant Shift 45, 55 semantics 7, 71, 76–81, 314 Schriftsprache 122, 160 scribal caste 4 scribal community 277–278 scribal schools 6, 11, 217, 219, 225 scribes 3–16, 21, 45–46, 61–62, 71–81, 85–89, 93–95, 102, 108, 121, 131, 159–179, 192, 217, 223, 227, 235, 237, 241–257, 261, 264, 272, 277–289, 292–293

328   

   Index

secretaries 13, 99–118, 164, 208 Shabbethai ben Isaac Sofer 227–238 social distance 13, 43, 53–64, 267 social ties 302 Solomon b. Judah 183, 187–196 Spain 185–186, 204 speaking 3–5, 7, 243–245, 261–263, 277–278 spelling 7, 9, 13, 28–29, 39–69, 71–82, 86–89, 106–111, 118, 134, 163–178, 194–195, 217–226, 242, 245, 251–255, 265, 280, 283–289 split intransitive alignment 316 standardisation 6, 9–12, 16, 59, 73, 159–166, 179, 229, 234–235, 283, 288 status 4, 13, 25, 36, 64, 102, 177, 183, 191, 195–196, 209, 301–302, 318 subject marking 314–318 substandard 3, 11, 261, 275 Sumerian 9, 85, 92–95 supralocalization 162–166, 172, 176, 178, 281 syntax 6–8, 30–31, 41, 5051, 57, 62, 64, 71–72, 76, 78–81, 93–95, 127, 229, 242, 245, 247, 250, 256, 264–266, 273, 280, 295, 297–298, 303 Tarjumo 15, 291–323 threshold 7, 15 traders see merchants

translation 6, 10, 21–37, 57, 61, 63, 127, 167–168, 186, 192, 254, 291, 294, 300 Tudor period 99–119 Umayyad period 199, 205–206, 212 uniformitarian principle 100, 282 upwardly mobile 13, 16, 100 utilitarian prose 4 variation 3–4, 40, 52–54, 76, 111, 121–122, 138, 141, 152, 155, 160–162, 176, 179, 217, 221, 245, 252–253, 277–290 – dialectal 292–295, 304–308, 319 – geographical 10–11 – individual 12, 222 – orthographic 251, 276 – register 13, 99–101, 112, 118, 280 variability 9 vernacular 5–8, 13–16, 21, 24–25, 39–69, 122, 127–128, 187, 189, 196, 230, 236, 261–275, 301 Verschriftung/Verschriftlichung 5–9, 14, 16, 54, 58–65 vocalisation 196, 229–232 West Midlands 125, 130, 142–143, 151–152, 155, 159–161, 167–178 West Saxon 122, 130, 132, 152, 160 word order 31, 33, 72, 80, 93–94, 122, 298 Yiddish 228, 233, 273