G
REEK
-L
ATIN
L
ANGUAGE
C
ONTACT AT THE
R
OMAN
Q
UARRY OF
M
ONS
C
LAUDIANUS
M.A. Thesis
Karel Stegeman (RUG) – s221107 CMRS – Classics
Supervision Prof. Mark Janse Prof. Gerry Wakker
3
P
REFACE
Mattock de trench maar zo ver mogelijk uit tot je de drain exposet, dan kan ik je wel helpen om die uit te trowelen. Als je finds hebt, moesten dacht ik in die groene plastic tray.
quote from Anna at the excavations of the Poulton Research Project (summer of 2016, Poulton, UK) After one and a half year I am very glad to present my work on the Mons Claudianus ostraca. Even though I often wished that the research and the writing would have gone faster and that I would not have had to face the often frustrating difficulties that I have had, in the end I am happy that I took the time to complete my thesis on my own terms. I am equally proud of the academic contribution (albeit small) that I have delivered and the personal challenges I have overcome.
The quote above is from my friend Anna, with whom I spent the summer of 2016 at the Roman excavations of the Poulton Research Project in Poulton, UK, supervised by dr. Kevin Cootes. I jotted her words down as soon as we came home after we spent a long and hot day digging in the British soil. After the lunch break I had forgotten the instructions Anna and me got from our supervisor; luckily Anna knew what the plan was and replied with the quote above, ‘Use a mattock to dig out the trench as far as possible until you expose the drain [i.e. a 19th century field drain that was in the way of our
excavation for Roman remains], then I’ll be able to help you uncover it with a trowel. If you happen to have any archaeological finds I believe you have to leave them in the green plastic tray.’ The similarities to the situation at Mons Claudianus, which I had been working on, struck me – us, two workers, learning about tools and techniques from foreigners, digging in the ground, happily and mindlessly code-switching and borrowing as we dug away.
This thesis would not have been possible without the extraordinary supervision of prof. Mark Janse from the University of Ghent. After I had spent a semester at his university with an Erasmus Grant, he offered to supervise my thesis remotely. His academic and personal guidance and investment have been invaluable; I hope that the work I present here at least reflects some of the knowledge and wisdom he shared with me.
I also thank prof. Gerry Wakker, my second supervisor. I know that as the dean of the Faculty of Arts she has a very full schedule, but I am very happy that she as one of the few remaining linguists of the University of Groningen Classics department gladly agreed to help Mark Janse in the supervision. As she said herself: ‘natuurlijk past dit eigenlijk niet, maar sommige dingen zijn gewoon te interessant om niet te doen’.
4
I would like to thank my family for their support; my brothers Wouter and Jan-Jaap, my mom for her everlasting interest in the (sometimes highly obscure) research field of me and my brothers, my dad for his personal advice that helped me through the first stages of my work. Lastly, I thank my friends Anna, Daan, Emma, Jeroen, Sanne and Tritia for their support and companionship during working days which helpd me tremendously. A special thanks goes out to Ruben who wrote a script for counting the Greek graphemes rendering different Latin ones; I was very happy to work with him and he saved me a lot of tedious but otherwise inevitable counting.
5
T
ABLE OF
C
ONTENTS
Preface 3
Table of contents 5
Chapter 1 Mons Claudianus and its ostraca 6 1.1 The Roman quarry at Wadi Umm Hussein 6 1.2 The personnel of Mons Claudianus 8
1.3 The ostraca of Mons Claudianus 9
Chapter 2 Language contact in linguistics 11 Chapter 3 Language contact phenomena at Mons Claudianus 22
3.1 Language choice 22
3.2 The semantic areas of lexical borrowing 27 3.3 Integration of Latin elements at Mons Claudianus 42 3.3.1 Orthography and phonology 42
3.3.1.1 General remarks 42 3.3.1.2 Vowels 45 3.3.1.3 Diphthongs 50 3.3.1.4 Semivowels 50 3.3.1.5 Consonants 52 3.3.2 Morphological integration 55 3.4 Morphological borrowing: the case of -άριος 58
3.5 Interference 61
Conclusions 62
Bibliography 64
Appendix I: Latin loanwords in Greek in O.Claud.
Appendix II: Roman names rendered in the Greek documents of O.Claud. Appendix III: Transcriptions of Latin words in Greek documents of O.Claud. Appendix IV: GSH application
6
C
HAPTER
1
Mons Claudianus and its ostraca
1.1 The Roman quarry at Wadi Umm Hussein
In the middle of the Eastern Egypt Desert we find now some of the most remote traces of structures erected during the Roman Empire at a place known at the time as Mons Claudianus. Anyone arriving in the harbour of Alexandria and wanting – for whatever reason – to travel to Mons Claudianus was to travel 500 kilometres south up the Nile river. From Kaine (mod. Qena) the traveller was then to leave the fertile valley for a couple of days worth of travel through the desert for around 150 kilometres east climbing a little under 700 metres before finally arriving at the quarry in the Red Sea mountains (Van der Veen 1998: 101). Certainly in ancient times this last part of the trip was no sinecure: apart from the extreme natural conditions – the dangers of dehydration, scorpion stings and sunstrokes never far away – one had to be on one’s guard for raids by Bedouin tribes.
In the last two millennia there have been two recorded groups of Europeans who have taken an outstanding interest in the area: the Romans who exploited the site for its granodiorite and provided the infrastructure to quarry the rock, shape it and transport it to Rome. The other group consists of the
7
archaeologists, papyrologists and historians who have attempted to uncover this infrastructure. The site was discovered in 1823 by Burton and Wilkinson (Bingen 1992: 10); however, the site was not excavated until the twentieth century. In the 1960s, a German team of researchers carried out a topographical survey, but they did not yet excavate the quarry itself (Bingen 2016: 9). The major excavation, which also yielded the ostraca that form the basis for the present study, was carried out by the French Institute for Middle Eastern Archaeology in Cairo from 1987 to 1993.1 The expedition was
led by Jean Bingen and yielded various publications, among which four corpora of each around two hundred ostraca, together amounting to around 10% of the total number of ostraca found on the site.
The Romans started quarrying the area during the reign of Claudius (41-54 AD; hence called Mons Claudianus). Their goal was to secure blocks of granodiorite, a granite-like rock, from the mountainous area. This material is also known as grey granite or by its Italian name granito del foro ‘granite of the forum’ – the latter name deriving from its use by Trajan in his forum in Rome. The quarry received a major impetus during his reign (98-117 AD) to supply the granodiorite for the pillars in his basilica Ulpia and quarrying operations were vastly expanded. The quarrying continued throughout the second century up until the first decades of the third century, until the reign of Alexander Severus. By the time the crisis of the third century hit the Roman Empire, the demand for the granodiorite from Mons Claudianus, a luxury building material, was waning; as Bingen (2016: 14) puts it, ‘the luxury of acquiring a rare granite reserved solely for the emperor’s use no longer corresponded to the mentality of the times’. The operation continued for some time, however, in isolation of the political, military and economic troubles elsewhere in the Empire, although our evidence indicates that activity ceased halfway the third century (Van der Veen 1997: 6f.).
The extraction of granodiorite was no sinecure. Quarrying was a matter of hard manual labour which workers at Mons Claudianus had to perform under the most challenging natural circumstances. The exceptional heat and drought were tremendous challenges for the work that was already extremely physically demanding. As Van der Veen (1998:101) mentions, ‘most of the Eastern Desert is classified as hyper-arid: mean annual rainfall is 5 mm, though this figure does not reflect recurrent rainfall but ‘accidental’ cloudbursts; indeed, some years see no rainfall (Zahran and Willis 1992). The present arid conditions have prevailed since c.3000 BC (Butzer 1961, 1976; Zahran and Willis 1992).’ Aside from the natural inhospitality of the region, Bedouin tribes were known to raid the area; the quarry with all its supplies in food and water thus was in need of defence by the Roman army, who had a headquarters on site. We know from the ostraca that around one thousand people lived at the quarry (Van der Veen
1 As the present study is mainly on the ostraca and their linguistic significance, the archaeology will not be discussed in
detail here; for discussions of the archaeology, see Bingen and Jensen (1993), Maxfield-Peacock (1997, 2001 and 2006), Van der Veen (1997 and 1998) and Bingen (2014).
8
1998: 101), who all needed food and water, neither of which was readily available in the vicinity. The Romans achieved this by installing posts along the 150km road leading from Mons Claudianus to the Nile river; a five day trip in ancient times, probably longer when travellers were bearing heavy loads (say, full sized columns). Workers not only quarried the stone but already cut the granodiorite into cylinders, the rudimentary shape of the columns they were destined to be so that their colleagues who had the duty of transporting the material to Kaine did not have to carry any excess weight.
One can thus imagine the huge operation the Romans had to run in order to supply the capital with its demand of granodiorite. But what demand was there, really? As Bingen mentions, the Romans could have (and have) easily found similar building materials closer to home (Bingen 2016: 16). It is, however, the very distance and toil needed for its extraction which Peacock (followed by Bingen 2016) assumes to be the main value of Mons Claudianus granodiorite (Peacock 1992). He investigated the use of Mons Claudianus granodiorite and concluded that it was used exclusively in imperial structures. The impressive operation of its extraction was the doing of a ruler of an Empire that spanned deep into the Egyptian desert. The arrival of the material from Egypt at the port of Ostia was a display of the impressive infrastructure required; there was no doubt that the emperor was in complete control of the Mediterranean. Additionally, since the emperor was in complete control over the operation, he was the only one who could use the material in his structures; other wealthy Romans could import all the stones they wanted for their porticoes, temples and houses, but the usage of Mons Claudianus granodiorite was an exclusive right to the imperial family (Bingen 2016: 16).
1.2 The personnel of Mons Claudianus
Apart from the extraordinary infrastructure required for the quarry activities, one might imagine that the human resource department would probably run into trouble finding volunteers to quarry gigantic rocks in the heat and aridity of the East Egyptian desert. We would imagine slavery or convict labour to be the most obvious solution. Thus, Flavius Josephus (Pol. 4. 418) writes: τοῦ δὲ λοιποῦ πλήθους τοὺς ὑπὲρ ἑπτακαίδεκα ἔτη δήσας ἔπεμψεν εἰς τὰ κατ᾽ Αἴγυπτον ἔργα (‘of the rest of them, [Fronto] put those older than seventeen years into bonds and sent them to labour in Egypt’2). In this
context, Whiston understands ἔργα as work in the mines3 and apparently his translation is followed by
Van der Veen (1997: 6 and 1998: 101) and Hirt (2010: 224). Although ἔργα ἀργυρεῖα is an attested
2 All translations are mine unless stated otherwise.
3 Whiston (1895) ‘…and as for the rest of the multitude that were above seventeen years old, he put them into bonds, and
9
phrase for ‘mining silver’,4 the isolated ἔργα is, according to LSJ, not attested in the specified sense of
mining or quarrying activities. Though it is possible that Josephus is thinking of the Roman mines or quarries in the Egyptian desert, he is not explicit about this.5 The Jewish Josephus may even be alluding
to the Jews’ earlier biblical slav status in Egypt in the time of Moses.6 The second source for convict
labour is Aelius Aristides (Log. 36.349), who visited Egypt in 142,7 who writes: [χωρία] ἅ γ’ἐστὶ
ξηρότερα καὶ ψεδνότερα παντὸς οὗ τις ἂν εἰκάζων εὕροι, ὥστ’ ἔν γε τῇ Ἀραβικῇ καὶ ἡ περιβόητος αὕτη λιθοτομία ἡ πορφυρῖτίς ἐστιν· ἐργάζονται δ’ αὐτὴν ὥσπερ καὶ τὰς ἄλλας δή που κατάδικοι (‘[the regions] which are more arid and bare than anything anyone could imagine, and in this very region of Arabia we find that famous quarry of porphyry; perhaps convicts work at that quarry as they do at the other ones’). Cuvigny (2000: 35) and Hirt (2010: 224f.), however, have serious doubts regarding the factual accuracy of Aelius Aristides’s statement.
Not only is the literary evidence scarce and uncertain in its sense, archaeologists have found no direct evidence attesting the presence of convicts at work at the quarries. Hirt (2010: 224f.) suggests that perhaps the σκόπελοι, mentioned frequently in ostraca may possibly be watchtowers that should be interpreted as structures guarding the convicts, but is very doubtful about this possibility. The watchtowers may very simply be protection against outside dangers such as Bedouin raids rather than inside dangers such as riots or escaping convicts.8 Cuvigny (2000: 36), too, is pessimistic; though she
does not reject the possibility of convicts being employed at Mons Claudianus completely, they have not left any trace in the ostraca.
1.3 The ostraca of Mons Claudianus
During the latest excavations, the international team of archaeologists working under the flag of the Institut Français d’Archéologie Orientale9 unearthed a total of around 9000 ostraca covered with
ink writing at Mons Claudianus (Cuvigny 2005:309). Of these ostraca, 896 have been edited by various papyrologists who worked on the excavation; four volumes of edited ostraca (Mons Claudianus.
4 LSJ s.v. A.3.c.
5 Thus Thackeray (1927) translates more reservedly ‘the works in Egypt’. 6 Ex. 1:8-22.
7 Although, as Cuvigny (2000: 35) points out, he did not visit the Eastern Egyptian desert.
8 Considering it being a 150 km walk to the nearest town in the Nile valley and the other nearby sources of food and
water the Roman garrisons along the supply route to and from the quarries in the desert, where would they run to anyway?
9 On the excavation and discussion of the archaeological data, see Bingen and Jensen (1993), Maxfield-Peacock (1997,
10
Ostraca Graeca et Latina) have been published thus far. In addition, several ostraca have been
published in separate articles; they are not included in the Mons Claudianus. Ostraca Graeca et Latina series (Cuvigny 2005; Cuvigny and Wagner 1986). Additionally, a small number of stone inscriptions have been found and published (Bingen and Van Rengen 1986; Cuvigny 1992).
The major part of the published documents deal with the administration of the quarry. The ostraca have proven to be extremely interesting from a historical point of view as they reveal in detail the everyday activities at an imperial quarry.10 The editors of the four main volumes of 896 ostraca
have done an excellent job in their systematic presentation of the texts and their contextualisation. The first two volumes include thematic contributions by seven different papyrologists; the final two are authored by Cuvigny and Bülow-Jacobsen. In their discussions, however, they make only scarce remarks concerning language use. None of the contributors attempt a systematic linguistic analysis even though the documents certainly demand one. The ostraca show Greek and Latin in use side by side in a small language community and consequent language contact phenomena, most notably (lexical and morphological) borrowing; additionally there is the matter of language choice as some documents are in Latin instead of Greek; we find cases of transcription of Latin in Greek letters and many Roman names rendered in Greek. In the cases where Latin loans are introduced in Greek we find morphological and phonological integration. The editors also mention quite regularly that the handwriting of some of the Greek scribes show influences from Latin writing. These observations imply that we can infer a lot about the linguistic reality at Mons Claudianus, the roles that both languages played at the quarry and about Greek-Latin language contact in general. In short, there is enough reason to take a new approach to the Mons Claudianus evidence: the material demands a systematic linguistic study with special regard to the attested Greek-Latin language contact phenomena.
10 E.g. the recent monography of Hirt (2010) on Roman imperial quarries draws heavily upon the recent Mons Claudianus
11
C
HAPTER
2
Language contact in linguistics
The present study will discuss a number of phenomena related to Greek-Latin language contact, in particular borrowing and code-switching. The field of linguistics has made great advances on these topics, from Haugen’s exploratory article (1950) up to extensive recent models (Romaine 1995; Thomason 2001; Croft 2002; Field 2002; Winford 2003; Matras 2009; Haspelmath 2009). These works draw on the large amount of linguistic research conducted in recent decades on various current situations of language contact around the world. Some of the concepts used in the field of linguistics have also been applied fruitfully in the study of corpus languages (cf. Adams, Janse & Swain 2002; Adams 2003; Dickey 2003; Mullen & James 2011; Leiwo et al. 2012).
In this chapter, I will provide a discussion of the concepts of borrowing and code-switching as used in modern linguistics. In the case of borrowing and code-switching, the distinction between the two phenomena has led to disagreement amongst scholars. For our work, then, it will be necessary to not only define these concepts, but also to review why the terms have sometimes proven to be difficult to use, what is at stake in the discussion, what solutions have been proposed and, most importantly, the arguments and evidence behind them. When appropriate, I will discuss how these concepts have subsequently been adopted and adapted by scholars of Greek and Latin.
12
First, we will discuss the matter of code-switching. As will prove the case for borrowing, the definition commonly used in linguistics is fairly clear-cut and uniform. The term is commonly used (as it shall be by me) to refer to the alternation of two languages (or language variants11) within one and
the same piece of discourse.12 This means that code-switching is per definition restricted to speakers
that are to some degree bilingual; i.e. they have enough knowledge to switch to another code. Theoretically, then, code-switching requires the speaker to make the utterance based upon their own knowledge of a second language; the language user is not copying a fellow user’s usage of the item. As Matras (2009: 147) puts it, ‘one of the important criteria for distinguishing codeswitching from borrowing is the replication of the item by monolingual speakers, if there are any, and in monolingual contexts’. In practise, this can be determined quite clearly for some items (e.g. a code-switch between sentences), and less so for others (e.g. tag-switching).
The phenomenon is commonly divided into three variants, namely tag-switching, intersentential switching and intrasentential switching. This typology is adopted mainly to distinguish the (syntactical) bilingual fluency of speakers in using them; in the order mentioned above, they involve increasingly greater risks of violating rules of either language (Poplack 1980: 589; Romaine 1995: 122f.).
Tag-switching refers to the insertion of a fixed tag into another language; these tags include exclamations and interjections such as English I guess or you see. From a syntactic point of view, it is important to distinguish this type of code-switching from intersentential and intrasentential switching as tag-switches are ‘subject to minimal syntactic restrictions’ (Romaine 1995: 122).13 As such, such
code-switches require little bilingual fluency in terms of grammar. For instance: English/Tagalog tag
The proceedings went smoothly, ba?14
‘The proceedings went smoothly, didn’t they?’
11 This nuance, adopted in Hoffmann (1991: 110) and followed by Adams (2003: 19), is of minor interest for the present
research as all demonstrable cases of code-switching in the corpus of Mons Claudianus concern the switch between languages, i.c. Greek and Latin.
12 Similar definitions can be found in recent literature, for instance ‘bilingual speakers (…) alternate between the two
languages in the same discourse, sometimes even within the same sentence or the same word.’ (Haspelmath 2009: 40), ‘Codeswitching is the term that is normally applied to the alternation of languages within a conversation.’ (Matras 2009: 101).
13 Similarly, Poplack (1980: 589) refers to them as ‘freely moveable constituents which may be inserted almost
anywhere in the sentence without fear of violating any grammatical rule’.
13 Dutch/English tag
Alright, we zien wel wat we halen. ‘Alright, we’ll figure out what to get.’
Adams (2003: 21-3) in his monograph on Latin bilingualism adopts this category not only for these kinds of interjections in dialogue but also for the formulaic tags found in epitaphs. Whenever such tags are used, this is done virtually without any syntactic restriction. Adams cites examples such as IGUR 294, which is an inscription consisting of ten lines of Greek followed by the common Latin funerary tag dis manibus ‘to the gods of the underworld’ and a number of cases from Noy (1993) containing the Hebrew tag םוֹל ָשׁ šālôm ‘peace’ transliterated as Greek σαλώμ.15
Intersentential switching refers to the switch into a different code at the end of a clause or sentence, or between speaker turns in dialogue (Romaine 1995: 122). One might make at least two further distinctions, namely switches at the end of a clause versus the end of a sentence; and the difference between switches at the border of two main clauses or of a subordinate clause. Switches at the border of a clause pose a greater risk of breaking the grammar of either language than switches and at the end of a sentence; furthermore, switches at the border of subordinate clauses (such as indirect speech or relative clauses) pose an even greater threat than those at the border of two main clauses. As Adams (2003: 23) points out, tag-switching as defined above technically falls under intersentential switching. The term intersentential switching, however, is commonly reserved for syntactically embedded elements in contrast to the exclamations and interjections which require virtually no syntactic adaptation. A famous example of this is the quote from a Puerto Rican informant in Poplack (1980), illustrating intersentential switching at the border of two main clauses (y ‘and’ marking the clause border):
English/Spanish
Sometimes I’ll start a sentence in Spanish [sic] y terminó in español.
‘Sometimes I’ll start a sentence in Spanish and finish in Spanish.’16
15 Noy (1993) nos. 48, 70, 71, 72.
16 Poplack used this interesting quote as the title for her 1980 article and noted on the double mentioning of Spanish
(instead of expected ‘Sometimes I’ll start a sentence in English...’) that the ‘lapsus provides ground for interesting speculation, which we leave to the reader’ (Poplack 1980 616 n.10). The quote is often quoted in its adapted form (e.g. Romaine 1995: 123) and has even led to confusion in bibliographies of major works in the field (e.g. Wei and Moyer 2009: 390).
14
Adams (2003: 24) provides an interesting example in ILCV 4463: Greek/Latin
Βηρατιους Νικατορας / Λαζαρίῃ καὶ Ἰουλίῃ / καὶ Ὀνησίμῃ κον φιλιους / βενε μερεντες / ὁ βίος ταῦτα.17
‘Veratius Nicagoras (erected this monument) for Lazaria and Julia and Onesima with her sons, who well deserved it. Such is life.’
The final type of code-switching within this framework is the type that occurs within the clause and is referred to as intrasentential. As this type of code-switching often concerns the switch of a single lexeme, one of the main problems faced in the study of borrowing is its distinction from (intrasentential) code-switching. This problem is especially relevant for the present study, as we are dealing with single lexemes that originate from another language. Whenever one encounters a speaker using one single word that originates from another language, can we make a useful distinction between the two phenomena?
This problem has been discussed by many different scholars and has proven to be one of the most difficult aspects of defining the two terms. A number of different concepts have been related to this distinction and although most studies do not commit to any clear-cut division between the two, they help us think about the most important aspects related to these concepts, such as integration, distribution and the relationship to the language user. My aim here is to explore what is at stake in the discussion and what features of these phenomena are relevant to it in order to develop a terminology to examine the corpus of Mons Claudianus. In the end the distinction aims to determine whether a linguistic item is likely to induce language change (i.e. in the case of borrowing) or that we are dealing with a code-switch that does not incur language change but is a case of language behaviour instead (Haspelmath 2009: 40).
It should be noted that the core of the problem of the distinction lies in the underspecification of the definition of what a borrowing is. Some works do not even give a clear definition of borrowing, but most works that do adopt a definition along the lines of that of Matras (2009: 146): ‘the replication of a linguistic structure, of any kind, in a new, extended set of contexts, understood to be negotiated in a different “language”’. The problem when applying this definition to any corpus is determining
17 There is obviously more to this inscription than the intersentential code-switch (such as the spelling Νικατορας for
Nicagoras (possibly a confusion of capitals Τ and Γ?), the transliteration of Greek into Latin or the Latin preposition cum (κον) here governing the accusative case); for a discussion, see Adams (2003: 23f.).
15
whether a certain item has indeed entered the group’s native language or not. Since linguists are often dealing with such a corpus – as is the case with the present study – more concrete criteria are required. Referring to loanwords specifically, Haspelmath mentions that the distinction between a single-word code-switch and a loanword is easy on an abstract level: ‘if [a lexical item] is part of the mental lexicon of the [native language] of the speaker, it is a loanword, otherwise it is a single-word switch’ (2009: 40). The crucial notion is that in the case of borrowing it is the code that is being mixed (and therefore this phenomenon belongs explicitly to the realm of language change), while in that of code-switching, we are dealing with an utterance consisting of two separate codes. He notes the difficulty that the researcher is unable to have a direct look into this mental lexicon as mentioned above. Therefore the notion of a mental lexicon is perhaps useful at an abstract level of defining what a borrowing is, but it does not provide much help in identifying a particular phenomenon.
This has led to doubts in recent literature whether such a clear distinction is even feasible in all cases. Romaine claims that there are ‘no unambiguous criteria which will decide in all cases what type of language contact phenomena we are dealing with’. Haspelmath does offer some criteria which we will review below, but also mentions in his discussion that the researchers working on the Loanword Typology project the practical but rather arbitrary instruction that ‘[o]nly established, conventionalized loanwords that are felt to be part of the language should be given, not nonce borrowings. This distinction is often hard to make (especially when there are no monolingual speakers), but authors should try as best they can.’ (2009: 41n.7). Matras (2009: 110), too, acknowledges the difficulty and proposes a continuum between a code-switching and borrowing, with similar criteria as Haspelmath (2009: 40-2) gives; in absence of an easy distinction between the phenomena, a continuum appears to be the best approach.
Let us review what the properties are of borrowed items and code-switches and on what end of the continuum they reside. We shall start with properties that are crystal clear in their outset, but that prove to be more difficult to test within a corpus of utterances; we will move on to those that are easier to identify in an utterance or in a corpus of utterances, but that, when identified, do not give exclusive evidence to the item being either a borrowing or a code-switch.
The most evident case of an item being a borrowing is by verifying whether it is being used by a monolingual. The ‘bilinguality continuum’, as Matras (2009: 111) coins it, draws on the key difference in availability of borrowing and code-switching. Code-switching, implying the usage of two separate codes per definition, requires a speaker to master two codes – thus only bilinguals can show this language behaviour. A borrowing, however, is available to both bilinguals and monolinguals as it has become part of the language a monolingual masters. English marriage, one of the many Medieval loanwords from French, is mastered by monolingual speakers of English, which by itself is reason
16
enough to identify it as a borrowing. Its ubiquitous usage by monolingual speakers of English attests to its place in the English language. However, the difficulty with this argument when dealing with a corpus is that it is hard to exclude the possibility that the speaker or writer was monolingual or bilingual at the time of the recording the utterance. A further problem is posed by the fact that we cannot blindly accept that the majority of any given population will be monolingual as some sort of standard in society.18 Unless we have explicit evidence that a speaker is monolingual, this criterion cannot be used
for determining whether an item is a borrowing or not.
The second key criterion is linked to the dichotomy between innovation and propagation. This notion is crucial to any approach of language change, but, according to Croft (2002: 5), is often overlooked. When dealing with two variants in any aspect of language the selection of a particular variant can be described as an innovation of at any particular point in time, corresponding in our case with a code-switch. This is an inherently different process from propagation: this signifies the acceptance of a variant across a speech community (including monolinguals). In this respect it is important to appreciate Myers-Scotton’s proposition of code-switching as ‘the gate for core borrowing’ (1993: 174-6).19 In this sense, some of the criteria on the continuum of borrowing and code-switching
we are discussing (such as frequency and adaptation, discussed below), can be viewed as a diachronic continuum of code-switches becoming borrowings.20 When addressing this aspect, however, we must
account for the restricted nature of our corpus and consider whether it is large and diverse enough so that if the innovation would be propagated, it would show up upon investigation; and even if it does not, it is dangerous to argue for a form to be an occasion of code-switching e silentio.
The notion of innovation and propagation is expanded upon by Matras (2009: 112) who notes the ‘special conversational effect’, more likely in the case of code-switches, versus borrowings being often the ‘default expression for the relevant concept’. This is, according to Haspelmath (2009: 40) a crucial notion in the discussion: whereas borrowing implies language change, code-switching is a form of language behaviour that spawns from language contact. There are exceptions to this tendency of borrowings being the default expression in a speaker’s language. We should pay special attention to core loans in this respect. As these loans exist alongside an older, native expression, their use implies a choice on the part of the speaker, induced by functional social factors. Matras (2009: 112) mentions for instance German Fernsprecher as a formal lexeme ‘normally reserved for official or institutional,
18 Note Thomason’s remark that ‘bilingualism is much more widespread than monolingual citizens of countries
traditionally dominated by English (…) tend to believe. The idea that monolingualism is the human norm is a myth.’ (2001: 31).
19 Note the specification of core borrowing (as opposed to cultural borrowing). On this distinction, see below. 20 As Gardner-Chloros (1987: 102) puts it: ‘[i]n short, a loan is a code-switch with a full-time job.’
17
written style, such as letterheads or business cards’, constrasting with Telefon, common in daily conversation. Incidentally we should be cautious not to confuse language behaviour and language change in the case of a loan: if a native equivalent for a core loan is available, speakers choose to use a loan or not after the language change itself has already occurred.
Essential, too, in this discussion is the distinction between cultural loans and core loans.21 As
pointed out by Myers-Scotton (1993: 168-76), these two different types of borrowings undergo different processes and must therefore be distinguished. A cultural borrowing is a loanword that is introduced into a language alongside a particular novelty, such as technological innovations (German
das Motherboard from English) or novel concepts (English apartheid from Afrikaans). A core
borrowing, in contrast, is a loanword that refers to a concept for which the recipient language already has a word in its lexicon (Dutch kids from English instead of native kinderen).22 The process of
borrowing is different in two ways, namely in motivation and propagation. The motivation for a cultural loan is obvious: there is a gap in the lexicon to name a particular novelty. In the case of core loans, given that a ‘viable equivalent’ (Myers-Scotton 1993: 169) exists, researches often point to the prestige of the donor language that motivates speakers to adopt the loanword (Matras 2009: 149f.). A problem with the term core borrowing or loan is the possible confusion with the Swadeshian concept of core vocabulary, i.e. a universal (and controversial) list of signifiés that are rarely borrowed in any language (and therefore useful for historical-comparative linguistics to determine affinity between languages), containing ‘basic’ signifiés such as I, eye, water, black (Swadesh 1955). As has often been pointed out, Swadesh’s list, though insightful, has ascribed greater truth value than it has or than Swadesh claimed it to have;23 as Haspelmath (2009: 36) summarises, Swadesh complied his list ‘on the basis of his
personal anecdotal knowledge and intuition, not on the basis of systematic cross-linguistic research’.24
21 Field (2002: 9) uses the terms additions and substitutions (for cultural and core loans respectively) instead. This
terminology seems to be less favoured; Haspelmath (2009) and Matras (2009), to name a few, prefer the core/cultural terminology. Moreover, the terms addition and substitution do not seem to cover the distinction cleanly: addition does not exclude the possibility that a viable equivalent already exists in the language and substitution implies that the older form in the recipient language is extracted while usually (and certainly in the earlier stages of a loan’s existence) both forms coexist.
22 Myers-Scotton (1993: 169n.5) poses the issue of ‘what, in fact, is a “viable equivalent”’ without specifying how this
issue may be resolved.
23 Note his careful article title ‘Towards greater accuracy in lexicostatic dating’.
24 Haspelmath (2009: 36) In practical terms, linguists often work with Swadesh’s (1955) list of non-cultural vocabulary,
which were intended by their author to be his best guesses as to which words are resistant to borrowing. But this list was drawn up by Swadesh on the basis of his personal anecdotal knowledge and intuition, not on the basis of systematic cross-linguistic research. The Loanword Typology project represents some of the research that would have been a prerequisite for Swadesh’s word-list-based historical-comparative linguistics.
18
In language contact, we generally refer to a lexical item as a core borrowing when the recipient language already has a lexeme to refer to the same signifié the borrowing refers to.
A third criterion commonly cited in distinguishing code-switching from borrowing is the degree of integration of the lexical item in question. In theory, a code-switch implies that there is a ‘full-blown switch from one language to another’ (Adams 2003: 19); this would imply that all aspects of the other code are switched to – not merely vocabulary, but also morphology and phonology. A single item code-switch, then, would retain all of its aspects from its language of origin and therefore show no integration into the other language of any kind. There are several issues we have to be aware of when relying on morphology or phonology as identifiers for the type of language contact phenomenon we are dealing with. Though integration offers extremely useful information for the study of language contact it is by no means a litmus test for its classification as a borrowing or a code-switch. Poplack, Sankoff and Miller (1988: 52) bring up the issue of items that originate from another language, are phonologically and morphologically integrated yet are not ‘well-established’. From their corpus of utterances by Canadian speakers they mention the following examples:
French/English
Je serais pas capable de coper [kɔˈpe] avec.
‘I couldn’t cope with it.’
Il est pas capable de firer [faiˈʁe] ses curés.
‘He can’t fire his priests.’
Coper and firer are integrated phonologically (French voiced uvular fricative [ʁ] in firer; contrast, too,
the vowels with Am. English [koʊp] and [ˈfaɪər] respectively; stress on last syllable) and morphologically (French infinitive marker -er [e]). These cases are problematic if we want to enforce a clear-cut distinction between borrowings and code-switching: they are not propagated through the speech community yet are integrated both morphologically and phonologically. Poplack, Sankoff and Miller (1988: 50) distinguish these items from the onset as ‘nonce borrowings’ since they resemble loans in their integration but are specified as being used only once. This separate category of borrowings, however, has been met with much criticism as it goes against the very definition of what borrowing is per definition: a foreign word that has found propagation throughout a speech community. The concept was criticised already by Myers-Scotton (1993: 163-207, esp. 181f.) and reiterated by Haspelmath (2009: 41), who claims that cases such as the Canadian examples above should be considered code-switches instead of ‘nonce borrowings’.
19
This conclusion, however, does not stand well with the definition of code-switching. If the speaker truly switches codes, why does he only switch in the lexicon and not in the adherent morphology or phonology? Haspelmath (2009: 41) claims that this phenomenon has to do with a ‘foreign accent’ more than anything; we are treading here into the realm of interference as explanation for the adaptation of the item. We should then actually not speak of a ‘nonce integration’ but rather of interference in the case of such a partial code-switch. Just as a loan and a code-switch share a diachronic notion (‘a loan is a code-switch with a full-time job’), we may see phonological integration of a loan, then, as the integration of interference phenomena.
We have seen that distinguishing between a borrowing and a code-switch is not always an easy task. Single lexemes that originate from a different language, which form a major part of the present study, show various aspects of either phenomenon that sometimes appear contradictory. Rather than making a clear distinction between loanwords and single lexeme code-switches,25 it seems more
sensible to adopt a continuum between the two as proposed by Matras (2009). On this continuum we judge single lexemes originating from another language based on (1) their exclusive use by bilinguals or integration in the lexicon of monolinguals, (2) their degree and nature of propagation into the speech community and (3) their morphological and phonological integration. This continuum can be used not only to distinguish the results of either process (language change in the case of borrowing and language behaviour in that of switching) can also serve as a diachronic sketch of how regular use of code-switches can result in the borrowing of the item in question. Considering our often fragmentary evidence that is more often than not unclear about the bilinguality of its author, it is still hard to assess these criteria effectively; we should consider that Haspelmath gave a warning to his researchers who were working on living languages26 and could map the bilinguality of any particular speaker in greater
detail than we will ever be able to reconstruct for the speakers of ancient languages.
Whenever a lexical item is borrowed from one system into another, it is snatched away from its original phonological context and inserted in another (in our case Latin and Greek respectively). When borrowed, a speakers sometimes retain the phonological shape of a word, but often it is adapted phonologically (Matras 2009: 148). Many languages tend to have overlap in some of their phonemes, in which case the phonological aspect of borrowing is straightforward and no change in pronunciation occurs. In the case of most phonemes, however, language systems differ and speakers can either (try
25 As Adams (2003: 19) does, who does discuss the problems involved but who adopts the principle of ‘generally
us[ing] the term [code-switching] to describe a full-blown switch from one language into another within one person’s utterance or piece of writing’.
20
to) reproduce replace the foreign phoneme of to replace the foreign phoneme with a native one; usually the native phoneme that substitutes the original phoneme approximates it in its properties (such as tongue position and roundedness in the case of vowels or phonation and manner of articulation in the case of consonants; Winford 2003: 46-48). We can find both strategies in Dutch garage [xaˈraʒə] borrowed from French garage [ga'ʁaʒ]. The phoneme [g] is foreign to Dutch phonology and is replaced with the native [x];27 foreign [ʒ], however, is reproduced in Dutch intervocalic position
Aside from phonological integration the second major form of integration is morphological integration. Words in different languages have different properties. Some languages mark gender in the morphology of words, some do not; some languages have only two genders, some have many. Thus we find English, marking no gender except in pronouns, French with two genders (masculine and feminine) and Latin, Greek and German with three genders (masculine, feminine and neuter). Sometimes phonological similarity is key to assigning a morphological class. Winford (2003: 48) gives the example of Swahili with its fifteen morphological noun classes: whenever a word is borrowed it is to be assigned in one of these classes. Each of these classes has distinct prefixes and there is a tendency in Swahili to assign foreign words into one of these classes on the basis of formal features: thus Arabic
kitab is assigned to class 7-8 because of the formal similarity to singular prefix ki- in for example native kiti ‘chair’.
In general, however, ‘borrowed words are treated like native stems of equivalent categorical status, and take the bound morphology and other properties appropriate to the class they are assigned to’ (Winford 2003: 48). This enables speakers to use the newly acquired words according to their own, native morphological system.
Not only lexical items can be borrowed among languages. As Thomason and Kaufman (1988:74-95) point out in their framework for linguistic borrowing, lexical borrowing is only a sign of ‘casual contact’. Further up the scale we find on the third level (of five, ranging from ‘casual contact’ to ‘very strong cultural pressure’) that ‘derivational affixes may be abstracted from borrowed words and added to native vocabulary’(Thomason and Kaufman 1988: 74). A prime example of this is the Romance -(a)ble suffix in English, derived from French -(a)ble (Winford 2003: 57). In Middle English a great wealth of borrowings from French was introduced into the English language. Among these loans was a subset that used the French suffix -(a)ble, such as credible, audible, legible, treatable, etc. This formed the first step in the borrowing of the derivational suffix. The second step was the
27 But contrast (more recent?) Dutch borrowings from English where word-initial [g] is retained, e.g. goal, game, gadget,
21
recognition of the -(a)ble adjective suffix as a separate morpheme denoting the possibility of an action being performed upon the noun it describes. It became productive, being added to other roots than those borrowed in the first step. This step yielded a plethora of derivational blends, combining native roots with the newly acquired suffix, such as lovable (< Germanic verb to love), doable (< Germanic verb to do), etc., thus adding new words to the lexicon.
22
C
HAPTER
3
Language contact phenomena at Mons Claudianus
In this chapter, the core of this study, I will provide an analysis of the actual ostraca. I will address four language contact phenomena in particular: language choice (3.1); lexical borrowing (3.2); morphological borrowing (3.3); the integration of lexical borrowings (3.4); and bilingual interference (3.5). As mentioned earlier in 1.3, the overwhelming majority of the ostraca are in Greek; therefore, the four latter aspects will be addressed in particular with respect to Greek. The rareness of the Latin material (only twelve ostraca are in Latin; one (788) contains both languages), for example, does not allow for a systematic and meaningful analysis of borrowing.
3.1. Language choice
In this section I will address the question of language choice and investigate the scarce evidence the ostraca yield on this point. It must be clear from the onset that Latin is found rarely in our material; of the 896 published ostraca, only twelve documents are in Latin and one is bilingual.28 Its scarcity
28 In this chapter, I refer to the Mons Claudianus ostraca through their number in the four Mons Claudianus. Ostraca
Graeca et Latina volumes (Bingen et al. 1992: O.Claud. 1-190; Bingen et al. 1997: O.Claud. 191-416; Cuvigny 2000: O.Claud. 417-631; Bülow-Jacobsen 2009: O.Claud. 632-896). Throughout the chapter, I will not write O.Claud. before the document number (so 788 = O.Claud. 788) unless confusion might arise (e.g. when Mons Claudianus ostraca are compared with other papyrus or ostracon collections).
23
may be read as either a sign of insignificance as its use was limited and only used by a minority; we must, however, consider the possibility that through its limited use, its use may appear all the more significant. As Adams (2003: 529) puts it, ‘although Latin statistically is insignificant, its importance sociolinguistically should not be underestimated’, being the language of Roman rule and in the light of possible H-L diglossia a ‘super-high political language’ (2003: 637). An example of this special position of Latin we find at the Colossus of Memmon. At this ancient site visitors came to behold and hear the supernatural Colossus ‘speak’; they could then leave an inscription remembering their visit, leaving their name and often a brief message such as ‘I, (…) heard the voice of the Colossus’. Most officials visiting the site and leaving such a message do so in Greek; however, Roman military officials sometimes choose to use Latin instead. This use of Latin is probably to stress the ‘Romanness’ and subsequent authority of the author of the message. Although Greek is frequent in inscriptions even from high-ranking officials, Latin can, according to Adams (2003: 546-55) function as a ‘super-high’ language when Roman military officials want to stress their Roman authority in the province.
In the Mons Claudianus ostraca, we are dealing with wholly different documents, with a much more mundane use than the Memmon inscriptions. There are twelve documents fully in Latin (2, 131, 135, 304, 305, 306, 308, 367, 843, 845, 846 and 847) and one containing both languages (788). As these Latin texts are least frequent in the corpus, their rarity makes their authors’ choice marked. If we want to find the motivation for the language choice between Latin or Greek for writing a certain ostracon, it would probably become most apparent in Latin texts, as they are the marked choice within the overwhelmingly Greek corpus.
As we discuss the Latin texts from the Mons Claudianus corpus, we should take into consideration the possibility that the author was a Latin monolingual. The language of the document then does not reflect choice but rather necessity. Adams (2003: 589) labels this ‘linguistic competence as a determinant of language choice’ but in the case of a monolingual author one cannot speak of any choice – the only choice he made was not to make the effort to get an interpreter to translate if he would deem this necessary.
Thus there are multiple motivations for the letter of Antistius Flaccus to Calinius (2) being in Latin. The editor does not mention any graphic interference which might reveal bilingualism; on the contrary, ‘l’écriture [est] eliée et posée, sinon élégante’; sometimes long /o:/ is marked with an apex (<ó>). Antistius Flaccus (or his scribe) uses the Greek loanword ydreuma (= ὕδρευμα, probably by loss of inital aspiration in (spoken) Latin29), but as there are no signs of the item being a code-switch it may
29 Väänänen (1963: 57); Adams (2013: 125-7). Bingen ad loc. proposes too (and primarily) the lack of aspiration in
Greek writing (as breathings are usually not written in papyri) as explanation for the spelling in this Latin document; this theory assumes borrowing through writing rather than speech and seems less likely.
24
simply be loan-word, not proving whether Antistius was either mono- or bilingual. We can therefore not exclude that either Antistius Flaccus or Calinius (or both) were Latin monolinguals in which case the message could only be transferred in Latin. Antistius Flaccus probably was a decurion.30 The name
Calinius is attested only once31 and may have been a decurion too; however, due to the lack of evidence,
we cannot be sure we are dealing with the same Calinius here. The military status of author and recipient, however, does not imply an obvious choice for Latin. Adams (2003: 599ff.) rightfully points out that it is a ‘persistent misconception’ that Latin was the official language of the army, citing a handful of scholars that make such assumptions, and that these authors do not always make clear what they mean with an ‘official language’. He defines it as the language in use for settings that are ‘not personal, but related to the operation of the army as an institution’ (2003: 600). In line with his observations, much of the military correspondence at Mons Claudianus (in documents which fit Adams’s definition of an ‘official setting’) is in Greek: 357-387, a series of military ostraca edited in the same chapter, are all in Greek save for 367, a Latin document which deserves discussion below. To conclude, the only discernible difference is that 2, from Antistius Flaccus to Calinius, might concern the conversation between two officers (if Calinius was indeed a decurion) in which case a choice for Latin might reflect their shared ‘Romanness’ in the context of a Greek-speaking province.
An interesting case where bilingualism is attested for the addressee we find in the archive of Successus. Successus probably was ‘a keeper of materials for the quarryworks’ (Bülow-Jacobsen 1992: 111) and because of the rarity of his name, there is little reason to doubt that the twelve documents in this archive (124-136) pertain to the same person. The identity of the addressee is significant because of the twelve documents addressed to him,32 ten are in Greek and two in Latin. Successus thus is a
perfect example of a Greek-Latin bilingual at Mons Claudianus who was able to do his job regardless of whether people addressed him in Greek or Latin.
Unfortunately, there are no clues as to what motivated the authors to write to Successus in either Greek or Latin. The Latin letters do not appear to pertain to a particular topic that would require the use of the language.33 131 is a request for zeucterias (= ζευκτηρίας, ‘yoke straps’)34; since the author
30 Bingen ad loc. mentions other attestations of this name where it is accompanied by either δεκουρίων or the usual
mark in Greek of an iota written above chi
31 Mons Claudianus inv. 3476 (unpublished but mentioned by Bingen ad loc.). 32 O.Claud 136 is written by Successus himself and is in Greek.
33 As Adams (2003: 591) too concludes.
34 The loanword is unattested and one might argue that it might even be more convenient to simply write the note in
25
probably knew Successus was a bilingual, he would assume that he would have no trouble understanding what he meant writing zeucteria. 135 concerns the delivery of three tunics (tunicas duas and a tunic(am) albam) and another cloak (palliol(um)); the topic, viz. Roman clothing, may be part of the motivation to write a Latin note instead of a Greek one.
Instead, we must probably assume that it was the proficiency in either Greek or Latin that determined the choice of language for a particular letter to Successus; he himself was assumed to handle either language. It is worth noting how these documents remind us of how the origin of names of language users does not necessarily indicate their nationality or language proficiency: the Greek notes in this archive are sometimes written by authors with Roman names such as Ponticus (124), Rusticus (133), Domitius (134) etc., where the two Latin notes are written by men with a Greek name – Athenor in 131 and Agathon in 135.35
Another interesting case of language choice is that of documents 366 and 367.36 The letters
appear to originate from the same author (Teres, curator) and to be sent to the same addressee (Anius Rogatius, duplicarius). In this case, too, it is difficult to reconstruct any motivation for Teres’ language choice. Note that in both letters the two names are identical (albeit the one in Greek, the other in Latin), but there are small differences in titles of the two men: in 366 Anius is called a duplicarius and Anius identifies himself as the curator of Raïma (a nearby praesidium); in 367 Anius bears the name Rogatus as well and Teres simply calls himself curator without stating his provenance. Apart from the names Van Rengen ad loc. points to similarities in style: Teres writes ταχύτερον in 366 and celerius 367 (‘rather quickly’) when stating his demands and uses two final clauses within one sentence both in Greek (εἵνα … εἵνα, ut … ut) embedded in each other. Van Rengen claims that there can be little doubt that Latin was Teres’s mother tongue but does not provide any grounds for his argument. One could argue that the spelling of the Latin is better than that of the Greek, but that probably simply has to do with the spelling capacities of the scribe. I have little doubt that both letters, one in Greek, one in Latin, have the same originator and addressee; there are, however, no clues as to why Teres chose for Greek in 366 and Latin in 367.
The final letter that bears Latin writing is 788; it is in fact the only bilingual document from the corpus. It concerns two requests for the delivery of tools; the document starts out in Greek and swaps to Latin about halfway through. Because of the significance and problems of the document, I will print it here in full:
35 For these points, see too Adams (2003: 591).
36 Note that there are two more (Greek) documents from Teres found at Mons Claudianus, listed as inv. 1226 and 2755;
26
Μᾶρκος (δεκουρίων) Σαβίνωι χ(αίρειν). δὸς Κρο- νίων<ι> πασσάλους β̅ σ̣ι̣δ̣η̣ροῦς καὶ
πέντε ξύλα εἰς κλίμ̣α̣κ̣α καὶ
σχυνία λ̣ε̣π̣τά. (hand 2) C̣ụbiṇo salut(em) 5 cados mit<t>e
p ̣alaẹaṣ V tenues uale
(first hand, in Greek) ‘Marcus, decurio, greets Sabinus. Give Cronion 2 iron pegs and five pieces of
wood for a ladder and thin ropes. (in a second hand, in Latin) Greetings to Sabinus(/Cubinus?). Send five old thin vessels(?). Take care.’
The document is of particular interest to us because it is our single most tangible piece of evidence for bilingualism at Mons Claudianus. The first request (in Greek) is a request from a certain Marcus to Sabinus. Bülow-Jacobsen (2007: 123) expects the decurio Marcus to be the same as the decurions attested in 365 as Μᾶρκος Κανίνιος and in 865 as Μᾶρκος; both of these documents are in Greek. Sabinus is known from ostraca 121 (a request to Sabinus), 126 (where Grata asks Successus to greet Sabinus) and 151 (a request from Sabinus to Zosimus to watch some of his παιδάρια37) and 786
(a request to Sabinus for water). Apparently Sabinus was an official at Mons Claudianus responsible for the distribution of goods. Bülow-Jacobsen (2009: 123) mentions that ‘the temptation to read Sabino instead of Cubino is high’; I think there can be little doubt that C̣ụbiṇo, the addressee, is the same Sabinus. Not only is *Cubinus an unattested name, the sigma of Sabinus’s name is written as a lunate sigma on the ostracon,38 which is quite reminiscent of Latin capital <C>. I suspect that the scribe of
the second half tried to copy the name from the Greek message and wrote <C> for the lunate sigma; alternatively, as Bülow-Jacobsen (2007: 123) mentions, ‘a lunate s in a Latin text would not be totally without parallel’. The <V> for <A> is harder to explain. Bülow-Jacobsen suggests that the scribe may have ‘simply turn[ed] the ‘A’ upside down’; we should also note the uncertainty of the reading by the editor (although the <V> on the photograph supplied in Bülow-Jacobsen (2009: 325) seems clear enough). I have little doubt that we are dealing with a scribal error and that the two addressees are the one and the same Sabinus.
37 Possibily some of his personal slaves, but the context of the document is obscure. 38<ϹΑΒΙΝΩΙ>
27
Whoever Marcus may be, it is all but certain that he is also the one responsible for the second half of our document in Latin. One could argue that he ordered both messages to be written, since they are written on the same ostracon and since there is no author mentioned in the second, Latin half of the text, which might imply that Marcus simply added an afterthought, written by another scribe.
The usage of p̣alaẹaṣ is also peculiar. No doubt it is a borrowing of Greek παλαιός, but if the text is complete, there are difficulties. Firstly, it does not agree with cados in gender; secondly, it is obscure what the sense of the order of a couple of old water jugs would be – perhaps the intention is recycling like the ἀσκὸν παλαι<ό>ν (‘old water skin’) used as greaves for the masons? Finally, why would he not use a Latin equivalent such as vetus?
It remains unclear why the second half of the document is in Latin; the contents surely do not seem to call for the switch. The reason may be as simple as the Greek scribe being on a break and a scribe who knew Latin wrote a brief afterthought on the ostracon – we cannot be certain. It does seem very probable, though, that Sabinus was the recipient of both messages (not in the least because they are written on the same ostracon) and that he was, like Successus, a bilingual official at Mons Claudianus who used both Greek and Latin in his communication.
Sabinus’s story is quite exemplary for the situation at Mons Claudianus on the whole. Given the overall ratio of Greek documents to Latin ones (ca. 100: 1) published thus far; of course, this does not necessarily correspond to overall language use at Mons Claudianus, but Greek seems to have been the most common language in both official and private documents. We find various kinds of documents in both Latin and Greek – lists and letters alike – and, likewise, various topics are discussed. We find no correlation between the names of the authors or recipients and the language they use; on the contrary, in the letters to Successus, we have found authors with Latin names writing Greek messages and vice versa.
3.2 The semantic areas of lexical borrowing
In this section I will discuss the plethora of loanwords we encounter in the ostraca of Mons Claudianus and indicate the semantic fields which they occupy. We mostly find lexical borrowings from Latin into Greek; however, we find a remarkable amount of Greek words in Latin documents. Due to the small amount of Latin documents, we cannot perform the same extensive research on the Greek loanwords or code-switches appearing in Latin texts, but we can show some tendencies for this direction of borrowing. The majority, however, will be dedicated to showing how the Latin language influenced the Greek at Mons Claudianus. I will consider what lexical domains show high frequencies of loanwords. In principle I will treat all terms originating from Latin which are not obvious transcriptions (such as the passwords in 309-332) and thus score extremely low on integration, but with
28
discussion of how they qualify on the borrowing continuum. To this end I will apply the three criteria mentioned in chapter 2: bilinguality, propagation and integration. Of these criteria bilinguality is nigh impossible to confirm: the only proof for a word being a borrowing on the basis of this criterion is by its use by a monolingual. Monolingualism of an individual is, however, extremely difficult to verify/prove?, and in our cases usually impossible. Propagation is easier to attest as a criterion of position of an item on the borrowing scale. I shall review both how frequent an item recurs and in what kind of documents; for this latter test I will make the distinction between lists, reflecting a highly formulaic bureaucratic context, and letters, which are by nature freer in composition and content. Adaptation is treated systematically in section 3.3 but will of course be alluded to here to confirm to what extent the word behaves as a Greek word by the language user and thus ‘scores high’ on the borrowing continuum.
Let us first consider the Greek loans in Latin. The following words are attested in Latin documents; they are given here with their attestation and are not added to Appendix I:
ὕδρευμα ‘well’ ydreuma (acc. sg.) 2
*ζευκτηρία ‘yoke-strap’ zeucterias (acc. pl.) 131
ἄγγαρος ‘courier ‘ ang(arius) 304
παλαιός ‘old’ p̣alaẹaṣ (acc. pl. f.) 788
λατομία ‘quarry’ lat(omia) (abl.; thrice) 845
latomia (abl.) 846
ν̅(όμαιος (sc. λίθος)) ‘a regular block of stone’ n̄(omaei) 847
Let us first consider the relatively high number of borrowings for the small sample of ostraca: within the thirteen documents that contain Latin (twelve solely in Latin, one Greek-Latin bilingual) seven different borrowings appear. Within our (admittedly small) sample of Latin texts there appears to be a fairly large number of Greek loansword. Some of these are not attested elsewhere (*zeucteria,
*palaeus, *n(omaeus)) yet most items show phonological and morphological integration where it
might be expected. Thus latomia receives the unmistakenly Latin ablative case in 846 (ex latomia
Sozuza ‘from the quarry Sozusa’) and Greek initial aspiration in ὕδρευμα is dropped according to the
tendency in either Greek or Latin.39
39 Bingens (1997) alternative suggestion ad loc. that the Greek spelling of ὕδρευμα in papyri as <υδρευμα> without any
sign for the aspiration seems far-fetched to me: Greek aspiration is very commonly rendered <h> when borrowed into Latin; furthermore, Bingens explanation assumes prominence of the written form <υδρευμα> over the spoken [ˈidrevma], which is debatable. Aspiration was from the early Empire onward dropped quite often in both Latin
29
The loanwords mainly concern the domain of technical objects and instruments in use at the quarry. Thus we find the hydreuma ‘well’, *zeucteriae ‘yoke-straps’ and cadi ‘jugs’. The usage of these loanwords probably reflects the common use of the Greek words by the work-force (arguably excluding the use of cadus, which, unlike hydreuma and *zeucteria, is a quite common loan in literary texts,40
albeit unattested in documentary texts). Three other terms appear to come from Greek administrative terminology. Cuvigny argues to read the abbreviation ang as ang(arius) from Greek ἄγγαρος, attested only in 304, a roster of workers. Equally, latomia (in 845 abbreviated as lat) is obviously borrowed as an administrative term from lists such as 839, 840 and 841 where λατομίᾳ is written (followed by the name of the quarry, e.g. λατ(ομίᾳ) Μάρωνος (‘in the Maron quarry), and a number), to designate where a particular number of blocks are located. Very interesting among these cases of ‘administrative borrowing’ is n̄ followed by the letter p41
and a number.42 Bülow-Jacobsen points to the parallel of ν̅,
extremely common in accounts of stone, most likely an abbreviation for ν(όμαιος λίθος), or a ‘regular sized stone’43 – for example a stone that was large enough to form one of the segments of a pillar, which
was one of the most common usages of Mons Claudianus granodiorite.
It is quite unfortunate that we have no other attestations of this use of n̄ as 847 is the only document of its kind in Latin in our corpus (though other such lists may remain among the ca. 8000 unpublished ostraca of Mons Claudianus). Yet although the document is quite damaged, n̄ is legible without difficulty and the parallel with ν̄ appears to be the only viable explanation for the otherwise unknown n̄. As such it can be easily explained within the common borrowing of administrative terminology in other documents, both Greek and Latin; it would be unique, however, in our corpus as a borrowing of a mere abbreviation (perhaps comparable with English A.M., P.S., R.I.P.,44 borrowed
abbreviations where often speakers may not know the exact origins but are of course still able to use the abbreviation without any problem).
Due to the small number of Latin texts it would be unwise to make any definitive statements about Greek borrowings in Latin at Mons Claudianus, but we can certainly see that Latin speakers at
(Allen 1965: 43-5; Väänänen 1963: 57; Herman 2000 [1967]: 38) and (koine) Greek (Allen 1974: 52-6; Gignac 1976: 133-8).
40 E.g. Plaut. Am. 427, Verg. Aen. 1.195.
41 Bülow-Jacobsen ad loc. notes the letter P regularly found as a quarry-mark on the stones in the quarries. 42 Here in 847 the number seven, probably denoting the amount of stone.
43 Bülow-Jacobsen ad 841.
44 These last examples perhaps being most apt for their highly formulaic usage in letters and epitaphs respectively, such
