The Language of the Vindolanda Writing Tablets: An Interim Report

by J. N. Adams
Citation
Title:
The Language of the Vindolanda Writing Tablets: An Interim Report
Author:
J. N. Adams
Year: 
1995
Publication: 
The Journal of Roman Studies
Volume: 
85
Issue: 
Start Page: 
86
End Page: 
134
Publisher: 
Language: 
English
URL: 
Select license: 
Select License
DOI: 
PMID: 
ISSN: 
Abstract:

THE LANGUAGE OF THE VINDOLANDA WRITING TABLETS:
AN INTERIM REPORT

By J. N. ADAMS

I. INTRODUCTION

The recent publication by A. K. Bowman and J. D. Thomas of The Vindolanda Writing Tablets (Tabulae Vindolandenses 11) (1994)~ provides another substantial corpus of Latin from a military outpost in the early Empire. The tablets take their place alongside such important military finds as the letters of Claudius Terentianu~,~

which are roughly of the same date, the ostraca from Bu Njem,3 and the ostraca from W2di Fa~ikhir,~

which again are dated to the firstisecond centuries.

The Latin of the Vindolanda tablets has recently been discussed by H. Petersmann as a specimen of 'Vulgar Latin', at a conference devoted to Vulgar and Late Latin.s But while the influence of spoken varieties of the language can be detected in some misspellings, and in aspects of the syntax, morphology, and lexicon of the tablets, one must resist the temptation to find 'Vulgar Latin' (however one defines that problematical term: see below, 1x.4) as the sole or principal element of the tabletse6 Many of the documents were not composed by free composition, but have a formulaic structure which made little or no demand on the linguistic competence of the writer (e.g, applications for leave (166-77), the daily reports of a type found at Bu Njem, which have certain distinctive features of syntax (155-6)). Accounts and lists (178-209) too may in their syntax and format, if not necessarily in their spellings, be heavily influenced by the conventions of their genre. Moreover record-keeping of this type usually falls to individuals with a degree of education and numeracy, and their writing may have little or nothing to reveal about the spoken language which they used or heard.

Scribes were responsible for the writing of many of the Vindolanda tabletse7 Since they are likely to have received some training in spelling from gramwzatici, they place a further barrier between the Latin spoken in the area of Vindolanda, and the modern reader attempting to identify colloquial patterns in written tablets. In this paper I will certainly be concerned with colloquial or 'vulgar' elements in the tablets, but these are only part of the story. The Vindolanda tablets deserve to be assessed as specimens of writing effected in a military setting in an outpost of the Empire. They may have something to tell us about, for example, standards of literacy within the garrison and its entourage, the training of scribes, and the influence of genre on syntax and lexical choice. The tablets should not be considered in isolation; we are in the position of being able to compare writing at Vindolanda with that in military environments in other parts of the Empire. A comparison of the orthography of the Vindolanda tablets with that of the letters of Claudius Terentianus and of the Bu Njem ostraca yields interesting information about the cultural level of the scribes operating at Vindolanda. I will be

I call this paper an 'interim report', because tablets The editors have found abundant evidence for the continue to be found at Vindolanda. See A. R. Birley, activities of scribes at Vindolanda. Tablet 234, for exam- 'Four new writing-tablets from Vindolanda', ZPE IOO ple, contains a dictation error (see the editors, 42). Some (1994)~ 431-46. Others have been read but are still letters are written by a first hand, with an appended final unpublished, for example Inv. 9311544, which contains greeting in a different hand, almost certainly of the author the address 'Ceriali regi suo' (with which compare Clau- (e.g. the letters of Severa, 291, 292, 293: see Bowman and dius Tiberianus, P. ,With. VIII.472, 'Longino Prisco Thomas, 256; see also the letter of Chrauttius, 310, with domin[o] et regi suo') . the comments of Bowman and Thomas, 289-90). A letter See H. C. Youtie and J. G. Winter, Papyn' and of Severa, wife of Brocchus, no. 292, is in the same hand as Ostraca from Karanis, Michigan Papyri VIII (1951). that of Brocchus, no. 246; both were no doubt written by See R. Marichal, Les ostraca de Bu ~Vjem, SupplC-the same scribe, associated with the household (Bowman ments de 'Libya Antiqua' VII (1992). and Thomas, 260). By contrast another letter of Severa See 0. GuCraud, 'Ostraca grecs et latins de 1' W2di (291) is in a different hand, which is however probably FawPkhir', BIFAO 49 (1942;), 141-96. also found in 243, 244, and 248, all letters by Brocchus

See H. Petersmann, Zu den neuen Vulgarlatei-(248 by Brocchus and Niger) (see Bowman and Thomas, nischen Sprachdenkmalern aus dem romischen Bri-256). The household of Brocchus made use of at least tannien. Die Tafelchen von Vindolanda', in M. Iliescu three different scribes. Similarly Bowman and Thomas and W. Marxgut (eds), Latin vulgaire -latin tardifIII. (199) tentatively identify the hand of 225-232 as that of Actes du III2me colloque international sur le latin vulgaire the prefect Cerialis himself, but they find four or five other et tardif(Innsbruck, 2-5 septembre 1551)(199z), 283-91. hands at work drafting Cerialis' correspondence.

See the pertinent remarks of Petersmann, op. cit.

(n. s), 284-5

emphasizing the conservative, even archaizing, tendencies of these scribes, tendencies which, rather surprisingly, are more prominent in the tablets than lapses into substandard spelling.

It is as well to dismiss at the outset any notion that the tablets might display a 'British' variety of Latin. The tablets are not the product of a local British population, but of highly mobile military personnel and (perhaps) their civilian entourage. Vindolanda was garrisoned by both Batavian and Tungrian units,8 and consequently the Latin written there might in theory have reflected that in use among Romanized natives of areas such as Gallia Belgica and Germany rather than Britain. One writer (344) explicitly identifies himself as a homo transmarinus. There are indeed some traces of Celtic influence on the Latin of Vindolanda (see below, 11.10, VIII. IS). If Batavians or Tungrians in some cases at least employed a 'broken' or 'foreigners' Latin', that would be virtually impossible to detect because of the part which will have been played by well-trained scribes in putting such speech into writing. Nevertheless there are one or two oddities in a letter by a certain Chrauttius, to which I will come below.

11. ORTHOGRAPHY

11. I. The Digraph ae

In the first half of the first century A.D. in the legal documents of C. I\.:ovius Eunus the digraph ae is hardly ever written (one false example; e for ae seventeen time^).^ In the Vindolanda tablets, by contrast, ae is correctly written with remarkable consistency. I have noted about seventy three correct examples of the digraph.

It has recently been suggested that the ae diphthong was preserved until fairly late,1° though the evidence of Eunus and indeed other evidence (see further below)ll makes this unlikely. Nevertheless one is obliged to consider the possibility that in speech in the region of Vindolanda the diphthong was maintained. Such a conclusion would be untenable, at least as a generalization intended to describe the usage of a whole community. In just two documents there are revealing clusters of examples of e for ae. In the account numbered 186 the (genitive) spellings ceruese (twice) andporcine occur, and in the long letter by the entrepreneur Octavius (343) we find not only arre, que, illec (= illaec), and male (= malae), but also the hypercorrect form mae for me. Remarkably, in 186 there is no case of ae, and in 343 only one correct example (uiae). The two tablets taken together thus show a strong preference fore (7:2, with one case of ae false), which is in striking contrast to the consistent preference for ae elsewhere in the tablets (71 :o in documents other than these two).

These facts suggest the following conclusion. The diphthong ae had been converted to a monophthong in the speech of at least some (probably most) of the community,12 but the cultural level of scribes associated with military personnel in the area was such that they were able regularly to write ae without error. The two tablets 186 and 343, which offer us a fleeting glimpse of a speech pattern which has otherwise been obscured by the successful preservation of a writing convention, may be treated as special cases. 186 is written in a hand described by the editors (146) as ugly and sprawling. It displays three times a remarkable phonetic misspelling (Februuar- for Februar-, with a glide [w] represented between two vowels in hiatus), which in this lexical item is unique not only at Vindolanda (where the word is correctly spelt about a dozen times), but perhaps in extant Latin as a whole. There is abundant evidence for loss of u in this word (after the consonant cluster -br-) :see CIL 1v.4182,0.BuNjem 74, 76, 77, 101Febran'as; cf. It. febbraio, Fr.fe'vn'er, etc.13 The writer of 186 must have countered such loss in his speech by the insertion of a glide.I4 The writer's employment of this misspelling, as well as his use of e for ae, would seem to place him outside the group of scribes responsible for the bulk of the extant texts.

It is of some interest that Octavius, like the businessman C. Novius Eunus a couple of generations earlier, fell into error on one of just two occasions when he attempted a digraph.15 Octavius' use of e for

On the ethnic origins of those garrisoned at Vin- 175-91. Coleman ('Vulgar Latin and Proto-Romance:

dolanda, see Bowman and Thomas, 30-2. minding the gap', Prudentia 25 (1993)~ 5) points out that

See J. N. Adams, 'The Latinity of C. Novius Eunus', Servius (GL1v.421.21) implies a monophthongal pronun- ZPE 82 (1990), 230-1. ciation of ae in contemporary educated usage.

lo See P. Flobert, 'Le tkmoignage Cpigraphique des l2 It is highly unlikely that there were just two speakers

apices et des I longae sur les quantitks vocaliques en latin at Vindolanda who pronounced the original diphthong as

impkrial', in G. Calboli (ed.), Latin vulgaire -latin a monophthong.

tardifII. Actes du IIime colloque international surle latin l3 See Vaananen, op. cit. (n. 11), 41.

vulgaire et tardif (Bologne, 29 ao2t -2 septembre 1988) '"or the.insertion of a glide [w] in hiatus after a back

('990)2 '05, vowel, cf. e.g. CIL x1.6289puuer =puer, 0.Bu Njem 86

l1 For the evidence from Pompeii, see V. Vaananen, Le duua = VL dua, Varro, :Wen. 290 clouaca = cloaca,

latin vulgaire des inscnptionspompe'iennes3 (1966), 23-5. Petron. 44.18 plouebat =pluebat.

See also R. G. G. Coleman, 'The monophthongization of l5 For Eunus' error, see Adams, op. cit. (n. g), 230.

lael and the Vulgar Latin vowel system', TPhS (1971),

88 J. N. ADAMS

ae, and of ae fore, is the first of a number of orthographic abnormalities (abnormalities, that is, by the standards of most of the other tablets) which we are going to see in his letter. The status of Octavius is open to question: was he a military contractor or civilian trader?16 His orthographic practices may have some bearing on this question, as we shall see.

In this section I will be offering a detailed comparison between orthography at Vindolanda and that in two other extensive collections of military documents (the contemporary letters of Claudius Terentianus, and the Bu Njem ostraca). The correctness of the Vindolanda tablets in the writing of ae vs. e stands in sharp contrast to the frequency of misspellings in Terentianus and the ostraca. In Terentianus ae outnumberse by only 21 :18.17 At Bu I\.:jem thee-spelling outnumbersae by 51 :14, and one example of ae is hypercorrect.18

A crude count of examples of ae vs. examples of e at Vindolanda might have led one to deduce that in the 'Latin of Britain'ae was still a diphthong. We have been able to cast doubt on that possibility from a consideration of 186 and 343. The practice of counting examples of misspellings in relation to correct spellings in inscriptions from different areas of the Empire in an attempt to find regional variation in Latin is, on this evidence, futile.lg A 'correct' spelling may reflect not a current pronunciation but adherence to an old writing convention.

11.2. Final -m
Another remarkable orthographic feature of the tablets lies in the accuracy with which -m is written in final position. There are about 393 cases of -m correctly written, but not a single certain case either of an omission, or of -m written where it does not belong.20 Final -m was scarcely, if at all, audible in final position, even in educated speech (note Quint. 1x.4.40, Velius Longus, CL ~11.54.4). Failure to articulate final -m was not, therefore, a 'vulgarism', but failure to write it betrays a lack of control over the spelling system, and can be taken as a sign that the writer had not had a full literary education. Recently published non-literary documents, including those from military environments, tend to display omissions more or less frequently. In the first-century legal documents written in the hand of C. Novius Eunus -m is left out nine times and written hypercorrectly once (whereas in the 'correct' versions of the same documents in the hand of scribes it is always written, and ~orrectly).~~

In the letters of Terentianus one accusative singular in five is without -m (29 out of 149 cases).22 In the Bu Njem ostraca -m is omitted as often as it is written (forty two accusative singulars without -m, roughly the same number with). In another recently published set of early substandard documents (from La Graufesenque) there are various cases of omission.23

As Marichal has recently remarked,24 'La frCquence relative de m final est rCvClatrice du degrC d'instruction d'un scripted. Scribes at Vindolanda would seem to have been of a distinctively higher cultural level than those responsible for various military documents written elsewhere in the Empire.

11.3. Gemination and Simplzpcation
The writing of double consonants at Vindolanda provides further evidence of scribal competence. There are about 190 examples of geminates (I include here proper names, a number of them of non-Latin origin, e.g. 182 Ircucisso, etc., but omit for the moment certain etymologically correct examples of -ss- which will be discussed separately below). These are almost invariably 'correct' in words of Latin origin (but see below on nissi). By contrast there is only one clear case of false simplification, Polionis (187). To this might perhaps be added comodati (180), but this term is somewhat obscure.25 Exercias (233, <exsarcio) also displays a form of simplification, exs (i.e. ekss) >ex. The editors also suggest (141) that -aliator at 184 may represent malleator.

Early evidence for simplification is to be found in the documents of Eunus (seventeen examples)26 and La Graufesenque, where there is haphazard alternation between geminate and single consonant in certain words;27 but these documents, from a Latino-Celtic milieu, are perhaps a special case. Terentianus is close to Classical norms in this respect.28

j6 See Bowman and Thomas, 30. 20 On the hypercorrect addition in substandard texts of l7 See J. N. Adams, The Vulgar Latin ofthe Letters of m where it has no place, see Adams, op. cit. (n. g), 236. Claudius Terentianus (P.Miclz. VIII, 46772) (1977), 12. 21 See Adams, op. cit. (n. g), 236. Is See J. N. Adams, 'Latin and Punic In contact? The 22 See Adams, op. cit. (n. 17), 22. case of the Bu Njem ostraca', JRS 84 (1994),, 103. 23 See R. Marichal, Les grafites de La Graufesenque,

l9 See, for example, P. A. Gaeng, An Inquzry znto Local Gallia, supplCment ~LVII(1988), 67. Variations in Vulgar Latin as Rejected in the Vocalism of 24 ibid., 67. Christian Inscriptions, University of North Carolina 25 See Bowman and Thomas, 125. Studies in the Romance Languages and Literatures 26 See Adams, op. cit. (n. 9),238. LXXVII (1968), J. Herman, 'Aspects de la difftrenciation 27 See Marichal, op. cit. (n. 23), 67. territoriale du latin sous lJEmpire', BSL 60 (1965), 53-70 28 See Adams, op. cit. (n. 17), 3j. (= idem, Du latin aux langues ro)nanes (~ggo), IC-28).

LANGUAGE OF VINDOLANDA TABLETS

a9

In addition to the correct geminates noted above, there is a substantial accumulation of examples of -ss- in environments in which the geminate is historically correct, though not usually written, i.e, after a long vowel (or diphthong) and deriving usually from assimilationlassibilation within the consonant clusters -tt-, -ts-, -dt-. The most notable cases of this sort are in the perfectum of mitto and its compounds:

256 remissens, 268 missi, 280 missi, 299 missit, 309 missi (twice), 31opromissit, 3 12misseras, 314 mijsi, 3 I 8 missi, 344 commississem. Note too :

I 80 ussus, 255 ussibus, 225 occa~~-nem, 225 [oc]cassionem. There is just one hypercorrect case of -ss- after a short vowel, and that, interestingly, is in the letter of Octavius (343.20 nissi) .29

In the perfectum of mitto there are thus eleven cases of gemination. Forms with a single consonant (-misi) also occur eleven times.30 The use of a geminate in this position was by no means standard orthography elsewhere. In the letters of Terentianus perfect forms of mitto occur six times, always with a single s.31 At Bu Njem there are twelve cases of -s-, but no certain example of the geminate.32 In the Vindolanda material no tablet which has miss- also has mis-, and no tablet which has mis- has miss-. Scribes would seem to have been consistent in their choice of one system or the other. Moreover 180, in which ussus occurs, is written in the same hand as 344, which contains commississen~.

Spellings of the -ss- type belong to the category discussed by Quintilian 1.7.20, who cites the spellings caussae, cassus, and diuissiones as in use at the time of Cicero and Virgil, but a little later displaying simplification (after the long v~welldiphthong).~~ They are well attested in laws and archaizing inscriptions of the kind which characteristically have a variety of old-fashioned spellings. In the Lex de Gallia Cisalpina of 49 B.C. (CIL 1".592), for example, there are cases of r-emeissem't, repromeissen't, repromeississet, andpromeissem't, alongside (e.g.) diphthongal spellings in ei, -xs- for x (on which see below, 11.5)~ and u for the'intermediate'vowel. Particularly relevant to the orthography of Vindolanda is the short verse inscription CIL 1".1216 (= CE 58, Rome), which not only has missit, but also a number of examples of -xs-, ei for i, and a third declension ablative in -i (as distinct from -e: see below, 1v.1.1); most of these spellings I will have occasion to mention below. Also revealing are C. Novius Eunus' determined efforts to write -ss- as often as possible. He 'correctly' writes promissi at xv111.2.9, but also constantly uses the geminate after long vowels or diphthongs in words in which it was not etymologically justified (notably Hessucus and Cessar = Caesar; I assume that the derivation of Caesar from caes(s)us was a popular etymology). Several times for good measure he even writes it after short vowels (in possitus, three times, and A~sinio).~~

Eunus' chaotic use of the geminate betrays an awareness that such spellings were appropriate to legal documents, but an ignorance of when they were historically correct. ---

Quintilian speaks as if the -ss- spelling was no longer in use by his day. He was, of course, almost contemporary with the Vindolanda tablets. The constant use of -ss- at Vindolanda (compared with its complete absence from the letters of Terentianus, of roughly the same date) suggests a taste for old-fashioned orthography among some scribes in the area, a taste which will be further demonstrated below. And whereas Eunus almost a century earlier was incapable of restricting the geminate to environments in which it was etymologically justified, the scribes of Vindolanda consistently used it correctly. The one exception in this respect, as noted above, was the writer of Octavius' letter. Octavius' nissi is exactly parallel to Eunus' Assinio and possitus.

The evidence thus continues to accumulate that at least some of the scribes of Vindolanda were of

some educational attainment, and that the letter of Octavius is a case apart.

11.4. The Aspirate
I have noted about I I I cases of h correctly written (sixty times initially, twenty seven times inmihi, twice in nihil; also coh(ors) twenty two times; note in addition 184 Huep-, 187 Huete-). The aspirate is never omitted at the start of a word. Mi occurs just seven times, but in six of these cases it is in the letter of Octavius (343). Also worth noting is chors (cho-tis) at 127 (cf. chor. at 396), a form which reflects the loss

29 For nissi (nessi), see Tab. Sulk 32.7, 14, 65.10, and 32 Transmisi at 76, 77, 78, 79, 86, 87(?), 101, 104, I 10,

R. S.0. Tomlin, 'The Curse Tablets', in B. Cunliffe 148(?),misi at 95, retnisi at 103. [Trlanni~seat 105 may

(ed.), The Temple ofSnlis ~Tfinerva at Bath 11. The Finds represent transmisisse, with a haplography, but the text is ,from the Sacred Spring, Oxford University Committee for very fragmentary.

Archaeology Monograph xv~(1988), 151, 199. For a few 33 On the phenomenon, see in general M. Leumann,

further early examples of hypercorrect -ss-in this environ- Lateinische Laut- und Fomzenlehre6 (1977)~ 181. For a

ment, see Vaananen, op. cit. (n. I I), 60. On such spellings few examples of -ss-in this environment at Pompeii, see

in Eunus, see below. also Vaananen, op. cit. (n. II), 59.

30 218, 252, 7-59, 7-71> 292, 295,300,3", 317, 320, 345. 34 Full details can be found in Adams, op. cit. (n, g), 31 P. j1,lich. ~111.467.27 , 29, 468.5, 8, I 5, 28. 239.

9O J. N. ADAMS

of h intervocalically (cohors >cors), and its restoration graphically in the wrong place.35 The only other case of omission is in exibe at 282.

With alica at 2-33 should be compared halica at 193. There may not have been an established 'correct' spelling. There was clearly some controversy among grammarians as to the correct form: note Charisius, GL 1.96.9, 'alicam sine aspiratione dictam Verrius tradit, et sic multi dixerunt'.

Finally, haue at 291.14 is the form most commonly found.

In the overwhelming majority of cases h is retained and is correctly written, particularly if one makes allowances for the letter of Octavius, with its accumulation of abnormalities. It follows that in 154 in is (six times) is likely to stand for in iis rather than in his, given that omission of h initially virtually never occurs, whereas -ii-is often contracted (see 11.9). The alternation of in is with ex eis in the same tablet may have something to do with considerations of euphony.

There is one other piece of evidence at Vindolanda relevant to the aspirate. At 234 in a letter of Cerialis et hiem is erased and apparently replaced by etiam. The editors observe that this must be an error caused by dictation. If so, certain deductions can be made. It is scarcely conceivable that Cerialis pronounced etiam with an aspirate after the t. If the scribe wrote et hiem for etiam, then it follows that

(a) he was used to hearing hiems without an aspirate; and (b) on hearing the word pronounced thus he was capable of writing it with the correct h-spelling. Thus we learn something both about the pronunciation of Cerialis, and about the learning of scribes associated with the officer class at Vindolanda. Cerialis' letter 225, which may well be in his own hand,36 reveals a man with some stylistic pretentions (see 1x.1); yet it would seem that he did not pronounce the aspirate in initial position. Scribes were obviously attuned to inserting the aspirate in writing when they had not heard it in speech. The consistently correct spellings in this respect throughout the whole corpus of documents suggest that scribes had a significant degree of training in the relationship between the spoken and written language.

The documents of Bu Njem present a radically different picture. If one leaves aside the official daily reports numbered 3-24? in which in his is correctly written twelve times, the aspirate is omitted eleven times,37 but written only twice.38 The contrast between the stereotyped reports and the rest of the tablets shows that exemplars were provided for those keeping records, and that these could have an influence on orth~graphy.~~

The general indifference to the aspirate in the letters found at Bu Njem (0. Bu Njem 74-1 17) stands in contrast to the correctness of the correspondence at Vindolanda (if one leaves aside the idiosyncratic letter of Octavius).

In Terentianus h is more often written than omitted, but there are significant variations from letter to letter which betray the varying practices of different scribes. In P.Mich. v111.467 h is written eleven times and never omitted. At the other extreme is 471, where it is omitted nine times (inc, mi five times, abuit, abiturum, nil), but written only three times (hoe, mihi, nihil). There seems to have been a fairly uniform literacy among scribes at Vindolanda which distinguishes them from the scribes to whom Terentianus had access.

xs is commonly written for x in the tablets:

181,uex~illari,284 exsigas, 301 sexs, 309 axses, axsis, axses, 333 exse-, 343 uexsaTe

Tablet 309, as well as having three instances of xs, also has missi twice (see further below on the

combination of these two types of spellings).

For x unaccompanied by s, cf. 161 Expeditu~, 182 Exomnius, exungiae, 185 axes, 190 axungiae,

2 I 8 exegeras, 225 amplexuj, 282 exibe .

To these examples might be added exercjus (~33)~ where the etymologically correct xs (ex-sarcio)

has been simplified. An anomalous spelling is -tel[excisse (229), perhaps for intellexisse.

The spelling xs probably derives from a feeling that a consonant cluster (x = [ks]) should be represented graphically by more letters than one.40 While the spelling may turn up anywhere (e.g. in tomb inscriptions uixsit and uxsor are constant),41 there can be no doubt that it had the status of a 'formal' or archaizing spelling, appropriate (e.g.) in laws which display a range of other archaizing spellings. For example at CIL 1'.582 one finds exsigito (cf. Tab. Vind. 11, 284, above), taxsat, lexs, proxsumeis, as well as diphthong spellings, the intermediate vowel u, and also third declension ablative singulars in -i(luuci, luci =luce), forms which, as we shall see (IV. I. I), are also attested at Vindolanda. Just as Tab. Vind. 11, 309 has three cases of axsis alongside two of missi (see above), so the Lex de Gallia

35 See F. Sommer, Handbuch der lateinischen Laut- 38 Hora (67), mihi (83).

und Formenlehre4 I (rev. R. Pfister, 1977)~ 154 n. 5, and 39 See further Adams, op. cit., (n. 18), 96.

on the history of the spelling, B. Lofstedt, Studien iiber 40 cf. Vaananen, op. cit. (n. II), 64.

die Sprache der langobardischen Gesetze. Beitrage zur 41 For xs spellings at Pompeii, see Vaananen, op. cit. friihmittelalterlichen Latinitat (1961), 77. (n. II), 64. 36 See Bowman and Thomas, zoo. 37 Annibal (32, 34, 68), mi (86), oralura (91, 103, 105,

II~),ordeum (97), Vrtato (II~),abes (116).

Cisalpina (CIL 1'.592) has on the one handproxsume, duxsen't, noxsiaeue, deixsen'tue, and on the other a cluster of -ss- spellings in the perfectum of (-)mitt0 (see above, 11.3). There is a consistency of practice in the lex: xs is always written, except in the prepositionlpreverb ex, which always has x alone. Also worth comparing with Tab. Vind. 11, 309 is CIL I".IZI~(see above, 11.3)~ which not only has missit alongside four examples of xs (uixsi, uixsere, senexs, saxso), but also thei-ablative. It is the fact that the xs spelling at Vindolanda is attested alongside other archaizing spellings which gives it its significance. Scribal practice at Vindolanda was not only predominantly correct, but also conservative.

The presence of xs at Vindolanda may be contrasted with the convention followed in the letters of Terentianus. There I have noted about nineteen examples of x (note, e.g. P. Mich. v111.467.8 uexillo,

468.18 sex), but not a single case of xs. In the same documents, as we saw, there are six examples of the perfectum of mitto, all of them spelt with a single s.

The Bu Njem ostraca follow the same conventions as the letters of Terentianus. The ostraca contain about ten cases of x, but just one of xs (78 sexsagi[nta). Similarly the perfectum of (-)mitt0 alway has a single s (see n. 32).

11.6.Final -ti-d

In monosyllables and some 'grammatical' words there was some confusion between t and d in final position (see Quint. 1.7.5 on atl~d).~~

Confusion arose because of a tendency for the final consonant to be assimilated in voice to the phoneme which followed.

There is again a high degree of accuracy in the spelling of such forms in the Vindolanda tablets, with only isolated deviations from the norm in certain documents. Ad is correctly spelt more than sixty times (seven times, for instance, in the expression ad te,43 in which the d was particularly subject to assimilation: see below). Ad is misspelt just twice, both times in the letter 292, in the expression at te. Id is correctly spelt three times, quid seven, sed three, aliquid three, quod fourteen (omitting descn'pta).

The only other misspellings comprise two in 248 (it, quot), and four in the letter of Octavius, 343 (quit twice, aliquit twice). Whereas in the documents as a whole correct spellings predominate overwhelmingly, in the letter of Octavius incorrect spellings predominate by 4:z (quod is correctly spelt twice). Octavius alone at Vindolanda misspells quid and aliquid, and he alone makes a distinction between (a1i)quit and quod. Here is further evidence for the abnormality of his orthography.

Incorrect spellings are rather more common in Terentianus (twenty two examples).44 Terentianus (or his scribes) often spells et and ut as ed and ud, two spellings which nowhere occur at Vindolanda.

Bu Njem again provides the sharpest contrast with Vindolanda. Whereas (as we saw) at Vindolanda ad te predominates over at te by 7:2, at Bu Njem ad te is never correctly spelt. At te occurs five times, and a te three times.45

11.7. vowels
The merger in Vulgar Latin of 2with rand6 with zi causes misspellings of the typee for original rand

o for zi,46 though early evidence of such mergers is neither extensive nor straightforward of interpre- tati~n.~'The

only possible case of a misspelling of this type at Vindolanda is debetorem at 250. Though I previously have interpreted this as a manifestation of the merger of 2 and i,48the absence of definite parallels in the full collection of documents now raises doubts about such an interpretation. It is possible that the word has been subject to a false analysis (debet-orem: influence of debet?).

The most interesting vocalic misspelling is tuGas in the account 180.20. The editors tentatively

suggest that this may be an alternative form to torta, which is attested a number of times in the Vulgate

(see the editors, ad loc.) as a name for a type of twisted loaf. This suggestion is undoubtedly right, and

paradoxically it is the misspelling with u which establishes its correctness.

Torta was no mere ephemeral oddity of the Vulgate. It survives widely in Romance languages with

much the same meaning as the Latin term (e.g. Fr. tourte, It., Sp., Pg., Prov. tort^).^^ It is of

significance that the Romance forms reflect not the expected CL 6 in.the first syllable,50 which would

have produced an open^ in Romance, but have rather a closeo ([o]) which was the Romance outcome of

42 See e.g. Adams, op. cit. (n. 17), 27-8, idem, op. cit. Adarns, op. cit. (n, g), 231, idem, op. cit. (n. 18), 103. On

(nip, 237. e for i at Pompeii, see Vaananen, op. cit. (n. II), 21-2

218 b~s, 226 his, 252, 263, 318. (considering the possibility of Oscan influence).

44 Adarns, op, cit. (n. 17), 27-8. 48 In A. K. Bowman and J. D. Thomas, Vindolanda:

45 0.BuiVjem 71, 76, 77, 78, 79; 86, 101, 104. the Latin Writing Tablets, Britannia Monograph Series IV

46 See especially Lofstedt, op. cit. (n. 35), 5666 (with (1983), 73.

extensive bibliography). 49 See W. Meyer-Lubke, Romanisches etymologisches

47 Neither in the documents of C. Novius Eunus nor at W6rterbuch3(193 j), 8802.

Bu Njern is there a certain case of such a misspelling: see The o of the participle of torque0 in CL was short.

92 J. N. ADAMS

the merger of CL 6 and zi.51 Here indeed in the Vindolanda example we see the u which would have produced o in Romance.

The u-spelling is readily explicable: r + consonant tended to close a preceding0 (or indeed e, as the early spelling stircus shows). Cf. Appendix Probi 25, 'formica non furmica' (but possibly a popular etymology, <fur);52 also curs, curtis for c(h)ors, cortis (< cohor~).~~

Cohors is reflected in Romance with o rather than q (e.g. It. ort to),^^ which shows that it developed on exactly the same lines as torta. For a further case of closing in this environment, see GLv. 575.7, 'cortina per o dicendum, non curtina'. Also of note is furnus alongside its near-synonym fornax, both words perhaps of the same root as fomzus -a -um (cf. Fomziae) .55The form furnax is attested at Vindolanda (I ~5.7)~

a spelling which might reflect contamination with fur nu^,^^ or an independent closing under the influence of r + consonant. Fornus at 0.Bu Njem 7.14 looks like an inverse contamination.

Turta is thus a phonetic misspelling which represents an intermediate stage between the Classical and Romance forms. The sequence of changes was to'rta > tzirta > torta.

11.8.Syncope

I have noted only five examples, three of them in titles or terms of address: 170 domne, 214 fraterclo, 21j comiclan'o, 196 subuclas = subuculas. Syncope is particularly common in terms used in the vocative or in titles habitually attached to names.57

11.9.Hiatus

Abundant evidence for the modification of vowels in hiatus contrasts with the relative absence of evidence for various other phenomena, and leads to some interesting conclusions. I begin with a classified collection of the data:

(i) Contraction of -ii- in hiatus

I 54in is (six times: on the interpretation of is, see above, 11.4), I 54 Con's, 164gladis, 175 Con's, 180 bubulcan's, 181 uexsillari, 185 Vinouis, 186 Febl~uan's, 189 Zunis, zjopetit =petiit, 266 Con's, 292 peti, 312 Con's, 343 necessan', 343fie[issem, 345 ali, 349propiti. TOthese twenty two examples might be added seven examples of mi, though orthographically (but not phonetically) there is a difference between the spelling mi for mihi andgladis forgladiis, in that some writers will have been aware of the place of h in the written form of mihi, and that awareness will often have helped them avoid the contracted spelling.

There is also a case of cho~is at I 27, which, as we have seen (11.4), reflects co(h)ors >cors (>chors).

But in this section I concentrate especially on the far more frequent contraction of ii.

I move on now to examples of ii which are uncontracted in writing. There are in total thirty one

examples, but twenty seven of these comprise the form mihi, which, as we saw, is a special case in that it

contains a letter to keep the written form of the vowels apart. Similarly there are two examples of nihil.

The only other uncontracted forms are: I 56 Mart@[s], 343 Ianuariis (cf. possibly 201 F<bruan'[is]).

Both examples are in the names of months; in the writing of dates the correct spelling may have been

more persistent. The second example is of special note, because it is in the letter of Octavius, in which

contraction is particularly frequent (see p. 95).

If milmihi and nihil are excluded, contracted forms outnumber uncontracted by 22:2.

The letters of Terentianus offer hardly any examples of iili, whether contracted or not, except in

words originally written withh between the vowels. The only other case isdis at P.Mich. v111.467. I 5, but

that is a standard spelling. What the letters do show is the persistence of uncontracted spellings in those

words which in their correct graphic forms have anh to keep the vowels apart. Mihi occurs sixteen times,

mi eleven, nihil four times, nil once. These bare figures obscure variations from letter to letter which

presumably reflect different scribal practices: in 467 mihi is preferred to mi by 9:o, whereas in 471 mi is

preferred by 5: I, and in 468 by 6:4.

In this case there is a close correspondence between scribal practice at Vindolanda, and that of the Bu Njem tablets. In the latter mi and mihi are not frequent enough to influence the picture. ii is

51 See Meyer-Liibke, op, cit. (n. 49), and especially 0. 56 See Adams in Bowman and Thomas, op. cit. (n. 48), Bloch and W. von Wartburg, Dictionnaire e'tyrnologique

73.

de la languefran~aise5 (1968), 642, S.V. tourte. 57 See V. Vaananen, Introduction au latin vulgaire3

52 See W. A. Baehrens, Sprachlicher Kommentar zur (1981), 42 on domnusldomna in contrast to other

vulgarlateinischen Appendix Probi (1922), 55; also Leu- Romance words which retain an original Latin sequence

mann, op. cit. (n. 33), 48. -min-: 'Quant i dominus, domina, la syncope tient

53 See Lofstedt, op. cit. (n. 35), 77-82. I'emploi de ce substantif comme appellation ou comme

54 ibid., 81. titre . . .'

55 See A. Ernout and A. Meillet, Dictionnaire e'tyrno-

logique de la langue latine4' (1959)~248.

LANGUAGE OF VINDOLANDA TABLETS

contracted twenty eight times (mi once, at 86), but correctly written only once, and that in mihi (83); perhaps too stationan'i at 28.

In various other aspects of orthography, as we have seen, the Vindolanda tablets are very much more correct and conservative than those of Bu Njem. That the two sets of documents should correspond so closely in this one respect reveals the extent to which the contraction of similar vowels in hiatus had progressed. Scribes at Vindolanda set out to employ a correct orthography. That they should so consistently have written i for ii except where h was present suggests that contraction was absolutely normal in speech, and that there was not a 'correct' pronunciation current, say, among particularly careful or educated speakers which might serve as the model for a 'correct' spelling. Contraction of ii is a phenomenon which strictly should be given no place in handbooks of 'Vulgar Latin', because there is every reason to believe that in the speech of all classes it was the norm from a relatively early period.58

I move on to some further misspellings found in different environments in hiatus.

(ii) Closing of e to i in hiatus after a consonant

Closing of this type is represented about seven times (I 59 hordiar[, 181 balniatore, 184, 207, 255 sagacia, I 89 uinias, 299ostn'a). In this environment there was undoubtedly a clearer perception of what was 'correct'. e is correctly written in hiatus about thirty six times. Some words are regularly spelt correctly (e.g. commeatus, at 168, 174, 175, 176, 177, hordeum, at 185, 213, and seven times in go), whereas sagacia (see above) is regularly misspelt. The familiarity in written forms of the language of certain lexical items (e.g. commeatus) will have enabled scribes to spell them correctly. Sagacia, on the other hand, was a new, specialized term (see v. I), which had not achieved a correct orthographical form.

The name Cerialis turns up constantly, spelt with an i (so too Genialis). Just once (261) it is spelt Cereali~.~~

(iii) Glide insertion

When two vowels of quite different quality stand together in hiatus, contraction is unlikely but the insertion of a glide ([j] or [w]) may take place. One would not expect much sign of this phenomenon at the level of spelling at Vindolanda, given that the orthography of scribes is conservative, but there are two interesting cases. In 186 Februar- is three times spelt with a u (Febmuar-) which represents [w]. In the speech of the writer (see 11. I) [w] must have been inserted to counter the loss of u after br in hiatus. This spelling sets the writer apart from the generality of scribes at Vindolanda.

Secondly, the Celtic(?) name Gauo is twice spelt Gauuone at the start of documents (192,207)~ and then at the end of both documents is written Gauonis (genitive), without the glide represented. This is precisely the type of word in which one would expect to find orthographic uncertainty, since there was no traditional written form to guide the writer. Presumably the glide would be heard, but was not consistently indicated in writing because the Latin spelling system did not regularly mark glides.

(iv) Omission of u before a back vowel One example, Iugenus (181) .60

11.10.Some Consonant Clusters

n is omitted before s just three times (187 Masuetus, 337 castresia, 344 trasman'num). The omission (in speech) had long been standard even in the educated language (Cicero said foresia, megalesia, and hortesia :see Velius Longus, GL v11.79.1-2) .61 Its omission in writing was non-standard, but the phenomenon is rare at Vindolanda. Conversely fomzonsa at 302 is hypercorrect; cf, Appendix Probi 75, 'formosus non formunsus'.

The assimilated spelling emtus = emptus occurs twice (181, 189). This assimilation no doubt reflects a widespread pronunciation; for an early Imperial parallel, one might compare sumtuan'um in the ostraca from Ma~ada.~~

The omission of n before a stop (136 renutium) occurs sporadically in early documents.63

This last misspelling is relevant to the interpretation of souxtum at 301.3:

souxtum saturnalicium
(asses) iiii aut sexs rogo frater
explices

58 C. Novius Eunus offers in the early first century A.D. 61 See Leumann, op. cit. (n. 33), 146. the spellings sestertis, medis, and isdem (see Adarns, op. 62 See Hannah M. Cotton and J. Geiger, Masada II. cit. (n. 9), 235), against one case of iis. Contrast iis The Yigael Yadin Excavations 1963-1965, Final Reports. sestertiis in the correct (scribal) version of TPIs. The Latin and Greek Documents (1989), 722.7, 14 (with

59 On the variable spelling of CerealislCenalis, see TLL 37 n. 33). Onom. 11.343.37ff (in inscriptions usually Ceri-); also 63 See e.g. Vaananen, op. cit. (n. I I), 67-8, Adams, op. Servius on Virg., Aen. I. 177. cit. (n. 9), 241.

60 See, e.g. Adams, op. cit. (n. g), 235.

94 J. N. ADAMS

Souxtum, the reading of which is certain, is a Celticized misspelling of su(m)ptum, with a typical Gaulish substitution of xt forpt (cf. sextametos 'seventh' at La Graufesenque, alongside Lat. septimus; also captiuus > Gallo-Latin *caxtiuus > Fr. che't~fl.~~

11.I I. Some Conclusions
(i)
In a number of respects the orthography of the Vindolanda tablets is largely correct: in the use of the digraph ae (11. I), the writing of final -m (II.~), and in the writing in the preservation of geminates (II.~), of the aspirate (11.4). The case of the dictation error et hiem (11.4) reveals the ability of a scribe to write h when he had not heard it. It has been shown that orthography at Vindolanda is superior to that in various other military documents.
(ii)
Alongside this general scribal correctness we have noted certain archaizing tendencies, notably the writing of doubles after long vowels (11.3). There is a distinction between the efforts of C. Novius Eunus in this respect, and of the scribes of Vindolanda, in that Eunus often writesss where it is not historically justified. It has also been suggested that the examples of xs display an archaizing taste (11.5). These two forms of archaizing are completely absent from Terentianus and at Bu Njem. In the discussion of -ss- after long vowels and -xs- for x it was pointed out that the artificial form was not universally preferred in the tablets. But those tablets which have -ss- do not haves, and likewise those that have xs do not have x. There were two possible orthographies in each case, and scribes seem to have adhered to the one or the other. The letter 309 is especially remarkable because it combines two examples of -ss- with three examples of -xs- (with no cases of either long vowel + s or of x). To scribes of the archaizing school the two spellings belonged together; and we will see below (1v.1.1) another artificial form which went hand-in-hand with -ss-.
(iii) There are, of course, spelling errors at Vindolanda, and particular attention has been drawn to contractions in hiatus (11.9). Whereas scribes got some things constantly right, they constantly contracted -ii-, unless an aspirate was present in the written form of the word to separate the vowels. Contraction must have been so firmly established in the speech of all classes that it was difficult even for a well-educated scribe to be sure when to write ii rather than i. Some further statistics will be given below.

(iv)
The letter of Octavius (343) displays an accumulation of spelling abnormalities when it is compared both with other letters, and with the documents, military and otherwise. Octaviususually fails to use the digraph ae (11. I), he shows a higher than normal rate of error in the use of the aspirate (11.4) and of tld in final position (II.~), and he alone produces hypercorrect examples of the digraph ae (mae) and the geminate ss (nissi). In general the incidence of spelling errors in his letter is higher than that in other letters and in documentary tablets. I will return to this final contention below. The spelling irregularities of Octavius do not prove decisively that he was a civilian rather than a military man, but they point in that direction. If many of the other tablets were written by military personnel or military scribes, it is possible either that Octavius did not have access to a scribe, which might mean that he was not in the army, or that, if he did, his scribe was not of the same cultural level as those associated with the army.
(v)
The relative incidence of various spelling errors prompts some tentative observations on the chronology and status of certain changes in spoken Latin. The contraction of vowels of similar quality in hiatus was clearly the norm in all social dialects. At the other extreme there is a total lack of clear-cut evidence for the vowel mergers Z with iand 6 with zi. This could mean that, even if these mergers were taking place, they were not so widespread that they could yet influence the orthography even of writers (such as Octavius) who were prone to various other types of spelling errors. The monophthongization of ae on the other hand could influence the spelling of Octavius. It follows that monophthongization of ae was further advanced than the vowel mergers, though it might possibly have been resisted in some social dialects.
(vi)
The spelling souxtum, if our interpretation is right, would appear to reflect a Celticized pronuncia- tion of a Latin term. The tablets were written in a Celtic-speaking area, but, more to the point, at least some of those garrisoned at Vindolanda probably originated in continental regions where Celtic was spoken. To explain the form of certain lexical items in Gallo-Romance (e.g. Fr. ch6tif) it is necessary to assume that substratum influence of this very type had caused a modification of the Latin term in Gallo-Latin. I take souxtum as one important piece of evidence for language contact and its influence on
64 For details, see J. N. Adams, 'The interpretation of souxtum at Tab. Vindol. II.301.3',forthcoming in ZPE.

95 the Latin which would have been heard at Vindolanda. On Celtic loan-words at Vindolanda, see below,

VIII.15 .

I turn now briefly to the distribution of spelling errors across the various types of documents surviving at Vindolanda (military documents, lists and accounts, correspondence, if one follows the classifications of the editors). By 'spelling errors' I mean phonetically inspired misspellings, as distinct from deviant spellings which are not phonetically determined but motivated by notions of archaizing 'correctness' (e.g. xs). The study of the distribution of such spelling errors in a corpus of documents may well reveal variations of authorship, as some of the data from Terentianus noted above show (see 11.4 on the writing of h in Terentianus, and 11.9 on contraction of vowels in hiatus).

The distinction made by the editors between 'military documents' (127-77) and 'accounts and lists' (178-209) may be rather arbitrary, as at least some of the latter may have been 'military'. Nevertheless I treat the two categories separately here. In the military documents there are about fourteen spelling errors of the type defined above, of which the great majority (ten) involve the treatment of vowels in hiatus. Six of the errors are in the expression in is in 154. In accounts and lists there are twenty four errors, fourteen of which are in hiatus. There is no clear distinction in the pattern of errors between military documents and accounts; a feature of both categories is the high incidence of errors in hiatus. The account 186 stands out for the frequency of its errors: seven, or 29 per cent of all errors in accounts and lists. The document is far from providing 29 per cent of the lines in accounts.

In the correspondence there are thirty four errors, almost half of which (sixteen) are in the letter of Octavius. To these may be added the hypercorrect formsmae and nissi in the same letter. Though 343 is long (more than forty lines), it is certainly not half the length of the full corpus of letters extant. In nineteen other letters which I have chosen for comparison because of their good state of preservation there are some 280 lines, i.e. six times as many lines as in 343.

Errors in hiatus are again well represented (eight examples in 343, fourteen examples in all in the correspondence; six of the seven examples of mi in letters, as we saw (II.~), are in Octavius' letter). There is one error which is restricted to the letters (eight examples), that is confusion between tld in final position. It is hard to know what importance to attach to this distribution. There are many examples of ad in military documents and accounts which might potentially have been misspelt.

The most significant feature of the distributions noted above is undoubtedly the rate of error in Octavius' letter.

111. PUNCTUATION

III. I. Medial Points

Interpuncts are not regularly used in the Vindolanda tablets,65 though there are one or two documents in which almost all words are marked off (most notably 345; also 323) The evidence of Vindolanda fits in well with the general picture of writing with interpuncts which has been constructed from other evidence. It had been regular Roman practice up to the first century to write with interpuncts (see Sen., Epist, XL.I I) but thereafter the practice faded At Vindolanda then one would expect little or only sporadic interpunction.

But once regular interpunction had disappeared, the way was open for occasional medial points to be used for syntactic or related purposes, to mark cola, pauses, constructions, et~.~~

There is some evidence at Vindolanda for the use of interpuncts to mark sense pauses, if only occasionally. The ethnographical fragment 164 has interpuncts enclosing each of a pair of parallel negative clauses:

Snon utuntur equites.nec residunt Brittunculi.

65 See Bowman and Thomas, 567. lerer Hohe fur die Bezeichnung von Sinnpausen zu ver-

66 Note also Cic., Mur. 25. See M. B. Parkes, Pause and wenden'. Miiller cites (367) P. Oxy. 1.32 (= R.
Effect. An Introduction to the Histoly of Punctuation in the Cavenaile, Corpus Papyrorum Latinarum (1958), ~49)~

a
West (1992), 10, T. N. Habinek, The Colomety ofLatin second-century letter of recommendation, where some
Prose (1985),43 with n., P. J. Parsons in R. D. Anderson, clauses are divided by interpunction. For such use of

P. J. Parsons and R. G. M.Nisbet, 'Elegiacs by Gallus medial points in early manuscripts, see Habinek, op. cit.
from Qasr Ibrim',rRS 69 (1979): 131 with n.43. (n. 66), 61, 82-3 (on the Medicean manuscript of Virgil).

67 See R. W. Miiller, Rhetonsche und syntaktische On lnterpuncts in the Bu Njem ostraca which correspond Interpunktion. Untersuchungen zur Pausenbezeichnung 'B une coupure logique', see Marichal, op. cit. (n. 3), 40 im antiken Latein (1964)~36: 'Erst das Verschwinden der (see 0.Bu ~Vjem 68, 71,72, 81; perhaps too 77, 78). Wortinterpunktion machte es moglich, den Punkt in mitt-

J. N. ADAMS

This is the most striking example of syntacticlrhetorical interpunction, but there are some other cases. Note:

205 ]vii K(alendas).Ianuarias.in singulos dies
] (sextarii) iiii.fiunt dies x!iii
m(odii) x s(emis)

Here the points (other than that after K(a1endas)) happen to correspond to the punctuation in the editors' translation:

26(?) December. Per day, 4 sextarii. Total for 42 days, modii 10%.

In the address of 260 interpuncts separate the name of the addressee from that of the writer:

Flauio Ceriali praef(ecto) coh(ortis).a Iustino col(lega).

For clauses or verb-phrases marked off, see:

cya~.benemane Vindolandam uenisut . .

242

266 uolo ueniat.ad me Coris.et accipiat

In 31 1 there is an absence of interpuncts except before a vocative expression:

cupio.homo inpientissime

Cf. 175 'rogo.domine', 345 '~te.rogo~frater.'.

It is possible that occasionally the use of interpuncts may be relevant to the phenomena of enclisis and proclisis. I take proclisis first. Prepositions were proclitic, in the sense that they formed a single accentual unit with the dependent term.68 No doubt for this reason prepositions are only sporadically divided from the dependent term by interpuncts in those texts which make use of points.69 A good example of a document which, despite its regular use of interpuncts, generally does not separate preposition from dependent term is the Lex de Gallia Cisalpina (CIL 1".592). At Vindolanda note:

211 dehacsre
315 et aliassad Vocusium
&fyicanum.praefectum
323

ex ratiunculis.(in a text in which interpuncts are regular).

Of rather more interest in the Vindolanda material is a possible tendency for unemphatic (enclitic) pronouns to be left undivided from the term to which they are attached. 345 has almost complete interpunction, one exception of only two being misitibis. On 315 the editors (301) state that 'Interpunct is found everywhere except after monosyllables and after camla'. This is a correct statement, but it should be noted that carrula is followed by a pronoun:

ut carrula uobis.dentur.et alias9ad Vocusium

Note too 339:

et suasit me.

I have not found any discussion of the possible relationship between interpunction and the accentuation of personal pronouns, nor can I find convincing parallels for the examples cited above.70 Unless more material is discovered and parallels found it would not be safe to take the above items as new evidence for the enclitic character of unemphatic personal pronouns.

68 See e.g. W. S. Allen, Accent and Rhythm. Prosodic 'O But see P. Berl, inv. I 1649, 'salutem tibi .dicunt, Features of Latin and Greek: a Study in Theory and nostril. Reconstmction(1973), 24-5.

69 See E. 0.Wingo, Latin Punctuation in the Classical &e (1972)~16.

LANGUAGE OF VINDOLANDA TABLETS

111.2.Apices

The editors (57-61) list and discuss occurrences of the apex mark. I add here a few further observations, attempting in particular to assess what linguistic significance if any may be extracted from the use of the apex. The editors' lists require some minor adjustments. Facihs should not be in the list of items (59)showing apices over short vowels. Conversely bptamus should not be in the list of apices over long vowels. The editors include in their list quite a few apices which are difficult to interpret, either because the mark itself is unclear, or because the text is so fragmentary that the length of the vowel on which the mark stands is uncertain. Accordingly I give my own modified and abbreviated lists, containing only cases which are relatively certain, on which some statistical remarks will be based.

Apices over long vowels

194 lateran'b, compendibn'um, 196 ci Tranquillo, ci (Broccho), Brocchb, 21 2 Verecundb, sub, 2 I 5 aequb (?),Cassib, 221 Flauio, 234 Octbbres, 239 Flauius, sub, 242 numercitioni, 243 sub, fhter, 245 -rb (probably part of a name), 248 tzi, 255 Flauib, 261 sub, 263 tub, 265 frciter, sacnficib, uoluercis, 291 rogb, interuenth, salutci, facias, 292 Brocchb, una, 305 Vettib, Seuerb, 307exorb,310 sub, 3 1 I cupib, putb, scrib6, rogb, nbmina, 3 19 Veranib, sub, 324 -inn& (probably a name in the ablative), 330 me6

I omit the (mostly doubtful) cases listed by the editors from 371-513.

Apices over short vowels

175 &go, 192 -nean'a, I 98 -bra, membrh, 207 saga, 248 bptamus, 265 Khlendarum, shcnficio, 29 I

Severa, facturh, 292 necessan'a

There are in these lists forty three instances of apices placed over long vowels, and eleven over short.

Flobert, in his study of apices in various inscription^,^^ has noted a marked tendency for an apex to be placed over vowels in stressed syllables. The placing of an apex on ashort vowel in a stressed syllable might possibly be related to a tendency for short vowels under the accent to be lengthened (and, indirectly, to a converse tendency for long vowels in unstressed syllables to be shortened). In the Vindolanda material, however, there is little coincidence between apex placement and the position of the stress accent. There is only one case of a short vowel with an apex in a stressed syllable (175 rbgo). As far as correct apices are concerned, these occur over vowels in stressed syllables in the following places:

194 compendiarum, 221 Flauio, 234 Octbbres, 239 Flhuius, 243 frater, 248 tu, 265 frater, 311 nbmina

Only about 16 per cent of the apices in the two lists are on stressed vowels. There seems to be little tendency to mark stressed vowels as such, and there is hardly any evidence here for the lengthening of short vowels under the accent.

Of the fifty four apices in the two lists, all but two (248 tzi, 291 interuentzi) are on the lettersa ando. It may be worthwhile to consider the place within the word of these letters. I exclude from consideration the monosyllable 6 (twice at 196).The thirty nine cases of a and o with a correctly placed apex (i.e. an apex over a long vowel) are distributed thus:

internal syllable final syllable
over a 6

5

over o 2 26

Thus thirty one instances (= 78 per cent) stand on a long vowel in final syllable.
Incorrectly placed apices are distributed thus:

internal syllable final syllable
over a 2

7

-

over o 2

op. cit. (n. IO),e.g. 104, 106.

98 J. N. ADAMS

The figures reveal a high incidence of apices on vowels in final syllables (76 per cent of instances in the two tables). The significant proportion of apices in this category might possibly be related to a tendency for long vowels in final syllables to undergo shortening. The final syllable of a Latin word never bore a stress accent, and for that reason vowels in this position were historically subject to modification of various types (e.g. loss, as in *homce >hunc, duce >duc, uidesne >uiden ;opening of i to e to counter the particular vulnerability of i, e.g. facili >facile; shortening of long vowels, e.g. aue, as Quintilian

1.6.21 seems to imply; note too the phonetic spelling seese twice in the lex incerta, CIL 12.582, apparently for sese). In the later period grammarians refer both to the lengthening of short vowels under the stress accent,,and to the shqrtening of long vowels (including those in final syllables) which were not accented: e.g. CZre's for CL Ce'rZs, Sacerdos GL v1.451.13, 'per immutationem accentuum, ac si dicas CCre's ce longa, cum breuis sit, et res breui, cum sit longa' (cf. Consentius, GL v.392.3, 11, 18).

Of the seven misplaced apices in a final syllable, every one is on an a. a is the only letter which in final position of a noun or adjective may carry different phonetic values of potential semantic significance (e.g, mens6 nom. vs. mensti abl.). In the case systeme in final position is always short, whereasi, o, and u are always long. Since, according to Quintilian (1.7.2), the apex was not to be placed over every long vowel, but used only to avoid confusion, there may have developed a tendency for apices to be written particularly over long a in final position in case endings. But if distinctions of length were being lost in this position, and if consequently scribes were often uncertain about the 'correct' length of the vowel, the conditions were ideal for the hypercorrect use of the apex where it did not belong historically.

I move on to apices over o. In six cases the apex is on the finalo in a first-person verb form (291 rog6, 307 exorb, 3 I I cupib, put6, scn'bb, rog6). I have listed these examples under the heading 'apices over long vowels' for historical reasons only; in fact, in some or all cases the vowel may have been short. Rogo (twice) and puto are iambic words, in which the long final vowel would long since have been subject to iambic ~hortening.~~

Moreover by the Augustan period shortening of -6 begins to appear in non-iambic words (e.g, in Horace and O~id).~~

Eventually in verb-endings, whether the structure of the verb was iambic or not, short o was to become the norm: note Charisius, GL 1.16.20, 'paulatim autem usus inuertit, ut in sermone nostro sc~ibo', dico', et item talibus, ubi o non solum correpta ponitur, sed etiam ridiculus sit qui eam produxerit' (the man who lengthens the o of scn'bo or dico makes himself ridiculus), Pompeius, GL v.232.21-37, 'nemo dicit dic6, sed dico' . .. omnis ergo o in prima persona semper corripitur exceptis monosyllabis' (i.e. d6, but reddo').

It is likely then that at the time when the Vindolanda tablets were written, final -o in first-person verbs was tending to be shortened. The six instances of apices written in this position I would not classify as mistakes. Rather, they probably reflect an attempt by careful and well-trained scribes to counter the habit which was spreading in colloquial speech; the apex demonstrated that the writer was aware of the correct length of the vowel.

Among the correctly placed apices listed earlier, there are, quite apart from those inputo and rogo, various other examples on the second (historically long) vowel of iambic words (sub six times, tub once, me6 once). Thus eleven of the apices on final -6 appear in words of iambic structure. Again it is possible that the writing of the apex was motivated by a desire to counter the process,of shortening which may have left a legacy in colloquial speech.

By the early Imperial period a short o begins to appear in final position not only in first person verbs of iambic and non-iambic structure, but even in non-iambic ablative and imperative forms (esto, uincendo) .74 I am inclined then to relate the high incidence of historically correct apices over final -6 in the Vindolanda tablets to the effects not only of an earlier iambic shortening, but also of ashorteningof -6 which was spreading even to non-iambic words. The scribes of Vindolanda aspired to an orthographic correctness, and hence they often were careful to mark the correct quantity of the -6.

The editors draw attention (6-1, with n. 55) to a long-standing convention whereby the apex was

employed in the address of letters, on the ablative ending usually of names. This convention is probably

unrelated to the shortening of long final vowels in the colloquial language, but is rather to be seen as a

form of orthographic formality appropriate to the address of a letter.

The motivation for the use of the apex in the documents is far from easy to determine. The picture

may become clearer as more tablets come to light. In the meantime I would tentatively suggest that the

placement of apices may be of some linguistic significance. The constant use of apices on final vowels can

be interpreted as a reaction against the shortening of long vowels, particularly -0, in that position.

72 See Leumann, op. cit. (n. 33), 109-10, Allen, op. cit. '3 Leumann, op. cit. (n. 33), 110

(n. 68), 179-85. 74 ibid.

LANGUAGE OF VINDOLANDA TABLETS

IV. MORPHOLOGY

IV. I. Nominal Morphology

IV.I. I. Ablative singular in the third declension

There are various anomalous (or superficially anomalous) third declension ablative singulars in the tablets, namely -i in consonant stems, and -e in i-stem names of adjectival origin:

(i) -i for -e in consonant stems

312 a Tullioni, 344 ua[let]udini

Moreover a document recently published by A. R. Birley (Inv. 9111022) provides an example of the phrase ab p~tri.~~

(ii) -e for -i in names

18I ab Vitale, 242 <a> Cenale, 263 ab Equest~e, 343 a Fatale, 349 Fatale

These examples contrast with the following:

(i) -e in consonant stems

18I ab Alione, I 8I balniatore, I 92, 207 a Gauuone, 263 centunone, 263 decurjone, 284 dqcu*ne,

343 Gleucone

(ii) -i in i-stems

These variations reflect the fact that there were two inherited singular endings in the third declension ablative, -e and -i,their historical distinction obscure to native speakers. The -e ending had long tended to be generalized to all roots in nouns, whether consonant- or i-stems. But the -i ending was kept alive partly by its persistence in certain i-stem nouns (e.g. febri alongside febre, ciuilciue, classilclasse, collilcolle, finiifine, igniligne, imbnlimbre), and also by the fact that it had come to be assigned to adjectives, particularly those in -is (omnis, omni, et~.).~~

I take first the -e ending in names of adjectival origin (ab Vitale, etc.). Though the adjective uitalis might be expected to have an -i ending, and though Sollemni (31 I) shows that a name derived from such an -is adjective might retain an -i ending, it is likely that in names the ending -e had become the norm. Note already at Pompeii, CIL 1v.3943 Cen'ale, with Charisius, GL 1,124.16, 'Cenale ablatiuo e terminabitur, si homo sit, cereali, si res sit, ut fr~ctus'.~~

All the apparent examples of -e for -i so far attested at Vindolanda are in names, and they probably should not be treated as aberrant at all. It would, of course, be another matter if -e for -i in an -is adjective were to turn up (as e.g. at CIL ~11.2366, 'animo f~rte').'~

It is the forms ualetudini andpatn' which are of particular interest. I would interpret them not as 'vulgar' in any sense (but see below) but as (false) archaisms, and I would base this assertion on the distribution of such ablatives. Since it was -i which, some classes of adjectives aside, was under threat (as, e.g., is shown by a form such as ciue alongside ciui), it may have acquired a reputation for being old-fashioned or more 'correct' than -e, even in consonant stems. It is particularly common in archaizing

inscription^,^^ as for example in the Lex Agraria (CIL 1'.585), which is heavily archaizing in its orthography: e.g.

XXIII ab eo herediue eius is ager locus testamento hereditati deditioniue obuenit

75 Birley, op. cit. (n. I), 441. thirty examples of -i ablatives. Some, but by no means all,

76 Whereas -i is normal in -is adjectives, there is some are in -i stem words. See also Leumann, op. cit. (n. 33),

variation in other types: e.g. par, pan, but uetus, uetere. 435, stating: 'Endungen: -e' vorwiegend bei Substantiven,

On the whole it is true to say that the language had set up und -i vorwiegend bei Adjektiven. Im Altlatein war die

an economical distinction between -e and -i, with the Verwendung nach Ausweis der Inschriften noch nicht so

former allocated to nouns, and the latter to many fest geregelt'. The situation in old Latin is difficult to

adjectives. determine, because most archaizing inscriptions are not

77 Cited by Vaananen, op. cit. (n. II), 84 n. 2. old. But there can be no doubt that -i in nouns of

78 See further Vaananen, op. cit. (n. II), 84. consonant stem was by some considered to be old.

79 See the index to CIL I", p. 819, quoting more than

I00 J. N. ADAMS

Note too in another law (Lex incerta reperta Bantiae, CIL 1'.582):

in poplico luuci . ..
palam luci in forum uorsus

This same lex also has six instances of xs, against three of x. For further examples of -i in archaizing inscriptions, see:

CIL I'.IZI~, monumentum indiciost saxso saeptum ac mannon'
CIL 1'.1615, nomen delatum Naeuiae L. 1. Secunda (sic), seiue ea alio nomini est

The inscription CIL 1'. 1216 (above) has already been cited for certain of its archaizing spellings, notably -xs- and the formmissit. If xs, -ss- in the perfectum of mitto, and -i in consonant-stem nouns all belonged in the eyes of the writer to the same level of (old-fashioned) orthography, then the presence of all three spellings at Vindolanda surely reflects an archaizing taste among some scribes there. Indeed the -i ablative and the -ss- spelling come together in the work of one scribe. The letter 344, which contains ualetudini, is in the same hand as 180, which has ussus. We have already seen (11.5) that another letter, 309, combines -ss- with xs.

Comparable ablatives in -i are also found in poetry: e.g. Enn., Ann. 324 Skutsch sorti, Catull. 68.124, Prop. 11.30.39 ca~iti.~O

Poets no doubt found it metrically convenient at times to use both -e in adjectives and -i in nouns which were not original i-stems.

While aberrant ablatives in -i may often in archaizing registers have been determined by the conventions of the genre, an archaizing tendency does not explain all such spellings. For the sake of completeness I mention two examples found in a verse inscription from Rome, possibly belonging to the end of the second century A.D. (CIL ~1.32808 (=CE 474)). Note first 1. 9:

ista prius triste munus posui dolon repletus

The -iof dolon'must be scanned as a short; the unstressed vowel in final syllable, which ought to be long, has been treated as short. Once distinctions of vowel length were lost in final syllables (see above, III.~), the close and half-close vowels i and e may scarcely have been distinguished in pronunciation in this position, and that might have opened the way for scribes to write indifferently i ore. In this connection

1.5 of the same inscription is even more significant:
dolon' maklno substentauit tempore longo

Here dolon magno is not ablative but accusative, with -m omitted, and .with the o of magno representing the VL merger of 6 and 2i as close q.The use of the -i spelling instead of the -e deriving from -em suggests that the scribe was not clear about the quality of the vowel in final position. Examples of -i for -e in ablatives in inscriptions of substandard Latinity abound,s1 and these may sometimes reflect not a hypercorrect ('archaizing') use of the i-stem ablative instead of the consonant stem form in -e, but a phonetically motivated uncertainty on the part of the writer about the quality of the short vowel in final syllable.

The scribes at Vindolanda were not given to vocalic misspellings (except in hiatus), and it is therefore more likely that forms such as ualetudini and patn arose from a transfer of the i-stem morpheme into consonant stems.

Subliganoncm (346) is a genitive form of subligar. From the plural -aria in such -ar(e) neuters a new nominative in -akum tended to develop by a back-formation. Subliganoncm is thus strictly genitive plural of *subliganum. A parallel is provided by the form cocleanum which emerged alongside cochlear(e) (Appendix Probi 67, 'cochleare non co~liarium').~~

so See 0.Skutsch, The AnnalsofQ. Ennius (1985), 500, naen (I~IO),see e.g. nos 242 funen', 536 pro piaetati,

Kroll on Catull. 68.124 (also on 68.99, on -e for -i in 1493genen, 1556Bautoni, 1566 adulescenti.

adjectives and i-stem nouns in poetry, for metrical s2 For further parallels, see J. N. Adams, 'The Latin of

reasons). the Vindolanda writing tablets', BICS 22 (1975),20.

s1 For examples in E. Diehl's Vulgarlateinische Insch-

LANGUAGE OF VINDOLANDA TABLETS

IV. I .3. Some fomns of the relative pronoun

I start with:  
234  qui feqamus !em- pestates [[et hiem]] etiam si molestae sint  

The preceding clause is not complete, but qui looks to be the old instrumental use of q~i,~~

which is particularly well represented in Latin comedy but survived into the Classical period. A collection of material can be found in OLD, S.V. Qui is common as an interrogative (= 'how') (for example in Horace as well as comedy: OLD S.V.I). For the relative use, see e.g. Ter., Andr. 512, 'multa concurrunt simul I qui coniecturam hanc nunc facio', Varro, De uitapopuliRomani I frg. up. Non. p. 853.1 L., 'cocula, qui coquebant panem, primum sub cinere, postea in forno' (cf. Lucr. 1v.615). Qui may be followed by a subjunctive verb (final I potential I generic), as apparently in the tablet: e.g. Plaut., Aul. 502, 'uehicula qui ~ehar',~~

Stich. 292, 'quadrigas 1 qui uehar', Varro, Rust. 111. 17.9, 'maritumum flumen inmisisset in piscinas, qui reciproce fluerent ips<a>e', Cic., Att. XI. I I .2, 'qua re id quoque uelim cum illa uideas, ut sit qui utamur' ('so that there might be (something) which we might use'). This last example has formulaic phraseology: note Plaut., T?i;n.355, 'habemus et qui nosmet utamur', Cato, Agr. 104.1, 'uinum familiae per hiemem qui utatur', Cic., Att. ~111.23.3, 'magis enim doleo me non habere cui tradam quam habere qui utar'. The relative 1 final use may have had a limited survival in the colloquial language, particularly in one or two formulae.

Qui is used for quis at 215.4: 'si qui uolet'. There is a difference of status between qui forquisbefore s, and that before other phonemes. The former is not uncommon even in literary and poetic texts, particularly in the expression qui sit (e.g. Cic., Vew. 11.188), but the collocation which we have here would probably have been considered substandard by some.8s Vitruvius, who was apologetic about his own Latinity (I. I. 18), used si quis twice in the first chapter of his treatise (1. I .5, I. I. 10), but he thereafter lapsed into si qui (eighteen times) .s6 Si qui occurs particularly often before uolo in Vitruvius, as in the tablet (e.g. 1.4.7, 11.7.5, n.8.8, 1v.3.3, v.5.6, etc.).

The letter of Octavius seems to have an example of the masculine form of the relative used for the feminine: 343.40, 'Frontinium Iulium audio magno licere pro coriafione quem hic comparauit (dena- rios) quinos'. The masculine form eventually subsumed the functions of the feminine (cf. e.g. Fr. qui, masc./fem.), and qui = quae is common in late Latin.87 This example (somewhat undermined by the obscurity of the text) is very early, but note in C. Novius Eunus, T.P. xv111.2.8f. (15 September 39), 'quem suma [= quam summam] iuratus promissi me .. .redtur~m'.~~

IV. I .4. A demonstrative form

The letter of Octavius (343.19) provides an example of the demonstrative illic (with the deictic reinforcement -c(e)),in the form illec = illaec (neuter plural). Illic was primarily an old Latin demonstrative form, very common in Pla~tus.~~

Though it scarcely survived into the late Republic literary language (see Lucr. 1v.1059, Catull. 50.5), there is a group of examples in the Pompeian inscriptions (e.g. CZL IV. 1691, 'qui illunc pedi~at'),~~

and another example of the same form illec in Octavius' contemporary Claudius Terentianus (P. Mich. v111.469.18). Terentianus offers four other cases of illic in various forms.91 The distribution of the word suggests unmistakably that it fell out of respectable educated use but lingered on in colloquial speech.92

IV. I .5. A superlative form

The superlative form (homo) inpientissime is used in the letter 31 I. In this case it would probably not be appropriate to call the usage substandard. The superlative form piissimus is well-attested, but it

s3 For which see Ernout-Meillet, op. cit. (n. 55), 556; See Adams, op. cit. (n. g), 243-4. most obviously found in quicum. See TLL v11.1.370.1off., Adams, op, cit. (n. 17), 45, s4 See W. M. Lindsay, Syntax ofPlautus (1907), 43. F. Neue and C. Wagener, Fomzenlehre der lateinischen

For a comprehensive discussion of quilquis, see E. Sprache"1892-IIJO~), 11, 428-9. Lofstedt, Syntactica. Studien und Beitrage zur histon- 90 See Vaananen, op. cit. (n. II), 86 for further schen Syntax des Lateins (1956), 11, 79-96. examples.

I have consulted L. Callebat et al., Vitruve, De 91 See Adams, op. cit. (n. 17), 45. Architectura, Concordance (1984), S.V. ;see also Lofstedt, 92 See also M. Jeanneret, La langue des tablettes d'exi- op. cit. (n. 85), 11, 92-3. cration latines (1918), 79 for examples in curse tablets.

See, e.g. E. Lofstedt, Philologischerkbmmentarzur Peregn'natio Aethenae (191 I), 132.

I02 J. N. ADAMS

caused unease (see Cic., Phil. ~111.43). Pius: pientissimus may be based on the analogy of beneuolus: beneuolenti~simus.~~

Pientissimus is particularly common in inscription^.^^ It is possible that inpien- tissimus at 31 I is the only example of the negative form attested.95 Its use in a letter is of some interest. Pientissimus occurs mainly in grave inscriptions, and it might accordingly have been assumed that the formation was an artificial one restricted to this special register. The example at Vindolanda suggests that the remodelled superlative was established in colloquial Latin.

IV. I .6. A change of prefix: exungia = axungia

An account (182.16) has the item: 'pretio exungiae (denarios) xi (asses ii)'. The misspelling of axungia is of some interest.

Axungia is a compound (ax-ungia), the first part of which consists of the root of axis 'axle' (= 'axle grease').96 But by a popular etymology this first element was reanalysed as the preposition a(b); hence the forms absungia, assungia at Mulomedicina Chironis 599. By a change of prefix exungia emerged as an alternative to absungia. The verb ausculto 'listen' underwent a similar series of changes. Ausculto was dissimilated to asculto (cf. Agustus for Augustus), a- was interpreted as a prefix and hence the form absculto appeared; finally ab- was replaced by ex-, a change which lies behind Fr. d~outer.~'

Changes, or confusions, of prefix in Vulgar Latin take two forms: (I) phonetically similar prefixes were often confused (e.g. prae- and ~ro-,~~

di(s) and de99); (2) the semantically equivalent, but phonetically dissimilar, prefixes ex- and ab-might also interchange. 'The spelling exungia at Vindolanda is remarkably early; previously the form had been attested in the manuscripts of late technical works, such as Theodorus Priscianus and the Mulomedicina Chironis (5159 887).

1v.2. Verb Morphology

1v.2.I. debunt

There is a substantial group of military reports with the heading renuntium, all of them of formulaic structure.loO In six texts (130, 134, 135, 139, 145, I 50) quidebunt is written. Should this be interpreted as representingquidebunt, qui (ui)debunt, or indeed something else?lol I would now opt forquidebunt. I base myself on the clause-structure of the document-type and on considerations of meaning.

The full format is now clear from an unpublished document (Inv. no. 1418) reported by Bowman and Thomas, 76: '. . . coh. vijii Batauorum. omnes ad loca quidebunt et inpedimenta renuntiarunt optiones et curatores .. . .'

Debunt for debent linguistically raises no problems. The second and third conjugations were conflated in Vulgar Latin, and -unt for -ent is attested, for example, in a letter of Terentianus contemporary with the Vindolanda tablets (P. Mich. v111.468.40 ualunt).lo2 What would be puzzling about debunt, if that is intended, would be its presence in a type of document which was clearly quasi-official and formulaic. The documents are all in different hands.lo3 The same phraseology would not have gone on repeating itself unless an exemplar had been provided for the use of the different optiones.lo4 But if an exemplar was in use, it is remarkable that it should have contained a substandard form.lo5 Even at Bu Njem a century later there was a model-letter form in use for recording the dispatch of goods, which, in its formulaic sections, contained no errors of orthography or syntax. And we have seen evidence earlier for the relatively high cultural standards of at least some of the scribes who were on hand at Vindolanda.

I turn to the interpretation of the immediate context. Ad loca (-urn) must, as the editors suggest

(75), mean 'at duty stations'. If quidebunt were taken as an abbreviation forqui (ui)debunt, the structure

of the passage might be: 'omnes ad loca. qui uidebunt et inpedimenta'. The sense roughly would be: 'all

(are) at duty stations. They will see also to the baggage'. There would seem to be two problems with this

interpretation. First, the use of uideo in this sense, though not impossible (OLD, s.v. 19), would

93 See Leumann, op. cit. (n. 33), 499. 99 See Lofstedt, op. cit. (n. 35), 294-7.

94 Neue-Wagener, op. cit. (n. 89), 11, 208-9. looSee Bowman and Thomas, 73-6 for a discussion of

95 At TLL VII.I ,620.49ff. the form impiissimus is illus- these documents.

trated, but not impientissimus; so at Neue-Wagener, op. 'OISee Bowman and Thomas, 75 on the problem of

cit. (n. 89), 11, 206. interpretation which this locution raises.

96 See Ernout-Meillet, op, cit. (n. 55), 62 (s.v. axis I). lo2See Adams, op. cit. (n. 17), 51.

97 See, e.g. Bloch-von Wartburg, op. cit. (n. 51), 211 lo3Bowman and Thomas, 74.

S.V.&outer. lo4Onoptiones as authors of the documents, see Bowman

98 Note, for example, propositus =praepositus at 0.Bu and Thomas, 74.

Njem 84,85 (see further J. Svennung, Untersuchungen zu lo5Note that the correct form debent is found in the

Palladius und zur lateinischen Fach- und Volkssprache account 181.

('935)B 378).

'03

certainly be somewhat unusual. But more worrying is the connective use of the relative qui. There is something of a literary flavour to the connective relative (= (et) hi), which is not appropriate to the document.

If on the other hand debunt is taken to be a substandard form for debent, good sense and syntax are introduced. Qui becomes a normal restrictive relative, with omnes as its antecedent: 'all who should be are at duty stations, and the baggage as well'. There is a special reason for taking the expression thus. When an infinitive has to be understood with debeo (as would be the case here: understand esse), it is particularly common for debeo to stand either in an ut- (= 'as') clause, or in a relative clause (e.g. Cic., Clzdent. 185, Cat. 1.16, Mur. 3, Q. Cic., Pet. 4).'06 Qui debunt = debent would therefore have a decidedly idiomatic look to it.

If it is the substandard form debunt which recurs in these texts, we may speculate about how it got into formulaic documents. The answer may lie in the editors' suggestion (74) that the documents were written by optiones themselves. The exemplar may have been drafted not by one of the well-educated scribes who served the high-ranking officers, but by an optio. An item of soldiers' vernacular might thus have entered the model document and been perpetuated in later documents because it represented standard usage among those of the rank responsible for dispatching such reports. It is of considerable interest that the form habunt forhabent has now turned up in an unpublished document (Inv. 9311544) written by a decun'o.'07

If the above speculations are along the right lines, they have important linguistic and other implications. That the form should have been repeated by a variety of optiones, with no-one ever substituting the 'correct' form debent, suggests that -unt for -eat, despite its poor attestation in writing, was standard usage in the army in the speech of under-officers; and it is worth recalling that in Egypt at the same time the soldier Terentianus said ualunt for ualent. It is highly unusual to be able to locate a substandard usage so precisely on the social scale. Secondly, it would seem to follow that not all exemplars for use in the army were drafted by professional scribes of superior education. Some initiatives were left to minor officers themselves; and literacy extended beneath the level of the social Clite.

117.2.2. rescnpsti (310.6)
I mention this form under verb morphology, but in fact it displays a commonplace haplol~gy.~~~

V. WORD FORMATION

In this section I discuss various terms classified by suffix.

Sagacia has now turned up four times in the tablets:

184.20  sagaciam (denarios) v (asses iii)  
207.3  a Gauuone   
 sagh n(umero) iji   
 sagacias n(umero) vii   
2553-9   sa
 gacias sex saga [c. 3 pallio-]  
 fa septem tu[nicas se]x   
Tab. Vind. 1.44 sagacias duas   

Sagaceus must in origin have been an adjective of the -aceus formation; hitherto only one example of the word used adjectivally has been attested (at Col. XI. I .21: see below). There are now sufficient examples at Vindolanda to make it clear that the word was also used as a feminine substantive signifying a type of garment; twice indeed it is immediately juxtaposed with saga 'cloaks', the term which provides its base. Whatever thesagacia was, it was presumably associated or worn with thesagum. The gender of the substantival use can only be explained from the deletion of a feminine noun (uestis, tunica?). The question arises what is the implication of the suffix -aceus in such a combination.

lo6 See TLL v.1.97.33ff. lo8 For examples, see Leumann, op, cit. (n. 33), 234,

lo' 1am grateful to Professor A. R. Birley for supplying 598, Neue-Wagener, op. cit. (n. 89), 111, 50-5. me with a text of the letter.

IO4 J. N. ADAMS

Most -aceus adjectives are adjectives of material.log The base is usually, but not always, a plant name (e.g. betaceus, fabaceus, rosaceus etc.). For different types of base, note coriaceus 'made of leather' (TLL1v.950.48ff.) and membranaceus (sometimes = 'made of parchment': TLL v111.631.82ff.). As far as can be deduced from the material assembled by Gradenwitz,ll0 such extended uses of the suffix with non-plant names as base were rare.

Could sagaceus be an adjective of material? Cloaks (saga) were often recycled and converted into different objects. ll1Could it be that sagaciae (tunicae?) were utilitarian tunics(?) converted fromsaga? This possibility cannot be ruled out.

But there is another possibility. Not all -aceus derivatives are adjectives of material in the strict sense, as I have argued elsewhere:l12 'The function of the suffix was .. . widened to mean vaguely "belonging to, pertaining to, suited to" ';l13 alternatively the suffix could denote the idea of resembling the object signified by the base. The one attested literary example of sagaceus suggests that it was indeed to this extended category that sagaceus belonged: Col. x1.1.21, 'frigoribus et imbribus, quae utraque prohibentur optime pellibus manicatis et sagaceis cucullis'. Bowman and Thomas quote the passage stating that 'the last two words must refer to cloaks with hoods attached to them'.l14 That may be the form of attire to which Columella alludes, but etymologically the phrase could not be given this meaning. The sense would have to be 'hoods associated with, worn with, saga'. It is then possible that sagaciae indicated items of clothing (tunicae?) typically worn withsaga. The juxtaposition of sagaciae twice with saga suggests that the two went together.

The OLD interprets sagaceus at Col. x1.121 as meaning 'made of the material of sagum' (see our first interpretation above). On the available evidence it is impossible to determine decisively which of the two interpretations of sagaciae offered here is correct. The key to an understanding of the type of garment to which sagacia refers undoubtedly lies in the feminine noun which determined the gender of the substantival adjective.

The suffix -adus was particularly productive at all periods, and is represented in all Romance languages (e.g. Fr. -ier, It. -ajo, Sp. -ero). It furnishes adjectives, masculine and feminine substantival adjectives designating persons whose occupation centres on the object expressed by the base, and neuter nouns. There is an abundant corpus of examples of all three types in the Pompeian inscriptions,l15 which bears witness to the importance of the suffix in everyday life. At Vindolanda likewise it is by far the best represented adjectival suffix. I first list examples and then discuss a few interesting cases.

(i)
adjectives:
carrarius, necessan'us, regionan'us

(ii)
nouns indicating occupations, both military and civilian:
military technical terms: benefican'us, corniculan'us, duplicanus, tesseranus, ~exilla~us
other :balnean'us(?), bubulcan'us, ceruesan'us, scutan'us, ueten'nan'us

(iii) neuters:
compendian'um, lateran'urn(2) , locan'um, ouan'urn, panan'um, sudan'um, superan'a

carrarius: found at Vindolanda twice as an adjective, in the expression ax(s)es carran'os (185.20, 309.5)~which must have been a technical term for 'carriage axles'. At 185 the axles belong not to a carrus, but to a raeda; it follows that the expression had a generic sense (= 'carriage axles' in general, as distinct from 'axles of a carrus'). Hitherto carrarius had not been attested as an adjective before the medieval period;l16 it was known as a noun, = 'carriage-maker, driver' (TLL 111.497.41ff.)."~ The Vindolanda examples reveal the true range of the term. One might compare raedan'us, which had both an adjectival and substantival use in the classical period (Varro, Rust. 111.17.7, 'raedarias . . .mulas', Cic., Mil. 29, 'raedarium occidunt'), and also cawucan'us (for the adjectival use, see Ulp., Dig. xx1.2.38.8, 'carruca- rias mulas').

lo9 See Leuminn, op. cit. (n. 33), 287. 112 Adamsap. Bowman and Thomas, op. cit. (n. 48), 74. 110 0.Gradenwitz, Laterculi Vocum Latinarum (~goq), 113 ibid.

482. 114 op. cit. (n. 48), 142.

11' I owe this information to Dr Wild. See J. P. Wild, 115 See VaanPnen, op. cit. (n. II), 91-5, 'Vindolanda 1985-89: first thoughts on new finds', in L. 116 See R.E.Latham, Revised Medzeval Latin Word-

B. Jdrgensen and E. Munksgaard (eds), Archaeological List from British and Irish Sources (1965), 73 (uia car-
Textiles in Northern Eurobe :Rebort from the 4th ArESAT raria. c. I 2 I 6).
Symposium 1-5. May ~gboin 'copenhagen (I~~z),72;
117 see also R. 0. Fink, Roman hlilitary Records on
idem, 'Vindolanda 1985-1988, The Textiles', in Vin-Papyrus (1971), 58.ii.6.
dolanda, Research Repovts, new series. 111. The Early
Wooden Forts, 89 n. 23 on a cloak converted to a tunic.

regionanus: probably at 250.8: (centunoni) ~egiona+. Perhaps the first example of the title centurio regionan'us: cf. RIB 152, CIL x111.zg58.l~~ bubulcan'us: hitherto attested only at CGL 11.259.44, where it is glossed by fiow.cqs 'ploughman'. See now 180.9, 'bubulcaris in siluam m(odii) viii'.

The meaning of the term is not absolutely certain. Suffixal derivatives in -an'us, as noted above, often designate an individual who deals in or makes or is in charge of or attendant upon the object denoted by the base (e.g. subsellan'us 'maker of subsellia', cemesan'us 'brewer', tabernan'us 'one who works in or is in charge of a tabema', sumptuan'us 'one who is in charge of household expenses', horrean'us 'superintendent of a horreum', etc.). If bubulcan'us stood to bubulcus as horrean'us to howeum, then the bubulcan'us might have been the superintendent of bubulci.

But the fact that bubulcatius is here in the plural raises doubts about such an interpretation: it does not seem plausible that more superintendents than one would be needed to supervise bubulci working in asilua. An alternative possibility is that bubulcan'us is synonymous with bubulczis. If so it would belong with a substantial number of purely augmentative terms in -an'us, in which the suffix merely lengthens the base-term without contributing any discernible semantic nuance. I list a few examples below: cataphracta~ius (late Latin, including four examples in the Historia Augusta) =cataphractus, 'mailed' (<xaz&+eaxzo~),which is classical (e.g. Sallust, Livy). Both cataphractus and cataphractan'us had a substantival (TLL 111.592.1ff., 55ff.) as well as an adjectival (591.71ff., 592.39ff.) use. manifestan'us (e.g. Plaut., Bacch. 918, 'moechum manufestarium'; cf. Poen. 862, 'manufesti moechi'). subitan'us (e.g. Plaut., Mil. 225, 'res subitaria est'; cf. Curc. 302, 'res subita est'). praesentan'us (e.g. Plaut., Most. 361, 'argentum .. . praesentarium'; cf. Poen. 89, 'praesenti argento').

There were good analogies for the formation of bubulcan'us. A number of terms denoting keepers of animals had this suffix (e.g, asinan'us, burdonan'us, iumentan'us, asturconan'us, camelan'us). These had the name of the animal as their base, but they could still have motivated the augmenting of bubulcus.

cemesan'us: found at 182.14 (Atrectus ceruesar[ius), where it can only indicate a person practising a profession, i.e. = 'dealer in, brewer of, beer'. Against the only example quoted by the TLL (CIL

XIII. 10012.7) it is observed 'siue de ceruesae potatoribus siue de coctoribus ...cogitandum est', but the first meaning can be ruled out. Bowman and Thomas (133) cite the revealing case negotiatorcemesan'us (AE 1928.183). Cf. Fr. cemoisier (since 1260)."~

uetennanus: the standard designation for a veterinarian in the early Imperial period,120 found twice at Vindolanda (181.7, 310.11); ~ompendia~um

: found at 1g4.B. I in a text which is 'a list of household objects which are almost all related to cooking, eating or drinking' (Bowman and Thomas, 162) ;the find-spot was probably a kitchen. Neuters in-an'um often signify receptacles, e.g.granan'um,poman'um, andat Vind~landapana~um

and ouan'u~z.'~~

One possibility is that a compendian'um was a receptacle for a compendium, 'savings, a saving'. What a compendium might have been in the context of the kitchen is not immediately obvious. locan'um: a significant new attestation of a very rare word: 185.24, 'Cataractonio locario (denarii) ~(emissem)'. There is only one example of locahum (neuter) noted at TLL ~11.2.1554.60 (Varro, Ling.

v. 15:see below), but the word is widely reflected in the Romance languages with the meaning 'rent', and it must therefore have been commonplace in spoken Latin (e.g. Fr, loyer, Cat. lloguer, Sp, aloguero, Pg. al~guer).'~~ with a place-name, can plausibly be given the

The Vindolanda example, juxtaposed

meaning 'rent'. The editors (141) note the series of place-names in the document, and suggest that it may

be an account of expenditure incurred on a journey.

The passage of Varro runs as follows: Ling. v.15, 'in<de> locarium quod datur in stabulo et

taberna, ubi consistant'. This passage has, I believe, been persistently misinterpreted in the lexicogra-

phical tradition.123 According to the TLL (~11.2.1554.60) locan'um here does not signify rent paid by

guests, travellers, but by tradesmen (hiring a stall in the market-place): 'pretium, quo taberna sim.

locatur (cauponibus sim. potius quam hospitibus devertentibus)'. So too Lewis and Short, 'rent paid for

a stall to sell goods from'; cf. FEwv.390. It is hard to see how the use of in +ablative would square with

this view. The OLD continues the tradition ('rent paid for a stall in a market'), but does allow an

alternative interpretation, 'or perh. = payment for accommodation in an inn or sirn.'. There is a similar

ambiguity about R. G. Kent's rendering (Loeb), '. .."place-rent", which is given for a lodging or a shop,

where the payers take their stand'. It is no doubt the presence of consistant (lit. 'take up a position')

118 See further Bowman and Thomas, op. cit. (n. 48), lZ2 See FEW, op. cit. (n. II~), v, 390.

110. lZ3 The editors, on the other hand, are undoubtedly 119 W. von Wartburg, Franzosisches etynzologisches correct in stating, 144: 'the word can be interpreted as a Worterbuch 11.1 (1940), 613 See also G. N. Olcott, charge for lodging or accommodation either for people or Studies in the Word Fomation of the Latin Inscnptions for animals'. See further J. Collart, REL 26 (1948), 61, (1898), 147f., giving the sense as 'brewer'. noting the two interpretations which I mention below,

lZ0 See J. N.Adams, 'The origin and meaning of Lat. and adding a third of his own.
ueten'nus, ueterinan'us', IF 97 (1992), 7-5.
lZ1 See too the examples of such terms at Pompeii collec-
ted by Vaananen, op. cit. (n. II), 95.

I06 J. N. ADAMS

which has led to the view that Varro was referring to the fee paid by market-traders. But consisto also possesses the meaning 'stay, break a journey': see OLD s.v. 2 'to break one's journey, make a stay': e.g. Cic., Vew. v.28, 'ex iis oppidis in quibus consistere praetores . .. soleant'. On this interpretation stabulum and taberna would signify different types of lodging-places,124 and the sense of locan'um would be exactly that of the Vindolanda tablet, and of the Romance reflexes. In stabulo in Varro becomes perfectly understandable. This is not the first case we have seen of a usage at Vindolanda falling into line with evidence derivable from the Romance languages (see 11.7 on turta; also below, v.6 onexcusson'um).

superan'a: used as a substantival feminine at 184.2 (superan'as (denan'os) xiii), deriving no doubt from ellipse of uestis (cf. perhaps sagacia) :note CGL IV. 180.I 5, 'uestis que superinduitur'. Superan'a had previously been well attested in glosses (cf. CGL v11.318, index s.v.), but the Vindolanda evidence now takes its coinage well back before the date of g10ssaries.l~~

Represented by caligan's(?), dextralis(?), legionan's, uentrale(?), umerale(2).

caligan's: in the expression claui caligares at 186.8. Clearly technical: note Plin., Nut. 1x.69, 'squamis conspicui crebris atque praeacutis, clauomm caliganum effigie', :Vat. xx11.94, 'si caliga~s clauzls . . . adfuerit'.

legionan's: at 180.22 militibus legionan'bus. An unambiguous example of a change of suffix, -an'us >an's. Legionan's is elsewhere hardly attested: TLL v11.2. I 109.60 cites only a variant reading at Caes., Ciu. 111.2.2. There was a good deal of interchange between the suffixes -an'us and -an's, both in military tern~inology and in other areas of the 1exic0n.l~~

Olcott notes that 'in the military language . ..-ans is readily forced into the position of -an'us, (alan's decurio, auxilian's miles, commanipulan's, . . This is undoubtedly true, as our example shows, but equally it is easy to find examples of -adus for -an's in military terminology. A notable case i~pnmipila~us

forpn'mipilans (see HA, Pesc. :Vig. 2.4, Did. Iul. 5.I), commented on atAppendix Probi 69. Note also (e.g.) duplarius for d~plaris,~~~sesquiplicam'us

for sesq~iplican's.~~~

The interchange operated in both directions, and not only in military language. For non-military examples of both types see, e.g, balnean's = balnean'u~,~~~peculiarius = pec~lian's,~~~ pulican'us =pulican's,132 simplan'us =simplan's (Dig. XXI. I .48.8), and urceolaris alongside urceolanus

(e.g. Pelagonius 39/37).

v.4. Diminutives

There are seventeen diminutives in the tablets: can-ulum, castellum, Jliolus, jlanzmula, frater- culus, gallicula, modiolum ,ofella, palliolum, pellicula, porcellus, ratiuncula, scutula, siluola, tensiun- cula(?), Brittunculi, ungella. I omit some conjectural terms. The frequency of diminutives is testimony to the productivity of the formation both in technical vocabularies (note, e.g. jlamnzula, 133 gallicula, modiolum, palliolum, scutula) and in colloquial speech (e.g.Jliolus, fraterculus). I offer comments on a few of these terms.

ungella: at 233A.3, in a list of foodstuffs; in the extant text the term has no further specification. Culinary terms, particularly those designating parts of animals eaten as meat, were often diminutives. Note Celsus 11.22. I, 'quad fere quidem in omni domestica fit, praecipue tamen in ungulis tmnculisque suum, in petiolis capitulisque haedorum et uitulorum et agnorum, omnibusque cerebelli~'.~~~

The force of ungella is difficult to grasp. Not infrequently, as in the tablet, the word occurs without specification of the type which might pin it down to any particular animal: e.g. Marc., Med. 20.27, 'statim dare debes lentem et betam coctam uel alicae ius uel oua apala uel ungellam discoctam uel cocleas elixas' (cf. 20.26). It seems generally to be taken to indicate the trotters of a pig.135 Grounds for assuming such a specialization might seem to come from Apic. IV.j.2, 'iecineraporcelli etgallinamnz et ungellas et ascellas [VE]diuisas'. The ascellae 'wings' would appear to belong to the gallinae, the ungellae to the porcellus, but the issue is confused by the fact that aucellas (Humberg) is an easy conjecture accepted by Milham (Teubner). At Apic. v11.1 ungella occurs twice (in the heading and at

lZ4 See T. Kleberg, Ho*tels, restaurants et cabarets duns Olcott, op. cit. (n. I ~g),164.

I'antiquite' romaine (rg57), 18-19 (stabulum), 19-20 133 Bowman and Thomas, 354 on no. 463.

(tabema). 134 For cerebellum as a culinary term, see J. AndrC, Le

lZ5See also Bowman and Thomas, 138 ad loc., citing 4. vocabulaire latin de I'anatomie (1991), 34-5, and for

Bruckner and R. hlarichal, Chartae latinae antiquiores capitulum, J. N. ildams, 'Anatomical terms transferred

(1954-), "1.204.4. from animals to humans in Latin', IF 87 (1982), 106. See

lZ6 See (e.g.) Baehrens, op. cit. (n. 52), 121. also below on codicula 'tail' of a pig, and labelli, found at

lZ7op. cit. (n. II~), Apicius VII.I.

183.
lZsSee Olcott, op. cit. (n. II~),151, TLI,v.1.zz58.gff. 135 See J. Andre, L'alimentation et la cuisine a Rome
lZ9See Olcott, op, cit. (n. II~),

168.

('981)s '37 n.43.

130 See TLL 11.1703.27ff.

131 Olcott, op. cit. (n. II~),
162.

1.5) in a chapter which seems to deal with pork, though there is no specification. The heading reads 'uuluae steriles, callum, libelli, coticulae et ungellae', where AndrC136 plausibly emends libelli to labelli and prints an earlier emendation codiculae ('tails') for coticulae. Vulua is the delicacy sow's matrix, and there is therefore a likelihood that the other terms indicate edible parts of the pig. Diocletian's Prices Edict (4.12) also provides a hint, although an equivocal one, that ungella may have been especially applicable to the pig: 'ungellas quattuor et aqualiculum pretio, quo car0 distrahitur'. Aqualiculus, though it came to be widened in meaning, was once the voxpropna for the pig's maw.137 But what is one to make of Apic. 1.9, 'callum porcinum uel bubulum et ungellae coctae'? Here callunz 'crackling', seen above in the passage apparently about pork, comes either from the pig or from the ox; are the ungellae exclusively those of the pig? And at Physica PliniiF1or.-Prag. 11.23.3 the ungella is surely not that of the pig: 'sero cum ungella aut cum pede de ansere'.

Further evidence which is possibly of relevance is provided by the terms acro and Etxeov6te~ov.l~~ Acro, which is qualified byporcinus at Pel. 85, Veg. 11.130.2, Pel. 463, Mul. Chir. 201, was equivalent to Gk. &x~ov&~tov,

as a comparison of Pel. 85 and Veg. 11.130.2 with Corpus Hippiatn'con~m Graecoruttz

11.73.11 &x~ov&~tov

indicated the foot, not only of the pig, but of at least one other small domestic animal, the kid, as can be seen from a comparison of Corp. Hipp. Graec. 11.106.1 kei+ou xe+crhqv pet&TQV Etx~ovcreiov with Veg. 1.56.17, 'caput haedinum depilatum cum pedibus suis' (also 111.8.1).140 AndrC therefore suggests (op. cit., 46) that acro (=&xeov6etov ) denoted 'le pied (i.e. 1'extrCmitC de la patte) de petits quadrupkdes domestiques, porc, mouton et chkvre'. Could it be that acro and ungella were synonymous? The trotters of various small animals may have been used in both culinary and medical recipes. It is possible that acro and ungella tended to be subtly distinguished: acro being the generic term, ungella tending to be specialized to the pig but capable of a more general meaninp.141

~

0

ofella: at 203.2 in a shopping-list or menu for a single meal (so the editors, 174). There is no doubt that originally ofella denoted a piece, a mouthful, to be eaten. There are revealing examples of the other diminutive of offa in veterinary texts. At Pelagonius 296 offula signifies a lump of wheat flour formed with the use of water. At Vegetius IV. 12.2 it is a lump of bread. And at Vegetius 11.134.7 an offula is the size of a nut, and composed of a variety of non-meat substances. It was the form of an offula which was distinctive, not its components, despite the statement at TLL 1x.z.530.56 that it usually signified a piece of meat. So it was with ofella, though in this case a specialized reference to meat can be accepted. The editors, following D~nbabin,~~~

tentatively translate as 'pork cutlet'. Dunbabin, basing himself on the recipes found at Apicius v11.4, says that 'ofellae were of pork . . .;they were cutlets, not chops, for they would be made from the paunch'. However, Dunbabin goes on to acknowledge that 'in some of the recipes there is nothing to indicate what meat was used; this may have been pork, veal, or lamb, but not fish, since fish are dealt with in a separate book'. Though an ofella might be of pork, that was not a defining characteristic. What was distinctive was that the ofella was a piece (of meat), which might be shaped: note Apicius v11.4.2, 'ofellasexossas, in rotundum complicas, surclas, ad furnum admoues' (the ofella is boned and rolled up). As Dunbabin notes, this instruction is suggestive of Martial's use of the adjective curua with ofella at x1v.221. I. Whereas an offula was a shaped ball of any edible substance, an ofella was a piece of meat which might be rounded. That an ofella was little more than a mouthful is suggested by its use at Juvenal I I. 144 and Mart. x.48.15, x11.48.17, in all of which passages it is implied that it was humble or insubstantial fare.

cal-mlum: in the plural at 315. Again the Vindolanda material allows an expansion of the early

Imperial lexicon, as the diminutive has hitherto been attested only in the Digest (xv11.2.52.15). Its

neuter gender shows that it was based on camm =canus, which is probably attested at 343.17, and was

undoubtedly widely current: note Bell. Hist. 6.2, and Nonius p. 287.24L., 'carra neutri generis esse

consuetudine persuasum est' (see further TLL 111.499.41ff.). Camus would have been drawn into the

neuter on the analogy of such neuters as uehiculum, plaustnim, carpentum, and petonitum.

modiolum :probably signifies the nave of a wheel at 309.4, in a series of technical terms to do with

vehicles (see, for the sense, Vitr. x.9.2, Plin., Nut. 1x.8, Edict. Diocl. 15.3). In thissense the word is well

represented in the Romance languages (e.g. OFr. moieul; particularly in Rheto-Romance and dialects of

Friuli: see FEU' (op. cit. (n. I 19), v1.3. I I). The neuter is scarcely found, and then only late (see the

editors ad loc.), but the base-word modizls is attested in the neuter as early as Cato (Agr. 58, 'salis

unicuique in anno modium satis est'). It is possible that masculine (inanimate) technical terms tended to

develop a neuter by-form. In the same document (309.7) radius occurs, probably for the first time in

extant Latin, in the neuter plural (= 'spokes (of a wheel)').

136 J. AndrC, .4picius, L'art culinaire, De re coquinana 140 ibid.

(1965), 186. 141 See also W. Heraeus, Meine Schnyten (ed. J. B.

137 See Adarns, op. cit. (n. 134)~100. Hofrnann, 1937), 24-5.

138 See J. AndrC, 'Notes de lexicologie', RPh 40 (1966), "= 2.L. Dunbabin, 'Notes on Lewis and Short', CR 49

4658. ('935)> '0.

139 See AndrC, op. cit. (n. 138), 46.

108 J. N. ADAMS

gallicula: at 197.2, in a fragment mentioning two items of footwear. Gallica is rather better and earlier attested than its diminutive (Cic., Phil. 11.76, Juv. 7.16 forgallica), which is mainly in late Latin; but see RIB 1.323.

There is one neologism in -ti0 in the tablets, coriatio at 343.40: 'Frontinium Iulium audio magno licere pro coriatione quem hic comparauit (denarios) quinos', 'Frontinius Iulius is asking a high price in return for the coriatio which he bought here for five denarii apiece'(?). The sentence is very difficult to interpret.143 Licere in this sense complemented by pro is unparalleled. Why is the masculine relative quem used when its antecedent seems to be the feminine coriatione? Is (denanos) quinos a sort of accusative of price (see below, v1.1.7)? The reading of coriatione is clear enough. As a verbal noun it must be based on a verb corio(r) 'make con'um, leather', which is not attested until the late medieval period.144 Like numerous abstracts in -tio, particularly in technical vocabularies, con'atio has apparently passed from an abstract to a concrete meaning ('leather making' >'leather goods').145 If it had already undergone a semantic change, it must have been in existence for some time, though hitherto nowhere attested. This case provides a salutary warning against making the assumption that extant literary Latin does justice to the range of technical vocabulary which must have existed at a subliterary level.

The letter of Octavius has, in addition to coriatio, another substantival suffixal derivative with a technical meaning hitherto unattested, namely excussorium =area, 'place where threshing takes place': 343.27f., 'ut possjm spicam habere in excussorio'. The sense emerges from the context, where both excutzo (1.25) andperexcutio (1.29) are used of threshing.146 The adjectiveexcussorius is found at Plin., Nut, xv111.108 in an unrelated sense. Our neuter nominal use of excusson'um belongs with a group of such neuters denoting rooms, places for specialized activities (e.g. dovmitorium, auditon'unz, gustato- riunz, etc.) The existence of excussorium and its association with threshing might have been deduced from certain Romance languages, 148 though its reflexes signify instruments for threshing rather than the place where the activity was carried out. Nevertheless excusso~um in the specialized sense attested at Vindolanda undoubtedly belonged to a living technical vocabulary.

VI. SYNTAX

VI.I.
Cases and Prepositions
VI.I.
I. A use of the locative
Various letters contain on the back an address of the form (310)

Londini
Veldedeio
equisioni co(n)s(ularis)
a Chrauttio

143 See Bowman and Thomas, 328.

144 See Petersmann, op. cit. (n. 5), 288, referringto J. F. Niermeyer, Mediae Latinitatis Lexicon minus (1976),

s.v. For con'atio Petersmann cites Latham, op. cit.

(n. I 16), s.v. An example of coriatio mentioned there is dated to the fifteenth century, but no reference is given. Greater detail can be found in the full Dictionary of Medieval Latin from British Sources (ed. R. E. Latham, fasc. 11, 1981), 491: 'coriatio, (?)covering with leather', citingoneexample (Fabr. York 27, A.D. 1404). It would be hazardous to assume that there could be a direct line of descent linking an isolated fifteenth-century British exam- ple of a noun not reflected in the Romance languages, and our second-century example. The medieval example may be a very late, neo-Latin coinage. In any case, the attesta- tion of the word in the fifteenth century does nothing to assist in the elucidation of the numerous difficulties which surround coriatione in the Vindolanda tablet.

145 For such abstractapro concretis in technical vocabu- laries, see e.g. A. Onnerfors, Pliniana. In Plinii Maioris Naturalem Historiam studia grammatica semantica cri- tics (1956), 12-13, J. B. Hofmann and A. Szantyr, Lateinische Syntax und Stilistik (1965), 749-50. Hofmann-Szantyr (749) cite as an early example habitatio = 'dwelling' at Plaut., Most. 498, Cato, Agr. 128.

146 For excutio in this specialized sense, see A. K. Bowman, J. D. Thomas, J. N. Adams, 'Two letters from Vindolanda', Britannia 21 (~ggo), 49-50. Excutio has Romance reflexes with this meaning (op. cit., 50).

14' See Bowman, Thomas, Adams, op. cit. (n. 146), 50; Leumann, op. cit. (n. 33), 301, on the formation. 148 See Meyer-Liibke, op. cit. (n.49), 2997, FEW, op. cit. (n. II~), 111, 286-7. 149 See in general Bowman and Thomas, 43-5.

Veldedeio is obviously the addressee, but what is the force of the locative? Does it signify the place where the addressee is, or the place where the letter was written? Bowman and Thomas (44-5) argue for the former, and they are undoubtedly right. They draw attention to some letters from Vindonissa where the locative is preceded by dabis;150 the address can therefore be seen as an instruction to the carrier to hand the letter over at a certain place.

To the arguments marshalled by Bowman and Thomas I would add one further, decisive, piece of evidence. Claudius Terentianus, writing to Claudius Tiberianus, at one point instructs his addressee as follows:

P. Mich. ~111.467.25 et si scr[i]bes mihi epistulam inscribas in libuma N[e]ptuni

If you write me a letter, address it "On the ship Neptune".

In libuma Neptuni is clearly the address which is to appear on the letter. It is a locatival expression (= 'at, or'), not a directional (= 'to'), and it unambiguously refers to the place at which the addressee, Terentianus, would be found.lS1 Here is unequivocal evidence that the address was expected to take the form of a locatival expression (indicating the place where the recipient would be), rather than a directional of the sort which might have seemed more logical to English speakers.

The convention may well have originated from the abbreviation of a fuller form of instruction to the bearer (e.g. trade, dahis + locative + dative of name: see above). A locative in an ancient letter may, of course, indicate the place at which the letter was written, but such a locative would be at the end of the letter proper, not on the address side. Alternatively an address such as Londini Veldedeio might just be derived from an adnominal locative of the structure alicui Romae, with the locative later promoted to focal initial position because the bearer had first to get to the place before finding the addressee. The editors cite (44n. 24) a Carlisle stilus tablet with the order alicuiR~mae,'~~

and to that might possibly be added CIL 1v.879, 'M. Lucretio flam. Martis decurioni Pompei(s)'.lS3

It is also worth recalling that locatival expressions tended to take on a directional meaning in Vulgar Latin. From this same period we haveAlexandn'e (= -ae) used a number of times by Terentianus in the sense 'to Alexandria' (P. Mzch. v111.471.15, 22, 25, 32, 33),154 and even in Cicero a locatival may come close to assuming a directional nuance. Note, Att. x1v.2.4, 'in Tusculanum hodie, Lanuui cras, inde Asturae cogitabam', where two locatives are in alternation with the directional in Tusculanum. As in the letter address, so too here the reader is left to supply the verb, which makes the locative less harsh (= 'I am planning to go to Lanuvium tomorrow and to spend the night there'). In the address the implication is 'destination to London, for Veldedeius who is at that place'. Notable in the Ciceronian example is the 1ocativeAsturae after inde; contrast Att. x1v.7.1, 'ut inde altero die in Puteolanum'.

I move on to some other structural features of addresses. If we leave aside the locative Londini which begins the address of 310, the rest of the address takes the form of a dative (= the recipient) followed by a + ablative (= sender):'Veldedeio equisioni cos, a Chrauttio fratre'. The structure of this address represents a reversal of the structure of the initial greeting, which names the sender first (in the nominative), followed by the recipient (in the dative), with salutem 'greetings' at the end: 3 10 (start), 'Chrauttius Veldeio suo fratri . .. salutem'.

The same variation between the ordering of the initial greeting and that of the address on the reverse recurs at Vindolanda: e.g. in 255, '[Cllodius Super Ceriali suo salutem . . . Flauio Ceriali praef, a [C]l[o]dio Sugero' (cf. 291, and probably also 260, 263, 292, 312, where the start is missing). The motivation of the reversal of order at the end is clear: in the address on the back the name of the recipient is promoted to the focal initial position, as is appropriate on the outside of a letter. It follows that a locative which precedes the name of the recipient is itself also bound to be focalized, i.e. to represent the destination of the letter. It is the very essence of such addresses that the name of the sender is demoted to final position, and it is accordingly structurally inconceivable that the place at which he was writing

. ... .

should be given prominence.

The address-structure of 255 and the other Vindolanda letters above was not a convention peculiar to Vindolanda. Note P. Hibeh 276 (= CPL (op. cit. (n. 67), 260)), which begins 'Iulius Repositus C1. Germano suo salutem', and on the verso has the familiar change of construction: 'Cl. Nidio Germano a Iulio Reposito coll(ega)'. A letter of Terentianus (P. Mich. v111.471) has on the back 'Claudio Tiberiano

lS0 SO Terentianus, P.Mich. v111.468.68 appears to have lS3 Cited by J. Svennung, Anredeformen. Vergleichende

the address trade Claudio Tiberiano (restoration). Forschungen zur indirekten Anrede in der dritten Person

lS1 For the type of genitive to which Neptuni belongs und zum Nominativ fur den Vokativ (1958), 23; also P.

(gen. definitivus), see Hofmann-Szantyr, op. cit. Cugusi, Evoluzione e fome dell' epistolografia latina

(n.1451, 62. nella tarda repubblica e nei pn'mi due secoli dell' impem

lS2 See M.W.C. Hassall and R. S. 0.Tomlin, 'Inscrip- (1983), 65 n. 104.
tions', Bn'tannia 19 (1988), 496, no. 32. lS4 See Adams, op. cit. (n. 17), 38.

I10 J. N. ADAMS

[pat]r[i a Cla]ud[io] Teren[tiano'. The start of the letter is missing, but it would undoubtedly have had the pattern Claudius Terentianus Claudio Tiben'ano . . . salutem (cf. nos 467, 468, 469).

VI. I .2. Ablativellocative

In classical Latin, first declension locatives have the -ae ending, second declension (singulars) -i.In the tablets the first declension form is intact:

190.38 Brigae man[serunt]

292.c.2 Brigae man~ural~~

In the second declension, however, there has been a shift to the (locatival) ablative singular:

154.9 Londinio
185.23 Isurio
185.24 Cataractonio

250.9 Luguualio

343.16 Cataractonio
Inv. 88.836, p. 364 Cataractonio

To these examples could be added:

154.6 officio Ferocis ( at the office of Ferox )

The old -i morpheme is found twice, at 310 (Londini:cf. Londinio above), and in a stilus tablet Inv. no. 575 Eburaci (see Bowman and Thomas, 44). The use of -o alongside the old locative -ae confirms a development of the 1ang~age.l~~

The frequency of the place nameRomae kept the -ae morpheme alive for much longer than -2. Although some Romance place-names reflect original locatives in -i (e.g.Brindisi <Brundisi, Girgenti <Agrigenti),the encroachment of -o on -i is well attested from Vitruvius onwards. Such a trend is particularly understandable in a region where place-names were of non-Latin origin and only recently assimilated to the Latin declension system; in an old name of Latin origin, on the other hand, the retention of -i might be supported by long-standing usage (e.g. Brundisi; perhaps too Londini).

v1.1.3 ab + names of towns (in the ablative)

Three times ab is used instead of the plain ablative of the name of a town, twice in conjunction with the verb mitto, once with scn'bo:

295.6 miseras a Bremetennaco
299.1.2 a Cordonouis amicus missit mihi ostria
225.24f. a Vindolanda scribo

With the first two examples, compare (e.g.) Cic., Verr. 11.19, 'Messana litteras Halaesam mittit', Att. I. 10.I, 'Roma puer a sorore tua missus epistulam mihi . . . dedit', Fam. III. I I. I, 'quas ad me Q. Seruilius Tarso miserat'. With the third, cf. Cic., Att. 1x.6.1, 'Roma scripsit Balbus', Att. xv1.6.I, 'scribam ad te Regio'. 157 Prepositions +the ablative in such collocations are not unknown, particularly in colloquial texts and the post-classical period (e.g. Plaut., Bacch. 389, 'ex Epheso huc ad Pistoclerum litteras 1 misi'; Bell. Hisp, x11.3, 'a Corduba ad Pompeium missi sunt'; even Cic., Phil. x1v.23, 'at misit postea de Alexandrea, de Pharnace') but the prepositional construction which appears to be the norm at Vindolanda was in general non-classical. It represents another step of the many by means of which prepositions replaced unaccompanied cases in the history of Latin. The emperor Augustus was not afraid to use prepositions with the names of towns in the interests of clarity (Suet., Aug. 86.1).

The classical use of the plain ablative of the name of a town is probably to be found at 266, 'uol~ ueniat ad me Con's'. The editors translate 'I want him to come to me at Con'a' (my italics), but I feel that in this case Con's is unlikely to be locatival. The usual Latin idiom equivalent to Eng. 'come to me at

lS5 See further the editors, 43-4 for two cases of Vin- lS8 See R. Kiihner and C. Stegmann, Ausfuhrliche dolande (= -ae) on a wooden leaf-tablet and a stilus tablet. Grammatik der lateinischen Sprache: Satzlehre (3rd lS6 See Lofstedt, op. cit. (n. 85), 11, 73-8, Hofmann-edn., rev. A. Thierfelder, 1955)~ 1,478, Anm. 4 (the uses Szantyr, op. cit. (n. 145)~ 145. of prepositions illustrated at Anm. 3, (b), (c) have various

lS7 See further G.Funaioli, 'Der Lokativ und seine special motivations). Auflosung', ALL 13 (~goq),

326-7.

LANGUAGE OF VINDOLANDA TABLETS I11

Coria' was ueni ad me Coria, 'come to me to Coria': e.g. Caes., Ciu. 11.20.8, 'Varro Cordubam ad Caesarem uenit'.lS9 Moreover ad me (te etc.) + ablative (= 'from X') is a standard collocation, in the expected order (cf. Cic., Att. xv1.6.1, Fam. 111.1 I. I, quoted above). It is true, as we have seen, that the preposition ab is usually used with place-names to express separation in the Vindolanda material, but it cannot be deduced from the limited evidence available that ab was absolutely invariable. If Con's is to be taken as locatival, it could only be a highly unusual (perhaps unparalleled) adnominal locatival attached to a pronoun. There seems to be an adnominal locative at 250.8, 'Annjo Equestri (centurioni) regionario Luguualio', but this is somewhat easier to accept, because (a) the preceding name is not in the accusative (with ad) of motion (a construction which would tend to occur in a 'to-from' opposition), and (b) Luguualio has a noun-phrase (centurioni regionano) to hang on.

I return to scribo ab in the Vindolanda material. In one sense, as we saw, the expression is largely non-classical, in that the place-name is accompanied by a preposition; but in another sense it is a continuation of the classical practice, in that scn'bo is accompanied by a separative complement, rather than a locative. Though the locative does occasionally occur in classical Latin withscn'bo, do etc. of the point at which the letter was dispatched, the ablative was preferred.160 There appears to have been some controversy about correct usage in this environment: note Varro up. Scaur. GLv11.32, 'scribunt quidam "litterae datae e Gallia", item "Roma", uitiose; nam dici oportet "in Gallia" et "Romae"; dantur enim in loco, afferuntur e loco'.

Some distinctive uses of ad are found particularly in accounts (notably 180). There is a good deal of potential variability to the implications of ad when it is used without expressed verbs in lists, accounts, etc., and some of the Vindolanda examples are difficult to interpret. I begin with some extraneous evidence in order to provide some points of comparison with the Vindolanda material.

First, in the daily reports from Bu Njem (0.Bu Njem 1-62) expressions with ad are constant: e.g.

2 ad aqua balnei

ad Bojnag

3 ad pyeposit[um]

5 ad balneu

ad st(ationem) camellar(iorum)
I3 ad porta

These phrases are followed by numerals, signifying the number of men dispatched. Sometimes a passive form of mitto is expressed, and such cases furnish the key to the understanding of the various expressions : e.g.

12 missus ad ESuba

36 missus ad lignu balnei

Ad often expresses destination. The noun dependent on ad indicates the place, person to which soldiers were sent (to carry out duties): praepositus, statio camellariomm, porta, balneum, Boinag, etc. Alternatively the noun may refer, rather elliptically, to the duty itself on which the men were sent: e.g. ad aqua balnei '(to supply) water for the bath', ad lignu balnei '(to supply) wood for the bath'.

Secondly, there are adnominal uses of ad-expressions, particularly in official inscriptions, 161which tend to take on the status of titles. These uses recall the more familiar use of ab in titles such as ab epistulis, a rationibus. Ad in this function is typically attached to a name + a noun (e.g. in -tor) designating a profession, or to a name on its own: e.g.

(i) Name + professional designation + ad: e.g.

CIL ~1.8450, D.M. T. Ael(io) Aug
(ustorum) lib(erto) Saturnin(o) ...
tabul(ario) a rationibus, tabul(ario)
Ostis ad annona(m)

lS9 See E. C. Woodcook,ANewLatin Syntax (1959), 4. 161 Examples may be found at TLL 1.528.1gff. 160 See Funaioli, op. cit. (n.157))325-6.

II2 J. N. ADAMS

Note here the juxtaposition of tabulan'o a rationibus with tabulano ad annonam.

CIL ~1.5197,Musico Ti. Caesaris
Augusti Scurrano disp(ensatori)
ad fiscum Gallicum prouinciae
Lugdunensis
CIL v1.8688, C. Iulio Basso Aemiliano
actori Caesaris ad Castor(is) et
ad loricata

For loricata, see OLD s.v. lon'catus -a -um : '(fern. as sb., sc, domus?) A building with a protecting wall; (spec., perh.) one containing the imperial accounts'.

CIL ~1.9383, Diophanthus exactor ad insulas
CIL x1v.20, C. Pomponius Turpilianus proc(urator) ad oleum

(ii) Name + ad: e.g.

CIL ~1.3985, Isochrysus Liuiae ad uestem

CIL ~1.3972, Syneros Ti. Caesaris ad imagines

CIL ~1.3973,Q. Annio Q. 1. Philocalo

Helenus Liuiae ad insul(am) ollam

dat

CIL ~1.7884,Phoebo Marciae Maxsimi ad margarita

With ad insulas (insula = 'block of buildings') above, cf. CIL ~1.3974,'<C>erdo insular(ius)', where insulan'us is no doubt interchangeable with ad insulas. These uses of ad can be readily derived from the final meaning of which the preposition is capable

(e.g.
procurator ad oleum, 'proc. for the purpose of, for supplying, oil'). In each case the noun dependent on ad indicates a substance, object, sim. which the person is responsible for or in charge of
(e.g.
CIL ~1.3985,'Isochrysus in charge of the wardrobe'). The original final nuance of ad is most obvious in those cases where the dependent noun is accompanied by a gerundive: e.g. CIL v111.10500, 'proc. Aug. ad census accipiendos'.
In a few cases the prepositional phrases, while still possibly final, could alternatively be interpreted as local, i.e. as indicating the locality, building sim. in which the official operates: e.g.

CIL ~1.8689,T. F1. Aug. lib. Martiali proc. Aug. ad Ca~tor(is)'~~
CIL ~1.8688,C. Iulio Basso Aemiliano actori Caesaris ad Castor(is) et ad loricata

The two senses, final and locative, are difficult to distinguish here. I turn now to the Vindolanda evidence, and begin with a revealing example which is clearly distinct from the cases discussed above:

185.2-1 axes carrarios
duos ad raedam

wagon-axles, two, for a carriage.

Ad here is final, and interchangeable with the dativ~sjnalis.'~~

For the same usage in a letter, see 309, 'missi tibi . . . axses ad lectum'. This example does not have the same structure as that seen in the inscriptions above, because here the ad-expression indicates the purpose to which the object signified by the head-noun (axes) is to be put; in the inscriptional examples, the head-nouns signify not things but persons (professionals) exercising some sort of control over the objects expressed by the prepositional phrases.

I move from here to the account 180, where there is a series of examples of ad, two of which have been translated by the editors as equivalent to the inscriptional examples cited above which have the force 'in charge of'. I cite all ad-expressions in the account:

'62 This example, quoted by P. R. C. Weaver, Familia 163 On this use of ad, see Kiihner-Stegmann, op. cit. Caesans.A Social Study ofthe Emperor's freed me?^ and (n. 158), I, 522. Slaves (1972)~268 n. 2 is filled out by him asad Castorem, but it is more likely that Castons is intended (see Lofstedt, op. cit. (n. 85), 11, 249).

LANGUAGE OF VINDOLANDA TABLETS

(i)
3 mihi ad panem [
(ii)
10 Amabili ad fanum m(odii) iii
(iii) 20 ad turtas tibi m(odii) ii

(iv)
27 Lucconi ad porcos [
(v)
33 patri [a]d i[uu]encos [
(vi)
37 item mihi ad panem m(odii) i[
Nos (iv) and (v) are translated by the editors as 'to Lucco, in charge of the pigs', and 'to father, in charge of the oxen'. But is this correct? Nos (i), (iii), and (vi) are without question straightforward final cases, much like ad raedam above. Thus (i) 'to me (so many modii), for bread' (i.e. for the making of bread), and (iii) 'for twisted loaves, to you, 2 modii'. Nos (iv) and (v) can obviously be taken in the same way, in keeping with the pattern of the rest of the account. Thus, e.g., (iv) 'to Lucco, for the pigs, (. . . modii)'. Adporcos on this view is a satellite not of Lucconi (which structure would make the expression parallel to e.g. Isochrysus ad uestem) but of modii (fmmenti) (cf. axes ad raedam). The account lists quantities of fmmentum; forfi-umentum used to feed pigs, see Varro, Rust. 11.4.6, 'hoc pecus alitur maxime glande, deinde faba et hordeo et cetero frumento'.

Ad in ad fanum (ii) may mean much the same ('for the fanum'), though here perhaps the phrase shades into a local meaning ('to Amabilis, at the shrine': so the editors). At 30 (Lucconiin ussos suos) the writer has substituted a final use of in. On bubulcan's in siluam, see v1.1.5.

The final use of ad in the structure identified here occurs a number of times in other accounts: I 83 Candido ad porcq[s . .., 190 ad sacrum .. . (four times); also domino ad stipes (showing the same juxtaposition of dative of the person, ad of the thing).

Finally, 155 presents some uses of ad identical to those in the daily reports from Bu lVjem (nos 1-62).

3 ~[trluctoye~ ad balneum xviii

Herestmctores are to be dispatched to the baths to carry out duties; ad balneum, as we have seen, is also found at Bu Njem. Similarly at 1. 7 (adfumaceq) an unspecified number of men is to go to the kilns. On the other hand 1. 4

[a]d plumbum uacat [

-would appear to parallel Bu Njem examples such as ad aqua balnei, ad lignu balnei. Men are presumably to be sent off to acquire lead.

155 might be compared with 156, which is a similar type of document. There, however, missi (perfect participle) is expressed, as sometimes at Bu Njem, and the spheres of activity are made explicit by the use (three times) of the construction ad + noun +gerundive (as distinct from the more elliptical ad + noun).

One final question is worth posing in this section. Can any difference be discerned between the use of ad in accounts (expressing the object on which money or some other substance was disbursed), and that of the dative? Six examples of ad from 180 were quoted above, to which could be added, in the same document, the incomplete entry (16)patm'ad. . . In addition we noted Candido adporcos (183), and, in 190, four examples of adsacrum and domino ad stipes. There is a total of thirteen examples of ad, nine of which are juxtaposed with a dative (of the pattern Candido ad porcos). It is only the expression ad sacrum in 190 which is not accompanied by a personal dative.

The dative instead of ad occurs in the following places: 181 lignis emtis, 182. I, I zpretio, 182 rebus minutis (twice), pretio exungiae, 185 faeci (five times), locario. The total is twelve examples, none of which is juxtaposed with a personal dative. The conclusion is clear. Ad was substituted for the dative when another dative (of the person rather than the thing) had to be used in the immediate context.

v1.1.5. A use of in + acc.

The account 180, as well as offering the uses of ad seen in the previous section, also has the following entry:

bubulcaris in siluam m(odii) viii 9

The force of in siluam is difficult to determine. A purely locative sense, as adopted by the editors (who translate 'to the oxherds at the wood') does not seem possible. Ad is apparently used with local meaning in the next line (Amabiliadfanum), but in +acc. in this sense would be anomalous (though in +ablative acceptable). Later in the same document, as we saw, in + acc. is used in a final sense:

30 Lucconi in ussos suos

114 J. N. ADAMS

The development by in + acc. of a final meaning can be seen from its intrusion into expressions of purpose containing a gerundive: e.g. Lex pag. Herculan. (CIL1'.682), 'utei in porticum .. .reficiendam pequniam consumerent', Livy xxv111.4 5. 18, 'Rusellani abietem (polliciti) in fabricandas naues'. 164 For other examples of in, without a following gerundive, which can be interpreted as final, see, e.g., Cic., Verr. 11.38, 'in eam rem iudices dentur', Livy 1x.24. I, 'nouisque cohortibus in supplementum adducti~'.~~~

TOjudge by the examples cited by Kiihner-Stegmann, the final use of in +acc. was largely post-Classical. I am inclined to take in siluam as parallel to in ussus suos and the uses of ad in the same document,

i.e. as a pregnant final use, = 'for the purpose of the wood', i.e. 'for use in the wood'. It is in the very nature of ad as it is employed in the document that its exact implication is variable. Adporcos implies 'for feeding to the pigs', whereas adpanem implies 'for the making of bread'. So too the function of in +acc. might in theory have varied with the context.

One might ask what bubulcahi were doing in a wood. The bubulcus was in charge of oxen, but since oxen were chiefly used to plough, bubulcus in effect refers to a ploughman (see Col. 11.2.25). If bubulcan'us means the same as bubulcus (see v.2), then it may seem rather odd that a wood should be the sphere of activity of ploughmen. Perhaps wooded terrain was being converted into ploughland. For this activity, see Col. 11.2.1 I, which describes two methods of clearing. Either the trees might be torn out by the roots and removed, or if they were not densely packed they might be cut down, burnt, and then ploughed under. Bubulcan'i might well have been engaged in this second activity (see further 11.2.28). Thefrumentum would be for their oxen working there.

It is an alternative possibility that in (+ siluam) is purely local, with no final nuance: = 'for the bubulcan'i (going) into the wood (to work with their animals)'. Could a local use of in + acc, be used adnominally in this way? It should be noted that on the other interpretation offered above, in siluam is not adnominal at all, but a satellite (with final sense) of an implied verb-phrase: frumentum (is given) to bubulcan'i,for the purposeslneeds of the silua. A directional adnominal expression (as distinct from the sort of locatival adnominal that ad fanum may be taken as) would be far from easy to defend.

VI. I .6. Remarks on the syntax of case in accounts and lists

Accounts and lists, for which there is an abundance of new evidence at Vindolanda, display forms of syntax without an expressed verb. Nouns may be juxtaposed with one another in a variety of cases: e.g. locative + dative + (accusative):

18 5.24 Cataractonio locario (denarii) ~(emissem)

Often the cases adopted can be explained in relation to an implied verb. The account 18 j appears to list a series of payments made on a journey, and a verb of giving, paying out can readily be supplied. Thus 185.24, '(I paid) at Catterick for lodgings, denarii '/2'.

But that is by no means the whole story. There are instances of case usage in the Vindolanda documents which do not become explicable if the expected verb is understood. I illustrate this contention first from the same account, 185:

18 5.20 axes carrarios
duos ad raedam (denarios) iii (s)emissem

The editors choose to interpret the denarius sign as standing for an accusative plural, but in fact its case is of no great significance. If one makes the assumption that case should be explained from an understood verb, then the denarii might be assigned either nominative case or accusative depending on whether the verb is taken to be active or passive. Of more interest is the accusative axes carran'os duos. It is obvious that a payment was made for carriage axles. Elsewhere in the account the object on which payment is made goes, predictably, into the dative case, which carries a final nuance. At 183.24 above locan'omeans 'forlodgings', and five times in the account an entry begins faeci (= 'for lees of wine': but the reading is in every case problematical). Why then is axes carrarios in the accusative? It would be over-subtle to attempt to relate the case to an understood verb-phrase, e.g. '(I bought) carriage axles by means of X denarii'. For one thing the verb would have to be different from that understood in the other entries, and for another the denarii would have to be assigned a case other than nominativelaccusative. The explanation seems to be that the accusative was a sort of unmarked case which in accounts and lists could be given to the nouns signifying the object(s) boughtisold, even in contexts in which that accusative could not readily be derived from the verb expected in such an account.

164 See TLL v11.1.765.6gff. 165 See further Kiihner-Stegmann, op. cit. (n. 158),I, 346,TLL v11.1.765.16ff.,

OLD S.V. in,22.

Exactly the same variation as that between faeci (dat.) and axes carrarios (acc.) can be seen in 181.3-4, part of a 'cash account recording sums received and debts outstanding' (Bowman and Thomas, 129):

lignis emtis (denarjos) yii

sticam (denarjos) iji.

In this part of the account receipts are recorded. Thus in each of the next four lines we find ab + name

(e.g. ab Alione uetem'nario) followed by a quantity of denarii, = '(received) from X, Y denarii'. Line 3 obviously means 'for timbers purchased, (received) 7 denarii'. The next line must describe a comparable transaction (= '(for) a cloak, (received) 3 denarii'), but here the writer has lapsed into a syntactically unmotivated 'accusative of the thing sold, acquired'.

I note in passing that (lardi) pernam at 182.7 is not the same sort of unmotivated accusative as sticam above. In 182there is variable case usage, but the variations can be explained in reference to verbs readily understood. Thus, 1. 3, '[relbus minutis (denarios) ii (asses ii)', = 'for sundries, (received) denarii 2, asses 2', or 1. j, 'Ircucisso ex pretio lardi (denarios) xiii ~(emissem)', ='Ircucisso (paid) as part of the price of bacon, 1~~12

denarii'. The next two lines differ from those just quoted, in that quantities bought rather than prices paid are stated :

67 Felicio (centurio) lardi p(ondo) xxxxv I item lardi pernam p(ondo) xv ~(emissem)

Felicio the centurion (bought) 45 pounds of bacon, likewise bacon-lard I j1/2 pounds.

Here there is a typical variation between a genitive (lardi) in 1. 6 dependent on pondo, and partitive apposition in the next line, with (lardi) pemam in apposition to pondo (cf., e.g. 191.5-6 for such variation in a list). Lardipemam can be construed as object of the verb of buyingiacquiring which is demanded by the quantity term pondo.

In 182 the persons acquiring goods from the unnamed trader who wrote the document are repeatedly expressed in the nominative case (at the head of the entry), followed by a statement of the sums paidlquantities bought. The nominative (of the name) is quite logical, in that it is a simple matter to supply a verb of which it is subject. Indeed in another document (181) the process of ellipse can be seen in action. Lines 11-1 j have nameslnominatives at the beginning of the entry, followed by an expression signifying a sum of money: e.g.

I I Ingenus (denarios) vii

In these entries, however, the intended verb is actually specified:

10 reliqui debent

We have identified then in the Vindolanda accounts a use of the accusative (of the thing), which may be difficult to relate syntactically to the verb-phrase which seems to be demanded by the nature of the account, and a nominative (of the person acquiring the goods), which can usually be taken as subject of an understood verb. I note finally that the two case uses, accusative of the thing and nominative of the person, come together repeatedly in the account 184. The account consists of a long series of entries of the following form:

20-1 sagaciam (denarios) v (asses iii)
Lucius scutarius

An accusative, expressing an object (bought) is followed by a monetary statement and then a name, in the nominative. Lucius must have bought the goods named, and the nominative is explicable in the sense described above.

Again, however, the accusative (of the goods) is not so easily amenable to logical syntax, because of

its unspecified syntactic relationship to the denarius symbol. It is reasonable to conclude that the

accusative was felt to be an appropriate case for the goods listed in an account, and that those using the

case in a list would not necessarily have considered its theoretical relationship to an implied verb-phrase.

The accusative had a long history of use in recipes, lists, and the like where no verb was expressed

(see, e.g. Cato, Agr. 128; also chs 12, 13, which consist entirely of accusatives, without a governing

116 J. N. ADAMS

verb).l'j6 The Vindolanda examples, and particularly those which display a switch from, say, a motivated dative to an unmotivated accusative, nicely illustrate this role of the accusative. For the same combination of nominative (of a name) and accusative of goods as that seen above in 184, note the Pompeian inscription CIL 1v.4227:

pan(e)m l(ibram) IS P. Catillus lib(e)ra(m) I u(n)cias q(u)i(n)q(u)e semu(n)c(i)a(m)

Quantum at Petron. 43.4 ('uendidit enim uinum, quantum ipse uoluit') has traditionally been taken as an 'accusative of price',167 though that interpretation has, quite justifiably, recently been questioned by H. N. Parker:168 there seems no good reason why the second clause cannot be translated 'as much as he himself wanted'. But while Parker may be right about Petron. 43.4, it is going too far to suggest that the construction belongs only to the fourth century and beyond. Parker can quote only literary examples, but there are two possible cases from the early second century in non-literary sources. Octavius' letter

(343) runs thus at 11. 38-41 :

Frontinium Iulium audio magno licere pro coriafione quem hic comparauit (denarios) quinos

The passage is notoriously problematical (see above, v.5 on coriatione, and the editors ad loc.). Nevertheless, the syntax of the qz.iem-clause seems to admit of only one interpretation: 'which he boughtiacquired for five denarii apiece'. (Llenan'os) qziinos would have to be an accusative of price.

In addition to this case there is also Terentianus, P. Mich. v111.469.17, '[mlerca minore pretium', a passage which is open to more than one explanation.l'j9

VI. 1.8.A use of the accusative(?) in a letter

The letter of Chrauttius (310) may be punctuated thus at 11. 4-9:

et rogo te, Veldei frater -miror

quod mihi tot tempus nihil

rescripsti -a parentibus nos-

tris si quid audieris aut

Quofm in quo numero

sit

I ask you, brother Veldeius -I am surprised that you have written nothing back to me for such a long time -whether you have heard anything from our 'parents', and Quofus- in what unit he is.

The obscure word Quotm must be a name (see the editors ad loc.), which ought to have been in the nominative as subject of sit.

'Unconstrued' accusatives are well recognized,170 but they do not represent a single phenomenon. 'Recipe-accusatives', for example, may sometimes be explained from ellipse of a verb. At Plaut., Amph. 1009 ('Naucratem quem conuenire uolui in naui non erat') and Poen. 644-5 ('hunc chlamydatum quem 1 uides, ei Mars iratust') the accusatives may perhaps be due to attraction into the case of the relative pronoun. 171

It is not easy to find parallels for the accusative in the letter of Chrauttius. Vaananen, however, distinguishes on the walls of Pompeii between unconstrued accusatives indicating materials and the like (op. cit. (n. I I), I 17), a use not unlike the recipe-accusative, and accusatives, usually of names, which he describes as an 'accusatif exclamatif'. Note particularly CIL 1v.352j, 'Puteolos Antium Tegeano Pompeios -hae sunt uerae coloniae'. Here the (virtually exclamatory) accusatives serve to introduce various places, and then follows a clause in which the names might have been expressed in the

166 See, e.g. Hofmann-Szantyr, op. cit. (n. 145), 29, '69 See Adams, op. cit. (n. 17), 4-2.

Svennung, op. cit. (n. 98), 185-7. For such accusatives "O See the bibliography cited inn. 166; also D. Norberg,(of materials, etc.) on the walls of Pompeii, unaccompan- Syntaktisclze Forsclrungen auf dem Gebiete des Spat- ied by any verb, see Vaananen, op. cit. (n. I I), 117, lateins und desfnilzen witt tell ate ins, Uppsala Universitets

'67 See E. LGfstedt, Spiitlateinische Studien (1908) Arsskrift 1943: 9 (1943)~ 92-6, A. Josephson, Casae 79-82, idem, op. cit. (n. 85), I~,271, idem, Ve~mischte Litterarum. Studien zum Corpus Agn'mensoncm Romano- Studien zur lateinischen Sprachkunde und Syntax rum (~ggo), 165-8. (1936), 173; also Adams, op. cit. (n. 17), 4-2. '"See Svennung, op. cit. (n. 98), 183.

168 'A curiously persistent error: Satyricon 43.4', CP 89

(1994), 162-6.

nominative. In the letter of Chrauttius a similar sort of accusative perhaps serves to introduce a new subject, which might instead have been in the nominative in the following indirect quotation. It is not impossible that Chrauttius admitted a second example of the same type of accusative in 1. I 7 (see the editors ad loc.) :

et rogo te, frater Virilis,
salutes a me Thuftenam
sororem. Velbutejum -
rescribas nobis cum . . .
se habeat

greet our sister Thuttena. And as for Velbuteius, write back to us how(?) he is.

If Velbuteium were object of salutes, the asyndeton bimembre would be odd, and habeat would appear not to have a subject.

VI .z. Paratactic Uses of rogo

A feature of the letters is the frequency with which, in requests, rogo is used paratactically with the (jussive) subjunctive, unaccompanied by ut (e.g. 233, 'rogo mittas mihi . . .'). Rogo (1st pers.) + subjunctive occurs twelve times definitely,172 and possibly in two other places where the text is fragmentar~.'~~

By contrast there are five places where rogo is followed by ut.174

It seems likely that rogo + subj. introducing a request was current conversational Latin. Martial, who was roughly contemporary with the Vindolanda letters, preferred rogo + subj. to rogo + 2~t.l~~ Petronius on the other hand preferred rogo (-amus) + ut to rogo + subj., but the latter is scarcely outnumbered;176 all three examples of rogo + subj, are in speeches, two of them by freedmen. The distribution of rogo utirogo + subj. in Cicero's letters is of some interest. In the letters ad Familiares rogo ut overall outnumbers rogo + subj. by 43 :IO, but there are variations according to authorship. Cicero himself prefers rogo ut by but in the letters by his various correspondents rogo + subj, is preferred by 8 :4.178 There is also one example of rogamus + subj. (Fam. x1.2.3), in a letter by Brutus and Cassius, compared with one example of rogamus ut (Fam. v. I 2.9), in a letter by Cicero. The relative frequency of rogo + subj. in these various letters by miscellaneous correspondents suggests that Cicero's own letters falsely imply a lack of currency for the paratactic construction in late Republican educated usage. Cicero's use of rogo (ut) with indirect commands is rather formal (and formulaic), in that he tends to place the ut-clause before rogo, and he tends to use the construction in a restricted set of expressions. There must have been current a less formal use of rogo + subjunctive.

In the letters to Atticus rogo ut is similarly preferred by Cicero (25:~).l~~

I mention finally that Terentianus always uses ut in conjunction with rogo (seven times); in the same corpus of documents, however, his correspondent Tiberianus has rogo + subj. (P. ,il/Izch. ~111.472.11).

The letters from Vindolanda thus provide new evidence for the frequency with which rogo must have been attached to a subjunctive in expressing a request. There are other cases of first person present verbs followed by a subjunctive in the letters. Note 266 uolo ueniat; also the more usual uelzm + subj. at 349.

In the Vindolanda letters the paratactic formula of farewellopto bene ualeas is preferred to the type with acc. + infin. (e.g. opto te bene ualere and variants). In other non-literary letters extant, while opt0 bene ualeas and also opto ut bene ualeas are found, the accusative + infinitive construction predomin- ates.lsO Opto . . . ualeas has so far turned up at Vindolanda nine times (215, 260, 289, 300, 309(?), 312,

17' 218, 233, 291, 301, 310, 311, 312 (three times), 314, 176 Rogol-amus ut: 64.1, 71.9, 75.8, 99.2, 134.II; + 326, 345. Similarly in the commeatus documents 167, subj.: 49.6, 75.3,, 137.6. 174, and 176 the expression rogo . . . dignum me habeas is 177 Rogo + sub] only at Fam. v.18.1 (rogo atque oro), more or less preserved, and it is likely that in the remain- x111.57.2. der, which are more fragmentary, it was also used. I have Forrogo + subj., seeFam. v.~ob (Vatinius), v111.11.4 not included the three examples just listed in the figure of (Caelius), x.21a twice (Plancus), x.24.8 (Plancus), XI. I .5 twelve. (D.Brutus), x1.9.1 (D.Brutus), x11.14.4 (Lentulus); for

173 257, 345. TO these might be added 342, where rogo ut, see v111.2.2, v111.9.4 (both Caelius), x.9.3 (Plan-
rogamus rather than rogo seems to have a plain subjunctive cus), x1.28.5 (Matius).
complement, but there are gaps in the texts. 179 Rogo + subj, at 1v.14.2, v.12.3.

174 250 (a rather formal letter of commendation), 255, lso For a collection of the evidence, see P. Cugusi,
313, 316 (twice). Covpus Epistolarum Latinarum (1992), 11, 37.
'75 Rogo ut never; rogamus ut at ~1.35.5.

Rogolrogamus + subj. seven times (-0 11.79.2, 111.95.3, v1.5.2, v11.95.18; -amus 1.35.13, v.80.4, v111.2.8).

I 18 J. N. ADAMS

316(?), 345, 353),ls1 against three examples of the acc. + infin. (248, 250, 258). Conventions no doubt developed among groups of scribes.

An alternative form of paratactic construction shows rogo used in conjunction with an imperative. There is only one example of the construction so far in the Vindolanda material, in the letter of Octavius (343.14-15 rogo . . . mitte). Parallels turn up sporadically, usually in texts of colloquial colour (Petron. 67.1, 137.4, Mart. 11.25.2, Terentianus, P. Mich. v111.469.17). There are a few examples in the correspondence of Cicero, one in a letter by D. Brutus (Fam. x1.26), and another in a letter from Quintus Cicero to the freedman Tiro (Fam. xv1.26.2).

Octavius uses various other paratactic constructions (following verb-forms other than the first person present indicative, discussed above): 343.16-17, 'scribe dentur',ls2 31-2, 'desiderabat coria ei adsignarem'.

VII. SOME SOCIOLINGUISTIC EVIDENCE

VII.I. Some Terms of Address

The terms of address frater and domine, which in the vocative had lost their full lexical meaning, ls3 are well attested in Latin, but the Vindolanda tablets are important in providing a substantial body of examples from one area and one time, which allow some deductions to be made about the implications of such addresses.

There are some significant contexts in which domine is used:

(i) In letters of application for leave (commeatus: 166-77)ls4 the officer to whom the request is put is invariably addressed as domine, with a vocative of the name (e.g. 166 domine Cqialis, 172 domine F'lauiane; ten examples in all ;no example of e.g, frater) .ls5 (ii)In an appeal to higher authority (containing the expression tuam maiestatem imploro) domine is used

(344.7).

(iii) Twice in a formal letter of recommendation Cerialis is addressed as domine (250). Then in the final greeting the writer (probably Claudius Karus: see the editors ad loc., 221) apparently adopts a less formal tone, using uale frater. The editors suggest that Karus may have been a fellow-prefect of Cerialis. Domine turns up elsewhere in letters of commendation, perhaps most notably at P. Oxy. 1.32. In this, a second-century document addressed to a military tribune, domine is used three times (e.g. 7peto domine ut .. .; cf. Tab. Vind. 11, 250.6 rogo ergo domine . ..). In another relatively early (first- or second- century) letter of recommendation (P. Ryl. 1v.608) note 6 rogo domine . . .hab[eas].ls6 (iv)Next I quote the fragmentary letter of gratitude, 332:

salutem
summas tibi domine gratias

There is no context, and the sender and addressee are unknown. What is interesting here is the use of domine embedded in a highly formal pattern of words. Tibi is in the so-called 'Wackernagel position' in second place in the clause, and along with the following vocative it separates the adjective summas from the noungratias. For this order, cf. the opening of Cicero's tenthphilippic: 'maximas tibi, Pansa, gratias

,

...

(v)Finally, Cerialis'(?) letter (225) to a man (Crispinus) who is clearly his superior requesting that he intercede for him with the governor Marcellus not only has the vocative domine; in the same sentence the accusative dominum rneum is used in apposition to te (4-6) :

amplexus s[um do-
mine salutandi te occas~ionem
dlominum meum

For expressions of this same structure (i.e. vocative +a form of tu +an appositional expression which might alternatively, with re-phrasing, have been in the vocative), see Cic., Mil. 44, 'te, Q. Petili, appello, optimum et fortissimum ciuem', Mil. 102,'quid tibi (respondebo), Quinte frater, qui nunc abes, consorti

lR1cf. 264, 'opto ... sis felicissimus' (so ~IO), 346, Iss

Similarly domine turns up at Bu Njem in letters '[olpto felicissimus uiuas'. addressed by soldiers to their commanding officer: e.g. 0. IR2 Bu Njem 76.4, 'transmisi at te domine . ..'

For parallels and bibliography, see Bowman, Tho-
mas, and Adams, op. cit. (n. 146), 48. la6

For a collection of such litterae commendaticiae, see lR3 Bowman and Thomas, op. cit. (n. 48), 106.

Forfrater and its implication, see 0.WiidiFawiikhir
2.6-9, HA, Did. Iul. 4.1, Hor., Epist. 1.6.54.

la4

On the type of document, see Bowman and Thomas,

77.

mecum temporum illorurn', CIL v111.9052.4, 'ate, Clodia Luciosa, uxore mea, sus~ep<i>'.'~~

It seems likely that forms of address of this type were highly formal.

Although most of the examples collected above from the tablets are in address to a superior, the use of domine does not inevitably imply a subordinate status for the writer. In this connection the letter of recommendation 250 ((iii) above) is revealing. Writer and addressee seem to have had equivalent status. It was the convention of litterae commendaticiae which dictated the use of domine in the body of the letter (whereas in the final greeting the friendship-term frater -on which, see below -could be substituted). Domine was a term of politeness or deference, with a formal tone. Though it was certainly suitable in address to a superior (see further below), its use might also be determined by circumstances requiring a posture of formality and deference. I mention finally 234, where Cerialis addresses a certain September. September was probably an equestrian officer (see Bowman and Thomas, 199; also 223, on 252), but he is unlikely to have been Cerialis' superior. If he was of subordinate status, then the use of domine must have been determined either by the nature of their relationship (formal rather than friendly), or by the context of the letter (which is fragmentary).

The evidence collected above showing domine used in an address by a subordinate to his superior may be compared with some contemporary material, the correspondence between Pliny and Trajan. Domine begins almost every letter from Pliny to Trajan,lss but Trajan never uses the term in reply (cf.

x. 16.I mi Secunde carissime, also x.20. I, x.50, x.55, x.62, x.80, x.89, x.95, x.99 ;cf. x.97 mi Secunde, x.44Secunde carissime, also x.91, x. I IS, x. 121). This non-reciprocity of address nicely establishes that domine was appropriate from an inferior addressing his superior. Sherwin-White, loc, cit., quotes further examples spoken by inferiors to superiors (both of free birth), though that, as we have seen, is not the whole story.lsg It is of note that Trajan responds to Pliny by means of friendship-terms (notably the mi-form of address, with frequent use of carissime). This would appear to be a form of condescension from superior to subordinate.

I move on to fraterlsoror. 233B is a letter from Cerialis to an equestrian officer Aelius Brocchus.

Cerialis addresses Brocchus as frater, in a request accompanied by the familiar expression si me amas.

243, conversely, is a fragment of a letter from Brocchus to Cerialis; this too contains the vocative(?) frater. In 248, this time from Brocchus and another equestrian officer Niger to Cerialis, frater occurs again. 345 is a letter apparently from one prefect (Celonius Iustus) to another.lgO The unknown recipient is addressed as frater, and also as frater et domine.

Further dbwn the social scale(?), in 310 we find a soldier with a Germanic name, Chrauttius, addressing his old mess-mate (contubemalis antiquus) Veldedeius as frater. In the same letter Chrauttius incorporates a message for the ueter~narius Virilis, who is addressed in the first person, again as frater.

--,

~

301 is a letter from a slave Severus to another slave Candidus. Frater is used in the vocative.

Comparable to the use of frater is that of soror, which is nicely illustrated by the two letters 291-2.

Just as letters survive which passed between Cerialis and Brocchus, so we have two letters from Claudia

Severa, wife of Brocchus, to Sulpicia Lepidina, wife of Cerialis. Soror occurs five times in the vocative in

the two letters, along with various endearments.

It is obvious then that frater and soror were used between speakers of equal status -prefect to

prefect, slave to slave, soldier to soldier. Frater was not restricted to the speech of any one social class.

Both frater and soror were at home in informal contexts. The tone of frater is revealed by (e.g.) HA, Did.

Iul. 4.1, 'unumquemque, ut erat aetas, uel fratrem uel filium uel parentem adfatus blandissime est', but

it is useful to have a new corpus of examples in letters composed by writers of different social classes. The

most significant examples of frater are those in the correspondence between Cerialis and Brocchus and

their families. The two families were on intimate terms, and fraterlsoror was clearly the appropriate form

of address.

Finally, for completeness I note that the attachment to domine of frater or karissime or both occurs

seven times in the Cerialis archive (domine frater at 252, 260, 289, mi domine karissime at 288, (mi)

domine fraterkan'ssime at 247, 255, 285(?); cf. zp), but nowhere else in the Vindolanda letters. Little

or nothing is known about the writersladdressees in these cases, and the letters are often fragmentary; it

is not therefore possible to draw any conclusions about the character of such addressees.

ls7 I am grateful to Dr F. Jones for supplying me with CIL v.1706 (aunt to nephew). This use represents a this last example. debasement of the deferential use into an empty (formal) lss See A. N. Sherwin-White, The Letters of Pliny. A form of address much the same ashfaster in English. Historical and Social Commentary (1966), 557. 190 See Bowman and Thomas, 334. ls9Domine could even be addressed to a child, as e.g. at

I

I20 J. N. ADAMS

VII .z. Female Terms of Address

The archive of Sulpicia Lepidina (291-2 especially; also 293-4), brief as it is, seems to confirm that endearments were a distinctive feature of female speech in Latin. 191In letters written by men, as we have seen, the standard terms of vocative address were domine, frater, combinations of the two, and combinations of either with kari~sime.'~~

Occasionally too there is a mi-form of address, with mi preceding either a name (242 mi Felici- ...ka+$~qe)l~~

or domine (247 mi domine frater karissime, 288 mi domine karissime). The mi-form was intimate and affectionate, but it was not an endearment. Carissime may loosely be described as an endearment, but it was a hackneyed and fairly empty epithet.

The two letters of Claudia Severa to Lepidina are remarkable not only for the accumulation of examples of soror (five times, as noted above, twice within three words at 291. I 1-1 z), but also for the use of anima (+ possessive +adjective) twice as a vocative endearment: 291.12, 'uale soror anima mea ita valeam karissima', q~.b.back, '?oror karissima et anzma ma desideratissima'. There are no comparable endearments in the far more extensive letters written by men.

In our only substantial source of information about 'female' endearments, Latin comedy, anima (voc.) does not occur, though mi anime is regularly put into the mouths of females (nine times out of twelve in Plautus and Terence). illea anima (in the plural) turns up twice in Cicero when he is addressing women (Fam. XIV. 14.2, XIV. 18. I). If anima mea had become established in the late Republic as a female endearment, it would not be surprising that it should also have been used by men addressing women;lg4 endearments associated particularly with women may be used between the sexes, but rarely by a man addressing another man. There is a later example of animae meae at Peregrinatio Aetherz'ae 19.19, addressed by a woman to some other women. The editors cite an example of anima (voc.) accompanied by the adjective dulcissima addressed by a man to a man (in Fronto, ad M. Cues. 11.10.3,

p. 30 van den Hout; Marcus Aurelius to Fronto; part of an exaggeratedly affectionate and deferential farewell greeting: 'desiderantissime homo et tuo Vero carissime, consul amplissime, magister dul- cissime, uale mi semper anima dulcissima'),lg5 but it seems in general to be true that anima (voc.) was used by or to women rather than in discourse between two males.lg6 It is not only the endearment as such that is of interest in the Lepidina archive; it is that endearments should be used twice in letters by a woman, but never in the more extensive correspondence composed by men.

The expression anima desideratissima occurs also (though not in the vocative, and conflated with the present participial form) at CIL ~1.21974.6-8, cited by Bowman and Thomas, 262 (also TLL

VI. I .710) : 'coniugi carissim[ae] animae desideran[tissi]mae'. This example fits the pattern proposed above, in that it is from a tomb inscription dedicated by a man to his wife. Desideratissima ('much missed') was no doubt highly emotive, since it could be applied to a deceased loved one.

,&la (followinganima) is perhaps a simple dittography, but alternatively it may be the reduced form ma =mea. &la =mea is attested at much this time in the letters of Terentianus (P.Mich. v111.471.34 materma; also 471.17patertus, 30sum negotium). There is a set of such reduced forms (mus, tus, etc.), with reflexes in Romance (Fr. mon, ton, ma, ta etc.). These may reflect the effects of an accentuation mehs for me'us, tubs for thus, et~.'~~

VIII. VOCABULARY AND STYLE

In this section I discuss the lexical diversity of the tablets. It is, I stress again, a mistake to approach the Latinity of the tablets as if they were a specimen of 'Vulgar Latin'. They contain, it is true, a stock of terms (some of which have been seen already: turta, locarium, excussonum) of the type which are rarely, if at all, attested in literary genres, but which survive in the Romance languages. These clearly belonged to spoken, as distinct from literary, varieties of the language, but it would probably be more accurate to classify them as 'technical' rather than 'vulgar'. There are at least two other components of the

lgl For the evidence of comedy, see J. N. Adams, lg6 Bowman and Thomas, 258 note that the expression 'Female speech in Latin comedy', Antichthon 18 (1984), anima dulcissima occurs on a gold-ring found in the

71

,-' fourth-century uicus at Vindolanda (see R. P. Wright and lg2 Homo inpientissime (311.5) belongs apart, as it is not M. W. C. Hassall, 'Inscriptions', Britannia 2 (1971)~301 deferential or friendly. no. 72) :could the ring have been presented by a man to a

lg3 The only other names in the vocative are in the letter woman? of Chrauttius, both accompanied by frater: 310.4 Veldei 19' See B. Lofstedt, 'Die betonten Hiatusvokale in Wor- frater, 15frater Virilis. Frater is not usually accompanied tern vom Typus pius, meus, tuus , Eranos 60 (1962)~ by a name when used in the vocative: see Bowman, 8-92, especially 89; also C. Lyons, 'On the origin of the Thomas, and Adams, op. cit. (n. 146), 37, 40. Old French strong-weak possessive distinction', TPhS

lg4 See the evidence collected by Adams, op. cit. 1986, 22-7. For another example of ma, see Onnerfors,

(n. I~I),71-2. 'Iatromagische Beschorungen in der "Physica Plinii

lg5 Marcus Aurelius seems deliberately to have used Sangallensis"', Eranos 83 (1985), 237 no. 4.
highly emotive language which might have been appro-
priately used by or to a woman.

lexicon at Vindolanda (in addition to technical terminology) which should be distinguished from 'Vulgar Latin' elements. I refer first to learnedlformaliliterary terminology, and secondly to usages characteristic of the particular genre in which they occur. Genre as a determinant of linguistic usage has already been illustrated in the section on syntax, where (e.g.) certain features of the syntax of lists were discussed. The material in this section is miscellaneous; I begin with an example of a generically determined usage.

VIII. I .item

180.10 Macrino m(odii) xiii
bubulcaris in siluam m(odii) v,iii
item Amabili ad fanum m(odii) iii

The editors ad loc. (125) observe that 'the force and referent of this word is unclear here and in several other entries'. But item tended to lose its comparative sense ('likewise, in the same way'), and to take on a purely additive force (= 'also', simul, praeterea): see TLL ~11.2.535.50 .This is particularly clear in an example such as Apicius 11.2.6: 'isicia de pauo primum locum habent ita si fricta fuerint ut callum uincant, item secundum locum habent de fasianis, item tertium locum habent de cuniculis, item quartum locum habent de pullis, item quintum locum habent de porcello tener~'.'~~

Here item has no other purpose than to introduce successive members of a list. This usage will have originated in lists, e.g. of medical recipes, remedies, where a recipe (or the like) so introduced had the same purpose as the recipe previously stated: e.g. Pelagonius 273, 'item. unctio Optati ad curam suprascriptam' (theunctio is for the same condition as that just mentioned, and item retains its comparative force). From its use in such contexts item came to be used mechanically to introduce the items in a list, even when no comparison was intended. Note, e.g. Pelagonius 169-71, 'item, ad clauum de mercurio ... item, ad cicatricem uel ut pilum ducat ...item, si dorsum motum erit'. Neither the diseases nor the treatments have anything in common, and item is purely a means of demarcation between one section and the next. This weakened use of item was as old as Cato in practical texts containing lists: e.g.Agr. 157.2, 'prima est "leuis" quae nominatur ...altera est crispa ...item est tertia, quae "leriis" uocatur . . .'(for the form of the list, with ordinals, see Apicius 11.2.6 above).

There is a curious case of item in one of the ostraca of Bu Njem which may show a further stage in the debasement of item: 86.3, 'trasmisi at te domine .item per puros tuuos gura duua semis'. Had the writer sent goods by another means, and is item accordingly comparative? 'Alternatively the word may have been used almost as a form of punctuation separating the formulaic verb-phrase transmisi ad te, domine from the equally formulaic prepositional expression.199 If this second interpretation is along the right lines, item no longer comes at the head of items in a list, and its original comparative function is completely lost.

Whatever the case, the use of item in the Vindolanda tablets is determined by the conventions operating in lists.

VIII. 2. per siluolas repto

This expression occurs in an incomplete letter (256):

adhuc per
siluolas repto tuti~r illo
futurus si remisserjs

The editors translate: 'I am still lingering in the thickets to be safer from him (?)if you release (?)him'. They are inclined to think that the writer is hiding from someone (228, 'Genialis appears to be admitting that he has behaved unacceptably towards someone and is afraid that he may have to suffer for it if his victim is released (or sent back) by Cerialis'). This is a lot to derive from a fragmentary context, and I would suggest that the implications of the expression as can be deduced from the (admittedly rare) literary occurrences of the verb (and silua, if not its extremely rare diminutive) would not support the idea that Genialis was in hiding.

Repto denotes slow movement. There is an example at Lucr. 11.318, applied to a flock of sheep as they slowly move along cropping the grass.200 Two other examples of repto are in contexts similar to each other, and closely comparable to that of the example from Vindolanda. At Horace, Epist. 1.4.4 repto is applied to the slow movement of a literary man (possibly the poet Tibullus) through a wood in silent

198 See Svennung, op. cit. (n. 98), 639 ('wertlosesitem'). Onthe appropriateness of the verb to this activity, see lg9 See Marichal, op. cit. (n. 3), 194ad loc.; also 48. C. Bailey, Titi Lucreti Can' De Rerum ~\'atura libn' sex

(1947)~ad lot.

I22 J. N. ADAMS

contemplation: 'taciturn siluas inter reptare salubris I curantem quidquid dignum . ..'An example at Plin., Epist 1.24.4 is remarkably similar. The subject is another man of letters, this time the biographer Suetonius, who wants to buy a farm (agellus) where he can take repose. Pliny says that literary owners (scholasticidomini) do not need much land: it is enough for them to wander though their land, observing the vines and shrubs (arbusculae). They are thereby refreshed: 'scholasticis porro dominis, ut hic est, sufficit abundare tantum soli, ut releuare caput, reficere oculos, reptare per limitem unamque semitam terere omnesque uiteculas suas nosse et numerare arbusculas possint'.

Reptare therefore was applicable to a leisurely stroll, taken for the purposes of pleasure and contemplation, on a country estate, amid woods or plantations (silua, arbu~cula).~~~

The linking of repto with per siluolas is strongly suggestive of leisure and contemplation. The image of an equestrian officer in hidingzo2 in the undergrowth is not convincing.

The diminutive siluula occurs at Col. v111.15.4 (though the word has no entry in the OLD).

To take illo as a masculine ablative dependent on tutior (= 'safer from him') is not compelling. 'Safe from' is regularly expressed by tutus + ab (Lewis and Short, s.v. A.P ) rather than tutus + able2031110 looks like the adverb, which usually answers the question 'whither?' (= 'to that place'). But it had acquired a secondary sense by the first century A.D., answering the question 'where?' (i.e. ='there') :see TLL ~11.1.385.7off., quoting e.g. Sen., Epist. 108.6, 'cui philosophi schola deuersorium otii sit. non id agunt ut aliqua ill0 uitia deponant'.

It has to be acknowledged that, though the implication of repto may be deducible from its use in various literary texts, the situation which lies behind the letter 256 remains obscure.

VIII .3. exsarcio

In the letter 233B Cerialis asks Brocchus to send him nets (for hunting) ('rogo mittas mihi plagas'), then in a fragmentary text asks that they should be repaired:

fortissime . .. frusta exercjas

The editors are undoubtedly right to interpret exercias as = exsarcio; for the form, see CIL ~1.4095.

What makes this example interesting is that it exemplifies for the first time the literal meaning of the verb. Sarcio seems originally to have meant 'mend, repair by sewing',z04 a sense attested for sarcio itself at e.g. Cato, Agr. 2.3 ('quae opera per imbrem fieri potuerint: dolia lauari .. .funes sarciri') and Juvenal

3.254 ('scinduntur tunicae sartae modo') and for resarcio at Terence, Ad. 121 ('discidit I uestem: resarcietur'); note too the figurative use of sarcio at Plautus, Epid. 455 ('proin tu alium quaeras quoi centones sarcias'), and sarcimen 'seam, stiching' at Apuleius, Met. IV. 15.~'~

But verbs of this root were early generalized, taking on the sense 'repair' in general (e.g. Plautus, Most. 147, 'non uideor mihi I sarcire posse aedis meas', where sarcire = 'repair', of a house).

Exsarcio has hitherto been attested only in the general sense 'repair' (Terence, Heaut. 143, 'opere rustic0 I faciundo facile sumptum exsercirent suom', Q. Cicero, Comm. pet. 45, 'aliis te id rebus exsarturum esse persuadeas', CIL ~1.4095; cf Paul. Fest. 71.9 L, but with no context), and then so rarely that it is described by Ernout-Meillet as 'archaique'. Cerialis however reverts to the etymological sense, = 'sew up' (the torn, cut pieces of a net). Frustum is characteristically used of objects which have been cut to pieces (OLD S.V. I).

If exsarcio was genuinely archaic, then Cerialis reveals a literary bent; but it seems more likely that the original sense of the term had been retained in the technical vocabulary of hunting and the maintenance of hunting nets; for damage to nets of the sort which would require such mending, see Hor., Cam. 1.1.28.

201 For a silua as the appropriate place either for con- 203 Lewis and Short, s.~. A.E.cite only Bell. Alex. 1.3for
templation or for learned discussion, see Cic., De orat, tutus + abl., and there the ablatival complement is a
111.18, Att. x11.15, and particularly Ovid, Am. 111.1 .I-6. non-personal noun.
In the last passage Ovid strolls (spatior is the verb used) in 204 See Ernout-Meillet, op. cit. (n. 55), S.V.

a wood (silua) seeking inspiration from his Muse. Notable 205 See further the detailed discussion of Ernout-Meillet,
in the passage is the association of asilua, slow movement, loc. cit.
and literary contemplation.

202 Cicero (Att. x11.15) talks of 'hiding himself away' (me
. . .abstrusi) in a wood for contemplation.

v111.4. membrum

~l/lembrumoccurs twice in lists in what appear to be similar contexts:

196 cerui[

membra n(umero) [

catacysen [

ansatam

198 ]br8 n(umero) x

1, membrA n(umero) v

The editors twice (ad locc., 170, I~I), speculating on the sense of membra, suggest that the only possible meanings appear to be 'branches'or 'parts of a catapult or ballista'. But a term of the generality of membrum could only take on such specialized meanings as a result of adjectival or some other form of contextual specification. The only possible specification in 196, which is a list of utensils and clothing (I leave aside 198, which is too fragmentary to be revealing), would come from the general contents of the list; 'part of a catapult or ballista' is out of the question, in a list of this type, and 'branches' hardly better. There is another, more general, use of membra which would be appropriate heree206 In the plural membra could signify either items (e.g, of equipment) making up a set, or parts making up a whole. For the first use, see Varro, Rust. 111.2.9, 'quid .. .est ista uilla, si nec urbana habet ornamenta neque rustica membra' (= 'items of equipment' used on a farm: 'de apparatu ut torculis, uasis uindemiatoriis sim.': TLL). For the second use, see Vitr. v11.3.7, 'solidescendo, in quibuscumque membris est formata', ('becoming solid with whatever parts it is formed'), Plin., Nut. XXXVII. 19, 'uidi . .. adnumerari unius scyphi fracti membra' (of the (broken) pieces of a single item).

It is the first use of membra which may occur in 196 (= 'items', presumably of the type referred to in the (fragmentary) previous line).

v111.5.A Lke of facio

ut scias me recte ualere

quod te inuicem fecisse

cupio

quod . . . fecisse picks up recte ualere ('I am well, a thing which I want you to have been'). It might appear odd to have fecisse taking over from a stative verb. It seems more natural to an English speaker when facio, as a transitive verb, is substituted for an earlier transitive verb (-phrase) :e.g. Cic., Tusc. 5.90, 'an Scythes Anacharsis potuit pro nihilo pecuniam ducere, nostrates philosophi facere non potuerunt' (cf. OLD s.v. 26, TLL v1.1.107.31ff.). But in fact it is not unusual forfacio to replace a stative verb (i.e, a verb or verb-phrase expressing a state rather than an action or activity): e.g. Nep., Chabr. 3.4, 'neque uero solus ille aberat Athenis libenter, sed omnes fere principes fecerunt idem', Hor., Sat. 1.1.64, 'iubeas miserum esse, libenter 1 quatenus id facit'. In these cases, as in 3 I I, facio is accompanied by a neuter pronoun. It is also worth noting Hor., Sat. 1.1.94, 'ne facias quod 1 Ummidius quidam'. Here facio does not replace an earlier verb, but is used in such a way that its subject is not agent of an action, but patient; facio is close in meaning topatior 'suffer'. Also related are examples such as Mul. Chir. 706, 'si quod iumentum clauulum in latus fecerit' (the animal does not cause the ailment, but suffers it) ;Mul.

Chir. 475, 'statim uitae periculum faciunt, quibus hoc c~ntigerit'.~~~

The subject of facio may have the role indifferently of agent or patient.

v111.6.renuntium

A formulaic set of reports (I 27-1 53) begins (after the date) with a noun renuntium (with dependent genitive) :e.g. 134 'renuntium coh(ortis) viiii Bafauorum'. Renuntius, as the editors note (74), is cited by the lexica only a few times, as a masculine (= 'reporter'). The form in -um is to be explained neither as a mistake of gender nor as an accusative standing as object of an implied verb. The uncompounded base-noun nuntius is usually in the masculine, whether it means 'messenger' or 'message, but there was

206 See TLL v111.643.73ff. Deroux (ed.), Studies in Latin Literature and Roman

207 See J. N. Adams, 'Some Latin veterinary terms History VI, Collection Latomus (~ggz), 486 n. 16, 493

relating to diseases of the back @ulmo, pulmunculus,

n. 43.

pantex, cancer frigidum, pispisa, pilupia, clauus)', in C.

I24 J. N. ADAMS

an apparently learned neuter nuntium the motivation of which was to distinguish a 'message' from a 'messenger': note Servius on Aen, x1.896, 'nuntius est qui nuntiat, nuntium quod nuntiatur'. Nonius (317 L.) notes that the neuter has no widespread authority, but is attested in some docti: 'neutro aput aliquos non receptae auctoritatis lectum est, sed doctos'. It seems to have been old (it is cited by Varro, Ling. v1.86 from the tabulae censon'ae), and it found its way occasionally into high literature (Catull. 63.75, Lucr. 1v.704).

~111.7.A Lke of interuenio (interuentus)

By chance two letters preserve comparable idiomatic uses of the verb interuenio and its derivative noun interuentus which are not otherwise attested. At 343.35-6 Octavius writes '. . . constituerat se uentumm nec interuenit'. It is almost as if the compound, following the simplex, has completive force: 'he said that he would come and did not do so'. Similarly at 291.5-8 interuentus picks up uenio; 'rogo libenter facias ut uenias ad nos iucundiorem mihi [diem] interuentu tuo factura', 'I ask you to come, to make the day more pleasant by your doing so'. But the prefix inter- does not seem to have expressed completive a~pect;~Os

one is obliged to relate these uses not to compounding in any general sense, but to the semantics of interuenio. Intemenio often means 'show up, put in an appearance', unexpectedly or by chance. The idea of unexpected, or chance, arrival is certainly not present in our examples. It would seem that the verb could now mean 'make an appearance' in general, whether unexpected or hoped for. This sense, so far as I can see, is an addition to the lexicon; it is significant that the two contexts support each other.

v111.8. caballus

Caballus occurs in an account (182.12pretio caballi). Caballus rather than equus was to produce the Romance terms for 'horse', and it is usually treated as the archetypal vulgarism. But two uses of the word in the classical period are to be distinguished. First, it often has a pejorative tone, signifying a horse despised for some reason or useful only for heavy work: e.g. Lucil. 163, 'succussatoris, taetri tardique caballi', Petron. 117.12, 'quid uos, inquit, iumentum me putatis esse aut lapidariam nauem? hominis operas locaui non caballi' (of an animal such as a draught-horse used for heavy labour). Secondly, it may be used neutrally as a synonym of equus: e.g. Hor., Sat. 1.6.103, 'nam mihi continuo maior quaerenda foret res i atque salutandi plures, ducendus et unus i et comes alter, uti ne solus rusue peregre<ue> 1 exirem, plures calones atque caballi i pascendi, ducenda petorrita' (the possession of horses is by implication a mark of status; it is unlikely in such a context that Horace meant by caballi 'despised horses'), Sat. 1.6.58, 'non ego me claro natum patre, I non ego circum me Satureiano uectari rura caballo' (Horace is not of distinguished birth; he does not ride around an estate on a saddle-horse, which by implication would be a mark of upper-class status) ;note too the magical incantation quoted by Palladius x1v.17.2, where caballus is neutral in tone and generic in meaning: 'quomodo istud iacto, sic iactentur uermes de caballo illius albo aut nigro aut cuius fuerit coloris'.

In the first use above caballus may have been almost a technical term, and as such it might even have

been acceptable in upper-class discourse in reference to an animal of conspicuously low quality. It is the

second use which is most definitely vulgar. The reference is not specifically to low-grade animals. In

higher genres equus would have been obligatorily used with this force.

In the account it is impossible to tell whether caballus was the writer's unmarked term for 'horse', or

whether the sale is recorded of a horse of poor quality. In view of the ambiguity it would not do to find in

the term further evidence for 'vulgarisms' in the tablets.

v111.9.A else of resido

The tablet 164 seems to describe the fighting habits of the Britons:

gladis non utuntur equi-
tes nec residunt
Brittunculi ut iaculos
mittant

In the context residunt can only refer to the act of sitting on horseback (to launch the javelin). The exact implication of the verb may be illuminated by the following points.

ZOs See Ernout-i\/Ieillet, op, cit. (n. jj),S.V. inter.

(i) Sedeo and derivatives are not infrequently used of sitting on the back of an equine animal (see e.g. Pe~eg~natio I 1.4 sedendo in asellis). The verb may be used absolutely in this sense:

Aethen'ae Pelagonius 168, 'si dorsum ab iniuria aut inperitia sedentis intumuerit' (= 'the rider'), 269.2, 'in sole calidosedentes exercemus' (= 'mounted'). For the compound resideo used of riding, see e.g. Ovid, Fasti 111.749, 'ut . . . piger . .. tergo residebat aselli' (also Apul., Met. VIII. 17). The example from Ovid is listed by the OLD S.V.I under the meaning 'to be or remain seated, sit'. The force of the prefix is weak or non-existent in this use.209

(ii) The distinction between sedeo and compounds, and sido and compounds is clear-cut. Sedeo is stative in meaning ('to be in a sitting position, sit'), whereas sido indicates the process whereby this state is arrived at ('to sit down') .210 Corresponding to the stative use of resideo illustrated above, there was a use of resido = (OLD S.V. I) 'to take one's seat, sit down', in which again the force of the prefix is feeble: e.g. Virg., Aen. v111.467, 'congressi iungunt dextras mediisque residunt 1 aedibus'.

Just as at Pelagonius 269.2 above in the context of horsemanship sedentes (absolute, stative) = 'mounted', so residunt in the same sort of context on the tablet is used absolutely of the act of mounting: 'nor do the Britons mount to throw their javelins'.

v111.10 . A Use ofego

A letter of Severa (292) begins, after the greeting:

ego, soror, sicut tecum locuta fueram et promiseram ut peterem a Broccho et uenirem at te, peti .. .

In Cicero's ad Atticum ego is often the first word of a letter.2H This opening is not so common in the more formal letters ad Familiares (VI. 10. I, v11.7. I, x11.24. I, x1v.4. I). It is not always easy to grasp the force of ego (e.g. Att. x11.5c, 'ego misi Tironem Dolabellae obuiam'), but with the opening of Severa's letter, cf. Cic., Att. x11.33. I, 'ego, ut heri ad te scnpsi, . . . Damasippum uelim adgrediare', xv.28, 'ego, ut ad tepn'die scnpseram, Nonis constitueram uenire in Puteolanum'. In all three places ego is followed immediately by an ut- (sicut-) clause which refers to a previous statement of the writer. In the Ciceronian examplesego introduces a confirmation of a previously stated wish or intention. In the letter of Severa, it introduces a confirmation of the fulfilment of a previously stated intention ('I did ask Brocchus, as I had promised . . .'). The usage may be classified as loosely confirmative, though the writer merely confirms something which he (or she) has said; he does not express agreement with something which the addressee has said. This latter (stronger) confirmative use212 is usually effected by ego uero, but for unaccompanied ego used thus in a letter to Cerialis, see 265 :

salutem
ego, frater, sacrificio diem
Kalendarum sic-
ut uolueras dedi-

Just as you wished, brother, I have consecrated the day of the Kalends by a sacrifice.

Similarly, when (as often) ego uero begins a letter to Atticus, it usually introduces a response to something which Atticus has said: e.g. Cic.,Att. x1.9. I, 'ego uero et incaute, ut scribis, et celerius quam oportuit feci' ('yes, I did act incautiously, as you say'). Cf. e.g. v. I. I, x.7 I, x111.3. I, x111.41 .I, x111.43.

VIII. I I. cubitom'us, cenaton'us

Tablet 196, a list of utensils and clothing, has, within seven lines, examples of cubitod[a and ~enato~a.

Both terms are also found in Petronius, the former hitherto exclusively so. Here is useful evidence for the currency in the early Empire of two technical terms to do with dining. The deverbative cubitoria is derived fromcubo, cubit~m~~~ B.2). At

in its sense 'recline for dining' (Lewis and Short, S.V. Petron. 30.1 I the term is adjectival (applied to uestimenta); it signifies 'reclining garments', of some

209 See H. Ronsch, Itala und Vulgata. Das Sprachidiom 211 Att. v1.6.1, vIII.1za.I (Pompey), 1x.4.1, 1x.19.1, der urchn.stlichen Itala und der katholischen Vul-x.2.1, x1.17a.1, XII.~~,

x11.34.1, x11.53; this list does not

gata(1875), 380 quotingexamplesof resideo =sedeo, e.g. include ego uero, on which see below.

Phaedr. 1.13.4. 212 On which see H. Thesleff, Yes and No in Plautus and

210 See Ernout-Meillet, op. cit. (n. 55), 610, S.V. sedeo: Terence (1960), 39-40.
Leumann, op. cit. (n. 33), 564. 213 See Leumann, op. cit. (n. 33), 300.

I26 J. N. ADAMS

luxury in the context. The example at 196 may be a substantivized neuter (deriving from ellipse of uestimenta). Cenatonk (n., subst.) at Petron. 21.5 (cf. 56.9) clearly refers to garments for dining: 'lassitudine abiecta cenatoria repetimus et in proximam cellam ducti sumus' (cf. Mart. x.87.12,

XIV. I 36).

VIII. I 2. Formulae, ClichCs, Formality of Style in Epistles

A letter (212) from an unknown writer to Verecundus is restored (1. 2) by the editors as follows:

occasion]em nactus sum scribendi

For this formula in epistolary contexts, see Cic., Fam. XII. I 7.2, 'me scito .. .quasi occasionem quandam . . .nactum scn'bere', Asinius Pollio ap. Cic., Fam. x.31.I, 'nunc uero nactus occasionem. ..scn'bam ad te'.214 The letters are highly formulaic, and not only in the addresses and greetings. The editors draw attention to the similar, but perhaps more formal expression amplector occasionem in the long letter which is possibly written in the hand of Cerialis himself:

225.4 liblenter amplexus ~[um do-
mine salutandi te occassjonem

This formula occurs at Plin., Epist. 11.13.1, ' et tu occasiones obligandi me auidissime amplecteris'. With libenter in Cerialis, cf. auidissime in Pliny. Also of note in the letter of Verecundus quoted above is the elegant disjunction occasionem . . . scn'bendi. There are two other notable instances of disjunction in the letters. First, in 225, note:

22 ut beneficio

tuo militiam [polssim iucundam

experiri

And at 332 (summas tibi dotominegratias) the separation of summas fromgratias recalls, as we have seen, the opening words of one of Cicero's speeches (see VII. I). While it may be possible to find 'vulgarisms' in some letters (see VIII.I~),

the extant correspondence is not of uniformly low style. Cerialis' control of a formal, even literary style, is not only observable in the phenomena noted above. In the same letter 225 the attachment, for example, of a genitive to a substantivized neuter plural adjective in the expression interpraecipua uoti (gf., = 'among my chief wishes') is a syntactic structure which belongs to the high literary language (poetry, particularly epic, historiography, particularly Tacitus; very rare in classical prose:215 cf., e.g. Tac., Ann. 1v.40.1, 'quibuspraecipua rerum ad famam derigenda'). In this case Cerialis, far from using a clichC, has produced a novel variant on an epistolary formula: cf., e.g., the formula at 0. f8di Fawikhir 2.2, 'opto deos ut bene ualeas que mea uota ~unt'.~l~

On the style of Cerialis, see further below, IX. I.

Another stock expression, in a letter from a slave to a fellow-slave, is ualde desidero at 347 (without context). Cf. Cic., Att. 11.25.2, 'ualde te exspecto, ualde desidero'; and for the collocation without a personal object, see Cic., Att. XIII. 13-14.2, 'uolo Dolabellae ualde desideranti'.

Note too 260.2-4:

i]I> no [ti
tiam tuam lubenfissi-
me perfero

The editors compare Plin., Epist. x.75.2, 'quod in notitiam tuam perferendum existimaui'. With 31 1.i.8-g ('puto me humaniusfacere qui tibi scribo . . .'), cf. Cic., Att, x11.44.1, 'fecit . humane . ..; tu . .. etiam humanius' (cf. Fam. x1.27.7, Phil. x111.36, Mart. 11.15.2).

VIII. I 3 Lexical Vulgarisms in htters

Outright lexical vulgarisms in the correspondence are few, and generally in significant letters. I return below (1x.4) to the variable meaning of the term 'vulgarism'.

See further TLL 1x.z.235.49ff. See Adams, op. cit. (n. 215), 72 n. I.
See Adams in Bowman and Thomas, op. cit. (n. 48),
72; Kiihner-Stegmann, op. cit. (n. 158), I, 433.

'27

In a letter possibly written by a slave (see Bowman and Thomas, 294), ne is unambiguously used in the sense of ne . . . quidem 'not even': 31 1.6, 'homo inpientissime qui mihi ne unam epistulam misisti'. This usage owes its interest to the fact that Quintilian (1.5.39), who was almost contemporary with the letters, notes it as a sol~ecism.~~'~\Te

is not, however, entirely absent from high literature, and its exact status is a matter for discussion (see 1x.4).

The same usage is found in the letter of Octavius (343.5). Octavius' orthography, as we have seen, has aberrational features, and his syntax and morphology too are not without colloquial or substandard elements (see above, 1v.1.4 on illec, v1.2 on parataxis).

The letter of Chrauttius to Veldedeius (310.5) has the remarkable phrase tot tempus (= 'for such a long time'), in which tot has been treated as an indeclinable singular.218 I have not been able to parallel this usage, which is possibly foreigners"br0ken Latin'. Chrauttius' own name appears to be Germanic, and his addressee and others mentioned in the letter (Thuttena, 'ilelbuteius) have non-Latin names. Chrauttius may have been a Batavian or Tungrian whose acculturation (unlike that of Cerialis) was not complete. His bizarre use of tot stands in contrast to the correct orthography of the letter, which contains the old-fashioned spelling promissit, the repeated correct use of -m, and no obvious errors. Chrauttius probably dictated to a scribe.

These are the only possible lexical vulgarisms that I have noted in the letters, but even these are not unequivocally vulgar on a strict definition of that term. I shall return to the problem of definition below. The usages in question are in letters by a slave (?),a foreigner, and the uncultured Octavius. Not much can be made of quo (= ~bi)~l~

at 215 ('si qui uolet uenire et quo lignum et materiem seruant aequo perferef'), because the letter is so difficult to understand; it does, however, have an example of qui =quis (see 1v.1.3)~ a usage which, if not vulgar, was considered substandard by some.

VIII. 14. Abliscellaneous Colloquialisms

In letter 242 note the expression bene mane: 'cyas bene mane Vindolandam ueni'. Cicero does not intensify mane in either his speeches or philosophical works, but in the letters cf. Att. 1v.9.2 'bene mane haec scripsi' (cf. x.16. I, ~IV.

18.I ;also Petron. 85.6, 'bene mane surrexi', Stat., Szlu. 117.9.48). This is the colloquial intensive use of bene, which was to survive in Romance. Cicero admits intensifying bene (both with mane and in other collocations) in the letters.220 Bene is particularly appropriate with mane, in which expression it perhaps retains a trace of its real force. iblane itself was of a root meaning 'good', the idea being that 'good time' is the morning (cf. Fr. de bonne heure).

Another letter (314) has the alternative expression primo mane (cf. Col. XII. I .3).

A request by Cerialis (233.3) is modified by the expression si me amas: 'si me amas frater rogo mittas ...'Sime amas is rare in Plautus (but see Trin. 243-4), but common in Cicero's letters (e.g.Att. v.17.5);~~~

cf. 287 ama nos (with TLL I. 1957.21ff.). In connection with Severa's unusual use of spero (291 .I I, 'sperabo te, soror') Petersmann (op. cit.

(n. s), 289) draws attention to Ter., Eun. 195, 'dies noctesque me ames, me desideres, 1me somnies, 1 me exspectes, de me cogites, 1me speres'. The usage was possibly colloquial.

VIII.I 5. Celtic I~an-words

That the tablets originate in a Celtic-influenced milieu is shown by the relative frequency of Celtic loan-words, most of them inactive in literary Latin. Such words need not of course have been taken over exclusively from British Celtic; it is likely that most had entered the Latin of Romanized natives of areas such as Gallia Belgica who were now serving in Britain.

Bracis, signifying a cereal used in beer making (three times in the tablets: see the editors on 343.25), was known to Pliny as a term in use in Gaul for which there was a more familiar Italian synonym: Nat. xv111.62 , 'Galliae quoque suum genus farris dedere, quod illic bracem uocant, apud nos scandalam'. Scandala, itself a non-literary and no doubt foreign word,222 must have been in popular use, at least in some areas: it has Romance reflexes in North Italy and the Iberian peninsula.223 Bracis, which to Pliny was merely a linguistic curiosity (a provincialism), was clearly in everyday use at Vindolanda. The

217 See Bowman, Thomas, and Adams, op. cit. (n. 146), 220 See Petersmann, op. cit. (n. ZI~), "4; also J. B.

46. To the bibliography cited there, add H. Petersmann, Hofmann, Lateinische Umgangssprache3 (19 51), 74,
Petrons urbane Prosa. CTntersuchungen zu Sprache und Hofmann-Szantyr, op. cit. (n. 145)~ 163.
Text (Sy-ztax) (1977)~ 231-2. 221 See TLL 1.1957.4ff., Hofmann, op. cit. (n. zzo),

218 See Bowman, Thomas, and Adams, op. cit. (n. 146), 127-8.
37-8. 222 See Ernout-Meillet, op. cit. (n. jj), S.V.
219 On which usage, see Svennung, op. cit. (n. 98), 383, 223 See Meyer-Liibke, op. cit. (n. 49), 7650.
Hofmann-Szantyr, op. cit. (n. 145)~ 277.

I 28 J. N. ADAMS

evidence of the tablets confirms the accuracy of Pliny's observation. Bracis is found in Irish and survives in Old French (a further hint of its Gallic origin).224

Ceruesa (seven times already in the tablets; note tooceruesanus at 182.14) was another term known to Pliny from Gaul: h'at. XXII, 164, 'ceruesia et plura genera in Gallia aliisque prouinciis'. Its currency in Gaul is suggested also by its survival in French (cerrnoi~e).~~~

The frequency of the word at Vindolanda suggests that the drink itself and brewing had been brought to Britain from Gaul by soldiers. The form at Vindolanda is consistently ceruesa, though Meyer-Liibke gives cereuisia as the base of the Romance reflexes; ceruisia is attested sometimes in Latin Bracis and ceruesa were clearly characteristic of regional dialects of Latin (though not exclusively that of Britain).

Bedocem occurs in an account (listing various textiles) at 192.2. There are two significant examples of (3k8oEin Diolectian's Prices Edict. That at 19.56 is qualified by the adjective l?ahh~x6~,

that at 19.58 by No~tx6~.

Noricum was a Celticized Alpine province, and given too the 'Gallic' associations of the wordiobject (so 19.58), there can be no doubt that bedox was a Celtic term.227 In the Latin version of the Prices Edict Lauffer restores the Latin form of (3880tas fedox; it is now clear that he should have written bedox.

Remarkably, there is a second Celtic word in the same account, tos~eas.~~*

Tossia (cf. Breton toos) had previously been known only from the Gallic inscription CIL x111.3162,11 found at Vieux and now at Thorigny. The inscription, dated A.D. 238, records a specimen of a letter sent by Tiberius Claudius Paulinus, governor of Lower Britain, from an unidentified British town T~mpium.~~~

Tossiam is given the epithet Bht(annicam). The account in which it occurs at Vindolanda lists items acquired a Gauuone. Gauo was probably a Celtic name (see Bowman and Thomas, 160); Gavo may have been a Celtic (British) entrepreneur supplying (among other things) traditional Celtic goods. It is in transac- tions of this sort that loan-words might have entered Latin in the region of Vindolanda. I would stress that there are two possible routes by which Celtic terms could have found their way into the Vindolanda tablets: some will have been brought from the Continent by soldiers transferred to Britain, while others may have been picked up in Britain through contact with the local population.

In the letter of Chrauttius (310) both the addressee Veldedeius and one of the persons mentioned in the letter (Velbuteius) probably have Celtic names (see the editors on 310.1, 17).

In this section I have restricted myself to Celtic terms inactive in mainstream Latin. It goes without saying that the tablets also contain some words of Celtic origin which had long since entered the Latin literary language (sagum, raeda) .

To the lexical evidence for Celtic influence on Latin collected in this section can be added the

Celticized pronunciation of a Latin word which lies behind souxtum (see 11.10).

IX. CONCLUSION : LINGUISTIC COMPETENCE, LITERACY AND
SOCIOLINGUISTIC VARIATION AT VINDOLANDA

The tablets are a corpus of non-literary Latin composed during a relatively short period in or near a single outpost of the Empire. There is evidence that the authors of the documents were not of uniform social (or military) status. Many of the soldiers at Vindolanda must have been of 'barbarian' (i.e. Batavian or Tungrian) origin. The term 'Vulgar Latin', implying as it does a unity of sorts, cannot without qualification be used of the output of such a disparate group. Like the ostraca of Bu Njem, the tablets raise the question of the degree of acculturation of foreigners serving in the army. In this section I attempt a general overview of the Latinity of the tablets, giving particular attention to the variables which have contributed to the mixed character of the language.

224 See J. AndrC, Les noms de plantes dans la Rome 227 SO too A. Holder, Alt-Celtischer Sprachschatz antique (1985), 37. (1896)~ 366. 225 SO too in the Iberian ~eninsula (Mever-Liibke, OD.. 22s See 1. Andre. 'Tossia "couverture de lit"'. ~tudes

>, ,

cit. (n. 49)) 1830). ~elti~ues"11 (1~64-~),409-12.

226 See TLL 111.943.71ff. A bilingual exercise has been 229 On the place and the inscription, see A. L. F. Rivet
plausibly attributed to Gaul on the grounds that it con- and C. Smith, Theplace-NamesofRomanBritain(1979),
tains both bracis and centisia:see A. C. Dionisotti, 'From

467.

Ausonius' schooldays? A schoolbook and its relatives', JRS 72 (1982), 123.

LANGUAGE OF VINDOLANDA TABLETS

129

IX.I. Foreigners (i.e, non-native speakers of Latin) at Vindolanda (?)

Flavius Cerialis, prefect of the Ninth Cohort of Batavians, is likely to have been a Batavian noble. According to Tacitus (Hist. IV. 12), Batavian units in the Roman army were commanded by their own nobiles. The cognomen Cen'alis was surely taken, as the editors suggest (25), from Petillius Cerialis, who suppressed the Batavian revolt of A.D. 69-70.~~~ Either Cerialis himself or his father must have received the citizenship for loyalty to Rome during the uprising. It is then likely that Cerialis was only a first- or second-generation Roman (but which?: see below), and, as commander of Batavi, a Germanic (or Celtic(?)) speaker.

Of the documents so far published, the draft letter 225, which the editors argue is in the hand of Cerialis himself. Dresents the most formal and literarv Latinitv. Its orthonra~hv is

, L nLJ

consistently correct, and it has two types of old-fashioned spelling (the etymologically correct -ss- in occassio, twice, and saluom) favoured by some of the scribes at Vindolanda. But more striking is its accumulation of formal literary phrases, word order, and syntax of a type which cannot simply be explained as manifesting hackneyed epistolary clichCs. I refer to the ablative absolute in 1. 2, the expression libenter a?nplexus sum occassionem +genitive of the gerund (see VIII.I~), the combination of the vocative domine with the appositional expression dominum meum (see VII. I), the expression spei compo~,~~~

the syntax of interpraecipua uoti (see VIII. 12), the phraseology quomodo uoles imple at 20,~~~

the use of instrmo at 22 (amicis ita instr~e),~~~

and the phraseology and word order of militiam [polssim iucundam experin' (23: see VIII. 12). It may be no accident that a number of parallels for the phraseology in the letter have been found in Pliny (see also above, VIII.I~), who was roughly contemporary with the Vindolanda tablets. If we can accept the plausible argument that 225 was by Cerialis himself, and that Cerialis was a Batavian noble, then it would seem to follow that he had been formally trained in the upper-class Roman literary culture.

Other features of the Cerialis archive suggest, albeit indecisively, that he was completely Romanized. He uses exsarcio (233) in its previously unattested etymological sense, either as an archaism or as a rare technical term. He admits the instrumental use of qui (234), which was surelv all but defunct bv this date. Bene mane (242')is good educated idiomatic Latin. as is the

,,, "

use of ego at 265. The correspondence between Severa, the wife of Brocchus, and Cerialis' wife Sulpicia Lepidina (291-2) is also consistent with assimilation to Roman culture. Severa's Latin is elegant, colloquial, and syntactically correct; on her use of ego, see above, v111.10. If Cerialis' had been trained in the school of a grammaticus and perhaps even a rhetor, he is unlikely to have been a first-generation Roman citizen. Literacy in Latin was not a normal accomplishment even of Germanic chiefs until at least the fourth century,234 but we do occasionally hear of schools for the education in Latin of the offspring of provincial Clites (see Tac., Agr. 21, Ann. 111.43, Plut., Sert. 14). If Cerialis' father had received the citizenship, the son mav have been Romanized in a ~rovincial school.

~hrauttius, author of the let& 310, must, in view of his name, have been of either Batavian or Tungrian origin. His addressee and two of the persons referred to in the body of the letter also have non-Latin names.

There is a contrast between the syntax of Chrauttius' letter, and its orthography (see VIII.13). Twice, for example, he appears to introduce a new person to the discourse by means of a syntactically unconstrued accusative (see VI. 1.8), yet the spelling of the letter is correct. Chrauttius was probably dictating to a scribe, who had been taught to spell correctly, but was prepared to keep Chrauttius' odd phraseology. The variation in the form of the name of the addressee, VeldeiusiVeldedeius, is also suggestive of dictation. The language of the letter is to some extent formulaic, but there is sufficient departure from mere clichCs to show that Chrauttius' Latin was not unidiomatic. The initial greeting suo fratri contubernali antiquo plurimam salutem displays a creative variation on the normal formula.235 The parenthetical miror quod-clause separating rogo te (4) from its dependent construction can on the one hand

230 See further A. K. Bowman, 'The Roman imperial 233 See Bowman and Thomas, op, cit. (n. 48), 131, citing army: letters and literacy on the northern frontier', in A. e.~.Plin., Epist. x.28.

K. Bowman and G. D. Woolf (eds), Literacy and Power See A. D. Lee, Infomzation and Frontiers. Roman in the Ancient World (1994), I 11. Foreign Relations in Late Antiquit)' (1993), 28. 231 See Bowman and Thomas, op. cit. (n. 48), 129, citing 235 See Bowman, op. cit. (n. 230), 124. Livy xx1x.zz.5.

232 On which see Bowman and Thomas, op. cit. (n. 48),
131, citing Plin., Epist. 1.20.25.

I3O J. N. ADAMS be paralleled in colloquial texts,236 yet on the other hand exhibits command of a fairly complex sentence structure. ..lparentibus nostris siquid audieris in idiomatic fashion has the expression a payentibus nostm's focused by its position outside the si-clau~e.~~~

The spelling yescrz3sti reflects the sound of the spoken language. And the use of (promissit) pyetio is idiomatic.

If we can assume that the scribe was taking down dictation rather than writing Chrauttius' letter for him, then we might conclude that Chrauttius was a speaker of an idiomatic, but non-standard, variety of Latin. Stylistically the letter is at some remove from letter 225, but it is more suggestive of 'Vulgar Latin' than of 'foreigners' Latin'. The only item which might possibly fall into the second class is the use of tot. If then Chrauttius was a foreigner, he had picked up in the army a form of colloquial Latin. Similarly the Africans at Bu Njem had acquired a type of Vulgar Latin, though their efforts at writing do display abnormalities which are no ordinary vulgarisms. Foreigners recruited into the army were expected to learn and use Latin, but there must have been varying degrees of mastery among common soldiers. Chrauttius was well assimilated, but perhaps in tot tempus one can hear the voice of a second-language learner. blass- and count-terms are often the subject of cross-language interference.

'Foreigners' Latin' at Vindolanda seems also to be reflected in the Celticized spelling souxtum = suptum (see 11. IO), in a letter written by one slave to another. Presumably the spelling represents the slave's pronunciation of the Latin word, whether he wrote the letter himself or dictated it to a scribe. This item is the best evidence that we have that there were speakers at Vindolanda whose Latin showed substrate influence.

Tablet 192 possibly introduces us to a foreigner of another type, Gavo. The document records the receipt of goods from Gavo. Two of the items have Celtic names (bedocem, tosseas), words scarcely attested in Latin. Could Gavo have been a local trader supplying the army with goods, some of them of British type (and name)? It is in commercial intercourse of this type, as we have suggested (~III.I~),

that Celtic loan-words might have found their way into Latin.

1x.2.Scm'bal Practices

Orthography has been discussed earlier, and I add no further detail here. I would stress on the one hand the degree of orthographic correctness in many of the documents and the presence of old-fashioned spellings, and on the other the aberrant character of Octavius' spelling, which gives support to the idea that he may have been a civilian trader without access to military scribes. There was an educated secretariat at Vindolanda. Scribes were employed both by the cultivated (e.g. Cerialis, who it seems sometimes wrote his own letters, and sometimes had scribes write for him), and by speakers of substandard Latin (e.g. Chrauttius). While scribes successfully avoided a number of types of spelling 'errors' (e.g. omission of h and final m, e for ae), they were incapable of avoiding misspellings involving the treatment of vowels in hiatus. From this I conclude that vowels in hiatus had been so radically modified in a variety of ways that deviations from what might be called the classical (literary) norm were no longer perceived as errors.

1x.3.Literacy Below the Level of the Social ~lite

It has been argued by Bowman (op. cit. (n. 229)) that there is evidence at Vindolanda for literacy among those who were below the level of the social or cultural Clite.238 A linguistic feature of the renuntium documents throws further light on this contention. These docu- ments, of standard format, come from a variety of hands, probably of optiones. The various writers repeatedly wrote debunt for debent, even though in an account (181) the verb retains its second-conjugation form debent. It was observed earlier (1v.2. I) that various deductions

236 See Bowman, Thomas, and Adams, op. cit. (n. 146), wrote poems (see Adams, op. cit. (n. 18), 112). The

37. military inn~a~eo;Apsyrtus was written to by decurions

237 For this phenomenon, see H. Pinkster,Latin Syntax for advice about the treatment of their horses (see Corpus

and Se>izantics (~ggo), 170. Hippiatn'corzim Graecon~nz I, passim).

238 It is worth noting that two centurions at Bu Njem

can be made from the repetition of the error. First, the original exemplar of the renuntia, though clearly of official or semi-official status, could not have been drafted by a member of the highly educated classes or by one of the scribes at Vindolanda. It is significant that the account 181, containing the correct form debent, was written by a scribe who was also employed to draft personal letters: the hand is the same as that of 344, a letter of appeal. 344 is a document which contains both an etymologically correct -ss-spelling (commississem) and an archaizing ablative in -i (ua[let]udini), both tell-tale traits of the scribal class at Vindolanda. Clearly that class said debent not debunt. Secondly, the form debunt for debent would not have been persistently perpetrated by a group of writers, without ever being corrected, unless it were the standard form in the speech of every member of the group. The change debunt to debent is so minor that a writer unhappy about the form debunt might well have changed the spelling without disturbing the standard format. The renuntia thus give us an intriguing glimpse of a social class (probably that of the optiones) who regularly used the substandard form debunt, yet were literate.

1x.4. Vulgar Latin, Technical Latin, the Influence of 'Genre '

Since a good deal of the Latin which has turned up at Vindolanda emanates from well trained scribes andlor the officer class, it should not be labelled mechanically as Vulgar Latin simply because it is non-literary. But the tablets do have material relevant to the study of Vulgar Latin (on the term see below) and of developments in the spoken language. It is not inconsistent on the one hand to assert that many documents come from the hand of scribes with a taste for formality, and on the other hand to seek information about the spoken language in the same documents. I would stress the following points:

(i)
A learned scibe might use correct orthography, but retain the lexical or syntactic errors which were dictated to him (note, e.g. the letter of Chrauttius (~Io), and also 31 I, with ne = ne . . . quidem).
(ii)
Not all writers at Vindolanda belonged to the class of educated scribes (note, e.g. the renuntia, the letter of Octavius, and the account 186).
(iii) Even professional scribes were not necessarily of the highest educational attainment. Scribes at Vindolanda are consistently correct in some respects, but consistently incorrect in others, and this would suggest that they did not belong to that literary Clite which might be capable of classing as a substandard deviation even a usage (such as the contraction of ii to i) which was deeply entrenched in the speech community.

The term 'Vulgar Latin' has been often criticized, and it is unsatisfactory, implying as it does that there was a single entity 'Vulgar Latin' distinct from another entity such as 'literary Latin'. Even a reasonable definition such as that of Coleman239 -'By Vulgar Latin is meant primarily that form of the language which was used by the illiterate majority of the Latin-speaking population' -runs into difficulties. The form debunt, for example, which was completely excluded from all varieties of literature, can be placed at its upper social level in the speech of under-officers such as optiones and decuriones, who were not illiterate. To describe -unt for -ent merely as a 'vulgarism' would be imprecise, because it was current above the level of the illiterate uulgus. Nevertheless it belonged to a variety or varieties of the language clearly distinguished from that of the educated Clite. It is perhaps best to think of Latin as a single language which embraced the usual types of sociolinguistic and dialectal variations. Those usages which were departures from the educated norms as represented in high literature might differ in the degree and nature of their unacceptability to the educated.

The case of ne for ne . . . quidem (twice in the tablets) is subtly different from that of -unt for -eat. It is true that Quintilian (1.5.39) singles out ne as a soloecism, and his remark shows that there were those among the literate who would have found it unacceptable. But grammarians did not necessarily look much beyond their own social class for usages to brand as soloecisms. When 'Sergius' says (GL 1v.517.24) 'nemo enim (dicit) ab ante', he probably means something like 'no-one should say ab ante', or 'none (of us, who are highly educated) says ab ante'. There is perhaps an implication that ab ante had penetrated, to the author's

239 R. G. G. Coleman, 'Vulgar Latin and Proto.
Romance: minding the gap', Prudentia 25 (1993), 2.

J. N. ADAMS

distaste, the social dialect of those around him. Aie, unlike -unt for -ent, is not excluded from high literature. It is used not only by Trimalchio in the Cena Trimalchionis (47.4), but even in the main body of Petronius' novel, at 9.6 in a remark by Encolpius. Typically, editors emend the text at 9.6, but with what justification? :lie for ne . ..quidem is also admitted in the novel of Apuleius, again in speeches (1.23, III.II).~~O

There are no grounds for classifying ne as a vulgarism in the strict sense; indeed it may even have originated not in lower-class speech, but in the speech of the educated. In this milieu, as an innovation, it was perhaps accepted by some, but frowned on by others (such as the purist Quintilian). The acceptability or otherwise of a usage was also influenced by genre. A usage avoided in some varieties of literature (or writing) might be acceptable in others. It is surely no accident that ne turns up in novels. The novel was stylistically less exclusive than some other forms of high literature. And what was acceptable in a novel would presumably cause no comment in a private letter.

The influence of genre in determining the acceptability of a non-standard usage may also be illustrated from the case of siqui =siquis. I leave aside siqui before s. Vitruvius, as we saw, began by using si quis then switched to si qui. In no meaningful sense could the Latin of the highly educated Vitruvius be described as vulgar. But technical writers in Latin did not necessarily adhere to the purist conventions of the higher literary forms. Siqui =siquis clearly had some currency among the educated, even if there were genres in which it was avoided. Again it is found at Vindolanda in a letter.

The use of item discussed at VIII. I will not be found in forms of high literature. But that does not mean that it was usually restricted to the speech of the illiterate masses. It belongs to a particular genre, in this case listsirecipes/accounts, and it is in these that it can be seen to have developed its special nuance.

Another non-standard, but not necessarily vulgar, phenomenon is the use of the accusative seen in accounts (VI. I .6). It was determined by the conventions of the genre, not by the speech patterns of the illiterate.

I move on to some usages which, at first sight at least, have more claim to be considered features of Vulgar Latin. In caballus Vulgar Latin and proto-Romance might seem to come together, because caballus survives in all Romance languages and must have become the standard term for 'horse' in the speech of the majority of the population across the Empire. But there are two possible determinants of its use in the account. The writer, influenced by popular usage, may have employed caballus as his generic term for 'horse'. But we have seen (v111.8) that there is evidence (from such literary genres as satire and the novel) that in the early period

(i.e. the Republic and early Empire) caballus was not yet exclusively a generic term, at least in the speech of the literate classes. It is used particularly of lower-quality animals of the sort which might be put to heavy work. In an account it is at least as likely to be a technical term as a vulgarism.

The reduced possessive form ma (if it is not simply a dittography) also has a superficial claim to be regarded both as a vulgarism and as an anticipation of a Romance form. Forms such asma, tus etc. are completely avoided in all forms of literature, yet reflected in Romance. They will have developed in, and been restricted to, the spoken language. But the spoken language of what class? Of the illiterate classes (the uulgus)? Not necessarily. The example at Vindolanda was employed by a member of the immediate circle of the prefect Cerialis. Ma is used by a woman, in an affectionate expression. It is possible that nzus, tus, sus were in use in the colloquial speech of all social classes, but never written. If so, they cannot be classified as vulgarisms. The term 'vulgarism' implicitly contrasts the practice of the uneducated masses with that of the educated. But the contrast in terms of which the register of ma may be explained is that between informal speech (of whatever class) on the one hand and formal speechiwriting on the other. A good deal of the vocabulary, morphology, and syntax which survived from Latin into Romance was not restricted to the speech of the hypothetical uulgus but was common to all social classes, and this fact undermines the notion that Romance developed out of Vulgar Latin. Much of the classical Latin verb-morphology, for example, passed into Romance.

240 See Petersmann, op. cit. (n. 217), 231-2.

The spelling turta = torta raises an issue of a different kind. Torta hardly turns up in literary texts, but that is not because it was a vulgar word, but because the object which it signified (apparently a sort of twisted loaf) was not the sort of thing which would inevitably come up in literary genres. But what of the spelling with u rather thano? The spelling reflects a current pronunciation, with r +consonant causing closing of the preceding vowel, but there is no means of knowing whether that pronunciation was standard in all social dialects, or only in the speech of the uncultivated. The spelling itself may justifiably be described as non-standard and as reflecting a certain lack of training in spelling on the part of the writer, but it cannot be argued from there that the speech of the writer would have differed from that of a better educated writer who might well have pronounced the word withu even though writing it witho.

I stress the distinction between substandard spelling and non-standard ('vulgar') pronun- ciation, because the two are constantly confused in handbooks of Vulgar Latin. A substandard spelling need not imply that the pronunciation which it represents was also considered substandard. The omission of final -m, for example, habitually finds a place in discussions of Vulgar Latin. But we know from the evidence of grammarians that -m was lost in the speech even of the educated class. What set the educated apart from those who had not been fully immersed in the literary culture was not necessarily the way in which they pronounced accusative singulars, but the form in which they wrote them.

Other spellings attested in the Vindolanda tablets which, while substandard as spellings, were probably based on widespread rather than distinctively 'vulgar' pronunciations are those which involve the treatment of vowels in hiatus. The spelling Februuar-, on the other hand (see 11.I), is decidedly aberrant. It may reflect an idiosyncratic hypercorrect glide-insertion in the speech of the writer, and that might be seen as a reaction against a tendency, not yet fully established outside vulgar speech, for the u to be lost in pronunciation.

Some termslforms which may have been vulgar in the restricted sense (i.e. characteristic of the uulgus rather than of the educated) are exungia and quem = quam. Neither is a mere phonetic misspelling. Both eventually found their way into texts, but not for some centuries. The status of tot (tempus) is impossible to determine: vulgarism or 'broken' Latin? The use of the accusative expressing price may have been restricted to lower social dialects. The demonstrative illic (represented by illec =illaec, neut.) seems to have died out in the Latin of the literary classes; there is growing evidence that it lingered on in lower-class speech.

The tablets throw up a number of words which are either extremely rare or otherwise unattested, at least in the senses in question. I draw particular attention to coriatio, excussorium, locam'um, exsarcio, cubitom'us, cenatom'us, and explico, which has not been discussed here (see also below, IX.~).~~~

TOapply the designation 'vulgarism' to any of these would be inappropriate: they belonged rather to technical vocabularies of the sort which leave little mark on higher literary genres.

I have sought in this section to express reservations about the expression 'Vulgar Latin', instead of using it as a blanket term to embrace a variety of phenomena. One phenomenon, the conflated thirdlsecond declension form debunt, we have been able to locate in a precise social milieu. It is usually impossible to classify socially an aberrational form, and in this respect therefore the Vindolanda evidence is very valuable.

IX. 5;. New Linguistic I!latem'al in the Tablets

There is a good deal of new evidence in the tablets relevant to the history of the language and its technical varieties. I mention in summary:

(i)
New words, or first attestations of words: excussom'um, coriatio, sagacia, bubulcarius, superaria ;
(ii)
Known words in new meanings: interueniolinteruentus, exsarcio, tot;
(iii) Anticipations of Romance: locam'um, turta (spelt with u rather thano), ma, the use of the masculine form of the relative pronoun for the feminine;

(iv) First attestations of abnormal forms of words :exungia, renuntium, Februuar-, legiona- ris ,carrulum ,modiolum ,radium (?).

241 See the editors, 324 (on 343.4).

I34 J. N. ADAMS

The abundant examples of the apex allow some tentative conclusions to be drawn about the shortening of long vowels in final syllables. The obsolescence of the -ilocative morpheme, and its replacement by a locatival ablative -0, are confirmed. And the numerous lists and accounts provide evidence for the case syntax of sentences without expressed verbs.

University of Reading

This paper is published with the aid ofgrants from The British Academy, The Administrators of the Haverjield Bequest, and The University of Reading

Comments
  • Recommend Us