¹ Marichal, R., Les ostraca de Bu Njem (Suppléments de ‘Libya Antiqua’ VII) (1992)Google Scholar.

² See R. G. Goodchild, ‘Oasis forts of Legio III Augusta on the routes to the Fezzan’ PBSR 22 (1954), 57; Marichal, Bu Njem, 107 (with map).

³ See AE 1976, 698, ‘Vexillatio leg(ionis) III Aug(ustae) P(iae) V(indicis) S(euerianae) quae at castra Chol. aedific(anda) uenit Muciano et Fabiano co(n)s(ulibus) VIIII kal(endas) Febr(uarias) et reuersa est Antonino II et Geta Caes(are) Augg(ustis) co(n)s(ulibus) VII kal(endas) Ian(uarias)’. See further Speidel, M. P., Roman Army Studies II (1992), 275–6Google Scholar, Le Bohec, Y., La troisième legion Anguste (1989), 441, n. 443.Google Scholar

⁴ Goodchild, op. cit. (n. 2), 58.

⁵ See Nos 28, 71, 72 (with Marichal's notes in the latter two cases), 147.

⁶ See the inscriptions (numbered 67–15, 67–89) published by Rebuffat, R., Deneauve, J., and Hallier, G., ‘Bu Njem 1967’, Libya Antiqua 3–4 (1966–1967), 97, 102.Google Scholar

⁷ See the inscription 67–15, in Rebuffat et al., op. cit. (n. 6), 97.

⁸ See Le Bohec, Y., ‘Un nouveau type d'unite connu par l'épigraphie Africaine’, in Hanson, W. S. and Keppie, L. J. F., Roman Frontier Studies 1979, Papers Presented to the 12th International Congress of Roman Frontier Studies (1980), 950Google Scholar.

⁹ See Marichal, Bu Njem, 65.

¹⁰ The daily reports which form part of the material (Nos 1–62) present between forty two and ninety six men in active service (Marichal, Bu Njem, 70).

¹¹ Bu Njem, 65–6.

¹² On Aemilius, see Marichal, Bu Njem, 58 with n. 3, 61. Aemilius is a common name in Africa. It has been (1976), 26. suggested that it may have been chosen by some of its African bearers because of its similarity to the Punic name Himilis: see Birley, A. R., ‘Names at Lepcis Magna’, Libyan Studies 19 (1988), 4–5.CrossRefGoogle Scholar

¹³ Marichal, Bu Njem, 64.

¹⁴ Marichal, Bu Njem, 65. On Syrians in the Roman army in Africa, see in general Le Bohec, Y., ‘Les Syriens dans l'Afrique romaine: civils ou militaires?’, Karthago 21 (1987), 81–92.Google Scholar

¹⁵ See Segert, S., A Grammar of Phoenician and Punic (1976), 26Google Scholar.

¹⁶ See, e.g., Aug., Epist. 66.2, 108.14, 209.3, Expositio ad Romanos inchoata 13.1.

¹⁷ See Brown, P., ‘Christianity and local culture in late Roman Africa’, JRS 58 (1968), 87–8.Google Scholar

¹⁸ See Millar, F., ‘Local cultures in the Roman Empire: Libyan, Punic, and Latin in Roman Africa’, JRS 58 (1968), 132Google Scholar; also Elmayer, A. F., ‘The re-interpretation of Latino-Punic inscriptions from Roman Tripolitania’, Libyan Studies 14 (1983), 86–95CrossRefGoogle Scholar; idem, ‘The reinterpretation of Latino-Punic inscriptions from Roman Tripolitania’, Libyan Studies 15 (1984), 93–105.

¹⁹ Millar, op. cit. (n. 18), 128–9. On the continued vitality of Punic/Libyan culture in Roman Tripolitania, see in general Mattingly, D. J., ‘Libyans and the “Limes’: culture and society in Roman Tripolitania’, Antiquités africaines 23 (1987), 71–93CrossRefGoogle Scholar, esp. 73–80; and on the survival of African languages, also Harris, W. V., Ancient Literacy (1989), 179–80Google Scholar.

²⁰ cf. e.g. CIL IV.221, cum sodales, 8976 cum iumentum.

²¹ See 8 cum kamellos, 67 cum litteras, 73 cum suriacas,103 cum epistulas duas (cf. 104, 105); cf. 28 cum Garamanti[bu]s, 37 cum asinis.

²² See Marichal, Bu Njem, 62.

²³ See TLL II. 2103.81ff.

²⁴ See Tekavčić, P., Grammatica storica dell' italiano II: Morphosintassi (1972), 568Google Scholar; Rohlfs, G., Historische Grammatik der Italienischen Sprache III (1954), 127Google Scholar (§887); also Meyer-Lübke, W., Romanisches etymologisches Wörterbuch³ (1935), 1028, 1208Google Scholar.

²⁵ Thomason, S. G. and Kaufman, T., Language Contact, Creolization, and Genetic Linguistics (1988), 168–70.Google Scholar

²⁶ See Green, J. N., ‘Romance Creoles’, in Harris, M. and Vincent, N. (eds). The Romance Languages (1988), 441Google Scholar, Holm, J. A., Pidgins and Creoles I: Theory and Structure (1988), 53Google Scholar.

²⁷ See Marichal, Bu Njem, 58.

²⁸ See Sankoff, G., ‘Variation, pidgins and Creoles’, in Valdman, A. and Highfield, A. (eds), Theoretical OrientationsGoogle Scholar

²⁹ See in general J. C. Richards, ‘A non-contrastive approach to error analysis’, in idem (ed.), Error Analysis: Perspectives on Second Language Acquisition (1974), 172–88.

³⁰ idem, 172–3.

³¹ idem, 174–8.

³² See Hakatu, K., ‘A case study of a Japanese child learning English as a second language’, Language Learning 26 (1976), 332–3.Google Scholar

³³ See, e.g., Ferguson, C. H. and DeBose, C. E., ‘Simplified registers, broken language, and pidginization’, in Valdman, A. (ed.), Pidgin and Creole Linguistics (1977), 108–9Google Scholar.

³⁴ Apul., Apol. 98.8–9, ‘loquitur numquam nisi Punice … enim Latine loqui neque uult neque potest. audisti, Maxime, paulo ante, pro nefas, priuignum meum, fratrem Pontiani, diserti iuuenis, uix singulas syllabas fringultientem’.

³⁵ On Secedi, see Marichal, Bu Njem, 62, 75, 106.

³⁶ See Segert, op. cit. (n. 15), §62.3 on determination in Punic; also §74.24–241.

³⁷ Segert, op. cit. (n. 15), §56.5; also §66.131, ‘The nota accusativi is used for connecting the determined … direct object to its verb’.

³⁸ Segert, op. cit. (n. 15), §74.241.

³⁹ See Sznycer, M., Les passages puniques en transcription latine dans le ‘Poenulus’ de Plaute (1967), 48Google Scholar; also 64, 83, 86.

⁴⁰ As an example of formulaic composition in a military context, I would cite the so-called renuntium documents from Vindolanda: see Bowman, A. K. and Thomas, J. D., ‘New texts from Vindolanda’, Britannia 18 (1987), 132–5CrossRefGoogle Scholar; eidem, The Vindolanda Writing Tablets (Tabulae Vindolandenses II) (1994), 73–6.

⁴¹ See Marichal, Bu Njem, 59.

⁴² For such formulae, see Cugusi, P., Corpus epistularum latinarum papyris tabulis ostracis seruatarum (1992), II 56.Google Scholar

⁴³ cf. Marichal, Bu Njem, 101.

⁴⁴ See Marichal, Bu Njem, 269 S. V.

⁴⁵ See, e.g., Claudius Terentianus, P. Mich, VIII. 468.8, ‘misi t[i]bi pater per Martialem imboluclum’, and Cugusi, P. P., Evoluzione e forme dell' epistolografia latina nella tarda repubblica e nei primi due secoli dell' impero (1983), 101, 278–9Google Scholar; idem, Corpus epistolarum latinarum, I, 24; II, 59 (on 73.4–5).

⁴⁶ See, e.g. 86, and the other examples cited by Marichal, Bu Njem, 269 S. V.

⁴⁷ See Marichal, Bu Njem, 265, S. V.

⁴⁸ See Marichal, Bu Njem, 60.

⁴⁹ See the material cited by Marichal, Bu Njem 60; also Cugusi, P.Corpus epistolarum latinarum, II, 315 (on 214.6–7).Google Scholar

⁵⁰ Note 5 ad balneu (cf. 7, 17, 34, 39), 5 ad prepositu (cf. 12, 13, 30, 32, 34), 26 ad fiscu, 36 ad lignu, 46 ad pretoriu, 71 seruu fugitiu, 94-atianu, 99 asinu (twice), 99 decimu.

⁵¹ It is worth stating that part of this inscription is now lost.

⁵² cf. G. Konjetzny, ‘De idiotismis syntacticis in titulis latinis urbanis (C.I.L. Vol. VI) conspicuis’, ALL 15 (1908), 330; Svennung, J., Untersuchungen zu Palladius und zur lateinischen Fach- und Volkssprache (1935), 253Google Scholar.

⁵⁴ But note the examples of usque + names in the ablative cited by Svennung, op. cit. (n. 52), 253 from an Episcoporum catalogus.

⁵⁵ On this construction in late Latin, see now Helttula, A., Studies on the Latin Accusative Absolute (1987)Google Scholar.

⁵⁶ See Helttula, 6.

⁵⁷ See Sommer, F., Handbuch der lateinischen Laut und Formenlehre I⁴ (revised by R.Pfister) (1977), 203Google Scholar.

⁵⁸ See Marichal, Bu Njem, 188 ad loc.

⁵⁹ See Löfstedt, E., Syntactica. Studien und Beiträge zur historischen Syntax des Lateins (1956), II, 79–96.Google Scholar

⁶⁰ P. Cugusi, Corpus epistolarum latinarum II, 152 takes ea as a nominatiuus pendens.

⁶¹ On this term, see Marichal, Bu Njem, 110–11.

⁶² See Marichal, Bu Njem, 70.

⁶³ On the interpretation of refuga, see Marichal, Bu Njem, 110.

⁶⁴ I cannot parallel precisely piciparis for principalis. The omission of n before the stop is not exceptional. Such spellings are common at Pompeii, in the documents of Eunus, and in other colloquial texts: see, e.g. Väänänen, V., Le latin vulgaire des inscriptions pompéiennes (1966), 67–8Google Scholar, Kiss, S., Les transformations de la structure syllabique en latin tardif (1972), 29–30Google Scholar, J. N. Adams, ‘The Latinity of C. Novius Eunus’, ZPE 82 (1990), 241. The suffix -alis must have assimilated to -aris (see Schopf, E., Die konsonantischen Fernwirkungen: Dissimulation, Fern-Assimilation und Metathesis (1919), 137Google Scholar for assimilation of l > r, but the examples which he quotes are regressive – e.g. cereberrimo for celeberrimo – rather than progressive), and then a dissimilatory loss of the first r must have taken place (i.e. r - r > ø - r: cf. e.g. castrorum > castorum, cited by Schopf, 150).

⁶⁵ Isolated cases of the nominative for accusative can be found admitting of various explanations (see further below): see e.g. Lofstedt, B., Studien über die Sprache der langobardischen Gesetze (1961), 215–17Google Scholar. Gaeng, P. A., ‘La morphologie nominale des inscriptions chrétiennes de l'Afrique’, in Iliescu, M. and Marxgut, W. (eds), Latin vulgaire – latin tardif III. Actes du IIIéme Collogue international sur le latin vulgaire et tardif (Innsbruck, 2–5 septembre 1991) (1992), 118Google Scholar, cites two alleged cases of nominative for accusative from African inscriptions, adding ‘je n'ai trouvé aucun exemple de ce remplacement dans d'autres provinces étudiées’. His first example, intus aque dulces (= Diehl, E., Inscriptiones Latinae Christianae Veteres (1925–1931), 785Google Scholar), is something of a curiosity. Intus is not a preposition, and in any case the words are correctly quoted from Virgil (Aen. I.167), as Diehl points out.

⁶⁶ Herman, J., ‘Recherches sur l'évolution grammaticale du latin vulgaire: les emplois “fautifs” du nominatif’, Acta Classica Univ. Scient. Debreceniensis 2 (1966), 109–12Google Scholar; = idem, Du latin aux langues romanes. Etudes de linguistique historique (ed. S. Kiss) (1990), 321–5. Here and later in this article I cite Herman's collected papers rather than the original publication.

⁶⁷ See Herman, op. cit. (n. 66), 322.

⁶⁸ See Elcock, W. D., The Romance Languages (1960), 95Google Scholar. For Vulgar Latin developments in the paradigm of the relative, see Väänänen, V., Introduction au latin vulgaire³ (1981), 125.Google Scholar

⁶⁹ See Adams, J.N., ‘British Latin: the text, interpretation and language of the Bath curse tablets’, Britannia 23 (1992), 3.CrossRefGoogle Scholar

⁷⁰ See, e.g., Schumann, J. H., ‘The acquisition of English relative clauses by second language learners’, in Scarcella, R. C. and Krashen, S. D. (eds), Research in Second Language Acquisition. Selected Papers of the Los Angeles Second Language Acquisition Research Forum (1980), 118–31Google Scholar, and Scott, M. S. and Tucker, G. R., ‘Error analysis and English-language strategies of Arab students’, Language Learning 24 (1974), 75, 87CrossRefGoogle Scholar (with tables showing the relative frequency of imperfect relative clauses).

⁷¹ Iddibal would be the συνοδιάϱϰης For such men, see in general Rostovtzeff, M., ‘Les inscriptions caravanieres de Palmyre’, in Mélanges G. Glotz (1932), II, 793–812Google Scholar (at, e.g., 806).

⁷² This suggestion was offered by a member of the Editorial Committee.

⁷³ I am grateful to Dr Healey for information on this point.

⁷⁴ See Friedrich, J. and Röllig, W., Phönizisch-Punische Grammatik² (1970), 104Google Scholar; Segert, op. cit. (n. 15), 111.

⁷⁵ It should, however, be noted that accusative plurals spelt -us do sometimes turn up in late texts. See, e.g., Löfstedt, op. cit. (n. 65), 86–8; J. Herman, ‘Témoignage des inscriptions latines et préhistoire des langues romanes: le cas de la Sardaigne’, in idem, Du latin aux langues romanes, op. cit. (n. 66), 188; Adams, J. N., The Text and Language of a Vulgar Latin Chronicle (Anonymus Valesianus II) (1976), 42–3.Google Scholar

⁷⁶ See Marichal, Bu Njem, 262.

⁷⁷ See Friedrich-Röllig, op. cit. (n. 74), 154, 103 (§219); Segert, op. cit. (n. 15), 180.

⁷⁸ For Greek and Latin, cf., e.g. O. Flor. 15.3, ἔπεμΨά σοι διὰ Κουίντον ἱπποιατϱοῡ; Cavenaile, R., Corpus Papyrorum Latinarum (1958), 303.4Google Scholar, ‘item per Dracontem amaxitem; 303.8, ‘misi tibe per Thiadicem equitem’; 304.13, ‘misi tibe per Arrianum equitem’; P. Oxy. XLIV. 3208, ‘Ohapim regium mensularium Oxsyrychitem’. Some of these examples are in similar contexts to those in our documents. For information on this point about Punic, I am grateful to Dr Healey.

⁷⁹ In the caravan inscriptions of Palmyra (see Rostovtzeff, op. cit. (n. 71)), caravan leaders are constantly given their name and then a patronymic expression.

⁸⁰ Marichal, Bu Njem, 47, basing himself on the quantities transported according to letters 76–9, argues that in 79 one camel was needed, but in 76, 77, 78, three, four, and two respectively. This line of argument is questionable, and it does not help with the syntax of per camellarius + name. If there were more camels than one in a caravan, there might only have been one camellarius, or alternatively if there were more camellarii than one, Aemilius need only have named the leader.

⁸¹ As I suggested, Adams, op. eit. (n. 64), 243. This and a few other similar examples are also taken by the TLL (forthcoming, S. V. per) as arising ‘errore scribae uel lapidarii’. The other examples cited are manifestly slips or special cases of other types: CIL III. 7791, ‘per Antiochu sacerdos loci’ (acc. Antiochu, followed by an appositional nominative at a further remove from the preposition); CIL III. 10515, ‘per Siti doretus patrem’ (= per Sit>tium Theo>doretu<m>, Mommsen(?); the mistaken nominative use is corrected in patrem); CIL III. 14184.27, ‘per L. Apronius Pium’ (again the slip is corrected in the next name); AE 1964, 251, ‘p(er) Qui(ntio) Prisciano et Iuli(o) Marco, IIviri quinquennales’ (another appositional nominative). I am grateful to Dr P. Flury for supplying me with material forthcoming in the TLL.

⁸² See Löfstedt, op. cit. (n. 59), I², 76, with additional bibliography.

⁸³ See, e.g. Ewert, A., The French Language² (1943), 130Google Scholar.

⁸⁴ See Löfstedt, op. cit. (n. 59), I², 76–80; also Hofmann, J. B. and Szantyr, A., Lateinische Syntax und Stilistik (1965), 27–8Google Scholar, with bibliography; Parera, J. Bastardas, Particularidades Sintacticas del Latin Medieval (1953). 23–4Google Scholar.

⁸⁵ Some of the Romance nouns which might seem to reflect Latin nominative forms (as distinct from expected accusatives) are indeed descriptive of occupations (e.g. Fr. pâtre beside learned pasteur, peintre, It. sarto (<sartor), curato (< curator)), but they are not normally accounted for by historical linguists from this semantic feature (see Ewert, op. cit. (n. 83), 130, Tekavčić, op. cit. (n. 24), 46). In these cases too (cf. Louis etc. above) it was probably the vocative (which was usually indistinguishable from the nominative) which was fossilized: see e.g. Bourciez, E., Eléments de linguistique romane⁴ (1956), 231, §216c.Google Scholar

⁸⁶ See Pinkster, H., Latin Syntax and Semantics (1990), 61Google Scholar.

⁸⁷ See in general Sas, L. F., The Noun Declension System in Merovingian Latin (1937).Google Scholar

⁸⁸ For this phenomenon as a feature of second-language learning, see Richards, op. cit. (n. 29), 177–8.

⁸⁹ Marichal, Bu Njem, 62.

⁹⁰ See n. 50.

⁹¹ Similarly in Terentianus -u occurs as an accusative singular ending, but -o does not (Adams, J. N., The Vulgar Latin of the Letters of Claudius Terentianus (P. Mich. VIII, 467–72) (1977), 23)Google Scholar, and in Eunus there is one case of -u (acc.) but none of -o (Adams, op. cit. (n. 64), 236). Contrast the material assembled from later texts and discussed by Löfstedt, op. cit. (n. 65), 226–33, where -o rather than -u is in alternation with -um. In the early period, represented for example by the Pompeian inscriptions, accusatives in -o (for which see Väänanen, op. cit. (n.64), 29) are likely to derive from the old accusative -om, whereas in the late period such misspellings would reflect the merger of ō and ῠ.

⁹² Adams, op. cit. (n.64),231.

⁹³ On this chronology see Adams, op. cit. (n. 64), 231, R. G. G. Coleman, ‘The monophthongization of /ae/ and the Vulgar Latin vowel system’, TPhS (1971), 185.

⁹⁴ Flobert, P., ‘Le témoignage épigraphique des apices et des I longae sur les quantités vocaliques en latin impérial’, in Calboli, G. (ed.), Latin vulgaire - latin tardif II. Actes du Ilème Colloque international sur le latin vulgaire et tardif (Bologna, 29 aûut - 2 septembre 1988) (1990), 105Google Scholar.

⁹⁵ Similarly at Pompeii ae is often replaced by e(Väänanen, op. cit. (n.64), 23–5), and in Terentianus the e-spelling is almost as common as ae (Adams, op. cit. (n. 91), 11–12). On the other hand at La Graufesenque ae occurs a few times, but there is no case of reduction to e (Marichal, R., Les graffites de La Graufesenque (1988), 59)Google Scholar. It is obviously likely that there were regional variations in the chronology of the monophthongization of ae.

⁹⁶ The stability of ē/ĭ and ō/û at Bu Njem might be seen as at variance with Herman's argument (‘Un vieux dossier réouvert: les transformations du système latin des quantités vocaliques’, in idem, Du latin aux langues romanes, op. cit. (n. 66), 217–31) that in Africa ‘l’allongement des voyelles accentuées brèves et l'abrègement massif des voyelles longues non accentuées semblent avoir été particulièrement précoces' (217).

⁹⁷ See e.g. Herman, ‘Evolution a > e en latin tardif? Essai sur les liens entre la phonétique historique et la phonologie diachronique’, in idem, Du latin aux langues romanes, op. cit. (n. 66), 209; Adams, op. cit. (n.91), 13–14; idem, op. cit. (n.64), 231; Marichal, op. cit. (n.95), 58.

⁹⁸ See further Adams, op. cit. (n. 69), 12, with bibliography.

⁹⁹ See Marichal, Bu Njem, index 274; also 46.

¹⁰⁰ See Carnoy, A., Le latin d'Espagne d'apres les inscriptions ² (1906), 28Google Scholar; Bourciez, op. cit. (n.85), 150; Löfstedt, op. cit. (n. 65), 22–8, 98–9; Adams, op. cit. (n.75), 40–1.

¹⁰¹ Löfstedt, op. cit. (n. 65), 22, 26. Löfstedt (27) offers an additional explanation of -cipit for -cepit: ‘Bei der Schreibung -cipit für -cepitmag auch das c einen Einfluss ausgeübt haben … Eine derartige Beeinflussung halte ich für sehr wahrscheinlich bei der regelmässigen Schreibung mercide statt mercede im Sangallensis des Edikts’.

¹⁰² See Adams in Bowman, A. K. and Thomas, J. D., Vindolanda: the Latin Writing Tablets (1983), 73Google Scholar.

¹⁰³ See Löfstedt, op. cit. (n.65), 71–2.

¹⁰⁴ See Löfstedt's list, op. cit. (n. 65), 69–70.

¹⁰⁵ See Adams, op. cit. (n. 64), 233–5; Marichal, op. cit. (n.95), 64–5 (on I longa and its significance at La Graufesenque); , Flobert, ‘Les graffites de la Graufesenque: un témoignage sur le gallo-latin sous Néron’, in Iliescu, M. and Marxgut, W. (eds), Latin vulgaire — latin tardif III. Actes du IIIème Collogue international sur le latin vulgaire et tardif (Innsbruck, 2–5 septembre 1991), 106Google Scholar; also Väänänen, op. cit. (n. 64), 33–41.

¹⁰⁶ For early examples of the phenomenon see Väänänen, op. cit. (n. 64), 39–40; Adams, op. cit. (n. 64), 235.

¹⁰⁷ See Marichal, Bu Njem, 53–4.

¹⁰⁸ For examples at Pompeii, see Väänänen, op. cit. (n.64), 36–7.

¹⁰⁹ For a few such examples, see Väänänen, op. cit. (n. 68), 45; idem, op. cit. (n.64), 49; Adams, op. cit. (n. 69), 10.

¹¹⁰ See Löfstedt, B., ‘Die betonten Hiatusvokale in Wörtern vom Typus pius, tuus, meus’, Eranos 60 (1962), 88–9Google Scholar; also Lyons, C., ‘On the origin of the Old French strong-weak possessive distinction’, TPhS (1986), 1–41Google Scholar, esp. 20–1.

¹¹¹ See Marichal, op. cit. (n. 95), 64.

¹¹² See Väänänen, op. cit. (n.64), 49.

¹¹³ See Marichal, Bu Njem, 64 n. 8.

¹¹⁴ See Väänänen, op. cit. (n. 64), 41.

¹¹⁵ See Adams, op. cit. (n. 64), 233–4.

¹¹⁶ See Adams, op. cit. (n. 69), 9.

¹¹⁷ See Adams, op. cit. (n. 64), 233 with bibliography.

¹¹⁸ At Vindolanda: see Tab. Vind. II, 186 (op. cit. (n.40)).

¹¹⁹ See Väänänen, op. cit. (11.64), 41.

¹²⁰ Found ten times: Marichal, , Bu Njem, index 266, S. VGoogle Scholar.

¹²¹ See Meyer-Lübke, op. cit. (n. 24), 7722.

¹²² See Gratwick, A. S., ‘Latinitas Britannica: was British Latin archaic?’, in Brooks, N. (ed.), Latin and the Vernacular Languages in Early Medieval Britain (1982), 23Google Scholar; Adams, op. cit. (n. 64), 235.

¹²³ On the bilabial fricative, see Väänänen, op. cit. (n. 68), 50.

¹²⁴ See Löfstedt, op. cit. (n. 65), 226–8 for statistics and discussion. Löfstedt (228) relates the frequency of -a to the transfer of neuter plurals into the feminine singular.

¹²⁵ But on the special case of the documents of La Graufesenque, see Marichal, op. cit. (n. 95), 68–70; Flobert, op. cit. (n. 105), 109–10.

¹²⁶ See Adams, op. cit. (n. 64), 238.

¹²⁷ See Kiss, op. cit. (n. 64), 75–6.

¹²⁸ Bu Njem, index, 275.

¹²⁹ See Sommer, op. cit. (n.57), 203; Adams, op. cit. (n. 91), 27–8.

¹³⁰ See Prinz, O., ‘Zur präfixassimilation im antiken und im frühmittelalterlichen Latein’, ALMA 21 (1951), 97–8Google Scholar; Kiss, op. cit. (n.64), 33. Such assimilations can, however, be found in medieval manuscripts.

¹³¹ Audollent, A., Defixionum tabellae (1904)Google Scholar; on the phenomenon, see Sommer, op. cit. (n.57), 186.

¹³² See Jeanneret, M., La langue des tablettes d'execration latines (1918), 55Google Scholar. Pirson, J., La langue des inscriptions latines de la Gaule (1901), 93Google Scholar, cites Iamlychus = Iamblichus from CIL XIII.2374, but I have been unable to find the correct inscription.

¹³³ See Sommer, op. cit. (n.57), 183–4; Leumann, M., Lateinische Laut- und Formenlehre⁶ (1977), 146Google Scholar.

¹³⁴ Bu Njem, 101 n. 14, 184, 274.

¹³⁵ Adams, op. cit. (n. 64), 241.

¹³⁶ See Adams, op. cit. (n. 64), 240, with bibliography.

¹³⁷ See Konjetzny, op. cit. (n. 52), 329 with n. 3; Svennung, op. cit. (n. 52), 252.

¹³⁸ = Diehl, ILCV, op. cit. (n. 65), 2349 adn.

¹³⁹ See further Diehl, ILCV, III, 304.

¹⁴⁰ See Konjetzny, op. cit. (n.52), 329, and cf. O. Bu Njem, 109 die Idus Nouemres (cf. Diehl, ILCV, III, 307, e.g. Nos 2459, 2610 adn.).

¹⁴¹ cf. Marichal, Bu Njem, 62.

¹⁴² See Konjetzny, op. cit. (n. 52), 330 n. 2, Svennung, op. cit. (n. 52), 252.

¹⁴³ See Marichal, Bu Njem, index 266, S. V.

¹⁴⁴ The alternation -iusl-is is not uncommon in Latin gentilicia: see Neue, F. and Wagener, C., Formenlehre der lateinischen Sprache I³ (1902), 211Google Scholar. In the early period the -is form reflects Oscan influence: see Leumann, op. cit. (n. 133), 423. In the later period Greek influence is a possibility, since ιоς and -ιоν tended to be reduced to -ις and -ιν, and these endings appear in Latin loan-words: in den griechischen Papyri (1927), 116. It is difficult to know what to make of such a spelling at Bu Njem.

¹⁴⁵ See Svennung, op. cit. (n. 52), 378.

¹⁴⁶ Meyer-Lübke, op. cit. (n. 24), 6722; and particularly von Wartburg, W., Französisches etymologisches Wörterbuch IX (1959), 303Google Scholar on the history of the word and its various meanings.

¹⁴⁷ Cited by Meyer-Lübke, op. cit. (n. 24). See further Bonnet, M., Le Latin de Grégoire de Tours (1890), 238 n. 5.Google Scholar

¹⁴⁸ But note ad Boinag (e.g. 2).

¹⁴⁹ See Adams, op. cit. (n. 91), 37–8, and in general Hofmann and Szantyr, op. cit. (n. 84), 277–8, with bibliography.

¹⁵⁰ See Adams, op. cit. (n. 75), 57–8.

¹⁵¹ An example such as Paul. Diac, Hist. Rom. XV. 15, ‘egressus igitur Constantinopolim’ (cited by Norberg, D., Beiträge zur spätlateinischen Syntax (1944), 53)Google Scholar, where the original accusative form has been fossilized as the place name, is only superficially aberrant, because separation is marked by the prefix of the verb.

¹⁵² See André, J., Lexique des termes de botanique en latin (1956), 132Google Scholar (S. V. faba); idem, Isidore de Séville, Étymologies Livre XVII (1981), 90 n. 197, on Isid. XVII. 8.9.

¹⁵³ See Wagner, M. L., Dizionario etimologico Sardo II (1962), 450Google Scholar S. V. surdzága; also Meyer-Lübke, op. cit. (n. 24), 8502.

¹⁵⁴ See Meyer-Lübke, op. cit. (n. 24), 8502; G. Rohlfs, ‘Span. judia, kalabr. suráka “Bohne”’, ZRP 40 (1920), 340; idem, Nuovo dizionario dialettale delta Calabria (1977), 704, S. V. suriaca.

¹⁵⁵ Faba is not (infrequently used in the plural, though it is normally singular: see TLL VI. 1.2.52ff.

¹⁵⁶ Rebuffat, R., ‘Notes sur le Camp Romain de Gholaia (Bu Njem)’, Libyan Studies 20 (1989), 155–67CrossRefGoogle Scholar, esp. 161–2. This reference was drawn to my attention by a reader.

¹⁵⁷ Note Rebuffat, op. cit. (n. 156), 162: ‘Il nous paraît très possible que le géomètre chargé de tracer le camp ait été de civilisation punique’.

¹⁵⁸ cf. Marichal, Bu Njem, 48.

¹⁵⁹ Noted by Marichal, Bu Njem, 48 (cf. 194).

¹⁶⁰ An alternative to the formula transmisi at te is worth noting: 81, ‘suscipies ab Glareo ašgatui dua semis, facent m(odios ) triginta’. The components of the letter are much the same as those of the type with transmisi at te, but the writer has made the recipient of the goods, rather than the sender, the subject of the main verb. I cannot find examples of thi s formula in Cugusi, Corpus epistolarum latinarum. The letter has a second deviation from the normal pattern: ‘]o d(ecurioni) preposito salutem ab Au]relio Donato mili(te)’. There is a clear symmetry between this form of greeting (for which see also 86, and P. Oxy. I. 32, Cugusi, op. cit. (n. 45), 55–6, idem, op. cit. (n. 42), II, 215), and the formula with suscipies seen above. In both cases the person sending greetings/dispatching goods is removed from the focal subject position. It is as if two alternative forms of letter describing the dispatch of goods had been provided.

¹⁶¹ On Latin as the language of the army, see Balsdon, J. P. V. D., Romans and Aliens (1979), 118–19, 131–2Google Scholar. On ‘language policy’, see Val. Max. II. 2.2 and the discussion of Gruen, E. S., Culture and National Identity in Republican Rome (1992), 235–41Google Scholar. For foreigners writing Latin in a military context, see Balty, J. Ch., ‘Apamea in Syria in the second and third centuries A.D.’, JRS 78 (1988), 102Google Scholar: ‘… the stone engravers who inscribed the texts were no doubt Apameans and not accustomed to write Latin words, as is indicated by numerous confusions between P and R, C and S, uncial C and E, and by the introduction of Δ in words such as Gordiana or decurio’. The unit in question was the II Parthica.

¹⁶² The poem, set up in the bath-house at Bu Njem, has been known since 1928. It has recently been republished and discussed in some detail by Rebuffat, R., ‘Le poéme de Q. Avidius Quintianus à la déesse Salus’, Karthago 21 (1987), 93–105Google Scholar. The poem dates from the period 10 December 202–9 December 203: see Rebuffat, op. cit. (n. 156), 155.

¹⁶³ On the poem of Iasucthan, see e.g. Rebuffat, op. cit. (n. 162), 93, 101, 103 (but without quotations). It may be dated to the beginning of 222 (Rebuffat, op. cit. (n. 156), 155).

¹⁶⁴ e.g. the use of the infinitive in indirect questions. Avidius' scansion of quamdium and aestuantis recalls the phenomena discussed above at VI.2; see also VI. 2 on harenacis = harenaceis.

¹⁶⁵ See Rebuffat, op. cit. (n. 162), 102.

¹⁶⁶ idem, 101, 102.

¹⁶⁷ See Rebuffat, op. cit. (n. 156), 167, listing two relevant articles as ‘sous presse’.