² During excavation directed by Dr Ann Ellison for the Committee of Rescue Archaeology in Avon, Gloucestershire and Somerset. For other tablets from the site see Britannia xxiii (1992), 310Google Scholar, No. 5 and n. 9. The present item is briefly noted in RSOT's interim report in A. Woodward and P. Leach, The Uley Shrines (1993), 130, where it is No. 86, but the text is published by him here for the first time.

³ PETRONIIVS could be understood as Petroneus, but the termination -ius is so stereotyped in personal names that a Vulgar confusion of e for i seems unlikely here; moreover it would be odd to find two forms of E in such a brief text. The scribe seems to have repeated I by oversight or by confusion with the preceding N. Petronius is a common Latin nomen, but it is noteworthy here since the name-stock of the ‘curse tablets’ from Bath and Uley consists almost entirely of Celtic personal names or commonplace Latin cognomina.

⁴ On the surface during field-walking. The finder made it available to RSOT through Jan Summerfield of English Heritage, after cleaning and conservation by Colin Slack. It will be deposited in Kegworth Village Museum. Three other tablets have also been found at Ratcliffe-on-Soar. (i) That published by Turner, E.G. as ‘A curse tablet from Nottinghamshire’, JRS liii (1963), 122–4.Google Scholar This is addressed to Jupiter. (2) A second tablet (Tab. Sulis, p. 61) found in c. 1963, also examined by Turner but ‘more baffling in that it doesn't seem to make any obvious sense’ (in lift, to RPW, 10 March 1964). It carries five lines of capital-letter text and is apparently complete; it has not been published, but a photograph and Turner's transcript survive. (3) A third tablet and a coin of 367/75, found in 1992 by Peter Reeves of English Heritage. It is an irregular piece of sheet lead, c. 110 by 110 mm, 2–3 mm thick, with what seems to be a nail-hole at each of the four comers. It has not been unrolled, but a mould taken of the interior by Colin Slack revealed no trace of writing.

⁵ With the exception of N, which is reversed as the first letter of line 1, but written orthograde thereafter. Writing from right to left caused the scribe to cramp his letters; in particular, L always extends below the next letter, and in 5 A is ligatured to V. Two letters accidentally omitted (in 3 and 4) were added afterwards. Mirror-image capitals were also used in one half of a text from London (Britannia xviii (1987), 360, No. 1)Google Scholar, although here too N was not reversed.

⁶ This is a letter-by-letter transcript from left to right, of a text written from right to left. Two uninscribed spaces between words have been indicated, but otherwise there was no word-separation except at the end of each line. Dotted letters are letters too damaged for certainty, but where the reading is regarded as probable. Where the traces are quite uncertain, the approximate number of letters is represented by stops. Stops enclosed by square brackets indicate the approximate number of letters lost altogether.’ ‘indicates a letter interlineated.

⁷ This is the text after word-separation, punctuation, capitalisation of proper names, and restoration of letters [ ] lost in the original. Letters which have either been changed or supplied to bring a word into conformity with the classical norm are both bracketed ( ); letters removed for the same reason are indicated by < >. In all these cases, for what the scribe actually wrote and for whether or not it is significant, reference should be made to the transcript and commentary above. Unresolved words remain in capitals.

⁸ Commentary. [We are grateful to Dr J.N. Adams for some valuable comments. Tab. Sulis is R.S.O. Tomlin, Tabellae Sulis: Roman Inscribed Tablets of Tin and Lead from the Sacred Spring at Bath (1988), reprinted from B.W. Cunliffe (ed.), The Temple of Sulis Minerva at Bath, 11: the Finds from the Sacred Spring (1988). Reference is by page or item number (No.).]

1. nomine: this is the first instance from Britain of a ‘curse tablet’ being written explicitly on behalf of someone else; nomen otherwise is regularly used to refer to the nameless thief (Tab. Sulis, 95–6,98–100).

Camulorigi(s): ‘Celtic’ personal name formed from that of the war-god Camulos and the common name-element -rix (‘king’); it is attested in Gaul (CIL xIII. 11216) and Britain, where it survived the Roman period (ECMW Nos. 27, 349, 403). The scribe has treated it as a second-declension noun in -us.

Titocun(a)e: ‘Celtic’ personal name apparently unattested, but formed from the female name Tito (CIL in. 9817) or Titto (III. 8319, XII. 95) and the common name-element -cunos/a (‘hound’). The (genitive) case-ending -e for -ae is frequent in Vulgar Latin.

molam: the object stolen, the reading guaranteed by its repetition in 3. It is not clear whether it is to be identified with PAVLATORIAM (5) and VERTOGN (7), and thus was stolen ‘from the house’ (7); see further, below. Mola is properly a ‘mill’ or ‘millstone’, which seems too large and heavy to be a likely object of theft, but if (say) a lava quern were worth importing, it would be worth carrying off. However, since the scribe was capable of confusing o with u (see n(o)men, below), it is possible that he meant mulam, a female mule or simply ‘mule': see J.N. Adams, ‘The generic use of mula and the status and employment of female mules in the Roman world’, Rheinisches Museum für Philologie cxxxvi (1993), 35–61. Confusion between mula and mola is rare but attested: TLL notes it in a graffito in Greek from Pompeii, CIL iv. 2204, and Vitae Patrum 111. 128 (Migne, PL lxxiii, 785), nisi pistor molae oculos operiret, the baker must blindfold the mule that grinds the corn. The latter is echoed by a false etymology of mulus from molendo and molas (Isidore, Orig. XII. 1.57), which implies that confusion was possible. Stolen animals are mentioned in the Uley ‘curse tablets': a beast of burden (iumentum) in Britannia x (1979), 342, No. 2Google Scholar, and four cows in xxiii (1992), 310, No. 5.

perdederunt: the verb is regularly used of loss by theft (Tab. Sulis, 65); this form of the perfect, a recomposition by analogy with the -do (-dedi) stem, occurs in Tab. Sulis No. 62.

2. in fanum dei: the stolen property is regularly ‘given’ to the temple of the god addressed, to ensure his interest in recovering it (Tab. Sulis, 68); for the improper use here of in and the accusative, compare Tab. Sulis No. 97, donat in templum Martis.

devovi: damaged by corrosion; the surviving traces resemble DADVS, but only the V is certain. lovis cannot be read, and the syntax requires a verb: the central traces could be taken as the letters EVO crowded together. Devoveo is used instead of the usual dono in Tab. Sulis No. 10.5, but of the thief, not the object stolen.

cuicumque: the initial letter cannot be read as Q; the scribe may have written C by mistake simply because of the identity of sound [kw], but since he wrote QVICVMQVE correctly thereafter, the dative should be seen as intentional. Quicumque (nominative masculine) would in any case present a problem, since nomen is neuter. See further, next note. The clause is an exception to the norm noted by Adams (Britannia xxiii (1992), 21–2)Google Scholar that in British ‘curse tablet’ texts verb follows object in subordinate clauses.

n(o)m[e]n: the scribe wrote NVM[E]N, but ‘divinity’ makes no sense, and this must be another instance of confusion with nomen (see Britannia xviii (1987), 361 No. 1Google Scholar, with note; and Tab. Sulis No. 102, numen furti). Cuicumque, a dative, should probably be taken in a possessive sense as if cuiuscumque: ‘whosoever's name stole it’; in other words, ‘whatever the name of the (unknown) thief. Compare RIB 306, with its curse of ‘anyone called Senicianus’, inter quibus nomen Seniciani. Versnel takes this to mean ‘the group to which Senicianus belongs’ (in C.A. Faraone and D. Obbink (eds), Magika Hiera (1991), 84 and 103, n. 122), but the old interpretation is supported both by the present text, where numen/nomen must be the subject of involasit, and by other instances of a thief's ‘name’ being cursed when his identity is not known (Tab. Sulis, 95–6).

involasit: S is certain, and the form is guaranteed by its repetition in 4, 7 and 8. Involavit and involaverit are frequent and formulaic in British ‘curse tablets’ (Tab. Sulis, 64, 69–70), but this is the first instance of involasit. Adams (in litt. to RSOT) identifies it as a sigmatic aorist subjunctive (optative) usually spelt with a double -ss- (see M. Leumann, Lateinische Laut- und Formenlehre (1977), 622), but note ILS 4911, sei quis violasit. This form is typical of archaic Latin legal language, and quicumque involasit may be a formula lifted from it.

3. mola(m): omission by oversight of the final -m which was evidently unpronounced; compare the correction of DIE to DIEM. Molam was written correctly in 1.

ut: as often after donavi (or similar); it can be understood here either as introducing a final clause or, as in the punctuation which has been adopted, as ‘the old use of free-standing ut introducing a wish’ (Adams, in Britannia xxiii (1992), 6).Google Scholar

sa(n)guin(em): the scribe wrote SAMGVIN, another variant of the word variously contracted in the Bath Tablets, for example Tab. Sulis No. 44. 6, sangu(in)em, where see note. Adams notes (in litt. to RSOT) A. Stefenelli, Die Volkssprache im Werk des Petron (1962), 117ff, from which it can be argued that SAMGVIN is either the archaic neuter sanguen (compare Petronius, sat. 59.1) or, by interchange of M and N, the Vulgar masculine accusative sanguem, from which the Romance forms have been deduced. The parallel with Tab. Sulis No. 44.6 makes the latter seem more likely.

mittat: the initial letters are damaged, but the scribe seems to have written MITTAT only, not AMITTAT (‘lose’). Adams notes (in litt. to RSOT) that sanginem mittere is a medical term also used colloquially (for example by Cicero, ad An. vI. 1.2), in the sense of to ‘let blood’, to ‘bleed’ someone; it is used however of the agent, not the patient, as here. ‘Blood’ is a favourite theme in British ‘curse tablets’, often as the currency in which debts are liquidated (Tab. Sulis, 67); the formulation here is new, but compare Tab. Sulis No. 44.6–7, sangu(in)em suum fundat.

3–4. usque diem quo moriatur. M was added afterwards above the line, a correction the scribe failed to make of mola (3). This periphrasis equivalent to quoad vixerit (Tab. Sulis No. 45) is unparalleled in Britain, but was probably formulaic: compare the ‘curse tablet’ found in a grave at Mariana in Corsica, which reads ut male contabescat usque dum morie[t]ur. (This is published by G. Moracchini-Mazel, Les Fouilles de Mariana (Corse), 6: la Nécropole d' 1 Ponti (1974), 18–19, with fig. 46; text improved by H. Solin, Arctos xv (1981), 121–2Google Scholar, but note also that QVICVMQVE should be read in 3–4.)

4. invo[l]a[sit]: the restoration is inevitable, although diagonal traces survive of 2–3 letters after O which may be a crossing-out; N is partly lost in the crack.

(f)urta: the first letter is certainly H, but since HVRTA is unknown, it would seem the scribe mistook ligatured FV for HV in the text he was copying. Furta might be taken either with invo[l]a[sit] or with moriatur, if with invo[l]a[sit] (and for the word-order compare 2–3 above), furta would be the objects of theft (‘stealing stolen things’), an inept variant of fraudem fecit (compare Tab. Sulis No. 32.5-6, with note); but if furta is to be taken with moriatur, it would be adverbial, a mistake perhaps for furto (death ‘by his theft’, ‘secretly'). The former has been adopted here, but neither possibility is altogether convincing.

moriatur. I was omitted by oversight and added afterwards, a vertical stroke extending from the line above.

5. PAVLATORIAM: V is ligatured to A. The reading seems inevitable, but no such word is known; it is evidently an object of theft. One possibility is that the scribe intended pabulatoriam (‘pertaining to fodder’), an adjective noted by TLL only in Columella (vi.3.5; xi.2.99, of a basket). If it is an adjective here, it would have defined the stolen mola; and since (green) fodder is not milled, it would seem more appropriate to the hypothetical ‘mule’ (see above, 1). But if it is a substantive, then perhaps it means an ‘(animal) eating fodder’. But this can only be speculation.

[illam]: the restoration is inevitable, given the scribe's usage elsewhere; there is enough space, granted that the second L would have been written above the first.

6. et ipse: strictly speaking, this would imply a second thief, and thus that PAVLATORIAM is another object stolen by someone else, but the scribe's use of pronouns is inept (compare illam).

<moriato> mo[ri]atur: unless the scribe thought there was a noun moriatum equivalent to mors, derived from morior, this is simply a copying mistake. He evidently wrote it twice in error.

7. VERTOGN: the reading seems certain, save that N could be read as IA (the cross-bar being lost in damage). The context requires this to be an object of theft like PAVLATORIAM, but there is no likely solution. Vertigo is excluded by sense, and perhaps also vertragus and its variants, a word of Celtic etymology meaning a type of hunting dog.

de (h)ospitio: compare Tab. Sulis No. 99.2-3 (with note), and Britannia xxiii (1992), 311Google Scholar, No. 5.5 (with note). The failure to aspirate is typical of Vulgar Latin.

vel vissacio: vel must introduce an alternative to (h)ospitio, either a virtual synonym or an alternative locale. The solution adopted here is to see vissacium as a variant of bisaccium (‘double bag’) which occurs uniquely in Petronius, Sat. 31.9: asellus Corinthius cum bisaccio, a bronze table ornament in the form of a donkey wearing a pair of saddle-bags. The word occurs in glosses as visaccia or vissaccia. (See TLL s.v. bisaccium.)

8. devo: compare RIB 306, devo Nodenti, where V represents a glide [w] between two vowels in hiatus.

mori(a)tur. the scribe wrote MORIOTVR, repeating O by oversight only, since he writes moriatur correctly elsewhere. However, it is symptomatic of the disappearance of the deponent verb from Vulgar Latin that he should understand a deponent verb here as if it were a passive.

⁹ By Alan Whitworth.

¹⁰ The figures do not seem to be letters. Note RIB 1389, an unique ansate centurial stone with an initial in each ansa, but this is not really comparable.

¹¹ By a walker, who gave it to the National Museums of Scotland. The exact find-spot is unknown, but the fort of Newstead (Trimontium) is nearby. Dr L.J.F. Keppie discussed it with Fraser Hunter, who sent full details.

¹² The lower (curved) letter must be C., G, O or Q, and the manner of its widening curve suggests the fragment should be aligned as in Fig. 4. The upper letter is anomalous, and it may be that its final ‘stroke’ is casual damage.

¹³ During excavation by the Milton Keynes Archaeological Unit directed by R.J. Zeepvat. R.J. Williams provided information and a drawing. For the site see Britannia xvii (1986), 399–401Google Scholar with fig. 20, and xviii (1987), 327 with fig. 14.

¹⁴ Presumably part of a personal name. The first I is shorter than the other letters and, although it appears to be deliberate, could be an abandoned first attempt at the upright stroke of the letter R which follows. The final letter, of which only the top of a stroke sloping downwards and to the right survives, could be part of a V or an A. If the reading given here is correct, a possible parallel would be lrrico (CIL II. 2843, Numantia), where however the reading is also uncertain.

¹⁵ During excavation for the Carlisle Archaeological Unit and the BBC directed by Ian Caruana, who made it available and discussed the reading. For the site see Britannia xxii (1991), 235.Google Scholar

¹⁶ The instrument used was probably a stilus sharpened to a fine chisel point. The scribe wrote II for E, as often in stilus tablets and graffiti, but not in ‘rustic capital’ texts on papyrus and parchment: compare the contemporary Vindolanda fragment of Virgil, Britannia xviii (1987), 130–2Google Scholar, and see the commentary by Bowman and Thomas ad loc.

¹⁷ What is transcribed here as XXX is cross-hatching which cannot be understood as letters, but may have been diagonal strokes superimposed on letters to cross them out. The text obviously echoes the votive formula votum solvit libens merito, and what may have happened is this: the scribe first wrote votum Fortunae libens, omitting solvit by mistake; then with solvit in mind he wrote merito, producing meruit by conflation; and concluded with a blundered merito.

¹⁸ During fieldwalking by Ian Caruana, who sent a drawing and full details. It is now in Tullie House Museum, Carlisle.

¹⁹ Caruana comments that the stamp is not RIB 2463.2, but could well be 2463.3; if so, it is the first example from Scalesheugh.

²⁰ During excavation by the Archaeology Section of Winchester Museums Service, directed by M. Morris, G. Scobie and J. Zant. See Britannia xx (1989), 318–19Google Scholar. Robert Foot sent details, with a drawing and photographs, to RSOT.

²¹ Retrograde Chi-Rho is unusual (compare RIB 2431.9), likewise the superimposed line and circle. The Rho like a ‘crook’ resembles that on a dish from the Water Newton treasure (RIB 2414.4).

²² During excavation by the Dover Archaeological Group directed by Keith Parfitt. For the site see Britannia xix (1988), 484.Google Scholar Mike Halliwell made the sherd available and provided full information.

²³ The sherd was identified and dated by Dr V. Rigby.

²⁴ Vas communis (‘common bowl’) occurs as a graffito on a samian bowl from Ospringe (M. Whiting, H. Hawley and T. May, Report on the Excavation of the Roman Cemetery at Ospringe, Kent (1931), 68), but Kajanto, Cognomina, and Mócsy, Nomenclator, show that the personal name Communis, itself derived from the adjective communis in the sense of ‘sociable’ or ‘obliging’, is relatively common, especially in Italy and Gallia Narbonensis. It occurs as a graffito on pottery in Britain: see JRS xliv (1954), 109, No. 37Google Scholar; JRS lii (1962), 199Google Scholar, No. 52 with liii (1963), 167 corrigendum. The preConquest date and context of this sherd is of interest and compares with inscribed sherds of a similar date from the King Harry Lane cemetery, St Albans (Britannia xix (1988), 501, No. 62)Google Scholar, and from Braughing, Herts. (Britannia i (1970), 313, No. 34Google Scholar (an Arretine sherd); x (1979), 349, No. 25; xvi (1986), 443, No. 52 = Potter, T.W. and Trow, S.D., ‘Puckeridge-Braughing, Herts.: the Ermine Street Excavations, 1967-1972’, Hertfordshire Archaeology x (1988), 147, No. 1).Google Scholar

²⁵ During excavation by the Leicestershire Archaeological Unit directed by J.N. Lucas. For the site see Britannia xx (1989), 286.Google Scholar Dr R. Pollard provided a rubbing and full information on this and the next four items. Three further graffiti made after firing from the site are too fragmentary for inclusion here, but will be illustrated in the site report.

²⁶ The second letter could be B or R, the third letter V.

²⁷ The personal name Sacer/Sacra is frequent in the Celtic provinces (see Mócsy, Nomenclator) and is generally thought to be Celtic: compare RIB 262, a tribesman of the Senones. But some examples may be Latin, since the adjective is so common (see Kajanto, Cognomina, 211).

²⁸ During excavation by the Upper Nene Valley Archaeological Society directed by Roy Friendship-Taylor, who made this and the next item available, and provided full details.

²⁹ With a metal detector by P. Bach, in whose possession it remains. D.F. Mackreth sent details to RSOT.

³⁰ The brooch is a rebus on the owner's name, Delfinus (‘dolphin’), written here with a ‘hyper-correct’ ae for e,f for Greek ph (as in RIB 2448.8), and the final s being omitted for want of space. For other rebuses see RIB 1821 (shrew) and RIB 2491.124 (bull). The name Delfinus is uncommon, but it does occur in Britain, on a tombstone (RIB 1620) erected by ‘Delfinus, son of Rautio, from Upper Germany’. This dolphin has the S-curve and dorsal fin usual in Roman art (e.g. RIB 1319); the tail, which must have echoed the exaggerated beak, has been lost. Mackreth, who dates the brooch to the second century, comments (in litt. to RSOT) that the form is unparalleled in his corpus of 12,000 Romano-British brooches; evidently it was a special order, and may even have entered the country with its owner.

³¹ With a metal detector by R. McDonnal, who made it available to the Jewry Wall Museum, Leicester. Peter Liddle and Patrick Clay provided information and a photograph. It is noted briefly in M. Todd, The Corilani (2nd edn, 1991), 148.

³² This might be taken as a personal sealing, G(ai) Corel(li), ‘(Property) of Gaius Corellius’: for the nomen Corellius compare RIB 684. But since Leicester, the capital of the civitas Coriehauvorum, is only 11 miles to the south, the abbreviated name of the civitas is more likely. This would be the first instance of such a sealing from Britain, but compare RIB 2411.40, a sealing of the R(es) P(ublica) G(levensium). For the name of the civitas see Tomlin, R.S.O., ‘Non Coritani sed Corieltauvi’, Antiq. Journ. lxiii (1983), 353–5CrossRefGoogle Scholar. The form Corel- here is probably due to the unstressed short i in hiatus being sounded as a semivowel [y] and thus coalescing with the e: for a comparable synizesis see pride for pridie in Tab. Sulis No. 94.

³³ With a metal detector by Mr C. Ayling, who has sold it to Rowley's House Museum, Shrewsbury. Early military metalwork has also been found there, but the nature of the site is not yet known for certain. Mike Stokes made the item available to RSOT.

³⁴ C, not G, must be read; but a single dot would have made the difference, and the common cognomen Niger must have been intended. Nicer is the name of a river (Neckar), but is not known to have been a personal name.

³⁵ During excavation by Birmingham University Field Archaeology Unit directed by Peter Leach: see Britannia xxii (1991), 279–81Google Scholar. The ingot is illustrated in P. Leach, Shepton Mallet: Romano-Britons and Early Christians in Somerset (1991), 22. Jane Evans sent photographs and full details to RSOT.

³⁶ Minnius is a personal name, that of the mould-maker or the lead-smelter. It occurs as a nomen in north Italy (CIL v. 1892 = ILS 5371); v. 7034; compare vi. 3884) and in Gaul (xIII. 1871); and as Minius, as a nomen in north Italy (CIL v. 1289; 7481) and a cognomen in Gaul (XIII. 5780). It is also cognate with Minianus, which occurs on a ‘curse tablet’ from Bath (RIB 154 = Tab. Sulis No. 4). Roman lead ingots or ‘pigs’ from Britain bear formal moulded inscriptions (see RIB 2404), often dated, and the Shepton Mallet ingot is unparalleled. However, many of the (British) lead ingots from the Ploumanc'h wreck off the north coast of Brittany, although much larger and differently inscribed, also bear personal names of similar form. (Information from the excavator, M. L'Hour, who cites his own contribution to Proceedings of the igth Conference on Underwater Archaeology (1986), ‘The Ploumanac'h wreck: first evidence of the maritime commerce of lead from Great Britain in Antiquity’.)

³⁷ During the same excavation as the previous item. It is illustrated ibid., 24. It will be fully published by Catherine Johns in the final report, and our description is based on hers. We are also grateful to her for making her draft report available, and for discussing it with us. Jane Evans sent information and a photograph to RSOT.

³⁸ This form of Chi-Rho is later than the ‘Constantinian’ form, X superimposed upon P, which is almost the only form found in Roman Britain. Perhaps the only other Romano-British instances of the later form are the Sussex brooch (JBAA ³ xviii (1955), 17Google Scholar, pl. 4, and J.M.C. Toynbee, Art in Britain under the Romans (1964), 344, with pl. lxxix (c)), which this pendant closely resembles, and the Fifehead Neville silver ring (RIB 2422.45). A few graffiti on tile and pottery have also been so interpreted, but they remain doubtful. Epigraphically therefore the Shepton Mallet pendant is almost unique in Britain, and is not to be dated before the late fourth century. Were it not for the secure archaeological context, its antiquity would be in question, both because of its almost unique form and epigraphy and because of its anomalous metal composition, which resembles that of modern sterling silver.

³⁹ During the same excavation as the two previous items. It is illustrated ibid., 22. Jane Evans made it available to RSOT.

⁴⁰ This cognomen is already attested in Britain (RIB 890); it is elaborated from the common Latinus which occurs twice at Bath (RIB 158; Tab. Sulis 98, 15) and survived the Roman period (CIIC 470, 520).

⁴¹ During excavation for the Department of the Environment directed by Dr D.J. Breeze, who made the sherd available. For the site see Britannia V (1974), 405.Google Scholar The sherd is evidently not that published ibid., 470, No. 63.

⁴² During excavation for the Department of the Environment directed by Dr V. A. Maxfield, who made this and the next five items available. For the site see Britannia Vii (1976), 300Google Scholar; viii (1977), 362–4; xiii (1982), 337. Dr Maxfield sent drawings of all 37 graffiti found, which will be published in her final report, but the other graffiti are omitted here, since they are of less than three letters.

⁴³ HB and R are differently aligned. Between them are two parallel scratches which could each be taken as I, but in view of their different alignment and the difficulty of understanding HBI, they are better taken as casual damage. After R there would have been space for a second letter, but there is no sign of it. The graffiti are presumably the initials of successive owners.

⁴⁴ VIIS, with or without M(odii), is the most common measure of capacity of a Dressel 20.

⁴⁵ B is of cursive form. The second stroke of II (i.e. E) is incomplete, and could be the first stroke of another letter (M, N, P, R, etc), but this is unlikely. There is no sign of any letter to the left of B, which suggests that it is the first letter, but this is not quite certain in view of the space taken by its leftward loop. Personal names in Ble- are rare, but note Blescius (RIB 1254).

⁴⁶ During excavation for Cadw, Welsh Historic Monuments, directed by R.J. Brewer and P. Guest. See Britannia xxiii (1992), 258–9.Google Scholar The tile is now in the National Museum of Wales, Cardiff, where Richard Brewer made it available to RSOT.

⁴⁷ The top edge is original and a small part of the right-hand edge survives. Line 8 may be the last line, since it contains the total of the numerals in lines 2–7. The style of lettering resembles that of stilus tablets from the period 75-125, so the text would be contemporary with the construction of the basilica in the early second century.

The six names are presumably those of workmen at the tilery, each responsible for ‘1’ or ‘2’, totalling ‘9’. A heading (line I) specified the units, clearly a feminine substantive in the accusative plural, but otherwise lost. To the left of AS is a leftward-curving stroke suggestive of O or possibly D, then the tail of S; but no restoration presents itself. It can hardly have been worth counting single tiles, so the unit must have been a quantity of tiles. On e possibility is daily quotas: compare Étienne, R. et al. , Fouilles de Conimbriga, ii (1976), 159Google Scholar, No. 358a (brick graffito): ex officin(a) Maelonis diarias rogatas solvi.

VAS (line 2) is also problematic: S cuts the F of FRONTO, so VAS must have been written afterwards. If vas, it would mean ‘vessel, dish’, but it may have been a correction or addition to line 1.

Sentius is a common nomen, and four of the other names are common cognomina, but Semilicinus seems to be unattested; possibly it is a derivative of the common cognomen Similis (compare RIB 505, Similina) in a variant spelling.

⁴⁸ Its discovery was noted by Samuel Johnson, a local historian, in a work published in 1861. See M.R. Fossick, ‘Roman Milestones from Castleford’, in Roman Castleford: Excavations 1074–85, forthcoming.