² Associated with Roman pottery in a Roman triple-ditched enclosure, during excavation by AC Archaeology, from where Naomi Payne sent details and a photograph. It will be deposited in the Royal Albert Memorial Museum, Exeter.

³ The lower edge is broken, and the bottom of the letters lost. Thus it is not certain they should be read this way up, but there is a horizontal line above IX which continues downward, cutting the arm of X, which must have been there already. This would have marked IX as a numeral, and also the end of the graffito. The incomplete first letter is probably L, made with an initial leftward ‘serif’ that continued downwards and then to the right, judging by the space before IX. It would have resembled the L in RIB 394 (‘LXX’).

L is not expressly marked as a numeral, and might otherwise be an abbreviation such as l(atum) (‘wide’) or l(ongum) (‘long’) found on some grave-markers, but the numeral which then follows is always identified as ‘feet’ (p(edes)). In view of the next item, L might even be l(aeua), ‘on the left’, but there is no explicit instance of this. Two of the Richborough fragments mentioned below (RIB 61 and 62) are inscribed with a bold L followed by a smaller IIII, III or XIV, which also raises this possibility.

⁴ During excavation by Northern Archaeological Associates before the upgrading of the A1 (Dere Street). Julie Shoemark sent details and photographs. It will be published with the graffiti from these excavations (see below, note to no. 34).

⁵ The bottom of the letters is now lost, and it is also unclear whether the inscription once extended further to the right. Since a sequence STT is almost impossible, the ‘T’-like figures must be the barred digits of a numeral. In Britain, some building stones carry a numeral (e.g. RIB 1326, 1370–2, 1379), but none is preceded by S. Instead, compare the building stone from Mainz inscribed SIIII (CIL XIII 11909, as drawn in Mainzer Zeitschrift 6 (1911), 126, no. 18) and especially the two fragments of marble casing from Périgueux (J.-P. Bost and G. Fabre, Inscriptions Latines d'Aquitaine (2011), 132, 133) inscribed S II[…] and S III, which the editors understand as s(inistra) II[…] and III. These recall the fragments of marble casing from Richborough (RIB 58–65) and Wroxeter (RIB III, 3142) inscribed with numerals on the reverse which must be a guide to assembly, like the groups of voussoir stones from Corbridge (RIB 1196), Halton Chesters (RIB III, 3289) and Vindolanda (RIB 1720; III, 3361–2) inscribed with numerals to mark their position in the arches they would constitute.

The annotation SI̅I̅[…], s(inistra II[…], may thus be understood to mark the stone's position in the structure into which it was to be lifted by means of the lewis hole; this was quite likely the river Swale bridge-abutment at Catterick, just as most of the large facing stones of the Chesters bridge-abutment are cut with lewis holes (D.J. Breeze, J. Collingwood Bruce's Handbook to the Roman Wall (2006), 192–3), although they are not numbered.

⁶ During excavation by the Faculty of Classics, University of Cambridge, directed by Rose Ferraby and Martin Millett, who made it available.

⁷ The stone is too worn and rounded to tell what it came from, whether (e.g.) a slab or squared block. There is no sign of another letter in the space to the left; the broken edge to the right almost meets the upper tip of C.

⁸ During excavation by the Vindolanda Trust directed by Andrew Birley, who made it available (sf<22170>).

⁹ F has a leftward bottom-serif, as quite often in cursive lettering and informal inscriptions. C is almost cut by the squared-off edge, which suggests that the name was originally complete but was truncated when the stone was trimmed for use: the letters would have been pecked on the quarry-face or a slab waiting in the mason's yard. Fuscus is a common Latin cognomen, and occurs in a list of names at Vindolanda, perhaps members of Cohors I Tungrorum (Tab. Vindol. II, 161.3; compare 206.back 5, fusci).

¹⁰ During excavation by the Boxford History Project (Britannia 49 (2018), 394–6), which sent photographs and a copy of A. Beeson, M. Nichol and J. Appleton, The Boxford Mosaic: A Unique Survivor from the Roman Age (2019). Anthony Beeson discusses the captions, and so does Stephen Cosh, who sent a draft of his paper ‘The Bellerophon and Pelops panel at Boxford: a possible Christian mosaic?’.

¹¹ Two captions, the first (a) being terminated by a vertical band (not a letter) immediately after […]IɅ (or possibly […]TɅ), which is a neuter plural or feminine singular termination; most likely the end of the name of the female figure below. The second (b) ends in […]NI (with a single tessera surviving from the diagonal of N), followed by a space; this is probably a genitive termination. In the scene below, a male figure seated on a throne gestures with open hand to a half-naked female to the left; to the right stands a male figure bearing a spear and shield, his right arm outstretched. Beeson identifies the latter as a guardsman like those on the Madrid missorium, a ‘scene setter for the solemnity of a ruler's court’; he links the king with Pelops and the chariot race below, and sees him as King Oenomaus of Pisa introducing his daughter to her many suitors, who included the Phrygian Pelops. But the gesture of the ‘guardsman’ is unparalleled on the Madrid missorium, and seems rather to be one of supplication; and Cosh points out that his shield is identical with that borne by Bellerophon (and typically borne by cavalrymen in sculpture and on the contemporary Gloria Romanorum coinage of Magnentius, although omitted from other mosaic portraits of Bellerophon, which is why Beeson sees it as a cloak). Cosh then relates the ‘audience’ to the scene of Bellerophon killing the Chimaera, and takes it to depict King Iobates bestowing his daughter Philonoe upon Bellerophon after he had killed the Chimaera.

¹² In line 1, C extends without a break into Ʌ, which itself extends into the horizontal stroke of L, as the letters were fluently inscribed in the ‘cartoon’ before the tesserae were laid. But since a name *Calpio is not attested, it is conceivable that the mosaicist mistook E for L, the honorand's name being the Latin cognomen Caepio, which in Britain has occurred once in RIB 1270. In line 2, the end of the caption is cramped by the leg of Pegasus, making the draughtsman enclose final E within G. Its vertical has no sign of horizontal strokes, for which there was no room in any case. But the mosaicist misunderstood this ligature, making GE (or GI) into a single letter, which resembles a reversed D.

The invocation vivas recalls that on ‘presentation’ silver, for example on spoons with the owner's name in the Thetford Treasure (RIB II.2, 2420.2, 5, 30, 35, 37). Although it is often qualified by in deo (or similar), it is not exclusively Christian. The wife's name (cum … coniuge) is lost except for the termination -ata and the ‘loop’ of P or R. Fortunata is only a guess, and requires the assumption that V was ligatured to N, which itself was finished with a top-serif (which otherwise would suggest T).

¹³ The spacing is peculiar, since LE is tucked within the first L, as if to save space, but followed by the upper tip of R (unless B or P) and part of the upper stroke of E, and then a space; there is no sign of the expected O. The next letter is F, with an exaggerated bottom-serif. ‘Bellerophon’ is thus spelt BELLEREFONS, as in the Malaga mosaic (Hispania Epigraphica no. 2882).

¹⁴ It is unclear whether the Chimaera was also captioned, as it is in the Malaga mosaic (as QVMERA). Just above its spout of flame is what looks like the lower right-hand corner of a caption panel, but the ‘Pegasus’ panel would leave too little room.

¹⁵ PELOBS is more likely a misspelling by the draughtsman than due to the mosaicist misreading P as B. The voicing of p to b is remarkable, since A.L.F. Rivet and Colin Smith (The Place-Names of Roman Britain (1979), 395) reject the derivation of Mabomi from ‘Maponus’ in RIB III, 3482, as p > b ‘would be unique in Britain in Latin, and could not be British at this date’.

The identity of the charioteer is uncertain. Standing to his left is the groom Myrtilus, holding the fatal wax linchpin: it is unclear whether he was captioned to the right of the charioteer's hand with what looks like the first apex of M followed by damage (fig. 5b, bottom-left corner), which Beeson interprets as the end of the whip. He also identifies the charioteer as Oenomaus, despite his Phrygian cap and short tunic, but Cosh sees him as Pelops, depicted twice as in other representations of the chariot race, which is perhaps more likely.

¹⁶ Portable Antiquities Scheme (PAS) ref. LEIC-A9ED3D. Inscribed objects from the PAS database were communicated by Sally Worrell.

¹⁷ The bottom edge is damaged, removing the first half of M and part of other letters, but there was no room for any more to left and right, so the inscription must be complete. A woman's name (‘Martia’) or the title of Legion XIIII seem less likely than a man's name abbreviated, the owner presumably. He might be Martianus, but Martialis is much more common.

¹⁸ During trial trenching by Nene Valley Archaeological Trust and its partners including Albion Archaeology, from where Mike Luke sent details and a photograph.

¹⁹ Despite the broken edge, the graffito is complete: the second stroke of A was extended to mark it as the ending. M is distinguished from the letter to its left and right by being made with four diagonals, the third parallel to the first (which was extended downwards to mark it as the first) and the fourth parallel to the second.

²⁰ PAS ref. LVPL-2A78F9.

²¹ Comparison with the one-pound weights in RIB II.2, 2412, marked with a vertical stroke (‘I’) would suggest that this is another, but it weighs much more than one libra (327.45 g); in fact, it weighs almost exactly one pound (libra) and four ounces (unciae), which would be 436.60 g. So the scratched graffito may have indicated the number of ounces in excess of one pound. M, whether it was a letter or only a zigzag, may have identified the owner.

²² With the next item during excavation by East Dorset Antiquarian Society directed by Lilian Ladle, who made them available. They will be published with two non-literate graffiti in her final report (in preparation).

²³ The peculiar R is made with a downstroke topped by the usual long diagonal, but then cut by a short diagonal at the foot, as if the letter were made twice. It is followed by the left tip of T in the broken edge. The cognomen Martialis, since it recalled ‘Mars’, was popular with soldiers but not confined to them; however, it is noteworthy at a villa on a pre-Flavian vessel when Dorset was still occupied by the military.

²⁴ The fourth letter is incomplete, but looks more like B than R. Possibly Torbius or Torbanius, both nomina uniquely attested in Italy (CIL VI 32526, XI 6283), but too rare to be restored here with confidence.

²⁵ The graffito is complete, the second υ being extended to mark this. It is published by Martin Henig under ‘Intaglios’ (Finds 1, no. 487) in M. Atkinson and S.J. Preston, Heybridge: A Late Iron Age and Roman Settlement, Excavations at Elms Farm 1993–5 (2015), who dates it to the second century. He compares the intaglio and its graffito with RIB II.3, 2423.9 (Colchester); both graffiti are in Greek letters, and surely identify the craftsman who cut the gem or set it, not the owner of the signet.

²⁶ With the next item during the excavation published in A. Woodward and P. Leach, The Uley Shrines: Excavation of a Ritual Complex on West Hill, Gloucestershire: 1977–9 (1993), in which they are noted on pp. 127–8 as nos 9 and 41, respectively. They both refer to the theft of a sheep, and will be published with fuller commentary in the final report being prepared on the inscribed lead tablets, which will include no. 58, the theft of ‘wool’ (lanam), and no. 83, the theft of ‘wethers’ (uerbeces).

²⁷ Notes line-by-line, (a) inner face

1. diu[o] Mercurio. The god is addressed as diuus (and again in (b)3), instead of the usual deus (as in 5), which may reflect the Celtic word for ‘god’ (deiuos); compare the use of deuo in Britannia 48 (2017), 462, no. 10 (Uley); RIB 306 (Lydney).

donaui (‘I have given’) regularly introduces a curse tablet (Tab. Sulis pp. 63–4).

2. ma[li] consil<l>i is the only instance, at least in a British curse tablet, of this unusual phrase (‘of evil intent’). In writing consilii, the scribe began with ce, but corrected e to o; he then doubled ll, the first of several misspellings.

3. desputauerit. The reading is certain, but disputauerit (as it should be spelt) makes no sense here. Taken with me, it would mean literally that the thief ‘has argued with me’, but if there had been an ‘argument’, the writer would have known who the thief was. desputauerit is evidently a rhetorical synonym for inuolauerit (4); perhaps he had despoliauerit (‘despoiled’) in mind.

ouem. The word's position before inuolauerit (4) would mark it as the object of theft, even if the reading were not certain; there is no possibility of reading (say) quem or bouem. Also compare the next item (no. 13).

5. interscia(t). Although the scribe had difficulty in writing r across the crack, there is no reason to doubt the reading. There is no sign of t at the end, and the downward extension of a may mark it as the final letter. The loss of final -t from a verb is a frequent ‘Vulgarism’, found, for example, at Lydney with perfera for perferat (RIB 306) and at Bath three times in a single tablet (Tab. Sulis 5).

There is no verb *interscio, but the conjunction with ‘blood’ suggests that its subject, the god (deus), was to make the thief shed blood for his crime (Tab. Sulis p. 67); perhaps the writer had interficiat (‘kill’) in mind, but committed another malapropism like desputauerit (3).

sangu(i)n[e]. The same ‘Vulgar’ misspelling is found at Bath, in Tab. Sulis 47.4, 65.11.

6. uili. The first letter was written over another, apparently b. This is the only instance of the adjective ‘vile’ or ‘cheap’ being applied to the victim's ‘blood’, instead of the usual possessive pronoun (suo). It is in keeping with the rhetorical nature of this text: if blood was the currency of retribution, it was evidently debased.

6–7. These mutually exclusive alternatives are very frequent (Tab. Sulis pp. 67–8). In writing puella, the scribe became confused between e, u and even l, which were all made with two continuous ‘hooked’ strokes, and he left out at least one letter. By contrast, he exaggerated the final letter a, to mark the end of the word and ‘page’.

(b) outer face

1. licet is probably a conjunction (‘even if’) and continues the sense from face (a), but the text is too damaged to see what this concession was.

qu[o]d [n]escio. The traces of the damaged letters are consistent with this reading, amounting to a clause which is probably parenthetical, ‘(something) which I do not know’.

2. aput. This preposition (‘at’) must have governed a place or attribute ‘of the god’ (3), his temple most likely, but although e and m survive in the damaged area which follows, the initial long descender makes it difficult to restore templum. The letters after m are crowded because space was running out: they suggest uican, as if for uican(um) (‘the village temple’) or a misspelling of uicin(um) (‘neighbourhood temple’).

3. diui ipseu[s]. The ‘Vulgar’ spelling of ipsius is due to confusion between unstressed i and e in hiatus, like illeus for illius in a London curse tablet (RIB 7). A short word has then been lost, since the line clearly ends with me (‘me’).

4. [n]essi is the only recognisable word in the last two lines, a common ‘Vulgarism’ for nisi (‘unless’), which usually introduces a provision that the curse is to apply unless the stolen property is returned or the thief pays with his own blood (Tab. Sulis p. 65, s.v. nisi).

²⁸ Notes line-by-line, (a) inner face

1. si is in capitals, as if the scribe at first intended to write a capital-letter text.

The god is not addressed by name (deo Mercurio), but is implied by the formulas cursing a thief including non permittas (4) addressed to ‘you’. This abrupt beginning is unusual, but Tab. Sulis 31 begins similarly with si (qui)s … inuolauit and does not name the goddess.

1–2. si quis <h>ouem … tulerit. The verb is a synonym of the usual inuolare (‘steal’), also used in Tab. Sulis 47.3 and the Caerleon tablet (RIB 323+add.).

<h>ouem. Between si quis and tulerit (the verb of which it is the object) is the property stolen. This word is damaged but begins with h- and ends in -m. Between these letters, in the broken edge, are two shallow ‘loops’, the second of which leads into a downstroke: apparently ur ligatured, but corrected to ue by over-writing r with a bold diagonal that leads to m. The resulting -uem resembles that in nouem (6). There is a similar confusion between r and e in bib(e)re (5). If the first ‘loop’ is taken as an incomplete o, houem can be read, which is ouem (‘sheep’) with a redundant aspirate due to hyper-correcting the ‘Vulgar’ tendency to drop initial h; compare haue for aue (‘greetings’) in RIB 1115; Tab. Vindol. 291.14.

1–2. [de pro]|prio. The second line begins with prio, which must be the end of a word from the first line, since it is followed by tulerit. It may be restored as [de pro]|prio, since proprium as a neuter substantive means ‘private property’ and the phrase de proprio (with variants) is frequent in epitaphs, in the sense of paying for the tomb ‘from one's own resources’. It is likely that the usage has been extended here, in the sense of locating the theft de proprio, ‘from (my) private property’. This would be a variant of the phrase de hospitiolo meo (‘from my house’) in Britannia 23 (1992), 310, no. 5 (Uley) and elsewhere.

2. Virilis cannot be the adjective uirilis (‘masculine’), since this is commendatory and there is no instance of it defining a thief; besides, he may have been a woman, as the formula si baro si mulier admits. Virilis is a common personal name and, despite its odd position after tulerit, should be so understood here, either in the nominative or the genitive case. si quis and the other formulas imply that the thief is unknown, so Virilis cannot be the subject of tulerit (‘has stolen’), as its position might suggest. It must be genitive, the owner's name awkwardly inserted (‘of Virilis’), rather like Tab. Sulis 99.2–3: qui Deomiorix de hos<i>pitio suo perdiderit (‘who has robbed Deomiorix from his house’).

2–3. These formulas, which ‘define’ the thief by mutually exclusive alternatives, are frequent (Tab. Sulis p. 67), as is the ‘Vulgar’ spelling serus for ser(u)us. si ser(u)us si liber has been repeated by oversight.

4. non [i]llis. The first two words are damaged by the break, but the reading is confirmed by the next word, permittas. Whether ut has been lost at the end of the previous line is uncertain, but si ser(u)us is set a little further to the left than the corresponding si ser(u)us in the line above, suggesting there would have been space.

The non permittas formula is frequent at Bath and elsewhere (Tab. Sulis pp. 65–6). The plural illis is inconsistent with its antecedent si quis (1), which is singular, but no doubt the writer was stringing formulas together without thinking syntactically. It is followed by nec, but the next word is lost; it would have been something else which was not ‘permitted’, such as ‘health’ or ‘sleep’.

5. bib(e)re (nec m)anducare. The scribe actually wrote bibre banducare, first by confusing r with re (the similarity can be seen at the end of banducare), then by repeating the b of bibere and omitting nec altogether; nec would have looked much like his letters r, e and the initial loop of b, so he probably thought (unconsciously) that he had written it. nec bibere nec manducare is a frequent formula at Uley (Britannia 23 (1992), 310, no. 5) and elsewhere, so its familiarity may have made him careless in copying or composing his text.

6. nouem (‘nine’). The context of this numeral has been lost, but it suggests the formula that the curse should work ‘within nine days’: for example the London Bridge tablet (Britannia 18 (1987), 360, no. 1), ante q(u)od ueniant die(s) nouem.

san(g)uinem suum. g has been lost by lenition, as in Tab. Sulis 46.7. The case is accusative, but the usual formula is that the thief must pay ‘with’ his blood (Tab. Sulis p. 67, s.v. sanguine suo), which would suggest that san(g)uinem here is the object of non permittas, followed by an infinitive such as habere to echo bib(e)re and (m)anducare: the thief is to be bloodless within nine days unless he makes restitution. This would equate his ‘blood’ with his ‘health’, as in Tab. Sulis 41.3, which requires the goddess to exact the price of stolen goods per sanguinem et sa[nitatem suam].

7. si uindictam meam. si probably completes [ni]si from the previous line: the thief is to enjoy no health ‘unless’ the stolen property is ‘vindicated’. uindictam is difficult, but the usual uindicas (etc.) cannot be read, especially since the word is followed by meam, requiring a noun not a verb. u is ligatured to i, and u was repeated over it; d was apparently ligatured to i, but has almost disappeared in the fold; and after the well-made c, t has only a slight rightward cross-stroke, like t in permittas (4). The London Bridge tablet requires vengeance ‘within nine days’, but uses the verb uendicas. The only parallel for the noun is a tablet from Verona (dfx 1.7.6/1) which names three persons, followed by uindictam de illis fas (‘take vengeance on them’). Like ‘vindication’ in English, uindicta is not just ‘vengeance’: it also has the sense of ‘asserting ownership’ and thus, in the present context, the recovery of stolen property.

(b) outer face

All three lines have lost their beginning, and most of line 1 has gone in the broken top edge. Lines 2 and 3 are damaged and corroded.

2. The sequence san can be read, but not san(g)uine. The line ends with -uit, the third-person singular termination of a verb in the perfect tense: probably a reference to the thief, but the verb (e.g. inuolauit) is illegible.

3 ends with -ine, and the final e is extended to mark the end of the text. To the left, the traces would suit san(g)uine as in (a)6, suggesting the formula already mentioned which requires the thief to make restitution ‘with his blood’.

²⁹ PAS ref. HAMP-13E77C. Martin Henig discussed the reading and suggested parallels.

³⁰ There is an incomplete ‘loop’ near one end of the vertical stroke, as if (Greek capital) Ρ were intended. Otherwise it is (Greek capitals) ΙΧ ligatured, presumably for ‘J(esus) Ch(rist)’. It recalls the six-pointed ‘stars’ often scratched on pottery as an illiterate owner's mark of identification, but Christian symbols such as Chi-Rho are frequent on ‘Brancaster’ rings: see Gerrard, J. and Henig, M., ‘Brancaster type signet rings’, Bonner Jahrbücher 216 (2016), 225–50Google Scholar. On the Hambleton ring (cat. 25), it has been reduced to a six-pointed ‘star’. The Chedworth slabs (RIB 128), Chi-Rho × 2 and ‘a Chi-Rho without the loop’, show that I͡X might co-exist with X͡P and bear the same or related meaning.

³¹ In a garden, and auctioned by Sothebys, 9 July 1987 (lot 234), after the finder declined an offer from the British Museum (letters, November and December 1977). It was examined by Martin Henig in 2019 at Wartski's, London, by courtesy of Kieran McCarthy, and is now in private ownership.

³² There is a leftward stroke at the top of L which is probably an exaggerated top-serif, not a ligatured letter. Both cult titles originate from Trier and are well attested in the Rhineland, Loucetius being also spelt Leucetius. In Britain, for Lenus see RIB 309 (Caerwent) and perhaps 126 (Chedworth); and for Loucetius see RIB 140 (Bath), dedicated by Peregrinus, ciuis Treuer.

³³ PAS ref. PUBLIC-FBA44F. Walter (Jo) Ahmet, Finds Liaison Officer for the PAS for Kent, sent details and a photograph.

³⁴ The first stroke of V was made twice, and the ‘cross-stroke’ of A is a long descender. Iunianus is a cognomen developed from the Latin nomen Iunius, which, however, is found by itself as a cognomen and may have ‘concealed’ a Celtic name element (Tab. Lond. Bloomberg 4, with note). There is a second graffito on the wall just above the foot-ring: the single letter T, which is the name of another owner abbreviated to its initial letter.

³⁵ PAS ref. LIN-79CAB5.

³⁶ Both Ts are inscribed with exaggerated bottom serifs formed by two diagonals like other TOT rings such as Britannia 41 (2010), 453, no. 22; 45 (2014), 444, no. 18. For another TOT ring, see below, no. 29; for many previous discoveries, see Britannia 50 (2019), 504, no. 16 (with note).

³⁷ In the same excavation by the Museum of London Archaeology Service (MoLAS) as the pewter tablets published as Britannia 30 (1999), 375, no. 1; 44 (2013), 390, no. 21; 46 (2015), 406, no. 42; 49 (2018), 446, no. 38. Jenny Hall made it available from the Museum of London (VRY 89 <254>).

³⁸ The script is transitional between Old Roman Cursive (ORC) and New Roman Cursive (NRC). It is rather square and upright, but most letters are still of ORC form, notably a, q and s; but d is vertical; r is sometimes NRC, notably at the end of (b)1; and m is sometimes NRC, notably at the end of (b)3.

³⁹ There are too many gaps and uncertainties for a satisfactory transcript, but words when recognised have been separated, without there being word-separation in the original. Full commentary including detailed discussion of the readings is not feasible here, but see ‘Intermittent fever: a Latin textual amulet from Roman London’ in A. Cain and G. Hays (eds), Omnium Magistra Virtutum: Studies in Honor of Danuta Shanzer (forthcoming).

⁴⁰ The key lies in line 7, the reference to cottidiana, a variant spelling of quotidiana, meaning a fever (febris, understood) which reaches its peak every 24 hours. This sense is guaranteed by the reference in lines 4, 5 and 6 to tertiana (a fever which peaks every third day by inclusive reckoning), and the reference in 2 to secundana (every other day), in 8 to quarta (for quartana, every fourth day) and in 9 to quinta (for quintana, every fifth day). ‘Intermittent’ fever is characteristic of malaria, its ‘periodicity’ depending on which species of the parasite Plasmodium has invaded the victim: see R. Sallares, Malaria and Rome: A History of Malaria in Ancient Italy (2002), chapter 2 (‘Types of malaria’), especially 10–11: ‘It is above all the characteristic periodicity of intense fever recurring on the second day, in the case of P. falciparum and P. vivax, or on the third day, in the case of P. malariae, which distinguishes malaria from other diseases.’

A group of curse tablets from Rome (dfx 1.4/8, 9, 10, 11, 12) afflict their victim with intermittent fevers: eripias salutem … tradas illanc febri quartanae tertianae cottidianae. But the London tablet, despite using the same terms, is not a curse: it was found with other pewter amulets, and pewter, as a substitute for silver, would be inappropriate for a ‘curse’; there is no sign of any ‘cursing’ formula; and it concludes by referring to ‘health’ (sanitatem, (b)3), this being preceded by retteli, which may be a misspelling of rettulit (‘has restored’). The amulet was written for a man called Maurentius, to judge by maur[e]ntio in (a)11 and maurentio in (b)2, both perhaps preceded by the preposition a (‘from’). A protective deity is suggested by (b)1, where nothing was written to the right of t[e]rra mater, as if it were centred as a ‘heading’: perhaps the amulet was addressed to Mother Earth (the goddess Tellus) because the Romans associated malarial fevers with low-lying, marshy ground (Sallares, op. cit. (n. 40), chapter 4.2, ‘Malarial environments’), although a metrical prayer to dea sancta Tellus (Loeb Minor Latin Poets, 342–44) makes no reference to fever. Amulets were certainly inscribed against intermittent fever (Sallares, op. cit. (n. 40), 50–5), but one in Latin is hard to find. It is unparalleled in Britain, but supports Robert Sallares’ conclusion (op. cit. (n. 40), 156–60) that ‘P. vivax probably already existed and operated … in Britain in classical antiquity.’

⁴¹ With the next item during excavation by MoLAS (Britannia 47 (2016), 335), where Julian Hill made them available.

⁴² Septembres was probably abbreviated. The date is possibly the Kalends (1 September), but much more likely to be a day in the second half of August, with K(alendas) preceded by a numeral; the highest numeral and earliest date would be XVIIII for 14 August. The graffito notes when the brick was made, presumably to monitor the drying time of a particular batch.

⁴³ Assuming the tile was c. 0.40 by 0.30 m, there would have been six lines of text written long-axis, of which part of the last three lines survives, about one-third of each. It is not the usual note of tile-making with date, numerals or maker's name. In line 1, the adjective humanus (‘human’ or ‘humane’) qualifies the subject of a verb, a masculine noun ending in -us or -is, which is probably the antecedent of qui (‘who’) in 2. This is followed by the preposition circa (‘about’, ‘towards’) before a noun in the accusative. In line 3, quod is probably a neuter relative pronoun (‘which’), whether preceded by the adverb semper (‘always’) or the preposition per (‘through’) or propter (‘on account of’). The next word begins with c, but the next letter cannot be n in this Old Roman Cursive, besides being different from the n of humanus. It is probably e ligatured to i, but the sequence ceil- does not suggest any Latin word or well-attested name.

⁴⁴ During trial excavation by Caistor Roman Project, in the third-/fourth-century fill of a ditch north of the Roman town. William Bowden sent details and photographs. It will go to Norwich Castle Museum.

⁴⁵ It was enclosed in a folded strip of silvered copper alloy, but its purpose is unknown. It is not a sealing, but perhaps an amulet or small pendant, or the bezel of a large finger-ring. The best parallel seems to be the triangular silver Chi-Rho plaques in the Water Newton Treasure (RIB II.3, 2431).

⁴⁶ With the next item during excavation by the Vindolanda Trust directed by Robin Birley. They are now stored by the British Museum (inv. nos 1995-0701-404 and -440), where RTI scanning of the many Vindolanda stylus tablets has begun under the direction of Richard Hobbs and Alex Mullen.

⁴⁷ The lettering does not respect the binding cord, so it was probably made before the tablets were tied up. D is ‘lower-case’ in form and O is damaged, but there is no reason to doubt either letter. The final letter is E made with two vertical strokes, e for ae (since they both sounded the same); this is a common ‘Vulgarism’, but rare in the Vindolanda ink texts.

⁴⁸ For a complete tablet reused in this way, see RIB II.4, 2443.10 (Carlisle). The present item is illustrated in Vindolanda Research Reports IV.4: Writing Materials (1999), 14, captioned as ‘Tab. 1617: the reverse side bears part of a scratched address or instruction’.

⁴⁹ There is no sign of a suprascript line above XXIIII to mark it as a numeral, but the preceding denarius symbol (a ‘cross’ with a leftward horizontal stroke) makes this certain. E is made with two vertical strokes (II). The first stroke of R is rather taller than II before it, but there is no trace of its second stroke; the scribe seems to have conflated its ‘tail’ with the lower curve of C. This line ends with another II, without any trace (or indeed space) for the two diagonal strokes of M; apparently a second omission. The second line reads ONII M, with a gap before the final M in which there is no trace of the expected V. But this gap would have respected the binding cord, as clearly does the first line, with its wide gap between XXIIII and INMII (where there is no space after the preposition IN). The ‘envelope’ was evidently inscribed after being bound up.

Although this resembles an ‘address’, it is not one. A tablet was broken in half and the pieces notched for binding together, so as to enclose 24 denarii as if in a purse. This recalls Tab. Vindol. III, 655, in which the writer says that he has sent denarii which the recipient is to put into individual wrappings: denariorum … quae si(n)gulas in cartas ?inse[ri coges]. The editors understand this as ‘wrapping money in papyrus’, citing POxy. XLIV 3189.7 (with note). It is illustrated by a third-century coin hoard found in Brittany, a sealed jar containing thousands of coins, mostly antoniniani, which were topped by 18 denarii wrapped separately in a papyrus letter (A. Provost, Plouhinec (Morbihan: Mané-Véchen Villa gallo-romaine (2000), 9–10).

⁵⁰ During excavation by the Vindolanda Trust directed by Andrew Birley, who made it available (sf<21733>).

⁵¹ The first letter might be the second half of M, but looks more like Ʌ. Then the stylus slipped making R, and the letter was repeated. There is then a gap due to the central concavity, followed by the left tip of T in the broken edge. The cognomen Martialis and its cognates are common: at Vindolanda, see Britannia 42 (2011), 453, no. 19, Martialis, and possibly Tab. Vindol. III, 609.b.back 3; but note also RIB II.8, 2503.327, […]MɅRTINV[…] and 328, MɅRT[…].

⁵² With the next three items during excavation by Allen Archaeology. Jane Timby made them available and provided details. The Dressel 20 sherds come from different amphoras, much fragmented but possibly complete, found in separate pits.

⁵³ It seems to conjoin a smaller sherd (as drawn) inscribed with a long descender, which is difficult to relate to the main sherd. The latter reads -auio, granted that the last letter is a terminal o, not q (not being followed by any letter, let alone u, and whereas q would have been made with a long diagonal). The fragmentary letter before a seems to have trace of a diagonal stroke in the broken edge, which would suggest n or u. Since the graffito was made before firing, and probably near the base of the amphora (the sherd is quite thin), it must be part of the potter's signature. This would usually be a personal name and perhaps a date as well, but neither is evident here.

⁵⁴ II is more neatly inscribed than the previous letters, which might suggest a numeral (‘2’), as if Dura owned two amphoras; but more likely, these two strokes were easier to make and simply represent the alternative form of E. An after-firing graffito on the shoulder would usually be the owner's name, but Dura is difficult. The adjective durus (‘hard’) is understandably rare as a cognomen, even if an officer called Laberius Durus was a casualty of Julius Caesar's invasion of Britain (BGall. 5.15.5). It seems more likely that Dura is related to the Germanic name Durio attested by RIB 2063; in the spelling Durra it is found in the Rhineland (AE 1920, 123).

⁵⁵ The letter before INI in graffito (a) is incomplete, but probably a badly formed R or T. In the broken edge to the left are the ends of other strokes, which may not all belong to the same graffito. Non-literate marks like graffito (b) are quite often found on the foot-rings of Samian vessels, but are excluded by RIB. These are too far apart to be a numeral (XI, ‘11’) like RIB II.7, 2501.859 and 860, and are best taken as marks of identification made by the owner, who may have been illiterate, to distinguish his dish from others which were similar. He may have made other such marks elsewhere round the foot-ring, but they are now lost. Graffito (a) probably identified another owner, whether he was previous or (more likely) subsequent.

⁵⁶ The first letter is incomplete, but from its angle is unlikely to be I or N. The most likely name is Privatus, but some less-common cognomina are possible, for example Servatus or Torquatus.

⁵⁷ PAS ref. 4A576B.

⁵⁸ Each T was made with four strikes of a short-line punch, and O with eight strikes of a round punch in a circle surrounding a ninth. For other TOT rings, see no. 17 above (with note).

⁵⁹ PAS ref. YORYM-283A9A.

⁶⁰ The reading of the reverse is by PAS, who may have overlooked a centurial sign; the reading cannot be checked against their photograph. The same centurion is attested by a sealing from Piercebridge (Britannia 49 (2018), 439, no. 20), but the obverse and reverse dies are different.

⁶¹ With the next item during excavation by MAP Archaeological Practice before redevelopment. Phil Mills made them available. The site is on the Roman road heading south-west for Tadcaster, with Roman features including pits and ditch enclosures with many pottery sherds, mostly dating from c. a.d. 120 to the end of the second century. The final report will include four other Samian sherds with ownership marks, a ‘cross’, a ‘star’, and (if letters) ‘L’ and ‘VI’.

⁶² V was made twice and cuts a vertical stroke, as if the scribe wrote I by mistake and then incised V firmly over it. A blundered VIIRI is possible (compare the next item), but Virilis is a common name. R has lost its ‘tail’ in the break, but a narrow B (suggesting no names) is unlikely.

⁶³ R is made with a square ‘loop’, no doubt because a curved stroke would have been difficult to make. Its ‘tail’ is broken by the edge, so that it is uncertain whether the name was originally complete or abbreviated to its first three letters like VER in RIB II.7, 2501.816 and 817, and VAR in Britannia 50 (2019), 518, no. 40. The most likely name is Verus and its cognates; but Verecundus, if abbreviated, is another possibility.

⁶⁴ By metal-detector, PAS ref. YORYM-2C0218. Richard Hobbs sent photographs and other details, including a copy of the report for the Coroner (Treasure Case: 2018 T854).

⁶⁵ The total weight of the fragment and the coins is 32.96 g, almost exactly one-tenth of a Roman pound (327.45 g), which is likely to be deliberate. Richard Hobbs comments that Hacksilber was often used with coins to create parcels of bullion that corresponded to fractions of a Roman pound.

⁶⁶ The third letter is damaged by the fold, but must be a vowel and looks most like a well-serifed I. The lettering is ‘professional’ and recalls the inscribed silverware presented by late Roman emperors to officers and officials as largesse (largitio), but does not suggest any imperial formula. The plate is thus more likely to be a private presentation piece, perhaps to mark a wedding, bearing the name(s) of the recipient(s). -alis is probably a name-ending, for example of Genialis or Vitalis.

⁶⁷ With the next 55 items during excavation by Northern Archaeological Associates at various sites along the A1 (Dere Street) before its upgrading. Rachel Cubitt made them available before they are deposited with York Museums Trust. They will be published with more details of provenance and fuller discussion in chapter 12, ‘The graffiti’, of S. Ross and C. Ross, Cataractonium: Establishment, Consolidation, Retreat (in preparation), except for the moulded glass (nos 36–8) published by Hilary Cool. This chapter will include almost 100 other graffiti that are non-literate or comprise fewer than three complete letters or numerals. For many previous finds, see P.R. Wilson, Cataractonium: Roman Catterick and its Hinterland. Excavations and Research, 1958–1997 1 (2002), chapter 11, ‘Graffiti’.

⁶⁸ Obverse and reverse duplicate RIB II.1, 2411.83 (Carlisle), from which the reading can be restored in full. Despite the evidence of cavalry in no. 69 below, neither this item nor the next need indicate the unit. Lead sealings, although sometimes found where a unit was stationed, are evidence rather of official consignments being opened. The ala II Asturum is first attested at Ribchester by RIB 586, but by the governorship of Ulpius Marcellus (a.d. 178–84) was at Chesters, where it remained. One of its sealings was actually found there (Britannia 26 (1995), 382, no. 15) and another at Corbridge (RIB II.1, 2411.82), neither of them duplicating this item.

⁶⁹ The obverse duplicates RIB II.1, 2411.91 (also Catterick), except that it carries no trace of the tail of Q, which presumably failed to register. The reverses are different, that of 2411.91 reading PRO, probably the initials of the issuing officer. The present item carries a moulded figure impressed by a signet-ring, now indistinct. The equites singulares served with the governor, but a s(ingularis) c(onsularis) dedicated the Catterick altar (RIB 725+add.) re-dedicated by a beneficiarius consularis, which suggests that they might be detached for special duties.

⁷⁰ Published in full with the next two items by Hilary Cool (see note above), who sent details including a photograph and drawings. Two further base-fragments read S[…] and ..[…].

⁷¹ From the same mould as RIB II.2, 2419.147 (York), if one allows for some distortion, according to Hilary Cool who has seen both. She notes a related base from Belgium in which the same letters (but rectograde) are divided by pellets, showing that they were abbreviated tria nomina. Although RIB transcribes 147 and 148 (a fragmentary duplicate from Silchester) as SIL, the maker's name was probably L(ucius) I(…) S(…).

⁷² Like the previous item, this might be read either way up. Neither sequence suggests a name, whether abbreviated or not.

⁷³ The letters are worn and corroded, especially in column (ii), where the first looks like I or E, but is perhaps a ligatured IF. The second is ‘square’ enough to be EL ligatured. Assuming also that V was repeated twice by mistake in column (i), it would seem that the cutter intended something like utere felix, an exhortation often found on small objects of personal use: in Britain, on silver rings (RIB II.3, 2422.41; Britannia 41 (2010), 448, no. 10), for example, and bronze brooches (RIB II.3, 2421.56, 57, 58). The variant uti felix is also found; in Britain, it is found on a silver spoon (RIB II.2, 2420.52) and a bronze handle (RIB II.3, 2433.3). A gold ring from Lydney (Britannia 47 (2016), 395, no. 10) inscribed V F might be either.

⁷⁴ One of 17 inscribed roundels or ‘counters’, lathe-turned discs of bone with a central chuck-mark in the obverse face, ranging in diameter from 15 mm to 24 mm. RIB II.3, 2440 with 381 entries, but none from Catterick, admits single letters, numerals and two or three lines intersecting to form ‘X’ (whether a ‘cross’ or the numeral ‘10’) or a ‘star’, but they have been excluded here. They include single letters, the numerals III, VI and IIIXI, seven roundels with ‘X’ (whether ‘10’ or a ‘cross’) and two with ‘stars’.

⁷⁵ Some letter-strokes trail off as they reach the edge, showing that they began near the centre of the disc, with the scribe drawing the stylus towards him and down to the circumference. In a few places, one stroke can be seen to cut another, which allows the sequence to be deduced and confirms that he was writing from left to right along the lower circumference of the disc.

Between the vertical strokes of the two detached letters, I and L, there is a very short diagonal stroke which was cut by the second vertical, apparently V ligatured to L, a device to save space. Iulius is often abbreviated to IVL. The main word clearly begins with MI, which is followed by two verticals close together, the second tending to the right; and then by two diagonals meeting at the top. This sequence of strokes can be understood as a cramped M, especially since it is followed by V and a badly made S, suggesting the word-ending -us. S was made with the usual downstroke tending left, but was completed by a second stroke that started too low and was curtailed because of IVL to its right. This word can be read as MIMVS, but Mimus is hardly attested as a personal name, so it is more likely that the scribe, cramped for space and probably confused, omitted ɅX by oversight because it resembled M. Maximus is a common cognomen. The other face is scratched with two lines intersecting towards their ends, probably a ‘cross’.

⁷⁶ The reading is complicated by other lines (drawn in outline), some of which underlie these letters and some which may be casual. Ulpius is an imperial nomen, but unlike Iulius is uncommon by itself as a cognomen.

⁷⁷ The potter's name can be restored with certainty since the identical signature has been found at Vindolanda (Britannia 43 (2012), 409, no. 22, fig. 19).

⁷⁸ The sequence of letters is deduced from the horizontal of the third cutting the second, which must be T cutting P, not D, since the sequence DT is impossible. Optatus is a common cognomen. Dressel 20 potters’ signatures are usually in cursive letters, but compare the next item and RIB II.6, 2493.47.

⁷⁹ The reading is not certain, but the first letter is sinuous, suggesting S. The second letter is inscribed close to the first but spaced away from the third, suggesting that room was made for the loop of P, now lost. The third letter is only the meeting point where a vertical stroke was crossed by a horizontal at right angles, but this would suit H. Sph[…] or […]sph[…] would be part of the potter's name.

Five other Dressel 20 sherds carry marks made before firing, but they are not necessarily letters. The most interesting is an impressed leaf, apparently an olive leaf: perhaps only by oversight, but possibly a deliberate hint of what the amphora would contain.

⁸⁰ Like the next six items. There are seven other Dressel 20 sherds with graffiti made after firing, including a rim sherd with VK; the others are fewer than two letters or numerals.

⁸¹ The barred V is the beginning of a numeral, probably V[I] (‘6’) or V[II] (‘7’), a note of capacity in modii, but the preceding ɅV is difficult. The graffito resembles RIB II.6, 2494.106 (York), A ^o V ^o V ^o | M[…], probably the owner's initials followed by a note of m(odii), but the two letters ɅV hardly suggest an abbreviated name. However, this seems the best solution, since a blundered M for m(odii) is unlikely in such a careful graffito.

⁸² RIB II.6, 2494.134 (Rudston) is another instance of IVLI incised on the handle of a Dressel 20. Iulius is an imperial nomen, but is often used by itself: see Tab. Luguval. 19 (with note). It is explicitly a cognomen in RIB 672 (citing ILS 4650 as another example), as indeed in the name of the governor Modius Iulius.

⁸³ Since the rim is complete, (b) is the only annotation there. In (c), the suprascript line above II extends to the tip of V, marking this graffito as a numeral, ‘7’ written retrograde. Within V are two short vertical strokes, now rather faint, which in view of (b) above can be understood as a subsidiary numeral, II (‘2’). The average capacity of a Dressel 20 was about seven modii, and is often expressed by two numerals, first ‘6’ or ‘7’ (modii) and then a second numeral for sextarii (RIB II.6, 2494, pp. 33–4). Taking (b) with (c) would also explain the isolated numerals I, II or III sometimes found on the rim of Dressel 20 (ibid., p. 34).

⁸⁴ There is just enough space either side to show the graffito is complete. The lower half of I is lost, but the letter cannot be L. Carus is a common cognomen, and the initial C is quite often replaced by K: see Britannia 50 (2019), 516, no. 36 (with note).

⁸⁵ The first two letters of line 1 have lost their upper half, but the first has a slight rake appropriate to A and the second must be I or T. This reading is confirmed by the four letters which follow. There is space to the right of final I, confirming that it is the end of a word; evidently a personal name in the genitive case ending in -erni. Not many names end in -ternus. Aeternus cannot be read; Fraternus is possible, but mostly found in Iberia; Maternus and Paternus are so common that it is almost certain to be one or the other.

This would suggest that two letters have been lost to the left of R in line 2. The next letter, K, is incomplete but cannot be a second R. K is uncommon within a word, but is equivalent to C. Maternus/Paternus was identifying himself in some way; perhaps only by his father's name, but more likely, if he was a soldier, by his century, for example (centuria) Marci as in RIB 415 and 428, but with K for C.

⁸⁶ The graffito is broken to the left, but there is enough space to the right to show that it ends here. G might be read as S, but is much more like G. This unusual letter sequence cannot be matched with any attested name whether Roman or Celtic. The termination -a might suggest the owner was a woman, but the unique cognomen of the consul of a.d. 141, M. Peducaeus Stloga Priscinus, advises caution.

⁸⁷ The shipper was called Q(uintus) C(ornelius) Cle(mens), to judge by the related stamps collected by M.H. Callender, Roman Amphorae (1965), 223, no. 1439; his fig. 14.31 shows an incomplete example of this stamp from Wroxeter (now CEIPAC 15737). A complete example has since been recorded from London in a context of a.d. 120–60 (CEIPAC 15739). The owner's name terminates in -SI, so is in the genitive case: ‘(property) of so-and-so’. But the preceding letters, except ST, are badly preserved and do not suggest any attested name.

⁸⁸ There is no suprascript line, but this is probably a numeral (‘6’), with just enough surface to the right to show that it is not VII (‘7’). This is the usual capacity in modii of Dressel 20 (RIB II.6, 2494, pp. 33–4), and was probably accompanied by a second numeral (now lost) which recorded the number of sextarii (compare no. 46 above).

⁸⁹ E was made with a downstroke curving right, with two horizontal strokes above, which determines which way up to read the graffito. It is perhaps the end of a note of capacity in sextarii, followed by the owner's name beginning with Ne[…], for example Ne[pos].

⁹⁰ These three incomplete lines occupy the mid-portion of the circle provided by the underside of the base, but it is not possible to tell whether there was any lettering above and below, containing, for example, the names of the donor and recipient. The third letter of line 1 is not M, to judge by the well-preserved M in 3, but N cut by the diagonal stroke of V, which makes don[um] much more likely than (say) dom[ine], don[o] or don[at].

⁹¹ AVO[…] may be restored as AVO[CO] (‘I call away’, ‘divert’) from CIL XIII 10018, 38 and 39 (Cologne), motto beakers with the legends AVOCO ME and TE, but, although many legends celebrate wine-drinking, the present item seems to be unique: it is not in CIL XIII 10018, RIB II.6, 2498 or Bös, M., ‘Aufschriften auf rheinischen Trinkgefässen der Römerzeit’, Kölner Jahrbuch 3 (1958), 20–5Google Scholar. An object of auoco is required: the four extant letters and their spacing occupy a little less than half the circumference, implying that some five letters have been lost. auo[co cura]m is thus possible, since other ten-letter mottos are found, for example CIL XIII 10018, 134, aquam parce (‘spare the water’). cura, meaning ‘anxiety (about anything), worry, care, distress’ (OLD), is the object of auocare in Columella, Res Rustica 11.1.24, when he warns a landowner not to let his manager's ‘attention’ be distracted by other interests: haec enim res auocat uilici curam.

⁹² Like the next 24 items, Samian graffiti from Catterick. Graffiti of fewer than three complete letters have been excluded, and also many ‘crosses’, ‘stars’ and other non-literate marks of identification, except a ‘trident’ (no. 72) and a ‘Tanit’ (no. 78).

⁹³ Only the first two letters are complete, and it is unclear whether A was ‘open’ or not. The next two letters are topped by a long horizontal stroke made later. It obscures the extent to which L curved leftward at the top, but the length of its downstroke (and its probable continuation to the right) suggests it was L rather than I. To its right, in the broken edge, is probably the upper part of A since the sequence abl- would preclude M or N. Finally, there is the horizontal stroke already mentioned, which runs into the edge; it must belong to a fifth letter, evidently T with a wide cross-stroke. Ablatus is not attested as a personal name, unsurprisingly, since it is typical of epitaphs in the sense of ‘taken away (from us)’. The magical palindrome ablat(h)analba is a possibility, but a remote one, since it would be expected on an amulet, not underneath a bowl like an ownership-inscription.

⁹⁴ There is space to one side, suggesting that the graffito began here, with Ʌ. Inverted, the reading would end in -VIV, which is unlikely as a name-ending. The horizontal stroke was made after the adjoining letter, which also suggests that the graffito should be read with that letter to the left. The first Ʌ was apparently made twice. The two succeeding letters were each made with double strokes. The long horizontal stroke which meets the broken edge is probably that of T, of which the vertical stroke is lost. The letters have been selectively abraded, presumably to ‘erase’ them when there was a change of ownership. There is a rare nomen Aiatius to which Aiatus might be related, but an attractive British parallel is the cognomen of the woman Aurelia Aia, buried at Carvoran (RIB 1828).

⁹⁵ The letters are all incomplete because of the broken edge. A was evidently made with three strokes, its ‘cross-stroke’ being a long descender. R is quite angular, but cannot be K before C. Only the lower half of C survives, but it can hardly be L. At first sight this looks like Arca[nus], which is well attested in the Vindolanda Tablets as a personal name, but in the left-hand edge the left stroke of A cuts a diagonal which must belong to a previous letter; its angle is appropriate to M, raising the possibility of Marcanus and Marcarius. Neither is common, and most instances of Marcanus are in the feminine form Marcana, which would also be possible here.

⁹⁶ To the left of B are three incised lines of a different character, two radii meeting at an acute angle, with the second crossed by a shorter line. This is probably a mark of identification, whether it prefaced the owner's name or was left by a previous owner. The name is probably Bellicus or one of its cognates, spelt with one L instead of two: compare Belicianus in RIB 375 (Caerleon) and II.7, 2503.206 (Housesteads). In the broken area to the right, there is possible trace of a curving incision, as if for C.

⁹⁷ I is rather close to A, but does not adjoin it; this would make N unlikely, even if A were not guaranteed by its ‘cross-stroke’, a short vertical stroke. Ligatured AN can also be excluded, both because this would be unusual in such a graffito and because the next letter is O, not D (for Candidus, as in the next item). The Celtic personal name Caio is uncommon, but is found in Noricum and Pannonia (CIL III 11592 and 10795).

⁹⁸ Candidus is a common cognomen, often found in Britain as a graffito on pottery (RIB II.7, 2501.117–21; II.8, 2503.216–19). It is more likely than some less-common names in Can- which include the derived Candidianus. Three incisions of different character (drawn in outline), to the left of C, below A and above N, are broader in cut and unrelated to these letters. It is difficult to tell whether they were made earlier or later, but they are presumably marks made by another owner.

⁹⁹ The first letter might be an ill-proportioned P. After B, there is only the bottom of a vertical stroke, like that of Ʌ and obviously a vowel, with the other possibilities being I or II. Although PAB occurs as a graffito on wall-plaster at Cirencester (RIB II.4, 2447.19), it is hardly possible to suggest a name. The sequence dab- would suggest the polite request dabis (‘please give’), but that would be inappropriate here.

¹⁰⁰ The graffito should be read this way up, to judge by L, and the way in which the diagonals of Ʌ and N cut the verticals to their left; and the second vertical stroke of N trails off towards the foot-ring, suggesting that it was made in this direction. The top of the third letter, a vertical stroke, has been removed by the broken edge, making it uncertain whether it was T with only a short cross-stroke or the first vertical of II (for E). The cognomen Latinus is common, but Laenas is quite well attested. Underneath the vessel is scratched the remains of another graffito, apparently […].IN[…].

¹⁰¹ The first I was cut two or three times, so as to splay at the top. O is not closed at the bottom, but can hardly be A. The angular R is certain. The concluding II is broken at the top. The two verticals are close together, so may represent E, but the ablative termination of a comparative adjective such as superiore (‘higher’) or inferiore (‘lower’) would be hard to explain. It is more likely to be the genitive termination of a name in -iorius, of which the only instance seems to be Maiorius. This has not occurred in Britain, but is quite well attested in Gaul.

¹⁰² The tail of the second R extends into the ovolo. A is made with a vestigial vertical stroke between the two diagonals. This theophoric cognomen (from ‘Mercury’) is quite common, being found in Britain at York (RIB 688) and Great Chesters (RIB 1742); as MIIR[…] it has already occurred at Catterick (RIB II.7, 2501.382), and at London abbreviated as MER (Britannia 13 (1982), 418, no. 65).

¹⁰³ The graffito is complete, and probably abbreviates the cognomen Pollio, which is quite common. In Britain it has occurred at York (RIB II.5, 2491.114) and Carvoran (RIB 1783). Less likely are the three initials of an abbreviated tria nomina or a name of Greek derivation such as Polybius.

¹⁰⁴ The loop of P was made with a short horizontal stroke as well as a large loop. R has a short ‘serif’ at its foot. There is just enough space to the right to suggest the graffito is complete. It might be the three initials of an abbreviated tria nomina, but is almost certainly Primus (or a derivative such as Primitivus), since this is such a common name.

¹⁰⁵ The tail of Q is neatly made, guaranteeing that the next letter is the lower part of V. The third letter is almost certainly I, although it is only the lower part of a vertical stroke with a very short horizontal stroke midway. This might suggest E, but there is no corresponding horizontal stroke at the bottom, and names in Que- are almost unknown. Perhaps it was made as the stylus moved to the next letter. This is part of a vertical stroke in the very edge of the sherd; it is enough to exclude A as the third letter, but not enough to decide between E, N and R. The most likely names are therefore Quietus and Quintus (like the next item), but there are other possibilities.

¹⁰⁶ The praenomen Quintus is often used as a cognomen: compare RIB II.7, 2501.462–4 and perhaps also the previous item. In the corner below V there is part of a diagonal stroke. Since a second name is unlikely just below the first, this must belong to the extended tail of Q.

¹⁰⁷ S is little more than a diagonal line, P and R are made with a small angular loop and A with a third short diagonal stroke. The personal name Superra is not directly attested, but must be a variant spelling (with geminated R) of Supera. This is the feminine form of the cognomen Super, which is quite common, and thus of interest as the name of a woman, presumably written by herself; compare no. 86 below.

¹⁰⁸ The second (vertical) stroke of V is continued by a ‘hook’ which links it to the first stroke of II: this is not a letter, but due to the scribe not lifting his stylus between V and II. There is just enough space to the right to suggest that the graffito did not continue round the circumference, but comparison with RIB II.6, 2494.181 suggests that it was completed in the missing lower half by the owner's name. This is another Catterick graffito, but on a Dressel 20 amphora, reading: T VIILO[…] | CAN[…], t(urma) Velo[cis] | Can[didus] or Can[didi], ‘the troop of Velox, Candidus’ or ‘(property) of Candidus’. It is unknown whether both graffiti were inscribed by the same man or by two members of the same troop (turma); but taken together, they suggest that a third Catterick graffito on Samian, TVOC[…] (RIB II.7, 2501.554), should be understood as t(urma) Voc[ontii …], ‘the troop of Vocontius, (property) of …’. It is rare for the owner of a Samian vessel to identify himself by first naming his troop, but RIB II.7, 2501.617 (Corbridge) is another example.

This item attests cavalry at Catterick in the Flavian period, and might be taken with a second-century sherd inscribed (centuria) M[…] (RIB II.6, 2497.5), and perhaps no. 48 above, to indicate that the garrison was a part-mounted cohort (cohors equitata). A connection with the ala II Asturum (no. 34 above) is less likely, since this unit is well attested elsewhere.

¹⁰⁹ The letters are all incomplete, but their stroke-endings are consistent with -alis, which would suit many common names such as Martialis and Vitalis.

¹¹⁰ The ‘cross-stroke’ of A is a short downward diagonal.

¹¹¹ The broken edge has removed the tops of the letters, and it is uncertain how much has been lost. It is not even certain that they should be read this way up, but the two strokes of Ʌ seem to be linked by an underlying drag-mark as if the scribe brought up his stylus to make the diagonal; and the long diagonal of R is topped by the trace of another diagonal in the broken edge, as if to finish the loop. The letter before it, an incomplete vertical stroke, shows a similar diagonal towards the top, but this is much slighter and perhaps only a drag-mark. The two vertical strokes after R suggest II (for E), but TI cannot be excluded. No name can be suggested, unless a blundered Ianuarius (compare no. 85 below).

Within the foot-ring is another graffito, a fork-like ‘trident’. Its prime position suggests that it was made by a previous owner, who was illiterate. RIB does not collect non-literate graffiti, but illustrates ‘tridents’ on a bone roundel (II.3, 2440.53) and wall-plaster (II.4, 2447.23); and for a coarseware ‘trident’, see Proceedings of the Cambridge Antiquarian Society 37 (1937), 68–71, pl. Ia. Although it symbolised Neptune and was borne by gladiators, there is no need to see a reference to them here; it was a simple geometrical figure, easier to make than the most remarkable non-literate graffito from Catterick, the ‘Tanit’ in no. 78 below.

¹¹² Only one letter (X) is complete, and it is uncertain which way up to read the graffito, but a few tiny marks on the inner surface of the foot-ring (as if continuing the line of the stroke as the stylus was lifted) suggest the scribe was writing at the bottom of the circumference. The sequence -emixi- does not fit any known name, but unless XI[…] is a numeral, which seems unlikely, this was perhaps a Celtic name in -ixius or -ixi (genitive) otherwise unattested.

¹¹³ T, although slight, is guaranteed by its long cross-stroke. It is preceded by the tip of a vertical stroke, I rather than N, and followed by the tips of two strokes meeting in an apex, Ʌ not N (which would be precluded by T). These traces, although slight, would suit the common name Vitalis, already found at Catterick (RIB II.7, 2501.626), but this is far from certain.

¹¹⁴ In the broken edge to the left of L is the extreme tip of a downstroke, probably I or II (for E). The incomplete letter after LI might be M, but is probably Ʌ especially since the sequence -ilia[nus] is typical of many names.

¹¹⁵ The first letter is incomplete, but is probably C, N or V. The second letter is also incomplete, but can be read as O made with two elliptical strokes joined by a short horizontal stroke; it is hardly A. The next two letters are complete, and probably C rather than L. The fifth letter must be a vowel, whether A, II (for E) or I. The sequence -occ- might suggest Cocceius, but a nomen would be unusual in such a graffito. Other possibilities are the Celtic names Voccio and Coccus, but there is no obvious candidate.

¹¹⁶ The second vertical stroke of N cuts the diagonal, confirming that the graffito was made this way up. The elliptical strokes of O and the first vertical stroke of N were made twice. A tiny nick in the broken edge to the right of I might be the tip of V, if it is not casual damage. The sequence -oni- or -oniu[s] would suit many names such as Antonius and Petronius, or (in the genitive) Senecionis and even Caionis (compare no. 58 above).

¹¹⁷ This graffito is non-literate, but is included here for possibly adding to the written evidence left in Britain by soldiers from North Africa, which includes RIB 653, 783+add.; II.8, 2502.14; III, 3445.

¹¹⁸ A is made with two diagonal strokes meeting at an acute angle, enclosing a shorter diagonal stroke parallel to the first which meets the second. The lower curve of C is rather faint and apparently encloses a very short vertical stroke, but there are no finishing strokes (repeated) like those of G. V is almost ligatured to S; its second diagonal is cut by the upper stroke of S, the lower stroke of which is now only a tail which terminates below V.

The name Camidigus is unattested, but the unique Germanic name Gamidiahus (RIB 2096, Birrens) might be a variant, supposing initial c were voiced to g and medial g lost before the back vowel u. It is preceded by the genitive termination of a masculine name in -anus, which is more likely to be that of the potter's master (whether he were a slave or only working in his officina) than a patronymic, which would surely have followed his name.

¹¹⁹ N and N are both incomplete, but the name Ingenuus is so common that they can be read with confidence. The first N was evidently made with a long diagonal that concluded in an upward curve. The second N has lost its third (vertical) stroke, but cannot be A (since the sequence -ngea- is impossible), and there is just enough surface left to exclude M.

¹²⁰ Like the next nine items. Twenty-eight coarseware graffiti made after firing have been excluded, since they are non-literate or preserve fewer than three complete letters or numerals.

¹²¹ Both graffiti seem to be complete, but they are unexplained. The bowl is much too small for ‘22’ to be a measurement of capacity in ‘pounds’ (librae) or ‘pints’ (sextarii). It now weighs 650 g, the equivalent of two ‘pounds’ (librae) or 24 ‘ounces’ (unciae) at 654.9 g, but another 50 g must be allowed for the missing pieces. Assuming it to be a hemisphere of radius 82.5 mm, its volume would be 1176 cc: the ‘pint’ was divided into 12 ‘cups’ (cyathi), but 22 would only amount to 1001 cc.

In graffito (b), the first letter (Ʌ) is larger than the letters which follow and set rather below, as if it were separate. The tail of R is extended. I is rather close to C, but it would be difficult to read these letters as O. ɅVRCI might be the genitive of *Aurcus, but this is not attested as a name, although it seems unlikely that someone who was called (say) Aurelius Civilis would abbreviate his name to AVR CI instead of writing CIVILIS.

¹²² The broken edge cuts the first P, and the graffito is probably not complete. A note of the weight when ‘full’ would have been preceded by a note of the weight when empty, as in RIB II.8, 2503.6–33, t(esta) p(ondo) with a numeral.

¹²³ Crescens is a common cognomen, well attested at York (RIB 652, 671, 679, 695) and Vindolanda (Tab. Vindol. II and III, Index II).

¹²⁴ The letters are rather small (c. 7 mm high) and not well defined. The final letter is two vertical strokes, the second topped by a downward diagonal. This cannot be II for E, since E has just been made quite differently with three horizontal strokes; it is presumably S since it concludes the name, and was clumsily made with three separate strokes, two of which (and not only one) should have been diagonal. The name Hermias, although of Greek derivation, is common in Italy and the western provinces, but this is its first occurrence in Britain.

¹²⁵ Initial I is exaggerated in height. N is broken between the sherds. The next letter is a downstroke tending to the right, as if L, but this would be unlikely after N. It might be V which has lost its second diagonal. After that, the surface is worn, with trace of one or two downstrokes, and then quite a marked downstroke which might be a concluding I. The name Ianuarius is quite common.

¹²⁶ The broken edge has removed the lower half of NIɅ, but, by continuing what is left of the strokes, the restoration is easy. Iunius, although a Latin nomen, is often found by itself: in Britain, see Britannia 47 (2016), 407, no. 24 (with note) and note especially RIB II.8, 2503.296 (Catterick), Iunia. The scripts are not distinctive, but the hand might be the same; quite likely it is the same woman.

¹²⁷ S was economically made with two overlapping diagonal cuts; a double curve would have been difficult to make. The cross-bar of T is the only cut which is not transverse. The technique is the same as that used for the numeral-like notches (III, V, etc.) often found on rims and under foot-rings (for example no. 27(b) above). Iustus and its derivatives is a common cognomen.

¹²⁸ The second R is almost complete, and with its exaggerated descender suggests that the previous letter is also R with an even more exaggerated descender. The scribe seems to have repeated letters for decorative effect, to judge by the neat zigzag to the left of RR, which must be M or ɅɅ. The personal name Marinus is quite common.

¹²⁹ The loop of P was partly repeated, and just enough remains of the sherd to exclude R. Prudens is the most likely name, but the sequence -pru- is found in others.

¹³⁰ By metal-detector, PAS ref. WREX-8D3982. It has now been bought by Wrexham Museum. Details, including a drawing, photographs and RTI scans, were sent by Susie White and Steve Grenter.

¹³¹ The transcript distinguishes between (i) true ligatures where letters share an element, for example R͡I, P͡R and T͡R͡E, by linking the ligatured letters with an inverted breve, and (ii) letters which have been reduced in size to fit within, above or below another letter, for example CaeS, Op and maxImo, by reducing the font-size. In three words, maGvl, fvSvm and maxImo, the mould-maker saved space by grouping reduced capitals symmetrically around a central letter of full size. He marked word division and abbreviation with a small triangular interpunct best seen after CaeS and above I͡N, but with traces elsewhere except after X and Op.

¹³² Neronis may have been preceded by IMP for imp(eratoris), but the only other Neronian lead pig (RIB II.1, 2404.24) omits it, as does the only Claudian lead pig (ibid. 1). The expansion of BRIT as (plumbum) Brit(annicum) is deduced from the Claudian pig's Britannic[u]m and confirmed by the Brit(annicum) ex arg(entariis) of Flavian and later pigs from the Mendips (VEB) and Derbyshire (Lutudarum). This might suggest the next word (after X) was ARG, but the lower letter is a complete A, while the upper (and first) letter is the left half of M, its entire diagonal excluding R and also too steep for N; the right half of this M has been lost in a gash. Nonetheless, the large X must introduce the source of the lead: it is possible that a leftward horizontal stroke has been lost in the damage, for ligatured (e)x, but this would have resembled a denarius symbol, and it is more likely that the mould-maker, unless he omitted E by mistake, reckoned that a bold X would do for ex.

The next word, abbreviated as MAGVL, is the source of the lead, and must correspond to the DECEANGL of Flavian lead pigs from the Clwyd field (2404.31–6). Initial mag- is a frequent name element (A.L.F. Rivet and C. Smith, The Place-Names of Roman Britain (1979), 405, s.v. *Magalonium), and MAGVL may represent the place-name *Magalonium inferred by Rivet and Smith from MACANTONION in the Ravenna Cosmography, which it apparently locates between Gloucester and Leintwardine. The end of the next word, FVSVM (‘melted’, ‘smelted’) is damaged, but is not X: despite damage due to the gash, it can be seen that the diagonals did not coincide; they belong to an almost complete V above M which has lost its left half. The expansion of OP as op(eribus) is conjectural, since it seems to be otherwise unattested; but in the context of in prov(incia) it makes sense. The governor's name and title, ingeniously condensed, complete the legend in the ablative case, for dating purposes. This too is unparalleled among British lead pigs.

¹³³ In material from the J.P. Bushe-Fox excavations by Philip Smither, who sent a photograph.

¹³⁴ A reference to the cohort is attractive since it was the garrison of nearby Reculver, but this was in the third century and the sherd comes from a mid-second-century context. Besides, it would be odd for a cohort to ‘own’ a jug. These difficulties were noticed by Roly Cobbett who told Philip Smither, who sent photographs. The drawing in RIB misrepresents the downstroke of L, which does not bend at the top, and exaggerates the cross-stroke of I (which Bushe-Fox read as T), which is probably a top-serif. For similar graffiti of LIB, see RIB II.8, 2501.289, 290.

¹³⁵ The new fragments confirm the reading of horrioli (7) and the intercolumnar addition on the left at right angles, which however remains illegible. mitte (9) repeats the abrupt imperatives of (i).4 and (ii).4. After it, Verecundus seems to have omitted the relative pronoun quae, perhaps by confusion with quam in 6 and 8, but possibly [d]ebuerat began a new sentence. There is no sign of the missing d, but the space is there.

¹³⁶ Information about this and the next four items was given to Scott Vanderbilt by Gill Woolrich, Archaeology Curator, Southampton City Council Cultural Services.

¹³⁷ Acc. no. A.1996.186.

¹³⁸ Acc. no. A.1996.187.

¹³⁹ Like the next item, it was on display at the Museum of Archaeology until 2011.