I Captivated by curses

Fig. 1. Drawing of the Moorgate curse tablet published in RIB I, drawn by R. G. Collingwood, with minor amendments by R. P. Wright.

Tretia Maria’s vivid curse appears, somewhat unexpectedly, as the seventh entry in The Roman Inscriptions of Britain Volume I: The Inscriptions on Stone (1965). Found in Moorgate, London, in 1934, it is not written on stone but on a thin sheet of lead inscribed in neat Old Roman Cursive script, pierced with nails driven through the uninscribed side. When it was discovered, it represented an unusual type of inscribed object, one of just a handful of so-called defixiones, or ‘curse tablets’, from Britannia. Auguste Audollent’s authoritative collection of eastern (except Attic) and western defixionum tabellae from 1904 had published only two texts from that province, the ‘Vilbia’ tablet from Bath in Somerset (DT 104; = RIB I 154) and the dedication devo Nodenti ‘to the god Nodens’ from Lydney Park in Gloucestershire (DT 106; = RIB I 306). Both, it turns out, were precursors to the large corpora recovered through excavation of the sanctuary of Sulis Minerva at Bath and, just across the River Severn from Lydney Park, at Uley.Footnote ² These curse tablets have captured imaginations from the start, sometimes luring commentators into imaginative interpretations. The tablet found alongside Vilbia’s at Bath in 1880 was published by Edward Nicholson in 1904 as a letter from Vinisius to Nigra, and heralded as the first documentary evidence of Christianity from the province.Footnote ³ This interpretation was brilliantly and wittily dismantled by Roger Tomlin: the scrawly New Roman Cursive had been misread and both sides of the tablet read upside-down.Footnote ⁴ These tablets are now all understood as curse tablets familiar from across the western provinces: texts (usually) on lead sheets which ‘intended to influence, by supernatural means, the actions or welfare of persons or animals against their will’.Footnote ⁵ Those found at the sanctuary at Bath, largely lists of names of potential culprits and texts demanding cruel justice for thieves, were inscribed into the UNESCO Memory of the World Register in 2014.

The bulk of the defixiones in Audollent’s corpus hail from the East, North Africa (especially Carthage and Sousse) and Italy. Audollent’s chosen descriptor, defixio, has since been supported by the appearance of the word defisonis in a late-antique prophylactic inscription found at Furnos Maius (Aïn Fourna, Tunisia) in 1923, and, more recently, by defictcsione in one of five curse tablets found near the vicus of the Roman fort of Abusina (Eining, Germany).Footnote ⁶ The English term usually deployed is ‘curse tablet’, and this features in the titles of all the volumes under review.Footnote ⁷ In Audollent’s collection there were slim pickings from the north-western provinces, and Greek content outweighed the Latin. In the past half-century, the picture has radically changed with the discovery of a small number of substantial caches of Latin tablets, particularly from Bath (c. 140 tablets with c. 110 inscribed, and more if ‘pseudo-inscriptions’ are included), Uley (c. 140 tablets with nearly 90 inscribed) and Mainz (34 inscribed tablets), and numerous scattered finds.Footnote ⁸ The inscribed objects provide extraordinary source material for exploring language, onomastics, literacy, everyday life, social dynamics, psychologies, the practice of ritual and magic, the nature of the economy, agriculture and textiles, inter alia. Scholarship has embraced them in various volumes dedicated to ‘magic’, several issues of the Religion in the Roman Empire series devoted to ‘curses in context’, and important recent monographs, including the three volumes under review here: Sánchez Natalías’s extensive collection of, and slim commentary on, non-Greek curse tablets from the Roman West; Tomlin’s eagerly anticipated corpus of the Uley tablets; and McKie’s theoretically-informed monograph on the practice of cursing in the Roman West.Footnote ⁹ In recent decades volumes such as these have rescued the study of western curse tablets from a relatively fringe status, drawn them out from the shadow of Greek binding spells, and created an energetic—though perhaps not always productively synergetic—research community.

II The triad

The three books reviewed here are marked by differences in approach. Sánchez Natalías produces a remarkable collection of curse tablets from the Roman West, another doctoral project from Zaragoza with epigraphic collection at its heart.Footnote ¹⁰ Volume 2, the sylloge, marshals and, in some cases, re-edits curse tablets published before summer 2018, excluding Greek language and all African examples. Its 575 pages eclipse Volume 1, which weighs in at just 75 pages and offers focused commentary on the nature of the assembled material, including the African texts. Sánchez Natalías underscores that, unlike others after Audollent, she has personally inspected many tablets and offers improved readings. In line with her keenness not to produce ‘just another compilation of texts’ (2022: 4), she provides drawings and occasionally images, and details of the iconography. She expresses the hope that this justifies her decision not to include the African material, as these ‘texts are highly visual, and publishing them without their iconography would be a serious shortcoming’ (2022: 83). She does not explain why György Németh’s publication of Audollent’s 86 sketches of the North African materials could not suffice, and she does include these African-origin tablets in her commentary volume.Footnote ¹¹ She impresses on readers that she has produced a sylloge and not a corpus: ‘the geographical dispersion and the sheer volume of the corpus have made it difficult to apply the best epigraphic criteria for every piece in the catalogue’ (2022: 83).Footnote ¹² In the Bryn Mawr Classical Review Eleanor Dickey takes issue with the lack of apparatus criticus, noting that ‘it is high time editors of curse tablets joined those of inscriptions and papyri in providing a proper record of where their readings come from’.Footnote ¹³ Sánchez Natalías has made numerous improvements to our understanding of the texts, including identifying some rogue lead tags and a label of an officina plumbaria, but she offers many minor improved readings which are easily overlooked in the absence of proper exposition, and conversely there are occasional slips in the presentation which could easily be mistaken for re-readings.Footnote ¹⁴

Tomlin’s long-anticipated volume of the Uley curse tablets presents a feat of epigraphic wizardry and a complete edition to replace previous publications. Since significant time has passed since these tablets were found, the epigraphist had already released several texts in his interim report within the Woodward and Leach volume of the Uley excavations and in his annual epigraphic round-ups in the journal Britannia.Footnote ¹⁵ Nevertheless, the combination of material and the care with which it has been presented to the reader—though again, like the Sylloge, he offers no apparatus criticus—make the volume a desirable, if not especially affordable, item.Footnote ¹⁶ In the introductory chapters Tomlin deliberately mirrors the content of the sections of his 1988 publication of the tablets from Aquae Sulis (Tab. Sulis), a decision with plenty to recommend it, particularly for those who wish to use both volumes side-by-side and to compare the two sites.Footnote ¹⁷ However, the Tabellae Sulis volume was a product of the scholarly environment of the 1980s, and Paul Russell remarks in his critical review of Tab. Uley that ‘time has moved on and the methodological frameworks within which these tablets are to be considered has changed significantly; as a result, the volume has a rather dated air about it’.Footnote ¹⁸ Despite this, Tomlin’s Uley edition is undoubtedly a work of art, reflecting deep epigraphic skill and experience, and infused with a compelling concern, even if not theoretically driven, for the local people who are the volume's protagonists.

Conversely, Living and Cursing in the Latin West is pervaded by the strides made in investigating lived experience, materiality and practice, particularly as associated with the Lived Ancient Religion and New Materialism movements.Footnote ¹⁹ McKie’s engagingly written volume takes the approach of ‘creative speculation’ (2022: 2), and uses the curse tablets as a springboard to enter the lives, psychologies, and social contexts of those involved. The deep engagement with theoretical perspectives, generally close attention to context and constant self-questioning reflect a researcher who has looked at the material from multiple vantage points and has become practised in defending his stance.Footnote ²⁰ The result feels productive and persuasive. McKie pushes back against previous scholarship that has taken narrowly philological and/or linguistic approaches to large collections, or approached the ritual importance of the material from a ‘pan-imperial tradition’. Though he argues that ‘it is essential to have a sense of the backdrop’ (McKie 2022: 12), ‘[s]uch macro-scale analysis is unsustainable, and fundamentally misses the point of curse tablets, which were individualized reactions to personal crises’ (2022: 3). He states that his work ‘represents a paradigm shift’ (2022: 3), bringing to the fore individual psychology, gossip and rumour, webs of human interactions and their embeddedness in physical contexts and with objects, arguing that work on Roman curse tablets has remained largely immune from ethnographic and anthropological developments (2022: 88).Footnote ²¹ McKie is arguably not shifting the paradigm, but rather following trends which have deeper histories, for example in the ‘psycho-social’ approaches pioneered, amongst others, by Richard Gordon, and in the focus on materiality.Footnote ²² Strikingly, an article by Greg Woolf published in the same year as Living and Cursing mirrors many of McKie’s ideas, though taking an ‘actor-network theory’ approach.Footnote ²³ McKie’s intervention is nevertheless lucid and important. He is up-front about the limits of the analysis, the randomness of finds, the patchiness of contextual information and our inability to know much of the social context, but squeezes a good deal from what we have, with cautious archaeological awareness.Footnote ²⁴ He notes that at Mainz, where the deposition of tablets occurred in brick-lined fire pits behind the cella of the temple of Isis and Magna Mater, at least three of the cursers, whose texts have not completely melted away, planned their curse aware of the next steps, including ‘sympathetic formulae transferring the burning and melting of lead into the bodies of the victims’ (McKie 2022: 55, on SD 488, 498, 499). It is disappointing that he did not have the space to interrogate in detail the depositional contexts of more of the tablets (2022: 51).Footnote ²⁵ His creative but sober and informed approach had the potential to develop our understanding further.

III Cursing practices

The Papyri Graecae Magicae — a fascinatingly eclectic modern collection of ancient advice for practising ritual drawing on Greek, Egyptian, and Roman traditions — states that the finished curse tablet should be rolled.Footnote ²⁷ Sánchez Natalías highlights the regional differences in the manipulation of the tablets and links this rolling to the standard manipulation of papyri (2022: 32). However, she also notes that at Sousse and Carthage the rolled tablets were often deposited into cinerary urns via the libation tubes, so the rolling could simply have been a practical solution. The thickness and the type of metal used will also no doubt have made different methods of manipulation more or less practicable.Footnote ²⁸ Sánchez Natalías notes that the African curse tablets are very often rolled (33 out of 88 instances), twice rolled and nailed, but are only folded in 8 cases, whereas the British sheets are much more commonly folded (72 out of 256 examples) than rolled (just 6), with 12 instances of nailing and 7 of folding and nailing (2022: table 5.1). However, there must be a problem with these British figures, since Tomlin makes it clear that most of the Uley tablets (both inscribed and not) had been rolled and flattened (2024: 7): 109 according to the excavation volume of Woodward and Leach (in addition 5 were rolled but not flattened, 9 had not been rolled, and for 17 no judgement could be made).Footnote ²⁹ In the Uley tablet presentations, however, Tomlin describes all the tablets in relation to the number of folds, or lack of folds for example, and only one tablet, Tab. Uley 78, is described as having been ‘rolled up’. Given Sánchez Natalías liaised closely with Tomlin in the creation of her volumes (2022: ix), these descriptions have perhaps led to her over-estimation of this aspect of the differences in practices across the Empire. Conversely, she usefully corrects the widely held misapprehension that nailing tablets was an iconic part of the cursing ritual. Only 72 tablets, 11% of the corpus of non-Greek western defixiones, were found ‘closely associated with nails’, with just 66 pierced (McKie 2022: 40; Sánchez Natalías 2022: 33). We should also remember that the nailing can happen before, or after, the rolling/folding. Examples of pre-manipulation nailing can be seen in the examples where the nail has pierced specific words, for example the curse of Anniola (Tab. Sulis 8), whose name is pierced on its two appearances, one on each side.Footnote ³⁰

Similarities in practice exist across the provinces which allow the definition of the defixiones as ‘physical remains of a series of actions, gestures and movements that could have included writing, speaking, folding, nailing and depositing’ (McKie 2022: 23), but there are also differences, not only in physical manipulation of the object as we have seen,Footnote ³¹ but, perhaps more strikingly, in text and context, carefully illustrated by Sánchez Natalías and McKie. The curses related to theft and similar matters, called defixiones in fures, calumniatores et maledictos by Audollent when this category was known through a mere five examples, now form the biggest sub-set when those of unknown function are excluded.Footnote ³² Debate has rumbled around the question whether these should form a separate group designated by Versnel as ‘prayers for justice’, and be split from the ‘main’ group of Graeco-Roman defixiones, which were traditionally divided into four main categories, agonistic, juridical, commercial and amatory, based on the Greek examples.Footnote ³³ These ‘prayers for justice’ present a series of features which mark them out from the broader defixiones group, but Sánchez Natalías and McKie make a strong case for seeing them as part of a continuum of practices. McKie (2022: 64) questions the suitability of Versnel’s term, preferring ‘demands for vengeance’ since the petitioners, at least in some cases, seek ‘the disproportionate punishment of perceived wrongdoers far beyond the normal limits of justice as defined by the legal system’.Footnote ³⁴ Sánchez Natalías (2022: 61) also identifies the primacy of ‘aggression’ and prefers Audollent’s original defixiones in fures. ‘Prayers for justice’, or Tomlin’s rather incongruously bureaucratic ‘petitions full of complaint’ (2024: 17), may indeed fail to capture the animus of many of the texts, particularly that from Groβ-Gerau (SD 483) which demands that worms, tumours and vermin invade the thief’s body after stealing a cloak.Footnote ³⁵ Howsoever we wish to refer to these curses concerning theft, it seems clear that treating them within a flexibly broad group of curses is important for capturing the wider picture of connections and influences, and the birth and development of clearly interconnected practices.

According to Sánchez Natalías’s analysis, just under half of the categorised non-Greek corpus from the West consists of these demands for justice, and, of these, 80% are British (2022: table 7.1).Footnote ³⁶ But is it true, as Tomlin states, that we have the excavators of Bath and Uley to thank for that?Footnote ³⁷ That is, is the distribution a reflection of ancient regional practice, or largely due to modern archaeological practices? 72 of 88 defixiones from Africa derive from funerary contexts, whereas Britain offers just one (Sánchez Natalías 2022: table 6.1). Given the high number of carefully excavated Roman funerary contexts from Britannia, it seems likely, if they had existed, that some defixiones would have been recovered. While we could argue that the corpus is simply too biased to discuss distributional patterns of aspects of cursing, McKie (2022: 84) and Sánchez Natalías both agree that the trends seem too strong to be random, though arguably the latter’s commentary and maps of distribution across the provinces over time in ch. 8 create an unwarranted sense of confidence in the data, whose dates, in particular, are often extremely insecure.

Perhaps as many as a quarter of all surviving curses in Greek are erotic, but Sánchez Natalías collects only 23 out of 623 within her non-Greek western corpus, with 14 from Africa alone (2022: table 7.1).Footnote ³⁸ Similarly, the agonistic curses, so well represented within the Greek corpus, are in the West almost entirely an African phenomenon, with 37 of 39 total examples (2022: table 7.1). McKie argues for flurries of activity in short spaces of time by specialists engaged in circus curses at Carthage, Sousse and Rome, presumably creating a buzz around the success of their services. He suggests that at Carthage the curse producers, who successfully created a market for charioteer-related texts, tried unsuccessfully to drum up a market in beast-hunting curses (McKie 2022: 28, 71), reflecting the sense that curse tablets were seen as ‘an evolving technology that could be adapted to changing social circumstances’ (Sánchez Natalías 2022: 16).Footnote ³⁹ Unlike the majority of the curses from the northern provinces, the African material seems to be founded on knowledge of Graeco-Egyptian magical tradition, and relied on experts for the evocation of the appropriate voces magicae and other features (McKie 2022: 28; Sánchez Natalías 2022: 16).Footnote ⁴⁰

McKie argues that the social character of cursing is fundamentally about taking control, a creative tool for the curser to define their place in society, rebalance skewed power relations and navigate issues, both short-term and ongoing. His lived ancient religion approach makes it easy to account for variation across time and space (2022: 130) and to see the practice as alive and populated with individuals within their different communities, not with lone practitioners. However, these texts are private by their very nature, a communication with the supernatural world to demand, for example, awful punishments, often folded/rolled and almost always deposited out of sight, commonly in cemeteries, sometimes popped down libation tubes. McKie does not quite resolve this acknowledged tension in his analysis, where the practice is simultaneously furtive and personal but with significant societal involvement and ramifications.Footnote ⁴¹ Sánchez Natalías’s vision of the cursing evokes a more traditional view of a private and illegal practice (2022: 5), but she also argues for the utility of applying Gordon’s description of ‘vernacular’ practice from Italy to the West.Footnote ⁴² The shared understanding which circulates and creates regional specificities clearly requires the practitioners to be connected in shared webs of knowledge and practice.

IV Writing curses in the Western provinces

Central to these shared webs of knowledge and practice are questions of literacy, and the involvement of scribes and perhaps ‘freelance experts’ with specific knowledge of the ritual practice in creating curses.Footnote ⁴³ These questions drive much discussion in our trio of volumes and the petitioner is at the centre of their analyses: making the curse perhaps exclusively orally in the cases where the folded tablets appear to be blank; producing ‘pseudo-inscriptions’ if illiterate; copying out texts produced by experts; relying entirely on others to produce the curse; and independently creating their own texts, with or without the aid of manuals.Footnote ⁴⁴

Presumably an expert was involved at some point in the process to compose the five nearly identical, lengthy and elaborate curses from Italy (possibly found just outside the Porta Salaria in Rome), targeting five different individuals and addressed to Proserpina (SD 10–14).Footnote ⁴⁵ Similarities in texts suggesting specialist practitioners can be found also at Hadrumetum (Sousse) (DT 272–274), and the spaces left for adding the names of the curser and victim in the text found in a grave at Carthage leave the involvement of an expert in little doubt (DT 230). Sánchez Natalías (2022: 6) argues that the use of such professionals is not attested before the second or third century c.e. in North Africa and before Late Antiquity for other provinces and Italy, though McKie (2022: 28) notes the issues with dating the tablets and implies a much more continuous history of engagement of cursers with specialist knowledge and experts. Certainly the choice of complex layout, with reversals of lines of letters or individual words for example, iconography, transliterations of Latin texts into Greek, use of voces magicae, magical characteres etc., is much more common in certain regions than others.Footnote ⁴⁶ 62 out of 88 texts from Africa have such features, compared to 14/59 in the Germanies or just 21/256 in Britain (Sánchez Natalías 2022: table 4.1).

The complexity, layout and length of some curse texts indicate that in several cases drafts must have been prepared for others to copy or that scribae wrote the texts themselves. Perhaps the clearest evidence for the former, apart from the slips across the corpus which can be attributed to copying errors, comes from Bath, with the phrase carta picta perscripta ‘the written page has been copied’ (Tab. Sulis 8). Scribal practices might also be detected in several of the curse tablets from Roman Britain thanks to the use of the ‘clerical’ abbreviation for supra scripta (barred SS) found several times at Uley, Caistor St Edmund, Ratcliffe-on-Soar and Eccles.Footnote ⁴⁷ Indeed, at Bath and Uley the switches between New Roman Cursive (NRC) openings and then the bulk of the text in Old Roman Cursive (ORC) (e.g. Tab. Uley 8 and Tab. Sulis 66) may suggest the petitioner’s involvement alongside a scribe.Footnote ⁴⁸ Perhaps Tomlin is right that the Saturnina, who is described as mulier, is unlikely to have written that text herself (Tab. Uley 2; Reference Tomlin2024: 43, n. 11). However, his detailed tabulations of script forms, which are fundamental in quantifying how many different hands may be at work, in scrutinising his readings and in building our appreciation of handwritten Latin forms across the imperial period, show that at Bath and Uley only two tablets (Tab. Sulis 95 and 96) seem to involve the same hand (Tomlin Reference Tomlin2024: 76).Footnote ⁴⁹ If scribae were used extensively for the final versions, we might expect less variety in handwriting and, given the personal and emotive nature of much of the content, perhaps autography in the direct petition to the deity was deemed more potent.

Tomlin considers that more duplication of texts might be expected at Bath and Uley if professionals were commonly involved in the drafting. Instead, there is clear understanding of the type of format of text that should be composed and of the deployment of formulae, but variation in precise formulation and spelling, in some cases reflecting spoken language. Tomlin suggests ‘“amateurs” at work’ (Reference Tomlin2024: 77). He cunningly proposes that one of the Uley cursers, Biccus, has used a handbook and accidentally mangled his selection of the standard British binary formulae (si vir si femina or si mascel si femina) to describe the thief, inadvertently producing si vir si mascel ‘whether man or male’ (Tab. Uley 4). In this case, given the several mistakes in the text and the uncertain capital script, we can be relatively confident in imagining a worshipper using a manual. The appearance of the si-formula in curse tablets from Italica in Spain and Aquincum in Hungary, and the verbal resonances between per tuam maiestatem te rogo oro obsecro uti vindices quot mihi furti factum est (SD 120) at Mérida, Spain, and rogo [s]anctissimam maiestatem tuam [u]t vindices ab his [q]ui [fraude]m fecerunt (SD 240) at Bath, might encourage us to think that handbooks circulated widely, though we cannot exclude coincidental creations based on broadly similar oral practices.Footnote ⁵⁰ For the British context at least, Tomlin affirms that usually these ‘[h]andbooks, if indeed they existed, were being used at second-hand. In other words, petitioners made inquiry and were told what to say’ (2024: 78).

There appears to have been ‘access to the same fund of formulae’ at Bath and Uley since Tab. Sulis 10 and Tab. Uley 43 are strikingly alike, and, despite the similar names involved, Tomlin argues that they are probably not the output of the same composers.Footnote ⁵¹ In south-western Britannia, then, perhaps freelance religious experts, or the priests of Sulis-Minerva or Mercury, helped at least some of the devotees to prepare and make the curse. In some cases, this may have entailed a translation process: a conversation in British Celtic, or in a mix of British Latin and Celtic, to determine the necessary details for the curse before a (usually) Latin utterance was produced. This may have been written straight out by the devotee, or the expert, or may have gone through a drafting phase and was subsequently copied by the devotee. In other cases, perhaps literate devotees consulted manuals directly. In others still they may have had the confidence (or lacked the time, connections, money, or willingness to share the details) to go it alone and drew on knowledge they may have gleaned from oral ritual practices or shared conversations.

The implications that the large number of hands might have for the levels and spread of literacy in Roman Britain are essentially ducked by our authors. We have a scatter of curse tablets (Tomlin Reference Tomlin2024: 5), often betraying local concerns and language, across the so-called ‘Lowland’ zone.Footnote ⁵² We then have two large caches at sites in the south-west, both of which seem to have been produced over a long period of time, from the second to the fourth centuries c.e., and involving around 200 different writers. The tablets take in numerous individuals, many of whom in the case of Uley seem to be from rural and farming communities, and the majority of whom seem unlikely to be high on the social ladder. There are community links and similarity of population groups between the two sites, reflected in the common naming stock.Footnote ⁵³ Of the at least 40 names at Uley, Tomlin notes that 18 are found at both Uley and Bath (2024: 42, 45). The fact that the petitioners turn to the gods for justice might indicate their relative powerlessness to achieve it through legal routes in the Roman provincial context.Footnote ⁵⁴ The choice of the majority of these individuals to use Latin and writing as part of this cursing practice tends to be taken for granted, especially now the tablets from Aquae Sulis are so well known. But given the linguistic and epigraphic context of the north-western Empire, we should check our assumptions; though of all the regions of Britannia, it is this part, namely the Chilterns–Cotswolds corridor, which the mass of writing equipment and other inscribed finds suggest may be attaining higher levels of rural literacy, perhaps particularly in the later Roman period.Footnote ⁵⁵

V Local languages and bilingualism

One fascinating aspect of the curse tablets is the insights they bring to local languages and socio-linguistic variation. The most striking previously unpublished texts within the Uley corpus are numbers 7 and 35, described by Tomlin as ‘Un-Latin texts’ (Fig. 2). This formulation reflects Tomlin’s scholarly caution, but, combined with a lack of detailed discussion of the language of the texts (see further below), the consequence may be that readers do not appreciate what is at stake.Footnote ⁵⁶ Though the transcription and interpretation are both uncertain, it is possible that these texts contain Celtic-language content.Footnote ⁵⁷ It has previously been argued that two tablets from Roman Bath, Tab. Sulis 14 and 18, are likely to be Celtic; the shorter complete example, with which this section opens, even renders a text that makes some sense (my liberty with the translation of a ‘not-good thing’ (andagin) may perhaps be permitted, given the cursing context).Footnote ⁵⁸ The conundrum has been whether the language of these texts from the sanctuary at Aquae Sulis is more likely to be British Celtic or Gaulish, the Celtic languages of Britain and Gaul respectively. The problem is in part caused by the lack of continuous written Celtic from Britain — we construct it primarily through names and the later testimony of the Brittonic languages — but also thanks to the (unsurprising) similarity of the languages either side of the Channel in this period. We cannot, in the current state of evidence, securely diagnose any linguistic features which may have distinguished the languages. The argument against continuous written British, as opposed to Continental, Celtic at Bath was always that it was otherwise unattested, whereas across the Channel Gaulish was used in several texts on metal tablets (e.g. at Chamalières, Chartres, Larzac), some of which can be interpreted as relating to the magico-ritual realm.Footnote ⁵⁹ Aquae Sulis was a cosmopolitan sanctuary site, with dedicators hailing from Gaul inscribing their origins in stone (e.g. RIB I 149, a lapidarius from the area of Chartres). Twenty years ago, with just the Celtic-language Bath tablets in play, it seemed wise to stay on the fence when it came to proclaiming what kind of Celtic they might contain.Footnote ⁶⁰

Fig. 2. Tab. Uley 7, side a. Photo: © Centre for the Study of Ancient Documents, reproduced with kind permission.

In line 2 of number 7 of the Uley collection, and then again on the other side of the tablet, Tomlin suggests that it is possible to restore LVCIVMIO, which happens to be the first word of Tab. Sulis 14, possibly repeated in lines 4 and 5, which may mean ‘I swear’ or ‘I ask’.Footnote ⁶¹ The striking possible echo in two tablets which have been identified as possibly being written in, or partly in, Celtic from two separate Roman-period religious sites feels irresistible and, if both the Bath and Uley tablets contain Celtic, the scales start to tilt in support of the four examples involving British Celtic rather than Gaulish. The temple of Mercury at Uley was a less cosmopolitan hub than the sanctuary at the thermal springs of Sulis Minerva; further, we assume, based on a range of clues, particularly the links to farming in the items stolen (e.g. vas apium in no. 24, ‘the first document of beekeeping in Britain’: Tomlin Reference Tomlin2024: 38) and animals harmed (Tab. Sulis 43, 76), that many of the users of the shrine were locals.Footnote ⁶² In this context starting to consider the Celtic of the tablets as British Celtic feels more plausible.

Indeed, curse tablets across the western Empire provide insights into sub elite-level local society and are a more fruitful, though still relatively meagre, source of indigenous languages than much of the epigraphic record. In addition to the four tablets which may be written in Celtic, there are possible glimpses of Celtic in the Latin examples. Tab. Uley 33, although damaged, contains the not-obviously Latin word aexsieumo in line 3, which may be a Celtic verb, though as yet no interpretation has inspired confidence.Footnote ⁶³ Russell has argued that Tab. Sulis 4 (RIB I 154), which has often been thought to bemoan the kidnap of a woman named Vilbia (unparalleled amongst curse tablets), might actually deal with the theft of a Celtic-named pointed object, perhaps a knife (linked to Middle Welsh gwlf) (rather disappointingly for fans of the Cambridge Latin Course, where the lady Vilbia features in a segment set in Bath).Footnote ⁶⁴ Similarly, Tomlin suggests there may be ‘a Celtic loanword’ in Tab. Uley 33 referring to some kind of hut, though Russell has recently noted that there is no obvious Celtic cognate and has suggested a re-reading to recover instead a Late Latin form retiis, perhaps denoting a net-like container.Footnote ⁶⁵ Another intriguing form appears in the Hamble estuary curse tablet, niske (line 8), though this might be a name.Footnote ⁶⁶ This echoes a similar form in the hard-to-follow tablets (now lost) from Amélie-les-Bains in the valley of the river Tech, which have turned out to be a multilingual playground for linguists, with one commentator even using them to invent a putative language, ‘Sorothaptic’.Footnote ⁶⁷

While we should be wary of the invention of new ancient languages, in many tablets from across the western provinces we do find evidence of mixing of languages and scripts. Sánchez Natalías notes that around 10% of all curses from the western provinces involve a combination of language and/or scripts (2022: 19), though some provinces show significant divergence from this average, with Britannia, for example, offering in the order of 2%. It has been a feature of scholarship on epigraphy generally, and cursing specifically, that Greek and Latin texts have been treated separately.Footnote ⁶⁸ Sánchez Natalías’s collection of curses from the Roman West includes Latin, Greek-Latin bilinguals, Etruscan, Oscan and Celtic examples, but excludes all Greek-only examples. In an ideal world all would be analysed together, the bilingual examples finding possible contextual support and resonances within the broader corpus of Greek examples. Sicily, for example, on the basis of these materials seems a quiet place for curses, but is home to the earliest written evidence for cursing, dating to perhaps c. 550 b.c.e., with many discovered in recent decades.Footnote ⁶⁹ This deeper history, and the questions of how creators of non-Latin curses in Italy might have played a role in their transmission to the Roman world and how the cursing practices link to other local rituals both in Italy and across the provinces, remain in the shadows in these volumes.Footnote ⁷⁰

Bilingualism has rarely been properly considered in analyses of these texts. Scholars have traditionally focused on trying to fit the texts into the ‘correct’ monolingual categorisation (and thus deciding in which corpus they should appear), and often do not consider the prevalence of bi- and multi-lingual contexts in which they might have been produced. Appreciating the nature and extent of bilingualism in the western provinces can help to assess the mixtures of languages in the defixiones, for example to discriminate between features which may have been taken from manuals, with or without knowing the language(s) involved, and those which are reflections of spoken realities. It is a striking feature of several of the possible curse tablets from Gaul that they may involve mixtures of Gaulish and Latin (e.g. Les Martres-de-Veyre L-102; Rom RIG II.2 L-103; Le Mans L-104; Paris L-105).Footnote ⁷¹

Similarly we ought to recognise that ‘Classical Latin’ is a construct, and regionality and social variation in the Latin spoken across the Empire were normal. While those who could write often grasped for established linguistic norms, we should embrace divergences from those norms as windows onto complex linguistic and educational realities, rather than seeing them as features to be corrected. Two chapters from the Uley collection touch on such matters. Ch. 9 on ‘Language and Spelling’ is essentially a list of examples of what Tomlin refers to as ‘Vulgar Latin’ (a term many linguists hoped had been retired). Colin Smith’s 1983 chapter on non-standard Latin in Roman Britain is used to order the features but there is little engagement with more recent resources, for example Adams’s Social Variation and the Latin Language.Footnote ⁷² ‘Morphology (etc.)’ takes in syntax and vocabulary, the examples for which are too few to inspire confidence. Tomlin separates ‘copying and other errors’ into ch. 10, but again there is limited analysis, and some examples need not appear in both lists (e.g. quattor/quator and viglet, Reference Tomlin2024: 67 and 71). There is more linguistic information in the commentaries on the individual tablets, but a coordinated discussion, which might have considered the evidence for non-standard forms of Latin, and possible influence from British Celtic, could have been usefully assembled here.Footnote ⁷³ The linguistic features are given attention across the volume, but Tomlin deeply misses the input and advice of his irreplaceable colleague J. N. Adams, highlighted in his wistful call for ‘a second Adams’ (2024: 65).Footnote ⁷⁴

VI Divining the future

The world of ancient cursing is constantly renewing itself, thanks not only to new scholarship but also to the steady stream of new finds. The excavations by the Service d’Archéologie de la Ville d’Orléans, begun in 2022 at l’Hôpital Porte-Madeleine in Orléans, France, have uncovered an approximately 5 m × 130 m strip of inhumation burials c. 1 km to the west of the centre of Roman Cenabum, and among the 80 skeletons in wooden coffins 22 curse tablets have been found to date.Footnote ⁷⁵ These are being analysed using Reflectance Transformation Imaging, and in some cases being imaged whilst still rolled, using medical scanning techniques.Footnote ⁷⁶ So far several seem to be inscribed in Latin and one, written in ORC but in Gaulish language, has recently been published.Footnote ⁷⁷ The burial has been dated to the middle or second half of the second century c.e. and the tablet was deposited between the legs of the skeleton (tomb F2199). The first full edition of this text by Pierre-Yves Lambert will be, as always, only the beginning of the history of its interpretations. Whether others will concur that the authors of this ‘charme magique’ seem to have deployed Gaulish to support ‘an anti-Christian ideology’, we wait to discover, but what is striking to any specialist of Gaulish is the astounding familiarity of the text.Footnote ⁷⁸ So similar are the structure, phraseology and forms to those of the Larzac, Chamalières and Chartres tablets that we must consider shared knowledge of Gaulish-language ritual practices across Gaul and perhaps even ‘the existence of Gaulish magical formulae, surely modelled on Greek and Latin’.Footnote ⁷⁹ As with other similar texts there are signs of bilingualism through possible Latin and Greek forms. Lambert considers that the names in the text imply a well-off and cultured environment, and that Gaulish language and culture had by this point become an ‘object of curiosity’ for educated ‘antiquarians’.Footnote ⁸⁰ Or we could instead see in this text the possible on-going vitality of the Gaulish language in religious and private domains.Footnote ⁸¹ The curse tablets still have so much to reveal about the communities and individuals that produced them; we need to continue to study them with a combination of multiple disciplines and by reaching across the artificial linguistic boundaries created around their corpora.

McKie also argues that interdisciplinarity is key to the development of the field; however, the nature of his promoted interdisciplinarity — essentially a mix of archaeology and anthropology — tends to reflect closed-off attitudes towards ‘more traditional’ and ‘less productive’ philological and linguistic analysis (McKie 2022: 10; Sánchez Natalías 2022: 4). But not only are these, and epigraphic, skills essential for the recovery of the curse texts themselves, the kinds of socio-linguistic approaches that have been increasingly pursued in the past couple of decades are also compatible with building a deeper understanding of the social contexts and the individuals involved in cursing practice. A resistance to the linguistic-philological realm has shut out potentially valuable complementary perspectives. What is needed is the integration of multiple disciplines, perfectly nested like the Russian-doll-style vessels with curses in the fountain of Anna Perenna (SD 24–7, 35–6), and combining a wide range of anthropological, archaeological, epigraphic, imaging, linguistic and palaeographic skills, which can realistically perhaps only be successfully achieved by multidisciplinary teams.Footnote ⁸²

The decipherment of the scratches on the metal objects might be sped up, and possibly made more complete, if modern digital imaging and computer science techniques are deployed. Tomlin’s preferred method of decipherment uses a combination of autopsy, plus scrutiny of four photographs captured with raking light, which he submits to ‘digital tracing’, essentially using the digital images in four layers to draw a fifth which combines the information from each using Adobe Photoshop.Footnote ⁸³ Constantly developing technologies, such as those used in the decipherment of the recently discovered tablets from Orléans or indeed those tested on Roman-period stylus tablets, hold the promise of more refined recovery of sometimes faint scratches on highly corroded surfaces.Footnote ⁸⁴ Modern computing technology can analyse the nature and depths of the surface changes, helping us to distinguish between, for example, damage and writing, and Handwritten Text Recognition (HTR), undertaken with the ‘human in the loop’, can offer possible readings for a team of experts to scrutinise.Footnote ⁸⁵ For British curses the scholarly community has for decades been primarily working from Tomlin’s own excellent drawings, which he recognises are inherently ‘interpretative’ (2024: 47) and ‘simplify what is visible’ (2024: 50). New imaging of defixiones and the availability of these on open access platforms would widen participation. One can appreciate that a static image as reproduced in Fig. 2, and found across Tomlin’s volume, cannot be usefully interrogated to the same degree as digital files of varying formats on a computer.

The value of open access collections leads us to another point. McKie presents in an appendix the curses cited in his book. Readers will find this useful, but the argument for hard-copy republishing of the texts cited by McKie, or the 309 texts in Urbanová’s Latin Curse Tablets of the Roman Empire (under half the total corpus, chosen because they fit specific parameters), may be weakening with easy-to-access online corpora.Footnote ⁸⁶ The risk of the proliferation of these printed sets of texts, shaped by the research focus of the authors and without apparatus criticus, is that the assiduous scholar has yet more corpora to wade through, and has to decide whether the minor unflagged changes are new readings or merely differences in presentation or typographical errors.Footnote ⁸⁷ Many of these texts are already digitally available, thanks to collections of curse tablets such as TheDefix (Thesaurus Defixionum), the inheritor of the earlier TheDeMa (Thesaurus Defixionum Magdeburgensis); Amina Kropp’s CD-ROM of 391 Latin tablets, known as DfX; the major aggregating Roman epigraphic online corpora; and, for Britain, Roman Inscriptions of Britain Online.Footnote ⁸⁸ In a world of established epigraphic ontologies, standardised vocabularies and Linked Open Data, these collections could be assembled with a few clicks and new data linked-in where necessary by the authors. With the digital resources at our disposal, there should be no need to spend time retyping or copying and pasting the basic metadata and texts. Rather than recollecting data authors should be able to focus on the essential quality control, modifications (e.g. changes to the readings), additions (e.g. new texts, translations, images), or classifying texts (e.g. by types of cursing formulae).

When it comes to the intensive sub-categorisation of curse typologies of the type proposed by Urbanová, we might feel sympathetic towards Sánchez Natalías’s exasperated declaration that ‘the debate in this case has gone too far, and the rigidity of these taxonomies is no longer helpful’ (2022: 60). While we might not, in the manner of the author of SD 119 which opens this section, want to drive a nail through the classificatory nomina, more thoughtfulness about the point of the proliferation, renaming or reconfiguration of categories and sub-categories would be helpful. This is particularly critical when the information is absorbed into digital datasets, since often the dividing lines become hardened and edge cases lose their ambiguity, intensive digital categorisation sometimes concealing more than it reveals. Indeed this is one strong argument for continuing to publish valuable collections such as SD and Tab. Uley in print, as well as digitally: we need the opportunity to browse without the distraction of search bars and filters, and to see the complexity and uncertainty of the interpretations without the sometimes concealed simplifications of a structured dataset.

These three books should be read together, so that their interconnections, similarities and differences can be explored. Each offers so much, but no one volume fully captures the vibrancy of ancient cursing practices and their modern receptions, and even taken together, much remains unresolved. The subtle divergences between the commentaries, for example on the use of experts, the furtive nature of the practice and the value of distributional data, require further interrogation. In this review I have highlighted some themes raised by the authors for which future research surely awaits, namely the nature and evolution of provincial literacy, local varieties of Latin and ‘indigenous’ languages, and bilingualism. Perhaps of even broader appeal may be aspects of cursing that lurk mostly in the shadows in these volumes, not least the beginnings, development and evolution of western provincial cursing, the intersection with existing local ritual practices and the interaction, or not, between the Graeco-Egyptian and Latin-language cursing. The foundations that these volumes have constructed mean that many will be inspired to build more of these fascinating worlds.