STONE-MASTERS workshop identification model

One of major aims our project had to tackle was the identification of workshops and individual stonecutters in a situation where we have no workshop/artisan’s signatures recorded on their works. The central question was whether it was possible to establish if a group of inscriptions come from the same atelier or were carved by the same “hand” judging only from their features and visual appearance. After a period of detailed research on the cemetery of Zoara/Zoora (modern Ghawr as-Safi in Jordna) a site particularly rich in dated formulaic epitaphs, we assume that such a study is possible, though with obvious limitations and different levels of certainty or “risk assessment” needed for such identifications.

Below, we present convenient sets of features which need to be examined in order to suggest that a group of inscriptions carved on stone (a separate model for mosaics will follow) share the same styles, and possibly are products of the same artisan or workshop. We developed this model for published evidence and for researchers who have limited contact with the sources, mainly through good or middle quality photographs. A wholesomely different procedure should be adopted by epigraphers working on archaeological sites or in museum collections as they have unique opportunity to approach the sources with greater details. In a manner proposed by Stephen Tracy and later discussed, for example by Patrcia Butz, through a close examination of the letter groves, traces of stonecutting tools, sampling pigments used to decorate the inscriptions, preparing photogrammetric documentation, and others, they can harvest much more information on the origin of collections which they close study. However, such research is not always possible: some stones are lost, others accessible only upon acquiring permission, or accessing and close studying them necessitates the use of significant amounts of resources (funds, time, people) which are not always available at hand. In other cases, large numbers of inscriptions need to be scrutinized in a limited amount of time, so no examination going too much into detail can be afforded. This is where the simplified model proves its value. Perhaps, further progress with the development of AI algorithms for the examination of stylistical features in photographs of inscriptions such as those put forward by Charlotte Tupman (Exeter University), and Stephen Tracy (American School of Classical Studies at Athens and the Ohio State University) will allow the epigraphers to rely on machine learning for these aims.

Our state-of-the-art model mixes our original considerations with ideas postulated by former researchers, and offers a controlled vocabulary. The results can be subjective and questionable but the idea has been around for decades, and the treat, if successful, would be truly worth the inherent risk. Epigraphers have been dreaming of identifying workshops of origin of inscriptions for quite a while. In the 1970s, Nancy Gauthier brilliantly pointed to the existence of at least four different workshops at Trier (RICG I, 27–36), while Mark Handley (2003, 27–32; 2001, 195–196) was able to transpose and further develop her methods for other sites in Gaul and Hispania (Trier, Vienne, Arles, Tarragona, Mérida). Other researchers made similar inquiries and proposed their own criteria, for example for the study of inscriptions from medieval monasteries of Northumbria (Higgit 1995), Akmoneia in Phrygia (Chiricat 2013, 211), and Kadoi (Lochman 1991). In Egypt, Tudor 2011 and van der Vliet 2020 were successful in bining particular epigraphical styles with monasteries rather than workshops. These researchers were analysing mainly already published inscriptions, subject to an optical and textual study, emphasizing: the analysis of the lettering, punctuation, abbreviations, ligatures, and peculiarities of detailed decorations over the textual features (where the existence of regionally used textbooks was, however, often assumed). The direct result was distinguishing of a specific styles, and, subsequently its attribution to a specific workshop. So far an in-depth discussion on the workshop/style attributions was lacking but rather it was assumed that a style (sometimes termed “house style”) can be equalized with one workshop. A more careful approach may, however, suggest that one workshop changed its style over the course of several decades, that several styles were offered by one workshop simultaneously, or that we are dealing with a different workshop for each style. They were also using their own, not standardized, vocabularies.

A special research strand was also the study of textual formulae and their usefulness in distinguishing particular workshops. Giancarlo Susini (1973, 46; cf. Handley 2003, 34) was advocating the theory that the ordinary customers of late antique workshops were choosing from a very limited number of formulae which were typically offered by the artisans, and that they either took their advice on the most suitable formula, or even just supplied the stonecutter with the name, and date of death of the deceased, and then a designated person from the workshop chose a relevant formulaic model from a textbook. Reconstructing such model textbooks used by workshops would be a challenging task but surely possible, as it was shown already by Edmond Le Blant (1890). He used extant inscriptions to point to a lost textbook from Briord in the territory of Vienne (pp. 70–71). This he did by rejecting general, omnipresent formulae, and focusing on repeatable phrases occurring only locally while Louis Robert (1955, 211) observed in several inscriptions from Delphi that incomplete texts (FD III/2 215 and 216) clearly implied through their layout that they would have had to be incised from a pre-existing model, since the stonecutter omitted large portions of text, but still knew where exactly to place specific words. As if he had a layout of the inscription before his eyes.

Earlier works, however, showed no systematic approach to identifying workshops and stonecutters. The research results were based on the experience of individual scholars or tackled very specific, geographically limited collections. Our model and the atlas, which accompanies it, aims to guide every research interested in pursuing such research for the Greek and Latin inscriptions from the entire Mediterranean. At present, the model is based on the evidence from Zoara/Zoora, as stated above, but it will be gradually expanded with features and respective vocabulary for other sites and regions as the project develops and catches the wave. At present, the model allows its user to establish similarity levels through the following criteria of the visual appearance and textual formulae of inscriptions. The criteria are divided into five parts. Part One: Features of the Support: 1. Shape of the stela; 2. Material of the support; 3. Epigraphic field; 4. Ordinatio; 5. The lettering of the ordinatio; 6. Whether the stone was reused or not; 7. The presence of a signature. Part Two: Features of the Decoration: 8. The shape of the decoration of the frame; 9. Decoration of the entire inscribed face; 10. Colour. Part Three: Textual Formulae: 11. Languages used in the text; 12. Opening formula; 13. Funerary Syntax (N: name, P: patronymic, O: occupation/profession); 14. Verb relevant to the actions described in the inscription; 15. Complementary descriptions of these actions; 16. Closing formula. Part Four: Script Features: 17. Primary writing features; 18. Secondary writing features; 19. Sections highlighted; 20. Script features; 21. Quality of script; 22. Presence of apices/serifs. Part Five: Palaeography (it is described in detail on a separate page of this website).

The identification model (which can also be built through attached xlsx sheets by other teams and projects) and the atlas are based on three levels – first enough records on the evidence, are needed. Then inscriptions sharing specific features within domains should be described as sharing the same styles. We assume that the same style can be shared by different workshops, can be a regional or global phenomenon. Furthermore, one workshop can consciously or not use more then one style at the same time. It can offer goods of lower quality for more affordable prices and higher priced but more refined ones, with better levels of execution. The proficiency in replicating an intended style may also play part. The complex production chain which existed in antiquity, with some work done by the apprentices, some by masters, the existence of forgeries, imitations, influence of some groups of artisans on others, makes it very problematic to distinguish between them. At least a partial answer to this complexity may be the search for characteristic errors (as in the studies of manuscript transmission where they are more telling of the manuscript’s origin and its belonging to a specific stemma than other features), or, more generally, spotting idiosyncratic elements in very tiny details patterned on the methods of study of Athenian pottery and vase paintings by John Beazley, patterned on the original ideas of “connoisseurship” of Italian renaissance painters Giovanni Morelli back in the nineteenth century, a method which is now pursued and developed, for example by Maria Villano on the ERC GRAPH-EAST project. A person willing to practically apply our model, should download the xlsx file posted at the beginning of this page. The model can be applied only to sites with good yield of inscriptions. Preferably, sites where comparable evidence of dated inscriptions allow one to be certain that some epigraphical production comes from the same year or from subsequent years, ideally within one decade. This chronological window gives some plausibility to the supposition that we are dealing with objects which could be made by the same people, influenced by the same traditions, trends or fashion in implementation. Hopefully, with the progression of the project, we will be able to establish how long particular trends remain in a given community and how long workshops pursue the similar styles. These were impacted by natural development of approaches to decoration patterns, textual formulae, but also generational changes, and flow of employees.
Within such sites, chronologically ordered inscriptions should be inserted in the first column of the form. Specific features should be then described in other columns using our vocabulary (see also below for the vocabulary breakdown). Note that this is a multiple-choice feature descriptions. Presence of one feature in a specific category, does not exclude coexistence with other features. In a filled-in form, one can then search and highlight datasets from approximately the same decades which reveal similar feature of the visual appearance of inscriptions and textual formulae. Whenever inscriptions share a number of features within the same criterium of similarity, we can say that they share the same “style” from the perspective of this criterium. When “style” similarity within different criteria accumulates in a specific dataset, we can carefully suppose that we might have works of the same workshop or artisan before our eyes. The decisive factor, should be, however, characteristic errors of idiosyncratic features, mentioned above, which highlight the dataset against the general stylistical tendency of the region and decade.

Below follows, the controlled vocabulary of the STONE-MASTERS’ workshop identification model. We divided it into several parts:

Part One: Features of the Support (1. Shape of the stela; 2. Material of the support; 3. Epigraphic field; 4. Ordinatio; 5. Ordinatio lettering; 6. Reuse; 7. Signature)

Part Two: Features of the Decoration (8. Decoration Frame; 9. Decoration (entire face); 10. Colour)

Part Three: Textual Formulae (11. Languages; 12. Opening formula; 13. Funerary Syntax; 14. Relevant verb; 15. Action complements; 16. Closing formula)

Part Four: Script Features (17. Primary writing features; 18. Secondary writing features; 19. Sections highlighted; 20. Script features; 21. Quality of script; 22. Apices/Serifs)

Part Five: Palaeography (Described on a separate page: Greek, Latin)