Begin with the object, not a desired reading
Record material, dimensions, shape, surface, technique, layout, damage, later reuse and present location before interpreting the text. A monumental decree, potsherd graffito, coin legend, seal and copper-plate charter were produced for different purposes and audiences.
Provenance and provenience are evidence
Document ownership and movement after discovery, and the precise archaeological location if excavated. Layer, feature, association and recovery date can constrain chronology and function. An unprovenanced object loses contextual evidence and carries a higher risk of modern alteration or forgery.
Preserve the discovery state
Images, field notes, coordinates, conservation records and custody history should be created before cleaning or moving the object. Cleaning can reveal strokes but can also remove pigment, deposits or tool traces. Conservation decisions belong in the publication history.
Use complementary recording methods
Raking light, calibrated photography, reflectance transformation imaging, photogrammetry, microscopy and three-dimensional scanning reveal different surface properties. Squeezes and estampages remain useful in suitable conditions. Processing settings and original files should be retained so enhancements do not become invisible evidence.
Separate marks from characters
A groove may be a deliberate stroke, damage, stone grain, tool preparation or later graffiti. Stroke order, cross-section, patina and relation to the surface help evaluation. Character segmentation is itself an interpretation, especially where signs touch, ligatures occur or the writing is fragmentary.
Script and language are different variables
A script is a graphic system; a language is represented through it. Brahmi-derived scripts wrote several languages, while one language can appear in different scripts. Identifying familiar letterforms does not automatically identify the language, date, author or community.
Establish direction, lines and layout
Reading order depends on line arrangement, writing direction, object orientation, columns, borders and relation to images. Portable objects may rotate; seals can be cut in mirror image; inscriptions may continue across faces. A diplomatic layout preserves these features before normalized presentation.
Transcription should expose editorial work
A diplomatic transcription reports visible forms and line breaks. An edited text may expand abbreviations, regularize spelling, join words and restore losses. Leiden-style conventions and EpiDoc encode the difference between surviving text, uncertain readings, supplied letters, deletions and editorial explanation.
Uncertainty belongs at character level
Editors should distinguish a damaged but probable letter, several possible readings, an unread trace and a wholly lost section. A dot or structured attribute can represent doubt more honestly than a fluent reconstruction. Confidence can vary within a single word and should remain machine-readable.
Restoration is a constrained hypothesis
Supplied text may be supported by surviving traces, formula, grammar, line length and parallels. Formulaic expectations can also circularly create the text an editor expects. Publish alternative restorations and the reasons for preferring one, especially when historical conclusions depend on missing letters.
Palaeography usually gives relative ranges
Letterforms change across time, region, medium, scribal practice and register. Archaizing monumental forms and conservative hands weaken simple evolutionary sequences. Palaeographic dating should use securely dated comparanda and express ranges rather than false precision.
Dating should combine independent clocks
Explicit regnal or calendrical dates can be powerful only after era and conversion are established. Archaeological strata, associated radiocarbon dates, coin sequences, historical titles, linguistic features and palaeography offer separate constraints. The inscribed object and the archaeological layer may not be contemporary if reused.
Translation is an argument
Translation requires morphology, syntax, vocabulary, genre and cultural context. Word division and grammatical analysis should be published alongside the result. Retain uncertain names and technical terms rather than forcing modern equivalents, and distinguish literal translation from explanatory paraphrase.
Names support prosopography carefully
Personal names, kin relations, titles, guilds and places can build networks when spellings, dates and locations converge. Identical or similar names do not prove identical people. Prosopographical links should carry confidence levels and alternatives, particularly across long spans or common titles.
Royal language is performative and selective
Eulogies, victories, genealogies and land grants present authority in conventional forms. They can preserve dates, offices and places while exaggerating reach or omitting failure. The fact that a ruler commissioned a claim is secure evidence of representation; the claim’s factual content still needs testing.
Small inscriptions reveal different social worlds
Donations, labels, graffiti, potters’ marks, ownership texts, coins and seals may record people missing from monumental histories. Their brevity and mobility complicate interpretation. A marked sherd found at a site need not have been inscribed there, and a name does not establish a resident community.
Absence is rarely simple evidence
Survival depends on material, climate, reuse, excavation and publication. Literacy practices may favour perishable media that vanish. The absence of inscriptions in a sample does not prove absence of writing or administration unless preservation and recovery probabilities are addressed.
Corpora make comparison possible
A corpus should provide stable identifiers, object metadata, findspot, images, diplomatic and edited texts, apparatus, translation, bibliography and revision history. The ASI series, Tamil Nadu publications, SIDDHAM and DHARMA preserve different parts of this record. Coverage and editorial histories must remain visible.
Digital editions should remain scholarly editions
EpiDoc uses TEI XML to encode text, materiality and editorial interventions. Validation, controlled vocabularies, persistent identifiers, licences and versioning support reuse. Search interfaces must not flatten uncertain restorations into certain strings, and images should link to the specific reading they support.
A reproducible epigraphic workflow
Publish the object record and find context; archive unprocessed and processed images; identify script and language separately; provide diplomatic transcription, edited text and apparatus; explain dating; give literal translation and commentary; cite parallels; state alternatives; license data where possible; and log corrections.
Evidence assessment
Strong: Reliable epigraphy requires documentation of the inscribed object, its context and the editor’s interventions as well as a readable text. Strong: Script, language, chronology, provenance and historical identity are distinct questions with different evidence. Strong: Palaeography is most reliable when calibrated against securely dated comparanda and combined with independent chronological evidence. Probable: Open images, structured uncertainty and versioned digital editions materially improve reproducibility and correction. Contested: Many individual restorations and identifications, especially where readings rely on formula and few surviving traces. Unsupported: The claim that one name, word, script or language in an isolated inscription establishes the ethnicity or political identity of an entire population.