Glossing Priscian: models and open questions

Abstract

This talk discusses the DH aspects of the inclusion of glosses to Greek portions of the text in the digital critical edition of Priscian's Ars grammatica, under preparation within the ERC PAGES project (Grant n. 882588). It presents the data model, focussing on the relationship between the "main" (i. e. Priscian's) text in the individual manuscripts and the glosses, as we modelled it within the Cadmus software framework.

Our project is devoting particular attention to the Greek portions of the Ars (Priscians' Graeca): only for them, (a) we are encoding diplomatic transcriptions of the readings of all manuscripts, even those including only minor or merely orthographic variants; and (b) we are recording all glosses commenting on them.

In our data model, each reading of a Graecum and each gloss have two properties (among others): a reference to a location in the main text (identified through word offsets), and the manuscript ID.

Software can thus connect a specific gloss in a manuscript with the specific reading of that manuscript. It can also identify relations between glosses commenting on the same text location.

However, a number of issues arise from the data model described above. Let us shortly discuss them, in order of increasing importance.

Main (critical) text as mere anchor. In our model, we only record (a) the connection between a location in the main (critical) text and the manuscript readings and (b) that between the main text and the manuscript glosses, whereas the only real connection given in the textual tradition is that between a manuscript reading and a manuscript gloss. This may create paradoxycal cases such as that in Ars 13, 17, 4, in which the critical text will probably have εὐτυχής, but the archetype and all manuscripts have ἅγιος. All glosses clearly comment on ἅγιος, however in our model we are connecting all glosses to εὐτυχής. The paradox is only apparent, though, if we consider εὐτυχής as a mere anchor, i.e. a mere location in the main text, identified by a unique word offset (e.g. "the 4th word of the 2nd line of the textual item"). The role of εὐτυχής, in this context, is to identify a location in the text (line 2, word 4), regardless of its content, to which a number of related glosses in different manuscripts point.

Granularity. In our model, the formal connection gloss/MS variant, as well as the connection between glosses of different manuscripts commenting on the same Graecum, require that they all point to the same main text offset. However, if a Graecum is constituted of many words (e.g. ἐμαυτοῦ σαυτοῦ in Ars 12, 25, 2), it is often the case that different glosses specifically point to different portions of the main text. In our example, MSS TbNWMko have glosses on ἐμαυτοῦ; bWMko on σαυτοῦ, vNYIk on both words; please note that most MSS have one gloss on a specific word, and another one on the whole Graecum (with k having three glosses, one for each word plus one on both). Even word boundaries may be challenged: the reading of Z is εμαιτοι ϲ[.]υτοι (εμοιτοι ϲουτοι post corr.), and it interestingly has four clearly distinguished glosses commenting on four groups of letters identified as separated words: mei (on εμαι/εμοι), huius (on τοι), tui (on ϲου), huius (on τοι).

Textual variability. The main text varies among the manuscripts, so using the critical text as base for anchors is not trivial. Think, for example, on a case in which a manuscript has extra text not present in the critical text, and a gloss in that manuscript comments on it. For example, in Ars 13, 22, 3 Hertz's text is ut eadem sit et agens et patiens, potest significare ἑαυτοῦ sui, but manuscripts have ut eadem sit et agens et patiens, quae ιδιοπαθιαν potest significare ἑαυτοῦ sui. If the editor decides not to include the words quae ιδιοπαθιαν in the critical text, we will not be able to anchor the glosses on ιδιοπαθιαν to a location in the main text.