La cohérence généalogique dans la Coherence-Based Genealogical Method : évaluation critique

 = Paper = 

The proponents of the Coherence-Based Genealogical Method expect the algorithm to reveal genealogical relationships between witnesses (understood as texts) from genealogical relationships between variants. However, it is an illusion to automatically associate objectivity and algorithm (or software program). Our demonstration with group 2138 in the Catholic epistles is indisputable in this respect. To infer from the CBGM that both the group 2138 and the Byzantine type are descendants of a common ancestor and that the Byzantine type is closer to the common ancestor is a fundamentally erroneous assumption in terms of the history of the text. The numerous biases that characterise the method are, in fact, fatal flaws: a data bias, an omitted variable bias, a selection bias. Because the genealogical coherence considers only the predominant orientation between two witnesses, a considerable amount of information is thereby omitted. The limitation of witnesses to Greek manuscripts alone also contributes to a selection bias. The failure of the CBGM to include versions (and patristic quotations) thus distorts textual groups other than the Byzantine text. What is the impact of 28 these biases on the text of the Editio critica maior? The initial text A is identical to the NA28 in more than 97% of the cases, but when it differs, it is Byzantine for almost 70%. The close proximity to the Nestle-Aland text (and to Codex Vaticanus) arises from a circularity of the initial text, and the extent of the Byzantine text in the changes is caused by an algorithmic convergence of the CBGM to this text-type. Consequently, in our opinion it is imperative to discard the use of CBGM in the Editio critica maior, and to abandon it altogether insofar as its various biases are at the very heart of the method’s structure and are beyond correction.