This post describes a now-published paper presented at the Pacific Northwest Regional Meeting of the Society of Biblical Literature, Portland, OR, U.S.A., in March, 2015. This paper presented a proof-of-concept of a statistical method to partition readings in Codex Bezae (= Bezae) by layer.
Historically and developmentally, there seem to be two main approaches to Bezae, one that considers its distinctive text the earliest attainable form (despite some corruption) and the other that considers it largely secondary (though with some initial readings perhaps). Naturally, the former approach tends to focus on the coherent aspects of Bezae’s unique text, while the latter focuses on indications of Bezae’s composite character.
The two approaches are not entirely at odds. Using Bezae’s text of Matthew, Michael W. Holmes highlights several coherent threads in readings Bezae shares with the Old Latin tradition.  But Holmes is able to isolate these threads only after first assigning Bezae’s readings to distinct “layers,” which are differentiated by the respective combinations of witnesses that tend to support readings in that given layer.
That Bezae’s readings are attested by various consistent combinations of witnesses will be apparent even if we regard its text as a coherent literary piece. It will also be observed the readings Bezae shares with particular witnesses have distinctive characteristics. For example, those shared with the versions tend to be larger and more exegetically significant than those shared with Greek witnesses. These phenomena can be interpreted as layering regardless of our historical model, though they will differ in the sequence of the layers.
In the gospels, Bezae’s layer shared with the Old Latin Gospels is the most recognizable. These are readings attested exclusively (or almost exclusively) by Bezae and Old Latin witnesses. Of course, Bezae shares readings also with the Byzantine and Alexandrian traditions, majuscules such as Θ, and minuscule families, such as Family 1 and Family 13. We should also recognize that Bezae’s textual character varies between gospels and, presumably, within the same gospel.
Bezae’s composite text introduces complications if we wish to profile it against other manuscripts with more homogeneous texts. Because of its variegated elements, Bezae will appear more divergent when its readings are evaluated together. But if the readings are first partitioned by layer, it seems the individual layers will not all diverge equally.
The idea behind layer extraction is to partition Bezae’s readings into more homogeneous components. The paper compares three methods of layer extraction using IGNTP transcripts from John 4:1–42 and finds that they largely corroborate one another.
 Michael W. Holmes, “Codex Bezae as a Recension of the Gospels” in Codex Bezae: Studies from the Lunel Colloquium, June, 1994 (ed. D. C. Parker and C.-B. Amphoux; Leiden: Brill, 1996) 123–160.