Mark G Bilby: Marcion's Gospel and Data Science

[vocesanticae]

Just want to say thanks to everyone for the engagement with the material in my LODLIB and the corresponding Youtube interview. Suggestions/corrections for the book are much appreciated, as are recommendations on future topics of discussion and future conversation partners/panelists on Youtube.

A special note to those of you inclined to edit Wikipedia articles: almost none of my scholarship on Marcion's Gospel is cited or even linked there, despite there being many links to 100 year old public domain stuff and religious propaganda. It would be nice for my JOHD articles and datasets--not to mention the LODLIB itself--to be brought to the attention of a wider array of readers. Because of Wikipedia conflict of interest policies, I can't add self-references, so I'm reliant on others who are passionate about this area of research to make it more discoverable.

[vocesanticae]

A few rejoinders to the extended response of Irish1975:

The commonplace use of "Matthew, Mark, Luke, and John," if used only as titles for the canonized forms of these textual formations, would be fine if that's all they were. But scholars (myself included) repeatedly and habitually slip into using these names as references to discrete authors. If each of these texts were in fact evolving compositions with at least two major editors in two major generations, then this authorial mode of reference needs to be discarded in favor of more accurate references to discrete strata/compilers. Shorthand references are commonplace in all areas of scientific inquiry, so I don't see why proposing and using simple and easy to understand abbreviations should be dismissed as problematic.

The lists of passages that I've compiled in regard to class and gender bias should be taken seriously, and the broader scholarship cited in both should be consulted before dismissing these out of hand. The earliest traditions (Qn) really do consistently take the perspective of poor people (ptwxoi/"beggars") and slaves, couching Jesus as a new Aesop from beginning to end, whereas the later canonical overlay (Lk2) consistently takes the perspective of the wealthy and slave/house-owners. Qn also really does have a woman (likely Miriam) as the one who first anoints Joshua/Jesus, i.e., makes him the messiah, and a group of women as the first witnesses of the resurrection. Once the two-source hypothesis is taken seriously for Marcion's Gospel, it becomes evident pretty quickly that there has been a centuries-old scholarly tendency to dismiss many of the passages that are harshest toward the rich and that feature women as seriously deserving of inclusion in the Q gospel. This should not come as a surprise, given that the field is dominated by males of European descent (like myself).

The claim that I've named/branded Marcion's Gospel as "the Gospel of the Poor" reflects a complete misunderstanding of my book at the most basic level and from beginning to end.

[mlinssen]

gMark only puts the women on stage so he can blame them for nobody knowing about the resurrection that he just invented himself; it's evident that Luke defends their position (them being from the Chrestian tradition) and shifts the blame to the disciples, and that Matthew ends it all by teleporting IS directly into their combined midst and uttering a pathetic "Greetings"

The poverty in Luke is redaction only, Mark: Matthew keeps sending the same message to the *Ev audience: belly up, surrender, turn the other cheek, give in. Make yourself poor, give all you have, and so on.
What is being done here is little less than immobilising the opponent; not only is the *Ev audience (apparently it all was overwhelming, so I'll neutralise the labels) now sent the new Christian message, but basically it also gets told / instructed to submit

Your datasets are a treasure, especially the morphological tagging of them - that's just priceless. And the openness, traceability, completeness: unprecedented.
Yet I agree with Irish that riding a high horse reflects negatively on what you've achieved

And also, noblesse s'oblige

[Leucius Charinus]

Just want to say thanks to everyone for the engagement with the material in my LODLIB and the corresponding Youtube interview. Suggestions/corrections for the book are much appreciated, as are recommendations on future topics of discussion and future conversation partners/panelists on Youtube.

I enjoyed learning about your project on the vid and made some comments. I will repeat the comments below. I still seek perspective on any answers. My background (now retired) was in large scale database management and engineering. Here's my cv:
http://mountainman.com.au/cv_in_it.htm

I want to provide this cv information because my next questions may appear confronting and way out on the fringe (as usual). Before I state the questions here are my comments and some responses on YT.
https://www.youtube.com/watch?v=quRv7Xg83vQ

Interesting approach to using ancient texts as data. Example given of different Marcion reconstructions / translations / "datasets". Stylometric analysis bolts in to this approach. This is getting towards the use of AI. However the differentiation between primary and secondary historical sources as they are utilised by the historical method was not mentioned in the first 30 minutes. This is IMO essential. Of course depending on what historical propositions are to be explored, various texts which were primary sources in other studies become secondary sources in others. And vice verse. For example we do not have any primary evidence text for Marcion. All these reconstructions are based on secondary evidence. Accounts which are at least one step removed from Marcion and, if we can believe the back-story, accounts which are supposedly hostile to Marcion. Anyway great interview and interesting approach using data science. Thanks !!

Thank you for such an insightful response. Those are excellent points of constructive criticism. I'll plan to address them in a future installment of the podcast.

IMO, this is an important comment. As one of my professors of Biometry stated, if you are either consciously or unconsciously biased in the design of your experiment, the choice of variables or in the collection of your data, even the most elegant or sophisticated data analyses cannot remove these biases to produce a valid result. Dr. Bilby has produced an impressive set of analyses. Dr. Bilby is, however, working with limited and biased data sources so how and why data sources were chosen would be interesting.

Re AI, I hope that Dr. Bilby reconsiders using it. AI has been used to identify previously unknown admixtures of Neanderthal and Denisovan groups in ancient populations of Asia and Oceania that could not be found by conventional analyses. This is not using a chatbot to write college papers. He might find consultation with statisticians in fields such as genetics useful. Again, this is impressive work.

Background: About "The Gospel of Marcion"

1) There is no primary source for this gospel.

2) It is being constructed from a set of hostile secondary sources (Justin Martyr, Irenaeus, and Tertullian)

3) The earliest extant manuscripts of these secondary sources establishing the Gospel of Marcion are from the 14th, 11th and 9th century respectively.


MY QUESTIONS:

1) Should the above three facts be spelled out and the assumptions made explicit? With respect to the historical method the primary evidence is far more valuable than the secondary evidence. Does this not necessarily imply that anyone so reconstructing this gospel cannot hope to reconstruct the value of the primary evidence? Assumptions must be made.

2) The secondary sources are hostile and as a result it should become more difficult to be confident that the sources are treating Marcion with integrity. How is this being addressed if at all? Again assumptions must be made.

3) The copies of the secondary sources themselves are many centuries removed from the original sources in antiquity. Again assumptions have to be made with respect to the integrity of the historical transmission of the manuscripts.


SUMMARY

Much of scholarship studying Christian literature is based on rafts of assumptions. The study of Marcion's Gospel is no different. These assumptions are rarely if ever stated. This can be a problem. If all your assumptions are the correct ones then you don't have a problem. However if any of your assumptions are wrong then you'll likely end up with a problem of GIGO.


NHL

Are you going to turn your analysis on the NHL? The NHL is nothing like the secondary sources of the hostile heresiological church fathers obtained from 14th century manuscripts of the church community archives. OTOH the NHL is a time-capsule direct from the mid 4th century and if its not primary evidence then it is only one step removed. Far less chance of GIGO - the input data integrity is of a much higher standard. As a result you will have a much greater chance (IMHO) of discovering ancient historical truths.

[vocesanticae]

Just want to say thanks to everyone for the engagement with the material in my LODLIB and the corresponding Youtube interview. Suggestions/corrections for the book are much appreciated, as are recommendations on future topics of discussion and future conversation partners/panelists on Youtube.

I enjoyed learning about your project on the vid and made some comments. I will repeat the comments below. I still seek perspective on any answers. My background (now retired) was in large scale database management and engineering. Here's my cv:
http://mountainman.com.au/cv_in_it.htm

I want to provide this cv information because my next questions may appear confronting and way out on the fringe (as usual). Before I state the questions here are my comments and some responses on YT.
https://www.youtube.com/watch?v=quRv7Xg83vQ

Interesting approach to using ancient texts as data. Example given of different Marcion reconstructions / translations / "datasets". Stylometric analysis bolts in to this approach. This is getting towards the use of AI. However the differentiation between primary and secondary historical sources as they are utilised by the historical method was not mentioned in the first 30 minutes. This is IMO essential. Of course depending on what historical propositions are to be explored, various texts which were primary sources in other studies become secondary sources in others. And vice verse. For example we do not have any primary evidence text for Marcion. All these reconstructions are based on secondary evidence. Accounts which are at least one step removed from Marcion and, if we can believe the back-story, accounts which are supposedly hostile to Marcion. Anyway great interview and interesting approach using data science. Thanks !!

Thank you for such an insightful response. Those are excellent points of constructive criticism. I'll plan to address them in a future installment of the podcast.

IMO, this is an important comment. As one of my professors of Biometry stated, if you are either consciously or unconsciously biased in the design of your experiment, the choice of variables or in the collection of your data, even the most elegant or sophisticated data analyses cannot remove these biases to produce a valid result. Dr. Bilby has produced an impressive set of analyses. Dr. Bilby is, however, working with limited and biased data sources so how and why data sources were chosen would be interesting.

Re AI, I hope that Dr. Bilby reconsiders using it. AI has been used to identify previously unknown admixtures of Neanderthal and Denisovan groups in ancient populations of Asia and Oceania that could not be found by conventional analyses. This is not using a chatbot to write college papers. He might find consultation with statisticians in fields such as genetics useful. Again, this is impressive work.

Background: About "The Gospel of Marcion"

1) There is no primary source for this gospel.

2) It is being constructed from a set of hostile secondary sources (Justin Martyr, Irenaeus, and Tertullian)

3) The earliest extant manuscripts of these secondary sources establishing the Gospel of Marcion are from the 14th, 11th and 9th century respectively.


MY QUESTIONS:

1) Should the above three facts be spelled out and the assumptions made explicit? With respect to the historical method the primary evidence is far more valuable than the secondary evidence. Does this not necessarily imply that anyone so reconstructing this gospel cannot hope to reconstruct the value of the primary evidence? Assumptions must be made.

2) The secondary sources are hostile and as a result it should become more difficult to be confident that the sources are treating Marcion with integrity. How is this being addressed if at all? Again assumptions must be made.

3) The copies of the secondary sources themselves are many centuries removed from the original sources in antiquity. Again assumptions have to be made with respect to the integrity of the historical transmission of the manuscripts.


SUMMARY

Much of scholarship studying Christian literature is based on rafts of assumptions. The study of Marcion's Gospel is no different. These assumptions are rarely if ever stated. This can be a problem. If all your assumptions are the correct ones then you don't have a problem. However if any of your assumptions are wrong then you'll likely end up with a problem of GIGO.


NHL

Are you going to turn your analysis on the NHL? The NHL is nothing like the secondary sources of the hostile heresiological church fathers obtained from 14th century manuscripts of the church community archives. OTOH the NHL is a time-capsule direct from the mid 4th century and if its not primary evidence then it is only one step removed. Far less chance of GIGO - the input data integrity is of a much higher standard. As a result you will have a much greater chance (IMHO) of discovering ancient historical truths.

1. Disagree. Every manuscript of canonical Luke is, to a greater or lesser degree, a *primary source* for Marcion's Gospel, and not just Papyrus 69. This is true whether one maintains Marcion's Gospel as an earlier version of Luke or a later abridgement of Luke. If you write a book and it goes through three drafts, but only the second one goes to a mass-market publisher, much of the raw data (quantitatively speaking) of the first and/or the third draft is present in the 2nd draft.

2. Mostly hostile patristic commentators (over 15 of them, giving more than 700 clear textual references) are crucial data sources, to be consulted closely. Factoring in their polemical tone is definitely important, and all major recent editors of the Evangelion are doing that.

3. The fact that many texts, from those by Plato to Euripides to Tertullian to Epiphanius, are only preserved in manuscripts from the Middle Ages does raise concerns about data contamination and data degradation. That is why rigorous computational linguistics works on contemporaneous corpora, as well as the earliest receptions of those texts, are crucial. Iphigenia among the Taurians doesn't appear in any manuscripts until the 14th century Codex Laurentianus, but its reception is all over the place in antiquity, from Aristotle on down, as well as in artwork. Even if we can only be confident in 95% of the data of the Laurentianus text, that is still highly valuable. Same goes for the canonical gospels, for which we cannot attain 100% confidence or level of data integrity, even for the canonical forms, because of manuscript variations and scribal fluidity and tampering. For the canonical forms of the text, we may well be at around 97-99% fidelity to the canonical forms. I'm interested in using computational linguistics and data science to get us from 50-60% fidelity (which we have in Roth, Nicolotti, and Klinghardt) to 80-95% fidelity for Marcion's Evangelion (which we will have in BeDuhn's Greek edition and in mine).

[Secret Alias]

Why are you wasting time with this mental case?

[mlinssen]

Iphigenia among the Taurians doesn't appear in any manuscripts until the 14th century Codex Laurentianus, but its reception is all over the place in antiquity, from Aristotle on down, as well as in artwork. Even if we can only be confident in 95% of the data of the Laurentianus text, that is still highly valuable. Same goes for the canonical gospels, for which we cannot attain 100% confidence or level of data integrity, even for the canonical forms, because of manuscript variations and scribal fluidity and tampering. For the canonical forms of the text, we may well be at around 97-99% fidelity to the canonical forms. I'm interested in using computational linguistics and data science to get us from 50-60% fidelity (which we have in Roth, Nicolotti, and Klinghardt) to 80-95% fidelity for Marcion's Evangelion (which we will have in BeDuhn's Greek edition and in mine).

I presume that confidence and fidelity are related, yet how can the 95% of an extant text be matched via a reconstructed one?
That sounds like wishful thinking, and a bit of overconfidence

[davidmartin]

Qn also really does have a woman (likely Miriam) as the one who first anoints Joshua/Jesus, i.e., makes him the messiah, and a group of women as the first witnesses of the resurrection

ah that's what I like to hear, what evidence do you see for that?
I do not think the women were inserted but reflect a tradition of female elders in the early church who presided over a prior incarnation of the movement. The orthodox acknowledged this but was reluctant to draw from this anything more than 'he lives', and whatever else they said was doubted
(eg the scenario found in the gospel of Mary)

This paradigm is exactly represented in John by the woman at the well. She descends from the mountain preaching, in effect evangelising - only for the narrative to dismiss her 'we don't need to hear from you any more'. The acceptance of a woman's testimony, then the dismissal and replacement while still using that testimony only to support the replacement.

So what were these women teaching? It it lost? Of course not
The Odes of Solomon appear to contain ideas that would fit very well, featuring female leadership (I would argue the main narrators voice is female)
To add some spice, the Odes are Jewish-Christian, ie, freaking early

All very simple. No computer algorithms needed.
To combine all this together, based on the theology of the Odes, there was a prior gospel that did not feature the cross as a sacrifice but did feature a messiah figure and all the key ingredients

[vocesanticae]

Iphigenia among the Taurians doesn't appear in any manuscripts until the 14th century Codex Laurentianus, but its reception is all over the place in antiquity, from Aristotle on down, as well as in artwork. Even if we can only be confident in 95% of the data of the Laurentianus text, that is still highly valuable. Same goes for the canonical gospels, for which we cannot attain 100% confidence or level of data integrity, even for the canonical forms, because of manuscript variations and scribal fluidity and tampering. For the canonical forms of the text, we may well be at around 97-99% fidelity to the canonical forms. I'm interested in using computational linguistics and data science to get us from 50-60% fidelity (which we have in Roth, Nicolotti, and Klinghardt) to 80-95% fidelity for Marcion's Evangelion (which we will have in BeDuhn's Greek edition and in mine).

I presume that confidence and fidelity are related, yet how can the 95% of an extant text be matched via a reconstructed one?
That sounds like wishful thinking, and a bit of overconfidence

The UBS/NA texts are themselves reconstructions. Their texts do not exist 100% in any manuscript. All manuscripts involve dealing with data alteration and degradation, with potentially extra/superfluous and missing/lacunose data. Canonical text criticism can and does take seriously patristic attestations, both favorable and hostile. Sometimes the presumably hostile and/or heretical witnesses (e.g., Arius on John's Gospel) can be more significant and reliable than the positive and/or orthodox witnesses regarding the earliest retrievable textual data.

Conversely, speaking in terms of quantitative signal data, all of the Marcionite texts do exist in manuscripts, both in the thousands of manuscripts of their respective canonical versions and in dozens of manuscripts of patristic commentators referencing the Marcionite scriptures. Klinghardt has shown this extensively, and Vinzent's forthcoming Apostolos also shows relatively high levels of correspondence between patristic attestations to the Apostolos (mainly Tertullian) and the bilingual Greek/Latin manuscripts 010 and 012. The Old Latin and Syriac witnesses are often more reliable than the Greek manuscripts in terms of these correspondences, preserving older forms of the text.

The figures I threw out (97-99% data fidelity/integrity in the canonical forms, and a range of 50% to 90% for the Marcionite forms) are admittedly educated guesses, but also helpful heuristic devices to move the discourse and analysis in a scientifically grounded and viable direction. Given that Roth and Harnack have Evangelion texts that are around 4000 words, but Klinghardt has around 13000 words and Nicolotti 11500, and BeDuhn and I around 7000 words, I think we have a helpful picture of the lower bound and upper bound for a data science approach to restoring the text, as well as a reasonable via media estimate for its actual size, which should be a target for restoring the text to its maximal possible levels of fidelity and lowest possible levels of canonical contamination.

Vinzent's forthcoming text of the Apostolos, which is tremendously rigorous in quantitative lexicographical and word frequency analysis, and which I just finished translating and Jack Bull is now editing, has a little over 8000 words. It's about 30% of the size of the canonical Paul. To put it differently, the canonical redactor tripled the size of the earlier form. The difference between the earlier/shorter Marcionite Evangelion and the canonical form (19500 words) likely evidences roughly comparable increase (2.5 times).

[gryan]

By 10 and 12 do you mean
010 Augiensis
012 Boernerianus
??