The results in summary, not including those texts not tested or too short to say anything meaningful.
102 of 119 (# matches to Origen) / 85.7% | Certainly Origen |
40 of 52 | Contra Celsum |
4 of 4 | De principiis |
16 of 18 | Commentarii in evangelium Joannis (extant books 1-32) |
18 of 19 | Commentarium in evangelium Matthaei (books 12-17) |
3 of 3 | Exhortatio ad martyrium |
6 of 6 | De oratione |
15 of 17 | Philocalia |
6 of 7 (# matches to Origen) / 85.7% | Probably Origen |
2 of 3 | Homiliae in Lucam |
1 of 1 | Epistula ad Africanum |
1 of 1 | De engastrimytho (Homilia in i Reg. [i. Sam.] 28.3-25) |
1 of 1 | Homiliae in Exodum |
1 of 1 | Selecta in Exodum, Fragmenta ex commentariis in Exodum |
1 of 4 (# matches to Origen) / 25% [cumulative 83.8%] | Possibly Origen |
1 of 1 | Homiliae in Leviticum |
0 of 1 | Dialogus cum Heraclide |
0 of 1 | In Jesu Nave homiliae xxvi (fragmenta e catenis) |
0 of 1 | Libri x in Canticum canticorum (fragmenta) |
8 of 15 (# matches to Origen) / 53.3% [cumulative 80.7%] | Possibly Origen, Mixed Authorship, or Author Unknown |
6 of 11 | In Jeremiam (homilies 1-20) |
2 of 4 | Fragmenta in Lucam (in catenis) |
4 of 30 (# matches to Origen) / 13.3% | Likely Not by Origen, Author Unknown |
2 of 20 | Fragmenta in Psalmos 1-150 [Dub.] |
1 of 4 | Commentarii in epistulam ad Romanos (I.1-XII.21) (in catenis) |
1 of 4 | Fragmenta ex commentariis in epistulam ad Ephesios (in catenis) |
0 of 2 | Scholia in Apocalypsem (scholia, 1, 3-39) |
8 of 10 (# matches to Gregory) / 80% | Not by Origen, Possibly by Gregory Nyssenus (or Someone Else) |
8 of 10 | Fragmenta ex commentariis in epistulam i ad Corinthios (in catenis) |
10 of 11 (# matches to Clement) / 90.9% | Not by Origen, Possibly by Clement of Alexandria (or Someone Else) |
2 of 2 | Fragmenta in Jeremiam (in catenis) |
5 of 5 | Expositio in Proverbia (fragmenta e catenis) |
3 of 4 | Fragmenta in Lamentationes (in catenis) |
The basis for deciding whether a text was "possibly Origen" or "likely not by Origen" is mentioned in this post (using Fisher's exact test).
This doesn't seem to contradict scholarship on Origen or on the patristics severely, in general terms, which is good. The ones that "should" have been identified as certainly Origen's, according to scholarship, have been identified as certainly Origen's. The ones that "should" be considered "[Dub.]" are considered dubious. While there may be some expansion of the "dubious" category, the texts claimed are not among those asserted to be part of the 'core canon' of Origen's texts; they frequently are attested only through the catenae, which have something of a reputation for occasional misattribution.
On the other hand, there are some issues with the possible rewriting of Origen (especially in excerpted fragments), differences arising from certain selection biases when a continuous text of the author is not available (again, for excerpted fragments), or possible differences due to genre (in the homilies for example) or the difficulty of finding where the quotes end and begin in an efficient manner (especially scriptural quotations), so it's possible (possible) that a few more texts end up in the "dubious" categories than possibly (possibly) should be.
The suggestions that Clement of Alexandria or Gregory Nyssenus could be responsible for some of this text are intriguing and deserve further investigation.
(the nitty gritty details...)
Using a list of 33 words, the study begins with a leave-one-out analysis of the Contra Celsum, using samples approximately 2500 words in length. The direct quotations are removed first. A total of 52 samples are made this way. 40 out of 52 point to the author of the rest of the Contra Celsum, instead of the author of the Praeparatio Evangelica of Eusebius, as the author (76.9% accuracy). When including 31 controls as well, 40 out of 52 still point to the work itself ('Origen') as the best match (76.9% accuracy).
Four samples were taken from the De principiis, between 3000 and 3500 words in length. The author of the Contra Celsum was treated as the 'Origen' candidate. Using the same list of 33 words, 'Origen' was selected as the best candidate out of 33, including controls, 4 out of 4 times (100% accuracy).
Ten samples were taken from the Commentarii in evangelium Joannis (books 1, 2, 4, 5, 6, 10, 13), 3500 to 4000 words long. The direct quotes were removed first. The author of the Contra Celsum was treated as the 'Origen' candidate. Using the same list of 33 words, 'Origen' was selected over Eusebius 10 out of 10 times (100% accuracy). When including the 31 controls, 'Origen' was selected as the best candidate 8 out of 10 times (80% accuracy).
Three samples were taken from the Exhortatio ad martyrium, using samples approximately 2500 words in length. The direct quotes were removed first. The author of the Contra Celsum was treated as the 'Origen' candidate. Using the same list of 33 words, 'Origen' was selected as the best candidate out of 33, including controls, 3 out of 3 times (100% accuracy).
Six samples were taken from the De oratione, using samples approximately 3000 words in length. The direct quotes were removed first. The author of the Contra Celsum was treated as the 'Origen' candidate. Using the same list of 33 words, 'Origen' was selected as the best candidate out of 33, including controls, 6 out of 6 times (100% accuracy).
Five samples are taken of the homilies known as 'In Jeremiam' 1-11. The samples are taken along homily divisions: 1st Homily (3355 words), 2nd-4th Homily (2726 words), 5th Homily (4047 words), 6th-8th Homily (3866 words), and 9th-11th Homily (3922 words). Direct quotes were removed first. The author of the Contra Celsum was treated as the 'Origen' candidate. For all five, Origen rated more highly than Eusebius. The second sample (2nd-4th Homily), the fourth sample (6th-8th Homily), and the fifth sample (9th-11th Homily), however, were rated more highly for one of the controls. One possible explanation of this data is that the two longer homilies were genuinely Origen's (the 1st and the 5th), while the shorter ones were added to the collection but were not Origen's.
The work called the 'De engastrimytho (Homilia in i Reg. [i. Sam.] 28.3-25)' measures 3574 words in length. Using the same 33 words used in every other test with Origen so far, and using the Contra Celsum as the sample representing the author 'Origen', Origen is identified as the best candidate author for the sample, out of 33 candidates, including the controls.
The text known as the 'Homiliae in Lucam' divides into three samples of approximately 3000 words each. Quotes are removed first. The author of the Contra Celsum was treated as the 'Origen' candidate. When compared against Eusebius, 'Origen' is picked as the most likely candidate 3 out of 3 times (100% accuracy). When compared against 33 candidates, including the controls, Origen is selected as the most likely candidate 2 out of 3 times (67% accuracy).
The 'Dialogus cum Heraclide' measures 3613 words long, after the direct quotes are removed. The same list of 33 words was used. The author of the Contra Celsum was treated as the 'Origen' candidate. Between Eusebius and Origen, Origen is selected as the better candidate. However, among 33 candidates, Origen was edged out slightly by one of the controls. This might just be a false negative. Or, because the text itself records the substance of a dialogue between Origen and Heraclides, referring to "Origen" in the third person, the text may have been written by one of Origen's students, someone else who attended the affair, or some other author.
The compilation known as the 'Philocalia' was broken into 17 samples, each approximately 3500 words in length. The same list of 33 words was used. The author of the 'Contra Celsum' was considered the 'Origen' author candidate. Between Origen and Eusebius, Origen was selected as the better author candidate 16 of 17 times (94% accuracy). When 31 controls are added as candidates, Origen was selected as the best candidate, out of 33 total candidates, 15 out of 17 times (88% accuracy). The ultimate author of the Philocalia, said to be compiled by Basil of Caesarea and Gregory of Nazianzus as excerpts from the Alexandrian church father's works, clearly seems to be Origen.
Six samples are taken of the homilies known as 'In Jeremiam' 12-20 (continuing the numbered homilies on Jeremiah). The samples are taken along homily division: 12th-13th Homily (4195 words), 14th Homily (3791 words), 15th-16th Homily (3612 words), 17th-18th Homily (4283 words), 19th Homily (2187 words), and 20th Homily (4516 words). Direct quotes were removed first. The same list of 33 words was used. The author of the Contra Celsum was treated as the 'Origen' candidate. Between Eusebius and Origen, Origen is picked over Eusebius 6 out of 6 times. Origen is picked over all the controls for the 12th-13th Homily, the 17th-18th Homily, the 19th Homily, and the 20th Homily. A control is selected over Origen for the other two samples (14th Homily Homily, 15th-16th Homily).
In Jeremiam | Control Group (Results in Authentic Works) | |
Positives | 6 | 75 |
Negatives | 5 | 17 |
Chi-square value: 4.2567
degrees of freedom: 1
one-tailed p-value (chi-squared test): 0.0195 (this is <0.05)
one-tailed p-value (Fisher's exact test): 0.054 (this is >0.05)
It is plausible to assume that all twenty of the homilies 'In Jeremiam' came from Origen, but it doesn't seem very likely.
Nineteen samples were taken from the 'Commentarium in evangelium Matthaei' (books 12-17), each approximately 4000 words in length. The direct quotations were removed first. The same list of 33 words was used. The author of the Contra Celsum was treated as the 'Origen' candidate. Between Eusebius and Origen, Origen is selected as the better candidate 19 out of 19 times (100% accuracy). When the 31 controls are included as well, Origen is selected as the best candidate 18 out of 19 times (94.7% accuracy). Origen clearly seems to be the author.
After quotations are removed, the extant letter of Origen to Africanus ('Epistula ad Africanum') is 3146 words long. The same list of 33 words is used. 'Origen' is represented by the text of the Contra Celsum. Origen is selected as the best candidate out of 33 candidates, including the controls.
Eight samples were taken from the 'Commentarium in evangelium Joannis' (books 19, 20, 28, 32): one comprising the 19th book (4090 words long), three from the 20th book (lengths of 3243, 3207, and 3233 words), two from the 28th book (2732 and 2665 words), and two from the 32nd book (3861 and 3930 words). The direct quotations were removed first. The same list of 33 words was used. The author of the Contra Celsum was treated as the 'Origen' candidate. Origen was considered the best candidate for authorship, out of 33 candidates including the controls, 8 out of 8 times (100% accuracy).
There are extant some 'Fragmenta in evangelium Joannis (in catenis)'. Once quotations are removed, it is 1610 words long. It may be too short in length to expect a reliable result. It is also an amalgam of several different authors making citations. And there may be some errors in discerning where the direct quotations actually are.
It does not seem that a reliable result can be obtained for the 'Fragmenta in librum primum Regnorum (in catenis)', at 1865 words, either.
Clement of Alexandria has been moved from 'control' to 'author candidate' (control #31 takes the place that Clement had as control #25, which becomes author candidate #3), and Cyril of Alexandria has been added as an author candidate (#4). Clement of Alexandria is represented by his 'trilogy' (Protrepticus, Paedagogus, and Stromata--without quotations). Cyril of Alexandria is represented by 'Against Julian' and his 'Commentary on John' (without quotations).
The controls now are: #1 Mark, #2 John, #3 1 Corinthians, #4 Hebrews, #5 Revelation, #6 Letter of Aristeas, #7 1 Maccabees, #8 2 Maccabees, #9 Polybius, #10 Diodorus Siculus, #11 Dionysius Halicarnassus, #12 Strabo, #13 Plutarch, #14 Arrian, #15 Herodian, #16 Herodotus, #17 Thucydides, #18 Xenophon, #19 Epictetus, #20 Galen, #21 Lucian, #22 Philostratus, #23 Josephus, #24 Acts, #25 Gregory Nyssenus, #26 Justin Martyr, #27 Theophilus of Antioch, #28 Athanasius, #29 John Chrysostom, and #30 Basil.
The 'Fragmenta in Lucam (in catenis)' is a tough nut to crack. If divided into two samples of 6000 words each, it goes to Cyril of Alexandria (once) and then Clement of Alexandria (once). If divided into three samples of 4000 words each, it goes to Cyril of Alexandira (once), Origen (once), and Clement of Alexandria (once). If divided into four samples of 3000 words each, it goes to Cyril of Alexandria (once), Origen (twice), and Clement of Alexandria (once). If divided into six samples of 2000 words each, it goes to Cyril of Alexandria (twice), a control (once), Origen (twice), and Clement of Alexandria (once). A bit mind boggling, right? Makes you want to understand it, right? To tease through and figure out what it means? Well, good luck. Even starting from 12,000 words, there may sometimes be no very definite results, despite how tantalizing it is to 'make a guess' for our pattern-seeking minds (yes, I am guilty of this also). If I had to make a guess, none of these three men were the author (or maybe Origen was after all), but we can't be sure of that either.
The 'In Jesu Nave homiliae xxvi (fragmenta e catenis)' measures 3138 words, after quotations are removed. The same list of 33 words was used. The best candidate selected for authorship is Clement of Alexandria, and the second-best candidate is Origen. With only one sample from which to get a reading, these results are inconclusive. The text may have been written by Clement of Alexandria, by Origen, or by a third party.
The 'Libri x in Canticum canticorum (fragmenta)' measures 2688 words, before quotations are removed, and 1829 words, after quotations are removed. The same list of 33 words was used. Clement of Alexandria is selected as the best candidate before quotations are removed and as the second-best candidate (just behind one of the controls) after quotations are removed. Because only one sample can be tested, it is not very conclusive. The text may have been written by Clement of Alexandria, by Origen, or by a third party.
The 'Fragmenta ex commentariis in epistulam ad Ephesios (in catenis)' divides into eight samples, about 2000 words (or into four samples, about 4000 words long). The direct quotations were removed first. The same list of 33 words is used. With eight samples, Origen is closest 2 times, Clement of Alexandria is closest 2 times, Cyril of Alexandria is closest 2 times, and a control (Gregory Nyssenus) is closest 2 times. (With four samples, Cyril is closest 2 times, Origen is closest 1 time, and Clement is closest 1 time.) With results like these, either the text is of mixed authorship or, if it is not of mixed authorship, it is unlikely that any of the candidates is the actual author.
The 'Commentarii in epistulam ad Romanos (I.1-XII.21) (in catenis)' breaks into four samples of approximately 2500 words in length. When using four candidate authors and thirty controls, the closest match appears to be Origen (once), Clement of Alexandria (once), and Gregory Nyssenus (twice). Although this makes it less likely that Origen is the actual author of the entire text, other than that, no very sure conclusions can be reached on this basis.
Two samples are made of the 'Scholia in Apocalypsem (scholia 1, 3-39)' that are approximately 2500 words in length. They match best with two different controls. This indicates against the hypothesis that Origen wrote it all, but no further conclusion can be reached.
Twenty samples are made from 'Fragmenta in Psalmos 1-150', which are marked by the TLG itself as '[Dub.]', with each sample being approximately 4000 words in length. Direct quotations, where marked by the editor, are removed first. The same list of 33 words is used. The same 4 author candidates and 30 control candidates are used. Among all 34 candidates, Origen was selected as the closest 2 of 20 times, Clement was selected as the closest 4 of 20 times, Cyril was selected as the closest 5 of 20 times, control #25 was selected as the closest 3 of 20 times, and control #26 was selected as the closest 6 of 20 times. If the text is not a composite work, this is sufficient to reject Origen (and all the candidates) as being unlikely to be the actual author of the text.
The work known as the 'Homiliae in Exodum' measures 5070 words long, after quotations are removed. The same list of 33 words is used. There are four author candidates and 30 control candidates. Among all 34, Origen is selected as the closest match by a considerable margin.
The work known as the 'Homiliae in Leviticum' measures 1319 words long, after quotations are removed. The same list of 33 words is used. There are four author candidates and 30 control candidates. Among all 34, Origen is selected as the closest match by a slight margin. This does not, however, provide a secure conclusion, due to the short extent of the text and the existence of other, potential near-matches.
The works titled the 'Fragmenta ex commentariis in Exodum (= In illud: Induravit dominus cor Pharaonis)' and the 'Selecta in Exodum (fragmenta e catenis)' are combined into one text of 5070 words, after quotations are removed. The same list of 33 words is used, as well as the same 34 candidates, including 30 controls. Origen is selected as the closest match by a considerable margin.
The rest of these works ascribed to Origen in the TLG are considered too difficult to work with (or just too short) for the time being.
Fragmenta de principiis
Fragmenta alia de principiis
Fragmentum in Lamentationes (in catenis)
Fragmentum in librum primum Regnorum (in catenis)
Fragmenta in Lucam (in catenis)
Homiliae in Genesim (in catenis)
Homiliae in Ezechielem
Commentariorum series in evangelium Matthaei (Mt 22.32-27.63)
Fragmenta ex commentariis in evangelium Matthaei
Fragmenta in evangelium Matthaei
Epistula ad Gregorium Thaumaturgum (e Philocalia)
Commentarii in epistulam ad Romanos (I.1-XII.21) (in catenis)
Commentarii in epistulam ad Romanos (e cod. Vindob. gr. 166)
Commentarii in Romanos (III.5-V.7) (P. Cair. 88748 cod. Vat gr. 762)
Commentarii in Romanos (cod. Athon. Laura 184 B64)
Epistula ad ignotum (Fabianum Romanum)
Epistula quibusdam qui ei obtrectrabant (ad Alexandrum Hierosolymitanum)
Scholia in Apocalypsem (scholia 28-38)
De resurrectione libri ii (fragmenta)
Commentarii in Genesim (Fragmenta)
Selecta in Genesim (fragmenta e catenis)
In Ruth (fragmentum)
Homiliae in Job (fragmenta in catenis, typus I) (e codd. Paris.)
In Canticum canticorum (libri duo quos scripsit in adulescentia)
Fragmenta ex commentariis in Ezechielem
Selecta in Ezechielem (fragmenta e catenis)
Fragmentum ex commentariis in Osee
Fragmentum ex homilis in Acta apostolorum
In epistulam ad Hebraeos homiliae
Adnotationes in Genesim
Adnotationes in Exodum
Adnotationes in Leviticum (fragmenta e catenis)
Adnotationes in Numeros
Adnotationes in Deteronomium (fragmenta e catenis)
Adnotationes in Jesu filium Nave (fragmenta e catenis)
Adnotationes in Judices
Homiliae in Job (fragmenta in catenis, typus II)
Excerpta in Psalmos [Dub.]
Scholia in Canticum canticorum
Scholia in Matthaeum
Scholia in Lucam (fragmenta e cod. Venet. 28)
Homilae in Job (fragmenta in catenis, typus I II) (e codd. Vat.)
Part of this is just fatigue. I might be able to do something with some of this later.