7 April 2022

In the last blog entries we listed all quotations based on Isaiah. In this entry we attempt to draw some conclusions on the way how the authors of the New Testament quote the Old Testament.

We found 7 literal matches (class 1), 15 almost literal matches (class 2), 4 matches where the introduction is preceded by the quoted text (fully or partially, class 3), 5 matches where a passage is skipped from the quoted text (class 4), 4 matches where some texts were added in the quotation (class 5), and 16 more complicated matches that were classified as “class 4” (10:22, 58:6 and 61:1-2; 2 matches), “class 5” (40:3) or a mixture of “class 4” and “class 5” (29:10, 6:9-10 in two passages, 8:14 and 28:16, 28:11, 28:16, 42:1-4, 52:5, 52:7, 53:12, 54:13, 55:3, 56:7; 13 matches). If we count the first three classes as (quasi-)literal, we get 26 such matches. The other matches (6 of “class 4”, 5 of “class 5” and 14 of “class 4/5”) are in 25 quotations. Altogether we have an overview of 51 quotations, based (fully or partially) on Isaiah.

Now let us have a look on the quoted texts. In the following table all of the 56 quoted texts are listed:

It is remarkable that the getrefs algorithm finds almost all matches (92%, 52 of 56). An effective fuzzy search could probably find Isaiah 29:10 in Romans 11:8, or a more fitting similarity algorithm may report a shorter distance between Isaiah 28:11 in I Corinthians 14:21 – but the other two quotations are too short to be found even with better algorithms. (The quotations that cannot be found mechanically are colored with a yellow row in the table.)

We need to admit that the getrefs algorithm sometimes finds false positives, but this is quite rare. For some prominent examples we found Isaiah 53:12 in Luke 22:37 and Isaiah 54:13 in John 6:45. It remains a research project to learn how often a false positive is found – for the whole Bible.

Another important fact is that sometimes the Greek translation of the Old Testament does not contain all the information mentioned in the New Testament. This seems to be the case when referring to Isaiah 42:1-4 in Matthew 12:18-21. My personal opinion is that this reflects that the Jews are set aside only temporarily by God (cf. Romans 11). Even if most of the LXX texts appear surprisingly accurately in the SBLGNT, there is still a need to study the Hebrew texts as well to get a more detailed picture and understanding on God's will. This comment supports Perry's opinion on the “authorship” of quotations: “The doctrine implies that the Spirit would quote, allude to or echo its own writing… our answer is that the Spirit does not quote the Septuagint; it quotes its own writing which was in Hebrew and Aramaic. It did this in Greek and multilingual individuals were involved.” (See Andrew Perry: Did the NT writers “quote” the Septuagint? Christadelphian EJournal of Biblical Interpretation, 7(2):59–78, 2013.) On the other hand, my opinion is similar to Gleason's criticism on the accuracy of the best available Hebrew texts: The Greek text (that is, LXX) contains specific pieces of information on Christ that are no longer available in the Hebrew manuscripts. Either by mistake or on purpose. Anyway, the Greek texts are unavoidable to study to get the whole picture.

The previous blog entries may have shown some evidence and confirmed Pfeiffer's statement who claimed in 1948 already that about 80% of the New Testament quotations are based on the Septuagint (Robert Henry Pfeiffer: Introduction to the Old Testament, Harper, New York, 1948). My impression after summarizing the Romans, the Psalms and Isaiah, is that this number is even closer to 100%. There are, however, a couple of quotations that are difficult or impossible to match with any texts from the known manuscripts (for example, Matthew 5:33 seems to be like that – it is said to be a quotation of Psalm 50:14, but the Greek texts substantially differ; a similar question can be raised for Isaiah 64:4 and I Corinthians 2:9 with a similar answer given). But I think that better text matching strategies could automatically detect almost all quotations, and, some fine methods could even filter out false positives.

An important remark is that the quotations of Isaiah 6:9-10 in Mark 4:12, Luke 8:10, John 12:40 and Acts 28:26-27 are somewhat different, and Luke uses two variations in his Gospel and in the Acts: one is very accurate (on 267/269 characters, with a difference of 2%), and the other is quite inaccurate (on 40/41 characters, with a difference of 50%). We also noticed that the inaccurate text has some extra context and the importance of a literal match may be overridden by other aspects. This warns us researchers: “He has made us competent as ministers of a new covenant—not of the letter but of the Spirit; for the letter kills, but the Spirit gives life” (II Corinthians 3:6). The final message of God's word goes far beyond textual searches and algorithms, it is much more than a just mechanical comparison.

Anyway, the matches we found in the previous entries can be considered of an evidence that the quotations in the New Testament are not just random ideas. They reflect a very precise work of several authors. The 56 passages of Isaiah are cited by Matthew, Mark, Luke, John, Paul, the author of Hebrews, and Peter, 7 authors of the New Testament; in 12 books (Matthew, Mark, Luke, John, Acts, Romans, I and II Corinthians, Galatians, Ephesians, Hebrews and I Peter). This shows not just the importance of Isaiah in the 1st century Christian literature, but the accuracy of the quoted texts is also remarkable. Accurate citation, or just a sign of the attempt to be accurate, shows that God's word played a very important role among the Christians. Jesus himself emphasizes that even the smallest portion of the Law may be relevant: “For truly I tell you, until heaven and earth disappear, not the smallest letter, not the least stroke of a pen, will by any means disappear from the Law until everything is accomplished” (Matthew 5:18).

May God give wisdom to study His word as accurately as possible, but look behind the letters and follow the Spirit's guidance at the same time! Now I close this series of blog entries until I can find substantial improvements on the algorithms. I would like to focus on the following topics as next steps:

1. Finding an effective and fast fuzzy match algorithm. This plan is quite promising, because theory of data mining offers quite good strategies if 2-shingles and the Jaccard distance is used. You may, however, ask the question: Why the usual indices in the well-known Bible programs are not adequate enough? My answer: They are based on words, and the original texts do not contain spaces or punctuation.
2. I think there is, however, a way to automatically detect words in a long text that has no punctuation and spaces. I mean here a text that is written in an unknown language (for the machine). Clearly, several shorter parts of the text have repetitions – a clever way to detect these patterns seem to be able to identify most words automatically. This idea may help in inventing better algorithms that are unbiased in regard to punctuation and spaces. After such an unbiased auto-punctuation an unbiased auto-indexing could be performed and used in all algorithms.

Final remarks

• I mentioned that the online database of the Wuppertal project was unavailable in mid of January. Now it can be accessed again. I wrote that the Wuppertal database contains more entries than mine, but I would like to revise my statement because the methods we use differ in some sense. For example, the Wuppertal database was created on a per verse basis: each verse is considered a different entry – in my approach the verses have no significance because the manuscripts are not split into verses at all. Also, the Wuppertal database uses a per word indexing – in my approach there are no words assumed because they are not present in the manuscripts either. I wanted to be unbiased in my own research by not assuming any information that is not present in the manuscripts.
• I wrote that it was impossible to access other digitalized versions of the Septuagint and the Greek New Testament than LXX and SBLGNT when I started this project. This is no longer the case. I recently learned about the wonderful CNTR project that aims at providing the best possible version of the Greek New Testament, based on computer science, statistics and mathematical logic. The beta version of their database is available here. Hopefully there will be a Sword module soon ready to use inside a web page as well.
• The latest version of the database I maintain is available online in the repository of the bibref project, in the folder docs/common. A Makefile is provided. A full run requires several command-line Linux utilities to be installed (see also the .yaml file that describes GitHub actions to be performed on a git push). Here I show just one view on the current state of project: How many books of the Greek Old Testament are quoted in the Greek New Testament? The following table shows the number of passages cited from the Old Testament:
  
  

  LXX book	Number of quoted texts
  Genesis	18
  Exodus	14
  Leviticus	5
  Numbers	1
  Deuteronomy	18
  II Samuel	1
  I Kings	2
  Psalms	42
  Proverbs	3
  Ecclesiastes	1
  Isaiah	42
  Jeremiah	3
  Ezekiel	1
  Daniel	2
  Hosea	6
  Joel	1
  Amos	2
  Micah	1
  Habakkuk	2
  Haggai	1
  Zechariah	3
  Malachi	1
  Total	170

  
  Altogether, these are 22 books (out of 39 canonized books, according to most protestant churches). For the reference, here is which New Testament books refer to these texts:
  
  

  SBLGNT book	Number of quotations
  Matthew	44
  Mark	22
  Luke	20
  John	12
  Acts	25
  Romans	51
  I Corinthians	10
  II Corinthians	8
  Galatians	10
  Ephesians	4
  I Timothy	1
  II Timothy	1
  Hebrews	28
  James	4
  I Peter	7
  Total	247

  
  Altogether, these are 15 books (out of 27 canonized books, according to most protestant churches). I think these numbers clearly confirm the very close relationship between the texts of LXX and SBLGNT, or, in more general, the inner consistency of the Bible.