5 February 2022

You are out of your mind, Paul! Your great learning is driving you insane.
— Porcius Festus, told by Luke the Evangelist (Acts 26:24)

In a couple of former blog entries we listed Paul's quotations in the Romans. In this entry we attempt to give some conclusions on the way how Paul quotes the Old Testament.

It turned out that 20 passages contain only literal matches, with longest match on 96 characters (in Romans 4:7, quoting Psalm 32:1). The shortest match is difficult to count, because there are very short parts that match (even on 2 characters), so it is better to count the minimum of the longest literal matches in a quotation. By using this concept the shortest match appears in Romans 9:9, on 20 characters (quoting Genesis 18:10), but at the same time there is another quoted text in the same quotation (from Genesis 18:14).

This leads us to notice that a quotation may skip some words from the Old Testament. It seems that Romans 2:24 selects certain expressions from Isaiah 52:5 but some parts of it are not copied by Paul. Also, in the previous example, a long part of Genesis 18:10-14 in the middle (on 386 characters) is skipped from the quotation in Romans 9:9. A similar situation is present in Romans 10:19 (where 5 characters are not copied from Deuteronomy 32:21), Romans 13:9 (where 49 characters are skipped from Exodus 20:13) and Romans 15:12 (where 15 characters are not used from Isaiah 11:10).

Among the non-literal (or not fully literal) matches we find a similar concept. Romans 9:27 has literal copies of Isaiah 10:22, but the quoted text from Hosea 1:10 is non-literal. From Isaiah 10:22, however, some parts are not copied. Also, Romans 9:33 skips some characters from Isaiah 28:16. Similar structures can be identified in Romans 10:6, 10:15, 11:3, 11:4 and 11:8.

We also notice that Paul sometimes inserts his own words to explain the quoted text, that is, some extra text is added. Examples are 2:24 (on 7 characters), 9:9 (on 8 characters), 9:26 (on 54 characters), 10:19 (on 1+3 characters) – among the literal matches. Among the non-literal ones 3:10 (on 10 characters), 9:17 (on 26 characters), 9:27 (on 3 characters), 9:33 (on 8+18 characters), 10:5 (on 28 characters), 10:6 (on 20+27+68 characters), 10:15 (on 5+3+2 characters), 11:4 (on 6+7 characters), 11:8 (on 6+5+5+7 characters), 12:19 (on 4+3 characters) and 14:11 (on 16 characters). It is also possible that these additions match other manuscripts than the ones used in the digital version of LXX. But, in my opinion, here is a high number of additions that can be hardly explained just by pointing to unidentified manuscripts. So, in my opinion, Paul intentionally adds some texts to explain the quotes from the Old Testament, or to extend their meaning by his own thoughts. On the other hand, the following quotations do not contain substantial changes and seem to be quasi-identical to the quoted texts: the 20 literal quotations, plus Romans 1:17, 3:4, 4:3, 8:36, 9:13, 10:20, 10:21, 11:3, 11:9, 11:26, 11:34, 12:20, 15:9, 15:11 (14 quotations). After some dispute we may add some further quotations to this set, maybe 3:10 as well (since 10 characters between the quotes from Ecclesiastes 7:20 and Psalms 14:2 may come from Paul, or from a variant of the quoted texts), but even if we stop counting here we already have 34 quotations that are (quasi-)identical. I think this is a great number and confirms that Paul's study is a very well-edited text and not just a random collection of thoughts. In other words: without any doubt the Romans uses a scientifically clear intention to quote several passages from the Old Testament. Paul obviously behaves as a scientist and makes a huge effort to give explanations based on the Old Testament.

Our study in this sequence of blog entries focuses on mechanical identification of quotations. Paul's quotations seem to be, on one hand, quite literal, but not completely, so some hard work may be needed to find possibly all quotations, even if they are not always literal.

First of all, the Jaccard distance seems to be a possible way to find non-literal matches. A challenging question is to compute the expected value of the Jaccard distance of randomly chosen texts (see some nice attempts to answer this challenge via Google) and search for matching candidates whose distance is far enough from the expected value, by automation, quickly enough. There are promising algorithms in this direction, so we leave this question for a later time.

On the other hand, in some quotations the Jaccard distance is indeed a high value. For example, in Romans 11:3 and 11:4 quite high numbers of differences are obtained (38%, 42%, 42%, 80%). Either we give up to obtain such quotations mechanically, or we should find a better concept to safely detect matches like these.

We recall that the getrefs algorithm exploits the existence of unique passages in the Old Testament. In fact, almost all quotations contain at least one literal part that can be identified with the getrefs algorithm. The table below summarizes this remark.

First of all: note the incredible high number of the different Old Testament books used by Paul: they are (in the order of appearance) Habakkuk, Isaiah, Psalms, Ecclesiastes, Genesis, Malachi, Exodus, Hosea, Leviticus, Deuteronomy, Joel, I_Kings and Proverbs: 13 books! This is an evidence of how wonderful theological background Paul had. But also, it shows that these books were widely used in the ancient times and were accepted as a source of reliable knowledge.

It is clear that 45 of the 50 quoted texts are appearing at least as one getrefs chunk, that is, at least one unique part of almost each passage can be found in the Old Testament. This is, again, a promising fact, and gives us the hope that the getrefs command will be useful as a global tool to find most of the quotations mechanically (here we found 90% of them!). As already mentioned, some duplicated passages in the Old Testament can make things here a bit difficult:

• The Ten Commandments has a duplicate in Deuteronomy (see Romans 13:9).
• Romans 10:15 skips a couple of words, inserts a few other characters and also modifies the text a bit, so a literal search is ineffective. A part of the passage in Romans reads “ποδες των ευαγγελιζομενων τα αγαθα” – some key words of this can be highlighted and put into a Bible search program, for example, BibleTime (it is also based on Sword), in the form “ποδες ευαγγελιζομεν* αγαθα”. (The program bibref does not support such types of searches, at least for the moment.) This searching phrase will explicitly find the quoted text in Isaiah 52:7.
• Romans 11:8 clearly references Isaiah 29:10, but any literal chunks are not unique. On the other hand, the BibleTime search “πνευμα* κατανυξεως” on LXX gives the only match.
• Romans 12:19 refers to Deuteronomy 32:35, but the literal chunks are, again, not unique. Here the BibleTime search “εκδικησ* ανταποδωσω” is useful, but there are still two matches found: one of them (Hosea 4:9) is however a false positive, so it can be removed from the candidates. But for this final check some further steps may be required that are eventually not trivial.
• Finally, Romans 3:10 begins with a reference to Ecclesiastes 7:20 (“ουκ εστιν δικαιος”), but this is not a unique match, there is another passage in I_Samuel 2:2. To check this match, one needs to carefully read and understand the text, and learn that in I_Samuel 2:2 there is a somewhat different context.

After all, you don't necessarily have to study the Bible just mechanically, but rather with reason. God's ultimate plan is to communicate with us in a personal way, and therefore, understanding is finally unavoidable. “I am not insane, most excellent Festus,” Paul replied. “What I am saying is true and reasonable.” (Acts 26:25)

In the next entry we will continue how the getrefs algorithm can be applied from a different point of view: we will systematically search for quoted texts in the Book of Psalms.