# How are statistical comparisons between bible translations calculated?

(This arose from this question which was in turn triggered by this answer.)

Quoting Jack Douglas quoting Wayne Grudem:

In "The advantages of the ESV” by Wayne Grudem, he states that the 2001 edition of the ESV is "92% RSV, 8% modified, or 60,000 words" (from the RSV).

I'd like to know more about the formulae that are used to make such comparisons. Specifically:

• Is punctuation counted?
• How are differences in word order dealt with?
• Are meaningful variations weighed differently than less meaningful variants? 1
• Is there more than one way of making such calculations, and have people compared which are the most useful?

1. For instance, in comparing the NIV and RSV, a quick survey of Genesis 1 shows that the RSV tends to preserve the "and...and...and" from the Hebrew whereas the NIV often finds other ways of expressing sequence in English. This would be considered an insignificant variant in my mind. On the other hand, the notorious Isaiah 7:14 “young woman” vs “virgin” may represent a more significant variation between these two translations.

• This paper presents an overview of text similarity measures in general, as used in computer science. I have no idea whether Grudem's measure conforms to one of these (I somehow doubt it). Commented Apr 3, 2017 at 17:24
• Why is this question important? Wouldn't it be better to spend the effort in learning the languages required to read the manuscripts rather than to compare the translations? If comparing translations already, then how does the statistical difference in German and French translations compare with the statistical differences in English translations? Does it matter?
– user17080
Commented Sep 9, 2017 at 17:40

• Here is an online tool that compares two strings using "fuzzy" string comparisons.

• Here is a detailed description of the algorithm, published in 1985. The basic principle is described in the first sentence of the work.

The edit distance between strings a1 … am and b1 … bn is the minimum cost s of a sequence of editing steps (insertions, deletions, changes) that convert one string into the other.

I performed a comparison of the RSV and ESV 1.

### Results

• the texts are 93.89% similar, with punctuation.

• the texts are 94.2% similar, without punctuation.

• punctuation accounts for (94.2 - 93.89)/93.89 = 0.0033 (less than 1/2 of 1 per cent).

I noticed that the RSV gave numbers using words, whereas the ESV gave them using digits. For example,"All the days of Noah were nine hundred and fifty years; and he died.", against "All the days of Noah were 950 years, and he died.". This, like punctuation, would be of relatively little significance.

### Update

I have created a tool of my own so the similarity of texts can be investigated using various options. Differences are displayed using green for inserts, and red for deletions -- for those, like myself, who like to know. Selecting all the options returns 96.09% similarity between the RSV and the ESV 1.

Notes:

1. The first 9 chapters of Genesis (the most whole chapters of the ESV allowable from biblegateway)

I use the Scriptrue4all.org Interlinear Scripture Analyzer (ISA2)*. Literal Hebrew translation doesn't include "surely" or "freely".

KJV - Gen. 2:16 - And the LORD(3068) God(430) commanded(6680), (5921) the man,(120) saying,(559) Of every(4480), (3605) tree(6086) of the garden(1588) thou mayest freely eat:(398), (398)

ISA - Gen. 2:16 - And~he-is-(m)instructing (u~itzu 6680) Yahweh (ieue 3068) Elohim (aleim 430) on (ol 5921) (-) the~human (e~adm 120) to~to-say-of (l~amr 559) from~any-of (m~kl 3605) tree-of (otz 6086) (-) the~garden (e~gn 1588) to-eat (akl 398) you-shall-eat (thakl 398).

KJV - Gen. 2:17 - But of the tree(4480)(6086), of the knowledge(1847) of good(2896) and evil,(7451) thou shalt not(3808) eat(398) of(4480) it: for(3588) in the day(3117) that thou eatest(398) thereof(4480) thou shalt surely die.(4191), (4191)

ISA - Gen. 2:17 - And~from~tree-of (u~m~otz 6086) the~knowledge-of (e~doth 1847) good (tub 2896) and~evil (u~ro 7451) not (la 3808) you-shall-eat (thakl 398) from~him (mm~nu 4480) that (ki 3588) in~day-of (b~ium 3117) to-eat-of~you (akl~k 398) from~him (mm~nu 4480) to-die (muth 4191) you-shall-die (thmuth 4191).

• Although ISA3 Basic is the first version seen on the website of Scripture4all.org, it won't be fully functional until ISA3 Pro comes out. ISA2 can still be downloaded from the ISA3 Basic download page. ISA2 also has a concordance that shows when different words are used for the same word form. And incredible program, to me. "Errors in ISA2 will be corrected in ISA3 Pro."

Comparing English Bible versions has limited value, to me. I find it far more instructive to look at commentaries that emphasize Hebrew meaning (ie Pulpit, Keil-Delitzsch, etc.).

I begin with ISA2, then go to scholarsgateway.com (gives me parsing and root for each Heb. word). Then to pealim.com - that dictionary allows me to plug in the root and find all the forms. The generosity of scholars who share their work is simply amazing.

For anyone interested in comparing English versions and the methods used, the following links may be useful.

1.This link includes the OP quote at the top of p. 3 (and explains other aspects).

"The ESV was first published in 2001. It changed about 8% of the RSV, or about 60,000 words. The remaining 92% is the RSV, much of which is simply “the best of the best” of the KJV tradition."

-- To me, this appears as a simple count of added/deleted words.

1. JOURNAL ARTICLE - Biblical Texts and Statistical Analysis: Zechariah and Beyond

http://www.jstor.org/stable/3260311?seq=1#page_scan_tab_contents

1. "The statistical trade-off between word order and word structure – Large-scale evidence for the principle of least effort" - Alexander Koplenig , Peter Meyer, Sascha Wolfer, Carolin Müller-Spitzer -- Published: March 10, 2017https://doi.org/10.1371/journal.pone.0173614

http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0173614

--- This one has calculations and many References at the end. Plus, specifically addresses 6 books of the NT further down.

1. This forum has a post (10th down - Joe Fleener) that says Bible Works 6.0 has a 'text comparison' program. Whether it gives stats or not, I don't know.

https://www.bibleworks.com/forums/archive/index.php/t-713.html

1. This link gives "Translation Methods" for several versions (and the translators involved).