Scripture Leaf 🍃
Research
Explore study
Read BSB

Scripture Leaf · Translation Study

Five translations. Two independent AI reviewers.

A source-aware comparison of BSB, ESV, NASB 2020, CSB 2020 and NIV 2011 across a frozen 160-passage corpus, reviewed independently by two AI systems. Translation dimensions are kept separate rather than collapsed into one accuracy score.

160passages
120challenge
40controls
5 × 2translations × AI reviewers
How to read this study

The challenge corpus deliberately concentrates difficult translation problems. Results describe this corpus; they are not whole-Bible prevalence estimates. AI Reviewer A and AI Reviewer B remain separate throughout. Agreement is a robustness signal; disagreement is reviewer sensitivity, not a penalty against a translation.

Study map

Five research questions, kept separate

RQ1

Translation profiles

What profile does each translation show across the frozen constructs?

RQ2

Translation choices

What kinds of semantic, grammatical, ambiguity and explicitness choices appear?

RQ3

Philosophy context

How do observed profiles sit alongside each publisher's own stated philosophy?

RQ4

Frozen strata

How do patterns look within eight prospectively frozen challenge categories?

RQ5

Source text or translation?

When translations materially differ, how does each reviewer classify the cause?

No overall translation rank was created. No composite accuracy score, weighted winner or clarity-fidelity blend was created.

In plain English

What this means for a Bible reader

Broadly close on source meaning

Across the five translations and two independent AI reviewers, semantic-preservation rates on the frozen semantic opportunities sit in a relatively narrow band of about 89–94%. The study does not support a simple faithful-versus-unfaithful split among these five versions.

Most visible differences are translation choices

The versions separate more in how they express meaning: grammatical restructuring, explicitness, ambiguity, lexical choices and English style. Those differences are real, but more literal or more explicit is not automatically better.

BSB is not an outlier in this corpus

BSB falls within the same broad semantic-preservation range as ESV, NASB, CSB and NIV and receives many high English-clarity ratings. The relationship view below places its translation-choice profile closer to CSB and NIV than to ESV or NASB under this post-hoc descriptive measure.

Reviewer disagreement means caution

The two independent AI reviewers often agree on broad semantic, morphosyntactic and lexical groupings, but differ more on exact clarity scores and some subtle cause judgements. The study treats that as uncertainty rather than deciding that one reviewer is right.

Scope matters. These conclusions describe a deliberately difficult 160-passage study corpus. They are not whole-Bible percentages and they do not identify a best translation.

Study corpus

The 160 passages studied

The frozen corpus contains 120 deliberately difficult challenge passages and 40 controls. Each reference below opens BibleHub at the first verse of the study passage; ranges continue from that starting point.

120 challenge passagesDifficult translation decisions, variants, syntax, ambiguity and related issues
  1. CH001Genesis 1:26
  2. CH002Genesis 1:27
  3. CH003Genesis 2:4
  4. CH004Genesis 4:1
  5. CH005Genesis 5:2
  6. CH006Genesis 37:3
  7. CH007Exodus 3:22
  8. CH008Exodus 4:21
  9. CH009Exodus 12:11
  10. CH010Exodus 21:22-25
  11. CH011Exodus 34:6
  12. CH012Leviticus 18:18
  13. CH013Numbers 23:19
  14. CH014Deuteronomy 7:9
  15. CH015Deuteronomy 15:12
  16. CH016Deuteronomy 17:5
  17. CH017Deuteronomy 32:15
  18. CH018Joshua 24:19
  19. CH019Judges 5:7
  20. CH020Judges 11:39
  21. CH021Ruth 3:9
  22. CH022Ruth 4:14
  23. CH0231 Samuel 10:27-11:1
  24. CH0241 Samuel 14:41
  25. CH0251 Samuel 17:4
  26. CH0261 Samuel 25:22
  27. CH0272 Samuel 15:7
  28. CH0282 Samuel 21:8
  29. CH0292 Kings 2:9
  30. CH0302 Kings 24:8
  31. CH031Job 19:26
  32. CH032Job 26:5
  33. CH033Job 42:6
  34. CH034Psalm 2:2
  35. CH035Psalm 2:12
  36. CH036Psalm 8:5
  37. CH037Psalm 23:5
  38. CH038Psalm 51:5
  39. CH039Psalm 68:11
  40. CH040Psalm 73:26
  41. CH041Proverbs 31:10
  42. CH042Ecclesiastes 3:11
  43. CH043Ecclesiastes 11:1
  44. CH044Isaiah 6:10
  45. CH045Isaiah 7:14
  46. CH046Isaiah 14:12
  47. CH047Isaiah 42:1
  48. CH048Isaiah 50:11
  49. CH049Isaiah 53:5
  50. CH050Isaiah 53:11
  51. CH051Jeremiah 31:20
  52. CH052Jeremiah 31:31
  53. CH053Ezekiel 36:26
  54. CH054Ezekiel 37:14
  55. CH055Hosea 11:1
  56. CH056Amos 4:6
  57. CH057Amos 9:12
  58. CH058Micah 6:8
  59. CH059Zechariah 3:2
  60. CH060Zechariah 12:10
  61. CH061Matthew 1:23
  62. CH062Matthew 5:17
  63. CH063Matthew 11:12
  64. CH064Matthew 22:32
  65. CH065Matthew 26:28
  66. CH066Matthew 27:9-10
  67. CH067Mark 9:29
  68. CH068Mark 11:22
  69. CH069Mark 16:9-20
  70. CH070Luke 7:47
  71. CH071Luke 14:26
  72. CH072Luke 16:8
  73. CH073Luke 20:17
  74. CH074John 1:5
  75. CH075John 3:5
  76. CH076John 3:8
  77. CH077John 5:3-4
  78. CH078John 12:38
  79. CH079Acts 2:17-21
  80. CH080Acts 7:42-43
  81. CH081Acts 8:37
  82. CH082Acts 15:16-18
  83. CH083Acts 16:34
  84. CH084Acts 22:16
  85. CH085Romans 5:12
  86. CH086Romans 6:17
  87. CH087Romans 8:15
  88. CH088Romans 8:29
  89. CH089Romans 11:26-27
  90. CH090Romans 16:1
  91. CH0911 Corinthians 7:36
  92. CH0921 Corinthians 9:9
  93. CH0931 Corinthians 11:5
  94. CH0941 Corinthians 15:29
  95. CH0951 Corinthians 15:54-55
  96. CH0962 Corinthians 3:18
  97. CH0972 Corinthians 5:21
  98. CH098Galatians 3:13
  99. CH099Galatians 3:16
  100. CH100Galatians 5:12
  101. CH101Colossians 1:15
  102. CH1021 Timothy 3:11
  103. CH103Hebrews 1:8
  104. CH104Hebrews 2:11
  105. CH105Hebrews 3:7-11
  106. CH106Hebrews 8:8-12
  107. CH107Hebrews 9:15
  108. CH108Hebrews 11:1
  109. CH1091 Peter 1:2
  110. CH1101 Peter 2:6
  111. CH1111 Peter 4:6
  112. CH1122 Peter 1:1
  113. CH1131 John 2:2
  114. CH1141 John 5:7-8
  115. CH115Revelation 1:7
  116. CH116Revelation 3:14
  117. CH117Revelation 12:17
  118. CH118Revelation 13:8
  119. CH119Revelation 13:18
  120. CH120Revelation 22:15
40 control passagesDeterministically selected comparison controls
  1. C001Deuteronomy 28:61
  2. C002Numbers 12:5
  3. C003Deuteronomy 27:2
  4. C004Leviticus 16:9
  5. C0051 Samuel 2:19
  6. C0062 Samuel 5:7
  7. C007Judges 9:19
  8. C008Ruth 2:16
  9. C009Proverbs 17:14
  10. C010Psalms 7:2
  11. C011Psalms 10:17
  12. C012Proverbs 10:7
  13. C013Jeremiah 48:20
  14. C014Ezekiel 17:8
  15. C015Isaiah 10:11
  16. C016Isaiah 46:6
  17. C017Micah 2:2
  18. C018Joel 1:2
  19. C019Hosea 5:13
  20. C020Habakkuk 1:9
  21. C021John 8:8
  22. C022John 15:16
  23. C023Luke 2:5
  24. C024Mark 15:12
  25. C025Acts 18:10
  26. C026Acts 2:34
  27. C027Acts 3:24
  28. C028Acts 28:23
  29. C0292 Corinthians 8:1
  30. C0301 Corinthians 7:13
  31. C031Ephesians 5:14
  32. C032Romans 14:13
  33. C0332 John 1:9
  34. C034Jude 1:10
  35. C0351 Peter 5:13
  36. C0361 Peter 4:13
  37. C037Revelation 1:19
  38. C038Revelation 21:14
  39. C039Revelation 11:9
  40. C040Revelation 14:1

Post-release descriptive extension

How similarly do the translations behave?

These matrices are derived only from the already-frozen study data. They were not part of the original measurement instrument, so they are presented as a transparent post-hoc descriptive view rather than a confirmatory finding. Similarity here means a similar distribution of recorded study outcomes; it does not mean identical wording, common editorial ancestry, or higher quality.

Translation-choice profile similarity

Semantic representation (SEM), morphosyntax, explicitness, ambiguity, intertextual visibility and lexical continuity; equal weight by construct.

VersionBSBESVNASBCSBNIV
BSB100%same translation87.4–87.7%reviewer range85.8–86.0%reviewer range91.2–92.2%reviewer range88.8–91.3%reviewer range
ESV87.4–87.7%reviewer range100%same translation86.9–89.3%reviewer range83.8–87.1%reviewer range79.2–81.1%reviewer range
NASB85.8–86.0%reviewer range86.9–89.3%reviewer range100%same translation82.5–90.0%reviewer range82.3–83.4%reviewer range
CSB91.2–92.2%reviewer range83.8–87.1%reviewer range82.5–90.0%reviewer range100%same translation87.6–87.9%reviewer range
NIV88.8–91.3%reviewer range79.2–81.1%reviewer range82.3–83.4%reviewer range87.6–87.9%reviewer range100%same translation

English-clarity profile similarity

Syntactic clarity, referential clarity and idiomatic naturalness; equal weight by clarity dimension.

VersionBSBESVNASBCSBNIV
BSB100%same translation76.5–87.3%reviewer range75.8–79.4%reviewer range85.2–90.0%reviewer range89.6–92.3%reviewer range
ESV76.5–87.3%reviewer range100%same translation90.0–93.8%reviewer range81.0–89.2%reviewer range71.9–82.7%reviewer range
NASB75.8–79.4%reviewer range90.0–93.8%reviewer range100%same translation81.0–85.8%reviewer range71.5–74.4%reviewer range
CSB85.2–90.0%reviewer range81.0–89.2%reviewer range81.0–85.8%reviewer range100%same translation83.8–84.6%reviewer range
NIV89.6–92.3%reviewer range71.9–82.7%reviewer range71.5–74.4%reviewer range83.8–84.6%reviewer range100%same translation
How to read a cell

Each off-diagonal cell shows the range between AI Reviewer A and AI Reviewer B rather than averaging them. For each reviewer separately, similarity is calculated as 1 minus total-variation distance between the two translations' normalised outcome distributions. The translation-choice matrix gives equal weight to six already-frozen constructs; the clarity matrix gives equal weight to the three already-frozen clarity dimensions. A higher percentage means the two profiles behaved more similarly on this corpus; it is not an accuracy score.

RQ1 + RQ2

Translation profiles

The cards stay in the frozen display order: BSB, ESV, NASB, CSB, NIV. Percentages are construct-specific descriptive rates, not an overall quality score.

BSB

BSB

Two independent reviewer-system profiles. Display order is fixed and is not a ranking.

AI Reviewer Aindependent AI review
Semantic preservation
88.9% n=208
Material loss/change
3.4%
Morphosyntax preserved / restructured
100.0%
Lexical units: same family
100.0% n=11

English clarity distributions1 lower · 5 higher · no mean calculated

Syntactic
1124394465100
Referential
11222329434574
Idiomatic
11212333484530
AI Reviewer Bindependent AI review
Semantic preservation
92.8% n=207
Material loss/change
2.4%
Morphosyntax preserved / restructured
100.0%
Lexical units: same family
100.0% n=11

English clarity distributions1 lower · 5 higher · no mean calculated

Syntactic
1129315453582
Referential
11227344453535
Idiomatic
1129340471539
RQ2 · Translation-choice inventory

Counts describe how the frozen instrument classified choices. More explicit, more literal or more repetitive is not automatically better.

AI Reviewer ASemantic representation: 208 determinateExplicitness: 160 items
AI Reviewer BSemantic representation: 207 determinateExplicitness: 160 items
ESV

ESV

Two independent reviewer-system profiles. Display order is fixed and is not a ranking.

AI Reviewer Aindependent AI review
Semantic preservation
93.8% n=208
Material loss/change
1.4%
Morphosyntax preserved / restructured
100.0%
Lexical units: same family
100.0% n=11

English clarity distributions1 lower · 5 higher · no mean calculated

Syntactic
10212318445585
Referential
10221340441558
Idiomatic
11218353471517
AI Reviewer Bindependent AI review
Semantic preservation
93.8% n=208
Material loss/change
1.4%
Morphosyntax preserved / restructured
100.0%
Lexical units: same family
100.0% n=11

English clarity distributions1 lower · 5 higher · no mean calculated

Syntactic
10219337459545
Referential
10247353443517
Idiomatic
1023036545659
RQ2 · Translation-choice inventory

Counts describe how the frozen instrument classified choices. More explicit, more literal or more repetitive is not automatically better.

AI Reviewer ASemantic representation: 208 determinateExplicitness: 160 items
AI Reviewer BSemantic representation: 208 determinateExplicitness: 160 items
NASB

NASB

Two independent reviewer-system profiles. Display order is fixed and is not a ranking.

AI Reviewer Aindependent AI review
Semantic preservation
88.9% n=208
Material loss/change
2.9%
Morphosyntax preserved / restructured
100.0%
Lexical units: same family
81.8% n=11

English clarity distributions1 lower · 5 higher · no mean calculated

Syntactic
1129321464565
Referential
11219341439560
Idiomatic
1322236846057
AI Reviewer Bindependent AI review
Semantic preservation
93.3% n=208
Material loss/change
1.4%
Morphosyntax preserved / restructured
100.0%
Lexical units: same family
81.8% n=11

English clarity distributions1 lower · 5 higher · no mean calculated

Syntactic
11221344455539
Referential
11241352454512
Idiomatic
1123736245654
RQ2 · Translation-choice inventory

Counts describe how the frozen instrument classified choices. More explicit, more literal or more repetitive is not automatically better.

AI Reviewer ASemantic representation: 208 determinateExplicitness: 160 items
AI Reviewer BSemantic representation: 208 determinateExplicitness: 160 items
CSB

CSB

Two independent reviewer-system profiles. Display order is fixed and is not a ranking.

AI Reviewer Aindependent AI review
Semantic preservation
91.8% n=208
Material loss/change
1.0%
Morphosyntax preserved / restructured
100.0%
Lexical units: same family
100.0% n=11

English clarity distributions1 lower · 5 higher · no mean calculated

Syntactic
12217318451572
Referential
12222329432575
Idiomatic
17215342466530
AI Reviewer Bindependent AI review
Semantic preservation
91.3% n=208
Material loss/change
1.9%
Morphosyntax preserved / restructured
100.0%
Lexical units: same family
100.0% n=11

English clarity distributions1 lower · 5 higher · no mean calculated

Syntactic
19226329435561
Referential
12239339455525
Idiomatic
19214344462531
RQ2 · Translation-choice inventory

Counts describe how the frozen instrument classified choices. More explicit, more literal or more repetitive is not automatically better.

AI Reviewer ASemantic representation: 208 determinateExplicitness: 160 items
AI Reviewer BSemantic representation: 208 determinateExplicitness: 160 items
NIV

NIV

Two independent reviewer-system profiles. Display order is fixed and is not a ranking.

AI Reviewer Aindependent AI review
Semantic preservation
91.3% n=208
Material loss/change
1.4%
Morphosyntax preserved / restructured
100.0%
Lexical units: same family
90.9% n=11

English clarity distributions1 lower · 5 higher · no mean calculated

Syntactic
11283114305110
Referential
11220321442576
Idiomatic
11211323471554
AI Reviewer Bindependent AI review
Semantic preservation
93.2% n=207
Material loss/change
2.9%
Morphosyntax preserved / restructured
92.9%
Lexical units: same family
81.8% n=11

English clarity distributions1 lower · 5 higher · no mean calculated

Syntactic
12210323439586
Referential
11224342454539
Idiomatic
1226331465556
RQ2 · Translation-choice inventory

Counts describe how the frozen instrument classified choices. More explicit, more literal or more repetitive is not automatically better.

AI Reviewer ASemantic representation: 208 determinateExplicitness: 160 items
AI Reviewer BSemantic representation: 207 determinateExplicitness: 160 items

English clarity

Clarity is measured separately

Why distributions rather than a single number?

The study scores syntactic clarity, referential clarity and idiomatic naturalness from 1–5. The public package deliberately does not calculate a clarity mean or combine clarity with source-relative fidelity. The distributions above let you see, for example, how BSB performs in English without turning readability into an accuracy bonus.

RQ3

Translation philosophy, from first-party sources

Philosophy was added only after the result layer was frozen. It is descriptive context, not a quality proxy and not a causal explanation for the measured outcomes.

BSB3rd Printing

modern English study translation

The Berean Bible describes the BSB tier as a modern-English study translation and places it within a transparent multi-tier process linked to the original languages. Its official material says the BSB aims to maintain core meanings while producing high-literary-quality English, and the wider tier system is intended to preserve access to both word-for-word and thought-for-thought renderings where languages do not map one-to-one.

Translation Tiers ↗Translation Process ↗Berean Bible homepage ↗Free Downloads ↗
NASBNASB 2020

formal equivalence

The Lockman Foundation describes the NASB as a formal-equivalence translation that seeks word-for-word correspondence from the original languages as far as possible while recognising the need for readable, understandable English. For NASB 2020 it states that the longstanding standard remains while language, vocabulary, grammar and sentence structure were modernised to improve clarity and readability.

New American Standard Bible — Accuracy, History, Updates ↗About NASB 2020 ↗NASB 2020 official store collection ↗
CSBCSB Text Edition 2020

optimal equivalence

Official CSB material describes optimal equivalence as pursuing linguistic precision to the original languages together with contemporary-English readability. It states that word-for-word rendering is used where it communicates clearly, while more dynamic rendering may be used where a word-for-word construction would obscure meaning.

CSB Pew Bible introduction / translation philosophy ↗CSB Text Revisions 2020 ↗
NIVNIV 2011

balanced approach

Biblica describes the NIV approach as balancing transparency to the original with clarity of meaning in contemporary English. Its stated goal is accurate communication in clear, natural English rather than mechanically preserving source-language form or moving so far from form that source nuances are unnecessarily lost.

NIV Bible Translation Philosophy ↗NIV Bible Translation Process ↗

RQ4

Eight frozen challenge strata

Each challenge stratum contains 15 passages. The 40 controls remain a separate control cohort, not a ninth stratum.

Theological Vocabulary15 challenge passages
Gender Number Referent15 challenge passages
Difficult Translation15 challenge passages
Hebrew Idiom Poetry15 challenge passages
Lexical Ambiguity15 challenge passages
Textual Variant15 challenge passages
Ot In Nt15 challenge passages
Greek Syntax15 challenge passages

RQ5

Textual basis versus translation or interpretation

AI Reviewer Aprimary route only
76route units
59no material difference
17material differences
Textual basis
5 · 29.4%
Translation / interpretive
10 · 58.8%
Mixed
2 · 11.8%
AI Reviewer Bprimary route only
67route units
60no material difference
7material differences
Textual basis
6 · 85.7%
Translation / interpretive
1 · 14.3%
Mixed
0 · 0.0%
The reviewer systems diverge here.

AI Reviewer A classified 17 primary material differences and AI Reviewer B 7; their cause distributions also differ substantially. The study therefore does not average them or publish a consensus claim about which cause predominates.

Reviewer sensitivity

Where the two systems agree, and where they do not

90.2%Semantic representation agreement938 / 1040
98.6%Morphosyntax derived-group agreement69 / 70
98.2%Lexical outcome agreement54 / 55
85.0%RQ5 cause agreement102 / 120

The two AI reviewer estimates remain separate. Differences between reviewer systems are sensitivity evidence and are not averaged, adjudicated or converted into translation penalties.

Evidence

Ten worked cases

These are deterministic examples from predeclared construct cells, not a list of the “worst” passages or examples selected to favour one translation. Restricted translation wording is not reproduced; each case links to BibleHub for comparison.

V2CASE-01

Deuteronomy 28:61

SEM:Functional Equivalent
Translation
BSB
Reviewer
AI Reviewer B
Construct
SEM
Recorded outcome
Functional Equivalent

Represent the semantic content carried by the frozen source-language units for the primary span Deuteronomy 28:61; functional English restructuring is permitted and formal correspondence is not required.

Source evidence
  • Frozen source span · Deuteronomy 28:61 · OSHB
Compare translations on BibleHub ↗
V2CASE-02

2 Samuel 5:7

SEM:Narrowed
Translation
CSB
Reviewer
AI Reviewer A
Construct
SEM
Recorded outcome
Narrowed

Represent the semantic content carried by the frozen source-language units for the primary span 2 Samuel 5:7; functional English restructuring is permitted and formal correspondence is not required.

Source evidence
  • Frozen source span · 2 Samuel 5:7 · OSHB
Compare translations on BibleHub ↗
V2CASE-03

1 Samuel 2:19

SEM:Broadened
Translation
BSB
Reviewer
AI Reviewer A
Construct
SEM
Recorded outcome
Broadened

Represent the semantic content carried by the frozen source-language units for the primary span 1 Samuel 2:19; functional English restructuring is permitted and formal correspondence is not required.

Source evidence
  • Frozen source span · 1 Samuel 2:19 · OSHB
Compare translations on BibleHub ↗
V2CASE-04

Ruth 2:16

SEM:Omitted Or Reassigned Or Materially Changed
Translation
BSB
Reviewer
AI Reviewer A
Construct
SEM
Recorded outcome
Reassigned Or Materially Changed

Represent the semantic content carried by the frozen source-language units for the primary span Ruth 2:16; functional English restructuring is permitted and formal correspondence is not required.

Source evidence
  • Frozen source span · Ruth 2:16 · OSHB
Compare translations on BibleHub ↗
V2CASE-05

Genesis 1:27

MORPH:Functional Restructuring
Translation
NIV
Reviewer
AI Reviewer A
Construct
MORPH
Recorded outcome
Functional Restructuring

The singular collective ha-adam is followed by male/female and plural pronouns; English must move between humanity person and plural human beings

Source evidence
  • Frozen source span · Genesis 1:27 · OSHB
  • BDB · Brown-Driver-Briggs Hebrew and English Lexicon
  • HALOT · Koehler-Baumgartner-Stamm, Hebrew and Aramaic Lexicon of the Old Testament
  • WALTKE_OCONNOR · Bruce K. Waltke and M. O'Connor, An Introduction to Biblical Hebrew Syntax
Compare translations on BibleHub ↗
V2CASE-06

Exodus 21:22-25

AMB:Retained Or Favoured But Open
Translation
BSB
Reviewer
AI Reviewer A
Construct
AMB
Recorded outcome
Retained

The rare noun ason in the injury law is crucial to whether the harm refers to the woman child or any fatal injury

Source evidence
  • Frozen source span · Exodus 21:22-25 · OSHB
  • BDB · Brown-Driver-Briggs Hebrew and English Lexicon
  • HALOT · Koehler-Baumgartner-Stamm, Hebrew and Aramaic Lexicon of the Old Testament
  • NET_NOTES · NET Bible Full Notes Edition
Compare translations on BibleHub ↗
V2CASE-07

Numbers 23:19

AMB:Resolved Or Displaced
Translation
BSB
Reviewer
AI Reviewer A
Construct
AMB
Recorded outcome
Resolved

Nacham spans regret relent comfort and change-of-course senses and here occurs in a denial of divine human-like changing

Source evidence
  • Frozen source span · Numbers 23:19 · OSHB
  • BDB · Brown-Driver-Briggs Hebrew and English Lexicon
  • HALOT · Koehler-Baumgartner-Stamm, Hebrew and Aramaic Lexicon of the Old Testament
Compare translations on BibleHub ↗
V2CASE-08

1 Samuel 14:41

TB:Recognised Alternative Or Mixed
Translation
BSB
Reviewer
AI Reviewer A
Construct
TB
Recorded outcome
Recognised Alternative

Ancient witnesses preserve a substantially longer form of Saul's prayer for the sacred lot

Source evidence
  • Frozen source span · 1 Samuel 14:41 · OSHB
  • BHS_APPARATUS · Biblia Hebraica Stuttgartensia
  • TOV_TCHB · Emanuel Tov, Textual Criticism of the Hebrew Bible
Compare translations on BibleHub ↗
V2CASE-09

Numbers 12:5

EXPL:Resolves Live Alternative Or Expansive Source External
Translation
NIV
Reviewer
AI Reviewer A
Construct
EXPL
Recorded outcome
3 Resolves Live Alternative

Assess interpretive explicitness against the complete frozen source evidence for this passage; greater explicitness is descriptive and is not automatically better or worse.

Source evidence
  • Frozen source span · Numbers 12:5 · OSHB
Compare translations on BibleHub ↗
V2CASE-10

1 Samuel 14:41

RQ5:Textual Basis Vs Translation Or Interpretive Cause
Translation
BSB + NASB
Reviewer
AI Reviewer A
Construct
RQ5
Recorded outcome
Textual Basis

This RQ5 example shows how a material pair difference was classified without publishing restricted target wording.

Source evidence
  • Committed private evidence reference · identity-safe SHA-256 commitment
  • Committed private evidence reference · identity-safe SHA-256 commitment
  • Committed private evidence reference · identity-safe SHA-256 commitment
  • Committed private evidence reference · identity-safe SHA-256 commitment
  • Committed private evidence reference · identity-safe SHA-256 commitment
Compare translations on BibleHub ↗

Glossary

What the study's terms mean

Textual basis
Which recognised source-text reading underlies a translation. A recognised alternative reading is descriptive, not automatically an error.
Semantic representation (SEM)
Whether frozen source meaning is preserved, functionally represented, narrowed, broadened or materially changed.
Morphosyntax
Meaning-bearing grammar and syntactic relationships. Functional English restructuring can preserve the source relation; word order itself is not rewarded.
Interpretive explicitness
How much interpretation is made overt in English. More explicit is not automatically better or worse.
Ambiguity handling
Whether a live source ambiguity is retained, favoured while left open, resolved, displaced or replaced by a new ambiguity.
Lexical continuity
How a frozen source sense is rendered across linked passages. Contextually justified variation is allowed; literal repetition gets no automatic bonus.
Intertextual visibility
How visibly an Old Testament / New Testament relationship remains available without requiring maximal verbal identity.
English clarity
Separate source-blind ratings of syntactic clarity, referential clarity and idiomatic naturalness. It is not folded into fidelity.
Reviewer sensitivity
Differences between the two independent AI reviewer outputs. They are reported as uncertainty, not converted into translation penalties.

Method and limits

What the study can and cannot say

Difficulty-enriched corpus

120 passages deliberately concentrate recognised translation challenges; 40 controls provide contrast. This design supports stress-testing, not whole-Bible prevalence claims.

Independent AI-assisted review

Two AI systems independently reviewed the frozen instrument. Their outputs are labelled AI Reviewer A and AI Reviewer B on this page and are not averaged or adjudicated into a single reviewer.

Separate constructs

Semantic representation, morphosyntax, explicitness, ambiguity, lexical continuity, intertextual visibility, reader disclosure and English clarity answer different questions.

No best-Bible claim

The study does not identify a best translation, create an overall accuracy percentage, or replace expert human textual and translation scholarship.

Relationship view is post-hoc

The similarity matrices were added after the study results were frozen. They reuse frozen outcome distributions only, keep reviewer estimates separate as a displayed range, do not alter any study score, and must not be read as a new quality ranking.

Reproducibility

Frozen publication package

Reader guide Data dictionary Public projection JSON Worked cases JSON Manifest Validation record

The public package contains no restricted target wording and no private blinded target identifiers. Historical V1 research files remain retained separately in the repository.