get_cross_references

USE THIS whenever you are explaining, exegeting, or arguing from a specific Bible verse — and for any theological or doctrinal question.

Two complementary modes:

By verse reference (most common). Pass reference="John 3:16" (or any canonical verse) to get the passages historically read alongside it. The database draws from four scholarly sources, returned in a three-tier ranking:

Tier 3 (top, "consensus and curated"): - CH — Harrison & Romhild's curated dataset (~58k links, OT-only as source). Hand-vetted; high-relevance pairs flagged canonical-direction. - Gage parallel — the tighter pairings from Warren Gage's John ↔ Revelation typological reading (Bradley/Gage, John/Rev only). - TSK ≥100 votes — TSK pairs with crowd-source consensus that strong are near-universal cross-references (top ~0.4% of TSK) and break through to compete with curated.

Tier 2 ("argued and acknowledged"): - Burnett — David A. Burnett's argued chain for the Gen 15:5 / Rom 4:18 "star-like seed" deification reading (JSPL 5.2, 2015). ~30 pairs. - Gage chiastic — the looser-typology sheet of Bradley/Gage (the source spreadsheet labels these "looser connections, just noting"). - TSK 20–99 votes — solid topical links acknowledged across commentaries (top ~5%).

Tier 1 (long-tail): TSK <20 votes — surface only when explicitly raising limit for exhaustive study.

Use this BEFORE drawing any theological conclusion from a single verse — results frequently surface the texts the original verse is quoting, the fulfilment passages, contested parallel readings, and the chain of NT authors who picked the verse up.

By theme. Pass theme="atonement" (or salvation_by_grace, deity_of_christ, resurrection, holy_spirit, justification) to get a hand-curated chain of foundational passages for that doctrine. Use this when the user asks a broad theological question without anchoring to a specific verse.

Important caveat about coverage. TSK is built on R.A. Torrey's 19th-century index, which catalogues topical/thematic connections — not necessarily direct quotations or verbal allusions. Consequence: a verse with few cross-refs here is NOT necessarily a verse with few biblical echoes. Famously, Revelation shows surprisingly few links to OT prophetic books even though it is saturated with OT symbolism, because Torrey indexed by subject and Revelation's subject is "apocalyptic". The CH dataset partly compensates (it leans toward NT-quotes-OT linking), so when you suspect a quotation/ allusion is being missed, retry with source="ch" or use find_similar_passages to catch verbal parallels the topical index would skip.

Adaptive default — limit is a CAP, not a target. Default limit=8. The tool returns rows in tier-then-strength order and SUPPRESSES tier-1 noise (low-vote TSK) by default whenever the verse has at least 3 rows from tier 2+. So:

• Signal-rich anchors return 6–8 strong refs spanning curated, scholarly, and consensus-TSK sources.

• Signal-poor anchors return only what passes the bar — fewer rows is the correct answer, not a bug. Don't pad your reasoning with weak refs.

To pull the long tail (only when the user explicitly asks for exhaustive study): pass source="tsk" (returns all TSK including tier 1) or min_strength=0 (treats as explicit "I want some long-tail too"). Either disables tier-1 suppression. Pair with a higher limit (20–30).

How to interpret the scores you get back. Each row carries type (the dataset), relevance (its native strength signal), and where applicable a tsk_votes side-channel showing the TSK count for that pair. You MUST read these before using a ref:

TSK vote scale (full corpus distribution): ≥ 500 votes — extraordinary; near-universal cross-reference (top 0.01%, only 35 pairs) 100-499 — very strong; the link tradition reflexively makes (top 0.4%) 50-99 — strong; well-established parallel (top 1.3%) 20-49 — solid; real connection acknowledged across commentaries (top 5%) 10-19 — moderate; one of many recognised links (top 12%) 5-9 — weak; thematic stretch, use with caution (top 33%) 2-4 — very weak; mostly noise floor (62% of TSK) 0-1 — noise

CH (curated — all CH refs carry signal, but the tag tells you weight): "canonical direction" (rel=3 or 2) — Harrison's flag for the canonical direction of the pair, often part of a thematic circle (top 78% of CH) no tag (rel=0) — present in CH but unflagged (still hand-curated)

Gage (John ↔ Revelation typology): relevance=3 ("parallel" tier) — tighter pairings from the parallel-reading of John 1 ↔ Revelation 1 relevance=1 ("chiastic" tier) — looser thematic echoes across the full John-Revelation chiasm; the source spreadsheet flags these as "looser connections, just noting" The note field carries the thematic tags + commentary + per-row attribution (Bradley vs Gage). Treat as canonical-typology, not topical.

Burnett (single-paper argued chain): All Burnett rows are at relevance=5 by convention — they're explicit claims from one scholar's published argument, not graded by strength. The note field carries the JSPL citation and which step of the argument the pair belongs to. Cite Burnett by name when surfacing these to the user; they're a scholarly proposal, not consensus.

When the top results are weak, SAY SO. If the strongest ref returned has only 5-15 votes, do not present it with the same confidence as a 200-vote parallel. Caveat the answer: "this verse isn't strongly cross-referenced in the topical index — the closest link is X with only N votes, suggesting tradition didn't treat this as a major thematic anchor." When TSK is thin, try source="ch" — Harrison's curated set leans toward NT-quotes-OT links and may catch what a topical index missed. Or fall back to find_similar_passages for verbal/semantic parallels Torrey wouldn't index.

You may pass source="ch" or source="tsk" to restrict to one dataset — useful when CH alone gives too little (e.g. an obscure verse with no CH coverage) or when you want the dense TSK long-tail.