Appendix F — Replicability Note

Enable any reader to reproduce the corpus observations that underwrite the claim of grammatical licitness and distributional under-attestation.

Data Sources (publicly checkable)

• Biblical Hebrew base text: BHSA/ETCBC (morphologically annotated Hebrew Bible).

• Secondary witnesses (for cross-checking verse shapes): MT (BHS/BHQ), DSS transcriptions where available, and major English translations for quick alignment only (not for analysis).

• Grammars used for category labels: Holmstedt; Arnold & Choi; Waltke & O’Connor; Joüon–Muraoka; Davidson; Driver; BHRG (Van der Merwe). See Notes/Bibliography.

Why BHSA/ETCBC? It provides a morphologically and syntactically annotated Hebrew Bible, allowing us to query clause constituents (verb, subject NP, direct objects) and test for the presence/absence of ’et and waw. In this appendix, annotated / tagged / labeled refers to features assigned by the database; distributionally under-attested refers to patterns that are grammatically licensed but infrequent or presently unmatched in the surviving corpus.

Tagging Rules (what each column means)

• Frame. S + qatal / S + participle / S + predicate / S + yiqtol, etc., following ETCBC clause function labels; “w+S+…” indicates prefixed waw on the clause or verb.

• NP/definiteness per item. INDEF (bare noun), DEF[ART] (with ה-), DEF[ET] (governed by ’et), DEF[SUF] (possessive suffix), PROP (proper name), PREP (prepositional object—not counted as DO).

• ’et (by item). ✓ means item is marked with ’et / wᵉ’et; — means ’et-unmarked.

• Waw. final (only before last item), each (before each item after the first), pairs (grouped couples), irreg (mixed), none (no coordinating waw on items).

• Asyndeton/Additives. Note when no waw appears between adjacent items (e.g., “1–2”) or when additive particles (e.g., גם) or repeated prepositions (e.g., ל־, מן) run the chain.

How to Verify (step-by-step)

1. Open the base text. Use an interface to BHSA/ETCBC or a morphologically tagged MT.

2. Locate the verse. Confirm Hebrew word order and identify the target clause (avoid subordinate/appositive spill-over).

3. Mark constituents. Identify the verb, the subject NP (if explicit), and each direct object. Distinguish DOs from prepositional complements.

4. Record definiteness & markers. For each object, record: presence of ’et; article ה-/pronominal suffix; whether it is a proper name; any כל- constructions.

5. Record coordination. Note where waw appears (each item vs. final-only), and any asyndeton (adjacent items without waw).

6. Classify the pattern. Assign the legend codes (NP/def., ’et, Waw, Asyndeton) and add a short functional note (e.g., “category → subtypes,” “household roster,” “polysyndeton”).

7. Cross-check. Compare the classification with at least one reference grammar for that construction type and glance at two parallel examples in our tables.

Tip. For mixed-marking cases, double-check whether the “’et-unmarked” but definite NP is actually definite by article or suffix (e.g., כל-NP with article) rather than truly indefinite.

Query Templates (conceptual)

• S-initial narrative frames. Find clauses with explicit subject NP before verb (e.g., waw + NP + qatal/participle/copular). Sample query idea: clause where role = Subj precedes role = Predication/Verb; domain = narrative.

• Lists with DOs. Find verbal clauses that have ≥ 3 direct objects; then filter where obj1 lacks ’et, obj2/obj3 have ’et, and coordinator waw appears only before obj3 (final-only).

• Negative controls. Pull “all-definite multi-’et” lists and “all-indefinite” lists to contrast with the target shape.

Replication caveat. Different databases may segment clauses and phrase boundaries differently; visible object series can therefore appear either as one complex Objc phrase or as several linked clauses. Always verify against the visible Hebrew.

SHEBANQ as a Replication and Refinement Tool

The project developed an extended, reproducible SHEBANQ workflow that materially improved candidate harvesting and clarified the limits of the underlying annotation model. In addition to SHEBANQ’s online query interface, exported CSV results were further processed with a custom VBScript developed collaboratively during this project. The script aggregated phrase- and clause-level annotations to both clause and verse levels, distinguished free from suffixed object-marker forms, and assigned provisional candidate labels (for example, appositional control, coordination only, and internal-mix near-miss). This extended workflow proved more informative than raw online query results alone, especially where BHSA encodes visible object series as a single complex Objc phrase or distributes them across multiple clauses.

In the book of Genesis, verse-level provisional labels yielded the following counts: COORDINATION_ONLY 160; APPOSITIVE_CONTROL 277; INTERNAL_MIX_NEAR_MISS 21; NOISE 1075. Clause-level provisional labels yielded: COORDINATION_ONLY 176; APPOSITIVE_CONTROL 279; INTERNAL_MIX_NEAR_MISS 12; NOISE 2222. The verse-level set proved more useful for screening, since it recovered cases whose relevant structure is distributed across multiple clauses; the clause-level set remained more useful for tight citation and grammatical discussion.

Illustrative examples include: Gen 1:1 as a possible core target; Gen 1:16 as an appositive control; Gen 1:21 as a verse-level internal-mix near-miss; Gen 1:29 as an internal-mix near-miss; Gen 2:7 as a mixed clause-level artifact requiring caution; Gen 2:19 as coordination only; and Gen 6:10 as a repeated-’et triad that is still not a true Gen 1:1 match.

Conclusion. SHEBANQ 2021 proved highly valuable for harvesting and structuring candidates, but its annotation model does not directly recover all the fine-grained object-chain distinctions relevant to Gen 1:1. Visible object series may be annotated as a single Objc phrase with internal substructure, or distributed across multiple clauses. Its strongest use in this project is therefore not direct proof, but candidate harvesting, annotation diagnosis, and replicable screening.

Versioning & Cross-Refs

• This appendix syncs to: Appendix B (v4) — Unified Table (swept, 2nd-pass summary). Key sections for verification: “Negative Controls (compact),” “Second-Pass Summary & Legend,” and the “Comparanda” note.

• Core pointers: Method box includes the “Apposition vs. Coordination — Quick Diagnostic (Gen 1:1)” pointer; the Core’s “Replication status” paragraph mirrors the language here.

• Change-tracking: If Appendix B increments to v5+, update this section header list so replicators land on the right subsections.

• Resources. Working CSVs and supporting scripts referenced in this appendix are available via the website Downloads menu.

What Counts as a Match (for the Gen 1:1 test)

A single verbal clause whose three direct objects form: INDEF (no ’et) → DEF (’et) → DEF (waw ’et), with no waw between items 1–2 and a single final waw before the last item. Apposition, resumptive lists, and prepositional chains do not count.

Current status. Our unified tables (see Appendix B (v4)) display all components (mixed ’et-marking; final-only waw; asyndeton). After a second-pass sweep with tighter apposition detection, no exact BH exemplar of INDEF → DEF[’et] → DEF[’et] with asyndeton 1→2 and final-only waw has been found. We therefore describe the Gen 1:1 coordination profile as grammatically licit but distributionally under-attested, pending broader corpus sweeps. More specifically, the presently unmatched feature stack is: an ’et-unmarked/category first object, two definite ’et-marked objects, final-only waw, and an asyndetic 1→2 break.

Box F.1 — Falsifiable Counter-Reads (How this proposal could fail)

Purpose. Make the claim testable. The subject-reading of בראשית (as nominal subject) is permissible but distributionally under-attested. The following findings would undercut the proposal; conversely, none of these is assumed true by the argument.

Clause-shape & coordination falsifiers

1. No S-initial licensing. A corpus-level demonstration that BH narrative does not license the S-initial frames we cite (e.g., the Appendix-A exemplars are misparsed or genre-bound such that Gen 1:1 cannot analogize) would remove the word-order plank.

2. Coordination ban. Evidence that a three-item direct-object chain cannot take the shape INDEF → DEF(’et) → DEF(’et) with final-only waw and 1→2 asyndeton (i.e., a principle-level prohibition, not just rarity) would collapse the coordination plank.

3. Apposition is forced. A formal cue in Gen 1:1 showing אֱלֹהִים is head + appositive renaming ("namely the heavens and the earth")—or a semantic argument that gods/powers and cosmic domains are necessarily co-extensive in BH usage—would defeat the simple-coordination analysis.

B) Lexeme & onomastics falsifiers

4. Onomastic impossibility. A strong, rule-based case that prepositional/adverbial bases (ב/כ/ל - forms, rēʾšît-type) cannot function as names/titles in BH—i.e., that our nominalizing comparanda are illusory—would remove the name-value plank.

5. Local אֱלֹהִים constraints. A grammatical constraint in Gen 1:1 specifically that forces proper-singular Elohim as subject (agreement, pronominal binding, prosodic evidence) would preclude reading אֱלֹהִים as a common plural object ("gods/powers").

C) Textual/para-textual falsifiers

6. Pre-Masoretic disambiguation. Early Hebrew witnesses (e.g., pre-MT orthography/accents) that unambiguously encode a subject-first “God” reading, beyond translator smoothing, would weigh against the alternative parse being live in antiquity.

What does not falsify the claim (controls)

• Frequency alone. Rarity alone does not equal ungrammaticality in a finite corpus.

• Translation traditions. LXX/Targum choices show public stabilization, not the only licit BH parse.

• Re-use of “בראשית.” The lack of reuse as a title is not decisive; incipit-epithets can be unique.

Claim (A) note — distribution of rēʾšît/berēʾšît

The proposal does not depend on frequent nominal use of בראשית elsewhere. Outside Gen 1:1, berēʾšît typically occurs in construct-like contexts ("at the beginning of X"), and rēʾšît functions as a common noun ("first/primacy"). Our argument is narrower: BH onomastics does allow nominalization/epithet-like titles, so treating בראשית as a proper/nominal subject in Gen 1:1 is possible in principle; the question is distributional plausibility, handled by the clause-shape and coordination tests above.

Quick Access — Data & Replicability (compact)

If you want the shortest path to replicate the tables and claims:

1. Base text/interface. Open BHSA/ETCBC (e.g., SHEBANQ) and display clause, phrase, and word features. In this workflow, database features are treated as annotations to be checked against the visible Hebrew; they are a screening aid, not a substitute for philological judgment.

2. S-initial frames (Appendix A). Filter clauses where a Subject phrase precedes the verbal/predicative phrase (e.g., S + qatal / S + participle / S + predicate). On ETCBC this means finding a phrase with function=Subj that occurs before a phrase with function=Pred (or a verbal Pred). Then sample verses from Appendix A to confirm the pattern.

3. DO-list extraction (Appendix B). In verbal clauses, collect direct-object phrases (ETCBC function=Objc). Record, per item: presence of ’et (token = אֵת/וְאֵת), article (ה- at word level), suffix/pronoun, and whether the item is a proper name. Note coordination by observing waw on items; mark asyndeton where adjacent objects lack waw.

4. Target triad test (Gen 1:1 shape). Among clauses with ≥3 direct objects, check for the sequence INDEF (no ’et) → DEF[’et] → DEF[wᵉ’et], with no waw between items 1–2 and final-only waw before the last item. Exclude apposition/resumption (ETCBC often marks apposition with rela=Appo at phrase/word links).

5. Negative controls & comparanda. Also pull:

• All-definite multi-’et chains;

• All-indefinite chains;

• Clear apposition sequences (e.g., repeated ’et with proper names) to contrast with simple coordination.

6. Definiteness basis (legend). Use: ART (article), SUF (pronominal suffix), ET (’et-marked direct object), PROP (proper name), BARE (none of the above). See the legend in Appendix B (v4) for quick codes.

Minimal MQL sketch (conceptual) — not copy-paste ready, but a guide for SHEBANQ:

• S-initial: select clause c where exists phrase p1 (function=Subj) before phrase p2 (function=Pred or verbal) within c.

• Obj-triples: select verbal clauses where count(Objc) ≥ 3; harvest the three left-to-right object phrases; test each for ’et token and article/suffix; check inter-item waw/token sequence.

Pointers. Appendix B (v4) holds: Negative Controls (compact), Second-Pass Summary & Legend, and Comparanda; Appendix F (this page) states what counts as a match and falsifiable counter-reads. The Core’s Method box links to the Quick Diagnostic for Gen 1:1.

Notes & Bibliography

See Appendix D — Reception & Composition Notes for the consolidated “Core Grammars” list and other works cited.