The method

The study reads every text by strict syntactic textualism. It takes the official words, sets out their syntax, shows their meaning at ratification from the evidence of the time, and reads each clause with the whole text. Each canon is named where it is used, and four layers are kept apart.

1. The official text. The Constitution is quoted from the National Archives’ transcription of the parchment, which says: “The spelling and punctuation reflect the original.” The Sixteenth Amendment is quoted as ratified: “Passed by Congress July 2, 1909. Ratified February 3, 1913.” The Code is quoted as in force, and each section as first enacted, from the Statutes at Large, with the bill it came from.

2. Syntax first. For each clause the study states the subject, the verb and the object, and what each modifier and each comma attaches to. Two examples carry the whole argument:

• “The Congress shall have Power … To coin Money, regulate the Value thereof”. The subject is The Congress. The verb is shall have, and its object is Power, whose reach the infinitives name. Thereof points back to Money. The clause names who holds the power and what it reaches, and nothing else.
• “All Bills for raising Revenue shall originate in the House of Representatives”. The subject is All Bills, and the participial phrase for raising Revenue restricts which Bills. The verb is shall originate, and in the House of Representatives says where. Revenue stands only as the object of raising, inside the modifier. The clause commands where a class of Bills begins.

3. Original public meaning. The meaning is what the words meant to an ordinary informed reader when they were ratified: 1787–88 for Article I, 1909–13 for the Sixteenth Amendment. It is shown from the evidence of the time and never asserted:

• the dictionaries: Webster’s American Dictionary (1828), the first American dictionary, forty years after the Convention; Black’s Law Dictionary (2d ed., 1910); and Webster’s Revised Unabridged (1913);
• the Convention’s records (Madison’s Notes) and The Federalist;
• the statutes of the time: the Coinage Act of 1792, the Act of 1834, the Gold Standard Act of 1900 and the Revenue Act of 1913;
• Hamilton’s Report on the Establishment of a Mint (1791).

A later court’s phrase is not evidence of what the words meant in 1913. Johnson’s Dictionary (1755) and Webster’s New International (1909) are listed and were not opened, so no grade rests on them.

4. The whole text. Each clause is read with its neighbors. In the Constitution these are Article I, § 1; § 2, cl. 1; § 7, cls. 1–2; § 8, cls. 1, 2, 5, 6 and 18; § 9, cls. 1, 4 and 7; § 10, cl. 1; Article II, §§ 1 and 3; Article VI; and Amendments V, VII, X and XVI. In the Code they are 26 U.S.C. §§ 1(f), 61, 1001, 1011, 1012 and 1016, 31 U.S.C. § 5101, and 12 U.S.C. §§ 225a, 263 and 411.

5. Canons, named where used.

• Surplusage: “It cannot be presumed, that any clause in the constitution is intended to be without effect; and therefore, such a construction is inadmissible, unless the words require it” (Marbury v. Madison, 1803).
• Whole text and harmonious reading: a clause is read with its neighbors, so that they stand together.
• Expressio unius: to name one thing is to leave out another.
• Consistent usage: a text that changes its word changes its meaning.
• Ordinary meaning: words carry the sense an informed reader of their time gave them.

6. Four layers, each labeled, none passed off as another.

• The text: what the words say, and what the evidence of their time shows they meant. This layer answers on the ratified words alone and never uses the phrases text 01’s constraints name.
• The facts: the data at their primary sources, each with its date.
• The courts: every holding that bears on a link, quoted from the opinion and set beside the link, for or against. A holding is weighed, never used in place of the words.
• The logic: what follows, checked by the solver.

Where a premise’s grade on the text alone differs from its grade on the whole record, both are given.

7. Receipts at witness level. A receipt is the exact words of a primary source and where they stand: the clause, the statute, the opinion, the dictionary, the data release. A summary can lead to a receipt but never is one. “A quotation cannot be repaired by inference” (Gavin, 2026-10-09). A broken quotation is reopened at its source, or the claim that used it is cut.

8. The grades and the rule. The grades are ranked from strongest to weakest: stated and it follows, then strong support, moderate support, one reading, left open, weak support, not shown and ruled out. A conclusion takes the grade of the weakest premise it needs. Where it has more than one route, it takes the grade of its strongest route. A conclusion whose needed premise is ruled out is not shown by that link. It is ruled out only where a valid link proves its negation from premises that are stated or follow. The guard checks each of these rules on every conclusion.

9. The solver. Every inference is set out as numbered premises and a conclusion, in plain words and in logical notation, and checked with the Z3 solver. A conclusion follows when the premises together with its denial cannot all be true. Each check is run again with each premise dropped, and a premise counts as needed only when the conclusion then fails. A check that cannot fail proves nothing. Every either-or is tested for completeness: a construction counts as permitted when the words, entered as written, are consistent with it. All 88 checks give their expected result. The solver shows what follows, and the grades say whether the premises are true.

10. Gavin’s rules. R1: there is one correct conclusion, and where the words are ambiguous the Supreme Court must rule. R2: every link is attacked from both sides until one outcome is left standing. Under R1, where the words read with their meaning at ratification decide a question, that decision is the conclusion. Where they leave it open, the Supreme Court’s holdings decide it. Where the Court has not ruled, the grade is left open, and the study says so.

11. The record. The study lists 150 authorities and opened 148 of them. It quotes 365 passages, each copied from its page in spans of 130 characters or fewer and rejoined where the spans overlap. Each of the 253 passages a grade or a finding rests on was read back whole against its page and matched. Every source quoted, 140 in all, also carried a planted error the reader had to catch, and all 263 plants were caught. In the first pass, eight readings came back as no match: seven found a real difference, and each of those quotations was corrected at its source; one was a hyphen at a line’s end, and the passage matched when read again. A second pass read the opinions of the Supreme Court on the United States Reports as the Government Publishing Office scans them. Of the 123 passages quoted from them, 78 now follow those pages, 34 of them corrected where the scanned page differs from Justia’s text by a word or a mark. The other 45 follow Justia’s text: 23 from the five decisions from 2020 on; 9 from four long opinions whose scans the reader could not reach at the passage; and 13 where the scan’s text garbles the passage itself (a date run together, a quotation mark read as a number).