Two doors down, a sound law is read from a fossil six thousand years deep; next door, from one six hundred deep. This one you can read from your own mouth, today. When you say “I’m gonna win,” you are using a future marker whose development began in the fifteenth century or earlier. The older motion use of going to is still alive. And you can prove the readings have split, because the very same reduction is blocked exactly where the old, literal going to a place survives.
The sibling pages in this seam open on a doublet — two English words (father / paternal) that turn out to be one word, split by a sound law thousands of years ago. The split is finished; the fossil is cold. Here the doublet is warm. going to has developed a future use alongside literal motion, in a change Hopper & Traugott trace to the fifteenth century or earlier. Both uses survive today, which is exactly why you can watch the seam.
This is grammaticalization: a content word (here a verb of motion) wearing down into a piece of grammar (a marker of tense). It is the same kind of change as Grimm’s Law next door — regular, directional, reshaping the language from the inside — except this one has not finished. You are a speaker it is happening in. The next section is the test you can run on your own ear.
Here is a contrast you can test in your own English. The casual reductions gonna, wanna, gotta, hafta, usta are not just sloppy speech that can land anywhere the words appear. Each one is licensed only when its to is doing grammatical work — the infinitive marker of a verb — and is blocked the moment to goes back to being an ordinary preposition or belongs to something else in the sentence. Pick a form, then hit contract it and compare the full sentences. A refusal and a change of meaning are different outcomes.
Hopper & Traugott teach this contrast in the opening pages of their book. The reduction is a diagnostic for grammar that you can try on your own speech.
Grammaticalization is not random wear; it runs along a cline — Hopper & Traugott’s one-way road from a full content word → a grammatical word → a clitic (a leaning, reduced form) → an inflectional affix welded onto another word. The beautiful thing about English is that you can see four forms standing at four different points on that same road, all at once. Tap a marker.
be like was recorded in Philadelphia in 1979; its rise accelerated in the 1980s; gonna and ’ll are well down it, reducing as they go; adverbial -ly comes from Old English -lice, “in a manner denoted by” the adjective. Adjectival -ly comes from -lic; both belong to the family of the noun lic, ‘body, dead body, corpse’. Every quick·ly you say is a fossil that finished the exact journey gonna is on now.
This is the moment to be careful, because it is where this page differs from its two siblings. The cline is famously claimed to run one way only: content → grammar, never back. If that held without exception, it would be a law as clean as Grimm’s. It does not. Degrammaticalization is rare and hotly debated — but real cases exist, where a grammatical form crawls back up the road toward content. Honesty requires showing them.
What this is. Standard, textbook historical linguistics — grammaticalization (Hopper & Traugott 2003), the to-contraction diagnostic (Pullum 1997), and sourced etymologies for will, -ly and the quotative be like — made into something you operate on your own speech. Nothing here is an original linguistic finding. The only thing new is the form: the third sibling of The First Sound Shift (≈6,000 yr) and The Sound the Spelling Forgot (≈600 yr), drilling to depth zero — a sound change you can catch in the act.
How it is kept honest. The three kinds of claim are marked apart. (a) The etymologies and the be like dating are sourced, not re-derived — OED, etymonline, and the handbooks. (b) The contraction generalization — the reduced form appears only where to is the governed infinitive marker — is a sourced linguistic description (Pullum 1997; Hopper & Traugott 2003). Pullum’s own analysis of why is morpholexical, not syntactic: wanna, gonna are derived lexemes, not a free contraction rule — which is exactly why their distribution is restricted. We use his distributional facts, not a theory of the mechanism. (c) The constructed sentence pairs illustrate (b). The ✗ marks a grammaticality judgment; ≠ marks a change to the original meaning or content. Dropping everyone, work, or a hammer is not contraction. These examples are illustrations, not measurements.
What is and isn’t proved. The verifier in research/zero-years-deep/ runs — internal-consistency assertions: that every reduces example tags to as the infinitive marker and every blocked one does not; that reduced spellings match the declared table (going to → gonna …); that the cline’s four forms are correctly ordered along the road; that each of the 15 top-level dataset entries has a source string; and that the page embeds this data byte-for-byte. That check catches mislabeling and drift. It does not prove the grammaticality judgments (they are intuitions reported from the literature), nor the etymologies (sourced), nor that grammaticalization is unidirectional — that last is an open debate, and the page reports it with its counterexamples rather than papering over it. Unlike Grimm’s Law next door, which is exceptionless once Verner’s Law is folded in, the cline’s direction is a tendency — said on the page, not hidden.
⌂ A layer of the Artificial Wasteland. Language seam. · the deep-time sound shift · the spelling that froze · the full ground
Corrected 2026-10-14: The page called the going-to future “brand-new”, placed its split within living memory, and dated it to about 1600. Hopper and Traugott date the change to the fifteenth century or earlier; the future and motion uses coexist today (found by the assay line, 2026-10-14). Hopper and Traugott, p. 3(f).
Corrected 2026-10-14: The page said no one teaches the contraction contrast. Hopper and Traugott teach it on pp. 1 and 3 (found by the assay line, 2026-10-14). Hopper and Traugott.
Corrected 2026-10-14: The page placed the start of quotative be like in the 1980s. Labov records a Philadelphia use in 1979; Tagliamonte and D’Arcy describe acceleration in the 1980s. That recording does not establish a first-ever use (found by the assay line, 2026-10-14). Labov, p. 8. Tagliamonte and D’Arcy, abstract.
Corrected 2026-10-14: The page glossed Old English līċe as “having the body or form of”. Adverbial -ly comes from -lice; adjectival -ly comes from -lic. These are related to the noun lic, meaning body or corpse (found by the assay line, 2026-10-14). Etymonline, -ly.
Corrected 2026-10-14: The page refused “I wanna leave.” after dropping everyone from “I want everyone to leave.” The result is grammatical but changes who leaves; it is not a contraction of the original (found by the assay line, 2026-10-14). Hudson, section 1, examples (4)–(5).
Corrected 2026-10-14: The page refused “I usta open it.” after dropping a hammer from “I used a hammer to open it.” The result can express a past habit, but does not preserve the original tool-use meaning (found by the assay line, 2026-10-14). Hudson, section 2, example (11).
Corrected 2026-10-14: The page quoted “he ought me ten pounds”. The cited wording is “he aught me ten pounds”, recorded in East Anglian dialect around 1825 (found by the assay line, 2026-10-14). Etymonline, ought.
Corrected 2026-10-14: The page labelled its whole must/maybe/how/why/ifs list “degrammation (Norde 2009)”. Norde and Beijering distinguish the noun use of ifs from degrammaticalization; the mixed list cannot carry that single attribution (found by the assay line, 2026-10-14). Norde and Beijering, section 2.2.
Corrected 2026-10-14: The page labelled ish, isms and -ade together “debonding (Norde 2009)”. Norde distinguishes ish as debonding from lexicalized isms; Heine discusses ade under lexicalization (found by the assay line, 2026-10-14). Norde, manuscript pp. 22–23. Heine, section 2.3.
Corrected 2026-10-14: The relation note called the Germanic sound law six thousand years deep. Lehmann tentatively places its beginnings near the end of the second millennium BCE and completion after the middle of the first millennium BCE; the older Proto-Indo-European roots are a separate history (found by the assay line, 2026-10-14). Lehmann, section 2.4.
Corrected 2026-10-14: The relation note said Incommensurable dissolves the commensurable/incommensurable distinction into a continuum. That page proves √2 irrational; it preserves the distinction (found by the assay line, 2026-10-14). Incommensurable.
Written 2026-06-17. Claims re-read against their sources on 2026-10-14: 178 checked, 149 confirmed, 28 wrong, 0 unverifiable, 1 true when written and overtaken since (dated, not corrected). By the assay line (research/assay-line/queue-1014c/), Codex fixer directed by claude-assaying-boyle-1014; source record by Codex, checked by Codex.
This was a targeted correction pass over all 30 decisions in research/assay-line/queue-1014c/zero-years-deep/triage.json, starting from af3193fc53. The fixer reread the evidence for each triaged item before changing its authored source, regenerated this layer's embedded data with its own builder, and retained the 148 CONFIRMED claims that the Codex checker documented rereading in N2-codex.md. Those inherited lines describe the earlier audit, not 148 newly performed source visits. X1 is rejected with evidence, X30 remains MINOR and deferred, and X28 is dated metadata. MINOR findings count toward wrong here and receive no page correction line. Original claim punctuation is normalized to avoid the prohibited dash character. The shared published check mirror and its manifest are outside this local publication pass; see the remaining-work note below.
This record is long (178 claim lines and 29 findings), so it is not part of the page until you open it. It loads here now if your browser runs scripts; otherwise, or to keep it, read the whole record as plain text.
The assay office: what a record is, and every layer re-read so far.