Pre-Registration — Chord vocabulary in charting popular music, 1958–1991: does the harmonic vocabulary narrow?

Registered: 2026-09-27, before any chord annotation was read. What was done first, and is disclosed here: the corpus documentation was read for coverage, sampling scheme and annotation count (§3), and the sample-size problem in §4 was identified from that documentation alone. No chord label from the corpus has been counted, and no window has been assigned a vocabulary size.

Why this document exists: MUSIC_NOTATION_NARROWING_PLAN_2026-09-27.md argues that The Narrowing — the vocabulary-compression model built for language — holds phase for phase in music. Three of its four phases rest on the documentary record and were verified in §7 of that plan. The fourth, the digital era, made a measurable claim: that popular music's harmonic vocabulary has contracted. A claim of that shape must be registered before the data is touched or it is not evidence, it is illustration.

Status: NOT YET RUN. Revisions require a new dated document; both remain public.

---

1. Hypotheses, one direction each

H1 (vocabulary contraction). The number of distinct chord qualities used per five-year window declines across 1958–1991, measured at equal sample size (§4). H0. Distinct-quality count shows no monotonic trend across windows beyond sampling variation.

H2 (transition contraction). The number of distinct chord-to-chord transitions per window declines across the same windows, at equal sample size. A vocabulary can hold steady while the grammar joining it narrows; this asks that separately.

H3 (concentration, secondary). The share of all chord tokens taken by the three most common qualities rises across windows. This is the "fewer words doing more work" version of the claim, and it is possible for it to hold while H1 fails — the vocabulary stays available but stops being used.

All three are one-directional because the thesis is directional. An increase is a refutation, not a result with the sign flipped, and §6 says so.

2. What the corpus cannot be asked

The chord annotations end in November 1991. This is the single most important constraint and it is stated before anything else, because it is the one a reader of the finished exhibit would otherwise assume away.

That means: