Read Shakespeare’s couplet — proved / loved — and the rhyme is right there on the page and gone from your mouth. A rhyme is data about how a word was said; when the saying changes, the rhyme stops being heard and becomes merely seen. This is a poem’s music lost to time — and you can hear it close back up. Every line is verbatim; every reconstructed sound is sourced and attributed; the contested ones are flagged, because the most famous example is the one scholars still argue about.
A rhyme chimes when two words share their stressed vowel. Below, each pair is sounded twice — at its modern value, where the words clash, and at its reconstructed period value, where they ring together (or, for the contested cases, much closer). The synthesiser sounds the stressed vowels only — the seat of the rhyme — not the whole word. Tap the boxes and listen for the difference: the same line, two distances apart in time.
A rhyme that breaks loses its music. A pun that breaks loses its meaning — a whole second sense that lived in the sound. When two words that once fell together drift apart, the joke (or the menace) simply stops arriving. Here the word on the page is the one you still hear; the one beside it is the ghost.
It is tempting to blame the Great Vowel Shift — the great raising of the long vowels, c. 1400–1700 — for all of this. But that is the easy, half-true story. Most of these rhymes were broken by other changes entirely: the FOOT–STRUT split that split love from prove (first described in 1644), the LINE–JOIN realignment that pulled join off divine, a plain 18th-century shortening of wind. “Sound change broke the rhyme” is true. “The Great Vowel Shift broke it” usually isn’t. Each change, and which pairs it claimed:
So we reconstruct. The work is real and careful — but listen to what its leading practitioner says it is. David Crystal, who built the standard Original Pronunciation of Shakespeare, refuses the word definitive:
What is exact, and what is inferred. Three kinds of claim are kept apart here. (1) The text — every quoted line — is verbatim from a named edition, and the verifier confirms each rhyme word actually occurs in its passage. (2) The reconstructed period sound is the cited, attributed work of others (Crystal’s Original Pronunciation; Kökeritz 1953; the OED), never asserted as bare fact; where authorities disagree the pair is tagged disputed and the dispute is named on its card. (3) The inference that a rhyme “worked” follows from (2) and inherits its uncertainty.
What the verifier proves. That every “broken chime” genuinely encodes vowels that matched in the reconstruction and differ today; that every sounded vowel resolves to a published formant value (so the audio is honest); that each rhyme cites a real sound change, and that most of those changes are not the Great Vowel Shift; and a drift guard that the data embedded in this page is byte-identical to the source of truth. It does not re-derive the phonology — that is two centuries of philology, demonstrated and cited here, not re-proven.
What is sounded — honestly. The voice is formant synthesis: a buzzing ~120 Hz source filtered to each vowel’s first two formants F1/F2. Those numbers are published means — Catford (2001, 2nd ed., p. 154) and Peterson & Barney (1952, Table II). This is an acoustic sketch of a vowel target, not a recording — least of all of a 1600s speaker; no playable recording of any human voice predates 1877. Where a reconstructed value is a diphthong the synth can only approximate (the period PRICE [əɪ]), it uses the nearest sounded target and says so. Your ears judge the timbre; the verifier checks only the vowels.
The backing data, the verifier, and the full source list live in research/the-rhyme-the-sound-forgot. ◆ A combine in the Language seam: the sound-machine of The Sound the Spelling Forgot turned on the question of the Translation-Criticism Venue — the untranslatable, here across time.