Sound change alters phoneme inventories and processes over centuries through mechanisms including phonetically-motivated reanalysis, analogical leveling, and substrate effects from language contact. Reconstruction uses the comparative method and internal reconstruction to infer historical sound changes and test phonological principles.
Compare cognate sets across related languages to reconstruct proto-phonology; identify sound correspondences and formulate sound change rules; test for exceptions indicating different change periods.
Sound changes are systematic, not random; even when phonetically abrupt, changes proceed through gradual frequency shifts in variation before categorical adoption.
From your study of phonological systems and language variation, you understand that every language has a structured inventory of phonemes with rules governing their distribution — and that speakers vary in their realizations of those phonemes. Diachronic phonology asks how those systems change over time: why does the English vowel in *moon* have a different quality than in Chaucer's English, why does Spanish *hijo* correspond to Latin *filius*, and how can linguists reconstruct sounds no living person has ever heard? The answers require understanding both the mechanisms that generate change and the methodology for inferring change from surviving evidence.
Sound change mechanisms fall into several categories. Assimilation occurs when a sound becomes more like an adjacent sound — Latin *septem* (seven) becomes Italian *sette* because the /p/ assimilated toward the /t/ following it. Dissimilation is the opposite: two similar sounds in a word diverge to reduce articulatory interference — hence the *l/r* alternation in English borrowings of Latin words with two /l/ sounds (*colonel* pronounced with /r/). Lenition (weakening) is the pervasive tendency for consonants to weaken in intervocalic position: the /p/ in Latin *lupum* (wolf) voiced to /b/ in Spanish *lobo* via a stop-to-fricative trajectory. Chain shifts occur when changes in one part of the vowel space trigger compensatory shifts elsewhere to maintain phonemic contrasts — the Great Vowel Shift in Middle English (ca. 1400–1700) was a systematic upward rotation of all long vowels that produced many of the irregularities between English spelling (which preserved pre-shift forms) and modern pronunciation.
The comparative method is the tool for reconstructing historical sound states. When you find cognate sets across related languages — English *father*, German *Vater*, Latin *pater*, Sanskrit *pitā*, Greek *patēr* — you can observe systematic sound correspondences. The *f*-*v*-*p* pattern for what is clearly the same root word is not random: it reflects regular processes (Grimm's Law) that affected Proto-Germanic consonants in a predictable way. By applying the comparative method across many cognate sets, linguists reconstruct proto-forms like \*ph₂tḗr in Proto-Indo-European and formulate the rules that generated each daughter language's reflexes.
The critical insight — and the content of the misconception — is that sound changes are exceptionless in their application to the environments that trigger them, even though the changes themselves proceed gradually through the speech community via variation. Neogrammarian linguists in the 19th century controversially claimed "sound laws have no exceptions," meaning that every word containing the affected sound in the affected environment undergoes the change. Apparent exceptions always have explanations: loanwords borrowed after the change completed, analogical leveling where morphological pressure kept a form stable, or dialect borrowing where a form was imported from a dialect that hadn't undergone the change. This systematic regularity is what makes historical reconstruction possible — if sound change were random, we could never confidently project backwards to proto-forms.