Grimm's law (also known as the First Germanic Sound Shift) is a set of sound laws describing the Proto-Indo-European (PIE) stop consonants as they developed in Proto-Germanic in the 1st millennium BC. First systematically put forward by Jacob Grimm but first remarked upon by Rasmus Rask, it establishes a set of regular correspondences between early Germanic stops, fricatives, and the stop consonants of certain other centum Indo-European languages.
Grimm's law was the first discovery of a systematic sound change, and it led to the creation of historical phonology as a separate discipline of historical linguistics. The correspondence between Latin p and Germanic f was first noted by Friedrich von Schlegel in 1806. In 1818, Rasmus Rask extended the correspondences to other Indo-European languages such as Sanskrit and Greek, and to the full range of consonants involved. In 1822, Jacob Grimm put forth the rule in his book Deutsche Grammatik and extended it to include standard German. He noticed that there were many words which had different consonants from what his law predicted, and these exceptions defied linguists for a few decades, but they eventually received explanation from Danish linguist Karl Verner in the form of Verner's law.
This chain shift (in the order 3,2,1) can be abstractly represented as:
Here each sound moves one position to the right to take on its new sound value. Note that within Proto-Germanic, the sounds denoted by ⟨b⟩, ⟨d⟩, ⟨g⟩ and ⟨gw⟩ were stops in some environments and fricatives in others, so b? -> b should be understood here as b? -> b/?, and likewise for the others. The voiceless fricatives are customarily spelled ⟨f⟩, ⟨þ⟩, ⟨h⟩ and ⟨hw⟩ in the context of Germanic.
The exact details of the shift are unknown, and it may have progressed in a variety of ways before arriving at the final situation. The three stages listed above show the progression of a "pull chain", in which each change leaves a "gap" in the phonological system that "pulls" other phonemes into it to fill the gap. But it is also conceivable that the shift happened as a "push chain", where the changes happened in reverse order, with each change "pushing" the next forward to avoid merging the phonemes.
The steps could also have occurred somewhat differently. Another possible sequence of events could have been:
This sequence would lead to the same end result. This variety of Grimm's law is often suggested in the context of the glottalic theory of Proto-Indo-European, which is followed by a minority of linguists. This theoretical framework assumes that "voiced stops" in PIE were actually voiceless to begin with, so that the second phase did not actually exist as such, or was not actually devoicing but a loss of some other articulatory feature such as glottalization or ejectiveness. This alternative sequence also accounts for the phonetics of Verner's law (see below), which are easier to explain within the glottalic theory framework when Grimm's law is formulated in this manner. Additionally, a change from aspirated stops to fricatives is known to have happened in the transition between Proto-Indo-European and Proto-Italic, so represents a plausible potential change from Proto-Indo-European to Proto-Germanic.
Once the changes described by Grimm's law had taken place, there was only one type of voiced consonant, with no distinction between voiced stops and voiced fricatives. They eventually became stops at the beginning of a word (for the most part), as well as after a nasal consonant, but fricatives elsewhere. Whether they were plosives or fricatives at first is therefore not clear. The voiced aspirated stops may have first become voiced fricatives, before hardening to stops under certain conditions. But they may also have become stops at first, softening to fricatives in most positions later.
Around the same time as the Grimm's law adjustments took place, another change occurred known as Verner's law. Verner's law caused, under certain conditions, the voicing of the voiceless fricatives that resulted from the Grimm's law changes, creating apparent exceptions to the rule. For example:
Here, the same sound *t appears as *þ /?/ in one word (following Grimm's law), but as *d /ð/ in another (apparently violating Grimm's law). See the Verner's law article for a more detailed explanation of this discrepancy.
The early Germanic *gw that had arisen from Proto-Indo-European *g (and from *k? through Verner's law) underwent further changes of various sorts:
Perhaps the usual reflex was *b (as suggested by the connection of bid < *bidjan? and Old Irish guidid), but *w appears in certain cases (possibly through dissimilation when another labial consonant followed?), such as in warm and wife (provided that the proposed explanations are correct). Proto-Germanic *hw voiced by Verner's law fell together with this sound and developed identically, compare the words for 'she-wolf': from Middle High German wülbe and Old Norse ylgr, one can reconstruct Proto-Germanic nominative singular *wulb?, genitive singular *wulgij?z, from earlier *wulgw?, *wulgwij?z.[failed verification]
Further changes following Grimm's law, as well as sound changes in other Indo-European languages, can occasionally obscure its own effects. The most illustrative examples are used here.
|Proto-Indo-European||Meaning||Non-Germanic (unshifted) cognates||Change||Proto-Germanic||Germanic (shifted) examples|
|*p?ds||"foot"||Ancient Greek, (poús, podós), Latin: p?s, pedis, Sanskrit: p?da, Russian: (pod) "under; floor", Lithuanian: p?da, Latvian: p?da, Persian: (pa), Serbo-Croatian: "pod" ("under; floor") and "peta" as in "heel"||*p > f [?]||*f?t-||English: foot, West Frisian: foet, German: Fuß, Gothic: f?tus, Icelandic, Faroese: fótur, Danish: fod, Norwegian, Swedish: fot|
|*tréyes||"three"||Ancient Greek? (treîs), Latin: tr?s, Welsh: tri, Sanskrit: tri, Russian: (tri), Serbo-Croatian (tr?), Lithuanian: tr?s, Polish: trzy, Albanian: tre||*t > þ [?]||*þr?z||English: three, Old Frisian: thr?, Old Saxon: thr?e, Gothic: þreis, Icelandic: þrír|
|*?wón- ~ *?un-||"dog"||Ancient Greek (ký?n), Latin: canis, Welsh: ci (pl. cwn), Serbo-Croatian: ku?e, Persian: (sag), Russian: (sabaka)||*k > h [x]||*hundaz||English: hound, Dutch: hond, German: Hund, Gothic: hunds, Icelandic, Faroese: hundur, Danish, Norwegian, Swedish: hund|
|*k?ód||"what"||Latin: quod, Irish: cad, Sanskrit: kád, Russian: (kahk), Lithuanian: kas, Serbo-Croatian (Torlakian dialect)vo (kvo), Serbo-Croatian (Kajkavian dialect): (kaj)||*k? > hw [x?]||*hwat||English: what, Gothic: ?a ("hwa"), Icelandic: hvað, Faroese: hvat, Danish: hvad, Norwegian: hva|
|*h?éb?l||"apple"||Lithuanian: obuol?s, Gaulish abalom, Serbo-Croatian (j?buka)||*b > p [p]||*aplaz||English: apple, West Frisian: apel, Dutch: appel, Icelandic: epli, Swedish: äpple, Crimean Gothic apel|
|*dé?m?t||"ten"||Latin: decem, Greek (déka), Irish: deich, Sanskrit: da?an, Russian: (desyat'), Lithuanian: de?imt, Polish: dziesi||*d > t [t]||*tehun||English: ten, Dutch: tien, Gothic: taíhun, Icelandic: tíu, Faroese: tíggju, Danish, Norwegian: ti, Swedish: tio|
|*gel-||"cold"||Latin: gel?, Greek: (gelandrós), Lithuanian: gelmenis, gelumà||*g > k [k]||*kaldaz||English: cold, West Frisian: kâld, Dutch: koud, German: kalt, Icelandic, Faroese: kaldur, Danish: kold, Norwegian: kald, Swedish: kall|
|*g?ih?wós||"alive"||Lithuanian: gyvas, Russian? (?ivoj), Sanskrit: j?vá-, Serbo-Croatian: (?iv), Polish: ?ywy||*g? > kw [k?]||*kwi(k)waz||English: quick, West Frisian: kwik, kwyk, Dutch: kwiek, German: keck, Gothic: qius, Icelandic, Faroese: kvikur, Danish: kvik, Swedish: kvick, Norwegian kvikk|
|*b?réh?t?r||"brother"||Sanskrit: bhr?t?, Ancient Greek: (phr?t?r) ("member of a brotherhood"), Latin: fr?ter, Russian, Serbo-Croatian (brat), Lithuanian: brolis, Polish: brat, Old Church Slavonic: (bratr'), Latvian: br?lis, Persian: (barádar)||*b? > b [b ~ ?]||*br?þ?r||English: brother, West Frisian, Dutch: broeder, German: Bruder, Gothic: broþar, Icelandic, Faroese: bróðir, Danish, Norwegian, Swedish: broder|
|*méd?u||"honey"||Sanskrit: mádhu, Homeric Greek (methu), Lithuanian: medus, Russian: (mjod), Serbo-Croatian: (med), Polish: miód||*d? > d [d ~ ð]||*meduz||English: mead, East Frisian: meede, Dutch: mede, German: Met, Danish, Norwegian: mjød, Icelandic: mjöður , Swedish: mjöd|
|*steyg?-||"walk, step"||Sanskrit: stighnoti, Ancient Greek? (steíkhein)||*g? > g [? ~ ?]||*st?gan?||Old English: st?gan, Dutch: stijgen, German: steigen, Icelandic, Faroese: stíga, Danish, Norwegian: stige, Gothic steigan (all meaning "ascend, climb")|
|*ans-||"goose"||Latin: anser < *hanser, Ancient Greek: (kh?n), Sanskrit: hamsa ("swan"), Lithuanian: sis (older ?ansis), Russian (gus'), Persian: (?az), Serbo-Croatian? (guska), Polish: g||*g? > g [? ~ ?]||*gans-||English: goose, West Frisian: goes, guos, Dutch: gans, German: Gans, Icelandic: gæs, Faroese: gás, Danish, Norwegian, Swedish: gås|
|*seng-||"sing"||Homeric Greek: ? (omph?) "voice"||*g > gw 
|*singwan?||English: sing, West Frisian: sjonge, Dutch: zingen, German: singen, Gothic: siggwan, Old Icelandic: syngva, syngja, Icelandic, Faroese: syngja, Swedish: sjunga, Danish: synge, sjunge|
This process appears strikingly regular. Each phase involves one single change which applies equally to the labials (p, b, b?, f) and their equivalent dentals (t, d, d?, þ), velars (k, g, g?, h) and rounded velars (k?, g?, g, h?). The first phase left the phoneme repertoire of the language without voiceless stops, the second phase filled this gap, but created a new one, and so on until the chain had run its course.
When two obstruents occurred in a pair, the first was changed according to Grimm's law, if possible, while the second was not. If either of the two was voiceless, the whole cluster was devoiced, and the first obstruent also lost its labialisation, if it was present.
Most examples of this occurred with obstruents preceded by *s (resulting in *sp, *st, *sk, *sk?), or obstruents followed by *t (giving *ft, *ss, *ht, *ht) or *s (giving *fs, *ss, *hs, *hs). The latter change was frequent in suffixes, and became a phonotactic restriction known as the Germanic spirant law. This rule remained productive throughout the Proto-Germanic period. The cluster *tt became *ss (as in many Indo-European daughter languages), but this was often restored analogically to *st later on.
Examples with preceding *s:
|Non-Germanic examples||Change||Germanic examples|
|Latin: spuere, Lithuanian: spjáuti||*sp||English: spew, West Frisian: spije, Dutch: spuwen, German: speien, Danish, Norwegian, Swedish: spy, Icelandic: spýja, Faroese: spýggja, Gothic: speiwan|
|Latin: st?re, Irish: stad, Sanskrit: sta, Russian? (stat'), Lithuanian: stoti, Persian: ? (istâdan)||*st||English: stand, Icelandic, Faroese, Norwegian: standa, Gothic: standan; West Frisian: stean, Dutch: staan, German: stehen, Danish, Swedish: stå|
|Lithuanian: skurdus||*sk||English: short, Old High German: scurz, Icelandic: skorta|
|Irish: scéal||*sk?||English: scold, Icelandic: skáld, Norwegian: skald; West Frisian: skelle, Dutch: schelden, German: schelten|
Examples with following *t:
|Non-Germanic examples||Change||Germanic examples|
|Ancient Greek (klept?s), Old Prussian: au-klipts "hidden"||*pt->ft||Gothic: hliftus "thief"|
|Latin: atta, Greek (átta)||*tt->tt||Old High German: atto, Gothic: atta "father"|
|Ancient Greek (okt?), Irish: ocht, Latin: oct?||*kt->ht||English: eight, West Frisian, Dutch, German: acht, Gothic: ahtáu, Icelandic: átta|
|Irish: anocht, Latin: nox, noct-, Greek: , ?- (núks, nukt-), Sanskrit: (naktam), Lithuanian: naktis, Hittite (genitive): nekuz (pronounced /nek?ts/)||*k?t->ht||English: night, West Frisian, Dutch, German: Nacht, Gothic: nahts, Icelandic: nótt|
The Germanic "sound laws", combined with regular changes reconstructed for other Indo-European languages, allow one to define the expected sound correspondences between different branches of the family. For example, Germanic (word-initial) *b- corresponds regularly to Latin *f-, Greek p?-, Sanskrit b?-, Slavic, Baltic or Celtic b-, etc., while Germanic *f- corresponds to Latin, Greek, Sanskrit, Slavic and Baltic p- and to zero (no initial consonant) in Celtic. The former set goes back to PIE *b?- (faithfully reflected in Sanskrit and modified in various ways elsewhere), and the latter set to PIE *p- (shifted in Germanic, lost in Celtic, but preserved in the other groups mentioned here).
One of the more conspicuous present surface correspondences is the English digraph wh and the corresponding Latin and Romance digraph qu, notably found in interrogative words (wh-words) such as the five Ws. These both come from k?. The present pronunciations have undergone further sound changes, such as wh-cluster reductions in many varieties of English, though the spellings reflect the history more; see Interrogative word: Etymology for details.