In historical linguistics, the High German consonant shift or Second Germanic consonant shift was a phonological development (sound change) which took place in the southern parts of the West Germanic dialect continuum in several phases, probably beginning between the 3rd and 5th centuries AD, and was almost complete before the earliest written records in the High German language were made in the 9th century. The resulting language, Old High German, can neatly be contrasted with the other continental West Germanic languages, which mostly did not experience the shift, and with Old English, which was completely unaffected.

General description
The effects of the shift are most obvious for the non-specialist when we compare Modern German lexemes containing shifted consonants with their Modern English or Dutch unshifted equivalents. The following overview table is arranged according to the original Proto-Indo-European phonemes. (G=Grimm's law; V=Verner's law)
(Notes: Old English fæder, "father"; English has shifted d→th in OE words ending in -der).

Overview table

The four phases in detail
The first phase, which affected the whole of the High German area, has been dated as early as the fourth century, though this is highly debated. The first certain examples of the shift are from Edictus Rothari (a. 643, oldest extant manuscript after 650). According to most scholars, the Pre-Old High German Runic inscriptions of about a. 600 show no convincing trace of the consonant shift. It saw the voiceless stops become geminated fricatives intervocalically, or single fricatives postvocalically in final position.
p→ff or final f
t→zz (later German ss) or final z (s)
k→hh (later German ch)
Note: In these OHG words, <z> stands for a voiceless fricative that is distinct somehow from <s>. The exact nature of the distinction is unknown; possibly <s> was apical while <z> was laminal.
Examples:
Old English slǣpan : Old High German slāfan (English sleep, Dutch slapen, German schlafen)
OE strǣt : OHG strāzza (English street, Dutch straat, German Straße)
OE rīce : OHG rīhhi (English rich, Dutch rijk, German reich)
Note that the first phase did not affect geminate stops in words like *appul "apple" or *katta "cat", nor did it affect stops after other consonants, as in words like *scarp "sharp" or *hert "heart", where another consonant falls between the vowel and the stop. These remained unshifted until the second phase.

Phase 1
The second phase, which was completed by the eighth century and concentrated on the Upper German area, saw the same sounds become affricates in initial position, when geminated, and when following a liquid consonant (l or r).
p→pf (also spelled <ph> in OHG; after a liquid this later became f)
t→tz (in Modern German often spelled <z> and pronounced /ts/)
k→kch (pronounced /kx/; this step has not been completed by standard German).
The Southern Austro-Bavarian dialects of Tyrol is the only dialect where the affricate /kx/ has developed in all positions. In High Alemannic, only the geminate has developed into an affricate, whereas in the other positions, /k/ has become /x/. However, there is initial /kx/ in modern High Alemannic as well, since it is used for any k in loanwords, for instance [kxariˈb̥ikx], and since /kx/ is a possible consonant cluster, for instance in Gchnorz [kxno(ː)rts] 'laborious work', from the verb chnorze.
Examples:
OE æppel : OHG aphul (English apple, Dutch appel, German Apfel)
OE scearp : OHG scarpf (English sharp, Dutch scherp, German scharf)
OE catt : OHG kazza (English cat, Dutch kat, German Katze)
OE tam : OHG zam (English tame, Dutch tam, German zahm)
OE liccian : OHG lecchōn (English to lick, Dutch likken, German lecken, High Alemannic schlecke/schläcke /ʃlɛkxə, ʃlækxə/)
OE weorc : OHG werk or werch (English work, Dutch werk, German Werk, High Alemannic Werch/Wärch)
In the following combinations, however, the shift did not take place: sp, st, sk, ft, ht, tr.
OE spearwa : OHG sparo (English sparrow, Dutch spreeuw, German Sperling)
OE mæst : OHG mast (English mast, Dutch mast, German Mast[baum])
OE niht : OHG naht (English night, Dutch nacht, German Nacht)
OE trēowe : OHG [ge]triuwi (English true, Dutch (ge) trouw, German treu "faithful")

Phase 2
The third phase, which had the most limited geographical range, saw the voiced stops become voiceless.
b→p
d→t
g→k
Of these, only the dental shift d→t finds its way into standard German. The others are restricted to Swiss German, and to Austrian and Bavarian dialects. This shift must have begun after the first and second phases ceased to be productive, or else the resulting voiceless stops would have shifted further to fricatives and affricatives. We are therefore thinking of the 8th or 9th century.
It is interesting that in those words in which an Indo-European voiceless stop became voiced as a result of Verner's law, phase three of the High German shift returns this to its original value (t→d→t):
PIE *māh₂ter- → Germanic *mōder → German Mutter
Examples:
OE dōn : OHG tuon (English do, Dutch doen, German tun)
OE mōdor : OHG muotar (English mother, Dutch moeder, German Mutter)
OE rēad : OHG rōt (English red, Dutch rood, German rot)
OE biddan : OHG bitten or pitten (English bid, Dutch bieden, German bitten, Bavarian pitten)
It is likely that pizza is an early Italian borrowing of OHG (Bavarian dialect) pizzo, a shifted variant of bizzo (German Bissen, "bite, snack").

Phase 3
Other consonant changes on the way from West Germanic to Old High German are included under the heading "High German consonant shift" by some scholars who see the term as a description of the whole context, but are exluded by others who use it to describe the neatness of the three-fold chain shift. Although it might be possible to see ð→d, ɣ→g and v→b as a similar group of three, both the chronology and the differing phonetic conditions under which these changes occur speak against such a grouping.

þ/ð→d (Phase 4)
The West Germanic voiced velar fricative /ɣ/ shifted to /g/ in Old High German in all positions. The same change occurred in Old English, but with the important exception that next to a front vowel it instead experienced Anglo-Saxon palatisation and became /j/. Dutch has retained the original Germanic /ɣ/, though as Dutch spells this with <g>, the difference between it and the English and German consonant is invisible in the written form.
Dutch goed (ɣu:t) : German gut, English good
Dutch gistern (ɣistern) : German gestern, English yesterday
This change is believed to be an early one. As the existence of a /g/ in the language was a prerequisite for the south German shift g→k, this must at least predate phase 3 of the core group of the High German consonant shift.

ɣ→g
West Germanic [v], which was an allophone of /f/ used in medial position, shifted to Old High German /b/ between two vowels, and also after /l/.
OE lufu : OHG liob (English love, Dutch lief, German Liebe)
OE hæfen : MHG habe (English haven, Dutch haven. For German Hafen see below)
OE half : OHG halb (English half, Dutch half, German halb)
OE lifer : OHG libara (English liver, Dutch lever, German Leber)
OE self : OHG selbo (English self, Dutch zelf, German selbe)
OE sealfian : OHG salbon (English salve, Dutch zalf, German Salbe)
In strong verbs such as German heben (heave) and geben (give), the shift contributed to elimiating the /v/ forms in German, but a full account of these verbs is complicated by the effects of grammatischer Wechsel by which [v] and /b/ appear in alternation in different parts of the same verb in the early forms of the languages. In the case of weak verbs such as haben (have, Dutch hebben) and leben (live, Dutch leven), the consonant differences have an unrelated origin, being a result of the Germanic spirant law and a subsequent process of levelling.
NB: a similar v→b shift occurred in German between a vowel and /r/, but as a separate process much later, in early modern times: Middle High German varwe > German Farbe, colour, cf. Dutch verf.

v→b
High German experienced the shift /sp/, /st/, /sk/ → /ʃp/, /ʃt/, /ʃ/ in initial position:

German spinnen (/ʃp/), spin.
German Strasse (/ʃt/), street.
German Schrift, script.

s→ʃ
Other changes include a general tendency towards terminal devoicing in German and Dutch, and to a far more limited extent in English.

General
Since, apart from þ→d, the High German consonant shift took place before the beginning of writing of Old High German in the 9th century, the dating of the various phases is an uncertain business. The estimates quoted here are mostly taken from the dtv-Atlas zur deutschen Sprache (p. 63). Different estimates appear elsewhere, for example Waterman, who asserts that the first three phases occurred fairly close together and were complete in Alemannic territory by 600, taking another two or three centuries to spread north.
Sometimes historical constellations help us; for example, the fact that Attila is called Etzel in German proves that the second phase must have been productive after the Hunnish invasion of the 5th century. The fact that many Latin loan-words are shifted in German (e.g. Latin strata→German Straße), while others are not (e.g. Latin poena→German Pein) allows us to date the sound changes before or after the likely period of borrowing. However the most useful source of chronological data is German words cited in Latin texts of the late classical and early mediaeval period.
Precise dating would in any case be difficult since each shift may have begun with one word or a group of words in the speech of one locality, and gradually extended by lexical diffusion to all words with the same phonological pattern, and then over a longer period of time spread to wider geographical areas.
However, relative chronology for phases 2, 3 and 4 can easily be established by the observation that t→tz must precede d→t, which in turn must precede þ→d; otherwise words with an original þ could have undergone all three shifts and ended up as tz. The phenomenon that an early phase of a sound shift leaves a gap (in this case voiceless stops) which a later phase then fills by means of a chain shift is familiar enough; Grimm's law proceeds in a similar sequence. By contrast, as the form kepan for "give" is attested in Old Bavarian, showing both ɣ→g→k and v→b→p, it follows that ɣ→g and v→b must predate phase 3.
Alternative chronologies have been proposed. According to a not widely accepted theory by the German linguist Theo Vennemann, the consonant shift occurred much earlier and was already completed in the early 1st century BC. On this basis, he subdivides the Germanic languages into High Germanic and Low Germanic.

Chronology
Roughly, one may say that the changes resulting from phase 1 affected Upper and Central German, those from phase 2 and 3 only Upper German, and those from phase 4 the entire German and Dutch-speaking region. The generally-accepted boundary between Central and Low German, the maken-machen line, is sometimes called the Benrath line, as it passes through the Düsseldorf suburb of Benrath, while the main boundary between Central and Upper German, the Appel-Apfel line can be called the Speyer line, as it passes near the town of Speyer, some 200 kilometers further south.
However, a precise description of the geographical extent of the changes is far more complex. Not only do the individual sound shifts within a phase vary in their distribution (phase 3, for example, partly affects the whole of Upper German and partly only the southernmost dialects within Upper German), but there are even slight variations from word to word in the distribution of the same consonant shift. For example, the ik-ich line lies further north than the maken-machen line, although both demonstrate the same shift /k/→/x/. Furthermore, the exact line can move over a period of time. Since German reunification, a northward movement of the eastern end of the Benrath line has been observed.
The subdivision of West Central German into a series of dialects according to the differing extent of the phase 1 shifts is particularly pronounced. This is known in German as the Rheinischer Fächer ("Rhenish fan"), because on the map of dialect boundaries the lines form a fan shape. Here, no fewer than eight isoglosses run roughly West to East, partially merging into a simpler system of boundaries in East Central German. The table on the right lists these isoglosses (bold) and the main resulting dialects (italics), arranged from north to south.
For a map of the boundaries of a number of key sounds, see a general map and the Rheinischer Fächer.

Geographical distribution
Some of the consonant shifts resulting from the second and third phases appear also to be observable in Lombardic, the early mediaeval Germanic language of northern Italy, which is preserved in runic fragments of the late 6th and early 7th centuries. Unfortunately, the Lombardic records are not sufficient to allow a complete taxonomy of the language. It is therefore uncertain whether the language experienced the full shift or merely sporadic reflexes, but b→p is clearly attested. This may mean that the shift began in Italy, or that it spread southwards as well as northwards. Ernst Schwarz and others have suggested that the shift occurred in German as a result of contacts with Lombardic. If in fact there is a relationship here, the evidence of Lombardic would force us to conclude that the third phase must have begun by the late 6th century, rather earlier than most estimates, but this would not necessarily require that it had spread to German so early.
If, as some scholars believe, Lombardic was an East Germanic language and not part of the German language dialect continuum, it is possible that parallel shifts took place independently in German and Lombardic. However the extant words in Lombardic show clear relations to Bavarian. Therefore Werner Betz and others prefer to treat Lombardic as an Old High German dialect. There were close connections between Lombards and Proto-Bavarians: the Lombards settled until 568 in 'Tullner Feld' (about 50 km west of Vienna); some Lombard graves (excavated a few years ago when a new railway line was built) date after 568; evidently not all Lombards went to Italy in 568. The rest seem to have become part of the then newly formed Bavarian groups.
When Columban came to the Alamanni at Lake Constance shortly after 600, he made barrels burst, called cupa (English cup, German Kufe), according to Jonas of Bobbio (before 650) in Lombardy. This shows that in the time of Columban the shift from p to f had occurred neither in Alemannic nor in Lombardic. But Edictus Rothari (643; extant manuscript after 650; see above) attests the forms grapworf (throwing a corpse out of the grave, German Wurf and Grab), marhworf (a horse, OHG marh, throws the rider off), and many similar shifted examples. So it is best to see the consonant shift as a common Lombardic — Bavarian — Alemannic shift between 620 and 640, when these tribes had plenty of contact.

Sample texts
The High German consonant shift — at least as far as the core group of changes is concerned — is an example of a sound change which permits no exceptions, and was frequently cited as such by the Neogrammarians. However, modern standard German, though based on Central German, draws vocabulary from all German dialects. When a native German word (as opposed to a loan word) contains consonants unaffected by the shift, they are usually explained as being Low German forms. Either the shifted form has fallen out of use, as in:
Hafen (harbour): Middle High German had the shifted form habe(n), but the Low German form replaced it in modern times.
or the two forms remain side-by-side, as in:
Wappen (coat of arms): the shifted form also exists, but with a different meaning: Waffen (weapon)
Further examples of common German words in this category include:
Lippe (lip); Pegel (water level); Pickel (pimple)
However, the vast majority of words in Modern German which contain consonant patterns which would have been eliminated by the shift are loaned from Latin or Romance languages, English or Slavic:
Paar (few), Ratte (rat), Peitsche (whip).