A language spoken in prehistoric northern Europe, which gave rise to the Germanic language family, including English, Dutch, German, Icelandic, Danish, Swedish, and Norwegian. It was never written, and is reconstructed by linguists from its descendants and by comparison with more distant branches of the Indo-European family, such as Latin, Greek, and Sanskrit. It is sometimes called Primitive Germanic.

By about the first century it was splitting into three branches. These might represent dialects within Proto-Germanic. They are the North, East, and West Germanic. North Germanic has the earliest writing, a few runic inscriptions from the 300s; this branch gave rise to Old Norse, the language of the sagas, and to the Scandinavian languages. East Germanic is also early, in the form of the Gothic bible, which however was written in what is now Bulgaria, the Goths having migrated far and wide from their original home in Scandinavia. Gothic and the East Germanic branch are extinct. West Germanic includes English and German; some Old English texts date from the 600s. Between them we can work out what their common ancestor was like.

The main phonetic difference between Proto-Germanic and other Indo-European languages is the change known as Grimm's Law, stated by Jacob Grimm in 1822. Where Latin or Greek has d, Proto-Germanic had t; where Latin or Greek has t, Proto-Germanic had th, and so on through a whole range of other consonants -- read about it in Grimm's Law. One example is tooth compared to Latin dent-, Greek odont-.

In some cases Grimm's Law didn't apply. Some of these made sense easily—after s the change didn't take place, e.g. stand = Latin sto—, but other exceptions had to wait till 1877, when Verner's Law was proposed. This showed that the Grimm's Law change was altered by a subsequent one if the accent was on the following syllable. This discovery gave a depth of history to Proto-Germanic. In Proto-Indo-European, as also in early Greek and Sanskrit, and still in a few modern languages like Latvian and Slovenian, accent was a variable pitch that could be on any syllable. It was clear that this was true of early Proto-Germanic, and allowed Verner's Law to apply. Then, another major phonetic shift occurred: Proto-Germanic took on a stress accent on the first syllable. This is generally the case with modern Germanic languages.

One possible influence for this is Uralic languages, its neighbours up there in the north. The most familiar Uralic language is Finnish. They have initial stress. There was certainly interplay between Germanic and Uralic at a very early period. Finnish borrowed words like kuningas 'king' and rengas 'ring'. These still contain the ancient ending -az (there was no Z in Finnish so they borrowed it as S). This ending corresponds to the familiar Latin -us and Greek -os. In Gothic it had lost its vowel and become -s, e.g. fisks 'fish'. In Old Norse it had changed to -r, e.g. konungr. Only in the earliest runic inscriptions do we see the original ending. It disappeared in all other branches, e.g. Old English cyning, German König.

The change of short O to A, illustrated in that ending, is the most prominent vowel change between Proto-Indo-European and Proto-Germanic: Latin octo = German acht 'eight'. But confusingly, long A changed to O, e.g. Latin frater = brother.

Grammatically, Proto-Germanic reduced the Indo-European tenses to just two, present and past. It then innovated an entirely new series of so-called "weak" past tenses, with a d ending, of unknown origin. This is where we get our walk - walked. Elaborations like making compound tenses with "will", "have" etc. arose later.

A huge amount of Germanic vocabulary is of unknown origin. As much as one third of it hasn't got any cognate in other Indo-European branches, including very common words like sea, earth, blood, hand, evil, little, sick, bring, run. In the prehistoric period in (perhaps) Scandinavia when Proto-Germanic was formed, the speakers' neighbours were Saami and Finns, maybe some Slavs, to the north and east, and Celts to the south, but none of those language families explains the huge amount of new material. There must have been some substrate language which has completely disappeared from history. Some people have suggested calling this Folkish, because the word folk is one of the unexplained ones.

Note that German is not especially close to the ancestor; English is in no way descended from German (Deutsch). They are parallel descendants in the Germanic group (German name Germanisch, which means more like Teutonic).

Much of this refreshed in my mind from: W.B. Lockwood, Indo-European Philology, Hutchinson, 1969.

There's an interesting discussion of the original homeland and date of split (neither of which are known with any certainty) at:
http://www.ling.su.se/staff/oesten/papers/Theorigin.pdf