The emergence of Germanic

« previous post | next post »

From their origins to the present day, speakers of Germanic languages have been distinguished by the high degree of their mobility on land and on water:  the Völkerwanderung during the Migration Period, Goths, Vikings, the British Empire on which the sun never set, Pax Americana….  From antiquity, they ranged far and wide, so it is not surprising to see them popping up all over the place and, in their travels, to come in contact with an enormous number of different ethnic and linguistic groups.

Before setting out on their multitudinous journeys, they had to have begun somewhere, and — on the borders of their original homeland — they had to have been in contact with other ethnic and linguistic groups.  I asked a colleague where and when they might have arisen, and who their neighbors were.

He replied:

The usual date for Proto-Germanic / first sound shift is c 500 BCE. Germanic was, at one time, probably connected with proto-Balto-Slavic but also may have come into contact with Italo-Celtic; if you look on Don Ringe's trees, it has always been a problem with the possibility of multiple ancestral relationships.

Ringe's trees may be found here:

Don Ringe [principal author], Tandy Warnow, and Ann Taylor, "Indo-European and computational cladistics", Transactions of the Philological Society, 100 (2002), 59-129.

And Ringe replied to my questions about the origins and associations of the Proto-Germans thus:

That's hard to say.  One of the things you have to bear in mind is that PGmc., as reconstructed from the well-attested daughters, is a single dialect with little internal variation; you'll see how that's relevant below.

There's a consensus that PGmc. belongs in the Jastorf culture and its successors in southern Denmark and northern Germany.  But early Jastorf, ca. 750 BCE, is too early for our reconstructed PGmc., given how similar Gothic (ca. 350 CE) and Early Runic (roughly contemporary) still are.  Later Jastorf and its successors occupy too large an area for its population to have spoken anything like a single dialect (see above). The only possible solution is that PGmc. was *one* of the dialects spoken in the Jastorf area, probably not before ca. 500 BCE, possibly a bit later.  It follows that most of the tribes identified by the Romans as Germans spoke not our reconstructed PGmc. but closely related sister dialects that have left no descendants.

There's good evidence for borrowing of Celtic words into (pre-)PGmc., and Celtic and Gmc. also share a good deal of vocabulary which might be loanwords or might be common inheritances.  It seems clear that the Germanic tribes got some technology (e.g., iron) and some cultural items (e.g. *ri:k- 'king, ruler', from Proto-Celtic *ri:g-, with the distinctive Celtic shift of long *e: to *i:–cf. Lat. re:g-) from the Celts. There are also shared items between Gmc. and Balto-Slavic, but fewer, and there are definitely some early Gmc. loans in Baltic, Slavic, and (non-IE) Baltic Finnic.  In other words, the ancient linguistic map corresponds roughly to the historical one, except that there must have been population groups that were absorbed and languages that died out–including all the near sisters of our reconstructed PGmc.

For extended treatment of these topics, see Donald Ringe, From Proto-Indo-European to Proto-Germanic, 2nd ed. (Oxford:  Oxford University Press, 2017).

The closeness of Celtic and Germanic is important in terms of patterns of cultural dispersal that I observed decades ago during my mummies research project.  The parallels with the dispersion of Indo-Iranian peoples, languages, and cultures are unmistakable.  Think where the latter began and where they have ended up.  People who have horses, wheeled vehicles, and herded animals, and who are adept with watercraft do not sit still.  By their very nature, they are highly mobile, and where they go, so go their languages.


  1. Andrew Usher said,

    February 28, 2019 @ 8:40 am

    These are indeed interesting questions but summarising them by asking when or where was 'Proto-Germanic' badly misses the mark. 'Proto-Germanic' is a reconstructed 'average' dialect, and it would be a miracle if it actually matched (to within the accuracy of reconstruction) any one dialect actually even spoken. It is true and important to know that most languages have left no descendants (and, before writing, most often no discernible trace) e.g. the modern Romance languages alone would not support postulating any Italic language not descended from Latin; only direct evidence shows us. This is where language evolution most closely parallels biological evolution.

    So you can rightly ask when the Germanic sound shift started, or when the last common ancestor of Gmc. and any other living IE language was, but 'when was PGmc.' is an ignorant question that obscures, rather than enlightens, the history of the Germanic-speaking peoples.

    k_over_hbarc at yahoo dot com

  2. David Marjanović said,

    February 28, 2019 @ 6:22 pm

    So you can rightly ask when the Germanic sound shift started, or when the last common ancestor of Gmc. and any other living IE language was, but 'when was PGmc.' is an ignorant question

    Am I understanding this right? Are you saying that asking when the last common ancestor of Gmc. and any other living IE branch was spoken is not an ignorant question, but that asking when the last common ancestor of the attested* Gmc. languages was spoken is an ignorant question?

    * The last common ancestor of the living Gmc. languages was Proto-Northwest-Germanic, which was not ancestral to the extinct East Germanic branch. The one identifiable and attested East Germanic language is Gothic.

  3. Nelson Goering said,

    March 1, 2019 @ 3:39 am

    I have to agree with David Marjanović – asking when Proto-Germanic proper (the 'most recent common ancestor') was is not ignorant or useless, and is in fact a very interesting question. We might face any number of difficulties in trying to answer it, but it's still a good thing to ask.

    We can also ask when Grimm's Law occurred, but I think it's important to stress that this is a completely different question. For much of the twentieth century, there seems to have been this peculiar idea that Grimm's Law 'defines' Germanic, and that 'Proto-Germanic' is some sort of phase lasting from then until the breakup into daughter branches. This is not a very useful way of looking at things. Grimm's Law was one of very many innovations characterizing the Germanic branch, and it was neither the first nor the last. It was a very important change, certainly, but there's no rigorous sense in which it 'defines' Germanic: Germanic had already begun to diverge as a distinct branch of IE before that, and there were a number of further changes which needed to follow after before we reach 'Proto-Germanic proper' (i.e. the MRCA). The tree in Don Ringe's book shows one possible relative chronology of Germanic innovations, and you can see there just how embedded GL is (I can't check the page reference right now).

    As for the actual absolute dating of GL, I suspect the 500 BC guess is much too late. Not that sound changes occur at an even pace, but on the whole it does seem to be one of the earlier Germanic sound changes rather than later (again, it's worth looking at Ringe's table to get one sense of this). Saying 1500 BC would be just as arbitrary, but maybe a little more likely to be near the mark.

    The only other thing I'd mention is that linguistic phylogeny is probably not of much interest here. All branches of IE diverged early on, and any truly phylogenetic relationships probably just reflect the very earliest patterns of dialectal innovation. Later relationships (now technically of a contact sort, even if between closely-related dialects) quite possibly cross-cut that. This is obvious with much later similarities, such as those with Celtic, but probably equally true of potentially earlier similarities with Baltic (and Slavic). Those interested in the external histories of prehistoric languages often point to linguistic trees, but this is probably only useful in those contexts where it's clear that a given subgroup developed for a good while before in turn diverging again – e.g. with Indo-Iranian within IE. With Germanic, we're not going to get very much out of its phylogenetic status.

  4. maidhc said,

    March 1, 2019 @ 6:14 am

    I happened upon this documentary which is somewhat relevant to the discussion:

    It is more about culture than about language. For the experts here I'm sure it is elementary stuff that they would not deign to look at. But I thought it was rather interesting, although I have have some reservation about a few of the things that were said relative to some books I have. For those of us who are striving to keep up, it might be of interest.

    Unfortunately it has advertisements, you will have to click through.

  5. David Marjanović said,

    March 2, 2019 @ 1:25 pm

    Saying 1500 BC would be just as arbitrary, but maybe a little more likely to be near the mark.

    Most, possibly all, of the Celtic loanwords that were already present in Proto-Germanic have – where applicable – undergone Grimm's law. This argues for a late date, because contact between speakers of Celtic and Germanic continued into Roman times.

    When contact between Celtic and Germanic began, on the other hand, is a difficult question… at one extreme, they could have formed next to each other; at the other extreme, it has been suggested that the Hallstatt culture spoke Italic, and Celtic only expanded into the vicinity of Germanic with the La Tène culture.

    (Italic was clearly spoken north of the Alps at some point; there are a few Italic loanwords in Proto-Slavic.)

    Unfortunately it has advertisements

    Does it? Firefox with Adblock Plus leaves me with no way of knowing. :-)

    The video is a bit confused in some places, a bit out of date in others, but generally good. Perhaps the most important criticism I have is that riding is a very recent invention, distinctly younger than the chariot.

  6. Andrew Usher said,

    March 2, 2019 @ 4:23 pm

    I was I suppose a bit too strong in apparently condemning the search for Proto-Germanic; yes, if you define it as the most recent common ancestor it is fine. But such questions are often conflated with others (like Grimm's law, here) and improper assumptions. We must remember that that Proto-Germanic and its general line of descent coexisted at all times with other Germanic-like dialects that have died out, some of which probably never underwent Grimm's law at all, making the loanword evidence questionable.

    The time period of evolution of the peculiar features of Germanic, that is, between the two common ancestors mentioned, was probably no less than two thousand years, of which there is very little evidence of linguistic value. Believing that our best reconstruction exactly captures any one moment during that time is rather a stretch, as I pointed out; it must not be assumed that the PGmc. we reconstruct _is_ the common ancestor.

    As to the last, while riding horses _in battle_ is post-Bronze Age (and perhaps of questionable worth at any time), I think riding in general is older, and probably (assuming the usual dating of PIE) common Indo-European.

  7. Nelson Goering said,

    March 3, 2019 @ 3:58 am

    The evidence of Grimm's Law certainly does not point unambiguously to a largely pre-GL stratum of _borrowing_. There are a few examples which may show GL operating on Celtic loans, such as *rīks (unless this was adapted from the Celtic nominative, or underwent a sound-substitution, if Germanic *k was felt to be closer than *ɣ in a post-GL phonology), *tūna- (assuming the PIE source really is with *dʰ-, rather than this being a common adaptation from a non-IE source), and *aiþa- (though this could be an early parallel formation with a morphologically-transparent suffix rather than a strict loan; there are a fairly large number of words of this sort, many of which do clearly show GL, but whose status as loans is _highly_ uncertain). Others are ambiguous, and can only predate GL _if_ they were also at an early enough date that (pre?) Celtic retained the PIE breathy-voiced stops: *gīsla-, *rīdan-, *brunjōⁿ. If PGmc *krumba- is a borrowing from Celtic (which is not sure), then this clearly postdates GL, though there aren't too many examples of this sort. This is very difficult to really pin down, since only a very few words actually show any distinctively Celtic innovations which could secure their status as Celtic borrowings into Germanic (rather than shared vocabulary from any number of other causes: early Western IE dialectal area developments; mutual borrowings from a lost third source, which could be non-IE or a now-lost branch of IE; shared archaisms whose retention was aided by an early shared cultural system).

    Off the top of my head, the testimony of the assuredly-Celtic loans is mixed. We have *rīks, which _could_ show GL (though this isn't as secure as people sometimes make out), and *ambahta-, which probably does not (unless, again, it was early enough to have been borrowed with *bʰ).

    A further complication, of course, is that GL itself was not a single change. Most likely, the phases conventionally labelled GL 1 and GL 3 are rather older than GL 2. How much older, we don't know. In between these two phases we should probably place both Verner's and Kluge's Laws.

    The upshot of this is that I wouldn't place much faith in the use of Celtic for dating GL. I mean, it can maybe rule out putting GL in the very earliest stratum of pre-Germanic in Europe, but we could probably assume that anyway. The Celtic evidence carries very little probabilistic weight for distinguishing between 1500 BC and 500 BC (say).

    (It's too bad the word for 'iron' doesn't contain an obstruent stop.)

  8. David Marjanović said,

    March 3, 2019 @ 2:51 pm

    As to the last, while riding horses _in battle_ is post-Bronze Age (and perhaps of questionable worth at any time), I think riding in general is older, and probably (assuming the usual dating of PIE) common Indo-European.

    Riding, for any purpose, is mentioned a single time in the Rgveda, and that, IIRC, in one of the youngest songs. It's all chariots all the time.

    if Germanic *k was felt to be closer than *ɣ in a post-GL phonology

    That's a very good point; it's actually strange that I've never seen it mentioned before.

    Most likely, the phases conventionally labelled GL 1 and GL 3 are rather older than GL 2.

    I've seen them numbered in two different ways, so I'm not sure which you mean… I prefer putting aspiration first, then devoicing, and then the aspirates (voiceless and voiced) become fricatives (with positional exceptions for the voiced ones). In any case, a thorough, recent discussion of the relative dating of the parts of Grimm's, Verner's, Kluge's and what is doubtless going to be called Kümmel's laws can be found in these slides presented at a conference in 2014.

    (It's too bad the word for 'iron' doesn't contain an obstruent stop.)

    It does contain a *s, which underwent Verner's law. But, as you point out, dating Verner's in relation to the parts of Grimm's is not trivial.

  9. David Marjanović said,

    March 3, 2019 @ 2:53 pm

    (Kümmel presented his laws here. And of course the *s of iron isn't a stop, I only pointed it out because it might contribute to dating Verner's law.)

  10. Nelson Goering said,

    March 3, 2019 @ 4:53 pm

    I'll take a look at Kümmel's slides tomorrow – I think I know what scenario he favours, but I should refresh myself on the details.

    The conventional numbering is:

    GL 1: *T > *Þ
    GL 2: *D > *T
    GL 3: *Dʰ > *D/Đ

    I prefer the following chronology:

    1) *T > *Tʰ (aspiration of voiceless stops)
    2) *Tʰ, *Dʰ > *Þ, *Đ (fricativization of aspirates and breathy voiced stops; not necessarily a single change)
    3) *Þ > *Đ when following a non-accented syllable (Verner's Law)
    4) *ɣʷ > *w word-medially (only supported in this point in the chronology by one word)
    5) *Đn, *Dn > *DD when following a non-accented syllable (Kluge's Law)
    6) *D > *T ('GL 2')

    Any hardening of *Đ > *D which was genuinely before the breakup of Proto-Germanic (probably word-initially for *ð and *β, but not *ɣ; probably after nasals for all stop types) followed after or was concurrent with 6. I've also got some thoughts on just what 6 actually meant in terms of phonological structure (since for a long time, in my view, we had a single stop series, so voicing would not have been contrastive). I'm also committed to the idea that *Dʰ was really a breathy-voiced series (though I'm also open to the idea that *D was preglottalized, which can also affect how we look at change 6).

RSS feed for comments on this post