When writing arrives, we find Slavic, Germanic, Celtic, as discrete language groups. Since they undoubtedly started out as being the same, the question is what breaches of structure and origins of structure brought about the divisions. If you look at a page of German and a page of Russian, as they are today, they are certainly very different. Since there are three Indo-European groups in Northern Europe, we are entitled to ask why it is exactly three, and not one or eight. Peter Schrijver, a professor at Utrecht, has published Language contact and the origins of the Germanic languages (2014). This contains a set of radically original theories on the influence of bilingualism and phonetic interference, which I expect we will be debating for the next hundred years. For him, the breach between Indo-European and Germanic is the impact of a phase of bilingualism in which a Balto-Finnic language influenced local Indo-european.
Indo-European has no double-length (geminate) consonants but Germanic does have them. This is an innovation. Schrijver describes in Finnish a set of alternations in consonants depending on the level of stress of the preceding syllable. This can affect which consonant is heard, but is often also an alternation between a sound and zero. Finnish has in fact two sets of rules affecting consonants, rhythmic grading and syllabic grading. (The first affects odd syllables, so first and third ones.) Thus it has a rich set of variations in the consonants of syllables other than the first one. Schrijver takes this arrangement and points out that in Germanic we have some alternations which can be explained as grading:
*dūb- Old Norse dúfa ‘to immerse’
*dubb- Norwegian dubba ‘to stoop’, Middle Dutch dubben ‘to immerse’
*dūp- Dutch duipen ‘to hang one’s head’
*dupp- High German düppen, Norwegian duppa ‘to nod’
*duff- Faeroese duffa ‘to bob up and down (of a ship)’
*dump- Norwegian, English, Danish dump ‘hole, pit, pond’ East Frisian dumpen ‘to dive’
This word-group will be familiar to those who read substrate studies, in the journal NOWELE and elsewhere. (If I am not mistaken, this table was compiled by Frans Kuiper.) It is popular in discussions of the so-called “geminates language” and some of the words show geminates. The study of it was rather frustrated by the fact that it was definitely an Indo-European root (meaning “deep”, and instanced in Sanskrit) although its final consonant behaved in ways which were, strikingly, non-Indo-European. Schrijver has offered a good explanation for this – Germanic arose as the product of Indo-European being spoken, 3000 years ago, by a population which had Balto-Finnic speech habits. That means they had “old stock” words which were, however, subject to Balto-Finnic patterns of variation as concerned consonants at the end of syllables.
The other “geminate” group which keeps cropping up is (in English) stub-stump – stock (and, probably, “stem”). I don’t have a view on the rather puzzling variations within this group, but consonantal gradation seems to offer a solution.
The variation in consonants in Finnish is correlated with syllabic stress, and there is a biomechanical reason why unstressed syllables coincide with weakened consonants. This means also that vowel gradation would be, in an Indo-European language, occurring in the same pattern, so that different ablaut grades would appear with different consonant grades, in words from the same stem.
It is not instantly clear to me why the same word, in the same case, should end up in different (but related) languages with different grades and conditioned by different syllabic grades. I do not get why the words would not all have the same consonantal grade. However, this is a genuinely new idea and I feel that it will open up new routes for Germanic etymology. As appears from the back catalogue of substrate studies, this class of words is rather small. (Although one scholar claims the total is several hundred words, while denying that there is any substrate influence involved in the first place.) I would like to mention at this point the word dumble. This refers to the low ground around a stream and is used locally to me – for example, Lambley Dumbles, two miles away from me. It appears that this means deep in the sense of a hollow stream, which has worn a deep bed between high banks. (Or, low-lying pasture which is flooded seasonally.) Thus dumble actually means deep (or low). Dump is originally a pit – again, it means a “deep place”, and originally often meant a pit in a river bed, a patch of deep water. The contrast in the word ending between dump and dumble is the kind of thing which Schrijver is talking about.
Schrijver’s book also deals with Insular Celtic languages, as part of the structure within which he explains the shifts which differentiate Old English from closely related Continental dialects of Germanic. He states that Irish was identical with proto-Welsh as late as 150 AD and was the speech of migrants from Britain to Ireland at that date. “it seems safe to say that an Irish arrival in Ireland close to or in the first century AD is much easier to unite with the linguistic evidence than an arrival around, say, 500 or 1000 BC.”
Thus he posits Gaelic as the product of a late migration from Britain. As follows, the language which the Anglo-Saxons encountered, on disembarking, resembled Gaelic (as well as Welsh). But recently, Ranko Matasovic posited that there was a period of bilingualism around 400-600, connected with Irish invasions of Western Britain, which produced an assimilation between the two languages, so that the most obvious shared features between the languages (lenition and compound prepositions) are actually late, and their separation was much earlier. I am having difficulty balancing these two views. Matasovic does not propose a social mechanism for the spread of these features outward from the bilingual zone, on the Western shores of Britain, to the whole of Ireland (and the whole of Wales and Cornwall). It is certain that these features are not part of the inheritance of the Insular languages, so it must follow that the two languages innovated in parallel – attaching this to a bilingual zone is perfect. Since we actually have a credible bilingual zone, connected with Irish settlements on the “yonder” side of the Irish Sea, the theory is rather robust.
I am attracted to elements of Schrijver’s idea. If you look at the map, it seems unlikely that anyone would colonise Ireland directly from Spain, Aquitaine, Normandy, Holland, etc. Further, the Atlantic is a large and stormy sea. But, migration from England to Ireland looks easy, even for people with limited sea technology. Meanwhile, we have exactly two Celtic languages in the British Isles. It is attractive to link this duality with two large islands, separated by the Irish Sea. So, a pattern in which speakers of an ancient Celtic language started in Gaul, colonised Britain, and then colonised Ireland from there, and their language community split into two parts, separated by the sea, and the two parts evolved from unity into exactly two languages, sounds pleasing even if it is hard to find concrete evidence that it was like that.
Schrijver’s method is based on the shapes of phoneme structures. This reduces any language to 30 or 40 phonetic elements, which occur in rows and can be conveniently recorded or memorised. The sound which change in any period of history are even fewer. Patterns can be recognised in these simple datasets which are distinctive and rather objective. He relies absolutely on this method. Of course language contact also shows up in vocabulary, but he barely uses this as evidence. So, someone else will have the task of seeking the postulated chronologies in vocabulary, and seeing whether the ideas are confirmed by it. Of course vocabulary is very extensive and demands much more time and more pages.
He states that Irish is uniform in early manuscripts, whereas Welsh is dialectally differentiated. This is a basis for arguing that Irish was homogeneous in 600 AD (when manuscripts start), and so a recent arrival from overseas. I am very surprised to hear that there are dialect differences in the scant records of mediaeval Welsh, so far as I know the opposite is true and there is a “national” language for literature which blanks out any regional differences. To be sure, Wales is not all that large a place. The poems supposedly written by Taliesin and Aneirin, in a sub-Roman 6th century, have come down to us in a form indistinguishable from standard Welsh, although archaic; linguistic differences would be pure gold, to a scholar, but the editors of these texts do not point to any such differences. Of course we only have late and normalised manuscripts. No, my suspicion is that Old Irish is standardised because scribes were taught to apply rigid standards, and this is connected with the wish of the Church to see holy texts transmitted in a way which did not even suggest that human fallibility applied to them. They learnt to write Latin without personal variation and Irish without such variation. Of course literacy spread outwards from the Church, which initially used writing for Latin texts. It does not follow from consistency in spelling and vocabulary that Ireland was a linguistically uniform space.