The Question of a Hamito-Semitic Substratum in Insular Celtic and Celtic from the West

Steve Hewitt s.hewitt@unesco.org

https://www.academia.edu/6701071/The_Question_of_a_HamitoSemitic_Substratum_in_Insular_Celtic_and_Celtic_from_the_West

1.

Introduction

=m-

Striking structural resemblances have long been noted between the Insular Celtic languages (divided into Goedelic Irish Gaelic, Scottish Gaelic, Manx; and Brythonic Welsh, Breton, Cornish) and various Hamito-Semitic (more broadly, Afro-Asiatic) languages. Few of these traits appear to have been noteworthy of the now extinct Continental Celtic languages (Gaulish, Celtiberic, Lepontic, Galatian), and they are not generally found in other Indo-European languages. This fact has led some to raise the possibility of a Hamito-Semitic substratum in the Celtic Isles to explain their origin. Among the most prominent shared features are: VSO order; singular (apersonal) verbmarking with plural post-verbal lexical subjects; the Semiticists construct state a [HEAD [the-DEPENDENT]] genitive construction; conjugated prepostions; and oblique relatives with pronoun copies, each illustrated below in Breton and Arabic.

Breton VSO order e skrivas Yann AFF wrote.PRET John ul lizher a letter

Arabic kataba Yay risla wrote.PRET.M John letter John wrote a letter

Singular (apersonal) verb-marking with plural post-verbal lexical subjects e skrivas ar merched AFF wrote.PRET the girls katabat al-bant wrote.PRET.F the-girls The girls wrote

Construct state [HEAD [the-DEPENDENT ]] genitive construction ti ar roue house the king bait al-malik house the-king The house of the king, the kings house

Conjugated prepositions gant, ganin, ganit, ganta, ganti, ganeomp, ganeoch, gante with, with me, with you.SG, with him, with her with us, with you.PL, with them ma, ma, maak, maik, mauh, mah man, makum, mahum with, with me, with you.SG.M, with you.SG.F, with him, with her with us, with you.PL, with them

Oblique relatives with pronoun copies ar gwele a meus kousked enna the bed.M AFF I.have slept.PP in.it.M as-sarr al-ladh nimt fh the-bed.M the-REL.M I.slept in.it.M

The bed I slept in, the bed in which I slept

2.

Main authors

Morris Jones (1900) reviews startling similarities between Welsh and Egyptian, including (1900:625-6) the periphrastic conjugation be + preposition + verbal noun: In Welsh and Irish, although these languages retain many of the Aryan tenses, this construction is extremely common ... The three prepositions commonly used for this purpose in Egyptian are em in, er to, for, er above [= Loprieno (1995:80) m in, r towards, r on] indicating the present, future, and perfect respectively. These correspond in use with the Welsh prepositions yn in, am for, wedi after. He also notes the surprising parallels between Welsh yn and Egyptian em: (1) preposition in, (2) in + verbal noun = progressive, and (3) predicative and adverbializing in. Other similarities are noted with Berber, but no Semitic language is examined. He concludes (1900:639) that the resemblance
seems to involve an intimate connection of some kind between the two families of speech in the prehistoric period, though they are probably not actually cognate. It is with Hamitic, however, rather than Semitic, that Celtic syntax is in agreement; for, as we have seen, it agrees with Berber where the latter differs markedly from Arabic, as, for instance, in the shifting of the pronominal suffix from the verb to a preceding particle ... Is the influence of a Hamitic substratum to be discovered in the simultaneous development on the same analytic lines of French, Spanish, and Italian in their use of infixed and postfixed pronouns?

Wagners main work on the Insular Celtic / Hamito-Semitic question (usefully summarized by Gagnepain (1961)) is (1959), especially the third part, entitled The linguistic geography position of the Brittonic verb, with sections on Celtic, Berber, Basque, English and French as representatives of a North African Western European linguistic stratum, exemplified by the structure of the verb, The Berber verb system and Remarks on the Semitic verb system. He describes his basic position (see also 1981) as follows (1987:19-20):
Between the fourth and sixth centuries A.D. Insular Celtic suffered revolutionary changes As a result of it Insular Celtic developed features and grammatical categories hardly found in other Indo-European languages. They have, however, close parallels in Berber and Egyptian, the Hamitic languages of Northern Africa, as well as in Basque The linguistic structure of Insular Celtic compels me to assume that, long before the arrival of Celtic or Belgic tribes, these islands were populated by people, who spoke languages or dialects which, from the point of view of E. Lewys typology could be described as Hamito-Semitic, languages not necessarily connected with but of a similar type as Berber and Egyptian and, somewhat more distantly Hebrew and Arabic. For my latest position on this subject, cf. my articles of 1976 and 1982. When Celtic was adopted by pre-Celtic populations, the structure of their original language(s) began to impose itself on the language of the Celtic invaders. The result was a linguistic revolution which led to the making of the mediaeval and modern Celtic languages.

Hewitt (1985) was written before I became aware of the Pokorny-Wagner tradition; I assumed that the resemblances could only be typological. The similar features reviewed in that article include head-dependent typology, VSO~SVO word order (main and subordinate clauses), verb-subject agreement/non-agreement, collective/singulative, conjugated prepositions, expression of have, the construct state genitive, compound construct-state adjectives, double topic subject sentences, relatives (restrictive, non-restrictive, on prepositional objects, on possessives), the dummy sentential pronoun, and circumstantial subordinating and. The chief shared feature omitted is the verbal noun, most certainly historically speaking a verbal noun in Breton, but in modern Breton it behaves like an infinitive (accusative rather than possessive object pronouns).

Gensler (1993) examines 12 Insular Celtic or Hamito-Semitic languages plus a random sample of 58 other languages from all over the world. He identifies a set of 17 exotic structural features shared by Insular Celtic and Hamito-Semitic, but not common among languages worldwide. Assigning scores for each feature in each of the 70 languages, he concludes (1993:426):
On the basis of the sample used in this study, nothing remotely close to the full-blown Celtic/Hamito-Semitic [CHS] linguistic type recurs anywhere else in the world. The relatively few languages which are best matches actually rather poor matches are scattered all over the globe, from the West Coast of North America to the Caucasus and New Guinea. However, the continental average score for Africa is higher than for any other continent, and drops only slightly when the CHS languages Egyptian and Berber are omitted; West Africa scores especially well, and appears especially hospitable to several of the CHS features (adpositional periphrastic, word-initial change, kin terms, inter alia). Conversely, Europe has one of the lowest average scores, and when Welsh and Irish are excluded its score drops far below that of any other continent. Celtic is thus radically out of place in a European landscape, whereas the Hamito-Semitic languages simply intensify a structural trend seen over much of Africa. A weak form of the CHS type, then, would appear to have a natural home in Africa, in particular Northwest Africa. Within Afroasiatic, the highest-scoring languages are on the Mediterranean; scores fall away in every direction, but the Chadic language Hausa (in West Africa) scores much higher than Cushitic Afar (in East Africa). The diachronic evidence, too, argues that the (weak) CHS type is something quite old in Africa: the African and Arabian case studies all show stronger CHS-ness further back in time. All this, in conjunction with the blood-type agreement between the British Isles and Northwest Africa, argues for some sort of prehistoric scenario specifically linking these two regions.

While Gensler does not claim to have proved the Hamito-Semitic substratum hypothesis for Insular Celtic, he does appear to be saying: in the face of such statistical results, what else can it be? He takes great care with the weighting (0, , 1) of scores for individual features, but no account is taken of the centrality or frequency of each feature within a particular language, such that pervasive features such as the genitive construction (feature 8) have the same weight as more marginal ones like the idiomatic genitive kinship constructions (feature 17). Furthermore, the languages that score highest in Genslers table, thus most strongly exemplifying what he calls the Celtic/Hamito-Semitic type, are the Insular Celtic languages rather than the Hamito-Semitic languages, i.e. the languages that are said to have been influenced by the Hamito-Semitic substratum which purportedly gave rise to the type. Another drawback is that possible typological explanations for some of the features (cf. remarks on the genitive construction below) are not envisaged or examined. Nevertheless, with his thorough analysis and the sheer wealth of linguistic evidence he has marshalled, Gensler has certainly put the Hamito-Semitic substratum hypothesis back on the map; all authors on the subject will henceforth need to take due account of his arguments.

Jongeling (2000) provides an excellent, lengthy introduction to the history of the subject. The features he surveys include VSO, head-dependent order, numerals, nominal clauses, circumstantial subordinating and, relatives, the verbal noun, conjugated prepositions, and the lack of a verb have. He proposes (2000:149-50) an interesting variant of the substratum hypothesis:

This scenario not only explains the congruity in syntax of Welsh and Hebrew but at the same time gives a reason for the lack of lexical correspondences not only between Welsh and Hebrew, but in general between Afro-Asiatic and Insular Celtic.

Supposing that the explanation of certain peculiarities of Insular Celtic are due to substratum influence, one might suppose that the same or a similar substratum has influenced some subgroupings of Afro-Asiatic [Hamito-Semitic] In short, this scenario would mean that we should consider Western Europe and North Africa as an old coherent area of VSO-character. The influence on the three northern Afro-Asiatic groups, Semitic, Egyptian and Berber is comparable to the influence on the Celtic sub-grouping of Indo-European ... one might suppose that Western Europe and Northern Africa once formed one great contiguous VSO area. This area was split by the incoming IndoEuropeans. The proportion of Indo-Europeans on the continent was so great that any influence of a pre-existing language was blotted out, while the number of pre-IndoEuropeans inhabitants on the British Isles was such that their influence there was felt long after they were gone from memory.

Vennemann sees a Hamito-Semitic substratum as having influenced Insular Celtic, and through Celtic, English. In (2001:351) he claims: The European Atlantic Littoral was, at the dawn of history, explored and colonized by Mediterranean, probably Palaeo-Phoenician seafarers. The main features examined in this article include the verbal noun and the related progressive construction, the English Northern Subject Rule reminiscent of Semitic and Celtic verb-subject agreement, and the replacement of the sympathetic dative by the internal possessor construction (Jean sest cass le bras vs John broke his arm, see feature 22 below). Explaining why what he calls the Atlantic type arose only in Middle English, he reiterates (2001:364) an argument of Pokornys and Wagners: substratal influence originates in the lower strata of a society and usually takes centuries to reach the written language, and regularly only after a period of social upheaval. In (2002), in which he draws attention to circumstantial subordinating and, and the prevalence of tensed verb/auxiliary responsives in Celtic and English, Vennemann states:
In my view the case is closed, the thesis of a Hamito-Semitic substratum underlying Insular Celtic being one of the most reliably established pieces of scientific knowledge there is in any empirical discipline. As Gensler has shown, the substratum really was not simply Hamito-Semitic, which is a huge family including hundreds of languages in Africa and Asia (which is why it is also called Afro-Asiatic or Afrasian), but more specifically Hamito-Semitic of the Mediterranean type, which includes Libyco-Berber, Ancient Egyptian, and Semitic. In order to stress the similarity of the substratum to this particular manifestation of Hamito-Semitic, I sometimes refer to it as Semitidic or simply Semitic.

Celtic influence on English. The idea that certain features of English may be attributable to Insular Celtic is particularly in vogue among Finnish and German scholars, cf. articles by Filppula, Klemola, Vennemann, and especially Filppula, Klemola, Pitknen (eds) (2002) and Tristram (ed.) Celtic Englishes I, II, III, IV (19972006). This theory presupposes that the Anglo-Saxons were (thinly) superimposed on a British-speaking population which eventually shifted to Anglo-Saxon, leaving subsequent substratal structural traces in English. However, a recent study by Capelli et al. (2003) has found genetic evidence to support the more traditional picture of massive population shifts (2003:979): By analyzing 1772 Y chromosomes from 25 predominantly small urban locations, we found that different parts of the British Isles have sharply different paternal histories. Coates (2004) casts doubt on the influence of British Celtic on Anglo-Saxon. Typological approaches. Not all authors dealing with similarities between Insular Celtic and Hamito-Semitic languages subscribe to the substratum hypothesis. Recent papers adopting a purely typological stance include Borsley (1995), Isaac (2004), and Roberts (2004). An important methodological critique of the substratum approach is Isaac (2008).

3.

Shared features

The table on pp.7-8 shows the main features shared by Insular Celtic and HamitoSemitic, according to author and language. The first 17 features correspond to Genslers (1993) shared features. These are followed by additional features identified by Morris Jones (1900), Pokorny (1927-1930, 1959), Wagner (1959), Hewitt (1985), and Jongeling (2000). Features marked with () in the column for Gensler are mentioned by him, but are not central to his thesis; features 35-39 from Hewitt (1985) are given by no other author, and are purely typological in nature. 1. Conjugated prepositions 2. Word order: VSO, head-dependent, prepositions 3. Invariable relative clause linker, not relative pronoun 4. Relative clause copying, not gapping: the bed that I slept in it 5. Special relative tensed verb form 6. Subject and object marking in verb 7. Object marker: preverb-infix-V/V-suffix 8. Genitive construction: def. art. on dependent only: house the-man 9. Non-agreement of verb with plural noun subject 10. Verbal noun, not infinitive (object in genitive, not accusative)

11. Predicative particle: he is in a farmer 12. Prepositional periphrastic: he is at singing 13. Periphrastic DO: he does singing 14. Circumstantial clause and S PRED (subordinating and) 15. Nonfinite possible instead of finite main-clause verb 16. Word-initial phonetic changes (mutations), various syn. functions 17. Idiomatic genitive kinship constructions: son of X 18. Nominal clause (absence of copula) 19. Amplification of negative by noun after verb: French pas 20. Numerals followed by singular 21. Prepositional expression of have 22. Possessive he broke his arm rather than dative il sest cass le bras 23. Preference for parataxis (Pokorny: anreihend stringing along) 24. Basic unit word group rather than single word 25. Subjectless sentences (impersonal constructions) 26. No present/active participle

 27. Distinction between essential and contingent BE (is/t) 28. States/relations expressed with N (PREP-O) PREP-S T scilling agam ort 29. Welsh yn, Egyptian m in: predicative, locative, progressive 30. Old Irish infixed pronoun -d- identical to Berber -d 31. Comparatives (and superlatives) predicative only, not attributive 32. Initial focus clefts 33. Yes/no responsives with auxiliary/verb/entire clause 34. Prepositional relative: fronting of bare preposition the bed in I slept 35. Unmarked collective, derived singulative 36. Virtual complement clause VSO; factual complement clause SVO~VSO 37. Construct state bahuvrhi adjectives (Ar. improper annexation) 38. Topic subject (double subject sentences) 39. Yes/no dummy sentential pronoun: I dont know and he they came

4.

Main shared features

VSO order (feature 2). This apparently concerns both families, but there are numerous other VSO languages worldwide, even if the proportion of VSO languages is low (around 15%). Furthermore, there has been drift of varying degrees towards SVO in both Hebrew and Arabic, and also in Breton. Welsh (and to some extent even Irish) could be analysed as SVO if the predicative function is taken as being more important than the tense-bearing function for V. In the World Atlas of Language Structures Online, VSO does not cluster well with feature 4 (pronoun copies with oblique relatives), where many other languages partial to that strategy are strongly SVO. Finally, analysis of Arabic suggests that for that language at least, and possibly for other Hamito-Semitic languages, but clearly not for any Insular Celtic language, the basic order of the verb and its arguments is not based on the syntactic status of the arguments (subject, object: VSO), but rather on their information salience (given, new: VGN), the VGN principle economically accounting for all observable verb-initial orders, including the most frequent verb-initial order VSO.

Construct state (feature 8). The [house [the-king]] the kings house [HEAD [theDEPENDENT]] structure is certainly one of the most striking features shared by the two families. This involves: head-dependent order; availability of a definite article; relation-marking on the dependent only; relator mechanism of simple adjacency. Historically in both families, the relation was marked on the dependent with the genitive, and to some extent continues to be so in Arabic and Irish and Scottish Gaelic. The [HEAD [the-DEPENDENT]] structure becomes crucial in defining the genitive construction only with the loss of case (Colloquial Arabic, Hebrew, Welsh, Breton). Finally, Hamito-Semitic allows absolutely no adjectives to be placed between the head and the dependent, whereas all Celtic languages do so freely. The two families thus appear to have arrived quite independently at this construction, which is not so exotic as might be thought: Jamaican Creole, which has lost all case-marking, has [[di-king] hoos], which is merely the dependent-head inverse of [house [the-king]].

Non-agreement with plural post-verbal subjects (feature 9). Greenbergs (1966) Universal 33 states that: When number agreement between the noun and verb is suspended and the rule is based on order, the case is always one in which the verb precedes and the verb is in the singular. The tendency in Semitic (Arabic and Hebrew) has been to relax or lose non-agreement with plural post-verbal subjects; in the older stages of the Celtic languages (Old Welsh, Old Breton, Old Irish) number agreement was general; non-agreement has come in since the old periods, and in Irish, for instance, is still not complete, agreement being common in the southwestern dialects of Munster. Thus the trends appear to be going in opposite directions in the two families. In the negative, Breton functions exactly like Standard Arabic in both affirmative and negative: number agreement with preverbal subjects; non-agreement with post-verbal subjects.

Conjugated prepositions (feature 1). The apparent conjugations are probably historically, in both families, simply the result of a morphological process of incorporation of post-prepositional pronominals. Pronoun copies with oblique relatives (feature 4). In addition to Insular Celtic and Hamito-Semitic, the World Atlas of Language Structures Online shows this strategy to be common in numerous languages of central sub-Saharan Africa, but also in Persian, Eastern Kayah Li (Thailand, Myanmar), Paamese (Vanuatu, South Pacific) and Guaran (Paraguay, Brazil). Verbal noun rather than infinitive (feature 10). The distinction between the two is a matter of degree, the chief polar features being: gender: yes/no; definite article possible: yes/no; pronominal object: possessive/accusative; lexical object: genitive/accusative. Only Irish and Arabic have the first option (verbal noun) for all four features; other languages are less clear-cut. Breton (mainstream KLT dialects) and Modern Spoken Welsh have reached a point at which little distinguishes the untensed citation form of the verb from the infinitive of French and English. Arabic is not moving in this direction at all; as in Georgian, for instance, its madar is fully nominal. Circumstantial subordinating and (feature 14). I am unaware of any other languages apart from Hamito-Semitic and Insular Celtic which have this feature.

The apparent correspondence of Egyptian m in, r towards, r on and Welsh yn in, am for, wedi after expressing, respectively, predicative / locative / progressive, future, perfect (Morris-Jones and feature 29) is quite tantalizing, but to this level of detail concerns these two languages only, and is not widely shared by the rest of the Insular Celtic and Hamito-Semitic languages (tar is after with a verbal noun may express the perfect in Irish, and war on with the verbal noun/infinitive may be used for a prospective in Breton). For none of these prominent shared features is a substratal explanation demonstrably more plausible than a typological explanation or mere coincidence.

=mThe existence of striking structural similarities between the Insular Celtic and the Hamito-Semitic languages is beyond question. However, the matter of whether this is to be attributed to substratal influence through prehistoric contact or to typological tendencies and correlations remains unresolved. Genslers statistical approach (the low likelihood of such clustering of exotic features in two genetically unrelated families) is in itself skewed: by focusing on the shared features, he loses sight of the bigger picture, including all the features that are not shared by the two families. Furthermore, he has no way of accounting for the relative frequency or centrality of his various features in the languages concerned.

Authors inclined to a substratal explanation for the shared features appear to be prey to a kind of substratum frenzy, as if prehistoric contact must be the only possible explanation for un-Indo-European traits in an Indo-European language. They pay scant attention to the possibility of typological explanations, even though, as we have seen, such explanations are perfectly plausible for many of the shared traits. With lexical items, the number of possible phonetic sequences is so vast that any significant accumulation of lexical similarities between two genetically unrelated languages can hardly be anything other than a sure sign of contact and borrowing. With structures, however, the range of possibilities across languages is far more limited there are, for instance, only so many ways of expressing a genitive relation , so it is less surprising for unrelated languages to possess analogous structures. It is therefore important always to bear in mind and investigate thoroughly the possibility of a typological explanation. A major problem with the substratal explanation is the precise identity of the substratum. A subsidiary puzzle is the special affinities noted between Welsh and Hebrew (several authors, most recently Jongeling), Welsh and Egyptian (Morris Jones), Irish and Berber (Morris Jones, Pokorny and Wagner), and Breton and Arabic (Hewitt probably typological: both SVO~VSO topic-prominent languages); in each case, however, the number of features concerned is low enough for the special affinity to be coincidental.

Surprisingly, Genslers scores suggest that it is the Insular Celtic languages which are most typical of the Celtic/Hamito-Semitic type rather than the Hamito-Semitic languages, and this is borne out by our table of shared features by author and language, where the various features are more consistently present in Insular Celtic than in Hamito-Semitic. This is the reverse of what one would expect if the shared features really had their origin in Hamito-Semitic. Rather than positing some Berber Urvolk, or Phoenician settler ghosts (who have somehow managed to leave no archaeological traces), substratalists (and it should by now be clear that I am sceptical) might take a cue from Jongeling (2000), who moots a single prehistoric substratum to both Hamito-Semitic and Insular Celtic. Such a substratum might have been centred on north-western Europe or even the British Isles, where it might have affected the incoming Celtic languages strongly, but the more distant Hamito-Semitic and North African languages less so. The identity of such a substratum would, however, perforce be so shrouded in the mists of prehistory as to be quite unknowable.

Celtic peoples during the second Iron Age

5. Celtic from the West Punic-substratalized Tartessian spreads to the Isles by maritime route?
It is known that Punic was well established at various points along the southern Iberian coast, including, in particular, at Gades (modern Cdiz), the major Punic settlement nearest the area in which Tartessian emerged, from before, possibly well before 700 BCE, and continued to be spoken through the Roman imperial period. It is therefore plausible now, for the first time, to imagine that Punic may have provided a substratum to Tartessian. If so, and if Tartessian is indeed Celtic, this could conceivably solve the problem of the hitherto unidentifiable substratum language which might account for non-Indo-European traits in Insular Celtic. This would involve several steps: Find typical HS/IC substratal features in Tartessian; Find sources/correlates for such features in Punic or Phoenician; Imagine a scenario (maritime route?) which would get such a substratalized Tartessian to the Celtic Isles entirely bypassing Gaul, where none of the putative HS/IC substratal features have been attested;

 Suppose that this substratalized Milesian language was Q-Celtic, and became prevalent both in Ireland and in Britain; Suppose that P-Celtic arrived in Britain subsequently from Gaul, and acquired its HS/IC substratal features from the Punic-substratalized QCeltic already present; This would resolve the Insular Celtic/Continental Celtic vs Q-Celtic/PCeltic conundrum by allowing both fundamental divisions to be maintained simultaneously; However, long shot on multiple counts:
Consensus on Tartessian being Celtic? Paucity of clear relatable evidence from Punic; Timeframe of transmission of Milesian Q-Celtic to Isles? Possibly before Tartessian inscriptions? Likelihood of Q-Celtic in Britain as well as in Ireland? Need to bear in mind possible typological explanations for non-IndoEuropean traits in Insular Celtic (cf. above).

So, sit on several fences: Insular Celtic non-Indo-European features due to substratum or typology? If substratum, pre-Celtic inhabitants of Isles, or Punic > Tartessian? Tartessian: Celtic or not? Tartessian the source of the Milesian language brining Q-Celtic to the Isles? Milesian language to both Ireland and Britain, and P-Celtic from Gaul only later? Teleological fallacy: one should ideally avoid positing conclusions before investigating the available evidence; But difficult to avoid when attempting to imagine plausible scenarios on the cusp between prehistoric and palaeohistoric linguistics.

Tibetan for dog, pronounced /ki/, cf. Welsh ci, Breton ki, Irish c.

=m-