1 IntroductionFootnote ¹

When taking stock of the extensive literature on Old English historical phonology, one may well ask what is the use of an article about the Celtic influence on Old English phonetics and phonology, for the consensus is that no evidence exists for such influence.Footnote ² Taking a chronologically and thematically broader view of possible Celtic influence on the English language in England, we find that the many articles that argue for extensive morphosyntactic influence surfacing only in the Middle and Early Modern English literary languages are matched by only one article known to me about phonological influence appearing in the written record in the later medieval period (Laker Reference Laker2002 on the dialectal replacement of Old English cw- and hw- by Middle and Modern English χw-).Footnote ³ The mismatch is striking, for morphosyntax and phonology/phonetics are the two areas in which a population shifting from British Celtic to English could betray itself.

There is one good reason to address the issue of the phonological−phonetic influence of British Celtic on Old English, and that is that we now know more than we did a decade ago about the linguistic landscape which the Anglo-Saxon settlers engaged with when they first set foot in Britain. That knowledge enables us to reassess the nature of the influence that British Celtic may have had on Old English.

2 Languages in Britain around AD 400

The traditional picture of the linguistic map of Britain by the end of the Roman period is

1. – that British Celtic was spoken all over Britain south of Hadrian's Wall;Footnote ⁴

2. – that spoken Latin was present to some extent as well, probably more in the southern and eastern Lowland Zone than in the western and northern Highland Zone, and more in towns, military camps, and among higher social classes than in rural areas and among lower and less mobile social groups.

I have attempted to modify this picture (Schrijver Reference Schrijver2002; Reference Schrijver2007) by arguing that in the Roman period British Celtic was divided into a Lowland and a Highland variety, and that Latin was much more widespread and may well have been the dominant if not the only language spoken around 400 in the Romanized Lowland Zone. The argument runs as follows.

Knowledge about British Celtic is derived almost exclusively from the British Celtic languages Welsh, Cornish, and Breton, which are the only ones to survive as literary languages in the medieval and later periods. The history of these languages can be reconstructed through the early medieval and Roman periods to the Proto-Celtic language of circa 700 BC. All three British Celtic languages derive from the west and north of Britain, to which archaeologists and historians conventionally refer as the Highland Zone. During the Roman period, the Highland Zone was socio-economically much less Romanized than the southern and eastern Lowland Zone, where a Roman civil society developed, which included such features as towns, the rural villa system, and a dense network of roads. So what we know about the history and structure of Celtic in Britain is essentially limited to Highland Zone Celtic. When studying the contacts of Latin with Highland Zone Celtic, we observe a striking discrepancy between the Roman period and the early post-Roman period. During the Roman period, Highland Zone Celtic adopted large amounts of Latin lexemes but, as far as can be judged, underwent no phonological−phonetic or morphosyntactic influence. That is typical of relations between high-prestige and lower-prestige languages: lower-prestige languages borrow from high-prestige languages, and the process of borrowing begins with vocabulary. Between the fifth and seventh centuries, however, the adoption of Latin vocabulary in Highland Zone Celtic all but came to a halt. Instead, Highland Zone Celtic was affected by a host of phonological and morphosyntactical developments that were typical of the spoken Latin of late Antiquity. These developments dramatically changed the phonological structure of British Celtic. This state of affairs is typical of language shift: a population shifts to speaking a new language and by doing so introduces phonetic, phonological, and morphosyntactic features of its old language into the new language, but hardly any or no vocabulary.Footnote ⁵ In this case, accordingly, speakers of Latin shifted to speaking British Celtic but did so with a strong Late Latin accent. Apparently, this latinate variety of British Celtic prevailed over the original variety and became the attested Welsh, Cornish, and Breton languages, not because the latinate variety was of higher prestige – in that case we would expect to find Late Latin loanwords in British Celtic, which we do not – but because the shifting population was very large in proportion to the original speakers of Highland British Celtic. This linguistic scenario evokes images of large numbers of Latin-speaking refugees from the Lowland Zone entering the Highland Zone before the gradual advance of Anglo-Saxon settlers in the period between the fifth and seventh centuries. All this suggests that by the end of the Roman period the Lowland Zone was home to a large population of Latin speakers,Footnote ⁶ with obvious implications for the question of Celtic influence on Old English: it is probable that in the Lowland Zone, and along the eastern and southeastern coast in particular, Anglo-Saxons primarily met with a Latin-speaking rather than a Celtic-speaking population. If that population shifted to speaking Anglo-Saxon, it would potentially introduce Late British Latin rather than British Celtic phonology and phonetics.Footnote ⁷

The large-scale presence of Latin in the Lowland Zone by the end of the Roman period was itself the result of a language shift from (at least mainly) British Celtic to Latin. We know about the presence of Celtic speech in the pre-Roman and Roman Lowland Zone through the evidence of Celtic place names and personal names, but these are obviously uninformative as to the morphological and syntactic structure of Lowland British Celtic. The onomastic material is even uninformative as to the characteristic phonetic and phonological structure of British Celtic, as the names appear essentially in their Old Celtic forms, which are uniform across much of the Celtic-speaking world. One of the two British Celtic inscriptions from Roman Britain, the so-called ‘Bath pendant’, offers a glimpse into the phonology of what was probably a variant of Lowland British Celtic. The short text shows diphthongization of the long vowel of the first-person singular verbal ending in adix-ou-i < *-ū mi (where *-ū is the original verbal ending and *mi is the clitic first-person pronoun) and diphthongization of *ō to ua in the name cuamiinai (< *kōmignāi < *koimignāi, literally a diminutive derivation of the word for ‘dear’, hence ‘deary, darling’). These diphthongizations are absent from Highland British Celtic, but they are characteristic of the language of the northern Gaulish inscriptions of Châteaubleau and Baudecet, which date from the second or third centuries AD. The rules involved are:

1. – earlier Celtic word-final *-ū, *-ī become -ou, -ei, respectively (cf. Bath adixoui);

2. – earlier Celtic *ō (< Proto-Celtic *eu, *ou, *oi) and *ē (< Proto-Celtic *ei) become uo/ua and ie/ia, irrespective, it seems, of the phonetic context of the vowel (cf. Bath cuamiinai).Footnote ⁸

Comparable diphthongizations are characteristic of Late Spoken Latin, and I have argued that Celtic adopted them where Celtic was in intensive contact with Latin and a bilingual Celtic-Latin-speaking population was gradually shifting to become monolingual Latin speakers and in the process adopted the Late Latin phonetics of the area (which may itself have been influenced by local Celtic phonetics).Footnote ⁹ If we project the data extracted from this one single short text from Bath onto all or most of Lowland British Celtic, conscious of all the hazards involved in that extrapolation,Footnote ¹⁰ we conclude that at least as far as the vowel system was concerned Lowland British Celtic was less similar to Highland British Celtic than to contemporary Northern Gaulish; and we hypothesize that that similarity across a vast area is explainable by the intimate contact of Lowland British Celtic and Northern Gaulish with local varieties of Latin. Hence, where in eastern and southern Britain Anglo-Saxon came into contact with British Celtic rather than British Latin, British Celtic was of a variety whose phonetic−phonological system was different from that of Highland British Celtic and closer to that of the Late Spoken Latin of the northwest of the Roman Empire.Footnote ¹¹

All this is relevant to an assessment of the phonological influence which the language of the pre-Anglo-Saxon inhabitants of eastern and southeastern Britain may have had on Old English: even if we knew nothing at all about the linguistic situation in those areas around AD 400, it would be foolhardy to claim to settle the question on the basis of phonetic and phonological comparison of Highland British Celtic and Old English. That procedure is even less satisfactory now that we know that Lowland British Celtic was probably not identical to Highland British Celtic and that Latin may well have replaced Lowland British Celtic in most of the Lowland Zone. The question we need to answer is: was Old English influenced by Lowland British Celtic and/or Late Spoken British Latin, bearing in mind that we know next to nothing about those languages (for they have no surviving offspring, having been replaced by Anglo-Saxon) and that Highland British Celtic and Gallo-Romance must serve as imperfect proxies? It should be clear that under these circumstances the traditional claim that Old English was not influenced by the language of the pre-Anglo-Saxon inhabitants of Britain is just as unwarranted as the claim that it was. It would be fairer to say that we simply do not know and, perhaps, cannot know.

Rather than stop here, however, I propose to reopen the dossier and adopt the following procedure. After a reconstruction of the North Sea Germanic phonemic system which the Anglo-Saxons probably brought with them to Britain, the development of that system on British soil will be sketched up to the earliest surviving Old English records of the eighth century. The innovations which characterize Old English on British soil and the phonemic system to which they give rise will then be confronted with developments and systems in nearby Celtic and Romance in order to see whether there is a match. If there is a striking enough match, that is evidence for a (Celtic or Romance) language shift underlying Old English.

3 The prehistory of the Old English phonemic system: vowels

In view of ancient agreements of all Old English dialects with Old Frisian, Old Coastal DutchFootnote ¹² and, partly, Old Saxon, the vowel system of pre-conquest Anglo-Saxon probably was as follows:

In this system, *œ corresponds to North Germanic, Old Saxon, Franconian, and Old High German*ā, all of which derive from Proto-Germanic *œ (Nielsen Reference Nielsen1985: 126–7). Long *

developed from *a + nasal before fricative and from Proto-Germanic *œ before nasal. It ultimately merged with *ō but must have been a separate phoneme for some time, as the product of its shortening merged with *a rather than *o (Campbell Reference Campbell1968: 51 fn. 1). The presence of these two long open vowels and the absence of *ā renders the early North Sea Germanic long vowel system identical to the system of British Celtic c. AD 400–500,Footnote ¹³ which suggests influence exerted by a variety of Celtic similar to British Celtic but spoken on the continental side of the North Sea (Schrijver Reference Schrijver1999). Short *a may already have begun its phonetic fronting towards *œ as this is found in Old English, Old Frisian, and Coastal Dutch, but details differ (Nielsen Reference Nielsen1985: 128; Hogg Reference Hogg1992: 80 on Old English). Short *o, which started life as an allophone of *u, may still have been an allophone at this early stage.

There is at least one early phonological isogloss separating the Anglo-Saxon dialects: West Saxon and probably early Kentish (Hogg Reference Hogg1992: 62) kept *œ and *ē apart, but the Anglian dialects (Mercian and Northumbrian) merged both as *ē. The merger in the Anglian dialects preceded Old English breaking and was shared by Old Frisian. It may therefore even date to the period before the Anglo-Saxon settlement of Britain (Campbell Reference Campbell1968: 110 § 257; Nielsen Reference Nielsen1985: 232–5). Also preceding breaking and therefore early is the development of *ai to ā, which is shared by all dialects but was probably not shared by Old Frisian (Campbell Reference Campbell1968: 109; Nielsen Reference Nielsen1985: 130–4; Hogg Reference Hogg1992: 79–80).

By the early eighth century, Old English had transformed the inherited vowel system to a very different system, whose components were identical in all dialects, with one exception: West Saxon ie does not exist in the other dialects.Footnote ¹⁴

The arrangement of the vowels and diphthongsFootnote ¹⁵ in these systems reflects interrelationships which arose as a result of processes affecting prehistoric Old English (cf. Hogg Reference Hogg1992: 76–173):

1. (a) Unrounded front vowels ~ front-to-back diphthongs:

   1. – long and short io, eo, ea arose from long and short i, e, œ respectively, by the processes of breaking (before *r,*l,*χ followed by consonant, as well as before single *χ and *w) and a/u-umlaut (also known as back umlaut, before *a, *u in the following syllable);

   2. – conversely, long and short œ e, i arose from long and short ea, eo/ea, io, respectively, as a result of smoothing before (*r, *l +) *k, *g, *χ, which affected only Anglian.

      Long īo, ēo, ēa also arose from the Proto-Germanic diphthongs *iu, *eu, *au, respectively.

2. (b) Back rounded vowels ~ front rounded vowels

   1. – i-umlaut: long and short rounded back vowels *o, *u became long and short rounded front vowels ø, y, respectively;

   2. – i-umlaut: long and short front-to-back diphthong *iu via *iy became ie only in West Saxon; the other dialects showed no i-umlaut of *iu (> io); *eu did not occur in i-umlaut contexts.

3. (c) Back low vowels ~ front low vowels ~ front mid vowels

   1. – i-umlaut: long and short *a became œ (long *ā had arisen previously from Proto-Germanic *ai);

   2. – i-umlaut: short *œ became e;

   3. – i-umlaut: long and short *ea became ie (West Saxon), e (other dialects);

   4. – retraction and restoration of a: long and short *œ > a (under specific conditions).

The developments that affected the inherited North Sea Germanic vowel system can be described in more general terms as follows:

1. (1) Backing in back contexts: long and short front vowels *i, *e, *œ became opening front-to-back long and short diphthongs io, eo, ea before back contexts (breaking, a/u-umlaut).

2. (2) Opening of diphthongs: with these new diphthongs were merged the reflexes of the inherited diphthongs *iu, *eu, *au, which underwent a radical opening of the second element to become īo, ēo (part of Northumbrian ēa) and ēa, respectively. The development of *ai to ā, which created a new long vowel phoneme, may be described in terms of opening of the second element as well: *ai > *a œ > aa = ā. As a result of the merger, the inherited diphthongs were systematically tied into the simple vowel system.

3. (3) Fronting in front contexts: long and short back vowels (*u, *o, *a) and diphthongs (*io, *eo, *ea), were fronted by *i and *j in the following syllable (i-umlaut).

It is striking that the very similar vowel systems of the Old English dialects arose by similar processes, but from a slightly different starting point (vowel system 2) and, more importantly, with many differences in the detail of applying those processes: as is well-known, dialectal differences are rife in the application of breaking, the second fronting of *a and *œ in the Mercian dialect of the Vespasian Psalter, the influence of initial palatovelars, i-umlaut of diphthongs, a/u-umlaut, and Anglian smoothing. This created a situation in which cognate lexemes could be phonemically quite different, which cannot have failed to impair mutual comprehension. The following examples illustrate both typical differences and the irregularity of vowel correspondences:Footnote ¹⁶

Hence we observe that the Old English dialects all underwent highly similar and distinctive vowel changes leading up to an almost identical vowel system in all dialects. But we also observe that the slightly different application of these changes in the various dialects led to a potentially confusing array of differences between the dialects on the level of individual lexemes. The inevitable consequence of these observations is that the similarity of the vowel systems and of the processes leading up to them cannot be explained just on the basis of contact between the speakers of the Old English dialects, as the processes produced such different results that they impaired rather than improved mutual comprehension. It is as if the dialects adopted the same phonemic and phonetic system without contact with one another. That, I submit, is one of the best possible indicators of a common substratum, more explicitly, of a language shift by a non-Anglo-Saxon population to Anglo-Saxon in the period between c. 450 to 700.Footnote ¹⁷

4 The prehistory of the Old English phonemic system: consonants

Old English changes to the North Sea Germanic consonantal system were but few. The inherited system was as follows:

All consonants except *β and *γ occurred single and geminate. In all of West Germanic, the voiced fricative *ð had become *d. In North Sea Germanic, *z had become r or (word-finally) zero. Intervocalically, *f, *θ, *s became voiced to *v, *ð, *z; *v ultimately merged with *β. Between vowels, between liquid and vowel and between a vowel and a liquid, *χ was lost in prehistoric Old English.

Word-initially, *(s)k and *g were palatalized before front vowels, including œ < *a but excluding the front vowels which arose by i-umlaut, which puts the date of the palatalization between the fronting of *a to œ and i-umlaut. The palatalization agrees closely with Old Frisian and Coastal Dutch (Nielsen Reference Nielsen1985: 140–3); its phonetic roots may therefore predate the Anglo-Saxon settlements.

Since it changed so little between 450 and 700, the consonantal system cannot by itself provide convincing evidence for a language shift.

5 Identifying the substratum

Having established, in section 3, the probability that the development of Anglo-Saxon phonology and phonetics was fuelled by language shift, we need to address the question of which language or languages provided the template on which Old English restructured itself.

First of all, we must take account of the well-known fact that within a Germanic context neither the rise of opening front-to-back diphthongs, nor i-umlaut were exclusively Old English innovations. Old Frisian has a highly restricted development of *i to iu before cht (riucht ‘right’) and velar + back vowel (niugun ‘nine’), and it develops *eu into ia (biada ‘to offer’); Old Norse shows *e > ja (> jǫ) by a type of a/u-umlaut. However, the scale and systematic nature of the diphthongizations in Old English is incomparably more elaborate than in any other Germanic language, including Old Norse. Similarly, all of West and North Germanic shows i-umlaut, and one may plausibly argue that, at least phonetically, back vowels had started to become fronted in i-umlaut contexts already in pre-settlement North Sea Germanic. What requires an explanation, however, is that the phonemicization of i-umlaut and the rise of phonemically rounded front vowels had taken place before the earliest sources of the eighth century in Old English, Old Frisian, and Old Coastal Dutch, whereas Low Franconian, High German, and Old Norse lagged centuries behind. In order to account for the early phonemicization of rounded front vowels and i-umlaut, I have argued that contact with Celtic of the type represented by Highland British Celtic (Schrijver Reference Schrijver1999) and/or with Late Spoken Latin of the type represented by Old French (Schrijver Reference Schrijver2002) was responsible. Both Highland British Celtic and pre-Old French show spontaneous fronting of long rounded back vowels, and Highland British Celtic has i-umlaut of short vowels, which had been phonemicized by the sixth century at the latest. For the sake of quick reference, compare the North Sea Germanic vowel system (cf. system 1 of section 3; here I add the allophones of back vowels fronted by i-umlaut) with reconstructed Highland British Celtic and Northwestern (Picardian) pre-Old French systems:

If we wish to account for the changes affecting the Anglo-Saxon vowel system between stages 1 and 3 of section 3, however, it is apparent that neither Highland British Celtic nor pre-Old French is capable of providing the template on which the Anglo-Saxon dialects remodeled themselves. As I have indicated in section 2, that is not unexpected, for Highland British Celtic and pre-Old French are at best close comparanda for the lost language of the people who shifted to Anglo-Saxon, not that language itself.Footnote ¹⁸Footnote ¹⁹

There is another language, however, which shows developments that are indeed very similar to those observed in the Anglo-Saxon dialects. That language is Old Irish. Let us put aside for the moment the very valid question where, when, and why Anglo-Saxon and Old Irish might have been in contact and concentrate first on the similarities between the two languages. These similarities are obscured if we compare the phonemic systems but on a phonetic level they are immediately apparent. The Irish counterpart of Anglo-Saxon backing in back contexts and fronting in front contexts is called palatalization: simplifying somewhat, consonants between front vowels and consonants before (long or short) *i become palatalized; additionally, any *i or *e lost by apocope and syncope causes palatalization of the preceding consonant. It is the palatalization of consonants that by the Old Irish period (AD 600–900) had become phonemic, but phonetically the vowels flanking palatalized consonants are affected, too. Here are some examples, in which approximate phonetic representations accompany the Old Irish phonological forms:

1. A

The Old Irish counterpart of Old English a-umlaut and breaking is (phonetic) velarization: any consonant that is not palatalized (because the consonant or consonant group resisted palatalization by *i or because it was followed by the back vowels *a or *o) affects a preceding front vowel by producing a velar off-glide:

1. B

In the case of one reconstructed vowel, *ē (< Proto-Celtic *ei), diphthongization before a nonpalatal consonant was phonemic in Old Irish:

1. C

The Old Irish counterpart of Old English u-umlaut is (phonemic) u-infection:

1. D

It is true that the subphonemic presence of rounded front vowels and front-to-back diphthongs in A and B above is not directly attested in Old Irish and can only be extrapolated from Modern Irish. However, since all Modern Irish and Scots Gaelic dialects agree on this point and since it is virtually impossible to pronounce these forms without recourse to such phonetics, their subphonemic presence in Old Irish is guaranteed.

Thus, on a phonetic level Old Irish possesses long diphthongized and short rounded front vowels, and long and short front-to-back diphthongs, which arose under the influence of a front vowel (generally *i) and a back vowel or nonpalatalized consonant(s), respectively, in the following syllable. The similarity to Old English is obvious. Where Old Irish and Old English differ is on the phonemic level: with the exception of u-infection and the diphthong ía, Old Irish phonemicizes consonantal features (palatal and nonpalatal consonants), whereas Old English phonemicizes vowel features (rounded front vowels and front-to-back long and short diphthongs).

Within this general pattern, there is an even more specific correspondence. In Old English, vowel + *u + *i in subsequent syllables became vowel + *i + *i, and the vowel underwent i-umlaut. This is the so-called ‘double umlaut’ (Campbell Reference Campbell1968: 82–3). In prehistoric Irish, vowel + *u + *i became vowel + *i + *i, and if the *i in the second syllable was syncopated, it palatalized the preceding consonant (Greene Reference Greene1973: 134). It is the only original back vowel to do so. Examples:

• OE: *lat-umista- > *latimista- > lætemest ‘last’

• OIr.: *Lugudikos > *Luγiðeχ > Luigdech /luγ´ð'^eχ/ [lyγ'ð'^eχ], genitive singular of the personal name Luguid

Other, less specific features that the prehistory of Old English and Old Irish have in common are: initial stress, apocope, syncope, and a much richer vowel system in stressed than in unstressed syllables; and the presence of the voiceless fricatives *χ, *θ and *f (inherited from Proto-Germanic in Old English; reflecting lenited *k, *t, *sp/*sw, respectively, in Old Irish).

In conclusion, prehistoric Old Irish and prehistoric Old English, although separated by thousands of years of language evolution along separate branches of the Indo-European family tree, appear to share a common phonetic basis. The correspondences in the vowel system in particular are nontrivial and highly specific. They strongly indicate a causal link, which requires specification. That is the subject of the next section.

6 Explaining the phonetic similarities between Old English and Old Irish

Based purely on linguistic arguments, the simplest scenario would be that Old English and Old Irish share a prehistoric common phonetic basis because Old English developed on a substratum of a form of Primitive Irish Celtic (not British Celtic or British Spoken Latin). Translated in geographical terms, the simplest scenario would boil down to the idea that Old English either developed in Ireland and/or Scotland, or that a form of Irish Celtic was spoken in the British Lowland Zone during the Roman period. The former idea, which would imply that the Anglo-Saxon settlers made a stopover in Ireland or Scotland before turning to eastern and southern Britain, is so blatantly contradicted by the historical and archaeological record that it can be upheld only if there are compelling linguistic reasons to back it up, which I cannot see at the moment. The idea that Old English developed on an Irish Celtic rather than a British Celtic or British Latin substratum in the Lowland Zone of Britain does not receive prima facie support from the historical or archaeological record either, but the question is whether the historical and archaeological record has anything in principle to contribute to a discussion about the linguistic affiliation of the Celtic dialects of the Lowland Zone. Common sense tells us that it has not. Hence I propose to take this idea seriously and to further explore its merits and defects.

7 The origins of the Irish language

I have argued elsewhere that the Old Irish language is the result of a relatively recent language spread in Ireland, probably involving a language shift from a non-Celtic language to Celtic, and that the most probable window for the coming of Celtic to Ireland is as recent as the first century AD.Footnote ²⁰ Apart from a few trivial historical and geographical assumptions, the argument is completely of a linguistic nature and runs as follows.

1. (1) The earliest datable evidence for the presence of a Celtic language in Ireland is found in names attested in Ptolemy's Geographia (AD 150), e.g. the tribal names Brigantes (‘High’ or ‘Noble Ones’), Ouenniknioi (‘Family Descendants’?), Robogdioi (‘Very Poor Ones’). For reasons of historical phonology, the name for Ireland and its inhabitants itself, which is attested much earlier, is not linguistically Celtic (e.g. OIr. Ériu; Latin genitive plural (gens) Hiernorum (Avienus), I(u)verna (Pomponius Mela, Juvenal), Hibernia (Caesar); Greek Iouernias nēsos (Ptolemy), Iernē (Strabo)). But even if the name were Celtic, it may have been an exonym, which the Celtic inhabitants of Britain and Gaul used, rather than an endonym, which Celtic-speaking inhabitants of Ireland themselves used. Hence the terminus ante quem for the arrival of Celtic in Ireland is as late as the early second century AD. Needless to say, the arrival may be much earlier, but the point is that it does not need to be.

2. (2) The linguistic ancestors of Irish and Highland British Celtic are phonologically and grammatically identical up until the first datable isogloss separating British from Irish Celtic (*ai > British *ɛ, first century AD; Jackson Reference Jackson1953: 694). The only possible exception is the development of *k^w to *p in British Celtic and its preservation in prehistoric Irish, but that isogloss is now generally believed to be trivial, and British Celtic, just like Gaulish (p-Celtic generally but not universally, cf. Sequana, eqos, Ucuetin), may well have had pockets of speakers who preserved *k^w. The linguistic identity of Irish and British Celtic up to so recent a period is irreconcilable with the idea that the ancestor of Irish and the ancestor of British Celtic had been geographically separated by the Irish Sea for any length of time, for had that been the case, at least some linguistic differences, however trivial, should have arisen.

3. (3) The arguable presence of a non-Celtic language in Ireland from which Irish borrowed words at least as late as the sixth century ADFootnote ²¹ and the fact that the Irish Ogam inscriptions show radical phonological changes being phonemicized in the period between 400 and 600 combine to support the idea of a language shift to Celtic around that period, which resulted in Old Irish.

4. (4) Old Irish is a monolithic language, and Early Old Irish of the seventh century is identical to Proto-Irish: spoken Modern Irish, Scots Gaelic, and Manx descend from that language, and no isogloss can be shown to predate it. In this respect, Irish contrasts sharply with all other languages which entered the written record during the Middle Ages: Old English, Old High German, Old Norse, and Slavic, for instance, show dialectal differences as soon as they enter the written record. The exceptional nature of Old Irish is explainable if the language resulted from a recent spread in Ireland (see 2 and 3). Alternatives are conceivable. One might consider a top-down scenario according to which a standardization of the literary language successfully wiped out all pre-existing dialectal differences. However, the prime example of a successful top-down spread of uniformity, Classical Latin, in no way wiped out all traces of the earlier dialectal diversity of Latin in Latium, so I seriously doubt whether the top-down scenario is at all plausible. Alternatively, a population bottleneck may have left only a small, linguistically uniform community of Celtic speakers in Ireland as the sole survivors of a previously diverse linguistic landscape. Although this scenario is conceivable, it is obviously less capable of explaining the closeness of Irish and British than is a recent Celticization of Ireland.

A historical scenario that would account for the recent linguistic Celticization of Ireland would be that Celtic was introduced into Ireland in the wake of the Roman occupation of Britain, perhaps in connection with the destruction of Brigantian independence in AD 74, which arguably led to a movement of at least some British to Ireland (cf. the Lambay Island finds, Raftery Reference Raftery1994: 200–3).

The relevance of this excursion into the history of Irish is obvious: if Irish descends from a Celtic dialect spoken in Britain in the first century AD, it is reasonable to assume that a reconstruction of the phonetics and phonology of Old Irish holds the key to unraveling the phonetics and phonology of the dialect of British Celtic that is responsible for the phonetic and phonological features of Old English that are the subject of this article. Old Irish is probably an even better source for the reconstruction of British Lowland Celtic phonetics than is Highland British Celtic because the latter underwent extensive influence from spoken Latin in the early medieval period (section 2).

There is a test we can perform in order to assess the relation of Old Irish to Lowland British Celtic. What little we know of Lowland British Celtic points to the conclusion that it shared two diphthongizations of long vowels with Northern Gaulish (section 2):

1. – earlier Celtic word-final *-ū, *-ī become -ou, -ei, respectively (cf. Bath adixoui);

2. – earlier Celtic *ō (< Proto-Celtic, *eu, *ou, *oi) and *ē (< Proto-Celtic *ei) become ua and ie/ia, irrespective, it seems, of the phonetic context of the vowel (cf. Bath cuamiinai).

Neither diphthongization can be dated exactly; their terminus ante quem is the third century AD, which is the approximate date of the Châteaubleau tile.

Irish certainly did not undergo the former diphthongization, for in that case word-final *-ū > *-ou would have caused lowering of a high vowel in a preceding syllable and would have failed to cause u-infection. As it is, however, neither happened: cf. dative singular fiur ‘man’ < *wirū, n-stem nominative singular íriu ‘land’ < *īweriū, both forms lacking lowering and showing u-infection (hence not **fer < *wirou, **íre < *īweriou). But Irish does have a close counterpart of the latter development: *ē (< Proto-Celtic *ei) became the diphthong ía [ia] before a nonpalatal consonant, and *ō (< Proto-Celtic *eu, *ou, *au) became úa [ua] or ó in an arbitrary manner, it seems (Thurneysen Reference Thurneysen1946: 39–41; McCone Reference McCone1996: 134). Both ía and úa were phonemic in Old Irish, and their phonemic nature appears in the written record by the eighth century.Footnote ²² That is the terminus ante quem for the rise of phonetic diphthongization, which may or may not be much older.Footnote ²³ As long as this diphthongization was nonphonemic, any newly arising closed *ō or *ē would have joined the diphthongization. That would have been the case with British *oi > *ō (in Bath cuamiinai < *kōm- < *koim- ‘dear?’, a development it probably shares with northern Gaulish, cf. Châteaubleau muana < *moinā ‘gifts?’ but not with Irish) and with Irish *au > *ō (úaithed ‘small number, little’ < *au-tīto-).

By this single token, it appears that the phonetics of the language of the Bath inscription, together with the northern Gaulish inscriptions of Baudecet and Châteaubleau, is closer to the phonetics of Old Irish than it is to the phonetics of Highland British Celtic, which shows no signs of diphthongization. In other respects, however, such as the development of *oi to *ō (> ua) and the addition of the clitic *mi ‘I, me’ to the first-person singular verbal ending *-ū, Bath and ChâteaubleauFootnote ²⁴ agree with Highland British Celtic against Old Irish. If all this is put into a chronologically consistent framework, phonetic diphthongization of *ō and *ē would have arisen in pre-Roman Lowland Britain and northern Gaul and was shared by Irish when its ancestor was still in Britain, whereas *oi to *ō (> ua) and the addition of the clitic *mi belong to a later period when innovations spreading from Gaul were shared by Celtic dialects within the Roman Empire.Footnote ²⁵

It may be well to point out explicitly that it is not the purpose of the argument to insist that the diphthongization in Irish and Lowland British/Northern Gaulish must be a shared development that cannot have occurred independently. The purpose is to test the phonetic similarity of Old Irish to what little we know of Lowland British Celtic. The result of that test is that by one single token, diphthongization, Old Irish is phonetically more similar to Lowland British Celtic than is Highland British Celtic.

8 Synthesis

As was discussed earlier, the phonology of the language of the Germanic speaking settlers who spread across southern and eastern Britain was changed drastically between c. 400 and 700. In all dialects of Old English, from northern Northumbrian to Kentish and West Saxon, almost the same phonemic−phonetic system arose, but in such different ways that mutual understanding was impaired. The simplest explanation of this state of affairs is that phonetic−phonological developments were fuelled by a language shift of a pre-settlement population which stretched all the way from the Tweed to Wiltshire and which in its original language had that same phonemic−phonetic system. The shifting population introduced its traditional phonetics and phonology into the language of the Anglo-Saxon settlers, where those features survived and were perpetuated. The sociolinguistic scenario that gave rise to the spread of this particular ‘accent’ but at the same time blocked the spread of lexicon and, at least initially, syntax into Anglo-Saxon remains to be determined.

The original language of the shifting population can now be identified as a variety of Celtic which was ancestral to Old Irish. Old Irish is the product of a linguistic colonization of Ireland which had its roots in a variety of British Celtic spoken in Britain sometime in the late first or early second century AD. Descendants of the speakers of this variety of Celtic who stayed behind in Britain eventually became Roman citizens. Some of them, especially those living in the eastern and southern Lowland Zone and being part of a highly Romanized society, stood a good chance of becoming bilingual Celtic-Latin speakers. Some ultimately shifted to Latin altogether. Given the evidence for phonetics surviving multiple language shifts,Footnote ²⁶ these new Latin speakers to a lesser or greater degree held on to their particular Celtic phonetics. The phonological history of Old English indicates that these, either Celtic speakers or Latin speakers with a strong Celtic accent, were the people whom the first Anglo-Saxon settlers met and who left a characteristic trace in Old English phonology and phonetics.

9 Addendum: British Latin

I have argued (Schrijver Reference Schrijver2002) that Highland British Celtic was phonologically influenced by British Latin in the period between approximately 400 and 700. In the same period, the Germanic dialects of the Anglo-Saxons were phonologically influenced by British Latin based on ‘Irish’ Celtic phonetics or by ‘Irish’ Celtic itself. The question arises why the influence of Latin worked out so very differently in both cases: phonologically and phonetically, Old Welsh, Cornish, and Breton are very different from Old English. Various possibilities can be raised, some of which I shall formulate rather than explore here.

Perhaps Old English was influenced mainly directly by ‘Irish’ Celtic rather than by British Latin, while Highland British Celtic was influenced mainly by British Latin: different languages produce different results when populations shift. For this scenario to work, however, we need to assume that the phonetics of British Latin was different from the phonetics of ‘Irish’ Celtic, which is implausible if we assume that British Latin resulted from a language shift by a population speaking ‘Irish’ Celtic.

On the other hand, the linguistic situation is bound to have been more complicated: on a large island such as Britain, British Latin may have evolved in different regional varieties and social registers, partly corresponding to and partly irrespective of the different regional varieties and social registers of Celtic. Ultimately, there is enough leeway to allow one variety of British Latin to influence British Highland Celtic and quite another to influence Old English. Out of a multitude of possibilities, one might arbitrarily select one according to which it was city folk and the rural elite that fled to the Highland Zone, because they had the means to do so or because the collapse of towns left them no alternative. Their comparatively upper-class Latin would then have influenced Highland British Celtic, which became Welsh, Cornish, and Breton. The rural poor, small farmers and agricultural laborers, may have stayed on, hoping to strike a deal with the new powers, and in so far as they succeeded, they would have imported into Old English a comparatively lower-class Latin accent, phonetically similar to the original Celtic. Further study, combining the linguistic with the archaeological and historical record, should be able to advance our knowledge concerning such questions.

10 Addendum: from which part of Britain did the ancestor of Irish come?

While the area in which the ‘Irish’ phonetic and phonological system influenced Anglo-Saxon must have been extensive, stretching along the eastern and southern coast from Hadrian's Wall to the Isle of Wight, there is one particular section in the northeast where the match between Old Irish and Old English phonetics is so close that it stands a better chance than any other of being the area of origin of Irish. The Northumbrian dialect of the glosses in the Lindisfarne Gospels and the Durham Ritual contains a number of forms in which the Proto-Germanic diphthongs *eu and *au merged as ēa (as against ēo and ēa, respectively, in other dialects: lēaf, dēath as opposed to lēof ‘dear’, dēath ‘death’ elsewhere). Similarly, breaking and a-umlaut of long and short *œ and *e merged as long and short ea, too: hearg, hearta (cf. ea, eo elsewhere, as in hearg ‘temple’, heorta ‘heart’). Only by u-umlaut and by the influence of immediately following *w do long and short *e become long and short eo: seofo ‘seven’, heofon ‘heaven’ (Campbell Reference Campbell1968: 118 §278 (b)). The fact that in this dialect *e becomes ea by a-umlaut and breaking but eo by u-umlaut is strikingly reminiscent of the Old Irish developments which gave rise to the difference between such forms as nert ‘strength’ /nʹert/ [nʹeǝrt] < *nertah < *nertos and its dative singular neurt /nʹeurt/ [nʹeurt] < *nertū. The geographic location of the dialect that shines through in these Northumbrian sources coincides approximately with the area that was targeted by the Irish mission from Iona under Áedán. This fact raises the possibility that the distribution of eo and ea in our Northumbrian dialect reflects a seventh- to eighth-century Irish superstratum feature.Footnote ²⁷ That idea is problematic in the sense that phonetics and phonology generally surface as substrate rather than superstrate features.

Faced with that difficulty, we may look elsewhere for an explanation: one cannot fail to notice that the Northumbrian localization fits in rather well with the idea advanced above that Celtic may have been introduced into Ireland by first-century refugees from the Brigantian tribal area.