1 Introduction
In McWhorter (Reference McWhorter2002), I argued that English inherited less of the Proto-Germanic legacy than any other Germanic language, and that this contrast with its sisters is stark enough to demand an explanation in interrupted acquisition at some point in its time line. My proposal was that Scandinavian Viking invaders acquired an incomplete version of English, reducing or eliminating features such as grammatical gender (entirely absent in no other Indo-European language in Europe but English), directionally marked adverbs (here/hither), V2 syntax, and inherent reflexive marking (German sich erinnern ‘to remember’), and that this rendition was prevalent enough to be passed on to future generations of native English speakers.
Various scholars, however, while concurring that the grammatical contrast between English and its sisters is too vast to be a mere matter of chance, argue that the intervening factor was the Celtic languages spoken by the original inhabitants of Britain. In contrast to traditional arguments that the indigenous Celts were largely eliminated by the Germanic-speaking invaders, adherents of the Celtic hypothesis (henceforth CH) suppose that Old English was deeply transformed when used as a second language by speakers of Brythonic Celtic languages, specifically Welsh, Cornish, and possibly the extinct and sparsely documented Cumbric. These scholars point to numerous grammatical parallels between Celtic and English that do not exist between Celtic and other Germanic languages, such that English would be a deeply Celticized Germanic language just as Indo-Aryan languages are agreed to be imprinted by Dravidian.
Signature arguments begin with early work by scholars such as Keller (Reference Keller1925), Dal (Reference Dal1952), G. Visser (Reference Visser1955), and especially Preusler's (Reference Preusler1956) broad-ranging argument, and continue with a modern revival school exemplified by Poussa (Reference Poussa, Adamson, Law, Vincent and Wright1990), Klemola (Reference Klemola1996, Reference Klemola2000), Tristram (Reference Tristram1999), Filppula (Reference Filppula2003), Hickey (Reference Hickey1995), Vennemann (Reference Vennemann2000, Reference Vennemann and Brinton2001), van der Auwera & Genee (Reference van der Auwera and Genee2002), White (Reference White2002), and Poppe (Reference Poppe2003). The contributions of Hamp (Reference Hamp and Drachmann1975) and even J. R. R. Tolkien (Reference Tolkien1963) have also been useful.
The CH has been regularly dismissed by most scholars of English diachrony. A major objection has been that the purported Celtic influences do not appear in the historical record until the Middle English period despite contact with Celts beginning several centuries before this. Another has been that English has so few Celtic lexical items. I myself was previously convinced by such objections (McWhorter Reference McWhorter2002). However, upon further engagement with the CH literature, I have been moved to revise my views.
My reading of the evidence is that what drew English grammar away from the Proto-Germanic template was two developments. One was that Norse-speaking invaders simplified the grammar. The second was that Celtic speakers infused the grammar with new constructions as the result of language shift over time.
I will focus this presentation on a single construction, periphrastic do, which is especially useful in demonstrating that the usual objections to the CH fail upon examination. Our interest will be in do-support: the use of do as a semantically empty carrier of tense in negated and interrogative sentences:
I walk.
I do not walk.
Do I walk?
As ordinary as these sentences are in English, one would be hard pressed to name a language one had learned or was even aware of that used do in this particular way. That is, unless one is a speaker of, or familiar with, a Celtic language, in which case one is confronted with a close parallel.
The intended contribution of this presentation is to reinforce other CH accounts of periphrastic do by calling attention to the urgency of the CH argument in view of cross-linguistic facts, by bolstering counterarguments to the objections about issues of timing of occurrence and lexical contribution, and by showing that traditional accounts of periphrastic do's emergence cannot be seen as preferable to a CH account in the scientific sense.
2 Celtic models for periphrastic do
Various CH adherents have pointed to Welsh and Cornish models for periphrastic do (Preusler Reference Preusler1956; Poussa Reference Poussa, Adamson, Law, Vincent and Wright1990; Klemola Reference Klemola1996; van der Auwera & Genee Reference van der Auwera and Genee2002). Their tendency has been to designate Cornish as the most likely source specifically, given that periphrastic do appears first in southwestern texts dated to about 1300 and Cornish is the Celtic language that was spoken in this region.
Cornish's auxiliary do presents a thoroughly plausible model for English's periphrastic do, and not simply in its auxiliary status – plenty of grammars use do as an auxiliary – but in using this auxiliary in semantically neutral fashion as a syntactic place-filler with all verbs.
Do in Cornish is used both in finite form and in a complex with the verbal noun (VN in the glosses), in which do occurs in the third-person singular form after a particle (PT in the glosses):
(1)
‘I (do) love.‘
(Jenner Reference Jenner1904: 136, 116)
and in the negative and the interrogative:
(2)
‘I did not see you.’
(Jenner Reference Jenner1904: 160)(3)
‘Do you love?’
(Jenner Reference Jenner1904: 117)
The do auxiliary occurs not only in the present but the past:
(4)
‘You (did) love.’
(Jenner Reference Jenner1904: 137, 116)
The usage of auxiliary do conveys little or no semantic difference (George Reference George1993: 454). In fact, Jenner (Reference Jenner1904: 117) documents the do auxiliary as used not just alternately, but usually, in the first- and second-person singular. Cornish's do is also the usual choice in dependent clauses:
(5)
‘..if the man would love.’
(Jenner Reference Jenner1904: 117)
Of course our interest must be in Cornish as it was encountered in its earlier stages. The fact is that even the earliest substantial attestations of Cornish, termed the Middle Cornish stage, already display the same auxiliary usages of do as later Cornish. Middle Cornish used auxiliary do in the present:
(6)
‘I eat him.’
in the negative:
(7)
‘You do not stay here.’
in the interrogative:
(8)
‘Who will finish building your whole house?’
and in dependent clauses:
(9)
‘if he sent. . .‘
(Lewis Reference Lewis1990: 46)
As such, the usage of auxiliary do as model for English's periphrastic do in Modern Cornish was ancient.
Welsh, on the other hand, presents a slightly less close parallel to English. As in Cornish, the do verb occurs with the verbal noun, conveying no significant semantic difference from constructions without do:
(10)
‘I open.’
(King Reference King2003: 189)
The do auxiliary is used in Welsh in the negative and interrogative as well:
(11)
But in Modern Welsh the similarity to English's periphrastic do fades beyond the present tense. Welsh does not use do as an auxiliary in the past (rather, only in the northern dialect, ddaru ‘to happen’ is used thus (King Reference King2003: 189–90)). Then, in the future, do is used only when a specific meaning of intentionality (in King's terminology) is intended, rather than as semantically neutral:
(12)
‘Let's speak to him later.’
(King Reference King2003: 193)
As it happens, this incongruity of distribution regarding tense was somewhat lesser in Middle Welsh. Like Cornish, it used the do auxiliary in a semantically identical alternate construction with a particle, and here, do occurred in the past as well:
(13)
‘He sat.’
(Watkins Reference Visser1993: 327)
(Today this construction is a less semantically central one, connoting emphasis.) However, only Cornish, early and modern, prefers auxiliary do in dependent clauses, or has entrenched it even further in using auxiliary do semantically neutrally in the imperative: Gwracara ‘Love!’ (Jenner Reference Jenner1904: 141).
Thus while Welsh would hardly be an implausible source for English's periphrastic do, Cornish is the closer match, which is one of many suggestions supporting the view that the English construction's origin in the Cornish-speaking region was due to transfer from Cornish.
3 CH criticism number one: Celtic's auxiliary do was due to contact with English
Given that Cornish and English have existed in contact for a very long time, we must consider whether English developed periphrastic do internally and then passed the construction on to Cornish speakers using English bilingually.
To be sure, strictly speaking, attestations allow us to trace Cornish's usage of auxiliary do only so far back. The Middle Cornish stage began in 1200, and the aforegoing citations date to the 1400s or later. Thus, technically, there is still room for the surmise that Cornish before this had no auxiliary do, and that do developed on its own into an auxiliary in English around, as attested, 1300, with Cornish merely inheriting the construction subsequently from contact with English.
As it happens, the history of Celtic gives us a window on its grammatical nature long before its substantial attestations, and precisely when Germanic speakers invaded England in the middle of the 400s. Namely, Celtic speakers migrated from southern England to northwestern France in the 400s and 500s, seeding what is now the Breton-speaking community. As such, grammatical features that Breton shares with Cornish, tracing back to the last time they were the same language, must be taken as evidence of the nature of Cornish just when the Angles, Saxons, and Jutes would have encountered it.
If English lent periphrastic do to Cornish, then we would expect that Breton would be a language with nothing resembling periphrastic do – like most of the world's languages. This is not what we find. Modern Breton uses auxiliary do in the affirmative:
(14)
‘You listen to the wind.’
(Press Reference Press1986: 128)
It also uses it in the negative, but only for emphasis according to Press (Reference Press1986):
(15)
‘I shall not sleep.’
(Press Reference Press1986: 128)
McKenna (Reference McKenna1988) cites it in the interrogative:
(16)
‘Do you eat well?’
(McKenna Reference McKenna1988: 151)
Sources agree that where periphrastic do occurs in Breton, it is with verbs across the board, quite frequently (Hemon Reference Hemon1975: 158), and making no identifiable semantic contribution.
Unremarkably, contact between Cornish and Breton speakers continued after Breton branched off (Hardie Reference Hardie1948). However, a surmise that Cornish inherited periphrastic do from English and then Cornish passed it on to Breton via trade relations must be evaluated in terms of economy. This scenario would require first that English passed a particular syntactically disruptive feature, periphrastic do, on to not just one but two Brythonic Celtic languages, Welsh and Cornish – which would alone be a happenstance. Then, Cornish passed this same particular and syntactically disruptive feature on to Breton, despite the fact that this kind of syntactic transfer is typical of intimate contact of the kind that occasions widespread bilingualism (Thomason & Kaufman Reference Thomason and Kaufman1988: 83–100), rather than passing trade relations. That is, ‘contact’ alone, unqualified, cannot stand as grounds for a case for syntactic transfer. The result, under this scenario, would also be unusually faithful transfer occurring three times, rather than, say, periphrastic do across the board in Welsh, in only the affirmative and connoting emphasis in Cornish, and used only with a subset of verbs in Breton.
My claim is that Welsh and Cornish shared a genetic feature of Brythonic Celtic, this feature was passed on to a descendant of Cornish, and that later, Cornish passed the feature on to English. Supporting this is that, quite simply, English is the only Germanic language with a semantically empty syntactic place-filling do. Occam's Razor renders this account preferable.
However, it bears mentioning that Stephens (Reference Stephens1993: 394–5) specifies that in Breton periphrastic do occurs only in the affirmative, and that McKenna (Reference McKenna1988: 150) states that auxiliary do occurs in the interrogative but not in the negative.
Meanwhile, Middle Breton sources only give us affirmative citations, such as:
(17)
‘I swear to you.’
(Lewis Reference Lewis and Piette1990: 41)
Breton, then, allows two possible interpretations as to the presence of auxiliary do in Cornish when Anglo-Saxon speakers arrived.
One is that when Breton branched off from Cornish, auxiliary do occurred in the affirmative, negative, and interrogative as it does in Modern Cornish, and that today, Breton's negative and interrogative usages are pragmatically or dialectally restricted because Breton speakers have long been bilingual in French, which has no analogue of auxiliary do and thus may have led to the narrowing of the construction's grammatical application.
Alternately, we might take a more parsimonious route and suppose that the evidence shows that when Breton branched off from early Cornish, auxiliary do only occurred in the affirmative. This would mean that the Cornish spoken when Germanic speakers invaded England used auxiliary do only in the affirmative.
The fact is, though, that this alone would have been a plausible model for English's current use of do. In its first stage as an auxiliary in English, do was used in the affirmative as well as the negative and interrogative; i.e. as the do alternately treated as causative, habitual, perfective, etc. Only later did its affirmative usage disappear, leaving only the negative and interrogative usages. If Cornish provided a model only for an affirmative usage, then it is unproblematic to reconstruct that English first incorporated this usage, and then generalized auxiliary do into the negative and interrogative in line with the syntactic nature of auxiliaries: it is not usual that syntax bars the expression of an auxiliary in the negative and interrogative.
It is also relevant that periphrastic do occurs in English first in the affirmative, with negative and interrogative usages some decades later (Denison Reference Denison1993: 265). While this may be an accident of documentation, it would also be compatible with Cornish's having only the affirmative usage when English encountered it – or even by the 1300s − such that we could allow that its occurrence in Cornish sources in the negative and interrogative in the 1400s and afterwards was due to influence from English's having generalized periphrastic do to these contexts. Crucially, however, the ‘default’ affirmative usage was established in Cornish before English speakers came to Britain.
In sum, however interpreted, the evidence does not allow that Cornish began as a language without a periphrastic do construction and only later inherited it from English, leaving a mistaken impression to modern observers that Cornish was the borrower. Cornish had, at the very least, a periphrastic usage of do in the affirmative, which would have been a ready model for an English equivalent that was used in negative and interrogative sentences as well. And then, only the marginality or absence of periphrastic do in the negative and interrogative in Breton suggests the possibility that Old Cornish lacked these: Middle Cornish uses periphrastic do in all relevant contexts, and thus so may have Old Cornish.
This suggests that a scientifically responsible version of the CH on periphrastic do need not leave the pathway of influence flexible and reconstruct a general Sprachbund of which do-periphrasis is one trait, with its family of origin unknowable due to the sparseness of its documentation in Old Celtic sources, as Tristram (Reference Tristram, Ramisch and Wynne1997) has it and van der Auwera & Genee (Reference van der Auwera and Genee2002) concur with. Given that Breton split from Cornish starting in the 400s, then even in the absence of copious texts in Old Brythonic varieties, Breton's usage of periphrastic do confirms that Celtic spoken in the south at that time had already developed a periphrastic do before contact with English speakers.
4 CH objection number two: periphrastic do is a Germanic family trait
It has also been claimed that English's periphrastic do is less a local oddity than an unremarkable development, found in many Germanic varieties (e.g. Tieken-Boon van Ostade Reference Tieken-Boon van Ostade1988; Eroms Reference Eroms, van Ostade, van der Wal and van Leuvensteijn1998). For example, in nonstandard German dialects tun ‘to do’ is used in sentences such as (rendered with Standard German lexical forms):
(18)
‘He writes that.’
(Abraham & Conradie Reference Abraham and Conradie2001: 83)
Similar constructions with do are found in nonstandard Dutch dialects (Cornips Reference Cornips1994), earlier stages of Frisian (van der Auwera & Genee Reference van der Auwera and Genee2002: 287), Old Norse (Ellegård Reference Ellegård1953: 27), and elsewhere.
However, English's periphrastic do behaves quite differently than these nonstandard Germanic usages. While English's do is a syntactic place-filler, Abraham & Conradie (Reference Abraham and Conradie2001: 80–93), in a discussion more extended than the brief flaggings of the German and Dutch do-sentences in descriptions of nonstandard varieties, analyze nonstandard German's tun-construction as primarily a pragmatic strategy encoding focus:
(19)
‘Maybe I'd better tear up your piece of paper.’
(translation mine) (Abraham & Conradie Reference Abraham and Conradie2001: 86)(Standard German: Ich tue dir deinen Zettel am besten zerreisen.)
(Leipzig dialect)
(20)
‘Maybe I'll (just) cut the bag open.’
(translations mine) (Abraham & Conradie Reference Abraham and Conradie2001: 87)
Bavarian:
Importantly, we cannot treat this usage as a potential precursor of English's periphrastic do, despite the fact that nothing would rule this out as a plausible grammatical development. This is because do was never used in this pragmatic fashion in any earlier stage of English; i.e. no analyst has ever proposed such an analysis of early occurrences of auxiliary do in English. Germanic languages that use do in this way have come to do so independently of English.
The pragmatic analysis does not cover the entire domain of the usage of auxiliary do in nonstandard Germans, in that the tun-construction's usage in some dialects has been analyzed differently, such as in Hessian where it serves as an analytic subjunctive:
(21)
‘I wouldn't do it.’
(Durrell & Davies Reference Durrell and Davies1990: 235)
(Thuringian has a similar usage; Spangenberg Reference Spangenberg1990: 284.) Yet this is also a usage quite distinct from that of English's periphrastic do.
One might nevertheless suppose that these represent a mere difference in grammatical outcomes in different languages: certainly we would not expect do's usage as an auxiliary to be identical from one sister language to another. One might propose that Proto-Germanic lent an auxiliary usage of do that had differing fates in the daughter languages, including English.
However, it must be recognized that the particular fate of auxiliary do in English is, in the cross-linguistic sense, extremely peculiar (see van der Auwera Reference van der Auwera, Tops, DeVriendt and Geukens1999). To be sure, because of the general nature of its semantics, it is hardly uncommon for do to be generalized into semantically bleached functions. Commonly, for example, grammars can express many verbal concepts with compounds composed of do and a nominal element. Persian, for example, has only a few hundred simple verbs, otherwise making heavy use of combinations of nominals with kardan ‘to do’: u-ra bidar kard ‘he woke him up’ (Windfuhr Reference Windfuhr and Comrie1987: 538). Similar is Japanese's use of suru:
(22)
‘Taroo travels to Tokyo.’
(Miyamoto Reference Miyamoto1999: 1)
the usage of which can be so semantically distant from its core meaning of ‘to do’ as to encode possession and existentiality:
(23)
‘Taroo has a wound.’
(Miyamoto Reference Miyamoto1999: 5)(24)
‘There is a strange smell.’
(Miyamoto Reference Miyamoto1999: 4)
Yet in neither Persian nor Japanese has do extended to usage with all verbs and become a purely syntactic place-filler. We could say that this is barred in these languages because morphology makes it clear in both that the nominals are not verbs, whereas the barrier is lesser in English given its poor inflectional apparatus.
But this only raises another question: why no periphrastic do in Germanic varieties almost as inflectionally denuded as English, like Dutch, Swedish, Norwegian or Danish? The question is especially urgent given that all of these languages include numerous nonstandard varieties, often significantly divergent grammatically from the standard ones, in which we would expect that over so many centuries sheer chance would create a periphrastic do in at least some of them. Indeed, do occurs as a place-filler in Mainland Scandinavian in cases of inversion and topicalization, as in Danish:
(25)
‘As for reading, he never does it.’
(Allan, Holmes & Lunskær-Nielsen Reference Allan, Holmes and Lunskær-Nelson1995: 511)
But this is what English diachrony specialists typically term ‘substitutive do’, plausibly treated as a Germanic family trait – but our question is whether Germanic languages typically recruit do as a required dummy auxiliary element in negative and all inverted sentences such that interrogative ones are subsumed. More precisely: why no English-style periphrastic do in just one of the dialects of these several languages? Again, English's usage of do is extremely odd.
Many Australian languages have a usage more similar to English's periphrastic do, in that it occurs with verbs instead of nominals, and not in substitutive function. In these languages, do is one of as few as three simple verbs, and verbal concepts other than these few are obligatorily expressed in coverbial constructions with the simple ones. In Jingulu, for example, do is used thus:
(26)
‘You can see me.’
(Pensalfini Reference Pensalfini and Peeters2000)
But even here, do is not the only verb with this behavior, in contrast to periphrastic do's unique status in English as applying to all negated and interrogative sentences. In Jingulu, come and go exhibit the same behavior; meanwhile, in many Australian languages, this set of simple verbs is as large as thirty (Dixon Reference Dixon2001: 71), and in none of them is there only a single such verb, do or any other.
Thus worldwide, certainly do is uniquely ripe for grammaticalizations of various kinds, and for lexicalizations in verbal compounds with nominals. This is predictable from its cognitively central semantics as denoting basic, unspecified action. However, a search for languages in which do has devolved into semantically empty syntactic conjunction with all verbs is challenging indeed. Surely it is significant, then, that Celtic languages like Cornish – precisely the languages spoken by the indigenous inhabitants of Britain who adopted English – are among the precious few of the world's languages that use do in just this way.
5 CH objection number three: periphrastic do emerged via ordinary process of internal change
The sheer weirdness of English's periphrastic do in the cross-linguistic sense sheds light on the various accounts of its origin by traditional specialists in English diachrony. These authors suppose that periphrastic do can be traced to ordinary processes of internal grammaticalization, but upon analysis, their accounts raise as many questions as they answer. Fifty years after Ellegård's (Reference Ellegård1953) foundational crack at the problem, the origin of periphrastic do remains unclear, with analysts undecided as to which account is most appropriately subscribed to. Traugott's (Reference Traugott1972: 141) assessment that ‘the debate still continues and may never be resolved’ remains resonant decades later – while the simple fact that the Celtic languages English was in contact with had their own periphrastic do sits dismissed.
5.1 Earlier accounts
I will assume that certain earlier accounts have been conclusively refuted. Ellegård proposed that periphrastic do has its roots in the causative usage of do (Henry . . . þe walles did doun felle ‘Henry felled the walls’), with do developing a semantically empty interpretation because of cases where verbs were used with causative meaning without do, such as in the second clause of the aforegoing citation: Henry . . . þe walles did doun felle, þe tours φ bette he doun. But Denison (Reference Denison1993: 278) notes that causative do is too rarely attested to be a viable candidate for a grammatically central development like periphrastic do. Similarly, Visser's (1963–73: §1413–14) account based on the generalization of ‘substitutive do’ fails in that, as Denison (Reference Denison1993: 277) notes, there are too few citations of the relevant construction, do used in conjunction with an infinitive.
5.2 Habituality?
However, to derive periphrastic do from more recent presentations is similarly difficult. For example, Garrett (Reference Garrett1998) proposes that do was first recruited as a habitual marker, used first with deverbal nominals (do reste ‘do some rest’) which were reinterpreted as verbs when their verbal sources (here, resten) were phonetically eroded such that they were identical with their deverbal derivates (i.e. the verb, too, was now reste).
However, the problem is why once the nominals were reinterpreted as verbs, habitual do devolved into a syntactic place-filler instead of retaining its habitual semantics. Garrett proposes a semantic reanalysis beginning with do used in negated sentences. The immediate connotation of a sentence like Cows do not eat nettles is the habitual one, that eating nettles is something that cows do not do. But it could also be taken as meaning that not eating nettles is something typical of cows – a descriptive rather than durative statement. Under Garrett's analysis, when negation acquires scope over the entire proposition (neg-raising) in this way, it left the habitual marker semantically empty: now the sentence simply negated the concept of cows eating nettles, with habituality left by the wayside. Thus sentences that were ‘assertions of general nonexistence’, Garrett states, ‘were strengthened to become assertions of nonexistence’ (Reference Garrett1998: 317).
Garrett's account, formalized in principles of theoretical semantics, is ingenious and elegant in itself. Yet, it does not explain why this is not a typical pathway of habitual markers worldwide, much less why it seems to have only happened in English. Crucially, it does not explain why it could not have been that the new neg-raising interpretation could not have preserved the habitual semantics of do regardless, such that a sentence like We do not eat dinner early on Saturday would be reinterpreted to mean the still habitual ‘It is not the case that we are in the habit of eating dinner early on Saturday’ rather than the habituality-neutral ‘The act of our eating on Saturday does not exist.’ Why, that is, did do not retain habitual semantics in such sentences even if subsumed to negation? Garrett's analysis seems to imply a certain inherent evanescence of habitual marking, which is perhaps tempting from the perspective of Modern English in which its indication in the present affirmative is a zero-marker (We eat dinner early on Saturday), but would be less so from the perspective of other languages that have overt habitual markers clearly and eternally central within their batteries of aspectual morphemes.
In that light, we also wonder why the habitual semantics of do in the affirmative, robust despite whatever tendencies threatened the negative usage, would not have blocked its semantic bleaching in the negative. Presumably, while neg-raising semantics was threatening do in the negative, in affirmative sentences, semantically central to human expression, do's habitual semantics would still have been transparent and reinforced by constant usage.
Also, Garrett's analysis does not conclusively demonstrate that early auxiliary do was in fact a habitual marker. He acknowledges that out of the 195 pre-1400 citations of periphrastic do that he bases his argument upon, no fewer than 39 percent of them (i.e. virtually 40 percent, which could be parsed as ominously close to half) are not habitual. Even after accounting for some of the exceptions by surmising that conditional sentences carry an inherent negative inference that would encourage a similar reanalysis of do, and sidelining others as idiosyncratically advanced ones already behaving like modern periphrastic do in negative, interrogative, and inverted sentences, the volume of exceptions remains problematic. Garrett notes that ‘almost 75 percent’ are not (1998: 320), but then it could also be put that one in four of the citations in his corpus does not fit his hypothesis – for reasons that he leaves to ‘future research’.
Finally, on the subject of habitual do, Garrett's supporting proposition that do is also used habitually in German dialects must be seen as founded upon too brief a reference to the data to be conclusive. He notes this Low German sentence as ‘apparently habitual’:
(27)
‘They used to call him Nats.’
(Grimme Reference Grimme1922: 126 in Garrett Reference Garrett1998: 313)
But this is only one sentence. And while his citing habitual descriptions of the usage of auxiliary do in two other German dialects must be acknowledged, these are passing comments in sources written at the turn of the previous century. In this light, it bears mentioning that while do in these dialects gives ready appearance of being a habitual marker upon passing analysis of the kind that these early grammarians lent the construction, more extended and concentrated studies such as Abraham & Conradie (Reference Abraham and Conradie2001), as well as modern descriptions of the dialects such as those in Russ (Reference Russ1990), give no indication that a habitual analysis captures the essence of the behavior of German dialects’ auxiliary do. Garrett does not engage modern sources on German dialects’ usage of do, and in general only addresses the issue briefly.
There looms, then, a possibility that the nexus between negative habitual and universal semantics, and the theoretical plausibility that the latter could have overtaken the former, were in fact not the source of periphrastic do. Rather, it is equally plausible theoretically that periphrastic do was modeled on, for example, periphrastic do already existing in another language such as Cornish.
Importantly, if the rendition in English of Cornish's do did have even a leaning towards habitual semantics, this by no means belies its source in transfer from a Celtic item whose behavior was determined by syntax.
Transfer, after all, is almost never precise between languages. Cushitic languages have a causative double suffix; Ethio-Semitic, influenced by Cushitic acquirers, has a causative double prefix (Thomason & Kaufman Reference Thomason and Kaufman1988: 135). As to a syntactic construction being rendered in a borrowing language as a semantic one, as English would have rendered a Celtic periphrastic do as a habitual marker, Saramaccan Creole's rendition of the Niger-Congo Kwa language Fongbe's reduplication is an analogue. In Fongbe, property items (adjectival concepts) are syntactically obliged to occur in reduplicated form when used attributively:
(28)
‘The black bag.’
(Lefebvre & Brousseau Reference Lefebvre and Brousseau2002: 358-9)
Saramaccan, with a syntax deeply parallel to Fongbe's given that a preponderance of its creators spoke that language natively, has reduplicated verbs as attributive adjectives as well, but only as a semantic construction. In Saramaccan, with stative verbs the reduplication encodes counterexpectational semantics, such that dí baákabaáka sáku would mean ‘the blackened bag’ (as by fire), i.e. ‘that bag that is, against what you would expect, black’. ‘The black bag’ is simply dí baáka sáku, whereas in Fongbe the reduplication is syntactically required.
Here, then, is a living example of a mechanism via which English would incorporate Cornish's periphrastic do that is both theoretically plausible – like Garrett's account – but also cross-linguistically typical, which Garrett's account is not. We could propose that English took on Cornish's periphrastic do with a semantic leaning, but only that, towards habituality – encouraged by the action-focused semantics of the verb do, although the particular Cornish rendition happened not to be colored by such and was solely syntactically conditioned. We must choose between Garrett's ingenious but studied semantic pathway to periphrastic do – unparalleled anywhere else among the world's languages – and an account that stipulates that English simply incorporated periphrastic do from a language that had it already.
5.3 Perfectivity?
Denison's (Reference Denison, Eaton, Fischer, Koopman and van der Leek1985, Reference Denison1993) account of periphrastic do as a semantic bleaching of what began as a perfective marker leaves similar questions. Denison (Reference Denison1993: 264) refers to sentences such as His sclauyn he dude dun legge ‘He laid down his pilgrim's cloak’ as indicating this perfective function, but, like Garrett, is faced with quite a few exceptions. Namely, in his corpus of references, 30 sentences are activities or states as against about 400 that conform to his analysis (Denison Reference Denison, Eaton, Fischer, Koopman and van der Leek1985: 54), such that its support in the 1200s and 1300s qualifies in his own estimation as only ‘broad’ (Denison Reference Denison1993: 281). Interestingly, the exceptional 30, with their imperfective semantics, could support Garrett's habitual analysis instead, and in general, the historical data's amenability to accounts based on aspectual concepts as distinct as the habitual and the perfective can be taken as suggesting that the very quest to treat early auxiliary do as expressing any single aspect is futile. We return to the case that English's usage of do was simply modeled on the semantically empty usage in Cornish.
In any case, Denison otherwise appeals to speculations that are theoretically questionable. An example is his claim (1993: 281) that do developed its perfective function as compensation for English's loss of Germanic derivational prefixes like be-, for-, and ge- that had carried considerable functional load in expressing aspectual distinctions. This is a brilliant analysis – except that, as Denison readily indicates, we must reconstruct that this perfective usage of do for some reason did not last, and devolved into the semantically neutral periphrastic one that concerns us. Denison supposes that do's compensation for the loss of the prefixes was a passing ‘experiment’ – but we must ask just why it was. That is, why didn't do remain a perfective marker until today? There is no theoretical basis in the tenets of diachronic syntax or semantics for a process in which constructions arise as ‘experiments’ only to mysteriously evanesce. The derivational prefixes did not return, and as such we wonder what general principle would have motivated that do, purportedly entrenched among certain generations of English speakers as a perfective marker, would then strangely bleach into a semantically empty syntactic place-filler even though the derivational prefixes remained gone forever. Just why did the ‘experiment’ fail? Importantly, Denison, unlike Garrett, does not propose a theoretical semantic pathway via which this devolution would have been predictable or even natural.
Also relevant: Old Norse also lost the derivational prefixes, and while in Mainland Scandinavian they were restored by contact with Low German (Braunmüller Reference Braunmüller, Bandle, Braunmüller, Jahr, Karker, Naumann and Telemann2002: 1037–8), Icelandic and Faroese remain lacking them. And yet do has not compensated for their loss in these languages by becoming a perfective marker, despite Old Norse having used its do-verb as an auxiliary in other uses (albeit not analogously to periphrastic do).
Denison, perhaps aware of such problems, brings in a reinforcing factor, proposing that the causative usage of faire ‘to do’ in French influenced English's do (Denison Reference Denison1993: 279). However, this contravenes the magisterial argumentation of Thomason & Kaufman (Reference Thomason and Kaufman1988: 306–15) that French speakers were never numerous enough, or in intimate enough contact with most English speakers, to affect English grammatically. Namely, for French grammar to have affected English, there would have to have been a critical mass of Francophones mastering English incompletely and passing on their rendition to the Anglophone masses. There is no evidence that this was the case: rather, French was spoken by a small elite rarely encountered by most English speakers, such that this elite's renditions of English, however incomplete, would have had no appreciable effect upon English grammar over time.
Thomason & Kaufman are mainly concerned with refuting the idea that French speakers developed an English creole, but their argumentation also refutes the idea that French speakers could have had even less dramatic influence on English grammar. Sociohistorical circumstances in Norman-ruled England were not ones that would have conditioned language-shift effects; rather, French affected English only in the superficial layer of lexical incursion, and this largely via the written medium. Denison developed his analysis before Thomason & Kaufman's study existed (and given the realities of academic publishing, even his 1993 article would have been written long before a book released in 1988 became an established reference). But the fact remains that today, French influence must be judged as incapable of supporting an account of English grammatical diachrony.
5.4 A reinforcement for SVO order?
Kroch, Myhill & Pintzuk's (Reference Kroch, Myhill, Pintzuk, Tuite, Schneider and Chametzky1982) explanation for the emergence of periphrastic do has a potential explanatory power less evident in Garrett's and Denison's presentations – but still runs up against cross-linguistic generality. They propose an explanation for why in modern English periphrastic do is limited to negated and inverted sentences. This is useful, in addressing an important discrepancy between Brythonic Celtic and Modern English: e.g. Cornish uses do in affirmative sentences as well. While Kroch et al. are not concerned with the CH, the eclipse of the affirmative usage stands as a question under the CH, and their account offers a solution.
Namely, they propose that do's usage in the negative and interrogative brought the relevant sentences in line with the SVO order that became regular in Middle English in contrast to Old English's SOV order in subordinate clauses and frequent VS order elsewhere. As such, in a newly SVO language, do-support in negative and interrogative sentences would iron out the cognitive dissonance of sentences like Alas wherefore lighteth me the sonne?, by displacing the main verb after the subject (Alas, why does the sun light me?). We might suppose that English first inherited Cornish's use of periphrastic do across the board, but then trimmed its usage to the negative and interrogative for reasons of processability amidst the grammar's transition into SVO word order.
As useful as this idea is, the question still remains: why was a semantically empty dummy do available for such recruitment in the first place? If this account were valid by itself, then presumably, worldwide, there would be various languages which, after a transition from SOV to SVO word order, would have brought a do-verb (or maybe a be-verb?) into such a usage. For example, Latin was primarily SOV. Wouldn't we expect, then, that just one of the countless grammars that have evolved from it, all of them today SVO, would have a reflex of Latin's facere as a dummy place-filler in negated and inverted sentences? I am aware of a single case: a few dialects of Italian such as Monnese have a dummy usage of do used only in questions – but not negated sentences (Benincà & Poletto Reference Benincà and Poletto2004). Yet there would appear to be nothing of even this sort in French or Spanish, the most nonstandard dialect of Portuguese on the Iberian peninsula or the obscurest and most untutored dialect of Romanian, and in no village dialect of Occitan, Catalan, or Rhaeto-Romance. The famous SOV−SVO transition in West Africa (Hyman Reference Hyman and Li1975) yielded no languages documented to date that English diachrony specialists could point to as showing that do typically becomes a syntactic place-filler due to exigencies created by word order change.
Kroch, Myhill & Pintzuk's argument alone is, then, a just-so story. It ‘explains’ only a particular situation via what, very technically, could have been, rather than what actually tends to be.
The argument instead can be taken as, in fact, entailing that English was in contact with another grammar in which do had already become, by chance, a dummy auxiliary. To put it another way, the fact that Cornish already had such a place-filler cannot be taken as irrelevant to English's developing one, nor can it be taken as a mere ‘possible factor’ – rather, it must be seen as a crucial factor. This is because innumerable grammars making the transition from SOV to SVO do not bring in do to obligatorily disambiguate syntactic patterns that disturb SVO order – this including not some but all Germanic grammars except English.
The question becomes: what scientific basis justifies dismissing Cornish as the source of this recruitment of do, or even marginalizing it as merely a ‘possible influence’? What other factors qualify as more or even equally possible? Which other factors are so theoretically compelling as to disqualify Cornish as anything less than pivotal to the scenario?
Obviously English did not incorporate all of Celtic's grammatical features. Kroch, Myhill & Pintzuk's article can be taken as a principled explanation of why English incorporated one of them, periphrastic do. It is hardly unscientific to propose that periphrastic do was useful for processing reasons in a language moving from SOV to SVO. But this also requires that we treat Cornish as the source of the construction.
6 CH objection number four: periphrastic do appears too late to have been modeled on Celtic
Periphrastic do first appears in Early Middle English, in southwestern texts of c. 1300. English speakers had first encountered Brythonic-speaking Celts in the 400s, almost a thousand years before, and yet there is no evidence of periphrastic do in the vast Old English literature.
This has been seen by many as counterevidence to a Celtic account. As Ellegård (Reference Ellegård1953: 119) had it regarding Celtic influence, ‘I do not see any reason why it should have been stronger in the 13th century than during all the previous centuries that the races had been in contact.’
The observation is well-taken, but various CH adherents have argued that we can expect that constructions like periphrastic do, if developed by Celtic speakers, could easily have been restricted to the vernacular register for a long period, considered ‘low’ speech unsuitable for the print medium. An especially succinct rendition was by Dal (Reference Dal1952: 113), for which, because it was written in German, I offer a translation:Footnote 1
However, the primary justification for our take on the issue is that the sociohistorical relationships in question indicate that we would, precisely, not expect ample usage of syntactic Celticisms in the Old English literature. The Celts were the oppressed group; to the extent that they expressed themselves in the English language, this carried the stigma of the speech of the vulgate, of a sort that was to be avoided in refined literary language. Certainly, then, it is hardly unexpected that constructions of the quotidian language of the masses could thrive for centuries without appearing in the written register.
(See also Vennemann Reference Vennemann and Brinton2001; Klemola Reference Klemola2002: 207.) In this section I would like to flesh out this point: Dal was correct that if Celts contributed periphrastic do, then its delayed appearance in texts was nothing short of ordinary. For it to have occurred in Old English texts would have been a stark contradiction of sociolinguistic realities typical of written languages worldwide since time immemorial.
6.1 Dating periphrastic do to 1100: the nature of Early Middle English texts
In assessing whether Celts contributed periphrastic do to English on the basis of the documentation, it must be held front and center that Early Middle English texts are an especially approximate reflection of the reality of the spoken language at the time, such that periphrastic do may very well have been widespread, at least in a certain region of Britain, earlier than 1300. Namely, the nature of the texts of the period in no sense establishes that the construction existed only by 1300, and indicates nothing implausible in the construction's occurring at least as far back as the beginning of the Middle English period in 1100.
First, documentation of English from 1100 to 1300 is notably sparse (Laing Reference Laing2000: 103–4). This means that to the extent that periphrastic do first appears in two 1300 texts, it could easily have been accidental that no text was written, or survived, in which the construction occurred as early as 1100.
In addition, English documents at this time are highly variable in terms of dialectal diversity (Dobson Reference Dobson, Davis and Wrenn1962; Milroy Reference Milroy1992; Smith Reference Smith2000). English was re-emerging as a written medium after centuries of replacement by French, and its transcription reflected the changes that the spoken language had undergone during this time (and likely, as agreed upon by all analysts in reference to the extent of inflectional erosion, even during the Old English period). As with all primarily oral languages spoken over vast distances, there was considerable dialectal divergence, and Middle English would only be standardized starting in the 1400s. This means that if periphrastic do emerged in a particular dialect or dialect group, then its attestation at all in a multidialectal corpus from 1100 to 1300 is a matter of chance. Its absence in texts written before 1300 in other dialects cannot be taken as indication it had not yet arisen in the southwest before this.
6.2 Dating periphrastic do to the fifth century AD: the conservativity of the written medium
The next question is why periphrastic do is entirely absent in Old English texts. Here is where observations such as Dal's are relevant. The nut of the case is that writing the way you talk is a modern idea: for English speakers in 1300 and before, writing was an elite, codified, and static activity.
We have concrete evidence of this in that even when written English was ‘reborn’ as Early Middle English, the centuries-deep tradition of Old English writing exerted a conservative pull on how the language was rendered on the page. Middle English between 1100 and 1300 varies between the new variety reflecting spoken norms and more antique varieties reflecting the Old English tradition. In fact, most Early Middle English is distinctly conservative (Laing Reference Laing2000: 105; Sweet Reference Sweet1892: 211). The Owl and the Nightingale text of c. 1200 retains grammatical gender marking on definite articles and nouns; the Ancrene Wisse text of c. 1215 retains Old English features such as the genitive plural, present-tense plural, and infinitive endings.
Thus arguments such as Dal's and others’ that Celtic features could have been established in spoken English for centuries before they were used in written English should be received not as particular and athletic surmises referring to English alone, but as applying to English diachrony a typical phenomenon of register differentiation in language use worldwide since the invention of writing. The differentiation has always been especially robust in societies that have not known the democratic impulses in the wake of the Reformation and the Enlightenment, which encouraged the ushering of vernacular varieties into the written realm.
It could be ventured that specialists in the history of English are skeptical of the ‘delay’ hypothesis of CH adherents partly because they live in societies in which, because of these very democratic impulses, a relatively close correspondence between spoken and written language is ordinary. In contrast, in the Early Middle English speakers’ world, speakers of what we know today as the Romance languages still encountered Latin on the page as the closest written equivalent to their everyday speech – while as often as not, the English speaker found on the page (if in an essentially preliterate society they ever engaged with writing at all) an inflection-heavy tongue alien to what they had learned on their mother's knee.
In this light, in a rigidly hierarchical pre-Magna Carta England in which writing was restricted to formal subjects and literacy was the province of a small elite, a gulf between spoken and written language equivalent to that between Moroccan and Modern Standard Arabic would not be a surprise: that the gap was lesser almost qualifies as a quandary.
For example, Persian is another language whose middle stage of attestation reflects the emergence in writing of what had previously been relegated to the spoken register. Old Persian was a typical early Indo-European language, bristling with three genders, six case inflections dividing into declensional paradigms and displaying complex ablaut patterns, a proliferation of tenses and aspects expressed with conjugational paradigms, and so on. But just three centuries after the last attestations of this language whose grammar was akin to Sanskrit's, in the 300s BC Middle Persian attestations begin, in a language as anomalously light on inflection and other complexities as English is in comparison to its sisters (Khanlari Reference Khanlari1979: 231). Certainly a change this vast did not occur in a mere three hundred years, before which the language had retained a typical Indo-European morphological battery for countless eons.
Iranian specialists suppose that these changes had already begun during the Old Persian period, but that the highly formal contexts in which Old Persian had been written discouraged the transcription of the spoken reality. Indeed, late Old Persian reveals the very beginnings of the breakdown of the case system (Schmitt Reference Schmitt and Schmidt1989: 60), although nothing remotely approaching the vast simplifications evident in Middle Persian. It is assumed that the cracks in the dam evident in the latest Old Persian data were due to transcribers by then writing Old Persian as a different language, their everyday language being an inflectionally light one like Middle Persian. This latter kind of Persian only appears in writing in full bloom after a profound sociological dislocation, when an empire had been overthrown by a new regime. There is no reason to suppose that the situation was different between Old and Middle English.
In many modern languages there is a gulf between written and spoken language that reflects the situation before the emergence of Middle Persian and Middle English; i.e. the very definition of diglossia. Standard Finnish is spoken everyday by no one: the colloquial language includes various ‘evolved’ constructions such as (Karlsson Reference Karlsson1999: 245–8):
(29) vienkö minä > vienks mä ‘Shall I take?’
he tulevat > ne (those) tulee (sing. ending) ‘they come’
autolla > autol ‘by car’
me sanomme > me sanotaan (passive) ‘we say’
The last construction above, in which the colloquial version is an alternative (a passive usage) rather than a phonetic erosion, is useful as a means of showing that spoken language can differ from written not only in abbreviations, but in substitutions – of the kind such as, in Early Middle English, using auxiliary do plus a main verb as an alternative to sentences with matrix verbs alone.
In the Finnish case, we see the passive, not a simplification but a substitution, used in place of the indicative. This involves variation between constructions available within a single language. But elsewhere, we get closer to our proposition re a Celtic-induced periphrastic do: ‘substitution’ modeled on a substrate language, never reaching print except in academic linguistic publications.
An example is the influence of Quechua on Spanish in Ecuador. The most prominently covered language variety here is Media Lengua, a ‘mixed’ or ‘intertwined’ language that uses Spanish lexical items with Quechua morphological affixes, phonology, and syntax. However, Ecuadorians also speak Spanish dialects with varying degrees of influence from Quechua less extreme than in Media Lengua (ocupado estoy busy I-am ‘I'm busy’, reflecting Quechua's SOV word order (Lipski Reference Lipski1994: 251); de Juan su sombrero ‘of John his hat’ based on Quechua Huwan-pa ch'uku-n Juan-GEN hat-his) – Spanish with moderate grammatical interference from Quechua. These varieties reach the page only in academic works by linguists: in the consciousness of Ecuadorians they are strictly spoken language, with no written presence over countless centuries.
Another example is the impact of Berber languages on the spoken Arabic of Morocco. Berber nomads relocated to this area soon after Arabic speakers reached it, to the extent of constituting a majority of the population of settlements there, and Berber dynasties ruled from the end of the tenth century into the fifteenth (Abun-Nasr Reference Abun-Nasr1987; Versteegh Reference Versteegh1997). Today, Moroccan Arabic contrasts with other colloquial (i.e. largely unwritten) Arabic dialects in its especially pronounced loss of short vowels and reduction of long ones. This is surely due to influence from the notoriously vowel-shy Berber languages (Kaye & Rosenhouse Reference Kaye, Judith and Hetzron1997: 265). Yet, this variety of Arabic, the native language of millions for a millennium-plus, is and always has been all but absent from the written medium.
We might imagine a thought experiment, in which at some point in the future, in Morocco Moroccan Arabic replaces Modern Standard Arabic as the written variety. It is plausible that sociological reality would at first mean that for many or most writers, writing Moroccan Arabic would mean nudging the dialect towards standard – that is, purportedly ‘good’ – Arabic by restoring vowels that Berberization has eliminated, out of a lay sense that the elided vowels were ‘sloppy’ subtractions. This would be entirely plausible given the lay assumption that language is words rather than grammar. It is equally plausible that over time, as public memory of the earlier place of Modern Standard Arabic faded and the relentless spoken reality of Moroccan Arabic had its effect, this ‘revocalization’ would gradually disappear. As such, Moroccan Arabic would only come to be written in phonetically accurate form long after contact with Berber speakers conditioned that very variety, before which a Berber-influenced Arabic had been spoken as a living home language for centuries.
In which case: imagine future linguists supposing that this meant that the elided vowels were not conditioned by Berber because they appeared in documents so long after Berber speakers encountered Arabic. Naturally many linguists would craft deft accounts via which the vowel elisions were internal developments – but leaving as a question why this did not happen in colloquial Arabics elsewhere, and neglecting that Moroccan Arabic was created by speakers of languages notoriously given to reducing and eliding vowels.
To wit: if Celtic influence were rife as early as Old English texts, then this would be so peculiar as to be nothing less than problematic for the CH. Faced with grammatical Celticisms in Old English documents, a reasonable strategy would be to search for evidence that the Celtic influence had occurred in northern Europe via continental Celts, before Anglo-Saxon was brought to Britain. Pushing the Celtic influence back to this extent would allow for a typical extended period during which vernacular varieties are considered ‘bad’ speech, which would be the only coherent reason why, by the time it was put on the page, Old English would already have Celtic grammatical influence. That is, we would suppose that English's transplantation to British soil would have encouraged an honest rendition of a variety that in northern Europe had long been considered ‘tainted’ by aliens. The last thing we would expect is that in Britain, an indigenized, non-native rendition of English that emerged on that island would have been readily transcribed on the page as soon as it emerged.
7 CH objection number five: there are too few Celtic loanwords in English
Certainly a natural expectation is that if Celts significantly affected English grammar, then they would also have affected its lexicon, lexicon being the most readily exchanged feature in language contact. As such, it would appear problematic for the CH that, for example, Kastovsky (Reference Kastovsky1992: 318–19) lists only fourteen Celtic borrowings in English – most obsolete – beyond place names and mostly defunct ones introduced by Christian missionaries from Ireland.
However, the problem is in fact illusory. When speakers shift to a new language, they may well incorporate L1 lexical items in their rendition – but they may just as well do so to only a marginal extent, acquiring the L2 lexicon essentially in its entirety. As Thomason & Kaufman (Reference Thomason and Kaufman1988: 117) have it:
We can be confident about only one retrospective generalization in this area: if the language of a shifting population did not contribute lexicon to the target language, other than a few words for local natural and cultural items, then we can conclude that the shifting population did not enjoy much social or political prestige.
Put another way, of course, this is to say that a group without prestige (i.e. the Celts under English rule) can not contribute significant lexicon. This has been the case with various languages that are the product of language shift.
For example, the Uralic influence on Russian grammar is uncontroversial, although opinions differ on specifics. Yet there are only a couple of dozen Uralic loanwords in Russian, most of them marginal (Kiparsky Reference Kiparsky1969: 23). Similarly, despite the rich Dravidian influence on Indo-Aryan grammar, while the contact situation lent many Dravidian loanwords in early Sanskrit, the number decreases in later Sanskrit, is even lower in Sanskrit's descendant Prakrit, and then, despite a millennium of subsequent contact between Dravidian languages and Modern Indo-Aryan ones, there have been almost no new Dravidian loanwords (Burrow Reference Burrow1955: 380–6).
Thus: Russian marks objects as genitive in negative sentences just as Finnish marks objects in negative sentences as partitive; the Russian possessive is a nonverbal one marking the possessor as an oblique (u menja kniga at me book ‘I have a book’) just as in Finnish (which uses the dative) while other Slavic languages use a have verb; and so on. The Uralic impact on Russian's grammar cannot be denied. Yet there is no significant battery of Finnic or Uralic lexical items in Russian.
This means that it is theoretically unproblematic that English could have Celtic grammatical influence, such as periphrastic do, without a battery of Celtic loanwords. The useful schema is Russian/Uralic versus English/Celtic. The CH stipulates that the extent of grammatical influence is equivalent, and in evaluating that, we must keep in mind that the Russian/Uralic relation entails no significant lexical influence. If we accept that Russian contains grammatical peculiarities due to Uralic despite having little Uralic lexical influence, then we cannot dismiss the CH because of the paucity of Celtic loanwords in English.
Finally, Celtic loanwords beyond the ‘cultural’ have not been as utterly rare in English as traditionally implied. Breeze (Reference Breeze2002 and elsewhere) compellingly presents hitherto neglected Celtic etymologies of numerous words such as brag, brat, curse, and baby. It must be mentioned that these four are among the very few of the words Breeze adduces that lasted beyond Middle or even Old English: he refers mostly to archaisms now unknown. What Breeze attributes this mysterious evanescence of Celtic borrowings to, in contrast to the hardiness of most Scandinavian lexical items in English, is unclear. However, to the extent that one is skeptical that language shift could leave essentially no loanwords of significance at all, Breeze's findings will be useful: at least in the past, English picked up plenty of Celtic words beyond names of towns.
8 Why not just Celts?
Of course, the strongest version of the CH is that both the simplification and the mixture were due to Celtic influence: many CH arguments note Proto-Germanic features lacking in English that also happen to be absent or weak in Celtic grammar (e.g. Vennemann Reference Vennemann and Brinton2001 on the external possessor construction, Tristram Reference Tristram2002 on inflection). However, this kind of widespread elimination of features absent in the L1 is not, actually, typical of cases of language shift over long periods of time. Supposing so perhaps underestimates the capability of shifting speakers to acquire unfamiliar L2 features, especially given that such cases occur over multiple generations amidst rich opportunities for exposure.
For example, Dravidian languages mark biological gender and animacy but not grammatical gender. Indo-Aryan languages are replete with features inherited from Dravidian to an extent that indicates that Dravidian speakers acquired Indo-Aryan varieties in vast numbers. Yet Indo-Aryan languages tend strongly to retain grammatical gender despite its absence in Dravidian, with Marathi being an Indo-Aryan language that retains three genders despite eons-long contact with Dravidian Kannada to the south. That is, Dravidian speakers acquired Indo-Aryan grammatical gender. Or: Altaic languages have no tones, and yet indigenized varieties of Mandarin that its speakers have created are richly tonal (cf. Lee-Smith Reference Lee-Smith, Wurm, Mühlhäusler and Tryon1996).
Widespread abbreviation of grammatical machinery is typically associated, rather, with rapid acquisition within a single generation. Pidginization is the most extreme example, but the parallels between pidginization and fuller but still incomplete acquisition are well known (Andersen Reference Andersen1983; Veenstra Reference Veenstra and Plag2003). Here, the Scandinavian invaders fit the profile, being newcomers who rapidly assimilated into the pre-existing society. The Celts, indigenous inhabitants coping with Anglo-Saxon settlement over several centuries while retaining their native tongues, do not. Rather, we would expect that they would have acquired English in relatively full form, but infusing it over generations with Celticisms in additional fashion, creating an Anglo-Celtic hybrid along the lines of the blending of Australian languages described by Heath (Reference Heath1978) or Amazonian languages as described by Aikhenvald (Reference Aikhenvald, Aikhenvald and Dixon1999).
In addition, what is known about the emergence and spread of the features distinguishing English from its sisters further supports the idea that the Norse speakers subtracted from the grammar while the Celts added to it. It is well established that the elimination of grammatical gender began in the north and spread southwards. This has also been demonstrated for the loss of V2 syntax (Kroch & Taylor Reference Kroch, Taylor, van Kemenade and Vincent1997) and there is suggestive evidence along the same lines for some other constructions (McWhorter Reference McWhorter2002: 255–6). There is known to me no argument that a case of simplification proceeded from any region but the north.
In contrast, Celtic impact on English traces to various regions, and conditioned additions rather than subtractions. The emergence of do as an auxiliary is agreed to have begun in the southwest by all specialists, regardless of their position on the CH. But the Northern Subject Rule (in which third-person plural agreement (i.e. zero-marking) occurs only with (a) the subject pronoun rather than full NPs and (b) only when the pronoun and the verb are adjacent, such that Birds flies but They peel and boils them) occurs to this day only in northern Englishes.
This is exactly what we would expect of Celts shifting gradually to English across a vast expanse of space: sociohistorical and even fortuitous particularities establishing different L1 transfers in different locations. An analogy would be Finnic influence on Russian: northern Russian dialects have inherited a syllable-initial stress pattern, but central and southern ones instead likely inherited a different Finnic feature, the akan'je transformation of [o] to [a] or [ə] in unstressed syllables from Finnic languages of that area (Thomason & Kaufman Reference Thomason and Kaufman1988: 241, 244–5). What would be peculiar in Britain is if the transfers all occurred in a single region and spread from there, in which case the question would arise as to what suppressed L1 transfer in all but that given region.
We would expect, then, that if Celts were the sole or principal agents in the transformation of English, then the result would have been a Britain-wide hybrid of Germanic and Celtic grammar with no significant degree of simplification in comparison to languages of either original family. When Ethio-Semitic was influenced by Cushitic languages, the result was not that Amharic, Tigrinya et al. were oddly grammatically streamlined compared to Arabic and Hebrew. The grammatical mixture described in South America by scholars like Aikhenvald (Reference Aikhenvald, Aikhenvald and Dixon1999) does not yield languages less grammatically complex than their relatives.
What actually resulted in English suggests two processes. One group was infusing the grammar with L1 strategies, creating, as it were, a Celticized Anglo-Saxon variety, while another group equally influential, the Scandinavian invaders, was acquiring and passing on an incomplete rendition of the grammar.
I am skeptical, therefore, of the longer lists of proposed Celtic inheritances in English that include things absent both in Celtic and in English. It is not impossible that Celts learning English may have eliminated an English feature they were unfamiliar with here and there. But general tendencies in language contact suggest that these cases would have been exceptions. Cases in which Celts transferred an L1 feature into English are therefore more compelling in arguments for the CH.
9 Standard of proof: what's good for the Balkan Sprachbund. . .
As White (Reference White2002) valuably notes, to the extent that contact accounts such as the Balkan Sprachbund are accepted despite ever lingering questions and ambiguities, the evidence allows more than that Celtic ‘may have’ been ‘an influence’. It was the direct cause.
For instance, in the Balkan Sprachbund, the replacement of infinitives with dependent constructions begins in Greek as far back as the Hellenistic koine and is also found as far afield as German (Ich will, dass er geht ‘I want him to go’). It is considered sound to venture that the Sprachbund may once have extended further, albeit without direct evidence of such. Kinks in the Celtic-English situation are hardly more damaging to the CH, in view of the weight of evidence pointing to a single conclusion.
For example, Garrett (Reference Garrett1998: 295–6) argues that the CH account of periphrastic do is refuted by the fact that Celtic lexical influence is strong in the southwest only in the small region of the West Cornwall finger. For him, since in the southwest beyond this, lexical influence is slight, this means that certainly we would not expect deeper influence of a grammatical nature, such as the incorporation of an auxiliary do.
But as we have seen, grammatical transfer can occur without significant lexical influence. We can reasonably propose that Cornish speakers could have lent their English a periphrastic do even without retaining Celticisms such as whidden for ‘smallest pig of a litter’, and that the retention of such items in West Cornwall was due to its relative geographical isolation.
Periphrastic do emerged eons ago in a society in which writing was an elite activity. Thus our sources constitute an approximate refraction of the spoken reality of their times. Inevitably we will encounter odd wrinkles in the data. Our interest will be in whether English and Brythonic's likenesses are likely to be accidental in light of their typological oddness. To the extent that we do not consider the Balkan Sprachbund a dubious case short of demographic and linguistic data obviously lost to history, we must allow that a CH account be deemed to stand under equivalent conditions.
10 Conclusion
It is reasonable, therefore, to assume a direct line of influence from Celtic to English. That is, without the Celtic model, English could easily make no use of do as an auxiliary at all, or would use do in the fashion of German dialects, probably in substitutive or perhaps habitual function, and quite possibly as a permanently nonstandard feature, varying in its usages across dialects.
Technically, to subscribe to the CH account of do leaves a question: why would this be the only significant grammatical parallel between Brythonic and English? Put otherwise, if periphrastic do is a borrowed construction, then we must predict that Celtic contributed other constructions. Under this analysis, CH accounts of other features merit more attention than they have traditionally received.
It remains to be seen whether we need go so far as to treat English as virtually another Celtic grammar, as some CH accounts imply, and I for one doubt that this conception will prove valid. However, we must question the outright rejection of the CH, which includes allowing that Celtic may have had merely ‘some’ influence, where this means – in practice – continuing to document the diachrony of English with Celtic playing anything less than a crucial role.
I remain agnostic as to whether resistance to the CH is rooted in British chauvinism (cf. German Reference German2000). I might note that my embrace of the CH is not based on any previous commitment to Celtic languages or culture. In that light, my Celtic surname is an inheritance from a nineteenth-century slave-owner; I am a black American, not Irish.
It is more empirically obvious that dismissal of the CH is sparked especially by the fact that periphrastic do and other proposed Celticisms do not occur in Old English texts. It is also encouraged by the sheer familiarity of core linguistic concepts such as habitual aspect and the INFL node, which founds so much work on English diachrony.
On the basis of this article, I submit that in view of the actual nature of the evidence for the CH, this dismissal qualifies as unscientific.
