Peter Schrijver = _ ee THE CELTIC CONTRIBUTION TO THE DEVELOPMENT OF THE NORTH SEA GERMANIC VOWEL SYSTEM, WITH SPECIAL REFERENCE TO COASTAL DUTCH 1. Introduction! The medieval and modern Dutch dialects of West Flanders, Zealand and the coastal areas of Holland show a substantial number of com- mon features. Many of these link them to the other North Sea Ger- manic languages (English and Frisian) and separate them from the Low Franconian dialects that constitute the majority of Duteh dia- leets, including the modern standard language. Such features are u- sually termed Ingvaeonic or North Sea Germanic (NSGm.). In as far as they occur in Dutch I shall label them Coastal Dutch (GstDu,; the notion will be defined in section 8). The scholarly literature on Ing- vaeonic is vast? In this article I would like to suggest a new explanation, according to which a considerable number of early Ingvaconisms can be attrib. uted to the presence of a Celtic, more particularly a British Celtic, substratum. 2. Four Types of Ingvaeonisms The terms Ingvaconie and North Sea Germanic cover a wide range of developments of different antiquity and geographical extension. Thus, for instance, the Ingvaeonie loss of a nasal before "f *, *p with compensatory lengthening of the preceding vowel, as in *uns 'us! > “is > MDu. Fl. wus /7o/ is also found in OFris, is, OF tis, OS us, MLG iis beside wns, and even reached HG in some words, e.g. MoHG Stiden ‘south’ < “sunp- (the development also occurs in ON)..F1Du. normally sides with HG in preserving the nasal (MDu, ons, OHG uns ‘us’, but ef, forms like suthon, hlothun in the eastern OFrDu. Wach- tenclonck Psalter’), |PETER SCHRLIVER Geographically more limited and chronologically later ia the Ing- vaeonic unrounding of *y. i-umlaut of PGm. *w produced “y in North and West Germanic, including the NSGm. dialects and FrDu, ‘This *y was unrounded in a more limited area, comprising CstDu. (pit, pet as opposed to put ‘pit! < *putjas < Lat, puteus"), OF ris, (pet ‘pit’) and OB (pytt > pit), In English, the unrounding can be dated to the 9th to 11th c., depending on the dialect (Kentish seems to be the earliest: around or before 900)*, Oris, generally has unrounded e but there is evid- ‘ence for i in Insular North Frisian.’ The Frisian place name Bruggi- heim 1944, copy 1150-1158] (now Britsum) still shows /y/ (*bryggi < *brugié ‘bridge’; but the spelling may have been remodeled by the German seribe, cf. heim). In Coastal Dutch, unrounding is found in Lith and 12th ¢. place names: Roriks-pit (NH), Pet-heim (= Petten NE) < *pyt ‘pit’, Wuduis-hil (Z) < *hyl 'hill'.® A possible earlier exam- ple is Uuarmelde [918-948, copy end 11th el < *-muldjé, the old name of Warmond (SH; Schénfeld/van Loey 1970: 52). An early WF, in- stance is Crumbrigha [960] beside Crumbrugge [966], = Crombrugge near Meirelbeke (*hruggi- ‘backs, ridge! Mansion 1924: 116) or *brugié 'bridge’). Chinicwirde (786-787, copy early 13th ¢.] (= Kenwerd, Gro- ningen) would be particularly early if it contains the equivalent of OFris. wurd > werd ‘occupied elevation’ < *wurpi-, as seems likely. Blok (in Kinzel et al, 1988; 205) labels the spelling unclear, perhaps because the -/- considerably antedates other instances of unrounding. OF. Locuuirde [791] (Gent) may contain *waripa- ‘land near water’ @u, waard, OHG werid; thus Mansion 1924) rather than *wurpi-. Gysseling (1960a: 78) uses both forms to date the unrounding in Coa- stal Dutch and Frisian to before 786. Another Ingvaeonic development is the monophthongization of PGm. “au tod, whieh is common to OstDu. and Fris. but is not found in the other Gm. languages, including OE. In Du., this a occurs almost ex- clusively in toponyms: ef, Rethehesclat [1130-1161] (SH, near Ocgst- geest), Vronanslat [1125-1150] (SH, in Westland), Meneslata [end 12th-early 13th c.] (= Mientsloot, water in NH), Ekkerslato, Ekerslate 11405-1120] = Akersloot (NH) < *slauéa- ‘ditch' (OFris, slat, MoDu. slot; ef. latinized abl.pl. Sclautis in OF I, A.D, 696; Mansion 1924: 208). The earliest examples are Nordcha [889, copy about 1520] be- side Norhego (918-948, copy end 11th .] = Noordwijk (SH), which con- tains “gawi, “gaujo- ‘district!, and even earlier, OF, Datnesta [820- 4‘THE CELIIC CONTRIBUTION 822, copy 941] beside Dotnesta, lit. ‘dead nest’ (later Dodeman, a for- est near Drongen, Bast Flanders), which presumably contains *daup- ‘dead’ (Mansion 1924; 114; Gysseling 1960; 80), Only very few lexical non-onomastic items show the development, eg. MDu. baken, MoDu. baken ‘beacon’ < *baukna- (cf, MDu, béken ‘sign’, which shows the normal MDu. development of “az), MDu, sade (F., NH) ‘sod’ (cf, MLG sdde, OF ris. scitha, OF séad < “saupé, but ME sade, MoDu. zo (Gooree) < *s1po). Finally, yet another type of geographical spread is shown by the latinized personal names Eodbertus (< *ax) and Korneacus (< *e /_r0), both from Flanders, which show a vowel development that is highly reminiscent of OF. These names have also been labelled Ingvacon- isms. In view of the specifically English character of their vocalism, these are doubtlessly straightforward OF names.” Itis evident that these four Ingvaeonisms have different histories. The loss of nasal before voicsless fricatives belongs to the whole Northern WGi, area, and related phenomena occur in North Gm., OHG and Gothic; itis prehistoric and may therefore well antedate the arrival of the Anglo-Saxons in Britain in the fifth c. A.D.'The unrounding of *y probably occurred a couple of centuries later and has a western WG. distribution, the details of the development differing from dia- lect to dialect. The development "au > d is geographically and lexi- cally even more limited, shared as it is by CstDu., OF ris, and, partly, OS. It has been claimed that the CstDu. development should be con- nected with a Frisian superstratum. Finally, there are Ingvaeonisms that look like straightforward introductions from English. ‘Various historical scenarios have been proposed in order to account for the Ingvaeonic phenomena.” According to Kuhn (1956), they are linguistic deposits of a commercial traffic community across the North Sea that arose after the Anglo-Saxon invasions in Britain in the fifth , A.D. They could be remnants of a large northern and west- ern West Germanic linguistic centre or centres which receded under the thrusts of Franconian expansion (Frings 1944). According to Heeroma (1965), Ingvaeonisms represent conventionalized features of "careless" speech at the northern frontiers of the Frankish empire. Other theories claim that the CstDu, Ingvaconisms all or at least in part reflect a Frisian superstratum that is connected with the Frisian domination of the coastal areas in the second half of the first millen- 5PETER SCHRIJVER nium A.D. (Miedema 1970, 1971; Bremmer 1997: 68-73). I do not sub- scribe to the theory that Ingvaeonisms in Dutch all derive from Fri- sian and had arisen during the period of Frisian expansion along the Dutch coast: there is reason to doubt that the Frisian expansion stretched all the way along the Flemish coast and even penetrated so far into the interior as to produce the Fast Flemish place name Dat- nesta (cf, Taeldeman 1982: 288). Moreover, a number of developments usually ascribed to CstDu. are not found in Frisian (e.g. i-umlaut of *6, “t yielding i rather than (OF ris.) é, fronting of *ij; cf. Taeldeman 1982: 288-289, contra Miedema 1971). I do not of course deny that some Ingvaeonisms could be due to direct Frisian influence but main- tain that each item should be judged on its own merits. Nor do I wish to divorce CstDu. developments completely from developments in Fri- sian, which if not identical are usually remarkably similar; but here one should consider the possibility of shared development rather than. borrowing. One may also recall that if the genesis of NSGm. may be dated to the fourth-fifth ¢ A.D, (thus e.g. Kuhn 1955, 1956; this seems to be the consensus), we cannot really speak of a linguistic en- tity 'Frisian' (as opposed to CstDu. or English) when the ‘Frisians! started their southward expansion in the fourth and fifth ¢, A.D. ‘Thus, when we speak of developments taking place as early as the fourth to sixth centuries, the labels ‘English’, 'Prisian' and 'Ingvaeo- nic Dutch' mean very little, as NSGm. had not yet developed into en- tities that wore sufficiently distinct to merit these names." An Eng- lish superstratum in Flanders, which has recently been advocated by Nielsen (1992), has been held accountable for southern Ingvaconic Dutch phenomena. One should appreciate the fact that these scenarios are not necessarily mutually exclusive, 8. About This Article This article deals with the development of the vowel system from PGm. to the Ingvaconic CstDu, dialect or dialects until approximately 1200. Here CstDu. is defined as the early medieval Gm. dialect or dialects, closely related to the other NSGm. dialects, OF and OFris., that were spoken along the Dutch and Flemish coast and in which originated the Ingvaeonisms that are found in MDu, Franconian texts 6‘THE CELTIC CONTRIBUTION (FrDu.). CstDu. underwent strong franconian influence, and MoDu. dialects along the coast have been thoroughly franconized, showing only a limited number of CstDu. features. An essential condition for studying the development of the CstDu. vowel aystem is that one should be able to establish this vowel system descriptively. This is no easy task for several reasons. First, we do not possess medieval CstDu. texts, only FrDu. texts with supposedly CstDu. features. The CstDu. vowel system cannot therefore simply be described but must be reconstructed. Second, as we have seen, the Ingvaeonic features are historically and geographically diverse, so that it is sometimes unclear whether and to what extent a feature belonged to the CstDu. system. The de- velopment of “aw to d, for instance, has been claimed as Frisian rather than CstDu., and the fronting of * to *y could be FrDu. rather than OstDu. (see section 5.2.5-6). Third, since Ingvaconic developments in CstDu. are recognized by the fact that they differ from developments in FrDu. and since Ing- vaeonisms tended to be lost under Franconian pressure, it is difficult to determine whether CstDu. shared developments with FrDu.: seem- ingly shared vowel developments of modern CstDu. dialects (all of which are franconized) and ¥'rDu. may well reflect franconization of the CstDu. dialects rather than genuine correspondences (this is prob- ably true in the case of PGm. *ai > approximate é in modern CstDu, dialects; see section 5.2.6). Thus, the fact that *é became 7 in FrDu, and in CstDu. dialects does not necessarily mean that this is the origi- nal CstDu, development. More complicated, because it involves an interpretation of the potentially ambiguous orthography, is the case of PGm. *au. We have seen that “au became a sound that was written ain forms of CstDu, extraction (-slat ‘ditch’, In FrDu. “au became 6, as in *slauta- > sloot ‘ditch’, Medieval and modern CstDu. dialects generally show a similar reflex 6 or 09. There are (at least) two ways to explain this situation historically: (1) “aw originally became @ in CstDu., but this disappeared completely under Franconian influence, the modern dialect reflexes representing a developed state of Fran- conian 6; (2) “au originally became approximately *s: or *a:, for which early Dutch orthography for want of a better sign sometimes used a. This *o: subsequently shared the development to approximate *5 with Franconian. A form slo(o}t appearing in a medieval Dutch text 7PETER SCHRIJVER may well be CstDu. by scenario (2) but there is no ‘way to prove that it is CstDu. rather than Franconian. Early place names in presumably CstDu, areas such as Zealand usually show “au > o(0), e.g. Osborch (949; Oosthurg), Osteapellam (end 12th e.-1207; Oostkapelle), Osten- de (1189; on Walcheren), which contain o(o)st ‘east! < taust-. These forms may be genuine CstDu. but they may equally well be Franco- nian (which gradually encroached on CstDu.) or franconized CstDu. The following procedure will be adopted in order to reconstruct the CstDu. vowel system. As CstDu, are regarded those phenomena which are proper to coastal areas of Dutch and which diffor from FrDu, and agree with NSGm. phenomena (these are identified as Ingvaeonic in the literature), These phenomena should be relatively old: they must be attested in MDu. texts and/or early CstDu. place names, As to the separation of truly CstDu, phenomena from OFris. intrusions, an Ingvaeonic development can be more reliably identified as CstDu, if it is attested all along the Flemish and Dutch coasts than if it is limited to a small area, and also ifit is attested only in West Flanders than if itis found only in North Holland. In the absence of evidence to the contrary, OstDu, is assumed to have undergone the same develop- ments as the neighbouring FrDu. dialects, especially if these develop- ments are attested in early western Dutch placo names. Finally, in the absence of evidence to the contrary, the vowel system of CstDu. is considered to be monolithic all along the coast from West Flanders to North Holland. This is of course an abstraction but as far as I can see only on the phonetic, not on the phonological level. It goes without saying that there are certain problems, such as the one confronting *au, that will not be solved unequivocally, After these preliminary remarks we can move on to the actual topic of this paper, which is a study of the direction in which the NSGm. and CstDu. vowel system developed during late Antiquity and the early Middle Ages. T would like to suggest another factor that could account for many Ingvaeonic phenomena, which as far as I know has not been noticed before, viz. that of a Celtic, more specifically a British Celtic substrat- um. In order to bring out this point, the following issues will be rais- ed. First, the historical and linguistic plausibility of a Celtic, more specifically a British Celtic Presence along the Flemish, Dutch and Frisian coast will be addressed (section 4). Subsequently, the develop- 8‘VE CELTIC CONTRIBUTION ment of the PGm. vowels via NSGm. into CstDu. and a rough chronol- ogy of events will be established in section 5. Section 6 is the key section, in which the British Celtic vowel system is compared with the vowel systems of NSGm. and CstDu. 4. Celtic Along the Dutch Coast? There is no doubt that when the Anglo-Saxons settled in Britain in the 5th c. A.D, they entered into intensive contact with people who spoke a British Celtic language (Jackson 1953: 194-261). Since it is hard to believe that the Anglo-Saxons killed or expelled all Celtic speakers whom they encountered, we must assume that, the disap- pearance of British Celtic from most parts of Britain was effectuated by a process of language shift: speakers of Br. acquired a WGm, dia- lect as their second language and in the course of time became WGm. monoglots. Although English must have had a Br. substratum, actual linguistic traces of this substratum, apart from onomastic material, are few and far between. Loanwords that can be dated to the second half of the first millennium, for instance, are extremely scarce.” Yet the historical and place name evidence is such that the presence of this substratum cannot be doubted. ‘There are reasons to assume that a similar situation occurred in West Flanders and Zealand, Approximately between Dunkirk and the estuary of the Maas, there lived a Belgic people known to Caesar as the Menapii. The Belyae spoke a p-Coltic language, ‘Tribal nameg such as Morini 'Sea-people', Eburones 'Yew-people', Atrebates ‘Sett- lers', Caleti ‘Tough Guys' leave no doubt: as to their Celticity.° Lin- guistically, the Belgae were probably closely akin or all but identical to the British Celts. Classical sources and a study of tribal names bear out that there was extensive Belgic immigration in southern Bri- tain, and itis altogether possible that the speech of these immigrants 4s directly reflected in the surviving southern British languages, Cor- nish and Breton. A direct piece of evidence for the type of Celtic that the Menapii spoke is probably provided by the toponym Helinium, the Roman name of the estuary of the Maas, which most likely reflects Proto-Celtic *sel- 'marsh' and shows the typically British Celtic gen- eralization of lenited *s- > h-, cf. MCornish “heyl ‘estuary', MoWelsh 9PETER SCHRIVER, hel ‘river meadow'."” The element hel- found in Helinium probably survives in modern toponymy as a designation of low-lying, marshy areas, which, significantly, is common in West Flanders (e.g. Hel- ackere, het Helleke). While the assumption that a British type of Celtic was spoken in West Flanders and Zealand at the beginning of our era seems reason- able enough, a greater effort should be produced in order to convince the reader that this was also the case further to the north, more par- ticularly among the Frisians who were situated on the Rhine estuary around 100 A.D." In Holland, the Cannenefates of Roman times may have spoken Germanic (Schrijver 1995¢) but they are known to be recent immigrants, who, together with the Batavi, split off from the Chatti, perhaps after Caesar's time. They may or may not have met Celtic speakers upon their arrival at the Dutch coast. As to the Frisii (Frésii), none of the etymologies explaining the name as Germanic exceeds the status of a vague possibility (cf. Ram- at 1976: 37-88; Neumann 1996: 3-5). Hence there is no good reason to think that the name is Germanic, or even Indo-European. As far as I know Celtic does not offer a plausible alternative etymology of this name and in this respect it fares as badly as Germanic, There is no formal reason, however, why it could not be Celtic in origin or why, if the name has a non-Indo-European origin, it could not have been adopted by a Celtic language in the attested form Frisii or Frésii, The fis certainly no obstacle; British Celtic developed a word-initial f- from "s- before r, as in Welsh frivd ‘stream’ < “sru-tu-, There are clear Continental Celtic parallels for this development: (Schrijver 1995a: 448-44), However this may be, in general, of course, the origin and etymology of a tribal name say little about the language spoken by the members of the tribe. The fact that the present-day Parisians have a Gaulish tribal name, Parisii, which itself doubtlessly has pre- Celtic roots, does not imply that the Parisians speak Gaulish. Thore are throe indications that a Celtic linguistic element was once present in Frisia: (D) The toponymic element hel- ‘low land, marsh’, which may be con- nected with Br. *hel- 'marsh, estuary’, is common in Friesland (de Hel, Lytse Hel ete,);" (2) The names of the Frisian noblemen Verritus and Mal(Lorix (1st c. A.D.) can be interpreted as Celtic names (note especially Ver- < “uper-, 10‘THE CELTIC CONTRIBUTION ef, Gaul. Ver-cingetorix ete.; one could translate them as ‘Super-Run! and 'Praise-King’, respectively); it must be admitted that these names could easily have been borrowed; (8) The old name of the island of Terschelling® was Wicsile [first half Lith ¢], later (latinized) Wuxadia, Wicsile may reflect PBr. *jxsel- ‘high’ with regular NSGm, unrounding of the rounded front vowel and pre- servation of the original rounding as an on-glide w-."' PBr. *jasel- < PIE *oupselo- ‘high’ survives in Welsh uchel, Breton whel, Old Irish dasal and in the Gaulish toponym Uxello-dunum ‘High Town’. Wiesile would then mean ‘height’, referring to the island emerging from the sea. ‘On the basis of this evidence it seems at least possible that Frisia in the Roman period was (partly?) a Celtio-speaking territory. If so, the names with Hel again suggest a British type of Celtic, It may well be that the ‘Celtic’ Frisians were germanized at some point during late Antiquity:”” there is no doubt that the Frisians we encounter in the second halfof the first millennium spoke Germanic, although we cannot exclude the possi- bility that isolated pockets of Celtic speakers still survived.” We may conclude that the presence of British Celtic speakers is certain in the case of Britain, probable in the case of West Flanders and Zealand, and possible in the case of Holland and even Frisia, Even if we admit a British Celtic presence so far north, it docs not imply that we expect a sizeable British Celtic impact on the Germanic languages of the area, After all, there seems to have been very little impact in the case of OE. I shall now return to my main concern: the CstDu. vowel system and its genesis. 5. The Coastal Dutch Vowel System ‘We shall now turn to a reconstruction of the CstDu, vowel system, in order that it may later be compared with that of OF ris. and OE and confronted with the British Celtic vowel system (section 6). 5.1. The Proto-Germanie short vowels in Coastal Dutch As is well-known, the PGm. short vowel system consisted of the fol- lowing vowels: *i *e “au, iPETER SCHRIVER 5.1.1, *u and #0 PGm. *u had an allophone *o which occurred if the following syllable contained a non-high vowel unless u was followed by nasal + conso- nant or ‘j, in which case *u was retained. Cf, *gulpa- 'gold' > *golpa- > OF OFris. OS OHG gold; “tungon ‘tongue! > OF OF ris, tunge, OS tunga, OHG zunga. There are many exceptions, which are probably partly due to paradigmatic leveling and partly to the influence of neighbouring sounds: ef. OF OFris. OS full, MDu, (CstIu.) vul (with “u > [5)) vs. OHG fol < *fullas. As a result of the loss of qualitative oppositions in final syllables and apocope, a phonemic contrast arose between *u and *o, While in many Dutch dialects this opposition be- tween *u and “o was again eliminated, it was retained in OstDu., whence a number of words with u entered FrDu., eg, MDu, buse 'for- est! < *buska- beside MDu, bosch, and MDu, vul mentioned above (van Loon 1986: 70, Schénfeld/van Loey 1970: 93-95, Franck 1910: 69). At a certain date, probably in tandem with the fronting of PGm. = WG. * (see 5.2.5 below), this u was fronted. In later OFrDu., pre- sumably also in CstDu., fronted u was lengthened to g in open sylla- bles, ef. western MoDu, veugel ‘bird’ < *fujlas, MoDu. zeug ‘sow! < “st16. Short fronted *u became a rounded central mid vowel /3/, which it still is in MoDu. By i-umlaut, PGm *u beeame *y, which in CstDu. was unrounded to t, presumably already in the eighth e. A.D. or earlier (see section 1). In closed syllables, this became “7, The reflex of this was written i ore in MDu. (as was the reflex of PCim, * > 1, Francle 1910: 65-68), and it was short (pit, pet 'pit's rig, reg 'back'), In western MoDu. dia- ects, predominates in Flanders and Zealand, while e is commoner in North Holland. It cannot be determined whether this reflects an old isogloss within OstDu. or whether eis an OFris, import, In open sylla- bles (ike "kruplis > *krypl > “krypal), the unrounded *y > *i waa re- gularly lengthened toé, ef. western MoDu. dialect krépel ‘cripple; nar- row passage’, Crepel-fliet [12th c., copy + 1420] (NH), cf. MoDu. hreupel, OF crypel > cripple; cf. also PGm. *ubilas > *ypil > *ipil > MDu. evel ‘bad! (beside MoDu. ewvel, MoHG iibel) (Schénfeld/van Locy 1970: 50). In inherited Gm, words, “0 could not be followed by the umlaut fac- tors “i and 'j. The combination seems to have entered the language in loanwords, The following place name indicates that *o could undergo 12‘THE CELTIC CONTRIBUTION umlaut in OCstDu. and that the expected umlaut product of *o, viz. *p, merged with *y and was affected by the same unrounding as *y: Gallo-Romance *brogilos 'hedged-in plot of land' > OCstDu. *broyi > OCstDu. “bryyl- > *briyl- > Den Briel (1257, SH], De Brele [1297] This form shows the Ingvaconic loss of *-} and the subsequent con- traction of *1 or, with ODu, reduction of the unstressed vowel, “za, to MDu. “é > *7. The etymon must be compared with French breuil < *brogilos, Medieval Latin brogilus, ultimately from Gaulish *brogis ‘and, border’, Welsh, Cornish, Breton bro.” 5.1 ‘a In western WGm., PGm, *a had an allophonic fronted articulation in most phonetic contexts, as its most common reflexes in OE (e) and Oris. (¢) show, 'Traces of this can be found in CstDu. forms in MDu. and in modern western dialects, e.g. dek beside dak ‘roof, step beside stap ‘step’. It cannot be determined whether and how this fronting was conditioned in CstDu. ‘Traces of rounding of short "a before nasals, as occurred in OF ris. and OE, are not found in CstDu., but this may be accidental, the ex- pected *o having been eliminated under FrDu, pressure, i-umlaut of a in WGm. produced a closed %, which in OHG and MHG was phonologically distinct from the more open PGm. “e (Braune- Eggers 1975: 30). In MDu. these two types of ¢ cannot be distinguish- ed on the basis of rhyme (Franck 1910: 38°") but they must have been distinct: a number of modern Dutch dialects show different reflexes (Goossens 1980: 204-5), 5.1.3, “i and *e In early Germanic, "i became *e if the following syllable contained a middle or low vowel, a rule to which there are many exceptions, e.g, PGm. *fiskas > MDu. visch, MoHG Fisch ‘fish’, Short * was lowered to *r during the ODu. period, as is shown by the fact that in open syllables the product of late ODu. lengthening of *i became é, not i, and in closed syllables the reflex of *i was spelt i or e (Franck 1910: 18PETER SCHRUVER 65-67), which may represent // and/or dialect differences.” CatDu. presumably took part in “j > % since the product of unrounding of “7 (a clear characteristic of CstDu.) was spelt with e and i, 5.1.4. Summary of short vowel developments The PGm, short vowel system, which consisted of i, e, a, u, was modi- fied in CstDu. as follows: as a result of i-umlaut and a-umlaut, the system was enriched by o (a-umlaut of 2), ¢ (i-mlaut of a) and *y (- umlaut of *z and *o). y was subsequently unrounded to i (but perhaps toe in NH dialects, which may betray Frisian influence), "2 became fronted. Like in FrDu., "i was lowered to rand fronted *u was lowered to *5, At least in some phonetic contexts, *a was fronted to *w, which in MDu. was written and in MoDu. merged with e: or e, depend- ing on whether it occurred in an open or closed syllable, 5.2. The Proto-Germanic Jong vowels in Coastal Duteh PGim. had the following tong vowels: *f, 82 (= %6,), "8 (= 2), 86, 8.” 5.2.1, 87 “i was retained as a monophthong (usually short) in the modern coastal dialects of Weat Flanders and Zealand, but it became a diph- thong (approximately [e1]) in many parts of FrDu, in the course of the MDu. period (van Bree 1977:167, Schinfeld/van Loey 1970:91), In view of the geographical distribution of these reflexes of “7 in western Dutch dialects it seems likely that non-diphthongization was characteristic of CstDu. while diphthongization was rooted in western FrDu. 5.2.2, %8 In FrDu., PGm, "@ passed through a diphthongal stage “ia > tie > *ia before becoming MDu. i, spelt ie (Gyaseling 1992; 25-35; van Bree 1977: 4‘THR CELTIC CONTRIBUTION 162-63). Franconian-based Standard MoDu. has short i < %2. A number of western MoDu, dialects have i < PGm. *é vs. i < PGm. *i, which probably reflects the OstDu. situation (e.g. Zealand /ri:m/ ‘belt! < *é vs. /rim/ 'rhime' < i), 5.2.3. ee *g retained a fronted articulation in OstDu. while it became a in FrDu. Cf. PGm. “muépa- ‘hay field’ in carly place names in Zealand and West Flanders, e.g. Birne-mede, Bret-mede, Hol-met [1181-1210]; Voplinmet, Woburgmet [1171] (Gysseling 1992: 23-24), In western MoDu, dialects, FrDu. c has conquered WEI, while most of Zealand and part of South Holland retain “and part of North Holland has é, ei (van Bree 1977: 156). There is no CstIDu, evidence for *# > 6 before nasal, as in OFris. ména ‘moon! < *min-. 5.2.4, *6 *6 became a close vowel in FrDu., which gradually developed into *i © Standard MoDu. short u) via a diphthongal stage *uo or *uo. There is no evidence that “@ in OstDu, passed through a diphthongal stage, In western Du, dialects, the process was not completed until MoDu, times." Along the coast, 12th , spellings with (beside ) of place names in Zealand such as Berten-huc [1177-87] (< “hdka- ‘corner’, Culant [1189] (< *k6 'cow'), Ost-mur (1181-1210] (< *méra- ‘marsh’) suggest that CatIDu, had a close articulation by that time. 5.2.5. "i “ii was fronted to % in western FrDu, The dialects of Holland and western Brabant subsequently developed a diphthong (approximate- ly ley] or foil) by the end of the MDu. period, In modern dialects, this diphthong is found all through Holland, while Zealand and West Flanders have retained the monophthong, usually as short y, There are western Du. place names which seem to have retained the non- fronted articulation of “a, as in Souburg (locally pronounced (suborx]) < Sutburch (1162) < *stip- 'south', This phenomenon is regarded as an 16PETER SCHRISVER Ingvaconie survival by some scholars (e.g. Schinfeld/van Loey 1970: 86), but W. de Vries (1983) cast serious doubts on this interpretation: forms with @ in place names in North Holland may be due to direct Frisian influence, while those more to the south are unreliable for various other reasons. ‘There are structural arguments to support a CstIu. origin of the fronting of a and plead against. Schénfeld/van Loey, who angue that retained i may be Ingvaeonie: the undisputedly CstDu, character of the fronting of short *u (aoe 5.1.1 above) is structurally connected with the CstDu. shift of older *y (by i-umlaut) to *s, whether by a push ora pull chain, In CstDu,, the long *y that was the product of i-um- laut of *a shifted to "7 (see 6.2.9 below), The parallelism of this devel- opment to that of its short counterpart strongly suggests that the fronting of *i to * belongs to the same linguistic system and should therefore be labeled CstDu. ‘The westera FrDu. fronting of *% may therefore well have originated in CstDu. 5.2.6. Monophthongization of diphthongs There is CstDu. evidence for *au > d, which was discussed in section 1, and for “ai > G. Apart from examples in place names like Haam- stede (Haemstede (1229 copy 14th c,] Z) < *xaima- (OHG heim, MDu. ‘hém) this development is found in a few common words like MDu. lad(der(e) < *hldddr < *hlaidr. (OHG leitara, Oris. bladder, OB hledder), MDu, eldver(e) < *klaifr- (MLG kléver, OF cldfre).® All modern coastal FrDu. dialects generally show up reflexes of the FrDu. developments *au > 6 and “ai > é (van Breo 1977; 143-144, 147), Western FrDu. dialectal forms like MoDu. toon 'toe', MDu. toe < “taihwo (beside MDu. tee with Frano, é) reflect CstDu. *ta(n), which seems to have developed in the same direction as OE td > MoE toe.® An interesting alternative to this questionable OE-CatDu. isogloss is the idea that ¢oon and similar words are historically incorrect fran- conizations of CstDu. “tdfn). Bither possibility presupposes CstDu. “ta(n). Other instances like toon are MoDu. roop ‘rope’, roof ‘ball of thread’, moot ‘slice’, oot ‘wild oats’ One might wonder whether the development of *ai > d to modern coastal dialect é, i is continuous, in other words, whether CatDu. @ 16‘THE CELINC CONTRIBUTION developed into é (> 2), This is probably incorrect for the following two reasons: (1) ladder, klaver, Haamstede etc, have « not e (but one might argue that d in these words entered FrDu. and was preserved there, thus yielding the attested forms which, like so many other forms, entered the modern coastal dialects). More importantly: (2) the primary reflex of *ai in CstDu. is @ while that of PGm. * is @. In other words, *ai developed into a lower vowel than did *# in CstDu. In modern coastal dialects, the situation is reversed: the reflexes of “ai are generally higher (6, i) than those of *é, (, 6). In view of the fact that phonetic developments take place along a continuous scale (@> & > é), it would be implausible if CstDu. a > é "jumped" across #@ < e, without affecting it, One could avoid this conclusion if one were to assume that “ai > *d followed a route towards é that did not cross *#, for instance *a > *d > é, Whereas this possibility cannot be wholly excluded, it calls for special pleading which cannot be backed up by any attested forms showing such an intermediate stage. Under the circumstances it seems simpler to assume that *ai > é, i reflects the FrDu. development (“ai > “é) which ousted the original CatDu. ré flex.** In the case of “au > CstDu. dvs. approximate 6 of modern coast- al dialects, the continuity of the reflexes can be neither proved nor disproved (FrDu. has “az > 0, so modern coastal 6 can easily be a FrDu. intrusion), While PGm. “iu and ‘eo (< *iu by a-umlaut) remained distinct in most Dutch dialects (> MDu. ie and uu > ui, respectively), they ulti- mately merged in WF'L., where both yielded MDu. ie /i/.* It is possible that the development to ie took place throughout the whole CstDu, area: In Holland and Zealand, there are instances of “ix > ie [i] before r (dier ‘dear’ beside duur < *diur, which may be CstDu, relics; Schén- feld/van Loey 1970: 80-81, van Bree 1977: 153). hiet ‘calf of the leg’ (beside huit < *kiut-) is common not only in Flanders but also in Zea- land. In OF rDu,, the development of *eo to A/ probably passed through the intermediate stages io > diphthongal ie ({ie] or [iol), which have also been assumed for the area around Gent (Bast Flanders).*” It seems unlikely that *eo and *iz in CstDu. became i through a com- mon intermediate stage *io™ since CatDu. retained the distinction be- tween PGm. “uv and “o, What did happen to "iz is difficult to ascertain in view of the lack of reliable CstDu. evidence for the intermediate stages of *iu and "eo > é. If *iu became *iy along with the rest of Du. WwPRYER SCHRIJVER (van Bree 1977; 152), it may subsequently have yielded OstDu. "ii >i 4&8 a result of the CstDu. unrounding (*piudisk- > “piydish- > MCstDu, dietsc Dutch’), *eo, on the other hand, could not have shared this de- velopment. It may have become i/é in the same way as it did in Fran- conian, thus merging with the reflex of PGm. “é. Along the coast, forms with PGm. "eo can already be spelt as , in 10th-11th e. toponyms such as Dipanha [976], Depena [1003] > Diepene labout 1210; 2) < *deop 'deop’, MoDu, diep and (Suthera) Suthflita (976; Z] < “fleot- ‘stream’, MoDu. vliet (of. Gysseling 1992: 75). The spelling variation suggests that "eo merged with %é, before becoming i. It has beon argued that the development of *eo toi may have spread from Merovingian Gaul, where the evidence for monophthongization stretches back to the 7th-8th e. (Gysseling 1992: 74-75). Under the circumstances it seems safest not to draw any conclusions about the exact CstDu, developments before the 10th c.”” 5.2.7, Vowels before nasal + fricative In all WGm. dialects with the usual exception of HG and FrDu., a short vowel in *VIVX (¥N = any nasal, *X =/; p, s) was lengthened and the nasal lost; all Gm. languages underwent this development in the case of *Vnx. If the vowel was a, it also became rounded in NSGm. (OS has @ beside 6), The loss of the nasal is also found in CstDu., e.g. in MDu. uus < uns ‘us. The development “a > 4 in this position is presupposed by MDu. brochte ‘brought’, dochte ‘thought! < *bréchte, “dochte < *branat-, "panxt- (beside MDu, bra(elchte, dachte) and di- rectly attested in western dialects: boee ‘part of a cow-shed’ < *bds < “bans-, cf. LG banse ‘wheat barn’; goes ‘goose’ < "gds < “gans-, ef. MoDu. gans (Schinfeld/van Loey 1970: 27). These forms show that “an > "6 merged with PGm. "6 and became MoDu. ti . 5.2.8. i-Umlaut of long vowels In CstDu,, long vowels underwent j-umlaut, while in western FrDu, they did not (on FrDu, see Goossens 1980; 178ff}), The following ex- amples may be cited: 18