The dialect position of the Old English Orosius.

My good friend the mediaeval Welsh literary historian Dr. Andrew Breeze contends in several recent articles (1991, 1992ab) that the translator of Orosius's Historia adversus Paganos into Old English was a Cornishman. I do not think for a moment that his arguments hold water; but he may justly claim to have raised a reasonable question which Old English specialists have for a long time ducked. What dialectally was the place of origin of the Old English Orosius? The methods traditionally used in Old English dialect study offer little prospect of placing sub-dialects within early West Saxon (cf. Campbell 1959 [section] 20), those pioneered by Gneuss (1972) and consolidated by Hofstetter (1987) if anything less (cf. Kitson 1993: 1, 47). The only recent editor, in what is in many ways a splendid edition, discusses "The language of the manuscripts" purely in a frame of reference of early West Saxon as a whole (Bately 1980: xxxix-1v), declining to draw any finer conclusions "since we know so little about the sub-dialects of eWS" (Bately 1980: xxxix). So Dr. Breeze, whether one agrees with his findings or not, actually holds the field at present.

All the same, his reasoning is not credible. It depends entirely on phonetic details in a small number of proper names. That is no valid basis for argument about the work as a whole, since the names are all foreign ones, not likely to have had normal forms in the translator's dialect whatever it was (accepting as a working hypothesis Dr. Breeze's assumption (1991: 153-4; 1992a: 271) of a single "author" for the Old English Orosius). He might have altered phonetic patterns in particular items to conform better to those he was familiar with, or he might have left what to him were weird forms in a written exemplar severely alone. Forms of proper names not belonging to the language of a text may have a variety of relations to those of the text at large. That is particularly true of this one. For it is precisely for aspects of the phonology of proper names that there "remains unchallenged" (Bately 1980: cix) after more than a century (1) a consensus that the source was oral dictation to a native Old English-speaker by someone who was not a native Old English-speaker.

The treatment of intervocalic stops indicates to Bately (1980: cix-cxvi; 1966: 261-267, 270-280) specifically that the dictator was Welsh, sharing a linguistic origin with Asser bishop of St. David's rather than with one of the continental scholars King Alfred also attracted to his court (Asser [section]78). The whole ensemble of phonetic peculiarities not plausible as ordinary scribal error find explanation if the dictator was a Welshman; if he was a Romance or High German-speaker only some of them do (Bately 1966: 294 etc.). Breeze (1991: 153) reasonably points out that "The peculiarities of dictation mentioned by Janet Bately apply equally to Welsh, Cornish, and Breton." He cites as circumstantial evidence of the contribution of the latter two nationalities to Anglo-Saxon scholarship in Alfred's time some of the considerable number of manuscripts containing Breton and Cornish known to have been in England in the tenth century, and would have it (1992b: 432) that since "such names as Ualentinianus, Ualerius, Uespasianus, Uitellius, or Wascan 'the Basques' show no trace" of the strengthening of articulation of initial w- to gw- characteristic of all the P-Celtic languages but completed earliest in Welsh and latest in Cornish, the dictator is likeliest to have been a Cornish-speaker.

That may be so, but the non-existence of initial gw- in Old English, acknowledged by Breeze, would make an Old English-speaker taking dictation from a Welsh-speaker likely to ignore the latter's strengthened articulation of initial W-anyway. The famous eleventh-century Gospatric writ, cited by Breeze (1992b: 432) together with a form Cwoespatrik from 1254 as showing what Anglo-Saxons would be likely to write for Old Welsh (in this case Cumbric) Gw-, is a red herring, not only because of the wassenas in the same writ, but because of the fundamentally different linguistic situation in the two cases. The relevant words in the Gospatric writ came from a Celtic vernacular, and were known as such to Anglo-Saxon scribes who wrote down how they sounded. The first four names in Breeze's list are Latin names. Scribes could not have failed to know this from the context; and it is highly likely that scribes used for such a purpose would have had known that U- was and Gu- was not an initial consonant(-sequence) in Latin, and would have corrected if necessary for a Welsh accent accordingly. It is intrinsically likely, though this is not provable, that Anglo-Saxon scribes writing Latin, even ones who had not learnt the language, would have been taught or picked up pretty quickly the differences between Latin and Old English spelling-conventions, such as that u, in normal Old English spelling always a vowel, in Latin could be a consonant, and in initial position followed by a vowel practically always was. It is even possible that Anglo-Saxon scribes had their own strengthened pronunciation of initial U-, or at least were acquainted with something like the modern dichotomy between [w-] and [v-] in English pronunciations of Latin, since on the rare occasions when u is used as a consonant in Old English (Beowulf 1799 hliuade the most famous) it replaces voiced f standing for [v]. As for Wascan, that name from whatever ultimate source is naturalized Old English, as the initial consonant suggests and the grammatical inflection conclu sively shows, so a dictator's exact pronunciation has no bearing on it.

Moreover that section of Dr. Breeze's argument rests anyway on improbable assumptions about use of scribal manpower. He assumes without discussion (1992b: 432) that the dictator responsible for the Celticized name-forms was identical with the translator into Old English; but that is neither proven nor likely. In a land where Latin learning was so rare that scholars had to be brought in from abroad to restore a basic level of competence, as King Alfred famously describes it in the preface to his translation of Gregory the Great's Cura Pastoralis, teaching Latin is what they would be used for. It would make sense for them to be involved in the discussions about meaning which were the preliminary stage of translation, as the king says they were in the Pastoral Care; it would not make sense for them to be responsible for the details of translation into a tongue not their own, and it would be a complete waste of their time to have them dictate for copying by scribes a full-length book in that tongue. The rational assumption must be that the circumstance in which Celticized forms entered the scribal tradition was dictation by a native Celtic-speaker to Anglo-Saxon scribes of the Latin text. Several copies of it would have been rather quickly needed if the stage of preliminary discussion described by Alfred for the Pastoral Care took place also for the Orosius. Since as Bately (1970) has proved, Alfred was not himself the translator of the Orosius, arrangements for it may have been different; but either way, by his own account he set about organizing multiplication of basic texts in their original Latin (in which both import and new copying of manuscripts played a part) before it occurred to him to start his programme of learning in Old English. The underlying contrast between U- spellings in the Orosius and forms such as Asser's Guuihtgaraburhg, to which Breeze (1992b: 432) draws attention, is not between Cornish and Welsh pronunciation or spelling but between names which were not, and were, written down by native spea kers of a P-Celtic language whichever.

The point about scribal manpower was raised in correspondence soon after the publication of Bately (1966) by Professor P.A.M. Clemoes, who aired the possibility of a subtler interaction of nationalities than any so far mentioned. He asked "Would an Englishman who had learnt his Latin under Welsh influence produce the same phenomena? There is a general probability that an Englishman would have been chosen to dictate a long and complicated work in English if a suitable Englishman were available." Professor Bately replied that such evidence as there is for carly Welsh educational practices is that a reformed classicizing pronunciation of Latin had spread from Carolingian France before the second half of the ninth century. "Thus a Welshman would learn to pronounce intervocalic d as [d] but might occasionally accidentally give that symbol the value it had in Welsh; an Englishman taught by Welshman might be expected not to make that mistake." The sporadic nature of the substitutions in the Orosius must mean that ev en if 'Welsh' pronunciation of Latin was still current "the dictator (whether Welsh or English) deliberately substituted the 'reformed' type of pronunciation when dictating, occasionally slipping back into the 'Welsh' type he had originally learned (a Welshman would perhaps make more mistakes in this than a Welsh-trained Englishman), or that the scribe was familiar with the Welsh convention and able to write d where the dictator used the sound [d], p when he used [b] etc., only occasionally forgetting to do so." Secondly, and "fairly conclusively, certain forms would seem to be explicable only in terms of a native Welsh speaker--notably forms involving alterations to initial consonants, where mistakes would appear to be due to differences in points of articulation between Welsh and English (when you would expect sound-substitution on the part of an Englishman) or to Welsh sentence phonetics (which would certainly not influence an Englishman reading a text in his own language)" and presumably not in Latin eith er. (1a)

Bately (1966: 301-303; 1980: cxv-cxvi) argues that the dictation which gave rise to the peculiar spellings was of the Old English not the Latin text; I do riot think even that is likely. Her reason is that in other Anglo-Saxon manuscripts whose language is Latin, substitution of d or p for th of classical proper names is extremely unusual, in ones whose language is the vernacular less so. But the substitution in the Orosius being ford not th is as she notes practically unparalleled in manuscripts in either language, so that argument from relative probability is unusable anyway. (It is analogous to a [chi square] test two of whose cells have values of 2 and 0.) In my opinion this is one of the respects--there are several--in which it just has to be accepted that in its linguistic background the Orosius stands apart from all other Old English texts. (1b) The obvious explanation of it lies in the circumstance of dictation itself. Anglo-Saxon scribes copying Latin texts from exemplars did not substitute native sy mbols for Roman ones, but writing from dictation they simply were not capable of consistently substituting for a sound that was phonemic in their own language and was only expressible in native symbols a sound, also phonemic in their own language, that contrasted with what they heard. If a dictation so abundantly affecting proper names really had been of the Old English text it would be bound to have left some traces in Old English words, but none have been found. The only common noun adduced with any is the exception that proves the rule, since it is not English but a Roman official title dictator spelt as tictator six times (Bately 1980: cxci). (2) The very consistency of tictator points to its being, however improbable to our eyes, a taught spelling, since names appearing so often, whether or not affected by the dictation, are not usually spelt quite consistently. (A likely reason why, in a Welsh oral milieu, special attention might be drawn to that word, is that -ct- 1S a non-Welsh sound-sequence, Old Wel sh written -ct- being an archaism, usually for -ith- but sometimes for other sequences.) (3)

The three names on which Dr. Breeze mainly bases his argument that the dictator was a Cornishman specifically all fail as evidence, because all the phonetic details to which he points could readily arise within Old English. Explanation invoking nationality of a non-native Old English-speaker is simply uneconomical. Thc first is Ercol 'Hercules', to which Breeze (1991: 152) contrasts Erculus in Alfred's Boethius, one of many differences between the two texts in the treatment of Latin names and technical terms (Bately 1970: 440-442) and in vocabulary generally (Bately 1970: 442-450). He assumes the vowel o to need special explanation, and finds that in a regular development of "British and Latin u...to o...common to Cornish and Breton, but not Welsh". The real explanation is that where King Alfred treated the name of Hercules gingerly, with what readers would recognize as a Latin nominative inflection (if not actually the right one), to the Orosiustranslator it was familiar enough to have been more or less natu ralized. His normal forms are an Old English endingless nominative (ed. Bately 727, 731) and accusative (ed. Bately 3015) Ercol and the corresponding Old English genitive Ercoles (ed. Bately 99, 15, 2021, 7211). Only twice (ed. Bately 8020, 8112) is the ending -es used for a case, accusative, to which it is not appropriate in Old English. (There it is not appropriate in Latin either (regular accusative Herculem); but neither in the Latin is accusative, one is nominative (Sweet 1883: 1499, 1514), and the two are so close together as to suggest a trace either of Homer nodding or of the activity of a less skilled assistant on that page.) Now what Breeze has not taken into account is that in unstressed syllables OE u where it occurs interchanges freely with o (Campbell [section][section] 355(5), 373), with o much the commoner, and before liquid consonants, and to a lesser extent nasals, where the vowel might either be pronounced as such or merely indicate a syllabic sonant, u/o interchanges fairly freely with e, the outcome being conditioned as much as anything by vowel harmony (Campbell [section][section] 363, 381, 385). The mid vowel o is more harmonious with e than the high vowel u is; so Ercol is simply what is to be expected as a naturalized Old English form. The last two syllables of Erculus show vowel harmony in the other direction; the non-involvement of the stressed syllable in that is one aspect of what I have called Alfred's gingerliness in handling the name.

The uselessness of a single form of such a variable item as a dialect criterion is graphically illustrated by the forms of the word stapol 'pillar', fairly common in the boundary descriptions of land charters. (4) They are: endingless accusatives and nominatives stapol Warks S55, Worcs 5726, S786(xv), Gloucs S7869x), Wilts S492(i), S635, S767, S275(i), S229(i), S1588, Hams S754, S378(i), S944(i), stapul Devon S255, Wilts S891(i), S393(i), Hants S381(iii), stapel Gloucs 5467, Som S292(i), S292(u) x 2, Dorset S656(i), S277, Hants S619, S811, Sussex S562(i), Oxon S1028(i); datives in -e stapole Warks S55, Worcs S726, S786(xv), Gloucs S179(ii), S1556 (one MS), Wilts S492(i) x 3, S1811, S493, S635, S767, S766(i), S1588, Hants S268, S412 x 2, S463 x 3,5 S619, S693(u), S754, S381(i) x 3, S800, Kent S1215, Beds S772 x 2, Berks S591 x 2, stapule Gloucs S1556 (both MSS), Wilts S1811, Hants S268, Berks S577, S761, S964 x 2, stapele Worcs S1174, Gloucs S179(ii), S1556 (one MS), Dorset S419(ii) x 2, S442, Wilts S1586, Han ts S811, Sussex S562(i), Oxon Surrey S1165(i) x 2, S1028(i), stapile Dorset S442, steaple Kent S293(ii); (6) accusative plurals in -as stapolas Gloucs? S1862(i), stapulas Devon S830, S1003, stapelas Worcs S579, staples Som S509; dative plurals in -um and weakened variants stapulum Worcs S579, stapulon Devon S1003, stapelum Bucks S138 (ii), stapelan Hunts S566, staplum Devon S830. Fluctuation between S393(i) stopul and S275(i)/S229(i) stapol in bounds derived textually from Wilts S891(i) stapul, and occurrence of stapole beside stapule in Wilts S1811 and Hants S268, stapele beside stapole in Gloucs S179(ii) and stapele/stapole beside stapule in Gloucs S786(x), even stapel beside stapole in Hants S619, are especially telling. Most of the southwestern -e- forms are in late corrupting cartularies, as are the obvious corruptions staples and stapile. Such regional patterning as there is when corruptions are removed seems to be that -u- is commonest in the extreme south-west and -e- in the extreme south-east and/or in very late Old English and post-Conquest texts. If the charters show anything as the form to be expected in Old English speech from Cornwall it is -ul not -ol. There are no instances of stapol from Cornwall, but in the ones from Devon, all in good texts, -u- is the only relevant vowel, in three different phonetic contexts, S255 stapul, S830 stapulon, S830, S1003 stapulas. However, since these include none with a dative -e, we are not really in a position to weigh the probabilities; and it is clear that -ol(-) was the normal West Saxon form. The Orosius-translator uses it also in Escolapius and Escolafius for Aesculapius (ed. Bately 328, 764), and in Daedolas for Daedalos (ed. Bately 723). The relevance of vowel-harmony may be seem in his keeping -ul in Nuchul (ed. Bately 119), his changing Pelorum or Peloris to 214 Polores (cf. Bately 1980: 427), and with west midland o for a before nasal 2728, 29 nominative Tontolus, genitive Tontolis 'Tantalus'. Charter forms are not yet as generally accessible for philol ogical purposes as might be wished, but these forms from the Orosius are readily found in Bately's fine glossary of proper names. I do not think Dr. Breeze has taken the requisite pains over this part of his argument.

His second item again features an unstressed vowel before a liquid consonant, this time r. It is dative Ligore Loire'in a single instance (ed. Bately 1828). To it is obviously to be compared Ligor 'Liguria', likewise in a single instance (ed. Bately 10928). Dr. Breeze does not make this comparison, presumably because Bately (1980: 421) notes that several of the Latin manuscripts most closely related textually to the Old English Orosius have Ligor- for 'Liguria', but 'Loire' is consistently Liger- in the "related" Latin textual tradition. Both Bately (1980: 204) and Breeze (1992a) discuss this item as if it should reflect changes in pronunciation of particular names; but it involves no more than variation in spelling of reduced vowels in unstressed syllables as before. Positive evidence against the view that that variation reflects sound-change is furnished by the Anglo-Saxon Chronicle's forms for 'Leicester', 917 Ligeraceaster, 942 Ligoraceaster, of which the later is etymologically "correct" for that name. I f as is likely he stressed the initial syllable in the Old English way, the Orosius-translator uses -or- for other -Vr- comparably in Falores for Phalaris (ed. Bately [33.sub.35]); the divergent contrast [44.sub.18] Damaris beside [44.sub.34] Dameris 'Thamyris' is already present in related Latin manuscripts according to Bately (1980: 415). Instability of unstressed -Vr- followed by a vowel is seen also in [31.sub.31] Omarus for Homerus 'Homer' (see n. 9 below). When a consonant followed he normalized a different way: Carthaginians called Hasdrubal, for which Bately (1980: 419) notes a common Latin manuscript variant was Hastrubal, are consistently Hasterbal in more than a dozen instances. Probably to be compared is [110.sub.8] Hungerre 'Hungarians', gens Hungarorum in the chronicle of Regino of Prum sub anno 889 quoted by Bately (1980: xc). Possibly to be compared, with -VIC-, is Hime(o)lco 'Himilco' (two instances), for which however Bately (1980: 419) notes that Himelco is a common Latin manuscript variant .

Dr. Breeze's third item is slightly more interesting. It is the name of the Danube, in seven instances consistently Donua (indeclinable: see n. 21 below). Breeze (1992b: 431) notes that the QE [Don-ea.sup.*] postulated by Forster (1924: 2) would not account for the u in the Orosius form; Forster (1924: 4) explained that as a blend by the translator with Danubius in the Latin original. The name is derived from early Celtic [Danouia.sup.*], the Welsh reflex of which is Donwy, exhibiting the regular development of the suffix -[ouios/-ouia.sup.*] in Welsh. That obviously would not do as a source of Donua, nor would regular Cornish -ow, Old Breton -oe, but Breeze (1992b: 432) finds in a single Old Breton gloss with -uiu for the masculine version an excuse "to propose a form [Donuia.sup.*] from the British feminine [Danouia.sup.*], dictated to an Anglo-Saxon scribe, who reproduced in Anglo-Saxon orthography what he heard"; and "records of Old Cornish are too few for certainty" that this rare alternative did not exi st in Cornish as well. Readers may agree with me that this is an impossibly tenuous chain of speculations. It is not even adequate on its own terms, since Donua does not represent any reasonable pronunciation of "a form Donuia" in Anglo-Saxon orthography. The -i- would surely be regarded as significant by an Old English hearer, and represented in spelling, in a river-name most likely with -e-, since a river-name of the form postulated would be likely to strike such a hearer as a compound of ea 'river.

The Cornish theory is also not necessary, since OE Donua can be quite adequately accounted for within Germanic, as it is by Bately (1980: cxiv, 416), quoted by Breeze. Breeze's following statement that "Forster's account...also notes difficulties in taking it as a purely Germanic form" (1992b: 431) misrepresents the gist of both authors. If we are to suppose, as Forster, Bately, and Breeze all assumed, that Donua was exclusively a name for the Danube, then the hypothesis indicated would be to take it as Bately (1980: cxiv) does, as one of a group of "names of people and places that the author of Or. could have known in their contemporary form", in this case as spoken by people like Alfred's helper Grimbold the Old Saxon, as a loan from the Old Low German antecedent of MLG Donowe, Dunowe (Forster 1924: 2, 1941: 141 n. 1), to which correspond Old High German Duonowa, Tuonouwe (Forster 1924: 2, Bately 1980: 416). The phonetic pattern of a long syllable followed by -owV is not Old English, and some simplification would be likely (cf. mutatis mutandis Campbell [section][section] 392, 345-6, 351-2, 405, 468, 470). As a purely Old English naturalization [Donwa.sup.*] might be expected, by analogy with inherited wa-stems (Campbell [section] 594), but the continued availability of the continental pronunciation as a model would favour selection of the syllabic variant, thus the attested Donua.

That is a workable hypothesis, and so much more economical than Dr. Breeze's of a Cornish loan that the latter must be forthwith dismissed; but I doubt if it is the whole story. The stability of the Orosius's form suggests that it was not in fact just naturalization of a recent loan but had a history within Old English, as is intrinsically likely for the name of such a large river from the Germanic homeland. (7) The regular reflex of Celtic [Danouia.sup.*] borrowed early into either Germanic or Old English would in Old English as Forster (1924: 3; cf. 1941: 606-8 n. 7) points Out be [Doneg.sup.*]. Since in Old English eg means "island" and ea means "river", substitution of the latter in such a compound would be rather likely. Forster gives instances of such substitutions both in Old English and in continental Germanic. That brings us to Forster's postulated [Donea;.sup.*] whence the -u- in the attested form? The answer lies, I think, in two things not focussed on by Forster at this stage in his argument nor b y Dr. Breeze at all, the freedom with which Old English can use in parallel simplex river-names and compounds in ea, and the existence in England (including what is now southern Scotland) of a largish number of rivers Don, occasionally Doon, from the same Celtic stem Danu- (cf. Forster 1941: 145-148, Ekwall 1928: 126-128). The former is conveniently illustrated from the name of another river with a long first syllable ending in n. Gloucs S896 goes innan Cyrne; andlang Cyrne and eft on Cyrne; up andlang Cyrne 'into the Churn; along the Churn...again into the Churn; up along the Churn', whereas S1556 goes innan cyrnea. 7 lang ea and S202 on cyrnea a ongean stream 'into the Churn River; along the river' and 'along the Churn River continuously against the stream'. The relevance of the rivers Don is that as Forster himself later points out (1924: 19-20), the name would have been borrowed from Brittonic as [Donu.sup.*] and remained as such as long as OE -u survived after long syllables (cf. Campbell [section][secti on] 345-346); flodu as an epigraphic archaism on the early eighth-century Franks Casket may well be the latest example. Any compound of ea formed up to that time would contain a u; and in a compound [Donuea.sup.*] with a triphthongal sequence it would naturally be the middle vowel that with the passage of time was simplified out of existence. (Lapse of time is a crucial difference between this and Dr. Breeze's dictation theory.) Stress and length are regularly lost early even in ordinary lexical items in monosyllabic second elements of at all reduced meaning (Campbell [section][section] 356, 88). Since in terms of referential meaning [Donuea.sup.*] adds nothing to [Donu.sup.*], and since place-name compounds notoriously are reduced faster than lexical compounds, all the conditions for reduction must be taken to apply here. Ordinary Old English words are not well furnished with triphthongal sequences for exact comparison, but the regular change niui->-ny- (Campbell [sections] 265) is very loosely comparable, a nd it is of course the same kind of reduction which makes contemporary BBC announcers talk about Northern [a:l""nd] for standard English [ai""l""nd] 'Ireland'. I suggest it was because a Donua arising in this way was already in the Orosius-translator's language as a name for one or more English rivers that he used consistently that particular form as an approximation to Low German [Donoue.sup.*]

We have seen that the arguments for a Cornishman as translator of the Orosius, ingenious as they are, all fail. A defect of method in all of them was that their advocate did not look closely at the possibilities within Old English. We are left to seek a dialect position from scratch by the methods of rational dialectology. These involve, obviously, taking the text as a whole, and looking for characteristics in it which map coherently in texts of known local origin, roughly as has been done by McIntosh (1986 etc.) and Samuels (1963 etc.) for Middle English, by Dees (1980, 1985, 1987) for Middle French, and for some other languages by other contributors to Fisiak (1995). With Old English there is the additional complication that most of the anchor texts, which are charter boundaries, are not extant in contemporary manuscripts, so one has to keep a weather eye open for possible contamination by cartulary copyists; but that is not usually a problem, because cartularies up to the mid-thirteenth century on the whol e copy tenth- and eleventh-century texts more accurately than do tenth-and eleventh-century literary manuscripts. With the Orosius there is the additional complication that we do not know from external sources to what extent it is the work of a single man or of a committee. With all early West Saxon there is the problem of telling when 'Anglian' or 'Mercian' items that appear do so as part of a genuine dialect mixture and when as a result of the Mercianizing scribal tradition that affects most ninth-century charters, even grants by West Saxon kings of land well south in Wessex, such as the famous one by AEthelwulf in 847 of land om Homme or as a good West Saxon should have written ymb Hamme "around Ham" the South Hams in Devon, S298. But having taken cognizance that these questions exist we can meet them as and if they arise. Comparable questions arise after all for many texts of most periods. The findings of McIntosh and his collaborators for Middle English, where the materials for testing are much more copi ous and have been much more fully investigated, that most texts approximate much more closely either to completely accurate copies or to complete dialectal translations than to half-way houses with several different significant dialectal components (McIntosh 1986 I 32-33, Benskin--Laing 1981: 79-84 etc.), may reasonably be anticipated to apply also to Old English, and do apply to the only anonymous Old English text for which their applicability has yet been seriously tested, the Life of Machutus (Kitson 1993: 35-40). A control for two of the variables will be provided, some of the time at least, by the Parker Chronicle, which not only is agreed to be a product of the same scriptorium, supposedly Winchester, its second scribc, who depending on whose palacographic eye you believe wrote the annals 892-912 (Batcly 1980: xxxix) or nearly the whole of 891-924 (Sprockel 1965: was the same who wrote the main ("Lauderdale") manuscript of the Orosius (8) (Bately 1980: xxiii-xxiv). Linguistic features in which the Orosi us as a whole disagrees with the Parker Chronicle must derive from its own textual tradition not the scriptorium which produced the extant manuscript, and are likely to be evidence for the dialect of a single author or redactor of the work as a whole. Further control may be furnished for some items by Middle English, since though little work has been done on the extent to which relevant isoglosses changed over time, in at least one text for which detailed investigation has been made Old and Middle English dialect criteria point quite strongly in the same direction (Kitson 1992a: 30-34).

Several items in the Orosius have prima facie significance for dialect as traditionally studied by the grammarians, though it has not been admitted in standard accounts of this particular text. They have noted that the Orosius differs from other early West Saxon in that it has an appreciable amount of "late West Saxon smoothing" ea > e (Campbell [sections] 312). They have abstained from noting hat it is likely to be dialectal, indicating a northerly position within Wessex, The main environment of the smoothing, before palatal and velar consonants, is identical with the main environment of the earlier "Anglian smoothing", though there are differences in that the West Saxon change occurs also after palatals and the Anglian one affected more diphthongs, and the phonetic outcome of the Anglian change could be ae as well as e (Campbell [sections] 222). Still the similarity is great enough for it to be intrinsically likelier, from a diachronic dialectological perspective, that the West Saxon change represents a spr ead, with modification, of the already existing Anglian dialect feature than a totally independent new development. The proposal of Hogg (1992: 170) to rename the West Saxon change in a way that obscures its similarity to the Anglian one seems unwise. The fact, which troubles the grammarians, that West Saxon smoothing is only sporadic in the texts in which it occurs, tends to support the view of it as the spread of a feature, as ceritainly does the fact that the main "late West Saxon" text involved is the Abingdon version (MSS B and C) of the Anglo-Saxon chronicle.

So, above all, does the fact that so-called "late West Saxon" smoothing yields the majority form for some words, such as peh for peah 'though, in the "early West Saxon" text of the Orosius, yet it does not in the "late West Saxon" texts exhibiting it. (Hogg [sections][sections] 5.120-1 gives some exact figures.) To take these phenomena as "northern West Saxon" is a significantly more economical account of them than the conventional chronological one, unless contradicted by more definite data on other items. I do not see any. On the contrary, very many of the divergences of the Orosius from other "early West Saxon" texts observed by the grammarians and assembled by Bately (1980: xl-xlix) point the same way. I shall mention here only two of the most striking. The strong preference for -ad- over -od- as formative suffix for second class weak verbs is as Bately (1980: xlvii) acknowledges Anglian' (Campbell [sections] 757). (9) And charter boundaries show retention of rounding in such words as aepel, soelest (10) (standard West Saxon epel 'homeland', selest 'best'), even if "schwerlich acht ws." as Bately (1980:xliii) quotes from Cosijn (1883-1886 I [sections] 65), to be present in that part of geographical Wessex where there is most admixture of 'Anglian', west Gloucestershire (both banks of the Severn) and extreme north Somerset (Kitson 1995 map 12). The charter evidence on that point is not fully comparable, on the one hand because the word efes, efisc 'eaves' which provides it has a short vowel, from which rounding would presumably be lost more easily, on the other because the following labial would certainly make retention more likely; (11) but it is certainly a significant pointer. (Frequency of verbal forms can obviously not be tested from charter boundaries, and regrettably none of the frequent words offers environments for direct testing of the distribution of smoothing.)

A more important respect of phonology in which charter boundaries bear significantly on a difference between the Orosius and other "early West Saxon" texts is its showing appreciably more of the "late West Saxon" sound-change iey (Campbell [sections][sections] 300-1) A westerly origin of the change icy within Wessex is indicated by forms mainly of the word 'well', meaning "spring", in charters. The only tenth-century bounds with wielle are from Hampshire in the first decade of that century, 909 S378(i)(iii). In charters from the 920s on wile is nearly as common as wylle in Hampshire, including the earliest texts there; from Wiltshire west tenth-century texts have consistently wylle (cf. list of forms in my Guide [sections] 6.20.1.2(ii)). Positive evidence for the origin of the sound-change points specifically to north-west Wessex; however, the texts involved are (or may be deemed to be) problematic. Gloucs 854 S1862(i)(ii) have one boundary feature spelt uuielle but four spelt wylle. The manuscript is tenth-c entury not ninth-century but preserves archaisms in other items (including notably b for medial [v] in gabul, beber for gafol 'tribute', befer 'beaver'), and there seems no good reason to think it has modernized four out of five instances of 'well'. But my localization of it in the Badminton area of east Gloucestershire is based on vocabulary and phonology and a single river-name, not on a full solution of the bounds, and some readers may consider it suspect for that reason. Somerset S237(i) has two features spelt uuylle. Consonantal u(u)-is an archaism, and the survey's sub-quarterly form and slightly flowery Latin language fit a date not later than the first half of the ninth century, and possibly earlier. But it is anyway a forgery for its purported date 682, and a case could be made for considering it an eleventh-century reworking of earlier materials. Worcs 849 S1272 has accusatives woellan and wyllan, with respectively Mercian and West Saxon vowels, for a single feature. This kind of dialect mixture is actually typical of tenth- and eleventh-century south-west midland charter boundaries (Kitson 1992a: 35 n. 34), but so routinely have Anglo-Saxonists explained it away as due to influence of "standard West Saxon" scribal habits that some will refuse to believe it. However, the eleventh-century Worcester cartulary which preserves S1272 certainly is not systematically biassed toward West Saxon, and the weak ending is a west midland very un-West Saxon form (but again typical of 'well' with the West Saxon vowel in the south-west midlands). If either vowel is scribal not authorial it is more likely to be the Mercian one, and its cause ninth-century Mercianizing scribal habits, such as produce four features woelle in the South Hams charter already mentioned, beside one oewielme which keeps its genuine Devon vowel because oewielm 'river-spring' was a West Saxon word not possessing a Mercian form (Guide [section] 6.20). S298(i) shows that in the mid-ninth century iey was a sound-change proper only to north-west not to south-west Wessex. S1272 contains one other innovative y, in Byrnhelmes 'Beornhelm's', which maps nicely as part of a genuine sound-change (Guide [section] 8.3.9.8). I am disposed to take it as confirming what I think is the evidence of S1862(i)(ii), that for the word 'well' the change iey was largely complete at the Severn Valley end of Wessex in the mid-ninth century. Of course the context between w and l is about as favourable to retraction and rounding as one can get (cf. modern sub-standard English pronunciation (12) of will as a homophone of wool), so this is not evidence for the exact speed of the change in other contexts; but it should mean that the Orosius belongs dialectally significantly nearer the Severn Valley end of Wessex than the Parker Chronicle and Pasoral Care do; conversely free interchange between i and ie (Campbell [section] 300) in the Pastoral Care as represented by the Hatton MS is one of the reasons for associating it with Hampshire or further east, others being the much greater proportion of io to eo than in the Orosius or Parker Chronicle (Campbell [section] 1296; cf. Bately 1980: xliii) and probably the rare betweoxn 'between' (Kitson 1993: 25, 43 n. 116).

An awkwardness in this line of argument is that the word 'well' itself in its only two occurrences in the Orosius is spelt wielle and wille (ed. Bately [98.sub.28] and [131.sub.1]). But I think the agreement between it and the other evidence discussed in this article confirms that there is a significant correspondence between charter wylle and the Orosius's yie in other words, and those two spellings must be explained away enough to conform to it, wielle probably as literary conservatism (cf. on -um and beorg below), wille possibly as scribal (cf. eas nonce-genitive of ea below), though two N. Gloucs charter boundaries, S550(ii) and S1551, use that spelling repeatedly (the latter round Deerhurst, a few miles north of Gloucester). Both Orosius instances are nominative, so they provide no firm information on gender. A reasonable guess is that the translator's declension for the word was like his vowel(s) consciously West Saxon, i.e. strong masculine, even though writing anywhere in Gloucestershire except the far south or the far north-east he would be in an area where the vernacular was weak feminine (Kitson 1990: 208 map 8), even in combination most of the time with the West Saxon vowel.

Contrast in some details with the surrounding vernacular is intrinsically likely for a writer trained in a definitely West Saxon or Mercian tradition in Warwickshire, Worcestershire, and Gloucestershire, the region of the Hwicce, who are accepted since Stenton (1927: xiv-xxii; cf. 1971: 44) as having been of mixed Anglian and Saxon origin. Such contrast may be present in the Orosius's consistent pwyres for older West Saxon pweores 'obliquely'. This maps with beautiful neatness as a feature of 'Thames Valley Saxon', excluding the south-east (Kitson 1993: 18 map 6); excluding also Wares and Gloucs, where in extant charters Mercian pweres is consistently used. But the only positive datum for Gloucestershire is from the Mercianizing period, so it is not evidence that a West Saxon writer there would not have used the West Saxon form. The charter evidence, including a truncated pwe from north Somerset, may well however mean that he would have had to make a conscious choice between the two. (13)

Turning to matters of vocabulary, the certainty that "early West Saxon" texts used "Anglian" words because of Mercian literary influence means that argument had better be based on words for which charter boundaries show geographically coherent variation within West Saxon, and/or show definitely that an item is non-Mercian or non-West Saxon, and/or topographic vocabulary where importation of alien dialectal forms is excluded by contradictions of meaning. Most of the items treated by Bately (1978), and most of those treated as dialectal by the school of Gneuss, are likely to be stylistic not dialectal. For one clear exception, the words for 'island' (Bately 1978: 104, 117-119), names recorded in charter boundaries belong mostly to a stage of the language too much earlier than the literary texts for their testimony to be very useful anyway. It is plain from literary texts that ealond was Mercian, igland West Saxon, and that they must have replaced in normal use Anglian eg and "Thames Valley Saxon" ige common in charter names (the latter itself being a reshaping of the form ieg given as normal "West Saxon" by the grammars). It is not clear whether the presence of names in eg in south Somer- set and in the south-east, including Hampshire, should imply that the Chronicle's preferred ealand reflects the vernacular of one of those counties, or whether it is just a West Saxonizing of the Mercian word and igland and igod were the only words for 'island' current in any part of Wessex (and whether, if so, the sole or main difference between that pair was one of meaning, igod smaller and/or exclusively riverine, or dialectal, igland proper mainly to central Wessex, igod to the south-east and south-west). Very frequent igland in the Orosius is no close evidence for localization, since it could occur as far south as Dorset in charters and as far north as north-east Gloucestershire at least in localizable literary dialects (it is frequent in the Life of Machutus).

The hypothesis of a north-westerly, Severnward position of the Orosius within Wessex is supported by its frequencies and forms of words for 'between'. Old English has three main words for the concept, in descending order of frequency in literary texts as assembled in the Toronto Microfiche Concordance (detailed Kitson 1993: 11) betweox, betweonum (usually with reduced -an), and betweoh, all with several phonetic variants. Charters show betweox the exclusive word in Wessex from north Devon through mid-Somerset and all but the extreme south of Wilts to NW Hants and W. Berks; betweonan the exclusive word in Worcestershire and the west/north midlands north from there (Kitson 1993: 13 map 4). A scatter of both are also found in other areas, but those are the ones that mainly concern us. Betweoh predominates in charters of Middx--E./S. Surrey--mid-Hants--S. Wilts; it is also attested in N. Gloucs. In a west midland area from mid-Somerset, Wilts and S. Oxon north and from W. Nhants west the -eo- tends to go to -u- i n betweox in late texts, (14) elsewhere in all these words to -i-, with patches of -y- in the south-west midlands and south-east (Kitson 1993: 15 map 5). The samples in both the areas of apparently exclusive usage are small enough to leave open the possibility that the words charters show in them were not exclusive, merely predominant, likewise in this labial context -eo- > -u- as a sporadic change must be reckoned a possibility in any part of the country where OE y > ME u was, and perhaps more widely; (15) but samples in all areas west and south of Watling Street are large enough for it to be unlikely that actually predominant forms do not appear. Obviously the exact positions of my mapped isoglosses are not to be relied on (cf. Kitson 1993: 17), but the datum-points constraining most of them do not allow vast scope for alternatives. The exclusive betweox area explains why that word predominates in literary texts, since N. Wilts and adjacent parts are the heartland of what I call "Thames Valley Saxon", which demonstrably means the heartland of literary "West Saxon", (Kitson 1995 maps 3-5 and discussion), especially Elfric's (Kitson 1993: 24).

The majority word for "between" in the Orosius is betweonum [49 including variants], with about equal minorities of bet(w)uh [19] (16) and bet(w)ux [17], all but one of the 36 instances of the latter pair having the vowel u and over 90% of them dropping -w-. The combination of betweonum as majority word and the extremeness of the tendency to a would seem to rule out any area of Wessex except the one for which there is no positive charter evidence between the exclusive betweonum and betweox areas, Gloucestershire west of about Cirencester and the extreme north of Somerset. The numerical predominance of betweonum and the fact that the main West Saxon word is the least common of the three would point by dead reckoning to the north end of that range, that is to Gloucester or its immediate vicinity. However, dead reckoning from the bare figures will not do here, because with one exception in the contents-list betweonum is used solely for set expressions 'between them(selves)', 45 Out of 49 the postpositional formu la him betweonum (Bately 1970: 449; cf. Appendix below). The meaningful word for between' in contexts of giving new information is bet(w)ux, and there is strong reason, discussed below, to believe that bet(w)uh was either not part of the main translator's active vocabulary at all or a very much less significant part of it than appears from the raw figures. So what is basically present here is a phenomenon much discussed by Samuels (1972: 97-103, 111-125), a dialectal subsystem on the border of two main dialects using elements of both. As with the phonology, it is West Saxon with a strong Mercian colouring, not vice versa. By adjusted dead reckoning this points to south Gloucestershire rather than north. The substantial Anglian component is not a problem however far south: in many items of vocabulary most or all of Gloucestershire, even including the northern fringe of Somerset, contrasts with most or all of Wiltshire (Kitson 1995 maps 4, 8, 11, 14, 21, 1993 map 11, etc.). It seems the Cotswolds were a physica l feature having more effect on Old English dialects than most. Charters show too quite a clutch of 'Anglian' phonetic forms in the far south-west of Gloucestershire, e.g. in more than one boundary of Stoke Bishop just west of Bristol, so this evidence would consort fairly comfortably with anywhere in the county west of a line from about Cheltenham to Bristol.

The extremeness of the tendency to a, much greater than one would have anticipated from the charters, could be partially explained if it were the result of a common articulatory tendency with other changes of rounded front vowels to u. It is then satisfying that Gloucestershire is where in Middle English the west midland-centred u-region for OE "stable" y (Jordan 1906 [section][section] 39, 42) and u-spellings for OE "unstable" y, which tend to be Thames Valley-centred (e.g for cyrice McIntosh 1973: 56-57, 1986 IV 249-255) principally overlap. It is, too, intriguing that occasional -u- forms for 'between' words in the Parker Chronicle occur only in the stint of the Lauderdale Orosius scribe (Sprockel 1965: 52), whom there is more reason to think native to the same general area as the Orosius-translator than to the supposed area of the two works' common scriptorium at Winchester.

The sequence twu- is so rare that in seeking parallels for the loss of w in it we may more profitably use the broader definition of loss of w before a back vowel in the onset of a stressed syllable. It is then again in the south-west midlands that most parallels arc found, in all three periods of English. Inorganic gain and loss of initial w in modern English is substantially commoner there than elsewhere (Wright 1905 [section] 236). McIntosh (1986) maps 1183-4 show the same in Middle English for [+ or -]with -u- preferred to -o- following. And in charter boundaries the word 'weald' or 'wold' lacks initial w- in NW Herefs S677, NW Gloucs S1551, and S. Wores S1322(ii), all its west midland occurrences (Guide [section] 6.7.2.1). It fits too that the Orosius has the "nur westsachsisch" combinative u-mutation before velars in cucu for cwicu 'alive' and wucu for wicu 'week'. (17) Austere Neogrammarians may reject all the facts mentioned in this paragraph as not evidence. Practical dialectologists will probably acc ept them as circumstantial evidence; and they obey the great rule of circumstantial evidence as the best detective-stories tell us (Freeman 1939: 219), they all point to the same conclusion. Reduction of -w- in the sequence -twu- occurs elsewhere in these words, (18) but at much lower frequency, e.g. 10% of the 39 -u- spellings in the works of Alfred as transmitted (Kitson 1993: 25).

A further point which needs explaining is why the Orosius with only one exception has unreduced -urn in betweonum, beside an appreciable scatter of reduced -un, -on, and -an in nouns. Since prepositions have lower sentence-stress, the disparity ought to be the other way round. There are two possible lines of explanation. One is that the Orosius-translator's dialect had been so conservative with this word that only recently had it become consolidated there as a single word, as opposed to two be...tweonum with the noun governed between them, which is its etymological origin but which in extant Old English is almost confined (10 out of 13 instances) to the archaizing language of poetry. If that is so, we should wish for some reason why his milieu might be especially isolated dialectally, e.g. being on the west side of the Severn. The other is literary conservatism, that he had been taught a correct spelling of betweonum, whereas -um in dative plurals in nouns was too fully current to need teaching. This might ac count too for the spelling of wielle discussed above.

The -n spellings for dative plurals in nouns are not actually very numerous, (19) but they are more so than in other "early West Saxon" literary texts to an extent that invites comparison from charters. The question we should ask is where -m was most reduced to -n in DPs earlier than -u- was reduced to schwa; which resolves into the question, what is the proportion of spellings -um to -un and -on. It must be emphasized that the following figures are not to be relied on at all closely, because most dative plurals in charters are reduced to -an (some even further); the samples with unreduced vowels for individual counties are mostly too small to be statistically reliable, they are not strictly comparable because their chronological distributions are different anyway. Yet they furnish the possibility of some guide, which is better than arguing in vacuo. The proportions of -un and -on spellings as percentages of -um, -un and -on altogether (20) in those counties which have any of the three, followed in square bra ckets by the sample-size per county, are as follows: Staffs 100 [1], Warks 40 [10], Gloucs 44 [18], Som 63 [8], Corn 100 [1], Devon 43 [7], Wilts 33 [40], Hants 47 [32], IoW 0 [2], Sussex 20 [5], Kent 0 [8], Surrey 33 [3], Middx-Essex 100 [3], Herts-Suffolk-Beds 0 [4], Bucks 50 [2], Berks 34 [32], Oxon 100 [17], Hunts 0 [2], Nhants 86 [7]. It looks from these figures as if a "standard West Saxon" text with a really high score for -un and -on might be expected to come from Oxfordshire or Northamptonshire. As within geographical Wessex and the mainly unreduced inflections of "early West Saxon", if any trust at all is to be put in these figures they put the Orosius as before closer to Somerset and Gloucestershire than the other texts.

Topographic vocabulary whose interpretation is clear tells the same story. Swelgend, meaning perhaps 'whirlpool' and/or perhaps 'swallow-hole', is on the charter evidence feminine in south Wessex and the south-east, neuter in mid- and west Wessex, and masculine in north Wessex from NE Somerset north (Kitson 1990: 211-212). The samples for genders other than feminine are too small to be relied on as representing the original populations coherently, but if they are coherent, that is the way they point. Swelgend in literary Old English means 'glutton', in which sense it might be expected to be masculine anyway. But the single instance in the Orosius, which is masculine, though applied to a person translates gurges miseriarum, Alexander the Great as a 'whirlpool of miseries' for the Orient (Bately 1980: 257). Alexander was not known as a glutton, so if the translator was operating competently here his masculine gender is a significant link with the north-westernmost charter instance. And rather nicely, charters s how this topographic sense to be West Saxon not Anglian.

Conversely, the valley in which were Sodom and Gomorrah is called a doel 'dale'. (This seems to be the only 'valley' mentioned in the Orosius.) Charter boundaries show doel to be a strongly Anglian item, contrasting as a word with southeastern dell and 'Thames Valley Saxon' crundel (Kitson 1990 map 1, 1994 map 4). As a general word for a large valley, which it seems to be in context (ed. Bately [23.sub.7]), choice of doel in preference to common Old English denu is even more remarkable, since dell/doel/crundel are words specifically for steep-sided valleys, including ones made by prehistoric quarrying, and doel for large dales in place-names is quite strongly northern (Gelling 1984: 94-96). The south-westernmost doel in charters, the only one in Wessex, is in south-west Gloucestershire. Granted that the Orosius is West Saxon, this item seems to have very much the same dialectal implications as the 'between' words.

There are three more striking items of topographic vocabulary, one inflectional, two semantic. The words concerned are ea 'river', beorg 'mountain', and clif 'cliff'. Clif denoting large coastal features such as the white cliffs of Dover or the equivalent brimclifu seen gleaming in Beowulf is common Old English. Charter boundaries use it also for much smaller inland features, some but not all with water at the bottom. That usage is specifically 'Thames Valley Saxon' (Kilson 1995 map 3), absent in charters of Gloucestershire except the extreme north-west and S553 in the far south. What is peculiar about clif in the Orosius is that it seems to be applicable to any sea-shore. The opening geographical description twice refers (ed. Bately [11.sub.4,13]) to the clif of the Red Sea, translating litus 'shore'. At [112.sub.10] it is translated by Bately "cliff", and an alleged height of two miles does suggest that, but phrasing on doem soes clife with definite article implies again that a clif is something you know a sea has even when you do not know the context. [119.sub.12], lava from Etna burning up ealle da clifu pe neah poem soe woeron, likewise works better the more widely inclusive, 'all the coasts that were near the sea'. (It is presumably not the ground that was consumed but the living things and artefacts on it.) Conversely the word for the precipitous, 'cliffy' coast of Norway is cludig [15.sub.23]; a town in India is set about with precipices cludum [72.sub.7] (not 'rocky' and 'rocks' as Bately translates); and in earthquakes cludas fall off mountains [135.sub.6]. Clud is much rarer in charters than one would have guessed from literary texts; there is only one, an inland stanclud whose referent is not identified in NE Somerset S508(iii). Stoed, the normal Old English word for 'shore', denotes in the Orosius the shore of of a lake at [16.sub.34] and a river-bank at [146.sub.17]. It is used for a sea-shore once only, at [26.sub.24], and that is in a section that will be shown below to be not all the work of the main translator. If I am right to think that clif in his usage means sea-shores rather generally, it might be a spread of the 'Thames Valley Saxon' usage to coastal features and/or a sign of formative years spent where a whole coast was cliffy (as it might be between Portishead and Clevedon in the extreme north-west of Somerset, or Lydney on the west bank of the Severn in south Gloucestershire). Possible snags are that nowhere are low seashores called clif in place-names (Dr. Margaret Gelling, pers. comm.); and though the translator is not likely to have had access to information (or pseudo-information) about cliffs along the Red Sea, it is possible that he was making an intelligent deduction from context, which has the Nile rising near the Red Sea but flowing away from it. He may have reasoned. that for such a long river that would be impossible unless the initial shore were pretty high. So it is uncertain whether clif has implications to our purpose; but if it does, it points to the area of Bristol rather t han Gloucester.

Beorg is the inherited Germanic word for 'mountain', but it was replaced as the most general word for natural hills very early in the settlement of southern Britain by the loan-word from Celtic dun. Thus Somerset S311OE on midneweardne del poere dune pe man Hoetbeorg nemd 'to the middle part of the hill that is called Heathberrow'. Beorg came to be used more of prehistoric tumuli, whence modern English 'barrow'. In the Orosius beorg not dun is the main word for 'mountain', though dun is also used. (There is even in [10.sub.5] plural beorhtte a derivative adjective *beorgiht (Bately 1980: 349), formed with what charter boundaries incidentally show to have been a distinctly 'Anglian' suffix.) The Orosius is the only substantial prose text in which beorg is the main word for mountain. Now there is nowhere in Wessex where one would confidently say from charter boundaries that that usage still existed. Soberer scholars like Grinsell (1953 etc.) and myself admit that beorg in Wessex appreciably often names natural hills; the argument has been with those like the late G.B. Grundy and Dr. Della Hooke, who assert that it practically never does. But no-one disputes that that meaning is commoner in the midlands than Wessex, or that barrows constitute the majority of features in Wessex called beorg. The argument is bedevilled rather by a substantial number of instances where there is an artificial barrow on the summit of a natural hill, or where the point that would make best sense of the boundary circuit is on a natural hill but there is a barrow a furlong or two away, or where there is evidence for the former existence of lost barrows, or when lost barrows are postulated even where there is not evidence. There is certainly no area where phrases "the beorh" with a definite article, not recapitulating a proper name, refer oftener to natural hills than to barrows. But it is clear that beorg denoting natural hills was commoner in the far west, from mid-Gloucs through Somerset and Devon to the eastern fringe of Cornwall, than e lsewhere in Wessex, and any part where it was still current as a general word must lie in that region.

I should say there were two candidates. In much of Devon, with E. Cornwall and NW Somerset, natural hills may well be the commoner referents (though no doubt Dr. Hooke would dispute this). But given the evidence above, that is hardly of interest for the Orosius; and in Devon dun is also used. The other candidate area would be Gloucestershire from about Gloucester south. The western two-thirds of this contains only 3[2] features, but they are both natural hills. Of the features on the eastern fringe, some are natural, some artificial, and several disputable. The sample is, then, too small to erect a theory about this area on the basis of charter beorg. But what may make it significant is the complete absence of dun in it, not paralleled in any other comparably hilly area with a comparable sample of charters. That is an effect which might be produced by beorg being still current for the normal meaning of dun. Gelling (1984: 149) finds something like this area conspicuous for place-names in dun named apparently from much smaller features than usual, which would fit, but not devoid of normal dun-names, which might be the place-name analogue of the Orosius. Altogether, if the conservative use of beorg in the Orosius comes from anywhere in Wessex, charters show southwestern Gloucestershire as distinctly the best bet.

Ea in normal Old English is a feminine with indeclinable singular (hence, in my view, the indeclinability of Donua above). (21) What is special about the Orosius is that it has the dative and genitive singular ie (cf. Campbell [section] 235.3). It is the only literary text which does. (22) Precisely one charter boundary has ie, in two datives, SW Somerset 882 S345 nord to ie; ponne nord upp of poere ie 'north to the river; then north up from the river'. It is doubtless significant that this is one of the rare ninth-century charter boundaries; the much commoner tenth-century ones never have ie, including ones covering the same area as S345. As with beorg, the Orosius has a conservative usage. Austere Neogrammarians might then reject this as evidence altogether, on the grounds that an obsolescent form might be kept with equal probability in patches anywhere. Practical dialectologists would dispute "with equal probability", admitting possible plurality of patches but expecting their number to be rather small. The anticipated model would be something like Samuels' (1972: 101) map of combined use of hit and it in Middle English, with the difference that since ie was only ever a West Saxon, not Anglian, form the total area over which they might occur would be much smaller. And use of ie for genitive involves an innovation anyway, because the fronted form can only derive from a prehistoric locative or dative. (23) Levelling to genitive must be analogical, quite likely originating in pre-Old English as part of a process of differentiation from Germanic near-homonyms meaning 'eye' and 'reptile' discussed by Cubbin (1979: 231-233). Forms corresponding to ie beside ea are reported in Old Frisian e(e) beside a 'river' (von Richthofen 1840: 585; on phonology Heuser 1903 [section][section] 20, 25.V), and by van Helten (1890: 158) also in Middle Low German (specifically Lower Franconian), but not as constituting a coherent grammatical system with the unfronted ones, in fact the opposite. (24)

Systematic use of ie in the Orosius-translator's dialect favours the Bristol end against the Gloucester end of the area already deduced as its probable home. Not much weight is to be placed on the particular geographical position of S345; we can with reasonable probability conjecturally reconstruct the larger area in which ie had been current. It is likely to be reflected in modern south-western river-names Yeo. That form with its long o descends not from the dative/genitive but from a stress-shifted fronted version of the nominative/accusative OE ea > ia > ia > ME io. By the charter evidence comparable frontings of (-)ea(-) could occur randomly almost anywhere (Guide [section] 8.3.8.1). But only occasionally; while commoner in Devon than elsewhere, in no other word were they nearly as frequent as the names Yeo suggest for ea. Analogy with the vowel of the dative/genitive might well be what motivated the fronting process in that word. Well, the northernmost Yeo I am aware of is the Land Yeo debouching into th e sea at Clevedon in NW Somerset a few miles due west of Bristol. (25) It is also the northernmost registered by Ekwall (1928: 480-481), whose more than a dozen streams and rivers Yeo are all in Somerset and Devon, and who following Wright (1898-1905) calls it "simply the dialectal form of OE ea in these counties". We shall not be far wrong to assume that that was the area within which dative/genitive ie was ever current.

This may be a reason for thinking in terms of the hilly part of Somerset, from the Axe and the Mendips north, rather than even the south of Gloucestershire. However, it is intrinsically likely that there would have been some currency of ie outside the area preserving fronted forms of ea in place-names. If absence of dun in charter boundaries is a criterion for possibility of the Orosius's usage of beorg, it halves the possible area of Somerset to that north of the Yeo. An item which should restrict it still further is wyrttruman meaning 'roots' (ed. Bately [26.sub.2]). That seems unlikely in the area where the identical word wyrttruma meant 'wood-bank', as it did in most of north Somerset, and the extreme south-west and south-east of Gloucestershire (Kitson 1995 map 12). But the instance in the Orosius comes from a section where there is appreciable dilution of the author's vocabulary (on which more below), and we cannot be certain that a word occurring only there just once is his. Incidence of "Anglian" item s in charters falls off steeply within a few miles of the north border of Somerset, so the balance of evidence still probably favours south Gloucestershire rather than north Somerset, but either way it seems to be pointing to somewhere remarkably close to Bristol.

This seems the moment to introduce another piece of charter evidence hard to evaluate because isolated. The Orosius has a compound prepositional usage wid...weard + genitive 'toward, in the direction of'. The only other author with repeated wia...weard at all is AElfric (Mitchell [section] 1217), in whose usage it governs accusative (Kitson 1993: 5-7). Just one charter boundary has wid...weard, governing genitive, Gloucs S1346, one of the charters already mentioned of Stoke Bishop just west of Bristol. To judge by Bately's glossary, wid in a spatial sense in the Orosius governs more often genitive than it does accusative or dative, usages proper in charters respectively to Wessex, the west midlands, and counties north of the Thames, one of the latter two down to Somerset. That again is hard to evaluate because most use of wid in the Orosius is not spatial and because of very large gaps all round S1346 on the relevant charter map (Kitson 1993: 6 map 1). Still, presence of all three cases for spatial wid is int eresting, and if there is one county for which the charter map should predict it it is Gloucestershire.

Not so predictable is use as case governed by compound prepositions wid nordan etc. 'N/S/E/W of' of dative, in charters mainly proper to south Wessex though also a patch of Worcs and Warks, more than accusative, normal elsewhere. If it could be upheld as part of the dialect of the Orosius-translator, a likely corollary would be that the ambiguous cases in Somerset for simplex wid (Kitson 1993: 6 map 1) are dative not accusative. But the samples are so tiny, compared to be nordan etc. in the same meaning, and the distributions are so suspicious as to suggest that all derive either from source-scribes or from subsequent scribes not the author. This is another point returned to below.

It seems from all the above that though most indications would fit anywhere in west Gloucestershire between Gloucester and Bristol, and might fit the part of north Somerset immediately south of Bristol, persistence of GD ie strongly favours the southern end of that range. Any circumstantial deductions possible from clif and wid...weard do so too, and that end is neater for pwyres. All of the main divergences between the Orosius and other "early West Saxon" texts seem to be accounted for. So I draw the provisonal conclusion that the language of the Orosius is very largely a unitary dialect; if it belongs to a major town that town is Bristol, and if not from Bristol it is from very close to there. (26) Bristol does not seem to find its way into historical sources until the eleventh century, but it was large and important enough then (Heighway 1987: 146-151) for there to be nothing implausible in its already having been an important centre in the ninth (Heighway 1987: 149), with the possibility of origins much earlier (Heighway 1987: 152) The relative closeness to AElfric in north Wiltshire (Kitson 1993: 22-24) is satisfying from the point of view of their agreement, not complete but not shared at all by other authors, in some rare items like wid...weard. Another deserving mention in an article with so much about rivers is an adjective flede 'in flood', which occurs thrice in the Orosius and nowhere else. All three instances describe rivers, all in the nominative singular, two correctly feminine fledu. The related verb oferfledan likewise occurs only in the Orosius, a related noun fleding only in the second series of AElfric's Catholic Homilies. (The MED has no reflex of flede but reports the simplex verb fleden from the other end of the Hwiccean region in Lazamon's Brut, as well as St. Margaret which it lists as "south-western".) (27)

Concluding that the Orosius has a unitary dialect does not mean that there are not some particular items in the Lauderdale manuscript which we can identify as not belonging to that dialect. An important clue to them is the declension of ea. The dative/genitive ie so characteristic of the translator's dialect does not obtain equally throughout the whole work. It is consistent at the beginning and in the second half, but a large section in between has mainly the standard Old English usage. Exact figures are as follows, rounding to whole pages of Bately's edition. Book I: 1-12, G 6, D 9, all ie; 13-17, D 1 ea, G 1 eas! (but still feminine!); 18-35, G 3 ea, D 1 ea, 1 ee. Book II: 36-52, G 2 ea, D 1 ie. Book III: 53-82, no data. Book IV: 83-112: D 6, all ie. Book V: 113-132, D 1 ie. Book VI: 133-156, G 1 ie. So the first part of book I and book VI have for both these cases only ie; books IV-V probably go with them but contain no genitives of ea. book III neither genitives nor datives; book II has ie dative but ea genitive, the second part of book I ea for both cases but the anomalous genitive eas and dative ee. The cause of disruption of the norm of ie for both cases is probably not the same in the two books that show it. In Book II it is reasonable to suspect the participation (whether as assistant writing a first draft or scribe making a final copy) of someone with a less broad version of the translator's own dialect, or a very closely neighbouring dialect, at his own place (though not implausibly the translator might choose to work with a scribe of as nearly as possible his own dialect even if he were working at a different place). In Book I the disappearance of ie begins precisely with the voyages of Ohthere and Wulfstan (pages 13-17 of Bately's edition), a famously obvious insertion from some other source. It would seem that the reason there is use of material from other writers imperfectly assimilated to his own dialect by the main translator. (This means, alas, that we must discard before raising it the romanti c hypothesis that Ohthrre and Wulfstan in their English dealings were Bristol merchants.) A section from [15.sub.1] to [28.sub.11] of Bately's edition is actually not extant in the Lauderdale manuscript but only in the Cotton one. The correlation is definitely not with that but begins earlier, in the same place as Bately (1970: 439) notes that other changes of usage do, and ends later. (I have therefore not scrupled to follow other writers in counting, for purposes of vocabulary, that section from Cotton in with the Lauderdale manuscript. Whatever its exact dialectal position, or mixture, Cotton is clearly more 'West Saxon'/'south-eastern' and less 'Anglian' than Lauderdale; yet the part only in it contains the single most Anglian' piece of topographic vocabulary, doel. On the other hand some difference is present between the language from the beginning of Ohthere's report to the end of the section only present in Cotton and the final part of book I, as will be shown below.)

The grossly anomalous genitive eas offers a broad hint where Ohthere spoke to his lord king Alfred. For as Mic. Conc. helpfully reveals, the only other text with a genitive eas for ea, which has it repeatedly, is the Parker Chronicle (plus the version of one of its annals in Chronicle manuscripts CD). Since of the annals where eas occurs, those for 896, 918, 919, 922, and 924, one or all (depending on which palaeographer you believe) fall in the stint of the scribe who wrote the Lauderdale Orosius, the inference must be that the masculine-looking genitive for ea was the usage at that time of his scriptorium, usually identified with Winchester; and, less surely but still plausibly, that the Ohthere-Wulfstan material was supplied to the translator from there. A Winchester or other south-eastern source would be a possible origin for the dative ee, otherwise only found in appendages to the grants in two Kentish charters, 863 S332 and [s.ix.sup2] S266; Sprockel (1965: 66, 182) notes once ei for dative of ea in the earliest section of the Parker Chronicle, and eae once and genitive e twice in the stint of the Lauderdale scribe. Alternatively it might be part of the contribution of someone with a less broad version of the translator's own dialect, who might well be the Lauderdale scribe himself; the genitives e might well be that scribe's personal dialect's contribution to the Chronicle.

The fact of collaboration might mean that linguistic forms found only in that part of book I from the voyage of Ohthere to the end should be discounted as evidence for the dialect of the main translator. Of those discussed above that only affects doel, which as already mentioned is pretty certain to be genuine, and wyrttruma. It does, however, distort the samples of some plurally occurring items, e.g. stoed/clif as already mentioned, and the 'between' words. The sole spelling betwuh and eight of the sixteen instances of betuh come in the second part of book 1. The remaining instances are once randomly in book IV ([106.sub.19]), once randomly in book V ([123.sub.12]), and six times in book VI, not randomly but clustered two at [137.sub.7,9], three at [149.sub.1,6,7]. Either the main translator's taste in words was subject to sudden yawing or these are the traces of collaborators, whether from first drafts before or copying after the main composition. The 'between' words actually show that the second part of bo ok I is two sub-parts, the division correlating with the end of the section only present in the Cotton manuscript. For the eight instances of betweonum in book I are all in the final sub-part as so defined (the first at [28.sub.18]); and in the previous sub-part there are two phrases with a different 'between' word, [23.sub.13] betuh him and [27.sub.15] betuh him selfum, for what in the main translator's main usage, as established from the final sub-part on, would be him betweonum. (Of these betuh him is strong evidence of discontinuity of production, betuh him Selfum not, given that [95.sub.19] betux him selfum in book IV and [145.sub.13] betuh him selfum in book VI rival in number the two instances of betweonum him selfum (in books II and VI) and single him selfum betweonum (in book II).)

Another group of items whose distribution obviously correlates with divisions in production of the material are the phrases wip norp/eastan/sudan/westan as opposed to be norpan/eastan/sudan/westan 'N/E/S/W of'. Altogether there are 149 of the latter, 7 of the former, occurring 60+4 in the first half of book 1, 85+3 in the second half, 4 all with be in the rest. Those with wip in the first half of book I are the first of nineteen 'east of', the first of twelve 'west of', and two successive ones out of twenty-one 'north of'. Those in the second half are the first two of eighteen 'east of' and the first of twenty-two 'south of'.

Whether we are to interpret these distributions in terms of raw material presented to the translator with locutions he does not like but only becomes aware of with repetition or scribes automatically correcting the translator's work to their idiolect and only disciplining themselves to stop when repetition brings home to them the inappropiateness of it is a nice question. In this instance I should say the former, but perhaps not in every comparable instance. The vowel in betuh bespeaks Gloucestershire even if the word doesn't. Did a south-eastern scribe alter betux or did the author naturalize a south-eastern source's betweoh? it is curious that 'land boundaries' are Beowulfian landgemircu thrice on the first page, consistently londgemaere thereafter. Mearc 'boundary' is a strongly south-eastern word; such rare derivatives have a south-eastern flavour. Landgemircu might be a blend of source (land)mearc with home dialect londgemaere. Such blends and traces of other hands are exceptionally likely in Book I, bec ause the greater part of it is a collection of facts about the world; it would require less intervention with authorial decisions where to shorten text or to adjust the cultural or ideological viewpoint than the others would; also perhaps the author was still establishing his style, taking time to decide which forms from outside his dialect or idiolect were acceptable, which not. (Him betweonum looks like an item of style which he did not establish straight away.) If I am right about input from the scriptorium supposed to be Winchester about Ohthere and Wulfstan, such input may be present more thoroughly worked over in other sections. There is linguistic variation in what one would assume must be the main translator's work too. The dative plural of ea is thrice reduced ean, once unreduced eaum. The phrase in all four is "between the two rivers". `Between' in the first three is betux; with unreduced eaum, for whatever reason, betwux also unreduced.

There is doubtless much more to say on such minute details, not all of which has been said by older scholars. But it will not change the main picture. I suspect that the 4% of directional phrases pinpointed above as likely to be non-authorial are likely to give a fair index of the amount of non-authorial language in the extant text as a whole, closer anyway than the 12% to 22% of 'between' words stigmatized as suspicious. There is nothing to prevent our concluding, and I do conclude, that the dialect of the Lauderdale Orosius is in its essentials not a scribal chimaera but a genuine Old English dialect. The area within which one could make on the indications above, allowing one or two to be taken not quite at face value and allowing for the deficiency of charter evidence positive or negative on the west side of the lower Severn, some sort of case for the origin of that dialect corresponding roughly in seriousness to the outer thick lines on Kitson (1993) maps 8 and 10 would be roughly that within a line from Clevedon to Congresbury to Midsomer Norton in north Somerset, north thence to Mangotsfield in Gloucestershire and north-east to Stroud, then west to the western boundary of Gloucestershire and down to the Severn. The small area of high probability corresponding to the shaded one on Kitson (1993) map 8 would be maybe a dozen miles north-south and half that east-west on the Gloucs-Somerset border centred on Bristol. I offer it as a reasonable working hypothesis that the Orosius represents to a first approximation the late ninth-century dialect of Bristol. Wherever exactly it originates, from the essential integrity of the dialect as within Old English it follows that any dictation by a Celtic-speaker, Cornish or otherwise, affecting the forms of proper names lies back in the antecedent Latin textual tradition. The author or translator, call him what you will, of the Old English Orosius was not a Cornishman.

APPENDIX: THE SYNTAX OF 'BETWEEN'.

It was said above that 45 out of the 49 occurrences of betweonum in the Orosius are in the postpositional formula him betweonum". The possibility deserves testing that such a usage might have a somehow significantly different distribution, dialectally or otherwise, from betweonum in other contexts. Whether it does or not, one would like to know what the wider context of such a striking authorial preference might be. Mitchell (1985) does not discuss distinctions of usage among 'between' words, nor does the Toronto Old English dictionary, (28) which in addition goes badly astray in dividing the forms between them, (29) so I have gone again to the Toronto Microfiche Concordance. For this purpose I have examined its spellings betweonum and betweonan, 228 items comprising 50% of the total sample for betweonum, betweox and betwux, 610 items comprising 77% of the total sample for betweox, betweoh and betwuh, 227 items comprising 40% of the total sample for betweoh, 85 items of 24 spellings comprising what I think is its total sample for betweon and 36 items of 13 spellings comprising what I think is its total sample for betweoxn. (30) The figures arrived at in the present exercise are as in the following table.

The significant patterning in this material is according to whether 'between' governs a noun or non-personal pronoun (i.e. usually the definite article) on the one hand or a personal pronoun on the other. Instances of 'between' used adverbially are to minimize complication counted with the former. They are not frequent enough to distort comparison significantly, except for betweon where a substantial minority of items in the "governing nouns" column are adverbial (mainly in the spelling bituien, including some glosses turning single Latin words by more than one English word). It is next necessary to discard instances of complex pronominal phrases ("between me and you", "between him and his thanes", "between ourselves", and the like), because there 'between' is nearly always preposed whatever the word. The appropriate phrases for comparing word-order are where the word governed is a single personal pronoun. Texts with postposed betweonum have it sometimes at a distance, separated by (an)other word(s) from the pronoun governed.

What the figures mean is that him betweonum is not either really a formula or specific to the Orosius, but is part of a tendency shared by users of the word betweonum generally to postpose it when it governs personal pronouns; and that is bound up with a tendency only to use that particular word for 'between' when a personal pronoun is what is governed. The reason why him betweonum in the Orosius looks like a formula is that, especially in narrative historical material, the third personal pronoun is very much oftener than first or second personal pronouns governed by words for 'between'. That the Orosius agrees in this with the generality of other texts which use betweonum shows that formulaic usage does not here cut across dialect distribution, and leaves the inferences for dialect drawn above (and 31 in my 1993: 41-45) (31) essentially valid.

The descending order of frequency of postposition with single personal pronouns, betweonum 94%, betweon 57%, betweoh 17%, betweoxn 11%, betweox less than 1%, constitutes an 'Anglian'/'West Saxon' dine (albeit these words are not mentioned in such standard works on the diWalects as Jordan (1906) or Hofstetter (1987)). Half the instances of betweon are in the rare Northumbrian texts (the Lindisfarne and Rushworth gospel-glosses and the Durham Ritual gloss); betweoh is the main alternative to betweonum in the mainly Mercian texts which use the latter. The proportion of postposing for betweonum is still so much higher than for the other words as to suggest that there is involved not only a difference between general tendencies in the dialects but, within 'Anglian', specifically influence of usage of the word betweonum on that of the other words. That betweoxn is closest in proportion to betweox is interesting in view of my suspicion (1993: 43) that its distribution is Middlesex-centred; interesting too is that it is nevertheless next lowest after betweonum in proportion of personal pronouns versus other words governed. Some two-thirds of the occurrences are from texts where betweox is the majority word, so this is definitely a fact about usage of words not authors.

It can be fairly stated from there being but a single counterexample for betweox that postposing was a definitely non-'West Saxon' usage. The one exception has very strongly the air of deliberate rhetorical patterning. It comes from the second series of AElfric's Catholic Homilies: Hit gelamp on sumum doege, de da godes englas comon, 7 on his gesihde stodon, da woes eac swylce se scucca him betwux 'It happened on a certain day that when God's angels came and stood in his sight, then was likewise also the devil among them'. Some significant patterns of distribution within and between texts can also be seen in particular spellings of other words. All 10 instances of betweoxn governing a single pronoun are in the Pastoral Care. The one with betweoxn postposed is the first of them. Does the change represent increased efficiency of dialectal translation? Then again, though the samples (except for bitwion) are too small to make much of it, there is prima facie an interesting contrast between the preference for prep osing in the Rushworth Gospel of Matthew with betwion and Lindisfarne Gospel of John with betuien and postposing in the Rushworth Gospels of John and Luke with bitwion and Old English Bede with betweohn. It is to be remembered that Matthew alone of the Rushworth Gospels is a Mercian not Northumbrian text (Campbell 1959 [section] 11).

Diachronic implications too are latent in the above figures. It seems reasonable to relate the fact that betweonum occurs four or five times less commonly with words other than personal pronouns than betweox and betweoh do to its origin as two words be...tweonum with the word governed in between, which would tend to select for short words. Most pronouns, unlike many nouns and any phrase of noun preceded by definite article, are monosyllables. It is hardly coincidence that all three prose instances of be...tweonum have it governing a monosyllabic personal pronoun. (32) 9 Out of the 10 instances in Beowulf and other poems govern a long monosyllable in the formula be soem tweonum 'between the seas'; in the tenth, Andreas 558, the word governed, werum 'men' (dat.), is metrically equivalent to a long monosyllable.) This preference survives to the very latest Mic. Conc. citation from this group of words, inc tweonan "between you two" from the Holy Rood-Tree whose cast of language as extant is mid- to late twelfth- century. (33)

Texts with betweonan are on average later than those with betweonum; so are those with betwuh than betweoh (and betwux than betweox). Figures for the individual forms show a spread of 'West Saxon' preference for preposing even with single pronouns: all the instances of postposing of betweoh use the earlier form, and all but one of the instances of preposing of betweonum with single personal pronoun use the later form.

It is hard to believe that some turn-of-the-century German scholar has not already discovered at least the main lines of all this; but if so it seems to have disappeared completely from the literature. Professor Bately, who when working on the Orosius perused authors like Wulfing much more thoroughly than I have, tells me she does not know of a general account of this syntactic pattern either; and it clearly was not known to the Dictionary of Old English editors or such scholars as advised them on these words. So I hope it will be of some use to give it here.

TABLE: Usage of words for 'between'

                                Immediately  govering

                        Noun or non-
           Occurences     personal         Personal pronoun
                          pronoun
                                           Any       Single

betweonum      86            14             72         69
betweonan     142            22            129        110
                       (inc. 3 daer)

betweonum     228            36            192        179
                            16%            84%

bitwion        14            0              14         14
betweohn       10            8              2          2
betweon        12            6              6          2
bituen         6             2              4          4
betwion        5             2              3          3
Others         38            28             10         5

               85            46             39         30
                            54%            45%
betweon
betweoh       167           142             25         23
betwuh         60            38             22         17

              227           180             47         40
                            79%            21%
betweoh
betweoxn       12            1              11         10
betweoxan      5             1              4          4
betweoxen      2             1              1          1
Others         17            13             4          4

               36            16             20         19
                            44%            56%
betweoxn
betweox       167           140             27         24
betwux        443           337            106         94

              610           477            133        118
                            78%            22%
betweox




                   Single personal pronouns

           between P               P between

betweonum      1               68 (inc.  distant)
betweonan     10              100 (inc. 9 distant)


betweonum     11                      168
              6%                      94%

bitwion        5                       9
betweohn       0                       2
betweon        1                       1
bituen         2                       2
betwion        3                       0
Others         2                       3

              13                       17
              43%                     57%
betweon
betweoh       16                       7
betwuh        17                       0

              33                       7
             82.5%                   17.5%
betweoh
betweoxn       9                       1
betweoxan      3                       1
betweoxen      1                       0
Others         4                       0

              17                       2
              89%                     11%
betweoxn
betweox       24                       0
betwux        93                       1

              117                      1
             99.2%                    0.8%
betweox

(1.) Beginning according to Bately with Schilling (1886: 56-60) and Pogatscher (1888: [section][section] 247n, 310, 317, 329, and 340n).

(1a.) I thank Professor Clemoes for making this unpublished correspondence available to me and Professor Bately for permission to use her so long forgotten utterance.

(1b.) One slight exception is the Lambeth Psalter, where in the single word pider Lindelof (1914: 82) notes that spellings -d-l-p- outnumber -d-. This is curious in view of geographical closeness apparent in n. 26 below.

(2.) I do not see anything material in Cosijn 1 [section]148 cited generally for further possibilities by Bately.

(3.) E.g. in the Old Welsh poem Gododdin (ed. Williams 1938) 738 rector and 731. ractaf correspond to rheithor and rhagddaf in modern spelling. The sound-changes involved are discussed by Jackson (1953: 404-411).

(4.) Charters are cited by the numbers of charters in Sawyer (1968), refined as Kitson (1990: 186-187) where there is more than one boundary to a charter.

(5.) Two of the three stapolz, with the interchange of e and z which is the characteristic quirk of the main scribe of the Winchester cartulary.

(6.) This form which presupposes second fronting is, not surprisingly, in a charter of the ninth century with other Mercianizing features: med [or med, bergas for beorgas. On Mercianizing charters generally cf. Kitson (1993: 20-21 n. 72), on bergas as an extreme Mercianism Kitson (1993: 19 n. 64) and in much more detail Guide [section] 8.3.92.

(7.) Bately's (1980: cxiv) "conceivably forming part of the vocabulary of the Anglo-Saxons from an early date" seems too weak to me. Her further remark that "...Donua is a form not found outside Or." should not have been taken by Breeze (1992b: 431) as a counter-indication, since the only ready example of naming the Danube at all in Old English outside the Orosius is Danubie in lines 37 and 136 of Elene by the Latin-literate poet Cynewulf. His learned borrowing is no more evidence against the existence of an inherited form of the name than is the affectation of some modern English scholars of listing books as published at Munchen against the use by their countrymen of the inherited form Munich.

(8.) I.e. British Library Additional Manuscript 47967.

(9.) This is not phonctic variation but a generalization of the vowels descending from different personal forms (Campbell [sections][sections] 331.6). The possibility deserves mention however that continuance and/or degree of selection for it correlated with preference for -a- forms in other words where they are phonetic, e.g. famously margen for morgen and/or dative mergen (campbell [sections] 156), warhte for wrohte (Vleeskruyer 1953: 101), which though widespread, mainly in Anglian dialects, as sporadic occurrences, have a well-known bias toward the south-west midlands (Vleeskruyer 1953: 100-102; cf. d'Ardenne 1961: 188). These items may well be relevant to Omarus cited above.

(10.) The former spelt oe- thrice, with the rune ae once, the latter oe- twice beside e- once (the fourth citation in Bately's word-index is a passage not in the Lauderdale manuscript).

(11.) For this reason the much wider spread in southern England recorded by Wright (1898-1905, s.v. oaves) is not evidence for an underlying presence of rounded forms of generally in sub-literary West Saxon; and at least some of it looks likely to be post-Old English (Guide [sections] 6.25.1). Halliwell's (1850: 595) Devon-centred ovvis and the gloss weor [or wer "man" reported from Cornwall by Le Duc (1979) (whose eo before a single consonant can hardly be breaking, so presumably is a spelling for r of the same kind as noted in ninth-century Surrey documents by Campbell [sections] 291 after Ekwall 1923) may well however be evidence for retention of rounding in favourable phonetic environments down the whole Bristol Channel littoral.

(12.) And even one delicious spelling in a Guardian political report a few years ago, predicting if memory serves that someone "wool" be a Minister of the Crown.

(13.) The need to choose between West Saxon and Anglian forms of individual words may even be why the scribal centres whose writings exhibit most history of precision of house style at the level of particular words, Bodley 343 in the first three quarters of the twelfth century and the "AB language" in the second half of the twelfth and first quarter of the thirteenth century, are in the south-west midlands, possibly Hereford and Wigmore (see Kitson 1992 and Dobson 1976 respectively). But this may be an illusion of scholarship, in that precision more minute than that of whole sound-changes has hardly been sought in Old English texts, and the manuscripts most likely to give clear evidence for it if it existed, the copies of AElfric's homilies least close to AElfric's own practices, mostly have not been separately edited nor their readings more than cursorily reported in editions of AElfric.

(14.) I would emphasize that we are dealing with long vowels, and an interchange between eo and u rather as in second class strong verbs, not at all as Campbell [section] 338 n. 1 posits with shortening of io in low sentence-stress and its retraction. In betweoh the syllable in question is a stressed syllable, unlike Campbell's examples of retraction of io nanwuht "nothing", fulluht "baptism".

(15.) That isogloss is also the one least constrained by data, except in Berks/Oxon (Kitson 1993: 16). One or both of these points must be borne in mind in interpreting the large minority of -u- forms in the works of Alfred (Kitson 1993: 25), 48% of "between" words as a whole, comprising 29% for betweoxn and a small majority, 52%, for words other than that, which accords very well with the indications that betweoxn originates in a south-eastern stratum.

(16.) A betuh was missed at Kitson (1993: 17) owing to duplication in one Mic. Conc. citation.

(17.) This in combination with the charter facts reported Kitson 1992b [section] 29 suggests that if the "nur westsachsich" combinative u-mutation had a distribution enclosable in a neat isogloss its centre would be in southern Oxfordshire, north of the Thames. A more probable inference would be that its distribution could not be expressed in that way, but only as a dine of percentages in different areas as for dative plurals below.

(18.) And in others, notably tu beside twa "two" neuter. There the change is pre-Old English (Campbell [section] 122); but the parallel adds to the reasons, which are already sufficient (Kitson 1993: 16 n. 56, and cf. mutatis mutandis pp. 19-20), for insisting that the vowels in these words remained long.

(19.) If Bately's presentation (1980: xlv) suggests otherwise it is misleading, as the statement on the preceding page that "by the time the manuscript was written the unstressed back vowels u, o, and a had largely coalesced in a single unaccented back vowel, and that this was becoming--or had become--confused with unaccented e" is on the observable frequencies very highly misleading. There is significant levelling before n and in some particular grammatical categories, but the supposed evidence for levelling of e with back vowels is too sparse to be evidence of any general tendency in the language and must be put down to random scribal error. On the need for greater discipline in argument from scribal variants to linguistic tendencies than has prevailed in studies of Old English unstressed vowels in the last sixty years cf. Kitson (1992: 38, 46, 57-58, 77-78).

(20.) Including scribal errors that belong obviously to one of the three, i.e. -ym under -um, -ond and -one under -on. These occur once each, the first in Warks, the last two in corrupting cartularies in Wilts. Spellings -u with suspension indicating a final nasal but not specifying which are excluded, but here is a further possible source of distortion in the figures: editors treat them inconsistently, and I have not checked all the manuscripts. But nasals are shown by suspension much more rarely in manuscripts of charters than in late Old English literary manuscripts, so any distortion is probably not great.

(21.) Forster (1924: 4) took the indeclinability as favouring his then theory of a learned blend, but later he recognized that a substantial number of ordinary Old English river-names are indeclinable, putting the figure at 40% (Forster 1941: 314-342, whence Campbell [section] 628.6 n. 2). That is also the figure I would draw for indeclinable feminine from a sample of 250 simplex river-names in charter boundaries, compared to 26% weak feminine, 24% strong feminine, 9% masculine/neuter (presumably masculine). Forster's figures for the other categories from a sample of some 200 names of what he calls "die Masse der schon altenglisch belegten FluBnamen" (many patently from charter boundaries, not all actually early) were 33% weak feminine, 27% strong feminine; he did not acknowledge masculine names as a category, but they comprise at least 1% or 2% of his names (1941: 320, 331-2). There is some passage of individual names between categories, also recognized by Forster (1941: 332-338), 50 the figures cannot be pr essed beyond the nearest percentage point, and perhaps not quite there. But they do suffice to show that of foreign river-names in the Orosius the treatment of Donua as indeclinable does not require special explanation, and that of Rin "Rhine" as strong masculine does.

(22.) Campbell [section] 628.4 saying "West Saxon has frequently also g. and d.s. ie" is inaccurate on this point. Sievers-Brunner [section] 284 n. 4 is even less focussed.

(23.) Since non-Germanic cognates (Latin aqua etc.) show the word was originally an ordinary a-stem (so Campbell; Sievers-Brunner's treating it under minor consonant-stem declensions, whether on the basis of the fronted genitive or of the etymology mentioned in the next note, is erroneous.

(24.) Van Helten regards ee as originally a consonant-stem answering to Sanskrit aep-. That etymology is incredible, given the absence of a final consonant anywhere in the attested forms (whether regular Germanic f or p(p) from derivative formations as in other names cited by Pokorny I 51-52). Since he adds that it behaves like an i-stem, i.e. effectively indeclinable, and since the sources for it are all place-name material, we should see it as an original dative fossilized in place-names in exactly the same way as Old English dative phrases like aet paere ea are fossilized in the Middle and modern English river-name Ray and Rea. The word is not indexed, nor apparently mentioned, by Lasch (1914).

(25.) Amusingly, despite Middle English distortion in both, rivers Yeo and Rea reflect a West Saxon/Anglian contrast.

(26.) This means that the Lambeth Psalter-gloss, which has occasional pweres beside pwyres (Kitson 1993: 46), can happily be from further north in Gloucestershire. Depending on whether the isogloss implied by the position of the Orosius in Bristol for the southern end of the distribution of betweonan was more like Kitson (1990) map 1 or map 8, a provisional placing indicated might be respectively Gloucester, close to the later mediaeval provenance of Llanthony in terms of which I rather unconvincingly discussed it (1993: 42) or somewhere further east such as Cirencester. Exclusively masculine gender of hyll (Kitson 1990: 219; overlooked in my 1993) would consort better with the latter.

(27.) Also figurative uses in the thirteenth-century John Mirk (also west midland) and a 1250 Lofsong Louerdes (Nero), and a derivative adjective fledlyk translating decursurn in the fifteenth-century glossary Medulla.

Back in Old English, an ea inflede in Genesis 232, which being the Tigris is presumably no more than "full of water", and one in Andreas 1504, where "in spate, flooding" would be highly appropriate, are likely to be poetic diction not significant of anything dialectal. But a noun oferflewedness in the Regularis Concordia gloss may just be in significant contrast to oferflowedlicness common in AElfric and found in the Benedictine Rule and oferflow(ed)(lic)ness occasional in other authors; as both with their underlying weak past participle must be with the strong past participle oferfleowen in the homily Belfour XII (MS Bodley 343, from ?Hereford) and oferflowen in Mic. Conc.'s homily "S12" (MS Hatton 114, from Worcester).

(28.) Where despite statements at the heads of the entries for betweoh, betweonan, and betwux that they are organized in the same way, a category "with preposition in postposition" appears for betweoh (as A.4.a.) alone of the three. Betweonan, for which alone it is significant, does net have one. Worse than that, be...tweonum with tmesis, which constitutes a significant section in itself, is not given one, but is split between four widely separated sections according to shades of meaning.

(29.) Betweoxn is treated as a sub-type of betwux, and the spellings for betweon are divided between betweoh and betweonan.

(30.) Comparison with the Dictionary of Old English, not available to me when I produced my former figures (1993: 11), reveals that I missed a few of the weirder deviant spellings; also I counted a few more items than I can now find for betweox. The concept of percentages of sample is complicated anyway by the Dictionary's listing some spellings not in the Concordance, most but perhaps not all in variant manuscripts of texts already there, presumably most but not necessarily all representing single items. The figures 1 would now offer are betweox, 787 items from 27 spellings, plus 11/12/13 not in Mic. Cone. (two unresolved discrepancies for particular items); betweoh, 563 items from 24 spellings, plus 7 not in Mic. Conc.; betweonan, 452 items from 42 spellings, plus 7 not in Mic. Conc.; betweon, 85 items from 24 spellings, plus 1 not in Mic. Conc.; betweoxn, 36 items from 13 spellings, plus 7 not in Mic. Conc.

(31.) One thing there I would now retract, though, is the tentative guess of N. Berks or E. Oxon as provenance of the archetype of the West Saxon Gospels. The Corpus manuscript, likely to be the closest extant manuscript to the original text (Liuzza 1994: lxxiv etc.), was written in Bath (Liuzza 1994: xxvi), which actually better fits the high frequency of u in "between" words. Usage in them is distinctly similar to that of the Orosius, with prepositional betwux the norm but postpositional betwynan after personal pronouns a fairly common variant. A local document quoted by Liuzza (1994: xxx) from another Bath manuscript has in and habbe we us geraedd betweonan to ure saule pearfe "and we have agreed between us for our souls" "benefit" a nice example of distantly postpositioned betweonan after a personal pronoun. The discrepancy in the Gospels between apparent indications from numerous "between" words and the single item of unsyncopated foranongen is then to be dealt with either by slicing a bit off the forage an isogloss or by postulating a difference between educated and uneducated usage for that item. Much more minute work would be needed to distinguish any second layer of dialect, which might reflect the archetype, from that of the extant Corpus manuscript.

(32.) Be us tweonum in Vercelli Homily XIV (line 115 in the new EETS edition by D.G. Scragg), be us tweonum and be him tweonum in Blicking Homily XIII ([143.sub.12] and [139.sub.23] in the EETS edition by R. Morris (1880), nicely contrasting with betuh preceding noun phrases twice at [143.sub.18] in the same homily).

(33.) [8.sub.16] in the EETS edition by A.S. Napier (1894).

REFERENCES

d'Ardenne, S.R.T.O. (ed.)

1961 [1970] e liflade ant te passiun of Scinte Iuliene. (Early English Text Society OS 248). Oxford: Oxford Unversity Press.

Bately, Janet (ed.)

1966 "The Old English Orosius: the question of dictation", Anglia 84: 255-304.

1970 "King Alfred and the Old English translation of Orosius", Anglia 88: 255-304.

1978 "The compilation of the Anglo-Saxon Chronicle 60 B.C. to A.D. 890: vocabulary as evidence", Proceedings of the British Academy 64: 93-129.

1980 The Old English Orosius. (Early English Text Society SS 6). Oxford: Oxford Unversity Press.

Benskin, Michael - Margaret Laing

1981 "Translations and mischsprachen in Middle English manuscripts", in: Michael Benskin - Michael Samuels (eds.), 55-106.

Benskin, Michael - Michael Samuels (eds.)

1981 So meny people longages and tongues: Philological essays in Scots and mediaeval English presented to Angus McIntosh. Edinburgh: Middle English Dialect Project.

Breeze, A.C.

1991 "Cornwall and the authorship of the Old English Orosius", Notes and Queries 236: 152-154.

1992a "Cornish Ligore, 'Loire', and the Old English Orosius", Neuphilologische Mitteilungen 103: 271-273.

1992b "Cornish Donua, 'Danube', and the Old English Orosius", Notes and Queries 237: 431-433. Campbell, A.

1959 Old English Grammar. Oxford: Oxford University Press.

Cosijn, P.J.

1883-1886 Altwestsachsische Grammatik. 2 vols. The Hague: Nijhoff.

Cubbin, G.P.

1979 "A case of homonymic clash in Germanic", Indogermanische Forschungen 84: 226-236.

Dees, A.

1980 Atlas des formes et des constructions des chartes francaises du 13e siecle. Tubingen: Niemeyer.

1985 "Dialectes et scriptae l'epoque de l'ancien francais", Revue de Linguistique Romane 49:87-117.

1987 Atlas des formes linguistiques des textes litteraires de l'ancien francais. Tubingen: Niemeyer. Dobson, E.J.

1976 The origins of Ancrene Wisse. Oxford: Oxford University Press.

Ekwall, E.

1923 "On the Old English fracture of a before I followed by a consonant", English Studies 5: 57-63.

1928 English river-names. Oxford: Oxford University Press.

Fisiak. J. (ed.)

1995 Medieval dialectology. Berlin: Mouton de Gruyter.

Forster, M.

1924a "Der Name dcr Donau". Zeitschrift fur slavische Philologic 1: 1-25.

1924b "Zum Donau-Namen". Zeitschrift fur slavische Philologic 1: 418.

1941 Der Flussname Themse und seine Sippe. Munchen: Verlag dcs Bayerischen Akademic der Wissenschaften.

Freeman, R. Austin

1939 Thc stoneware monkey.

[1987] Reprinted New York: Dover.

Gelling, M.

1984 Place-names in the landscape. London: Dent.

Gneuss, H.

1972 "The origin or Standard Old English and /Ethelwold's school at Winchester", Anglo-Saxon England 1: 63-83.

Grinsell. L.V.

1953 The ancient burial-mounds of England. (2nd edition.) London: Methuen.

Halliwell, J.O.

1850 Dictionary of archaic words. London: Smith

[1989] [Reprinted London: Bracken.]

Healey, A. diPaolo et al.

1985 Dictionary of Old English. Toronto: Pontifical Institute for Medieval Studies.

Heighway, C.

1987 Anglo-Saxon Gloucestershire. Gloucester: Alan Sutton.

Helten, W.L. van

1890 Altostfriesische Grammatik.

[1982] [Reprinted Schaan, Liechtenstein: Sandig.]

Heuser, W.

1903 Altfriesisches Lesebuch. Heidelberg: Winter.

Hofstetter, W.

1987 Winchester und der spatahenglische Sprachgebrauch: Untersuchungen zur geographischen und zeitlichen Verbreitung altenglischer Synonymne. (Munchener Universitats -Schriften, Philosophische Fakultat, Texte und Untersuchungen zur Englischen Philologic 14.) Munchen: Fink.

Hogg, R.

1992 A grammar of Old English. Volume I: Phonology. Oxford: Blackwell.

Jackson, K.H.

1953 Language and history in early Britain. Edinburgh: Edinburgh University Press.

Jordan, R.

1906 Eigentumlichkeiten des anglischen Wortschatzes. (Anglistische Forschungen 17.) Heidelberg: Winter.

Kitson, P.R.

1990 "On Old English nouns of more than one gender", English Studies 71: 185-221.

1992a "old English dialects and the stages of the transition to Middle English", Folia Linguistica Historica 11: 27-87.

1992b "On the chronological and geographical spread of Old English combinative u-mutation", studia Neophilologica 64: 3-23.

1993 "Geographical variation in Old English prepositions and the location of AElfric's and other literary dialects", English studies 74: 1-50.

1995 "The nature of Old English dialect distributions, mainly as exhibited in charter boundaries. Part I: Vocabulary", in: J. Fisiak (ed.), 43-135.

forthcoming Guide to Anglo-Saxon charter boundaries. English Place-Name Society.

Kurath, Hans -- S. M. Kuhn et al. (eds.)

1954- Middle English Dictionary. Ann Arbor: University of Michigan Press.

Laing, M. (ed.)

1989 Middle English dialectology. Aberdeen: Aberdeen University Press.

Lasch, A.

1914 Miuelniederdeutsche Grammatik. Halle: Niemeyer.

Le Due, G.

1979 "Une glose anglo-saxon glosee en Brittonique", Etudes Celtiques 16: 261-262.

Lindelof, U.

1914 Der Lambeth-Psalter. Acta Societatis Scientiarum Fennic"" 43. iii. Helsinki: Societe Neophilologique.

Liuzza, R.M.

1994 The Old English version of the Gospels. (Early English Text Society OS 304.) Oxford: Oxford University Press.

McIntosh, A.

1973 "Word geography in the lexicography of medieval English", Annals of the New York Academy of Sciences 211: 55-66. (repr. in Laing 1989).

McIntosh, A. et al.

1986 Linguistic Atlas of Late Mediaeval English. 4 vols. Aberdeen: Aberdeen University Press.

Mitchell, B.

1985 Old English Syntax. 2 vols. Oxford: Oxford University Press.

Napier, A.S.

1894 History of the Holy Rood-tree. (Early English Text Society O.S. 103.) Oxford: Oxford University Press.

Pogatscher, A.

1888 Zur Lautlehre der griechischen, lateinischen und romanischen Lehnworte im Altenglischen. Strassburg: Trubner.

Pokorny, J.

1958-68 Indogemanisches etymologisches Worterbuch. Berne: Francke.

Richthofen, K.F. von

1840 Altfriesisches Worterbuch. Gottingen. [No publisher.]

Samuels, M.L.

1963 "Some applications of Middle English dialectology", English Studies 44: 81-94.

1972 Linguistic evolution with special reference to English. Cambridge: Cambridge University Press.

Sawyer, P. H.

1968 Anglo-Saxon charters: an annotated list and bibliography. London: Royal Historical Society.

Schilling, H.

1886 Konig Alfreds angelsachsische Bearbeitung der Weltgeschichte des Orosius. Halle/S.: Niemeyer.

Sievers E. -- K. Brunner

1965 Altenglischen Grammatik nach der angelsachsischen Grammatik von Eduard Sievers (3rd edition.) Tubingen: Niemeyer.

Sprockel, C.

1965 The language of the Parker Chronicle. The Hague: Nijhoff.

Stenton, F.M.

1927 Introduction to The place-names of Worcestershire, ed. A. Mawer and F.M. Stenton with F.T.S. Houghton (eds.). (EPNS v. IV) Cambridge.

1971 Anglo-Saxon England. (3rd edition.) Oxford: Oxford University Press.

Stevenson, W.H.

1959 Asser's Life of King Alfred. (2nd edition.) Oxford: Oxford University Press.

Sweet, H.

1883 King Alfred's Orosius. (Early English Text Society O.S. 79.) Oxford: Oxford University Press.

Venezky, R.L. - A. diPaolo Healey

1985 Microfiche concordance to Old English. Toronto: Pontifical Institute of Mediaeval Studies.

Vleeskruyer, R.

1953 The Life of St. Chad: An Old English homily. Amsterdam: North-Holland.

Williams, I.

1938 Canu Aneirin. Cardiff: University of Wales Press.

Wright, J.

1898-1905 The English dialect dictionary, 6 vols. Oxford: Oxford University Press.

1905 The English dialect grammar. Oxford: Oxford University Press.