[94] Breca, Finn, Fitela, Hengest, Hnaef, Hrethel, Scyld, Weland, Guthhere, Geat, Gifeca, Heoden, Helm, Wald, Beaduca, Frithla, Secca, Gifeca—to which we may add Waelse in Walsingaham. Widia is also to be added, if it is not allowed for the Liber Vitae. For the list of place-names (not the personal names) I am dependent upon Binz, Beitr. XX 141-223.
[NOTE II. ON THE DATING OF CERTAIN SOUND-CHANGES IN ANGLO-SAXON.]
In my Studies in Old English[95], published in 1899, I endeavoured to formulate a scheme for dating approximately the chief sound-changes which took place in English during the first few centuries after the invasion of Britain. In the course of these investigations I was led to the following conclusions (pp. [117], [253] ff.): i. that 'palatal umlaut' in Northumbrian and the dialect of the Vespasian Psalter took place before 650; ii. that the change from ǣ to ē (in all dialects except West Saxon) was in operation about 650-680; iii. that the loss of intersonantal h (in all dialects) belongs to the same period or a little later; iv. that contraction through loss of intervocalic h may be dated roughly between 680-710; v. that the loss of final -u after long syllables and in words of the form ⏑⏑⏑ took place in all dialects at a time approximately contemporaneous with the operation of palatal umlaut in Northumbrian (i.e. before 650).
Prof. Morsbach in his paper Zur Datierung des Beowulfepos[96] has dealt with several of the same problems and come to conclusions which differ widely from those at which I arrived. The chief differences are as follows: i. that after a long syllable which bore the chief accent -u was not lost before the end of the seventh century, though after a long 'nebentonig' syllable the loss was somewhat earlier (p. [261] f.); ii. that intervocalic h was lost in Kentish by about 680, but in Mercian and Northumbrian the same change cannot be shown to have taken place before about 700 (p. [264]); iii. that postconsonantal h (before vowels) was retained in Kentish in 679; its loss, at least in Mercian and Northumbrian, may be dated about 700, but after the loss of -u (p. [265]). In summarising the results of his discussion (p. [273]) he gives "about 700" for the loss of postconsonantal h and "shortly before 700" for the loss of -u. Incidentally he follows Bülbring (Elementarbuch, §§ 146, 528) in dating the origin of ē (from 'West Germ.' ā) before the breaking, and in placing the loss of h before l (in neolaecan) anterior to the operation of 'palatal umlaut' (monophthongisation).
The importance of this discussion for our present purpose[97] lies in the dates proposed for the loss of -u and of postconsonantal h. Prof. Morsbach concludes that Beowulf cannot have been composed before 700, since it contains a number of half-verses which would have been metrically impossible before the operation of these changes, e.g. ofer fealone flod(u), to widan feor(h)e. I have already expressed scepticism as to whether such inferences are really justifiable. This applies more particularly to the verses affected by the question of postconsonantal h, which are quite few in number. The date which Prof. Morsbach himself (p. [274]) proposes for the composition of Beowulf is 700-730. He finds no difficulty in reconciling this with the statistics (given above, p. [45]) for the use of the article. This seems to me rather strange; yet the opinion of a scholar who stands in such deservedly high estimation cannot lightly be disregarded.
Now let us examine the evidence on which these conclusions are based. First it will be convenient to take the loss of h. Prof. Morsbach holds that in Kentish intervocalic h was lost before postconsonantal h. The evidence is derived from a single charter issued by King Hlothhere in 679 (Birch, Cart. Sax. 45), which contains the place-name Vuestan ae beside the personal name Velhisci (Latin Gen.). But surely conclusions of this kind are admissible only when a number of examples can be adduced. On the same principle we might argue from the name Irminredi (in the same charter) that the change ǣ > ē had taken place and also from the name Aedilmaeri (again in the same charter) that it had not. And what should we do with the earliest East Saxon charter (Birch, 81), in which the grantor is called both Oedelraedus and Ho(di)lredus? Again, it is clear that Bede wrote his own name Baeda; but will anyone venture to hold that this represents the current pronunciation of his name in 731—or indeed for some half a century earlier? In personal names we must clearly allow for traditional orthography. The form Irminredi may no doubt be used as evidence for the change ǣ > ē, and similarly the form Vuestan ae may be used as evidence for the loss of h. But forms, especially personal names, like Velhisci and Aedilmaeri, which must long have been in use, may very well show an antiquated orthography—one which correctly represented the pronunciation of thirty or forty years previously. A single instance of such a kind is totally insufficient ground for supposing that the Kentish dialect treated postconsonantal and intervocalic h differently.
Next we must consider the date given for the loss of h in Mercian and Northumbrian (p. [263] f.). I find some difficulty here in following Prof. Morsbach's line of argument. In Northumbrian there is, admittedly, no evidence at all for the preservation of h, while cases of its omission are numerous in Bede's History (written in 731), in addition to one or two instances in probably earlier authorities. For Mercian[98] we are dependent on the Epinal, Erfurt and Corpus glossaries, the archetype of which is placed before 700 by Prof. Morsbach. In my Studies, p. [232], I came to the conclusion that in this archetype the cases of retention and omission of h were probably about equal in number. Prof. Morsbach replies that there is no necessity for such a conclusion, since all the extant glossaries themselves date from times when h was already lost. He himself decides[99] against the loss of h in the archetype for two reasons: (1) because postconsonantal and intervocalic h are treated alike in the glossaries and the former was still retained in Kentish when the archetype was written[100]; (2) because the assumption of such an early date for the loss of h would be incompatible with his own date for the loss of -u. The first of these arguments, it will be seen, rests upon the dating of the loss of postconsonantal h in Kentish, on which enough has been said above. The second depends upon a hypothesis which we shall have to consider presently.
My reason for concluding that the loss of h occurred in the archetype was that in at least eight entries (probably several more) all three glossaries agree in showing forms without h. It is to be remembered that in these glossaries we are dealing not with independent documents but with copies made, more or less mechanically, from one original. This remark applies of course much more to Epinal and Erfurt than to Corpus; for in the latter the materials have been rearranged, as well as augmented from other sources, while incidentally the forms have been modernised to a considerable extent. In Epinal such modernisation is not unknown, but it is restricted within very narrow limits, as may be seen from the use of b and f[101]. Further, it is to be remarked that we have no ground for assuming the language of the archetype itself to have been consistent. The occurrence of numerous Dative forms and of expressions containing more than one word (e.g. per anticipationem—þorch obst) shows that the materials were derived largely from glosses in books, just as in the Leiden glossary[102]. Many of these glosses may have been written a generation or more earlier than the compilation of the archetype glossary. Whoever bears these facts in mind and at the same time compares the evidence for forms with and without h with that for f and b will, I think, be forced to the conclusion that the forms with h do not represent the pronunciation of the compiler of the archetype, but that they were taken over by him from earlier sources.