Now we may take the evidence for the proposed date of the loss of -u. After enumerating (p. [253] ff.) a list of cases which have been suggested by various scholars Prof. Morsbach comes to the conclusion that the only certain example of a form in which -u is retained[103] is the word flodu in the inscription on the Franks casket. As the whole theory largely depends on the value attached to this form we must consider it carefully. In my Studies, p. [156], I suggested that it should be regarded as an archaism and at the same time pointed out that -u is lost in another word (unneg) in the same inscription. Prof. Morsbach rejects this explanation of unneg, which he connects, rightly, I think now, with O. Sax. nah. But this form can come perfectly well from *nāh(w)u (earlier *nēhwō), if not from *nāhwa (cf. Goth. nehwa, nehw). That Ang.-Sax. neah, neh has lost an u-sound is, I contend, shown not only by the Gothic forms, but also by neolaecan and neowest (cf. Lind, genehwaþ, etc.). Prof. Morsbach rejects my explanation of these forms also and adopts that of Prof. Bülbring, as stated above. But the latter is untenable; for if ē (from ā) had come into existence before the operation of breaking, we should never find forms with ae, which as a matter of fact are fairly common in the earliest texts of all dialects. In particular we may note that the three extant coins of the Mercian king Aethelred (675-704) all have -ræd. Hence the change ǣ > ē can hardly have taken place much before 650.
Apart from unneg[104], there are three forms on the recently discovered right side of the casket which may show loss of -u. Prof. Napier (An English Miscellany, p. 375 f.) is inclined to regard the forms sær and dœn (?), if not also hos, as Nom. sg. fem. But the interpretation of this part of the inscription is still uncertain in many details.
I confess that since the discovery of the new side I am less inclined to regard the form flodu as an archaism than as a mere blunder. Even in the more intelligible parts of the inscription we find a number of forms which present serious difficulty: Romwalus, Reumwalus[105], gasric, Giuþeasu. The last of these still seems to me to present the best illustration of flodu. If the one is due to the loss of some letters—e.g. su for su(mæ)—the same may be the case with the other[106]. At all events an inscription which presents so many difficulties cannot be regarded as a safe authority on which to base a scheme for the chronology of sound-changes.
The only other instance of -u on which Prof. Morsbach lays any weight is the form aetgaru in the Erfurt glossary. He speaks of it as 'nicht unwahrscheinlich' (p. [257]), though 'fraglich' (p. [264]). Many scholars cite it as an example of -u without reserve. Now in order to form a just estimate of the value of any form which occurs in the glossaries it is obviously necessary to take it in connection with the forms which the other texts show in the same entry. The entry in question (framea—aetgaru) occurs in the Epinal and Corpus glossaries, as well as in that of Erfurt, though the two former have aetgaeru (ætgaeru) for aetgaru. There can be no doubt that the relationship between the three glossaries is as follows:
x (Arch. I)
|
|--------------------|
| |
y (Arch. II) |
| |
|--------------| |
| | |
Epinal Erfurt Corpus
though we do not know exactly how many intermediate stages lie between the extant texts and the original archetype. It will be seen that the question at issue is whether the Erfurt glossary or the other two have kept the original form—for if the archetype had contained both forms we may assume that some trace of the double entry would have been preserved (as in other cases). Now the Erfurt glossary is the latest of the three, it is the work of a foreign scribe and it is very carelessly copied. Moreover, no letter or combination of letters has suffered more than ae. Most usually this combination has been reduced to e; but the loss of e is not infrequent, e.g. smal, hrad, nacthegelae. In view of these facts it is unintelligible to me how anyone can uphold the evidence of the Erfurt glossary against the other two. But in this case it is used to prove the existence of an archaic form for which none of the glossaries elsewhere present a parallel. Lastly, we may remark that though it is frequently assumed that the word gar was an u-stem (*gaizu-) no evidence worth consideration has ever been adduced to prove it[107]. On all grounds therefore we are brought to the conclusion that the evidence for the preservation of -u in aetgaru is not merely open to question but entirely worthless.
We must now notice certain early documents in which -u is clearly lost. Prof. Morsbach mentions the form felt in the earliest East Saxon charter (Birch, 81), which dates from 692-3. Here then -u was lost before the date in question[108]. But we can get back further than this, for the place mentioned is called Vuidmundes felt. Some considerable time must have elapsed after the loss of -u before a noun, even a proper name, could change its inflection and adopt the endings of a-stems.
Next let us take the Northumbrian evidence. The form -gar on the Bewcastle column can prove nothing. In the same inscription however we may find an example in Cyniburug; for in view of hnutu etc. it is probable that consonantal stems used what was originally the Accus. sg. form also for the Nom. sg. (as in Old Norse). I cannot see any probability in the suggestion that this monument may date from some considerable time after 670[109], if it has been rightly interpreted.