is quoted by Ordericus Vitalis, in his Hist. Eccles., in Book XII. chap. 47. (a Book apparently completed in 1136 or 1137).

(3). 1138. Two Grants, made by Archd" Walter to Godstow, on occasion of the Dedication of the Church of St Giles (2nd April, 1138). The first Grant is witnessed (among others) by Radulphus de Monem," and "Magister Gaufridus Arturus", and the 2nd Grant by the same: "Magistro Galfrido Artdo" and "Radulpho de Monumuta".

By the end of the year 1138 the first edition of the Historia was finished; for it was seen by Henry of Huntingdon in the Abbey of Bec (near Rouen) in January 1139. — Robert of Gloucester (to whom the Historia was dedicated) was styled in his youth Robert of Caen; and it was probably he, who sent it to Bec. Huntingdon gives an Abstract of the Historia, in a Letter addressed to one "Warinus Brito" (see my pp. 210211), styling it "librum grandem Gaufridi Arturi”.

Archdeacon Walter was a man of some official importance under Henry I. (see my page 218). After Henry's death (1135) we lose sight of him; but he lived till 1151; and he probably had at least some access to his new Diocesan, Robert de Chesney, (Bp. of Lincoln in 1148): It was just then that Geoffrey found it necessary to turn to a new Patron; and he accordingly dedicated his metrical Vita Merlini to Bp. Robert, begging him to show more favour to the Author, than had been shown by his predecessor, (i. e. Bishop Alexander): "Ergo meis ceptis faueas. uatemque tueri Auspicio meliore uelis. quam fecerit alter. Cui modo succedis merito promotus honore": (see my vol. I. p. 280).

Geoffrey probably hoped for a good canonry in Lincoln: such as fell (some 30 years later) to the lot of another Welshman, Walter Map. But he had to content himself with a Welsh Archdeaconry, (probably in succession to his Uncle), and with his chaplaincy to Earl William (son of the great Earl Robert) of Gloucester; till he was consecrated Rp. of St Asaph on the 24th Feb. 1152 (see my p. 205); and even then, according to the Gwentian Brut. (see my p. 204) "he never entered upon his functions";

(North Wales being

then in dire confusion); "but he died in his house in Llandaff in 1154".

We may conjecture that Geoff. took his Degree (probably at Paris) between 1129 and 1138. But the evidence is not conclusive; for the names of the Witnesses are crowded together in our copy of the Oseney Cartulary, and the Scribe may have omitted "Magistro" for want of room.

So much for the original documents! If I might add a very small amount of guess-work, I should regard them as telling the following story. Archdeacon Walter of Oxford brought home from Brittany an Old-Welsh Ms., containing many British genealogies and several historical glosses. He had not leisure (perhaps not skill enough) to translate these into Latin, and arrange them. — He naturally turned to South Wales, where Robert of Gloucester was Prince of Glamorgan, and Urban was Bp. of Llandaff. - One of these (perhaps both of them) recommended Geoffrey, the nephew and fosterson of Uchtryd, at that time Archdeacon (and afterwards, in 1140, Bp.) of Llandaff. Geoff. did his work, from first to last, under the sanction of Archdeacon Walter; and he especially notes the historical assistance afforded by the Archdeacon, when he is approaching the last battles of Arthur: (Book XI. chap. 1).

HAMPSTEAD. LONDON N. W.

HARRY L. D. WARD.

OE. RÆSN, REN ÆRN, HRÆN HÆRN.

It is customary to parallel OE. ærn 'house' and Go. razn 'house'. Not long ago it occurred to me that OE. ræsn 'ceiling', 'timber', must also be associated, and on looking it up I found that Ettmüller long ago suggested the equivalence of Go. razn and OE. ræsn but was not aware of the relation of OE. arn to the other words. Now that the association of Go. razn and OE. ærn is generally accepted, Ettmüller's suggestion seems to have been largely lost sight of, though it is still to be found in Fick and Skeat (under ransack).

It is clear that we have to assume Gc. rás-na- raz-na-. The former is preserved, so far as I know, only in OE. ræsn. The latter appears in Gothic as razn and in Old Norse as rann, with the regular assimilation of zn, rn to nn (Brugmann 2 I. p. 778, Noreen Altisländ. Gr. § 208). In Old English this assimilation is not normal, and we should expect *rearn < *rærn, *rarn, cf. leornian, Go. lais. But the word never appears with both r's and it is evident that this is due to the working of dissimilation. The case is nearly parallel to that of *hrærn (= ON. hronn < *hrarnō-, *hraznō-) which appears in Old English as hræn and hærn.

It will be necessary for us to consider carefully the form of the differentiation. It might be suggested that it was of the type r-r > r—, that is, that the second r was dropped as in OE. cwearternes > cweartenes (Hpt. 513, JGP. II. 361), OHG. prart > prat, Lat. aratrum > Sp. and Prt. arado, and such names as Northrup > Northup, Purmort > Purmot, etc. This explanation would hold for OE. ærn, which is clearly due to metathesis (Sievers, § 179) of *ræn, older *rærn =

Go. razn,
became
r+consonant.

r;

provided the dissimilation took place after a, a, before a nasal and earlier than a became ea before The dissimilation could not have had the form r - r > for in that case the a would have stood before rn from the beginning and this early *arn would have broken to *earn. Though the dissimilation r-r>r might thus have taken place in ærn, it could not have taken place in hræn and ren, unless we assume, which we are not justified in doing, that the dissimilation took place later than the loss of -u and i after long syllables (Sievers, § 132). For hræn(n) is an ō-stem1) (= ON. hrọnn) and ren(n) is an i-stem. If *hrærnu had hecome *hrænu, with the absolute loss of the second r, the final -u would not have disappeared after the short syllable and we should not find hræn(n). And had *rærni become *ræni with absolute loss of the medial r, the final would not have disappeared and we should not find ren(n) but *rene. We must, therefore, assume that the dissimilation was of the type r rr- n, that is, that one r changed to n, cf. OE. ortÿdre > ontydre. This is particularly likely to happen in words that (like *rærni and *hrærnu) already have a nasal; cf. Lat. meretrix > menetris (Lindsay, The Latin Language, p. 96, § 104), OE. grorn > gnorn, hordernum > hondernum (Vesp. Hy. 7, 51). This gives us as prehistoric OE. forms *rærna-> *rænna- and *hrærnō*hrænno. As we shall see directly that ren is an outgrowth of *ræn ærn, it is now incumbent upon us to assume the same form of dissimilation in the latter as in ren and hræn, namely, r―r > r―n, and not the dissimilation (r r>r) admitted above as possible in the case of arn considered alone. That we have thus to deal with early *rænna- and *hrænnōis made even more certain by the metathesized forms ærn

1) Sievers suggests (§ 89 A4) that ærn and hærn are from *ranni and *hranni with mutation checked at the a-stage by the metathesis. In the case of arn this would a priori be possible (it implies: *rarna- > *rarni> *ranni > *ronni > *rænni > *rænn > ærn); but I shall show directly that it is not necessary or likely. As much may be said of hærn; but, as there would be no reason for the checking of the mutation in the case of the unmetathesized form (cf. wrenna by the side of warna), *hranni would give us a *hren(n), whereas we find hræn(n). There is therefore no reason to suppose that the word ever changed its feminine gender and ō-declension.

and hærn; for, as Sievers points out (§ 179), such metathesis of prevocalicr takes place before nn and before s-combinations (misprinted "r-verbindungen" in the new edition), but not before single n.1)

We have seen that *hrænno- became hræn(n) and by metathesis hærn; similarly *rænna- would regularly appear as *ræn(n), the a coming to stand before n after the change of a to g. One would infer from Sievers that both *ræn and ren were found in the glosses. We find, however, only ren(n). That this is an i-stem is clear. Moreover, we know (Sievers, § 267) that it was not unusual for neuter a-stems to become i-stems. In fact, the word we are considering throws light on the process by which this change of declension came about. Sievers has shown that the early form of the instrumental of a-stems in Old English ended in i in contrast to the -æ of the dative, and he suggests that this instrumental is really an old locative with i, Gc., IE. ei, from the e of the stem and the locative ending i; PBB. VIII. 324, Angelsächs. Gram., § 237 A 2. cf. also Jellinek, HZ. XXXIX. 130, and Streitberg Urgerm. Gram., p. 229. The development of our word fits this theory perfectly. The old locative of the a-stem *rænnawould be *rænni. The locative of the word for 'house' or 'home' was, of course, much in use, and it was through this fact that the form became associated with the i-stems and that the rest of the inflection of the word ultimately came to accord with the i-declension. As I have shown in an article. on English beach, beck, pebble, a similar frequency of the use of the case (whether in its original function of a locative or in its later function of an instrumental) must be assumed as the cause of the passing of other nouns from the a-declension to the i-declension.

We thus have in the OE. hræn hærn, a primitive *hrarnō(= ON. hro̟nn); in OE. *ræn ærn, a primitive *rarna- (= Go. razn); in OE. ren, an i-stem that arose out of the locative of *rarna-; and in OE. ræsn, a primitive *rasna-.

It is natural to ask: Does the co-existence of rás-na- and

1) Sievers suggests that cornuc may be for *cronuc and thus constitute an exceptional case of metathesis before single n; but, as I shall show in detail elsewhere, cornuc is from *cronnuc.