as gigantic monsters rather than as Northmen and certainly lends some support to the conceptions found in the late versions of our story of the monstrous inhabitants of the islands as descendants of Cain or Ham.98 It must be admitted, however, that the author may have been thinking of the Northmen when he described the martyrdom of the ship's company on the Isle of the Catheads. Neither Poem nor Prose A refers to Ham or Cain in the account of the Swineheads, and traditions of actual experiences of Ionan clerics among the heathen Norse are not improbably reflected. This does not mean, of course, that the author had any specific, particularly satiric, purpose of identifying the Norse with the monsters as such.99 Zimmer's attempt to identify the sources of Sections 5 and 8 seems needlessly complex; 100 but does not require discussion here.

1905) for text and translation of Duald mac Firbis' sixteenth-century treatise "On the Fomorians and the Lochlannachs."

"An interesting Oriental parallel may be found in a quotation from Legends of the Koran (quoted by Bassett, p. 298): the plague-stricken descendants of the maker of the golden calf dwell on an island in the Red Sea, whither Moses had sent their guilty ancestor. When a ship approaches, they run to the beach to cry, "Touch me not."

"The dogheads in Oriental legend are called idolaters (Yule, Marco Polo, loc. cit.). Grendel and his dam in Beowulf are heathen and are descendants of Cain.

100 The elements which Zimmer traces to the fragmentary LU version of "The Two Sorrows," to Imram Hua Corra, and to Dicuil were all present in traditional vision literature accessible to the author (Fís Adamnain itself seems as old as our imram, Windisch, Irische Texte, 1, 167), and in current conceptions of terra repromissionis. Bird-islands appear in Imram Maelduin, Episodes 3, 18, 19 (cf. Ep. 10); in Navigatio Brendani, Sec. 10, and elsewhere. On the possible geographical basis for the accounts of bird-islands in early Irish literature, see D. Du Noyer, Proc. R. I. A., vm, 429 ff.; Schröder, Perigrinatio S. Brend., p. 40, note 6; Westropp, Proc. R. I. A., xxx, 237; as well as Dicuil. The tree with birds as a feature of otherworld descriptions seems common to pagan and Christian literature, A. C. L. Brown, [Harvard] Studies and Notes, VIII, 82 ff.; Plummer, VS, 1, clii; and Boswell, Irish Precursor of Dante (London, 1908), p. 85. The notion of souls appearing in the form of birds is also common to pagan and Christian tradition (Plummer, p. cxlvii), and is a widespread motive, especially common in Celtic: see T. P. Cross, R. C., XXXI (1910), p. 437, note 2.

In an earlier article,101 I presented, in condensed form, evidence of the extremely close similarity between Ionan tradition recorded in Adamnan's life of Colum cille and the imram tradition, and called particular attention to the career of Cormac mac Liathain,102 itself almost a complete imram. I also surveyed some of the evidence which indicates not only that the sixth and seventh centuries constituted a heroic period in Irish church history and were the era par excellence of the sea pilgrim, but that all other imrama have their settings in this early period. These facts, I think, offer strong corroborative evidence that the original time-setting of the legend of Snedgus and Mac Riagla, clerics of Colum cille, was this heroic era rather than late eighth century. The chronological error of making Colum cille a contemporary of early seventh century kings is chargeable, I believe, to the original author of the tale, who did not think in terms of centuries and who did not scruple to make Colum cille a contemporary of the son and grandson of royal persons with whom the great saint did have intimate personal relations.

The University of North Carolina.

101 "Clerical Sea Pilgrimages and the Imrama," Manly Anniversary Studies in Language and Literature, pp. 276-283.

103 For the text, see Adamnan, Vita Sancti Columbi, Lib. I, Cap. vi and II, xlii (Reeves or Fowler edition-Skene's edition in "Historians of Scotland" series has different chapter numbers).

THE OLD FRENCH RETENTION OF LATIN a IN

UNACCENTED SYLLABLES

BY HENRY DEXTER LEARNED.

A statement long dignified by the name of "Darmsteter's Law"1 calls for a division of a Vulgar Latin polysyllable into two more or less equal parts, one containing the tonic syllable together with the final, the other including whatever syllables may precede the tonic; so canta-torem, quadri-furcum; both parts following parallel courses of development, each with an accented and an unaccented syllable, the vowels of each unaccented syllable being eventually dropped in French. Two exceptions are noted: (1) any vowel will be retained as "mute" e, in either part of the word, if its loss would produce an awkward consonant group; and (2) the vowel a is retained, as "mute" e, in any case. Thus canta-torem gives chante-dor, quadri-furcum gives * quadre-forc, French carrefour, partem gives part, portum gives port, but portam gives porte.

The explanation proposed for the retention of a where all other vowels are lost is that a possessed certain heaviness or sonorous quality, which is, conversely, denied to the others.

This explanation has long enjoyed unquestioned acceptance. It simply postpones the ultimate solution of the problem; we must still ask how the a acquired or why it possessed this quality. Stability was not characteristic of a in Latin or Romance in unaccented syllables other than final. Even there, e and o, in the other languages, are as persistent as a. A Greek loanword like paxava (Attic unxan) becomes Latin machina; the compounds of cado and sacro, for example, with short a, have i : accido, or e: consecro, while the short o of rogo remains in irrogo, arrogo, and the long a, e, in invado, irrepo suggest that length may have something to do with stability in some cases, making allowances, of course, for the date of the new quantitative accentuation in Latin (cf. caedo: occido). Again, there is no tendency to distinguish a from the other

1 Rom. v, 140.

vowels in the syncopation of Vulgar Latin proparoxytones such as colaphum: col-po: French coup.2

It is pertinent to inquire at what period, or under what circumstances the Vulgar Latin of Gaul developed this characteristic pronunciation of a which was unknown to the language at the time these changes were taking place.

Scrutiny of the Old French words which retain Latin unaccented a in an internal syllable provokes some doubt as to whether the treatment of a here is really different from that of any other vowel similarly placed. The Old French forms in which a is retained may, in the cases that have been cited in support of Darmsteter's law, be due either to the fact that the loss of a would produce an awkward consonant group, as in sacramentum: OF sairement (i. e., sai-rə-ment): F serment, or else to the clear analogy of some other form of the word in question, in which the syllable containing a was accented or final. The words which are supposed to show retention of accented a in an internal syllable fall into several fairly distinct groups:

3

66

1. The class of sacramentum: OF sairement, *caccabella: OF c(h)achevelle, in which e representing Latin a is clearly a vowel of support" as in carrefour from quadrifurcum;

2. Nomina agentum in -tor-em, with which may be included feminines in -tura: canta-tor-em, furca—tura, aura-tura, OF chantedour, fourcheure, oreure, with e for Latin a.

Any one of several influences would satisfactorily explain the retention of a here, without the necessity of attributing extra heaviness to it: the nominative singular in -ator, where a was accented (OF -edre) might have affected the oblique cases with unaccented a; or again the consonant group might have been awkward, as -nt-t- in cantator-em, causing the retention of a "vowel of support"; or the influence of the simple (primitive) word might easily be sufficient to keep rarer derivative forms in line;

The two most recent studies in this direction: E. Philipon, L'a médial posttonique dans les langues romanes, Rom. XLVIII, 1-31, and E. Seifert, Die Proparoxytona im Galloromanischen, Beiheft 74 zur ZrPh 1923, fail to show any general or constant peculiarity in the development of the unaccented a in French in internal syllables.

This point has been made before. So A. Thomas, Rom. xxi, 7.

3. Words like orationem: oraison, oreison, oroison; venationem; ligationem, may have been strongly influenced, not only by their own nominative singulars and by simpler forms like orare, etc., but also by the sound of commoner words like raison, saison, maison.ʻ Whether or not we have sufficient evidence in support of a special treatment of unaccented a in the first part of the word, there is no question that the a in final syllables stands out clearly differentiated from all other vowels: it is consistently retained as "mute" e, the others are consistently dropped unless required for the support of a consonant group. Here, then, we have more material in which to study the origin of the peculiar quality evidently to be assigned to final syllable a in Old French.

It is clear that the disappearance of a vowel cannot come about suddenly, but must proceed by gradual weakenings, that is, shortenings, until no length is left and the vowel ceases to be pronounced at all. For various reasons the Germanic languages offer the best documentation of the process; but the same thing must have happened, and must still be happening, in any language which, like Germanic and Latin, has an expiratory (stress) accent. In primitive Germanic a post-tonic short vowel (one mora) loses one mora and disappears; a vowel originally long (two mora) becomes short (one mora), so that I.-E. long ō corresponds to P. Germ. short a; while a vowel originally hyper-long (three moræ, indicated in Greek by the circumflex accent) becomes simply long (two more), so that I.-E. hyper-long ō corresponds to P. Germ. long ō. Gothic preserves this state of affairs, while in younger dialects these remaining vowels lose length progressively, one mora at a time, until a practical disappearance of inflections results.

The disappearance of any vowel in Latin, at any period, must similarly have been preceded by one or more stages at which the vowel was progressively shortened, the last stage before complete disappearance no doubt having the "mute" e sound perhaps to be read in very early double spellings like op (i) tumus: optimus. As we should expect, long vowels resist change better: incido but invado, colligo but irrepo, surgere but surrexi.

The word allemand, Latin alamannus must be a special case in several respects; REW suggests Italian influence.

Cf. Rydberg, Geschichte des frz. a 1, § 13, page 23.