fytmll WLmvmxty ptarg 
 
 BOUGHT WITH THE INCOME 
 - FROM THE 
 
 SAGE ENDOWMENT FUND 
 
 THE GIFT OF 
 
 Hcmij W* Sage 
 
 1891 
 
 ^4..,...JL.&..0....7..7-7--. l£.MA.f... 
 
 35I3-J 
 
The original of this book is in 
 the Cornell University Library. 
 
 There are no known copyright restrictions in 
 the United States on the use of the text. 
 
 http://www.archive.org/details/cu31924027253131 
 
A CONTRIBUTION 
 
 TO THE HISTORY OF THE 
 
 Unaccented Vowels 
 in Old French 
 
 INAUGURAL DISSERTATION 
 
 PRESENTED TO THE PHILOSOPHICAL FACULTY OF THE 
 
 UNIVERSITY OF HEIDELBERG, FOR THE DEGREE 
 
 OF DOCTOR OF PHILOSOPHY 
 
 — BY— 
 
 WILLIAM PIERCE SHEPARD 
 1897 
 
 EASTON, PA., U. S. A. ; 
 
 CHEMICAL PUBLISHING COMPANY. 
 
 1897. 
 
A CONTRIBUTION TO THE HISTORY OF THE UNAC- 
 CENTED VOWELS IN OLD FRENCH. 
 
 INTRODUCTION. 
 
 In all languages possessing a well developed expiratory ac- 
 cent there is found a tendency to weaken the syllables which 
 stand on the lower stages of accentuation. The energy devoted 
 to the production of the syllable on which the principle accent 
 rests makes necessary a reduction in the force of the expiration 
 of the other syllables of the word. The vowels of these sylla- 
 bles then show a loss of sonority ; and are liable to be reduced 
 to that quality which demands the least amount of expiratory 
 force for their articulation. Only that part is left which is ab- 
 solutely necessary for the existence of the syllable. Or, in other 
 cases, the reduction may go still farther. Then the weaker 
 syllables disappear entirely ; the energy once expended on their 
 production goes to swell the stress given to the more highly ac- 
 cented syllables, and they lose their independent existence. To 
 observe the effect of these tendencies, we have only to compare 
 a language with a chromatic — or musical — accent with one pos- 
 sessing a strong expiratory stress. In the former, all the 
 vowels are articulated distinctly and generally preserved through 
 long periods of development ; in the latter, they are first reduced 
 in force, their articulation is slurred or hasty, and they often 
 disappear entirely. For example, in ancient Greek, which had 
 undoubtedly a tone- accent, there are almost no examples of the 
 syncope of unaccented vowels.. The only cases of loss are due 
 to the subsequent contraction of two vowels standing in hiatus 
 after the outfall of an interlocal j or w. But in the Teutonic 
 branch, on the other hand, the vowels of the unaccented sylla- 
 bles are constantly weakened ; from the earliest period this ten- 
 dency may be observed, and its operation is unchecked at the 
 present day. In modern English these vowels are continually 
 slurred in pronunciation, and are frequently weakened to the 
 so-called " irrational" vowel (the sound of u in but), which is 
 oftentimes a mere voice-glide, without decided articulation. 
 
 Latin, it is probable, possessed an expiratory accent, although 
 
the stress was not so great as in the other languages of the West 
 Indo-European branch. The exact nature of the Latin accent 
 is still a subject of discussion, but there is little doubt that it 
 was partially, if not entirely, expiratory, {cf Eindsay: The 
 Latin Language, Chap. Ill; Seelmann, Aussprache des Lateins, 
 p. 22, ff.) Throughout its entire history we find instances of the 
 syncope of the unaccented vowels. Examples of the syncope 
 under the earlier accent law are : 
 gavideo — gaudeo ; 
 Old Eat. aevitas — aelas, 
 Old Eat. primiceps — princeps, 
 *osinos, ornus, cf. Norse aski ; Sl.jasika. 
 
 The laws governing this vowel loss are for the most part ob- 
 scure, but it suffices for my purpose here to call attention to the 
 fact. On the other hand, the weakening of unaccented short 
 vowels, in syllables with slighter stress, according to the early 
 accent law, is much more extensive and regular. All are re- 
 duced to e, and in open syllables still further, to i. {cf. Eind- 
 say, p. 185, ff.) 
 
 Turning now to the Romance languages, we find there this 
 principle still active and extended much farther than in classical 
 Eatin. Already in Vulgar Eatin inscriptions, and from the tes- 
 timony of the grammarians, the existence of the shorter synco- 
 pated forms is manifoldly proven. (For a list of examples from 
 the inscriptions see Schuchardt : Vocalismus des Vulgarlateins, 
 II, p. 381, ff.) It is especially marked in the loss of the unac- 
 cented vowel of the penultimate syllable in proparoxytones, 
 some examples of which go back to the Vulgar Eatin of Plautus 
 (domnus). In Romance, this reduction is naturally more ex- 
 tended. 
 
 In Italian, for example, we find instances of the loss of a pro- 
 tonic vowel, especially in the neighborhood of a liquid, as in 
 cerebellum — cervello, verecundia — vergogna, etc., and the syncope 
 of the penult in proparoxytones is regular and thorough for a 
 large number of consonant groups, (cf. Meyer-Eiibke, Ital. 
 Gram., p. 169, ff.) But it is in French that this principle finds 
 its fullest expression. In that language all pretonic and post- 
 tonic vowels are affected by it, with the result that all fall with 
 the exception of a. This widely extended reduction has been 
 
traced to the influence of the primitive Keltic dialects which 
 were supplanted by the "lingua romana rustica. " (cf. Schu- 
 chardt: Ztschr. f. Rom. Phil., IV, 142 ff. ; Meyer- I^iibke : ibid. 
 VIII, 240; and Rom. Gram., I, 538.) Keltic had a very strong 
 stress accent, and this was tranferred by the Iyatin-speaking 
 Kelts to the latter language, and thereby the accented syllable 
 was strengthened at the cost of its neighbors. Therewith 
 stands also in undoubted connection the diphthongation of the 
 vowels under the tonic accent, a process much more extended 
 in French than in the other Romance languages. 
 
 For Diez and the earlier masters in Romance philology, the 
 development of the unaccented vowels was a division of the 
 science given over largely to the action of chance. Diez recog- 
 nized, to be sure, that the outfall of the penultimate vowel in 
 proparoxytones was the rule in French and Provencal, and 
 that the ultima, when retained, resumes the form of mute, or 
 feminine, e. Farther than this he did not go. 
 
 An unsuccessful attempt to formulate the laws governing the 
 protonic vowels was made by Brachet (Jahrb.f. Rom. u?id Engl. 
 Sprache u. Litt., VII, 301), who tried to establish the signifi- 
 cance of the l,atin quantity for the French development. Ac- 
 cording to him, the long vowels are retained, while the short 
 fall. This theory was completely refuted by Darmesteter, in his 
 classic article soon to be mentioned. 
 
 The first to demonstrate clearly and convincingly the laws for 
 the ultima was Zupitza ( Jahrb. f. Rom. u. Engl. Sprache u. Litt. , 
 XII., p. 180 ff.) His general results remain uncontested. He 
 showed that when the ultima vowel is a, this remains in French 
 under the form of mute e. All the other vowels fall, but after 
 certain groups of consonants difficult of pronunciation, a 
 " euphonic" e is developed. The principal groups indicated by 
 Zupitza are, — cons. + r; cons. + e ; .s — m, pt, and the sound 
 g=d; zh moreover in original proparoxytones, like comi- 
 tem conte ; hominem — ome. Worthy of note is Zupitza 's view 
 that the mute or "feminine" e in these cases is not the 
 representative of the I^atin vowel of the ultima in a weakened 
 state, but is a sound developed in French, and due to " euphonic" 
 considerations. He did not express himself in regard to the 
 chronology of the development. 
 
Concerning the last point, Stengel {Ztsckr. /. rom. Phil., I, 
 p. 1 06) endeavored to prove a difference in the date of the loss 
 of e and i in the ultima. He emphasized the difference between 
 such iorms as facit— fait, dicit—diton the one hand, and placet— 
 plaist, tacel—taist, on the other, and concluded that i disap- 
 peared in Gallic Vulgar Latin, earlier than e, the former before 
 and the latter after, the assibilation of the guttural. This view 
 has since been adopted by Schwan (Afranz. Gram, II., p. 70). 
 
 By far the most important contribution to the literature of the 
 subject was the essay by Darmesteter on the vowels of the pro- 
 tonic syllable (Phonitique francaise — La protonique non initiate 
 non en position. Romania, V, 140 ff.). Applying the rules 
 traced by Zupitza for the ultima, Darmesteter showed that, in 
 the development from Latin into French, the tonic accent may be 
 considered as dividing the word into two equal parts ; and that 
 the final vowel of the first half is subject to laws of the same 
 nature as that of the second half. For example, in a word like 
 bonitatem, with two syllables before the tonic accent, there is a 
 secondary accent on the initial syllable, while the second is un- 
 accented. The phonetic laws governing this vowel are : that, 
 if it is a, it is retained under the form of e "mute ; " if e, i, 0, u, 
 it falls, unless protected by a group of consonants which precede 
 or follow it. The principal groups are nt—gr, r—gr and the 
 " mouillees " consonants /, n, which always demand protonic e 
 before them. Furthermore, Darmesteter showed how the action 
 of the law is implicated by analogy ; firstly, in the futures of 
 verbs of the 4th conjugation, where the longer forms (parlirai, 
 etc.), are due to the influence of the infinitive ; secondly, in de- 
 rivatives (can be either substantives, adjectives, or verbs, like 
 amoros, felonie, coroner) , on which at all times the influence of 
 the simplex made itself felt ; thirdly, in words like sentiment, 
 faisebr, bateure, baie'is, etc., where the preservation of the pro- 
 tonic is due to a species of suffix-exchange, viz., for these suf- 
 fices, mentum, torem, turam, ticium, the stem of the first conju- 
 gation has been generalized, so that in Gallic Vulgar Latin the 
 suffix forms are amentum, atorem, etc., which are attached to 
 verbal roots of all four conjugations. Darmesteter leaves uncon- 
 sidered the words with the protonic in position before two con- 
 sonants, although he seems to believe that it is retained. Those 
 
Words, not numerous, with more than two protonic syllables are 
 not treated, although he assumes for them a secondary accent 
 on the second syllable before the tonic (aspiritatem) , indica- 
 ting a system of binary accentuation. I may also call attention 
 to Darmesteter's explanation of double forms like soverain-sov- 
 rain, beverage-bevrage, etc. The shorter forms are the primi- 
 tive and normal ; the longer are due to a tendency to ' ' adoucis- 
 sement," which is however not absolute. Darmesteter's work 
 has formed the basis of all later research. 
 
 The next most important contribution was that of Meyer- 
 Liibke, on the development of proparoxytones (DieBehandlung 
 tonloser Paenultima im Romanischen : Ztschr. f. Rom, Phil., 
 VIII, p. 205 ff.). His principal results, so far as French is 
 concerned, are the following : Syncope of the penult is older 
 than the reduction of the tenues to mediae (so-called " I^aut- 
 abstufung " ) , and also antedates the development of a toe in 
 open syllables, but is later than the action of the laws govern- 
 ing the final vowel. Nevertheless, each particular case must be 
 judged for itself. So, for instance, a has hindered syncope in 
 monachus : moine, *Jetacum : ftrie. Again the ending icum loses 
 the i of the penult later than tea, so we have nache, manche, on 
 the one hand, siege, age on the other. In the first, syncope 
 has taken place before the weakening of the tenues, in the lat- 
 ter, after this process. In general, according to Meyer-Liibke, 
 syncope in proparoxotones (apart from some very old cases 
 like caldus, frigdus, virdis) is a comparatively late process, 
 which set in after the dialectal separation of the Romance 
 languages. It is, however, older than the earliest literary 
 monuments, and mostly antedates the individual peculiarities 
 of the particular languages. In French it may be traced back 
 to the substitution of the strictly expiratory accent of the Kelts 
 for the chromatic expiratory accent of. Latin. 
 
 A later treatise on the proparoxytones in general is the short 
 
 work by Andersson ' ' Zum Schwund der nachtonigen Vocale im 
 
 franzbsischen " in Sprakvetenskapliga Sallskapets For handling ar . 
 
 Upsala, 1893. In general, his results are not different from 
 
 those of Meyer- Iytibke. 
 
 Another highly important contribution was made by Karsten 
 (Zur Geschichte der altfranzbsischen Consonantverbindungen. 
 
Freiburger Diss., 1884). In this work, entirely new objective 
 points of view were set up. The part played by the principles 
 of " Satz-phonetik " was for the first time insisted on, with the 
 result of unifying the treatment of many phenomena which had 
 hitherto seemed incongruous. Especially worthy of mention is 
 the explanation of the double forms of super (p. 42 ff.); accord- 
 ing to Karsten the shorter forms sor, sur, are explained by the 
 character of the word as a proclitic preposition. Another fertile 
 suggestion of Karsten's is that relating to the development of the 
 secondary groups icus, idus, which led him to a revision of the 
 old notion of "supporting groups" of consonants. He shows 
 that phonetically we cannot speak of supporting groups at all, 
 and that the preservation of the ultima after such groups stands 
 in no direct relation with them. Where long groups arise 
 by syncope, the language has other means at its disposal for 
 simplifying them. So the relation jbetween consonant group 
 and final vowel is not one of cause and effect, rather it is due to 
 a coincidence between the groups and the longer forms of the 
 sentence-doublet. 
 
 In connection with this work of Karsten, I may name the kin- 
 dred one of Kauffmann {Die Geschichte des consonantischen 
 Auslants im Franzosischen. Freiburger Diss., 1886), although 
 this does not enter so much into the pre-literary development of 
 the final vowels. The general theme of the work, that the ten- 
 dency in Old French was always in favor of final consonants, 
 not vowels, must always be taken into consideration in discuss- 
 ing the history of the last. 
 
 The last important contribution is that of A. Thomas {La 
 Loi de Darmesteter en Provencal ', Rom . XXI, 7 ff.), although it has 
 more direct importance for Provencal than for French. He 
 treats especially the Provencal forms of words with the suffixes 
 mentum, torem, ticium, etc. These show a retention of the pro- 
 tonic, which is contrary to the law of Darmesteter. There are in 
 Provencal three series of words with these endings, one in amen 
 ador, always associated with verbs of the first conjugation • one 
 in imen, idor, with verbs of the fourth conjugation ; and one in 
 emen, edor, with verbs of the second and third conjugations. If 
 Darmesteter' s explanation — levelling to the stem of the first con- 
 jugation — which he established for French alone, holds good 
 
here, we should expect amen, ador for all three series. Thomas 
 accordingly takes recourse in the old notion of the influence of 
 the simplex on the derivative ; these suffixes were constantly 
 associated in the minds of the people with the verbs to whose 
 stems they were attached, and they thus maintained a quasi- 
 independent existence. Accordingly, the protonic vowel is 
 retained as legitimately as the initial vowel is always retained in 
 independent words. Thomas hints at the application of this 
 theory to the French forms, but does not carry it out in detail. 
 
 The results of Meyer- I^ubke {Rom. Gram., I), and of Schwan 
 {Afranz. Gram., II, 55 ff.), will be referred to so constantly in 
 the course of the following discussion that there is no need to 
 summarize them here. 
 
 A recent work of great importance for comparison considers 
 the history of the unaccented vowels solely in names of places. 
 Lindstrom. — " Anmarkingar till de obetonade vocalernas bort- 
 fall i nagra norn-franska ortnamn." Upsala, 1892. His results 
 agree in the main with those of previous writers. As this con- 
 tribution will cite for the most part only appellatives, I/ind- 
 strbm's conclusions will be mentioned only in the case of striking 
 agreement or difference. In general it seems to me that place- 
 names should be used with more caution than common sub- 
 stantives in investigations of this kind, being greatly ex- 
 posed to the action of folk-etymology, as Ijndstrom himself 
 points out. In other cases, the mediaeval I^atinization of the 
 name probably represents only approximately the original Keltic 
 or Teutonic sounds. See also the reviews of this work by 
 Vising, Littbl.f. Germ. u. Rom. Phil., XIV, 288 ff. 
 
 I propose, in the discussion of the history of the unaccented 
 vowels in Old French, to consider, first, this development of the 
 ultima vowel in (original) paroxytones ; second, the treatment 
 of original proparoxytones ; third, the protonic vowels. 
 
 Before passing on to the consideration of the development of the 
 unaccented vowels in detail, I wish first to touch upon some 
 general principles which are of importance for an understanding 
 of this development. 
 
 First and foremost, we have always to bear in mind that the 
 laws for the unaccented vowels, as for all other sounds, apply 
 only to the genuinely popular words of the language, the " Frb- 
 
IO 
 
 worter," and not to the words which were borrowed from the- 
 classical Latin at different periods, the " Lehnworter. " Now 
 the treatment of the unaccented vowels is often of great import- 
 ance as a means of distinguishing these loan-words. In most 
 cases we have other criteria to help us. For instance, the word 
 castum-chaste betrays itself as a loan-word by the preservation of 
 the s not less than by the retention of the final vowel. So fragt- 
 lem-fraile compared with soticulum-soleil, shows a retention of the 
 ultima after the secondary group, guttural + /; this, com- 
 bined with the non-palatalization of the /, enables us at once to 
 pronounce it a loan-word. I shall have occasion many times in 
 the course of this discussion to call attention to the distinction 
 between inherited or popular words, and loan-words. 
 
 Furthermore, in abstracting a phonetic law from the develep- 
 ment exhibited by a certain group of words, we must look apart 
 from all those which are subject to the disturbing influence of 
 analogy. All forms which stand in a "system," which are as- 
 sociated in the mind in groups or categories, are liable to a pro- 
 cess of levelling, which often suspends the action of the phonetic 
 laws. So all verb-forms are to be regarded with more suspicion 
 in deducting such a law than substantives or adjectives, which 
 lead a more isolated existence. For instance, Willenberg in 
 his study of the present subjunctive of the first conjugation in 
 Old French (Historische Untersuchung iiberden Conjunctiv prae- 
 sentis der ersten schwachen Conjugation im Franzbsischen . Rom. 
 Stud., Ill, 373 ff.) proposes the law that the endings rm, m, de- 
 maud a "supporting" vowel, and cites the form retornes {Rose 
 2334) as an example. But an examination of the substantives 
 with the same ending, jorn, ivern, etc., contradicts the assumed 
 law, and shows that retornes is one of those subjunctive forms 
 which assumed early the analogical e. Considerations like to 
 this will always make their appearance. 
 
 Of great importance for the development of the unaccented 
 vowels is the principle of " Satzphonetik." The position of the 
 word in the sentence is of great weight in the treatment of the 
 sounds composing it. In the ordinary spoken language, a word 
 is not isolated, but stands in the most intimate connection with 
 its fellows in the sentence. The syllables composing a sentence 
 arrange themselves into certain groups, the divisions of which 
 
II 
 
 do not always correspond with the (written) words. These groups 
 consist of a greater or less number of syllables, bound together 
 by a unity of respiration. The syllables that are weaker in ex- 
 piratory force are subordinated to those with greater stress, and 
 group themselves about the latter. These " stressgroups" 
 ("Satztakte," see Sievers, Grundzuge der Phonetik, 4th ed., p. 
 214 ff.) form a higher unity in the sentence of which the sylla- 
 bles are the elements. Now it is evident that when a syllable, 
 which may be an entire word, stands in one of the subordinate 
 positions in a stressgroup, the sounds composing it will be treated 
 like the sounds of those syllables having a lesser stress, — that is, 
 like the unaccented syllables of words which form stressgroups 
 by themselves. In the stressgroups, too, the final or initial 
 sounds of one word are conditioned by the sounds of the words 
 which precede or follow them. In this way double forms arise, 
 due to the varying positions a word may assume in the stress- 
 group. A syllable on which the main accent of a group rests 
 will assume one form ; placed in a position where a lesser ex- 
 piratory force is given to it, it will assume another. The final 
 sound of a word varies according as the following word begins 
 with a consonant or a vowel. These principles are in active op- 
 eration at every stage of the language, and are of especial im- 
 portance in connection with the minor parts of speech — preposi- 
 tions, conjunctions, adverbs, etc. {cf. for this whole subject, 
 Neumann : Ztschr. f. Rom. Phil., VIII, 243 ff. ; Schwan : Ibid, 
 XII, 192 ff. and Karsten : op. cit.) 
 
 As said above, we can see this principle at work in all periods 
 of the history of the language. Even in the classical Latin traces 
 of its influence are not wanting. Examples of the differentiation 
 into ante-vocalic and ante-consonantal forms are seen in con- 
 junctions like atque, ac, neque, nee, etc. And it has been sug- 
 gested that the shorter syncopated forms often met with (lar(i)dus, 
 cal(i)dus, audac(i)ter, etc.) are due primarily to the same cause, 
 namely, the reduction of the unaccented syllables in more rapid 
 pronunciation in the stress-group {Schnellsprechformen : cf. Ost- 
 hoff: Wolflins Archiv, IV, 355 ff.). When the great wave of 
 syncope swept over the language later, in Gallic Vulgar Latin, 
 it is probable that this principle played a great part ; the results 
 are difficult to trace in all cases. Under the greater force of the 
 
12 
 
 expiratory accent given to the tonic syllable, the vowels in the 
 syllables with slighter stress were weakened or dropped. Now 
 the stress of these syllables would vary according to the position 
 of the word in the sentence. In this way double forms arose; after- 
 wards the reasons for their origin being forgotten, they would be 
 used promiscuously. Then the one generally supplanted the 
 other, through causes which are often difficult to determine ; 
 sometimes it was due probably to chance alone. The results of 
 these various processes lie before us in the earliest literary 
 French. 
 
 A word here as to the variations in stress, especially in pro- 
 paroxytones. Karsten (op. cit., p. 33) assumes as possible sen- 
 tence-doublets in the case of the suffix aticum the forms ateke, 
 atek, atke, atk. Now the second of these forms presupposes a 
 greater degree of accentuation on the penult than on the final 
 syllable. Our knowledge of the degrees of accentuation in Latin 
 is extremely limited; the position of the secondary accent even is 
 a matter of contention. But if we take into consideration the facts 
 that syncope of the penult is met with from a very early period, that 
 such syncopated forms are frequent in the inscriptions, while apo- 
 cope of the ultima in proparoxytones is very rare ( the only examples 
 arefamul, mascel, etc; for these see Brugmann : Indg. For- 
 schungen, IV, 220 ff . , who explains them by a ground-form, *fam- 
 los, with subsequent syncope and development of a " svarab- 
 bhaktic" vowel, *J 'amis, famul; famulus is due to analogy), and 
 that the few French words which show a loss of the final sylla- 
 ble may most probably be considered loan-words (cf. infra, p. 
 87). The accentuation dtclum would seem to be unknown to the 
 language at any period. And this is natural. The syllable 
 standing next to the chief accent suffers more from the tendency 
 to reduction than syllables more remote. The great stress given 
 to the tonic syllable reduces to a corresponding degree the energy 
 of expiration expended on the next syllable. The sequence of 
 accentuation, chief accent, secondary accent, unaccentuation, 
 -£- -^- — seems to be unknown in Latin or Romance. Such an 
 accentuation is not, however, impossible, as the relations in 
 Old High German show. There a short vowel in an open sylla- 
 ble is syncopated after the secondary accent immediately follow- 
 
13 
 
 ing the chief accent, as *mdhtiigro, mahtigro {cf. Paul in Paul 
 u. Braunes Beitrage, VI, 144 ff.). 
 
 A last question remains to be considered before proceeding to 
 the discussion of the "groups." What was the quality of the 
 ' ' indefinite ' ' vowel to which the others were reduced ? Meyer- 
 Iyiibke {Rom. Gr., I, 245), gives e "gutturalen Laut," as the 
 vowel peculiar to North French. This sign e represents the 
 sound of the modern Fr. "e femmin." According to Sweet 
 (Primer of phonetics, p. 86) , this is a front vowel, like b inpeur, 
 but partially unrounded. Schwan (Af. Gr., 56) represents this 
 sound as ' ' dumpf es " e ( " Stimmtonlaut " ) ; on the following 
 page he sets it as equal to b. Apart from the earliest literary 
 monuments, the Strassburg Oaths, and much more rarely, the 
 Chanson d' Alexis, this sound is always written e. In the Oaths 
 we find besides e {fazet ) , also a (cosa, fradra), and (poblo, 
 karlo) ; the last may be a I,atinism. I am inclined to believe 
 that this indefinite vowel had always a more anterior articula- 
 tion ; it was a " front" vowel. For the representatives of I,at. 
 e, i, a, this view presents no difficulty, but for u, o, a slight 
 shifting of articulation must be assumed {a was already a front 
 vowel in L,atin. cf. Seelmann, p. 169 ff.). Here the fact that 
 already in the classical language u, could be reduced to e or i, 
 is of importance. Compare for instance, novitas: Greek vsorrjg, 
 agimus: Greek ayo^isv, lacrima: Old Latin, dacruma, ilico, from 
 insloco, anno, but artniculum , socio, but societas, etc. This in- 
 dicates that the ' ' guttural ' ' vowels had a tendency to be shifted 
 farther front under a weak accent stage. The fact then that in 
 French e is always chosen as the sign of the weakened vowel, 
 and that this e has in Modern French a "palatal " quality, are 
 farther proofs. So that the assumption that the indefinite vowel 
 to which all the others were reduced was a front vowel seems to 
 be well supported. This view removes a difficulty in the devel- 
 opment of guttural consonants before u of the ultima, as in att- 
 aint age, *pedicum: piege, etc. {=adzhe,predzhe) . If the ^retained 
 its "back" quality, the shifting of the consonants from a 
 "back" mute to a " blade-point" spirant is difficult to account 
 for. We should expect adge (g = German g in sagen). 
 Before a front vowel e, however, the shifting of g to zh is natural 
 and is parallel to that of initial £- before a. 
 
H 
 PART I. 
 
 The final vowel in original paroxy tones . 
 Final a. 
 
 a is always retained under the form of e (=?). The only 
 exceptions are found in the Strassburg Oaths and the Alexius, 
 where a seems to be retained as such. It was probably only a 
 sign for e, as we find a also used to express this sound where it 
 is not etymologically justified, for example in fratrem : fradra. 
 {cf. Koschwitz. Commentar zu den altesten frambsischen 
 Sprachdenkmalern, p. 8). Examples of the retention of a in the 
 Oaths are dunat, aiudha, cadkuna, cosa. It is unnecessary to 
 give further examples of the rule. 
 
 The retention of the vowel a in distinction from the others, is 
 due undoubtedly to its greater sonority. A weakening process 
 which affected the other vowels to such an extent that they lost 
 all syllabic function, would still leave a perceptible voice for a. 
 So much more respiratory energy is necessary for its production 
 that it is the most capable of all of withstanding total reduction 
 through lessening of the stress. 
 
 Exceptions. — Most interesting are certain cases of sentence- 
 doublets which some adverbs show. So, for example, hachora: 
 ore (s) and or, and its compounds, encore, encor, aores, aor, in 
 which e from a has been absorbed by the voice of the r {adho- 
 ram), desores, desor ; oncore, oncor. 
 
 But we find only elores, (compare Modern French alors), 
 quore (Marie de France, cf. Godefroy.) qua kora. 
 
 Possibly also the doublets bor, bore, mar, mare, {bona hora, 
 mala hora?) belong here. Cf. Meyer- I^iibke, Rom. Gr., I, 
 522. The earliest example of the shorter form is found in the 
 Passion, L,. 366 and the I^eodegar I,. 5, both before vowels ; the 
 first purely French monoment in which they occur is the para- 
 phrase of the Canticles or 1,. 57, encor ~L,. 40. Meyer- Ltibke re- 
 jects the view that these are the ante-vocalic and the ante-conso- 
 nantical forms respectively, owing to the fact that already in 
 Alexius or is used before consonants. He explains the short 
 forms as the result of the frequent use of the word. But in the 
 Passion and Leodegar they are still rightly distinguished. 
 And nothing hinders the assumption that already at an early 
 
15 
 
 date a promiscuous use of the two originally distinct and charac- 
 teristic sentence-doublets had set in. 
 
 Another example of doublets is found in the adverbs, umquam: 
 onques, and one, numquam: nonques, and none. The latter 
 undoubtedly represent the ante-vocalic development. I^ater, 
 according to this model, a number of doublet forms for adverbs 
 were formed, so avec — aveques ,aluec — alueques , poruec — -porueques, 
 iluec — ilueques, done — donques. cf. Meyer- Lubke, Rom. Gr., II, 
 642. puis is probably from *postius, not post-ea. cf. Schuchard: 
 Ztschr.f. rom. Phil., XV., 240. But compare the form puisse-di. 
 (Chev. au cygne). 
 
 Besides the usual hors,fors, a few examples of fores are found. 
 (cf.Godefroy, Dictionnaire, one ex. from the Chans, de Renaud.) 
 This is undoubtedly from foras (cf. Prov. foras) ; the usual 
 forms fors, hors, may then be early generalized forms, in which 
 e from a has been absorbed by the voice of the r instead of from 
 foris , as is generally accepted. 
 
 Further doublets are mica: mic, and mi (Raoul de Cambrai), 
 the word was used as a negative particle; interea, therefrom 
 dementieres — dementiers (Marie de France cf. Godefroy); ende — 
 mentieres-endementiers . (Renart.) ilia: elle, and el (cf. Marie 
 de France, Fraisne, 36). Possibly el arises by a process simi- 
 lar to that which causes the loss of the final syllable in image, 
 siege, etc. French has an inherent antipathy against propar- 
 oxy tonic stress-groups. So in combinations like dtt-elle fit-elle 
 the paroxytonic accentuation was restored by the fall of the 
 final syllable. Compare the accent-shifting in donni-je, etc., 
 which is due to the same cause. I owe this suggestion to Prof. 
 Neumann. 
 
 Another group of words showing double forms, with apparent 
 loss of final a in one, are certain feminine substantives of the third 
 declension, in -tatem, which are more or less irregular in other 
 respects; so civitas-cit (never cite) obi citi, paupertas: poverte and 
 povert, potestas: poeste, and poest, tempestas: tempeste and tempest. 
 cf . also jovente, jovent, from juventus-tas . 
 
 Of these the first, cit, may be explained by its use as a pro- 
 clitic before names of cities. It represents the shorter doublet 
 with loss of final e from a. The others are rare and probably 
 due to analogy. When in place of the original paupertas-tatem, 
 
i6 
 
 a *pauperta of the first declension had been formed (see Meyer- 
 Liibke, Rom. Gram. II, p. 22) according to the analogy of 
 juventa, juventas, it was not difficult to go a step further and 
 create a paupertus on the model of juvenilis. 
 
 Most of the other cases of the apparent loss of final a are due 
 to a change or confusion in gender. I indicate only a few of 
 them. This kind of analogical change occurs, as is natural, 
 most frequently with names of animals. Thus, besides the old 
 Fr. daime<C_dama, we find a masc. form dain, which has alone 
 been preserved in mod. Fr. So also formica: formie and formi, 
 lusciniola: rossignol. 
 
 Another quite frequent case occurs with names of plants and 
 trees. The latter have all become masculine in gender, and 
 with this change was sometimes associated one in form ; thus 
 alba spina: albespin, (Simon de Pouille) but mod. Fr. aubepine. 
 The form is remarkable because of the perservation of the pro- 
 tonic a in the adjective. The simplex espin is also found (lyes 
 I/oherains) . 
 
 Betula: bole, and bol (cited by God. from a chart of 1215). 
 (Probably a loan word ; *beille would be the regular form.) 
 
 Cliva, olive, and olif (I^es IyOherains) . 
 
 Many of these examples are probably new formations in 
 French, and do not go back to Vul. ~L,a£. ground forms, spinus, 
 etc, as albespin shows. 
 
 The same change of gender, carrying with it a change of 
 form, may be noticed in other words. — Ex. Balaena : baleine, 
 and balein, furca: forche, and also fore (Macaire) from *furcum. 
 costa, coste, and cost (Voyage de Charl., 211). Medulla, moelle, 
 and mo'el. 
 
 Plica, pli, and ex-aura, essor are probably post-verbal forma- 
 tions from the verbs plier and essorer. 
 
 Sagina, sain, has been explained through a suffca change in 
 Vulg. Iyat ; imen took the place of ma. See Cohn Snffixvand- 
 lungen, p. 57. 
 
 l,ike:wise pastindcem, pasnai for pastinaca (cf. Cohn p. 202) 
 The vowels z, e, 0, u in the ultima. 
 
 Zupitza's law, that these vowels fall unless supported by 
 "euphonic" considerations, viz., after a group of consonants 
 difficult to pronounce, undoubtedly corresponds to the external 
 
17 
 
 facts of the case. But does it explain equally the underlying 
 principles ? In the first place, the explanation that the devel- 
 opment of the final vowel is due to the desire for " euphony" 
 must be rejected, so far as it assumes a conscious striving after 
 harmonious effects of sounds. The phonetic laws always work 
 without the consciousness of the speech community in which 
 they act ; they are due wholly to physiological causes acting 
 unknown to the will of the individual speaker. 
 
 There is, however, a general tendency at work in all lan- 
 guages to simplify the longer consonant combinations that arise. 
 The assimilatory processes that are so marked in the develop- 
 ment of Romance are one expression of this. What then ex- 
 plains the partial loss and partial preservation of the final vowel ? 
 
 In connection with this subject the principles that govern the 
 division of syllables should be considered. A single expiratory 
 syllable can last only from the implosion of the mute which pre- 
 cedes the sonant of the syllable to the formation of the stop of 
 the next following mute (cf. Sievers, Phonetik, p. 186). With- 
 in these limits the sounds are arranged according to the degree 
 of sonority. The most sonorous sound is the real syllabic, the 
 others are non-syllabics. The sounds are classified according 
 to sonority as follows : vowels, vowel-like consonants (of these 
 the liquids are more sonorous than the nasals) , voiced spirants, 
 unvoiced spirants, voiced mutes and unvoiced mutes. Therefore, 
 at the end of a syllable we can have the succession vowel, liquid, 
 mute, or vowel, spirant, mute, but not the reverse. A combina- 
 tion of two mutes at the end of a syllable is physiologically impos- 
 sible. If, however, we do find written groups like apt, akt, etc., 
 we ignore in such the existence of a slight secondary syllable, or 
 voice-glide (" nebensilbe") formed between the two explosives. 
 These voice-glides must be formed after every such combination 
 of stops. 
 
 In Latin, if we may accept the testimony of the grammarians, 
 such groups were not uncommon at the beginning of a syllable, 
 (cf. Lindsay, p. 125 ; Seelmann, p. 139.) They divided in this 
 way no-ster, prb-pter, a-gmen, no-dem. How much these 
 grammarians were influenced by the theories of Greek phonet- 
 icians it is impossible to say ; it is probable, however, that they 
 extended this system of syllable- division farther than the facts 
 
i8 
 
 warranted, (cf. Kaufmann, op cit., p. 15; he divides ag-men, 
 prop-ter noc-tem). At any rate, when under the influence of 
 the strong expiratory tonic accent, the final syllables began to be 
 weakened, these groups, which had hitherto stood at the begin- 
 ning of the final syllable, would come to stand at the end of the 
 preceding tonic syllable. In this position, so long as each con- 
 sonant maintained its full value, a slight " nebensilbe" would 
 always be necessary for pronunciation. The sonant of this 
 ' ' nebensilbe' ' would be the original final vowel in a reduced 
 state. But the tendency of the French language was irrevoca- 
 bly toward consonantal finals ; and with this was associated an- 
 other, equally strong, toward the reduction by assimilation of 
 long consonant groups. When the assimilation had been com- 
 pleted, and the double consonant had been reduced, so that 
 there was but a single stop after the sonant of the tonic syllable, 
 then the weakening process would reassert itself and the 
 " nebensilbe" would be lost. The expiratory force hitherto ex- 
 pended on its production, would go to swell the stress of the tonic 
 syllable. So I would assume the development stages of a word 
 like nodem (accepting Thomsen's theory of the palatalization of 
 the consonants) noc-tem, noc-te, noc-tee, (e here represents the 
 voiced element following the explosion of the second stop,) 
 not'-te, noit-e, nuit. 
 
 Somewhat different would be the development of a final sylla- 
 ble containing a liquid or nasal. Take, for example, pa-trem. 
 The development of the accented vowel shows that the combi. 
 nation tr was never regarded as closing the syllable. This is 
 undoubtedly due to the peculiar nature of these sounds, the 
 liquids and nasals. They are capable of functioning either as 
 vowels or consonants (syllabicsor non-syllabics, more properly); 
 and this distinction is a purely relative one, depending on the 
 nature of the surrounding sounds. When now the final sylla- 
 ble in pa-trem was weakened, the tendency would be toward the 
 reduction of the syllable to that element of it demanding the 
 least expiratory force. The sonority, and consequently the 
 stress energy is much less in the case of the vowel-like conso- 
 nants than in that of the real vowels. Accordingly, the final 
 syllable would be reduced to the mute + the liquid, serving as 
 syllabic A?- dr. Or under certain conditions where the stress was 
 
19 
 
 greater, a " svarabhaktic" vowel would be developed from the 
 voice of the r, pe- dre. That this represents the real nature of 
 the development, and that the final vowel was not retained as 
 such, but that the e oipedre, vendre, etc., was only a sign to ex- 
 press syllabic r is, I think, proved by the development of such 
 words as major: maire, minor: moi?idre,grandior: graindre, pejor: 
 pire, etc. Schwan (Afr. Gr., 57), on the contrary, assumes 
 metathesis of the r for these cases (on the ground of Ital. sempre, 
 quaitro) . But the laws governing the final vowels and conso- 
 nants are quite different in Italian. There is no weakening of 
 final syllables there as in French. In my opinion, the loss of 
 the final vowel in these cases is evident ; the spelling re can 
 only express vocalic r. 
 
 A real exception to the rule that final e, i, 0, u fall, is fur- 
 nished by those words — not numerous — where these vowels 
 stand in hiatus after the tonic. In these the contraction of the 
 two vowels to a diphthong took place early, probably long be- 
 fore the action of the law for the finals. As a result of this fus- 
 ing, the originally final vowel is naturally retained as part of 
 the diphthong, (cf. Meyer- L,iibke: Rom. Gr., I, 255.) Examples. 
 — Final e, air, air ; final i, fui: fid, mei: *miei, mi, tui: toi, sui: 
 soi, *dui: dui, cui: ciii, *illui: illui, etc. 
 
 Final*?, ego, ( with early loss of g) , eo, eo,*stao (for sto. cf. 
 Meyer- I/iibke : Rom. Gr., II, 258), *esto: estois, duos: dous, 
 deus. 
 
 Final u, deum: dieu, pium: piu, judaeum; judieu. Hebraeum: 
 Ebrieu, Matthaeum: Matieu. 
 
 The apparent exception in meum: mieu and the other personal 
 pronouns is due to their character as proclitics. 
 
 All other examples of the retention of final e, i, 0, u, after a 
 simple consonant, must be considered either as loan-words, or as 
 due to some analogical influence, apart from a single category 
 not very extensive, where the effect of the position of the words 
 in the sentence makes itself felt. Most of these show double 
 forms. They have been discussed in extenso by Neumann : 
 (Ztschr.f. Rom. Phil., VIII, 382 ff.), and there is no need here 
 to recapitulate his conclusions. 
 
 Examples are focum: fou,jocum:jou, locum: lou, fagum: fou, 
 *traucum: trou, paucum: pou, caecum: ciu , graecum: griu , clavum: 
 
20 
 
 clou, *blavum: blou, *sclavum: esclou ; adjectives in -z'vum, iu 
 (only in the Eastern dialects). Andegavum: Anjou, Pictavum: 
 Poitou, etc. 
 
 All other cases are either loan-words, or new formations. 
 IyOan-words for example, rams: rare, avarus: avare, (O. Fr. 
 aver.) ; ritus: rite, vasem: vase, vivacem: vivace, cf. *vivacius: 
 viatz, canus: O. Fr. chanes " cheveux blancs," molem: mole , in- 
 famies: infame, etc. 
 
 Other cases may be explained through the exchange of suf- 
 fixes, thus, junicem: genisse, fomacem: fornaise, icia, atia for 
 icem, acem, mordacem: mordache, is an Italian loan-word. cf. 
 Cohn : Suffixwandl, p. 295. *abelamen, aveline. 
 
 Other instances may be ascribed to the analogical formation 
 of new feminine forms in Vulgar Latin or in French itself. 
 Examples. — * excarduus: escharde, Old French also eschard. 
 
 Gruem: grue, from a V. L,. *grua, faginus: fa'ine, intybus: en- 
 dive, vocalis: voiiel, but mod. Fr. voyelle. Ramum: rain and 
 raime. See Meyer-Liibke: Rom. Gr., II, 434. 
 
 Very frequently the two forms, masculine and feminine, ex- 
 ist side by side in Old French. The majority are original neu- 
 ters. Such are allium: ail and aille (cited by God. from Th. de 
 Turbeville) augurium: eur and eure (Berte as grans pieds), capi- 
 tium: chevez, and ckevece ; *bestialium: be stall and bestaille ; bra- 
 chium: braz, and brace; *farsum; fars, and farse; floccum: floe 
 and floche , : fustem: fust, and fuste; *capellum: javil and javelle ; 
 justitium: jtistis and justice, a learned word. 
 
 Meretricem: meretris and mereirice, probably a loan-word. 
 
 Nardum: nard and narde ; orbum: orb, and orbe ; pomum: pun 
 (Floire et Blaudreflor) and pomme, pastum: past and paste , 
 *roccum: roc and roche ; servitium: servis and service ; *sonium: 
 soin and soigne. Perceval, Solium: sueil and sueille ; super- 
 cilium: sorcil and sorcille ; oleum: uil and uille ; ostium: uis 
 and uisse, etc., etc. The final vowel after a group of consonants. 
 
 Sonant + consonant. This combination, standing at the close 
 of a syllable, is perfectly pronounceable. There is no break in 
 the expiratory current in passing from the vowel-like consonant 
 to the spirant or mute, and the synthesis is effected without a 
 vowel-glide. So there is no need of a " Nebensilbe " after the 
 explosion of the mute, and the final syllable is weakened to 
 
21 
 
 nul, with loss of its vowel. All exceptions must be accounted 
 loan-words. 
 
 n -f- mute. This was a group already very frequent in Eatin. 
 Its French development offers no difficulty. Examples : ab, 
 ante: avant, montem: mont, centum: cent, tantum: tant, de, unde: 
 dont, quando: quant, graudem: grant, rotundus: reont; excep- 
 tions : carpentum: charpente, is probably a post verbal deriv- 
 ative of charpenter, sequente, soventre, which show a widening 
 with re, also found in Rhaeto-romance and Provencal ; it is 
 probably due to the analogy of other adverbs in re, like sempre, 
 dementre, etc. cf. Meyer- Eiibke: Rom. Gr., II, 643. hirundo 
 aronde, probably from a Vulgar ~L,aXm*hirunda, cf. Prov. aronda. 
 
 Eoan- words are mundum: monde, Old French also mont, 
 glandem: glande, lendem: lende. 
 
 n + s . Here the nasal had already fallen in Vulgar Latin . 
 Examples: *defensum: defois, mensem: mois , francensem: francois , 
 trans: tres. 
 
 All exceptions are loan-words, like penso: pense, densum, 
 dense, immensum: immense, etc. 
 
 n -f- guttural. In this group the nasal had already in Eatin a 
 "velar" or guttural character {cf. Eindsay: op. tit. p. 65). 
 The guttural character of the nasal was probably preserved in 
 French till its final assimilation with nasalization of the preced- 
 ing vowel. When c, g before e, i were assibilated, a similar 
 change probably occurred in the nasal. The regular loss of the 
 final indicates an always homorganic group. 
 
 Examples, —juncum : jonc, sanguem: sam mane, cum: mane, 
 *flancum : flanc, longum : long, etc. 
 
 For n +pal. c, g-\- e, i, apart from these cases where the c, i 
 is in hiatus. I find only the following: instances : 
 
 Longe : loin ; and verb forms, plangit : plaint, jungit : joint, 
 frangit : fraint ; ungit : oint. In these n was assimilated to the 
 palatal g after the shifting of the latter, and both were palatal- 
 ized, after which the stop of the g was loosed and palatal n alone 
 remained. Probably no voice-glide was ever necessary. For n 
 + palatal c there are found only the following examples, both of 
 which are liable to suspicion, bilancem : balance ; lyncem : I'once. 
 The Vulg. Eatin ground forms were probably *bilancia, *lyncia; 
 cf.ltal.bilanciajoma; Prov. balansa (see Schwan : Af.Gr.,-p.tf). 
 
The Latin group nf is not found in final syllables. 
 
 m + consonant. The only Latin combination is m + labial. 
 Here, as in the case of nt, the synthesis of the labial nasal with 
 the homorganic stop is completed without any break of contin- 
 uity, and the final vowel falls. Examples, umbo : ams, rhom- 
 bum : romb, campum : champ, tempus : tens, plumbum : plon, 
 ricmpit : ront. 
 
 Loan words are : bombus : bombe, limbus : limbe. 
 
 I -j- consonant. Here likewise the ultima is reduced to nul, 
 synthesis bring too close to admit the formation of a " Neben- 
 silbe" after the explosive. Examples. /+ dental, altum : 
 halt ; multum : molt, pulsus : pols, *malaldum : marald ; saltum : 
 salt. Exceptions, celsum : cese (see Scheler : Lexique,^. 77). 
 *salsum: sake (feminine) : voltum : volte (feminine). The ety- 
 mology of the first word is doubtful ; if correct, it must repre- 
 sent a non-popular development. 
 
 / + guttural. This group should be especially noted, for 
 comparison with the same as a secondary group in proparoxy- 
 tones. Here the final vowel is always lost. 
 
 calcem : ehalz ; falcem : falz. Examples oil -\- c, g before 0, 
 u are not numerous. It occurs mostly in words of Teutonic 
 origin, like fulco : folc; balco : bale, falco : falc is Latin. 
 
 Exceptions, remulcum : remorque ; valgus : vouge (cf. Hoff- 
 mann : Rom. Forsch, II, 360) ; both are probably new feminine 
 formations. 
 
 / -\- labial. salvum : sal/ ; solvit : solt. Exceptions, album: 
 albe ; calvum : chalve. These two adjectives, with apparent 
 preservation of the final vowel, are probably original feminine 
 forms which have been generalized, a process not uncommon. 
 cf. riche, sage, O. Fr. roit but N. Fr. raide ; O. Fr./ers but N. 
 Vx./erme, etc. Loan words, such as balbns : balbe, bulbus: bulbe; 
 ko'Xtios : golfe : gouffre. 
 
 r -f consonant. A very frequent combination, and one that 
 offers no difficulty as regards the weakening and loss of the final 
 vowel. 
 
 Examples, r -\- Dental, artem : art, hortuvi : ort, fortem : 
 fort, mortem : mort, surdum : sord, tardum : tard, etc. r -\- s. 
 In the following, the r has already been assimilated in Vulgar 
 Lat. sursum: susum: sus, deorsum: deosum: jbs or jus, extrorsus- 
 
23 
 
 estros. Farther, cursus : cors ; ursus: ors ; sparsum : e spars ; 
 morsus : mors. Exceptions, burdo : bourde (also bort, regular) 
 is probably of learned origin. O. Fr. certes : beside cert is 
 probably derived from certas. 
 
 r -f- guttural. *quadri-furcum: carrefour ; arcum : arc ; por- 
 cum : pore. 
 
 largum : large is a generalized feminine. O. Fr. cerche, (Ar- 
 tur, etc.) " cercle, rond" probably represents *zlrca for circum. 
 
 r -f- Eabial. servum: serf, airbus: corb (but N. French courbe, 
 fern.) ervum : ers, corpus: cors, etc. Exceptions, morbus : 
 morve. The etymology is doubtful (see Schuchardt : Ztschr.f. 
 rom. Phil., XI, 494. The word is feminine in 0. Fr.) Eoan 
 words, serpens', serpe ; verbum: verbe. 
 
 Spirant + mute. The combinations, si, sp, sc are the only 
 ones found in Eatin. They could begin, as well as end, a syl- 
 lable ; and the grammarians always divide no-ster, i-ste, etc. 
 As there is no stop in these groups before that of the final explo- 
 sive, there would be no break of continuity when under the Fr. 
 accentuation they came to stand at the end of the syllable ; so 
 the final vowel falls regularly. 
 
 Examples : s-\- t, iste : est, christus : critz, hostem : ost, tostum : 
 tost, etc. 
 
 Willenberg (/. c, p. 279) is certainly wrong when he cites the 
 conjunctive forms, aprestes, ostes, etc., as proof that this group 
 demands a "supporting" vowel. They must be early examples 
 of the analogical e. Eikewise he cites laisses, oses, where the e 
 occurs after ss «x) and s, before the .r of the personal ending. 
 cf. the substantives, bois, nes. 
 
 Exceptions. — All loan-words, as the preservation of the s in 
 N. Fr., shows castum : chaste, robustum : robuste, tristem: triste, 
 vastum : guaste, regestum : regeste : regestre. 
 
 s -(- guttural. The same result so far as the final vowel is 
 concerned, pasco: pais, discum: dois,fascem: fais, cresco : crois, 
 *friscum: frois, N. ~B.,fraiche, a generalized feminine, liiscum : 
 lois ; N. Fr., louche, *lascum : lasche, had already generalized 
 the feminine in O. F. s-\-p. *crispu : cresp, N. Fr. crdpe, diaspre 
 from *jaspis : idem is a loan word. 
 
 Mute + mute. The only Eatin combinations are ct, pt, and 
 very rarely gd, bd. 
 
24 
 
 Guttural + dental. I have already discussed fully the history 
 of this group. The final vowel is never retained after it. Exam- 
 ples are very numerous. Factum: fait, ledum: lit ! , fructus: fruitz ', 
 noctem:nuit, etc. Occasionally without the production of parasitic 
 2, in loan-words, but still with loss of the final vowel; so contractus : 
 contrat, fluctum : flot, subjectus : sujet, ructus : rot. 
 
 With this I may treat the group net. Like ct, this shows 
 early assimilation, net : net': n't': int. The assimilatory pro- 
 cess was probably completed before the last weakening of the 
 final syllable. Examples, etnetum : ceint, *ftnctum : feint, sanc- 
 tum : saint, unctum : oint, etc. 
 
 A remarkable exception for the group ct is enoites, " cette 
 nuit," beside enuit (St. Bernard). Connection with noctem can- 
 not be doubted. Has the syllable es been attached according to 
 the analogy of other adverbs of time like ores, onques ? 
 
 The group gd seems to occur only in smaragdus : esmeraude, 
 a word that is not entirely popular. See Gutheim. Ueber Con- 
 sonant-Assimilation im Franzbsischen, p. 22. The preservation of 
 the final vowel indicates that the assimilation gd > ud had not 
 been completed when the law for the ultima went into effect. 
 This is in keeping with its character as a loan-word. 
 Labial -\- Dental. 
 
 This primary group should be especially remarked, for com- 
 parison with the corresponding secondary combination. In the 
 majority of examples, the final vowel is not retained, thus : 
 
 corruptus : corrots, deceptus : deceit (cited by God. from P. de 
 Langtoft) . subtus : sotz, septem : set, ruptum : rout, cruptus 
 (for ciypta): crot (cited by God. from Mace 'dela chariti). 
 Exceptions, aptus : ale, and its compound *ad-aptus : aate. 
 This adjective has been already in the earliest Fr. levelled in 
 favor of the feminine form. 
 
 I need not enter into the manner of the assimilation of this 
 group {cf. Karsten, op. cit., p. 17; Gutheim, op. cit., p. 11). It 
 was a comparatively early process, and is found in all the 
 Romance languages, a fact which at once shows that it was 
 earlier than the loss of the ultima in French. Until the assimi- 
 lation was completed, the final syllable would be retained at 
 least as a vowel-glide following the explosion of the t. This as- 
 
25 
 
 similatory process was completed early in Vulgar Latin; the date 
 is fixed by Schuchardt between ioo and 300 A. D. ( Vocalismus, 
 I, 104.) Examples from the inscriptions are ottimo, setemb. 
 (Soluthurn, 219 A. D.) 
 
 The group mpt, like the corresponding combination net, show 
 no retention of the final syllable. The assimilation and outfall 
 of the middle consonant must have been completed before the 
 final weakening of the ultima ; thus promptus : pront, *rumplus: 
 ront. 
 
 Mute -\- Spirant. 
 
 The group cs (x) always began a syllable, as the spelling of 
 the inscriptions, A-XIS, etc. shows. In French an assimilation of 
 the same nature as for ct has occurred. Both the consonants 
 were palatalized ; then the c was assimilated to the s and a para- 
 sitic i was developed. Inscriptions with s for x occur very early, 
 before 100 A. D., according to Schuchardt (Vocalis, I, 104. 
 cf. also I/indsay, op, tit., 108). There is no doubt that the 
 process was completed early in French ; and as a result the 
 final vowel falls as it does after a single consonant. Exam- 
 ples: axem : ais, buxus: bois, sex: sis, traxi ; ttais, etc. 
 
 For the group est, the only example seems to be sextus : sistes ; 
 also bis-sextus : bissHre. The last is obviously borrowed. For 
 sistes we should expect *sists : *sitz. The assimilation of c be- 
 fore s must have been at least as early as for intervocal cs ; and 
 the group st would remain, after which the final vowel falls reg- 
 ularly. Possibly sistes is explained by a levelling to the dissyl- 
 labic form of the other ordinals, like disme, nuefme, setme. (cf. 
 the relation between faimes and faites, dimes and dites, etc.) 
 
 The group ncs, like net, is assimilated to n's', after which the 
 
 ultima falls as it does after original ns. Examples only in the 
 
 perfects of verbs in ngere, cinxi : ceins, unxi : oins, ftnxi : feins, 
 
 tinxi: teins, etc. 
 
 Labial -j- s. 
 
 I/ike pt, this group shows assimilation in all Romance lan- 
 guages except Roumanian; isse for ipse, was already a feature of 
 the colloquial pronunciation of Latin, and examples of « for ps 
 are found on the inscriptions. (See Schuchardt, Vocalismus, I, 
 148.) The final vowel is never retained after it in French. 
 Examples are not numerous. *volvsi: vols, scripsi: esc? is, 
 
26 
 
 *solv-si: sols, *excarpsum (for exeat ptum): eschars, and com- 
 pounds of ipse, like nec-ipse : nets; ad id ipsum : ades inter hoc 
 ipsum: entrues, etc., gypsum (Gk. yvipov) ■ O. F. gif gipse is 
 a loan-word. 
 
 For the group mps, which is treated, so far as the final vowel 
 is concerned, exactly like ncs, the sole example seems to be 
 *readempsi : raens. 
 
 Mute-\-r. 
 
 I have already considered to some extent the history of these 
 groups. The apparent preservation of the ultima was due to the 
 vowel-like character of the liquid ; and it seems to me probable 
 that the orthograph re in O. Fr. was but a sign for sonant r. 
 The fact that after the final assimilation of the dental mute 
 (which occurred in the eleventh century, long after the loss of 
 the original final vowels) final e is retained, seems to contradict 
 this theory. But this might happen through the preservation of 
 the dissyllabic character of the word, the peculiar period for the 
 reduction of unaccented syllables being past. Then from the 
 voice of the r an indefinite vowel sound could be easily devel- 
 oped. This process occurs frequently with protonic r and is 
 common in all languages. It took place in the preliterary period 
 in I/atin itself, in the final syllables of ri and ro stems. Com- 
 pare, for instance, Gk. a^pos : Sanskr. ajras with Lat. ager: to 
 explain the latter, the following development has been assumed 
 — *agros\ *agrs: *agr : ager, with a "svarathaktic" vowel de- 
 veloped from the voice of the r. The development of Fr. pere was 
 probably parallel to this ; pa-trem : pe-dr : pe-dre : pe-re: (or pos- 
 sibly pe- dr: pe-rr: pe-re) . The question is not of very much 
 importance, because at every stage of the language an indefinite 
 vowel could be developed from the voice of the r under the in- 
 fluence of slightly greater stress. 
 
 Of value as proving the theory presented above are the words 
 with original final r. A great number of these is preserved in 
 French. 
 
 super : sore, marmor : marbre, inter : etttre, minor : moindre 
 *melor : mieldre, major : maire, calor : chalre, semper : sempre 
 turtur : tortre, robur : rouvrejulgor : foldre, pgjor : pire, and all 
 nominatives in or of the third declension, like antecessor: ances- 
 tre, *custor: costre, *tropator: trouvere, etc. 
 
27 
 
 The explanation of Schwan, that all these underwent early 
 metathesis of r, as in Ital., quattro, sempre, will hardly suffice. 
 Metathesis is a process which is essentially sporadic, and has 
 not the fixed and constant character of a phonetic law, which 
 the development of these forms indicates. It rests essentially on 
 mispronunciation and generally affects but a few words, {cf. 
 Paul. Principien der Sprachgeschichte, II, 59.) The view that 
 this development is due to a real loss of the original final vowel, 
 with the r then assuming its place as the syllabic, seems at once 
 simpler and more in accord with the general tendencies of the 
 language. 
 
 Examples for dental + r are very numerous. I give only a 
 few: aratrum : arere, Petrum : Pierre, fratrem : frere;iter : oire, 
 etc. After another consonant, inter: entre; capistrum : cheves- 
 tre ; dextrum : destre, feretrum : fertre, *pinctor: peintre, quattuor: 
 catre, iurtur : tortre. For dr, quadrum : querre. 
 
 Interesting is one» exception, due to the action of " Satz- pho- 
 netic;" retro shows double forms, riere : rier, (Ogier); also its 
 compounds arriere : arrier, (Jean Bodel) ; dederriere : dederrier, 
 (Artur). The shorter form is undoubtedly developed in the 
 position before vowels ; cf. the doublets or : ores, and super: 
 sore : sor. 
 
 Godefroy also gives one example of petrum; pierre, usually, 
 but pier " monnaie a 1 'dffigie de St. Pierre" from a chart of 
 Liege, 1249. This is so exceptional that it must be ascribed to 
 the neglect of the scribe. 
 
 Further, the ■ pronouns nostros ; nos, vostros; vos represent the 
 shorter protonic development of sir in proclisis. They have suf- 
 fered very heavy reduction. 
 
 Guttural + r. cr is not a common group in Latin in final syl- 
 lables. Examples are acrem: aigre, aldcrem; aliegre, macrem: 
 maigre. All seem to be loan-words. We should expect com- 
 plete assimilation of c, with production of parasitic i, as in pro- 
 paroxytones like facere : /aire. The development of the final 
 syllable is regular, lucrum: lucre, lavacrem: lavacre, etc., are 
 entirely of learned origin. O. Fr. gieres : giers: gier, has been 
 variously explained, a.s irova. de hac re(Suchier, Ztschr.,/. r.p. I, 
 42,i)orixom.igitur, (Cornu in Rom., X, 399). Its etymology is 
 so doubtful, that it can hardly be made the subject of any dis- 
 
28 
 
 cussion. It shows, however, the double forms already frequently 
 noticed with adverbs. 
 
 gr. First, examples — agrum: aire; nigrum: noir, *flagrum: 
 flair, integrum : entir. Schwan, going out from these exam- 
 ples, proposes the law that primary gr demands no "Stiitz- 
 vocal." But noit and entir are adjectives, and may be explained 
 as newly formed masculines, due to the analogy of adjectives 
 with a separate form for masculine and feminine, flair is un- 
 doubtedly a past-verbal derivative of flairer {cf. Meyer-I^ubke, 
 Rom. Gr., II, 445). There remains the substantive aire, which 
 has always final e; and is, moreover, of masculine gender. As 
 there can be no ground for assuming analogical influence in this 
 case, d/'remay be accepted as exhibiting the normal develop- 
 ment of gr, which then harmonizes with the other groups with 
 mute-p-?"- 
 
 aegrem : heingre, podagrum : pouagre, are loan-words. 
 
 I,abial+r. A common group in I/atin.. The mute is re- 
 duced to the spirant r, and r remains as the syllabic. Exam- 
 ples. — fabrum: fevre, robur: rouvre, coprum : cuevre, super: 
 sovre, etc. After other consonants, arbor: arbre , semper: sempte, 
 scalprum : eschalpre, sulfur : solfre, etc. ; bl has been substituted 
 for br in acer arbor : erable; cribrum : crible. Exceptions. — The 
 preposition super shows a triple development : sovre : sore : sor 
 (N. Pr. sur). The last is undoubtedly the unaccented pro- 
 clitic antevocalic form of the word. The earliest examples oc- 
 cur in the Passion Christi: see Karsten, op tit., where this de- 
 velopment is worked out in detail. 
 
 Some new r-groups arise through the fall of the ultima-vowel 
 before final r. Such are l-r, *melor: mieldre, calor: chalre, 
 nugalior: nuadre (Aiol). Of the two sonants, r possesses the 
 greater sonority, so that the synthesis could not be effected 
 without a perceptible glide. Hence we find the development of 
 the glide-consonant d in most dialects, and everywhere the re- 
 tention of the r in its full syllabic quality. The same occurs in 
 the group Ig-r, fulgur : foldre, with assimilation of the g on the 
 dental environment. 
 
 A similar treatment is experienced by the groups n-r, nd-r 
 s-r, m-r, minor : moindre, senior: seindre, *juvenior: joindre 
 grandior: graindre, antecessor: ancestre, marmor: marbre. 
 
29 
 
 The group/: r presents a difficulty. Examples are p$jor: 
 pire, major: maire, se{n)ior : sire , vismajor: vimaire. 
 
 The pronunciation of / in the earlier period of Latin was that 
 of a semi-vowel i. If j preserved this sound, there would be no 
 reason for the retention of the final syllable as such ; as i pos- 
 sesses more sonority than r, the latter would preserve its conso- 
 nantal character after the outfall of the final vowel. But in the 
 Vulgar Latin of the imperial period, a shifting in the articula- 
 tion oij took place. It became a sibilant spirant (cf. Lindsay, 
 p. 46); this took place undoubtedly before the Gallic vowel- re- 
 duction. When, then, this occurred, the r after the shifted j 
 would assert its value as a sonant, and would serve as syllabic 
 of the ultima. The reduction of j to "parasitic" i must have 
 occurred after this. Had it occurred earlier, and the diphthong 
 ai been already formed, there would have been no cause for the 
 retention of r in its sonant function. The orthography in Prov- 
 encal mdjer, confirms this hypothesis. 
 
 Group r — r. It occurs only in the word soror : suer. The 
 final vowel falls and the two r's are reduced. This develop- 
 ment is noteworthy for the sake of comparison with the corres- 
 ponding secondary group, as in currere. 
 
 mute -\- I. 
 
 Groups of this nature were not common in Latin final sylla- 
 bles. Only pi seems to occur. It shows the same development 
 as pr, except that p is only reduced to b ; I remains as the syl- 
 labic of the ultima. dicplum : doble, triplem : treble. After a 
 consonant, pi naturally remains, exemplum: essemple, templum : 
 temple. 
 
 A new group m-l is formed in insimul : ensemble. Here the / 
 being of greater sonority than m, retains its syllabic function ; 
 and a glide-consonant b is developed. But how are the by-forms 
 ensenes, insiame to be explained ? {cf. Ital. insieme, Prov. en- 
 sems) . If we may assume for insimul + adverbial s a proclitic 
 position, with rapid pronunciation in the stress group, then we 
 have here a pendant to the development of nostros : nos ; / was 
 lost in a group of four consonants, where it could not retain its 
 syllabic character, owing to the atonic position of the whole 
 complex. 
 
3Q 
 
 mute -\- nasal. 
 
 The group gm seems to occur only in words with final a. 
 
 gn. — This group was pronounced in Latin with the guttural 
 nasal ygn (see Brugmann : Grundriss, I, 54). It differs from 
 all other combinations with a final sonant in that the final sylla- 
 ble is never retained after it. The assimilation of the two con- 
 sonants is reciprocal, the end result being palatal n with com- 
 plete loosing of the£--stop (cf. Gutheim : op. cit., p. 6, ff). This 
 process must have been completed before the final reduction of 
 the ultima. It occurs in all the Romance languages, so that it 
 must have occurred comparatively early. 
 
 signum : sin, dignum: din, signum: seing, lignum: lin (cited 
 by Godefr. from G. de Marchant) . pugnum : poing, *praegnis: 
 prains, permagnus : parmaing, regnum: ren, stagnum : estaing, 
 malignum : malin. 
 
 In loan words the ultima is retained, although the palatal «is 
 also produced, as in magnum : magne, agnum : aigne, cyqnum : 
 cygne, dignum : digne. 
 
 Therefore, the assumption of Willenberg {Rom. Stud., Ill, 
 373), that the subjunctive forms which show final e after this 
 group are regular, is inadmissable, notwithstanding the early 
 date of some of them, degnet, Eulal. 26, regne, Rois, 58. These 
 can only be analogical forms ; enseint (Brandan, 129) shows the 
 regular development. 
 
 Sonant -\- sonant. 
 
 Groups rm, rn. After these the final vowel is lost, r is much 
 more sonorous than m, and hence the transition was effected 
 without a break in the expiratory current ; r in this position 
 probably formed a close diphthongic union with the stem-vowel, 
 at mum: arm, vermen : verm, firmus: fers. The ioxra. verme 
 (Jonas, 14), Koschwitz {Commentar, p. 132) explains as a lev- 
 elling to words like terme, charme, where the retention of the 
 final syllable is regular, germen: N. Fr. germe is most probably 
 a loan-word. F 'or rn, alburnum: alborn, cornu : corn, furnum: 
 forn, hibernus: ivern, diiirnum : jorn: *quaternum, caem. 
 Exceptions like alborne and viburnu : viorne, are new feminine 
 formations. The following are loan-words : externum : esterne, 
 modernum: moderne, brnus : orne. The etymology of some 
 
31 
 
 (possibly saturnum, cf. Korting : Lat. Rom. WbrterbucK) is 
 doubtful. 
 
 The assumption of Willenberg (/. c, p. 375) that these 
 groups m, rn, demand "euphonic" e is therefore unjustified, 
 as his own examples retor, a tour, show. 
 
 Im. This group apparently demands a " stiitz-vocal." 
 
 Examples. — oilmen: colme , scalmum: eschalme , psalmum: salme, 
 ulmum : orne, helmu : helme, *calmis ( ' ' lande, ' ' cf. Thomas : 
 op. tit., p. 8): ckalme. The two consonants are not homor- 
 ganic, but as / possesses more sonority than m, the group is pro- 
 nounceable at the end of a syllable. Meyer- Ltibke {Rom. Gr., 
 I, 251) assumes that before the loss of the ultima aim had be- 
 come aulm ; in this group / could only function as a consonant, 
 and the nasal became sonant in character, aul cm ; though accord- 
 ing to the law proposed by Meyer-Liibke, the final vowel was 
 retained as in all proparoxy tones. But is the stage aulm possi- 
 ble ? The vocalization of the /before consonants consists in the 
 gradual loosing of the obstruction formed by the articulation of 
 the point of the tongue, while the position is retained {cf. 
 Meyer- Iyiibke, I, 403) ; how then can a parasitic ^-element be 
 formed before the /, and this retain its character as conso- 
 nant ? Cannot the development be explained by the peculiar 
 nature of the Latin I? This sound was, as we know now, 
 " guttural," that is, pronounced with a well-marked off-glide. 
 {cf. Lindsay, p. 90. Osthoff, Dunkles und Helles I im Latein, 
 in Transactions of the Amer : Phil. Assoc, XXIV, 50.) If this 
 off- glide became more distinct then the syllabic union between / 
 and m* would be broken, and m become syllabic. This expla- 
 nation harmonizes to a certain extent with that of Meyer-Lubke, 
 who also assumes the development of a " Neben-silbe' ' between 
 / and m. (Here I may call attention to the fact that in Ameri- 
 can-English dialects this group likewise loses its monosyllabic 
 character ; and a syllable is developed from the voice of the m. 
 Thus elm is pronounced el-m ".ellum" with two very distinct 
 syllables. ) 
 
 Exactly the same process holds good for In, in alnum : alne. 
 In baldjnium, I was already lost in Vulg. Lat. {cf. Ital. bagno, 
 Sp. bano, Prov. bank. 
 
 mn. This group differs from most of those hitherto consid- 
 
32 
 
 ered in that it shows progressive assimilation ; the end-result is 
 m. Owing to the complete closure of the mouth-passage, in 
 forming the m, the transition to the ^-position cannot be effected 
 without a perceptible glide. Therefore, if the assimilation was 
 late, the retention of the final syllable, as m, is to be expected. 
 Here, however, a considerable variation is found, damnum : 
 dame and dam, sommum : somme, and som (Epistles of St. Bern- 
 ard), scamnum : eschame, solemnis: solemes, Interamnes : En- 
 frames. Loan words are, autumnus: automne, dictamnun: 
 didame (O. Fr. ditain : ditaim, which indicates apparently a 
 Vulgar Latin *dictamus): solemnem: solemne, etc. Meyer- Liibke 
 assumes the forms with e as showing the regular development, 
 explains dam as a verbal substantive. ( cf. Meyer-Liibke, Die 
 Strassburger Eide und das gattische Anslautsgesetz, Ztschr.f. Rom. 
 Phil., XII, 520. G. Paris, Romania, XVIII, 520. Koschwitz, 
 Commentar, p. 9) . The chief question is as to the date of the 
 assimilation. If it were completed when the law for the ulti- 
 ma went into effect, there would be no cause for a preservation 
 of the final syllable. Pompeius, grammarian of the fifth cen- 
 tury, mentions colunna as a barbarism (Lindsay, p. 69), but on 
 the other hand, spellings like mpn in the inscriptions do not 
 favor the hypothesis of early assimilation. If this process was 
 completed late, then the retention of the final syllable may be 
 explained ; consonantal m -f- n cannot be pronounced together 
 without one of them assuming a sonantal character. The by- 
 forms dam and som (Meyer- Liibke' s theory is inadmissible for 
 som) must then represent in some way sentence-doublets, they 
 are probably ante-vocalic forms ; damme -f- voc : ~^>*dapin 
 *^>dam but damnu -f- cons. ~^>amn ~^>ddame ; that the longer 
 form is more common is due to the greater prevalence of initial 
 consonants. The same variation is found with the secondary 
 group m, n. 
 
 m + consonant u. The only example of this group, outside 
 of verb-forms, seems to be the adjective tenuem, which appears 
 in O. Fr. as tenve, tanve, teneve, tenvre. Hiatus u after nasals 
 remains in Fr. as v, cf. Januarius: Jenvier, annualem: anvel. 
 
 The u lost syllabic value early in Vulgar Latin (see Lindsay, 
 p. 52), and was reduced to a consonantal w' sound; the exact 
 nature of the sound at the period of the weakening of the ultima 
 
33 
 
 is doubtful. If it had already become spirant v, the total loss of 
 the final syllable would be probable. If, on the other hand, the 
 pronunciation was still tenue, with u as a semivowel, then the 
 ultima would probably be preserved, as u is more sonorous than 
 n. The latter supposition seems more probable. It is, how- 
 ever, seemingly contradicted by the forms of the ui — perfects, 
 tenuit : tint ; venuit : vint; analogy has played so great a part in 
 their development that it is perhaps better to accept the adjective 
 tenuis as showing the phonetic law. For place-names with sim- 
 ilar development to tenve ; see I^indstrom, op. tit., 56. 
 
 Apart from the w/-perfects, the only other words with conso- 
 nant -f- v final are viduum : veuf and antiquum : antif. The 
 imperfects with mute -f- u all show assimilation of the mute, and 
 subsequent union of u with the stem-vowel, with loss of the ultima. 
 cf. habui, Picard au, placui : plau, debui : dui, potui : pou, etc. 
 These represent undoubtedly the regular ante-consonantal de- 
 velopment, and indicate that the assimilation was completed 
 early ; veuf and autif, with u preserved as f are levelled to the 
 form of the feminine. 
 
 Consonant + hiatus — i. The reduction of i in hiatus to con- 
 sonant y, with loss of syllabic value, took place very early in 
 Vulg. Latin ; already in Plantus the ending ius, iu, is 
 counted as only one syllable. This consonant y then generally 
 palatalized the preceding consonant, and was itself lost. (cf. 
 for all these groups, Waldner : Die Quellen des parasitischen i im 
 Fransosischen. Herrig 1 s Archiv., XXVIII, 453 ff. Meyer- 
 Iyiibke, Gr., I, 423 ff.) I am here concerned only with the fate 
 of the final vowel. Till the palatalization of the preceding con- 
 sonant was completed, and the y sound absorbed, the final sylla- 
 ble must have been retained, asjj/ is of greater sonority than any 
 of the consonants ; after this process was ended, then the reduc- 
 ion of the final syllable could occur as after a simple consonant. 
 We find in general a complete loss of the ultima after these 
 groups, indication that the palatalization was early completed. 
 In loan-words, on the other hand, the ultima is always retained ; 
 and the hiatus i appears &sg{ = dzh), or is "attracted" and 
 united with the stem-vowel. * 
 
 Dental -\-y. The popular development of this group is indi- 
 ■cated by the following examples : ty becomes is, dy is reduced to 
 
34 
 
 parasitic y, both with loss of the following final vowel : 
 •poteo : puis, palatium : palais, pretium : pris, *retium : rots, 
 radium : rai, modium : mui, medium : mi, etc. 
 
 In loan-words the ultima is retained ; ty is reduced to voiceless 
 s, without the development of parasitic i. servitium : service, hos- 
 pttum : hospice, *mutium : mousse, justitium : justice, also justis, 
 (regular), arbuteum :. arbousse, capitium : chevece, vttium : vice, 
 cf. Cohn, Suffixwandl., p. 37 ff. avisstruthio, autruche : O. F. 
 austruce, must also be a loan-word, dy in loan-words sometimes 
 shows a change to r, with retention of the ultima, homicidium .- 
 homecire, Aegidium : Girris, remedium : remire, studium : estuire. 
 cf. Meyer-Liibke, Gr., I, 432, 487. Sometimes also /, homici- 
 dium : homicilie, aegidium : Gilles, inyidia : envilie. 
 
 Lastly dy sometimes gives dz (g), podium : poge, gladium : 
 glage, alsoglai, regular, and glaive, a remarkable form, certainly 
 not popular. 
 
 After other consonants, ty undergoes assimilation (on the 
 earliest Vulg. Latin forms, see Lindsay, p. 83). So far as the 
 final vowel is concerned, its treatment is exactly similar to that 
 of consonant + original J. It always falls, nty, nuntius : 
 noinz, (N. Fr. nonce is an Ital. loan-word, see Littre.) *antius : 
 ainz;rty, terttus : tiers, martins : marz; cty, ductium : dois ; stt,*pos- 
 tiu .-puis, ostium: uis, also uisse, supra, p. 7. cf. further the 
 subjunctives esteez, esforz, quoted by Willenberg : loc. cit.,p. 377. 
 He believes that ' ' assibiliertes ' ' c demands euphonic e, and 
 quotes the early form annunce. Cam. Ps., 105 : 5, as proof. But 
 this view is not borne out bj' the substantives cited above. 
 
 One group of words show an apparent exception to this law : 
 the ultima is retained. They are seemingly nominatives of the 
 third declension, which have been preserved contrary to the 
 general tendencj'. cf. Suchier in Grober's Grundriss, I, 638. 
 captio : chace, destructio : destruce, *tractio : trace, buteo : buse, da- 
 tio : dace. Of these dace and buse, are undoubtedly borrowed ; 
 the others are explained by Meyer-Liibke, Gr., II, 445, as post- 
 verbal substantives. 
 
 Examples of dy after consonant are very rare ; hordeum : orge, 
 *thelandium : chalande : chalandre. Compendium : Compiegne. 
 
 Of these, orge may represent a popular development. We know 
 that dy was assibilated like ty in Vulg. Lat., and in late inscrip- 
 
35 
 
 tions z was often chosen to express this sound, {cf. Seelmann : 
 Aussprache, p. 323) ; this could indicate equally a sound dzh. In 
 this case, after the double consonant, di was not reduced to par- 
 asitic i, like interlocal di, but was developed like initial di. 
 Since the final vowel is retained, there must have been com- 
 plete closure in articulating the d, then followed the explosion, 
 and the articulation of the sibilant, succeeded by a weak vowel 
 sound, as the syllabic *ordzhe. But the word may also repre- 
 sent an original feminine (neuter plural) hordea ; the gender is 
 prevailingly masculine in O. Fr. chalande and Compiegne are 
 probably not popular words. 
 
 For sy examples are rare ; they show loss of the final, segu- 
 sium : seus, basio : bais. 
 
 Guttural -j- y. ci gives regularly is (z) in O. Fr. gy, a para- 
 sitic i. After both the vowel of the ultima is lost. Examples : 
 facio : faz : fais ; laqueum : laz, bracchium : braz ; panicium : 
 paniz ; glaciem : glas ; solatium solaz, etc. 
 
 Exceptions are quite numerous. All must be classed as loan- 
 words, delicius : delices, fabricium : fabrice, licium : lisse, judi- 
 cium : juise, (alsojuiz, regular) / arcaceum : arcasse, bisaccium : 
 besace, pellicium : pelisse, convicium : convice, vinaceum : vinasse. 
 Others show ch (= ish) ; they are probably loan-words from 
 Italian. *posiicium ■: postiche, *pisaceum : pisache, *spinaceum : 
 espinoche. gy ; popular words are : exagium : essai, rcfugium : 
 refui. In loan-words, gy generally produces dzh (g) ,with'retention 
 of the ultima. *dis/ugiuum : deffuge, horologium : horloge, refu- 
 gium : refuge, elogium : eloge ; navzgium : navire, shows a change 
 similar to that of dy. 
 
 After other consonants ci and gi are treated analogously ; the 
 ultima is not retained, trunceum : irons, *romancium : romanz, 
 (cf. Meyer-Eiibke, I, 252) , *barchio : bars, longius : loins, nes- 
 cius : nice, is surprising ; we should except *nis. It may be a 
 generalized feminine form. 
 
 Whether Georgium : forge is a popular form, indicating a re- 
 tention of the ultima after rgy, as in orge, is, in the lack of other 
 examples, difficult to decide. It may be a Greek-loan word. 
 
 Labial + hiatus *. This is the most difficult of these 
 groups. Schwan (Afr. Gr., p. 88) accepts the theory of a 
 
36 
 
 double development, (i) py, £y>parasitic «", which fuses with the 
 stem-vowel; (2) py, by>tsh, dzh {ch. z). 
 
 The last is certainly regular before final a. Before other 
 vowels we find, on the one hand, sapio : sai, debeo : dot, habeo : 
 at, *quadruvium : caroi (also caroge) ; and on the other propius : 
 proche, apium : ache, rubeum : rouge, *sabium : sage (also savie : 
 saive, borrowed) . rabiem : rage, trivium : trege, niveum : neige, 
 vtduvium : veouge, alveum : alge, diluvium : deluge, levium : liege. 
 
 Of these, ache, rage, neige, alge, are feminines, and may go 
 back to Vulg. Lat. feminine forms with final a. 
 
 Proche, rouge, sage are adjectives, and may show levelling to 
 the feminine, a fact frequently observed, deluge 'is a loan-word, 
 as ii^>u shows. Their remain trege, rouge, and liege, which are 
 masculine substantives. The etymon of trege is somewhat 
 doubtful. Schuchardt (Ztschr. Rom. Phil., IV, 125) proposes a 
 Keltic origin. Liege is irregular in its stem-vowel, which may, 
 however, be due to the analogy of levat : lieve. 
 
 All things considered, I believe, with Waldner, that sai, ai, 
 represent the genuine popular development of this group. 
 These show an assimilation of the consonant beforejy after which 
 the ultima falls. If, however, the forms with^-, ch, are popular, 
 the retention of the final syllable shows that the assimilation of 
 p, 5 before^, which had become a blade-point spirant, sh, zh, was 
 completed late. 
 
 The following are undoubted loan words : *Jlovium : fluive : 
 pleuve, * avium : aive, ingluviem : e?igluive. 
 
 Sonant + hiatus i. ly produces a palatalized I, after which 
 the ultima is not retained, as the assimulation probably was one 
 of the earliest. Examples : allium : ail ' , polium : fueil, consilium : 
 conseil, extlium : eissil, etc. Willenberg again puts / as one of 
 the consonants which call forth " stiitz-^." But his examples 
 (consoilles, etc.) are certainly analogical, cf the organic con- 
 junctive forms cited by him (p. 381). merreilt, Ox. Ps., 120-3; 
 consaut Ogier, 12442. 
 
 Exceptions generally show the pure lingual /, with retention 
 of the final syllable ; they are all loan words : * solium : suele 
 (beside sueil, regular) , pallium : pdlie, nobilium : nobile. 
 
 Meyer-Eiibke (Gr., p. 252) counts pal. / as one of those 
 sounds after which final vowels persist, through retention of the 
 
37 
 
 vocalic character of the i. Elsewhere, p. 439, he cites paile and 
 uile as loan-words. O. Fr. uil, masculine, is also found. 
 
 Occasionally in loan words we find a change of / to r. con- 
 cilium : concire, evangelium : evangire, trifolium : triforie, etc. 
 A sound-substitution of this nature speaks for the non-popular 
 character of these words. 
 
 n -\- y ; This .group produces early palatal n after which 
 the final vowel falls. coneum : coing, junium : juin, scri- 
 nium : escrin, ingenium : engin, etc. Loan-words frequently 
 show the sound dzk after n, as a result of the reduction of hiatus i 
 to a consonant, without assimilation of the n. The ultima 
 naturally persists in these, extraneum : estrange, lineum : linge, 
 laneum : lange. Others show dental n, without palatalization. 
 progeniem : progene, idoneum : idonie : aoine, dominium : demeine, 
 capitaneum : chevetaine, preconium : preone, etc. 
 
 r -\- hiatus i. This group shows likewise loss of the final syl- 
 lable after it, with early "absorption" of the hiatus i. varium : 
 vair, imperium : empir, corium : cuir, dormitorium : dortoir. Loan- 
 words again may be divided into two groups : those in which 
 the hiatus i is retained a.sg ; and those where it is attracted. 
 In both the final is retained : * 'excoreum : escorge(cf. cuir) , soro- 
 rium : serorge, cireum : cierge, salhorreum : salorge, eboreum : 
 ivSrie : ivoire, ciboreum : cib6rie, {mar) *moreum : moire, magis- 
 terium : maestire ; also, maesteyr (Alexander- fragment), which 
 may be regular. Further monastire, battistire, avoltire, *bur- 
 rium : buire, empire beside empir, etc., etc. 
 
 m -j- hiatus i. The only example of this group, apart from 
 those words which have ultima a {vendange, losange, etc.) is 
 simium : singe. The Romance cognates, Prov. sin/a, Ital. scim- 
 mia, Sp-Jimia, indicate a feminine etymon *simia, but the word 
 is masculine in New French. Singe shows that the hiatus i 
 became a spirant zh ; so long as the m preserved its labial articu- 
 lation before this, amonosyllabic pronunciation without a ' 'neben- 
 silbe" would be impossible. The date of the assimilatory pro- 
 cess by which m>n is unknown ; if it took place after the re- 
 duction of the final syllable, the dissyllabic form is explained. 
 A similar result appears in the group mny, with only one exam- 
 ple somnium : songe. The possibility of analogical influence by 
 the verb songer is present in this case, so that it is difficult to 
 decide if the retention of the final syllable is completely regular. 
 
38 
 
 PART II. 
 
 The Unaccented Vowels in Proparoxy tones . 
 
 The vowel of the penultimate syllable of proparoxytones in 
 Latin is always i or u ; a appears only in Greek loan-words, 
 mouachus, pampanus, orphanus, calamus, Lazarus, etc. From 
 the earliest period of the language a tendency toward syncope is 
 observed. We find, for instance, domnus, ardus in Plantus, as- 
 pris in Vergil : Aeneid, II, 374, soldus in the Lex municipalia 
 of Caesar. (For other early examples see Lindsay, p. 184, ff.) 
 In the vulgar speech this tendency was much more marked and 
 extensive. It probably was at all times a peculiarity of the 
 ruder, uncultivated language spoken by the people. Undertheetn- 
 pire, shorter forms are mentioned more and more frequently by the 
 grammarians, and the appendix. Probe has a long list of such 
 syncopated proparoxytones. Schuchardt's list of examples from 
 the inscriptions ( Vocalismus, II, 403 ff.) shows that the words 
 most commonly syncopated were those with a liquid or nasal in 
 the penult or ultima. Most common is the syncope in the case 
 of the suffix — cuius, of which Schuchardt gives no less than 27 
 examples. It is questionable, however, whether the longer 
 forms of this suffix ever really belonged to the popular language. 
 In this the middle vowel u is unoriginal, being developed in 
 classical Latin from the voice of the /. It represents the Indo- 
 European suffix — tlo. (cf. Lindsay, 146, 302.) Plantus gener- 
 ally has the shorter paroxy tonic form. Apart from these com- 
 binations with liquids or nasals, the only examples given by 
 Schuchardt are those with the secondary groups b-t, grabtas, 
 c-t,filicter„ g-t, dictus, g-d,frigdo, s-t, postum {cf. pp. 143, 144). 
 Whether all these indicate a real syncope for the earlier period, 
 or how far they are due to mistakes of the stone-masons it is 
 impossible to determine. 
 
 To explain these syncopated forms, various attempts have 
 beeu made. Stolz (Indg. Forschungen, IV, 234) , following von 
 Planta (Grammatik de? oskisch-umbrischen Sprache) connects it 
 with the quantity of tne final syllable. If this is short, the pen- 
 ultima vowel is preserved; if long, then syncope occurs. For 
 example: calidus ; calidum, from original caledos, caledom but 
 
39 
 
 caldo, caldi, from original caledod, caledi. Then caldus by anal- 
 ogy, Osthoff (Wolfram's Archiv, IV, 355 ff), on the other hand 
 explains the shorter forms as " Schnellsprechformen," that is, 
 they are the natural reductions which we observe in the ordinary 
 rapid speech of every day. When we articulate rapidly, there 
 is always a tendency to slur or drop entirely the unaccented 
 shorter syllables. This explanation is accepted also by Soemsen 
 (Studien zur lateinischen Lautgeschichte \ p. 159 ff.) and by 
 Skutsch, (Forschungen zur lateinischen Grammatik und Mettik, 
 pp. 47, ff. 57). 
 
 In French, this tendency is carried out to its legitimate conclu- 
 sion. All penultimate vowels fall, even a. Just when this 
 process was completed, it is impossible to say, but probably very 
 early in the history of Gallic Vulgar l,atin. It was before the 
 dialectic separation from Provencal, as it occurs also in the lat- 
 ter. 
 
 First to be considered is a group of words which show appar- 
 ently a total loss of the final syllable. These have been last dis- 
 cussed by Horning (Ztschr. f. Rom. Phil., XV, 493 ff.), who 
 gives a long list of such examples, especially from the eastern 
 dialects. In addition to the examples given by Horning, I may 
 cite the following : in icus, apostolicus: apostolie (Alex, 61): apos- 
 toile, arsenicum: arsoinc, (d'Aubigne in God.), basilicum : basile, 
 (Vie de Ste. Christine in God.), cf. basoche, magicum : mage : 
 maige (Wace), remedicum : remire (Benoit), umbilicum: omble, 
 persicum : perse : pierse (Ph. Mousk£), rusticus : ruste : ruistre. 
 * exmanicum: esmain (cited by Godef. from a chart of 1565). cf. 
 manche. dalmaticum: dalmaire (Dial. St. Greg, in Godefr.) cf. 
 dalmage. artem magicum: artemaire cf. artemage. grammati- 
 cum : grammaire. In icem, lubricem : lubre, simplicem : simple, 
 codicem : code, itnbricem: limbre, sardonycem : sardoine. 
 
 Further we find verginem: virgene: verge, *antiphona: antefie: an- 
 teife (Trad. deBelet. in Godef) . cf. antiennes, marginem : matge, 
 organum: orgue, turbinem: trombe, tympanum: timbe: timbre, can- 
 onem : cane, *arganum : atgue, imaginem : imagene : image, 
 Carthaginem : Cartage, cophinus : coffe (cited by Godef. from a text 
 of 1472) : coffre, supplicem: souple, pallidum : pale, aspidem : aspe, 
 ■mucidus: mucre, pecudem: peque : pec, rancidus : ranee, salmacidus : 
 salmache, senopida : sinople, *diaspidem : diaspre, consolida : consire 
 
4° 
 
 {Oliv. de Soir. in Godef.), <f. solde, *ruvidum : rove, flaxidus : 
 flasque, expavidus : espave, magida : mate : mire ; {cf. maide), 
 cupidus : cove, sapidus : sape {cf. sade) , *digita : doze {cf. doit), 
 *aboculus: avucge {cf. avuele, aveugle), scandalum : eschande {cf. 
 eschandle) , fragilem : frage (cited by Godef. from D. Anton 
 Chroniq), {cf.fraele), mugilem: muge (Baif. in Godet.),pagina : 
 pagene : page, vasculum : flasche (?), *situlum : set (Renart in 
 Godef.): sete {cf. settle), scedula : cede (cited by Godef. from a 
 text of 1530), *beryculus: berique, *amandola : amande, ange- 
 lum: angele: ange, clavocymbalum : clavecins , cicerem : cese (cited 
 by Godef. from B. de Gord. Pratiq.), cf. ceire), cancer em : cran- 
 che (Mirede St. Eloi in Godef.), borboros: bourbe,dactylus:datte, 
 colera : cole (Anc. Th. frc. in Godef.), {cf. colre, ckolere) , passa- 
 rem : passe {cf. paistre) , *cercita : sarce (?), semita : sent {cf. 
 sente), spiritum : espir (cf. esperit), subitum : sups (cited by 
 Ste. Pallaye from a text of 1373), cf sonde, principem: prince. 
 
 Horning considers these forms, at least so far as concerns 
 those in icus, idus, as genuinely popular, remnants of a very 
 early period in the history of the language. He believes that 
 for Eastern dialects the syncope has occurred in every case, and 
 that the final forms are the results of various processes of assim- 
 ilation. In icus the c was reduced through £■ tojp, then syncope 
 took place, after which y disappeared without affecting the pre- 
 ceding consonants. For the other words, in idus, cited by him, 
 he proposes various possibilities, into which I cannot enter more 
 nearly. 
 
 Meyer- Liibke, with the exception of a few words with a in 
 the penult, like monachus : moine , fecatum : foie, anatem : ane, 
 {Gr., I, 262, 447), or with^- as the initial consonant, doie (448), 
 classified these as learned formations. Neumann: {Ztschr.f.Rom. 
 Phil., XIV, 550), puts them all in the latter category. I be- 
 lieve that the last is the true explanation of these remarkable 
 forms. We find by the side of nearly all other forms which show 
 the regular development with syncope and subsequent consonant 
 assimilation. To assume with Horning that these forms are 
 popular, is to assume two or more developments of one and the 
 same sound group, an assumption which can be accepted only 
 when the relations of " Satz-Phonetik " come into play, an 
 assumption not considered by him. At least for the central dia- 
 
4i 
 
 lect from which the literary language developed, such forms can 
 only be considered as abnormal and non-popular. How far dia- 
 lectal differences may occur, it is impossible to say. I do not 
 understand from Homing's essay that these words represent the 
 sole forms under which the proparoxytone endings appear in the 
 Eastern dialects ; if such be the case, then we have here a real 
 instance of dialectal differentiation. For central French, how- 
 euer, this possibility is excluded ; and the assumption that they 
 are words of late and learned origin seems justified. Proparoxy- 
 tones are unknown to the language, and when such were intro- 
 duced artificially, it sought to adapt them to the prevailing 
 accentuation in various ways. In the earlier sheath of learned 
 proparoxytones the whole final syllable was thrown off ; after 
 their adoption they underwent various modifications, some of 
 which are not easy to understand, as the origin of the diphthong 
 in grammaire, tieve, etc. In proparoxytones introduced later, the 
 tonic accent was generally shifted to conform with the usual 
 paroxytonic accentuation, as in fragile, rustique, mobile, etc. 
 
 A striking illustration of the different ways of treating a 
 learned proparoxytone is found in the word fragilem ; ( i ) fraile, 
 earliest form, with the syncope, but without development of 
 palatal I, and with retention of the final vowel ; (2), frage, with 
 loss of the whole final syllable ; (3) , fragile, with accent shifting. 
 
 There is one possibility, however, which demands attention. 
 May not some of these words tuiste, dole, foie, etc., which are 
 very old, represent by-forms having a different accentuation in 
 the stress-group ? The degree of stress given to a single sylla- 
 ble varies, and the relative position of the various subordinate 
 accents varies with this. The usual reduction of proparoxy- 
 tones in Romance implies an accentuation with a secondary 
 accent on the ultima, and with least stress on the penult, / ^ ^ . 
 Could, under conditions which are now obscure, this accentua- 
 tion be changed to one where the penult receives the second- 
 ary accent ^ W.1 If this supposition be correct, then the loss 
 of the final syllable is easy to account for, as that syllable with 
 the weakest stress will always fall. One fact, however, seems 
 to decisively negative this supposition, namely, the development 
 of the vowels in the accented syllable. If the above accentua- 
 tion really prevailed in some instance, then we should expect to 
 
42 
 
 see the accented vowel develop as in an open syllable. This only 
 occurs in the East French forms manicum : men{e,) granica : 
 grene, rapidus : reffe; and in O. Fr. tieve (In dole , foie , etc., the 
 diphthong may be due to the following palatal consonant) , and 
 O. Fr. avuege. In comparison with the whole number, these 
 form a very small percentage, and therefore the assumption of 
 sentence doublets is probably to be rejected. 
 
 Meyer-L,iibke's hypothesis, that the vowel a in the penult 
 hindered syncope for a time, is based on moine, foie, ane, and 
 one or two other words. It is in keeping with the greater 
 sonority of the vowel a. But why, if Meyer-L,iibke's series, 
 anatem : anade .- anede : anee : ane (Gr., I, p. 447) be correct and 
 a be retained till after the fall of interlocal d (end of the elev- 
 enth century), is not the tonic vowel developed to ai, as in all 
 open syllables {cf. panem -.pain). It is moreover contradicted by 
 other forms with syncope of a in the penult, cf. cannabis: 
 chanve, petalon : poile, scandalum : eschandle, bufalus : buffle, bal- 
 samum : balsme, platanus : plane, Stepkanus : Estienne, calamus : 
 chalme, Lazarus : lasdre, colaphum : colp, etc. 
 
 It will have been noticed that nearly all of the words in -nicus 
 appear only in the learned form ; so, moine, canoine, arsoine, 
 maine, vetoine, etc. The ecclesiastical and medical character of 
 these is evident. The popnlar development seems to be repre- 
 sented by manica : manche, manicum, "handle" : mange, (Gir. 
 de Ross) : later manche, *granica : granche, manichus : mouge : 
 canonicum : canonge : Santonicus : Saintonge. Why is the final 
 vowel retained in these words ? cf. the primary group nc, banc, 
 mane, etc. As in the case of the suffix -aticum, the syncope 
 could not have occurred till after the shifting of c, through g, to 
 a spirant zh before the weakened vowel e. Then in the passage 
 from the dental n to the blade point spirant zh a slight ' ' neben- 
 silbe" or glide would be necessary ; this, aided by the original 
 greater stress of the ultima in proparoxytones has preserved the 
 final syllable. The commoner forms, moine, canoine, etc., I 
 should explain by the rejection of the whole final syllable with 
 its consonant, and the retention of the penultimate vowel as 
 consonantal i, with later "attraction" canonicu : canonie ; 
 canoine. cf. the forms monie, Roland; munie, St. Brandan, 
 etc. 
 
43 
 
 Similarly apostolicu : apostolie : apostoile, antiphona : antefie : 
 anteife. 
 
 The older language has a few examples of the apparent reten- 
 tion of proparoxytones, mostly in borrowed words; thus virginem: 
 virgene : later vierge, imaginem .■ imagene : image, ordinem : or- 
 dene : ordre (cf. ome) , diaconem ■ diacene : diacre, pagina : pagene: 
 page, orphanum : orfene : orfe,juvenum : jovene : juefne, Stepha- 
 num : Estevene : Estienne, anima : aneme : ante, angelum : an- 
 gele : auge, utilem ; utele : utle, schandalum : eschandele : eschandrf, 
 titulum : titele : titre, idolum : idele (Roland 3664) : idle. All of 
 these count for but two syllables in the metre, and in most 
 cases, the e of the penult is probably only a graphic sign, to 
 express the consonantal character of the v (written u) as in 
 jovene, or the spirant character of the g, as in virgene. In other 
 cases, titele, etc., it may represent a slight secondary syllable 
 developed from the voice of the / or n. In slower pronunciation 
 the vocalic character of these sounds would assert itself, and 
 produce an evident voice-glide after the consonant, *idele, or- 
 dnne, etc. 
 
 In discussing the pretonic syllables, I shall have occasion to 
 come back to this tendency of the vowel-like consonants for 
 more extended consideration. (cf. on these forms, Paris. 
 Etude sur le r6le de V accent, p. 24 ff. andTobler, Versbau, 28.) In 
 a few instances this svarabhaktic vowel is developed in original 
 paroxytones ; cf. tenuem: teneve (generally tenve) fabrum : fevere 
 (generally fevre) . But in these the e may be only orthographi- 
 cal. 
 
 Treatment of the Proparoxytones with Regular Syncope. 
 
 I omit all proparoxytones with original final a, and the late 
 borrowed ones with accent shifting. 
 
 Sonant -f- consonant. 
 
 r-\- dental, syncope in all the Romance languages. It was 
 probably very early, the group being already frequent in Latin 
 as a primary combination. Consequently the final vowel falls in 
 French, luridum: lort, laridum: lart, viridem: vert, experitum : 
 espert*exaritum: essart, horridum: ort; fascem- luridum: falourde, 
 is an exception difficult to explain. The etymon is not certain. 
 cf. Kbrting. Lat. Rom. Worterbuch. 
 
44 
 
 r -f- Guttural, r-c seems to occur only in the word clericum: 
 clerc, named by Schwan {Afr. Gr., p. 27) a loan-word, on ac- 
 count of the c. But would not the final guttural be unvoiced, 
 even though it had been reduced to g before the syncope ? cf. 
 vert. The ultima then falls as it does in arc, pore, etc. The 
 character of the word, however, renders it liable to suspicion, ex- 
 pressing as it does a purely ecclesiastical idea. If this view be 
 correct, then the by-forms clergue, clerge, represent later and 
 learned importations. 
 
 The development of r-g in Bituriges-.Bourges militates against 
 this view. In this case, the retention of the ultima must be ex- 
 plained by the possibility of a greater stress on the final syllable 
 of proparoxytones than of paroxytones, enough to preserve it in 
 favorable positions in the stress-group. Here the final vowel is 
 always anti-consonantal, and a voice glide is necessary to effect 
 synthesis between rdzh and .f. asparagus-.asperge is an Ital. loan- 
 words. Proparotoxytones with guttural after r-groups occur 
 but rarely, forge iromfabrica seems to have been influenced by 
 the verb forger. We should expect *forche, in accordance with 
 the law established by Neumann. *tenebricus:tenerge offers dif- 
 ficulty in its stem-vowel, and like forge may stand in a system 
 with the verb tenergier. After such groups, the retention of 
 the ultima would probably depend on the late or early assim- 
 ilation of the labial before r. {cf. however, Meyer- Liibke, Ztschr. 
 f. neufrz. Spr. und Litt., XV, 87). 
 
 r -\- labial occurs only in *scarabus: escharbe, which is hardly 
 a popular word. 
 
 / + dental. This group was syncopated early, {cf. Zt. caldo, 
 etc. ) , and the final vowel falls as it does after primary It. cali- 
 dum: chalt, Curio ssoltiae: Corseult, *falktum: fait, *solutum: solt, 
 soljdum: solt, also sol: sou. The last word shows generally lev- 
 elling to the feminine, solde. 
 
 /+ guttural. / in union with velar c, g does not occur. After 
 1 + palatal c, the final vowel is, as a rule, retained, pollicem .- 
 poke, ilicem : eke : yense, pulicem : puce, salicem : sake, with the 
 by-forms sals (Ps. of Oxford) . These words are adduced by 
 Meyer- I^iibke in support of his theory that the syncope does not 
 occur till after the fall of the ultima. If this were the case, why 
 should not the accented vowels be developed as in open sylla- 
 
45 
 
 bles ? This is never the case, and seems to constitute a fatal ob- 
 jection to his theory. How then is the preservation of the ultima 
 to be accounted for ? The form sals occurs in one of the earliest 
 monuments, cf. also the subjunctive forms cdlligit: emit, collocet: 
 culzt, *caballicet : chevalzt. The same singular retention of the 
 ultima after r-c is seen in *mellaricem: meleze, if the etymon be 
 correct. To account for the retention of the ultima in these and 
 semilar proparoxytones ( pance, comte, etc, ) , as contrasted with 
 its fall after similar primary groups, may we not suppose that 
 the original higher grade of accent on the ultima in proparoxy- 
 tones has preserved the vowel? In a word like picllcem, the stress 
 of the final syllable would be greater than in the dissyllabic 
 cdlcexa, where the ultima follows immediately the tonic syllable. 
 If, after the syncope, this greater stress was maintained to a cer- 
 tain extent, the final syllable would show itself more capable of 
 resisting the weakening process than otherwise. The greater 
 force given into the accented syllable — the chief cause of vowel- 
 reduction in French — could act directly on the final only after 
 syncope. This theory has already been put forth by Iyindstrom, 
 op. cit. p. 46. (cf. also Vising, in his review of I^indstrom's 
 work; Littbl. fur Germ, und Rom. Phil., XIV ,289.) The by-forms 
 sals, chevalzt, etc., are probably due to relations in the stress- 
 group, where the stress of the whole combination was weakened. 
 Verb forms especially are more weakly accented in the sentence 
 than substantives. 
 
 / -f- labial, in the one example, colaphum • colp, shows early 
 syncope and loss of the final vowel, polypus : poulpe, N. Fr. 
 pieuvre, is probably a loan-word, as the inorganic r in N. Fr. in- 
 dicates. 
 
 l-\- labio- velar qu is found only in aliquod : alques. This longer 
 form may be due to the analogy of other adverbs like donques, 
 onques, etc. It is remarkable that *alcs is never found. The 
 form may possibly be derived from aliquas. 
 
 n -\- dental. This group is rare. The final falls as after pri- 
 mary nt. repoenito : repent, genitus : gent, sub-monitus : semont; 
 cognitus : cointe, seem to be an exception, cf. the primary group. 
 It is cited by Meyer-Liibke in support of this theory of 
 late syncope. But in the compound acoint the shorter form is 
 the rule: cointe would then be a generalized feminine form, as in 
 
4 6 
 
 many other adjectives, lasche , chalve , etc. Namnetes .-Nantes, 
 Vannetes : Vannes are further exceptions. 
 
 n + guttural has already been considered. I may also men- 
 tion the word donique : done, and donques, which follow the ana- 
 logy of one ■ onques. 
 
 n + labial is not found as a primary group. Examples : can- 
 nabis : chanve : chanvre, sinapis : sanve. 
 
 Here the preservation of the final vowel seems to be due to 
 the disparity in position of n and p, b ; the transition from 
 one position to the other could not be effected without a decided 
 glide. As compared with colaphum : colp, both show late syn- 
 cope after the weakening of the intervocal labial to v. 
 
 m -\- mute, m as the initial consonant in the ultima of 
 proparoxytones only occur before t and palatal c; both unknown 
 as primary combinations. After each the final vowel is reg- 
 ularly retained, comitem : conte, fremitus : friente, *finiitus : 
 fiente, * comitem : jante, tarmitem : arte (?), domitum : donte, 
 *pumicem .-ponce, rumicem : ronce, *adrumicem : aronce, also aronz, 
 (cited by Godef. from chasse de Gaut.), ramicem : ranee, *cami- 
 cem : chanse. For further instances in place-names, cf. Lind- 
 strom : op. cit., p. 54. 
 
 In addition to the principle of greater stress noted above, 
 another one may have been at work here. Until the assimila- 
 tion, m^>n, before t, or ts ~ c was accomplished, a perceptible 
 voWel glide would be necessary to pronunciation. This would 
 tend to preserve the final vowels. arons, masculine, is proba- 
 bly a new masculine formation ; it occurs comparatively late. 
 *cremitus : orient, may also be a new masculine formed according 
 to the analogy of other strong participles. 
 
 m-d seems to be found only in quo modo, com and comme. The 
 shorter forms occur first in the Passion. The word shows strong 
 reduction in all the Romance languages. Ital. come, Sp. como 
 (com) , a reduction probably due to its procletic use. The dis- 
 appearance of d is difficult to explain. Meyer- L,iibke {Gr., I, 
 505), assume a Vulg. Lat. quomo. 
 
 s -\- mute. This occurs only in the combination s-t. There 
 are very early examples of the syncope. As with primary st, 
 the ultima falls always. positum : post, praepositum .- prev6t, 
 *acquaesitum : aquest. 
 
47 
 
 mute + j, seems to be found only in the proper nouns, Vidti- 
 casses : Vieux, Tricasses : Troyes, which show a remarkable diver- 
 sity. I^ack of material does not enable to decide which repre- 
 sents the regular development. The difference may be due to a 
 difference in the pronunciation of the original Keltic words, 
 which the Latin transcriptions represent only approximately. 
 
 Mute -+- mute. 
 
 We find a few examples of homorganic mute groups arising 
 through the syncope ; all are cases of dental + dental. The 
 final vowel falls as it does after original double consonants. 
 peditum : pet, putidum : put, madidum : mad. Participle-sub- 
 stantives like *fendita : fente, vente, tonte, etc., are all based on 
 the feminine, quitte is probably best explained with Suchier as 
 a learned legal form of quietus, not from *quitidus. 
 
 Guttural + dental. Like the primary group, after this com- 
 bination the ultima falls. Syncope, though peculiar to Gallic 
 Vulgar Latin, was probably early, and the resulting assimila- 
 tion was completed before the final reduction of the last syllable. 
 placitum: plait, solltcitum : souloit, rugltum : ruit, rigtdnm : roit, 
 magidem; maid (cf. maie) , adplicitum : aploit, digttum : doit, vic- 
 ginti : vint, *voatus : vuit. The same reduction occurs after a 
 consonant in gurgitem : gort, de-ex-pergitum : despert. Excep- 
 tions, Jlaccidus : flaiste, (cf. Ztschr. f. Rom. Phil., XI, 254). 
 *muccidum: moiste, raucidum: roiste. These are probably loan- 
 words ; they show, contrary the examples given above, late 
 syncope, after the assimilation of ^ before z. The verb-forms 
 faites, dites, estes, are undoubtedly due to analogy. 
 
 Dental + guttural. This group is not found as a primary 
 combination. In the case of relar c, it shows syncope after the 
 reduction of the tenues to mediae. Then ensued reciprocal 
 assimilation, and the sound dzk(g) is the end-product. This 
 assimilation was completed late. Schwan and others suggest 
 that the sound dzh was, in early French, unpronounceable 
 finally. Picard forms, like sench, cuich, 'dolch, etc., seem to con- 
 tradict this although here the sound tsh has a different origin, 
 and a dialect distinction may be present. The sounds d and 
 zh (point and blade point respectively) are not exactly homor- 
 ganic, and a slight glide is always heard after the combination ; 
 more especially if the stop of the d is marked. Examples : 
 
4 8 
 
 aticum : age, *pedicum : piege, medicum : mege, *huticum : huge, 
 sedicum : siege, porticum : porche, domesticum : domesche, levisti- 
 cum : livesche, toxicum : tosche. 
 
 Dental + palatal c likewise shows retention of the final vowel. 
 In this case, we shall have to recur to the explanation given 
 above for l-c; the two consonants are homorganic after the assib- 
 ilation of the c, and the so-called " affricate " ts, dz, occur fre- 
 quently as a final. *excorlicem : escorce, sedecim : seize, *dodecim : 
 doze, ondecim: onze, tredecim : ireze, panticem : pause, quindecim: 
 qninze. judicem: juge is due to the analogy of juger (see Meyer- 
 Lubke, Ztschr.f. Rom. Phil., VIII, 232.) 
 
 Guttural -+- L,abial. A very rare combination. It seems to 
 occur only in episcopum: evesque, a word that is not genuinely 
 popular. With the regular syncope we should expect a form 
 with assimilation of the middle consonant in the group, possi- 
 bly *evespe (cf. Span, obispo) The word seems to belong to the 
 category first treated, which show a loss of the whole ultima syl- 
 lable. So likewise Jacobus : Jacques, and Jaimes ; the last may 
 show early substitution ; Jacomus for Jacobus ; if so, then it is 
 regular, (cf. however, Meyer- Liibke. Gr., I, 494, and Karsten, 
 Modern Language Notes, VII, 343). 
 
 Labial + Guttural. This group is still rarer. One example, 
 sarcdphagus : sarcos, sarquieu. Meyer-L,iibke, (Gr., 443), as- 
 sumes reduction of agus to aus, as Vulg. I,at. *sarcoJus ; and 
 some such form seems necessary to explain the later diphthong- 
 ation, which can only occur in an originally open syllable, (cf. 
 also Grober, ALL, 2, 459, and Iyindstrom : op. tit., p. 46, where 
 reductions of a similar nature in place-names are cited) . 
 
 Labial -+- palatal c is more common. Most of the examples 
 show retention of the final syllable; thus : Jbrjicem : Jorce, hir- 
 picem : herse. In this triple consonant group, the middle conso- 
 nant falls, or is assimilated. The process could not have been 
 completed till after the loss of the final syllable. So long as 
 both mutes were pronounced with a full stop, the weakened 
 ultima would be retained as the necessary vowel-glide, after the 
 explosion of the last. On the other hand, EburSvices ; Evreux 
 indicates an earlier assimilation of the weaker spirant v ; at a 
 period before the final weakening of the end-syllable. 
 
 Dental + Dabial. This group does not seem to be found. 
 
49 
 
 Labial -\- dental. A common group ; the final vowel is gen- 
 erally retained after it. sapidum : sade, subito : sonde, rapidum : 
 rade, *malehabitum : malade, cubitum : coude, vapidum : fade, am- 
 bitum : onde, computum ; conte, tepidum : tedde : Hide, hispidum : 
 hisde, emputum (Gr. i'fi<j>vroy) ; ente, impetum : ende, *dubitum : 
 doute, (verbal subst.), hospitem ; oste. {cf. for the primary group, 
 set, sotz, etc.) I believe this distinction to be due to a difference 
 in the chronology of the assimilation. The assimilatory pro- 
 cess in the primary group was completed early, as we have seen, 
 probably before syncope in proparoxytones. After this latter 
 process, a period of assimilation would ensue. We are in 
 ignorance of the rate with which this process was consummated 
 or when it was completed. But the retention of the ultima in- 
 dicates that it was late. The combination bd was unfamiliar, 
 and the complete closure of the lips in the b articulation, with 
 subsequent explosion and formation of the d position would ren- 
 der a weakened final " Neben-Silbe " necessary. Still some in- 
 stances of the loss of the final are found. Atribates : Arras, *de- 
 servitus : de-sert, ad-compulum .- acont, (cited by God. from a 
 texte of 1298), debitum : det ^Guiot, Bible in God.) : deit, redu- 
 bitum : redot, (Chrest, Cliges, 5450), *reputum ■ ret, (Gamier, 
 Vie de St. Thomas in God.), terra, *movitum : terremot, (Dial 
 de St. Greg, in God.) cf. further the form dout <^ dubito and 
 dubitem, often found in the earlier monuments. No conclusion 
 is here possible, since dout may be due to the analogy of other 
 1 sq. pres indicative and subjunctive like aim, mont. cf. desir. 
 
 Some of these, ret, redot, acout, are probably late formations, 
 verbal substantives. On the other hand, for Arras and those 
 words with original v, we must assume that the assimilation 
 was completed earlier. This is not surprising, especially when 
 we consider that the reduction of intervocal b to v was one of the 
 earliest developments in Vulgar Latin. The combination v-\- 
 mute is pronounceable without a vowel-glide. *levitum : leude 
 is not entirely popular, as the vocalization of the v shows. 
 
 Mute + r. As after the original combination, the final sylla- 
 ble is here retained, either as sonant r, or as the indefinite e, de- 
 veloped from the voice of this. It is difficult to decide which 
 was present in the older language. Examples are .very numer- 
 ous : I cite only a few. 
 
5° 
 
 Guttural + r. In secondary groups, gr always shows reten- 
 tion of the final, facere : faire, ragere : raire, cicerem : coire, 
 ligerem : hire, lucere : luire, adaugere : aoire, etc. The same 
 development occurs for gutt. + r after other consonants, so far 
 as.the final vowel is concerned. Various processes of assimila- 
 tion occur, into which I need not enter here. The groups nc-r, 
 ng-r, rc-r, rg-r, Ig-r show an assimilation of the guttural to the 
 dental environment. (See Neumann, Littbl. f. Germ. u. Rom. 
 Phil., 1885, p. 244); Karsten, Mod. Lang. Notes, I, p. 219; Gut- 
 heim, op. cit., p. 55). Examples : jungere • joindre, tingere ; 
 teindre, adirigere : aetdre, tingere : ceindre, frangere : fraindre, 
 vincere : veintre, torquere : tordre, parescere : paroistre, cognoscere: 
 conoistre, crescere : croistre, carcerem : chartre, siirgere : sordre, 
 tergere : terdre, mulgere : moldre, cdlligere : cueldre. 
 
 Words that show a retention of the guttural are learned, thus : 
 socerum : suegre : suere, augurem : ogre, ancora : ancre, Mercu- 
 rius: merer e, Bulzarus : bolgre. The only exceptions are some 
 verbs which have undergone a change of conjugation ; they have 
 taken the ending of the fourth conjugation by analogy, maledl- 
 cere : male'ir, benedicere : beneir, colligere : cueillir. The first two 
 are moreover learned words. 
 
 Dental + r. Examples are very numerous. All show reten- 
 tion of the final syllable, claudere : clore, credere : croire, tadere: 
 rere, ridere : rire, rendere : rendre, perdere : petdre, presbytet : 
 prestre, mittere : mettre, duminterim, dementre : dontre (Leode- 
 gar) *butyrum ; burre, dexterum . destre, martyrem : martre. 
 
 The secondary group s-r intercalates d with retention of the 
 final syllable, consuere : cosdre, Lazatum : lasdre, *tisera (for 
 sicera, cf. Meyer- Liibke, Gr., I, 446) : cidre, miserum : mesdre. 
 Exception : the form proveir from presbyterum, generally pro- 
 vaire, is found in a late text, (cited by Godef. Psalms. Brit. Mus.) 
 The form desir, from desidero, or desiderem, is certainly due to 
 analogy, (cf. supra, dout.) 
 
 Eabial -j- r. Equally common piperem : poivre, pauperum : 
 povre, lupare : Louvre, sequere : sivre, asperum : aspre, vesperum : 
 vespre. The group Iv-r assimilates the v to the dental environ- 
 ment, solvere: soldre, volvere : voldre , pulvet em : poldre . 
 
 Sonant -j- - r - None of these groups occur as primary combi- 
 nations. Since r is the most sonorous of all consonants, the re- 
 
5i 
 
 tention of the final syllable is the rule. The transition from the 
 nasal and I to the more sonorous r is effected by means of a 
 glide-consonant. In every case r remains as the syllabic of the 
 ultima, cinerem : cendre, generem. : gendre, *manere : maindre, 
 ponere : pondre , cucumerem, ; concombre, numerum : nombre, *abe- 
 mere : *aviembre : aveindt e ; fremet e : friembre : freindre, *despri- 
 mere despriembre ■ despreindre, gemere : giembre : geindre, tre- 
 mere : criembre : craindre, *colyrus (for corylus) : coldre, molere .- 
 moldre, tollere : toldre, vellerem ■ velre, colere : coldre, etc. In 
 cammarus : jamble, sound-substitution in a loan-word has oc- 
 curred. 
 
 In the group h-r, in the infinitives struhere : struire, trahere : 
 iraire, c has been substituted for h in Vulg. I^at. according to 
 the analogy of the participles, stradus : tradus. See Grober in 
 Wolflin's Archiv, VI, 131. 
 
 A secondary combination of the same consonant could not need 
 a " supporting vowel." Therefore the retention of the ultima 
 in currere : corre, quaerere : querre, is surprising, (cf. soror:suer) . 
 They are probably due to analogy. Such infinitives as the 
 regular *cor ; *quer, would be entirely isolated in the older lan- 
 guage ; and so, according to the model of dorai : dore, to corrai 
 the infinitive corre was created. 
 
 Consonant -f- /. These groups, especially common in L,atin 
 in the ending -ulus seem to have been syncopated very early. 
 Indeed it is doubtful if the popular language ever possessed 
 -cuius, where the penultimate vowel arises through "svarabhakti" 
 in classical I^atin. Syncope must have been in general so early 
 as to admit of various processes of assimilation before the outfall 
 of the ultima. For I^atin examples of syncope before / see 
 Lindsay, p. 176 ; poplom, where the u is original, is met with in 
 Plautus. 
 
 Dental + /. -cuius is the regular representitive of -lulus in 
 Vulg. Iyatin. This secondary -cuius : -clus is treated like 
 other cl groups. Assimilation must have been completed very 
 early, with production of palatal / as the end-result. After this 
 homorganic sound, the final vowel falls as it does after a simple 
 consonant. vetulum : veclum : vieil, acetulunt, aisil : radula : 
 raille, situla : seille. . 
 
 In loan-words the ultima is always retained, while the / is 
 
52 
 
 never palatalized. They may be divided into several categories, 
 according to the period of their reception into the language. 
 
 (i). tl, dl>l, 11. {cf. Gutheim, op. cit, p. 44. This assimila- 
 tion could not have been completed when the reduction of the 
 finals occurred, modulum: moule, rotvlum : role, petalon : poele, 
 spatula : espalle, etc. 
 
 (2). The group is retained as such, or a substitution of the 
 more usual tr, dr occurs, idolum : idle : idre, utilem : utle, titu- 
 lum : title : titre : Wandalum : Vandle, capitulum : chapitle : chap- 
 ilre, apostolum : apostle : apostre, etc. 
 
 After the group ct-l, the final vowel seems to be retained, al- 
 though a palatal / is produced, ductilem. : doille. If this is reg- 
 ular, it would indicate a development, ductile : doctle : docfli . 
 doit'le : doile, the last assimilation not occurring until after the 
 fall of the ultima. 
 
 In *pestulum : pesle, simple reduction of the tri-consonantal 
 group has taken place, with / retained as the syllabic after j. 
 
 Guttural -\- I shows the same development in popular words, 
 as secondary clu from tuhcm. Examples are very numerous. 
 oculum : oeil, *piculum : peil, *dagulum : dail, soliculum : soleil, 
 etc. It should be observed that new feminine forms are fre- 
 quent, especially in the older language, for words in aculum, cf. 
 murail : muraille, cenail : cenaille, travail : travaille, ventail : ven- 
 taille, etc. 
 
 After the consonants n, r, the guttural seems to be retained ; 
 / then appears as the syllabic of the final syllable, angulum ; 
 angle, avunculum : aoncle, dracunculum : draoncle, circulum : cer- 
 cle, cooperculum : covercle. But the development in margila : 
 marie, excites a doubt as to whether cercle : covercle are genu- 
 inely popular, torculum : treuil shows early metathesis, calai- 
 lum : chail shows early syncope, with simplification of the result- 
 ing lcl>l'l > /'/>/', before the loss of the finals. 
 
 The group sc-l exhibits loss of the c with retention of the /, 
 as the syllabic of the final syllable after the .?. Is this popular ? 
 masculum : masle, musculum : musle, osmium ; oscle, vasculum : 
 vascle, are probably loan-words. Exceptions fall into several 
 categories, but all are loan-words, fragilem : fraile, gracilem : 
 graisle, show by the non-palatal / and the assibilation of the c in 
 graisle, that they are borrowed. *secale ; settle and segle. Other 
 
53 
 
 loan-words have a vocalization of the c, g to u, or retention of 
 the group, as such, saeculum : seule, siegle, siecle, *aboculum : 
 avuelle : avuegle, tegula teule .- Hule, *buculum .- bugle, cf. also 
 binocle, vesicle, vericle, spectacle, miracle (besides O. Fr. mirail), 
 etc. See Waldner : op. cit., p. 440. 
 
 The group s-l is preserved, and shows retention of the final 
 syllable, pensile : poisle, *grisulum ; grisle, pessidum .- pesle, (N. 
 Fr. pesne : pene). 
 
 Labial -4- /. Here a certain doubt arises as to the treatment 
 of the final vowel in popular words. The group presents, as is 
 well known, a double development. The first exhibits very 
 early reduction in Vulg. Latin of abulus to aulus : olus ; proba- 
 bly through the intermediate stages avulus : avlus, with gradual 
 cessation of the stop. The date of this vocalization is not 
 known, but as it occurs in other Romance languages, it must 
 have been early. The examples show generally a retention of 
 the final vowels, tabula : tole, stipula : esteule, populum ; poule, 
 peule, *gravulum : grolle, {cf. Meyer- l,ubke, Ztsckr.f Rom. Phil., 
 X, 172), mobilem : meule (generally mueble), Stabulae ; Etaules, 
 abilem : aule. But on the other hand, *parabolo -.parol, *para- 
 bolet • parolt, {Rol. 1206). The last two forms are open to the 
 objection of analogical influence, (cf. desidero : desir, for desire , 
 dubito : dout for *doude) . Still it is difficult to assume that in 
 words with such an early assimilatory process the final syllable 
 should be retained, grolle, meule, aule, may be due to contami- 
 nation with the other borrowed forms. A decision can hardly 
 be reached in the scarcity, of material. 
 
 The second, half-learned development shows retention of the 
 
 labial ; after this the / naturally remains as the syllabic of the 
 
 final syllable. debilem : doible, ebulum : ieble, capulum : citable, 
 
 flebilem .-foible, insubulum : ensouple, mespilum : nefle, etc. escu- 
 
 eil represents a Vulg. Latin *scoculus for scopulus. 
 
 After the secondary group m-l, the final syllable is preserved. 
 A glide-consonant b is generally intercalated in the majority of 
 the dialects, although in Picard and Walloon the group remains 
 as such, cumulum ; comble, simulo : semble, humilem : homble, 
 hunle, famulum : f amble : fanle. For n-l the sole example is 
 spinula : espingle, with final a. 
 
 More surprising is the preservation of the ultima after the 
 
54 
 
 group r-l. Examples : Carolus : Charles, gerulum : gerle, *oru- 
 lum : orle. Here again the peculiar nature of the Latin- French 
 / is seen. The synthesis of the dento-lingual r and the guttural 
 / could not be effected without a well-marked glide, which re- 
 tained a syllabic character, owing to the vocalic nature of the 
 consonants. 
 
 Mute -j- nasal. The group g-n shows the same development 
 as the corresponding primary group. The syncope is regular 
 and apparently early, and the resulting complex gn has under- 
 gone reciprocal assimilation, with production of a palatalized n. 
 After this simple consonant the ultima falls. Suffix -aginem, 
 -ain'va plantain, fusain, ondain, etc. rubigvnem : ruin, N. Fr. 
 rogne, * ad-vertiginem : avertin, etc. 
 
 All exceptions are either loan-words, with accent-shifting, 
 as in origine, etc., or are due to a change of suffix, as in origi- 
 nem : orine, caliginem : chaline, where the ending -iginam has 
 been substituted for -iginem. (cf. Cohn, Suffix-wandl, p. 175). 
 
 For g-n after consonants, I find one example, Langones . 
 Langres, where the more common grouper has been substituted 
 lor gn. We should expect secondary group c-n to show the 
 same treatment, a difference in the date of syncope being un- 
 likely. The only examples, however, exhibit assibilation of the 
 i, after which the n remains syllabic, as after original s. acinum: 
 aisne, cicinum ; cisne. Both are early loan-words, as L,at. z = i 
 shows. 
 
 For c-n after other consonants one example is found, circinum : 
 cerne ; here c must have been preserved in the triple consonant 
 group till after the reduction of the final syllable. 
 
 c-m. The regular development of this group is doubtful, ow- 
 ing to certain words with primary gm (all with a in the ultima), 
 where gm becomes urn, (fleume, piument, etc.) The secondary 
 group c-m, however, shows development of a parasitic i, with re- 
 tention of the final syllable, facimus : faimes, dicinius ■ dimes, 
 decimus : dismes, with .y from the analogy of dis. diaconum : 
 diacne : diacre is a loan-word, inguenem • aigne, N. Fr. aim; 
 in this word the triple group ng-n has apparently preserved the 
 final. 
 
 Dental -\- n, m. After this group, like the corresponding d, 
 t-\-r, the final syllable is preserved, although the dental has dis- 
 
55 
 
 appeared before the earliest literary monuments, abrotonum : 
 aurone, Redones : Raines, Rhodanum : Rhone, platanum : plane, 
 maritimum : marenne. After other consonants, cardinem : 
 charne, ordinem : otne, (more commonly ordre, loan-word), pec- 
 tinem : peigne, with palatal n, due to the preceding palatal group. 
 septimum : sedme, *octimum : uitme, *tetrinum ; terne. In this 
 word the reduction of tr-n to r-n must have occurred earlier than 
 intervocal tr> r; a dissillabic *tedrne is impossible, owing to 
 the sonority of the r. tudinem : tume, in costume, enclume, engue- 
 tume, etc. In these words the suffix -«««'»«« ox-umina has ap- 
 parently been substituted for udinem. cf. Cohn, Suffix-wandl., 
 p. 264 ff. Meyer-Liibke, Gr., II, 450. 
 
 5 + nasal. As the nasals are of greater sonority than the 
 spirant, the final syllable is retained. Whether the syllable 
 was reduced to syllabic sonant m, n, or whether the vowel was 
 preserved as e, is scarcely to be decided. Probably the relation 
 in the stress-group would determine the phonetic character of 
 the end-syllable in every case. In slower, more protracted pro- 
 nunciation a vocalic after-sound would be heard, sme, or sm"; in 
 more rapid speech this would be reduced to sm- asinum : asne, 
 resinum : reisne , fraxinum : fraisne, maximum : maisme, *pras- 
 simum : prasme, proximum : proisme, * seximum : sisme, esimum : 
 esme. 
 
 Labial -\- nasal. The labial was ultimately assimilated, al- 
 though this process was late, the earliest monuments showing 
 retention of the labial before n. After such a complex, the final 
 syllable is retained, juvenem : juefne : juene, Stephanum : Es- 
 tiefne : Estienne, *novimum : nuefme, avimus : ames, galbinum : 
 jalne, *carplnum : charme. Some loan-words show substitution 
 of the more frequent group, labial + r, for the original ; pam- 
 panum : pampre, timpanum : timbre, *cofinum : coffre. The 
 secondary group, l-m, shows the same line of development as 
 the primary. The final vowel is retained, calamus : chalme, 
 regalimen (for regimen, see Cohn, Suffix-wandl, p. 64) .• royaume. 
 
 r + nasal exhibits a remarkable variation from the corres- 
 ponding primary group, if we may judge from the two exam- 
 ples ; eremum : erme, erlmus : ermes. But for r-n, Turones : 
 Tours. Meyer-Liibke (Gr., I, 261) cites erm, but I am unable 
 to find examples of this form. It is, however, hardly a populai 
 
56 
 
 word, ermes may follow the analogy of other first person plurals. 
 In the lack of other material, and in view of the surprising dif- 
 ference from primary rm, I am indisposed to assume as a law 
 that after r-m the final vowel is preserved. 
 
 m-n. Like the primary group, after secondary m-n the final 
 vowel seems to be in general retained ; but here also some ap- 
 parent exceptions are found. Examples. — hominem : ome, canni- 
 mim : charme, *abominum : abosme, terminum : terme, viminem : 
 vime, legume, bitume, etc., are learned words. On the other 
 hand, dominus : dames and dans , Jarciminum : farcin. 
 
 The last probably represent the effect of the action of the re- 
 lations in the stress group. It is not improbable that doublets 
 existed originally in every case, although the reasons for the 
 preservation of the longer forms are difficult to discover, dame 
 and dans both show in the accented vowel that they were treated 
 ltke a proclitic, but this does not remove the difficulty, the laws 
 for the retention of the protonic vowel being the same as those for 
 the ultima ; cf. the doublet damoiselle, dancelle. But we shall see that 
 double-forms are found much more frequently in protonic sylla- 
 bles than in finals. The prevalence of the longer forms with 
 these words may be due to the greater frequency of initial con- 
 sonants in following words. That, however, such doublets 
 really existed, even in the case of non-proclitic substantives, is 
 proved by farcin. The final consonant n, for regular m, is due 
 to the analogy of the earlier nominative form with s, as in dans. 
 
 The reverse group, n-m, is found only in two examples, mini- 
 mum : merme, Hieronymus : ferome. The group shows regres- 
 sive assimilation, with retention of the final vowel. The same 
 causes would operate to preserve the ultima, as for m-n. Till 
 the ^-articulation ceased entirely, the following m must have 
 been heard as m, or me. And, as with mn, this dissyllabic 
 character was retained after the reduction of the final syllables 
 in general. 
 
 PART III. 
 Development of the Pretonic Vowels. 
 The general principles governing the weakening and outfall 
 of the vowels preceding the tonic accent are the same as those 
 
57 
 
 for the post-tonic. This is the main result of Darmesteter's 
 essay already cited, and has been accepted by all later investi- 
 gators. After the secondary accent which (at least in words of 
 only two pretonic syllables) always rested on the initial syllable 
 of the word, as after the chief ictus, unaccented syllables with 
 the vowel remain ; all others are subjected to syncope. I shall 
 take up first only those words (the great majority) which have 
 but two pretonic syllables, like bonit&tem, reserving those where 
 the chief accent is preceded by three or more syllables. In the 
 latter, the consideration is complicated by the different views 
 prevailing as to the position of the secondary accent. 
 
 Pretonic syncope was not common in classical Latin, most of 
 the apparent cases being due to the action of the earlier accent 
 law, according to which the chief ictus rested on the initial syl- 
 lable. (4/". Lindsay, 171, ff). During this earlier period, e, t, 
 in an open syllable immediately following the accent were, as a 
 rule, syncopated. Examples are, *ambi-colos : anculus, *formi- 
 ceps : forceps, *vini-demia : vindemia, aperio-apricus , Aprilis. 
 {cf. Lindsay, p. 178.) 
 
 Examples of true pretonic syncope under the later accent law 
 are rare ; such are, possibly, aridus, but ardere, ardorem, bene- 
 ficium : benficium, cale-facere ; calfacere, discipulus : discipllna, 
 figulus ; figlina. 
 
 In Vulgar Latin such shorter forms were undoubtedly more 
 numerous,, though the development of the non- Gallic Romance 
 languages shows that this pretonic syncope was far from being 
 so general as the post-tonic in proparoxytones. In Italian, for 
 example, such syncope occurs only after liquids and nasals, 
 and between a few other consonants like s-t, s-c, d-c. cf. Ital. 
 cervello, belta, bonta, costuma, riscare, dozzina, andare, etc. (See 
 Meyer- Liibke, Gr., I, 275, and Ital. Gr., p. 85.) In the exam- 
 ples from the inscriptions collected by Schuchardt, syncope is 
 most common in the protonic of words with only two pretonic 
 syllables, one of which contains a liquid. ( Vocalismus , II, 423 
 ff.). The most interesting of Schuchardt's examples, in refer- 
 ence to French, are Domnin, cf. O. F. dancel, maldicto, cf. O. F. 
 maldit, pulcare, cf. O. F. espulcer, vercundus cf. O. F. vergogne, 
 virdiario, cf. O. F. vergier. ■* 
 
 Accepting the suggestion of Darmesteter that the tonic accent 
 
58 
 
 divides the word into equal parts, we find that the initial sylla- 
 ble always persists, and that the second, distinguished as the 
 pretonic, is usually syncopated. There is no doubt that the in- 
 itial syllable had a secondary accent, probably preserved from 
 the original accentuation (cf. Lindsay, p. 159). In this case 
 we should expect that the pretonic syllable would have in gen- 
 eral the same fate as the ultima in paroxy tones. The protonic 
 vowel is, in almost all cases, the final sound of the syllable, the 
 succeeding consonant being drawn over to the following, where 
 it was possible. (I leave entirely out of consideration the cases 
 where in classical Latin the protonic syllable was formed by i or 
 u in hiatus, as in medianus, Januarius. Such vowels were re- 
 duced to consonants at a very early date, and can never be 
 counted as a syllable in Vulgar Latin. ) Before a single conso- 
 nant, then, or before a group which could commence the follow- 
 ing syllable, according to the laws of syllabic division in Latin, 
 we should expect, a priori, the reduction of the preceding pro- 
 tonic syllable to be regular and complete in every case. As, 
 however, the principles of Latin syllable-division presented by 
 the grammarians, have been disputed, I will reserve the in- 
 stances where the protonic stands before a group of consonants 
 for separate discussion. 
 
 The regular and effective action of the law of protonic syn- 
 cope is very often disturbed by various analogical influences. 
 These effects are naturally more conspicuous in the ease of the 
 protonic syllable than in that of the ultima. All derivatives of 
 adjectives from substantives (cf. amoros : amor) or the reverse 
 (cf. amertume ; amer) show unmistakably the results of these 
 influences. So also all verb forms, where the stem-accented 
 forms are generally levelled for the whole system. This princi- 
 ple has already been fully recognized by Darmesteter and others, 
 and it is not necessary to insist upon it here. I shall have oc- 
 casion many times to call attention to the results of analogy in 
 this discussion. In general, I shall leave out of consideration 
 all words which show plainly a levelling process, like amoros, 
 for example ; where the simplex is present, it is unnecessary to 
 call attention to its influence on the derivative, or to consider 
 the latter as a real " exception" to the laws of protonic syncope. 
 Furthermore, all words which are due to the creative power of 
 
59 
 
 the French language itself, which are not derived plainly from 
 Vulg. L,at. substrata, but which are made with the help of 
 various suffixes during a later period of development, need not 
 be considered in this discussion. They will be taken into ac- 
 count only so far as they exhibit the action of the phonetic laws, 
 or show striking deviations from them. 
 
 A further limitation must be established for most words com- 
 pounded with prefixes, prepositions, etc., especially verb-forms. 
 The prefix has been treated in nearly all cases as an indepen- 
 dent word ; the verb-stems after it underwent the usual develop- 
 ment without regard to the form of the foregoing prepositions. 
 Only such words as show evidently that the feeling for the com- 
 pound had been lost in Gallic Vulgar Latin, like amputate : con- 
 ter, *ab-secare : oschet , are entitled to be considered as simple 
 words. 
 
 Of great importance in the history of the pretonic vowels is 
 the theory of sentence-doublets, especially as applied to those 
 forms which contain one of the vowel-like consonants before the 
 tonic vowel. I have already shown, in connection with the final 
 syllable, that the liquids and nasals, owing to their peculiar 
 nature, are capable of functioning either as vowels or conso- 
 nants. They are true syllables, and from their voice an indefi- 
 nite vowel, especially in slower, more deliberate utterance is 
 often developed. Conversely, in certain cases, where there is 
 in the stress-group a quicker, more rapid pronunciation, there re- 
 sults absorption of the unaccented vowel by the vowel-like con- 
 sonant, which in its turn, if the surrounding sounds favor it, re- 
 sumes the r61e of a pure non-syllabic. These two processes 
 work reciprocally, and each reacts on the other. That at any 
 period doublets existed for all words with a protonic liquid or 
 nasal is difficult to affirm definitely. Such might have been 
 the case for the unconscious, spoken language, and the by-forms 
 have left unmistakable traces in the orthography even in modern 
 French. Compare for instance the modern souverain (superanum, 
 O. Fr. rarely sovrain), with the modern serment, earlier sairement 
 {sacramentum) . Both apparently contradict the strict law of 
 syncope, one in the retention of protonic e, the other in the loss 
 of protonic a. We shall see that such doublets have left greater 
 traces in the O. Fr. orthography than is generally supposed. 
 
6o 
 
 The physiological causes on which these apparently contra- 
 dictory tendencies rest, are easily apprehended. The syllable 
 with the tonic accent received under the Gallo-latin accentua- 
 tion a compound " zweigipflige " accent, as a result of the 
 greater stress given to it, which ' resulted in the corresponding 
 reduction of the unaccented syllables. From this original 
 accent character has been explained the spontaneous diphthong- 
 ation process prevailing in Old French. When now such a syl- 
 lable contains a consonantal liquid or nasal, and a compound 
 stress accent is given to it, the single sounds of the syllable may 
 be so divided that one of the moments of greatest stress 
 ( ' ' Nebengipfel ' ' ) falls on the liquid or nasal. If this moment 
 of greatest stress falls on the beginning of the formation of the 
 liquid or nasal, then will follow a period of slighter stress before 
 the second " Nebengipfel " falling on the vowel of the syllable. 
 Now let these moments be more sharply divided ; the first part 
 of the liquid then becomes semi-syllabic in character, while its 
 close remains consonantal, functioning as the non-syllabic of the 
 weak syllable thus formed. If then the narrowness of the /, r 
 contact be widened, a syllabic vowel-glide is easily developed 
 between the foregoing consonant and the following /, ? , which 
 may be widened till a complete syllable with a new vowel re- 
 sults. Whether in French, however, this "svarabhakti " ever 
 went farther than the development of a sonant liquid or nasal 
 from original consonantal /, r, m, n, is doubtful; the orthog- 
 raphy er, el, etc., may at all periods have expressed nothing 
 more than /, r, (soverain = sovrrain. The development in the 
 case of serment is exactly the converse of this process. Here 
 the stress of the original vowel was transferred to the preceding 
 liquid (sairrment) , which in turn lost all syllabic function, as it 
 stood after a syllable of weak stress, {cf. for all these processes 
 Siever's Phonetik, p. 270 ff.) Probably at first all these devel- 
 opments have been conditioned by the position of the syllable in 
 the stress-group ; if it had a more independent character, with 
 greater stress, then " svarabhakti " resulted; if, on the other 
 hand, it stood in closer union with neighboring syllables, and 
 had lesser stress, then non-syllabic /, r, n would be the rule. {cf. 
 the explanation of the future forms donrai, menrai, etc., by 
 Neumann: Littbl.f. Rom. u. Germ. Phil., Ill, 467). 
 
6i 
 
 This explanation, whereby the non-syncopated (with original 
 e , i, o, u) , and syncopated (with original a) , doublets are classed 
 as mutually reciprocal forms due to relations of sentence-accents 
 and stress in the French period, harmonizes with the general 
 theory of syncope. I would assume that for pretonic e, i, o, u, 
 the syncope was originally complete. This accords with the 
 theory of Darmesteter, who assumes that the shorter forms are 
 the primitive and normal, and that the longer are due to a later 
 " adoucissement." The longer forms, like soverain, are then 
 due to later relations of accent in the stress-group, which proba- 
 bly prevailed to a greater or less extent at all periods of the lan- 
 guage. The grammarians of the sixteenth century attest that 
 protonic * ' feminine " « in the neighborhood of /, r, was but 
 weakly pronounced, and give many examples of doublets, with 
 and without syncope. Such are, for example : ouvriere : ouve- 
 riere : ouverage, chevereul : N. Fr. chevreuil, chevron : cheveron, 
 lappreau : lappereau, couretier : courtier, carreler : carter, hobe- 
 reau : hobreau, horeloge : horloge, ourler : ourreler, armelin : 
 armlin, etc. cf. Thurot, La prononciation franc , I, p. 143 ff). 
 
 In view of these peculiar relations, I propose to treat those 
 words and word-forms with protonic liquid or nasal in a separate 
 section. These processes, svarabhakti or syncope in the neigh- 
 borhood, of the vowel-like consonants, are found in-all languages. 
 The effects, however, as in French, are manifoldly disturbed by 
 levelling and analogy. For examples of Svarabhakti (Anapty- 
 xis) in the Indo-European languages, see Brugmann, Grund- 
 riss, I, p. 469 ff. Examples which are strictly comparable with 
 the absorption of a vowel before or after /, r, n in French, are 
 found in many words in New High German, where the initial 
 vowel has been lost before a liquid or nasal, as in 
 
 Glaube : O. 
 
 H. G. giloubo, 
 
 gleich : " 
 
 " " gilih, 
 
 glied : " 
 
 " " gilid, 
 
 Glimpf : " 
 
 " " gilimpf, 
 
 Gluck : M. 
 
 H. G. gelucke, 
 
 Gnade : " 
 
 " " ginade, 
 
 grob : " 
 
 " " gerob, 
 
 bleiben : " 
 
 " ■"' biliben, 
 
 draus : 
 
 daraus, etc. 
 
62 
 
 Examples of the " svarabhakti " vowel, which occur in Old 
 High German, are to be distinguished from the cases in French 
 in that this development occurs at the end of the accented sylla- 
 ble, instead of the beginning, cf. for instance Old High German 
 alah, beraht, from alh, berht, with Fr. sovrain : soverain. 
 
 In Latin such double forms must have existed in great num- 
 bers, although as a rule the conservative literary language has 
 preserved but one of them. I have already cited the develop- 
 ment of a " svarabhaktic " vowel in the suffix -cuius, I. E. tlo. 
 The frequency of the ' ' syncopated ' ' forms in the inscriptions 
 (cf. Schuchardt : Vocalismus, II, 404), indicated for the popular 
 language the prevalence of these doublets. Other examples are 
 populum, Old Latin, poplom, and especially in Greek loan-words 
 mina, Gk. /.ira ; techina, Gk. rix^rj, drachuma, Gk. Spax^rj. 
 Notice moreover Latin developments like *en-cri-tas, en-cr-tos, 
 incertus, *acridhos: acerbus, *facli : faculi, *stabli : stabuli, *pend- 
 los : pendulus. (cf. for these forms Brugmann, Indg. Forsch, IV, 
 225 ff.) 
 
 In Vulgar Latin such examples increase in number, and are 
 found side by side with the original forms, without the " svara- 
 bhaktic " syllable, (cf. Schuchardt, Vocalismus, II, 398 ff. ; 
 Lindsay, pp. 71, 93, 145, 175; Stolz, Lat. Gram., p. 277). 
 Compare moreover inscriptional forms like magist(a)ratum, 
 libfejras, (=libras), confijsul, nucCujleum, T(e)rebonio, cfijrib- 
 rum, cfajlaudianus, Mith (a) ridaticis, etc. 
 
 The Vulg. Lat. accent shifting in original proparoxytones 
 with mute + liquid in the final syllable is to be traced back to 
 this development of a "svarabhaktic" syllable ; thus tonitru be- 
 came t5nitrru, after which the accent was shifted in accordance 
 with the prevailing accentuation to tonitrru, from which Fr. 
 tonnerre. And the diphthongation of accented vowels before mute 
 + liquid, as vapatrem : pere, has been ascribed by Meyer- Liibke 
 to the same cause, (cf Rom. Gr., I, 251). I have already 
 called attention to the fact that the opposite process, syncope in 
 the neighborhood of liquids and nasals, is common in Vulg. 
 Lat. Thus of the examples of pretonic syncope given by 
 Schuchardt, pp. 421-443, eighty-five are instances of vowel- loss 
 in the neighborhood of a liquid or nasal, only ten in the neigh- 
 borhood of pure consonants. 
 
63 
 
 The traces that these processes have left in French will be 
 cited in detail later. Here, however, I may call attention to 
 instances where the development of a " svarabhaktic " vowel 
 from the voice of the vowel-like consonants is clear and unmis- 
 takable. Such are, for example, many loan-words from the 
 Teutonic languages with initial mute + liquid or nasal. 
 
 Low German slop : French salop, 
 
 Dutch sloep : " chaloupe, 
 
 Dutch smak : " semaque, 
 
 I<ow German snau : " senate, 
 
 High German schnapphahn : " chenapan, 
 
 Teutonic hnapp ,- " hanap, 
 
 Teutonic hring : ' ' harangue, 
 
 Anglo-Saxon hripan : ' ' hureper, 
 
 Modern High German kliben : " galipot (?), 
 
 Old Norse knifr : ' ' canif, 
 
 Modern IyOw German knijpe ' ' guenipe, 
 
 Swedish vranger : " vatangue, 
 
 High German willkommen : " wilecome. 
 
 Note also Gk. yva8os : granache, Gk. nparroa ■ bareter. (cf. 
 Korting, Lat. Rom. Wbrterbuch). Arab, almatrdh : materas, 
 Pers. Kaschmir : cachemir. 
 
 I pass now to the consideration of the various protonic 
 groups. 
 
 Words with protonic a. As in the case of the ultima, protonic 
 a always persists in Old French under the form of " mute" e. 
 For examples see Darmesteter. I shall consider here only such 
 words as exhibit apparent "exceptions" to this phonetic law. 
 As in the case of final a, the preservation of protonic a is to be 
 explained through its greater sonority, as contrasted with the 
 other vowels. 
 
 All words with protonic a retained as such, not weakened to 
 e, are either borrowed, or due to some analogical influence. 
 Examples of the first class are creatorem : criateur, apparatus : 
 apparat, fondatorem : fondateur, cyparissus : ciparis, also O. Fr. 
 ciperis, regular, panaricium : panaris, * cingulatorium : cingladoir 
 from the Provencal. Ital. splanata : esplanade, etc. 
 
6 4 
 
 Examples of the last class, to choose a few out of many, are 
 caballarium ; chevalier, cf. cheval, *companaticum : companage, cf. 
 compaign, *ex-sarritare : essarter, cf. essart, *coraticare : coragier, 
 cf. corage, * battaliarium : bataillier, cf. bataille, etc. 
 
 Here I may call attention to the fact that several words which 
 in New French are treated as compounds, in the older text often 
 show the development of simple words. For example, M. F. 
 apanage is a derivative of the old verb apaner, which goes back 
 to a V. L. *ad-panare : apaner is the form of the compound word, 
 panare being the simplex. But, besides these, we find (cited 
 by Godef. from Amadis de Gaule) apener and apenage (from a 
 text of 1297 in Godef.), where the preposition has united firmly 
 with the verb and the a, become protonic, is regularly weak- 
 ened. Compare also the old form compenage (Renart) beside 
 companage, the last due to analog}', the first regular. 
 
 It is surprising that certain words in O. Fr. which are com- 
 pounded with the apparent prefix *cata (from the verb capiare : 
 catar, according to Kbrting, Lat. Rom. Wbrterbuch. This verb 
 does not exist in French, except in the compound acheter, which 
 shows regular weakening) , preserve as a rule the second a un- 
 weakened. *cata-lectum : chaalit or kaalit, cata-*faltum (from 
 Ger. balko) : chaafalt (also more regular chadefalt) . An expla- 
 nation is difficult, unless we suppose that this prefix was bor- 
 rowed. Another difficult form is * scarabellum : escharavel, N. 
 Fr. echaraveau. It shows throughout a regular development, 
 except for the protonic vowel alone. Is this also a Prov. loan- 
 word ? adamantem : aimant, with change of protonic a to z, is 
 hardly a popular word. cf. Thomas I.e. 
 
 I take up now the consideration of words with protonic a, 
 which show a complete loss of the last. This occurs in the 
 neighborhood of a liquid or nasal, and is to be ascribed to the 
 tendency to absorption of the vowel-element by the vowel-like 
 consonants in a more rapid tempo of pronunciation. In most 
 cases, the shorter forms are the younger, and are first formed in 
 the specific French development ; in others {vaslet, etc.) the 
 syncope has existed from the earliest period. Doublets are 
 present in the majority of instances. 
 
65 
 
 Examples, ambas-partes : ampars, which may be influenced by 
 the form andous, which in later O. Fr, was used for both gen- 
 ders, sacramentum . sairement : sairment, which first appears in 
 the Thirteenth Century ; and its compound aserementer : aser- 
 menter, etc., alabastrum .■ albastre : alabastre, also late, but</. the 
 Vulg. Lat. albastro, in Schuchardt's Vocalismus, II, 427. *alba- 
 retum : alberoi (Amis et Amiles, 1099), albroi (from a text of 
 1344 in Godef.); here the shorter form is the earlier. Gr. n parr 00 : 
 bareter : barter (cited by Godef. from a text of 1373) *arca-balis- 
 tum : arbelest : ablett (cited by Godef. from Scheller Lexiq, 95), 
 atramentum : airement : airment, illalamella : alamelle (for la 
 lamelle) . almelle (cited by Godef. from G. Durant). cuppa -f- 
 lettus : coupelet : coplet ( Aliscans) . cupa-retta : cuverete ■ cuvrete 
 (cited by Godef. from a text of 1462), *denariata : deneree : 
 denree ; comparare : comperer : comprer, separare : severer : sevrer ; 
 these may be due to the analogy of the stem-accented forms, 
 comparo : compre cf. Meyer- Liibke, Rom. Gr., II, 620). Darme- 
 steter assumes a Vulgar Latin seperare ; in this case the words 
 should be placed in the category with " svarabhaktic" protonic. 
 *curatarius : coretier : cortier. cf. Horning (Ztschr.f. Rom. Phil., 
 XIII, 325), who regards it as a derivative of currere. But 
 coretier and Prov. coratier favor the etymon of Diez ; likewise 
 coreterie : corterie, * ' excaracionem : escharecon : escharcon, quar- 
 tarantia : *quarterance : quartrance (cited by Godef. from a text 
 of 1274), so quarteron : quartron, quarteruel : quartruel, etc. 
 *fabariola : faverolle : feuvruelle, * deretranarius : derenier : der- 
 nier (should give *dertrenier, cf. infra deerain). ipsa-illa-hora : 
 epslor, mirabilia : mereveille : merveille. The first is very rare. 
 Darmesteter proposes *miribilia by the analogy of mirificus, but 
 the form mereveille contradicts this. cf. Meyer- Liibke, Rom. 
 Gr., I, 275. Compare further the derivatives mereveillable (Gir. 
 de Ross), mereveilliere (Benoit), sinapillus : seneveil : semveil, 
 cf. Cohn, Suffixwandl, p. 51). *balsamare : balsemet : balsmer. 
 This may be influenced by the stem-accented forms balsamat : 
 balsme, calamellum : chalemel : chalmel, cf. N. Fr. chalumeau, 
 regular, *calaminus : chalmin, never *chalemin. *fimaretum : 
 femeroi : fembroi, *vassalettum : vaslet, never vasselet, but cf. 
 vasselage. 
 
66 
 
 ferramentum : ferrement .-ferment, (cited by Godef. from a 
 text of 1391), *fortalitia ■: forteresse (also N. Fr.) : fortresse. 
 Here the shorter form is earlier (Dial. St. Greg, in Godef.) 
 *loraminarius : lotemier : lormier, (Dit. de la Queue de Ren. 
 in Godef.) Possible this example does not belong here ; it will 
 be discussed later under those with three pretonic sylla- 
 bles. Kelt, garra, therefrom .jarretier : gartier (Froissard), 
 and the verb esjareter : esjarter, (chart of 1474 in Godef.) 
 O. H. G. masar : masdre ; as derivatives maderer : madrer, 
 maderiner : madriner ; the shorter forms may have been influ- 
 enced by masdre. * putidarellus : puterelle : putrelle (Nic. de 
 Troyes in Godef.), hasta-rellus : hasterel : hastrel (Warrin, 
 Chronic d' Angl in Godef.) ; *materiamen + aticum : mairen- 
 age : marnage, (cited by Godef. from texts of the Fourteenth 
 Century) ; materiamen + arius : marrenier : marnier, (cited by 
 God. in a text of 1507); mappa : nappe, therefrom naperon : 
 napron, (cited by Godef. from the Recreat. des divers amours) ; 
 *paramentum .-perement : pairment, and its derivative parementer : 
 parmenter, (from a chart of 1241 in Godef.) ; *saccalettum : 
 sachelet : saclet, (from a text of 1268 in Godef.) ; costa-*rettum : 
 costeret : costret, (cited by Godef. from a text of 1458) ; *thymia- 
 monium : timoine; *terraticum -f- are : terragier : tergier, (cited 
 by Godef. from Picard texts of the Twelfth Century). The 
 shorter form is here the earlier. It is interesting as showing 
 that, even when sustained by analogy, a could be absorbed. 
 
 *bovariolum .• N. Fr. bouvreuil, calidaria : chaldiere, therefrom 
 ckalderon : N. Fr. chaudron. Such double forms are especially 
 frequent in words with the accent suffix ia, which has been added 
 to substantives in arius. A new suffix, erie, is formed thereby, 
 {cf. Meyer- Liibke, Rom. Gr., II, 453), the longer, unsyncopated 
 form of which is generalized in New French. But in the old 
 language we find correterie : cortrie, (from a text of 1281 in 
 Godef.) from core tier ; fusterie .-fusttie, (Froissard) from fustier ; 
 fovrerie : fourrie, (from a text of the Fourteenth Century in 
 Godef.) from fouvrier, courrerie ; courrie, (texts of the Fifteenth 
 Century in Godef.) from courrier ; avouterie : avoutrie, (Frois- 
 sart) from av outier ; battetie : battrie, (cited by Godef. from a 
 text of the Thirteenth Century) from battier; plorerie : plorie, 
 
67 
 
 (Gaufrey in Godef.) from plorier ; poeterie .- poehie, (cited by 
 God. from Boece, de Consolation) ; rusterie : rustrie (Rabelais) 
 from rustier; sarreterie : saretrie (Rabelais) from sdrretier ; nov- 
 ellerie : novelrie (from an Angl. Fr. text in Godef.) from novel- 
 lier ; maladerie : maladrie (Trahiss. de France, in Godef.) ; 
 mananderie : manandrie (Chev. au Cygne, in Godef.) from 
 manandier. 
 
 The same phenomena are met with in words with the 
 suffix N. Fr. -eresse, the feminine to -eor, {cf Meyer- Liibke, Gr., 
 II, 414) ; the e in eresse represents the stem-ending a generalized 
 in a-torem, a-trissa, etc. Chanteresse : chantresse, corteresse : cor- 
 tresse (from a Bible transl. of the Fifteenth Century in God.); 
 donateresse : donatresse, (text of 1472 in God.); orfeveresse : 
 orfevresse, (text of the Sixteenth Century in God.); revenderesse : 
 revendresse, (text of the Fourteenth Century in God.); semoner- 
 . esse : semonresse , (text of 13 12 in God.); flateresse : flatresse, 
 (D'Aubigny in Godef.); fondateresse : fondatresse ; moiteresse : 
 moitraice, (texts of the Thirteenth Century in God.); recom- 
 anderesse : recomandresse \ (text of the Fifteenth Century) . For 
 other examples in names of place, cf. I^indstrom, op. cit., p. 62. 
 
 It will be seen from the above examples that this syncope, or 
 absorption, in words with liquids, is generally late, belonging to 
 a distinctly French period of development. Nevertheless, some 
 words (vaslet, merveille, fembroi, etc.), go back to the earliest 
 texts, showing that the tendency was present from the begin- 
 ning, {cf. Meyer- Liibke, Gr., I, 275). It is probable that many 
 more of these doublets existed than the fixed orthographical tra- 
 dition admitted into the texts. We shall see that the same vari- 
 ation exists in words with protonic e, i, o, u, though here the 
 syncope was probably original and regular. The shorter forms, 
 with syncope of e, from a, are undoubtedly due to the position 
 of the word in the stress-group, and to the tempo of the spoken 
 language. To the same principle is due the shortened form of 
 futures of the first conjugation, like donrai, menrai, larrai, etc. 
 {cf. Neumann, I.e., Bnd differently, Korting, Formenbau des 
 franz. Verbums, p. 257) . Compare also the iormfrai, iorferai, 
 itself due to a reduction oifarai, in the stress-groups fhrdi {cf. 
 G. Paris, Rom., XXII, 570.) 
 
68 
 
 All other words without liquid or nasal, which show syncope 
 of protonic a, are probably due to analogy. (I do not take into 
 account those words with pretonic e <C a, in hiatus which 
 have undergone a comparatively late reduction. For these see 
 Hossner, Zur Geschichte der unbetonten Vocale im Alt- und Neu- 
 franzbsischen. Diss. Freiburg, 1886.) Such examples are very 
 rare. I may cite : *gabatellum : jadeau, which may go back to 
 a Vulgar Latin *gautellum; cf.gabata : *gauta : joue, though in 
 that case the retention of the dental is difficult to account for ; 
 more probable to me appears Merkel's {Die germ. Elemente in 
 der franzosische Sprache, p. 57), etymon, Frank, gabita ; a *ga- 
 bitellus would give regularly jadeau. 
 
 *usatile : ostil. This word is otherwise irregular (initial u >• 
 0). See also Meyer- Liibke, Gr., I, 278, who sets, as the Vul- 
 gar L,atin ground form *usitile. Perhaps it has been influenced 
 by the related verb ostillier (possibly < *usatiliare) , where the 
 fall of the a in the second pretonic syllable may be regular {cf. 
 infra) . 
 
 Interesting is the doublet puisse-di : puisdi ; in this compound 
 the first element is evidently postea. It supports Karsten's (l.c) 
 theory of the origin of puis (ante- vocalic form of postea) in op- 
 position to Schuchardt's, {Ztschr. Rom. Phil., XV, 240), who 
 assumes Vulgar Latin postius as the etymon ; monasterium : 
 mostier. Darmesteter assumes a Vulgar Latin form *moniste- 
 rium, which has followed the analogy of ministerium ; this ex- 
 planation has been generally accepted. 
 
 A last question connected with protonic a concerns the date of 
 the weakening to e. In words with the suffixes -tionem, 
 -sionem, etc., in which the palatal group develops a parasitic i 
 before the consonant, the a of the stem generally appears un- 
 weakened, and unites with the parasitic a to form the diphthong 
 ai. Examples are very numerous. — occasionem : ochaison, vena- 
 tionem : venaison, ligationem : liaison, separationem : sevraison, 
 etc. {cf. for this suffix, Meyer- Iyiibke, Gr., II, 539. Cohn, 
 Suffixwandl., p. 123 ff.) 
 
 Is this conclusion correct ? We have no data for the chro- 
 nology of the weakening of a ~^> e, nor for that of the develop- 
 ment of parasitic i, both being antecedent to the earliest monu- 
 
6 9 
 
 ments. (On a for e in the Strassburg oaths, cf. supra.) Unac- 
 cented a is not weakened in Provencal or Franco-provencal, 
 nor in the extreme southwestern French dialects, phenomena 
 which indicate a comparatively late date. If the reduction had 
 not been completed at the time when parasitic i was developed, 
 then the resulting diphthong, with its greater sonority, would 
 probably escape the weakening which affected simple a. Never- 
 theless, there are some French forms which seem to contradict 
 this hypothesis. Especially interesting is the word *bellatiorem . 
 bettezour of the Eulalia. If the above supposition be correct, we 
 should expect *beUaisor, a form which is not found, though later 
 we find belleisor, etc. (cf. Koschwitz, Commentar, p. 70). More- 
 over, in certain forms with protonica before intervocalic palatals, 
 which is treated exactly like ti, we find the same weakening ; for ex- 
 ample: Bellovacensis : Belveisis : Beauvoisis, Cameracensis : Cam- 
 broisis, also cf.palatiu-\-inus:palesin, beside/a/aw'??, a learned form. 
 How is this apparent contradiction to be explained ? I believe 
 that it arises from the fact that we have here to deal with a suf- 
 fix. Thomas (I.e.) has shown that, for Provencal at least, some 
 suffixes had a quasi-independent existence ; and underwent the 
 same modification as simple words. In that language the 
 suffixes -amentum, -ementnm, -imentum, having become respec- 
 tively -amen, -emen, -imen, were preserved in this form and 
 attached to verb-roots. I shall consider this hypothesis 
 more in detail in connection with other suffixes ; here it seems 
 to remove a difficulty. When -tionem is attached to stem- 
 ending in e or i, instead of the regular syncope (as in boisson, 
 parcori) in the majority of cases an uncontracted suffix-form 
 -oison, -ison appears. This can only be explained with the aid of 
 Thomas' theory. Applying this hypothesis to a-stems, if the 
 suffix -ationem, generalized for stems of the first conjugation in 
 the same way as -amentum, -atorem, etc., in Prov. developed in- 
 dependently, we should expect this (protonic) a to be treated 
 like initial a; and this is the case. (cf. rationem : raison) . 
 Moreover, the form -aison is not the only one which appears at- 
 tached to verb-stems of the first conjugation; by-forms in -oison 
 are much more common and wide-spread ; -ison appears to be 
 
7 o 
 
 mainly confined to the Eastern dialects. I have collected the 
 following examples of by-forms in oison that are found with 
 original «-stems. 
 
 abitoison, acordoison, afaitoison, airoison, arestoison, assembloison, 
 avooison, chacoison, chaploison, charpontoison , coltivoison, consir- 
 oison, coutoison cornoison, cortoison, costoison, covroison, crimson, 
 delitabloison , delivroison, demoroison, dementoison, desbaretoison, 
 desertoison, doloison, donoison, dotoison, enchaitivoison, enchantoi- 
 son, encombroison, envenimoison, escusoison, esgardoison, fioison, 
 foloison, formoison, getoison, livroison, mangeoison, membroison, 
 mostroison, ploroison, rovoison, semoison, sevroison, trenchoison, 
 vengeoison. 
 
 Of by-forms in -ison I have found the following : araisnison, 
 atargison, enchalcison, espargnison, herbergison, membrison, meii- 
 rison, mostrison, plorison, vengison. (For citations see Godef.) 
 
 The by-forms in -oison may be regular (cf. Beauvoisis, etc.), 
 with weakening of a to e, before formation of the diphthong 
 ei-oi ; but this does not account for the forms in -ison. It seems 
 perhaps simpler to assume that the forms peculiar to the differ- 
 ent conjugations have been confused and appear indifferently, 
 without regard to the original ending of the stem. This view is 
 confirmed by the fact that aison is found with stems of the 
 second, third, and fourth conjugations, where it must be due to 
 the analogy of stems of the first. See Cohn (Suffixwandl, p. 
 123, ff.), who admits the possibility of a regular reduc- 
 tion of -aison to -ison (with reference to Neumann, Laut. und 
 Flexionslehre, p. 53), but in view of the fact that such reduction 
 is never met with except for these suffixes and for verb- endings 
 {cf. raison, saison, etc., always with ai) this supposition is little 
 probable. 
 
 Some words which have this, or a similar suffix, but by the 
 side of which we find no corresponding verbs, seem to present 
 difiiculties. For example : occasionem : ochaison : ochoison : oche- 
 son : ochaison, etc. *lunationem : lunaison : lunoison. corbison, cf. 
 corbe (from *corba for corbem) . guambaison, guambison, (from 
 Teut. wambeis.) The variation in form of ochaison shows how 
 great the confusion in the various suffixes was. Beside lunai- 
 son, although no verb, *luner is present, we find other deriva- 
 
7i 
 
 tives like lunatus : luni, so that the a-form of the suffix was es- 
 tablished by analogy, corbison and guambaison are undoubt- 
 edly late French formations. 
 
 An isolated form of the suffix occurs in the word, *pedationem: 
 peason, cf. Prov. peasos. Another difficult form is escharegon : 
 eschargon. The simplex is escharas, for which a vulgar Latin 
 etymon *excarratium is usually given, {cf Korting, 2906) from 
 Gk. XapaS. This, however, should give *escharais, cf. palais. 
 So a Vulgar Latin * excarracium is preferable ; from this *ex-car- 
 racionem would give regularly escharegon, from which eschargon by 
 absorption of the protonic e by the liquid. A by-form escharisson 
 shows suffix-change. 
 
 The hypothesis presented above is based on the supposition 
 that protonic a was weakened to e before the production of a 
 parasitic i from the palatalized consonants. The possibility is 
 not, however, excluded that the latter process ra&y have occurred 
 first, but more exact data as to the chronology of the devel- 
 opment of parasitic i are needed. Other words which show pro- 
 tonic a intact, or in union with i (sometimes merely graphic, as 
 before /) are due to analogy. Such are, for example, bataillon, 
 bataillier, etc., cf. bataille ; compare this with *nugaliosus : nueil- 
 los, with regular weakening. *companionem : compagnon, cf. 
 compaign (=compan.) Compare O. Fr. compenage and compan- 
 iensem : champenois. *araneata : araignSe, cf. O. Fr. araigne, 
 * stiff ractosus : soffraitos, cf. soffrait, etc. 
 
 e, i, 0, u, protonic. 
 
 I. Before simple consonants, not nasals or liquids. In this 
 case the syncope occurs with the utmost regularity and com- 
 pleteness. The preservation of the protonic vowel before a sin- 
 gle consonant indicates invariably either borrowing from the 
 classical Latin or an analogy of some sort. Also after double 
 consonants and consonant-groups, the fall of the protonic is the 
 rule, in exact correspondence with the history of the final vowel 
 in the same position. As we have seen, the ultima regularly 
 falls, except after consonant-groups whose second element is a 
 liquid or nasal ; in which case this element persists as the sylla- 
 bic of the final syllable. The protonic syllable in the same cir- 
 cumstances is retained. The cases (not numerous) where the 
 
72 
 
 protonic vowel follows the group-consonant + sonant will ac- 
 cordingly be reserved for separate discussion. 
 
 The same considerations of assimilation, which applied to the 
 preservation of the final syllable, are also applicable here, and I 
 shall not enter more nearly into them, except to confirm the fact 
 of the reduction of the protonic after such consonant-groups. 
 In each case after groups which could not be pronounced with- 
 out a distinct glide (ct, pt, ps, etc.) the protonic was probably 
 retained till after the assimilation of the first mute. 
 
 Examples will be classified according to the consonant before 
 the protonic vowel. 
 
 i. after /. *alibanus : albain, *follitatem : folti. *alicunus : 
 alcun, malefatus : malfe, bellitatem : belli, maledicere : maldire, 
 *bullicare : bolgier, *maletolta : maltote, calidaria : chaldiere, mal- 
 evatius : malvais, *callidellum : ckaldel, *puellicella :pucelle, *colli- 
 bertus : culvert, pullicenus : pulcin, *culicinus : colsin, *palitonem : 
 palton, delicatus : delgii, saliceta : salsoie, eleemosyna : almosne, 
 solidare : solder, *filicaria : felgiere, sollicitat : solcie, *filicella : 
 ficelle, sollitanus : soltain, *solitivus : sol/if, pollicaris : polchier, 
 vilitatem : villi, *?e-ad-validare : ravalder, *vallicella : valcele. 
 
 Exceptions. — The following verbs have been levelled to the 
 stenr-accented forms : *al-lucare : alucher, saliitare : salusr, elu- 
 sare : eluser, pilucare : peluchier,follicare : foloiier ; these verbs in 
 -icare have undergone an accent-shifting, fdllico^follico, accord- 
 ing to the analogy of compounds like implied, with recomposi- 
 tion in Vulgar Eatin, cf. Darmesteter, I.e. For another explana- 
 tion, Vulgar Eatin idiare, <Gr. S,iv, see Schuchardt, Littbl.f. 
 Rom. u. Germ. Phil., V, 62 ; Meyer- Eiibke, Gr., II, 610. This 
 accent-shifting occurred in many verbs in -icare, in some of 
 which double forms are found ; so also *dolicare : doloiier. 
 
 Later, specially Fr. formations, are foleti, vileti, etc., with the 
 suffix -ti attached to the feminine form of the adjective. *vilu- 
 tellum : veluel, levelled to velu. *allevamen : alevain, is diffi- 
 cult. It may have been influenced by lever. 
 
 The following are words of purely learned origin : diligent, 
 olifant, filigrane, filipendule, militaire, politesse, politi, qualiti, 
 velUiti, etc. maledicere ■ maleir belongs to the same category ; 
 
73 
 
 although it was borrowed at an early date, before the outfall of 
 intervocalic d. molochinus (Gr. yuolo^zV??) : molequin is appar- 
 ently a loan-word ; the later form morquin shows regular syn- 
 cope. 
 
 2. After r. arepennis : arpent, aurichalcum : archal, caerefolium: 
 cerfuel, *aurifaber ; orfevre, claritatem ; clartS, caritatem ,- cherte, 
 *carricare : chargier (also charoiier, with accent-shifting), cleri- 
 caius : clergiS, cerebellum : cervel, corrogata : corvee", feritatem : 
 
 fiertS, *naricare : narguer, puritatem : purtS, *viridura : verdure, 
 * ex-sarritare : essarter, *veredate : verder, vreder, veritatem . 
 vertS, vretS, Verodunum : Verdun. 
 
 Exceptions. — The following verbs are levelled to the stem- 
 accented forms : *arredare ; arroier, *arrosare : arroser, *arri- 
 pare : arriver, maritare : marie? , *arrisare : arriser, so mariage, 
 etc. claricare : claroiier, with accent-shifting and levelling ; 
 durete, ra.retSa.re late and probably learned. Pure loan-words are 
 *barigeldus : barigel, diriger, charite, charitage, veriti. territo- 
 riunt : terreoir, is probably levelled to terra ; * auri-pellem : 
 Oftpeau is apparently half- learned ; cf. Prov. aurpels. 
 
 3. After a nasal, bonitatem : bonti, *clinicare : dingier, 
 plenitatem : plentS, comitalus : comtS, primicerius : princier, domi- 
 tare : donter, *limitarius : lintier, *cannicella : chancelle, *ramicel- 
 lus ■' raincel, *comitores : contors, rumigare : rongier, enecare : en- 
 gier, sanitatem : santS, *im-ptu?iutare : emprunter, * semitarius : 
 sentier, *limitellum : lintel, tinnitare : tenter, penicellum : pincel, 
 *trinicare : trenchier, *ramicarius : ranchier, *vanitare : vanter. 
 
 Exceptions. — Verbs with levelling are : *admetare : amoier, 
 *banicare : banoiier, *mamcare : manoiier : manier. Loan- 
 words. — benedicere : beneire, beneistre ; cf. maldire : maleir ; 
 this word has been conjectured to rest on a Vulgar Latin benadi- 
 cere {cf. Ital. benadetto), but in view of the further irregularity in 
 the endings, it seems to me preferable to regard it as a loan- 
 word ; so beneoit, beneance, beneisson, etc. poenitere : pene'ir, pen- 
 eance, etc. According to Cohn, Suffixwandl., p. 78, regular, 
 but cf. repentir. Like the foregoing it is an early loan-word. 
 coemeterium : cimetiere, and likewise fumiger, manigance, primi- 
 tif, unitS, vanitS, etc. innovdre : ennover is either levelled or 
 felt as a compound; the latter is more probable, primu-saltum : 
 
74 
 
 primesaut, but O. ~Pr.prinsa.lt, regular. In N. F. the learned 
 form of the adjective is restored, in accordance with the simple 
 prime, fumisterrae : fumeterre; ni this the gender and form of the 
 first element of the compound seem to have been influenced by 
 the second. *monetarius : moni'er, may have been influenced 
 by moneta : monnoie. 
 
 Examples of the secondary development of a protonic syllable 
 from the voice of r, I, m, n, in this position are exceedingly rare. 
 I have been able to find only the following instances : vergon- 
 deux : inveregondeux , (from texts of the Sixteenth Century in 
 Godef.) probably learned. pargon : paregon (Froissard in 
 Godef.) from partitionem. Here the protonic e is evidently due 
 to the voice of the liquid ; if it had followed the analogy of 
 other words with -tionem, it would have sounded *partoison. 
 Possibly also, O. Fr. galvardine : gale vardine (Rabelais) 
 belongs here. The etymology is unknown. 
 
 All other apparent doublets like cherti : chereti, folti : foleti, 
 purti : purete", saute" ■ saneti, etc., are illusory; the longer forms 
 are the later, but represent new analogical structures with the 
 suffix -Padded to the feminine form of the adjective. 
 
 The contrast between the strict syncope exhibited by the pro- 
 tonic in this position, after the vowel-like consonants, and the 
 abundance of doublets and apparently non-syncopated forms 
 when the liquid or nasal stands immediately before the tonic 
 vowel and after the protonic, is very striking. The fact is in ac- 
 cord with the explanation of the double forms with the svara- 
 bhaktic vowel given above. In the former the vowel-like conso- 
 nant belongs to the initial pretonic syllable ; this syllable has 
 only a secondary accent, and never exhibits the phenomena of 
 diphthongation shown by the tonic syllable. Its stress there- 
 fore is not only less in force than that of the latter, but is also 
 simple, uninterrupted ("eingipflig,") so that the vowel-like con- 
 sonant must retain at all times its non-syllabic function as the 
 final consonant of the syllable. After it the syncope is as regu- 
 lar and thorough as after any consonant in this position ; and 
 there was no chance for the production of doublets through 
 " svarabhakti," which is only produced in syllables with a com- 
 
75 
 
 pound accent. On the other hand, when the vowel-like conso- 
 nants stand immediately before the accented vowel, they share 
 the accentual character of that syllable compound, and thus 
 often assert themselves as a syllabic. 
 
 In fact, the tendency to syncope, except under these last con- 
 ditions, is so strong that the protonic vowel is frequently 
 "absorbed " after a liquid or nasal in some by-forms, while it 
 is otherwise generally retained, as derived from a, or as the re- 
 sult of analogy. A glance at the list of forms showing syncope 
 of pretonic a confirms this ; in very many the pretonic vowel 
 stands between a liquid and another consonant {cf. sairment, 
 albastre, chalmin, cortier, etc.). Notice also the following forms, 
 strictly secondary where a protonic syllable, originally retained 
 in the derivative through the analogy of the simplex, has been 
 lost after a liquid or nasal ; *corrupUosus : corrogos (cf. corrots) ; 
 corregos : corgos (Fierabras). The shorter form may be original, 
 as it is undoubtedly regular, but the earlier monuments have 
 corogos alone. (See the examples in Godef.) These doublets 
 appear side by side in the same text ; cf. the following exam- 
 ples quoted by Godef. Trait Cur(e)cuse, dunt le pung fu dor- 
 rez. Otinel 115, but Quant Courroucouse li est es poins torn£e. 
 Ibid, 550. 
 
 Notice further, cabareteur : cabarteur (cited by God. from a 
 text of the Fifteen Century) ; cabaretresse : cabartresse, both are 
 derived of carbaret,fleurctte, ihereiromf/euretts : fleutis,fieurcter : 
 fleurtir. pareil, therefrom repareillage : reparlagi, repareilleur : 
 repaHeur, repareillier : reparlier, the earliest example of the 
 syncope given by Godef. is in a text of 1260. 
 
 When the protonic syllable begins with the group mute -f- 
 liquid or nasal, we should expect, following the analogy of the 
 finals, that the syllable would be preserved, the liquid or nasal 
 serving, probably, as the syllabic of the syllable. This seems 
 to be generally the case, though the tendency to syncope, so 
 strong when the vowel-like consonants stood in the true protonic 
 syllable, frequently asserts itself later. This, however, can only 
 occur after the complete assimilation of the consonants standing 
 before the liquid or nasal. Examples : latrocinium : larrecin, 
 
7 6 
 
 larem. The earliest examples given by Godef. are from the 
 Thirteenth Century. Likewise larronesse (by analogy to lar- 
 rori) : larrenesse : larnesse, lartonie : larnie (Dolopathos), nutri- 
 tura : norreture : norture (one example in Godef. from the Lois 
 de Guil., I, d'Angleterre). Likewise, nutritionem : norreson : 
 norchon (from a text of 1407 in Godef.), norrecier : norcier 
 (1407). In this case the unsyncopated forms have been pre- 
 served inN. Fr. See also Cohn, Suffixwandt, p. 126, who propo- 
 ses as ground-iovm* nutrectura, etc., which is undoubtedly correct. 
 Patriciacum •' Perrecy, * quadrifurcum : carrefour, never synco- 
 pated, petroselinum : peresin : persin (N . Fr.persiZ). Earliest 
 examples given by Godef. in the Chev. au Cygne. *nigritatem ■ 
 noireti : noirtS, nigrescere : noircir. These have been influenced 
 by noir. lacrimate : larmer, is levelled to the stem-accented 
 forms. *putritura : podredute : porreture. N . Fr . pourriture by 
 the analogy of pourrir. See Cohn, op. cit., p. 126. *vitrinare : 
 vernir, never verrenir ; see Korting, Lat. Rom. Wortenbuch. 
 vitrum-glacies : verglas, but cf. Wereglas in L,ittre\ 
 
 The group br- cons.- underwent vocalization of the labial 
 element, apparently before the reduction of the protonic, so that 
 the latter is generally lost, fabricare : forgier, but notice favre- 
 gier, (cited by God. from Vie St. Alexius) : favorgier. *excollu- 
 bricare : escolorgier. Interesting is the doublet *adrestiare : a?es- 
 sier : arser ; the last appears only very late (Regnier). 
 
 Learned words naturally exhibit retention of the protonic 
 vowel always ; the consonants are generally not assimilated. 
 Such are, for example necromantia ■ nigremance, ingremance , 
 rubricare : rebrechier, sobreti, aigremoine,segreier, etc. On in- 
 strumentum : estrument, cf. infra. Verb-forms like the follow- 
 ing are due to analogy : emploiier, oublier, abregier, souplier, 
 etc. 
 
 For the protonic group gn the same law holds good as for the 
 ultima ; the syncope is regular. Assimilation to n was com- 
 pleted before the reduction of the protonic syllable : cognitare : 
 cointier, dignitatem : deintie, *stagnicare : estanchier. Exceptions 
 are all loan-words, like the by-forms dineti, digniti. Likewise 
 significat : senefie. Compare the various forms, popular and 
 learned, of signum. 
 
 Similarly regular is the syncope of the protonic after rm ; ex- 
 
77 
 
 amples firmitatem : ferti, later, half learned fermeti. dormito- 
 rium : dortoir ; also dormitoir, learned, cf. beside this dormebr : 
 dormitorem, which shows the usual treatment of the suffix-fc>m«. 
 4. After a dental, a. Simple dental, ad-id-ipsum ■ ades, 
 *lodivarium : lovier, * buticellum : bocel, medicina : mecine, *de-ex- 
 pedicare : dipecher, medicare : megier, *flattitare : flatter, *nauti- 
 care : nochier, fodicare : fogier, *putidana : putaine, judicare : ju- 
 gier, radicina : racine, matutinum : matin, *rodicare ; rugier, 
 *pediticulat : petille, *sedicare : segier, *peditonem : pieton, *niti- 
 dicat : nettoie, radicalem : regiel, *radicare : rachier, Autosiderum : 
 Auxerre. 
 
 b. After consonant + dental, antecessor : ancestre, monticel- 
 lum : moncel, *blandicare : blangier, pendicare : penchier, *ex-pan- 
 dicare : espanchier, *ponticellum : poncel, manducare : mangiet , 
 vindicare : vengier, Andegavum : Anjou, *particella : parcelle, 
 *tardicare : targier, *vertobellum : vervel, *cordubensis : corvois, 
 *ad-lecticare : allechier, *coacticare : cachier, coctitare : coitier, *flec- 
 ticare .-flechier, ex-corticare : escorchier, *fusticellum : fuissel, mas- 
 ticare : maschier. 
 
 Exceptions. — The following verbs show levelling: castigare : 
 chastoiier, *ardicare : ardoiier, mendicare : mendier, *festicare : 
 festoiier, *undicare : ondoyer, *urticare : ortier, etc. The following 
 are apparently more or less learned in origin : pracdicare : pree- 
 chier, borrowed early, before outfall of intervocalic d. pente- 
 costa : pentecoste, testimonium : testemoigne, cf. tesmoin, *vertobella : 
 vertevelle, cf. vervel. custodire : costeir, (Roland 2962); these last 
 two forms are peculiar in showing a weakening of protonic to 
 e ; they were probably borrowed very early, codicarium : coier ; 
 likewise early, aedificat : edefie, certificat : certefie, * aestivaticum : 
 estiage; the retention of the i is peculiar, particularly since no 
 traces of the adjective aestivus : *estifare found. The retention 
 of the protonic vowel in such words as vestement, chastei, saintei, 
 certei, vendeor, rieor, etc., will be considered later, in dealing 
 with the suffixes -mentum, -torem, -tatem, etc. 
 
 5 After J. positura : posture, *clausitura : closture, *lassitatem : 
 lasti, visitare : visder, *quaesitare : quester, *pinsitare : pester, 
 *consutura : costure. After consonant + s. *laxicare : laschier, 
 *taxitare : taster, *taxicare : taschier, *axicellum : aissel, *ex- 
 acquaculat : esgaille, *absecare : oscher. 
 
78 
 
 Exceptions. — Later formations with -tatem like falseti, lassetS, 
 etc. Loan-words, like usiti, visiter, deciper {dissipare), etc. 
 
 6. After a guttural, acque sic : aissi, figicare : fichier, acque 
 talem : itel, placitare : plaidier, *cogitare : cuidier, *ex-plicitare : 
 esploitier, *culcitinus : colsin (irregular should become *colsdin.) 
 frigidare : froidier, *digitarius : doitier, *longitanus : lointain, 
 digitatum : doitie, sanguisuga : sangsue, digitalem : deel ; this word 
 has apparently lost intervocalic g before the syncope, {cf. mais- 
 tre); *calcedonia : caldoine, learned, but with syncope. 
 
 Exceptions. — The following verbs show levelling : *excautare : 
 ichouer, *argutare : arguer, *glacicare : glacoiier, recusare : ten- 
 ser. The group of verbs with -cipere , parcevoir, recevoir, decevoir, 
 presents difficulties. If levelling had taken place, we should ex- 
 pect to find the diphthong of the stem-accented forms *parcoi- 
 voir, etc. Probably in the consciousness of the folk, a simplex 
 -cipere, -cevoir, was felt as the organic element of these com- 
 pounds. Loan-words are crucS-figere : crucifire, francheti, largeti, 
 richetS, etc. 
 
 7. After a labial, capitettum : cadet, *movitinum : mutin, capi- 
 talem : chadel, navicella : nacelle, *cavicare : choyer (with avi ^> 
 Vulgar Latin au) , civitatem : citS, repulare : reter, *cloppicare : 
 clochier, navigare : nagier, dubitare : douter, subitaneus : soudain, 
 * ex-pavitare : espauter, *bibitoria : boitoire, gravitatem : grieti, 
 *clavicella : clacelle, *in-debitare : endetter, *debitatum : deti, *probi- 
 care : prochier, debitorem : detor. After consonant + labial. 
 *bombitire : bondir, ambitare : hanter, * carpinetum. : charmoi, 
 *plumbicare : plongier, computare ; conter, imputare : enter, *am- 
 bitarius : andier, *orbitaria : ordiere, * servitellum : serdele, berbica- 
 rius : bergier, *ex-torpidire : estordir, *berbica(l)ium : bercail, hos- 
 pitalem : ostel, suspicare : soschier, * hospitaticum : ostage, *hispido- 
 sum : hisdos. 
 
 Exceptions are verbs with levelling. *convitare : convier, info- 
 dere : enfouir, *advlvare : avier, navigare : navoiier, refusare : re- 
 fuser, etc. Eoan- words are advocatus : avoui, *capitaneum : 
 chevetaigne, also, butmore rarely, chapdaine, capitettum : chapiteau, 
 *clavo-cymbalum : clavecin, *cordubanum : cordoan, cf. corvois, 
 obedire ; obeir, debiteur, envtier, refuter, servitume, diviser, etc. 
 
 Here I may conveniently discuss certain suffixes which begin 
 with a simple consonant. These are especially the suffixes 
 
79 
 
 -mentum, -torem, -turam, ticium, etc. (cf. Darmesteter l.c, and 
 Formation des mots nouveaux ; Meyer- l,ubke, Gr., II, 489, 
 529. 533. 535; Cohn, op. cit.,-p. 102 ff.) The protonic vowel 
 is retained before these suffixes, almost without exception, 
 so that the O. Fr. forms are -ement, -ebr, -eure, -eiz, which are 
 attached to the roots of verbs. Darmesteter assumes that the 
 form peculiar to stems of the first conjugation, that is, with pro- 
 tonic a, has been generalized for all other conjugations. Or as 
 Cohn puts it, the suffixes -itorem, -etorem, -utorem, etc., have 
 yielded to -atorem, etc. This explanation accords well with the 
 facts of the case, and has found general acceptance. The sister 
 language, Provencal, however, shows a remarkable discrepancy. 
 Thomas' comprehensive essay, already quoted, has demonstrated 
 that in the latter these suffixes are treated are treated as inde- 
 pendent words, and that they are fixed in three forms, namely : 
 -amen, -ador, etc., for stems of the first conjugation; -emen, 
 -edor, for stems of the second and third ; and -imen, idor, for 
 those of the fourth. 
 
 As examples I may cite : 
 
 I. 
 
 donamen . 
 
 ■ donador. 
 
 II. 
 
 movemen 
 
 • movedor. 
 
 III. 
 
 batemen : 
 
 batedor. 
 
 IV. 
 
 partimen 
 
 : partidor. 
 
 Thomas suggests that this same principle may have been ac- 
 tive during the pre-literary period in French. If so, we should 
 expect -ament, -ement, -iment, as in Provencal ; -ament, however, 
 is never found, and iment is rare. But for some other suffixes, 
 notably -tionem, Thomas' hypothesis is admissible. 
 
 Likewise, in my opinion, the prevailing form of the suffixes 
 -tatem, -tudinem, is best explained by assuming that they have 
 developed independently. If such words as claritatem : clarti, 
 *lassitatem : lasti, *bellitatem . belti, etc., followed the strict chro- 
 nological laws for syncope, we should expect to find forms like 
 *clarde", lasdS, etc., since the protonic syncope did not occur till 
 after the weakening of the tenues to mediae, cf. ordiere, visder, 
 etc. If, however we assume, with Thomas, that in the speech- 
 consciousness of the Gallo-Romans, the suffix -tatem developed 
 
8o 
 
 as an independent word, the preservation of the tenues is at once 
 clear. So likewise for -tudinem, -tume. 
 
 On the other hand, for the suffixes first mentioned, it seems 
 impossible to escape from Cohn's and Darmesteter's conclusion, 
 that eventually the forms -amentum, -atorem, etc. of the first con- 
 jugation stems have been generalized with regular development 
 of a as a protonic. This distinction from Provencal probably 
 stands in accord with another difference, namely that in French 
 the participle ending -antem of the first conjugation has been 
 generalized. This levelling to the ending -amentum, etc., may 
 have been helped by the fact that under Thomas' law, -ementum, 
 -tmentum of the second and third conjugations would equally give 
 -ement. 
 
 Traces of an ending -iment, indicating an independent devel- 
 opment of -imentum, are found in O. Pr. always associated with 
 -tr verbs. In some cases they may be later adaptations to the i 
 stems of these verbs. To the list of substantives in -iment given 
 by Cohn, p. 103, I may add O. Fr. amortiment, afieriment, 
 amaistriment, baniment, blandiment, bleciment, empa?timent, esco- 
 piment, encheriment, esperiment (experimentum, probably learned) , 
 finiment, pariment, muniment, etc. Nearly all have by-forms in 
 -ement, and many verbs in -ir have only forms in -ement, cf. for 
 instance dormement, roissement, tolement, etc. The majority are 
 further widened by the infix -isc > -iss, as in the present stem of 
 the verb. 
 
 Derivatives of the I,atin ending -umentum are exceedingly 
 rare in O. Fr. monumentum : monument (Passion), also mone- 
 ment, moniment, is probably learned, as the signification would 
 indicate, instrumentum : estrument ; here the by- form storment 
 would indicate that the word was treated as in Italian, namely, 
 that the initial ifnjs was regarded by false analogy as the pro- 
 thetic e-i before s + consonant ; cf. Ital. strumento. This pro- 
 thetic e has no significance in the development of initial or pro- 
 tonic vowels, owing to its origin through relations in the stress- 
 group after final consonants ; it is frequently omitted in the ear- 
 liest monuments, and probably never had, inO. Fr., the secondary 
 accent of initial syllables, strumentum is also found in the in- 
 scriptions. The form dorment, beside dormement, is probably 
 secondary, with later absorption of protonic e in the neighbor- 
 
8i 
 
 hood of sonants. It may, however, show original syncope, cf. 
 likewise acesmement : acesment. 
 
 For the other suffixes -torem, -ticium, -turam, we find a few by- 
 forms showing original syncope, where the protonic is 
 treated regularly ; such are, for example, debitorem : detor, *con- 
 siitura ,• costure, molitura ■ molture, *dausitura : dosture, positura : 
 posture, *rasitura : rasture, * dausiticius : clostiz, *voluticium : 
 voltiz. 
 
 More frequent are derivatives of L,atin strong formations 
 without a stem-vowel ; as for example, datorem : dator (learned) , 
 ditdorem : duitor and duieor, scriptorem : escritor and escrieor, 
 textorem : tistor, fadorem : faitor, undorem : oindor (with d from 
 oindre) , *pinctorem : peintor, also peignebr, pastorem : pastor, sar- 
 torem : sartor, pistorem : pes tor, also pestebr, apertuta : aperture, 
 factura .-faiture, *f,ndura .-feinture, ruptura : roture , aventirium: 
 aventiz , faditium : faitiz. 
 
 Instances of derivatives in tor, iure, etc., seem to be unknown. 
 The N. Fr. forms in time are borrowed {cf. Meyer- Liibke, II, 
 536) cribr, beside cricor (cited by Godef. from Aimeri de Nar- 
 bonne) to the verb crier, quiritare is puzzling. It possibly goes 
 back to a Vulgar Latin qu(i)ritorem. Difficult are also traior 
 and traditor, <^ traditorem ; cf. the nominative traitre. 
 
 The forms of the suffix -tatem have already been mentioned. 
 In the older language it is appended in the form -tii, -te, with 
 regular syncope, to the masculine form of the adjective, cf. 
 amerti, cherti, deintie,fert£,feinite,foltei,grieti, ivelti,last£, malti, 
 meurti, nalti, pitie, plenti, salfti, seurti, vilti, verti, (or vreti, with 
 metathesis) . Sometimes to the stems of substantives cf. amis- 
 tie, beneurti, eurti, mendistie, vevete". I^ater it is attached to the 
 feminine, lasseti, pureti, saleti, foleti, chasteli, etc. cf. Meyer- 
 I/iibke, Gr., II, 537. Certain older forms, however, showing 
 loss of the initial t of the suffix and retention of the protonic e 
 are more difficult of explanation. Such are, for example, ares- 
 tee <C* are 'stitatem, chastee : castitatem, netei : nitiditatem, ordei ■ 
 *horriditatem, putei : * putiditatem , quitei : * quitaditatem , saintei: 
 sanctitatem. 
 
 The by-forms chastS, quite, sainte", are also found, though they 
 are generally later. Meyer- Liibke {I.e.) regards the retention of 
 the protonic as due to the fact that otherwise the final consonant 
 
82 
 
 of the stem and the initial consonant of the suffix would coin- 
 cide. At any rate they are not strictly organic developments, 
 but have arisen through some analogical influence, as the by- 
 forms show. 
 
 The Protonic before Liquids or Nasals. 
 
 I have already discussed the peculiar nature of the combina- 
 tion, consonant + sonant in the position before the tonic syllable, 
 and shown how doublets easily arise ; and how this was due to 
 the compound (zweigipflig) stress of the accented syllable. The 
 question arises, whether this development was in all cases sec- 
 ondary, the new protonic syllable developing from the voice of 
 the vowel-like consonant, or whether it represents the original 
 protonic in a weakened state ; in other words, whether syncope 
 was originally complete or not. This question cannot easily be 
 answered, owing to the limited extent of the early monuments. 
 It seems probable, however, that the vowel of the protonic sylla- 
 ble was regularly so weakened as to be absorbed by the voice of 
 the following vowel-like consonant, and that then doublets ex- 
 isted showing an evident protonic syllable with r, I, n, or not, 
 according to the tempo of the pronunciation, and the relations of 
 the stress-group. The orthography would indicate only one of 
 these doublets, generally the ' ' syncopated ; ' ' nevertheless, as 
 the following examples will show, the longer by-form has been 
 quite frequently preserved ; and in some cases this new protonic 
 syllable is retained till the present day. Whether the orthogra- 
 phy el, er, en, adopted for the protonic syllable of these longer 
 forms, expressed simply a sonant liquid or nasal followed by the 
 homorganic consonant, fr, 11, nn, or whether an indefinite vowel 
 e was pronounced, it seems impossible to decide. The difference 
 in pronunciation between rr and er, e r is so slight that the ques- 
 tion is not of great importance. 
 
 It is necessary to distinguish carefully between such cases 
 where the preservation of the protonic syllable is due to analogy, 
 and where this " svarabhaktic " development really occurs. I 
 cite first those examples with regular syncope ; second, those 
 generally syncopated, but with by-forms having the apparent 
 preservation of the protonic; third, those where the longer forms 
 are the rule and syncope the exception. 
 
83 
 
 i. Examples with regular syncope; without by-forms, aesti- 
 mare : esmer, *cingulare : cingler, ambulare : ambler, circinare : 
 cerner, *ad-costurare : acostrer, *tircinella : cernelle, * ad-vesperare : 
 avesprer, * chresimalem : cresmel, *baculare : bacler, coagulare : 
 caillier, bajulare : baillier, blasphemare : blasmer, *buccularius : 
 boclier, desiderare : desirer, *bodinaticus : bonage, considerare : con- 
 sirer, cincturare : ceintrer, also ceinturer, with levelling ; *disju- 
 nare : disner, decimare : dismer, *ex-fondulare : esfondrer, *im- 
 pashirare : empastrer, also empasturer, with levelling ; fibulare : 
 fubler, *frixidare : fresler, *farsurare : fastrer, *fraxinire : frais- 
 nir, *iterare : erret , *im-pulverare : empoldrer, leporaritis : levrier, 
 machinarium ; maignier, manfurinum : mandrin, *interalia : en- 
 traille, *martoretum : martroi, involare : embler, *oculettus : ocil- 
 let, nubilare : nubler, marginare : marner, pedinarius : peignier, 
 *pectorina : poitrine, *pisturinum : pestrin, piper ata .-peuree, *pas- 
 tinaceus : pasnais, *proximarius : proismier, *rasculare : rascler, 
 sabulonem : sablon, sanguinare : saignier, sanguilentus : sanglant, 
 satureja : sarroie, sifilare : siffler, *seculare : sillier, septimana : se- 
 maine, *stabulire : establir, *resinare : resner, strangulare : es- 
 trainler, *taxomaria : taisniere, testimonium : tesmoin, * vicinaticus : 
 visnage, etc. 
 
 Examples of syncope where the protonic stands between two 
 vowel-like consonants are : camerare : chambrer, *cineratum : 
 cendri, * ad-terminare : atermer, *baronare : barner, *colyretum : 
 coldroi, *colerina : corine, horologium : horloge, luminare : lumer, 
 in-taminare : entamer, *intenerire : entendrir, *ilunata : aunSe, 
 meliorare • mieldrer, molinata : molnie, memorare : membrer, mo- 
 linarius : molnier, minimare : mermer, numerare : nombrer, semi- 
 nare : semer, *tremulare : trembler, ruminare : rumer, ululare : 
 hurler, carminare : charmer, simulare : sembler, verminare : ver- 
 mes, etc. 
 
 2. Examples with, in general, regular syncope, but which 
 occasionally show by-forms with " svarabhakti." angulata : 
 anglie, likewise anglos, etc., but *angulettus, anglet, anguelet, 
 (cited by Godef. from Molinet, Chron.), N. Fr. anguleux is 
 learned. *angelettus, anglet, angelet, (cited by Godef. from Ger. 
 de Coinci) . * arboraticum : arbrage, etc., but arboraria : arbrier : 
 arbenere (G. de. Coinci in Godef.) *asperclla : asprele : aspe- 
 relle (cited by Godef. from Ee grant Herbier), N. Fr. prile, 
 
8 4 
 
 always aspresse, etc., ad-titulare : atitler : atileler (Rois.), a 
 learned form, ad-temperantia : atemprance : atemperance, (cited 
 by Godef . from Pierrot de Neele) ; likewise atemprement : atem- 
 perement (from J le Fevre in Godef.) These are possibly 
 learned. *biberate : bevrer, etc., but bevrage : beverage : bever- 
 esse, (from a Gloss, lat. fr. in Godef.) beverie (Petit Plet. in 
 Godef.), beverant (Poeme moral in Godef.), embeverer (Trad, de 
 Rob. de Lincoln in Godef.), *buculare : bocler : boukeler (Les 
 Loherains in Godef.), capulare : chapter : chapeler (from texts of 
 the Fourteenth Century in Godef.) ; likewise, chapeleiz (from a 
 text of the Fourteenth Century), dechapeler (from a text of 
 1461), carcerarius : chart? ier : charteriere (Adenet le roi.), circu- 
 lare : cercler : cerceter (cited by Godef. from Palsgrave) , coperire : 
 covrir, but notice covrel : eouverel " herse." * cooperculatus : cou- 
 verclg : couvercele" (cited by Godef. from L. de Beauveais), *de- 
 liberare : delivrer, but delivrement : detiverement, delivrable : deliv- 
 erable (very late, from a text of 1633 in Godef. ) , * de-abanterianus : 
 devantrien : devanterien (from a text of 1277 in Godef.), de-popu- 
 lare : depeuplier : depeupeliet , half-learned (from the Geste des 
 dues de Bourgogne in Godef.) *im-pauperire : empovrir : poverir 
 (from an Anglo. Fr. text of the Fourteenth Century in God.). 
 indirectum : endroit : endereit (Vie de St. Thomas de Canterbury 
 in God.), *in-tabulare : entabler : entaveler (from a text of 1395); 
 likewise entableure : entabeleure (from a text of the Fourteenth 
 Century). These may be levelled to tavelle. labrum : levre, 
 therefrom esbaulerver : esbauleverer (cited by Godef. from the 
 Destruc. de Rome), ex-*cooperare : escoverer (cited by Godef. 
 from Gui de Warwick), Germ, stapul : estaple, therefrom esta- 
 plage : estapelage (cited by Godef. from the Cont. de Bouillon), 
 fabulare : fabler, also fablel, fablebr, etc., but j ublet : Jabelet. 
 *joculare : jogler, but jocularem : jogler : jogeler ; jogelerie, etc., 
 (cited by Godef. from the Hist, de Fitz. Warin), *lectorinum : lec- 
 trin, but *lectorile. leitril : leiteril (Villehardouin), liberare : liv- 
 rer, but livrance : liverance (Tristau). *misculare : mesler : 
 meseler, *marmorinus : marbrin : marberin, the last is compara- 
 tively common in O. Fr.; cf. the examples in God.; O. Fr. mar- 
 tre ; therefrom martrine : marterine : martrin : marterin, minis- 
 terialem : menestrel : menesterel (from J. le marchant in God.), 
 *ministerarius : menestrier : menesterier (from text of 1402 in 
 
85 
 
 God.), operare : ovrer, but overaigne : ovraigne ; the first is com- 
 mon in O. Fr. texts. That here a real protonic syllable was 
 present, not a mere orthography, is shown by such lines as 
 overayne ne put durer (Merlin), etc. *purpurinus : porprin ; 
 porperin (Tristan) , pauperinus : povrin : poverin (Alex.), also 
 povrement : poverement (Horn), recuperare : recovrer : recoverer 
 (Marie de France) / likewise recovrement ; recoveremenl (Rob. 
 Grosseteste in Godef.).- recovrier : recoverier (G. Gamier in 
 God.); O. Fr. ruifle ( < O. H. G. hruf), therefrom rufflel : 
 ruffelel (cited by Godef. from a text of 1481), rufflet ruffelet. 
 sa.rcula.9e : sarcler, but sarcleresse : sarqueleresse (cited from a text 
 of 1578 in God.) , *sufferire : soffrir, but sofrance : suferance (cited 
 by Godef. from H. de Valenc.) sofrable : soferable (from a late 
 Angl-Norin. text in God.), tabulata : taulie, tabile, etc., but, as 
 half-learned forms tavelet (from a text of 1366), taveleret (1443) .• 
 tavelier (1377). All may have been influenced by tavelle. 
 *temperare : temprer : temperer (Jean Bodel), temprif : tem- 
 perif (cited by Godef. from L,. d' Etaples. Bible), titulare : 
 (tiler: titeler, learned, O. Fr. tetre (*ter-trum, cf. Kortiug Wbr- 
 terbucK), therefrom tertrel : terterel (Maugis d' Aigremont). 
 
 Examples of this development where the protonic stands be- 
 tween two consonants are : *burrulare : borler : burreler ; this 
 may be influenced by borel. cameraria ; cambriere : ckamberiere, 
 cumulare : combler, but comblete : combelete, * hominaticum : om- 
 mage ; omenage (Benoit, etc.) ; also omenois : ornenegie. This 
 example shows that this development in the protonic syllable is 
 very old, before the assimilation of mn ^> m. 
 
 3. Examples where this apparent retention of the protonic 
 syllable is the rule, but which show occasional syncope. These 
 are much less numerous than the forms just considered, where 
 the shorter sentence- doublet in general prevails. This is not 
 surprising when we consider how strong the tendency toward 
 syncope was. I shall cite here also such words as show appar- 
 ent reduction to e of an originally different protonic vowel, in 
 which the possibility of analogical influence seems excluded. 
 Some of these may be partly learned. 
 
 angelorum : angelor (Alexis) , never *anglor, ventilare : ven- 
 teler,cf. N. Fr. ventiler a pure loan-word. *famulentus .-famelent, 
 Mid. I<at. gobettnus : gobelin, rarely goblin. imperator : empe- 
 
86 
 
 rere; and likewise emperebr, empereriz,ttc. This word-group was 
 difficult for Darmesteter, and has been explained as of partly 
 learned origin. Armstrong (Mod. Lang. Notes, X, 356) as- 
 sumes that the word was felt as a compound im-perator (from 
 parare, but this etymology is doubtful) in which case the reten- 
 tion of the protonic syllable is regular. But why can we not re- 
 gard it as a case where from the beginning — from some reason 
 now difficult to trace — the longer doublet was generalized, cf. 
 the Vulgar Latin form impratori, in Schuchardt, II, 431. Com- 
 pare the words souverain and sevelir, which show the same gen- 
 eralization of the longer form, except that with these the synco- 
 pated form has been by chance preserved. 
 
 inimicnm : ennemi. Darmesteter suggests a Vulgar Latin ety- 
 mon *inamicus, but Thomas, because of Prov. enemi, ini- 
 mi, rejects this and suggests a partly learned origin. It, how- 
 ever, can be placed under this category. ordinare : ordener ; 
 *orner (cf. or?ie) seems not to be preserved. The necessity of 
 distinction from ornare : orner has probably led to the early gen- 
 eralization of the longer form in this case. N. Fr. ordonner, 
 has been influenced by donner. *pimpinella : pimpenelle ! , later 
 pimprenelle . Syncope would produce *pimpnelle : *pimnelle : 
 *pimelle, which is not preserved, pompholygem (Gk. no)Acp6\.vS): 
 fanfelue :fanfrelue. This word is not entirely popular, debitor- 
 issa : deteresse, sepelire : sevelir, rarely sevlir (Rom. d'Alecsan- 
 dre). The reason for the preservation of the longer form in N. 
 Fr. is difficult to discover, superanum : soverain, rarely sovrain 
 (Rom. de Thebes) . On the development of this word according 
 to its position in the stress-group, cf. Karsten, I.e. *tertiolettum: 
 tiercelet, abstinentia : astenance ; this may have been felt as a 
 compound ; cf. contenance, etc. in-tumulare : entemeler ; partly 
 learned, as #>« shows; */elonosus : felenos ; this word may 
 have been influenced by felon ; but in that case we should 
 expect the full vowel felonos, which in fact is the more 
 common form. *gentilicium : gentelis; the possibility of in- 
 fluence by gentil seems excluded by the quality of the apparent 
 protonic vowel. *cardonetum : chardenoi (cited by Godef . from 
 a text of 1256.); also chardonoi, ckardonel, etc., which have been 
 levelled to chardon. *miserinus : meserin; also meserel, never 
 *mesrin. Compare mesre. Peculiar to the Eastern dialects. 
 
87 
 
 *reveranus : reverain (from a text of 1356 in Godef.) Probably 
 partly learned in form, cf. N. Fr. rivirer, rivirend, etc. *sapor- 
 osus : saveros (J. Bodel) beside savoros, which is levelled to savor. 
 *rubiolentum : rouvelent; cf. Cohn : op. tit., -p. 207. *populinus: 
 popelin: cf. peuplier, kortulanum : ortelain, *hortulaticum : orte- 
 lage, etc., never syncopated. For similar examples in place- 
 names, see Lindstrom, op. tit., p. 61. 
 
 The following examples show the results of various processes 
 of levelling, so that they do not belong in this category. *am- 
 orosus : amoros, cf. amor, * annulare : anneler, cf. annel, *alenare: 
 haleiner, cf. haleine, * astellarium : astelier, cf. astel, *ad-sationare: 
 assaisoner, cf. saison, *caminata : cheminer, cf. chemin, *caveo- 
 lare : cajoler, cf. cajole, *cellerarius : celleriir, cf. cellier, *cocinare: 
 cuisiner, cf. cuisine, coronare : coroner, cf. corone, dolorosus : do- 
 loros, cf. dolor, * ex-quartulare : escarteler, cf. quartel, farinarius : 
 farinier, cf.farine, choraulare : caroler, cf. carole, candelabrum : 
 chandelarbre, cf. chandelle, instaurare : estorer, cf. estore, *mesu- 
 rare : mesurer, cf. mesure, carbonarius : charbonnier, cf. charbon, 
 
 Borrowed words are, altirer, anguleux, ancolie {aquileja?), 
 artimaire {artem magicatn); acciMrer, centenier, delti (delicatus ; 
 cf. delgie) diorS<C decoratus (?) cf. advocatus, avoui. Can this word 
 have undergone, through folketymology, the influence of aura- 
 tus, ori{?); dominer, reculer, luminer, gramuler, labourer, liter- 
 air e, modirer, nebuleux , geniral, etc., etc. 
 
 I may call attention also to the following doublets — all de- 
 rived from Godefroy — which exhibit likewise the effect of the 
 principles of "svarabhakti " or of the syncope, according to the 
 tempo of the pronunciation, abrotonum : aurone : averoine 
 (Gloss, lat. frc. de Glasgow), denier, therefrom adenerer, but 
 also adnerer (from a text of i^yj),almande (amygdala ?) : alem- 
 ande (Rose) , alevoire : alvioire, sorte de bateau (from texts of 
 1255 an <i J 355 respectively. Ety ?) tenuem : tenve, therefrom 
 atenvir : atenevir, atenver : atenever. O. Fr. tropel, therefrom 
 atropeler : atropler (Kassidor), aventure, therefrom aventuros, 
 but aventreus, with regular syncope (Dit. de la rebell, d'Engl.), 
 aviron, therefrom avironer : avroner, the older form (St. Bran- 
 dan) , baronie : barnie : baroner : barner, the longer forms are 
 levelled to baronbletron (Gaimar) .• bleteron (Lancelot), Fty ? 
 0. Fr. bot, boter, therefrom botrel : boterel (Girbert de Metz), 
 
chevron {*capronem, according to Eittre), therefrom contre cheve- 
 ronner terre, therefrom desosterain : desostrain (Froissard) , dour- 
 deret : dourdret " sorte de monnaie." Etymology ? *canthellus : 
 chantel, therefrom enchanteler : enchantler (Partan), chapel, there- 
 from chapeleure : chapeleure (from a text of 13 12) endenvene : en- 
 demrie " qui ne peut pas rester en place." venin, therefrom en- 
 venimer, then, by dissimilation envelimer : envlimer (Gloss, lat. 
 fr.), *ex-renare : esrener : errener : erner (Bai'f), O. Fr. gabel 
 (Germ, gab) , therefrom gabeler : gabler (Earliest ; paraphase of 
 canticles), O. Fr. gabelle (Germ, gaful), therefrom gabellage : 
 gablage (text of 1399). O. Fr. hamel, therefrom hamelet : ham- 
 let (Angl-Fr. text of the Thirteenth Century). O. Fr. literon, 
 N. Fr. litron. Ety ? palemail ■ palmail " jeu de mail," O. Fr. 
 peterelle (from Petrus, learned), N. Fr. petrelle. petrinet : petrin- 
 gneth (Cambridge Ps.) "le plus petit." piffler : pifeler, "fouler 
 aux pieds." Ety? pulmentum : polment : polement (My s.du 
 vielTest.) publicanus : poplican : popelican, a common form. 
 Eoan-words pulvrin : pulverin (Sixteenth Century), rabobeliner : 
 rabobliner " racommoder " (Anc. Thcat. frc). sublin : subelin 
 (Rabelais), tres fin, subtil." Is this connected with subtilis, 
 perhaps Vulgar Eatin *subtilinus. subgrunda O. Fr. souronde, 
 N. Fr. severonde. Got. manvjau : manvlr : manevir : the latter 
 is the common form. 
 
 I may call especial attention in the above list to the forms 
 averaine, atenevir,cheveronner,pulement, severonde, manevir. These 
 show plainly the secondary development of a " svarabhaktic ' ' 
 protonic syllable from the voice of the vowel-like consonants, 
 and thus confirm the view that the doublets are due to this 
 cause, and not to the retention of the original protonic vowel. 
 
 In direct connection with this subject stands the syncope of 
 vowels before liquids in original initial syllables. This syncope 
 arises through relations in the stress-group. When the word 
 stands in close syntactical relation with another, the two are 
 uttered with the utmost possible economy of force ; they are 
 bound together with a single expiratory impulse, and so the 
 originally initial syllable, with its secondary accent, sinks to a 
 lower degree of accentuation, and undergoes the normal devel- 
 opment of the protonic. The form endereit, quoted above, com- 
 pared with the usual simplex droit, illustrates the intimate con- 
 
8 9 
 
 nection existing between syllable-loss and sentence-position, 
 and the disturbing effects on this relation through mutual anal- 
 ogy. These relations were active at all periods : some words 
 with this contraction appear in the earliest monuments ; in others 
 the loss is late and the rule only in new French, cf. especially 
 G. Paris, explanation of the relations in the future forms farai, 
 ferai, frai. Romania, XXII, 570. 
 
 Examples, many of which are well known, are : *beryllare, 
 driller, bi- ramica, branche, O. Fr. beloce, therefrom belocibr, blo- 
 cier, (cited by Godef. from Vaug. Idill.) O. Fr. beluter deriva- 
 tive of bulre, buriunt, N. Fr. bluter, likewise belutoire, blutoire. 
 *bis-rotetta, O. Fr. berouette, N. Fr. brouette, O. Fr. beluette, di- 
 minutive of bellue : *bisluca, N. Fr. bluetle. corrosus : cros : 
 creux, *corrotulare : crouler, directum : droit, cf. endereit, supra. 
 directiare : dresser, corrigia : courroie, therefrom O. Fr. coroiete : 
 croete. *de-rationare : deraisnier, also draisnier (Vie. de St. 
 Thomasde Canterbury )*ferritta : frete (Diez' etymology, doubt- 
 ful ) . filum lana : flaine (?) , furunculum : furoncle : froncle ( cited 
 by Godef. from R. Estienne) quiritare : crier, O. Fr. pelaud, de- 
 rivative of poil, therefrom O. Fr. pelauder : plauder (Bou de 
 Peviers in Godef. ) . Other derivatives of poil show the same 
 doublets, pelts : plis, pelicon : plicon, O. Vr.peloter, "ballotter, 
 repeter, " ploter (Farce du Cuvier in Godef.). O. Fr.peloton : 
 ploton (cited by Godef. from texts of the Sixteenth Century) , O. 
 Fr. pelottoir : plottoir (cited by Godef. from a text of 1469) . O. 
 Fr. peluc : pluc, N. Fr. peluche : pluche, piliicare : pelucher : 
 plucher ; N. Fr. iplucher, *perustulare : brusler (?) veracus : 
 verai : vrai. 
 
 The Protonic Vowel before Two Consonants. 
 
 Before considering the cases where the protonic is really " in 
 position," a small group of words demands attention, where the 
 protonic vowel is followed by mute + liquid. The vowels are 
 treated generally before this group is in an open syllable ; and 
 accordingly we should expect outfall of the protonic to be the 
 rule, as it is before a simple consonant. But many examples 
 seem to contradict this. 
 
 peregrinus : pelerin, *integrinus : enterin, also entrin, already 
 in St. Brandan. *deretranus : deerain, dererain : derrain. The 
 
9° 
 
 last is the earliest, multeplicat : molteplie, catedralem : chaerel. 
 On the other hand, ossifraga : osfraie : orfraie, *pulletranum : 
 poltrain. Learned words are, fenu-graecum ; fenu-grec, filu-gra- 
 num : filigrane, tenebrosus : tenebros, integrare : entigrer, gene- 
 tricem : genetris, meretricem : meretris, also mertris (cited by 
 Godef. from I,es Loherains). consobrinus : cousin offers so 
 many irregularities that it may be left out of consideration. See 
 Meyer-Liibke, Gr., I, 521. 
 
 The above list shows, especially before the group gr, a ten- 
 dency toward retention of the pro tonic. Many of the words may 
 be explained as the result of analogy ; thus chaerel to chaere, 
 poltrain to poltre, etc., but for pelerin such an explanation is im- 
 possible (Schwan, Afr. Gr., p. 79, holds pelerin and enterin for 
 learned words) . I am of the opinion that here the same principles 
 have been at work that were noted above for the simple liquids. 
 A svarabhaktic syllable was developed from the voice of the r 
 before the assimilation of the guttural, cf. the similar process in 
 integrum : integrrum : intigrrum : entier. Apparent retention of 
 theprotonic would naturally result. That, however, these forms 
 are the result of phonetic relations in the stress-group, and not of a 
 well-defined law by which the protonic is preserved before gr, as 
 Darmesteter assumes (cf. also Meyer- Liibke, Gr., I, 275), the 
 early doublet enterin ; entrin shows. 
 
 The Protonic in Position. 
 As stated above, Darmesteter left this case — where the pro- 
 tonic stands before two or more consonants — out of considera- 
 tion, although he hinted that it undergoes a different develop- 
 ment. See especially Lindstom, op. cit., chap. IV. I^indstrom 
 assumes as a law that protonic e, i, 0, u is always retained be- 
 fore two or more consonants. But his own examples do not 
 fully substantiate this conclusion, cf. Anvers, etc. As the 
 Gallo-latin vowel-syncope is the result of a greater stress laid on 
 the tonic syllable, with a corresponding weakening of the unac- 
 cented, it is difficult to see why this result should not have ob- 
 tained in close syllables as well as open ones. It must be re- 
 membered that the date of the final outfall of the protonic is 
 comparatively late, probably after most of the processes of con- 
 sonant-assimilation had been completed, so that originally 
 closed syllables had become open. 
 
91 
 
 That a closed syllable offers perhaps more resistance to the 
 process of weakening may be shown by the Latin vowel-reduc- 
 tion. There a short vowel is reduced to e originally, and re- 
 mains at this stage in closed syllables ; but in open it sinks still 
 further, to i. (See Lindsay, op. tit, p. 188, ff.) But it is 
 doubtful if this resistance could be effective against the much 
 stronger tendency toward weakening that prevailed in Gallic 
 Vulgar Latin. Moreover, as Karsten, /. c, argues, when 
 several consonants come together, they may be reduced by as- 
 similation ; a " Stutz-vocal" for the sole purpose of easing the 
 pronunciation is not necessary. (Why, however, assume 
 *soschon as the regular form of suspectionem ? The group cti can 
 only give voiceless s in central French, so *sosson would perhaps 
 be more correct.) Combinations of two or more heterorganic 
 mutes may in fact be pronounced without a particular voice- 
 glide between them, if the speech organs be brought into po- 
 sition for the second, before the contact of the first is loosed. 
 In this case the first consonant will consist of a simple stop, 
 without the usual following explosion. On the other hand, 
 when the stop of the first consonant is exploded before the con- 
 tact of the second is formed, then the synthesis cannot be ef- 
 fected without a perceptible voice-glide, which will probably 
 constitute a " Nebensilbe." So that the retention or non-re- 
 tention of the protonic syllable would depend on relations of 
 tempo ; but these relations are so disturbed by analogical in- 
 fluences that it is difficult to trace them. 
 
 A double consonant made ' ' position' ' in Latin and was un- 
 doubtedly pronounced with two force-impulses, as in Italian. 
 The consonant has been simplified in French, but seem to have 
 left traces in the preservation of the protonic syllable. Exam- 
 ples, where all possibility of analogy is excluded, are rare. 
 Possibly the following may serve to exemptify the law : abel- 
 lana : avellaine, *forcillatus : forceli. Mid. Lat. oppellanda : 
 houppelande. These examples contain the double liquid /, and 
 so aie treated like those which have the protonic in position be- 
 fore liquid + consonant. In these, the retention of the protonic 
 syllable is regular ; so the same may be assumed for the above. 
 The development would be abellana : avellana, avl-laine avelatne , 
 exactly like the development in paupertatem : po-ver-tet : po-vr-tf. 
 
92 
 
 All other instances which show the retention of the protonic 
 vowel before a double liquid or nasal are exposed to the suspi- 
 cion of analogical influence. Such are all verbs like rebellare : 
 reveler, which may be levelled to the stem-accented forms. Such 
 cases are very numerous ; cf. chamelin : chamel, chanceller : chan- 
 cel, chodeller : chodel, maiselliere : maiselle, fiaeller : flael, pais- 
 seler : paisselle, rasteler : rastel, taveler : tavelle, chevelu : chevel, 
 eschantellon : chantel, chastellain : chastel, porcellaine : porcel, 
 avorir (abhorrire) : avor, empenner : empenne, etc. 
 
 I have been able to find no certain examples where the pro- 
 tonic stands before a double mute or spirant. In the following 
 the stem-accented forms have evidently been influential : *extup- 
 pare : estoupper, * sangluttare : sangloter, *ingluttire ; englotir, 
 *quatottare : cahoter, */ringuttire : /rang otter, *embuccare : em- 
 boucher, *ingressare : engresser, etc. 
 
 The Protonic before Sonant -\- Consonant. 
 
 As indicated above, the protonic syllable is retained in this 
 position. The liquid or nasal can never be lost in a consonant- 
 group ; therefore, when stress was withdrawn from the unac- 
 cented protonic, these would assume syllabic function, — the syl- 
 lable being thus preserved, although the original vowel was 
 probably weakened or lost. The variation in spelling that is 
 found in povreti : poverti, covreture : coverture, etc., show that 
 phonetic value of the protonic syllable in these cases was that 
 of sonant r. Examples which seem free from analogical in- 
 fluences are, paupertatem : poverti : povreti, coopertorium : cover- 
 toir : covretoir, likewise covreture, etc. nocturnalem : nuiternel : 
 nuitrenel (Oxford Psalter). Notice that protonic ur has be- 
 come er = r; voluntatem : volenti. (On the spelling volanti, 
 Besangon, etc. cf. Forster, Ztschr. Rom. Phil., XIII, 535; G. 
 Paris, Rom., XVIII, 352.) 
 
 Forster's law (protonic on ]> au) can only hold good for pro- 
 tonic on that has been maintained by analogy. The retention 
 of the full vowel in the protonic syllable is unlikely, — and un- 
 known except by analogical influences. Later nasalizing of the in- 
 definite vowel developed from the voice of the n would give en~^>an. 
 *expaventare : espoenter, calumniare : chalengier, dum interea : 
 dementiers, redemptionem . raengon, possibly levelled to rdent. 
 
93 
 
 contentionem : contenson, levelled, *amentinum : amentin, *repen- 
 taculum : repentail, probably felt as a compound. *extendar- 
 dum : estendmt, * solemnalem : solennel, possibly a borrowed 
 word, cf solemne. Other examples which are probably due to 
 levelling are *recentiare : recenciet , *agentiare : agencer, talentare : 
 talenter, adversatius : aversier, hibernare : iverner, charpentier, 
 ferntenter, entercier, etc. Examples with the protonic vowel 
 preserved unweakened are in all cases the result of analogy ; 
 thus, ascolter : ascolte, immondice : immonde, reondel : reondir : 
 reond, remorquer : remorque, aombret : aombre, aplommer :plom, 
 etc. * quaterniolum : careignol, is difficult ; it probably shows 
 early metathesis. 
 
 Before spirant + mute, protonic syncope is the rule. This is 
 to be expected, since sc, st begin the syllable in Latin ; therefore 
 the protonic is treated as before as a simple consonant. Exam- 
 ples, ministeiium : mesher ; also menestier, which must be the 
 learned form {cf. G. Paris, Alexius, 41). Koschwitz (Com- 
 mentar, p. 59) , holds that menestier is the regular form for the 
 period of the Eulalia, e being " geschiitzt " by n-st ; and the 
 syncope is later. But minsterium is already found in Vulgar 
 Latin. Plautus, Pseud. 772, cf. Lindsay, op. cit., 202, and com- 
 pare Ital. mestiero. avem struthio : austruche, *monisterium : 
 mostier, magistatem : maistie (dissyllabic). Other examples 
 with the protonic preserved before st are either learned or levelled; 
 thus, canistellum : chanestrel, cf. chanestre, magisterium : ma'estire, 
 learned, cf. maistie, supra, honestatem : onesti, likewise onester, 
 cf. oneste. *conquistare : conquester, cf. conquest, potestatem : 
 poestS, poester, etc. cf. potestat : po'este. *pedestare : peester, cf. 
 peestre, *tempestatem : tempests, cf. tempeste, angustatem : angusti, 
 late and learned. *aristarium : arestier : aresios, etc., cf. areste. 
 agustare : aoster, cf. aost. *capistare : chevester, cf. capistat : 
 cheveste. 
 
 For the group sc examples are not numerous, canescere : 
 chancir, clarescere : cs-clarcir, nigrescere : noirch , here the reten- 
 tion of the protonic after gr would be regular, but compare 
 noir. *rogiscellum : missel ; apparent exceptions are *cognos- 
 centia : conissance, *vermiscellum : vermissel, * ramiscellum : 
 rameissel : tamissel. The last two words show probably an inde- 
 pendent development of the suffix -iscellum, in accordance with 
 
94 
 
 Thomas' theory ; influenced also by arbroissel : arbrissel, where 
 the retention of the protonic after br is regular. The by-form 
 raincel is generally traced back to *r amice Hum, but it can also 
 be from ramiscellum, with syncope and levelling of the initial 
 syllable to ramus : rains. 
 
 For the protonic before the groups d, pt, I find no certain ex- 
 amples, unless acceptorem : austor be such. The form shows syn- 
 cope, but is otherwise irregular. Apparent exceptions, like de- 
 lectare : delitier, cf. delit. *adruptare : arouter, cf. aroute, respectare : 
 respitier, etc., are levelled, eledeur, leduare, refleier, etc., are 
 learned. 
 
 Before a consonant + hiatus i, which constitutes "position," 
 as the development of the accented vowels shows, syncope of 
 the protonic is also regular, although the lawful reduction is 
 often hindered by analogy, or other causes. 
 
 For t -\- hiatus 2, examples *sanitiare : sancier, * com-initiare : 
 comencer, *funditiare .-fancier. 
 
 More frequent is the levelling to the stem-accented forms, cf. 
 *minutiare ■ menuisier, *cantitiare : chantisier, *sortitiare ■ sorti- 
 cier, * ex-lactitiare ; esleecier, cf. leece, * 'ad-titiare : attisier, *ad-prisi- 
 tiare : apprivoiser, *in-vitiare : envoisier. 
 
 For derivatives with the suffix -tionem, with regular syncope, 
 cf. venditionem : vengon, *bibitionem : boisson,partitionem : par (on, 
 * maturitionem : meurson (from a text of 1258 in Godef.), luison, 
 by the side of luoison, and duison, conduison, also probably belong 
 here, representing Vulgar Latin *lucitionem, *ducitionem. Strong 
 formations *ludionem, ductionem, would give lugon, dugon, cf. 
 f actionem : fagon. 
 
 By the side of these syncopated forms are found (and much 
 more frequently) by-forms in -oison, -ison, that is, with the stem- 
 vowel unsyncopated. As examples, I may cite aombrison : aom- 
 brir, aparoison ■ aparoir, avertison : avettir, batoison : battre, 
 \* battetionem for *battutionem, cf. Thomas, l.c.),compliso?i : com- 
 plir, creoison : croire (*creditionem) , faillison : faillir, fendoison : 
 fendre , finison : finir , fondoison : fondre , gemison : gemir, guari- 
 son : guarir, luoison ; luire, marison ; marir, partisan : partir, be- 
 side pargon, plorison : plorir, repentison : repentir, vestison : ves- 
 tir, vendison : vendre, beside vengon, trahison : trahir, etc. 
 
 Cohn {op. at., pp. 124, 133), assumes that these longer forms 
 
95 
 
 are regular, apparently believing that the protonic is not syn- 
 copated before t + hiatus i ; he therefore assumes as Vulgar 
 Latin substrata of the syncopated forms *biptionem, *ventionem, 
 *partionem. But this view can hardly be demonstrated, espe- 
 cially when we consider verb-forms like sancier, fonder ; vendi- 
 Honem : vengon, and */unditiare .-fonder axe completely compar- 
 able, and both show regular syncope. All difficulty is removed 
 when we assume with Thomas that the suffix forms for each of 
 the conjugations were -ationem, -ttionem, -etionem, and -itionem, 
 respectively, and that these were treated independently and at- 
 tached later to verbal roots. That in a few instances the con- 
 sciousness of the suffix was lost, with strict syncope as a result, 
 is not surprising ; it was probably favored by a slight shifting in 
 the meaning, compare pargon "partage," but partison " action 
 de partir." See moreover Meyer-Iyiikbe, Gr., II, 395, who also 
 recognizes in vengon, etc., the regular forms. 
 
 I have already called attention to the fact that these suffixes 
 were often confused, the form peculiar to the stems of one conju- 
 gation being carried over to those of others, cf. for instance, 
 batoison, batison, vendison, N. ~Bx.fleuraison,pendaison,pondaison: 
 cudllaison, etc. An interesting form is *minutionem : menuison, 
 with by-forms in -oison, and -ison ; it shows a Vulgar Latin suf- 
 fix -utionem, all traces of which have been lost in Provencal, ac- 
 cording to Thomas. 
 
 Some words in which the suffix -tionem appears as -egon are 
 more difficult to explain. Such are nutritionem : norregon, 
 *nepotionem .-nevegon, * summitionem : somegon, and anglecon. Ac- 
 cording to Cohn, op. dt.,-p. 126, the first represents Vulgar Latin 
 *nutredionem ; if so it is regular with preservation of the pro- 
 tonic syllable after tr. From it a suffix -egon might be extended 
 to other words. For somecon, the shorter somece, -summitia, 
 served as foundation. A similar -egon if found for -donem. cf. 
 also Meyer- Liibke, Gr., II, p. 500. 
 
 Before c -\- hiatus i we find the following examples of protonic 
 syncope, * polliciarius : polder, *pulidare : pulder, N. Fr. ipucer, 
 *berbidare : berder, with the suffix -donem, *palicionem : palgon. 
 In the other hand, as apparent exception, through analogy, 
 *convidare : convider, cf. convice, parochianus : paroissien, cf. par- 
 oisse, *pelidare : pelider, pelicon, etc. cf. pelts. Another excep- 
 
9 6 
 
 tion is formed by a group of words with the suffix -ionem, *ham- 
 icionem : hamegon, *iricionem : ire g on : erigon, senecionem : senefon, 
 *orecionem : Oregon. Of these words, hamegon and iregon have 
 been influenced by *hamice, *eris ; for the others, such primitive 
 formations may perhaps be assumed, {cf. Meyer- Lubke, Gr., 
 II, 500 ; Cohn, op. tit., p. 30). Possibly also, the extinct nom- 
 inatives *hamez, *senez, etc., if we may suppose their existence 
 for a short period in Vulgar I^atin may have contributed to pre- 
 serve the protonic. Such nominatives are apparently preserved 
 in some cases, cf. supra, dace, chace, trace, etc., but only in a 
 learned form. cf. moreover O. Fr. achaise, possibly occaseo. 
 
 For the protonic before s -\- hiatus i the only example with 
 syncope is pertusiare : percier. But all exceptions are easily ex- 
 plained by analogy, or as loan-words, cf. occisionem : occision, 
 borrowed. Derivatives in -isan from -ensianus, like partisan; 
 courtisan, artisan are loan-words from the Italian. We also find 
 *artensianum, O. Fr. artoisan, levelled to Artois, and N. Fr. 
 artisien, which can hardly be a popular word. That -anus after 
 palatals can become ten in loan-words is shown by chrestiien, 
 maximien, etc. 
 
 For the protonic before d -(- hiatus i examples with syncope 
 are viridiarius : vergier , fastidiare : faschier ; faschos, etc. Anal- 
 ogous forms are *apodiare : apuiier, cf. apuie ; *apodialem : 
 apoiail, etc. *inodiare : ennuiier, cf. ennuie. sordidiorem : sor- 
 deior, cf. sordeis, desidiare : deseiier, cf. the stem-accented forms, 
 meridianus : meriaine is a loan-word. Before^ -f- hiatus i, ex- 
 amples with syncope, *corrigiata : corgie, *corrigionem : corjon. 
 For the protonic before a labial + hiatus £, no examples with 
 syncope are found, but all are evidently levelled, cf. aleviare : 
 alegier, cf. alege, adpropiare : aprochier, cf. proche, vindemiare : 
 veiidanger, cf. vendange. 
 
 Before two consonants -+- hiatus i no examples show syncope, 
 but nearly all are plainly levelled, cf. *adauctiare : aoisier, cf. 
 aoise. *adrectiare : aressier, cf. arser, supra, maledictionem : 
 maldisson • maleigon, benedictionem : benetgon, cf. maldit, maleit, 
 beneit. * destrictiare : destrecier, cf. destrec. *corruptiare : corros- 
 sier, corogos, etc., levelled to corros, but cf. corsos, supra. *delec- 
 tiare : delicier, cf. delices. *electiare : elicier, cf. elice, similarly 
 electionem : eligon, angustiare : angoissier, cf. angoisse. suspec- 
 
97 
 
 tionem : sospecon. This word is generally regarded as proving 
 that the protonic, before two or more consonants, was not syn- 
 copated, but (leaving aside Karsten's hypothesis of a nomina- 
 tive *sospece) we find in O. Fr. a verb suspediare ; suspecier ; 
 and the substantive may have been levelled to the stem-accented 
 forms of this, suspediat : sospece, etc. 
 
 Before /, n + hiatus i, the protonic seems to be regularly 
 maintained, and Darmesteter and Meyer- Iyiibke, Gr. % I, 274, as- 
 sume this to be a phonetic law. Nevertheless, some significant 
 exceptions occur; thus *juveniorem : joignor (Wace, etc.), 
 which on account of the o cannot be from junior em, also jovetior: 
 levelled to jovene. * dominionem : donjon, dominiarium : dan- 
 gler, *laciniaria : lasniere, *gratilionen : greslon. 
 
 On the other hand, examples with the pretonic syllable pre- 
 served are, Avenionem : Avignon, * campinionem : champignon, 
 also O. Fr. champignol, etc. lusciniola : losegnol : rossignol, 
 *catenionem : chaeignon, cf. chaeine, * cydoniarius : cooignier, cf. 
 cooing, * quatrenionem (iox quaternionem ) : careignon, *ingeniare: 
 engignier, cf. engin. * ex-graphiniare : esgraftgner, humiliare : 
 timelier, cf. humiliat-umeille , *lorbilionem : tourbillon, cf. torbille, 
 *artiliosus : artillos, likewise artillier, etc., cf. *artiliat : artille, 
 *barbuliare : barbouiller, cf. barbouille, dispoliare : despouillier, cf. 
 despouille. *inoliare : enoillier, cf. enoille, *hostiliare : ostillier, cf. 
 ostille, *consiliare : conseillier, cf. conseil, *fusiliare : fusilller, cf. 
 fusil, papilionem : O. Fr. paveillon, *scopilionem : O. Fr. escouveil- 
 lon , frictiliare : fretiller, cf fretille, * ex-cantilionem : eschantillon, 
 gurgulionem : O. Fr. gourgeillon, *vestiliare : vestillier, cf. ves- 
 tille, * quadrilionem : O. Fr. careillon : carillon, * subtiliare : sotil- 
 lier, cf. solille. 
 
 Of these examples, the majority may be easily explained by 
 analogy, but for some, paveillon, quareignon, carillon, champig- 
 non, rossignol, this explanation is inadmissable. I believe that 
 here again the peculiar nature of the vowel -like consonants is to 
 be seen. Before the period of syncope, /, n, were palatalized by 
 the following i : in slower utterance this /, «, would retain their 
 semi-syllabic character, especially until the hiatus i was com- 
 pletely absorbed. The few instances where syncope is found 
 confirm this. The i, which appears before n, I in the protonic 
 syllable is a new French development, and may be due to the 
 
9 8 
 
 palatal character of the consonants ; in O. Fr. we find in nearly 
 every case by-forms spelled -eil, -eign, -egn, etc. 
 
 Before r + hiatus i, the syncope is regular. Examples, *gut- 
 turionem : goitron, * ntateriamen : merrien, N. Fr. merrain, *im- 
 pasturiare : empaistrer. antirieur, superieur, etc., are naturally 
 pure loan-words. 
 
 Summing up the results for the protonic in position, we may 
 say that the syncope is regular, except where the effect of the 
 phonetic laws has been disturbed by analogy. This conclusion 
 extends the action of the law of Darmesteter, and establishes 
 its invariability. An apparent exception presents itself only 
 where the vowel-like consonants follow the protonic ; their voice 
 has in most cases preserved the protonic syllable. This also 
 harmonizes with the results obtained where the protonic stands 
 before a simple liquid or nasal. In words with more than two 
 pretonic syllables a new question arises as to the position of the 
 secondary accent. This has lately been comprehensively treated 
 by Armstrong {Mod. Lang. Notes, X, 350, ff.). His results, 
 which seem to be conclusive, demonstrate the correctness of 
 the view held by Meyer- I/iibke, Rom. Gr., I, 273, that the sec- 
 ondary accent in all uncompounded words rested on the initial 
 syllable, cf. for contrary views, Darmesteter, who holds that 
 the law of binary accentuation prevailed, aspiritdtem ; and 
 Schwan, Afrz. Gr., p. 29, according to whom the secondary 
 accent follows the same law as the primary; namely, that it rests 
 on the second syllable before the main accent if the same is long; 
 it is short, on the third, heredtt&re, arboricillus. 
 
 Accepting Armstrong's view, which is moreover in harmony 
 with the laws of Latin accentuation, in which the main ictus 
 rested originally on the initial syllable and the classical accent- 
 law was the result of a shifting in position of the former primary 
 and secondary accents, we should expect to find that the pretonic 
 half of the word treated exactly like the post-tonic — that is, the 
 original protonic syllable would be subject to the same laws as 
 the final in proparoxytones. This, however, is the case but 
 rarely, as the following examples will show ; in many instances 
 it is the protonic syllable which disappears, while the second 
 pretonic syllable, corresponding to the penult in proparoxytones, 
 is preserved. Armstrong accounts for this by a regressive ac- 
 
99 
 
 tion of the main ictus ; the syllables standing immediately be- 
 fore or after the tonic syllable were more liable to be weakened 
 through its greater stress than those farther off. Can this view 
 be justified? 
 
 The relation of accent-grade among the different unaccented 
 syllables in L,atin is very obscure. The stress given to each 
 particular syllable would probably vary according to its position 
 in the stress-group ; and tbe influence of related words would be 
 important for their cognates. Thus, in amhrittido, for instance, 
 the a in the second pretonic syllable would probably at all 
 periods be more highly accented than the protonic i, through the 
 analogy of amdrus, so that we have O. Fr. amertume instead of 
 a possible *ambretume. This principle, in my opinion, suffices 
 in nearly all cases to explain the preservation of the second pre- 
 tonic syllable. On the other hand, in isolated words without 
 allied shorter cognates, the gradations in degree of stress in the 
 pretonic syllables would probably follow the rule for the post- 
 tonic, namely, a secondary accent on the initial syllable, fol- 
 lowed by two weaker syllables, of which the last or protonic 
 has the greater stress. 
 
 Certain phenomena of syncope found even in classical I,atin 
 confirm this view ; cf. for instance, altrinsecus with alteri, pos- 
 tridievnth posteri, officina-with. opificina (Plautus). cf., moreover, 
 the vulgar I,atin forms in Schuchardt like opermenti, autortate, 
 immenstate, but, on the other hand, impratori, virdiario, terben- 
 tinae, deposione. Schuchardt' s examples show syncope of the 
 second pretonic syllable in all cases where analogy does not 
 come into play. 
 
 In these instances, the syncope affects the syllable imme- 
 diately following the initial ; not the one standing next to the 
 tonic accent under the later accent-law. Similarly, in French 
 we find many cases where of three pretonic syllables the second 
 regularly disappears. One exception may possibly be admitted. 
 Pretonic a, in whatever syllable it may have stood, may have 
 received at all times greater stress than the neighboring vowels. 
 Bt the scarcity of material hardly permits a decision on this 
 point ; paraver'edus : palefroi seems to be only certain instance. 
 
 Hiatus i in the second protonic syllable, like all other hiatus 
 vowels, suffered early reduction to consonental i, and the follow- 
 
IOO 
 
 ing protonic is treated like others. Syncope is the rule. (The 
 examples where hiatus i stands in immediate hiatus to the tonic 
 vowel, I have already discussed.) Here it may be noted that 
 cases of syncope like partitionem : parcon, *bibitionem : bois- 
 son,fonditiare : f oncer, pretusiare • percer, etc., refute Schwan's 
 assumption for protonic accentuation, according to which they 
 should have a secondary accent on the long vowel, cf. his ex- 
 ample comparationem : comparaison. Bxamples of pretonic hiatus 
 i are : *puteolenlus ; pullent, * badiolettus : baillet, *modiolonem : 
 motion, meliorare : mieldrer, *rationare : taisnier, mansionarius: 
 maisnier, *mansionata : maisnier, *mustionellus : moisnel, *pas- 
 tionaticum : pasnage, curiosolitae : corseult, societatem ; soistie, also 
 soyesti, a half learned form, medietatem ; moitii, *medietarius: 
 metayer, is difficult ; we should expect *moitier. The Prov. 
 form meytadier would indicate a Vulgar I^atin ground torm*media- 
 tadarius ; cf. *medietadanus : mitoyen ; but even then the devel- 
 opment seems not to be entirely regular. *tertiolettus : tiercelet, 
 *truncionare : troncener, with apparently irregular preservation 
 of the protonic are both due to the principle already enunciated- 
 development of a weak syllable from the voice of the vowel-like 
 consonants, cf. ordener and hortelain. 
 
 The same law holds good for hiatus u before the protonic 
 vowel, both are syncopated, *mansuetinus ; mastin, cf. G. Paris 
 Rom., XXII, 597. consuetudinem : costume. Exceptions like 
 inquietudinem : enquetume, mansuetudinem : mansuetume, etc., 
 are not genuinely popular words. 
 
 Examples of the retention of the second protonic syllable are 
 the following. I omit all compounds like encombrer, esploitier, 
 empostrer, aproismier, etc., where the consciousness of composi- 
 tion was evidently felt. I cite in each case cognate forms which 
 may account for the seeming irregularity, amaritudinem : amer- 
 tume, cf. amer, desiderare : desirer, cf. desire, *formiculionem : 
 formillon, cf. formi, heredetare : ereder, cf the stem-accented 
 forms hereditat : erede, etc., also ireter, heriter, heritage, etc. 
 * ex-colliibricare ■ escolorgier ; cf. * ex-colubricat : escolorge, with 
 fabrica : forge, suavitudinem : soatume ; this word is hardly 
 popular; cf. suavem : soef. * de-ab-interanus : derantrain, cf. 
 entraigne, * ex-acquaculare : esgaillier ; cf. *ex-acqudculat : es- 
 gaille. Notice the development as an uncompounded word. 
 
IOI 
 
 ar. 
 
 interrogare : enterver < cf. enterve, but the word is also irreguL 
 adulterare : avoltrer, cf. avoltre, communicare .■ commungier ; cf. 
 comm&nicat : communge. The by-form comengier may be regu- 
 lar. In that case I should regard the spelling en as the graphic 
 form ; it is a parallel to the spelling er for r often met with in 
 the protonic syllable ; cf. for instance, poverti : povreti, nuiter 
 nel, asperti, etc. * ex-lucidare ; esluidier ; cf. esluire, infirmita- 
 tem : enferti, cf. enferm, *quadragesimalem : caresmel, cf caresme. 
 *ad-quaesitare : aquester, cf. aquest. famelicare .- fameillier ; we 
 should expect *famelgier, but metathesis seems to have occurred; 
 *famecilat would give fameille regularly. caballicare : cheval- 
 chier ; cf. cheval, recuperare : recovrer, cf. recovre, ex-corticare • 
 escorchier; cf. escorche, *juvencellus : jovencel, cf.jovence, ministe- 
 rialem : menestrel, * minister arius • menestrier. These last two 
 words are often cited as examples of the preservation of the 
 second protonic syllable. Regular forms would be *mestrel, 
 *mestrier, menstrel, is actually found in a chart of I^iege of 1355. 
 The longer forms are probably influenced by ministrum : men- 
 estre, or by the stem- accented forms of the verb ministrare : men- 
 estrer ; crudelitatem ; crueltS : crualti, cf. cruel : crual. 
 
 likewise for all derivatives in -t£, by the side of which are 
 found substantives or adjectives which serve as simplex, cf. 
 amerti : amer, malheurti : malheur, naite : nais, nientS : nien, 
 ptiorti : prior, novelti : novel, fealti : feal, royalti : royal, amis- 
 tie : amis, etc. Those with the protonic syllable also preserved 
 are undoubtedly later formations ; thus jolivetS, laideti, fraileti, 
 moieneti, prochaineti, seureti, roideti, veveti. Some forms with a 
 stem-ending in a liquid or nasal show doublets ; thus, asperita- 
 tem : asperti : aspteti ; this probably indicates only different 
 spellings for syllabic r, as has been already mentioned, cf. further 
 prospertS : prospret£,foibleti, umbletS. In all of these the two pre- 
 tonic syllables have been reduced to the vowel-like consonant 
 functioning as syllabic, and present perfect parallels to the 
 theory of final liquids and nasals given above. 
 
 Remarkable are the N. Fr. forms privauti, communautS , prin- 
 ceauti, primauti, and papauti, since no adjectives *privau, etc., 
 exist by the side of them. They have probably been recently 
 borrowed, and modified on the model of royauti, which itself is 
 partially learned. Another group is formed by the old words, 
 
102 
 
 *horriditatem : ordei, *putiditatem : puteS, nitiditatem : netei, cf. 
 Meyer-L,iibke, Gr., II, 537. 
 
 domesticare : domesckier, cf. domesche, cupiditare, convoitier : 
 convvilise, etc. qf. cupiditat : convoite. Why is e diphthongized 
 in this word ? 
 
 One exception must be made to the rule that of two pretonic 
 syllables, it is the second from the tonic which is syncopated. 
 This concerns all words which contain further derivatives of the 
 suffixes, -cuius, -tulus ; namely all verbs in -culare, all adjectives 
 in -culosus, etc. In these it seems probable that in all periods 
 in the popular language, the " svarabhaktic " vowel was sup- 
 pressed. Of course, the influence of the primary substantives 
 and adjectives in ulus was effective here as elsewhere ; but ow- 
 ing to the fact that even when the simplex does not appear in 
 French, the preserved derivative shows loss of the «, it may be 
 assumed that the Vulgar I^atin subtrata never kept the second- 
 ary vowel, if indeed it was ever a feature of the popular lan- 
 guage. Examples are very numerous. *catabolare : caabler, 
 * cattuculare : chatoillier, *graticulare : graellier, fodiculare : Jooil- 
 lier, * ex-auriculare : essoriller, *semiculare : semillier, *trabucu- 
 lare : travaillier, * ante-oculare ; antoillier, * dormiculare : dormil- 
 lier, adtitulare : atillier, *paricularius : parettlier, * ' pediticulare : 
 petillier, pediculosus : peouillous, * seticulosus : seeillos, etc. 
 
 All of these may have been influenced by the primary forms of 
 the verbs, etc. But one example is found where the possibility of 
 analogy is excluded : mat?-icularius : marreglier. In this the 
 second pretonic syllable persists, since it contains a liquid. I,ater 
 syncope occurred, marglier : marlier, after the complete assimila- 
 tion of the dental, as in serment, etc. The peculiar N. Fr. marg- 
 uillier may be explained by the secondary development of a 
 svarabhaktic syllable from the voice of the / before the tonic 
 vowel. 
 
 As examples of this same syncope in Vulgar I^atin I may cite 
 the following from Schuchardt, II, p. 429: cubiclarius, coaglari, 
 capitlares, commanuplari . 
 
 As has already been stated, protonic a, in whatever syllable it 
 may stand, seems to resist the tendency toward syncope ; owing 
 to its greater sonority, the syllable containing a had at all times 
 a greater stress than the surrounding. Examples. — paraveredus : 
 
103 
 
 palefroi, *loraminarius : loremier, later lormier, (Sixteenth Cen- 
 tury), *ligaminarius : liemier, *materiamenaticum : marrenage, 
 *acqualificare : egalgier. On the other hand, *usatiliare : ostil- 
 lier, the etymon is, however, doubtful, cf. also *terraticare : 
 tergier, cited above, which is earlier than terrag'xer (from Picard 
 texts of the Twelfth Century in Godef). This shows that the 
 law for pretonic a given above is not absolute. Examples of 
 the loss of the immediate protonic syllable which are more diffi- 
 cult to explain are, * cooperticellum .- covercil, Eburodunum : Yver- 
 dun, cominitiare : comencer, experimentat : espermente (Rois) ; 
 but experimentum ; esperement, according to the rule for the 
 suffix -mentum. 
 
 In all these the pretonic syllables are reduced to r so that the 
 preservation of the second pretonic is only apparent, not real ; 
 the spelling covrecel is also found. Likewise cominitiare, by re- 
 duction of all the pretonic sonant elements to that demanding 
 the least stress gives comncer : comencer. fructtfzcare : 
 frotigier, acdiflcare : aigier ; these are hardly popular words ; 
 fructificat should give regularly *fruiteche. Besides frotigier, 
 we also find frougier (cited by Godef. from the Vie de St. 
 Thomas) with regular syncope. 
 
 Examples with regular syncope of the second pretonic sylla- 
 ble, *frigidulosus : frilos, also froidelos (cited by Godef. from a 
 translation of the Fables of Ovid) .■ frieillos ; the last probably 
 secondary forms with Svarabhakti. *animalina : almeline, 
 fructificare : frugier, cf. frotigier, nidificare : nichet , *acuculentus: 
 aigle'nt, aurelianensem : Orlenois, *inde minare : emmener, mor- 
 sus gallinae : morgeline, Corduvensarius : Corvesier, antecessorem . 
 ancessor, * arboriscellum : arbrosseil, * arboricellum : arbroisel. 
 
 In these the retention of the immediate protonic is regular 
 after rb-r ; the diphthong arises through later union with para- 
 sitic i. *ad-subitaneare : asoutener, * annotinensem : antenois. 
 (/.Thomas, Rom. XXI, 506. *dominicare : donoieier ; but cf. 
 domiriicat : donoie. Likewise, *tenoricare : tenroiier, *ad-mmiori- 
 care : amenroiier, *auctoricare : otroiier, etc. * dominicellum : 
 damoisel : dancel. This doublet is a parallel to dame : dans. 
 
 In the more highly accented slower pronounced form the pro- 
 tonic syllable is not entirely reduced. On the other hand, in 
 the more rapidly pronounced proclitic form, m and n are early 
 
104 
 
 assiminated, and the synthesis of the consonantal nasal and the 
 following stop is effected without a voice-glide. *saturejetta : 
 sarroiette, later sarriette. oripelargus : orpres, * salicineta ; sal- 
 noie. This form is interesting for comparison with the treat- 
 ment of the secondary group l-c- as a final in proparoxy tones. 
 {cf. puce, pouce, etc.) It shows that the retention of the ultima 
 in these words is not really due to the consonant combination, 
 else we should have *salsenoie here, but to the amount of stress 
 given the syllable. The protonic i in shlicinita stands in the 
 more weakly accented part of the word and hence had less stress 
 than the final e in sdlicem. cf. moreover frugier, nicher, with a 
 similar reduction of the protonic after long consonant combina- 
 ,tions, d-fd-f, which would apparently need a " Stiitzvokal" if 
 final. 
 
VITA. 
 
 I, William Pierce Shepard, was born in Utica, New York, 
 U. S. A., June 9, 1870. I received my primary education 
 in the public schools of Utica. In September, 1887, 1 entered 
 Hamilton College, Clinton, N. Y., where in 1892 I received 
 the degree of A. B.. and in 1893 that of A. M. In October, 
 1893, I matriculated at the University of Heidelberg, as a 
 student of Romance Philology, where I remained five sem- 
 esters, during four of which I was a member of the Romance 
 Seminar. While there, I attended the courses given by Pro- 
 fessors Braune, Neumann, Osthoff, and by Doctor Schnee- 
 gans, to all of whom, and especially to Professors Neumann 
 and Osthoff I am indebted for much kindly aid. Professor 
 Neumann not only suggested the subject of this dissertation, 
 but his friendly counsel and encouragement have been un- 
 failing during its preparation. 
 
Cornell University Library 
 PC 2837.S54 
 
 Contribution to the history of the unacc 
 
 3 1924 027 253 131