We may put the situation in Old French in this way: every a in a final syllable, long or short, persists as though, for some reason, Old French had from the earliest period pronounced every final syllable a somewhat longer than i, e, o, u, similarly placed.

Unaccented final syllables occur in the following types of Latin words:

1. Nouns and adjectives (including pronouns, with a few rare exceptions), in which the final syllable was the inflectional ending of the nominative or accusative, singular or plural, of one of the first three declensions to which the mass of inflections was reduced in early Vulgar Latin;

2. Verbs, whose personal endings and mood forms were kept fairly intact;

3. Adverbs and prepositions, a small list having more than one sylla ble, as ante, super, contra, retro, heri;

4. Numerals (except unus, duo, tres, which fall in with the adjectives); *quattro, *cinque, septem, octo, novem, decem, viginti, *trinta, *quarranta and so on to centum and mille.

Since the force of analogy is strong in Vulgar Latin, the relative importance of these groups in the spoken language is of great moment. It may be roughly estimated for each group on the basis of its relative frequency of occurrence. For instance, a small group of nouns and adjectives belonging to the "i-stems" in the third declension is not kept distinct from the much larger number which have no special forms in their endings.

Analysis of the first five chapters of Tacitus' Agricola (about 650 words) reveals the fact that of all words containing more than one syllable, a safe majority (something over 75%) fall into the noun-adjective declensions; that is to say, of all final unaccented syllables, 75% are inflectional endings of one or another of the noun-adjective declensions (first, second, third). In compiling the data, all nominatives were counted as found, except that feminine nominative plurals in - were counted under -a; all oblique cases were reduced to the accusative, as in Vulgar Latin; neuter plurals were counted, of course, in the same class as feminines of the first declension. Tabulation of the results follows:

The consistency of these proportions is sufficient evidence of their .accuracy. The style of this particular document is narrative, differing, of course, from ordinary conversation (aside from the formality of the language) in a dearth of verbs in the first person and first person pronouns. Also, the passage examined is short. But the essential ratios between the frequencies of the groups are so clear that they must be trustworthy.

It would seem probable, from these figures, that the Vulgar Latin treatment of unaccented final syllables in general would be strongly influenced by the treatment of the inflectional endings (reduced to the nominative and one oblique) of the noun-adjective declensions, in that the development of the noun-adjective endings would become typical for all final syllables, no other single group of which is even one-fourth as large.

"was

These Latin noun-adjective endings originally contained important quantitative distinctions which do not appear very clearly in historic times. In Indo-European the "first declension" marked by the addition of long a to the root of the noun or adjective before the case ending; the "second declension " by the

addition of short -o; while the "third declension" added the case ending directly to the root of the noun or adjective, with no connecting vowel. These case endings in Indo-European were: nominative singular -s (which was omitted in the a-declension feminines and in some Latin second and third declension nouns and adjectives), accusative singular -m; nominative plural -es (lengthened in the Latin third declension, apparently after the analogy of "i-stems" like *tre-i-es > trēs, and entirely replaced in the first and second declensions by a feminine dual and a masculine pronominal ending: -ai, -oi>i); accusative plural -ns (in which n would regularly disappear in Latin with compensatory lengthening of the preceding vowel).

The noun-adjective endings would thus appear as follows in early Latin:

At this point we may remark that it seems not impossible that Plautus' occasional short -es forms in the nominative plural of third declension nouns may have persisted in the vulgar pronunciation; that is, the relatively rare i-stems, though favored by the poets, may not have extended their nominative plural forms to all other consonant stems in the colloquial speech, as has been supposed. It would seem especially unlikely that an inflectional ending should acquire additional length from the analogy of a much smaller group of endings, in the face of a persistent initial stress accent of which the language at least as late as Terence and Plautus shows clear traces. This a priori unlikelihood is further supported by 6 Brugmann: Grundriss.

the fact that a confusion arises as early as 150 A. D.' between the nominative singular -is and the accusative plural, where long -és is required phonetically (for *-ens). This would seem to point to a fairly early shortening of the original long accusative plural ending -ēs, hard to reconcile with a slightly earlier lengthening of an originally short nominative plural -ěs.

If we could trace or justify a retention of original quantity in these inflectional syllables through the republican period in the sermo plebeius, as Greek retains it, we should have a logical phonetic explanation of the differential treatment of final syllables containing a in Old French. If these relative quantities had been retained while vowels in final syllables were becoming progressively shorter, then, no matter how much final syllables were shortened, those containing a would be longer at every stage than those containing e or o; and when at last e and o had lost all length and therefore completely disappeared, there would still be enough left of a to be heard, at least as a "mute" e. A diagram will make this clearer:

(1) Unaccented i, u could not hold length as well as e, a, o, as a matter of general phonetic experience.

"Evidence summarized and bibl. in Grandgent, Vulgar Latin 244; 80 also Probus 95 apes non apis."

66

(2) Here the ō would early become indistinguishable from a close o so that the accusative plural and nominative singular would fal! together, as in fact they did in VL.

(3) The change is complete by the third or fourth century; cf. Grandgent, Vulgar Latin 174, 210.

(4) There is no need to assume retention of "circumflex" quantity in this form after the shortening of -ōs in the accusative plural masculine.

(5) Complete by 150 A. D.; bib. in Grandgent ib. 244.

(6) Compare the development of unaccented close e elsewhere in OF, as securum, levare, to e.

(7) This would be the weakest ending of the whole feminine series, tending to disappear, and early supplanted by an ending identical with the accusative plural (-as). A (surviving?) nominative plural feminine -as is found in Oscan in the first century B. C., cf. Lindsay, Early Latin Verse ш, 30, p. 155.

(8) This final -s is analogical, introduced into OF before our earliest documents appear.

(9) Analogy of the second declension causes loss of historical and phonetic -s here in OF.

The typical final syllable a would then carry with it all the other cases of a in the same situation in other sorts of words, many of which had a long a historically, as cantās, *trintă; that is, every final syllable a would be pronounced with sufficient length to preserve it.

The correctness of this reconstruction rests upon the validity of a single new assumption: that the original length of the connecting vowel in the Indo-European long a declension was not entirely lost in the nominative and accusative singular and accusative plural by the end of the republic, but still made itself heard in the sermo plebeius, and was carried into Gaul.

We have no direct evidence in support of this assumption. In fact, what little we do find seems quite contrary. In the Latin poets, even in Plautus, there seems to be no trace of -ā or -ām in nominatives and accusatives. The grammarians do not notice extra length here, as they notice the changing quality of other vowels; but this is not surprising: an over-long final a whose quality had remained unimpaired would not provoke criticism, and could easily

* Lindsay, Latin Lang. II, 40; Early Latin Verse пI, 4.