guarantied by Comparative Philology. It is no doubt true that progress in the knowledge of language is to be attained only, as in other sciences, by the constant action and reaction of theory and observation; of the comparison of phenomena in different languages with the special investigation of each for itself. I have chosen the latter part of the work, without supposing that all the secrets of Latin etymology could be discovered by so limited a view. But it is true all the same, that if one's eyes are but armed or practised (and some study of Comparative Philology alone can arm them), a closer and. longer gaze detects something which might otherwise be overlooked. Lastly this is a Grammar of Latin from Plautus to Suetonius. That is to say, I have confined my statements of facts and lists of words or forms (except with distinct mention) to the period from the commencement of Latin literature to the end of the silver age, i.e., roughly speaking, to the three centuries from cir. 200 B.C. to cir. 120 A.D. There are but few inscriptions before 200 B.C. What there are I have of course taken into account. On the other hand, the imperial inscriptions which come within this period are not yet conveniently accessible in trustworthy texts. The silver age I take to end at latest with Tacitus and Suetonius1, and I am convinced that this is as real a division with the line drawn at the right place, as literature admits of. It is quite remarkable how many forms and words are wholly confined to later writers or are used in common with only one or two rare instances in Pliny the elder, Suetonius, &c. Nor can any subsequent writer be fairly regarded as within the pale. The literature of the second century p. Chr. is but small. Aulus Gellius and Fronto are near in time, being indeed contemporaries of Suetonius' later life, but their claims are vitiated by so much of their language being conscious antiquarianism. The lawyers Pomponius, Gaius, Javolenus, Julianus, &c. have perhaps the strongest claim, for they naturally, as lawyers, use a somewhat older style than ⚫their age would imply. Their inclusion however would not noticeably affect the statements. But it is intolerable to find frequently given in modern Grammars, without a word of warning, forms and words which owe their existence to Apuleius or Tertullian— imaginative antiquarian Africans, far removed indeed from insig

1 Suetonius' Lives of the Casars, date about 120 A.D., though he lived to cir. 160 A.D, Teuffel, Gesch. Röm. Lit. § 324.

nificance, and not at all wanting in interest, but certainly not representative of the ordinary or normal language of the Romans. Some other writers, e. g. Justin, Florus, &c. are of too uncertain an age, and too unimportant to be worth considering. Writers of the third and fourth century, however good, are quite inadmissible. Nor am I at all disposed to attach weight to a mention of a word or form in Priscian or other Grammarians, unless accompanied by a clearly intelligible quotation from an author before 120 A.D., or thereabouts. I do not mean that distinct proof can or need be alleged e.g. for every person of every tense of an ordinary verb; but any typical form not shewn to have been used in the period here taken, ought to be excluded from a Grammar of Classical Latin, or mentioned only with the authority affixed. E.g. indultum is usually given as the supine of indulgere, but neither it nor its kin (indultor, &c.) are found before Tertullian; and this fact is seen to be important when it is observed that they deviate from the regular analogy of stems in -1g (§ 191, 3), and that their occurrence is in fact contemporaneous with the use of indulgeri as a personal passive. Again, I have said in § 395 that quercus has no dative singular or dat. abl. plural. But Servius uses (and the form seems right enough) quercubus (Neue, i. p. 376). It should be understood therefore that a statement in the following pages that a form or word is not found, does not necessarily mean more than that it is not found within the classical period. A form or word first found in subsequent writers may be legitimate enough, and the absence of authority for it may be only accidental, but in such cases, the subsequent use does not appear to me to add anything to the evidence for its legitimacy, i.e. it does not make it more probable that Cicero or Livy, or Horace, or Quintilian, or even Plautus might have used it. The character of the formation and the probability that, if no objections had been felt to lie against it, it would have been used by some now extant author, who wrote before 120 A.D., form the real turning points of such a discussion. And to gain a firm basis for the discussion we must have the facts of the normal Latin usage clear from later and inferential accretions. Corssen has made his wonderful collection of facts much less useful than it might have been, by not distinguishing always between later and earlier forms. Of course an exclusion of the later forms from a book like his is not at all

to be desired; but it is thoroughly misleading to put together words first found in the 4th century of the Christian Era, along with well-known words belonging to the ordinary language of the Romans. To take one instance (hundreds might be given); he adduces (Beitr. p. 107; Ausspr. i. § 77) nine substantives in -ēdîn (ēdön, as I call it), which he says are from verbs with -e stems, and stand beside six adjectives in -1do, from six of the same verbs. Now the six adjectives are all well accredited. But of the nine substantives, two only (torpedo, gravedo) are well accredited; one more (pingvedo) occurs once in Pliny the elder, and then not again till the 4th century: one other (frigedo) is quoted by Nonius from Varro; three others are first found in Apuleius, two more not until the 4th century p. Chr. Now these last five words are probably mere creations of a later age in conscious imitation of the earlier words, and, it may be, imitating them, because they were rare. But as soon as we get to conscious imitation by literary speculators, the value of the words as evidence of the proper development of the language is gone.

My authorities then are the writers of the classical period as above defined; and I have not knowingly admitted, without distinct mention, any word which they have not used, or made any statement which their writings critically examined do not justify. But, Donat and Priscian have so long reigned over Latin Grammar, and Latin Grammar has so impregnated literary speculation, that it is next to impossible, if it were desirable, to emancipate oneself from their influence. Still it is important to decline to recognize them as authorities for the grammatical usage of classical Latin, except where they may be taken to be witnesses to facts. They no doubt had access to some writings which are now lost, and they often transmit the theories of older grammarians; but they no doubt also sometimes misunderstood them, they avowedly regarded Greeks as their supreme authorities, they lived when Latin had long ceased to be pure, and they probably would have regarded a statement by Cæsar or Pliny of what ought to be said, as of more importance than the actual fact of what Cæsar or Pliny did say. But it is to the usage, not to the grammatical theories, of good writers that we should look for our standard of right. And for my part, if canons of grammar are to be laid down, I prefer Madvig to any

Roman whatever, and believe Ritschl and Mommsen know a great deal more about the Duellian inscription (§ 467) than Quintilian did.

The arrangement adopted requires a few words.

In Book I. I have thought it important to give a sketch, however slight, of the analysis of vocal sound and of the laws of phonetic change. The special Latin phenomena are treated at some length; but I have been desirous rather that the instances given should be tolerably certain, than that all possible instances should be included. In most grammars these phenomena are collected and arranged under the heads of Omission, Contraction, &c. If any one desires such an arrangement, he can make it for himself, by simply turning to those heads under each letter. But as the primary division of the matter it seems to me much more natural and fruitful to make each particular letter the centre of discussion. Whether it be changed or inserted or absorbed must ultimately depend on the sound it represents and on the relations of this sound to others. The ordinary procedure is the same as if a treatise on chemistry arranged all the phenomena of chemical action under such heads as Explosion, Solution, Combination, &c. Schweizer-Sidler's arrangement by the affections of groups of letters is rational enough, but not, I think, very convenient.

I have distinguished with some care between instances of correspondence and representation (see note on p. 24). The distinction of these two classes of phenomena is ignored in many of the earlier grammars, and is still not unfrequently forgotten. Yet the distinction is of great moment. In questions of pronunciation representation gives very important evidence, while correspondence witnesses at most to the pronunciation of primæval or at least præ-historical times. On the other hand, in discussing the affinities of language, correspondence bears the whole weight of the argument, and representation can only mislead.

The arrangement of the letters has been adopted as the one which best brings into connection allied sounds. Gutturals have a tendency to pass into dentals, and dentals into linguals; and these classes should therefore come in this order. Labials form a class somewhat apart from the rest, and I have therefore put them first, out of the way.

The relations of the nasals are on the whole

more with the labials, gutturals, and dentals respectively than with one another. The order of the vowels is that given by Ritschl, and is the same to a great extent as that given by Corssen. It is without doubt, so far at least as it is common to these two authors, the order of development in the history of the language. Any one referring to Bell's Visible Speech (p. 73), will see that the order has a physiological side also, in so far that the vocal cavity of the mouth is progressively diminished from a in this order to i.

I have not followed Schleicher and others in the treatment of Latin vocalization according to what for brevity I may call Sanskrit principles. This method applied to Latin seems to me to fail both in basis and result. Corssen's elaborate treatment of vowel-intensification in the first volume of his new edition is not more satisfactory; and on this point I can refer to Curtius (Studien, I. 2, p. 294) who commenting on Corssen's sanguine view of the result of his medley collection of long vowels in root-syllables, suffixes and endings, points out that vowel-intensification is "after all only a name for the fact that we often meet with a long vowel, when we expect a short one.' The parts of my Grammar which deal with contraction, hiatus, change of vowel quantity, &c., are far from being what I should like, but there is a great difficulty in arriving at any satisfactory conclusions, owing to our ignorance of the precise quality and quantity of the vowels, which were, or may be regarded as having been, the components of the long vowel or diphthong, at the time when the long vowel or diphthong first arose. Our knowledge of the language begins at a later period, when this process was already over, and we have therefore not facts enough for the historical method. I have little right to speak on such a matter, but I venture to think that the greatest light upon this branch of philology is now to be expected from strengthening the theoretical side of this investigation, but strengthening it not so much by the study of literature and grammar as in Sanskrit, but by a more accurate study of the physiological conditions, and by a closer contact with nature as exhibited in groups of dialects of living tongues. But the application to Latin must in any case be difficult.

In Book II. I have regarded the main division as twofold only, Nouns and Verbs. Adverbs, Prepositions, Conjunctions, have place