Now adgretus and egretus are of course, if genuine, simply instances of the preservation of the starting-point common to all theories: comestum is, I believe, the only instance of a perfectly possible way of dealing with the double dental in these forms, but may be equally well regarded as the sister, not the mother, of comesum: ostenta may be an instance of the same, but is probably to be regarded as belonging to ten-ere, while ostenga belongs to tend-ĕre. (On infestus, &c., quoted by Corssen, see my note on p. 220. On hæsum, hausurus see below.) It can hardly be said that there is any evidence for the change of st

into s or ss. My second objection to Corssen's theory is this. All the verbs which form their supine in -sum may be divided into three classes; the first, stems of which the final consonant is a dental mute, viz. t, d; the second, stems in which the final stem consonants are 1 or r preceding g, 1, or r; the third consisting of a miscellaneous list of verbs, all of which are however characterised by the active perfect (if they have one,) being in si (§ 705). Now this third class, not a large one, (lapsum, jussum, pressum; parsurus, mulsum from mulcere; fixum, fluxum; mansum; censum, hæsum, hausurus), but containing stems ending similarly to other verbs which have -tum, exhibits probably the result of various laws, and the -sum may be partly due to a kind of attraction exercised by the perfect. At any rate no light on its origin is derivable from Corssen's theory. second class, which is quite as numerous, is pervaded by words of the character named have the supine in -sum. this is utterly alien from Corssen's theory. No one will suppose thạt sparg-tum, mulg-tum, fall-tum, curr-tum became spar-sum, mul-sum, fal-sum, cur-sum by passing through the stage of spar-stum, mul-s-tum, fal-s-tum, cur-s-tum. (Torreo with stem torsmade torstum, then tostum, and there stayed.)

a

But the

law: all And yet

Two other grounds for hesitation in accepting Corssen's theory may be mentioned. (1) It supposes a progressive assimilation, whereas this is very rare indeed in Latin1. But I admit that it is possible. Its probability however is very small. Moreover (2),

1 It is rare in other cognate languages too, if I may judge from Schleicher's Compendium. (For the definition of progressive and regressive assimilation see note to page 12. Kühner (Ausf. Gr. Gr. § 64 Vol. I. p. 209, ed. 2) reverses the usual application of these terms.)

the fact should be noticed that stems ending in s originally do not follow the change prescribed by Corssen; e. g. ges-tum does not become ges-sum. But there are three verbs in which such a change

Now censeo is origiperhaps a secondary

appears to be found; censeo, haurio, hæreo. nally a t- stem (comp. Kevréw), and is derivative from census, which would in that case be the participle of a lost consonant verb. Hausurus is quite anomalous. I find it only in Verg. A. IV. 383, and an imitation in Stat. A. 1. 667, twice in Silius, and possibly once in Seneca (see p. 247). Hausturus (Cic.) and haustus show the regular supine stem. Hæreo is,

I suspect, an r stem (comp. aipéw, though h in Latin does not usually correspond to an aspirate in Greek), and owes its supine stem (hæsurus, hæsito) to the long penult (comp. curro, verro).

The theory which I oppose to Corssen's is, as applied to the dental stems, that tt, dt became first ts, ds, and then ss or s. This theory fulfils the really necessary conditions of truth as completely as Corssen's fails. For the first step is equally applicable to all stems, inasmuch as it supposes the suffix only to be primarily affected: the second step is inevitable if the first be admitted. Ts, ds are utterly unstable (in Latin), and must become entirely sibilant.

The only objection which I can see to this theory is that the phonetic cause of the change of dt, tt, lgt, bt, &c. into ds, ts, lgs, ps, &c. is not apparent. But neither is the phonetic cause of the change of at to st. Possibly a good phonetician, like Mr M. Bell or Mr Ellis, may find a reason for the change where others cannot. I content myself with referring to the fact that in Greek before T frequently changed to s (e. g. onơi, Dor. parí), páσis by the side of φάτις; εἴκοσι, Dor. εἴκοτι; ἀναισθησία from ἀναισθητός (Schleicher, Vergl. Gr. § 148; Curtius, Gr. Gr. § 60); and sometimes before v; e. g. σú, Dor. tú, Lat. tu (but also σoû, σoî); and to the word ipse, which is generally taken to be for ip-te; noxa for noc-ta; capsa, a box, from capere; &c. (See App. A. xxvi.) A double t is found in a few words; e.g. Attius, blatta, cette, Cotta, cottidie, quattuor, gutta, matta, sagitta, vitta, futtilis, littera, Mettius, mitto, and, it may be, some others. The question is perhaps one rather of spelling than of pronouncing. But, considering the frequency of suffixes commencing with t, the paucity of words with double t is striking. As I hold, the Romans had two ways of avoid

ing it; they changed tt either into ts or into st. The first produced an unstable combination, and therefore passed on into ss or s. This was the course usually adopted. The second produced a stable combination, but was (in Latin) used only where the first would occasion further difficulties of pronunciation. (Comp. §§ 151. 2; 152. 3.)

(b) The double s in the superlative form of most adjectives is also supposed by Corssen (see esp. Ausspr. II. 550 sqq. 1022, ed. 2) and many philologers to have arisen out of st. The superlative is formed from the comparative stem in ios, by adding, as is supposed, tumus (or timus), and compressing ios to is, as in magis for mag-ios (Ausspr. II. 215). My objection to this is the same as in the former case, the extreme improbability of a stable combination like st changing to ss. Corssen states that his theory of this change rests principally on his explanation of the two words sinistimus, sollistimus, as being superlatives. But, even granting that these are superlatives, the fact would only show that a superlative might, not that it must, be so formed. Here again, as I conceive, the Latins had two modes of forming the superlative; either by a suffix -umo, or by a compound suffix -tumo. It is agreed that some superlatives are formed by the simple addition of a suffix (without a dental) -mo or -йmo to the positive or comparative stem (see § 754; Corssen, I. 780); e.g. primus, minimus, plurimus, &c. I think that the easiest way of explaining the formation of the ordinary superlative is by adding the same suffix -umo to the comparative, and regarding the doubles as accentual and phonetic only, i.e. as representing the length of the syllable and the sharp sound of the s. It is possible to explain the superlative of words in 1 or r, as having the same suffix, but attached to the stem of the positive, and to give a similar explanation of the double 1 or r. But the suffix may have been appended to the comparative, and then a wig contraction have taken place, facilios-imus, facilīsumus, facilsimus, facillimus. This presumes indeed a progressive assimilation. But 1 and r seem to me the only sounds in Latin which show signs of such an influence; comp. velle for volere, turris by the side of Túpois (Corssen, Beitr. p. 402), and the evident incompatibility of 1 and r with a final s; e.g. consul for consuls, pater for paters. Corssen considers miserrimus to be for misersimus, and that for misertimus; 'for t after r and 1 is frequently changed to s' (Ausspr. 11. 552);

but with this I cannot agree. I do not remember any instance of t after a simple r or 1, preceded by a short vowel, becoming s. The instances are after rr, 11, and haurio, hæreo; on which see above, p. lx

Corssen accounts for the s in ordinal numbers on a similar hypothesis to that which he applies to the superlative forms; e.g. vigesimus for vigenstimus, for vigentitimus. As in the case of the supine form I suppose nt-t to have become nt-s, and then necessarily ns-s, and easily ns or s only. (Cf. §§ 757. c. 168. 3.)

Jxlix, 1.

Besides the above-named classes of formations we find ss also (c) in the old futures; e. g. prohibessit, levassit; (d) in arcesso, capesso, &c.; (e) in verrucossus, &c.; (f) in essem, amavissem, amavisse, &c.; (g) in assis, bessis, tressis, &c.; (b) to represent Greek C.

The cases of double s in prohibessit, &c. are explained in § 622; and here I have the authority of Corssen in supposing the double s to be due to accentual considerations (Ausspr. II. 563, ed. 2). For arcesso, &c. see § 625; for verrucossus, §§ 168. 3. 813. Essem, &c. are, I imagine, cases of a natural union of s with a suffix beginning with s, §§ 609, 610. Assis is of uncertain origin, unless it be a masculine formation from the same stem as æs, but retaining its s and therefore defending it with a double s.

The doubles used to represent the Greek ċ, at least in early times (§ 189), was perhaps really from the Dorian σơ. Whether this ss represented the sound of s or sh (which Curtius once attributed to σo) is not, as it seems to me, certain. Sh appears to my tongue and ears a more natural result of ky, ty, which are regarded as the origin of σo in Greek, than s (Curt. Gr. Gr. § 57). But see above, pp.

It appears. to be generally conceded that the sound of sh was not unknown in Italy. The Etruscans had it (Mommsen, UnterItal. Dial. p. 6); the Greek alphabet of Cære had it (p. 15); the Umbrian had it (p. 22); perhaps also the Sabellian (p. 24) and Oscan (p. 26). Moreover the Romance languages have the sound, Ital. and Wallach. c before i and e being=tsh; Spanish and Provencal ch=tsh; Portuguese and French ch=sh. (See Diez Gram. Vol. 1.)

On the assumption made in the above explanations that ss may stand merely for s, or at least for s when some letters or syllables have been extruded, I will only observe that the fact that the early Romans wrote no double letters (§ 58) seems to me a very important one. For, when first the Romans took to writing them double, what clue did they follow? It is possible that they followed the pronunciation, as an Italian now makes a difference between such sounds as ǎto and atto- -a difference which Englishmen do not make. (Comp. also Ellis, Early Engl. Pron. p. 56.) I am not sure whether Quintilian's language (quoted in note to p. 58) should be taken to imply a real difference in pronunciation, though the word dixerunt looks like it1. But, when the practice of writing double letters came in, it is surely very probable that they were guided, at least to some extent, by etymological theories; and thus, though I regard the supposition that levasso arose by assimilation from levaviso as unsound, I think it by no means unlikely that the notion of a syllable being dropt justified to the popular apprehension the spelling levasso".

On the vowels, especially o and e.

The exact determination of the quality of the vowels is a problem which scarcely admits of satisfactory solution. Descriptions of vowel sounds are worth very little, and the ancients had no full list of customary or possible vowels, derived, either from observation of provincial pronunciations, or from analysis of vowel

1 Mr Munro (in his note on Lucr. III. 545) quotes Servius on En. 1. 616: applicat; secundum præsentem usum per d prima syllaba scri'bitur: secundum antiquam orthographiam, quæ præpositionum ultimam 'litteram in vicinam mutabat, per : secundum vero euphoniam per a 'tantum ;' and adds, 'i. e. only one was sounded. In this, as in so 'many other points, it is clear that the artificial modern Italian pronunciation is directly contrary to that of the old Latins, with whom causa 'and caussa, excisus and exscissus, were identical in sound.' In his note, Lucr. III. 504, he points out the striking instances of mamma, mămilla; offa, Ŏfella; tintino, tintinnabulum; Porsenna, Porsĕna; Catillus, Catilus; and perhaps currus, curulis; quattuor, quăter; littera, litura. See also on III. 1044. And comp. Ŏmitto, Ŏperio, § 784. "A similar account may be given of dissicio, porricio for dis-jicio, por-jicio cf. § 144, 2 and 143. (I doubt these being analogous to aλos for alius, &c. on which see Curt. Gr. Etym. p. 592 sqq. ed. 2.)