On s.

Corssen maintains (Ausspr. 1. 294) that s had in Latin three sounds:

(1) Sharp (i.e. a hiss) as initial, and medial before and after other consonants, except n.

(2) Soft (i.e. flat = Engl. z) between two vowels, as now in the Romance tongues, and after n.

(3) Dull and faint at the end of words.

Of the sound of s as s sharp there is the strongest possible proof. For (a) it maintains its place before sharp consonants in st, sp, sq, sc, and it does not maintain its place before flat consonants, e. g. d, m, n, 1, r (§ 193. 2). (b) It changed a flat consonant preceding it to a sharp. It may be said that consul, mons, ars show flat consonants preceding. But consul was abbreviated cos, which shows the evanescence of the n. Mons, ars (from stems monti-, arti-) are instances of the refusal of the Romans, when sacrificing something, to sacrifice all. The ti has already gone: it was necessary at least to write and r to preserve the individuality of the words. But the pronunciation is a different thing. I conjecture that both n and r were in these cases whispered, not voiced (cf. App. A. viii.—x.). This necessity made the Romans unwilling to permit the retention of n and r, when there was no further reason. A whispered r exists in Icelandic (written hr, Ellis, p. 544). A similar whispered r may be presumed in words like prorsum, sursum, which became prosum, susum, by r assimilating to s. But that r as a general rule was voiced, appears clearly from its pathology and influence.

The third sound, attributed by Corssen to s, is inferred from the frequent omission of s in writing, and from its non-pronunciation in early verse (§ 193. 5). I do not know what precise sound Corssen means to give it, nor what it could have, different from s or z, but, this difficulty over, I have nothing to object.

But the second sound seems to me very doubtful. I cannot estimate properly the value of the argument from the Romance languages1. Their list of sounds is not so closely accordant with that

1 Mr Payne (Phil. Soc. Trans. 1868—9, p. 419) doubts the 8 between vo vowels having a z sound in French in the 13th or 14th centuries.

attributed either by Corssen or myself to the Romans, as to render it necessary to suppose any identity of pronunciation in this case. In Italian particularly s has a very different character from what it had in Latin. Witness the combinations sb, sm, sg, sd, sn, sl, sr, &c. There remain three other arguments which appear to me, if they prove anything, to prove that s written was 8 sharp.

(1) The fact that r supplanted s in many words is justly adduced (p. 280) as a proof that s was in these words pronounced like z. But why this should prove that s was pronounced as z in other words, in which this change did not take place, is far from clear. I draw exactly the opposite inference. If s had in these words been pronounced like z, it would have passed to r as in other words. This rhotacism swept over the language like an epidemic, and seized those instances of s as its victims which were predisposed to it by the sound; and it is surely most probable that it seized all such. Reason for discrimination I see none.

(2) Another argument (p. 281) is that an 8 between two vowels, which in some forms was changed to r, in other forms of the same stem was omitted. I cannot see what this proves, except that the flat s which changed to r was sometimes omitted. But the question is, what was the sound of an 8 which was not omitted, and which did not change to r?

(3) The last argument brought by Corssen (p. 284) is that s after n was pronounced, in certain words at least, as if between two vowels, the n being omitted, and consequently, it would have the ordinary sound of s between two vowels, i.e. z (see § 168. 3). On this matter I would refer to the extract from Mr Bell given in App. A. § v. It will be remembered that Cicero tells us that ns, nf lengthened the preceding vowel (§ 167). Now s and ʼn agree in being voiceless continuous consonants. And voiceless consonants are just those before which ʼn is so short, as scarcely to add any appreciable sound to the syllable.' I conclude from these facts that s was a voiceless consonant in this case also; that the n was scarcely audible; but that to compensate for this, the Romans lengthened the preceding vowel, i.e. dwelt longer on the preceding vowel, to signalize the fact of the syllable being more than the vowel +8.

Greek transcriptions show that it was the vowel, not merely the syllable, that was lengthened.

Curtius has made the origin of the long vowel in the nominative case of certain classes of Greek substantives the subject of an instructive essay (as indeed all he writes is instructive), Studien, II. 159-175; and has put forth a theory of the pathology of these cases, which has considerable bearing on the Latin long vowel before ns, a fact which he naturally notices in this connection. I am not sure that in setting out the different moments of the change from e.g. πατέρs to πατήρ, γνώμονς to γνώμων, φέροντς to φέρων, σαφέσs to σαφής (p. 169), he means to imply any chronological interval, even the smallest, between the stages of the change. But there seems to me to be an unreality about it, which makes me unable, at least without explanation, to adopt his theory. He describes the process thus: 'n, r, s before a final s inake the preceding ' vowel long; and thereby becoming weak themselves, run a risk of 'passing, as it were, into the preceding vowel, as is the case in the 'accusative plural (μovoas, musas, for μovoavs, musams). But in'stead of so doing, they draw the following s to themselves, assi'milate it to themselves, take, as it were, the duration of the s, and so recover their full sound of n, r, s.' The unreality of this lies first in speaking of n, r, s, as going through successive stages of weakness and recovery, and secondly in the notion of assimilation itself. The fact is, I suppose, that in Greek and Latin the vocal n and voiceless s were incompatible. The Latin generally, after its wont, and the Greek in some instances (e. g. povσas, doús), made the former of the two give way to the latter; n became voiceless, and the vowel was lengthened by the involuntary dwelling upon it in consciousness of the obscuration of one of the normal sounds of the word. In Greek generally the n won the day, and the so-called assimilation of ns to nn is, in reality, the voice dwelling longer on then and not uttering the s at all, the previous vowel having been, as before, prolonged in anticipation of the loss. If I may use

metaphorical language, the voice may be said to lengthen the vowel just as a leaper presses the ground more firmly before a spring. The speaker is aware of a difficult combination of sounds approaching, and instinctively spreads the time required for the vowel +n+s over two of them, because he knows he cannot apportion it strictly and preserve them all.

In Latin homons became homos; the s was then dropped, partly perhaps, because otherwise a confusion with the acc. plur. of o nouns would be imminent, partly from the slight hold a final s had in the early language. But in by far the majority of -on stems (cf. §§ 449, 450) the o was naturally long. The stems in -r and -1 (which were voiced consonants) repudiated the sharp nominative sign s. Stems in -s with a short preceding vowel and not neuter are rare. In some we have a long vowel in the nominative (e.g. Cerēs, arbōs), in others a short vowel (e.g. venůŭs, vetés, lepés, cinis). Of stems in -t with short vowel, only abies, aries, paries lengthen it in nom. sing.

Origin of ss.

It passes now for a recognized and certain theory among most philologers that ss is in many words the result of a progressive assimilation (§ 31. n.). This assimilation is alleged in (a) the supine stem; and (b) in superlatives and ordinals. It is always assumed by Bopp, Curtius', Corssen, L. Meyer, Schleicher.

(a) Corssen, who especially has defended this theory against all comers (Beitr. 419, 426 sq.), holds that, e.g. tond-tum became tons-tum, and then tons-sum, afterwards tonsum; and that in such cases as mer-sum, lap-sum, &c., where there is no dental at the end of the stem, the change of t to s is due to a false analogy.

Now to this theory there are, as it seems to me, two fatal objections: (1) tons-tum is a perfectly stable sound, and if this form had once arisen, no further change (except perhaps to tostum) would have occurred; (2) there is a whole class of stems forming their supine in -sum (§ 191.3), for which the theory utterly fails to account.

I have not a word to say against the possibility of Corssen's first step; viz. that tond-tum became tons-tum. Neither Greeks nor Romans tolerated two dental mutes coming together. It was important to show the existence of the suffix, and yet tond-tum, if left to the ordinary law, would have become simply tont-tum, and the double 1 Curtius says, 'σT is in all Greek dialects one of the most favourite groups. I only know at the most of one instance of its passing into The passage of st into s is in Latin as frequent, as it is strange in 'Greek.' (Studien, I. 1, p. 241—2.)

σ...

6

t would have been sounded like one only. The Greeks therefore softened the former of the two mutes into s; e. g. ἀνυτ-τός, ἀνυστός; ἀδ-τέον, ἀστέον; πειθ-θῆναι, πεισθῆναι. (Gurt. Gr. Gr. § 46. See also $ 50.) The Romans also adopted this course in cases in which it was important to preserve the t (e.g. in est for edit (edt), eats); and where an r follows immediately, because an 8 immediately before an r is hardly pronounceable; e. g. tond-trix becomes tons-trix (not tond-srix, tonsrix); and compare mulctrum with mulsum. There is therefore nothing against, but much to favour, the first step of Corssen's theory, if dental stems only were concerned. Tond-tum, mit-tum may well have become tons-tum, mis-tum; but why should any further change have occurred? If, as Curtius says (Erläut. zu S147), 'it is a prevalent law of speech that difficult combinations ' of sounds are more bearable, if they have arisen from others yet more difficult, language setting limits to the change of sounds in ‘order to make their origin more patent,' still less is it likely that, when change has secured an easy combination, a causeless further change should be made. The combination st is one of the commonest in the language1; e. g. fastus, festus, § 787; arista, costa, prætexta, &c. § 788; angustus, funestus, &c. § 789; astus, cæstus, &c. § 800; fustis, hostis, pestis, &c. § 803; agrestis, &c. § 808; egestas, potestas, § 811; equester, pedester, &c. § 903; post, ast; ostendo, abstineo, sustento, &c.; est, estis; venisti, audisti, &c.; stella, sturnus, sto, sterno, antistes, &c. Nor so far as I am aware is there a single clear instance of st passing into ss or s2. There is no necessity whatever for assuming that os, ossis, or the old form ossu, are formed from "σreov. The root of os may have had a d or t; in which case this instance would only exemplify the very same difference between the Greek and Roman method of dealing with double dentals, which we are here discussing. Corssen indeed brings forward adgretus, egretus, quoted by Festus as old forms of the past participles; ccmestum is also found in one or two places as well as comesum; and ostensa is found as well as ostenta.

Leo Meyer calls it 'a combination for which our languages (i.e. Greek and Latin) have a general and strong predilection.' (Vergl. Gr. I. 243.)

2 Even in the later imperial times there seems little evidence of such a change. See Schuchardt, 1. 145. III. 75.