Lastly, the same involuntary 8, which has expelled its parent y, is assumed by Curtius to explain patronymics in -Sa, names of beasts in -dev, nominal bases in -ad-, and --: that in these last the & was no essential part of the suffix is proved, he thinks, by the double forms, e.g. μήν-ι-οs and μήνιδος, σφράγιν and σφραγίδα, &c.: he believes the suffix to have been originally long; it then necessarily parted into y before case-suffixes beginning with a vowel, as is regularly the case in Sanskrit (e.g. bhî, bhiy-as); and so, as elsewhere, a & sprang up before this y. These are the principal cases in which Curtius assumes his parasitic S, Corssen controverts some of his results1, not, I think, on very strong grounds: first, because such a 8 could not have arisen after a consonant, e.g. in ypaßdŋv, þúpdŋv, &c.; I am not sure of this; besides, such words might be formed on the analogy of others; there can be no doubt that dnv (whatever its origin) established itself as an independent suffix: secondly, because the forms in -Sinu may be formed from do with the secondary suffix to: thirdly (and this is the strongest argument), that these Greek formations cannot be separated from the Latin adjectives in -do: and that forms like èλπí(8)-s, þvyá(8)-s, are parallel to lau(d)-s, here(d)-s, &c., Kfovídns to Alf-id-iu-s (beside Alfius). It is quite true that there is no trace of the generation of d before y in all classical Latin; but, if these formations be really identical, and not (as seems to me quite possible) the result of independent processes in the two languages, it cannot be said to be impossible that such a principle may have been in operation in GraecoItalian, and afterwards checked altogether in Italian. We have seen that a certain weakness of the d-sound belongs to the Latin as much as to the Greek and such a weakness leads to the wrongful insertion of a sound in some places, as well as to its omission in others.

The strongest argument for Curtius' view is well stated. by Schleicher: "In the stem-formations of the Indo2 Comp. 216,

1 II. 305,

CH. IX.

Estimate of the pre▾

bability

of these

changes.

CH. IX.

.

Germanic, y is an extraordinarily common, d is a rare
element, so that there is hardly another possible way
of bringing these Greek formations into harmony with
those of the kindred languages." This consideration must
at least prevent us from regarding the assumption as
merely arbitrary: and it is much more improbable that
in every case & should be weakened from 7, a weakening
for which the Greek shews no special liking. The argu-
ment brought against the theory, that it is improbable
that one and the same sound should appear in so many
different forms, is answered, I think, with great force by
Curtius. He says: "The less we regard as probable an
isolated deviation with no apparent reason from the path
of regular substitution in the case of those sounds which
remain to all time in common use in a language, so much
the more decisively may we allow sporadic variation for
those sounds in it which we perceive to be from the very
beginning vanishing out of it." Such a sound is especially
y
in Greek: in the earliest records of the language we find
only the imperfect substitutes for it: and it is certainly
not improbable that at a yet earlier period, when it was
still heard, the imperfect attempts to pronounce it may
have produced by its side a letter which is itself indis-
tinctly sounded in Greek, and so in process of time, out of
these two indistinct sounds, one distinct sound may have
arisen. At all events, as Schleicher prudently sums up,
"what every one allows in some cases (Švyóv, Svyór, and
xoés for ghyas) is also possible in others?."

1 Gr. Et. 580.

The instances of the same process in modern languages are well known: e.g. Ital. diacere for iacere, &c.; see Curtius p. 569, note; Ferrar, p. 85. The d heard before an English j, e.g. in John, is nearly parallel; and supplies the strongest argument for supposing that had that sound. See page 368.

CH. IX.

4. Parasitic y.

This is principally found after 8, and therefore pro- Parasitic duces the same results as parasitic & before original y; but y after d. they are much fewer, for y, a sound difficult to a Greek, was not very likely to spring up involuntarily, and clearly could only do so in that early prehistoric time when y had not yet vanished out of the language; so the traces of it are few. It seems to occur in Copκás', the dialectical form of Sopás, where the 8 is original; and, rather oddly, in the same word the y seems to have expelled the 8 and then vocalised itself, for we have the third form oрк-es (nom. plur.) in Hesychius. On the strength of a gloss in Hesychius, δείκηλα, εἰκόνες, and the form δείκελον, which occurs in this sense, Curtius believes that the original form of the common νικ, whence εἰκών, ἔοικα, ἴκελος, &c. was δικ, which produced a parasitic y and then vanished; so that οὐδὲ ἔοικεν in Homer should be scanned οὐδὲ ψέψοικεν, not FéFolkev: this I think is very probable, for there is nothing in the cognate languages to justify a digamma in the word. That & can fall out before (y) seems clear from the well-attested iwrn (√diwk), where, as Curtius has pointed out, the is certainly radical: this loss is only a further extension of the corruption of the dentals which turned diá into Ca. It is not necessary however to assume that laivo is for Staivw. Lastly, as dy, where the y was ἰαίνω διαίνω. radical, was sometimes assimilated to 88, so here also we may explain the peculiarities connected with √d, "to fear." Thus eddeσev is frequent in Homer, and frequent too is the lengthening of a previous short syllable, as μéya тe δεινόν τε, and οὔτε τί με δέος ἵσχει. These become quite intelligible on the supposition that y was produced

1 Herod. IV. 194.

2 Anth. Pal. v. 260.
5 Il. XII. 10.

2 Gr. Et. 607, &c.
4 As Geldart does, p. 32.
6 Il. v. 817; Gr. Et. 607.

CH. IX.

Involuntary aspiration of hard letters; found in Sanskrit and in Greek.

involuntarily after 8, so that Séos was dyeos: and so we need not suppose that eddeσev is a merely mistaken formation, like eμμale, &c.1: the fact that this apparent metrical license is not confined to the verb, but is found also with the nouns, speaks strongly for some real sound being heard after the 8. Further examples (not perhaps equally certain) may be found in Curtius 3.

5. Aspirating unaspirated letters.

This takes place to some extent in Greek. The new h is clearly parasitic, when it is initial; when it occurs in the middle of a word (almost exclusively after hard sounds) it might be regarded as merely a case of substitution; for the aspirate, as has been already said, is a weaker sound than the unaspirated letter: it arises from the stoppage being so short that a portion of the breath has not been appreciably checked; and so it makes itself heard after the cheek is removed. But I have preferred to treat of the whole subject together: though some of these cases have been mentioned incidentally in the account of the Greek aspirates, the theory of their origin is considerably confirmed by the occurrence of this involuntary aspiration. A similar phenomenon is found in Sanskrit; and it may sometimes happen that the same word has been aspirated in the two languages. But there can be little doubt that such coincidences are accidental, and that each language pursued its own course separately in this respect. The aspirates thus found in Sanskrit corresponding to the Greek are always hard ones.

The commonest cause of this parasitic h is the influence of an adjoining liquid or nasal, or a preceding σ*.

1 See Curtius, Erl. p. 46; and above, p. 346.

" Benfey however (G. W. II. 224) supposes the root to be DVI connected with duo, whence Sanskrit √dwish, to hate.

Gr, Et. 604, &c.

4 Gr. Et, 456,

Thus we find φροῦδος from πρό, ἐπίβα-θρον instead of the common suffix -τρον, with κλεῖθρον, λύθρον, and some others: the fact that we have sometimes τρον and sometimes Opov, with no apparent reason for the difference, shews how thoroughly sporadic the change is. We have Tép-pa from √tep, whence Latin tep-eo; the Sanskrit keeps original a in tapas : λύχνος from λυκ-, ἐξαίφνης from ἐξαπίνης, where the nasal is the cause; so also ἔγχος, (where the nasal seems to be intensive, as the root is probably ΑΚ,) αἰχμή from the same root, τέχνη from Ντακ, ρύγχος but ῥέγκω, ομφή from Fεπ; and several others given by Curtius (l. c.). The spirant has acted in oxila from skid, Latin scindo; probably in olé-vw, if this be a strengthened form of STA, which in Sanskrit becomes Astha; in σχελίς by σκελίς and σκέλος, σφυρίς by σπυρίs, and many others. The form σφαλ has been already discussed at length.

In other cases it is not easy to assign any further cause than mere laziness; which operated of course irregularly, but yet affected some words permanently. Such are βλέφαρον, σοφός, and σαφής : the two latter are from

sap, in Latin sapio, &c. Curtius rejects Benfey's explanation that the h may be caused here by a v; that ßλéφαρον is for βλεπ-Fapo-, for -vara is at least a Sanskrit suffix; and σopós is similarly for σoπ-Fo-s: it seems to me not improbable, at least for the first two, and oapns may have been formed on the analogy of copós. I know of no reason for the certain change of /dex into déx-oμai, ντυκ into τεύχω, of νπτυκ (from original γπυκ) in πτύσσω into TTX. The change of the labial is much more common; thus äp-evos is the Sanskrit ap-nas, which is Vedic, but occurs in apnas-vant, "efficacious'," and the Latin op-s, &c.; contrast in-ops with apnas-vant; copia is coορία. From (λιπ comes ἀ-λειφω, from σκαπ σκάφος, and τρέφω may be only a secondary form of τρέπω : further examples may again be found in Curtius.

1 Benfey, Lex. 8.V.

P. E.

2 Gr. Et. No. 653.

26

CH. IX.