II. p. 120). But the most interesting is -λevegos compared with the Latin liber and the Sanscrit g-ridh and g-ribh (Benfey, l. c. p. 140), because this furnishes us with another example of the principle under consideration, and leads to the important identification of the Latini and Lavini with the Lithuanians, whose name contains both forms, broken up again into the Lettonians and Livonians (Varron. p. 61). The aspirate presumed in the έ- of ¿-λɛúdɛgos is transposed in the Welsh rhydd, and the same is the case in rhwyf compared with hreblo. To return to lac, the Welsh equivalent is not only blith, but llacth, and there are very many instances in which a prefixed guttural or labial, or even a compound sound, is represented by this double l, the most familiar being the surnames Floyd or Fludd and Fluellen, as representatives of the Welsh Lloyd, Llwyd, and Llewellin. And this leads us to another interesting comparison of ethnographic import. For while we may compare the Aéλeyes (Lleges) with the Alyves (Varron. p. 63; Cambridge Essays for 1856, p. 35), we have a trace of the original prefix in the Dhéyves or Dheyvaι of Boeotia. The extent to which this double I may represent a compound prefix is shown by the Welsh lid compared with the Latin lis (lit-), originally stlit-, Old High German strît, Old Norse strida, Anglo-Saxon flytan, and Greek -o(d)s. The analogy of stlatus (stlatarius) and stlocus for latus and locus, seems to show that t is only an euphonic insertion, and it is omitted in the epitaph of Cn. Cornelius Scipio, where we have sl. judik. for slitibus judikandis (Varron. p. 224). That the s is the residuum of a compound articulation, which, besides the sibilant as a representative of the guttural, contained a labial element, may be inferred from a comparison of stlocus with the Breton leach, the Lithuanian plecus and the Lancashire pleck. The constant interchange of and r, of which stlit and strît furnish an example, will hardly allow us to distinguish between the cases which are now under consideration, and those in which r is preceded by a moveable articulation. Of these cases there are many examples in Greek. Thus, for ρίζα we have the Holic form βρίζα or βρίσδα, and this may be compared with the German vaurts, the Welsh gwraidd, and the Sanscrit bradhna. If the represents an original y, we may farther compare σφριγάω and ἀσπάραγος. Similarly, we know that ῥήγνυμι had the digamma, for Fontis is quoted from Alcaus; and this is confirmed by the analogy of the Latin f-rango, German b-rechen. On the whole we are disposed to regard this "law of divergent articulations" as involving a principle of scarcely less importance than Grimm's rule for the interchange of mutes of the same order, to which it may be regarded as a necessary supplement. The Greek grammarians were content to designate all interchanges of consonants by the general name of uerάhnis (Heraclid. ap. Eustath. 1654, 19). Scientific etymology requires us to regard as entirely distinct cases the passage of one cognate sound into another, e. g. that of p into b; and the appearance of incongruous articulations in words confessedly of the same origin; and while we recognise the possibility of a direct transition in the former case, we must say of the latter, that "none of the known forms are strictly speaking original, but that all have branched out of some still older element, capable, according to known phonetic laws, of producing them all*." = It is right to add that the same phenomenon is observable in forms in which the complex sound is not the original articulation, but has arisen by some subsequent process of consonantal mutation. Thus in Corfu and Negripont, the corrupt modern names of Kέoxvoa and s τὸν Εὔριπον, we have the labial and guttural elements respectively of the more recent articulations Korkura (according to the modern sound of v=v) and Evhripon Egeripon. Similarly in the French we have cage from cavea, Dijon from Dibio, rage from rabies, sache and sage from sapiam and sapiens (Varron. pp. 241, 244), where the complex sound, thus represented by one of its elements, has arisen from a mere synizesis of the vowel e or i, which has been thrown back on the preceding labial, has combined with it the palatal j, and afterwards superseded the original middle-sound, or converted the head of the firm into a sleeping partner. Something of this kind must be the true explanation of the fact that rudere, which has the first syllable short in the classical poets of the best age (Virg. Georg. m. 374; Æn. VII. 16; Ovid, Fast. I. 433; vi. 342; de Art. Am. ш. 290), has the u long in Pers. II. 90. The perfect rudivi shows that there must have been a by-form rudio like rugio, and it is probable that rudere was pronounced rudjere or rugere. The interchange of aspirates of different organs we have before explained (above, § 108). *This description of our principle is taken from the valuable paper by Mr. Garnett, "On certain initial letter-changes in the European languages” (Essays, p. 253). Mr. Garnett (p. 242) states it to be the object of his paper "to bring further evidence in favour of the general correctness of the above theory from some collateral sources of illustration, which it did not enter into Mr. Donaldson's plan to notice;" and he may be regarded as having demonstrated scientifically the proposition, which was briefly stated in the original edition of this work, not merely under the head of the digamma (above, § 110), but also in the present section. Mr. Garnett mentions that "nearly the same view of the subject," as far as the digamma is concerned, has been taken by Höfer in his Beiträge zur Etymologik, a work which we have never seen, but which was published soon after the first appearance of the present book. 122 We conclude this Chapter with a table of the consonants which correspond in related words of the Sanscrit, Greek, and Latin languages. If the reader desires to see this table immediately confirmed by examples, he may consult Pott's Etymologische Forschungen, 1. p. 84 and following. The cerebrals have few if any representatives among the Greek and Latin letters. APPENDIX To BOOK I. CHAP. V. § 110. (A) The Digamma as it appears in Inscriptions. THE words, which exhibit the digamma in extant inscriptions. have been collected and arranged in the order of the dialects by J. Savelsberg (de digammo ejusque immutationibus dissertatio, Aquisgrani, 1854). They are as follows: I. Æolic Inscriptions. a. Bæotic: Fuxías (C. I. n. 1562, 1563, 1564); Fiбotεliav (n. 1562, 1563); Γάρνων nom. pr. (n.1569,am.); Γελατίη nom. urb., Γελατιήν, Γέτεα, Είκατι, Γάστιος (n. 1569, a. Ι.); Ράδων nom. pr. (n. 1574); ἀΕυδός=ἀοιδός, in compounds in ῥαψαΕυδός=ῥαψῳδός (n.1583); BaxɛúƑau nom. pr. (n. 1639); Filαozióvτov (Leake, Travels in Northern Greece, II. n. 31); Fάoxov (Ib. n. 33); Fuxías (Rhein. Mus. new series п. p. 107, n. 7); Faótívios (Ib. p. 108, n. 8); Faótvμudóvtios and perhaps 'ExFáλtas (Ulrichs, Reise in Griechenl. Vol. 1. p. 247); EvFága and Fάorv (Eckhel, Doctr. Num. II. 196). b. Elean: Γράτρα, Γαλείοις, ΕὐΓαοίοις, Γέτεα, Γέπος, Γώργου, Féras (Böckh, C. I. n. 11; Staatshaush. II. p. 390). II. Doric. a. Crissæan: Exoi xλéƑos äñditov alFɛí (Böckh, C. I. n. 1; Ulrichs, Reise, I. p. 31). b. Argive: Fixa[tı] (Böckh, C. I. n. 18); ñɛdάƑoxoi (n. 19); AiFi (n. 29). e. Spartan: dédoFa (n. 15); Fixati (n. 1511); Faóóτvozo (n. 1520; Ahrens, dial. Dor. p. 100). d. Cretan: Fixadiovos nom. propr. (C. I. n. 2598); Favžíwv (n. 3050); Faglov (Eckhel, Doctr. Num. IV. p. 388); Feλyávos (Bull. del Inst. Arch. 1841, pp.61, 174, cf. Hesych. Feλxávos, ô Zevs nagà Κρησίν). e. Corcyræan: Fάoo[tv] (C. I. n. 20); ThaoiaFo (cf. the form of the gen. in the Phrygian inscription, above, § 92), лоúževƑos, goƑaiói, στον Γεσσαν and perhaps ἀριστεύοντα and ἀδυτάν (Trans. Philol. Soc. Vol. I. n. 14; Zeitschr. f. vergl. Sprachf. 1. p. 118). f. Italian (in Magna Græcia): AlFas (Monum. inedit. del Inst. Arch. Rom. Tom. I. tab. VIII. Ix.); τᾶς Ηήρας Μικρός ἐμι τᾶς ἐν πεδίωι· Θύνισκός με ἀνέθηκε, Ὄρταμος Γέριον δεκάταν (on a votive axe, Ephem. Archæol. n. 61; Bullet. Napol.n. 18, p.422); Fétos, Féteos, Γέτει, Γέτη, Γετέων, Γεξήκοντα, Γεξακάτιαι, Γιδίαν, Ειδίαι (ἐν τῇ Γιδίᾳ γᾷ, ἐγδηληθίωντι (ἐξειληθῶσι), δίκατι, δικατίδειον, δικατίTedov, Feinatı, Fé§, Fέxta (Tabula Heracleenses); Foxíav (C. I. n. 4). III. Ionic: [τ]ο αυτου λιθο εμι ανδριας και το σφέλας (Böckh, C. I. n. 10), i. e. ταὐτοῦ λίθου εἰμὶ ἀνδρίας καὶ τὸ ςφέλας, “of the same stone I am a statue and its basis" (Bentley, Correspondence, ed. Wordsworth, pp. 589, 598). (B) Extracts from BENTLEY'S MS. on the Digamma. BENTLEY first quotes the following authorities: Dionys, Halic. 1. 20. Servius, ad Æneados VI. 359. Julianus, Orat. xI. p. 71; and the following passages from "Grammaticæ Latinæ auctores antiqui; edidit Putschius." Diomed. Gramm. p. 416. Priscian, pp. 546, 7 (where, on the words-inveniuntur etiam pro vocali correpta hoc digamma illi usi, ut Alcman: xai xɛîμa não te dάFiov1.—F digamma Æolis est, quando in metris pro nihilo accipiebant, at ἄμμες δ' Γειράναν τὸ δὲ τἄρ θέτο μῶσα λιγεῖα. — Sciendum tamen quod hoc ipsum Eoles quidem ubique loco aspirationis ponebant, effugientes spiritus asperitatem3. Hiatus quoque causa solebant illi interponere F digamma, quod ostendunt etiam poetæ Æolidæ ; uti Alcman: xai xɛîμa não te dáƑiov et epigrammata quæ egomet legi in tripode vetustissimo Apollinis, qui stat in Xerolopho Byzantii, sic scripta: 4EMOПIHOF ON1, AAFOKOFON. Nos quoque hiatus causa interponimus U loco (tov) digamma F, ut DaUus, ArgiUi, PaUo, OUum, OUis, BoUi85— he remarks (1) Si locus sanus est, errat Priscianus: nam in iambico dimetro, cujus hæc forma est: ἐρῶ τε δῆτα κοὐκ ἐρῶ καὶ μαίνομαι κοὐ μαίνομαι- SáFiov facit longam primam syllabam. Sed sæpe apud Homerum correpta syllaba est, ut 'OFicov, "OFias, &c. (2) Non pro nihilo hic F accipitur, sed pronunciatur dFɛgávav, ut nos possumus DWIRANAN. (3) Non ubique et in omnibus Æoli inserebant F, sed in certis tantum verbis, ut ex Homero constabit. Et sic Dionysius supra, nolλá non návτa; et Sergius infra, "in quibusdam dictionibus:" et Terentianus Maurus "nominum multa." (4) Sic Codex MS. Cantabrigiæ, non |