clear to me. 163.3: That ecce is from em (e)+ke is very attractive, but phonetically not convincing. 164.2: Osc. -en with the ablative may rather be identical with Latin inde, cf. Brugmann IF 24.75-9. 165 ingēns: I am inclined to equate approximately with Eng. uncouth, separating it from magnus; cf. Conway CR 26.255, also Mackail ib. 251-4. 166 ignis: *egnis, varying with *ognis, would explain all the forms; on the length of the vowel, see my remark to 153. 170.35 Osc. faamat: see rather the new inscription, cf. Buck CP 17.111-8. 172 farferum: Muller fails to call attention to its dialectal quality. 188 fremo: Muller derives from bhr-, rather than from the usual mr-; I see no gain therein. 194 gelu: the irregular e before velar l is overlooked by Muller; it is dialectal or by influence of gelidus. 196 gemini: *genomenoi, with haplology; very unlikely.

201.17: That Umbrian naratu and other forms of the word have only one r, is not significant, since doubled consonants are rarely written, cf. von Planta, Gram. 1.54. 209 gravis: the transposition of sounds from *garuis is more easily explained as after levis, brevis. 213 Umb. -uomu: that this word, with initial y, is equatable with Greek ßwuós must be based on borrowing from Latin; but the word is not found in Latin. Muller's suggestion of error in writing or special phonetic development is weak. 216 ieiunus: Muller's explanation is weak phonetically. 219 imus: from infimus, with syncope; possible, despite Walde's objection. Muller fails to credit Stowasser, Wb., with the suggestion of this etymology. 221 ipse: on this, as well as on ille 218 and iste 222, I cannot agree with Muller in all details. 224 iubeo: the old ioubeatis SCDB should be recognized as the original vocalization of the present, and not treated as an error or analogical to the old perfect. 225 iuvencus: the retention of e before palatal n needs an explanation, given in Walde. 232 lectus: the caption lextos shows a curious consonant-group, which could not have persisted in just that form. 237 Umb. veskla: rather than separate this from uasor, I should attribute the vowel to the influence of vestikatu 'libato'. 238 ligula: the last three lines of the article are not clear to me. 238 līmen: this caption should have -ei-, unless the caption is to be Latin and not old Italic; the same applies to the caption of limes 239. 239 Osc. limu 'hunger': I cannot see why Muller objects to borrowing from Greek Xiuós. Eng. face comes from Latin through French, and has no surviving synonym from AngloSaxon. With such a word as hunger, taboo might easily cause the suppression of the native word.

250.23: I am inclined, despite Muller, to think that Osc. mais etc. are

based on Ital. mag-. 251 magnus: given by Muller without mark of length, despite maximus 250.19; but cf. my note on 153. 252 mālus 'mast': for d by influence of pālus, rather than by Sabine influence; good. 253 mamma 'breast': see rather Walde. 266 medulla: see rather Walde. 268 mille: quite unsatisfactory; cf. TAPA 42.69 ff. 270.36 pōmērium: for the e, see rather Walde. 273 mulier: a better interpretation in Sturtevant, PAPA 50. xiv, as a borrowing from a Mediterranean language. 277 mundus 'world': from *mouendos; good. The short vowel evidenced by Ital. and French can be a later development. 288 numquam: Muller's ně+oinom-quam would produce u; possible, but I incline rather to Walde's view, s.v. unquam (this word not in Muller). 291 nummus: Muller's nomzos appears to me phonetically impossible; I prefer borrowing from (not cognation with) Greek vópos. 293.6: the reading noine 'noni' on the Duenos Inscr. has been discarded by virtually all scholars for thirty years, and should no longer be mentioned even with a 'vielleicht'. 295 occa: from *oketā, with metathesis; hard to believe, despite testimony of cognates. 295 octāvus: a change of ōy to ay if the (initial) accent precedes, does not explain the a in flāvus, adduced among the parallels; also perhaps rāvus and nāvus. 299 omnis: better to ops, with Brugmann (see lit. in Walde, who however does not approve this view).

301 umbra: better in Walde. 302 opinor: here the opeinod of CIL I2 547 must be explained away, on account of the diphthong; I follow Kretschmer, Zts. f. öst. Gymn. 57.497, who explains op einod 'on account of this'. It is therefore not opinor, with d by graphic assimilation to the following devincam ted. 305.36: Osc. urust may, despite Muller, be borrowed from Latin ōro, after rhotacism had taken place; then ōro is denominative to ōs. 307: that urbs and orbis are doublets from the same original, orbis being a non-phonetic Verjüngung, is hardly conceivable. 309 os 'bone': most easily explained as from *ost-s-, zero grade of suffix -es-, cf. Johansson BB 18.24. 313 pālus: the caption pak(s) los> pallos gives a wrong development; the k was lost first, then the s. 313 paelex: good. 316 palla: see rather Walde. 317.21 vespertilio: not in index; for history, see rather Walde. 327 pessimus: the superlative -timus comes into Italic not as -tomos, but as -temos from a reduced vowel, cf. Sommer Hdb.2 457. 329 pello: pultāre should be cited as evidence for the original participle. 331 quinque: the long vowel is not marked, nor is any mention made of the peculiarity, either by Muller or by Walde; but the length is well established by the evidence.

355 possum: to derive potui from *pot[i]fui needs some added explana

tion; and the Oscan present seems to be a direct denominative to potis, shifted to the second conjugation, not a back-formation to the perfect. 361.14 Umb. iepru: another explanation at 218.15, without crossreference at either place. 363 prope: from prog"e; the assimilation to *q*- as in quinque is avoided, not because of the following r (the product would have been *croque; unobjectionable, cf. coquo from *quequò, *peqō), but to avoid separation from proximus. 363 prospěrus: explanation involves some difficult vowel shortenings; but it is an obscure word. 364 proelium: better in Walde. 364 provincia: good. 370 quantus: equation with Greek was very dubious; better *quam-tos, as in Walde. 371 trucido: 'cut into four pieces', good; support for tru- from *qu(a) tru is found in Oscan trutum 'quartum'. 371 quattuor: for the -tt- see LANG. 3.12-4. 372 quartus: caption and explanation are at variance. 372 quadragintā: the prior element is *quaturā, an Italic neuter plural, from which the long vowel spread to the subsequent multiples of ten; but the -d- is inexplicable, cf. my note on 154. 373 querquedula: better in Walde. 377 colo: if the o-grade of the root is original, the verb ought to be a denominative of the first conjugation. 379 ut: originally with adverbial ta; unlikely. Rather as in Walde. 382 rectus: Muller accepts Lachmann's Law; this seems impossible to me, and I explain the lengthened vowels as products of paradigmatic analogy. 384 rima: the caption reik (z) mā indicates a wrong chronology of the development; see my note to 313. 384 rīvālis: I prefer the connection with rīvus, disapproved by both Walde and Muller. 392 rumex: from rug+m-ak-s, according to Muller; but why then the short u, which he does not explain, though he points out the difficulty? Walde says that the g is lost by dissimilation against the following k. The u is short in Romance, despite Muller.

419 sincerus 'aus einem Wuchs': cf. cera from *crērā 85; good. 419 semel: doublet of simul, with adverbial -s; not quite satisfactory, for final el is rare and needs explanation. 420 sine: from sě-ně in proclitic position: good. 421 senex: relation of nom. to sen-is is not explained. 431 līmāx: an obvious borrowing from Greek λeiuag, though Muller does not hint at the possibility. 432 mitto: from smīd-ětō, after Brugmann; I cannot accept this, for the indic. mitat of the Duenos Insc. (cf. LANG. 2.213) has now been found in another insc. 434 sodālis: to *sodos 'road' Greek dôós; good. 447 stinguo 'extinguish': not identical with stinguo 'stechen', but by wrong division from ex-tinguo; good.

465 sublimis: in support of sub limen 'up to the lintel,' cf. Bährens

Glotta 15.53-60. 476 tenebrae: the n from m is not yet adequately explained; perhaps by popular etymology to tenere, quod tenebrae tenent? 479 tenuis: why e here from IE reduced e, but a in maneo? Muller makes no comment; Walde has a suggestion, not very cogent; but perhaps *tanyis became tenuis after *lexuis lěvis, associated in meaning. 482 testa: Walde seems to me to exercise sounder judgment among the possibilities. 483.12 territorium: rather a rime-word to some form in -tōrium, cf. Leumann-Stolz Lat. Gr. 213. 483 terreo: Muller notes that e instead of o in a causative is abnormal, but suggests no reason; it may have followed terror, to avoid confusion with torreo 'parch'. 487 tonitru: contamination of tonitu- and tonitro-; good, but the i is irregular, cf. triquetrus fulgetrum feretrum. The i comes from tonitu-, the other word being properly *tonetru-. 491 taberna: from tråbernā; yes, but why posit the prior r in the caption-form, from which it was early lost by dissimilation? 493 ter: this and the following caption are out of alphabetical position. On accented -ri- becoming -er-, cf. my note on 112. Ter is for *tris after quater: terque quaterque. Tertius followed in Latin and in Umb.; testis (from *tri-stis 495) followed in Latin, but not in Oscan. 494.2 trěcenti: with è in Romanic; yes, by influence of tredecim. 497.9 trucido; the explanation here differs from that on page 371, and there is no cross-reference at either place; this shows the value of a complete index.

ī

502.31 facētus: from stem in facies; see rather Walde. 504 falx; to me the article is in some details not clear. 505-6 filus: from fēlios; but some forms of the root in Sanskrit show a diphthong, and in general I distrust Latin i from earlier e. 507 firmus: with dialectal i, not commented on by Muller; see reff. in Walde. 508.35 Fal. fifiked, flifliqod: quoted wrongly fifiged fifiqont 509.10; see Herbig IF 32.71-8. 511 frāgum 'strawberry': to root of frigeo 'be cold'; curious semantically, see rather Walde. 515 fūsus 'spindle': from pur[e]sos, with loss of r before s; hardly, for in prōsa and suasum the combination is -rss-. 517 vervactum: the medial a, not commented on by Muller, denotes either a short vowel retained by analogy to some other word, or a long vowel; cf. also Robson CR 30.69-70. 519 caption valemos: should have ē, whatever the precise history. 525 via: either pre-form given by Muller meets with phonetic difficulties, cf. also Walde s.v.; apparently it must go back to older *vijā. 531 venēnum, 532 Venus: Walde's semantics seem to me preferable. 535 virga, 536 virgo: on the radical vowel, see rather Walde. 541 vespillo: from ue-spel-, cf. sepel-io; in part like Pott and Pauli (v. Walde); good. 543 vestibulum: Muller

takes up and defends the older view, that it is from the root yes- 'dwell'; good. 549 vindex: not surviving in Romanic; no, but vindico and vindicta survive, and this is not mentioned.

550 vīnum: from earlier form with Italic 7, and therefore not borrowed from Greek; good, but the conclusion is then inescapable that Italic and Greek borrowed the word from different but related nonIE Mediterranean dialects. 558 lupus: I cannot accept a pre-form (y)lukyos; on this and volpes 561, see LANG. 2.188-9. 566 religio: from (u)red+ligiōn 'das hemmende Band', to ligāre; this bases it on taboo and explains the long initial syllable found in poetry, disappearing in standard Latin by the law of mámma-mamilla.

The number of my comments is not to be taken as an unfavorable judgment of Muller's work, but as testimony to the interest which it has aroused. No two etymologists will agree in all details, or on all words and problems. I wish that Leumann's revision of Stolz's Lateinische Grammatik, Laut- und Formenlehre (in von Müller's Handbuch) had appeared in time for Muller to use the chapter on word-formation; but that was not destined to be. All in all, Muller's work takes its place alongside Walde's as indispensable to every serious student of Latin etymology; the two mutually supplement each other, and we are grateful to Muller for this additional assistance to the understanding of the Latin language.

ROLAND G. KENT

A Course in the Analysis of Chinese Characters. Pp. 384. By R. B. BLAKNEY. Shanghai: The Commercial Press, 1926.

This book is intended for students in the early stages of their study; we cannot say for absolute beginners. Due to the dearth of such books and the supreme disregard for the student which most of those that exist show, it will be very welcome.

Under 178 headings Blakney groups a total of some 1300 characters, which he discusses etymologically. The order of these headings is from simple to difficult. Similar forms, although not etymologically related, are noted, and cross-references throughout the book make it possible to analyse each character (radical, phonetic, or borrowed) completely without too frequent repetition.

Though the arrangement is pedagogical, the analysis is etymologic, depending in the main on Chinese authorities, especially on the Shuo Wên and the Seal forms of the characters. Western scholarship, however, is not neglected; Wieger is referred to frequently. Unfortunately,