Let us be aware that phonetic laws are the outcome of phonetic tendencies and that the latter-no less than other social habits and customs are subject to the dictates of fashion.24 In the domain of language, as elsewhere, the range of fashion is by no means unlimited. As a rule, we find fashions everywhere moving between two extremes, preference being given to the one or the other, as opportunity offers. In the case of phonetic tendencies the two extremes we have to reckon with, are those of vigorous and of restrained articulation. The difference. between the two may be illustrated, e.g., by that of a vigorous handshake as the expression of a cordial feeling and a delicate touch of the hand, meant as a refined substitute for the ordinary way of salutation. At times or in certain surroundings preference may be given to the one, at other times or in other surroundings to the other. In the evolution of phonetics the tendency on the whole runs in the direction of increased refinement. Sounds giving the impression of hisses or noises are softened or entirely suppressed, the complete suppression of a sound marking the ne plus ultra of delicacy. The present pronunciation of the word knight in English as compared with its spelling or with its counterpart Knecht in German may serve as an example. The substitution, in many languages of voiced s for the voiceless s either between vowels (Eng. risen) or at the end of a word (is, his) or as in Modern German, even at the beginning of a word before a vowel (sagen, sehen) should be looked at from the same point of view. In contrast with such changes the phenomena generally designated by the term of Grimm's Law are plainly the outcome of a tendency towards vigorous articulation, the impression of vigor being effected partly by using an abundant amount of breath, partly by adding to the muscular effort. No greater contrast, e.g., than that between Italian strada (a more delicate pronunciation of the Latin strata, and OHG strazza (the identical Latin word modified in accordance with Grimm's Law) can be found. It will readily be seen that, while the mountain climate may favor the tendency toward energetic articulation, it cannot be maintained that either one of the two modes of articulation is dependent exclusively on climatic conditions.

"As to the dependence of phonetic laws on fashion I may be allowed to refer to my review of Osthoff and Brugmann, Morphol. Untersuchungen, vol. I in the Anz. f. deutsches Altertum 5.320-21 (1879). This view goes well with the principle of imitation, on which due stress is laid by E. H. Sturtevant, 'Phonetic Law and Imitation,' JAOS 44. 39-53 (1924).

ON SOME ANIMAL NAMES IN ITALIC

ROLAND G. KENT

UNIVERSITY OF PENNSYLVANIA

Heretofore, in etymological investigations, in the Indo-European field at least, several processes have been overworked, being used to explain quite too many instances of irregularity. These include the following: (1) ablaut variation within the permissible grades, but without any indication of the reason for the variation; (2) cognation with words found in remote branches of the Indo-European family, but without any near congeners; (3) special phonetic laws, set up to explain some two or three examples, and so narrowly defined that they can apply to very few examples.

The necessity of such dubious procedure can be very largely removed, if the etymologist will realize that many words in the various Indo-European dialects were secured by borrowing from non-Indo-European neighbors, many of them no longer traceable, and that many other peculiarities of the vowels and consonants are due to the manifestations. of analogy, especially in the particular form which is known as wordcontamination, the influence of one word upon another. For while word-contamination has been operated with for many years, the wide extent of its application has been realized by but few etymologists. True, H. Güntert has devoted an entire volume to it in Indo-Iranian and Greek,2 and O. Jespersen emphasizes its value in English studies;3 but these and similar results may be yet a long while in gaining a full admission to the etymological handbooks. For this reason I wish to discuss here the apparent irregularities in the Latin names for certain animals, in which word-contamination is the most important factor contributing to the true interpretation-though not the sole one.

1 A. Meillet, Introd. à l'étude comp. d. langues indo-eur. 343.

2 Über Reimwortbildungen im Arischen und Altgriechischen, 258 pp. [1914].

3 Language: its Nature, Development, and Origin, 312-13 [1922].

It seems to be rather better represented in the recent [1926] fifth edition of F. Stolz, Lateinische Grammatik: Laut- und Formenlehre, revised by M. Leumann, in I. v. Müller's Handbuch d. class. Altertumswissenschaft II 2.

1. Latin aper 'boar' had originally an initial e, as is shown by the Teutonic cognates: OHG ëbur, NHG Eber, OE eofer, etc. But the Italic word got an a by the influence of caper 'buck, male goat', as was suggested by Skutsch. The probable early date of the contamination is testified to by the Umbrian, which also has a: apruf abrof 'apros'; by Marsian-Latin Aprufclano;" by Oscan Abella-in Abellanúís etc. if, as is probable, it is from *A pro-la- and means 'Boar-town' rather than from *Agno-la- 'Lamb-town'. The agreement makes the contamination at least common Italic, if not primitive Italic.

2. Latin änser 'goose' was originally *hanser, as is shown by Skt. hqsa- 'waterfowl', Gk. xýv 'goose', OHG gans, OE gōs 'goose'. The loss of the initial h is occasioned by anas ‘duck','as the two kinds of birds are naturally associated. But the final -er, although it has received several explanations,1o still remains puzzling; I am inclined to offer the following: Most earlier monosyllabic nominatives received in Latin new and longer nominative forms, as juvenis for *juvõ, senex for *sõ, canis for *quō, mensis for *mēns, nāvis for *nāus. So also *ghāns, attested in Greek xý as a consonant stem, has in Italic taken a nominative form *hansis, based on the feminine nominative seen in Skt. hąsi, even as in canis," which is discussed later, and in some other nouns, as well as in original

11

"It seems needless to give long lists of the cognate words, which may be found under the appropriate caption words in A. Walde, Lat. etym. Wörterb., and in E. Boisacq, Dict. élym. de la langue grecque. In the same works, ample references to the secondary literature also will be found.

• Krit. Jahresb. i. d. Fortschritt d. rom. Phil. 5. 1. 67. A different contamination is found in Slavic, with a prefixed w: OCS vepri, Lett. vepris 'boar'. This is derived presumably by contamination with the word seen in Lith. verszis 'calf,' Lett. versis 'ox', Skt. vṛša- 'bull', etc., Latin verres 'boar'; as the Baltic words from this root are i-stems, it is possible that they have contributed to vepri not merely the initial consonant but also the shift from o-stem to i-stem. Pedersen, Kuhns Zeitschrift f. vergl. Sprachforschung 38. 311-12, gives no real light, as Meillet says in his Etudes sur l'étym. et le vocab. du vieux-slave 410.

7 But cf. W. Schulze, z. Gesch. lat. Eigennamen 111, 124-5: von Grienberger Indogerm. Forschungen 23. 348-9.

R. von Planta, Gram. d. oskisch-umbrischen Dialekte, 1. 300, 336, 548, where earlier references are given.

• This interpretation appears in K. Brugmann, Grundriss d. vergl. Gram.d. indogerm. Sprachen 12 679 [1897], which seems to be its first occurrence.

10 Brugmann, Grundriss II1 2. 727-8, II2 1. 527 n., makes *hanser- a contamination of *hans- and *hanes-, occurring in different parts of the paradigm; A. Ernout, Eléments dialectaux du vocabulaire latin 109-10, regards the -er as suffixal, of the same origin as -ar(e), and compares NE cat-er-waul and gand-er.

11 H. Jacobsohn, KZ 46. 55-6.

u-stem adjectives and in present participles."2 From *hansis 'goose' and *anatis 'duck' came *ansis and *anats, and the relation of *anates genitive to *anats nominative induced an *ansis-es genitive to *ansis nominative. When genitive anseris developed from *ansis-es, it reacted on the nominative and gave änser in that case.13

3. Latin aries 'ram' corresponds to Umbrian erietu 'arietem'; the Greek cognate shows short e in the initial syllable, those of BaltoSlavic show long e. That the two Italic representatives show an ablaut variation :,14 is unlikely, for the words are otherwise identical in Latin and Umbrian; the cognates in the other languages are equipped with different suffixes, and show a variation e:e, an ablaut series in which a has no place. I believe therefore that the Italic word has *ĕr-, which survived unchanged in Umbrian, but was changed to ar- in Latin by the influence of the associated words caper 'buck, male goat', aper 'boar'.

15

4. Latin canis 'dog', with an alternative nom. canes occurring a few times. The diminutive cani-cula shows by its gender and probably also by its long vowel that the original feminine form of the nominative, inherited in Skt. çuni, played a part in the development of the Latin word. 16 The difficulty lies in the vowel of the first syllable, for Greek

has kuwν, acc. Kuva, gen. Kuvós, and Sanskrit has çvā çvānam çunas; from κύων, κύνα, κυνός which we extract an IE paradigm *kwō *kwonm *kunos (and *keuns). This should yield a primitive Latin *kwō *kwonem *kunes etc., and a Latin *quo *conem17 *cunis. The unduly short *kwō was however re

" The same point recurs in the discussion of canis and of lupus, later in this article cf. footnote 29 for references.

13 The case ending has been drawn into the stem in sper-äre, denominative to spe-s (assisted perhaps by the influence of jūs jūrāre, F. Stolz, Lat. Gram.", revised by M. Leumann, 246); in vir-ès, plural to vi-s (assisted by the analogy of mōs mõris, according to Leumann-Stolz, 1.c.; or by that of vir?); possibly in pūbēs pūber, gen. pūber-is, cf. pūbēs, gen. pubis.

14 Despite R. von Planta, Gram. d. osk.-umbr. Dialekte 1. 283; C. D. Buck, Grammar of Oscan and Umbrian 65.

15 Instances of canes are listed in Neue-Wagener, Formenlehre d. lat. Spr.. 1.284-5. On the gender of canis and canes cf. id. 920-1. It is wrong to regard canis as normally masc. and canes as normally fem. in Latin; F. Sommer, Krit. Erläuterungen z. lat. Laut- u. Formenlehre 192-4.

16 H. Jacobsohn, KZ 46. 55-66 [1914]; against his view, Sommer, Krit. Erl. 1.c. Though the long vowel in canicula might reasonably have a merely metrical basis, as Sommer argues, the gender of the word is decisive, for in Latin the diminutives have the gender of their primitives; Brugmann, IF 19. 215-16, Leumann-Stolz, Lat. Gram. 215. But the abl. sing, is always cane, and the gen. pl. always canum, both distinctive consonantalstem formations, so that the i-stem influence is limited.

17 On this loss of w, cf. Sommer, Hdb.2-3 157-8.

made to *kwonis *conis, even as in the similar extentions of mens-is juven-is nāv-is; and in this the feminine nominative, which after the Sanskrit we may assume to have been (Italic) *kuni-s,18 may have exerted an analogical influence.

But the a of the initial syllable is yet to be accounted for. It is unfair, in my opinion, to set up a phonetic principle which has no other closely similar example, as Walde does;1o and it is forcing the matter to attribute the a solely to the influence of catulus 'puppy', as Osthoff does,20 or to the influence of the verb canere 'to sing', used of many animal voices, or even to derive the word from canere, as Holthausen does,21 though both these lexical associations were made by the Romans.22 A valid explanation seems to have been found by H. Hirt,23 in an alternative ablaut form of the unaccented stem, in which the n and not the w maintained the vocalic function. He shows that in all probability the vocalic nasal before a vowel, however we may write it, developed in primitive Latin (or primitive Italic) into -an-; cf. maneō and μévw, both meaning 'I remain'. There would be therefore in Italic a weak stem which would appear not as *kun-, but as *kwan-, and the paradigm of the first four Latin cases, in the usual order, would be *kwō, remade to *kwonis after the feminine nom. *kwanis; *kwanes, *kwanai, *kwonem. With the regular change of kw to k before short o, and an extension of this loss of w to the other forms, we have *konis *kanes *kanai *konem; fem. nom. *kanis, acc. *kanim. The vowel of the greater number of forms has spread to the others, giving a uniform a: canis canis canī canem etc.; perhaps association with catulus and with canere helped the triumph of this vowel over that of the nominative-accusative, which is normally more potent.24 Such, I fancy, is the phonetic history of this troublesome word.

18 Cf. footnote 12.

19 Op. cit., s.v. caenum.

20 Parerga 1. 251 ff. (not accessible to me; quoted at second-hand).

21 IF 39.66-7 [1921].

Varro, LL 9. 74 ut est cista cistula cistella et canis catulus catellus. Varro, LL 5. 99 catulus a sagaci sensu et acuto, <ut Cato> Catulus; hinc canis, nisi quod ut. tuba ac cornua (quod], signum cum dent[e], canere dicuntur, quod hic item et noctulucus in custodia et in venando signum voce dat, canis dictus. Cicero, ND 2. 14. 38 Chrysippus omnia in perfectis et maturis docet esse meliora, ut in equo quam in eculeo, in cane quam in catulo, in viro quam in puero...

.....

23 IF 21. 167-9 [1907]; Indog. Ablaut §37 [1900]. Despite Walde's objections, Wtb. s.v. canis, I feel that Greek yuh and (Boeotian) Bavá 'woman', Greek kúkλos and Skt. cakram 'wheel', are fair parallels.

24 But not so in dīvos, hiems, perhaps in mel.