¹ MDOG 61.26 (1921).

² For this list of articles cf. Lang. 15.11, fn. 2 (1939).

³ Cf. e.g. SHG 30–2.

⁴ SHG 30.

⁵ Paper read by Sturtevant at the annual meeting of the Linguistic Society of America at Philadelphia, 1939.

⁶ Hittitisch und die anderen indoeuropäischen Sprachen (Kongelige Danske Videnskabernes Selskab, Hist.-filo. Meddelelser, Bind 25, Copenhagen 1938). Cf. also Sturtevant's review in Lang. 14.290–2 (1938).

⁷ Cf. e.g. Luw. hišhiyanzi ∷ Hit. išhiyanzi, with Luwian retention of an initial laryngeal already lost before the Hittite period. Luwian does not generalise the palatalisation of dentals before front vowels, although a tendency in that direction is already observable (cf. the relative pronoun, pl. kwinzi ∷ Hit. kwes).

⁸ The Indo-European character of Lycian has most recently been argued by Piero Meriggi, ‘Der Indogermanismus des Lykischen’, Hirt Festschrift ii, 257–82. For the earlier literature and the texts, cf. Johannes Friedrich, Kleinasiatische Sprachdenkmäler 52 sqq. For Lydian cf. P. Meriggi, ‘Der indogermanische Character des Lydischen’, Hirt Festschrift ii, 285–90. For the earlier literature and the texts, cf. Friedrich, op. cit. 103 sqq. For the problems of hieroglyphic Hittite see P. Meriggi, ‘Sur le déchiffrement et la langue des hiéroglyphes hittites’, RHA 2.1–57 (1932); B. Hrozný, Les Inscriptions hittites hiérogly-phiques, essai de déchiffrement, suivi d'une grammaire hititte hiéroglyphique en paradigmes et d'une liste des hiéroglyphes (Prague, 1937 sqq); I. J. Gelb. Hittite Hieroglyphs, i sqq. Chicago, 1931 sqq.; Gelb and G. Bonfante, ‘The position of “Hieroglyphic Hittite” among the Indo-European languages,‘ JAOS 64.169–90 (1944), and the criticisms of A. Goetze and W. M. Austin, ibid. 65.51–3, 63–4 (1945).

⁹ Thus both palatal and velar k appear in Hittite as k, and the labiovelar series develops precisely as in the other languages of the centum-group, cf. SHG 111–16.

¹⁰ Cf. especially J. Kurylowicz, Études indo-européennes, i; É. Benveniste, Origines de la formation des noms en indo-européen; H. Pedersen, op cit.; W. Couvreur, De hettitische ḫ; E. H. Sturtevant in Lang. 16.81–97, 179–82, 273–84 (1940); J. A. Kerns and B. Schwartz in JAOS 60.18–92 (1940). For the earlier literature see especially F. de Saussure, Mémoire sur le système primitif des voyelles dans les langues indo-européennes; H. Möller, Die semitisch-vorindogermanischen laryngalen Konsonanten, and his Vergleichendes indogermanisch-semitisches Wörterbuch.

¹¹ Repeated, with slight modification, from Kerns-Schwartz, op. cit. 182. Möller's system is difficult to fit into current concepts of the laryngeal hypothesis.

¹² According to Sturtevant-Sapir (cf. e.g., Lang. 17.1 [1941]), ‘a glottal stop of palatal color.‘

¹³ Cf. especially §§24–5, Ch. II below. Sturtevant-Sapir, ‘a glottal stop of velar color.‘

¹⁴ Sturtevant-Sapir, ‘a voiced velar spirant.‘

¹⁵ Annual meeting of the American Oriental Society, New York, 1940; summer meeting of the Linguistic Society of America, Ann Arbor, 1940; annual meeting of the American Oriental Society, Chicago, 1941.

¹ The laryngeals were not generally recognized until long after Brugmann's day, and Hirt seems never to have taken any account of them.

² See most recently, E. H. Sturtevant, Lang. 17, 11. H stands for any voiceless laryngeal, Z for any voiced laryngeal, t for any voiceless stop, and d for any voiced stop.

³ Cf., e.g., Kuryłowicz, EI, 54.

⁴ Not all! Cf. §8.2, and §9.

⁵ WP i, 60, s.v. 2.ank-. It will be noted here, and throughout this chapter, that IH roots are cited without regard for their probable initial laryngeals. Surh initials (in cases of roots with apparent IE initial vocalism) will be treated in the succeeding chapters.

⁶ WP i, 61, cited as an alternant for 2.ank-.

⁷ For the relation of this group of words to a -kh-root-final, cf. Sturtevant, Lang. 17.4 (1941); the material is assembled in WP i, 180, s.v. onogh-.

⁸ WP i, 218, s.v. u̯a-n-q-.

⁹ WP i, 218, s.v., u̯a-n-g-.

¹⁰ WP ii, 19.

¹¹ Cf. WP ii, 23.

¹² WP ii, 26, s.v. pent-.

¹³ Material assembled in WP ii, 368, s.v. reth-.

¹⁴ Sturtevant, l.e., makes this word contain a representation of Iranian -th-. His thesis that IE -th- is represented by Latin -d- intervocalically is both attractive and very possible, and is probably to be extended to Germanic. Nevertheless, unless we are compelled to resort to the equation Iranian -th- = Latin -d-, it seems safer to refer words of this type to a sequence with root-final -d-.

¹⁶ Bois. 561, s.v. λαφύσσω; for the assembled material cf. also WP ii, 383, s.vv. lāb-, labh-(?), lap(h)-. In view of the principle evolved here, it is now clear that the Slavic words must come from a root-final -6–.

¹⁶ Cf. §2 above, and fnn. 2 and 3.

¹⁷ WP i, 131 bot., s.v. emb(h)-.

¹⁸ Cf. WP i, 46, s.v. 2.ap-. Note the -b-alternations, which belong to group 1,d here. The West European a-vocalism is the reduced grade. Lith. ù- represents IE a or o.

¹⁹ WP i, 131.

²⁰ Hirt. IG i, §348.

²¹ WP i, 756, s.v. 2.trep-.

²² So Bois. 919. In the absence of non-Greek cognates, however, it must be admitted that the root-final might just as well have been -ph-. In any event, the relationship between the members of this triplet is certain.

²³ WP ii, 393.

²⁴ WP ii, 403.

²⁵ The Germanic words take on an especial significance in view of Smith's analysis of the phenomenon of Verschärfung, Lang. 17.93 ff. (1941). It will be noted that our words fit perfectly into that analysis, as an instance of a stop plus a laryngeal.

²⁶ WP ii, 403, s.v. leibh-; Bois. 42; Hirt IG i, §348; WH i, 811.

²⁷ That the 6 of Lat. ob is original and cannot represent IE bh is confirmed by the absence of *of in P-Italic, and by the presence of Osc. úp, op, cf. WP i, 122, bot.

²⁸ Lat. opācus < *op- plus formant -āko- is decisive for the voiceless nature of the stop here.

²⁹ WP i, 122, s.v. epi, opi, pi.

³⁰ WP i, 124, s.v. (ebhi?) obhi, bhi.

³¹ Cf. WP i, 777, top.

³² WP i, 776.

³³ WP i, 777, where besides the words meaning ‘tickle, prickle’, several words for pricking, stinging insects are cited, cf. esp. MIr. dega, NE tick.

³⁴ WP i, 115, s.v. eg(h)om.

³⁵ It is, therefore, no longer necessary to explain Arm. es by appeal to some obscure analogy (cf. Meillet, Esquisse², 37, 57, 163), especially in view of the Baltic and Hittite correspondents.

³⁶ Cf. SHG §66. Double writing of the stop corresponds to an IE voiceless stop, single writing to an IE voiced stop.

³⁷ It will be noted that no appeal to analogy is necessary here.

³⁸ WP ii, 257, s.v. meĝ(h)-.

³⁹ Thus clarifying the double writing of the stop in this word (cf. SHG §6Sa), otherwise inexplicable.

⁴⁰ WP ii, 127.

⁴¹ WP ii, 125.

⁴² Cf. WP ii, 126, bot., where, however, Skt. bhād- is denied a place within this root. The length of the vowel in the Sanskrit root is due to the reduplication of the initial syllable of the root, cf. §§30, 31, below.

⁴³ WH i, 522 makes this an extension of *bhōi- etc., ‘fear’, while WP i, 877 makes it an extension of dheu- ‘blow’, etc. The relation of all three members of this triplet to the root ede ‘stink; hate’ (cf. iv, 24 below) seems certain.

⁴⁴ Cf., however, WH i, 499, who connects this word with Lat. fimus ‘excrement’ and *dheu-(WP l.c).

⁴⁵ WH i, 522, s.v. foedus, where connexion with *πιθoς ‘hateful’ is actually suggested.

⁴⁶ WP i, 83.

⁴⁷ WP i, 84.

⁴⁸ I.e., IH -tH- > IE -th-, IH -tZ- > IE -d-, cf. §2, above.

⁴⁹ Cf. Kerns-Schwartz, JAOS 60, 60.

⁵⁰ Already partially conceded by Sturtevant, Lang 17.11 (1941).

⁵¹ The original nature of the root-final stop is, of course, also revealed if an IE root occurs only with a root-final aspirated stop.

⁵² Cf. §le, above.

¹ Cf. §22 below.

² Cf. Bois. 766, 780; WP ii, 660, s.v. (s)pen-, and cf. also pan- WP ii. 5. Lat. pānus is presumably a phonetic reinterpretation of *pannus < *pъn-nó-; Theocritean πāvíσδϵται, is possibly a hyper-Dorism; all the forms cited in the etymological dictionaries can be derived from IE pen-.

³ WP i, 722, Bois. 947.

⁴ WH 262 (the ‘prolonged’ grade of the vowel in the Latin word is due to reduplication of the root with an initial laryngeal, cf. §§30, 31 below), WP i, 398; cf. also i, 390.

⁵ WP onos- i, 132. Simplices will be cited without their initial laryngeals until §24.

⁶ WP 6.pel- ‘flow, pour, wet’, ii, 54; and 7.pel- ‘swamp’, ii, 55.

⁷ Cf. Bois. 901, WP ii, 642.

⁸ IE k̑el- postulated by Petersson; cf. WP i, 496, lines 11 sqq.

⁹ With varying accent in the Hittite and IE words, as IH 'elé-yeu̯e > Hit. lahw-, IH 'elé-yêu̯e > IE lou- ‘wash’ (cf. WP ii, 441).

¹⁰ Cf. WP 4.el- i, 152.

¹¹ WP ii, 18; ii, 19.

¹² WP k̑ad- ‘fall’ i, 339; the reduced vocalism is probably due to IH ge-'ete-yé > IE k̑ed-.

¹³ WP at- ‘fall; go’ i, 41.

¹⁴ Cf., e.g., Hirt IG i, §367.

¹⁵ Cf. Kerns-Schwartz, JAOS 60.185, fn. 8 (1940).

¹⁶ Cf. Kuryłowicz, EI 55: ‘On aperçoit en même temps la différence entre ₂ [= x], qui aspire, et ₃ [= y], qui sonorise’. Again, 254, ‘. . . ₄ aspire les occlusives’.

¹⁷ Cf., e.g., such cases as those suggested in §22 below.

¹⁸ The Hittite word is not exactly equivalent to the IE words, but is probably derived from the IH pé-yere-, with accented prefix; cf. §22 below.

¹⁹ WP 1.bher- ‘carry’ ii, 153.

²⁰ WP 2.dher- ‘hold’ i, 856.

²¹ Connected with the Greek words and Toch. A tsar, B šar in SHG §75 (the Tocharian words, however, possibly come from a guttural with eventual palatalisation). The equation Hit. kessar ∷ Gk. χϵίρ < *χϵσρ is denied in MS 67 (later granted by Pedersen, Hitt. 41) on the ground that this would result in ē in the Greek words; cf. Bois. 1054, where dialect-forms indicate pre-Greek *χϵρ-. The Hittite word was probably recast on the analogy of the -sar-suffix forming action-nouns, itself an analogical creation, cf. SHG §160.

²² WP 4.ĝher- ‘grasp, seize’ i, 603.

²³ That the root-final -k- is not necessarily part of the root is indicated by the paradigm of this verb, viz., 1 sg. harmi, 2 sg. harsi (harti is also attested in KUB 9.31.2.57 = HT 1.2.29 and KUB 9.31.3.1. = HT 1.2.36), 3 sg. harzi, 1 pl. harweni, 2 pl. harteni, but 3 pl. harkanzi. The preterite shows a -k- in the 1 sg. and 3 pl., as do the 2 sg. and 3 pl. of the imperative. Beside this, cf. the paradigm of the verb hark- ‘be destroyed, be lost’, with 3 sg. harakzi, 2 pl. harakteni, 3 pl. harkanzi, etc., with -k- throughout. MS 50 otherwise, but cf. the sb. harharān ‘container for liquids’ (a holder!). But even if the -k- is not an anlogical intrusion within the paradigm, we simply have a suffixed root instead of the simplex, cf. also Lat. arceō ‘enclose, shut up; hold or keep off’, Pedersen, MS 50.

²⁴ The IE words are roots with suffix, e.g., yeré-pe > yrep-. The rough breathing of Gk. ἀρπáζω is thus directly explained without appeal to an alleged initial s- (cf. WP ii, 501). The confirmation of W. M. Austin's thesis (Lang. 17.83–92 [1941]) on the alternation in Greek of ‘prothesis’ with rough breathing, is all the more interesting because the root-initial is a voiced laryngeal; cf. Austin l.c., (fn. 2, and 89, fn. 10), otherwise.

²⁵ WP bhen- ‘best, wound’, ii, 149.

²⁶ WP 3.dhen- ‘beat, push’, i, 853.

²⁷ For the labiovelar here, cf. §27 and table, below. Cf. WP i, 679.

²⁸ The Hittite word is not to be connected with Hit. henk- ‘fix, assign, devote, fasten (a door)‘, etc.

²⁹ WP 1.bhel- ‘shine’, ii, 175; 2.bhel- ‘bloom’ ii, 176; 5.bhel- ‘whitish plant’ ii, 180; all of these are very probably ultimately related. The long radical vowels of some of the forms cited are due to reduplication of the initial laryngeal of the root, cf. §31.

³⁰ WP 2.dhel- ‘light, yellow’, i, 865.

³¹ WP ghel- ‘shine, gleam; yellow, green, grey, or blue’ i, 624; ĝhel-ou- ‘reed (of yellow colour)‘ i, 631.

³² WP elu- 'yellowish'i, 159.

³³ Cf. §7 above.

³⁴ The writer holds that every IE o is the result of qualitative vocalic alternation, cf. Kerns-Schwartz, op. cit., 184, fn. 6.

³⁵ L. H. Gray, 'The Indo-European Base-Type *do-, *do-ie-, *do-u̯e-, *do-əe-' in AJP 62.476–84 (1941), considers this a monosyllabic base with original ŏ-vocalism. The writer considers Gray's extensions to the admittedly IE monosyllabic root to have been earlier infixes, e.g., de-ie-ye (cf. §31 below), while the short forms under the accent, e.g., Skt. dátra- ‘gift’, Gk. δóσις id., come from forms with reduced vocalism and secondary shift of accent.

³⁶ Just a few exceptions to this principle are all that have come to light in all the forma examined in Chap. IV.

³⁷ Of course, prefixation need not necessarily be involved in this type of alternation; an infixed laryngeal would have produced the same effects under unfavorable accentual] conditions without involving the accentual principle observed in the stop-prefixes.

³⁸ Cf. e.g., Lang. 17.1, 11, and fn. 44 (1941).

³⁹ Under Sutrtevant's phonetic interpretation, every such sequence as tx- would have to yield IE tha-, hence every IE ta- would necessarily be considered as of secondary origin; cf., however, such roots as those rubricated in WP; pan- ‘mud, mire, damp’ (ii, 5); panq-‘swell’ (ii, 6); takē- ‘be silent’ (i, 703); 2. tap- ‘press, press together’ (i, 705); stag- ‘trickle, drip’ (ii, 612), 1. qar- ‘abuse, punish’ (i, 353); qars- ‘scratch, curry’ (i, 355); 2. qal- ‘beautiful; sound, healthy’ (i, 356), among many others rubricated, where the cognate forms show no evidence of any vocalism other than a-.

⁴⁰ Cf. e.g., WP bak- ‘staff’ (ii, 104); bal- ‘whirl, turn’ (ii, 109); dak̑ru- ‘tear’ (i, 769); gabh- ‘goat, lamb’ (i, 533); 2.gal- ‘call, cry’ (i, 538), among many others rubricated.

⁴¹ Whether or not the laryngeals exercise vocalic effects, or are found in the neighborhood of certain vowels, is largely an academic distinction. For descriptive purposes, at least, a laryngeal (or the traces of one) seen in contact with a palatal vowel is called palatal; if found in the neighbourhood of a velar (or back) vowel, it is termed velar.

⁴² WP i, 334.

⁴³ WP ii, 76 has a root pou- ‘be alarmed, afraid’, but the alleged cognates are semantically unconvincing, and the root is actually questioned l.c.

⁴⁴ WP i, 38.

⁴⁵ WP top-, i, 743.

⁴⁶ WP i, 457, s.v. qob-.

⁴⁷ For the length of the vowel cf. §30 below.

⁴⁸ WP 1.ap- ‘fit, fasten’ i, 46.

⁴⁹ WP qaus- ‘draw lots’ i, 332.

⁵⁰ WP i, 27. The h- in the Latin word is secondary, cf. WH i, 667.

⁵¹ For the infix, cf. §30 below.

⁵² WP bhad(h)- ‘feel disgust’ ii, 130; bh(o)idho- ‘foul, disgusting’ ii, 186.

⁵³ WP i, 527, with the suggestion that the goat is ‘the springing one’; the animal is notorious, however, for its foul odour.

⁵⁴ WP 1.od- ‘reek’, 2.od- ‘hate’ i, 174, with the suggestion that the roots are probably connected.

⁵⁵ WP 2.bhag- ‘sharp’ ii, 128, and 1.bhag- ‘apportion’ ii, 127.

⁵⁶ WP dhāg- ‘sharpen’ i, 823.

⁵⁷ Cf. WP i, 38. IE ak̑- ‘sharp’, which also belongs here, represents our simplex with a voiceless laryngeal suffix.

⁵⁸ WP ii, 135 ‘bird of prey’.

⁵⁹ Cf. WP ghos- ‘eat, devour’ i, 640.

⁶⁰ WP i, 536 ĝhans-, but some of the satəm-forms point to the velar; contamination of some sort probably underlies those requiring the palatal. For the literature cf. WH i, 52, s.v. anser.

⁶¹ Festus: oscines, aves auspicium ore facientes; there seems to be no reason to reject this explanation in favour of the alternative *obs-cen, EM 716.

⁶² The initial h- is nowhere attested, cf. EM 56 and WH l.c., and while this does not exclude the possibility of Italic *hanser, it is no longer necessary to postulate this form in order to connect the Latin word with the common Indo-European word for ‘goose’. As to the connection with the words for ‘feed, devour’, etc., the goose is a notoriously gluttonous animal.

⁶³ Cf. Kerns-Schwartz, op. cit., §6 B3.

⁶⁴ The dittography of intervocalic stops, e.g., -tt-, corresponds etymologically with IE -i-; the single writing -t- corresponds to IE -d-.

⁶⁵ There is no evidence that the aspiration of stops is other than a specifically Indo-European development; in Hittite, where -h- is the continuant of at least two of the four laryngeals, the -h- does not appear in post-occlusive position except (1) in cases where paradigmatic pressure is involved, e.g., 1 sg. spant-hi, and (2) in the word-group 3 sg. tethai ‘it thunders’, etc., for which no explanation is offered here.

⁶⁶ Every total vocalic reduction presupposes a period during which the particular vowel was being reduced. Precisely as in historical Indo-European, the process of vocalic reduction was sometimes arrested, sometimes halted, and sometimes the reduced vowel was restored to full grade. By ‘partial reduction’ is meant that period or process during which an unfavourable accentual position was shortening a full grade vowel. cf. Kerns-Schwartz, op. cit., §8, and fn. 27.

⁶⁷ The chronological development was probably somewhat as follows: such a sequence as kъX (X representing any laryngeal) > qъX > q^u̯X, whereupon the now juxtaposed laryngeal affected the labiovelar as a unitary phoneme in respect to voice and IE aspiration. Every IH labiovelar would, therefore, imply late total reduction, but the three series for Indo-European as such is accepted here (otherwise Kurylowicz, EI i, 1 sqq.). The guttural series would thus be reduced to a single series in PIE against the two suggested by Kurylowicz, l.c., (who explains, in effect, the origin of the historical labiovelars as due to the effort of the centum languages to pronounce a velar k- before a palatal vowel), by Meillet, Intro.⁷ 91 sqq., and also by Hirt IG i, §§200 sqq. (although with a very different distribution). It is interesting to note how closely Hirt approaches the hypothesis presented here in Hirt-Arntz, Die Hauptprobleme der indogermanischen Sprachwissenschaft (Halle 1939), 220; ‘Wahrscheinlich hat eine noch ältere Zeit nur zwei k-Reihen gekannt; denn das nicht palatalisierte k tritt in der überwiegenden Zahl der Fälle vor a und dunkeln Lauten auf, während das palatale besonders vor e steht. In vorgeschichtlicher Zeit scheint also k vor e, i palatalisiert worden zu sein’. Translated into terms of the laryngeal hypothesis as interpreted in this study, the PIE guttural was palatalised by normal regressive assimilation when it stood immediately before a palatal laryngeal.

⁶⁸ For the significance of the Hittite orthographies cf. fn. 64. Every Hittite t corresponding to an IE (voiceless) t was written double in real or apparent intervocalic position.

⁶⁹ Otherwise Sturtevant, Lang. 16.273 ff. (1940), who has the medial collision -x- develop a glide > -kx- > IE -k- (but if this equation is true, and if a glide -k- is developed, how could the resultant IE -k- escape becoming -kh-?).

¹ By an IE ‘original’ long vowel is meant a vowel whose quantity is not due to later syllabic loss, later compensatory lengthening due to following tautosyllabic loss, alleged lengthening under the accent, lengthening by contraction etc., but the term refers solely to cases where the quantity of the vowel has hitherto been considered ‘long by nature’.

² It is interesting to note that the genetic relationship of these roots has already been suggested, cf. WP i, 122, top.

³ Bois. 322. The rough breathing is analogical, probably borrowed from IE sed-. Hirt's alleged *ē-sdtai < *sed- ‘sit’ is completely disproved by the Hittite form.

⁴ The semantic kinship of ‘be’ and ‘sit’ was challenged by Sturtevant when the suggestion was made in a paper read by Kerns and Schwartz at the summer meeting of the Linguistic Society of America, at Ann Arbor, July 1940. It may be noted, however, that Grassmann, Wörterbuch zum Rigveda, col. 188 sq., recognises the meanings ‘seinen Sitz haben, thronen’, ‘an einem Ort weilen’ (his mg. 3); in the last case also of abstract subjects (his mg. 4), cf. esp. 825.11 yátrā 'nandś ... sate, and other passages (e.g., 19.6, 105.11, 168.3, 727.2 = 737.4, 965.2); for Greek cf. πρòς ἐμ ψυχ- θάρσoς ἡσται Eur., Alc. 604. In all these passages the meaning of the verb is little more than a strong ‘be’. Confirmation of a semantic kinship between ‘be’ and ‘sit’ is strikingly afforded in a linguistic milieu far more primitive than even reconstructed Indo-European. Alf Sommerfelt, La langue et la société, in an analysis of the Aranta language of Australia, states, 81; 'Nous avons déjà vu que la racine na signifie “être assis”, et qu'elle s'emploie aussi d'une façon que nous devons traduire par “exister” . . .' Cf. the ‘paradigm’ of the verb ‘to be, exist’, 59 sqq. ‘Be’ and ‘sit’ are certainly not incompatible semantically. In Indo-European the meaning ‘sit’ was eventually reserved exclusively for the reduplicated form of the PIE root; but even here the meaning ‘be’ was never quite completely lost.

⁵ WH i, 65, connects the Latin word with āreō ‘be dry, dried out’ < IE ās- ‘burn, glow’, but this is quite unsatisfactory semantically. WP i, 79 rubricates a root āro- ‘das Freie, Weite; Raum(?)‘.

⁶ Gray, op. cit., operates with infixed u̯e- and i̯e-.

⁷ Cf. WP ed-, i, 118.

⁸ The length of the initial ū- in cognates from this root is not due to an ‘original’ long diphthong, but to a specifically IE partial reduction eu- > ъu- > ū-.

⁹ Cf. § 24.1 c, above.

¹⁰ It will be noted that the prehistoric nasal infix, in sharp contrast to its probable descendant (the historic presential nasal infix), runs throughout the paradigm.

¹¹ Cf. IV. 17 below.

¹² Cf. IV. 31 below.

¹³ Cf. IV. 49 below.

¹⁴ KZ 37.387 sqq. (1900).

¹⁵ IG i, § 346, and cf. § 1 c, above.

¹⁶ Zupitza, op. cit., 391. Interestingly enough, he offers a number of ‘triplets’ which fit nicely into our scheme: ‘Gelegentlich sind alle drei Stufen, media, media aspirata und tenuis, bezeugt . . .’ His explanation, however, 393, is that the earliest Indo-European probably made no distinction between the three ‘grades’, and that all three were pronounced indifferently.

¹⁷ WP i, 342.

¹⁸ s.v. qap- WP i, 344; cf. also ghab(o)lo- ‘fork’ WP i, 533.

¹⁹ WP 3. (der-) etc., ‘run, tread’ i, 791.

²⁰ Perhaps in 4. ter- ‘cross, over, through’ WP i, 734, and 5. ter- ‘over’ i, 735. Extensions in tragh-, tregh- i, 752 and treq- i, 755.

²¹ Cf. WP i, 874.

²² Whether or not this (like some other roots) was in imitation of the sound of sheep or other animals, the root became subject to modification as soon as it passed into the language.

²³ Cf. WP ii, 120.

²⁴ Cf. 8. bhel- WP ii, 812.

²⁵ WP i, 447.

²⁶ WP 1. gal- i, 537.

²⁷ Monier-Williams 338, col. 2.

²⁸ WP 1. (s)teg- ii, 620.

²⁹ But not necessarily, since Greek exhibits both initial and medial σθ.

³⁰ WP postulates a root o(d)ego- i, 175.

³¹ Cf. ozdos WP i, 186.

³² Or with suffix, ‘e-se-ge-Ze > IH ‘esgZ- > IE ezgh-, s.v. ozgho- WP i, 185.

³³ WP 1. perd- ii, 49.

³⁴ WP ii, 68.

¹ WP i, 256.

² Otherwise, e.g., Bois. 270, s.v. ἄπω, and WP ii, 487, s.v. sep- with quite unsatisfactory semantics.

³ WP i, 257, Feist 561.

⁴ WP ii, 524.

⁵ WH i, 356. The i for u in dis-sipō, etc., is not phonological, cf. Leumann-Hofmann 85.

⁶ WP 2. rebh- ii, 371.

⁷ Bois. 277.

⁸ WP 1. rebh- ii, 370.

⁹ WP top- i, 743, Bois. 975.

¹⁰ WP ii, 370, EM 861.

¹¹ WP i, 721.

¹² WP i, 741.

¹³ WP ii, 502, Bois. 283.

¹⁴ WP s.v. at- i, 41.

¹⁵ WH i, 51, Feist 62 s.v. aþ-nam.

¹⁶ WP i, 118.

¹⁷ Contra Bois. 242.

¹⁸ WP ii, 486, but cf. Bois. 685, who connects this root with sed- ‘be seated’.

¹⁹ WP i, 251, Bois. 293.

²⁰ WP i, 364.

²¹ WP s.v. i, 339, and cf. § 13.3b above.

²² WP ii, 49, Bois. 771.

²³ WP ii, 68, EM 748.

²⁴ WP ii, 691, WH ii, 74.

²⁵ WP i, 467, Bois. 530 s.v. κυθνόν.

²⁶ Cf. WP i, 114, WH i, 25, Bois. 324.

²⁷ WP i, 105.

²⁸ Cf. WP i, 133 s.v. enq-.

²⁹ WP i, 634, Bois. 154.

³⁰ WP ii, 477, Bois. 255, WH i, 702 s.v. inquam.

³¹ WP i, 204; cf. also WH i, 716 and Bois. 304. For the alternation of the palatal with the labiovelar cf. §§ 27–8 above.

³² WP i, 110, Bois. 300.

³³ WP ii, 414.

³⁴ SSS 449, WP ii, 44, EM 794.

³⁵ Cf. WP tolq- ‘speak’ i, 744.

³⁶ WP 2. leuĝh- ii, 414, Feist 334, and WP ii, 415, Feist 333.

³⁷ WP ii, 496, Feist 419.

³⁸ WP i, 245, Bois. 269.

³⁹ WP i, 169 s.v. oq^u̯-.

⁴⁰ The Hittite word better here because of the labiovelar; actually, the initial s- is due to a prefixed or infixed se-.

⁴¹ WP i, 169, s.v. oq-.

⁴² WP ii, 366.

⁴³ WP ii, 381, Bois. 585.

⁴⁴ SSS 463.

⁴⁵ WP ii, 659, EM 962, SSS 448.

⁴⁶ WP i, 510, Bois. 950.

⁴⁷ WP i, 456 s.v. koktu-s.

⁴⁸ SSS 450.

⁴⁹ Cf. WP i, 34, WH i, 60, EM 64, Couv. 89. The IE a- vocalism must stand for a reduced vowel. For a discussion of the relationship of ‘drink’ and ‘water’, cf. MS 49.

⁵⁰ Bois. 1080.

⁵¹ Bois. 997, cf. WP i, 248, s.v. u̯eg^u̯.

⁵² EM 865.

⁵³ Feist 397.

⁵⁴ WP i, 726, Bois. 947, EM 1041.

⁵⁵ WP ii, 300, EM 610.

⁵⁶ WP i, 306.

⁵⁷ SSS 484.

⁵⁸ WP ii, 515 s.v. s(u̯)aq^u̯o-s.

⁵⁹ WP i, 747 s.v. tu̯aq-.

⁶⁰ WP i, 759.

⁶¹ WP ii, 280 s.v. meregh-, Bois. 133.

⁶² WP i, 306.

⁶³ WP i, 41, s.v. aĝhl(u)-, Bois. 103.

⁶⁴ WP i, 185, Bois. 725.

⁶⁵ WP ii, 361.

⁶⁶ WP ii, 48.

⁶⁷ WP ii, 308, Bois. 646.

⁶⁸ WP ii, 346.

⁶⁹ WP i, 124, WH i, 400, where cf. suggested ‘rhyme-roots’ i̯em- and nem-.

⁷⁰ WP ii, 330, Bois. 662, Feist 375.

⁷¹ WP i, 207.

⁷² WP i, 458, s.v. k̑om; cf. also WH i, 251, Couv. 177, Bois. 420, LS 832, SHG § 111.

⁷³ Cf. WP ii, 493.

⁷⁴ WP i, 260, where the root is confused with words for ‘desire, love’.

⁷⁵ WP i, 727.

⁷⁶ WP i, 724–5, Feist 494.

⁷⁷ WP i, 513, Bois. 766.

⁷⁸ WP i, 726.

⁷⁹ MS 48.

⁸⁰ WP i, 182 s.v. ōr-, or-, Bois. 72.

⁸¹ WP ii, 676, Feist 445.

⁸² WP ii, 527, s.vv. 1. su̯er- ‘speak’ and 2. su̯er- ‘whisper’.

⁸³ MS 74.

⁸⁴ WP i, 283, Bois. 229.

⁸⁵ WP ii, 313.

⁸⁶ WP i, 823, Feist 554, EM 1088.

⁸⁷ WP i, 142, Bois. 270.

⁸⁸ WP i, 285 s.v. 10. u̯er-, Bois. 297.

⁸⁹ WP i, 142 s.v. 5. er-, EM 852.

⁹⁰ WP ii, 356 s.v. reu̯os, Feist 401, EM 879.

⁹¹ Cf. Feist 25 and EM 75, where it is pointed out that there is no trace of dissyllabism of the historical root, and that it is, therefore, independent of, e.g., Lat. arō ‘plough’ and its IE cognates.

⁹² WP ii, 671, Bois. 894, Feist 446, EM 964, where, curiously enough, the structural similarity of sternō is called to attention.

⁹³ WP ii, 638, EM 1034 s.v. terra and 1021 s.v. tellūs.

⁹⁴ EM 959, cf. also Bois. 927 s.v. σφαραγϵσθαι.

⁹⁵ EM 776, Bois. 792.

⁹⁶ Cf. EM 974 and 528.

⁹⁷ Feist 456, EM 989.

⁹⁸ Cf. WP i, 143 s.v. erdh-.

⁹⁹ WP i, 136 s.v. 3. er-, Couv. 93–98, Feist 401, Bois. 714, EM 712.

¹⁰⁰ WP ii, 497 s.v. 1. ser-, EM 932.

¹⁰¹ WP i, 149 s.v. eras-, ers-, ras-, WH i, 416.

¹⁰² WP i, 155, Bois. 239, WH i, 25, Feist 38.

¹⁰³ WP ii, 668 s.v. 1. sp(h)er-, and II, 52 s.v. 4. pel-, Bois. 889, EM 934, also Bois. 744 s.v. πάλλω, and EM 750 s.v. pellō.

¹⁰⁴ WP i, 727 s.v. 1. ter-, esp. as base for such sequences as tr-em-, tr-ep-, tr-es-, q.v.

¹⁰⁵ WP ii, 671 s.v. sp(h)er(e)g- and ii, 48 s.v. 2. perg-, SSS 450, Feist 146, Bois. 927.

¹⁰⁶ WP i, 756 s.v. 1 trep-, Bois 979, EM 1054.

¹⁰⁷ WP i, 758 s.v. trem-, Bois. 982, Feist 501, EM 1054, SSS 442.

¹⁰⁸ WP i, 760 s.v. tres-, Bois. 984.

¹⁰⁹ EM 1035, where connection with tremō is suggested.

¹¹⁰ Bois. 287, 720.

¹¹¹ WP ii, 675, Bois. 895.

¹¹² WP ii, 675, Feist 444.

¹¹³ WP i, 139, Bois. 76.

¹¹⁴ WP i, 268 s.v. 3 u̯er-, EM 1137, Bois. 729, SSS § 44 e.

¹¹⁵ WP ii, 507 s.v. selos-, Bois. 246.

¹¹⁶ Bois. 729.

¹¹⁷ WP ii, 530 s.v. 1 su̯el-.

¹¹⁸ WP ii, 66, Bois. 789.

¹¹⁹ WP ii, 50.

¹²⁰ WP i, 488 s.v. kros-no-, Bois. 515.

¹²¹ WP ii, 441 s.v. lou-, Bois. 587, WH i, 773. Couv. 189–94.

¹²² WP i, 495 WH i, 239, Bois. 473.

¹²³ WP i, 152, Bois. 243, Feist 8.

¹²⁴ WP i, 144 s.v. 1. erē-, rē-, Bois. 283.

¹²⁵ WP ii, 371, Bois. 328, Feist 398.

¹²⁶ WP i, 144 s.v. 2. erē-, Feist 394.

¹²⁷ WP i, 740.

¹²⁸ WP i, 143 s.v. 1. erē-, Bois. 275, EM 859.

¹²⁹ WP ii, 665, EM 960.

¹³⁰ WP i, 431, Bois. 448.

¹³¹ WP ii, 48.

¹³² WP i, 102, SSS 423, 458.

¹³³ WP ii, 241, s.v. 3. mei-, WH ii, 73.

¹³⁴ WP i, 228, Bois. 367. MS 73 regards this root as a pR.

¹³⁵ WP i, 106 s.v. 1. eis-, WH i, 717, Bois. 362.

¹³⁶ Cf. MS 62.

¹³⁷ WP i, 361, WH i, 213, Bois. 462.

¹³⁸ WP i, 748.

¹³⁹ WP ii, 348, Feist 393.

¹⁴⁰ WP i, 103, Bois. 369.

¹⁴¹ WP i, 104, Bois. 694.

¹⁴² SSS 425.

¹⁴³ WP ii, 12, s.v. pēu-: peu-: pū; cf. EM 743, Bois. 740; the difficulties raised in these passages are easily removed by the application of the principle of the alternation i̯-/u̯-.

¹⁴⁴ WP ii, 615.

¹⁴⁵ WP ii, 553, s.v. (s)k̑eu-.

¹⁴⁶ WP ii, 659, Bois. 895, EM 820.

¹⁴⁷ Cf. fn. 144 and Bois. 989.

¹⁴⁸ Cf. fn. 144 and Feist 451, EM 1064 and 990.

¹⁴⁹ Cf. fn. 144 and EM 990, Bois. 991.

¹⁵⁰ WP i, 755, EM 1060, Feist 535.

¹⁵¹ WP i, 254 s.v. 1. u̯edh-.

¹⁵² WP ii, 556, Feist 9.

¹⁵³ WP i, 160.

¹⁵⁴ Cf. § 30.2 and fn.

¹⁵⁵ SSS 444.

¹⁵⁶ Cf. Couv. 327.

¹⁵⁷ WP i, 168.

¹⁵⁸ WP i, 168 s.v. ōus-.

¹⁵⁹ WP i, 403 WH i, 198.

¹⁶⁰ WP i, 96, Couv. 89, 102.

¹⁶¹ Cf. Couv. 102.

¹⁶² Bois. 692, EM 1128, Feist 24.

¹⁶³ WP i, 452, WH i, 192.

¹⁶⁴ Feist 285, Bois. 972.

¹⁶⁵ WP ii, 488 s.v. 2. sem-.

¹⁶⁶ Cf. WH i, 1.

¹⁶⁷ WP i, 47, Bois. 69, Feist 3.

¹⁶⁸ WP i, 52, Bois. 105.

¹⁶⁹ WP i, 824, Bois. 333.

¹⁷⁰ WP ii, 229, WH ii, 49, EM 595.

¹⁷¹ WP i, 166 s.v. ord-, WH i, 16, EM 16.

¹⁷² Lit. ‘self-stink’. Cf. WP ii, 521 s.v. su̯oid-, Bois. 366.

¹⁷³ WP ii, 420 s.v. leq-: leg-, WH i, 743, Bois. 555.

¹⁷⁴ WP i, 246, WH i, 268.

¹⁷⁵ WP ii, 130, Bois. 781.

¹⁷⁶ WP i, 38, EM 29 s.v. agrippa.

¹⁷⁷ WP i, 28, WH i, 7, EM 9, Bois. 32, Feist 17.

¹⁷⁸ Cf. WH i, 7. Feist 17 b, bot.

¹⁷⁹ Feist 54.

¹⁸⁰ WP i, 7, Bois. 31.

¹⁸¹ WP ii, 127, 128, Bois. 1010.

¹⁸² WP i, 823, Bois. 343.

¹⁸³ WP i, 12 s.v. airā, Bois. 28.

¹⁸⁴ WP ii, 131 s.v. bhares-, WH i, 461, cf. also 455, 457.

¹⁸⁵ WP i, 606 s.v. 3. gher-, Bois. 1051, LS 1980.

¹⁸⁶ WH i, 619, Feist 220.

¹⁸⁷ WH i, 616, Feist 220.

¹⁸⁸ WP ii, 135, WH i, 96.

¹⁸⁹ WP i, 86 s.v. 2 al-, WH i, 31, EM 37, Feist 34.

¹⁹⁰ WP i, 90, WH i, 33 (where relationship with alō is suggested).

¹⁹¹ Bois. 41.

¹⁹² WP ii 176 s.v. 2 bhel-, WH i, 523, Bois. 1041; there is no basis for postulating e-vocalism in this root. The Greek and Latin cognates both indicate IE o- < a-/o-alternation.

¹⁹³ WP i, 825, Bois. 332.

¹⁹⁴ WP i, 705, Bois. 939.

¹⁹⁵ The ‘set’ base of 2 bhel-, WP ii, 176; cf. WH i, 518, Feist 100.

¹⁹⁶ Bois. 41.

¹⁹⁷ WP i, 92, Bois. 48.

¹⁹⁸ MS 63, SSS 424, WP i, 2, Bois. 28.

¹⁹⁹ MS 63.

²⁰⁰ WP 1, 2 s.v. 3 ai-, Bois. 105.

²⁰¹ WP i, 376, Bois. 393; the forms cited exhibit i̯-/u̯-.

²⁰² WP i, 665, Bois. 1011.

²⁰³ WP i, 4, Bois. 23, WH i, 20.

²⁰⁴ WP i, 17, WH i, 80.

²⁰⁵ WP i, 6, Bois. 31, WH i, 21, Feist 30.

²⁰⁶ WP i, 293, EM 1132, Feist 576.

²⁰⁷ WP i, 284 s.v. 9. u̯er-, EM 1089, Bois. 709, Feist 551.

²⁰⁸ WP i, 635 s.v. ghou-, Feist 207.

²⁰⁹ WP ii, 316, Feist 372.

²¹⁰ WP i, 832, Bois. 335; cf. also Feist 153, s.v. filu-deisei.

²¹¹ WH i, 80, Bois. 29.

²¹² WP i, 12.

²¹³ Cf. Couv. 144–146.

²¹⁴ Bois. 72.

²¹⁵ WP i, 705.

²¹⁶ WP i, 457, s.v. qob-.

²¹⁷ WP i, 46, WH i, 75.

²¹⁸ LS 141.

²¹⁹ WP i, 352, WH i, 51, Bois. 405.

²²⁰ Bois. 933.

²²¹ WP i, 44, WH i, 11, Feist 60.

²²² WP i, 189, Bois. 997, Feist 537.

²²³ WH i, 75, cf. esp. the discussion of the semantics; Feist 63.

²²⁴ Feist 46, where the connexion with Hit. anda is quite inadmissible semantically as well as phonologically.

²²⁵ WP i, 65, Couv. 110, WH i, 43, Bois. 64, Feist 46.

²²⁶ WP i, 24, Feist 65.

²²⁷ WP i, 38, s.v. agen-: ang-, Bois. 6.

²²⁸ WP i, 24, cf. also WH i, 84, s.v. aulla, olla.

²²⁹ WP i, 487, 594, Bois. 523.

²³⁰ WP i, 9, WH i, 16.

²³¹ Cf. Bois. 42.

²³² WP i, 133, WH i, 701, Bois. 12.

²³³ EM 663, Bois. 666.

²³⁴ WP i, 7, Bois. 30, Feist 30 (q.v. for Low Gmc. forms without the extension exhibited in Greek and Gothic).

²³⁵ WP i, 182 s.v. orĝh-, Bois. 721.

²³⁶ Monier-Williams 220.

²³⁷ WP ii, 3.

²³⁸ WP i, 484.

²³⁹ WP ii, 317, Bois. 670.

²⁴⁰ WP i, 25, Bois. 104.

²⁴¹ WP i, 62, WH i, 47, Bois. 10, Feist 13.

²⁴² Feist 487.

²⁴³ Like the precisely identical tamenk-, which also belongs to this root; -men- must be a local type of infixation.

²⁴⁴ WP ii, 480.

²⁴⁵ WP ii, 481, Bois. 302, Feist 419.

²⁴⁶ WP ii, 482, Bois. 862.

²⁴⁷ WP i, 704, Bois. 944.

²⁴⁸ WP ii, 2, Bois. 777, EM 721, Feist 134. There is no evidence for ĝ- in this root, especially since the Balto-Slavic cognates alleged for it are semantically unsatisfactory.

²⁴⁹ WP ii, 5, Feist 142.

²⁵⁰ WP i, 500.

²⁵¹ Cf. Feist 142 s.v. fani.

²⁵² WP i, 78.

²⁵³ Cf. Pedersen, Hitt. p. 177.

²⁵⁴ WP ii, 500, EM 896, Bois. 81.

²⁵⁵ WP i, 744 s.v. torianā.

⁵⁵⁶ WP i, 355, see also 2. qers-, i, 429, ‘draw furrows, plough up’; cf. SHG 137, Couv. 325.

⁵⁵⁷ EM 1086, Feist 586, Couv. 119, SSS 51.

²⁵⁸ WP i, 220, Bois. 17.

²⁵⁹ Bois. 102.

²⁶⁰ WP ii, 85, Bois. 798.

²⁶¹ WP i, 13, WH i, 79, EM 1078.

²⁶² WP i, 312, Bois. 366.

²⁶³ WP i, 146, and cf. Bois. 74.

²⁶⁴ WP i, 847, Feist 121.

²⁶⁵ WP i, 852, Bois. 334.

²⁶⁶ WP i, 866.

²⁶⁷ WP ii, 328, EM 662, Feist 371.

²⁶⁸ WP i, 771, Bois. 180 (but Lat. redimiō suggested here is better analyzed as red-imiδ; cf. WP i, 207).

²⁶⁹ WP i, 134.

²⁷⁰ WP i, 256, Feist 211.

²⁷¹ WP ii, 152, Feist 93.

²⁷² WP i, 608, Feist 90.

²⁷³ WP i, 323, s.v. ṇgni-s, EM 473, WH i, 676 (q.v. for the evidence against assumption of an IE ṇgnrs).

²⁷⁴ WP i, 181, s.v. ong-, Grassmann 14.

²⁷⁵ WP i, 717, Bois. 936.

²⁷⁶ WP ii, 187 s.v. bhōgō.

²⁷⁷ Bois. 1029, WH i, 510.

²⁷⁸ WP ii, 186 s.v. bhok-, WH i, 521.

²⁷⁹ WP i, 849, WH i, 466.

²⁸⁰ Cf. Bois. 346, 1032.

²⁸¹ WP ii, 495.

²⁸² WP ii, 326; Couv. 122–37 connects the Hittite word with IE aig-: ing- ‘sick’, etc., but this is semantically, as also phonologically, unsatisfactory. Again, since the Hittite word always has the meaning of ‘death; pestilence’, it seems quite unlikely that it belongs to the verbal root henk- ‘fix, assign, devote, fasten (a door)‘.

²⁸³ Cf., however, MS 50, Couv. 111, where the k of Hit. har(k)- is considered part of the root, and the root is equated with IE ark- in Lat. arceō, etc.

²⁸⁴ WP ii, 369, EM 852.

²⁸⁵ WP i, 280, Bois. 297.

²⁸⁶ Cf. Bois. 1054.

²⁸⁷ WP ii, 172, Feist 75–6.

²⁸⁸ WP i, 155, Bois. 1081.

²⁸⁹ WP ii, 159 s.v. 3 bher- ‘sharp tool, cut, bore, scrape’, and ii, 162 s.v. 6 bher- ‘protrude, edge, point, corner’; cf. also WH i, 481 s.v. feriō and i, 526 s.v. forceps.

²⁹⁰ WP i, 602, Bois. 1051.

²⁹¹ Cf. WP i, 629 s.v. ĝhel- ‘cut’, and Feist 215 s.v. gilþa.

²⁹² WP ii, 196.

²⁹³ Essentially an alternant of the preceding; WP ii, 194.

²⁹⁴ WP ii, 174, Bois. 771.

²⁹⁵ WP i, 159 s.v. clu-.

²⁹⁶ WP i, 321, Bois. 319.

²⁹⁷ WP ii, 175 s.v. 1 bhel- ‘shine’ and ii, 180 s.v. 5 bhel- ‘whitish plant’; cf. also Bois. 1013.

²⁹⁸ WP i, 865 s.v. 2 dhel-.

²⁹⁹ WP i, 624, cf. also i, 631 ĝhel-ōu- ‘tortoise (of yellow colour)‘.

³⁰⁰ WP ii, 166.

³⁰¹ WP ii, 214.

³⁰² WP i, 108.

³⁰² WP i, 546, WH i, 645, Bois. 1053.

³⁰⁴ WP i, 167, EM 718, Feist 70. In Hit. hwitar, the -tar is a noun-forming suffix, cf. SHG § 156.

³⁰⁵ WP i, 306, SSS 468, Feist 567, and cf. JAOS 60.185 fn. 8 (1940).

³⁰⁶ WP i, 111, Feist 97.

³⁰⁷ WP ii, 140, WH i, 557.

³⁰⁸ WP i, 835.

³⁰⁹ WP i, 843, Feist 121.

³¹⁰ WP i, 642, Bois. 344.

³¹¹ WP ii, 171, WH i, 551, Bois. 1017.

³¹² WP ii, 147, Bois. 776.

³¹³ WP ii, 252 s.v. 2 meu-, SSS 456, and cf. Bois. 57 s.v. άμύνω.

³¹⁴ WP ii, 323, Bois. 665.

³¹⁵ Cf. WP ii, 470 s.v. 2 seu-.

³¹⁶ WP i, 834, Bois. 342.

³¹⁷ WP ii, 144, WH i, 556, Bois. 1022.

³¹⁸ WP i, 257.

³¹⁹ WP i, 255.

³²⁰ WP i, 249, Feist 212; Hit. weh-, wah- ‘turn, fall, versari’ does not belong here either phonologically or semantically.

³²¹ WP i, 249.

³²² Cf. 16 above.

³²³ Instances of Verner's Law in initial position were claimed by S. Bugge, PBB 12.399 (1887); 13.311–2 (1888), and discussed by A. Noreen, Abriss der urgermanischen Lautlehre, 125 sqq. (Strassburg, 1894), and by E. Prokosch, Comparative Germanic grammar, 68 (Philadelphia, 1939). Bugge, however, could not have used this example, as the requisite IE initial t- could not have been equated with the initial d- underlying Gmc. *tō, e.g., OE tó. In this study, however, the alternation of t and d finds a ready explanation. Somewhat different is the explanation of J. Schmidt (KZ 26.355–6 [1882]) and B. Delbrück (IF 26.356–7 [1907]) of du as a proclitic form of Gmc. *tō (cf. Feist 127), which would involve a consonantal change parallel to, but distinct from, Verner's Law.

³²⁴ WP i, 639.

³²⁵ WP i, 358.

³²⁶ Possibly certain other instances of vacillation between single and double writing in Hittite may reflect a similar state of affairs, or at least analogical imitations thereof.

³²⁷ Cf. WP i, 826, WH i, 360. For PIE the essential structural difference between the members of this group of verbs lay in the presence or absence of a formative suffix; this continues to be the case, though with some phonological obscurations, in Hittite; here the sg. 3 of the verbs falls together, and the two can be distinguished semantically only when in context. In IE, where the different ‘orders’ of the stops early came to be stereotyped as unitary phonemes, the essential difference lay between initial d- and dh-; minority forms with the ‘wrong’ initial (e.g., the pl. 3 of deye, where deyé-nte > IE *dhenti, but was analogically reformed as the apophonic var. donti), were analogically reformed with the ‘right’ initial; by way of further distinction, the o-grade came to be used universally in the continuant of deye, only rarely in that of deye-ye. Over a large part (not all! cf. Lith. dúomi) of the IE area, reduplicated presents of both verbs came to replace the older unreduplicated ones; the new reduplicated present had a somewhat different accentuation, and in some IE dialects secondary reductions were thus brought about, e.g., reduplicated pl. 3 dé-dontr > dé-dṇti > Skt. dádati; cf. also Skt. pl. 1 da-dmás, probably for an earlier *dá-dmas. So, too, in substantive derivatives from this root, instances of deye continuants in IE with a short vowel can have achieved the short vowel only by contamination or analogy. Louis H. Gray, AJP 62.476–84 (1941), on the other hand, operates with a short monosyllabic base do-.

³²⁸ WP i, 645–58, contains no entry of IE *ĝr- or *ĝhr-, but only gr-, ghr-; so, too, Brugmann, Gdr. i, §§ 596–629. On the other hand, k̑r- occurs frequently.