¹ I interpret such Hittite orthographies as e-eš- as meaning 'es-, i.e. glottal stop + vowel + consonant. My reasons for this I hope to develop in a later paper.

² In this and subsequent papers on IE phonology I reconstruct in terms of the four IE ‘laryngeal’ consonants: ' = glottal stop with fronted timbre (*;e- = usual *e-; tautosyllabic *-e' = usual -ē, not to be confused with *ē = usual *ē when lengthened grade of e);; = glottal stop with velar timbre (*;a- = usual *a-; tautosyllabic *-a; = usual -ā, not to be confused with *;ā = usual *ā when lengthened grade of;a); x = velar voiceless spirant (*xa- = usual *a-; tautosyllabic *-ax = usual *-ā, not to be confused with *xā = usual *ā when lengthened grade of xa); γ- = velar voiced spirant (*- = usual *o- [*å-], not to be confused with * 'o-, *;o-, *xo-, *γo-, = usual *o-, respective o-grades of *'e-, *;a-, *xa-, *γ-, = usual *e-, *a-, *a-, *o- [å-]; tautosyllabic *-eγ = usual *-ō, not to be confused with *γ = usual *ō when lengthened grade of γọ, nor with *'ō, *;ō, *xō, *γō, = usual *ō when lengthened o-grades of *'e,;a, xa, γ). Certain ambiguities of reconstruction make it necessary to use cover-symbols; A = any 'laryngeal'; Ạ =; or x, causing a-timbre. In this system there is no place for , ī, ū, , , , . These correspond to _eA (= _e', _e;, _ex, _eγ), iA (or y_eA), uA (or w_eA), (or m_eA), ń (or n_eA), (or l_eA), ŕ (or r_eA). Properly speaking, i and u should be expressed as and . Better yet, all cases of 'i', 'u', '',;', '', and ‘‘ should be analyzed as _ey or y _e, _ew or w_e, _em or m_e, _en or n_e, _el or l_e, _er or r_e, according to whether they are reduction products of ey or ye, ew, or we, em or me, en or ne, el or le, er or re, while i (or ), u (or ), , , and might be reserved as cover-symbols for those cases in which our evidence does not allow us to choose between type _ey and type y_e. Much neater than such orthographies as _ey, (or i) and y_e would be _oy, , y_o, in which _o is a symbol for schwa (murmur-vowel) and syllabification of semivowels, nasals, and liquids at one and the same time, an identification that seems phonemically sound for Indo-European. The true IE vowels, according to the proposed system, are therefore :

1. A.

   A. Full grade

   1. e-type: e, a,

   2. o-type: o

2. B.

   B. Lengthened grade

   1. ē, ā,

   2. o-type: ō

3. C.

   C. Reduced grade

   1. Syllabic:

   2. Non-syllabic: zero

Most Indo-Europeanists will find it awkward, at first, to think from such accepted entities as *dhē- and *wī- (*ī-) to *dhe'- and *w_eyA- or *wy_eA- or *wỵA-, but in the long run they will, I believe, be surprised to find how much more regular, simple, and phonologically satisfactory the new system is than the old. It may be added, by way of preliminary justification of the proposed reconstructive orthography, that IE x actually occurs in Hittite as (= ₁) and, under certain circumstances, in Phrygian-Armenian (e.g., as Phrygian -k in βονό-κ 'woman' < *g^w_ená-x = *g^w_en-; Arm. x in, e.g., sxalem ‘I go wrong, wander’, Skt. skhalatē ‘he stumbles, goes wrong’, Gk. σφάλλομαι ‘I stumble, go wrong’ < IE *sk^wxal-, an e-grade base) and that IE γ occurs in Hittite as (= ḫ ₂) and, after nasals, as -k-(I shall return to this at a later time), is preserved in Armenian under certain definable circumstances as -k- (IE γ and g, g^w probably leveled to g and shifted to k), and, when not initial, seems regularly preserved in Tokharian as -k-. The other two ‘laryngeal’ phonemes, ' and;, are not so directly demonstrable but, in the long run, are just as inescapable. It was necessary to go into this long, and certainly not completely satisfying, preliminary statement because otherwise the schema of IE words for 'tear' can not be presented with due symmetry.

³ See, e.g., Walde-Pokorny 1.769 sub dak̑ru. See also E. Sapir, On the Etymology of Sanskrit áśru, Avestan asru, Greek dákru in Spiegel Memorial Volume 156–9, Bombay, 1908. The present paper supplements and in large part corrects my earlier one. The interpretation there proposed of *dák̑ru as 'biting (water)' (cf. Gk. δάκνω) is untenable because 'to bite' is IE *denk̑- (Gk. δάκ-νω < *d-), to which dak- of dák̑ru is unrelated. This was pointed out to me years ago by Professor R. G. Kent.

⁴ See Trautmann, Baltisch-Slavisches Wörterbuch, s.v. ašarā- (14). ašarà is the more archaic form accentually and ãšara is obviously due to leveling, in certain Lithuanian dialects, with those case forms, e.g. dat. ãšarai, in which the originally oxytone accent was transferred to the first syllable because of certain accent-shifting processes peculiar to Baltic in general or to Lithuanian in particular. The word is listed by Wiedemann, Handbuch der Litauischen Sprache 261, as a circumflected proparoxytone with invariable stress (his class 4, of circumflected sub-type); see also 47. For the original oxytonesis of feminine -ā-stems of three-syllabled words see T. Torbiörnsson, Die litauischen Akzentverschiebungen und der litauische Verbalakzent (Slavica 9, pp. 12 [type ašakà], 47–8, 53 Heidelberg, 1924); H. Hirt, Indogermanische Grammatik 5.257–60; J. Kurylowicz, Le Problème des Intonations Balto-Slaves (Rocznik Slawistyczny 10.56–59 [1930]).

⁵ See Brugmann, Grundr.² 1. §627.

⁶ Op. cit. §522, 1; also Hirt, Indogermanische Grammatik, 2. 85, 86.

⁷ For Baltic -aras, -ara for original *-eras, *era, see Brugmann, op. cit., 2.1. 357 and 1. 238, also Wiedemann, Handbuch der litauischen Sprache §59, 1, and Endzelin, Lettische Grammatik §167.

⁸ One might, of course, say that the expected Lith. *ãš(t)rus could not hold its own because it would have conflicted with the nearly homonymous adjective aš(t)rùs ‘sharp’, hence a transfer to a fem. *aš(t)rà and subsequent assimilation in form to the common type -arà. All of which sounds a little made to order and far from plausible in view of the obvious antiquity of the word and the specific formal agreement of Lithuanian and Lettish.

⁹ Note that in the apparently analogous Lith, vãkaras ‘evening’ we deal with a genuine IE form in *-ero- (cf. OChSl. večer, Lat. vesper, -er-is, -er-ī, Gk. ; there can be no reasonable doubt, in spite of apparent difficulties, that these all belong together).

¹⁰ For examples of such fully reduplicated and secondary broken forms in Indo-European see Brugmann, op. cit. 2. 1, §§70–74; Hirt, op. cit. 4. 6–9. There is, instead, a possibility of connecting the Baltic words and, less plausibly, the Slavic words for ‘tear’ with the Hittite form. See 4 below.

¹¹ There is no reason to think that reductions of etymologically nontransparent elements in compounds would be mechanically identical in form with reductions of morphologically parallel elements when uncompounded. Contrast, in English, board and -board of cupboard, wife and -y of hussy < hūs-wīf. It is, therefore, perfectly conceivable and even probable that w and y, and very likely still other consonant-phonemes, might disappear in Indo-European (or dialectically) when found in semantically obscure syllables immediately before other consonants. That no one has as yet formulated a phonetic law which would allow an old *wdr-' to simplify to *dr-' in certain cases and an old *w_odr-, in apparently parallel instances, to maintain itself as *udr- merely means that the more intricate reduction processes of Indo-European are still in the main to be unraveled. A parallel case is probably that of IE *snusú-s 'daughter-in-law', in which many have felt that *snu- represents a hyper-reduced form of *sunu- ‘son’, in our terms an old *s^wnu-', which lost its -w- largely because in an untransparent compound of this type non-syllabic semivowels caught between consonants might be expected to drop out without a trace.

¹² This simple etymology seems preferable to those based on a supposed *drou-or *drow- ‘wood, oak’, with unlikely ablaut-form of first element. See Boisacq, Dict. étym. de la Langue Grecque, s.v. δροíτη; Walde-Pokorny 1. 804.

¹³ See Boisacq, s.v. λάξ for other words belonging to this group. The resemblance of δρíλακε to δρΐλο ‘earth-worm’ can hardly be other than fortuitous.

¹⁴ See Walde-Pokorny 1. 805. There seems no warrant for the ‘*hölzernes’ in his rendering of drochta except a natural desire to connect the word with the well-known IE set of words for ‘tree, wood’.

¹⁵ See Thurneysen, Handbuch des Altirischen §§165; 281, 282; 292. -a < OIr. -(a)e.

¹⁶ See Pedersen, Vergleichende Grammatik der keltischen Sprachen 2. 47. Pedersen suggests an analysis into *druk-anto- ‘wood’ (of aberrant form in -k-) + formative element -anto- of unknown meaning = ‘beam’ > ‘bridge’. A ‘log’ (or 'pole') thrown over the water would seem to come a little nearer to yielding the desired 'bridge'.

¹⁷ See Walde-Pokorny 1. 806 and, far more clearly, Kluge, Etymologisches Wörterbuch der deutschen Sprache s.vv. Trog and Truhe ('chest'). Walde-Pokorny throw together a whole group of words that cannot possibly have anything to do with each other: 'fest, kräftig, gesund; Trog; Kiste, Truhe; eine Art Saum- oder Packsattel; eine Art Fischkorb; Hartriegel', all supposed derivatives of IE *dru- ‘tree, wood’. This is purely verbal etymologizing, without regard to cultural probability, at its worst. Even Kluge, who is more factual in such matters, remarks, sub Trog: ‘vorgerm. dru-kó-, das man mit Recht aus dem unter Teer behandelten idg. Stamme dru (dreu deru) “Baum, Holz” ableitet; vgl. skr. dru dâru “Holz”: Trog also eigentlich “Hölzernes”?‘

¹⁸ See Walde-Pokorny 1. 249.

¹⁹ Referred by Walde-Pokorny 1. 804, to *dru- ‘wood’.

²⁰ Cf. Grassmann, Worterbuch zum Rig-Veda, s.v. vána n., 9.

²¹ See Wackernagel, Altindische Grammatik 2.1.109, 110.