¹ In this paper, I shall not discuss in full the evidence of Hittite; it may be appropriate here to state my reasons for not discussing it. Briefly, I do not feel that the Hittite evidence proves the laryngeal theory. Even if there were no difficulties with the Hittite etymologies, they would not prove the theory correct; it would still be necessary to show that no other theory will do. But in fact the Hittite evidence for the laryngeal theory is not certain: at best it provides only feeble support. Furthermore, as we all know, the laryngeal theory had been around for a long time before the equation of Hittite with PIE /H/, so that it is justifiable to criticize the theory as a theory without taking into account developments in the daughter languages which tend to support it. O. J. L. Szemerényi, Trends and tasks in comparative philology 12 (London 1961), takes much the same position:

In other words, the evidence is far from being unequivocal—as one is usually led to believe. It is contradictory and raises the question whether the explanation of the Hittite sounds transcribed as h does not lie in an entirely different quarter. This does not mean that de Saussure's laryngeals must disappear; they are probably here to stay, but on a far less lavish scale than recent discussion would have us believe, and on purely structural grounds, not on the strength of Hittite ‘evidence’.

² For the sake of the argument I here assume that the triple reflex of in Greek is unoriginal, and that the regular phonological development of was Greek /a/. This means that thetós and dotós are analogical reformations of *thatós and *datós on the basis of the presents títhēmi and dídōmi respectively.

³ It is difficult to find forms with Greek /ī/ that have cognates with /ī/ in both Latin and Sanskrit, so pairs of cognates will have to serve. For Latin cf. vīs ‘force’ : Gk. îs ‘strength’, for Sanskrit cf. pīvan- ‘fat’ : Gk. ‘id.‘.

⁴ PIE also possessed the diphthongs /ei ai oi eu au ou/.

⁵ All phonemes and phonemic sequences are to be understood as appearing between consonants or initially before a consonant unless otherwise specified. A few notes to the table are necessary. (1) Some laryngealists prefer, for reasons not connected with the laryngeal theory, to include [i] and [y] in the phoneme /y/. They would therefore write /y w/ where I have /i u/. (2) I have here given a typical phonemic analysis of the PIE vowels in laryngealist terms, and have allowed only one phonemic writing for each segment. I have not indicated the origin of these phonemes and phoneme clusters. Since laryngealists frequently state their phonemic analyses in historical terms, it may be well here to indicate some of the possible origins of the sequences given in the text.

These may not be all the possibilities, and not all of those who operate with laryngeals would accept all these origins, but the table shows a number of ways in which a given phonemic sequence could originate. For much of the information in this footnote I am indebted to Warren Cowgill, who has been good enough to provide me with his comments on this paper.

⁶ Curtius Studien 9.367 (1878). Cf. de Saussure, Recueil des publications scientifiques 3 (Geneva, 1922) and Streitberg, IJ 2.205 (1914).

⁷ Möller's early laryngeal speculations, including his positing of /E/, occur in Englische Studien 3.150–1 with 151 fn. 1 (1880) and PBB 7.492 fn. 2 (1880). The scheme of Semito-Indo-European laryngeals here included appeared in his Semitisch und Indogermanisch: Erster Teil (Copenhagen, 1907). I have not seen this work, and derive my information about it from A. Cuny's review, REA 11.275–9 (1909).

⁸ Clearly /e/ is a phonetic guess if there was but one vowel, and Bopp's /a/ would serve as well. If there were two vowels, /e/ and /o/, we should probably describe them simply as front and back or rounded and unrounded, /e o/, then, would be simply shortcuts. Pedersen, Linguistic science in the nineteenth century 284–5 fn. 2 (Cambridge, 1931), would write ä and å.

⁹ I here merely parrot what Brugmann says, KVG §127. He points to Skt. dháyati ‘sucks’, a form which according to normal patterns of ablaut should show the reflex of , but which in fact seems to continue PIE /ai/. According to this reasoning , if it ever occurred, had merged with /ai/ already in PIE times so as to show the same reflex in Sanskrit as PIE /ai/. There is, therefore, no set of cognates which would allow us to reconstruct an for PIE. We should need a correspondence Gk. /ai/ = Skt. /ī/ .

¹⁰ Another possible phonemic solution would be to write the long vowels as geminates of the corresponding short vowels, i.e. /ee aa oo/. This would have the desired result of reducing the number of phonemes to six, but it would still leave quite isolated. Apart from this there is no objection to such an analysis on the phonemic level. It would cause certain difficulties in the morphology, however, as will become clear in the following discussions.

¹¹ I here assume that the symbols ē ā ō are taken by those who use them to be unitary phonemes and not /e a o/ plus a phoneme of length.

¹² Cowgill points out that this evidence does not really strengthen the analysis of [ē] as rather than , since these alternations might be of the type preḱ : prḱ instead of derḱ : drḱ. Better morphological evidence for the analysis given in the text is the parallel of the nasal presents, which force an analysis of , not because of , not .

¹³ This is overstated, for the zero grade of long diphthongs, phonemically , in some cases shows [ī]. The most efficient analysis of this seems to be . If this should prove to be in fact the best analysis, then we do have evidence for treating all examples of [ī] as . On this matter cf. W. P. Lehmann, Proto-Indo-European phonology 32 (Austin 1952), and W. Schulze, KZ 27.427 ff. (1885) = Kl. Schr. 49 ff. The opposite analysis is also supported by the morphology, for the optative shows an alternation [iē] ~ [ī] which would seem best to be analyzed as .

¹⁴ In a fuller treatment this solution could not be allowed to pass without more discussion; but since the question is not central to my argument, I have left it as is. Other explanations of Greek /ia/ = Skt. /ī/ however, are common. Brosman, Lg. 33.9–10 (1957), feels that the Greek development is phonologically conditioned in accordance with the Sievers-Edgerton law of the allophones of semivowels. J. Kuryłowicz, Apophonie en indoeuropéen 195 (Wroclaw, 1956), and W. Cowgill, Evidence for laryngeals 103 (Austin, 1960), feel that the Greek reflex is of analogical origin.

¹⁵ /a/ is still necessary, for we have not yet discussed the correspondence Gk. /a/ = Lat. /a/ = Skt. /a/.

¹⁶ As I have stated the allophones of /H/ there is no problem, for [h] is only one of the allophones and is another. In most laryngealist writings, however, unless I misinterpret them, is said to develop from /H/. So stated, problems of the zero grade are difficult if not impossible to explain. Granted that the difficulty is phonetic only, still it has considerable weight. How can [h] develop into a vocalic element? Cf. C. Watkins, Lg. 34.389 (1958); and on the whole matter of laryngeals and the zero grade see N. Collinge, Arch. ling. 5.75–87 (1953).

¹⁷ The assumption that an /h/ is to be restored to PIE rests on the belief that it is more economical to add a new symbol to PIE than to account for every instance of Hittite /h/ more or less ad hoc. But I am not a Hittitologist, and do not know whether this is in fact the case. In any event postulating a PIE /h/ seems dictated by the correspondences, and I therefore have no qualms about doing so, at least provisionally.

¹⁸ Cowgill informs me that appa is in fact not likely to be cognate with PIE *apó, and refers me to Friedrich's Hethitisches Wörterbuch 25 (Heidelberg 1952). There Friedrich prefers the connection of appa with Greek ópi- in ópisthen to that with apó. I agree that Friedrich and Cowgill are likely to be right, but not all laryngealists will, and hence I direct the argument in this paragraph only against those who accept as valid the equatio^u appa = apó and operate with /H₄/.

¹⁹ De Saussure, Recueil 603; Lehmann, PIEP 80 f.; Kuryłowicz, Apophonie 375 ff; Hoenigswald, EFL 13 ff.

²⁰ Recueil 231 ff.

²¹ Revue de phonétique 2.101–32 (1912).

²² EFL 94–5.

²³ EFL 109.