Published online by Cambridge University Press: 22 April 2026
Hermann Hirt, Idg. Gr. 1.336–7 (1927), in a treatment of initial consonant variation, presented eight sets of examples as evidence for an alternation gw [sic] : w. In general the etymological relationship of so much as lies to the left of the colon (e.g. Gk. bía, Skt. jayati, jināti) or to the right of the colon (e.g. Gk.
s, Lat. vīs) is admitted by the standard etymological dictionaries. Relationship between forms lying on the left and on the right is naturally not admitted, since Hirt did not develop his material to a point at which alternations of the kind in question could be used in support of new etymologies. We have in fact no right to take such equations for granted, though I hope in this article to show some reasons for regarding it as credible that certain forms with reflexes of /gw/ and of /w/ had an identical origin. In connection with the Latin forms it is well to mention that v, being a reflex of both IE /gw/ and /w/, cannot be used as unambiguous evidence for either. In §§2–9 I take up Hirt's equations in order, with brief discussion.
1 In §§2–9 I quote Hirt verbatim, except that I use Roman type in the Greek forms and capitalize at the beginning of each paragraph. In §32 and §33, where the discussion is more condensed, I do not quote Hirt verbatim, but in some instances I have retained his German glosses.
2 The u might be regarded as a phonemicization of the lip-rounding in gw, occurring in certain environments which favor the appearance of a ‘non-phonemic syllabic (vocoidal) segment’; see Hoenigswald, Evidence for laryngeals 22 (Austin, Texas, 1960), whose remarks, though bearing on the traditional ‘shwa secundum’ without special reference to labiovelare, would nonetheless be applicable here. The example given by Pedersen and Lewis-Pedersen, however, is not decisive; the u in OIr. guth is better taken as the zero grade of the diphthong reflected in Skt. gavate.
3 Since a very large proportion of the instances of alleged pure velars occur before a and r, those scholars are probably correct who believe that they became distinguished from the ‘palatals’ by phonemic split in Indo-Iranian, Armenian, and Balto-Slavic. See Hirt 1.232–6; Lehmann, PIEP 100–2.
4 A. G. Preobrazhensky, Etymological dictionary of the Russian language under vitb (New York, 1951).
5 M. Vasmer, Russisches etymologisches Wörterbuch (Heidelberg, 1953–8).
6 Brugmann, Gdr. 2 1.303–4, 306, 591, 620.
7 Pedersen, Vergl. Gr. d. kelt. Spr. 1.59; Lewis and Pedersen, Concise comparative Celtic grammar 11. Cf. also Pedersen 1.96, 108, Pedersen-Lewis 28, 34 for the appearance in certain British forms of gw- for g- < IE gh as well as gwh, resulting in partial merger with gw- < w-.
8 Cf. Hirt, Urgerm. 1.113–4; Prokosch 92–3; Lehmann, PIEP 36–46.
9 Cf. Tagliavini, Le origini delle lingue neolatine 2 199 (Bologna, 1952).
10 Mélanges Pedersen = Acta Jutlandica 9.55 (1937).
11 Even /b/, despite its evident rarity, must probably not be discarded from the PIE phonemic system entirely.
11a Frisk s.v. apolaúō rejects the connection of these two Sanskrit forms.
12 In Lehmann's system, PIEP 99, 111–4, the long vowels /e· o· a·/ in the final stage of PIE represent a merger of the lengthened-grade vowels and the vowels which became long upon the loss of following laryngeals before obstruents. Vowel plus laryngeal in word-final position may also be assumed to have developed into /e· o· a·/ by this stage, /i· u·/ were reflexes of the vocalic allophones of /y w/ when a preceding or following laryngeal disappeared (114).
13 Their allophones in postconsonantal position would be [i u] etc., in postvocalic position [y w] etc., forming, in traditional terms, the second elements of diphthongs, including the unstable long diphthongs.
14 A particularly frequent final; cf. Edgerton, Lg. 34.446 (1958).
15 Mélanges Pedersen 53–4.
16 Such juncture might possibly include juncture of the members of compounds, juncture before finite verbal forms in enclisis, and abnormally rapid utterance.
17 Cf. H. M. Hoenigswald, Language change and linguistic reconstruction 77 with figure 50 (Chicago, 1960); W. P. Lehmann, Historical linguistics 157–8 (New York, 1962).
18 His own symbols are gw, kw, ghw.
19 So, for example, in Buck, Comp. gr. of Gk. and Lat. 130; WH1; WH2; Boisacq. Comparison of vapor with Gk. kapnós and of ubi, u
, unde with qu-forms of the interrogative-relative pronoun was sometimes adduced as evidence for Lat. v- < kw or kw, but ubi, ut, unde are best explained by reanalysis of alicubi, sicut, etc. (so, for example, Kent, Sounds of Latin 3 122), while we should do well to discard the connection of vapor with kapnós. Hirt's Russ. kop seems to be without any attestation and may be merely an error for kopot' 'soot, lampblack', which Vasmer regards as related to Gk. kapnós, Lat. vapor.
20 WH3 also observe that the k in Gk. koîtai is not good evidence of a labiovelar.
21 Paleontologia linguistica = Annali della Facoltà di Lettere e Filosofia della R. Università di Cagliari 9:1.26 ff., note 20, as cited in Athenaeum 18.3 (1940).
22 So, for example, Falk and Torp, Norwegisch-dänisches etymologisches Wörterbuch (Heidelberg, 1911) under varm; perhaps also EM4 by implication, since they do not list the Germanic forms with Lat. formus. Vasmer admits connection of Ger. warm and Russ. varitb as a possibility.
23 His notation for our kw-. It should be observed that he treats k
- and
,- separately, but includes
along with
and
along with
while he has no instances of ![]()
24 According to Sturtevant, Lg. 6.224 (1930), Indo-Hittite had clusters kw etc. in contrast to Indo-European, which had labiovelar phonemes; Cf. Hirt, Idg. Gr. 1.232. I consider this view of the origin of the labiovelare very unconvincing, because other sequences of stop plus w did not produce single phonemes. It would be especially difficult, on this theory, to account for the difference in the history of the labiovelare (> Skt. k, c, Slav, k, č, Gk. p, t, ku, etc.) and that of kw, gw, ghw (Brugmann's
, etc.), which yield Skt. śv, Slav. sv} Gk. pp, ku, etc.