The University of Texas at Austin; College of Liberal Arts
Hans C. Boas, Director :: PCL 5.556, 1 University Station S5490 :: Austin, TX 78712 :: 512-471-4566
LRC Links: Home | About | Books Online | EIEOL | IE Doc. Center | IE Lexicon | IE Maps | IE Texts | Pub. Indices | SiteMap

Proto-Indo-European Phonology

Winfred P. Lehmann

5. PGmc. /g/ and /k(k)/ corresponding to PIE /w/

5.1. Evidence for the development of PGmc. /g/ and /k(k)/

A number of Gmc. words in which /g/ or /k(k)/ in some forms corresponds to /w/ in others have been much less frequently discussed than those in which /w/ and /y/ were lengthened. In 1888 Bugge had already given a full list of them and a phonetic description of the patterns in which the correspondence is found.1 Handbooks have subsequently repeated his data, adding or deleting occasional examples.2 The description given here differs somewhat from Bugge's, but his material will be discussed also.

Pre-Gmc. /w/ corresponds to PGmc. /g/ after /u/, and possibly, in two difficult words, after /i/: OIcel. bryggia ‘pier’, OHG jugund ‘youth’, OSwed. myggia ‘mosquito’, OE syʒel ‘sun’, OE suʒu ‘sow’, NSwed. dial. miggel ‘snow-ball’, OE niʒun ‘nine’.

Pre-Gmc. /w/ corresponds to PGmc. /k/ after /a/ and /ai/ in OE naca ‘ship’, OIcel. skeika ‘swerve’, OHG speihhila ‘spit’, OHG zeichur ‘brother-in-law’, and after /e/ in OIcel. leka ‘leak’, and to PGmc. /kk/ after /a/ and /i/ in OE cwic ‘alive’, OE hæccean ‘hack’, OE spic ‘fat’, and possibly OIcel. stakkr ‘pole’.

5.2. Proposed explanations

Although the data have long been accurately described, only two explanations have been suggested for the development of such /g/ and /k/. Bugge ascribed it to the position of the accent; according to him uw became PGmc. ug directly before an unaccented vowel; w became k after pre-tonic and ai or oi. But he was already dissatisfied with his formula; for some words, e.g. OE brūwe, OHG ingrūen retain w; and in others, e.g. OIcel. skugge, Goth. skuggwa, uw became uggw rather than ug; and Goth. áiws, hláiw, etc. retain w. As a result Bugge's suggestion gained little acceptance.

In 1946 Austin connected the development with the lengthening of w, explaining both with the help of the laryngeal theory. I have listed reasons why Austin's explanation seems untenable. (See 4.3.) But Austin's attempt at explanation with the laryngeal theory seems quite plausible; for some of the instances of w > g, k are found in words for which other IE dialects give clear evidence of laryngeals; such are OIcel. bryggia, Gaulish briva < /bhreXw-/, OE syʒel, Gk. ᾱ̓[Ϝ]έλιος < /saXw-/, OE naca, Gk. ναυ̑ς < /naXw-/. Consequently I shall reexamine the Gmc. words and their cognates, analyzing them on the basis of the laryngeal theory, in an attempt to provide an explanation.

5.3. The words in which /w/ became /g/

5.3a. OIcel. bryggia ‘pier’, OE brycʒ, OS bruggia, and OHG brukka, MHG brücke ‘bridge’ contrast with OIcel. brú ‘bridge’ in having g for pre-Gmc. /w/.3 From Gaulish briva ‘bridge’ we may reconstruct the PIE full grade form /bhreXwa-/. WP separate from this group the IE words for ‘brow’, Skt. bhrū́, Gk. ὀφρυ̑ς, OE brū, brǣw, OHG brāwa, etc. Such a separation is hard to justify, for the two groups are similar both in form and meaning. OE brǣw is a full grade form of PIE /bhreXw-/, Skt. bhrū́ a zero grade form. And the meanings of ‘brow, beak, bridge’ are found in OCS brъvъno and Gk. ὀφρυ̑ς. But with or without the words for ‘brow’ we have forms illustrating the various ablaut grades of PIE /bhreXw-/.

Both OIcel. bryggia and OE brycʒ may be derived from PIE /bhreXw-/. The reduced vowel /e/ regularly became Gmc. u, as in the second syllable of OHG zweinzug.4 I assume that these forms on the one hand and brú on the other developed from an alternation of the root before consonant or vowel of the ending, as that in Goth. mawi nom. sg., máujos gen. sg. The reflex of /bhreXw-/ became brug- in a position in which /w/ remained consonantal, but brū- when /w/ was vocalic. The following formulas illustrate the developments:

[bhreXw-] > [breg-] > [brug-]
[bhreXw̥-] > [brXw̥-] > [brū-].

5.3b. OE iuʒuð, OS juguð, and OHG jugund ‘youth’ have -g- for pre-Gmc. w; Goth. junda ‘youth’ shows the expected development. Extra-Gmc. cognates are: Skt. yúvan-, nom. sg. yúvā, gen. sg. yūnas, and Lat. iuvenis ‘youth’. WP assume an IE root yew-; but this does not account for the forms with ū. Because of these forms we must posit a root in ‘long diphthong.’ Such a root was suggested by Hoffmann, who connected Av. avi-yāo ‘grown’, Gk. αἰζηός ‘vigorous’ and assumed a root yēu- (/yeXw-/).5 I assume that the WGmc. forms developed from [yeXwn̥þi-] with secondary accent on [] as may be recognized from the reflex of -t-; the Goth. forms developed from [yeuntā́].

5.3c. Some of the Gmc. words for ‘mosquito, midge’, OSwed. mygg, OE muʒʒia, OS muggia, MHG mucke, have g < w, but OIcel. mýn does not show this development. The extra-Gmc. cognates, Gk. μυῖα, Arm. mun, gen. mnoy, like mýn, point to development from the zero grade; our only evidence for the full grade is a Swed. dialect form me-hank ‘mosquito’. From this I assume a root /meXw-/. From forms like the nom. /meXwī/ the g-forms were generalized, from forms like the gen. /meXwyōs/ the u-forms were generalized.

5.3d. The Gothic name for the rune s is sugil; OE syʒel ‘sun’ also shows the development of w to g. From the cognates, Goth. sauil, Skt. súvar, gen. sū́ras, Gk. ἠέλιος, Dor. Gk. ᾱ̓έλιος, we reconstruct PIE /saAwel-/. I assume that Goth. sugil developed from /seAwel-/, that Goth. sauil developed from the full grade.

5.3e. In no IE dialect do we find a full grade form of the word for ‘pig, sow’. Gk. ὑ̑ς, ὑός, Lat. sūs, suis, Toch.B. suwo, OIcel. sýr, sú, OE OHG all show zero grade. By comparison with Gk. ὀφρυ̑ς, OE brū, etc. we may assume a PIE root /seXw-/. I ascribe the variation between OE and suʒu to variation of w before following consonant or vowel, as in the words previously cited.6

5.3f. Bugge assumed development of w to g also in OIcel. þrúga ‘oppress’, OIcel duga, OHG tougan ‘avail’, OIcel. hrúga ‘heap’, and Goth. hugjan ‘believe’. Bugge's etymologies, however, are not without difficulty. For þrúga he cited no extra-Gmc. cognates. For OIcel. duga he assumed development from the root found in Skt. tavīti ‘have power’, with the somewhat dubious assumption that IE t became Gmc. d because the accent lay on the third syllable; consequently I assume that the etymologies cited by WP, Gk. τρῡ́ω, τρύχω and τυγχάνω are more plausible. OIcel. hrúga, like OIcel. hraukr, OE hrēac, Eng. rick seem to be developments of IE ker- ‘turn, bend’ with a suffix -ew- and determinatives. A number of etymologies have been suggested for hugjan, none of them convincing.

5.3g. The Swed. dialect word miggel ‘snow-ball’, which Noreen cites as an example of w > g is too doubtful to use as a basis for further theories. Development may be from the root of Skt. mīvāmi ‘grow fat’, as for OIcel. mývell ‘ball’; but, as Noreen suggests (Abriss 168), miggel may have developed from a form of this root with velar extension.

5.3h. Although g is well attested in OE niʒun, OS nigun ‘nine’, beside Goth. niun, OHG niun, we know too little about the PIE form to draw any conclusions about the WGmc. forms. From Skt. náva, Lat. novem, Hom. Gk. *ενϜα- in εἴνατος, PIE enewen has been reconstructed. But this fails to account for Gk. ἐννέ[Ϝ and OIcel. nío. Since the PIE form is unclear, we can draw no conclusions about the OE and OS forms.

I conclude that we can say with reasonable assurance that Xw > g only in OHG brukka, jugund, mucke, OE syʒel and suʒu. The u before g in all of these forms developed from the reduced vowel /e/.

5.4. The words in which /w/ became /k(k)/

5.4a. OIcel. nǫkkue, OE naca, OS naco, and OHG nacho ‘boat’ are cognate with Skt. naú, acc. nā́vam, Lat. nāvis, Gk. ναυ̑ς ‘ship’, νήιος ‘belonging to a ship’, OIcel. naust ‘boat-house’, nór ‘boat’, nóa-tún ‘ship-town’, and OHG ver-nawun ‘boats that carry wood’.7 From the Skt., Gk., and Lat. forms I posit PIE /naAw-/. In most IE dialects such combinations lost the laryngeal, e.g. Gk. βου̑ς PIE /gʷeXws/ ‘cow’, or the resonant, Dor. Gk. βω̑ς; but in Skt. we commonly find forms such as naú without any such loss. (Gdr.I.203 fn.) I assume that OIcel. naust maintained the form with au; OIcel. nóa-tún the form with ā; and that the Gmc. forms with k, such as naco, maintained the uncontracted root /neAw-ō/ > /naAw-ō/ > /naco/.

5.4b. OIcel. spékaldra, OHG speichaltra, Goth. spaiskuldra, OHG speihhila ‘spit’ are from a PIE root with many developments.8 WP posit a root sp(h)i ̯ēu-, spi ̯ū, spīu̯; in terms of Benveniste's theory of the IE root one would posit a root /(s)pey-/, with form I /spey-X-/, form II /spy-eX-/, and with a determinative, /spy-eX-w-/. Most of the IE forms developed from the last form; the Lith. verb spiáuti from the full grade; the Skt. ṣṭhī́vati and the Goth. speiwan from the zero grade /spyXw-/; the Gk. πτῡ́ω and the Lat. spuere from the zero grade /sp(y)-Xw-/. I derive the OIcel. and OHG forms given above from the o-grade of form I, to which the determinative w has spread; /spoyXw-/ became /speik-/. The Goth. noun parallels the OHG and OIcel. nouns except for the s after the root vowel; the only acceptable solution proposed for s is that it is a result of an error in writing.

5.4c. OIcel. kuikr, OE cwic, cwicu, OHG quec, queh ‘alive’, and OIcel. kweikja ‘make alive’ are cognate with Skt. jīvá, Lat. vīvus, OCS žīvъ ‘alive’, Gk. βίος ‘life’, ζώω ‘live’, fut. βείομαι. It is difficult to reconstruct a PIE root which will account for the diversity of development; WP posit guei ̯(ō)- and guei ̯ōu-. Because of the variety of development I assume a root similar in shape to that of /spey-/: /gʷey-/, form I /gʷey-X-/, form II /gʷy-eX-/ and with determinative /gʷyeX-w-/. I assume that the Gmc. adjectives developed from the same IE form as Skt. jīvá, that is, /gʷyXw-os/. In Goth. qius ‘alive’ the laryngeal was lost; in NWGmc. the combination /-Xw-/ became /-k-/.

5.4d. OE hæccean, OFr. tohakia and NHG hacken ‘cut, hack’ seem without question to be related to OIcel. hǫggva, OE hēawan, etc.9 Cognates in other dialects are Lat. cūdō < *cāudō ‘cut’, Lith. káuju, káuti ‘strike’, OCS kovati ‘hammer’. Hirt (Abl.102) assumed an IE root qou̯ā-, Brugmann (IF 6.99) qāu-. I posit a root /keX-w-/, from which some forms were made with X determinative. Such forms are Lith. káuti and the Gmc. words with lengthened w. OE hæccean, on the other hand, developed from the unenlarged root /keX-w-/ > /kaX-w-/ > /hak-/. The lengthened k before j is a regular WGmc. development.

5.4e. OE tācor and OHG zeichur ‘brother-in-law’ are clearly related to Skt. devŕ̥, Gk. δάηρ, Lat. lēvir, OCS děverъ, and Lith. dieverìs ‘brother-in-law’, all of which preserve evidence for -w-. WP posit IE daiwer, but with it we cannot account for some of the attested forms. Lat. lēvir, because of the vocalism, must be explained as a borrowing from an Italic dialect, the ā in the Gk. cognate as an Attic development. (Gdr.I.182.) It is much more plausible to assume PIE /deXywer/. Skt. devŕ̥ would derive from this, with -e- from /eXy/ as in dhenú ‘cow’. In Lat. forms the /y/ would have been lost, as in rēs < *rēis. The Baltic and Slavic developments are as expected.

Even from /deXywer/ the Gmc. forms are difficult to explain; in forms examined above we have found that Gmc. k developed from contiguous X and w. I assume that the Gmc. forms, like many of the r and n stems, show analogical leveling. The Hittite cognate of water is nom. sg. wa-a-tar, gen. ú-e-te-na-aš, and presumably represents the original r:n variation. In Goth. watō, watins the n was generalized; in OE wæter, OHG wazzer, the r; in all Gmc. dialects the a vocalism was generalized. For PIE /deXywer/ I assume a variation: I. -eXyw- > -aiw-; II. -eXyw- > -aXw- > -ak-.10 In Gmc. the vocalism of I was generalized, but -k-spread throughout the paradigm.

5.4f. OIcel. spic, OE spic, OHG speck, Skt. pī́van, pī́varī, Gk. πί̄ων ‘fat’ are cognate with Skt. sphāyate ‘grows fat’, sphīta ‘successful’; from these I assume a base /speXy-/. We cannot determine whether the Gmc. root vowel developed from PIE e or i. For the k, however, a development similar to that suggested for the k in tācor is probable.

5.4g. OIcel. skeika ‘swerve’ is a denominative weak verb derived from the root found in Gk. σκαι[Ϝ]ός, Lat. scaevus ‘left’. If we relate MHG schiec, schief ‘crooked’ we must assume PIE /skeXyw-/; (Gdr.I.207) the Gk. and Lat. adjectives developed from the unaccented root /skeXywos/. The nominal form from which skeika was made is nowhere attested in Gmc., but a development in it of /Xw/ to k similar to that in tācor may be assumed.

5.4h. Several other words have been cited to illustrate the development of Gmc. k from PIE w. Austin suggested that OIcel. leka ‘leak, trickle’, OE leccan, Eng. leach ‘wet’, Eng. latch ‘moisten’ were related to Lat. lāvit ‘washed’, Hitt. la-aḫ-ḫu-un ‘pour’. From the Hitt. form we may assume a PIE base /leXw-/.

5.4i. Austin also derives OIcel. stakkr ‘pole’, OE staca, Eng. stack, stake from PIE steA- ‘stand’ with w-determinative. The etymology is plausible, but uncertain, for steA- is enlarged by many other determinatives. The Gmc. words can be derived from such extended forms, as they have been in the past.

5.4j. Austin also suggested that OIcel. maka, OE macian ‘make’, OHG maccian ‘join, fit together’, OE ʒemæcca ‘mate’, Eng. match were cognate with Skt. mā́ti ‘measure’, and Hitt. me-ḫur ‘time’. He made no mention of the standard comparison with Gk. μαγη̑ναι ‘knead, paint’, OCS mazati ‘smear’; (WP 2.226-7) the Gmc. verbs are assumed to have been generalized in meaning from a term used in building to other types of construction. Since Austin finds difficulty in ascribing the phonological development in Gmc. to the laryngeal that must be assumed from me-ḫur, I conclude that his etymology is less plausible than that usually given.

5.4k. Some linguists have derived OSwed. knoka, OE cnucel, MHG knoche ‘bone’ from Gmc. knōw-, but it is more probably connected with the root found in Lith. gniáužiu ‘clench the hand’.

5.4l. Development of w to g and k(k) has been suggested for several other difficult words; among these is Germ. spucken ‘spit’, of which we have no record before the year 1716. I assume that such etymologies are no longer held, and that they do not need refutation.

Of the words examined I assume that OE naca, OHG speichaltra, quec, OE hæccean, tācor, spic, OIcel. skeika, leka, and possibly stakkr have k, kk from PIE Xw. Before k, kk are found the Gmc. accented vowels a, ai, i, e.

In all of these forms I assume that the original PGmc. development of -Xw- was -kk-. The -kk- was preserved only in forms with preceding short vowel, quec, hacken, and speck; after a long syllable kk became k in accordance with a regular PGmc. development. (See CGG 70-1.) The shortening of kk in OE naca and other WGmc. forms was apparently a later development, cf. OIcel. nǫkkue.

5.5. The conditions under which these developments took place

The development to g in some words, to k(k) in others I tentatively ascribe to a difference of laryngeals. For some IE etyma the difference is obvious; OE naca and cognates point to an a-colored laryngeal, OIcel. bryggia to an e-colored laryngeal. There is no further evidence to identify the laryngeals, for we do not consistently find cognates in Hittite; moreover, full-grade forms of some of the words involved are not attested. Of the five words with w > g, bryggia, iuʒuð, and muʒʒia point to an e-colored laryngeal, syʒel to an a-colored laryngeal, and the laryngeal of suʒu cannot be identified. Of the nine words with w > k, naca, hæccean, leka, and staca point to an a-colored laryngeal, speichaltra and spic to an e-colored laryngeal, and the laryngeals of cwic, tācor, and skeika cannot be identified. Until we have further evidence from other IE dialects to distinguish the different laryngeals, explanation of development to g in some words, to k(k) in others, because of a difference in laryngeals can only be tentatively proposed.

The material presented in this chapter is further evidence that PIE laryngeals did not contract with vowels when in the neighborhood of /w/ and /y/. The assumption found in various handbooks that PIE /eXw/ became PIE long vowel plus w was based on the development of PIE /eXt/ to long vowel plus obstruent, and does not hold. We have already noted that laryngeals were preserved when following PIE [w] and [y], and when standing before [y], and that a lengthened resonant resulted in Gmc. Examination of the forms cited in this chapter shows that laryngeals were also preserved when standing before [w], and that in this position /g/ or /k(k)/ resulted.


1 S. Bugge, Zur Altgermanischen Sprachgeschichte, Germanisch UG aus UW, PBB 13.504-15 (1888). Bugge lists cognates of the various words; the cognates listed below are taken from WP and other etymological dictionaries.

2 W. Wilmanns, Deutsche Grammatik3 I.155 (Strassburg, 1911); see the references there.

3 See Wilmanns 188ff. for the lengthened g.

4 E. H. Sturtevant, The Indo-European Reduced Vowel of the e-Series, Language 19.293-312 (1943), discusses the development of /e/ in the various IE dialects. The usual Gmc. development of /e/ is u, except when between obstruents; furthermore some Gmc. forms have i and a for /e/.

5 BB 15.62; see also Wackernagel, Aind.Gram. I.91.

6 Bugge also cites Norw. Swed. sugga ‘sow’ which he assumes to have -gg- from lengthening of w; I assume that the lengthened g here is a modern development, rather than a PGmc. one, for we have no other evidence for lengthening of w in this word.—Many attempts have been made to explain the long vowel of the PIE etymon; some Indo-Europeanists attempt to account for it in other ways from the one accepted here. Specht, KZ 59.281, assumes that a short u was lengthened; Kretschmer, Glotta 13.132, suggested origin in an interjection, supporting his view by pointing out the anomalous initial s in Gk. and saying that it might have been preserved by analogy with the interjection that was used to call pigs.

7 E. G. Graff, Althochdeutscher Sprachschatz II.1014 and 1109 (Berlin, 1834-42).

8 See WP 2.683 and Feist, Goth.Ety.Wb. under spaiskuldra for the wide variety of developments from this root; Benveniste's theory of the IE root is presented more fully above, 2.3f.

9 WP 1.330 derive hæccean from IE keg-, apparently because of the difficulty of w > k.

10 Our evidence is too small to permit accurate definition of the conditions under which resonants following a vowel and laryngeal (that is, final elements of long diphthongs) are lost. Brugmann, Gdr. I.203fn., suggests that they were lost when they stood last in the syllable.—For examples of such leveled forms in Gmc. see Streitberg, UG 291-2.