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

4. Lengthened /w/ and /y/ in the Gmc. Dialects

4.1. Evidence for lengthening in the various Gmc. dialects

The treatment of resonants is very complex in the Gmc. dialects. Of the six PIE resonants, /w/ and /y/ show the greatest variation in development and they too have been the subject of greatest discussion. Words with parallel developments in other IE dialects have Gmc. cognates which contrast markedly with each other. Intervocalic /w/ was lost in Gk. ὄις < *ὄϜις ‘sheep’ and κοέω < *κοϜέω ‘perceive’. Cognates of ὄις are OIcel. ǽr < *awiz and OHG ou ‘sheep’; cognates of κοέω, are OE scēawian, OS skauwon, and OHG scauwōn ‘gaze’. A similar contrast in the development of /y/ is found in Goth. fijaþwa ‘enmity’, Skt. piyāru ‘censuring’ and OIcel. Frigg, the name of Odin's wife, Skt. priyá ‘dear’. In one word the Gmc. dialects seem to preserve the single resonant attested in the other dialects, in the other they show lengthened resonants.

The developments in Gmc. of PIE /w/ and /y/ have long been accurately described.1 After a short vowel Gmc. /w/ and /j/ in some words were lengthened before vowels. The WGmc. dialects preserved the lengthened resonants. In Gothic and NGmc. they developed into clusters written -ggw- and -ddj/ggj-. These clusters are generally interpreted as velar stop plus [w] and palatal stop plus [j]. In Gothic and NGmc. the velar stop fell together with the g-phoneme, and was written -gg(w)-, as did the palatal stop in NGmc., written -gg(j)-; in Gothic the palatal stop fell together with the d-phoneme, and was written -dd(j)-. Most linguists follow Braune in assuming that the cluster of sounds which developed from Gmc. -jj- was originally pronounced the same way in Gothic and NGmc., but that it became fronted in Gothic and retained palatal articulation in NGmc.2

There are accordingly three stages in the development, and these may be represented by the following formulae: 1. -w- plus ?; -y- plus ?; 2. -ww-, -jj-; 3. [gw], [d/gj]. Stage three is attested in Gothic and NGmc., stage two in the WGmc. dialects, and stage one is nowhere attested but presumably was the PIE pattern.

While the ENGmc. developments remained clearly marked, the distinction between -w-, -j- and -ww-, -jj- is less obvious in the WGmc. dialects.3 Kögel made a detailed study of WGmc. documents and found the following orthographical evidence for a distinction.4

  PGmc.   OE   OS   OHG
      Medial Final   Medial Final   Medial Final
1 aww   ēaw ēaw   auw, auuu, auu au   auw, ouw au, ou
  aw   eaw ao, ēa   aw, auu ao, ō, ā   aw, ouu ao, ō
2 eww   ēow ēow   ēuw eu, euu   euw, iuw iu
  ew   ? eo   ? ?   ? io
3 uww         (evidence uncertain)      
  uw         (evidence uncertain)      
4 ajj   āj     aij ai   aij ai
  aj         ej ē   (ē) (ē)
5 ijj   ī     ī     ī  
  ij   eo              
6 ujj         (no evidence)      

4.2. Proposed explanations

Although the data in the Gmc. dialects were clear, no explanation for the development was attempted until Kluge in 1879 related lengthening to the IE accent.5 He suggested that PGmc. intervocalic w and j after short vowel were lengthened when the IE accent preceded. Thus j was not lengthened in Goth. saian ‘sow’ (Kluge assumed this to be sāian) because of the preceding long vowel, nor in Goth. niujis ‘new’ because of the preceding diphthong; and because of the following accent in IE *gʷiu̯ós, the w was not lengthened in Goth. qius ‘alive’. But -j- and -w- were lengthened in Goth. daddjan ‘suckle’ and triggws ‘faithful’ which Kluge reconstructed with accent on the first syllable. Schmidt in a review6 disposed of Kluge's explanation by analyzing differently the words which Kluge cited in favor of it, and by citing objections to it; of Kluge's four examples in favor of his theory only Goth. iddja remained, and exceptions like Goth. *þreis ‘three’, acc. þrins, Skt. tráyas were unaccounted for.

Schmidt also pointed out that the cognates of OIcel. egg, Crim. Goth. ada ‘egg’ have an original long vowel: Gk. ᾠόν, Polab. jojǘ, Serb. jáje from PSlav. /jajé/. Accordingly -j- was lengthened after long vowels too; later linguists have not taken account of Schmidt's remark concerning ᾠόν.

Bechtel then suggested that -w- and -j- were lengthened when the IE accent followed immediately;7 he cited more examples in favor of his theory than Kluge had. But like all linguists who relate the Gmc. developments to the IE accent, he cannot account for lengthening in strong verbs otherwise than by assuming regularization by analogy. Paul8 had earlier cited the OIcel. verbs

hǫggua   hió   *hióm   *háinn
búa   bió   bioggum   búinn
to illustrate the difficulty of explanation by analogy. Nonetheless Bechtel assumed an original Goth. conjugation
*bliwan   *blau   bluggwum   bluggwans
with bliggwan blaggw as later developments. Such regularization might be admissible in Gothic, but hardly in the Gmc. dialects which maintained grammatical change. And even if we admit it in the verb system, regularization is much more open to question for nouns and adjectives. Such a general objection is much more weighty in rejecting Bechtel's explanation than are exceptions like Goth. ajukduþs ‘eternity’.

After reviewing the attempted explanations, Streitberg rejected any possibility of relation between lengthening and the IE accent. (Germanisch 323-6.) In his opinion only the Gmc. accent can be related to lengthening. But if we accept Streitberg's statement and assume a causal connection between Gmc. accent and lengthening, we cannot account for many words with single resonant after the first syllable, such as Goth. awi-liuþ ‘thanks’, nor can we explain why w and j were lengthened in some words but not in others with the same accentuation, such as triggws and ǽr.

One inescapable conclusion seems clear from the many discussions of lengthening: it has no relation whatsoever to either the IE or the Gmc. accent.

Two further explanations have been offered that do not relate lengthening to accent. Brugmann suggested that lengthening is found in words which in PIE already had a combination of sounds that developed to -ww- and -jj- in Gmc.9 He derived: Goth. -waddjus, OIcel. veggr ‘wall’ from *u̯oi-i ̯u- ‘woven material’; OIcel. tyggua ‘chew’ from *kiu̯u̯ō < *kiuu̯ō; Goth. triggwaba, OIcel. tryggr, OHG gi-triuwi ‘faithful’ from Gmc. *treuʒuo, IE *drēu-quó-. It is difficult to cite cognates justifying these reconstructions; Brugmann's theory was accordingly not widely accepted. Osthoff's explanation10 that the -w- and -j- were lengthened in forte forms and unchanged in piano forms has likewise met with little acceptance. After reviewing all of the theories Streitberg finds none of them satisfactory. (Germanisch 325.)

4.3. Proposed explanations by means of the laryngeal theory

In 1941 Smith attempted a solution by means of the laryngeal theory.11 Smith ascribes to PGmc. a series of clusters which were made up of aspiration and (long) voiceless resonants; the aspiration was a reflex of IH laryngeals. These clusters, after the operation of Verner's law, developed to PGmc. [γw] and [γj], which in turn developed to the combinations found in the Gmc. dialects.

Various objections may be offered to Smith's theory. 1. Smith cites very few forms other than those in Gmc. dialects on the basis of which we can assume IH laryngeals where his formulae require them. 2. If as Smith says the clusters [-γw] and [γj] were already found in PGmc. one must explain how [γ] was lost before [w] and [j] in WGmc. with lengthening of w and j. This Smith fails to do. 3. Assumption of a relation between lengthening and IE accent brings up the difficulties cited by earlier linguists.

Sturtevant adopted Smith's theory but revised some details. (IHL 80-3.) He did not assume that -hw- or -hy- followed any short syllabic in IH or IE, nor did he restrict the development to positions immediately before the accent. Even the revision is open to various objections. 1. The phonetic development from reflex of laryngeal plus w and y to the sounds attested in the various Gmc. dialects is unclear. 2. As Sturtevant indicates, there are contradictions between Gmc. developments and the laryngeal theory; if the Gmc. short vowel before hw and hy came from IE ə, IH X was involved in two processes: IH ıXy became IE ə plus hy. Consequently we have to posit two laryngeals, or a lengthened laryngeal, in IH. 3. Gmc. forms from IE words with vowel other than ə before hy, hw are left without explanation.

Sturtevant gives reasons for connecting the Gmc. lengthening with IH laryngeals. He finds that 1. ‘in a number of cases the short vowel before the Verschärfung alternates with an IE long vowel’ and that 2. ‘on the other hand, a PGmc. short vowel of an initial syllable before short j or w does not alternate with an IE long vowel.’ (IHL 81-2.)

Eight examples are cited in IHL in support of the theory. In two of these, Goth. twaddjē, OIcel. tueggia ‘of two’ and OIcel. beggia ‘of both’, lengthening is found only in the genitive. If these words contained -hj-, one must explain why lengthening was restricted to the genitive; no such explanation has been offered. For another example, Goth. glaggwaba, there is no evidence for IH laryngeal, for glaggwaba is of doubtful etymology. At best then five forms are left in support of the theory.

Moreover, some Gmc. words with single w have cognates with long vowels, e.g. Goth. qius ‘alive’, Skt. jī́vati, Lat. vīvō, and OCS živati ‘live’.

Explanation by means of the laryngeal theory became even more difficult when Austin related another Gmc. development to the lengthening of w and j.12 A number of Gmc. words have g or k where cognates in other dialects have w. Austin suggested that those words with k < w developed from IH -hw- when the IE accent preceded; he set up the following patterns:

IE -hy-́ > Gmc. -ddj/ggj-
IE -hw-́ > Gmc. -ggw-
IE -′hy- > Gmc. -k-
IE -′hw- > Gmc. -k-.

Although the development of w to g or k had been investigated, this is the first time, to my knowledge, that it was related to lengthening of w and j. Austin supports this relation by citing a root in which both developments are found: OIcel. hǫggua ‘hew’ and OE hæccean ‘hack’.

Austin's corollary to Smith's formulation makes explanation by means of the laryngeal theory seem very precarious; for the combined formulations are open to more objections than were Smith's:

  1. We should expect in Goth. bliggwan grammatical change, giving us the forms: *blik *blak bluggwum bluggwans, as well as in the other strong verbs with lengthening. Assumption of regularization by analogy might be acceptable for Gothic, but open to serious objections for Old Saxon and other dialects which maintained grammatical change.
  2. There are no words showing a development -́hy- > Gmc. k. The formula is thus difficult to justify.
  3. Examination of the evidence is incomplete. None of the words in which w > g and only some of those in which w > k are cited. Moreover, some of the words cited for w > k have kk. No explanation is offered for the difference.
  4. Assumption of IE -hy′- > Gmc. -ddj/ggj- and -hw′- > Gmc. -ggw- is most improbable in view of the WGmc. forms.

I conclude that the solutions proposed for lengthening and development of w > k by adducing the laryngeal theory are as unsatisfactory as those proposed by relating the accent. On the other hand, Sturtevant, and to a lesser degree Austin, have cited evidence for laryngeals in some cognates of the Gmc. words under discussion. I therefore shall examine as complete a list as possible of the words in which IE w and y were lengthened, and their cognates, in an attempt to discover possible reflexes of laryngeals. Later I shall examine words in which the four other PIE resonants occur in the neighborhood of laryngeals, and try to establish the development of these in Gmc.

The discussion will be divided into four parts:

  1. lengthening of /w/ and /y/ in Gmc.;
  2. g and k(k) from w in Gmc.; (Chapter 5)
  3. development of r l m n and neighboring laryngeals in Gmc.; (Chapter 6)
  4. the OHG r-preterites. (Chapter 7)

4.4. The words in which lengthening is found

Complete lists of words in which w and y have been lengthened have been given by Kögel, Noreen, and Trautmann.13 Other linguists have discussed the cognates outside Gmc.; their work may be located most conveniently in WP. In the following list therefore are cited only forms to illustrate the distribution of the Gmc. words and only cognates pertinent to the discussion.

The list is divided into five parts depending on the PGmc. sequence: 1. -eww-; 2. -aww-; 3. -uww-; 4. -ayy-; 5. -iyy-.

4.41. -eww-.

4.41a. The principal parts of búa ‘dwell’ in OIcel. are Inf. búa pret. sg. bió, biugga, biogga pret. pl. bioggom, biuggum pret. ptc. búenn. OE būan, OS būan, and OHG būan are regular weak verbs with only two unusual forms, OHG 3d pl. pret. ind. biruun and 2d sg. pret. subj. biruwis. (See 7.5)

OIcel. byggja ‘inhabit’ and OE bēon ‘be’ are related verbs, both showing the regular development of PGmc. -eww-.

Prokosch has suggested that verbs of the seventh class are made from ‘heavy bases’ with reduced grade in the present and e-grade in the preterite. (CGG 178.) The agreement in development between OIcel. byggja and the preterite of búa supports his suggestion. The principal parts of búa derive from PIE /bhewX-/ /bhewX-/ /bhewX-/ /bhewX-/.

Analysis as a laryngeal base is borne out by the Skt. forms such as the infinitive bhávitum ‘be’ and the pret. ptc. bhūtá. See 3.6Ab and 3.6Bc.

Another wide-spread development from /bhewX-/ is OIcel. bygg, d.s. byggue, OE bēow ‘barley’, OS beuuod ‘harvest’, and OE bēor, OHG bior ‘beer’. Since these are PIE s-stems which have normal grade in the root, we have further evidence of development of -eww-.

From these Gmc. forms we can derive evidence for the development in Gmc. of PIE -ewX-. IHL, 70, suggests h or x from forms like the Lat. imperf. tense sign and Osc. fufans ‘they were’; but Meillet, Introduction 209, adducing Gk. φυη̑ναι, considers both -ā- (eA) and -ē- (eʔ) suffixes. If so, we cannot determine the laryngeal affecting the Gmc. forms.

There is such a variety of forms in Gmc. that the development cannot be explained as originating from one form and spreading by analogy.

4.41b. OE brēowan, OS breuwan, and MHG briuwen, and the OIcel. pret. ptc. brugginn, compare OS gibreuuan ‘brewed’ contain evidence for the development of -eww- and -uww-.

Lat. dēfrūtum ‘cider’ may have long or short ū; assumption of long ū seems supported by Thracian βρυ̑τος ‘type of beer’. If so, we may assume zero grade of a laryngeal base; such an assumption would be supported by the intonation of Lith. briáutis ‘push oneself ahead forcefully’.

For brēowan and cognates we have uncertain evidence of PIE -ewX- and -uwX-.

4.41c. OIcel. hryggua ‘make sorrowful’, OS hreuuan, pret. hrau, OHG riuuuon, pret. hrau, rou ‘rue’ and the OIcel. adj. hryggr ‘sorrowful’ show developments of PGmc. -eww- and -aww-.

The Gmc. words show verbal development of an IE base which is found in nominal formations with the meaning of ‘blood, bloody meat’, Skt. kravís, Gk. κρέ[Ϝ]ας ‘raw meat’. Because of Skt. -i- and Gk. -a- I assume PIE /krewX-/. Although the related adjectives OE hrēaw, OHG rouuēr, rō ‘raw’ seem to have single -w-, the verbal forms in Gmc. give us evidence for the development of -eww- and -aww-.

4.41d. Goth. triggws, OIcel. tryggr, OE trēow, OS treuua and OHG trēow ‘true’ are u-stem adjectives with -eww-.

While they are commonly related to IE derew(o)- ‘tree’, Gk. δόρυ ‘wood’, the closest cognate semantically is Lith. drútas ‘strong’. From this and OHG trūēn ‘trust’ I assume PIE /drewX-/; IHL, 82, suggests IH dréuh-.

The Gmc. adjectives give us evidence for the development of -eww-.

4.41e. Goth. bliggwan, blaggw, bluggwum, bluggwans ‘strike’ has always been difficult for linguists who explained lengthening by means of the accent, or considered it restricted to positions after certain vowels. Yet we find evidence for lengthening in other Gmc. dialects; OHG bliuwan, MHG blou, OS utbliuuid ‘struck’. Kögel cites NHG geblauen, which as well as the OHG spelling and Otfrid's use in meter, show that the -w- in the pret. pl. and pret. ptc. was long. It seems therefore most probable that -w- was lengthened in all forms. This assumption is supported by the lengthening in briuwan, hriuwan, kiuwan and related forms.

While bliuwan provides no evidence for origin of -ww-, I conclude from it that -w- was lengthened after any Gmc. short vowel.

4.41f. OIcel. tyggia, OE cēowan, and OHG kiuwan ‘chew’ show developments of -eww-.

Cognates in other dialects give us little evidence for determining the IE root; from the intonation of Lith. žiáunos ‘jawbone’ we may posit a final laryngeal.

4.41g. OHG spriu, pl. spriuuuer ‘chaff’ shows -eww-.

Cognates with a d-extension are Lith. spráudžiu, spráusti ‘force in’; the intonation points to laryngeal, but the root has so many extensions that we can only consider /(s)prewX-/ doubtful.

4.41h. OHG sou, gen., souwes, OE sēaw, Icel. söggr ‘something damp’ show development of -aww-.

Gk. ὕ-ει ‘it is raining’, Lat. sūgō, OIcel. súga ‘suck’ are zero-grade forms of a laryngeal base, but outside Gmc. we have no full-grade forms with evidence for a laryngeal.

4.42. -aww-.

4.42a. OIcel. dǫgg, dǫggvar, OE dēaw, OHG tou, touues ‘dew’ give evidence for -aww-.

Cognates in other dialects, Skt. dhavate ‘runs’, Gk. θέω ‘run’, θοός ‘rapid’ give us no evidence for assuming a laryngeal in the root.

4.42b. Goth. glaggwō, glaggwaba ‘careful’, OIcel. glǫggr, gløggr ‘strict, clear’, OE ʒlēaw, OS glauuuorro, OHG glauwēr ‘wise’ are related by WP to Irish gluair ‘clear’, but the etymology is uncertain.

4.42c. OIcel. hnǫggua, pret. ptc. hnuggen ‘push’ and OIcel. hnøggr, OE hnēaw, MHG nouwe ‘stingy, exact’ show developments from -aww-.

From the cognates Gk. κνό[Ϝ]ος ‘noise of wheel in axle’, κνῡ́ω ‘scratch’ and Lett. knūdu, knūstu ‘scratch’ we can derive only uncertain evidence for a laryngeal base.

4.42d. OIcel. hǫggua, hió (ONorw. hiogga), hioggom, hǫgg(u)enn like OE hēawan, OS hauwan and OHG houwan show lengthening in all forms; I assume an alternation -aww-, -eww- from -eXwX-, -eXwX-.

Assumption of a laryngeal is supported by the intonation of Lith. káuju, káuti ‘strike’.

hǫggua and related forms then give us evidence for the development of PGmc. -aww- and -eww-.

4.42e. OIcel. rǫgg, rǫggr ‘long coarse wool’, Swed. rugg ‘shaggy hair’, OE rēow ‘wild’ and OHG rouaz ‘crude’ are developments of -aww-.

From the intonation of Lith. ráuju, ráuti ‘tear out’, and the vocalism of OIcel. rýja, rúða and OS rūwi ‘rough hide’ I assume a laryngeal base /rewX-/.

4.42f. OE skēawian, OS skauwon and OHG scauwōn ‘gaze’ are developments of -aww-.

From Gk. κοέω ‘perceive’ and Lat. cavēre ‘take heed’ I assume a laryngeal base /(s)kowX-/.

4.42g. OIcel. snøggr, snǫggr ‘clipped, bald, bare’ show development from PGmc. -aww-.

Although WP derive these forms from the IE root qen-, Latin has a cognate novā-cula ‘razor’ from which we can determine the extensions of the root that are found in Gmc.; I assume these are -ewX-. A reduced grade of such a laryngeal base is found in Gk. ξῡ́ω ‘shave’.

4.42h. OE ðēaw, OS thau ‘custom’, and OHG kathau ‘discipline’ show development from -aww-.

The Gmc. forms have no close cognates, but have been related to Lat. tueor ‘protect, observe’. This gives us no evidence for laryngeals.

4.42i. OS tou, OFris. touw, tow ‘rope, tool’ show development from -aww-.

WP derive these words from IE dewā, that is, a laryngeal base /dewX-/, of which Skt. dū-rá- ‘distant’ shows zero grade. The Gmc. words are possibly o-grade forms of this root.

4.43. -uww-.

4.43a. Goth. skuggwa ‘mirror’, OIcel. skugge ‘shadow’, with OIcel. skyggua ‘overshadow’ are developments of -uww-.

From Skt. skunāti ‘covers’ I assume a laryngeal base /(s)kewX-/; OHG skūr ‘protection’ is a zero-grade form of this base.

4.44. -ayy-.

4.44a. OE clæʒ and OWFris. clay ‘clay’ indicate development from -ajj-.

From OIr. glenaid < /gly-n-eX-ty/, Welsh glynaf ‘adhere’ I assume an IE base /gleyX-/. The Gmc. nouns may be in the o-grade, like Lat. glūs < *glois ‘clay’.

4.44b. The OHG noun screi, screige ‘cry’, like the pret. sg. screi of scrīan ‘cry’ show development from -ajj-.

I find no evidence of a laryngeal root, other than the zero-grade forms, Gk. κρῑγή ‘gnashing (of teeth)’ and OIcel. hríka ‘gnash’.

4.44c. Goth. -waddjus and OIcel. veggr ‘wall’ show development of -ayy-.

From the intonation of Lith. výti ‘turn’ and the long ā of Skt. vā-na-, Lat. viēre ‘bind, weave’ I assume a laryngeal base /weyX-/.

4.44d. Crim. Goth. ada, OIcel. egg, OS gen. pl. eiiero, OHG dat. sg. eiie are developments of -ajj-.

The Gmc. words are related to Gk. ᾠόν, Lat. ōvum ‘egg’. But the initial vowel in the Gmc. words is short, in contrast with those of the forms in other IE dialects. Before analysis by the laryngeal theory such a contrast was inexplicable. I assume that Gmc. preserved the uncontracted PIE form /oXy-es-/ ; the other dialects show development from the contracted form /ōy-es-/. The uncontracted form became PGmc. /aXj-es-/ > /ajj -es-/ > /ej j -is-/; from this the attested forms developed.

4.44e. Goth. daddjan, OSwed. deggia ‘suckle’ show developments of -ajj-. Skt. dhāya ‘sucking’, dhītá ‘sucked’, Gk. θη̑σθαι, OHG tāju, tāan ‘suckle’ lead us to reconstruct a base /dheXy-/.

The relation of Goth. daddjan to this base is difficult, for the apparently closest parallel Skt. dháyati ‘sucks’ differs strikingly in meaning. Mikkola therefore proposed a connection with Sloven. dojíti ‘suckle’. If we accept this connection, daddjan gives evidence for the development of -oXy- to -ajj-.

4.44f. OHG hei ‘dry’, gihei ‘heat’ and arheijēn show development from -ajj-.

WP derive these forms from PIE kāi-, /keXy-/, which is found with a t-extension in Lith. kaistù ‘become hot’.

4.44g. MHG heie ‘hammer’ and MLG heien ‘strike’ show development of -ajj-.

Cognates with d-extension are Skt. khidáti ‘tears, pushes’ and Lat. caedō ‘strike’. From the aspirated stop of the Skt. form and the vocalism of the Lat. verb I assume an unextended base /keXy-, kXey-/, from which the Gmc. forms derive.

4.44h. OIcel. skeggia ‘ax’ shows development of -ajj-.

From the MIr. scīan ‘knife’, Welsh ysgīen ‘knife’ I assume a base /(s)keXy-/; this is found with d-extension in Lith. skíedžiu, skíesti ‘separate’.

4.44i. The inflected forms of the Gmc. words for ‘two’ contain many difficulties, among them the gen. pl. Goth. twaddjē, OIcel. tueggia, OE twēʒ(e)a, OS tweio, and OHG zwei(i)o. The genitive forms in all Gmc. dialects developed from -ajj-.

While the source of the Gmc. endings is unclear, it is apparent that they were closely associated with forms of Goth. bái ‘both’, of which Goth. baddjē and OIcel. beggja show lengthened j, again in the genitive.

From the cognates of bái, Gk. (ἄμ)φω, Lat. (am)bō I reconstruct PIE /bheXw/. To this and PIE /dwoXw/ the genitive was added in some way, possibly as Brugmann suggests, Gdr.2.2.10, 76-7, to a stem made with the collective; he compares Lith. dvejų̃, also a genitive. On the basis of such an uncertain set of forms we can draw no inferences about phonological developments; but I assume that twaddjē and baddjē are of the same origin, with respect to ending. The PGmc. gen. of the word for ‘two’ was apparently similar to the Skt. gen. du. dváyos, except that like the Gmc. words for ‘egg’ the laryngeal was preserved in Gmc., i.e. /twaXyēn, twaXyōn/.

OIcel. þriggia is the only form of the numeral ‘three’ in the Gmc. dialects with lengthening. Since the genitive forms in the other dialects, e.g. Goth. þrijē, seem closer to expected developments from the PIE forms, I assume analogy from OIcel. tueggia and beggia.

4.44j. OSwed. þrægge gen. sg. þræggia ‘covering, roof’ gives orthographical evidence of development from -ajj-.

The Gmc. words are derived with -y- extension from the laryngeal base found in Skt. trā́yate ‘protects’, Lat. intrāre ‘enter’; -ajj- then developed from -eXy-.

4.45. -iyy-.

4.45a. OIcel. Frigg, the name of Odin's wife, OHG friia, OS frī ‘free’ show development from -ijj-.

The closest cognate is Skt. priyá ‘dear’. From the related Skt. verb prīṇā́ti ‘is pleased’ I assume a base /preyX-/.

4.45b. Goth. iddja, OE ēode ‘went’, both of which point to -ijj-, are as perplexing as ever, even though we now have Hittite forms to assume a base ʔyah- (IHL 44). The Skt. forms based on a Skt. root ay- and those based on a root yā- can be derived from ʔyah-. But this is of little help for analysis of iddja and ēode; until we can relate the stem forms of these isolated verbs to forms of other verbs, we have no basis for determining their origin.

4.46. FORMS OF UNCERTAIN ETYMOLOGY.

The following Gmc. forms which contain evidence for lengthening are too uncertain in etymology to support inferences about the origin of the lengthening: -ww- OSwed. gnugga ‘scrape’; OIcel. gugna, gyggwa ‘scare’; OS leia ‘rock’, OHG leige ‘on the way’; OIcel. styggua ‘frighten’; -jj- OE cæʒ, OEFris. kei ‘key’; OIcel. gnegg ‘whinny’, gneggja ‘whinny’, OE hnǣʒan, OS tōnēhjan ‘whinny’; OE hwæʒ ‘whey’; OHG hwaijōn ‘neigh’, with spellings hwaiiot, uueigot.

4.5. The conditions under which /w/ and /y/ were lengthened

If we sum up our findings, we have the following evidence:

for eww < ewX 3; 5 uncertain; 1 unknown
  aww < o, a, ewX 5; 1 uncertain; 3 unknown
  uww < uwX 1
  ajj < o, a, eyX 2; 1 uncertain
  ijj < iyX 1; 1 uncertain
  ajj < o, a, eXy 6; 2 uncertain.

‘Holtzmann's Law’ may then be restated as follows: PGmc. -w- was lengthened after any short vowel when reflex of a laryngeal followed -w-; PGmc. -j- was lengthened after i when reflex of a laryngeal followed -j-, and after a when reflex of a laryngeal preceded or followed -j-.

On the other hand, -w- and -j- were not lengthened when there was no contiguous reflex of a laryngeal; for example, ǽr ‘sheep’, Gk. ὄις < /Xowis/, and Goth. áiz ‘bronze’, Skt. áyas ‘iron’ /Xayes/.

I assume that the evidence for this restatement of ‘Holtzmann's Law’ is so weighty that laryngeals can be assumed in pre-Gmc. forms when the Gmc. forms contain lengthened -w- or -j-. Lengthened -j- and -w- dating from PGmc. is then another development, which can be added to those in chapter 3, from which laryngeals may be assumed for the IE etyma.

Footnotes

1 A. Holtzmann was the first Germanist to point out that the lengthened resonants are found in words whose cognates in other dialects have unlengthened resonants; see the notes to his edition of Isidor's Epistolae ad Florentinam sororem 128-30 (Karlsruhe, 1836). He called the Gmc. development ‘doubling’ (Verschärfung); this term is still in use; the development is also referred to as Holtzmann's Law. Streitberg, Germanisch 323-6, gives a history of the scholarship.

2 W. Braune, Gotisches ddj und altnordisches ggj, PBB 9.545-8 (1884), compares modern Hungarian gy. Braune does not think that the ENGmc. spellings indicate a long stop, but rather that the letters were repeated to indicate a stop rather than a spirant. C. Marstrander, NTS 3.108 (1929), suggested on other than linguistic grounds that worship of a Gothic god Friddja, Friddjōs extended to the NGmc. area, where we find the name Frigg, Friggiar. This suggestion supports Braune's assumption of similarity in pronunciation of Gothic ddj and NGmc. ggj.

3 Some linguists seem to have been unaware of the evidence in WGmc.; see R. Trautmann's comments on Mikkola, Suum cuique, KZ 53.87-90 (1925).

4 R. Kögel, Ueber w und j im westgermanischen, PBB 9.523-44 (1884); illustrations of the spellings may be found below.

5 F. Kluge, Beiträge zur Geschichte der Germanischen Conjugation, QF 32, (Strassburg, 1879). See 127-30, Excurs über Gotisch dd und gg.

6 J. Schmidt, review of Kluge, Beiträge . . . , AfdA 6.117-29, esp. 127 (1880).

7 F. Bechtel, Ueber die urgermanische Verschärfung von j und w, GGN 1885.235-9. It might be worth noting, in view of contemporary explanations of phonological phenomena by means of the laryngeal theory, that after Verner's article in KZ 23 many developments were explained by the position of the accent. Parallels were cited from other languages to support such views. H. Zimmer, Keltische studien, KZ 32.153-240 (1891), cited on pages 218-9 a parallel Irish development of j to gg after accented vowel. F. Trautmann in a review in AfdA 35.105-9 (1911) on page 107 alluded to lengthening of w and y before the accent in Prakrit dialects which Pischel had demonstrated, KZ 35.140-50. J. J. Mikkola, Die Verschärfung der intervokalischen j und w im Gotischen und Nordischen, Streitberg Festgabe 267-71 (Leipzig, 1924) found the key to the Gmc. development in a similar development, p. 267: ‘Im Ungarisch-slovenischen heisst es bíje “schlägt”, aber pidjé (aus pijé) “trinkt”, d. h. j ist nach einer betonten Silbe geblieben, aber vor einer betonten Silbe in d′ verwandelt worden, s. Asboth, Archiv f. slav. Philologie, xxxiii, 322. Hier haben wir den Schlüssel zum Problem der sog. Verschärfung von intervokalischen j und w im Gotischen und Nordischen.’ Such citing of parallels from other languages in support of either of two views indicates the secondary use which ought to be made of such analogies.

8 H. Paul, Ausfall des j vor i und des w vor u im westgermanischen, PBB 7.160-8 (1880); in a footnote on 165 he states that Gmc. lengthening is independent of the IE accent.

9 Gdr. I.1.283 and 331. W. van Helten, Germanisches. LXVII. Zur entwickelung von altgerm. jj und ww, PBB 30.240-8 (1905), cites additional words in support of Brugmann.

10 H. Osthoff, Etymologische Parerga 1.138.

11 H. L. Smith, Jr., The Verschärfung in Germanic, Language 17.93-8 (1941). Smith has now modified his views in some respects, and may soon publish his revised theories. A modification of Smith's theory is suggested by E. Polomé, A West Germanic Reflex of the Verschärfung, Language 25.182-9 (1949). This, with other previous theories, is reviewed by F. van Coetsem, Le renforcement des semi-voyelles intervocaliques en germanique (j/jj > jj > gotique ddj etc.), Leuvense Bijdragen 39.41-78 (1949), who prefers the explanation suggested by Brugmann and Boer. See also the subsequent article of E. Polomé, Laryngeal-theorie en Germaanse Verscherping, Handelingen der zuidnederlandse Maatschappij voor Taal- en Letterkunde en Geschiedenis 4.61-75 (1950).

12 W. M. Austin, A Corollary to the Verschärfung, Language 22.109-11 (1946).

13 R. Kögel, PBB 9.523-44; A. Noreen, Abriss der Urgermanischen Lautlehre 160-2 (Strassburg, 1894); R. Trautmann, Germanische Lautgesetze in ihrem sprachgeschichtlichen Verhältnis (Diss. Königsberg, 1906). Mikkola gave a selected list in Streitberg Festgabe.