Proto-Indo-European Phonology

< previous section | Jump to: next section >

9. The Origin of PGmc. Long Close e

9.1. The forms in which long close e is found

From the evidence in the NWGmc. dialects it is generally assumed that the PGmc. phonological system contained two long e phonemes, one open, the other close. The type of articulation is clear from historical developments. Open long ē, also written ǣ in handbooks, developed to ā in NGmc., OS, OHG, and remained as ǣ in OE. Close long ē, usually written ē² in handbooks, remained except in OHG where it developed to a high front falling diphthong, ea, ia, ie. Gothic gives no evidence that there was a distinction between the two phonemes in EGmc.; in the Anglian dialect of OE too there is no evidence of a distinction between them.1

The source of Gmc. long open ē is clearly PIE ē; e.g. Goth. mana-sēþs ‘man-kind’, OIcel. sáð, OE sǣd, OS sād, OHG sāt ‘seed’, Lat. sēmen, OCS sěmę ‘seed’.

The origin of Gmc. long close ē has been the subject of much discussion.2 IE cognates have been found for many of the Gmc. words with ē², and the origin of the ē² in such words apparently established. But no consistent pattern of origin has been found; some ē²'s are assumed to have developed from PIE ēi, others from PIE ē in certain verb categories, others from PIE [iz]. The problem of the origin of ē² is not so much a problem of determining its origin, but rather a problem of explaining how the various combinations in these IE etyma can have developed to one phoneme in Gmc.

It would be difficult to deny the interrelationship between OE mēd ‘reward’ and Gk. μισθός ‘wages’, especially since an OE form meord is attested. And the likelihood of finding in Germanic and Lithuanian two unrelated verbs so similar in form and meaning as OHG liaz, OE lēt ‘left, let’ and Lithuanian léidžiu ‘let’ is small. The origin of Gmc. ē² in words of each of these is undisputed. But the development of such different sound patterns to one Gmc. vowel has found no plausible explanation.

The words in which ē² occurs have often been listed, and divided into groups by origin:3

  • A. nouns and adjectives in which ē² comes from PIE ēi, e.g. OHG fiara ‘side’, Skt. sphāra ‘spread out, wide’;
  • B. the preterite forms of seventh class verbs with presents in
    • PGmc. ai, e.g. OHG heizan ‘call’
    • PGmc. al, an plus consonant, e.g. OHG haltan ‘hold’
    • PGmc., ǣ e.g. OHG lāzan ‘let’;
  • C. nouns with cognates showing [iz], e.g. OE mēd, OHG miata ‘reward’, Av. miždəm, Gk. μισθός, OCS mъzda ‘reward’;
  • D. some pronominal forms, e.g. (Goth) hēr, OIcel. hér, OE, OS hēr, OHG hiar ‘here’;
  • E. borrowed words with Latin ē or e in the root syllable, e.g. OE bēte, OHG biaza from Lat. bēta ‘beet’, OHG fiebar from Lat. febris ‘fever’.

9.2. Proposed explanations

Of the attempted explanations only that suggested in 1891 by Jellinek has been widely accepted.4 Jellinek limited his explanation to the words of group A, deriving the ē² of these words from PIE ēi. He did not relate this development to that of [iz] to ē². And he did not attempt to solve the origin of the ē² in the preterites of seventh class verbs. Therefore Jellinek's suggestion, even if valid for the small number of words in group A, is at best an incomplete answer to the problem. Moreover, Jellinek did not attempt further identification of the PIE source; for his suggestion was made before the IE ablaut relationships were explored and formulated. Thus he failed to specify whether all PIE ēi, or only original ēi, /eXy/, as opposed to lengthened grade ēi, became ē².

Hirt proposed as objection to Jellinek's suggestion, HU 1. 34, that PIE ēi should have already become ē in PIE, and that this ē should have become PGmc. ǣ, citing as evidence Goth. lētan, OHG lāzan, Lith. léidžiu ‘let’. Hirt's argument cannot be upheld without analyzing the origin of the forms of the seventh class verbs in NWGmc.; he assumes tacitly that the NWGmc. present ablaut grade is that found in the Lithuanian present. But his objection indicates that Jellinek's theory must be refined or abandoned. The ablaut grade from which ē² originated must be determined. This entails an analysis of the origin of the various forms of seventh class verbs; for the forms of a sub-class of seventh class verbs contain in NWGmc. ǣ, e.g. the inf. OE lǣtan, OHG lāzan, and ē², e.g. the pret. 3d sg. OE lēt, OHG liaz. Phonetically, either of these may have developed from PIE ēi, the reflex of which is found in Lith. léidžiu. Only an analysis of the structure of seventh class verbs will permit us to answer this problem.

9.3. Substantives with long close e

Jellinek had already pointed out the presence of ī in cognates of words with ē². We should not expect to find ī if ē² had developed from lengthened grade forms of PIE ei. For IE ī developed from short i lengthened upon loss of a laryngeal; see above 3.6B. Moreover, normal grade forms would be much more likely than lengthened grade forms in o stems like schief ‘crooked’. Morphological evidence therefore points to derivation of words with ē² from the normal grade of ‘original long diphthongs,’ that is, in terms of the laryngeal theory, /eXy/.

When we examine cognates in other dialects we find evidence to support this deduction. Nine substantives are generally cited in group A: OHG fiara ‘side’, MHG kriec ‘resistance’, MHG Kriemhilt, a personal name, MHG schief ‘crooked’, OHG skiaro ‘clear’, OHG stiega ‘path’, OHG wiara ‘gold wire’, OE Wēland, a personal name, OHG ziari ‘ornamental’, to which possibly the name, Friesen, and wiege ‘cradle’ should be added. Although the etymologies of these have necessarily been doubtful, the following have been proposed.

9.3a. Uhlenbeck connected Goth. fēra, OHG fiara ‘side’ with Skt. sphāra ‘spread out, wide’; he supported his suggestion by citing a similar semantic relationship between OIcel. síða, OHG sīta ‘side’ and OIcel. síðr ‘long, wide’.5 Skt. sphāra is related to sphāyate ‘becomes fat’, sphīta ‘wide’, sphātí ‘big, strong’. On the basis of these we may assume PIE /(s)peXy-rē-/ as origin for the Gmc. nouns.

9.3b. The first component in MHG Kriem-hilt is assumed to be the designation ‘masked’ which was applied to Odin ; it is derived from an extended form of PIE /gher-/. We find the reduced grade of this extended form in Gk. χρί̄ω ‘anoint’ and OIcel. gríma ‘mask’; hence we may assume PIE /ghreXy-/.

9.3c. MHG schief ‘crooked’ is derived from PIE /skeX-y-/ with bho suffix. Cognates with different suffixes are Gk. σκαιός ‘left’ and OIcel. skeika ‘swerve’.

9.3d. OHG skēri, skiaro ‘clear, perspicacious’ is a cognate of MIrish scian ‘knife’ and OIcel. skeggia ‘ax’, cf. 4.44h, and like them is derived from PIE /(s) keXy-/.

9.3e. OHG stiega ‘step’ is generally connected with OHG stīgan, Skt. stighnoti ‘ascends’ and Gk. στείχειν ‘stride’. These cognates point to an IE root with short e. If, however, Lat. vestīgium ‘trace, step’ is connected with OHG stiega, some forms of the root may have contained a laryngeal, among these the OHG noun; but our evidence for assuming /(s)teXygh-/ is small.

9.3f. OHG wiara ‘gold wire’ and OIcel. vēl ‘artifice’, OE Wēland, Germ. Wieland, name of a mythical blacksmith, are developments from a PIE root which WP 1.223-7 assume to be /wey-/. /wey-/, like /wer-/ /weg-/, etc. is derived by them from the root /Xew-/ (au-), 1.16-7. Since seṭ-formations are made from this root, e.g. Skt. vāna ‘weaving’, and the pret. ptc. ūta ‘woven’, we must assume forms of the root with laryngeal suffix, /Xw-eX-/. I suggest that the Gmc. forms with y determinative were made from this form of the root, for we find reduced grade forms with ī, e.g. OE wīr ‘wire’, OE wīl ‘trick’. I thus assume PIE /weX-y-/ as the base from which wiara and vēl developed.

9.3g. OHG ziari ‘ornamental’ is derived from the PIE root found in Skt. ádīdet ‘shone’ ; cognates with laryngeal suffix and w determinative are Skt. dyaús, Gk. Ζεύς. I assume that OHG ziari is derived from this root with laryngeal determinative, /dey-X-/, followed by an r suffix.

9.3h. Of the eight substantives discussed, six have cognates which require a reconstruction /-eXy-/, one, stiega, may be derived from /-eXy-/, and one, ziari, from /-eyX-/. The origin of OHG chrēg ‘obstinance’, MHG kriec ‘resistance’ and of the name Friesen is quite obscure. Evidence for the ē² in wiege is not found until MHG times; OHG texts have the form wiga. Various etymologies have been suggested. Kluge, Ety. Wb. 689, connects wiege with the verb wegan and leaves the phonological relation unexplained. wiege has also been connected with Lith. vizgóti ‘waver’; according to this suggestion the ē² would have developed as did that of mēd. We cannot therefore draw any conclusions from kriec, Friesen, and wiege.

In all but one of the words in group A for which we know the IE cognates, ē² developed from a sequence, PIE consonant, eXy, plus consonant.

9.4. Seventh class verbs with long close e

The origin of the ē² in the preterite of seventh class verbs is even more difficult to establish from cognates in other dialects. For very few direct parallels in other dialects have been found. Prokosch said, CGG 151, that ‘of the diphthongal roots of this class . . . there is not one that has been explained with certainty.’ He does not, however, suggest why they are so perplexing. If, like Prokosch, we analyze the seventh class verbs as ‘heavy bases,’ we note that they have few direct parallels in other dialects because they are made from abnormal root patterns.

Almost none of the seventh class verbs are, according to Benveniste's theory of the IE root, see above 2.3, regular developments of PIE verbal roots. In accordance with Prokosch's suggestion that the normal grade of a ‘heavy base’ is found in the pret. 3d sg. we should derive OIcel. hét from PIE /keX-y-d-/; /keX-y-/ would be a normal form 1, but it could not be further enlarged. OIcel. hlióp is derived from a PIE root /kel-/, with -eX- suffix and w determinative; this is a possible form 2, but in hliōp it is further enlarged. According to Benveniste's theory, therefore, such verbs of the seventh class are not made from PIE verbal roots; we cannot then expect them to follow the patterns attested in verbal forms of other dialects; nor is the likelihood great that we shall find cognates of the same complex structure in other dialects.

Although we find few direct correspondences of seventh class verbs in other dialects, we may for many verbs cite cognates that give evidence of laryngeals, and thus uphold Prokosch's thesis that the seventh class verbs are from laryngeal (heavy) bases.

9.4a. The following are examples of seventh class verbs with a PGmc. stem vowel ai in the present:

OIcel. meita, OHG meizan ‘cut’ are developments of the root from which Gk. σμί̄λη ‘knife’, OIcel. smíð ‘artistic work’, OHG smīda ‘metal’ are made. We may assume a PIE root /meX-/, enlarged in these forms by a y determinative. The Gmc. preterites would have developed from an extended, and according to Benveniste's theory aberrant, form of this root, /meX-y-d/.

OE scādan, OHG sceidan ‘separate’ are extensions of the root from which OHG skēri and OIcel skeggia are made. The preterite forms would have developed from PIE /skeX-y-t-/.

OIcel. sueipa ‘throw’, OE swāpan, OHG sweifan ‘swing, sweep’ are derived from PIE /(s)weX-/ enlarged by y and b. MDu. swāien ‘sway’ shows the full grade without b enlargement; forms with different enlargements which show zero grade are Welsh chwil ‘whirling’ and MHG swīmen ‘move back and forth’. The preterite forms would have developed from PIE /sweX-y-b-/.

OE tǣsan, OHG zeisan ‘pull apart’ are developments from the PIE root /deX-/ with y and s extensions. Skt. dā́ti ‘separates’ shows full grade of the root; zero grade of the root with y determinative is attested in Arm. ti and OIcel. tīð ‘time’. The preterite forms would have developed from PIE /deX-y-s-/.

9.4b. We find even fewer cognates for verbs of the sub-class with Gmc. al, an in the present to support the suggestion that they developed from laryngeal bases.

OE bannan, OHG bannan ‘command’ are derived from PIE /bheX-/; the normal grade of this root is found in Dor. Gk. φᾱμί ‘say’ and in the nouns, Lat. fāma ‘report’, OIcel. bón ‘request’. Like Skt. bhánati ‘speaks’ the Gmc. verbs are made from this root with n extension. The preterite forms would have developed from PIE /bheX-n-/.

OHG spannan ‘stretch’ is derived from PIE /(s)peX-/ ; cognates are Av. spā- ‘cast off’, Gk. σπάω ‘draw’, and with y determinative OIrish séim < spē̆imi ‘thin’. The Gmc. preterite would have developed from PIE /speX-n-/.

Besides deductions based on morphological analysis we therefore have evidence to assume for some of the verbs with ē² in the preterite a pattern of vowel, laryngeal, resonant. Not all verbs in these groups show such a pattern; either the previous analysis has been inadequate and they must be analyzed as have been those listed above before cognates in other dialects can be found, or the pattern of such verbs as those listed above spread to verbs with similar structure, and led to the establishment of the seventh class. The bases for these deductions are necessarily weak. For it was apparently one of the characteristics of Gmc. to expand greatly the number of verbs made with more than two extensions of the root. We can only deal with this development as a Gmc. innovation and attempt to analyze it in accordance with our analysis of other PIE and PGmc. roots.

9.4c. The third sub-class of seventh class verbs, those with ǣ in the present, offer an additional difficulty. Verbs of this sub-class differ from those of the two sub-classes mentioned above in having a present tense vocalism that did not develop from a weak grade. It is therefore necessary to determine the source of the present tense vowel ǣ as well as that of the preterite.

From cognates we know that some of the verbs in this sub-class developed from roots with PIE /-eXy-/.

Gothic lētan, OIcel. láta, OE lǣtan, OS lātan, OHG lāzan, like Lith. léidžiu ‘let’, are from PIE /leXyd-/.

OIcel. ráða, OE rǣdan, OS rādan, OHG rātan ‘advise’ are assumed to have developed from PIE /reXydh-/; OHG reitun ‘advised’, attested in the OHG of Otfrid, apparently is a reflex of the normal grade, Goth. ga-ráidjan, OIcel. greiða, MHG gereiten ‘put in order, prepare’ apparently are reflexes of the reduced grade. Most cognates in other dialects have not preserved evidence of y in the root, e.g. Skt. rā́dhyati, rādhnóti ‘is successful’, OIrish rádid ‘says’. But we have sufficient evidence to assume PIE /reXydh-/.

According to the Brugmann-Wood theory the preterite forms of these verbs developed from the normal grade -eXy- (ēi); Prokosch here departs from the theory and assumes, CGG 182, that preterites of these verbs developed by analogy with first class verbs. It seems likely that the present forms in ǣ also developed from PIE -eXy- after IE loss of the /y/. If so, the divergent developments must be reconciled.

There is adequate material to show that such sequences, (original long diphthongs), may undergo one of two developments in PIE; the final element may or may not be lost, e.g. /eXy/ may become ē or ēi. But the material is very small for determining the conditions under which the loss of /y/ occurred. The most likely suggestion is that of Brugmann; according to him, Gdr. I.1.203 fn., the loss depended on the position of the syllable boundary. If ēi stood at the end of a syllable the i was lost, e.g. *lēi-dō became *lēdō; if a consonant followed in the syllable, the i was maintained, e.g. in *lēid-mi, *lēits.

If we assume PIE /eXy/, ēi, for the etyma of both the present and preterite stems of the verbs in sub-class 3, the differing vocalism in Gmc. would be in accordance with Brugmann's suggestion. For the PGmc. present indicative endings began with vowels; since the stems of all verbs in this sub-class end in one consonant, the /y/ would have been lost; PIE ē would regularly have developed to Gmc. ǣ. In the preterite on the other hand all endings but the 3d sing. began with a consonant; /eXy/ would therefore have been maintained in the preterite. The contrast between pres. lǣt- and pret. 2t- was then a regular development from PIE consonant, eXy, plus consonant in this and similar verbs.

I conclude that in the second group of words with ē², the preterite forms of three sub-classes of seventh class verbs, ē² developed from e and laryngeal before resonants.

As in the developments described in Holtzmann's Law and other Gmc. phonological developments, reflexes of laryngeals survived in the neighborhood of resonants also in the two groups of words discussed above. When they were lost, preceding e was lengthened. This ē was relatively high, and in most Gmc. dialects contrasted with the phoneme that had developed from PIE ē. ē² is thus the product of compensatory lengthening, similar to that in Skt. nīḍá ‘nest’ from PIE [nizdo-], and mīdhá ‘reward’ from PIE [mizdho-].

In forms with i following the ē², the i was normally lost (but survived in OHG reitun); other resonants and obstruents were preserved after ē². The loss of i is not remarkable; for after a long high vowel it would virtually have been absorbed. We have ample evidence that ē² was a very high vowel; Sievers, PBB 1.505, has pointed out that the skalds rimed it with i; Kögel, IF 3.285-6, cited OS and OIcel. forms written with i, e.g. OS hīr ‘here’ beside hēr. And since i was lost in the present in verbs like lētan, became the second part of a diphthong in verbs like áikan, and was never found in the present of verbs like haltan, by analogical influence from the present the i of the preterite would have been lost rather than maintained.

9.5. Forms with long close e from [iz]

While these developments then are similar, that in the words in group C appear to be quite different. Words in group C have ē² in some forms, the regular developments of [iz] in others. The [z] of PIE [mizdh-], cf. Gk. μισθός, OCS mъzda ‘reward’ is preserved in Goth. mizdō and regularly became r in the OE hapax legomenon, meord; OE mēd, OS mēda, OHG miata, however, have lost the z with lengthening of the preceding vowel.

In chapter 7 we have seen that in some verb forms reflexes of laryngeals were preserved in PGmc. as continuants. If ē² developed from e plus reflex of laryngeal (written Z below), at some time in early Gmc. there were found side by side forms with a pattern: leZd- and 2d. Besides these there were words of the pattern: mizd-; the allophone of /i/ before /z/ was apparently very low, see Twaddell, Language 24.147. The allophone of the /i/ in mizd- was then very similar to that of leZd. Thus on the pattern

leZd- : 2d
mizd- : x

alternate forms with 2d were made. As we might expect from such an analogical development, only a few such forms were made, and beside these survived the original forms. We find only those cited above, and OE cēn, OHG kien from *keznos, cf. Russ. sosná ‘pine’ and MLG hēde beside OE heorde, MDu. herde ‘tow’.

9.6. Long close e in borrowings and pronouns

Gmc. ē² accordingly developed from Gmc. e which was lengthened when a following consonant, usually the reflex of a laryngeal, was lost. I assume that these were the primary sources of ē², and that ē² did not develop in the two other categories of forms, D and E, until after the phoneme was established. After a contrast had been established between ē² and ǣ —after ē² had been established in the Gmc. phonological system as a comparatively high front vowel—other vowels coincided with this phoneme. Chief among these were high front vowels in the accented syllable of borrowings. from Latin into Gmc. Such are: OHG biaza, OE bēte, Lat. bēta ‘beet’; OHG mias, VLat. mēsa ‘table’; OHG riemo, Lat. rēmus ‘oar’; OHG ziagal, Lat. tēgula ‘tile’; OHG ziahha, Lat. thēca ‘case’; OHG briaf, Lat. breve ‘letter’; OHG fiebar, Lat. febris ‘fever’; OHG priestar, Lat. presbyter (OFrench prestre) ‘priest’; OHG spiagal, Lat. speculum ‘mirror’; and the proper names, OHG Trier, Lat. Trevir, OHG Kriach, Lat. Graecus, MHG Riez, Lat. Raeti.6 It is noteworthy that Lat. ē in unaccented syllables was not reproduced in borrowed words as ē², e.g. OHG munizza, OE mynet, Lat. monēta ‘coin’; Gmc. ē² was apparently a member only of the phonological system of accented syllables.

ē² in the last group of words is found in pronominal forms: OHG (Tatian) thie ‘he’; OHG (h)wia ‘as’; OHG hiar, (Goth.) hēr, OIcel. hér, OE, OS hēr ‘here’. Although attempts have been made to explain these as regular developments, see HU 33-5, our knowledge of the vowels in the PIE etyma of these pronouns, and of the development of final syllables in Gmc., is too inexact to permit us to consider these anything but secondary developments.

9.7. Explanation by means of the laryngeal theory

In this chapter I have attempted to show how we may provide a plausible explanation for the origin of ē² with a PGmc. phonological system containing reflexes of laryngeals. Some of our deductions must be based on morphological evidence. For we cannot expect handbooks which did not accept the laryngeal theory to provide us with evidence in its favor. If, as seems likely, the seventh class verbs developed from laryngeal bases, analyses of forms in other dialects with the help of the laryngeal theory may assist us in solving some of the phonological problems of words with ē² and in finding cognates for seventh class verbs for which no etymologies can be cited.

The simplicity of the explanation that can be offered for the origin of ē² with the help of the laryngeal theory is good evidence in its favor, as are the relic forms found occasionally in our old records. OHG reitun is a direct reflex of the normal grade form in the preterite. And I assume that anagelierzon, a preterite of lāzan recorded in an OHG gloss, has preserved in the r after the reflex of ē² evidence for a consonantal increment in the stem; survival of such an r as reflex of a laryngeal would agree well with the evidence presented above (Chapter 7).

The theory of the origin of ē² presented here is further evidence of the survival of reflexes of laryngeals into PGmc. in the neighborhood of resonants. Without investigations of phonological developments in other dialects we cannot determine whether such survival is peculiar to Gmc. In an attempt to answer this question, two crucial studies will be undertaken in the next two chapters.


1 Kossinna, in Festgabe an Karl Weinhold, 37ff. (Leipzig, 1896), tried to distinguish the Gothic reflexes of PGmc. ē² and ǣ; for a refutation see Streitberg, GE6 73. Streitberg, Germanisch 368, gives a bibliography of the work on the articulation of the two long e sounds, including earlier studies in which ē² was described as a relatively open vowel.

2 The bibliography is enormous; it has been reviewed by Streitberg, Germanisch 367-9; see also Hirt, HU 1.33-5.

3 Streitberg, UG 65-6; Hirt, HU 1.33.

4 Germanisch ē², PBB 15.297-301.

5 PBB 30.275 (1905); cf. WP 2.656-9. WP 2.40 cite other suggestions, and question Uhlenbeck's etymology on the unconvincing grounds that movable s is missing in the Gmc. cognates.

6 The source of the borrowing of OHG Kriach is disputed. Sievers, PBB 18.410, assumed that it was borrowed from Greek, with Γραικός becoming *krǣikaz becoming *krēikaz becoming krēkaz. Derivation from Latin, however, seems more plausible, cf. Hirt, HU 1.33.

< previous section | Jump to: next section >

  • Linguistics Research Center

    University of Texas at Austin
    PCL 5.556
    Mailcode S5490
    Austin, Texas 78712

  • For comments and inquiries, or to report issues, please contact the Web Master at