Proto-Indo-European Phonology

15. The Development of the PIE Phonemic System

15.1. Bases for suggesting earlier IE phonemic systems

In chapter 13 I have suggested a phonemic system of PIE which is based on a comparison of the phonemic systems of the dialects and analysis of the patterning of the dialect and PIE phonemes. This system is not the earliest one that we can determine. We know that the PIE phonemic system itself developed from an earlier, different system. For earlier phonemic patterns can still be detected in the PIE system, e.g. an interchange between /e/ and /o/ which correlates partially with location of the accent. Moreover, relic forms are found in various dialects, e.g. Gk. πότνια as compared with Skt. pátnī. Furthermore, the frequency of some phonemes is quite out of balance, e.g. /b/. By analyzing such characteristics of the PIE phonemic system we can determine at least partially earlier phonemic systems.

The earlier patterns can be detected most clearly in the vowel system. We can conclude either that this was the part of the system that underwent the greatest change, or that it would have preserved more clearly traces of an earlier system. The former alternative is preferable; apparently in pre-IE the obstruents underwent relatively little change as compared with the rest of the system. We have two indications of possible pre-IE obstruent changes: 1. the great infrequency of /b/; 2. orthographic evidence in Hittite for a labio-velar variation of [kw] before vowels, [ku] before consonants, e.g. a-ku-wa-an-zi ‘they drink’, e-ku-zi ‘he drinks’. Even if we should consider these indications that the labio-velars and /b/ are relatively late additions to the pre-IE obstruent system, such changes would be minor as compared with those found in the rest of the phonemic system. The reconstruction of earlier forms of the IE phonemic system is therefore restricted to the non-obstruents here.

15.2. Accent changes in pre-IE

The changes in the pre-IE non-obstruent system are associated with changes in the accent system. These were two: 1. phonemicization of pitch accent; and earlier: 2. phonemicization of stress accent. Both periods of distinctive accent were preceded and followed by periods of non-distinctive accent. In the PIE phonemic system pitch accent was non-distinctive; it could stand on any syllable of the word and was not correlated with vowel timbre or vowel length. In constructing the earlier vowel systems we must relate the changes of accent and accompanying vowel shifts with other vowel developments, notably those in the neighborhood of laryngeals.

15.3. Change of /e/ to /o/ and resultant disruption of the pre-IE phonemic system

When we examine the PIE material, we find an interchange between /e/ and /o/, /e·/ and /o·/, /a/ and /o/ which is generally accompanied by differences in pitch accent: e.g. Gk. ἄκοιτις ‘spouse’, κει̑ται ‘lies’; ἀπάτωρ ‘fatherless’, πατήρ ‘father’; ἀοιδή ‘singing’, ἀείδω ‘sing’; perf. ἔρρωγα, ῥήγνῡμι ‘break’; θωμός ‘heap’, τίθημι ‘place’; ὀξίνη ‘harrow’, ἄκρις ‘tip’; perf. τέθωκται, θήγω θᾱ́γω ‘sharpen’. Hirt, IG 2.172-87, gives further examples, and bibliography. The /e/ (/a/) vowels are characteristically accompanied by pitch accent, the /o/ vowels by lack of pitch accent. Although the /o/ vowels lack pitch accent, they occur where we would expect a normal grade vowel.

After various studies the conditions of change have been defined: /é/ /é·/ [é é· á á·], with phonemic pitch accent, became [ò ò·] when the chief accent was shifted to another syllable, and the syllable accented formerly received a secondary pitch accent.

e.g. /péte/ [péte] > /pèté/ [pòté]

The [o] allophone of /e/ developed regardless of the environment; evidence is found for it in the neighborhood of obstruents, resonants, and laryngeals, including A; it developed for long /e·/ as well as short /e/.

I assume that this phonetic change took place before the loss of laryngeals. For /Aé/ [Aa], which in PIE is /a/, interchanges with /Aè/ [Ao]. This assumption is supported by reflexes in Hittite; laryngeals are preserved there in the neighborhood of a, but the change of /e/ to [o] had already taken place. Since the change of [e] to [o] is the last of the ablaut changes, I conclude that all ablaut changes took place before any of the phonemic changes that the vocalic system underwent as a result of modification by laryngeals.

15.3a. We may then assume a pre-IE vocalic system with distinctive pitch accent and the phonemes: /e e· _e/. These segmental phonemes had various allophones: [ò ò·], [Aa, aA] [Aa· a·A] [A_{a a}A] and elsewhere [e e· _e]. /e·/ was found in syllables which had been open before the loss of a following unaccented vowel (Dehnstufe). Vowels with non-distinctive length were found in syllables in which laryngeal and consonant followed the vowel.

At some time before the establishment of the PIE phonemic system pitch accent became non-distinctive. With this accent shift vowels which had been non-distinctive variants of one phoneme became phonemes. Contrasts between
/tét/ and /tèt/     /tér/ and /tèr/   were shifted to
/tet/ and /tot/     /ter/ and /tor/.

When the pitch accent pattern was disturbed, when vowel timbre instead of pitch became the characteristic distinction between the members of the vowel system /a/, as well as /o/, became phonemic. As the contrast between
/tét/ [tét] and /tèt/ [tòt] was shifted to a contrast between /tet/ and /tot/, so the contrast between
/Aét/ [Aát] and /Aèt/ [Aòt] was shifted to a contrast between /Aat/ and /Aot/.

The disruption in pitch accent therefore initiated the development of the PIE vowel system reconstructed above, that is, the vowels /e o a e· o· a·/.

15.3b. To this development of the vowel system I ascribe the disruption of the laryngeal system. After a vowel system of three timbres had been established by the shift of pitch accent, contrasts which formerly were marked by laryngeal accompanied by vowel timbre were now marked distinctively by vowel timbre. The former contrast between
/pAét/ [pAát] and /pét/ [pét], with the laryngeal supplying the characteristic difference, now was between
/pAat/ and /pet/, with the vowel supplying the characteristic difference. In such environments, and in patterns like /peAt/ [pa·At], the laryngeal was lost. This loss was succeeded by loss of the two other laryngeals in the similar environments, and eventually by the loss of laryngeals in every environment.

15.4. Earlier development of /_e e·/

The patterns of an even earlier vowel system have left traces in PIE which have survived the changes listed above. /e/ and /e·/ are characteristically found in accented syllables, /_e/ in unaccented syllables. Interchanges of quantity like that between /e/ and /_e/ are found in languages with stress accent and correlate with the position of the accent. We may assume a similar correlation in an early stage of pre-IE; when stress accent was distinctive, [e] occurred only in stressed syllables, [_e] only in unstressed syllables, e.g. Gk. χθές : χθιζός ‘yesterday’, ἵστᾱμι ‘stands’ /'steA-/ ['steA-] : στατός /steA-'te-/ [st_eA-'te-].

The relation of PIE /e/ to /e·/ is not so readily apparent, but enough patterns remain to show us that /e·/ developed from /e/ when standing in an open syllable and when the following syllable lost its syllabic status. This development (Dehnstufe) occurred for example in the nom. sg. πατήρ; by comparison with the acc. sg. πατέρα it may be assumed that the nom. too formerly had a third syllable. (See Hirt, IG 2.36–51; the fundamental study is still that of Streitberg, IF 3.305–411.)

Since vowel quantity at the time when stress accent was distinctive was a function of the accent we may assume an early pre-IE vowel system with the contrasts:

In this stage of pre-IE [_e] occurred only when accompanied by minimum stress, [e] only when accompanied by maximum stress, [e·] only when accompanied by maximum stress and a fixed morphological structure. This vocalic system then was composed of one member. The investigations of various Indo-Europeanists have pointed to such a conclusion, but few if any have admitted a linguistic structure with a vowel system composed of one member. Debrunner in his review of EI, IF 56.55-8, noted that Kurylowicz' theories would leave us with only one vowel, but added that Kurylowicz specifically rejected this conclusion. The conclusion may have been generally rejected because of the vocalic structure of the IE dialects spoken today. To my knowledge no IE dialect, spoken today or formerly, has such a structure; and the well-known IE dialects have a comparatively ample vowel system. Consequently the conclusion that from a phonemic point of view a stage of pre-IE had a vowel system of one member seems foreign to the usual observations about IE linguistic structures. But other characteristics of PIE and pre-IE are adequately attested, and generally accepted, though they fail to fit subsequent patterns in IE dialects; such are the PIE variety of formation of present tense forms, the lack of system between present, aorist, and perfect forms. A further reason against such a conclusion seems to have been the peculiarity of such a vocalic system among those found in the well-known European and Asiatic languages. With the increased knowledge of languages the past few years we no longer find a phonemic structure odd though it has a small vowel system. Unless we find in our early IE material evidence for a different analysis of pre-IE in the stress accent stage we must accept the conclusions of the laryngeal theory and the theory of ablaut.

15.5. The pre-stress stage of pre-IE

We can construct one more stage of pre-IE, a pre-stress period. If stress gave rise to conditioned variants, [e _e e·], at an earlier period such stress must have been non-distinctive.

In accordance with its most common reflex the most open segment of the pre-IE syllable, the syllabic peak, is usually written e, as in Benveniste's reconstructions of the IE root. The syllabic peak does not, however, contrast directly with any vowel. Consequently an analysis of the pre-stress stage of pre-IE with a vowel phoneme e is misleading. If we find no phonemes in complementary distribution at the peak of the syllable, we cannot assume a segmental phoneme for this position. The peak of the syllable, syllabicity, must have been a prosodic feature. At this stage of pre-IE segmental phonemes were limited in occurrence to the slopes of the syllable. We may label this prosodic feature, syllabicity, with /∧/ in our reconstructions.¹

It is presumably this pre-stress stage that we have to use for attempts to establish further relations of pre-IE, such as with Proto-Hamito-Semitic. The pre-stress system suggested here could be checked only if we found some such cognate of pre-IE. And with the help of such a cognate we might find answers to problems in the pre-IE obstruent system, such as the origin of /b/, of the labio-velars, of the voiced aspirates, and further information on the allophones of the laryngeals.

15.6. The various stages of pre-IE

We may summarize our conclusions on the development of the pre-IE and PIE phonological systems by setting up a succession of phonemic systems beginning with the pre-stress system of pre-IE.

• A. The pre-stress stage of pre-IE
  1. Obstruents: p (b?) bʰ t d dʰ k g gʰ s
  2. Resonants: y w l r m n
  3. Laryngeals: h x γ ʔ
     A non-segmental phoneme /∧/, syllabicity.
• B.a. The stage of pre-IE with phonemic stress
  1. Obstruents: (unchanged)
  2. Resonants: (unchanged)
  3. Laryngeals: (unchanged)
     Three non-segmental phonemes:
     • ^||, maximum stress
     • ^|, minimum stress
     • ∧.
  With maximum stress /^||∧/ becomes segmental; allophone [e].
  With minimum stress /^| ∧/ is not segmental. It remains non-segmental between obstruents, /^|C ∧ C/, laryngeals, /^|X ∧ X/, obstruents and laryngeals, /^|C ∧ X/, /^|X ∧ C/. In the neighborhood of resonants it combines with segmental phonemes in simultaneous articulation: /^|C ∧ R/ /^|R ∧ C/ /^|X ∧ R/ /^|R ∧ X/; the R has the allophone [i u r̥ l̥ m̥ n̥].
• B.b. Some /^| ∧/ lose their syllabicity and a preceding /^|| ∧/ is /^|| ∧·/. Loss of syllabicity occurred in those syllables standing directly before or after the syllable with the chief stress accent. Hirt labelled these as syllables with zero grade vocalism, and those which retained syllabicity as syllables with reduced grade vocalism. This classification is misleading. For the original distribution depended on the position of the stress; syllabicity or lack of syllabicity was non-distinctive. The terms ‘reduced’ and ‘zero’ grade might be used to refer to an interchange, the result of a change; but they must be carefully differentiated from such terms as ‘lengthened’ or ‘o-grade,’ which represent a phonetic, and later phonemic, change rather than the result of such a change.
• C. The period of non-distinctive stress.
  The allophones of /∧/ become phonemes; the system is otherwise unchanged.
  This stage of pre-IE contains a 4th class of phonemes, which may be labeled: /e e· _e/. These have various allophones; an open vowel [a] in the neighborhood of /h x/; a close vowel [e] elsewhere.
• D. The stage of pre-IE with distinctive pitch.
  The first three classes of phonemes remain unchanged. Pitch-determined allophones of /e e· _e/ develop:

  /é/		[e]			/é·/		[e·]
  /è/		[o]			/è·/		[o·].

  /e e·/ now have three allophones:
  • /è è·/ have the allophones [o o·] in any phonetic environment;
  • /é é·/ have the allophones [a a.] when contiguous with /x h/ (a back allophone also develops for /_e/ in the neighborhood of /x h/.
  • /é é·/ have the allophones [e e·] elsewhere.
  
  When /e/ is followed by laryngeals and consonants its allophones have non-distinctive length.²
• E.a. The period of non-distinctive pitch. The first three classes of phonemes remain unchanged.³ The pitch-conditioned allophones of /e e·/ become phonemes. The resultant vowel system is: /e o a e· o· a· _e/.
• E.b. With this development of the vowel system the laryngeal system becomes disrupted. All laryngeals are lost after /e o a/ when before obstruents; the resulting vowels fall together with /e· o· a·/.
  • Laryngeals are lost also after, and before, the vocalic allophones of preconsonantal /y w/, and two vocalic phonemes result: /i· u·/.
  • Laryngeals also are lost when initial before vowels.
• F. The PIE phonemic system.

15.7. Changes between PIE and the dialects

Between PIE and the various dialects the system undergoes further changes. Many of the changes extend over only a part of the IE speech area.

General changes:

1. the laryngeals are lost as phonemes. Reflexes of them survive in various environments. These have been studied above.
2. the allophones of the resonants become phonemic.
3. reflexes of /_e/ fall together with the short vowels. (Restricted change: over an east-central innovating area the back vowels /a/ and /o/ coincide. Allophones of velar stops become phonemes over this area, and beyond it.)
4. diphthongs develop when vowels followed by resonants occur between consonants.

After changes 2 and 4 the phonemic system is made up of two classes:

• vowels—any phoneme that may be syllabic;
• consonants—the non-syllabics.

A complete description of the extent of these and less general developments would be a description of the phonological characteristics of the various dialects. Such descriptions would be moderately complex, and would be analyses of the phonological structure of the various dialects rather than of Proto-Indo-European.

Footnotes

¹ I am indebted to W. F. Twaddell for this observation. C. H. Borgstrøm, Thoughts about Indo-European Vowel Gradation, NTS 15.137-87 (1949), like L. Hjelmslev, Accent, intonation, quantité, Studi Baltici 6.1-57 (1936-7), esp. pp. 44ff., prefers to deal with a single vowel in an ‘older linguistic state’ than PIE. He reconstructs the etymon of Lat. sunt in this stage as *h₁äsä-nätä. On the basis of his phonological assumptions Borgstrøm makes some interesting observations about the development of the IE morphological system. I cannot, however, accept his fundamental assumption that ablaut was not caused by accent shifts. Moreover, his phonemic analysis implies that all non-vocalic phonemes fall into one class, the consonants; the later IE developments support the assumption of several classes, with laryngeals patterning differently from resonants, and both from obstruents. And if we apply the test of continuity in linguistic systems, see above 14.3, the change in linguistic ‘type’ from a language with a syllabic structure like that in *h₁äsä-nätä to the IE dialects is remarkable.

² A similar situation is found in American English. Vowels are longer before voiced stops and fricatives than before the corresponding voiceless sounds; compare bid [bI·d] : bit [bIt], bidder [bI·dr̩] : bitter [bIt̬r̩].

³ The labio-velars became phonemes at this stage of IE, or even earlier. No means has yet been discovered to determine their relative chronology.