The phonological system is presented in two sub-systems: the consonants & vowels making up the segmental sub-system, and the accentuation & basis of the syllabic structure making up the supra-segmental sub-system. Both sub-systems of Proto-Germanic, like other reconstructed languages, are determined largely by the comparative method. The method of internal reconstruction is also important in determining the earlier accentual system of Proto-Germanic, which then is supported by comparison with those of Indo-Iranian and Greek.
The development of the elements of a phonological systems is affected in part by the syllabic structure of the language. The syllabic structure of Proto-Germanic is still evident in the earliest Runic inscriptions, which show that many of the syllables were open; the closed syllables for the most part ended in resonants, as illustrated by the Gallehus inscription:
ek hle-wa-ga-stiz hol-ti-jaz hor-na ta-wi-do
The initial position of obstruents, g-, h-, t-, d-, illustrates why the first consonant shift was universally carried through except in clusters like st-: their position in the syllable was similar. The only syllable closed by an obstruent is the first. But its cognates, Latin egō and Greek égō, provide evidence that the early Proto-Germanic form was e-ga > e-ka > ek as supported by Runic inscriptions, e.g. on the Ellestad stone. The form ek in turn provides evidence that late Proto-Germanic had a stress accent and that unstressed final vowels were lost after the consonant shift had taken place, as illustrated also by words recorded in the dialects such as Gothic haurn. Further evidence for an initial stress accent in the dialects is provided by the poetry which, like the inscription above with its three syllables beginning with h, was alliterative.
The consonant system consists of ten obstruents, i.e. stops and fricatives, and six resonants. In the early stage of the language, each of the obstruents had the same pronunciation in its various locations, although the voiceless stops may already have been followed by aspiration except when after /f s χ/. Later, /b d g/ had fricative allophones when medial between vowels. The fricatives, notably /s/, may have developed voiced allophones in voiced contexts, and by late Proto-Germanic /z/ was phonemic. When between consonants, the resonants /w m n r l y/ had vocalic allophones in early Proto-Germanic, which developed to /u um un ur ul i/ in the language's later stages.
The vocalic system consisted of eight vowels and four diphthongs. The low back vowel, indicated below by the symbol a, is lower than that of the later dialects, as may be illustrated by the Gothic representation in haurn 'horn'.
Examples are as follows.
/p/ as in PGmc déwpaz 'deep', cf. Go. diups*, ON djūpr, OE dēop, OHG tiuf
/b/ as in PGmc bōks 'tablet' > 'book', cf. Go. bōka, ON bōk, OE bōc, OHG buoh
/f/ as in PGmc fōts 'foot', cf. ON fōtr, OE fōt, OHG fuoz
/t/ as in PGmc téuχanan 'to push', cf. Go. tiuhan, OE tēon, OHG ziohan
/d/ as in PGmc durez 'door' n.pl., cf. Go. daúr, ON dyrr n.pl., OE duru
/þ/ as in PGmc þrsúz 'dry', cf. Go. þaúrsus, ON þurr, OE þyrre, OHG durri
/k/ as in PGmc kŕnam 'grain', cf. Go. kaúrn, ON korn, OE corn, OHG corn
/g/ as in PGmc gárdiz 'garden', cf. Go. gards, ON garđr, OE geard, OHG gart
/χ/ as in PGmc χŕdiz 'wattle', cf. Go. haúrds, ON hurđ, OS hurth, OHG hurd
/s/ as in PGmc sunōn 'sun', cf. Go. sunnō, ON sunna, OE sunna, OHG sunno
[z] as in PGmc méyzaz 'more', cf. Go. máiza, ON meiri, OE māra, OHG mēro
/m/ as in PGmc maχtiz 'might', cf. Go. mahts, ON māttr, OE meaht, OHG maht
/n/ as in PGmc naχts 'night', cf. Go. nahts, ON nātt*, OE neaht, OHG naht
/w/ as in PGmc waganaz 'wagon', cf. Crim.Go. waghen, ON vagn, OE wægn
/r/ as in PGmc reχtaz 'right, straight', cf. Go. raíhts*, ON rēttr, OS reht, OHG reht
/l/ as in PGmc langaz 'long', cf. Go. lagg*, ON langr, OE long, OHG lang
/y/ as in PGmc yēra 'year', cf. Go. jēr, ON ār, OE gēar, OHG jār
/i/ as in PGmc witum 'we know', cf. Go. witum, ON vitom, OE witom, OHG wizzum
/e/ as in PGmc erþō 'earth', cf. Go. aírþa, OE eorþ, OHG erda
/a/ as in PGmc af 'from', cf. Go. af, ON af, OE of, OHG aba, ab
/u/ as in PGmc ufar 'over', cf. Go. ufar, ON yfir, OE ofer, OHG ubir, ubar
/ī/ as in PGmc swīnaz 'pig', cf. Go. *swein, ON svīn, OE swīn, OHG swīn
/ē/ as in PGmc sēþiz 'seed', cf. Go. mana-sēþs, ON sāđ, OE sǣd, OHG sāt
/ō/ as in PGmc flōduz 'flood', cf. Go. flōdus, Run. flodu, ON flōđ, OHG fluot
/ū/ as in PGmc fūla 'foul', cf. Go. fūls, ON fūll, OE fūl, OHG fūl
/ei/ as in PGmc steig-, cf. Go. steigan, ON stīga, OE stīgan, OHG stīgan 'climb'
/ai/ as in PGmc staig, cf. Go. stáig, OHG steig 'climbed'
/eu/ as in PGmc beud-, cf. Go. -biudan, OE bēodan, OHG biotan 'bid, offer, order'
/au/ as in PGmc baud-, cf. Go. báuþ, ON bauþ 'offered'
The Proto-Indo-European consonants are given here in the standard notation, i.e. p b bh. Proponents of the glottalic theory hold that PIE /b d g/ were glottalics, and that the other two sets had aspirated as well as simple allophones; a principal reason for the theory is the presence of relatively few examples of PIE /b/, a situation characteristic of languages with glottalics. But nonetheless the theory is not generally accepted; Szemerényi, for example, strongly rejected it (1996:151-154). Moreover, since the relationship among the three elements is maintained, the modifications proposed in the theory are sub-phonemic.
The principal changes that took place between Proto-Indo-European and Proto-Germanic involved the Proto-Indo-European stops. PIE /p t k/ > PGmc /f þ χ/, PIE /b d g/ > PGmc /p t k/, and PIE /bh dh gh/ > PGmc /β ð γ/ [b d g]. Formulated in this way by Jacob Grimm in 1822, the set of changes is referred to as Grimm's Law. In formulating his 'law', Grimm assumed three classes of consonants using Latin labels: Tenues (T) for the voiceless stops /p t k/, Aspiratae (A) for the compound stops /bh dh gh/, and Mediae (M) for the voiced stops /b d g/. This gave rise to the diagram:
The labels as well as the term 'law' were justified in Grimm's view because he assumed that the later consonant changes in Old High German were a continuation of the earlier changes. In the Old High German changes, voiceless stops (T) became affricates and fricatives, i.e. Aspiratae (A) as in the word Pfeife 'pipe', so that his rule still applies: Tenues change to Aspiratae.
Among other changes between Proto-Indo-European and Proto-Germanic is the loss of labio-velars, which became clusters in Germanic or else fell together with other phonemes.
For ease of comparison, the examples are listed below in the same order as those above, that is, by the order of the Proto-Germanic phonemes rather than Proto-Indo-European reconstructions. Cognates in the other dialects are provided when examples are readily available, and the same items are given here as in the chart of Proto-Germanic phonemes above.
PIE /b/ > PGmc /p/ as in dewpaz 'deep', cf. Lith. dubùs 'deep', Gaul. Dubno-rix
PIE /bh/ > PGmc /b/ as inbōks 'tablet', cf. Gk phāgós 'oak', Lat. fāgus 'beech'
PIE /p/ > PGmc /f/ as in fōts 'foot', cf. Skt acc. pā́dam, Gk gen. podós
PIE /d/ > PGmc /t/ as in teuχanan 'pull', cf. OLat. doucō, Lat. dūcō 'I lead'
PIE /dh/ > PGmc /d/, as in durez 'door', cf. Gk thúrā, Lat. foris 'door'
PIE /t/ > PGmc /þ/ as in þrsúz 'dry', cf. Lat. torrus, OIr. tur 'dry'
PIE /g/ > PGmc /k/ as in krnam 'grain', cf. Lat. grānum 'grain', Lith. žìrnis 'pea'
PIE /gh/ > PGmc /g/ as in gardiz 'yard', cf. Gk χórtos, Lat. hortus 'garden'
PIE /k/ > PGmc /χ/ as in χrdiz 'wattle', cf. Gk kúrtos 'cage', Lat. crātis 'wicker-work'
PIE /gʷ/ > PGmc /kw/ as in kwemanan 'come', OHG kweman, cf. Gk baínō, Lat. venio
PIE /gʷh/ > PGmc /gw/ as in singwanan 'sing', Go. siggwan, cf. Gk omφḗ 'oracle'
PIE /kʷ/ > PGmc /hw/ as in χwat 'what', cf. Lat. quod 'what'
PIE /s/ > PGmc /s/ as in sunōn 'sun', cf. Lat. sōl 'sun'
PIE /s/ > PGmc /z/ as in méyzaz 'more', cf. Gk meízōn 'more'
PIE [m] > PGmc /m/ as in malanan 'grind', cf. Lat. molere
PIE [n] > PGmc /n/ as in naχts 'night', cf. Skt náktam, Gk gen. nuktós
PIE [w] > PGmc /w/ as in waganaz 'wagon', cf. Skt váhanam 'vehicle', Gk óχos 'wagon'
PIE [r] > PGmc /r/ as in reχtaz 'right, straight', cf. Lat. rectus, Gk orektós 'straight'
PIE [l] > PGmc /l/ as in PGmc langaz 'long', cf. Lat. longus, Gaulish longo- 'long'
PIE [y] > PGmc /y/ as in yēra 'year', cf. Av. yārə 'year', Lith. jéras 'yearling lamb'
PIE /i/ > PGmc /i/ as in witum 'we know', cf. Skt vidma, Gk ídmen 'we know'
PIE /e/ > PGmc /e/ as in PGmc erþō 'earth', cf. Gk érā 'earth', Welsh erw 'field'
PIE /ə/ > PGmc /a/ as in fadēr 'father', cf. Gk patḗr, Lat. pater 'father'
PIE /o/ > PGmc /a/ as in aχtō 'eight', cf. Gk oktṓ, Lat. octō 'eight'
PIE /u/ > PGmc /u/ as in PGmc ufar 'over', cf. Skt upári, Lat. super 'over'
PIE /ī/ > PGmc /ī/ as in swīnaz 'pig', cf. Lat. suīnus, OCS svins 'of pig'
PIE /ē/ > PGmc /ē/ as in sēþiz 'seed', cf. Lat. sēmen, OPruss. semen 'seed'
PIE /ā/ > PGmc /ō/ as in brōþar 'brother', cf. Skt bhrā́tā, Lat. frāter 'brother'
PIE /ō/ > PGmc /ō/ as in flōduz 'flood', cf. Gk plōtós 'floating'
PIE /ū/ > PGmc /ū/ as in fūlu 'foul', cf. Lith. pū́lias pl. 'pus', Skt pū́tis 'foul'
PIE /ey/ > PGmc /ei/ as in steig- 'climb', Go. -steigan, cf. Gk steíkhō 'climb'
PIE /oy/ > PGmc /ai/ as in staig- 'climbed', Go. -staig
PIE /eu/ > PGmc /eu-/ as in beud- 'order', cf. Go. -biudan 'order', cf. Gk peúthomai 'examine'
PIE /ou/ > PGmc /au/ as in baud- 'ordered', Go. -bauþ 'ordered'
When Jacob Grimm stated his rules, he also noted words in which they did not apply; these have come to be known as exceptions, and there are three sets. In the first set, PIE p t k remained unchanged, as in Gothic speiwan, cf. Lat. spuō 'spit'. In the second set, Proto-Germanic voiced stops corresponded to stops rather than to aspirates in Greek and Sanskrit, as in Gothic dauhtar, Skt duhitā́ 'daughter'. In the third set, medial fricatives were voiced rather than voiceless in Germanic words with Proto-Indo-European voiceless stops, as in the past tense forms Old English tēah 'he pulled', tugon 'they pulled', cf. Lat. dūcō 'I lead'; modern English has the first form, though without h, in the present tense of the golfing term tee off, and the second in the present tense form with g in tug. Grimm did not account for the different reflexes; later linguists did, over a half century after the publication of the second edition of his grammar in 1822. The explanations for the exceptions illustrate the progressively greater understanding of the relationships within the Indo-European family, and also the greater understanding of language in general.
The first set of exceptions was accounted for relatively soon after Grimm's publications. Knowledge of and attention to phonetics was increasing. Various linguists then noted that the unshifted stops stood after Germanic voiceless fricatives /f s χ/. Examples were readily found, among them:
Their position in clusters rather than as independent elements blocked the change.
The basis of the second set of exceptions was noted by the young Rudolf von Raumer in 1837; he pointed out that, since Sanskrit does not permit aspirates in two successive syllables, a form like bódhati 'observe' might well come from a root bhudh. But the implications for the Germanic forms were not pursued. Only with Hermann Grassmann's article of 1863 did it become clear that the problem was not in Germanic but rather in Sanskrit and Greek, an observation that might well have shaken the generally held position that Proto-Indo-European should be largely reconstructed on the basis of data in these two languages. The first aspirate generally lost its aspiration, but so might the second, as in the Greek word for daughter, thugátēr. As in the examples below, PIE /bh dh gh/ had become the voiceless aspirates /ph th kh/ in Greek. Examples are:
The evidence was so clear that the statement that one of two aspirates in successive syllables in Sanskrit and Greek lost its aspiration came to be known as Grassmann's law; when deaspirated, the aspirates show up as simple stops. As a result of Grassmann's demonstration, linguists came to understand that they had to note successive syllables and entire words, not merely their individual elements.
In the third group of exceptions, Proto-Indo-European medial voiceless stops and medial s became voiced rather than voiceless fricatives in Germanic; the voiced fricatives from stops then became voiced stops in the Germanic dialects, as in the forms below.
The relationship to the position of the accent is established beyond question by noting the consonant variation in strong verb paradigms, e.g. in preterite forms:
|1 sg.||zeh||táh 'accused'||didéśa < *-déika|
|1 pl.||zigum||tigon||didiśimá < *-dikimá|
Verner (1875) accounted for these exceptions as follows: the Proto-Germanic voiceless fricatives /f θ s χ/ became voiced /v ð z γ/ in voiced surroundings if the preceding syllable did not have the primary stress accent according to its Proto-Indo-European position. This formulation, known as Verner's Law, found immediate acceptance, and influenced the views of linguists in numerous ways. Possibly most important, it directed their attention at supra-segmentals. Through their many studies of stress and pitch, and of metrical patterns based on these, the journals give abundant evidence of the impact of Verner's article during the last quarter of the nineteenth century. Verner's article also gave assurance that the Proto-Indo-European accent was one of movable pitch, and that it survived remarkably long in Germanic as well as in Sanskrit and Greek. Germanic, accordingly, was demonstrated to be more conservative than the thorough-going consonant changes had seemed to indicate.
With the last of the exceptions accounted for, younger contemporaries of Verner, labeled neo-grammarians by their older colleagues, concluded that sound changes take place without exception in carefully defined environments. This view enabled its holders to proceed to the solution of many phonological problems, and has subsequently remained the nucleus of historical linguistic methodology.
The time frame of the Germanic consonant change has been a matter of great speculation. From observations of other comprehensive sound changes, such as the Great English Vowel Shift, it may be assumed that the Germanic change took place over the course of centuries. Because pertinent evidence from borrowings into and out of early Germanic is lacking, there is no external evidence for the date of the Germanic change, and its beginnings can only be tentatively stated as possibly at the end of the second millennium B.C., with its completion after the middle of the first millennium B.C. when comparable shifts of accent were under way in Italic and Celtic. This dating receives support from the long maintenance of vowels in final syllables, as in -gastiz of the Gallehus inscription, and in the complex development of the Indo-European labio-velars in Germanic.
The Indo-European labio-velars were long maintained as units in Proto-Germanic. Even as late as the 4th century A.D., Wulfila selected distinct symbols for the voiceless velar kʷ in Gothic, that is, the symbol for the numeral 6, which is transcribed |q| in our texts, and also for the aspirated velar hʷ, that is, the symbol for 700, which is transcribed in our texts with a ligature based on hʷ. But he did not represent [gʷ] with either of the two unused symbols, those for 90 and 900. In time, the three labio-velars developed into velars or into labial resonants or into vowels, even in Gothic. The unit phonemes must then have been gradually modified to clusters and thereupon reduced or even lost in the dialects, as illustrated in the following examples:
Because the voiced variant, as in PIE gʷīwos, Go. qius, ON kvikr, OE cwic(u) 'alive', and in PIE gʷen-, Go. qino, ON kona, OE cwene 'woman', and in other words was treated like PIE g, holders of the glottalic theory face the problem of accounting for a glottalized labio-velar.
As illustrated by these examples, the reflexes of the labio-velars vary in the Germanic dialects, so that they must be accounted for in the dialects rather than in Proto-Germanic. It may be assumed that they became clusters only late, probably even at the time of the Germanic consonant shift. In that shift, some were treated like Germanic reflexes of the Proto-Indo-European clusters /kw gw gʰw/, as examples of these illustrate.
Treatment also varies, as before vowels. PIE kʷ and gʷ became PGmc hw, kw before front vowels, but h, k before high back vowels, as in ON kona beside Go. qino, OE cwene 'woman', and in OE cucu beside cwic(u) 'alive'. In verbs as well, the forms may differ from dialect to dialect in accordance with the variant generalized, as illustrated by the principal parts of 'see':
|Old High German||sehan||sah||sāhum||gisewan|
The final treatments must be accounted for in the descriptions of the different dialects, such as the -g- of the Anglian preterite form, and the -w- of the West Saxon forms. For Proto-Germanic we assume first unit phonemes, then clusters parallel to those that developed from sequences of velars and w, and finally variants of these in differing contexts.
The Indo-European resonants /m n r l y w/ were maintained as consonants in Proto-Germanic when initial, as illustrated by the following examples:
Here, as in other locations, the reflexes maintain the allophonic variation of the resonants in Proto-Indo-European that was determined by Sievers and subsequently refined by Edgerton; the basis of it is referred to as the Sievers-Edgerton law. Briefly, the law states that the resonants were consonantal initially and medially before vowels, vocalic medially between consonants, and vocalic plus consonantal medially after long syllables, e.g. /y/ [y i iy], /w/ [w u uw], /r/ [r ŗ ŗr] and so on. The distribution may be illustrated with examples of /y/:
The distribution was disrupted already in late Proto-Indo-European when, under some circumstances, /ī ū/ developed from /i u/ as in swīnos 'pig'. It continued into the dialects until the Indo-European situation was completely modified. Among other modifications, initial [w] was lost in Greek, as in oîda = Go. wáit 'I know'; there were similar modifications in all of the dialects.
The disruption was gradual in Germanic, as illustrated by examples such as the following. Go. skadus, OE sceadu, OHG skato 'shadow' must be reconstructed as PGmc skadwas, cf. OE sceadwian, OHG skatewen 'to shadow'. The vowel of the final syllable was not lost until final vowels underwent reduction in late Proto-Germanic; [w] thereupon became [u]. Similarly, PGmc knyam > kunyam > kunyã developed to PGmc kuni as in Go. kuni, ON kyn, OE cynn, OS kunni, OHG chunni 'race'. PGmc -ã < -an was lost relatively late; its reflex was still written in the early Runic inscriptions, e.g. Gallehus horna. When after its loss as in *kunyã the preceding -y- became [i], the allophonic variation must still have been current. Further evidence for the consonantal allophone [y] is provided by geminated forms like the -nn- of chunni. [y] must have been maintained until the time of the West Germanic consonant lengthening, as in the nominative-accusative form PGmc kunyam before the loss of -ã.
Metrical patterns also provide evidence for long maintenance of the earlier allophonic variation. In Old English and in Old Icelandic verse, l m n after short syllables as in OE setl, fæðm, dagn are metrically non-significant, but after long syllables as in OE sūsl, bōsm, bēacn they are counted as separate syllables. We may ascribe the difference in metrics to the earlier allophonic treatment, as between PGmc [setlas] and [sūsļlas]; metrical conventions treating the two patterns differently were apparently established at the time of the variation and maintained after the different treatment was eliminated from the language. In late Old English verse the linguistic patterns outweighed the metrical, so that the old conventions were abandoned.
Long maintenance of the Proto-Indo-European allophonic variation of the resonants is also indicated by a specific development. In a number of words, /w y/ are represented as lengthened in West Germanic and as clusters of stop plus resonant in North and East Germanic, as illustrated by the following examples:
The PGmc forms are reconstructed with lengthened y and w. The process has been treated in many publications; it is referred to as Verschärfung, or as Holtzmann's Law in accordance with the name of an early contributor (Lehmann 1952:36-46). Numerous explanations were proposed in the 19th century, as by position of the accent, but then dismissed because they did not apply consistently. Smith (1941) accounted for the lengthening by the presence of Proto-Indo-European laryngeals.
The evidence for the assumption of laryngeals in Proto-Indo-European and the early dialects that is pertinent for explanation of Verschärfung may be summarized briefly. Proto-Indo-European roots are assumed to have had the structure: CeC, as in roots like PIE teg- 'cover'. Roots reconstructed with an initial consonant followed by a long vowel were then analyzed with a final laryngeal, such as PIE dheʔ rather than dhē 'place', deγ rather than dō 'give', (s)teχ rather than (s)tā 'stand'. Similarly, forms that had been reconstructed with an original long vowel were reconstructed with short vowel plus laryngeal, e.g. Greek ōón with variant in Sappho ṓion, Latin ōvum 'egg', from PIE oH- followed by the resonant /y/ in Sappho's form, by /w/ in Latin. Similarly, Greek phū́ō 'grow' from PIE bhew 'grow' is assumed to have had a laryngeal suffixed in the variant bhu-H-, which is reflected in the Sanskrit participial form bhūta- 'grown' < PIE bhu-H-tó-. Accordingly, forms in other dialects provide evidence for laryngeals that may have been maintained in Germanic.
Other forms from such roots have i as reflex of the laryngeal, e.g. Skt bhavitum and the past perfect participles of the three roots above, dhitá, hitá, sthitá. Such forms with -i- then provide evidence for positing laryngeals in earlier forms of Germanic and other dialects. Further evidence is found in Sanskrit verbs of the ninth class, such as skunā́ti 'cover' from the extended root (s)kw-n-eχ-, in which -n- is infixed between the root and the laryngeal suffix. Ninth class verbs give evidence for the a-coloring laryngeal -χ.
Bases with laryngeals have acute intonation in Lithuanian verbs, such as káuti 'strike'; cf. Lat. cūdō 'strike'. Such forms provide evidence for the earlier presence of laryngeals in roots and bases with a contiguous w or y.
The only written evidence for laryngeals is found in the Anatolian languages where the reflexes of some of the laryngeals are indicated as -h- and -hh- (cf. Lehmann 1952:22-35 et passim and Lindeman 1987 for further discussion and bibliography).
Germanic cognates of forms with [w y] adjacent to laryngeals are attested with lengthened w and y in the West Germanic dialects. In Gothic and North Germanic, ww developed to ggw; yy developed to ddj in Gothic and to ggj in North Germanic. While these developments are of concern in treatment of these dialects rather than in a description of Proto-Germanic, examples of the late Proto-Germanic situation will be given here with cognates in a non-Germanic dialect that indicate evidence for a Proto-Indo-European laryngeal in contact with a resonant.
Among examples with lengthened -ww- < -wχ- are the Old Saxon form beuuod 'harvest' and ON bygg, dat. sg. byggue 'barley'. Cognates indicating the earlier presence of a laryngeal are Skt bhūtá- and Gk éphūn from PIE *bhew- 'grow' followed by a laryngeal.
Similarly, Lat. dēfrūtum 'cider' from PIE bhru- plus laryngeal indicates that OE brēowan 'brew', ON brugginn 'brewed' must be reconstructed with -ww- from -w- plus laryngeal.
OS hreuuan, ON hryggva 'rue' with -ww- in Proto-Germanic have the Sanskrit cognate kravís 'raw meat', in which -i- developed from the laryngeal in PIE kr-ew-X-.
Go. skuggwa 'mirror', ON skuggi, OE scuwo, OHG skuwo 'shadow' have as cognate Skt skunā́ti 'cover', in which the laryngeal is reflected in the long vowel ā of the suffix.
OE cēowan, ON tyggia 'chew' from PIE kew-X- have as cognate Lith. žiáunos 'jawbone', in which the laryngeal is indicated by the acute intonation in Lithuanian.
Lengthened -yy- developed further to Gothic -ddj- and North Germanic -ggy-. In contrast with -ww- from -wX-, -yy- developed from both -Xy- and -yX-. An example with lengthened -y- from a preceding laryngeal is Go. daddjan, OSwed. deggia, OHG tāju 'suckle', with cognates illustrating the root dhē in the Sanskrit aorist ádhāt, Gk thḗsato 'sucked' and the sequence dhʔ-y- in the Sanskrit past perfect participle dhī́ta. An example with -yy- from -y- lengthened by a following laryngeal is ON Frigg, the name of Odin's wife, and OS frī, OHG friia 'free' from PIE pr-y-X- as in Sanskrit prīnā́ti 'is pleased'.
Other examples may be found in the handbooks. Some of them have been disputed, as well as the hypothesis that laryngeals survived into Germanic and brought about the lengthening with further development to stops, though the evidence adduced is often slight. But for the items exemplified here, and many others, there is no question of the earlier presence of laryngeals.
While -ww- resulted from the combination of -w- plus a laryngeal, in the sequence of laryngeal plus w either -k- or -g- resulted. An example is ON nǫkkvi, OE naca, OHG nacho 'boat' which is cognate with Latin nāvis. The situation is complex because Germanic forms with -u- or -w- beside these are also attested, e.g. ON nōr 'ship', naust 'boat-house' and OHG ver-nawun 'boats that carry wood'. The difference in development may be ascribed to the treatment of the sequence -eXw- in Proto-Indo-European. As illustrated by the nominative forms Skt náus, Gk naûs and the accusative forms Skt nā́vam, Gk nêa, the laryngeal was lost in the nominative but maintained after -e- in the accusative, and also in the genitive as illustrated by Skt nāvás, Gk nēós, in which the laryngeal was maintained until it resulted in lengthening of the preceding vowel.
Among other examples are ON kuikr, OE cuic, OHG queh 'alive' with cognates illustrating the laryngeal, such as Skt jīvá, Lat. vīvus 'alive' < PIE gʷyXw-os. Similarly, ON spic, OE spic, OHG speck 'bacon' with cognates in Skt pī́van, Gk pī́on 'fat' < PIE spyXw-on. Also OE tācor, OHG zeichur 'brother-in-law-', cf. Skt devŗ́, Gk dāḗr < PIE daXw-. For other examples, see Austin (1946) and Lehmann (1952:37-42).
A number of examples in the Germanic dialects also give evidence of -g- < -Xw-:
In each of these, the sequence of -ug- resulted from a reduced vowel followed by a laryngeal and w. The reason for the difference from sequences resulting in -k- is unclear. Austin ascribed it to absence of chief accent on the preceding vowel; I proposed derivation from different laryngeals. The forms illustrate that, like those resulting from Verschärfung, combinations of laryngeal with w and y resulted in reflexes other than resonants in late Proto-Germanic.
By late Proto-Germanic, /i/ and /u/, which earlier were allophones of /y w/, had become separate phonemes. The frequency of /u/ was considerably increased by the addition of [u] from vocalic allophones of /l m n r/, but the short vowel system as a whole was symmetrical in consisting of four members:
PGmc /e/ and /a/ were reflexes of vocalic phonemes in Proto-Indo-European; /e/ was a reflex of PIE /e/, PGmc /a/ of late PIE /a o/. When PGmc /i/ and /u/ became phonemes, a fourfold system developed. But in spite of its symmetry in outline, the system was askew because of asymmetries in distribution. With its origin in /y/, PGmc /i/ was almost entirely restricted to syllables with weak stress; PGmc /e/ on the other hand was characteristically found in syllables with primary stress. Moreover, /i/ was found primarily before obstruents, whereas /e/ occurred before both resonants and obstruents. The two vowels were therefore in great part distributed complementarily. It has been suggested that they were allophones of one phoneme, and that a short vowel system of three members should be assumed for late Proto-Germanic (Twaddell 1948). Yet both occurred before /a u/ of following syllables, as in the near minimal pairs:
The assumption of a three vowel system is therefore untenable.
On the other hand, since /e/ and /i/ were virtually in complementary distribution, their coalescence would not have been unlikely. It may have occurred in the pre-Gothic vocalic system; in most words PGmc /e/ > /i/, as in itan 'eat', though not before /h r/ and finally, e.g. waír 'man'. PGmc /e/ was also maintained in the reduplicating syllables of Class 7 strong verbs in Gothic, e.g. laílōt 'let'. Interjections were unaffected, so that /e/ remained in waíla and /i/ in hiri 'come here'.
Redistribution is also notable in other dialects, as in:
On the other hand, PIE /e/ > PGmc /i/ before PGmc /i j/ regardless of the rest of the environment, as in:
In keeping with this change, PIE /ey/ > PGmc /ei/ > /ii, ī/ as in:
PGmc /e/ also became /i/ before nasal and consonant, as in:
The change took place sufficiently early so that, upon loss of the nasal, the root of the verb was associated with strong verbs of the first class, as in:
In syllables without primary stress, /e/ > /i/, as in:
Through such changes, the frequency of /i/ was greatly increased.
In the back area there was only one high vowel /u/, with two allophones, [u] and [o]; [u] stood before nasals, as in:
Elsewhere [o] stood before short and long /a/, [u] before other vowels and finally, e.g.
The phonological conditioning gave rise to subsequent variation, as in ON guð 'god' but goda-hus 'temple'. With such variation there was disruption of the original patterns, so that we also find ON goð beside guð and also OE god, OHG got; similarly, we find Go. fugls, ON fugl, OE fugol, OHG fugal, but also ON fogl, OHG fogal 'bird'. As in the high front area, there was a redistribution of [u o] in Gothic, with [o] (au) standing before /h r/ and finally. In the other dialects, /o/ came to be phonemic in a short vowel system of five members.
While the early long vowel system also consisted of four members, the articulation of the two lower vowels differed from those of the short vowels in being relatively low. In words with Latin ō, the vowel was raised to ū rather than lowered to ā, as in Go. Rumoneis, corresponding to Lat. Rōmānī, and OHG Rūma, ON Rūma-borg to Lat. Rōma. PGmc bōkān- 'beech' was represented by Caesar as (silva) Bacēnis 'Beechwood'. But at the end of the Proto-Germanic period the system was expanded with ā. As one source, [a] was lengthened on the loss of [n] before /χ/, as in the infinitive Go. fāhan, ON fā, OHG fāhan 'seize' beside the preterite ON fingom, OE feng, OHG fieng. The loss was late, because the older situation is still attested in the Burgundian name Hanhualdus; and nasalization in Icelandic was still apparent to the First Grammarian, cir. 1200 A.D. (Haugen 1950:33-34). Latin borrowings provided a second source as in OSwed. strāta, OHG strāza 'street' from (via) strāta, and OE pāl, OHG pfāl 'pole', from Latin pālas. /ō/ then was raised, as indicated by developments in Old High German and in Slavic borrowings; beside Go. Dōnawi 'Danube', OHG attests the form Tuonouwa, and the Germanic form is taken over in Slavic as Dūnāvi.
The early Proto-Germanic phonological system, like that of Proto-Indo-European, included no diphthongs. If vowels had formed diphthongal units with resonants, we would expect characteristic developments for them. Yet until late Proto-Germanic, vowels undergo the same developments before resonants as elsewhere; PIE /a o ə/ > PGmc /a/ before /y/ and /w/ as well as in other environments. Similarly, PIE /e/ > PGmc /e/ before /w/ as well as before obstruents.
When, however, the Proto-Germanic allophonic variation of resonants began to be disrupted, the lower short vowels came to have distinct allophones before resonants. As noted above, [e] > [i] before [y] and thereupon coalesced with it to yield /ī/. Before /w/ it became the first element of a diphthong, as did /a/ before /y/ and /w/, leading to the three diphthongs:
Without an /ey/ diphthong, the system was asymmetrical. The gap was filled for a time by the reflex of a number of sequences that may be represented as /ēy/ but soon was modified, on the loss of /y/, to a vowel represented in grammars as ē² (long e two). Its most frequent source is the sequence of /ĕ/ plus laryngeal reflex plus /i/. For example, Go. fēra 'side, part' and OHG fiara 'here' are derived from PIE sphēy- (Pokorny 1959:983-984). The largest number of examples occurs in preterites of verbs of the seventh class with -ai- plus consonant as the vowel of the present, and also -al/an- and -ē-. As noted at greater length in the treatment of these verbs in the section on morphology, the origin of the PGmc -ēy- in the preterite is from PIE -eXy-; as one example OHG sciet, preterite of sceidan, OE scēd, preterite of scādan 'separate' from PIE skeXy- plus dental, cf. Lith. skíedžu 'separate'.
Parallelism was lost in the early dialects as the span of the diphthongs was decreased, a process that Prokosch (1939:105-107) related to the strong primary stress. While discussion of these changes is a topic for grammars of the dialects, several examples may be noted here:
As exemplified above with OE hēr and cognates, the hypothetical diphthong /ey/ became monophthongal /ē/. Borrowings from Latin with /ē/ adopted this long ē rather than its earlier counterpart, as exemplified in Go. mēs, OHG mias, meas 'table' < Late Latin mēsa < mensa. The close articulation of both the new Germanic and the Latin vowels is indicated in the treatment of the Latin vowel in syllables with weaker stress in Germanic, as in OE eced, OS ecid < Latin acētum 'vinegar'. Moreover, with the addition of the new close ē, the inherited ē was lowered in many words such as OHG Saat 'seed'. The long vowels then were parallel to the short vowels, each set having five members.
The supra-segmentals may be determined by several criteria, especially by their use in verse in the dialects. Early verse is alliterative; metrical units consist of individual lines, like that of the Gallehus inscription. Lines are based on four highlighted syllables, two or three of which are further marked by agreement of their initial elements. Two, or perhaps one, are so marked in the first half-line, agreeing in alliteration with the major stressed segment in the second half-line. Examples may be cited from recurrent patterns, such as the following from the Old English Beowulf.
371 Hrōþgār maþelode, helm Scyldinga
Hrothgar answered, ruler of Scyldings
957 Bēowulf maþelode, bearn Ecgþēowes
Beowulf addressed him born son of Ecgtheow
All four stressed syllables may also alliterate, often with twofold alliteration, as in line 941 of the Old Saxon Heliand. As the alliterating syllables indicate, strong stresses typically fall on nominal elements in the clause. Followed as they are by less stressed syllables, a falling intonation may be assumed for declarative sentences as in line 941 of the Heliand:
sô mikilu is he betara than ic. Nis thes bodon gimaco
By so much is he better than I. There is no messenger with his ability.
From similar patterns in the contemporary languages, it may be assumed that the same intonation patterns applied already in earlier stages. English bétter, German bésser suggest an accent pattern of / ˊ ˇ /. The comparable intonation patterns of entire clauses may be similarly assumed, as indicated by the accents and the numerals following the sentences, with # indicating a final drop in pitch.
Hè is much bétter than Ì. 2 3 1 #
Èr ist viel bésser als ìch. 2 3 1 #
Further examples below illustrate more fully the intonation pattern of falling stress at the ends of lines.
The intonation pattern is also indicated by the bunching together of alphabetic symbols, as in line 1291a of the Monacensis manuscript of the Old Saxon Heliand, in which I indicate the weaker words by italicizing them:
satimthuoendisuigoda sat by himself then and was silent
The absence of spaces between words suggests a continuous pattern. The accent marks on thúo and suígoda in the Palatina manuscript, however, indicate position of accents, supporting the 2 3 1 # pattern; it may then be assumed for late Proto-Germanic after the accent change from pitch to stress.
Conclusions may also be drawn from the development of consonants and vowels. The development of the consonant differentiation in OHG uuas : uuāri and similar forms that will be discussed below in the morphological section gives us information about the position of the accent in early Proto-Germanic; in the pre-form of uuāri the pitch accent fell on the second syllable. And the incidence of umlaut on syllables with primary stress though not on final syllables also indicates the distribution of the accent in the early dialects.
Cognates in other dialects also point to the patterns of accentuation in early Proto-Germanic with reflexes in the dialects, e.g. Go. fidwor, ON fjōrir, OE fēower, and Homeric Greek písures, Attic Greek téttares, 'four'.
Examination of poetic lines provides information of the accentuation after stress replaced pitch. In Beowulf 1231, four principal stresses fall on the initial syllables of words:
drúncne dríhtguman, dṓð swā ic bídde
the drunken warriors do as I ask
A relatively important stress must also be assumed on the second syllable of drihtguman on the basis of Beowulf 99:
Swā ðā dríhtgùman, drḗamum lífdon,
So then the warriors lived in joy,
The assumption of a strong stress on the second elements of compounds is supported by comparison of the treatment of consonants by Verner's law. Gothic nauði-bandi 'fetter' and nauði-þaurfts 'need' have a voiced dental fricative, in contrast with a voiceless in forms of the verb nauþjan* 'force'. The voiced fricative may be accounted for by the accent of adjectival (determinative) compounds on their second element that was still maintained when Verner's law was in effect.
Three stresses may therefore be assumed for late Proto-Germanic, as well as breaks at the ends of sentences, and between words with those stresses. The earlier pattern of breaks after syllables is still apparent in the early Runic inscriptions, as noted above for the Gallehus inscription. And as in the Germanic languages today, pitch was maintained in conjunction with stress; normally the strongest stress was accompanied by the highest pitch, with pitch falling to ends of clauses and sentences.
As a result of the strong stress on words, final elements were lost; three such losses may be assumed for Proto-Germanic.
/n/ was lost in weakly stressed final syllables, as in the accusative singular horna of the Gallehus inscription. Initially the loss may have resulted in a nasalized vowel, for final vowels without -n were lost in the early inscriptions, as in Runic un-nam < *-nama. But the nasalized final a was also lost by the time of the individual dialects, as in the accusatives Go. dag, ON dag, OE dæg, OHG tag. Final n was maintained, however, in words with short vowels that could receive primary stress, such as Go. þan, OE þan 'then'.
Final /t d/ were lost after short vowels not under primary stress and after all long vowels, as in Gothic 3 sg. pres. wili, OE wile, OHG wili 'he wishes' < *welit; cf. Lat. velit. Also Go. undarō 'beneath' < *-ōþ; cf. Skt adharāt 'from below'. Comparison with the third plural preterite forms Go. bērun, OE bǣron 'they bore' < -nt indicates that -þ- < -t was lost later than -n, for -n was maintained in the third plural ending. Yet the change was early, for remaining long vowels like the -ō in Gothic were treated like other long finals. Final /t d/ were however maintained after short stressed vowels, as in ON þat, OE ðæt, OHG daz, cf. Skt tad.
Final open short vowels, PGmc e a, were lost except under primary stress, as illustrated by the first and third singular preterite of Go. wait, ON veit, OE wāt, OHG weiz 'I know, he knows', cf. Gk oîda, oîde, Skt véda.
Still other reductions and losses took place in the individual dialects.
The morphonology of Proto-Germanic vowels is best understood by reviewing the situation in Proto-Indo-European, which has been more closely maintained in Germanic than in any of the other dialects.
The basic situation is indicated in the vowels. It is assumed that the basic vowel of roots was /e/. In Proto-Indo-European one exchange was brought about by the change of /e/ to /o/, as in the root leg-, with reflexes like those of Gk légō 'read' and lógos 'word'. For this and other vocalic changes, Jacob Grimm introduced the term Ablaut 'sound away from'; it is often referred to by the translated term apophony. The term is used both for the original change and for its results in the dialects. The various vowels are referred to as grades (a translation of the German Stufe) in the ablaut system, e.g. e-grade, o-grade, etc.
Attempts have been made to account for this change, including explanation by position of the accent, by association with other elements like the resonants, and so on. None have been widely accepted. The change occurred so early, possibly even in late Pre-Indo-European, that the cause may be totally obscure. What is important, especially for Germanic, is the association of /e/ with reflexes of active forms of verbs and /o/ with reflexes of stative forms, as noted more fully below in the treatment of the verb system.
A later set of changes took place when Proto-Indo-European had a stress accent. At this time the basic vowel /e/ was lost when it was unstressed; this situation is referred to as zero grade. Reflexes of the loss in the root *sed- are found NE nest, Lat. nīdus from PIE ní-zd-os 'in (which a bird) sits' and Go. asts, NHG Ast 'branch' from PIE ó-zd-os 'on (which a bird) sits'. Complementary to zero grade is lengthened grade, in which the vowel is either ē or ō. The basis for the lengthening is unclear, although attempts have been made to ascribe it to loss of a vowel in the following syllable as a result of the accent on the lengthened vowel. It is found in specific forms, such as the nominative singular of monosyllables, e.g. Skt vāk, Latin vōx in contrast with Gk épos < PIE wékʷos 'word', or of nouns of a certain structure, such as Skt pitā́, Gk patḗr 'father'.
Sound changes, losses, and analogical changes between Proto-Indo-European and the dialects have obscured further the bases of the original ablaut situation. Yet enough ablaut relationships are evident, especially in Germanic, that knowledge of ablaut is highly important for understanding the phonological relationships between related forms like Go. sitan, 3 sg. pret. sat, and the causative ga-satjan 'set, place', OE nest, Go. 3 pl. pret. sētun, and OE sōt 'soot' as well as those among many other such related forms. The reflexes of PIE sed- given here illustrate the five grades: normal grade, deflected or o-grade, zero grade, lengthened-ē grade, and lengthened ō-grade.
Understanding of ablaut relationships was long unclear because laryngeals and their reflexes were not recognized. Even when Saussure (1877) posited the vanished consonants that came to be called laryngeals, their reflexes were not properly understood. Yet a difference was recognized between vowels involved in ablaut, such as those reflected in sētun and sōt, and those called "original long vowels," as in the widely evident roots PIE dhē- 'place' and dō- 'give' through their different morphological uses. But even after Kurylowicz (1927) determined that some of the forms with original long vowels correspond to Hittite cognates with vowel plus h, such as pa-ah-sa-an-zi, Latin pāscō 'protect', it took some time before the interrelationships among such reflexes was understood.
It is now clear that the structure of the "original long vowels" in Proto-Indo-European dhē-, dō-, stā-, etc. before the laryngeals were lost or combined with other phonological elements was comparable to that of the vowels and final consonants of roots like PIE sed 'sit', sew 'rain', sey 'bind'. The roots with such long vowels were then reconstructed with laryngeals; instead of PIE dhē-, the root was given as dheʔ- 'place', instead of dō-, as deγ- 'give', instead of (s)tā-, as (s)teχ- 'stand'. This clarification also led to understanding of the use of these roots with normal grade in the present tense forms of Greek as títhēmi, dídōmi, hístāmi, as well as their distribution in other dialects including Germanic.
Reflexes of Proto-Indo-European morphonology in Germanic may be exemplified by examining the verb system of strong verbs. Seven classes have been traditionally proposed, with the infinitive, preterite singular, preterite plural and past participle given as principal parts. These indicate the specific ablaut grades that were generalized in forms of the system that distinguishes a present tense, a preterite tense, and a past participle as well as indicative and subjunctive forms, and an imperative as well as distinctions of number and person.
It is further necessary to note the background of the distinction between the present and the preterite forms. In Proto-Indo-European the etymon of the present system of the first five classes indicated action, while the etymon of the preterite indicated state. As PIE dialects became accusative languages, the aspectual meanings were replaced by tense. But maintenance of the aspectual meanings in early Proto-Germanic led to the distinction between the first five classes and the sixth as well as the seventh. As Prokosch recognized (1939:150-151, 173-182), even before Active languages were understood, many roots and bases of the verbs of the sixth and seventh classes indicated state; their basic forms were accordingly taken as preterites in Proto-Germanic rather than as presents, for which a form with another grade was introduced.
In addition to this distinction into two groups by their meaning, the first five classes fall into two groups in accordance with their form. Those of the first three classes have bases with the vocalism e + a resonant (y, w, m, n, l, r) + a consonant, while those of the fourth and fifth have bases with e followed by a single consonant. An example from the fourth and fifth classes illustrates the ablaut grades in the principal parts. The forms will be more fully exemplified and treated in the section on strong verbs in the morphology.
|Infin. (present)||Pret. Sg.||Pret. Pl.||Past Ptc.|
The reflexes indicate that the normal grade of the root was maintained in the infinitive and the present tense in late Proto-Germanic. The a reflecting PIE o-grade marked the preterite singular. The preterite plural and optative, on the other hand, are based on the long-ē grade, for zero grade would have led to an impossible cluster. To escape this in the past participle, the e-grade of the present was introduced.
An example of the first class illustrates that the e-grade and the o-grade mark the present and preterite singular, as also in the second and third class of the strong verbs; the two other forms have zero grade, reflecting the position of the accent on the final syllable in Proto-Germanic.
An example of a sixth class verb illustrates that the base form with PIE -eH- > -ā- > PGmc -ō- was used throughout the preterite, and that an alternate form was introduced for the present and past participle:
Morphophonemic change is also found with consonants, as in the preterite of strong verbs. Before the accent was fixed, it fell on the root in the singular, but on the ending in the plural and optative. The second singular preterite in the West Germanic languages also shows the effect of the accent on the ending as in the plural. In keeping with the position of the accent, verbs have an f/v, s/z > r, þ/d, h/g interchange; the basis of the interchange may be illustrated by comparing the forms of the PIE base gew-s- in the Sanskrit perfect of its reflex joṣ- 'enjoy' and the Old English preterite of cēosan 'choose'.
|Sanskrit perfect||Old English preterite|
The interchange may be assumed for Proto-Germanic before the initial stress accent on the root syllable was introduced; its effects were maintained into the dialects.
Examples of the preterite third singular and third plural forms of verbs with the four fricatives may illustrate the morphophonemic variation, as well as analogical regularization in some forms which are indicated below in italics.
|Go.||þarf : þaúrbum||-laiþ : -liþom||kaus* : -kusun||-tauh : tauhun|
|ON||þarf : þurfom||leiþ : liþon||kaus : køron||----- : toginn|
|OE||þearf : þurfon||lāð : lidon||kōs : kurun||tēah : tugon|
|OS||tharf : thurbun||lēth : lidun||kōs : kurun||tēh : tugun|
|OHG||darf : durfun||leid : litum||kōs : kurum||zōh : zugun|
Further examples will be given in the treatment of the verb.
The same variation is found between forms of the causative and the simple verb, for the causative had its accent after the stem, as illustrated by Skt vártate 'turn' and vartáyate 'cause to turn'; cf. Go. fra-waírþan 'spoil' and fra-wardjan 'cause to spoil'; OHG ginesan 'be saved' and nerian 'save'.
Similarly, the variation is found in nominal forms, as in the Gothic adjective alþeis 'old' in contrast with the noun alds 'age'; taíhun 'ten' in contrast with -tigjus 'decade'. This led to the further distinction in modern English ten based on the Old English noun tīen and the suffix for the decade -tig > -ty.
The loss of nasals with compensatory lengthening led to a different type of change in OE bringan : brōhte, OHG bringan : brāhta 'bring, brought' where the loss of the nasal before the voiceless fricative in the preterite led to the lengthening of the preceding vowel.
The changes presented here illustrate the increasingly compact structure of words as opposed to distinct syllables in late Proto-Germanic, as is evident in the Runic name Hlewagastiz versus Old Norse Hliugast. They also demonstrate the effect of the fixation of dynamic accent on initial and root syllables as opposed to the earlier pitch accent, leading to weakening and loss of final elements in the dialects. Still other changes are introduced from dialect to dialect, leading to diverse modifications of their phonological and morphological systems.
As the sections above indicate, the phonological system of Germanic was highly conservative. The threefold relationship among the obstruents was maintained, while in Indic it was expanded to four, e.g. t th d dh, but reduced to two in Hittite, e.g. t d, or modified in other ways, as in Latin, e.g. t d f. The articulation of the threefold elements was however modified in Germanic, so that in the nineteenth century and later a large-scale change has been assumed. The assumption was based on phonetic criteria. From a phonemic point of view the obstruent system was maintained in early Germanic in contrast with the modifications in Indic, Hittite, Latin and other dialects. Similarly, the resonant and the vocalic systems were largely maintained for some time, and phonological developments like Gothic ddj, ggw indicate long retention of laryngeals. Only later, when pitch accent was replaced by stress, did extensive modifications take place in Germanic. At that time, some members of the fricative set were voiced, leading to a considerably different system.
The stress accent, located as it was on the base — often the initial syllable — had a strong effect on the vowels of syllables with weaker stress. Some of these were lost, as will be noted in the chapter on inflectional morphology; others were eventually reduced to schwa. In the course of time, the phonological structure of late Proto-Germanic and especially its dialects differed greatly from that of Proto-Indo-European.