Etymological Dictionary of Proto-Germanic Online

Note on the Structure of the Entries

The entries in this dictionary are composed according to the following model. Every entry starts with a Proto-Germanic reconstruction. The entries are ordered on the basis of this reconstruction. The alphabetization ignores vowel length, which means that long and short vowels (e.g. *e and *ē) can be found in random order. The thorn (þ) is placed between the letters t and u. After the head word, the grammatical category is given together with a reconstruction of the Proto-Germanic meaning. Since there is no methodology for semantic reconstruction, these proto-meanings are not necessarily factual, and are merely to be taken as an indication of the author’s intuition.

After the head word, the attestations in the Germanic languages are given. On the selection of these forms, see the preface. If feasible, the attestations are followed by a Pre-Germanic proto-form based on the Germanic attestations. The distribution of this etymological construct is given between brackets; its distribution may be purely Germanic (GM), non-Indo-European (NIE), Germanic/Balto-Slavic (NEUR), Germanic/Italo-Celtic (WEUR), European, i.e. with cognates in Greek or Armenian (EUR) or Indo-European (IE), i.e. when the etymology is based at least partly on cognates in the remaining languages. Obvious loanwords are recognizable by the abbreviation (LW). The extra-Germanic comparanda are then given after the long hyphen.

After the presentation of the material, the reader will usually find a discussion of the etymology of the word, drawing in phonetic problems or alternative etymologies. This is also where any possible internal Germanic derivations are given.

Introduction

1 The Proto-Indo-European and Proto-Germanic phonemes

1.1 Proto-Indo-European

In the present book, I have made use of the phonological system as envisioned by Beekes 1995 for all reconstructed Proto-Indo-European forms:

Proto-Indo-European had three series of stops, whose articulation in many ways continues to be debated (cf. Kümmel 2012). I have made use of the mainstream division into plain stops, voiced stops and voiced aspirated (breathy-voiced?) stops, but if the voiced PIE stops were actually pre-glottalized, as is assumed within the Glottalic Theory, the feature of aspiration becomes abundant.

Unlike Proto-Germanic, which typically had the stress on the first root syllable of the word, Proto-Indo-European had a free tonal accent: the accent could occur on the root, the suffix or the ending, and within nominal and verbal paradigms often shifted from one syllable to another. Stress-bearing elements were the vowels, the semivowels, the resonants and the laryngeals. The latter three types of phonemes could only receive the stress when they were in vocalic position, i.e. functioned as syllable-building elements. Since the vocalization rules of these phonemes in the individual Indo-European daughter languages are often incompatible with each other, they must post-date the parent language. PIE *h₂mb^hi adv. ‘around, about’, for instance, is realized as *h͓₂m̥b^hi in PGm. *umbi and Skt. abhí, but as *h̥₂mb^hi in Gr. ἀμφί. Similarly, the Proto-Indo-European word *h₃b^hruH- ‘eyebrow; bridge’ has no less than three different vocalizations, i.e. *h͓₃b^hruH- as in PGm. *brū- and Skt. bhrū́-, *h̥₃b^hruH- as in Gr. ὀφρῦς, and *h͓₃b^hr̥u̯H̥- as in ToB pärwā-ne (du.). The latter example is especially informative, because it proves that the vocalization of the *r was triggered by the vocalization of the laryngeal in Tocharian, just as the vocalization of initial *h₂ in Greek conditioned the non-vocalization of the *m in Gr. ἀμφί. It seems evident, in other words, that the vocalization of both the laryngeals and the resonants was phonologized in the individual daughter languages. I have therefore refrained as much as possible from indicating vocalization (i.e. h̥₁, h̥₂, h̥₃, m̥, l̥, n̥, r̥, as well as i̯, u̯) in PIE reconstructions, as this would inevitably lead to erroneous proto-forms, and have only used them to indicate vocalizations in proto-forms underlying specific IE dialects.

1.2 Proto-Germanic

Proto-Germanic phonology differs significantly from Proto-Indo-European. It acquired a number of new vowels, both short and long, and the stress was retracted to the vowel of the first root syllable of a word. Due to a shift of the Indo-European stops, Proto-Germanic also acquired a large amount of new fricatives, both voiced and voiceless. The phonemes reconstructed as *b, *d, *g in this dictionary also at least partly appear as *ƀ, *đ, *ǥ in the Germanic dialects. For instance, most languages have plosives word-initially, but *g emerges as a fricative in this position in both Saxon and Franconian. Since the distribution surfacing in the individual languages is divergent, this alternation is likely to have been subphonemic in Proto-Germanic. Another important innovation is the rise of phonemic consonant length. Due to a range of phonetically regular sound changes, Proto-Germanic acquired a geminated variant of practically any existing consonant, and this is perhaps one of the the most far-reaching phonological changes that the language went through. The resulting phoneme inventory can be summed up as follows:

2 From Proto-Indo-European to Proto-Germanic

2.1 The vowels

I will here discuss the most important sources of the Proto-Germanic vowels. As mentioned above, one of the striking innovations is that Proto-Germanic acquired a lot of new vowels, both short and long, that did not yet exist in Proto-Indo-European. These vowels developed from a number of dissimilar sources, mostly combinations of vowels plus laryngeals (HV, VH), vocalized laryngeals (H̥) and vocalized resonants (R̥).

2.1.1 Short vowels

2.1.1.1 PGm. *a

PGm. *a arose out of a merger of Pre-Gm. *o and *a. These vowels must nevertheless have remained distinct until after Verner’s law, as PGm. *gw from both PIE *k^w (*k´u̯) and *g^wh (*ǵ^hu̯) was delabialized before *j in roots with *a from old *o (see §2.2.4.). Pre-Gm. *o continues PIE *o, *Ho and *h₃e.

• *aita- m. ‘ulcer, pus’: OHG eiz m. ‘abscess, boil’ < *h₂oid-o-, cf. OCS jadъ m. ‘poison’, SCr. ȉjed m. ‘gall, poison, anger’
• *amban- m. ‘belly’: OS ambon (m.pl.), cf. Lat. umbō ‘boss (of a shield); protuberance’ < *h₃emb^h-on-
• *fadi- m. ‘lord’: Go. bruþ-faþs m. ‘bridegroom’ < *pot-í-, cf. Gr. πόσις, Lith. pàts m. ‘husband’

The direct sources of PGm. *a are PIE *H̥ and *h₂e:

• *aba adv., prep. ‘from; off’: Go., ON af, OE of, OHG ab < PIE *h₂ep-ó, cf. Gr. ἀπό, ἄπο adv., prep. ‘far away, away from’
• *bakan- s.v. ‘to bake’: ON baka, OE bacan, OHG bahhan < *b^hh̥₃g-e-, cf. Gr. φώγω ‘to roast’ < *b^héh₃g-e-
• *fader- m. ‘father’: Go. fadar, ON faðir, OE fæder, OFri. feder, OS fadar < *ph₂-tér-, cf. Skt. pitár-, Gr. πατήρ, gen. πατρός, Lat. pater, -tris, OIr. athir, athair m. ‘father’

An important issue concerns roots with *a < *H̥ that start with a resonant. It is widely assumed that the resonant rather than the laryngeal is vocalized in such roots (cf. Schaffner 2001; Müller 2007), but this is in disagreement with the facts (cf. Beekes 1988):

• *laka- adj. ‘weak’: ON lakr < *lh̥₂g-o-, cf. Gr. λαγαρóς adj. ‘weak’
• *lakjan- w.v. ‘to seize’: OE læccian, cf. Gr. λάζομαι ‘id.’ < *lh̥₂g-ie-
• *lata- adj. ‘lax, sluggish’: Go. lats, ON latr, OE læt, OS lat < *lh̥₁d-o-, cf. Gr. ληδεῖν ‘to be slow’ < *leh₁d-
• *magra- adj. ‘slim’: ON magr, OE mæger, OHG mager adj. ‘id.’ < *mh̥₂k´-ró-, cf. Gr. μακρóς adj. ‘long’
• *mahan- m. ‘poppy’: OHG maho m. ‘id.’ < *mh̥́₂k-on-, cf. Gr. μήκων ‘id.’ < *méh₂k-on-
• *natja- n. ‘net’: Go. nati, ON, OE net, OS netti, OHG nezzi < *nH̥d-io-, cf. Lat. nōdus m. ‘node’ < *noHd-o-; OIr. nascaid ‘to bind’ < *nHd-ske-

There are additional cases that seem to indicate that the vocalization of the laryngeals as *a does not change when the following obstruent is a consonantal resonant. A difficulty with the given examples is that their probative force is nullified by Dybo’s law, i.e. the regular shortening of pretonic long vowels before resonants. As a consequence of this law (for which see §2.1.2), the vocalization of *H to *a is technically unfalsifiable before resonants, because the *a can always have developed out of unstressed *eh_2/3, *oH or *ō in this position. Nonetheless, the vocalization *H before resonants can be ascertained on the basis of formations in which it is morphologically unlikely that the root had a full grade, such as, for instance, the PIE no-participles.

• *hanan- m. ‘rooster’: ON hani, OE hana, OHG hano m. ‘id.’ < *k´h̥₂n-on-, cf. Gr. ἠι-κανός ‘id.’ (< “morning singer”)
• *harōjan- w.v. ‘to sharpen’: MDu. haren w.v. ‘to sharpen’, derived from an adjective *k´h̥₃-ro- or * k´oh₃-ró-, cf. Arm. sur adj. ‘sharp’
• *namōn- n. ‘name’: Go. namo, ON nafn, OE nama, OHG namo ‘id.’ < *h₃nh̥₃-mén- (less likely *h₃neh₃-mén-), cf. Gr. ὄνομα ‘name’
• *nawi- m. ‘corpse’: Go. naus, ON nár < *nh̥₂u-i-, cf. Latv. nâve f. ‘death’ < *neh₂u-ieh₂-
• *wana- adj. ‘lacking’: Go. wans, ON vanr adj. ‘id.’ < *h₁uh̥₂-nó-, cf. Skt. ūná- ‘id.’ (vs. Lat. vānus adj. ‘empty, void’ < *h₁uéh₂-no-)

PGm. *a in a small number of words continues what looks like PIE *a, but many of these words have a limited European distribution, and it therefore seems unwarranted to project them back into the parent language (Lubotsky 1989). In reality, they are likely to continue Wanderwörter or were borrowed from now extinct European languages by the individual dialects after they had acquired *a by the vocalization of the laryngeals. This taken into account, very few potential instances of PIE *a remain, especially in comparison to the overwhelming evidence for the vowels *e or *o. This alone makes it highly unlikely that Proto-Indo-European had *a as a phoneme. But the prophecy is essentially self-fulfilling: anyone who accepts *a for Proto-Indo-European will start seeing it everywhere:

• *baunō- f. ‘bean’: ON baun, OE bēan, OS, OHG bōna < *b^hau-neh₂-, cf. Lat. faba f. ‘id.’, OPru. babo ‘id.’, OCS bobъ m. ‘id.’ < *b^hab^h-o/eh₂-
• *gait- f. ‘goat’: Go. gaits, ON geit, OE gāt, OS gēt, OHG geiz < *g^haid-, cf. Lat. haedus m. ‘young goat-buck, kid’ < *g^haid-o-
• *hafra- m. ‘billy goat, buck’: ON hafr, OE hæfer < *káp-ro-, cf. Gr. κάπρος m. ‘(wild) boar’, Lat. caper m. ‘he-goat, buck’ < *kap-ro-, OIr. gabor, W gafr ‘id.’ < *gabro-
• *hanipa- m. ‘hemp’: ON hanpr, OE hænep, henep, OHG hanaf, hanif < *kanib-, cf. Gr. κάνναβις f. ‘id.’ < *kannabi-, Ru. konopljá f. ‘id.’
• *paidō- f. ‘coat, shirt’: Go. paida, OE pād, OS pēda, OHG pfeit < *bait-éh₂-, cf. Thrac. βαίτη f. ‘coat made of pelt’

2.1.1.2 PGm. *e

The main sources of PGm. *e are PIE *e or *h₁e:

• *beran- s.v. ‘to carry’: Go. bairan, ON bera, OE beran, OFri. bera, OS, OHG beran < *b^hér-e-, cf. Skt. bhárati, Gr. φέρω, Lat. ferō, ferre, OIr. beirid ‘id.’
• *etan- s.v. ‘to eat’: Go. itan, ON eta, OE, OS etan, OHG ezzan < *h₁ed-, cf. Hitt. ez(za)zi, Gr. ἔδω, Lat. ēdō, ēsse ‘id.’

2.1.1.3 PGm. *i

PGm. *i developed directly out of PIE *i. It is sometimes assumed that it merged with *e in Proto-Germanic if the following syllable contained a, but this cannot be the case: unlike *e, *i is never affected by a-breaking in Old Norse, only by a-mutation.

• *fiska- m. ‘fish’: Go. fisks, ON fiskr, OE, OHG fisc m. ‘id.’ < *pisk-o-, cf. Lat. piscis < *pisk-i-, OIr. íasc < *peisko-
• *likkōn- w.v. ‘to lick’: OE liccian, OS likkon, OHG leckōn ‘id.’ < *liǵ^h-neh₂-, cf. Lat. lingō, -ere ‘id.’

A second common source for PGm *i is PIE *e: this vowel was raised before tautosyllabic nasals. This raising must have taken place relatively late, as it post-dated the assimilation of *-nu̯- to *-nn- (§2.2.5.6) as well as the assimilation of *-zm- to *-mm- (§2.2.5.4), which itself was posterior to Verner’s law. It is further evident that it was later than the merger of *o with *a, as *o would otherwise have been similarly raised to *u.

• *finþan- s.v. ‘to find’: Go. finþan, ON finna, OE fīðan, OHG findan < *pént-e-, cf. Skt. pánthās, gen. pathás m. ‘road, path, course’ < *pont-H-
• *bindan- s.v. ‘to bind’: Go., OE, OHG bindan, ON binda < *b^hénd^h-e-, cf. Skt. badhnā́ti ‘id.’ < *b^hn̥d^h-néh₂-

2.1.1.4 PGm. *u

The source for PGm. *u is twofold: it developed from PIE *u and from a PIE vocalized resonant (R̥ < uR). For examples of the former:

• *kula- n. ‘coal’: ON kol, OE col, OHG kol < *ǵul-o-, cf. OIr. gúal < *ǵoul-o-
• *kustu- m. ‘choice’: Go. kustus, ON kostr, OHG kust < *ǵus-tu-, cf. Lat. gustus ‘taste’

The second important source for PGm. *u consists of resonants that were vocalized between consonants:

• *fulla- ‘full’: Go. fulls, ON fullr, OE ful, OHG fol adj. ‘id.’ < *pl̥h₁-nó-
• *kwumþi- ‘arrival, coming’: Go. ga-qum(f)þs, ON -kund, OHG qhumft, chumft, chunft ‘id.’ < *g^wḿ̥-ti-, cf. Lat. con-ventio ‘convention’
• *tunþ- m. ‘tooth’: Go. -tunþus m. ‘id.’ < *h₃dn̥t-, cf. Lat. dēns < *h₃dn̥t-
• *uns pron. ‘us’: Go. uns, ON oss, OE ūs, OHG uns ‘id.’ < *n̥s, cf. Lat. nōs < *nōs

It is important to realize that resonants were also vocalized before a laryngeal that was later lost. It is consequently incorrect to state that laryngeals are irrelevant for the reconstruction of Proto-Germanic:

• *suma- pron. ‘some’: Go. sums, ON sumr, OE, OHG sum < *sm̥H-o-, cf. Gr. ἁμο- ‘someone’
• *ufuma- comp. ‘highest’: Go. auhuma < *up-m̥h₂-o-, cf. Skt. upamá-, YAv. upəma- supl. ‘id.’
• *fulan- m. ‘foal’: ON foli, OE fola, OHG volo < *pl̥H-on-, cf. Gr. πῶλος < *pōlH-o-
• *kuru- adj. ‘heavy’: Go. kaurus < *g^wr̥H-u-, cf. Lat. grāvis
• *furai adv. ‘before’: Go. faura, OE fore, OHG fora, fura < *pr̥h₃-oi, cf. Gr. πάρος, Skt. puráḥ ‘id.’ < *pr̥h₃-os

2.1.2 Dybo’s law

Another development involved in the creation of the PGm. short vowels is Dybo’s law. It was suggested by Dybo 1961 that in Celtic, Italic and Germanic long vowels were shortened pretonically. It is incorrect, however, to speak about this law as one single development. While in Italo-Celtic any long vowel seems to have been affected, the Germanic shortening applied only to long vowels before resonants. In this position, Pre-Gm. *ā, *ō, *ē, *ī (≮ PIE *ei) and *ū were shortened to *a, *e, *i and *u respectively:

• *alīnō- f. ‘elbow’: Go. aleina, ON ǫln, OE eln, OHG elena, elna, cf. Gr. ὠλένη < *h₃eHl-én-eh₂-
• *delō- f. ‘tit’: OE delu, OHG tila, cf. Gr. ϑηλή f. ‘breast’ < *d^heh₁-l-éh₂-
• *glana- n. ‘shine’: ON glan < *g^hloh₃-nó-, cf. *glōan- s.v. ‘to glow’: OE glōwan, OS glōian, OHG gluoan < *g^hlóh₃-e-
• *hula- adj. ‘hollow’: ON holr, OE, OFri., OHG hol < *k´uH-ló-, cf. Lat. cavus adj. ‘id.’ < *k´ouH-o-; Skt. śū́na- adj. ‘lack, absence’ < *k´uH-no-
• *stura- adj. ‘big’: OSw., Elfd. stur < *sth₂u-ró-, cf. Skt. sthūrá- adj. ‘big, strong, thick, massy’
• *sunu- m. ‘son’: Go. sunus, ON sunr, sonr, OE, OS sunu, OHG sun(u), cf. Skt. sūnú-, Lith. sūnùs, OCS synъ m. ‘id.’ < *suH-nú-
• *wira- m. ‘man’: Go. wair, ON verr, OE, OS, OHG wer, cf. Skt. vīrá-, Lith. výras, Lat. vir m. ‘id.’ < *uiH-ró-

In Pre-Germanic accentually mobile words, Dybo’s law may have given rise to paradigmatic length alternations. It is conceivable that the difference between OHG dūmo ‘thumb’ < *þūman- and OSw. þumi ‘id.’ < *þuman- arose in a paradigm *túH-mōn, gen. *tuH-mén-os, yielding PGm. *þūmō, *þumenaz. Since Dybo’s law affected the PGm. word for ‘egg’, whose *a through *o developed from originally long *ō, it is likely to have post-dated the change *ōu > *ō , for which see §2.1.5.

• *ajja- n. ‘egg’: Go. Crim. ada (n.pl.), ON egg, OE ǣg, OS, OHG ei < *oi̯ó- < *ōi̯ó- < *h₂ōu­ió-, cf. Gr. ᾠόν, Lat. ōvum, OCS aice n. ‘id.’

2.1.3 Long vowels

2.1.3.1 PGm. *ē

PGm. *ē deloped out of PIE *ē and *eh₁ and is usually reconstructed phonetically as [ǣ]. It yielded a close vowel ē in Gothic and Old Frisian, ǣ in Old English, and ā in Old Norse, Old Saxon and Old High German.

• *sēdi- f. ‘seed’: Go. seþs, ON sáð, OE sǣd, OHG sāt < *seh₁-tí-, cf. Go. saian, Lith. sė́ti ‘to sow’ < *séh₁-e-
• *nēþlō- f. ‘needle’: Go. neþla, ON nál, OE nǣþl, OHG nādala < *néh₁-tl-eh₂-, cf. *nēan- s.v. ‘to sow’: OHG nā(w)an < *neh₁-
• *kwēni- f. ‘wife, woman’: Go. qens < *g^wēn-i-, cf. Gr. γυνή f. ‘id.’ < *g^w(o)n-eh₂-

The question whether OE ǣ and OFri. ē (the latter with raising) reflects PGm. *ǣ directly or developed secondarily out of Proto-North-West Germanic *ā is difficult to answer. Latin loanwords such as OE strǣt ‘street’ < strāta are in support of the latter scenario. It has been argued that ā was substituted by ǣ in these cases because Anglo-Frisian did not have an ā at the time of borrowing, but there are additional indications that Anglo-Frisian raised older *ā. A small but relatively old, i.e. at any rate Proto-Northwest-Germanic group of n-stems displays ā/a-ablaut in the root, cf. OHG krācho m. ‘crook’ < *krākan- vs. ON kraki m. ‘id.’ < *krakan-. Since this type of ablaut was introduced analogically on the basis of n-stems with regular ī/i-ablaut, cf. OHG rīdo, dat. riten m. ‘fever’ < *hrīþō, dat. *hrideni < *kréit-ōn, *krit-én-i (Schaffner 2001: 549-51), it is likely that PNWGm. had *ā rather than *ǣ, since it thus would be more susceptible to the introduction of a pure length alternation. This *ā must then have been raised to *ǣ in Anglo-Frisian, and further to *ē in Proto-Frisian, as is supported by NFri. (Wiedingharde) krēk m. ‘hook on clothes’ (Kroonen 2011: 332).

2.1.3.2 PGm. *ō

PGm. *ō is the result of a merger of Pre-Gmc. *ō and *ā from PIE *ō, *eh₃, *oh_1/2/3 and *eh₂. Compare the following examples:

• *fōt- m. ‘foot’: Go. fotus, ON fótr, OE, OFri., OS fōt, OHG fuoz < *pōd-, cf. Gr. πούς, ποδός m. ‘id.’ < *pod-, Lat. pēs m. ‘id.’ < *pē̆d-
• *ga-nōga- adj. ‘enough’: Go. ganohs, ON (g)nógr, OE genōh, OHG ginuog < *kom-h₂nōk´-o-, cf. Skt. ā́naṭ aor. ‘reached’ < *h₁e-h₂nek´-t
• *dōma- m. ‘decision, verdict’: Go. doms, ON dómr, OE dōm, OHG tuom < *d^hoh₁-mo-, cf. Gr. ϑωή ‘punishment’ < *d^hoh₁-eh₂-
• *sōkjan- w.v. ‘to search’: Go. sokjan, ON sœkja, OE sǣcan, OHG suohhen, cf. Lat. sāgīre < *seh₂ǵ-ie-
• *mōder- f. ‘mother’: ON móðir, OE mōdor, OFri. mōder, OS mōdar, OHG muoter, cf. *meh₂-tér-, cf. Skt. mātár-, Gr. μήτηρ, Lat. māter f. ‘id.’

A third source for PGm. *ō is PGm. *ōu. The details of the underlying sound change are given in §2.1.5.

2.1.3.3 PGm. *ī

Two sources are available for PGm. *ī, i.e. *iH and *ei:

• *swīna- n. ‘pig’: Go. swein, ON svín, OE, OHG swīn < *suH- ‘sow’ + the suffix *-īna-, cf. Go. gait-ein n. ‘little goat’
• *stīgan- ‘to ascend’: Go. steigan, ON stíga, OE, OHG stīgan < *stéig^h-e-, cf. Gr. στείχω ‘to go, step’

2.1.3.4 PGm. *ū

Unlike PGm *ī, which partly developed out of *e + *i, the only regular precursor of PGm. *ū is PIE *uH, as the sequence *eu remains PGm. *eu.

• *mūs- f. ‘mouse’: ON mús, OE, OHG mūs < *muHs-, cf. Gr. μῦς, Skt. mū́ṣ-
• *sū- f. ‘sow’: ON sýr, OE, OHG sū < *suH-, cf. Gr. ὗς, Lat. sūs

Long *ū probably also arose secondarily, i.e. in analogy to the change of PIE *ei to PGm. *ī. For this development, see §2.2.5.2.

2.1.4 Diphthongs

Proto-Germanic had four diphthongs: *ai, *au, *eu and *ia. The vocalic elements of these diphthongs have the same origins as their corresponding short vowels, and can be traced back to the Indo-European proto-language accordingly. Likewise, the off-glides *i and *u go back to PIE *i and *u.

2.1.4.1 PGm. *ai

• *aida- m. ‘pyre, glow’: OE ād, OHG eit < *h₂eid^h-o-, cf. Gr. αἶϑος ‘fire’, Skt. édha- ‘firewood’
• *snaiwa- m. ‘snow’: Go. snaiws, ON snær, OE snāw, OHG snēo < *snoig^wh­o-, cf. OCS sněgъ m. ‘id.’

2.1.4.2 PGm. *au

• *auke conj. ‘then again, too’: Go. auk, ON auk, ok, OE ēac, OHG auh < *h₂eu-ǵe, cf. Gr. αὖ, αὖ-γε ‘id.’
• *rauda- adj. ‘red’: Go. rauþs, ON rauðr, OE rēad, OHG rōt < *h₁roud^h-o-, cf. Gr. ἐρυϑρός adj. ‘id.’ < *h₁rud^h-ro-

2.1.4.3 PGm. *eu

• *eudra- n. ‘udder’: ON júgr, OFri. jāder < *h₁euHd^h-r-, cf. Gr. οὖϑαρ < *h₁ouHd^h-r̥
• *keusan- s.v. ‘to try, choose’: Go. kiusan, ON kjósa, OE cēosan, OHG kiosan < *ǵéus-e-, cf. Gr. γεύομαι ‘to taste’

2.1.4.4 PGm. *ia

What is here reconstructed as *ia is traditionally referred to as so-called *ē². As opposed to *ē¹ (< PIE *ē, *eh₁), this second *ē has close reflexes throughout the Germanic dialects, viz. Go. e, ON é, OE ǣ, OHG ē, ie, ia, and is therefore generally assumed to have been a close-mid vowel [ē] in Proto-Germanic. It is especially frequent in Vulgar Latin loanwords:

• *bē²tōn- f. ‘beetroot’: OE bēte, OHG bieza, cf. It. bieta
• *brē²fa- m./n. ‘letter’: ON bréf, OHG briaf, cf. Lat. brevis
• *krē²ka- m. ‘Greek’: Go. Kreks, OHG Kriach, cf. Lat. Graecus
• *mē²sa- ‘table’: Go. mes, OE mēse, OHG mias, cf. Lat. mēnsa
• *rē²man- m. ‘oar’: OHG riemo, cf. Lat. rēmus
• *tē²gula- ‘tile’: OHG ziagal, cf. Lat. tēgula

It has been claimed that *ē² developed out of a PIE long diphthong *ēi such as, for instance, in *hē²r ‘here’ < *k´ēir (Streitberg 1896: §79; Prokosch 1939: 104). This development is not entirely inconceivable, although in view of the parallel change of PGm. *ōu to *ō (see §2.1.5) I would rather expect *ē¹ to be the outcome. In any case, a lengthened grade would be unexpected in the word ‘here’, since it is not attested anywhere else in the Indo-European language family. I therefore follow Kortlandt 2006, who suggested that *ē² at least in the case of *hē²r must be analyzed as deriving from *ia, *hi-ar consisting of the root *hi- < PIE *k´i- ‘this’ (cf. Lith. šis ‘this’) plus a locative suffix *-ar abstracted from *þar ‘there’ < *tor and *hwar ‘where’ < *k^wor. This *ia obviously merged with the diphthong *ea that is found in the reduplicated preterites of the class 7 strong verbs, cf. OHG erien ‘to plow’ < *arjan-, pret. iar, ier < *e-ar-, whence it spread to other originally reduplicating verbs. It further seems probable that the Gothic i-stem gen.pl. ending -e, which clearly spread to the other nominal stem classes (Vendryes 1927), developed from PIE *-ei-om (Kortlandt 2006) through an intermediate stage *-ea, i.e. *ē². On the basis of this evidence, I have decided to reconstruct *ē² as *ia throughout the dictionary, also in forms whose derivation or etymology is unclear, but it is not inconceivable, for example, that OHG sciari and ziari were formed by the addition of the adjectival r-suffix to the roots *skī̆- ‘to shine’ and *tī̆- ‘id.’:

• *skiari- adj. ‘bright’: OHG sciari < *skh₁i-or-i-, cf. Go. skeinan s.v. ‘to shine’ < *skīnan-
• *tiari- adj. ‘brilliant’: OHG ziari < *diH-or-i-, cf. Skt. dīdā́ya 3sg.perf. ‘shines’ < *diH-doiH-e

In West Germanic, additional cases of secondary PGm. *ē² arose by the occasional loss of *z after *i in some cases. The evidence for this loss is patchy, and the phonetic conditioning of the loss remains unclear. Perhaps it occurred only after i and before dentals.

• *liznōn- w.v. ‘to learn’: OE leornian, OFri. lirna, lerna, OS līnon, OHG lernōn < *lis-neh₂-
• mizdō- f. ‘reward’: Go. mizdo, OE mēd, meord, OFri. mēde, OS mēda, OHG miata
• *waizda- n. ‘woad’: OE wād, OFri. wēde, OS wēd, OHG weit

2.1.5 Osthoff’s law

Unlike the Indo-Iranian languages, the European branches of the Indo-European family, including Greek, Italo-Celtic, Balto-Slavic and Germanic, did not have primary long diphthongs, i.e. *ē or *ō followed by a sonorant. In order to explain this difference, it has been claimed that every long vowel that stood before a sonorant followed by another consonant was shortened in Proto-Greek (Osthoff 1884: 84-5). With the help of this vowel shortening, which later became known as Osthoff’s law, the difference between e.g. Skt. Dyáus and Gr. Ζεύς < *diēus and aorists such as e.g. Gr. ἔδειξα and Av. dāiš < *h₁e-dēik´-s-t ‘showed’ can satisfactorily be explained (cf. Beekes 1995: 68; 235-6). Osthoff’s law is now generally accepted for Greek and Italo-Celtic (cf. Ringe 2006: 75), but in Germanic the situation is actually fairly complicated. Unambiguous evidence for *ēi > *ei > *ī and *ēu > *eu is lacking, and we have to rely on long diphthongs with resonants as their off-glide:

• *fersnō- f. ‘heel’: Go. fairzna, OHG fersana, cf. Gr. πτέρνη, Skt. pā́rṣṇi- < *tpērs-n-
• *mimza- n. ‘meat’: Go. mimz, cf. Skt. māṃsá- < *mēms-o-
• *winda- m. ‘wind’: Go. winds, ON vindr, OE, OFri., OS wind, OHG wint < *h₂uéh₁-ent-o-, cf. Hitt. ḫuu̯ant- c. ‘id.’ < *h₂uh₁-ent-, Lat. ventus m. ‘id.’ < *h₂ueh₁-(e)nt-o-, Skt. vā́ta- m. ‘id.’ < *h₂ueh₁-nt-o-

An important issue is the outcome of PGm. *-ōu- that arose from both Pre-Gm. *-āu- < PIE *-eh₂u- and *-ōu- < PIE *-ōu-, *-eh₃u-, or *-oh_2/3u-. This long diphthong was affected by Osthoff’s law in just three cases, and only in closed syllables (i.e. before two consonants) or word-finally.

• *gōman- ~ *gauman- m. ‘gum, palate’: ON gómi, gómr, OE gōma, OHG guomo, gaumo < *gōmō, gen. *gaumnaz < *ǵ^heh₂-u-mōn, gen. *ǵ^heh₂-u-mn-os-, cf. Lith. gomurỹs m. ‘palate’, Latv. gãmurs m. ‘larynx, trachea’ < *ǵ^heh₂-mr-
• *nausta- n. ‘boathouse, boatshed’: ON naust < *neh₂u-sth₂-o-, cf. ON nór m. ‘id.’, Skt. náu-, Gr. ναῦς, Lat. nāvis < *neh₂u-
• *ahtau num. ‘eight’: Go. ahtau, ON átta, OE eahta, OFri. achta, OS, OHG ahto < *h₃ek´t-eh₃u, cf. Skt. aṣṭā́, aṣṭáu, Gr. ὀκτώ, Lat. octō ‘id.’

In open syllables, the diphthong was not shortened at all, but rather lost its labial glide (cf. Mahlow 1879: 29-34; Schmidt 1983; Streitberg 1892: 29-37). The material thus demonstrates that the intervocalic loss of laryngeals was coupled with compensatory lengthening at least before u. There is a large corpus of evidence for this change, of which I have given a selection here:

• *bōan- s.v. ‘to live, dwell’: Go. bauan s./w.v. ‘id.’ < *b^héh₂u-, cf. Skt. bhávati ‘to become, happen, come about’, Gr. φύομαι ‘to grow, arise, spring up, become’ < *b^héuh₂-e-, Lith. bū́ti, OCS byti ‘to be’ < *b^huh₂- (with laryngeal metathesis)
• *dōida- ptc. ‘vexed’: Go. af-dauidai (m.pl.) < *d^hoh₂u-i-to-, cf. OCS daviti ‘to suffocate’, Lith. dovýti ‘to make tired’ < *d^hoh₂u-éie-
• *fōr, gen. *funenaz n. ‘fire’: Go. for, gen. funins < *péh₂‐ur, *ph₂‐un‐ós, cf. Hitt. paḫḫur, gen. paḫḫuenaš n. ‘id.’ < *péh₂‐ur, *ph₂‐uén‐(o)s
• *lōma- m. ‘betrayal’: Icel. lómur < *loh₁u-mo-, cf. Go. lewjan, OE lǣwan w.v. ‘to betray’
• *sōel- n. ‘sun’: Go. sauil, ON sól < *séh₂-uel (gen. *sh₂-un-ós), cf. Gr. ἥλιος, Dor. ἀέλιος m. ‘id.’, Lat. sōl, sōlis m. ‘id.’, Lith. sáulė f. ‘id.’
• *stōra- adj. ‘big’: ON stórr, OE, OFri. stōr < *stéh₂u-ro-, cf. Skt. sthūrá- adj. ‘big, strong, thick, massy’ < *sth₂u-ró-, Skt. sthávira- adj. ‘broad, thick’ < *steuh₂-ro- (with laryngeal metathesis)

The sound law also seems to have been at work in the Gothic 1du. verbal ending -os (cf. Schmidt 1883: 11-13; Streitberg 1896: 322), whose derivation is often considered to be problematic (cf. Boutkan 1995: 319-20). In view of the Skt. thematic ending -āvas, attempts have been made to derive this -os from *-o-ues, but the intervocalic *u̯ would never have been lost in this position. I therefore reconstruct the ending as *-o-h₁u-es, with the element *-h₁u̯- as in Skt. āvám du. ‘we two’ < *n̥-h₁u-om (which no doubt developed from *-du̯- ‘two’). The resulting *-ōwiz then regularly lost its labial glide in Proto-Germanic times, and through *-ōiz developed into Go. -os.

The material presented here is of some importance because it proves that Osthoff’s law must have been posterior to the specifically Germanic loss of the labial glide. This implies that the law cannot have been identical to the parallel shortening of long diphthongs in e.g. Italo-Celtic and Greek, and must have taken place at a late stage within Germanic itself. In other words, there was no such thing as a common West-Indo-European innovation that can be brought under one umbrella. It is therefore better to consider the shortening of long diphthongs a linguistically trivial sound change that took place independently in the different Indo-European dialects at different moments in time. For more on the position of the change *-ōu- > *-ō- in Proto-Germanic relative chronology, see §2.2.5.7.

2.2 The consonants

The Germanic consonant system differs considerably from its Indo-European counterpart. One of the earliest changes in the Proto-Germanic consonant inventory was its centumization, i.e. the depalatalization of the Proto-Indo-European palatovelars *k´, *ǵ, *ǵ^h and the subsequent merger of *ku̯, *gu̯, *g^hu̯ with the labiovelars *k^w, *g^w, *g^wh. This development also occurred in other branches of the Indo-European family, e.g. Italo-Celtic and Tocharian. The most important, exclusively Germanic innovations are 1) the structural modification of the three series of stops known as the first and second Germanic sound shifts, and 2) the rise of consonantal length. Both developments are phonetically and chronologically complex, involving several different sound changes in often debated orders and interpretations, and can only be interpreted in a meaningful way by keeping track of the changes in the dynamics of the system as a whole. This is, in short, an overview of the most important changes that took place between Proto-Indo-European and Proto-Germanic.

2.2.1 Grimm’s law

Unlike the Germanic vowels, which do not radically differ from the vocalic elements in related languages such as Italo-Celtic or Balto-Slavic, the Germanic consonantism has evolved in an entirely different direction. This Lautstand has become one of the most striking features of the Germanic branch, and forms a major contrast with its closest relatives. It is, in other words, what to a large extent defines Germanic. The relationship between the Proto-Indo-European and the Proto-Germanic consonant inventories has been clarified by the discovery of Grimm’s law, which in the traditional view first lenited both voiceless *p, *t, *k, *k^w to *f, *þ, *h, *hw:

• *faiha- adj. ‘colored, colorful’: OE fāh, OS, OHG fēh < *póik´-o-, cf. OCS pьstrъ adj. ‘varicolored’ < *pik´-ro-
• *hamfa- adj. ‘maimed’: Go. hamfs, OS hāf, OHG hamf < *kómp-o-, cf. Lith. kum̃pas adj. ‘curved’ < *kmp-o-
• *hwaþera- pron. ‘who of two?’: Go. ƕaþar, ON hvárr, OE hwæðer, OS hwethar, OHG wedar, hwedar < *k^wó-ter-o-, cf. Skt. katará-, Gr. πότερος ‘which of two’

A consecutive stage consisted of the devoicing of the originally voiced stops *b, *d, *g, *g^w to PGm. *p, *t, *k, *kw:

• *inkwan- m. ‘lump’: Icel. ökkr, ökkvi m. ‘lump; hillock’, MDu. enke, inke m. ‘small wound’ < *eng^w-on-, cf. Gr. ἀδήν, -ένος f./m. ‘gland’, Lat. inguen, -inis n. ‘swelling on the groin; groin’ < *ng^w-en-
• *knewa- n. ‘knee’: Go. kniu, ON kné, OE cnēo(w), OS knio, OHG kneo < *ǵn-eu-, cf. Skt. jā́nu- n. ‘id.’, Gr. γόνυ n. ‘knee; joint of plants’ < *ǵonu-
• *paidō- f. ‘coat, shirt’: Go. paida, OE pād, OS pēda, OHG pfeit, cf. Thrac. βαίτη f. ‘coat made of pelt’
• *turhta- adj. ‘bright’: OE torht, OS toroht, OHG zoraht < *drk´-to-, cf. OAv. -dərəšta- adj. ‘seen, visible’

It is further assumed that lenition also turned the PIE voiced aspirates *b^h, *d^h, *g^h into the voiced fricatives PGm. *ƀ, *đ, *ǥ. The fricatives often surface as plosives, especially word-initially and after n.

• *banda- n. ‘band, bond’: ON band, OE beand, OFri. bend, OS band < *b^hond^h-o-, cf. YAv. baṇda- m. ‘bond, fetter’
• *berga- m./n. ‘mountain’: ON bjarg, berg, OE beorg, OFri. berch, OS, OHG berg < *b^herǵ^h-o, cf. Hitt. parku-, Arm. barjr adj. ‘high’ < *b^hrǵ^h-u-

Accordingly, PIE *g^wh (as well as *ǵ^hu̯) developed into PGm. *ǥw. This phoneme is not attested as such in the actual languages, except directly after a nasal, where it was a plosive:

• *lingwa- n. ‘heather’: ON lyng, OSw. liung < *leng^wh-o-, cf. OCS lǫgъ m. ‘meadow, underbrush’, Ru. lug m. ‘meadow’ < *long^wh-o-
• *sangwa- m. ‘song’: Go. saggws, ON sǫngr, OE, OS sang, OFri. song < *song^wh-o-, cf. Gr. ὀμφή f. ‘divine voice, oracle, emblem’ < *song^wh-eh₂-

PGm. *gw was delabialized under certain circumstances, especially initally before *u and *r:

• *guda- n. ‘god’: Go. guþ, ON guð, OE, OFri., OS god, OHG got < * g^whu-tó-, cf. OCS gověti ‘to revere’ < *g^whou-eh₁-
• *gunþī- ~ *gunþjō- f. ‘wound’: ON gunnr, guðr, OE gūð, OS gūdea < *g^whń-tih₂-, cf. Hitt. ku̯enzi ~ kunanzi ‘to kill, slay, ruin’ < 3sg. *g^whén-ti, 3pl. *g^whn-énti
• *grindan- s.v. ‘to grind’: OE grindan < *g^whrénHd^h-e-, cf. Lat. frendō, -ere ‘to grind one’s teeth’ < *g^whrénHd^h-e-

The default outcome of PGm. *gw seems to have been *w, however:

• *aiwiskja- n. ‘shame, disgrace’: Go. aiwiski, OE ǣwisc < *h₂eig^wh-isk-, cf. Skt. an-ehás- adj. ‘flawless’ < *n̥-h₂eig^wh-os-
• *neura/ōn- n./m. ‘kidney’: ON nýra, OHG nioro < *neg^wh-r-on-, cf. Lat. nefronēs m.pl. ‘kidneys, testicles’ < *neg^wh-r-on-
• *snaiwa- m. ‘snow’: Go. snaiws, ON snær, OE snāw, OS snēo, OHG snē(o) < *snoig^wh-o-, cf. OCS sněgъ, Lith. sniẽgas, Latv. snìegs m. ‘id.’
• *wambō- f. ‘womb, belly’: Go. wamba, ON vǫmb, OE wamb, OFri. wamme, OHG wamba < *g^whomb^h-eh₂-, cf. Skt. gabhá- m. ‘vagina’ < *g^whm̥b^h-o-
• *warma- adj. ‘warm’: ON varmr, OE wearm, OFri., OS, OHG warm < *g^whor-mo-, cf. Gr. ϑερμός adj. ‘id.’ < *g^wher-mo-

There is a small number of potentially convincing examples with PGm. *b as the outcome of Pre-Gm. *g^wh (Seebold 1980). The examples are too few to establish a phonetic conditioning, however, and since all instances with *b except *bedjan- have alternative etymologies, whereas the ones with *w have not, it seems best to suspend the implementation of this change until further notice.

• *banjō- f. ‘wound’: Go. banja, ON ben, OE benu, cf. OAv. bąnaiiən 3pl.inj. ‘to make ill, afflict’ < *b^hon-eie- or Hitt. ku̯enzi ~ kunanzi ‘to kill, slay, ruin’ < 3sg. *g^whén-ti, 3pl. *g^whn-énti.
• *bedjan- s.v. ‘to ask, pray’: Go. bidjan, ON biðja, OE biddan, OFri. bidda, OS biddian, OHG bitten, cf. Gr. ποϑέω ‘to desire, long for, miss’, OIr. guidid ‘to pray’ < *g^whod^h-éie-
• *beran- m. ‘bear’: OE bera, OHG bero, cf. Lith. bė́ras, Latv. bẽ̦rs adj. ‘brown’ < *bhērH-o- or Gr. ϑήρ, Lith. žvėrìs, OCS zvěrь ‘wild animal’ < PIE *ǵ^huēr- (Ringe 2006: 106)

2.2.2 Verner’s law

Verner’s law is the law that accounts for the ultimate merger of PIE *p, *t, *k, *k^w and *b^h, *d^h, *g^h, *g^wh into PGm. *b, *d, *g, *gw in non-initial, unaccentuated syllables. Proto-Indo-European was an accentually mobile language. Somewhere in Proto-Germanic, i.e. after Grimm’s law but before the stress was fixed to the root, Verner’s law caused voicing of *f, *þ, *h, *hw and * s everywhere but word-initially and directly after a stressed syllable, thus merging the former four of these fricatives with *ƀ, *đ, *ǥ, *ǥw.[1]

• *ahiz- n. ‘ear’: OE ēar, æhher, eher, OHG ahar, ehir < *h₂ék´-es-, cf. Lat. acus, gen. aceris n. ‘husk, chaff’
• *fader- m. ‘father’: Go. fadar, ON faðir, OE fæder, OFri. feder, OS fadar, OHG fatar < *ph̥₂-tér-, cf. Skt. pitár-, Gr. πατήρ, Lat. pater m. ‘id.’
• *hweula- n. ‘wheel’: ON hjól, MDu. wiel < *k^we-k^wl-ó-, cf. Skt. cakrá- ‘id.’
• *magra- adj. ‘slim’: ON magr, OE mæger, OHG magar < *mh̥₂k´-ró-, cf. Gr. μακρός adj. ‘long’, Lat. macer adj. ‘thin, lean’
• *uberi adv., prep. ‘above, over’: OHG ubar, G über < *h₁upéri, cf. Skt. upári adv. ‘above, over, upwards’

In a number of cases, Verner’s law also operated word-initially. It is generally assumed that this happened because those words predominantly occurred in clitic position and therefore had no stress.

• *ga(n)- perf. pref.: Go. ga-, OE, OFri. ge-, OS, OHG gi- < *kom-, cf. Lat. con-, com-
• *bi prep., adv. ‘by’: Go. bi, OE, OFri., OHG bī̆ < *h₁pi, cf. Gr. ἔπι, Skt. ápi adv. ‘on, at, by’ < *h₁epi

Verner’s law more often operated regardless of morpheme boundaries. Compare, for instance, the two following doublets consisting of an archaic Verner variant beside a restored form without it:

• *mati-sahsa- ~ *mati-zahsa- n. ‘knife’: OHG mezzisahs ~ mezzirahs, (a compound of PGm. *mati- ‘food’ and *sahsa- ‘knife’)
• *ga-fesjō- ~ *ga-besjō- f. ‘chaff’: OHG ga-vissa ~ ga-bissa, cf. OHG fesa f. ‘fiber’ < *fesō-

With the help of Verner’s law, the original position of the accent can sometimes be determined quite accurately in longer words with several consonants in different syllables. This is especially evident in some archaic comparatives, which, as opposed to their end-stressed positive counterparts, must have had the accent on the root syllable or – more accurately – on the antepenultimate.

• *alþizan- comp. ‘older’: Go. alþiza, ON ellri < *h₂él-t-i-son- vs. *alda- adj. ‘old’: OE eald, OS ald, OHG alt < *h₂el-tó-
• *junhizan- comp. ‘younger’: Go. juhiza, ON œri < Pre-Gm. *i̯uHúnkison- < *h₂i-Hń̥-k´-is-on- vs. *junga- adj. ‘young’: Go. juggs, ON ungr < Pre-Gm. *i̯uHunkó- < *h₂iu-Hn̥- k´ó-

Verner’s law seems to have preceded the resolution of the hiatus caused by the loss of intervocalic laryngeals. This is, at any rate, what follows from the following cases, which must still have been trisyllabic at the time of Verner’s law:

• *maizan- comp. ‘more’: Go. maiza, ON meiri, OE māra, OFri. māra, mēra, OS, OHG mēro < *méh₂-is-on-, cf. OIr. mór adj. ‘great’ < *meh₂-ro-
• *flaizan- comp. ‘more’: Go. flaiza, ON fleiri < *plóh₁-is-on-, cf. Lat. plūs, ­ris comp. ‘id.’ < *ploh₁-is-
• *winda- m. ‘wind’: Go. winds, ON vindr, OE, OFri., OS wind, OHG wint < *h₂uéh₁-ent-o-, cf. Hitt. ḫuu̯ant- c. ‘id.’ < *h₂uh₁-ent-, Lat. ventus m. ‘id.’ < *h₂ueh₁-(e)nt-o-, Skt. vā́ta- m. ‘id.’ < *h₂ueh₁-nt-o-

Contrary to the usual reconstruction, I derive *winda- from *h₁uéh₁-ent-o- (with generalization of the full grade in both the root and the suffix), not from *h₂uéh₁-n̥t-o- as continued by Skt. vā́ta-. The Proto-Germanic outcome of the latter ablaut variant would probably have been *wē(w)unda-, as follows from PGm. *ju(w)unþi- ‘youth’ < *h₂iu-Hń̥-ti- and PGm. *junga-, which is generally derived from *ju(w)unga- with a vocalized nasal (cf. Kluge 1913: 242; Ringe 2006: 83), and may like Go. junds ‘youth’ < *ju(w)undi- still have had a long vowel in Gothic. The general vocalization of resonants after laryngeals is also confirmed by the 1sg. subjunctive ending, cf. Go. -jau, which developed from PGm. *-jēu < PIE *-ieh₁-m̥. Also, even if *h₂uéh₁-nt-o- did develop into Pre-Gm. *u̯ḗnto-, it would have given **winþa-, not *winda-, and the same is actually true for the additional variant *h₂uh₁-ént-o-. Of course, it is still possible to start from end-stressed forms *h₂ueh₁-ent-ó- or even *h₂uh₁-ent-ó-, but given the fact that full grades typically take the accent, as for instance in Skt. vā́ta-, it is more attractive to reconstruct the word as *h₁uéh₁-ent-o-. Interestingly, this form would suggest that Osthoff’s law was posterior to both Verner’s law and the monophthongization of *-eh₁e-, which would confirm that it is a relatively late Germanic development unrelated to comparable vowel shortenings in other Indo-European languages, such as Greek and Italo-Celtic.

2.2.3 Epenthesis of *f

It is interesting to see that, in Proto-Germanic, m assimilated only to a following voiceless *d, not to fricative *þ. The latter appears to have triggered the rise of an automatic f in between, probably already within Proto-Germanic itself, and it has been argued in view of Go. anda-numt (see below) that the sequence *-mfþ- further developed into *-mft- (Rasmussen 1983).

• *samþu- m. ‘soft’: OE sēfte, OHG samfti, semfti < *sóm-tu-, cf. Skt. santya- adj. ‘belonging together’ < *som-tio-
• *tumþi- f. ‘agreement’: OHG zumft < *dḿ-ti-, cf. *teman- s.v. ‘to befit’
• *numþi- f. ‘taking, accepting’: Go. anda-numts, OHG numft < *nḿ-ti-, cf. *neman- ‘to take’
• *swumþi- f. ‘swamp’: OHG sunft < *suḿ-ti-, cf. ON swimma, OE, OHG swimman s.v. ‘to swim’ < *swimman-

The epenthesis of f still seems to have been automatic in synchronic Gothic in view of the doublet swumsl ~ swumfsl n. ‘pool’ < *swum-sla-, both variants of which occur in chapter 9 of the Gospel of John. It may follow from this that the f arose between m and þ at a relatively late stage, but certainly after the occlusivation of *đ to *d after nasals (cf. Rasmussen 1983), as there is no similar epenthesis of *b or *ƀ.

• *hunda- n. ‘hundred’: Go., OE, OS hund, OHG hunt < *dk´mtó-, cf. Lith. šim̃tas num. ‘id.’
• *skandō- f. ‘ashamed’: Go. skanda, OFri. skonde, OE scand, OHG scanta < *skom-téh₂-, cf. Go. skaman sik w.v. ‘to be ashamed’ < *skamēn-
• *sunda- n. ‘swimming; strait’: ON, OE sund < *sum-tó-, cf. *swumþi- (see above)

There are two cases that reveal an originally paradigmatic Verner alternation, which makes them particularly interesting:

• *kwumþi- ~ *k(w)undi- f. ‘arrival’: Go. ga-qumþs, OHG qhumft, kumft < *g^wḿ-ti- vs. *k(w)undi- f. ‘id.’: ON sam-kund < *g^wm-tí-[2]
• *grumþu- ~ *grundu- m. ‘floor’: ON grunnr, G Cimb. grumf < *g^hrḿ-tu- vs. Go. grundus*, OE, OFri., OS grund, OHG grunt < *g^hrm-tú- (for the root structure *g^hrem-, cf. ON grandi m. ‘isthmus’ < *granda-)

When Proto-Germanic still had a mobile accent, these ti- and tu-stems probably had root-stress in the nominative, and suffix-stress in the genitive, e.g. nom. *g^hrḿ-tu-s, gen. *g^hrm-té/óu-s. After the Germanic sound shifts, the nominative developed into *grumfþuz, whence G Cimb. grumf, while the genitive *grundauz ultimately served as the basis for Go. grundus and the aforementioned West Germanic forms. ON grunnr, on the other hand, goes back to *grunþuz, and appears to be a secondary variant with analogical n or þ. The fact that this analogy was possible proves that the paradigmatic Verner alternation must have remained intact until after the breaking up of Proto-Germanic and survived into Proto-Norse.

2.2.4 Delabialization before *j

As argued under §2.2.2, Pre-Gm. *gw was delabialized to *g under certain circumstances, especially before *u and *r. Another important position in which delabialization appears to have occurred is immediately before *j. The evidence suggests that this development was conditioned by the surrounding vocalism: delabialization is found in words where *gw was preceded by an originally round vowel.

• *dangjan- w.v. ‘to beat’: ON dengja, OE dencgan < *d^hong^wh-éie-, cf. OSw. diunga s.v. ‘to beat’ < *dingwan- < *d^heng^wh-e-[3]
• *sagja- m. ‘man, hero’: ON seggr, OE secg < *sok^wH-ió-, cf. Lat. socius m. ‘companion’ < *sok^wH-io-, Skt. sákhā, dat. sákhye m. ‘id.’ < *sok^wH-oi-
• *sagjan- w.v. ‘to say, recount’: ON segja, OE secgan, OFri. sedza, sidza, OS seggian < *sok^w-éie-, cf. Lith. sakýti, SCS sočiti ‘to indicate’ < *sok^w-éie- and Gr. ἐν(ν)έπω ‘to say, recount, announce’ < *h₁en-sek^w-
• *wulgī-, gen. *wulgjōz f. ‘she-wolf’: ON ylgr, ylgja < *ulk^w-íh₂-, *-iéh₂-s, cf. Skt. vr̥kī́-, Lith. vìlkė f. ‘id.’ (also cf. OHG wulpa f. ‘id.’ < *wulbjō- with the labial adopted from *wulfaz ‘wolf’ < *uĺk^w-o- prior to Verner’s law)[4]

When there was originally an *a in the root, we find the expected outcomes of PGm. *gw:

• *aujō- f. ‘wetland, island’: ON ey, OE īeg, OHG ouwa < *h₂ek^w-iéh₂-, cf. Go. aƕa, ON á, OE ē, OS, OHG aha f. ‘river’, Lat. aqua f. ‘water’ < *h₂ék^w-eh₂-
• *mawī, gen. maujōz f. ‘girl’: Go. mawi, gen. maujos, ON mær, gen. meyjar < *mag^h-u-ieh₂-, cf. Go. magus, ON mǫgr m. ‘boy’, OIr. mug, Corn. maw m. ‘servant’ < *mag^h-u-

The alternation of PGm. *dangjan- vs. *dingwan- is especially interesting because it provides a model for the original distribution of the *g and *w in Go. hneiwan, ON hníga, OE, OS, OHG hnīgan < *hnīwan- ~ *hnīgan- ‘to bow (down)’ and the pertaining causative Go. hnaiwjan, ON hneigja, OE hnǣgan, OS gi-hnēgian, OHG neigan, hneiken < *hnaiwjan- ~ *hnaigjan- w.v. ‘to make bow (down)’. It seems reasonable to assume that the labialization was regularly lost in the causative *hnaigjan- < *knoig^wh-éie-, but retained in the strong verb *hnīwan- < *knéig^wh-e-. In order to eliminate the root variation, Gothic generalized the *w and Northwest-Germanic the *g.

2.2.5 The rise of consonantal length

2.2.5.1 Assibilation of dental clusters

Unlike Germanic, Proto-Indo-European did not have long consonants. When two identical consonants collided across a morpheme boundary, the surface result was always a singulate, cf. PIE *h₁és-si ‘you are’ > *h₁esi > Skt. ási, Gr. εἶ. The only exception to this rule is when the colliding stops where dentals. The resulting dental clusters were not simplified in Proto-Indo-European, but received an automatic sibilant between the two segments, e.g. *-t-t-, *-d-t-, *-d^h-t- > *-tst-. The outcome of this cluster, which was retained as such only in Anatolian, varies across the different Indo-European dialects, but always yielded long *-ss- in Germanic:

• *kwessi- f. ‘consent’: Go. ga-qiss* < *g^wet-ti-, cf. Go. qiþan s.v. ‘to speak’ < *g^wét-e-
• *sessa- m. ‘seat’: ON, OE sess < *sed-to-, cf. Skt. sátta- ptc. ‘seated’, Lat. sessus m. ‘sitting’
• *wissa- adj. ‘certain’: Go. un-wiss (‘uncertain’), ON viss, OE wiss, OFri. wis, OHG gi-wis < *uid-to-, Skt. vittá- adj. ‘id.’, Gr. ἄιστος adj. ‘unseen’
• *wissi- f. ‘joint’: Go. ga-wiss < *(H)ued^h-ti-, cf. Go. ga-widan s.v. ‘to (con)join’, OIr. feidid ‘to lead, bring together’ < *(H)uéd^h-e-

Long *s may have been the first geminate to arise in Proto-Germanic. But as the result of a number of progressive and regressive assimilations, many others were to follow. Below is a summary of the most important ones.

2.2.5.2 Kluge’s law

A central Germanic innovation giving rise to Proto-Germanic long stops is Kluge’s law. According to the traditional formulation of this law, voiced *b, *d and *g were geminated to *bb, *dd and *gg by the assimilation of a following *n in a stressed syllable (Kluge 1984). These geminates were then devoiced to *pp, *tt and *kk together with old Proto-Indo-European *b, *d, *g and *g^w during stage 2 of Grimm’s law.

• *budmō, gen. *buttaz m. ‘bottom’: ON botn, OE botm, OS bodom < *b^hud^h-mēn, gen. *b^hud^h-n-ós, cf. Gr. πυϑμήν m. ‘id.’, Skt. budhná-, Lat. fundus m. ‘id.’ (with Thurneysen’s law)[5]
• *hwitta- adj. ‘white’: Du. wit < *k´uit-nó-, cf. Skt. śvítna- adj. ‘white, whitish’ < *k´uit-no-
• *þakkōn- w.v. ‘to touch, pat’: E þaccian < *th₂g-néh₂-, cf. Lat. tangō, -ere ‘to touch’ < *th₂g-néh₂- (again with Thurneysen’s law)

For obvious reasons, Kluge’s law had far-reaching consequences for the n-stems and the neh₂-presents: these grammatical categories developed paradigms with an alternation of geminated and non-geminated roots. The different dialects often resolved this allomorphy by leveling either the voiced or the voiceless consonant, a simplification process that paradoxically enough gave rise to a more complex phonological system by creating new, secondary geminates such as *ff, *þþ, *hh and *bb, *dd, *gg. Although these geminates can often be shown to go back to Proto-Northwest Germanic, it is not entirely certain whether they could already have been introduced in the Proto-Germanic period (but cf. Kroonen 2011: 80-2).

• nom. *hrīþō, gen. *hrittaz m. ‘fever’: OS hrido, OHG rīdo, rit(t)o < nom. *kréit-ōn, gen. *krit-n-ós, cf. OIr. crith, W cryd ‘id.’ < *kri-ti/u-
• 3sg. *lappōþi, 3pl. labunanþi w.v. ‘to lick up’, OSw. lapa, OE lapian, EDu. labben, lappen < 3sg. *lap-néh₂-ti, 3pl. *lap-n̥h₂-énti, cf. Lat. lambō, -ere ‘to lick’, Lith. lapènti ‘to drink greedily’ (of pigs)

Kluge’s law had a particularly strong impact on the verbal system. The PIE neh₂-presents, which often had iterative semantics, are an extremely productive category in Germanic. Browsing through this dictionary will reveal that practically every strong verb was accompanied by an iterative verb. In many cases, these iteratives seem to have been more primary than their pertaining strong verbs. It can be shown, at any rate, that many strong verbs were back-formed to their iteratives. When this happened, the root varation present in the iterative paradigm was typically exported to the strong verb, which as a result received a similar set of root alternants. Consider the cross-dialectal variation of the strong verb ‘to suck’:

• *sūgan- ~ *sūkan- s.v. ‘to suck’: ON súga, OE sūgan, sūcan, MDu. sugen, sucen, OHG sūgan, cf. Lat. sūcus m. ‘juice’ < *souk´-o-, OCS sъsati (sъsǫ) ‘id.’ < *suk´-eh₂-

Since the PIE root underling this verb was *seuk´-, not *suHǵ^h- or *suHǵ-, the alternation between root-final *g and *k must find its origin in the pertaining iterative:

• 3sg. *sukkōþi, 3pl. *sugunanþi w.v. ‘to suck’: OE socian, G Rhnl. sucken, Swi. (App.) sugə < 3sg. *suk´-néh₂-ti, 3pl. *suk´-n̥h₂-énti

There is an additional corollary to the frequent back-formation of strong verbs to iteratives. It seems evident that the long *ū of strong verbs such as *sūgan- ~ *sūkan- arose analogically as a result of the back-formation process (Kroonen 2011: 112-7). Parallel to strong verbs in *ī, which were accompanied by iteratives with short *i, the *ū of *sūgan- ~ *sūkan- must have arisen by the lengthening of the *u of the iterative allomorphs *sukk- and *sug-. For the shortening of the *kk after long *ū, see §2.2.6.

2.2.5.3 Nasal assimilation by resonants

The resonants *l and *r (possibly also *m and *n) were also lengthened by the assimilation of a following *n. Consider the following examples:

• *alla- adj. ‘all’: Go. alls < *h₂el-nó-, cf. Osc. allo (f.) ‘all, entire’
• *fella- n. ‘skin’: Go. þruts-fill n. ‘leprosy’ < *pel-no-, cf. Lat. pellis ‘id.’ < *pel­ni-
• *hulli- f. ‘hill’: OE hyl < *kl(H)-ni-, cf. Lat. collis ‘id.’ < *kolH-ni-, Lith. kálnas m. ‘id.’ < *kolH-no-
• *star(r)an- m. ‘starling’: ON stari, Icel. star(r)i, MDu. sterre, OHG star(o) < *h₂stór-ōn, gen. *h₂stor-n-ós, cf. Lat. sturnus m. ‘id’ < *h₂stor-no-
• *ster(r)an- m. ‘star’: OE steorra, OFri. stera, OS sterro, OHG sterro, sterno < *h₂stér-ōn, gen. *h₂ster-n-ós, cf. Hitt. ḫašter- c. ‘id.’, Gr. ἀστήρ, -έρος m. ‘id.’ < *h₂ster-
• *wellō- f. ‘wave’: OHG wella < *uel-neh₂-, cf. Ru. volná f. ‘id.’ < *ul-neh₂-
• *wullō- f. ‘wool’: Go. wulla, ON ull < *HulH-neh₂-, cf. Skt. ū́rṇā- f. ‘id.’

It is plausible that assimilation only occurred when the nasal was in a stressed syllable, especially since that would be parallel to the conditioning of Kluge’s law. It is probably significant that the lack of gemination in the following instances with *-rn- indeed seems to correspond to root-stress in the extra-Germanic cognates:

• *hurna- n. ‘horn’: Go. haurn, ON, OE, OFri., OHG horn < *k´rn-o-, cf. Skt. śŕ̥ṅga- ‘horn’ < *k´rn-go-
• *skarna- n. ‘dung’: ON skarn, OE scearn, OFri. skern < *sk´-or-no-, cf. Gr. σκῶρ, gen. σκατός n. ‘muck, excrement’ < *sk´-ōr, *-nt-ós
• *þurna- n. ‘thorn’: ON, OE þorn, OFri. thorn, OHG dorn < *tr-no-, cf. Skt. tŕ̥ṇa- n. ‘grass, blade of grass, herb’
• kurna- n. ‘corn, grain; kernel’: Go. kaurn, OE corn, ON, OS, OHG korn < *ǵrh₂-no-, cf. Lat. grānum, OIr. grán, OCS zrъno n. ‘grain’[6]

2.2.5.4 Long *m

It is not entirely certain whether geminated *m could arise by a parallel assimilation of *n, as the evidence is marginal. The cluster *-mn- usually seems to develope into *-bn-, although it is possible that this only happened in those cases where n was retained due to a preceding accent. In the Proto-Germanic word for ‘voice’ (see below), all three possibilities seem to be represented. Apparently, this nō-stem continues an older ablauting (m)n-stem in which the nominative *stemō alternated with a genitive *stimmaz and a dative *stemeni. Thematization into a nō-stem gave rise to several different variants. OHG stimma, for instance, seems to be built on the original genitive, while Go. stibna must continue *stebnō- or *stibnō-, which could have developed out of a secondary genitive *stemnaz or *stimnaz before the change *-mn- > *-bn-. OHG stimna, on the other hand, may have developed from yet another thematization posterior to this change.

• *stebnō- ~ *stimnō- ~ *stimmō- f. ‘voice’: Go. stibna, OFri. stemme, OS stemna, OHG stimma, stimna, cf. Hitt. ištāman- ~ ištamin- c./n. ‘ear’, Gr. στόμα n. ‘mouth’ < *stom-n-; Av. staman- m. ‘snout’ < *stem-n-; MW safyn f./m. ‘jawbone, mouth’ < *stm̥-n-

There further is compelling evidence for Proto-Germanic assimilation of a preceding *z:

• *gamman- m. ‘animal stall(?)’: ON gammi < *gazma-(?) < ?*g^hos-mó-(?), cf. Arm. gom ‘fold (for cattle)’
• *immi 1sg.pres. ‘I am’: Go. im, ON em (with e from the plural erum, eruð, eru) < *ezmi, cf. Skt. ásmi, ási, ásti and Gr. εἰμί, εἶ, ἐστί < PIE *h₁és-mi, *h₁ési, *h₁és-ti
• *kwramma- adj. ‘thawed, wet’: ON krammr < *kwramzma- < *g^wroms-mó-, cf. Lith. grim̃zti (grimztù) ‘to sink’ < *g^wrm̥s-ske, Ru. grjáznut’ ‘to sink into something sticky or boggy’ < *g^wrm̥s-ne-
• *mammōn- f. ‘flesh’ < *ma(m)zmōn- < *mo(m)s-mon-, cf. Go. mims, Skt. māṃsá- n. ‘meat’ < *mēmsó-
• *þammē dat.sg.m. ‘that’: Go. þamma < *tosmeh₁, cf. Skt. tásmai dat. ‘id.’ < *tosmōi

A problem is that the underlying *z of *immi and *þammē does not correspond to the initial accent in the corresponding Sanskrit forms, but it seems likely that Verner’s law operated in these two words simply because they often occurred in unstressed position (cf. Ringe 2006: 141). Incidentally, *immi seems to indicate that Verner’s law preceded the raising of *e to *i before tautosyllabic nasals. Assimilation did not affect *-sm-, as follows from the examples below:

• *bōsma- m. ‘bosom’: OE bōsm, OFri. bōsem, OHG buosum < *b^heh₂ǵ^h-smo-, cf. ON bógr m. ‘shoulder’, Skt. bāhú- m. ‘arm, forearm, forefoot of an animal’, Gr. πῆχυς m. ‘forearm, arm; cubit’ < *b^heh₂ǵ^h-u-
• *rusman- m. ‘rust’: OHG rosmo < *h₁rud^h-smon-
• *þaismjan- m. ‘sourdough’: OE þǣsma, OHG deismo < *teh₂is-mon-, cf. Ru. tésto n. ‘dough’, OIr. táis, W toes m. ‘id.’ < *teh₂is-to-

2.2.5.5 Long *l

Like long *m, geminated *l could arise by the assimilation of a preceding *z:

• *gilla-: Nw. dial. gjell m. ‘interrupted rainbow’ < *gizla-, cf. Icel. gísli m. ‘beam, ray’ < *gīslan-
• *krulla- adj. ‘curly’: MDu. crul, MHG krol < *kruzla-, cf. MHG krūs adj. ‘id.’ < *krūsa-

Similarly, long *l could arise by the assimilation of preceding *d. Apparent counter-examples such as OE īdel, OHG ītal adj. ‘void’ < *īdla-, may have been created with productive l-suffixes after the assimilation took place, or there may have been an a- or e-vowel before the l.

• *knulla- m. ‘lump’: OE cnoll < *knudla-, cf. OE cnoda m. ‘lump’ < *knudan-
• *stalla- m. ‘standing, stall, stable’: ON stallr, OE steall, OHG stal < *sth₂-d^hlo- or *sth₂-tl-, cf. Lat. stabulum n. ‘stable’
• *strullōn- w.v. ‘to gush’: MHG strullen, cf. OHG stredan s.v. ‘to seethe’
• *trullōn- w.v. ‘to pace’: MHG trollen, cf. Go. trudan s.v. ‘to tread’

2.2.5.6 Long *n

Long *n primarily arose by the assimilation of a *u̯ by a preceding *n. There are numerous examples of this change, including the following ones:

• *minna- adj. ‘small’: OE minn < *mi-nu-o-, cf. Lat. minuō, -ere ‘to diminish’ < *mi-nu-
• *þunnu- adj. ‘thin’: ON þunnr, OE þynne, OHG dunni < *tn̥h₂-u-, cf. Skt. tanú(ka)-, OCS tьnъkъ, Gr. ταναός, Lat. tenuis adj. ‘id.’
• *winnan- s.v. ‘to suffer; to labor; to gain’: Go. winnan, ON vinna, OE winnan, OFri. winna, OS winnan, OHG winnan < *uénu-e-, cf. Skt. vanóti ‘to win, defeat, procure’ < *u̯n̥-néu-

2.2.5.7 Holtzmann’s law

In a significant number of words, the PIE glides *-i̯- and *-u̯- emerge as PGm. *­jj- and *-ww-. The gemination underlying these long glides is referred to as Holtzmann’s law, after its discoverer Adolf Holtzmann (1835). In synchronic Proto-Germanic, the glides appear in intervocalic position, but only after short vowels. This constraint may be due to the original conditioning of the sound law, which is generally assumed to have operated only after short vowels. It is possible, too, that long glides from this law were simply shortened after long vowels along with the resolution of all other overlong syllables (see §2.2.6). In Gothic and Old Norse, *-jj- and *­ww- were further occlusified to -ggj- , -ggv- and ­ddj-, -ggw- respectively, a process that is generally referred to as the Verschärfung. It is considered to be an important Northeast Germanic isogloss, and is sometimes adduced to demonstrate a Gotho-Nordic versus a West Germanic division. Actually, it is more likely that Verschärfung only partly affected the Proto-Germanic dialect continuum, leaving the future West Germanic dialects untouched. Consider the following cases with PIE *­u̯- > PGm. *­ww­:

• *blewwan- s.v ‘to blow’: Go. bliggwan, OHG bliuwan < *mléu-e-, cf. Gr. ἀμβλύς adj. ‘blunt; dim, faint’ < *n̥-ml-u-, Av. mruta- adj. ‘crushed(?), weak’ < *mlu-tó-
• *brewwan- s.v. ‘to brew’: OSw. bryggia, OE brēowan, OFri. briouwa, brouwa, OS gi-breuwan < *b^hréuh₁-e-, Gr. Hsch. ἀπ-έφρυσεν aor. ‘brewed’ < *b^hruh₁-s-, Lat. de-frū̆tum n. ‘must’ < *-b^hruh₁-to-
• *gruwwa- n. ‘dregs’: Icel. grugg < *g^hruH-o-, cf. W gro ‘pebbles, coarse gravel’ <br/>*sawwa- m./n. ‘juice’: Icel. söggur, OE sēaw, OHG sou < *souo-, Skt. savá- m. ‘(Soma) juice’ < *sou-ó-, Lith. sulà f. ‘birch sap’ < *su-l-eh₂-
• *snawwa- adj. ‘quick’: ON snǫggr < *snouh₁-o-, cf. *snewan- ‘to rush’ (see below)
• *snawwa- adj. ‘bald’ (< ‘shaved’): ON snøggr < *ksnou-ó-, cf. Skt. kṣṇáuti ‘to whet, to sharpen’ < *ksnēu-, YAv. hu-xšnuta- adj. ‘well-sharpened’ < *ksnu-to-

The counter-examples to Holtzmann’s law are numerous, and this indicates that the scope of the law was restricted by some sort of conditioning. At present, it is widely assumed that gemination occurred only by the assimilation of a laryngeal (cf. Smith 1941; Jasanoff 1978b; Rasmussen 1999 [1990]). The material, however, contains many cases with long glides in roots that are usually reconstructed without a laryngeal (see above). Moreover, many of the forms in which gemination is lacking have roots that did have a laryngeal in Proto-Indo-European.

• *hrawa- adj. ‘raw’: ON hrár, OE hræw, hreaw, OHG rao < *krouh₂-o-, cf. Skt. krūrá- adj. ‘bloody, cruel’ < *kruh₂-ró-
• *knewa- n. ‘knee’: Go. kniu, ON kné, OE cnēo(w), OS knio, OHG kneo < *ǵn-eu-o-, cf. Hitt. genu- ~ ganu- n./c. ‘id.’ < *ǵén-u-, *ǵn-eu-
• *lewan- m. ‘scythe’: ON lé, obl. ljá < *leuh₁-on-, cf. Skt. lavítra- n. ‘sickle’ < *leuh₁-tro-
• *snewan- s.v. ‘to rush’: Go. sniwan < *snéuH-e-, cf. OE snūd m. ‘rush’ < *snuH-to-[7]

A similarly unclear distribution is encountered for the words with PGm. *-jj- from PIE *-i̯-, but the material is far more limited, and so are the counter-examples.

• *ajuki- adj. ‘eternal’: Go. ajuk-duþs (‘eternity’), OE ēce < *h₂oiu-g^wh₃-i-
• *dajjan- w.v. ‘to suckle’: Go. daddjan, OSw. dæggia < *d^hh₁-oi-éie-, cf. OCS doiti (dojǫ) ‘to breast-feed, nurse’[8]
• *ejjōn 1sg.pret. ‘went’: Go. iddja, OE ēode < *h₁ei-ōm, cf. Icel. iða, Gr. ἰτάω, Lat. itō, -āre ‘to go (here and there)’ < *h₁i-t-eh₂-ie-
• *twajjan num.gen. ‘of two’: Go. twaddje, ON tveggja < *duoi-om[9]
• *wajju- m. ‘wall’: Go. waddjus, ON veggr < *uh̥₁i-u-

In view of unambiguous cases such as *knewa- < *ǵnéu-o-, *sawwa- ~ Skt. savá- < *sou-ó- and *lewan- < *léuh₁-on-, I have decided to drop laryngeals as a factor in the rise of the long glides. With alternations such as *snewan- ‘to rush’ < *snéuH- vs. *snawwa- ‘quick’ < *snouH-o-, it seems out of the question, at any rate, that the laryngeals were the only factor at work. I have therefore chosen to adopt the more traditional formulation of Holtzmann’s law by Kluge 1879: 128, who defined it as plain and simple pretonic gemination. Although this conditioning is not without exceptions either, it seems to predict the Germanic material relatively accurately. Apparent counter-examples such as *brewwan- < *b^hréuh₁- can for instance be accounted for by leveling in a verbal paradigm *brewan-, pret.3sg. *brawe, 3pl. *bruwwun, ptc. *bruwwana-, or by assuming influence from the pertaining iterative ON brugga w.v. ‘to brew’ < *bruwwōn-. Most importantly, Holtzmann’s law seems to have been fed by Dybo’s law, which is another indication that the accent was located not on the root, but on the suffix or ending. This is especially clear in the PGm. word for ‘egg’:

• *ajja- n. ‘egg’: Go. Crim. ada (n.pl.), ON egg, OE ǣg, OS, OHG ei < *h₂ōu­ió-, cf. Gr. ᾠόν, Lat. ōvum, OCS aice n. ‘id.’

If we assume that Holtzmann’s law only affected pretonic glides, the Germanic form *ajja- must go back to earlier *oi̯ó-. This is exactly the form that is expected if the PIE form *h₂ōu-ió- was previously modified by 1) the regular change *-ōu- > *-ō- and 2) Dybo’s law. It follows from this chronology that Holtzmann’s law was triggered by a following accent, not by the assimilation of an adjacent laryngeal.9

2.2.6 Shortening of overlong syllables

One of the final sound laws leading up to the stage we reconstruct as Proto-Germanic caused the resolution of overlong syllables, i.e. syllables ending in a long vowel or diphthong plus a long consonant. In syllables like these, long stops were shortened, thus giving rise to a Proto-Germanic constraint on overlong syllables. The constraint is likely to have arisen at a relatively late stage, but it is conceivable that originally overlong syllables can be retrieved from the loanwords in the Finnic languages, which have no such phonotactic limitation (Mulder 2010). Anyhow, all geminates arisen according to the sound laws mentioned above were shortened in overlong syllables, as follows from the following examples:

• *aila- n. ‘fire’: OE āl < *ailla- < *h₂eid^h-lo-, cf. Skt. édha- m. ‘firewood’, Gr. αἶϑος m. ‘firebrand’
• *blōman- m. ‘flower’: Go. bloma, ON blómi, OS blōmo, OHG bluomo < *blōmman- < *blōzman- < *b^hleh₃-smon-, cf. OE blōs(t)ma m. ‘blossom’ < *blōsman- (without Verner’s law)
• *deupa- adj. ‘deep’: Go. diups, ON djúpr, OE dēop, OFri. diāp, OS diop, OHG tiuf < *deuppa- < *d^heub^h-nó-, cf. OIr. domain, W dwfn adj. ‘deep’ < *d^hub^h-ni-, Lith. dubùs adj. ‘hollow, deep, spacious’ < *d^hub^h-u-
• *dīka- n. ‘dam; ditch’: OE dīc, OFri., OS dīk < *dīkka- < *d^heiǵ^h-nó-, Gr. τοῖχος m. ‘wall’, Arm. dēz ‘heap’ < *d^hóiǵ^h-o-
• *hwīta- adj. ‘white’: Go. ƕeits, ON hvítr, OE, OS hwīt, OHG wīz, hwīz < *hwītta- < *k´ueit-nó-, cf. Skt. śvítna- adj. ‘white, whitish’ < *k´uit-no-, Skt. śvetá- adj. ‘white, bright’, YAv. spaēta- adj. ‘white’ < *k´uoit-o-
• *kīla- m. ‘wedge’: OHG kīl < *kīlla- < *kīdla-, cf. G dial. Keidel < *kīþla-
• *stōla- m. ‘chair’: Go. stols, ON stóll, OE, OFri., OS stōl, OHG stuol < *stōlla- < *stōdla- < *steh₂-tló-, cf. Lat. ob-stāculum n. ‘obstacle’
• *tōla- n. ‘tool’: ON tól, OE tōl < *tōlla- < *tōdla-, cf. Go. taujan w.v. ‘to do, make’
• *wīsa- adj. ‘wise’: Go. un-weis (‘ignorant’), ON víss, OE, OS, OHG wīs < *wīssa- < *ueid-to-

The shortening of long consonants in overlong syllables is especially apparent in those strong verbs that adopted their root-final consonantism from their pertaining iteratives or were back-formed from them. For this effect, see §2.2.5.2.

2.3 Flowchart

The relative chronology of some of the most important sound changes given above can be schematized as follows: