¹ E. Prokosch, A comparative Germanic grammar (Philadelphia, 1939), deals with the ‘semivowels’ on 90-93 and with the ‘consonants’ on 47-90, though a few pages of this section, 85-7, include comments on ‘nasals and liquids’, chiefly assimilatory changes in which they are involved. The proportion in H. Hirt's Handbuch des Urgermanischen I (Heidelberg, 1932) is similar; on p. 35 he discusses briefly the ‘syllabic semivowels’ and on pp. 109-16 the consonantal semivowels, while the obstruents are treated 79-105.

² W. Streitberg, Urgermanische Grammatik (Heidelberg, 1896), deals with the resonants 59-63, 67-9 and 138-43; the last section deals purely with assimilatory changes. F. Kluge, Urgermanisch (Strassburg, 1913), also discusses the various developments of the resonants at greater length (67-75, 107, 110-12), but he too is primarily concerned with combinatory changes. These four handbooks are adequate to indicate the previous treatment.

³ Cf. especially his detailed reformulation of Sievers' law in The Indo-European semivowels, Lg. 19.83-124 (1943), but also his earlier article, Lg. 10.242 ff.

⁴ See also W. G. Moulton, The phonemes of Gothic, Lg. 24.81-82 (1948).

⁵ A. Noreen, Altisländische Grammatik ⁴ (Halle, 1923), gives more information on these forms (136-37, 256-58).

⁶ See W. Krause, Handbuch des Gotischen 224 (München, 1953).

⁷ We have to reconstruct such patterns for early PGmc., however, in the etyma of Go. sōkeis < *sō-ki-yis and OIcel. hirðar < *hir-ði-yas, but the y in such positions did not long remain intervocalic.

⁸ See E. Sievers, Altgermanische Metrik §79.4, §156,4, and §179 (Halle, 1893); examples may be found in Beowulf 185, 781, 1195, 2013, etc. For the similar treatment of OIcel. kumbl, rausn, sandr, etc., cf. AM §39.

⁹ Cf. S. Feist, Vgl. Wb. d. got. Spr. 459 (Leiden, 1939), and bibliography listed there; the -ð- is preserved in the o-stem forms, OIcel. saðr, OE sōð ‘true’.

¹⁰ W. Streitberg, Gotisches Elementarbuch ^5-6 §146 (Heidelberg, 1920), and Krause, HdG 141-3, give the gen. sg. of nt. jo-stem forms as -jis regardless of the length of the preceding syllables—reikjis as well as kunjis—though for the masculine they distinguish between hairdeis and harjis. On 104, Anm. 3, however, Streitberg lists gen. in -eis which are found in forms with preceding long syllables, as does Krause §126.4. The information is given most completely in F. Dieter, Laut- und Formenlehre der altgermanischen Dialekte 573, Anm. 1 (Leipzig, 1900).

¹¹ Other examples may be drawn from the lists in the handbooks, such as Krause, HdG 225-27. In most of the forms of first-class weak verbs a -j- was generalized, because it was the predominant first-class suffix, in the verbs with IE denominative suffix -ye/o-, and in the causative hyper-forms.

¹² It may be objected that, in contrast with this interpretation, mawi was fitted into a pattern in which an alternation existed as a result of the earlier resonant variation. Difficulties with this view are that no other stems with a sequence -awi are found in the Go. jō-stems, and the variety of forms in which the allophonic variation occurs in words that had had PIE -k^w-.

¹³ For bibliography and discussion, cf. Feist, Vgl. Wb. d. got. Spr. 103, who gives the Gmc. form as -arja-,

¹⁴ Cf. Proto-Indo-European phonology 64 et passim. In the neighborhood of obstruents, on the other hand, the reflexes of laryngeals were lost, with compensatory lengthening of the preceding vowel, as in OE wēsten, OS wōstinnia, OHG wuosti ‘desert’ and Lat. vāstus ‘barren’.

¹⁵ In past treatments of Gmc. -ōw-, such as Trautmann's Germanische Lautgesetze in ihrem sprachgeschichtlichen Verhältnis 17 ff. (Kirchhain, 1906), the Gmc. developments were obscured because it was impossible to make an adequate analysis of the IE situation without the assumption of laryngeals. The Go. pret. ptc. ufbaulidai ‘puffed up’ seemed irregular because -au- was found before a consonant. But if our etymologies are correct, this is derived from IE bhew- ‘become’; the laryngeal therefore was suffixed to the root, so that we must assume bhew-X- not bheX-w- (bhōw-). -X- was lost between w and l, and -aw- > -au-. We account similarly for Go. skauda-raip ‘shoe-string’, which is related to Skt. skunāti ‘covers’ and must therefore be derived from a base skew-X-, not skeX-w-(skōw-). Similarly Go. gaunon ‘bewail’, OIcel. gaula ‘howl’, geyja ‘cry’ etc., which are related to Skt. havate ‘shout’, and therefore to be derived from IE ghew-.

¹⁶ Cf. Lg. 23.109. In drawing up a comprehensive description of the occurrence of the allophones, Edgerton gives the following definition of | (a slant-line in his symbolism): ‘The simplest possible definition of [/] would then be: (1) pause position (initial or final); (2) any consonant before consonant, or before vocalic semivowel, or final; (3) length of vowel before consonant.‘ By linguistic theory developed since 1943 these environments can be equated with junctural phenomena.

¹⁷ It is to this structural change that we may ascribe the highly complex rhythm of Germanic verse. I assume that the basic principles were established when quantity (and pitch accent) was significant in Germanic, and were handed down by tradition among poets into the time when stress accent was the most prominent suprasegmental feature of the dialects, though quantity had not yet been eliminated. The result has been a complicated rhythm which has led scholars, e.g. Sievers and Heusler, to set up various metrical systems. For a brief note on the changes in Gmc. rhythm, see J. Kurylowicz, Latin and Germanic metre, English and Germanic studies 2.34-8 (1948-9), who seems to follow exclusively Sievers' analysis.

¹⁸ See Noreen, Aisl. Gram. 132, §153: ‘Gegen die mitte der urn. zeit und im anfang der vikingerzeit, also etwa zwischen 450 und 900, wird allmählich jeder unbetonte kurze ... vokal synkopiert.‘

¹⁹ A. A. Hill, Juncture and syllable division in Latin, Lg. 30.443 (1954), has suggested that the allophonic variation described in the Sievers–Edgerton formulae was also preserved into Latin for /u/ and /i/.