¹ The following works are referred to by the name of the author: Leonard Bloomfield, Language (New York, 1933); Wilhelm Braune, Gotische Grammatik, 10th ed. (Halle, 1928); Hermann Hirt,Handbuch des Urgermanischen (Heidelberg, 1931); Max Hermann Jellinek, Geschichte der gotischen Sprache (Berlin and Leipzig, 1926); Ernst Kieckers, Handbuch der vergleichenden gotischen Grammatik (Munich, 1928); Wilhelm Streitberg, Gotisches Elementarbuch, 5th-6th ed. (Heidelberg, 1920).

Gothic forms are cited from Wilhelm Streitberg, Die gotische Bibel, 2d ed. (Heidelberg, 1919; glossary 1928), with translations from the King James version where possible. Although it is not customary to cite chapter and verse references when quoting Gothic forms, this has been done here out of caution (using Streitberg's abbreviations), since the standard handbooks seem occasionally to cite non-recorded forms to prove a given point. (Thus Braune §74 cites nom. alds ‘world’ as evidence that Gothic d remains between l and s. Acc. ald does occur, in t 4.10, but I could not find nom. alds.)

² Bloomfield 299–300.

³ Possibly also before /x/ in the pre-Germanic period, though it is fruitless to speculate on this, since the nasal was lost in this position: pre-Gmc. */fanxanan/ > PGmc. */fã:-xanan/ > ON fā ‘get’, OE fōn, OHG fāhan ‘seize’.

⁴ The symbol # = pause, or position at the beginning of end of a word.

⁵ I write /h/ rather than /x/ because of tradition, and because there is some slight evidence that /h/ was a glottal—rather than a velar—spirant. See footnote 15 below.

⁶ It is generally assumed that Gothic l, m, n, r were non-syllabic before and after vowels, but syllabic between consonants, or between consonant and # : fugls ‘bird’, tagl ‘hair’; maiþms ‘gift’, acc. bagm ‘tree’; taikns ‘sign’, rign ‘rain’; akrs ‘field’, tagr ‘tear’. (Cf. Braune §27; Streitberg §95, §99; etc.) If we accept this interpretation, these four phonemes have syllabic and non-syllabic allophones. The interpretation has, however, no support whatever from Gothic spellings, and is based on such West Germanic forms as OHG fogal ‘bird’, OS mêđom ‘treasure’, OHG zeihhan ‘sign’, OHG akkar ‘field’, where -al, -om, -an, -ar are very probably to be derived from (WGmc.) syllabic l, m, n, r. But this tells us nothing about Gothic. Unless the Gothic spelling gives a real clue, it does not seem helpful to guess about such phonetic details. Cf. Braune's skepticism §27 Anm. 1; and the very sound remarks on such speculation in general in Jellinek, Anmerkung pp. 33–4. Cf. also Bloomfield's statement (294) that ‘the phonetic values of the characters in ancient writings can never be surely known.‘

⁷ Cf. Streitberg §28.

⁸ Gustaf Kossina, Festschrift zur 50-jährigen Doktorjubelfeier Karl Weinholds 40 (Strassburg, 1896).

⁹ I have no reference for hlaifs; but Braune, §56 Anm. 1, says it occurs 12 times. /b/ is also replaced by /f/ before /t/: infin. giban M 27.58 ‘give’, but pret. gaft J 17.22 ‘thou gavest’.

¹⁰ Streitberg §35.8, 9, §105, accepts Sievers' conclusions that b, d, g represent voiced stops everywhere in Gothic, and is thus forced to assume spirantal allophones only for pre-Gothic. Sievers' conclusions are based on his studies of intonation in Gothic. Since his whole theory of intonation is a body of information which he alone controlled, and which is not successfully communicable to others, it must be placed outside the realm of science. It is therefore not considered in the present study.

¹¹ /z/ between vowels stands in a similar morphophonemic relationship to /s/ between vowel and # : dat. hatiza L 3.7, but nom. hatis E 4.31 ‘wrath’. The numerous spellings with b, d, z between vowel and #, or b, d between vowel and s (see Streitberg §25), may be interpreted as morphophonemic spellings (like those in modern German); or, where they occur before a following voiced sound, as retention of the voiced spirant in close juncture with the following voiced sound.

¹² Despite the usual confident assertion (e.g. in Braune §56, Streitberg §123) that lb before vowels does not alternate with lf before # and s, we cannot know whether this was so, since lb never happens to occur before # or s in our manuscripts. (Cf. the more cautious statement in Jellinek §67.) It is entirely possible—though perhaps improbable—that Gothic had such forms as nom. sg. masc. *halfs, neut. *half, in alternation with recorded acc. sg. fem. halba Mc 6.23, acc. sg. nt. halbata L 19.8 ‘half’.

¹³ This is, of course, the generally accepted view. Cf. Braune §54, §56, §72, §74; Jellinek §29, §67; Streitberg §123.

¹⁴ /g/ stands in morphophonemic alternation with /h/: pres. mag ‘he can’, pret. mahta ‘he could’; but this is inherited from Proto-Germanic. Cf. OE mæg, meahte; OHG mag, mahta; etc.

¹⁵ If the /g/ of acc. dag, nom. dags ‘day’ was a voiceless spirant, it contrasted phonemically with the voiceless spirant of jah ‘and’ and slahs ‘blow’. Possibly unvoiced /g/ was a velar spirant, while /h/ was a glottal spirant. The conjecture that Gothic h represented a glottal—rather than a velar—spirant is supported by the fact that it is occasionally omitted, or added where it does not etymologically belong (cf. Streitberg §27); that it is used to transcribe the Greek spiritus asper (cf. Streitberg §20.9); and that it is never used to transcribe the Greek voiceless velar spirant χ (cf. Streitberg §20.4). True, the letter g is also never used to transcribe Greek χ; but χ occurs only where, according to our conjecture, /g/ would have been voiced (cf. Jellinek §33).

¹⁶ For literature on these often discussed problems, see Hirt 1.43–50.

¹⁷ Gen.þarⁱhis M 9.16 ‘uncarded, new’, with added i in a doubtful passage, may be a scribal error for þairhrs. Literature in Streitberg §50 Anm.

¹⁸ In forms of urraisjan, urrannjan, urredan, urreisan, urrinnan, urrumnan, and of urrists, urrugks, urruns; see the glossary in Streitberg's Gotische Bibel.

¹⁹ For ur in words of non-Gothic origin, see Streitberg §52a.

²⁰ Further examples in Streitberg §49, §§210–2.

²¹ Cf. Streitberg §51b. In words borrowed from Greek, [e] and [o] occur freely before phonemes other than /h r/: acc. Aifaison t 4.12 “'Eφϵσoν', Saulaumon M 6.29 'Σoλoμών', aipiskaupus T 3.2 'ἐπίσκoπoς', etc.

²² The Gothic letter which we transcribe w had the shape and numeral value of Greek ϒ. In words borrowed from Greek (or one might better say: in the Gothic spelling of Greek words) it transcribes Greek v and oι (which were homonymous in 4th-century Greek): acc. swnagogen Mc 1.21 'συναγωγήν', in Lwstrws t 3.11 'ἐν Λύστρoις', etc. (Cf. Jellinek §35a.) This w represents, therefore, a sixth short vowel phoneme, though it occurs only in foreign words.

²³ For a more exact description of the distribution of these allophones, see Franklin Edgerton, Lang. 19.83–124 (1943), who bases his study primarily on Sanskrit. Beside the syllabic and non-syllabic sets of allophones, Edgerton assumes under certain conditions a third set: [ij uw ḷ1 ṃm ṇň ṛr]. He also works out, as far as it can be reconstructed, the distribution of allophones when two or more of these phonemes occur in juxtaposition.

²⁴ On such forms as acc. sg. fiand M 5.43 ‘enemy’, see footnote 27.

²⁵ Such forms as fraïtiþ k 11.20 ‘devours’, gaïbnjand L 19.44 ‘they lay even’, in which the syllabicity of i is specifically marked in our manuscripts, pose a problem. However, even though we can know little of Gothic stress, I believe we may assume that in these words the /a/ and the /i/ are not in juxtaposition, but are separated by onset of strong stress: /fra'itiþ/, /ga'ibniand/. These examples would then belong rather in position 7: # ()C.

²⁶ On such forms as saian L 8.5 ‘sow’, see §4.2 below.

²⁷ In some cases ij seems to alternate with i, with no distinction in meaning: acc. pl. fijands M 5.44 ‘enemies’, but in the preceding verse acc. sg. fiand ‘enemy’. (For further examples of this alternation, see Streitberg §30. Cf. also unbiarja Tit 1.12 ‘beasts’.) To this extent we must modify our statement concerning position 2, C()V: roots beginning with an initial consonant, or consonant cluster, have morpheme alternants with /i/ and /ii/ in free variation before vowels: /fi-/, /fri-/ in fiand, frion; but in free variation also /fii-/, /frii-/, in fijand, frijon. But after a non-initial consonant or consonant cluster we find only [j] before vowels: kunja. On such alternants as saiiþ and saijiþ, see footnote 38.

²⁸ CwC also in 2d sg. pret. saht J 8.57 ‘thou hast seen’. I interpret h and q as phonemically /hw/ and /kw/ for two reasons. First, h is the reflex not only of PIE k^w, as in saihan ‘see’ (PIE *sek^w-, as in Skt. sácatē ‘follows’, Gr. ἔπoμαi, Lat. sequor ‘follow’), but also of PIE kw, as in dat. sg. aihatundjai L 6.44, 20.37, Mc. 12.26 ‘bramble bush’ (lit. 'horse tooth') (aiha- from PIE *ekwo-, as in Skt. áçvas, Lat. equus ‘horse’). Secondly, by analyzing h and q as /hw/ and /kw/, we find that all but the labial stops and spirants form clusters with /w/: /tw/ in twalif Mc 6.7 ‘twelve’, /þw/ in þwahan J 9.7 ‘wash’, /dw/ in dwala M 5.22 ‘fool’, /sw/ in swa M 5.16 ‘so’, /zw/ in dat. pl. izwis ‘you’ M 5.18 (does not occur initially), /gw/ in bidagwa J 9.8 ‘beggar’ (does not occur initially); hence also /kw/ in qam M 5.17 ‘I came’, etc., and /hw/ in has M 5.39 ‘someone’, etc. Only the clusters /pw/ and /bw/ do not occur.

²⁹ CwV also in /kwam/ qam, /hwas/ has; see footnote 28.

³⁰ Cw # also in /sahw/ sah L 20.37 ‘I saw’, /sankw/ sagq L 4.40 ‘was setting’.

³¹ On such forms as staua K 11.29 ‘damnation, judgment’, see §4.2 below.

³² /wu/ before consonant also in /þlakwus/ þlaqus Mc 13.28 ‘tender’, /wu/ before # also in /ferhwu/ fairhu E 6.12 ‘world’.

³³ Scholars have surmised that [w] was some sort of spirant. See Jellinek §35a and references.

³⁴ But /CiwC/ once = [CiwC] in lasiws k 10.10 ‘weak’. Cf. Streitberg §90.

³⁵ The letter i between vowel symbols in the reduplicating syllable (afaiaik, anaaiauk) has already been interpreted as part of the orthographic symbol for /e/ (/afeaik/, /anaeawk/). See §2.2 above.

³⁶ Cf. Braune §22, §26; Jellinek §61.1, §62.4, §85–7; Kieckers 7–8, 19, 222–3; Streitberg §71, §75; and references given.

³⁷ The au of staua could conceivably be /au/, but this does not seem likely.

³⁸ Note that we seem to have two morpheme alternants of the verb ‘sow’, in free variation: /se:/ in saian, saiiþ, but /se:i/ in saijands, saijiþ.