¹ Fr. van Coetsem, Das System der starken Verba und die Periodisierung im ālteren Germanischen (Amsterdam, 1956); E. A. Makaev, ‘Ponjatie obščegermanskogo jazyka’, Voprosy germanskogo jazykoznanija 44–67 (Moscow-Leningrad, 1961), and ‘Ponjatie obščegermanskogo jazyka i ego periodizacija’, Sravnitel'naja grammatika germanskix jazykov 1.114–24 (Moscow, 1962; translations from Russian works are my own); W. P. Lehmann, ‘A definition of Proto-Germanic: A study in the chronological delimitation of languages’, Lg. 37.67–74 (1961).

² W. P. Twaddell, ‘The prehistoric Germanic short syllables’, Lg. 24.139 (1948); cf. also L. Bloomfield, Language 311 (New York, 1933).

³ E. Pulgram, ‘Proto-Indo-European reality and reconstruction’, Lg. 35.421–6 (1959).

⁴ R. A. Hall, Jr., Idealism in Romance linguistics 13 (Ithaca, N. Y., 1963).

⁵ Cf. H. Lüdtke's definition of a protolanguage: ‘ein in seinen Elementen möglicherweise unvollständiges, in seiner Struktur jedoch einheitliches und widerspruchsfreies Sprach-system, dessen räumlicher und zeitlicher Geltungsbereich gegen Null strebt’, in ‘Der Ursprung des germanischen ē ² und die Reduplikationspräterita’, Phonetica 1.160 (1957).

⁶ Cf. Lehmann 73–4; Hall 5–20; J. W. Marchand, ‘Was there ever a uniform Proto-Indo-European?’, Orbis 4.428–31 (1955); H. Penzl, ‘Zur Vorgeschichte von westsächsisch ǣ und zur Methode des Rekonstruierens’, Wiener Beiträge zur englischen Philologie 66.168–9 (1958).

⁷ In his study in VGJa., Makaev proposed the term ‘Proto-Germanic’ to designate the period from the dissolution of Indo-European to the establishment of a Germanic linguistic community, which was then followed by ‘Common Germanic’, the period from the establishment of this community to the appearance of individual dialect groups. In his contribution to the first volume of the Sravnitel'naja grammatika, Makaev gives the term ‘Common Germanic’ a new definition, according to which it includes ‘Early Germanic’ (comparable to his former ‘Proto-Germanic’) as well as ‘Late Germanic’ (comparable to his former ‘Common Germanic’). The need for a certain amount of ‘terminological hygiene’ (in Pulgram's words, 424 fn. 9) becomes abundantly clear from the problems faced by H. Fromm, ‘Die ältesten germanischen Lehnwörter im Finnischen’, Zeitschrift für deutsches Altertum 88.94–101 (1957–8).

⁸ From a purely objective point of view, there is little reason to prefer either ‘Proto-Germanic’ or ‘Common Germanic’ to designate the period of unity reconstructible for all the later dialects from internal evidence. The term ‘Common Germanic’, however, has been used in so many different interpretations, ranging from the period after the dissolution of Proto-Indo-European to features ‘common’ to the later dialects, that one must agree with Lehmann (67 fn. 2) that the term ought to be discarded entirely.

⁹ I am indebted to Charles F. Hockett for calling to my attention the important point that Pro to-Germanic must be reconstructed from internal Germanic evidence. The definitions by van Coetsem and Makaev, of course, do not meet this requirement.

¹⁰ J Fourquet, ‘Die Nachwirkungen der ersten und der zweiten Lautverschiebung’, Zeitschrift für Mundartforschung 22.33 (1954).

¹¹ Makaev's objections to Lehmann's definition of Proto-Germanic (SG 116–7) are based on the assumption that Lehmann denies the existence of a period of unity before the onset of Proto-Germanic. Such a period, however, is implicit in Lehmann's discussion of developments in pre-Germanic (74). Cf. H. Kufner's review of SG I in Lg. 39.509 (1963).

¹² In ‘The Proto-Norse vowel system and the younger fuþark’, Scandinavian studies 35.207 fn. 30 (1963), I rejected Lehmann's demarcation of the end of this period because I interpreted ‘the loss of /e a/ when final and weakly stressed’ to refer to absolute final position and it did not seem possible to fix the time of the loss of these vowels (cf. Makaev, SG 117). In the meantime, it has become clear to me that Lehmann was referring to final syllables and not to absolute final position exclusively, making it possible to equate the loss of /e a/ with the criterion I proposed as a substitute, the phonemicization of [o] > /o/. In addition, I believe I have now found internal evidence for the retention of PIE /e a o/ even in absolute final position into Proto-Germanic (see fn. 56 below).

^12a E. Hamp, ‘Final syllables in Germanic and the Scandinavian accent system’, Studia linguistica 13.29–48 (1959), also sees a possible correlation between the Germanic accent shift and a shift in junctural phenomena. According to Hamp, a pre-Germanic ‘secondary accent developed (or continued?), at first as an allophone of juncture, on those post-junctural syllabics which did not themselves bear a [Proto-Indo-European] primary’, whereupon ‘By a shift (perhaps of junctural phenomena), the old secondary became primary, and the old primary continued as a secondary’ (34). The loss of open juncture between such syllables would, of course, result in the so-called ‘loss of autonomy’ of the syllable in Proto-Germanic. It is a well-known feature of umlaut assimilations that they never occur across open juncture, but they can and do occur when that juncture disappears, as in the case of Otfrid's meg ih, werf iz for mag + ih, warf + iz; see W. Braune, Althochdeutsche Grammatik ¹⁰ §26 Anm. 3 (Tübingen, 1961).

¹³ W. F. Twaddell, ‘A note on Old High German umlaut’, Monatshefte für deutschen Unterricht 30.177–81 (1938), also Lg. 24.139–51 (1948); H. Penzl, ‘Umlaut and secondary umlaut in Old High German’, Lg. 25.223–40 (1949), ‘Zur Entstehung des i-Umlauts im Nordgermanischen’, Arkiv för nordisk filologi 66.1–16 (1951), also Wiener Beiträge 66.158–69 (1958); J. Fourquet, ‘The two e's of Middle High German’, Word 8.122–35 (1952), and ‘Perspectives sur l'histoire du vocalisme allemand’, Travaux de l'Institut de Linguistique 1.109–24 (1956); M. I. Steblin-Kamenskij, ‘Concerning the three periods in the Scandinavian i-umlaut’, Arkiv 74.105–11 (1959); C. F. Hockett, ‘The stressed syllabics of Old English’, Lg. 35.575–97 (1959).

¹⁴ See for example Twaddell, Lg. 24.143, and W. G. Moulton, ‘Zur Geschichte des deutschen Vokalsystems’, Beiträge zur Geschichte der deutschen Sprache und Literatur (Tübingen) 83.7–9 (1961), as well as §5 below.

¹⁵ Cf. L. L. Hammerich, ‘Über das Friesische’, Mélanges linguistiques offerts à M. Holger Pedersen 352 (Aarhus-Copenhagen, 1937), from Scandinavia southwards; E. Schwarz, Goten, Nordgermanen, Angelsachsen 261 (Bern-Munich, 1951), from the North Sea coast of Germany north, south, and west; and D. DeCamp, ‘The genesis of the Old English dialects’, Lg. 34.240 (1958), from ‘somewhere in southern Germany … northward to the Frisians … to Kent, and then, probably during the first half of the seventh century, spread throughout England’.

¹⁶ O. Höfler, ‘Stammbaumtheorie, Wellentheorie, Entwicklungstheorie’, Beiträge (Tüb.) 77.30–66, 424–76 (1955), 78.1–44 (1956).

¹⁷ It is primarily the fixation of the stress on the root syllable rather than its dynamic nature which should be noted here. Cf. A. Martinet's remark: ‘On se fait probablement une idée un peu trop simpliste et brutale de l'action d'un accent d'intensité. On y voit toujours une force aveugle à laquelle rien ne peut résister et qui saccage sans rémission les distinctions morphologiques ou lexicales qui ont le malheur de mettre en cause le vocalisme nonaccentué‘, Economie des changements phonétiques 169 (Bern, 1955). Martinet suggests the possibility that the stress accent in Germanic arose as the result of a transfer of energy from inflectional syllables to the root after the assimilation of the vowels in root syllables (umlaut) had rendered the endings functionally insignificant. Through this transfer of energy, the vowels of the inflectional syllables underwent syncope or apocope (199).

¹⁸ In spite of the arguments of such investigators as L. F. Leffler, ‘Bidrag till läran om i-omljudet’, Nordisk tidskrift for philologie og pædogogik NS 2.1–19, 146–80, 231–98, 299–320 (1875–6); A. Kock, ‘Der i-umlaut von e in den altnordischen sprachen’, Beiträge 27.166–90 (1902); and L. E. van Wijk, De klinkers der oergermaansche stamsyllaben in hun onderling verband 70–1 (Utrecht, 1936), some scholars still are reluctant to recognize the change of /e/ to /i/ as a form of i-umlaut; see e.g. K. Brunner, Die englische Sprache ² 1.242 (Tübingen 1960).

¹⁹ I have attempted to trace these developments for English and Icelandic in ‘Germanic umlaut anew’, Lg. 37.215–30 (1961), for East Norse in Scand. studies 35.195–207 (1963), and for German in ‘Zum Umlaut im Deutschen’, forthcoming in Beiträge (Tüb.) 86:2 (1964).

²⁰ A number of scholars present the Proto-Germanic short vowels in a rectangle with only two tongue heights, e.g. Makaev, VGJa 67, and Lehmann, ‘The conservatism of Germanic phonology’, Journal of English and Germanic philology 52.149 (1953). Such a presentation gives the impression of symmetry between the short- and long-vowel systems. It should be pointed out, however, that we must rely on internal evidence to determine the contrasts of the protolanguage. Since the spread vowels develop back allophones, the rounded vowels front allophones, and /a/ develops not only back and front allophones but also a central one, it is obvious that we are dealing with a short-vowel system which had the contrasts spread-rounded (with frontness and backness irrelevant) and three tongue heights (see Scand. studies 35.197–9). By the same token, in reconstructing the long-vowel system, we must posit two tongue heights with the contrast spread-rounded on the basis of the allophones which develop. It may well be that PIE /ā/ took up a position halfway between [ā] and [ō] during the course of its development toward PGmc. /ō/, but such a stage would have to be placed in pre-Germanic, since it is not reconstructible from purely Germanic evidence.

²¹ It should be stressed that the allophonic symbols in square brackets do not suggest absolute phonetic values. They do, however, suggest the relative values of the allophones with regard to the primary (nonassimilated) allophone of each phoneme. For typographical reasons, it has been necessary to substitute [ʌ] for the IPA symbol for the mid back spread vowel.

²² Concerning the later appearance of /ē²/, see §5.

²³ Cf. Twaddell, Lg. 24.139–51.

²⁴ Twaddell, Lg. 24.149.

²⁵ J. Kuryłowicz, ‘The Germanic vowel system’, Biuletyn Polskiego Toworzystwa Językoznawczego 11.54 (1952).

²⁶ J. W. Marchand, ‘Germanic *i and *e—two phonemes or one?‘, Lg. 33.346–54 (1957).

²⁷ Brunner 245; A. Noreen, Geschichte der nordischen Sprachen ³ §28c (Strassburg, 1913).

²⁸ Brunner 245; Noreen §29c; Lg. 37.225.

²⁹ Moulton 12; Scand. studies 35.197 fn. 8.

³⁰ See my discussion in Beiträge 86: 2 §6; cf. also Kuryłowicz 52 fn. 1, and M. Joesten, Untersuchungen zu ahd. (as.) ë, i vor u der Folgesilbe und zur 1, Pers, Sg. Präs. Ind. der starken e-Verben (Giessen, 1931).

³¹ In the case of the i-umlaut of PGmc. /eu/, we have an instance of the combined umlaut of the syllable peak /e/. The glide /u/ causes the retraction of /e/, while the /-i/ of the next syllable raises it, and the result is a higher back allophone of /e/, which I designate by . The diphthong is then realized as . Upon the attraction of the higher allophone of /e/ into the orbit of /i/, this higher back allophone also joins the /i/-phoneme, but since it is already a back variety (and continues to be conditioned by the glide /u/), it coalesces with the high back allophone [ɯ], producing the transition of > /iu/ = [ɯy]. See §7 below.

³² Lüdtke 171–2. A supposed South German origin was also proposed by L. Wolff, ‘Die Stellung des Altsächsischen’, ZfdA 71.129–54 (1934).

³³ Noreen §34.

³⁴ Here Lüdtke refers essentially to the list of Frankish names with e-spellings in Latin documents which O. Bremer presented in ‘Germanisches ē’, Beiträge 11.19–25 (1886). However, the research of S. Sonderegger, ‘Das Althochdeutsche der Vorakte der älteren St. Galler Urkunden: Ein Beitrag zum Problem der Urkundensprache in althochdeutscher Zeit’, ZfMaf 28.251–86 (1961), demonstrates the disparity between the traditional, official spellings of the documents and the actual forms of the spoken language, so that one must use utmost caution in employing such material for the purpose of establishing the chronology of phonological developments.

³⁵ Brunner 75 and 237.

³⁶ Penzl, Wiener Beiträge 66.158–69.

³⁷ W. Krause, Handbnch des Gotischen §61.5 (Munich, 1953).

³⁸ See S. Feist, Vergleichendes Wörterbuch der gotischen Sprache ³ s.v. stega and fidur-, fidwor (Leiden, 1939). The y of fyder can be considered the result of an i-umlaut if one posits PGmc. */feduri-,.

³⁹ Krause §62.3.

⁴⁰ Feist s.v. schuos and thurn. There are other Crimean Gothic forms which might be cited here (see Krause §22), but I have restricted myself to these, which, as Cowgill has suggested to me, are the only ones not likely to have been influenced by Dutch or German models.

⁴¹ Krause §55.2 and 22.

⁴² Cf. Kuryłowicz 50–2; Lüdtke 161–3; van Coetsem 35–40; and see fn. 49 below.

⁴³ See H. Moser, Deutsche Sprachgeschichte ⁴ 101 (Stuttgart, 1961).

⁴⁴ The chronological problem is analogous to the geographic one of establishing the boundary between High and Low German in the area along the Rhine (‘Rhenish Fan‘); see B. Martin, Die deutschen Mundarten ² 138–9 and 186 (Marburg, 1959).

⁴⁵ It is clear from the foregoing that I, following Lehmann, propose to define stages purely on the basis of structural data. The question of the point at which Gothic began its divergent development is not relevant to the issue at hand. The appearance of an independent daughter language cannot in itself determine the point at which we should consider the parent language to have ceased to exist. Such an argument would imply, for example, that Dutch must have ceased to exist, or at least to have ceased to be the ‘same’ Dutch as it formerly was, upon the appearance of Afrikaans. The family-tree theory encouraged such thinking by positing the sharp splits so justly condemned by Bloomfield. To be sure, the appearance of a daughter language affects the geographic distribution of the parent language, but not necessarily its structure.

⁴⁶ Concerning the consonantal system, see Lehmann, JEGP 52.140–52 (I have not included the laryngeals which Lehmann posits), and Moulton, ‘The stops and spirants of early Germanic’, Lg. 30.1–42 (1954). I am not completely convinced that it is necessary to posit /x/ and /h/ as separate phonemes, but the issue is not central to this paper.

⁴⁷ The development of these so-called diphthongs before their monophthongization in the various dialects or right down to the present day in High German and West Norse reveals that they were vowel clusters in which the constituents patterned in the same manner as the simple vowels (cf. van Coetsem 9). Since the first element of these clusters always represents the syllable peak in the periods under discussion here, I have not separately indicated the nonsyllabic nature of the second element. On the development of these clusters into Old High German, see Beiträge 86:2.

⁴⁸ See Moulton, Lg. 30.1–42.

⁴⁹ The retention of PIE /ey/ as a cluster into Proto-Germanic may be assumed on the basis of the following considerations. The forms Alateiviae in the inscription from Xanten and Teiva-varia(z) ‘Ziuverehrer, Beiname der Sueben’, cited by Rosenfeld from a gloss in the manuscript of the Wessobrunn Prayer, indicate it had not yet become /ī/; see W. Schulze, ‘Alaferhviae’, ZfdA 54.172–4 (1913), and H. Rosenfeld, ‘Die Inschrift des Helms von Negau’, ZfdA 86.258 (1955–6). The inscription on the Helmet from Negau is too controversial to be cited as evidence here; see R. Eggers, ‘Die Inschrift des Harigasthelmes’, Anzeiger der Österreichischen Akademie der Wissenschaften, Philologisch-historische Klasse 1959.79–91. The development of all the other Proto-Indo-European clusters parallels that of the single vowels, and there is no reason to assume that /ei/ constitutes an exception, since /e/ is retained in Proto-Germanic before /-i/ in following syllables, before nasal clusters and in certain weakly stressed syllables; see Lehmann, Lg. 37.71, and M. M. Guxman, ‘Sistem glasnyx fonem v germanskix jazykax’, Sravnitel'naja, grammatika germanskix jazykov 2.95 (Moscow, 1962). While we must assume that PIE /-e/ remained unchanged in absolute final position in view of the fact that not only [-i] and [-u] were retained but also /-a/ and /-o/ (see fn. 56 below), the presence of i-umlaut in forms with reflexes of PIE /-e/ before PGmc. /z/, e.g. OE menu < PGmc. */manniz/, indicates that in this environment /-e/ had become /-i/. It is also apparent from the presence of umlaut in the 2d and 3d sg. present indicative of strong verbs that there was an /-i/ in a following syllable in Proto-Germanic times. Whether this /-i/ was located in the third syllable, e.g. PGmc. */beudesi, -eþi/, or in the second, e.g. */beudis-, -iþ-/, cannot be determined on the basis of the umlauted root vowel (that an /-i/ of a third syllable could cause umlaut in the root is seen clearly from such forms as MHG mägede ‘maiden’ < OHG magadi < PGmc. */magudi-/; see Beiträge 86:2 §6). The problem deserves further investigation, but for the sake of brevity here, I cite the 2d and 3d sg. forms merely as root + /i/, e.g. PGmc. */beudi-/ ‘you, he bid(s)‘.

Van Coetsem has suggested that /ē²/ represents an a-umlaut of PGmc. /ei/ (35–40). His argumentation falters on the question of the analogical leveling of /ē²/ in favor of /ī/ in the infinitive and present participle of strong verbs of class I, as well as in the a-stem nouns and adjectives; see his ‘Zur Analogie im Germanischen’, ZfMaf. 29.216–27 (1962). It is indeed difficult to see how /ē³/ could have been removed without a trace in these forms. The situation is quite different, however, when we consider the development of /ei/ in the light of that of the other clusters. The glide /i/ has the effect of raising the syllable peak to [ẹ]. An /-a/ of the following syllable, however, affects the tongue height of the glide, lowering it to [į]. The cluster is thus realized in two variants, [ẹi] in most positions, but [ẹį] before /-a/ (cf. /eu, au, ai/ below). In the Proto-Germanic paradigm, then, we have the following forms: */beitan-/ = *[bẹįtan-] ‘to bite’, but */beiti-/ = *[bẹiti-] ‘you, he bite(s)’ and */beite/ = *[bẹite] ‘bite!'. When the reassignment of allophones takes place, the result is */bietan-/, but */biiti-/ and */biit/ (with /ii/ = /ī/; see Hamp, SL 13.32). The phonetic difference between /ie/ and /ii/ was, of course, extremely slight, as was the phonological burden of this contrast. Before retained /-a/ (as in the infinitive and the present participle) and before /-i/ (as in the 2d and 3d sg. pres. ind.), the contrast was neutralized. It was therefore significant only in those forms in which /-a/ or /-e/ had been lost, e.g. */biit/ ‘bite!’, but */hwiet-/ (< */hweita-/) ‘white'. That leveling should remove all traces of /ie/ from the verb paradigm is certainly not surprising, nor is it any more surprising that /ie/ should not have been able to maintain itself in the great majority of environments in forms not subject to analogical leveling. Where then do we find reflexes of [ẹį]? I believe we have them in those forms in Gothic and the other Germanic languages which have traditionally been considered to contain /ē²/. The influence of /r/ on the development of [ẹį] > /ē²/ is quite obvious, cf. OHG hiar, ziari, wiari, skiaro, fiara, and accounts for the low reflex in Goth. hēr and fēra. The presence of variant forms in different dialects, e.g. OE tīr, wīr, OIcel. tīrr, skīrr, and within the same dialect, e.g. OS hēr and hīr, need not surprise us either. Developments from original allophonic variants or from phonemes with neutralized contrasts are quite often inscrutable (cf. OHG wolf, OE wulf; OE spic, spec; OHG OE fisk; all a-stems). I also see no major obstacle to assuming that /ē²/ in the former reduplicating verbs represents a coalescence of the secondary clusters /ee/ and /ea/ which Lüdtke posits as an intermediate stage in North and West Germanic (164–5) with an already present /ē²/ < [ẹį]. According to this derivation, the rise of the prototype of /ē²/ as a conditioned variant would have to be placed at isogloss (1), the boundary between Pro to-Germanic proper and Late Proto-Germanic, while its emergence as a separate phoneme upon the loss of the conditioning /-a/ would coincide with isogloss (2), thus further strengthening the demarcation of the end of the Proto-Germanic period.

⁵⁰ This variant must be posited to account for the Old English development of /ai/ > /ā/; see fn. 56 below. I use the double symbol [ae æe] because it is not clear which of these approximate values should be assigned. It may well be that the actual value lay between the two.

⁵¹ Reflected in OIcel. ey, ONorw. φy, OGuthn. oy, and OSw. ODan. ; cf, Noreen §41c.

⁵² See Fromm 95–6; Guxman 93–4.

⁵³ See A. I. Smirnickij, ‘Otpadenie konečnogo z v zapadnogermanskix jazykax i izmenenie z v r‘, Trudy Instituta jazykoznanija AN SSSR 9.115–36 (1958).

⁵⁴ Although the geographically divergent distribution of the allophones of these phonemes may go back to Proto-Germanic times, it seems possible to posit for Northwest Germanic a pre-North-Germanic area in which /d/ was realized as [d] and [ð], and a pre-West-Germanic area in which the sole realization was [d]. Similarly, two areas of Northwest Germanic are defined by the realization of /b/ as [b] in pre-High-German, but as [b] and in the remainder of the community. The allophones of /g/, on the other hand, reveal a threefold division: [g] in pre-High-German, [ʒ] in pre-Low-German, and both [g] and [ʒ] in pre-English and pre-North-Germanic; see Moulton, Lg. 30.1–42.

⁵⁵ The opposition between NwGmc. /iu/ = [ɯy] and /eu/ ~ [ʌu] is neatly confirmed by the reflexes in Middle High German, where OHG /iü/ < NwGmc. /iu/ has not yet coalesced with OHG /iu/ < NwGmc. /eu/; see W. Braune, Althochdeutsche Grammatik ¹⁰ §49 (Tübingen, 1961), and Beiträge 86:2 §6.

⁵⁶ The positing of /ae/ and /ao/ for Northwest Germanic may seem somewhat schematic, but evidence for /ae/ is presented by the Old English reflex /ā/ < PGmc. /ai/ before /-a/ (contrasted with OE /ǣ ē/ < PGmc. /ai/ before /-i/). Compare PGmc. */daila-/ > NwGmc. */dael-/ > OE dāl ‘dole’, and PGmc. */daili-/ > NwGmc. */daili-/ > OE dǣl dēl ‘deal’; see F. Kluge, Etymologisches Wörterbuch der deutschen Sprache ¹⁷ s.v. Teil (Berlin, 1957). We must therefore conclude that OE wāt represents the normal development from PGmc. */waita/ > NwGmc. */waet/ ‘I know’, and we now have direct evidence for the retention of PIE /-a -o/ in absolute final position in Proto-Germanic, as Lehmann has suggested. Cf. Makaev, SG 117.

The fact that Old Frisian displays wēt for ‘I, he know(s)‘ may be interpreted to represent a different leveling in the paradigm, this time in favor of the 3d sg., PGmc. */waite/ > NwGmc. */wait/ = *[wæit], which we also find in all the other dialects: OHG weiz, OIcel. veit, OS OSw. ODan. wēt. The dialectal reflexes of NwGmc. /ae/ and /ai/ cannot, however, all be ascribed to leveling within paradigms, as can be seen from the forms of a-stem nouns: OE stān, sār (with the original distribution), but OIcel. steinn, sār, OHG stein, sēr, OFris. OSw. ODan. stēn, sār, OS stēn, sēr. It is obvious that secondary developments have obscured the original distribution in the latter dialects, leading to the coalescence of NwGmc. /ae/ and /ai/ in /ai/ in all positions in Old Saxon and apparently in Old High German (but note OHG /ē/, with the older spellings ae, ę, before /h, r, w/; Braune §43), whereas in Scandinavian and Old Frisian the lower cluster /ae/ is reflected in /ā/ before /h, r/ (Noreen §40b), and in Old Frisian also in open syllables before /w/; see W. Steller, Abriss der altfriesischen Grammatik §18 (Halle, 1928). A coalescence of these two clusters could hardly be surprising when one considers the slight phonetic difference between [ae æe] and [æi], the low yield of the contrast (until the loss of /-i/, the clusters contrasted only in forms formerly ending in /-a/ and /-e/), and the possibility of leveling in certain paradigms. We have many parallel developments in the later histories of the individual dialects, such as the eventual coalescence of NwGmc, /iu/ and /eu/ in New High German (see fn. 55 above), and in ODan. /ȳ/ (but OSw. /ȳ/ and /iū/; Noreen §160).

For the contrast between NwGmc./ao/ and /au/, the same conditions of phonetic similarity between [ɑo] and [ɑu], low yield, and the possibility of leveling (e.g. NwGmc. */baog/ < PGmc. */bauga/ ‘I bent’ and */baug/ < */bauge/ ‘he bent.’) have left little in the way of direct evidence. Nevertheless, I take the early Old English spellings aeo æo and aea æa for PGmc. /au/ (compared to iu and eu for PGmc. /eu/) to be an indication that in Old English the lower cluster won out over the higher, as was the case with NwGmc. /ae/ and /ai/ (Brunner 231–2). Similar evidence for a parallel development between NwGmc. /ae/ : /ai/ and /ao/ : /au/ may be seen in the fact that Scandinavian and Old High German have favored the higher cluster /au/ in most positions, but the lower one before /h r/; cf. OIcel. laukr < PGmc. */laukaz/ ‘leek’, but hōr < PGmc. */hauhaz/ ‘high’, and OHG lauh, but hōh, with the early Old High German spelling ao for later ō (see Braune §45, and Noreen §43).

⁵⁷ See Fromm 95–6; Guxman 93–4.

⁵⁸ Note that for Northwest Germanic we must posit distinct forms for the 1st and 3d sg. of strong verbs of classes I and II before the coalescence of the vowel clusters discussed in fn. 56. Thus, PGmc. */baita/ and */baite/ > NwGmc. */baet/ ‘I bit’ and */bait/ ‘he bit’; PGmc. */bauda/ and */baude/ > NwGmc. */baod/ ‘I bade’ and */baud/ ‘he bade’. The leveling of these contrasts in favor of the reflexes of /ae/ and /ao/ in pre-English, but generally in favor of /ai/ and /au/ in the other dialects, constitutes one of the isoglosses marking the transition from Northwest Germanic to the later dialects and therefore represents a dialectal variation which we can posit for Late Northwest Germanic.

⁵⁹ M. Adamus, ‘Mutual relations between Nordic and other Germanic dialects’, Germanica Wratislaviensia 7.115–58 (1962). Adamus speaks of ‘Common Germanic’ and ‘North-West Germanic’, the latter having ‘remained one single dialect for a fairly long time’ after the separation of Gothic (157). Another recent study by L. Rösler, Die Gliederung der germanischen Sprachen nach dem Zeugnis ihrer Flexionsformen (Nürnberg, 1962), presents a somewhat different view, but posits a long period of common development for Scandinavian and English. V. M. Žirmunskij, Vvedenie v sravnitel'no-istoričeskae izučenie germanskix jazykov (Moscow-Leningrad, 1964), demands that diachronic studies take full cognizance of extralinguistic evidence in tracing the development of the Germanic languages (7) and holds to the theory of a Gothic-Scandinavian linguistic community distinct from the West Germanic languages.

Since this paper is primarily concerned with the establishment of stages of development in prehistoric Germanic, I have made no attempt to give a complete picture of the dialect divisions within each stage. It might be noted here, however, that the isoglosses representing the different development of PGmc. /jj/ and /ww/ most likely demarcate a division of Pro to-Germanic into a pre-North- and East-Germanic dialect on the one hand and a pre-West-Germanic dialect on the other, while the same isoglosses continue to separate pre-North-Germanic from pre-West-Germanic in the Northwest Germanic period.