From the discussions in chapters 4 to 7 it is clear that the conventionally assumed PIE phonological system is incomplete as a source for the PGmc. phonological system. Most of the PGmc. developments have been explained on the basis of the conventional PIE system. But some wide-spread PGmc. developments remain obscure. To account for these I have assumed provisionally a modification of the PIE phonological system; the essence of this modification is the assumption of one or more laryngeals. With this assumption some unexplained phonological developments in PGmc. can be explained as developments from resonants and laryngeals; no other adequate explanation for them has been advanced.
The assumption of laryngeals or reflexes of laryngeals for conventional PIE is not essential because of the developments in any one dialect. Instead of revising the conventional PIE phonological system we may assume that the form of PIE from which PGmc. developed differed from conventional PIE; it may be an older stage of PIE than that from which PGk., PInd.-Ir., etc. developed. Neither alternative can be chosen on the basis of Gmc. evidence alone. Since our present Hittite material does not provide an answer, we can suggest one only by analyzing the reflexes of the laryngeals in other IE dialects and then reconstructing the phonological structure of PIE.
The investigations in chapters 4 to 7 were undertaken from the point of view of PGmc., with the assumption that at one stage of PIE the phonological system contained laryngeals. But I have not followed the current assumptions about the development of the laryngeals in the IE dialects. Those now held have been drawn largely from only a few IE dialects, especially (Vedic) Sanskrit and Greek, and from a preliminary, and consequently incomplete, investigation of these. One such is: (IHL 66) ‘Any laryngeal after a full-grade vowel and before a non-syllabic is lost in IE with lengthening of the preceding vowel.’ If one accepted this as a description of the treatment of laryngeals in that stage of PIE from which PGmc. developed, one would have to derive OHG nacho ‘boat’ from PIE *nāw-. No explanation for such a development has been given, and I fail to see any that might be at all plausible.
All of the Gmc. developments investigated above are found in the neighborhood of resonants. If for the time being, disregarding the current statements of IE phonology and of the laryngeal theory, we assume for PIE six resonants, each with three allophones, and two or more laryngeals, we can sum up as follows the PGmc. developments of : A. resonants; B. resonants in the neighborhood of laryngeals.
It is clear that the Gmc. developments can be explained more simply and completely if we assume that PGmc. developed from a form of PIE for which we assume six resonants and reflexes of two or more laryngeals.
Such a system will also help us better to explain the rather unusual developments of w and also j than would assumption of a compound reflex, e.g. hw, hj, of laryngeal and resonant in PGmc. For in Gmc. we can find patterns of phonological development similar to those of [w] and [y] in the neighborhood of laryngeals.
A pattern had been established for the interchange of velar stop and w in Gmc. A number of Gmc. words show an interchange of a velar stop g and w which developed from PIE labio-velars. The interchange is ascribed to the following vowel: velar stop before back vowels, resonant before front vowels; the interchange later was regularized in favor of either form. (CGG 72-4.) Examples are: Goth. magus, OIcel, mǫgr, OE mǣʒ, OS mago, OHG maga ‘boy, son’; Goth. mawi OIcel. mǽr, OE meowle ‘girl, maid’; Goth. hneiwan; OIcel. hníga, OE hnīʒan, OS OHG hnīgan ‘bow’.
This interchange of velar stop and w shows a twofold possible development from PGmc. gw. If somewhat similar clusters consisting of reflex of laryngeals and w had been maintained in Gmc., some of these may have developed to velar stop, others to a resonant in accordance with the pattern already established.
Lengthening caused by a following consonant is rather prominent in the Gmc. dialects though not attested in PIE. Some PGmc. pp tt kk are commonly ascribed to development from a sequence of single stop followed by n, e.g. OIcel, lokkr, OE loc(c), OHG loc ‘lock, curl’; compare OIcel. lykna ‘bend’. Furthermore, lengthening of consonants by a following consonant that later disappeared is very wide-spread in WGmc., though rare in NGmc. and not attested in Gothic. Therefore lengthening of j and w by a following reflex of laryngeal is quite in keeping with Gmc. developments.
It might not be wholly a matter of chance that lengthened w and j were preserved only in the WGmc. dialects, where lengthened consonants were relatively common. In OHG, for example, there are relatively many lengthened w's, some caused by following reflex of laryngeal, others by following j. These cannot be distinguished from one another, either by the orthography or by the later development. OHG tou, Mod. Germ. Tau ‘dew’ shows the development of lengthened w which was caused by reflex of laryngeal; OHG
The Gmc. developments which I have ascribed to preservation of reflexes of laryngeals are therefore in keeping with other Gmc. developments. We have further evidence pointing to the preservation in Gmc. of reflexes of laryngeals in the treatment of the so-called long vocalic resonants. This evidence, although negative, is convincing when found in support of the evidence cited above.
The Gmc. dialects, in contrast with other IE dialects such as Indo-Iranian and Greek, do not maintain the usually assumed PIE distinction between the short vocalic resonants [r̥ l̥ m̥ n̥] and the so-called long vocalic resonants [r̥̄ l̥̄ m̥̄ n̥̄]. We find the same development of n̥ in Goth.
Since chapter 7 has given us evidence that under some circumstances even reflexes of laryngeals were preserved in OHG, we have positive evidence, as well as evidence drawn from phonological developments, that in certain environments reflexes of laryngeals survived into Gmc.
We must therefore revise the phonological system that is assumed to be the source of PGmc. Since we cannot decide on the basis of PGmc. whether our reconstructions of PIE will have to be revised, or whether we will have to assume that PGmc. developed from an early stage of PIE, we must examine developments in extra-Gmc. dialects to find an answer. Before undertaking such investigations I shall test the conclusions given, by reexamining on the basis of a form of pre-Gmc. with reflexes of laryngeals one of the unsolved problems of Gmc. vowel development.
The phonemic system of this stage of Gmc. comprises four short vowels i e a u, four long vowels ī ē ō ū, three diphthongs ai eu au, and the consonants f þ x ƀ ð ʒ p t k s z h r l m n w j X1 X2. At least the chief allophones of these phonemes except X1 X2 are known; they may be assumed to be like the sounds symbolized by these letters in the notation of the International Phonetic Association; the allophones of the laryngeals will be discussed later.