Inventory of vowels

In the Ilkorin and Doriathrin words we find ă, ĕ, ī˘, ō˘ and ö. The lack of long ā and ē, as well as of ū˘ (either short or long), which exist in all the other Elvish languages, is probably due to the limited corpus we are working with.

The phoneme represented by ö is uncertain. In German this character is used for the umlauted o, representing a fronted /o/ (i.e. /ø/). But in Ilkorin it only occurs as a reflex of a in certain circumstances (in complementary distribution with long ō; see below). Therefore, it may rather represent an intermediate vowel between /a/ and /o/: the open-mid back rounded /ɔ/, or the open back rounded /ɒ/. Tolkien seemed to refer to the latter when he wrote about the open a-like ǭ that intervened in the development of ā > North Sindarin ō, diphthongized in standard Sindarin as au (XI:400) — but reduced to ō˘ in Sindarin unstressed syllables (cf. Gilson and Welden, 1978:119).

Little can be said of diphthongs, which are only represented in the corpus by ai (/ai̯/) in stain. But again this is most probably due to the small corpus; note that the same is true of Telerin, which we know to have other diphthongs, at least au (/au̯/) in T. austa summer and aurina hot weather (PE13:137, 160). The only thing that can be said is that preh. ei (/ei̯/) was not preserved, but transformed into ī (see below).

Vowel loss

There are several vocalic changes. As in Noldorin, original final vowels are normally lost, but in Ilkorin this phenomenon is more generalized: there is no word ending in a vowel in the Ilkorin corpus. Another interesting feature is that disyllabic words, either original (like preh. terar) or resultant from the loss of final vowels (like *peles < pelesa), undergo a syncope of the second vowel (þerr < *ter’r, fels < *pel’s, etc.). If þilf is not the result of a metathesis from *þlif, we could assume that such a syncope could occur even if the vowel was stressed, since the preh. form of that word is t’lḗpe, with long and stressed ḗ. This favors monosyllables ending in a consonant cluster; most words of the corpus are in fact monosyllables, and many show that pattern due either to the described syncope or to an original consonant cluster after the root vowel. We cannot know how the vowels of prehistoric forms closed trisyllables or longer words developed in Ilkorin.

Vocalization of sonants

On the other hand, as happens in other languages, old sonants ḷ and ṛ (and probably ṇ too) normally resolved into vocalic syllables. But while ṝ became Ilk. ar as in other languages (see the row for kark), ḷ became ol (see the rows for molk and þold), although in other languages, at least in Q., it became il or ul, cf. PE12:10, P14:70. The sonant ḷ could occur in Old Ilkorin as a result of the final vowel loss, has happens in þakl, takl (i.e. þakḷ, takḷ); but the development in Doriathrin was the same: þacol, tacol.

A-mutation

Some Ilkorin words show vowel mutations, but not systematically. The first one to discuss is ă > ö, ō: in swöt < swadwē; tök < dagāˊ; *hōþ < kantā; þōh < *tankā˘. As commented on above in reference to the phonetic value of ö, this mutation may be compared with Sindarin ā > au, ō˘, which was already a feature of Noldorin in the phase we are studying (cp. N. môr cp. Q. mā‑ra in PE13:125, or daw < dā in table 1, among others), and even of earlier Gnomish (cp. Gn. Edhofon < Eđusmānī‑ in PE11:31, or Aulas < Ālasso in PE13:125). However, in the Ilkorin instances the original ă was always short (there is no instance of original long ā in non-final position, which we might use to compare its development).

The vowel length, however, should have changed in an intermediate phase, causing the distinct mutation to either ö or ō: we can see that in the ō-cases the vowel was originally followed by a nasal cluster, which was later transformed into a single fricative (cf. the section Shift of voiceless stops below). This could have triggered a compensatory lengthening, so that ă mutated to a long *ȫ, and eventually this could have yielded a close-mid ō, while short ö kept its open-mid or open articulation point.

This mutation does not happen in all cases. In other instances the original ă is kept (Ilk. ank < *angā˘; *tak < dagā; þah < da’a;19 swat < swada; þakḷ, takḷ < dagla; and Dor. cath < kasla or þacol, tacol < dagla). If we look for differences between the prehistoric forms that mutate and those that do not, we may observe that in the first group there are three prehistoric words ending in a long vowel and one uncertain case (*tankā˘), while in the second group there is another uncertain case (*angā˘), and five cases in which the prehistoric word ended in a short vowel, and only one long final vowel (dagā). Incidentally, the entry of NW bearing the latter case happens to have been rewritten in the ND, resulting in one of the cases in which there was mutation (tök < dagāˊ). This would be consistent with a mutation triggered if the final syllable was longer or received the stress.

Other vowel mutations

There are other vowels that mutate in certain cases. The apostrophe in preh. t’lḗpe and t’lēpe, whence N. tlī and tlub, respectively, represents a missing vowel in unstressed position, that was kept in the other cognates. Qenya and Telerin, languages in which medial vowel changes were less generalized, show the form telpe, pointing to an original e. In Ilk. þilf (both entries, although the f in the second one is not fully visible in the typescript — see note 16), this e would have been raised to i. This mutation could have been favored by the absence of stress in that syllable, as occasionally occurred in Qenya (cf. PE12:9). In fact the stress is explicitly marked on the medial syllable in t’lḗpe, and other indicia confirm that this should have been the accentuation in the Ilkorin development, too (cf. Shift of voiceless stops). Other Ilkorin words like fels, helh and þerr show the original e unchanged. This difference could be caused by the distinct accentuation; but the instances are too few and their prosodic features too obscure, so we cannot infer any certain hypothesis about the origin of this mutation.

Conversely, i > e in Ilk. seht < siktā, although the original i is kept in fiss and migg. And the original ō is kept in Ilk. snōr, the only instance with preh. ō˘ in non-final position. There is no instance with preh. ū˘.

Finally we may note that diphthongs can change, too: ei > ī in slīw < sleiwa and smīg < smeigē; but ai is kept in stain < stainā.

Inventory of consonants

Table 2 shows all the consonants found in Ilkorin and Doriathrin words, arranged according to their phonetic distribution. The letters of Ilkorin orthography are used, which normally coincide with the IPA symbols of the implied phonemes, except þ, which represents IPA /θ/, and h, which albeit normally used in Elvish languages for its IPA value (a glottal fricative), in this case probably represents a velar fricative, i.e. /x/, at least originally; cf. the section Shift of voiceless stops below. In addition, the phoneme implied by the geminated gg might differ from a mere long /ɡː/, as explained next. On the other hand, Doriathrin orthography departs from Ilkorin, spelling /k/ as c, and perhaps in one case /θ/ as th.

Some consonants are missing probably due to the paucity of the corpus. Thus, the general pattern of consonantal shift (cf. Ilkorin Sound Shift) would require the existence of p at least. We could likewise expect the existence of its voiced counterpart b, although it might occur less frequently, as seems to happen with the other voiced stops.

In migg perhaps we should read ‑gg as a long velar voiced fricative, or even an affricate (/ɣː/ or /ɡɣ/). A long stop would be unusual in final position, and the only similar case that I have found is Gn. bordd fire place, that can be compared to N. bordh heat, rage or borth / bordh hearth, where the final consonant is fricative (PE13:116, 139).20 We don’t have evidence of other voiced fricatives, and we cannot be certain about their existence. It has been noted above that Ilkorin accounted for the existence of ð (dental voiced fricative) in contrast to z in Eldarin, but this does not mean that ð or other voiced fricatives were kept in Ilkorin, since they are a group of consonants that tended to change or vanish in other Elvish languages.

Likewise, we need not expect any nasal, lateral or vibrant phoneme different from those that are found in the corpus, since they suffice for other languages, although /ł/ could occur in *calh, if that form is assumed for published cath (cf. One dubious case: Doriathrin cath). More uncertain is the case of the approximant /j/ (normally spelt y by Tolkien): we have its labial or labiovelar counterpart w, but there are other languages that have w, too, like Noldorin and its conceptual relatives (Gnomish or Sindarin), and still have no word with y. However these languages do have a very similar sound, if not the same, in i when it forms a rising diphthong, sometimes spelt with the glide i̯ (cf. the N. pl. suffix ‑i̯ath in PE13:123). So we cannot say with certainty that any other Eldarin language lacked it. Our corpus has preh. y in the ending ‑ye, which disappeared or merged with other Ilkorin consonants; but it could have been preserved in other phonological contexts.

We find no hint of the palatal nor the labiovelar series either (except the approximant w, that I have assimilated to the labial series in table 2); in fact it is possible that Ilkorin did not have them at all, or only had their approximants. In other languages like Noldorin or Telerin labiovelar stops were generally transformed into labials (see in table 1 N. celeb, T. celpe < kelekwē, N. peth word < qettā˘ in PE13:164, etc.), and palatal stops into velar (N. corn < *kyurna cheese, T. alacha < alakya‑ *shield, ward off, protect, PE13:140, 158). And we can see the same development for fricatives in Gn. fui dark, murk related to the Qenya root ǶUY˙U (PE12:41, where ƕ = /χʷ/), and Gn. hanna, related to Q. hyanda (PE11:48). Our corpus shows that at least the original /kʷ/ in kelekwē is lost in Ilk. helh, too, although the development is not as in Noldorin and Telerin. But again we lack more cases to confirm whether palatals or labiovelars were preserved or generated in other circumstances. For instance, although original labiovelar stops were transformed into labials in Noldorin (and probably the same happened with fricatives, as in Gnomish), this language did have /ɡʷ/, extended from original w‑: see the entries for gwadh bark, skin, peel, gwedhion husband, gweg man, etc., in PE13:146, 162.

Consonant clusters

All our Ilkorin words are closed monosyllables. Thus, we can tell something about initial and final clusters, but we have no information about clusters in medial position. However, all initial and final clusters would likely be allowed as intersyllabic clusters, plus other combinations that would not exist in extreme positions.

Initially only clusters with s‑ (sl‑, sm‑, sn‑, st‑, sw‑) are observed, but if these clusters, mostly lost in the other languages, are allowed, it is probable that others were possible, too. In final position we find geminated consonants (‑gg, ‑rr, ‑ss), several clusters of a liquid followed by stops or fricatives (‑rk, ‑lf, ‑ld, ‑ls, ‑lk, ‑lh), one nasal cluster with the homorganic voiceless stop (‑nk), and one fricative followed by a heterorganic stop (‑ht).

Many similar combinations could be allowed finally, but not every one. It seems that liquids cannot share their syllabic units with a preceding consonant, at least in Doriathrin, where preh. *‑sl is resolved as ‑th in kasla > cath,21 and ‑kl in Old Ilk. þakl, takl yields ‑col, seemingly through */kḷ/. Likewise, final nasals with homorganic fricatives are reduced into the fricative alone (cf. this matter in the next section).

Ilkorin Sound Shift

Now we come to one of the principal features of the language: the mutations of stop — and perhaps fricative — consonants. This phenomenon is strikingly similar to the erste Lautverschiebung or First Germanic Sound Shift, generally known as Grimm’s Law. So I have labeled it Ilkorin Sound Shift, although we will also see some elements of other distinct historical consonantal changes, typical of Germanic languages. Cf. Krahe and Meid, 1967–9.

Other Germanic-like consonant shifts

If we use Grimm’s Law as a model of the Ilkorin Sound Shift, we should infer a change from original voiced aspirates (/bʰ/, /dʰ/, /ɡʰ/, /ɡʷʰ/) into voiced fricatives, eventually merged with voiced stops. But the early treatises on Eldarin phonology (SQ and QP) do not reveal the existence of primitive aspirates, although there were voiced spirants (i.e. fricatives). The chart in QP (PE14:63) has /β/, /z/, /ʝ/, /ɣ/, and accounts for the existence of /ð/ as distinct from /z/, according to evidence of Ilkorin, although in other languages these two phonemes were conflated into one; SQ in PE12:15 has all these voiced fricatives (including the conflating pair /ð/ and /z/, although Ilkorin is not mentioned there), plus /v/ as an alternative to /β/, the palatalized alveolar /zʸ/ as an alternative to /ʝ/ and the labiovelar /ɣʷ/. I represent them here with their IPA values for the sake of clarity, because Tolkien did not follow a consistent convention in these writings.28

Whether these voiced fricatives merged with voiced stops as in PGmc., or experienced any other kind of shift, we cannot know, because we lack clear instances with prehistoric voiced fricatives. But there is yet another phonological development related to Grimm’s Law that can be discussed in relation to Elvish languages in general and Ilkorin in particular. OHG shows a zweite Lautverschiebung or Second Sound Shift, which transformed voiced stops into voiceless (like the First Shift), and voiceless stops into geminated fricative stops or affricates in certain circumstances. According to some scholars, this is an effect of a general tendency of Germanic languages to repeat consonantal phenomena over time; in fact, modern languages of this family like English aspirate the pronunciation of voiceless stops at the beginning of tonic syllables, which can be compared to their fricativization or affrication in the First and Second Shifts (Ramal, 1998). And it is interesting to see that the Ilkorin Sound Shift also maintains some resemblance to a root strengthening process described in QP:

[In medial position] Originally probably only the voiceless stops were geminated (or lengthened) kk, tt, pp. The equivalent modification in the case (i) of spirants, was to stop them or nasalise them, giving an alternation between ʒ, j, z, w [= /ɣ/, /ʝ/, /z/, /β/] and ŋ, ni̯, n, m and g, d, b;29 (ii) of voiced stops, to unvoice them, producing a variation between k, t, p and g, d, b. (…) This strengthening of spirant to voiced stop, and voiced stop to voiceless stop, also clearly took place initially. (PE14:64)

This phenomenon, which I will call Root Medial Strengthening (although it could also occur in initial position), cannot be classified as a sound shift, because it did not occur systematically, but just in certain cases to create new roots; and unlike the Ilkorin Sound Shift, it affected nasals and liquids, too. But it is interesting to see that its effect on voiced stops was identical to the Ilkorin Sound Shift, and that the gemination of voiceless stops in the Root Medial Strengthening is not very far from their spirantization under the Ilkorin Sound Shift: at least in Gnomish and Noldorin geminated stops eventually became fricatives; cf. Gn. laf cp. Q. lappa loose end (PE11:52), N. noth < notta number (PE13:151), or N. crech spittle < kә̀rekka, inter alia. Notably, in the Root Medial Strengthening we have two different developments of Eldarin voiced fricatives, which can be used as a basis for discussing what happened to them in the Ilkorin Sound Shift.

If we compare the Root Medial Strengthening with historical echoes of Grimm’s Law, we can also find a special resemblance with the Second Shift of High German: voiced stops were unvoiced in both phenomena, and the geminated fricative or affricate reflexes of voiceless stops in OHG may be compared with the gemination of voiceless stops in Eldarin. However, the Root Medial Strengthening cannot be related with the Ilkorin Sound Shift as the First and Second Germanic Sound Shifts at all: as noted above, the Root Medial Strengthening is not even a regular sound shift, and moreover it should have been older than the Ilkorin Sound Shift, which was applied to formed words, not to their roots. Nonetheless, the parallelism between their effects on voiced and voiceless stops is remarkable, as shown in table 3.

This repetition of consonant shifts also provides a possible explanation for the problematic þ‑ < d‑ in Ilk. þah hush, be silent < da’a, dā. First, we must consider that the prehistoric form that lies behind the Ilkorin one should have been da’a, and not dā; a development from the latter would have required a shortening of the vowel and the rising of a new final consonant from nil; and especially the latter would be a rare development. More precisely, the actual prehistoric word whence þah came should have been the full form of da’a, including the missing sound indicated by the apostrophe. The apostrophe is a common symbol for the glottal stop (/ʔ/), but Tolkien normally uses it to represent any primitive phoneme, either vocalic or consonantal, that has been lost. Now, in this case between two vowels it should have been some consonant, and although we do not have any explicit indication of its value in this case, in Gnomish and Noldorin the missing consonant indicated by an apostrophe is usually g, which frequently disappears in medial position in the Eldarin languages (cf. PE12:16). And if da’a actually represented *daga, under the Ilkorin Sound Shift it would have normally yielded *tak(a) (with or without the final vowel, depending on the chronology of the phonological phenomena), and a repetition of the consonant shift in both consonants plus the final vowel loss would eventually produce *tak(a) > þah.

Of course, it is possible that the apostrophe represents some other lost consonant.30 But the final step of the proposed theory, *tak(a) > þah is especially attractive, if we compare it with PIE. */tak‑/ be silent > Goth. þahaiþ is silent (Ringe, 2006:132), nearly identical in phonology and meaning to our case.31 Why this word would have undergone a double consonant shift cannot be said. Perhaps it was favored to prevent the coalescence with preh. dagá > *tak high, which has also been cited above. Or it could be just a dialectal development.

Development of y

There are other consonant changes that cannot be accounted for by Grimm’s Law, and thus do not fall under what I have called the Ilkorin Sound Shift and related phenomena. First is the change of the approximant y, found in preh. pisye > Ilk. fiss and preh. míye > Ilk. migg. In both cases, the ending ‑ye has disappeared, and the new final consonant in the Ilkorin word is long: ss in pisye > fiss; and gg in míye > migg (for which however there was no original consonant before original preh. ‑ye).

Each development resembles a distinct Germanic sound change. In West Germanic languages, all consonants except /r/ were geminated before /j/, and this /j/ could disappear depending on the language and its phonological context, just as happens here in pisye > fiss.32 And in Old Norse (which is not a West Germanic language, however), sometimes /j/ was lengthened to /jj/ according to Holtzmann’s Law (Prokosch, 1939: §33c),33 and eventually hardened to /gg/, as here in míye > migg.

Similar sound changes may account for the development of Ilk. fiss and migg. The ending ‑ye could correspond to a rising diphthong (*‑i̯e) in Ilkorin history: cp. ‑i̯ā˘ in prehistoric Gnomish or Noldorin words, cognate with Qenya forms with ‑yā˘: Q. murya close, muggy cp. mburi̯āˊ and mbúri̯a hot, raging (PE13:139, 160), Q. purya fire place cp. búri̯ā (PE13:116), venya womanly cp. u̯eni̯ā˘ (PE13:118), Q. minya slender cp. miníi̯ā (PE13:164), etc. The semivowel *‑i̯ (phonetically the approximant /j/) could then cause the lengthening of preceding sounds, as in West Germanic. But when a vowel preceded it may be that the /j/ acquired a consonantal quality (Tolkien would spell it ẏ, cf. PE12 8, 11–3) and lengthened itself, as in Holtzmann’s Law for Old Norse. Alternatively, it may be that if the preceding vowel was i, this vowel assimilated the consonantal quality (cp. PE12:12, where the change ‑ī‑ > iy is discussed for Qenya). Then ẏ changed to g, as in Old Norse and in some West Germanic languages. Ilkorin words then lost the entire prehistoric ending ‑ye, as would be consistent with its proposed diphthongal character if we assume that the loss of final vowels could affect both its segments, perhaps in two different stages.

Thus, the development of these words may have been as follows: *pisi̯e > *pissi̯e (lengthening) > fiss (regular Ilkorin changes — loss of final vowels and Ilkorin Sound Shift); and *míi̯e > *miẏi̯e (consonantalization of i) > *miẏẏi̯e (lengthening) > migg (loss of final vowels and ẏ > g). Or if Ilkorin behaved like Old Norse, *míi̯e > *miẏẏe (Holtzmann’s Law) > migg (as above).

On the other hand, the gg of migg may have arisen by a stress-dependent development of y, like that observed in other Elvish languages. In pre-tonic position, that sound is kept (although it may become vocalic), but in post-tonic position it is transformed into ð in Noldorin and into r in Telerin (Rausch, 2008). Likewise, Ilkorin could have developed pre-tonic *pisyé > *pisi̯é, etc. as above, and a distinct post-tonic míye > *migge > migg. Notice that both Noldorin midh and Telerin mire may share a common ancestor *miðe (through loss of the final vowel in Noldorin and rhotacism of ð in Telerin, cp. the Qenya development of this consonant in PE12:24). Therefore, it is possible that the same form was also an ancestor of Ilk. migg. The shift from dental to velar consonants is rare in Elvish languages, but possible in certain environments (cf. *eð > eg before l in PE11:31 s.v. edh).34 In this case the long quality of the new consonant can be compared with the long fricative dd and the affricate dð in analogous Gnomish developments: búri̯ā > bordd fire place, and talii̯ēii̯un > ‑eii̯un > ‑oiðun > ‑uiddhon *mockery (PE13:116). If this comparison is suitable, the phonetic value of ‑gg could be long g (/ɡː/), long fricative (/ɣː/), or even affricate (/ɡɣ/); see the Inventory of consonants above. But in any case it may be assumed to be a velar consonant.

Development of w

The other approximant, w, is in the initial cluster sw‑ of preh. swada and swadwē kept unchanged in Ilk. swat and swöt; and it also occurs after d in the second instance, where it has vanished. The Noldorin cognate fadhw reveals an earlier form *swadu̯ē in which the second w was interpreted as a semivowel that was fully vocalized independently of the previous consonant. In Ilkorin this diphthong *‑u̯ē vanished like *‑i̯e in the examples of the previous sections (without effect on the previous consonant), and the rest of the word was developed according to previously commented processes.

There is also preh. kelekwē > Ilk. helh, which could be analyzed just as the previous case, but note that the cognates of other languages, N. celeb, Q. telqe and T. celpe, imply that kw was in this case a single labiovelar consonant (/kʷ/, preserved in Qenya and changed to a labial in Noldorin and Telerin, as usual). So either this consonant changed its point of articulation in the Ilkorin development — at least when it was final (cf. Shift of voiceless stops above), or it developed as ku̯‑ only in Ilkorin.

Yet another conclusion is that if the palatal approximant y behaved like w in other positions than the final syllable, initial *sy‑ might be kept in Ilkorin, although we don’t have any actual case to confirm it.

One dubious case: Doriathrin cath

The only phonological change that has not been yet discussed is ‑sl > ‑th (= ‑þ) or *‑lh in Dor. cath or *calh (cf. note 2). If cath is the correct reading, the relation between the resulting th and the original s is clear (both are voiceless fricatives with a near articulation point, normally assimilated in Qenya); but we don’t exactly know how the l disappeared. Perhaps it was assimilated to the previous consonant in the cluster (> *‑st, *‑sþ?), before both consonants merged into one. However, we have no parallel case to compare.

On the other hand, *calh would be closer to the development observed in N. caul and Q. kalla, where the preh. l is retained (and in Qenya it is lengthened). This form could be derived from *kals(a) or *kalþ(a) by early metathesis of kasl(a) (before the loss of the final vowel, or before the sonant ḷ was expanded to ol). In that case it would likely represent */kał(ł)/ rather than */kalx/. Note that in medial position (although not finally) Southern Sindarin also had */łł/ (spelt lh, lth, lþ or ll) from a similar environment, cf. Hostetter (2003).