Historical phonologies of Ilkorin, Telerin and Noldorin around 1923
Apr. 4th 2008
goldoth, guil(t), goluith/golthor, gelydh, goelaidh, goelidh, goeloeð, goeloeidh, geleidh, gelydh
|GL:13, PE13:117, PE13:120,145, SM:280,284,290, LR:377, WJ:364, PE17:139, Silm.|
After the writing of the Gnomish Lexicon (GL) and the Qenya Lexicon (QL) the next major stage of Tolkien’s linguistic creation was the Early Qenya Grammar (EQG) and the Early Noldorin Grammar (ENG) which he wrote some time around 1923 while he was at Leeds. The ENG is also accompanied by two compilations of vocabulary, the Noldorin Word-lists (NW) and the Noldorin Dictionary (ND). There is no such compilation for Qenya, although many Qenya cognates are listed in both NW and ND.
A new feature of this stage is the introduction of several minor languages: the Germanic-style Ilkorin and the Romance/Latin-style Telerin (beside the Celtic/Welsh-style Noldorin and the Finnish-style Qenya). There is also a mentioning of Doriathrin which is closely related to Ilkorin. These minor languages were hinted at in The Lost Tales and the Lexicons, but not actually developed. Since this stage of Qenya and Noldorin is commonly referred by the terms ’Early Qenya’ and ’Early Noldorin’, I will adopt the terms ’Early Telerin’ and ’Early Ilkorin’ whenever a fitting into the external development is required. Referring to the internal chronology the terms are ’Old Noldorin’, ’Old Ilkorin’ etc., according to Tolkien’s own terminology.
The EQG begins with some brief remarks on the historical phonology of Qenya and the Primitive Eldarin root system. The ENG begins with a discussion of Noldorin mutations. But as often, most details of the phonological development are left implicit within the accompanying wordlists. It should be also noted that the grammar clearly reflects an earlier conception, closer to Goldogrin – as it can be seen from retained initial cw- or the lack of lenition of m.
The aim of this article is therefore to give an overview of the sound changes from Primitive Eldarin towards Ilkorin, Telerin and Noldorin as they were conceived at that time. The corpora of Ilkorin, Telerin and Doriathrin will be listed completely, being sufficiently small (Ilkorin 25 words, Telerin 60 words, Doriathrin – only 2 words).
Citations from the main source Parma Eldalamberon #13 will be simply given by the page number in brackets. In addition the labels L and R will refer to the left or right column in the issue, since only the Noldorin entries are sorted alphabetically.
This diagram is drawn according to Tolkien’s account in PE14:60-62. I have tried to roughly account for the geographical distribution as well. It all begins with the march of the Elves towards Aman. Their language at that time is Primitive Eldarin. The main division is between Ilkorin in Middle-earth and Kor-Eldarin in Aman.
According to the mythological conception of that time there is an explicit connection between Arda and the ’real’ world. So Arda not only lies in Earth’s imagined past, but some Elves actually survive until today and live hidden in our world (all fictionally, of course). So Ilkorin splits up in many dialects and a form of it is said to be still spoken by Ilkorindi around Europe. Many human languages are also derived from a variety of Ilkorin. Doriathrin is not much different from Old Ilkorin and is spoken in Doriath under Thingol.
Kor-Eldarin yields Noldorin, Qenya and Telerin, the respective languages of the three Elvish tribes in Aman. Qenya and Inwian are not much different, the latter could be regarded as an elevated dialect (later Vanyarin Quenya). Tol-Eressean is spoken on the island of Tol-Eressea and derives from Old Qenya with influence of Telerin. A form of Telerin is also said to be spoken at the shores of England and Wales.
By the time the Noldoli (later Noldor) go into exile from Valinor they speak a slightly differing variety of Kor-Eldarin; and Noldorin is subsequently developed in Middle-earth. As Melko gains the upper hand during the war in Middle-earth the Noldoli are scattered and develop dialects according to their locations: Mithrim, Gondolin, Doriath, Nargothrond (Feanorian). Gondolic is the most archaic of them, not much different from Old Noldorin. Present-day Noldorin having been derived mainly from the dialect of Mithrim becomes a lingua franca for the Elves surviving in the lands of men.
An interesting question would be: Which languages are we actually dealing with – the historical or the present (20th century) Noldorin/Ilkorin/Telerin/Qenya? Tolkien imagined that an Anglo-Saxon mediator Eriol or Ælfwine was able to arrive at Tol-Eressea and brought the legends he learned there back to England. The title of the Gnomish Lexicon (see GL:3,6) suggests that it is supposed to be a compilation made by Eriol himself from what he had learned.
But nothing is stated about the internal history of the sources to be analyzed. We are probably dealing with the historical languages and their analysis founded upon the material that Eriol brought back with him. In the conception of the 30s, Ælfwine studied the Lhammas on Tol Eressea, an account of tongues made by Pengoloð (printed in The Lost Road).
The consonants of Primitive Eldarin are shown in PE14:63.
The evidence of primitive þ, đ is said to be derived from comparison with Ilkorin material, Qenya merges þ with s, and đ with r.
Primitive Eldarin probably had the five vowels a, e, i, o, u, long ā, ē, ī, ō, ū; and the schwa ə. With u̯ and ı̯ the diphthongs ai, ei, iı̯, oi, ui, au, eu, iu, ou, uu̯ were possible. Of them, iı̯ and uu̯ do not seem to appear in the original Eldarin roots, but they can appear in certain derivatives. For instance, -i- is a connecting vowel when a suffix beginning in ı̯- is appended to a long syllable. Qenya largely preserves this vowel system except for ei, ou (and iı̯, uu̯) (cf. PE14:41,71).
Apart from that, syllabic consonants ḷ, ṛ, ṣ seem to have been equivalent in use to the vowels (ṇ is also mentioned (PE14:56), but does not appear in any example) – this is a major difference from the later Common Eldarin system (as e.g. in The Etymologies). These syllabic consonants probably could also all be long, but only long ḷ and ṛ are attested for which Tolkien uses l:, r: or ḹ, ṝ.
See also the separate article about Early Ilkrorin by Helios De Rosario Martínez that treats many things in greater detail .
§ Noldorin Word-lists:
§ Noldorin Dictionary:
Ilkorin leaves the initial combinations sw-, st-, sr-, sl-, sm- unchanged; and probably also *sk-, *sp-, *sn-:
Medially, however, assimilation sr > rr is found in:
A striking change in the consonant system of Ilkorin is that of initial stops and spirants. Initial voiced stops become devoiced while voiceless stops are turned into spirants, hence:
§ *b-, d-, g- > *p-, t-, k-
§ p-, t-, k- > f-, þ-, (*χ- >) h-
In fact, the same sound change occurred between Primitive Indo-European and Proto-Germanic (Grimm’s Law). The unvoiced stops and spirants are shared by Germanic languages where other Indo-European languages usually keep voiced and voiceless stops. Compare for example Eng. call – Russian голос ’voice’ (< *gal-); Eng. tear, Gothic tagr, German Träne – Welsh deigr; Eng. field, German Feld – Russian поле, Latin planus; Eng. have, German haben – Latin capere ’seize’.
From this it can be seen that Early Ilkorin was imagined as a Germanic-style language, probably with a fictional influence upon the respective branch of Mannish tongues. See  for a more detailed comparison.
Furthermore we have evidence that the same sound change happened after nasals and r, l. Not all combinations are attested:
§ lb, ld, lg > *lp, *lt, lk
§ rb, rd, rg > *rp, *rt, rk
§ mb, nd, ŋg > *mp, *nt, ŋk
§ lp, lt, lk > lf, *lþ > -ld(?), lh
§ rp, rt, rk > *rf, *rþ, *rh
§ ŋk > *ŋχ > χ, nt > b (?) or *þ (?) [with change of preceding vowel – see discussion]
It is difficult to say whether the same is observed after vowels – *tagla > Old Ilkorin þakl ’axe’ (see 2.5) from EQG seems to indicate it. But in ND and NW only monosyllabic nouns are attested, as swada > swat ’bark’ (146L), dagā́ > tök ’high’ (161L). They may just show devoicing in final position – however, absent in smíg ’crumb’ (150L).
The apparent voicing tḷtā́ > þold may be due to a later change *lþ > ld (possibly only word-final). An irregularity appears to be Ilk. þah ’hush, be silent’ (142L) in relation to da’a, dā. The cognates N. daw, Q. tá, T. dā are more regularly derived. One may assume that Ilkorin develops -h from the glottal stop ’, but d > þ is unusual. Perhaps it is just derived from a hissing interjection associated with the order to be still.
In the case of ŋk the nasal is lost, probably with nasalization (or maybe just lengthening) of the preceding vowel. This nasalized ã or long ā then becomes ó. The same seems to happen in the case of hób ’a blow with an axe’; the primitive form for N. hant, T. scanta, Q. hanta is skantá, but since sk- would remain unchanged in Ilkorin, Tolkien put down the comment from k-- beside the Ilkorin form. This would lead to a reconstruction of *kant-, *kantá. However, the generation of -b in hób is somewhat peculiar phonologically, one would expect *kant- > *khanth- > *χãnþ > *hóþ. Actually, since Tolkien wrote the entries with a typewriter and had no key for þ, he for instance typed p and completed the character later by hand. Maybe it has just slipped his attention to change hób to *hóþ?
It should be noted that a similar development of ŋk is attested in later Sindarin, where nch > ch with lengthening of the preceding vowel (PE17:131,133).
However, more remarkable is that even though the later Ilkorin of The Etymologies is of a completely different style (it rather becomes Celtic-styled and very similar to Noldorin ), the same sound shifts seem to appear in the word Hwenti < kwendī (WJ:410) from one of the Avarin tongues. We see devoicing d > t after a nasal and probably initial spirantization kw- > *χw- > hw- (voiceless w [ʍ]) .
The earlier writings mainly focus on the Noldoli (Noldor) and there is not much description of the Ilkorindi who remained in Middle-earth. In The Lost Tales a Dark Elf called Nuin teaches the Ilkorin tongue to the first Men. Later on, however, the Sindar in Beleriand are a people very different from the other Dark Elves (Avari) which live further to the east and speak completely different languages. The Avari are those who first encounter Men and influence them in their speech. The Ilkorin language disappears altogether, but parts of it can be found in Sindarin and its dialects; and notions about Early Ilkorin seem to be reflected in Avarin Hwenti.
Other miscellaneous changes include:
§ -sy > ss
§ -ı̯ > -gg after the accent
The letters gg have probably indeed to be understood as the sound [gg], as the same sound change is attested for Old Norse, e.g. in the name *friı̯a- > Frigg . Other possibilities might involve [ʒ] (as in English ’azure’), [ʤ] or the velar fricative [γ], see  for more details. The position of the stressed vowel relative to ı̯ seems to be important, cf. the respective developments in Telerin 3.3.2 and Noldorin 4.1.2.
§ -kt- > -ht-
It would fit Tolkien’s general conceptions and spelling conventions if h represented a voiceless velar or uvular fricative (ach-Laut) ([x], [χ]) in -ht-, -lh, -Vh (where V is any vowel) and a voiceless glottal fricative (breath-h) initially in helh, hób.
In the development of the Ilkorin vowels one should at first point out ö in swöt ’parchment’ < swadwé (146L) and tök ’high’ < dagā́ (161L). The grapheme ö might stand for [œ], a rounded e, but it is difficult to explain how that sound would be derived from a in these particular examples and not in others.
But ö (or less ambiguously ǫ) is written in Old Norse for [ɔ], i.e. the short variant of the sound in English ’law’. If this is what was intended for Early Ilkorin, there are two possibilities of how this sound could have originated – maybe from long ā, as in the development of later Sindarin. But then swada > swat ’bark’ (146L) would have had a short vowel for some reason while swadwé > swöt ’parchment’ (146L) would have had a long one.
Another explanation may be the effect of u-mutation which is well-known in Old Norse. The vowel u adds roundness to all preceding vowels, hence a > ǫ, e > œ, i > y and o, u remain unchanged. We would therefore have swadwé > *swadu > swöt, but swada > swat whithout u. But then kelekwé > helh ’silver’ (140R) has to be explained by a early dropping of w before it could cause u-affection.
However, in the development of dagā́ > tök there is no apparent u either. However, swöt comes from the Noldorin Word-lists, where we find dagá > ta[k] ’high’ (141R) (with k lost at the margin) while tök is from the later Noldorin Dictionary. So maybe there was a change of conceptions – u-mutation in NW, *ā > ö in ND?
One sample of a-mutation (causing i > e) is clearly encountered (cf. the vowel mutation in Noldorin 4.2.1):
§ ei > í in:
§ possibly ē > *ī> i in:
All final vowels disappear in the known examples.
The attestations in this part are fragmentary, we only encounter short ḷ and long ṝ, so that the gaps are difficult to fill:
§ ḷ > ol
§ ṝ > ar
We are told in the discussion of the Elvish tongues in PE14:62 that Old Ilkorin is mainly the language of Doriath under Thingol preserved in records brought to Tol Eressea by Elwing and fugitives from Sirion, and in later days also recovered from the Thousand Caves. (PE14:62)
This idea of a very archaic speech being preserved in Doriath is essentially the same in all later writings except that it becomes an archaic dialect of Sindarin. Only one word is actually mentioned:
The only thing we get to know from this, is that Doriathrin shifts -Cl > -Col (where C is probably any consonant and -l might be syllabic), so it is not wholly pure Old Ilkorin after all.
It is somewhat peculiar that Tolkien kept the CE form dagla with initial d-, although the Ilkorin forms should derive from *tagla (as also their cognates, T. tagula << dagula, N. *tael, i·dael << *dael, i·dhael). The whole passage deals with possible stem variations in Primitive Eldarin as dag-, gak-, tag-, dı̯ag-, du̯ag-, ndag-, stag-, dagda- and what they yield in the respective languages. Maybe Tolkien has just overlooked to change dagla >> *tagla after he had changed the derivatives.
We find another word in NW, but in very faint writing, according to the editors:
With the preceding example we might have expected *cathl > *cathol.
§ Noldorin Word-lists
§ Noldorin Dictionary
§ Early Qenya Grammar
Tolkien’s transcription of Telerin seems to be influenced by Latin (according to the style of the language):
Long vowels are usually marked by acutes except for a macron in ūru. Compare .
Telerin leaves initial st-, sk-, *sp- unchanged as Ilkorin, but sl-, sr-, sm-, sn- > l-, *r-, m-, *n-; and sw- becomes su-:
§ ng-, mb-, nd- > g-, m-,*n-/*d-(?)
§ ku̯ > p as in Noldorin, but also gu̯, du̯ > b and hence probably *tw > p (unlike Noldorin, cf. 4.1.2):
But w disappears in narge ’pain’ < nr:gwé (150R).
The development of non-syllabic ı̯ or y is quite remarkable, as it seems to depend on stress for all the Elvish tongues treated here, and various consonants are yielded (cf. Noldorin 4.1.2).
In Telerin y > r after the stressed vowel, y > i (from non-syllabic to syllabic) before the stressed vowel:
§ -ky- > -ch- (= χ)
§ dı̯- > j-
There are several ways to explain alchíne – it might be a general shift before a front vowel ki > *chi, spirantization lk > lch (seems not very likely because of preserved -lp- in telpe) or simply an analogical formation from the verb base alacha.
As far as it can be figured out from the scarce evidence, Telerin seems to restrict the allowed final consonant clusters much like Qenya. At least we see that final -t > -s; and final -d seems to become devoiced, also merging with -s. Final -rn > -n. The examples below are however the only ones with a final consonant among predominant final vowels:
In NW, however, we encounter págant ’stern’ (152L) with final -nt and págas ’aft (on a ship)’ (152L) whose etymology is unclear. But they are mentioned as having been borrowed by Noldorin as poi and poiant respectively. The time of adoption must have been somewhere between Kor-Eldarin and Old Noldorin, i.e. after p > h, but before the vocalization of g (see 4.1.1 and 4.4). So págant, págas are likely older Telerin words.
EQG equals N. caifr ’flea’ < kamp’rū with T. camparon ’flea’. It may be that the Primitive Eldarin form is *kamparon, so preserved in Telerin; but Noldorin loses final -on > -ū.
In general there is not much change or assimilation seen among the Telerin consonants. It is the only one of the Elvish languages which allows intervocalic b, d, g and preserves original combinations like rg, lg. Some notable shifts include:
§ kt > tt
§ sk > x (= ks)
§ dl > ll
Vocalization of g is seen in gd (hence, such clusters of voiced stops do not seem to be allowed):
There is a sole example of rhotacism medially between vowels; it is not clear whether this is connected with stress:
Telerin vowels experience as little change as the consonants. Final vowels do not disappear, but become short; and syncope is sparse. No diphthongization is observed, but a notable tendency of monophthongization.
§ ei > í, i
§ eu > u
§ ou > ū
The only retained diphthongs are thus ai and au. No examples of oi, ui, iu is found.
With more attestations the pattern of syllabic consonants in Telerin is clearer than in Ilkorin.
§ short ḷ, ṛ > il, ir in NW
It seems that ḷ, ṛ > ol, or if before the accent (or there is a conceptional change from NW to ND):
§ ṝ, ḹ > ar, *al
Noldorin is a language profoundly based on Welsh and this influence is the greatest in this earliest design of it. So it is often very helpful to compare the Noldorin developments with those of Welsh. Tolkien first bought A Welsh Grammar, Historical and Comparative by John Morris-Jones in 1914; and I will provide some examples from this book.
Noldorin turns sp-, st-, sk- into the spirants f-, th-, h-; but sm-, sn- > m-, n-:
Mostly initial s- > h-, although s- is often retained:
The same variation is found in Welsh, compare W. hafal ’like’ with Latin similis, W. ham ’summer’ with Irish sam; on the other hand W. saith ’seven’, Latin septem.
§ sl- > lh- and probably sr- > *rh-; but sw- > f-:
A characteristic Noldorin change is that of initial p- > h- and similarly pl-, pr- > lh-, rh- (voiceless l, r). This might have been due to ku̯ > p (see 4.1.2) causing original ku̯ to merge with original p. To prevent the merging, original p would become aspirated at the same time via a chain shift, later a bilabial spirant and ultimately h-.
EQG mentions an early Noldorin divergence in Kor-Eldarin regarding the treatment of p, ú before the Flight of the Noldoli (PE14:61) and seems to refer to this change. This fits well with the adaptations of T. págas, págant (152L) that must have taken place while the Noldoli were still in Aman. So at first p becomes ph or h in the Noldorin dialect of Kor-Eldarin; after that T. págas, págant are adopted and become poi, poiant without further change of initial p-.
Initial l-, r- usually become unvoiced to lh-, rh-. This change is however as inconsistent as s- > h-, so that rán ’moon’, loloth ’poplar-tree’, luith ’magic, spell’ retain the voiced liquids (for luith a side form lhuth is given) in NW. The few entries in ND list luith ’magic, spell’, pl. luithar, but lung, lhung ’heavy’, pl. lhyng (only lhung ’heavy’ in NW).
§ mb-, nd-, ng- > b-, d-, g-:
§ ʒ- > g-:
ku̯ > p as in Telerin (see 3.3.2) with possible further lenition or spirantization; but gw is retained:
If y follows r, *l or i and comes after the stressed vowel, it becomes dh [ð]; no i-mutation is caused. Being before the stressed vowel, it causes i-mutation and is dropped or remains as part of a diphthong (see 4.2.1 for vowel mutations). Compare the Telerin development 3.3.2.
After the stressed vowel:
Before the stressed vowel:
However, y > ð does not happen after nd:
Stressed éy becomes ai in NW:
It may in fact be that the plural suffix -ir (cf. p.123) derives from -iya following the stress, as dágniya > deinir (141R), dagniı̯ > deinir (161L), pl. of dain ’height, summit’ suggests. Perhaps it matters that miníya > minedh has the stressed vowel in the adjacent syllable, hence maybe -íya > -edh, but - ́iya > -ir.
Compare also the identical Welsh development ı̯ > dd [ð] in iı̯, eı̯, aı̯ or after r, l if stressed or following the accent: W. arddhaf ’I plough’, Gothic arjan ’to plough’; Iu̯ér-ı̯on- > W. Iwerddon ’Ireland’, but *treı̯és > tri ’three’.
In initial combinations Cy- the glide is dropped:
After a vowel and before a consonant (mostly t) the stops p, k, *g are vocalized to ı̯, probably via becoming spirants *f, *χ, *ʒ. Thereby t becomes a spirant, hence Vkt > *Vχþ > Vı̯þ. It may be that p at first becomes k, so Vpt > *Vkt > Vı̯þ.
For ps, ks, kt spirantization to f, ch, th seems to be a competing change:
The nasals n, m are vocalized in the attested combinations -Vmpr- > -Vı̯fr-, -Vnkr- > -Vı̯χr- and likely in similar contacts of that kind:
The sibilant s may be vocalized as well, there is only one example before a consonant, where it changes to u̯. Alternatively, s is lost and the preceding vowel receives compensatory lengthening with subsequent ā > au:
Apparently es > eı̯, but otherwise s > h between vowels with further contraction:
Note however, that initial s is not affected by mutation according to the grammar (121). But neither is it in Welsh, where similar changes occur, e.g. esāk- > ehawc > W. eog ’salmon’, Latin esox.
§ -g- > -ı̯- medially in various positions:
Often g comes to stand in final position and probably becomes the spirant ʒ by lenition. After r, l it is vocalized to sg. -a, pl. -y and once *-rgw > -rw:
For an explanation one should once again compare with Welsh. There, final -g becomes non-syllabic y̯ towards Medieval Welsh. It is then assimilated in the North -ly > -l-l > -l and is lowered to -a in the South. So Medieval W. boly > W. bol, bola ’bag, belly’, pl. boliau.
If Noldorin follows this closely, we could have ONo. garg > *gary̯ > gara in the singular, but in the plural -y̯ becomes syllabic > gery.
After a vowel we find -ig > -iw; -eg > -e/-é; -ag > -á; -og > -ó
It appears that final -w or -u̯ is vocalized to -u, but is still written -w:
Another tendency for -w is to become a full spirant bh, mh (bilabial v [β]):
There seems to be evidence that final -n is lost in Noldorin. EQG equals N. caifr < kamp’rū with T. camparon without a comment. Maybe we have Primitive Eldarin *kamparon > kamp’rū > caifr in Noldorin (possibly at first -on > *-õ, a nasalized o) and > camparon in Telerin. In NW and ND we find andond- > N. ann ’door’, pl. ennyn (Q. andon, pl. andondi) (137R, 160L). On the other hand N. ailin ’lake’ corresponds to Q. ailin (136L, 158L); N. lhinn < pilind- to Q. pilin, making the picture less clear. Perhaps N. ann is rather derived from *annon with lost -n, while -nd > N. -nn.
At least, in the later Etymologies ai-lin- indeed yields N. oel rather than *oelin (Q. ailin).
§ -dt > -th
§ -dn > -n
Medially nc, nt, mp > nχ, *nþ, *mph > ng, nn, (*mm >) m in all positions, e.g.:
However, -nt- retained in hont ’trumpet, a trumpeting noise’, pl. hontath (analogical) (163L).
§ possibly -nl-, -rl- > -(n)lh-, -rlh- and -nr-, -lr- > *-nrh-, *-lrh-
Actually it appears that initial lh- is unchanged whenever in a compound after r, n. But since it derives initially from original voiced l-, it is likely that there is in fact the same change to lh medially unless the examples are all formed by analogy:
§ pp, tt, kk > *f, th, ch
§ ld > ll medially and finally, but occasionally -ld is devoiced to -lt:
After r, l and vowels, m apparently becomes a bilabial spirant [β] which is spelled mh or bh; and often further to v (now probably labio-dental). Note that m is not affected by lenition according to the grammar (121).
Somewhat unclear remains Gormagli ’Great Bear’ without mutation compared to Mornvegil ’Black Sword’ (149R).
The labial v (bh, mh) disappears after u and before a consonant; or after u in final position. Intervocalic v as well as -vr-, -vl- may be retained.
Even p (or lenited b) is seen vanishing after u:
Essential for the understanding of Noldorin vowels is the a-mutation and the i-mutation, in Early Noldorin they are slightly different from later Noldorin and Sindarin (discussed in detail in ). Especially the changes by i-mutation are very much apparent in the pluralization of nouns .
A-mutation is not only triggered by a word-final -a. This can be seen by the fact that the prefix ur- ’without, -less’ becomes or- when combined with a root vowel a, hence e.g. urfuin ’nightless’, urguil ’lifeless’, but ormast ’breadless’, orvab ’handless’ etc. (155R,156L) This conception was however preceded by the prefix um- which does not show mutation to either o or y (155L).
Rarely, a-mutation is totally absent: *kwissa- > pis ’whispers’ (for *pes(s)), 3rd masc. pisog (152L).
As it can be seen there are a lot of possibilities in the case of i-mutation, especially in the case of o, so that some explanation will be necessary.
The distribution of ei | ai is discussed in  with almost all the relevant examples listed: ai appears in the ultimate syllable and in monosyllables; otherwise ei. If a stands before r in the ultimate syllable, it mutates to e, ei and ei is preserved.
The vowel o probably at first mutates to oei | oe ([œı̯], [œ]). A later change oei, oe > ei, e causes it to merge with e(i) < e, a.
The mutation o > i, y may be compared to a similar Welsh change o > y, e.g. W. ystyr ’meaning’ < Latin historia, W. agor- ’open, expand’, 3rd sg. egyr. In Welsh, [y] changes to [ɨ] or [ə], but is still spelled y. It is unlear if y is merely orthographical in Noldorin as well.
From the Gnomish Lexicon Slips we get a hint that there was a first wave of i-affection responsible for fronting only: ornei > urnī > yrn (116), hence o > u and e > i (as proposed for later Noldorin and Sindarin ). If so, the pattern o > i, ui | y would naturally result because o merges with original u. To test this, an example of u > i would be required, but is not attested in plural formations. However, the 3rd sg. aorist form of dadnú *’sinks down’ is †dadní *’has sunken down’ (164R), in NW the past tense of *nuv- is ní (151L) and in the grammar 3rd sg. the aorist of *luv- ’wash’ is lhîf, lhîw *’has washed’. They are probably formed by i-mutation from an older ending -ı̯e .
The rare change o > y in the non-ultimate could thus be explained by a rare fronting of all vowels, loloth > pl. *luluthi > lylyth; *otok > pl. *utuki > ydig, while fronting only of the immediately preceding vowel is more common and leads to o > oe | e: *otok > *otuki > edyg.
This fronting would also have its exceptions, namely the patterns o > oi | ai, o > oei, oe | ei, e and e > ei | ai where it does not appear. But since the former two never occur by pluralization and ei | ai by pluralization is found only once (awest > ewaist), it might be triggered by a final -ī only. Also in the case of rotya- > rhoidia, mburyā́ > boir, boer a-mutation may have cancelled the effect of raising, if it took place later.
See also 4.2.5
Note that ı̯ > dh 4.1.2 occurs before i-mutation. Hence mbúryā > bordh has only a-mutation, but mburyā́ > *mborya > boir, boer (139R).
Diphthongs are rarely affected by vowel mutation, in a few cases au > oi is observed (in fact, as in German):
When au results from o it may be treated as original o, probably by analogy:
Note that the mutation oi > ui in the plural is probably only superficial and may be explained by the mechanisms outlined above, as all the examples show vocalized consonants:
§ o > u before nasals
This was at least a tentative conception. The grammar section mentions that u (from ŏ + nasal, ŭ) mutates to ui. This change seems to appear in ND gunn ’dragon’ < *gondo, Q. kondo (162R) and must have happened before a-mutation which undoes the effect: u > o, e.g. gonnas ’dragon’s lair’, gronn < *g-rondā, Q. ronda for *grunn. Somewhat unclear is gonn ’stone, rock’ on the same page; Q. hond-, pl. hondi is given, we would expect underlying *ʒondā.
In NW on the other hand we meet cunn ’dragon’, cunnas ’dragon’s lair’ (141L). Although it may be explained by o > u taking place after a-mutation (so that *kondas > *cundas > connas), an easier explanation would probably be that a-mutation is missing in cunnas and o > u does not take place before nasals anymore, cf. brond ’firm’ (140L), boron ’steadfast’ (139R) < *borond- (cf. Q. VORO- (QL:102)), gond, gonn ’rock’ (145L) < *ʒond-, thrond ’sky lord’ (154L) < *þorornd- (cf. Q. SORO (QL:86)) etc.
§ e > i might be an occasional change, possibly only before *r, l and nasals:
§ ā > au
§ ē > í
There seem to be no clear examples for long ī.
§ ō > ú (122)
§ ū > í (122) (probably via [y])
Compare the same change in Welsh: *cū ’dog’ > W. ci (cī), *rūn > W. rhin (rhīn) ’secret’, cf. Old Norse rūn, English rune.
However, it is unclear how túr ’power’ (154R) relates to this scheme, maybe it originally had a short vowel that was later lengthened. For several examples no etymology is given, like núd ’wet’, gú (145R). Also, N. Belaurin, Q. Palūrien (138R) seems to reflect the earlier Goldogrin shift ū > au  and its relation to the new ū > ī is also unclear.
Note that EQG mentions an early Noldorin divergence in Kor-Eldarin regarding the treatment of p, ú before the Flight of the Noldoli (PE14:61).
The diphthong ai changes to oi in Old Noldorin, but then back again > ai towards Noldorin proper, so that there is no change effectively:
Occasionally ai > e, as already met in Goldogrin with the same examples :
§ ei > ui (unstressed), ei > ai (stressed) in NW; but it seems that ei > ui generally in ND
Compare ei > wy ([uı̯] or [u̯i]) in Welsh, e.g. *u̯eid- > gwydd ’presence’.
§ oi > ui, but oi > ai if u̯ follows (and maybe if it precedes):
The same change oi > ai under influence of u̯ happened in Britonic, but ai > oe, oy towards Welsh, e.g. *gloi-u̯o-s > W. gloyw ’shiny, glossy’.
§ ui probably remains unchanged:
§ au is unchanged:
§ eu > ú
§ iw > i
§ ou > ú
§ ui | wi | y
ENG mentions that o mutates to ui, u, unaccented occasionally > wi (122). However, what can in fact be found, is a variation ui | y:
But wi > ui (although not unaccented) is probably seen in:
§ au | o
Although au remains unchanged, it later becomes o in the unstressed ultimate syllable:
On the other hand the alternation does not seem to be depending on stress in:
§ ai | ei
I-mutation and vocalization always lead to the diphthong ei that becomes ai in the ultimate syllable except if before r. See  for a discussion and a list of examples.
§ oi | ai | oe(i) | e(i)
In Old Noldorin i-mutation of o leads to oi beside oe(i) | e(i) and further oi > ai towards Noldorin (see 4.2.4). This sometimes still alternates with oi, probably by analogy from a basic form with unmutated vowels. Hence effectively the variation o > oi | ai:
Note that there is no such alternation when o > oi by vocalization since vocalization took place after Old Noldorin (at least as far as the conception in NW goes, see 4.4). Instead, oi is usually retained and in one example seems to become ui:
§ ai | e
Wa-ʒist > gwaist ’is aware, recognizes’ > past tense gwestaint might be a dissimilation from *gwaistaint; cf. occasional ai > e 4.2.4
§ wa | o
As in Goldogrin  (and in fact later Sindarin) the prefix gwá- ’together, co(n)-’ becomes go-’ if unstressed (146L, 162L).
§ ḷ, ṛ > li, ri with subsequent vowel mutation
§ ḹ, ṝ > al, ar
§ ṣ > is, *ṣ̄ > as
Throughout the wordlists Tolkien cites many historical Old Noldorin forms. They show us a preliminary stage in the development of Noldorin in some detail and allow us establish the order of some sound shifts, therefore a summary follows. The later Old Noldorin of The Etymologies may be compared .
Old Noldorin retains the combination nd (and most probably mb) that later changes to nn:
But initial combinations with s have probably already undergone the changes described in 4.1.1; one example is actually attested:
Here, hl may be a combination of h and l, later becoming unvoiced lh [ɬ]; or it is simply another indication of [ɬ]. Tolkien often used different spellings to create a visual separation of different languages (cf. ü below).
Similarly initial p- must have already changed to aspirated ph- or h- (see 4.1.1).
Initial nd- is still unchanged (most probably mb-, ng- as well):
It seems that there is yet no sophisticated system of mutations, cf. ONo. matgli ’honey-eater’ (149R) (later N. magli), but we see that aspiration in combinations of stops and nasals already takes place as well as m > mh, v:
The contact -k-t- appears as already vocalized ı̯þ in ND:
However, NW cite ONo. garg ’throat’, pl. geirg without any change of g. If vocalization took place after Old Noldorin, it would be a natural explanation of the diphthong oi in oif, oith (see 4.2.5). Possibly, Tolkien does not use ’ONo.’ to describe one exact point in the development of Noldorin and eith, heith have to be assumed to be later forms.
Primitive Eldarin ai > oi in Old Noldorin and again > ai in Noldorin, see 4.2.4 for examples.
Vowel mutation on the other hand is almost complete by the time of Old Noldorin, as one can see from nindyā > ONo. neinn > nainn ’aj. blue’ (164L). Mutated a in monosyllables and ultimate syllables always appears as ei, the change ei > ai happening towards later Noldorin (4.2.1, ):
The grapheme ü represents in several languages the close front rounded vowel, i.e. what in Noldorin is usually transcribed by y. So it seems that -w vocalizes in the plural to -y due to i-mutation in this example, although this is absent in deirw.
Note also that final vowels have already vanished in Old Noldorin.
update: Mar 4th 2009 reference added to Early Ilkorin Phonology]
update: Jun 15th 2009 development of ṣ and ei corrected
This document was translated from LATEX by HEVEA.