The disappearance of final vowels in Sindarin and Noldorin is extremely well known. It is an aspect of the phonetic development of nearly every Sindarin and Noldorin word. Based on evidence in The Etymologies, this final vowel loss took place in two phases. First, any primitive long final vowels were shortened in Old Sindarin or Old Noldorin, and then later the surviving short vowels vanished. The net result is that (nearly all) primitive final vowels (short or long) vanished in Sindarin, so that the majority of Sindarin words end in a consonant.

There are a couple of Sindarin/Noldorin sound changes that depend on original final vowels, namely:

• a-affection: short [i], [u] became [e], [o] preceding final [a]
• i-raising: short [e], [o] became [i], [u] in syllable before final [i]
• i-fronting: short [a], [o], [u] became [e], [œ], [y] preceding [i]

These sound changes would necessarily have occurred before the final vowel was lost, because otherwise the vowel triggering the change would already be gone. Further evidence that i-raising took place before final vowel loss can be seen in the phonetic development of Primitive Elvish words ending with the suffix -yā:

• ✶eryā > eir > S. air “lonely” (PE17/28).
• ✶elenyā > S. †elein “*of the stars” (PE17/139).
• ᴹ✶runya > rhoein [rhœin] > N. rhein “slot, spoor, track, footprint” (PE17/139).
• ᴹ✶ulyā > œil > N. eil “it is raining” (PE17/139).

If vowel loss was to occur before i-raising, then the y [j] > ı̯ would have become final in these examples. If so we would expect to see e > i instead of e > ei, and o > u > y instead of œ > œi > ei. It does seem there was some uncertainty on Tolkien’s part on the order of these sound changes, however, since we see an alternate etymology of S. air “lonely”:

• ✶eryā > erı̯a > erı̯ > irı̯ > S. îr (VT50/18).

Here i-raising took place after final vowel loss. This example seems to be the exception rather than the rule, however, and îr could represent a transient idea.

This same class of yā-words provides clear evidence that final vowel loss occurred before any i-intrusion, whereby a final [i] intruded into the preceding syllable. This is shown in the phonetic development of S. air above, and also in:

• ✶lisyā > liχı̯ā > leχı̯ > leich > S. laich (PE17/148).

This presents a problem regarding the timing of final vowel loss. In particular, if final vowels were generally lost first, then the primitive final -ī in most plurals would also have been lost before i-intrusion, but this is clearly not the case:

• ✶Abarī > †Evair, plural of S. Avar “Refuser” (WJ/380).

The likely phonetic development of this plural is: abarī > *everi > *eveir > evair. The Sindarin plural mutation of short a to ai in final syllables is well established, and is most easily explained by i-intrusion. Noldorin plurals are similar, but without the Sindarin change of ei > ai in final syllables:

• beir plural of N. bar “home” (PE22/36); likely *(m)barī > *beri > beir.

These obvious i-intrusions in Sindarin and Noldorin plurals present a puzzle for the timing of the loss of final i. If i was lost at the same time as other short vowels, these i-intrusive plurals could not have developed. One solution to the quandary, suggested to me in a private Discord chat by Elaran on 2018-08-25, is to assume that final i-loss took place later than other vowels, at the same time or shortly after i-intrusion. I find this idea very appealing, and adopted it wholesale. For symmetry, I also assume that u-loss was later as well (since it did not occur in all cases), and I discuss these two sound changes in a separate phonetic rule, whereby final [i], [u] generally vanished.

Conceptual Development: As discussed by Roman Rausch in his Historical Phonology of Goldogrin (HPG/§1.3), not all final vowels vanished in the Gnomish language of the 1910s. In a very early chart of Gnome Vowels, there is a somewhat cryptic statement under the chart saying “vanished all finally” which seems to imply all final vowels vanish (PE15/13). However, in the Gnome Grammar there appears the following statement (PE11/11):

For present purposes all those [nouns] that have lost their original final short vowels and come to end in a consonant are treated as consonant nouns, for any distinction has been obliterated by levelling of the “consonant-declension” endings.

• Declension A. contains those nouns that end now in a consonant.
• [Declension] B. nouns ending in a (from ā, ē, ō).
• [Declension] C. [nouns ending in] i, u.
• [Declension] D. monosyllables ending in a vowel.

A few pages later:

C. culu. brindi. urthu. gwilthi. Phonologically -u, -i only refer to [primitive] -ū, -ī (PE11/14).

Thus it seems that while primitive short vowels vanished in Gnomish, long vowels could survive in a shortened form, particularly -ā, -ī, -ū > -a, -i, -u. Based on the statements above, long final -ē, -ō developed into -a, though as pointed out by Roman Rausch (HPG/§1.3), there is evidence that -ē sometimes became -i: ᴱ✶Ou̯lē > G. Ôli, revised from Ôla (GL/62). Taken all together, it seems that in Gnomish short final vowels vanished first, and only later did long vowels shorten (sometimes changing in the process). This is the reverse of the order of these sound changes in later Noldorin and Sindarin, as pointed out in a post by Lokyt in the Aglardh forums on 2019-02-24.

Final vowels became much rarer in the Early Noldorin of the 1920s, but since we have fewer examples of Early Noldorin phonetic developments, it’s hard to say whether final vowel loss was completely universal at this stage. It does, however, seem to at least be fairly widespread in the 1920s:

• ᴱ✶aika > ᴱN. aig “high, steep” (PE13/158).
• ᴱ✶mburı̯ā́ > ᴱN. boir “heat, rage” (PE13/160).
• ᴱ✶ndŏre > ᴱN. dôr “land” (PE13/161).
• ᴱ✶wiqē > ᴱN. gwib “teors” (PE13/162).
• ᴱ✶loktu > ᴱN. luith “magic, spell” (PE13/149).

GS/§4.106 but apparently not always u: N. anw < ON. anu and N. guru < ON. nguru @@@