In Ancient Quenya, if two syllables had the same vowel twice and the second syllable was short (aCă, eCĕ, iCĭ, oCŏ, uCŭ), then the second vowel was frequently lost. This process is commonly referred to as the “Quenya syncope” and there are numerous examples of it:

• ✶glawarē > glaurē > ʒlaurē > alaurē “sheen of gold” (PE19/79).
• ✶nelede > neldë “three” (VT47/10).
• ✶min(i)k(e)we > minikwē > minquë “eleven” (VT48/7-8).
• ✶ñgolodō > Noldo (WJ/364).

The first example indicates this sound change was before voiced stops became spirants and thus occurred very early in Ancient Quenya. It isn’t always easy to tell when the Quenya syncope occurred versus cases where the vowel vanished (or didn’t exist at all) in Primitive Elvish, but the syncope can be confirmed by comparison with Sindarin. The second vowel is present in the Sindarin cognates of all of the above words: S. glawar, S. neledh, S. minib, S. Golodh. Tolkien described the Quenya syncope on numerous occasions:

You are of course right in seeing that the words for “silver” point to an orig. *kyelepē: Q. tyelpe (with regular syncope of the second e), S. celeb and Telerin telepi (Letter #347 to Richard Jeffrey, 1972, Let/426).

ala was an intensified and stressed form with stress on the first a, and consequent ancient shortening of *alā; al- was a reduction of this according to the Q. tendency to syncopate a short vowel after a stressed vowel of the same quality when the contact of consonants so produced was permitted in Q. (from notes on negation from the late 1960s, PE17/153).

In Quenya the second vowel was syncopated as usual with short unstressed vowels following a stressed vowel of the same quality: hence Q. kanta 4 < kanatā (from notes associated with The Rivers and Beacon-hills of Gondor from the end of the 1960s, VT42/24).

Quenya syncope operated in words of this pattern with 2 like vowels in sequence, unless the second was long or stressed or both in Eldarin. So ñgolodō: Q. ñoldo, S. golodh; but Orǭmē > Orome (from a draft of a letter to David Masson, 1955, PE17/153).

Perhaps the most detailed description of this phenomenon appears in the Outline of Phonetic Development (OP1) from the 1940s (PE19/58):

(1) In trisyllabic (or longer) words, but not in dissyllables; a short vowel of any quality, including ĭ, ŭ, was syncopated and lost before a single consonant, when preceded by a stressed syllable (main or secondary) that contained a vowel of the same quality: in other words there was a tendency to reduce surviving KALAT-forms to KALT(A)-forms. Thus *nélekī̀ "teeth" > nélkī; *ñgòlodṓ "Gnome" > ñgoldṓ; but *Òrōmḗ (name) remained.

This syncope occurred most regularly where the preceding syllable was short, and the intervening consonant one that (after syncope) would form with the next consonant a favoured group, sc. mostly before s, l, r, y, w and nasals. But it could also occur after a long vowel: as *tā̀rakā́ > tārkā́, tarkā́. And before stops and other non-sonant consonants: as *òkotā > oktā́.

Syncope of i, u occasionally occurs after diphthongs in ı̯, u̯ respectively, as áuluta- [>] áulta- > Q olta; Màilikṓ-r > Mailkṓ(r) > Melkor (cf. N Maeleg).

(2) After this syncope, but still before the Q. accent shift: Long vowels that remained in unstressed medial syllables were shortened, but not lost. So Òrōmḗ > Oromḗ (Q. Orome).

This section described in detail when the syncope did (and did not) occur:

1. The word had at least three syllables.
2. There were two sequential syllables with vowels of the same quality.
3. The second vowel was short and unstressed (but see below).
4. The resulting consonant cluster was a “permitted cluster”.

The quote from OP1 also indicates the syncope sporadically occurred when the initial vowel was long or part of a diphthong. For example, this could happen in the perfects of TALAT-stem verbs: atā̆latie > ataltie, (archaic) perfect of talta- “to slip down” (PE17/186). In the quote from OP1 Tolkien said the second vowel must be followed by a single consonant, so that VCVCV > VCCV. However, there are examples of the syncope occurring before a consonant pair ending in w or y:

• ✶min(i)k(e)we > minikwē > minquë “eleven” (VT48/7-8).
• ✶enetya > entya “central, middle” (VT41/16).

Clusters ending in y or w are the only permissible trisyllabic clusters in Quenya (PE19/82, item #iii.). Not every trisyllabic cluster ending in w or y was permissible: see the example atatya in the quote below (VT42/24) where the syncope was still inhibited.

The exacting timing of this sound change is hard to nail down. Some examples seem to indicate this sound change occurred quite early, but others seem to indicate a later syncope:

• ✶galadā > ʒalaðā > ʒalðā > alda “tree” (PE17/135).

This example indicates the syncope occurred after voiced stops became spirants, as opposed to the first example given at the beginning of in this entry, which shows a syncope before spirantalization. The relationship and timing between the syncope and stress patterns also isn’t entirely clear. In the OP1 quote above, Tolkien said that the syllable undergoing syncope must be both short and unstressed, and that the syncope took place before stress shifted forward in Quenya. Later, though, he said that the placement of primitive stress did not matter for the Quenya syncope:

Of a primitive atata the normal Quenya development was atta, while atatya remained because the second a was not syncopated, being in a long syllable. ... The placing of the accent would not affect Quenya since in PQ the accent became placed on the first syllable in all cases, except for words formed with still recognized prefixes (from notes associated with The Rivers and Beacon-hills of Gondor from the end of the 1960s, VT42/24).

palátā, Q. palta, T. plata, S. plad, meant “the flat of the hand, the hand held upwards or forwards, at and tensed (with fingers and thumb closed or spread)” ... In Quenya the position of the Common Eldarin stress was not important, since at an early period Quenya had shifted the stress back to the first syllable (from notes associated with “Eldarin Hands, Fingers & Numerals” from the late 1960s, VT47/8-9).

These quotes imply that a stressed short vowel could undergo syncope, because the stress would normally shift forward to the first syllable and then the syncope could take effect. This may represent a conceptual shift on Tolkien’s part, with the syncope occurring after the stress shift. If so, the revised conception introduces a new problem. In particular, this new paradigm would mean that sequential short syllables with like vowels should never occur in Quenya, but that is not the case:

• ᴹQ. alako “rush” (Ety/ÁLAK).
• Q. arata “high, lofty, noble” (PE17/49, 186).
• ᴹQ. naraka “harsh, rending” (Ety/NÁRAK).
• Q. Oromë (PE17/153, WJ/400).
• Q. pataca “consonant” (VT39/8).
• Q. silima, substance the Silmarils were made of (RGEO/65).
• ᴹQ. terene “slender” (Ety/TER).
• ᴹQ. tereva “fine, acute” (Ety/TER).

We do not know the primitive form of all the words above, but the ones we do know share an interesting trait, namely a medial long vowel:

• ᴹ✶álākō > ᴹQ. alako (Ety/ÁLAK).
• ✶arātā̆ > Q. arata (PE17/49, 186).
• ᴹ✶narāka > ᴹQ. naraka (Ety/NÁRAK).
• ✶Orǭmē or Arǭmēz > Q. Oromë (PE17/153, WJ/400).
• ᴹ✶terēn(ē) > ᴹQ. terene (Ety/TER).
• ᴹ✶terḗwā > ᴹQ. tereva (Ety/TER; EtyAC/TER).

This long vowel would later shorten if it was unstressed, as explained in the quote from OP1 above (PE19/58):

(2) After this syncope, but still before the Q. accent shift: Long vowels that remained in unstressed medial syllables were shortened, but not lost. So Òrōmḗ > Oromḗ (Q. Orome).

This is consistent with all of the examples above, with the exception of terḗwā, where Tolkien marked the stress in the middle syllable (EtyAC/TER). Conversely, there are a number of examples where Tolkien explicitly marked a short vowel as stressed but the syncope nevertheless took place:

• ᴹ✶berékā > ᴹQ. verka “wild” (Ety/BERÉK).
• ᴹ✶morókō > ᴹQ. morko “bear” (Ety/MORÓK).
• ✶palátā > Q. palta “flat of the hand” (VT47/8).

The simplest explanation is that a primitive stressed short vowel was not by itself enough to prevent the syncope; only a long vowel did so (stressed or unstressed). We cannot, however, simply assume that the syncope occurred after the stress shift, because if that were true the long vowels in words like Òrōmḗ would never shorten. The simplest explanation is that the Quenya stress shift took place in two phases:

• Phase 1: The stress shifted back to the initial syllable.
• Any now unstressed short vowels underwent syncope, but long vowels survived.
• Any now unstressed long vowel were shortened.
• Phase 2: The stress shifted forward to the heavy syllable closest to the end of the word.

The initial phase of this stress shift to the first syllable is indicated by the notes above: “the placing of the accent would not affect Quenya since in PQ the accent became placed on the first syllable in all cases” (VT42/24) and “at an early period Quenya had shifted the stress back to the first syllable” (VT47/9). Thus, the stress shift to the Latin-like stress pattern of modern Quenya (penult if heavy, otherwise antipenult) must have been later.

As a final note, there were some similar but earlier sound changes whereby short vowels ă, ĕ, ŏ were reduced in long compounds which superficially resemble the Quenya syncope:

As far as Quenya is concerned these very ancient effects of accent are the following: 1) obscuration of ă|ĕ|ŏ to murmured [ǝ] in entirely atonic syllables after the main stress. This occurred anciently in stressless final syllables of all kinds. Also often interiorly in long words with a far-retracted accent - for the most part in old compounds - where stressless ă|ĕ|ŏ were several syllables removed from the main stress - e.g. kwènedḗ “elf”, but móri-kwenedḕ “dark-elf” > móri-kwen(ǝ)dḕ > móri-kwèn(ǝ)dĕ. NB. This reduction did not occur medially before 2 consonants (including y, ı̯ | w, u̯) - except that the end of first element of old compounds were often treated as final syllables (PE19/57).

Although these results resemble the Quenya syncope, they are strictly speaking a distinct (and earlier) sound change, though perhaps this earlier phonetic development began the trend that lead to the general Quenya syncope. The mechanism of the Quenya syncope was probably the same: the vowel first reducting to [ǝ], then vanishing.

Conceptual Development: There are examples of the Quenya syncope dating all the way back to the Early Quenya of the 1910s and 1920s:

• ᴱ√ASAKA > [*aska] > ᴱQ. aksa “waterfall” (QL/29).
• ᴱ✶norokā́ > ᴱQ. norka “slave” (GL/31).

The basic rules for the syncope seems to have been established quite early, but as noted above the exact interaction was between the Quenya stress shift and the Quenya syncope is complex. There may have been some conceptual vacillations on Tolkien’s part for the timing of these two sound changes. The system I proposed above (a two-phase stress shift straddling the Quenya syncope) is the one the best fit the available evidence.