Primitive Elvish words allowed for initial nasal-stop clusters (PE18/43-44, 93), and in some cases those initial groups became syllabic. The exact conditions under which this syllabification occurred isn’t entirely clear, though sometime in the last 1950s or early 1960s Tolkien said “in some cases, usually words of strong or emotional meaning or important names, the nasal became syllabic” (PE17/124). In the 1930s he wrote:

The circumstances in which the nasal became syllabic are not certain. It is probable that phonetically the syllabic nasal developed (a) in absolute initial position (at the beginning of a sentence), especially before a stressed syllable; and (b) after a preceding word ending in a consonant (OP1: PE19/36).

These syllabic nasal clusters developed a preceding vowel in all child languages. In Sindarin the vowel produced was always a, but in Quenya the vowel varied depending on the character of the cluster: u before a labial or labialized cluster, i before a velar or palatalized cluster and a before a purely dental cluster. These syllabic developments can be seen as variant forms in the Comparative Tables of phonetic development from the 1930s (PE19/20) and Tolkien described these sound changes in both the Outline of Phonetic Development (OP1) from the 1930s and the Outline of Phonology (OP2) from the 1950s:

[OP1: PE19/36] As in the case of nasals before voiceless stops the nasals in certain cases became syllabic and vocalic. In these cases the consonant group was treated as medially:

• ṃ became um: as in umbart- “doom”.
• ṇ became an: as in ...
• ṇ̃ became iñ: as in Iñgolonde, but became uñ before kw, gw as in uñqale.
• ṇ < ṇ̃ before ty (< ky), dy (< gy) became in as in ...

[OP2: PE19/77] The consonantal nasal before voiced stops in some instances became vocalic and syllabic. In such cases the consonant group was treated as a medial group. The syllabic nasal developed an actual vowel dependent on the quality of the following consonant:

• before b, gw the vowel was u.
• before d the vowel was a.
• before g, gy (> dy) the vowel was i.

The ellipsis in the OP1 quote above represent unfinished sentences rather than omitted text. As indicated in the OP1, in the 1930s the syllabic nasal clusters could include either voiced or voiceless stops, and this remained true in the original, unrevised layer of OP2, but at some point in the 1950s Tolkien changed his mind, deciding that nasals + voiceless stops (syllabic or otherwise) could not appear initially in Primitive Elvish (PE19/76 note #40; PE19/77 note #46).

A representative sample of these developments include:

• ṃb- > umb-: ✶ṃbart- > Q. umbar “doom” (PE19/77; Ety/MBARAT).
• ṇd- > and-: ✶ṇdūnē > Q. andúne “sunset” (PE19/77; Ety/NDŪ).
• ṇ̃g- > iñg-: ✶ṇ̃golondē > Q. Ingolonde “Country of the Noldor” (PE19/77; Ety/ÑGOLOD).
• ṇ̃gw- > uñgw-: ᴹ✶ṇ̃gwā̆le > ᴹQ. ungwale “torture” (Ety/ÑGWAL).
• ṇ̃gy- > iñgy-/indy-: ᴹ✶ÑGYAL(AM) > ᴹQ. indyalme “clamour” beside variant yalme (Ety/ÑGAL).

As the last example shows, the combination ṇ̃gy still underwent the change whereby velars were dentalized before [j], so that the ultimate result was indy-. It isn’t entirely clear whether vocalization or dentalization occurred first, however. I am of the opinion that vocalization almost certainly occurred before initial nasals-stop clusters reduced to nasals more generally, that is: before nonsyllabic ñg- > ñ-, etc. But in The Etymologies of the 1930s, the variant development yalme indicates ñgy- > ñy- > y-; this must have occurred before dentalization because ny- would have survived.

The picture in OP1 and OP2 is more complex, because there Tolkien said that ñ(g)y- > ny- (OP1: PE19/34, 36; OP2: PE19/74, 76), removing the requirement that vocalization of syllabic nasals occurred before dentalization. Nevertheless, I still assume vocalization occurred first, just to keep the ordering of other sound changes easier in the data model.

Conceptual Development: In the Early Qenya of the 1910s and 1920s, syllabic nasals were part of the vowel inventory of Primitive Elvish, and could occur anywhere in a word, not just initially. The table of these early nasal vocalizations almost matches Tolkien’s later conceptions (PE12/10):

[Before]	q	k	k̂	t	p
short nasals	uŋ	aŋ	in	an	um

In the chart above, k̂ represents the palatal series: before such palatals the vocalized nasals produced the vowel i, whereas before labials and labialized velars (like q = [kʷ]) the vowel was u. Before dentals the vowel was a, but unlike the pattern from the 1930s-60s this was also the result before velars. This can be seen in examples like ᴱ√SṆKṆ > ᴱQ. sanka¹ “rend, jab” (QL/85).