In Sindarin and Noldorin, syllabic nasals at the beginning of words developed the vowel a, so that [ṃ-/ṇ-/ŋ̣-] became [am-/an-/aŋ-]. A clear description of this sound change occurs in the notes on ambar vs. umbar from the 1960s:

... in some cases, usually words of strong or emotional meaning or important names, the nasal became syllabic, and then in Quenya > um [from ṃ] (vocalic ṇ usually > an; vocalic ŋ̣ > iŋ except before gw where it > uŋ) ... In Sindarin a similar development (in similar cases) took place, but syllabic m, n, ñ all had the same vowel am, an, ang (PE17/124).

There are a number of examples of this phonetic development in both Sindarin and Noldorin:

• √NDŪ > S. annûn “sunset, west” vs. Q. andúnë (PE17/64).
• ᴹ✶ṇdūnē > N. annûn “sunset” vs. ᴹQ. andúne (Ety/NDŪ).
• ✶ṃbar > “primitive S” [?OS.] ambar “fate” vs. Q. umbar (PE17/104).
• ᴹ√MBAT(H) > N. amath “shield” vs. ᴹQ. umbas (EtyAC/MBAT(H)) [these forms were later deleted].
• ✶ñgolodō > S. Angolodh “The Noldo” vs. Q. Ingoldo (PM/360).
• ᴹ√ÑGYO(N) > ON. ango “grandchild” vs. ᴹQ. indyo (Ety/MORÓK).

Some of the above examples explicitly attribute the sound change to Old Noldorin or “Primitive Sindarin” [Old Sindarin]. David Salo also attributes this change to Old Sindarin in Gateway to Sindarin (GS/§4.40). Similar sound changes occurred in the history of Welsh, where short syllablic [ṃ] and [ṇ] from Primitive Indo-European evolved to [am], [an] by the Britannic period (WGHC/§62i). The Welsh changes were more pervasive, however, since syllabic nasals could appear medially in Primitive Indo-European words, not just initially.

Conceptual Development: In Gnomish and Early Noldorin, the development of syllabic nasals was different, in part because like Primitive Indo-European, Tolkien’s conception of Primitive Elvish in the 1910s and 1920s allowed them to appear both initially and medially in words, functioning as ordinary vowels. As described by Roman Rausch in Historical Phonology of Goldogrin (§3), short [ṇ] became [in] and long [ṇ̄] became [an]. For example:

• ᴱ✶fṇt > G. fintha- “to twist” (GL/35).
• ᴱ✶fṇ̄t > G. fant “crooked” (GL/34).
• ᴱ✶ṇdr > G. in(d)ra “long” (GL/51).
• ᴱ✶ṇ̄dai > G. annai “far away” (GL/19).

This can also be seen in the vowel gradations in some Gnomish present vs. past tenses, because in the primitive past tense the syllabic nasal was lengthened:

• ᴱ√QṆTṆ > G. cwintha- “to fill”, past cwanthi (QL/78, GL/28).
• ᴱ√YṆTṆ > G. intha- “to join to”, past ganthi (GL/37-38, 51).
• ᴱ√TṆTṆ > G. tinta- “to dance”, past tanti (GL/69-70).

As pointed out by Roman Rausch, apparently short [ŋ̣kʷ] > [uŋkʷ]:

• ᴱ√ṆQṆ² > G. unc² “ear, handle (of a jar)” (GL/75).

It is unclear whether [ŋ̣] > [uŋ] generally or only before labialized velars, but if I had to guess it would be specifically before labialized velars, with [ŋ̣] > [iŋ] more generally (as was the case with Quenya, see above). It is also possible that (like Quenya) [ṃ] > [um], but there are no examples. Apparently, however, this did not happen if the syllabic nasal was long:

• ᴱ✶ṃ̄t- > G. ant “cheek” (GL/19).

Based on these examples, it seems that syllabic nasals in Gnomish behaved very similarly to the corresponding Early Qenya developments, except that in Gnomish short [ṇ] > [in] whereas in Early Quenya short [ṇ] > [an]. There aren’t enough examples of syllabic nasals to determine the Early Noldorin development of the 1920s, but presumably it was similar to Gnomish.

See OS. initial syllabic [m], [n], [ŋ] became [am], [an], [aŋ] for discussion.