Starting in the late 1950s or early 1960s, Tolkien decided the ancient velar nasal ñ- [ŋ] was strengthened to ñg- initially in (Old) Sindarin and ultimately became g-. Tolkien seems to have introduced this idea in an etymology of the name of Galadriel written sometime between the first and second edition of The Lord of the Rings:

If the original form of Galadriel’s name was ñgal(a)tā-rig-el-, Quenya would be Ñaltariel and Galadriel would be correct. And only associated with trees in Lórien ... no never so associated, but by strangers. ñ- > ñ in Quenya, > ñg > g in Sindarin, only vanishes in Telerin [emphasis added]. Altarielle is Telerian (PE17/60).

In a note from the late 1960s Tolkien described a slightly different, development, however:

ŊAL (ŊGAL), “shine clear”. S gal-, rare in Quenya except in old words and names. Sense is coloured with √GALA, grow. S ŋ > ʒ > g (PE17/169).

Here is seems ancient ñ- became g- via ʒ rather than passing through ñg- first. Tolkien reaffirmed the idea that ancient initial ñ- ultimately produced g- in The Shibboleth of Fëanor essay written in 1968 but without specifying the intermediate form:

Galadriel ... was derived from the Common Eldarin stem ÑAL “shine by reflection”; *ñalata “radiance, glittering reflection” (from jewels, glass or polished metals, or water) > Quenya ñalta, Telerin alata, Sindarin galad (PM/347).

In an unrelated 1968 note Tolkien said:

This variety [of biconsonantal roots that lost their initial consonant] was largely increased in the descendant languages, notably Quenya, by the loss of older weak consonants initially: in Quenya, C.E. ʒ, h and g; in Telerin ʒ, ñ; in Sindarin h, were lost (VT48/26, Note 4).

Though Tolkien did not specify in this note how initial ñ- developed in Sindarin, the fact that it did not vanish supports the notion that he still imagined it ultimately becoming g-.

Conceptual Development: In Gnomish, various examples indicate that likewise initial [ŋ-] became [g-]:

• ᴱ✶ŋolđō > G. Golda “Gnome” vs. ᴱQ. Noldo (GL/41).
• ᴱ✶ŋaistă- > G. gaista- “to oppress” vs. ᴱQ. naista- (GL/37).

Since the lenited form of Golda was i·Ngolda (GG/8), it is clear that [ŋ-] > [ŋg-] > [g-]. The phoneme [ŋ] was actually fairly common in Early Primitive roots, even medially, and such roots frequently show g- or -ng- in Gnomish forms. For example:

• ᴱ√ŊAHYA > G. gaist “torment, oppression” (GL/37).
• ᴱ√GUŊU > G. gung “spider” (QL/98).
• ᴱ√EŊE > G. eng(a) “plain, vale” (QL/36, GL/32).
• ᴱ√RIŊI > G. ring “cold” (QL/80, GL/65).

This seems to indicate that [ŋ] became [ŋg] medially in Gnomish as well, at least where isolated or between vowels. Many corresponding Early Quenya forms also show medial ng, hinting that this might even be a phonetic rule in Tolkien’s early conception of Primitive Elvish. The early Qenya Phonology from the 1910s indicates this same medial modification to the velar nasal at an ancient (Kor-Eldarin) stage of Qenya: compare the charts on PE12/15 (ȵ) to PE12/16 (ng), though in those notes this sound change was still “after the departure of the Noldoli”.

However, there are a few examples that show other developments might be possible, for example G. rî “coolness, cool” (GL/65), also very likely derived from the root ᴱ√RIŊI and perhaps with a lost [ŋ].

By the 1930s, however, Tolkien changed his mind, deciding that [ŋ] vanished both initially and medially:

• Initially ñ vanished without trace (as in Kal[aquendian] generally) except that ñy > ny at the time that ky > ty and so remained in Q. ny but in N., T. > y whereas ñw > m in N., T. at time kw > p and so remained (Outline of Phonetic Development [OP1], 1940s, PE19/32-33).
• But already before adaption to Old Noldorin [the tengwa] h n had ceased to be consonantal, owing to disappearance of the nasals [ŋ, ŋ^u̯] for which they originally stood, and had become the signs for o, u̯ (w). This use persisted in ON (The Feanorian Alphabet, late 1930s, PE22/25).

This vanishing of initial ñ is reflected in phonetic developments in The Etymologies of the 1930s:

• ᴹ✶ñōle > N. ûl “odour” (Ety/ÑOL).
• ᴹ√ÑIW > ON. (g)yūta “hint” (Ety/ÑIW).

The second example seems to illustrate ñy- > y-. This primitive ñ seems to have been palatalized, perhaps ñiw+ta- > ñyiuta- > ON. yūta-. A description of this palatalized development appears within another description of initial ñ loss in the second version of Tengwesta Qenderinwa (TQ2) written around 1950:

[initial] ŋ̃ at first remained until ŋ̃w > m ... ñy > y not nil. This shows loss of y in ky-series was later than Noldorin, Telerin labialization of kw-series ... [medial] ñ later disappear[s] (with various vocalic effects) except (i) in combinations ñk, ñg, ññ (> ñg), (ii) in medial ñy, ñw > ñgy, ñgw > mb [more exactly: medial ñy > ñgy and medial ñw > ñgw > mb] (PE18/104).

This note helps establish the timing of this phonetic rule: after labialized velars became labials (since [ŋw-] > [m-] and not [w-]) but before [j] was lost after initial velars (since [ŋj-] > [j-] and not [ø-]). Furthermore, since ñ usually survived in Telerin (PE18/103), this change must have occurred in Old Sindarin rather than Ancient Telerin, as suggested by David Salo (GS/§4.38).

Tolkien seems to have retained this loss of initial ñ for some time after changing Noldorin to Sindarin. It can be seen in the following development in a document on names written in 1957:

• ✶ÑAR > S. ar- “by” (PE17/169).

However, Tolkien eventually restored a phonetic development similar to Gnomish whereby initial ñ- > ñg- (or ʒ) > g- as noted above (PE17/60, 169), likely motivated by his preferred forms for the names Galadriel and Gil-galad in the various Elvish languages. The first appearance of this restored phonetic rule appears in a document where he experimented with various possible etymologies and Quenya forms for Galadriel’s name before settling on the one involving ñ- > g- (PE17/60), which he retained thereafter.

GS/§4.38 @@@