In the Old Noldorin of the 1930s and 1940s, the voiced velar spirant [ɣ] evolved into [g] before nasals [m] and [n]; the same seems to be true of [ŋ], though Tolkien’s notation is unclear. This development is discussed in Tolkien’s notes on the Feanorian Alphabet from the 1930s:

The PQ combinations ȵm, ȵn (from ȵ, ʒ + m, n) appear to have given gm, gn, falling together with gm, gn from PQ k, g + m, n. Hence ON standard spelling xy x6. Later here x g became [ŋ], as bm, bn > mm, mn, dm, dn > nm, nn; but ON having no special sign for [ŋ] retained the gm, gn spelling. The nasality is indicated by the occasional spelling by b6 = ngm, ngn (PE22/26).

There is a very clear example of this phonetic change in The Etymologies of the 1930s:

• ᴹ√DOƷ/DÔ > dogme > dongme > ON. doume > N. daw “night-time, gloom” (Ety/DOƷ).

This phonetic rule helps explain why the sound survived before nasals rather than vanishing as [ɣ] usually did in Old Noldorin. At some later point the resulting [g] again weakened (to [ŋ] or [ɣ]) and ultimately vanished, but with different vocalic effects than during its earlier loss: becoming a vowel [i] or [u] and forming diphthongs in Noldorin, rather than lengthening the preceding vowel. Consider:

• ᴹ√DOƷ/DÔ [doɣmē] > ON. doume > N. daw “night-time, gloom” (Ety/DOƷ).
• ᴹ√taʒna [taɣna] > N. taen “height” (Ety/TĀ).

In the first example, had the [ɣ] been lost in Old Noldorin, the result would be vowel-lengthening to long [ō] then becoming [ū] (**dû) instead of a u-diphthong [ou] ultimately become [au]. In the second example, an early loss would result in lengthening to [ā] which would then become [ǭ] and ultimately [au] (**taun), not the i-diphthong [ai] which ultimately became [ae].

There is a similar example in (Old) Sindarin, but involving a liquid rather than a nasal:

• √TAƷ[+ra] > tagra > S. taer “lofty” (PE17/186).

As with Noldorin, the timing of the change may be detected by vocalic effects. If the [ɣ] were lost early the result would be [taɣra] > [tāra] > [tǭra] > **[taur], but the early strengthening to [g] results in [tagra] > [taɣra] > [taira] > [taer]. There isn’t enough evidence to determine the exact set of conditions under which this change occurred in Sindarin, since both [ɣ] and [ŋ] were rare medially in primitive roots. The strengthening to [g] could have applied to both [ɣ] and [ŋ], and could have occurred before nasals [m], [n] and liquids [r], [l], or any subset of these conditions.

It does seem that this sound change did not occur before liquids in the Old Noldorin of the 1930s, for example: ᴹ✶waʒrā [> wārā] > ON. wōra [wǭra] > N. gwaur and not **gwaer or gwoer (Ety/WAƷ). Furthermore, there is a note from the second version of Tengwesta Qenderinwa (TQ2) around 1950 which seems to indicate that Tolkien considered discarding this phonetic rule entirely:

[In Noldorin] Medially ʒ between vowels & before sonants vanishes very early with compensatory lengthening (PE18/104).

In this document, Tolkien uses the term “sonant” for consonants more commonly referred to as “sonorants”: nasals ([m], [n]), liquids ([l], [r]) and semi-vowels ([j], [w]). In this note, Tolkien indicates that ʒ was lost with the early vocalic affect of vowel-lengthening rather than the later vocalic effect of diphthong-formation, and thus could not have strengthened to [g]. It seems that Tolkien vacillated several times and in several ways on this particular phonetic rule, making it difficult to nail down his exact ideas.

Neo-Sindarin: The sound combinations involved in this phonetic rule are sufficiently rare that I find it very difficult to choose one set of rules to use for Neo-Sindarin phonology; for now I am assuming a broad change of [ɣ] to [g] before both nasals and liquids. Any inconsistencies can be chalked up to dialectical variations in Sindarin’s phonetic development.

@@@ possibly [ɣ] > [ŋ] before [m], [n] in Primitive Elvish: but DOƷ + mē > ᴹQ. lóme

See OS. [ɣ] became [g] before nasals and liquids for discussion.