In Sindarin (as well as Noldorin and Gnomish), the voiced velar spirant [ɣ] vocalized before various consonants, most notably liquids ([l], [r]) and nasals ([m], [n]). In Noldorin, the velar nasal [ŋ] vocalized in similar ways, perhaps by merging first with [ɣ] (see the Noldorin discussion below), but in Sindarin the vocalization of [ŋ] was distinct. The diphthongs resulting from the vocalization of [ɣ] in some cases had a distinct development from those of primitive diphthongs, indicating this was a comparatively late change, probably after the Old Sindarin/Old Noldorin period, as suggested by David Salo (GS/§4.87, §4.90). A similar change occurred in Welsh (WGHC/§104ii). In most cases it seems these developments were parallel to the vocalizations of the voiceless velar spirant [x], which can be used for additional clues for these phonetic developments.

In the case of nasals, Tolkien described this sound change for Sindarin in notes on Elvish numerals from the late 1960s (VT42/26):

In Sindarin voiceless stops (i.e. p, t, k) before nasals became voiced > b, d, g, and then together with the original voiced stops in this position became nasals before homorganic nasals (tn, dn > nn; pm, bm > mm), but before other nasals became spirants as generally medially (pn, bn > vn; tm, dm > ðm, later ðv, ðw; kn, gn > gn > in; km, gm > gm > im > iv, iw [emphasis added]).

As indicated by the later phonetic developments in this note, gn > in and gm > im. Most likely there was first an intermediate sound change of [g] > [ɣ], since voiced stops became spirants after vowels in Sindarin. Hence the most likely phonetic change was [Vgn] > [Vɣn] > [Vin], and similarly [Vgm] > [Vɣm] > [Vim] > [Viw]. There are a number of Sindarin words in which the sound changes described above seem to occur:

• ✶wagmē > S. gwaew “storm” (NM/237).
• ✶okma [> ogma] > S. oew “evil deed” (PE17/170).
• ✶logna > S. loen “soaking wet, swamped” (VT42/10).

In the first example, the underlying change of [ag] > [aɣ] > [ai] is obscured by the later diphthongal development of [ai] > [ae]. Similarly, in the last two examples, the sound change [oɣ] > [oi] is obscured by the later diphthongal development of [oi] > [oe]. There are a similar examples of phonetic developments in Sindarin words before the liquids [l] and [r] (with some probable intermediate changes added):

• √MAGA [> magla > maʒla > maila] > S. mael “well” (PE17/162).
• ✶keglē [> keʒle > keile] > S. cail “fence, palisade” (UT/282).
• ✶tagra [> taʒra > taira] > S. taer “lofty” (PE17/186).
• ✶ugrā > ogra [> oʒra > oira] > S. oer “nasty” (PE22/160).

Assuming the vocalization of [ɣ] before liquids followed the same pattern as it did before nasals (as was the case in Welsh), these phonetic developments also fit the rules described above, once you factor in obscuring changes like [ai] > [ae] and [oi] > [oe]. In the example of S. cail, the sound change is obscured by the fact that later [ei] became [ai] in final syllables in Sindarin. In the example of S. oer, a-affection also plays a role, but it is otherwise consistent with the sound changes [og] > [oɣ] > [oi] > [oe] seen in the phonetic development of oew and loen above. This last example also implies that the Sindarin vocalization of [ɣ] took place after a-affection.

In all cases, it seems [ɣ] vocalized to [i] before liquids and nasals in Sindarin.

Conceptual Developments: The vocalization of [ɣ] dates back all the way to the Gnomish of the 1910s, but the details shifted over time.

Gnomish (1910s) and Early Noldorin Developments (1920s): Roman Rausch discussed the analogous Gnomish and Early Noldorin changes in his Historical Phonology of Goldogrin (HPG/§2.6) and Historical Phonologies of Ilkorin, Telerin and Noldorin around 1923 (HPITN/§4.1.3), but there are not many examples to work from. In Early Noldorin there are a few of clear examples of [aɣ], [eɣ] > [ai], with [eɣ] presumably passing through [ei] before becoming [ai]:

• ᴱ✶dagnā́ > ᴱN. dain¹ “high, noble” (PE13/141, 161).
• ᴱ✶dágniya > ᴱN. dain² “height, summit” (PE13/141, 161).
• ᴱ✶dagla > ᴱN. dail “axe (blade)” (PE13/141; PE14/66).
• ᴱ✶tegna [> tein] > ᴱN. tain “straight” (PE13/153, 165).

There are a few more words in Gnomish that seem to show similar phonetic developments, but these examples are less clear:

• ᴱ✶sek’·ðana [> *seχðana > *seʒðana] > G. saidhon “noon” (GL/66).
• ᴱ✶foχo [> *foʒna] > G. fuin “secret, dark” (GL/36).
• ᴱ✶megđana > G. maidhon “midday” [deleted entry] (GL/56).
• ᴱ✶Baʒ̔- > G. bain “clad” (GL/21).
• ᴱ✶Baʒ̔- > G. bail “sheath” (GL/21).

The first couple of Gnomish examples are predicated on the assumption that χ became voiced ʒ before other voiced consonants, but this seems to be a reasonable assumption. The last two examples are a bit dubious, since the quote mark above ʒ̔ likely indicates that the primitive forms actually contained voiced palatal spirant [ʝ] instead of [ɣ]; the analogous root in the Qenya Lexicon for Baʒ̔- seems to be VAẎA “enfold, wind about” (QL/100), and Tolkien often used the symbol Ẏ in this document to indicate y-sounds that originated from palatal spirants.

Once we account for the obscuring changes whereby later [ei] became [ai] and [oi] became [ui], it seems that [ɣ] vocalized to [i] fairly consistently in the 1910s and 1920s, which was the conclusion of Roman Rausch as well. The analogous Gnomish and Early Noldorin vocalizations of [x] also fit this pattern, with a more complete set of examples.

Noldorin Developments (1930s): Tolkien partially described the Noldorin vocalizations of ʒ [ɣ] in notes on the usage of the Feanorian Alphabet from the 1930s (PE22/39-40). They resemble the apparent Sindarin pattern, but there are some differences:

The long diphthongs — ON diphthongs, diphthongization of ON ō, or new diphthongs from short vowel + vowel (in contractions or in contact with vocalized ʒ, χ), or from long vowels + epenthetic ı̯ ...

• [o͡u] ON au, ō or ŏ + ʒ, χ. archaic h. later > [au] q.v. ...
• [a͡e] older ai, as above; or ă + ʒ, χ. ]l later G[ondolic] ]Þ — usually ]Þ (beside ]l)...
• [e͡i] e + ʒ, χ or affected a + ʒ, χ; ... l`B l~B later G[ondolic] lÖ — usually lÖ (beside l~B)...
• [ui] ON ui; affect[ed] o, u + ʒ, χ ... .`B .~B later G[ondolic] .Ö — usually .Ö (beside .~B)...

Based on this note it seems that the Noldorin vocalizations of [ɣ] are:

• [oɣ] > [ou] > [au]
• [aɣ] > [ai] > [ae]
• [eɣ] > [ei]
• [uɣ] > [ui]

Although not listed, presumably [iɣ] > [ī]. These rules are mostly consistent with the phonetic development of words appearing in The Etymologies from the 1930s. Here is a representative sample:

• [aɣ] > [ai]: ᴹ✶magrā [> maʒra > maira] > N. maer “useful” (Ety/MAƷ|MAG).
• [eɣ] > [ei]: ᴹ√REG [> regne > reʒne > reine] > rhein > N. rhain “border” (Ety/REG).
• [iɣ] > [ī]: ᴹ√RIG > [> rigna > riʒna > rīna] > N. rhîn “crowned” vs. ᴹQ. rína (Ety/RIG).
• [oɣ] > [ou]: ᴹ√DOƷ/DÔ > dogme > dongme > ON. doume > N. daw “night-time, gloom” (Ety/DOƷ).

There are no clear examples of [uɣ] > [ui] in The Etymologies, however. Instead, we consistently see [uɣ] > [ū]:

• ᴹ✶lugni > N. lhûn “blue” (Ety/LUG²).
• ᴹ✶kuʒnā > N. cûn “bowed, bow-shaped, bent” (Ety/KUƷ).
• ᴹ✶suglu > N. sûl “goblet” (Ety/SUK).

There are similar variations in the vocalizations of [x], where the rules state [ux] > [ui] but many of the examples show [ux] > [ū]. It may be that Tolkien changed his mind about the development of this combination.

One of the examples above indicates the vocalizations of [ɣ] had some additional complications not present in Sindarin:

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

Here the primitive form seems to be *doʒme, but before vocalizations the ʒ first underwent a number of intermediate sound changes: oʒ [oɣ] > og > ong [oŋ] > ou. This is because the nasalization of stops before nasals was different in Sindarin and Noldorin:

• Noldorin: voiced stops became nasals before nasals in all cases (PE22/26).
• Sindarin: only [bm], [dn] became [mm], [nn], that is nasalization occurred only in homorganic combinations (VT42/26).

Based on this Noldorin development it seems that the vocalizations of [g] and [ɣ] before nasals first took a jaunt through the voiced velar nasal [ŋ] before vocalizing. It’s conceivable that the voiced nasal [ŋ] vocalized directly, but it seems likelier to me that it first became [ɣ] again, so that the full phonetic development of *doʒme > ON. doume was: [doɣme] > [dogme] > [doŋme] > [doɣme] > [doume]. This is a pretty roundabout way to arrive at the same destination. Perhaps Tolkien’s motivation for revising the nasalization of stops in Sindarin was to simplify these phonetic developments. In any case, the vocalization of the velar nasal [ŋ] before other nasals was different in Sindarin, since these surviving [ŋ] vanished with compensatory lengthening instead (PE17/44, PE22/149).

The ordering of the Noldorin vocalizations also seems to be different from Sindarin. The following two examples imply that the Noldorin vocalizations must have taken place before a-affection rather than after:

• ᴹ✶kuʒnā > N. cûn “bowed, bow-shaped, bent” (Ety/KUƷ).
• ᴹ√RIG > N. rhîn “crowned” vs. ᴹQ. rína (Ety/RIG).

Both of these examples seem to have primitive final ā, but if a-affection occurred before vocalization, they would have produced ✶kuʒnā > koʒna > kouna > **caun and ✶riʒnā > reʒna > reina > **rhain respectively. Since the vocalization occurred first, the resulting long [ī] and [ū] resisted a-affection. Compare this to the above S. oer < ogra < ✶ugrā (PE22/160), which implies a-affection took place first. However, Tolkien’s notes from the usage of the Feanorian Alphabet quoted above seem to imply yet another ordering of sound changes (PE22/39-40):

• [e͡i] e + ʒ, χ or affected a + ʒ, χ [emphasis added]; ... l`B l~B later G[ondolic] lÖ — usually lÖ (beside l~B)...
• [ui] ON ui; affect[ed] o, u + ʒ, χ [emphasis added] ... .`B .~B later G[ondolic] .Ö — usually .Ö (beside .~B)...

Here Tolkien states that the vocalizations of ʒ in combination with “affected a” has the same phonetic development as e and combinations with “affected o” and u also have the same phonetic developments. Since the only type of affection to modify a is i-affection, most likely this is what Tolkien means. But this is not consistent with the examples in The Etymologies, where the vocalizations of ʒ took place before a-affection, which itself preceded i-affection.

Summary of Conceptual Developments: In the Gnomish and Early Noldorin of the 1910s and 1920s, it seems [ɣ] vocalized to [i] before nasals and liquids (and possibly a few other consonants like [ð]). The same was mostly true in the Noldorin of the 1930s, but Tolkien said that [oɣ] > [ou] instead, and examples in The Etymologies indicate that [uɣ] > [ū] as well. By the Sindarin of the 1950s, it seems [ɣ] consistently vocalized to [i] once again. There does seem to be some vacillation on the ordering of this sound change, in particular whether or not it took place before or after (or between) a-affection and i-affection. In Noldorin, the voiced velar [ŋ] underwent similar vocalizations where it appeared before other nasals, but this was not true of Sindarin.

For further discussion of the timing of this sound change and its possible implications, see the notes on the phonetic development of Early Noldorin, Noldorin and Sindarin plurals in the entry on how [x], [ɸ] vocalized between a vowel and [θ].

GS/§4.87 GS/§4.90 WGHC/164:§104ii @@@ or earlier change from g