In Quenya an initial or intervocalic w became v. The same basic phonetic developments were discussed in both the Outline of Phonetic Development (OP1) from the 1930s and the Outline of Phonology (OP2) from the 1950s, but the timing and details of the sound changes were not the same. In the 1930s (OP1), this was a fairly recent sound change, occurring in the Tarquesta [TQ] period and possibly in Lindarin only [pre-Vanyarin]:

The signs for v and w remained distinct in classical PQ. In the normal (Lindarin) pronunciation of TQ v, w fell together in v, except in kw, (ñ)gw, tw, rw, nw (OP1: PE19/33).

The absence of the combination lw from the list above may be notable (see below). There is no mention of any special development for gw > ʒw [ɣw] > w, so likely this change occurred after [ɣ] from [g] vanished in the 1930s. There is no mention of bilabial ƀ [β], so this change was likely also after [ɸ], [β] became [f], [v].

The contemporaneous Comparative Tables of phonetic development show initial w, gw > w in Quenya but > v in Lindarin (PE19/18-19). The medial developments likewise show gw > uw in Quenya and uv in Lindarin (the u is an indication of the vocalization of ʒ before voiced consonants), but in Quenya medial w > both w, v as opposed to just v (or u) in Lindarin (PE19/23). This medial Quenya w/v variation may be explained by the following:

After long vowels y, w remained and w was strengthened to TQ v. So māya, rāwa (rava) (OP1: PE19/62).

Thus is seems that while w > v was only a universal rule in Lindarin in OP1, it could also occur in the more general Quenya dialect for w after a long vowel. Some contemporaneous notes on the Quenya use of the Feanorian Alphabet describe similar developments from Parmaquesta [PQ] to Tarquesta [TQ]:

Sound changes which occurred after the “classical” fixing of the spelling, and which sometimes resulted in spelling modifications (“incorrect” and unclassical) are the following:

1. Confusions of v/w. See phonology. The classical PQ retained the distinction of y v / n w according to etymology as a rule, but the TQ confusion often reflected either in use of y throughout for both — the cases of retained w (s.c. in tw, nw, ñgw, kw, lw, rw) being denoted by separate letters (z = kw, x = ñgw) or by the diacritic  ë, 1é 5è jè 6é = tw, nw, lw, rw (on b = nw, see below). The earliest distinction given up was that between lv, rv (< lb, rb) and lw, rw, which early coalesced in lw, rw. Here jè 6é for jn 6n is classical (PE22/45).

Again note the special developments for lw, rw. The change of w > v was sporadically applied in The Etymologies of the 1930s, with v/w variants often appearing in the same entry. One notable example is the root ᴹ√WIL “fly”, where Tolkien first showed wilwa > vilwa, but struck out the second form and also revised vilin >> wilin (Ety/WIL). The related word wilwarin “butterfly” appeared without v in The Etymologies, as well as in the Qenya Lexicon from the 1910s (QL/104), the early version of the Markirya poem from the late 1920s (MC/213) and the final versions of Markirya from the 1960s (MC/222).

The medial developments in The Etymologies mostly show w > v with two notable and interesting exceptions. First, the combination aiw consistently survives: ᴹQ. aiwe “bird” (Ety/AIWĒ), laiwa “sick” (Ety/SLIW), maiwe “gull” (Ety/MIW), vaiwa “wind” (Ety/WĀ), yaiwe “scorn” (Ety/YAY). The only example of aiv is derived from primitive aib: ᴹ✶laibē > ᴹQ. laive “ointment” (Ety/LIB²). These aiw survivals continue to appear in Tolkien’s later writings: Q. raiwë “lace” (VT42/12) as well as reappearances of Q. aiwë (UT/401), Q. maiwë (MC/222), Q. vaiwë (PE17/189); see further below.

Second, there is one example of medial w survival that was derived from medial gw: ᴹ✶yagwē > ᴹQ. yáwe “ravine” (Ety/LIB²). This indicates Tolkien may have been considering special developments for gw already in the 1930s. Likewise, the Quenya derivatives of the root ᴹ√GWEN show only initial w-: ᴹQ. wén, wenya, wende (Ety/GWEN), though the last of these has a variant vende derived from ᴹ√WEN(ED). These phonetic developments in The Etymologies may be an early sign of the sound changes described in the Outline of Phonology (OP2) from the 1950s:

But later the frictional element of w was increased where it stood initially or remained between vowels, and passing through bilabial v (ƀ) coalesced with v < CE b. This change was common to Vanyar and Noldor, and so began fairly early. It was consequently a feature of Tarquesta, and was usually also reflected in spelling (even affecting that of later writings in PQ). At the same time it was later than the reduction of gw > ʒw > w and affected w of that origin equally (but see gw: B 2 below) (OP2: PE19/72).

gw became [ʒw], and then initially a spirantal frictional w (with slight friction both at the back and at the lips); but already in PQ this w was weakened. The weakening was later than the beginnings of the change of original w > v, and w < gw remained as such and distinct from w, v for some time: in earlier PQ it is not represented by the sign for v; but in TQ it had become identified with w, v (OP2: PE19/72; 75-6).

These notes from the 1950s are a bit inconsistent, but it seems Tolkien’s intention was to pull the sound change w > v backwards in time, prior to the change of bilabial [β] to [v]. The full phonetic development was w > ƀ > v and this occurred early enough that the distinction was generally lost in PQ spelling. Furthermore, this change of primitive w was also before [ɣ] from [g] vanished, since gw > ʒw > w lingered longer and did not become v until the TQ period. These specialized developments for gw reappear in separate Quenya Notes (QN) from 1957 (PE17/145):

Q b-, w- > v-. ñw-, gw- > w-. medially b > v. w > u̯ or lost. gw > ¯w, u̯. mb, lb > b. ñw > ngw (PE17/154).

This note appears beneath a discussion of the root √GWAL whose Quenya derivatives consistently show w both initially and medially such as: Q. walmë “excitement”, walta- “to excite”, awalda “excited” (PE17/154, 189). This later note may indicate that Tolkien decided that w from gw survived all the way into spoken Tarquesta [TQ], but it is hard to say whether this was an actual conceptual shift or a transient idea; v/w variation continue to appear in the 1950s for words derived from primitive gw, such as wende vs. vende variations in the Quenya prayers of the 1950s (VT44/5, 12). It is likely that Tolkien introduced this special development for gw out of a desire to preserve certain early words like wilwarin “butterfly” and wingë “foam” which continued to appear as late as the 1960s (MC/222), albeit with occasional v- variants (MR/166, PM/370).

As described above, w generally survives in combination with other consonants both initially and medially, particularly ñgw, tw, nw, hw, kw, the last often written q or qu. The developments for rw, lw are more complicated, though, with various specialized developments noted above and elsewhere. See the entry on how [w] became [v] after [l], [r] for further details.

To summarize:

• In OP1 of the 1930s the sound change w > v occurred in TQ, and only fully in the Lindarin dialect. It occurred after gw > ʒw > w and equally applied to w of that origin. The Etymologies, however, show indications of the phonetics developments described in OP2.
• In OP2 of the 1950s the sound change w > v occurred in PQ before ʒw > w. Any w from gw did not become v until TQ and perhaps did not become v at all according to some later writings.
• As a general exception to these rules, the combination aiw (and oiw?) survives unchanged.

Tolkien did explicitly mention this aiw survival in OP2:

After r, l and the diphthong ai however w remained as such (OP2: PE19/72 note #22).

Tolkien then revised this sentence to add oi as another diphthong after which w survived:

After s (z), n, r, l, and the diphthongs ai, oi however w remained as such, and in N[oldorin] TQ the tendency was here to change v > w (OP2: PE19/72).

However, there are no Quenya examples of oiw past the 1920s, so it’s unclear whether or not this was a transient idea.

Conceptual Development: Initial and intervocalic w are fairly common in the Early Qenya of the 1910s and 1920s. The challenge with analyzing the early developments is that there is a much larger inventory of consonants in the Primitive Elvish of the 1910s, many of which could develop into either v or w. For example, in the charts appearing in the Qenya Phonology of the 1910s (PE12/15-16), the primitive labialized velar spirants x͡w, ɣ͡w [xʷ, ɣʷ] both could become Kor-Eldarin ẇ (v), and likewise the primitive labial spirants ꝑ, ƀ [ɸ, β] could become Kor-Eldarin v (w). In these charts primitive w [u̯] either survives in Kor-Eldarin or become u. Unfortunately, Tolkien does not accurately represent the most primitive forms of roots in the Qenya Lexicon, showing the Kor-Eldarin forms instead. This makes it very hard to determine the exact phonetic developments. However, I think the bulk of the evidence is that w > v did not occur in the 1910s and 1920s.

Neo-Quenya: Many Neo-Quenya writers go so far as to revise any initial or intervocalic w to v in all Quenya words. I used to be of the opinion that these v/w variations might be retained by assuming the survival of w from gw until TQ, but in 2023 I changed my mind and starting representing all Quenya words as beginning with v-. For those that feel otherwise, I indicate classical pronunciations with a following [w] in brackets.

@@@ late change only, move to TQ; fix rule: w direct to v