Where voiced spirants ƀ [β] and ʒ [ɣ] (themselves the result of the stops b and g) appeared before voiced consonants, they developed into vowel sounds. In the case of ƀ the result was always u, but the phonetic development of ʒ depended on the surrounding vowels and consonants. These vocalization could only occur before the liquids r, l or the semi-vowels y, w, since earlier on combinations of voiced stops were unvoiced and voiced stops became nasals before nasals.

Tolkien described identical developments in both the Outline of Phonetic Development (OP1) from the 1940s and the Outline of Phonology (OP2) from the 1950s. For ƀ [β]:

[b] at stage [ƀ, bilabial] became vocalized to u̯, sufficiently early for the resultant diphthongs to be treated in the same way as the inherited diphthongs (OP1: PE19/46).

b became ƀ and at that stage (before proceeding to v) became w, u̯, forming diphthongs with the preceding vowels. As no marked changes had at the period (AQ) yet affected the older diphthongs, the development of these diphthongs formed with vocalized b was the same in PQ and TQ. (OP2: PE19/94).

Tolkien also mentioned that, in those rare cases where v again became adjacent to a consonant in Parmaquesta [PQ] or Tarquesta [TQ], it often developed into u. One example appearing in both OP1 (PE19/45) and OP2 (PE19/90) is the reformed present tense auta of the verb apta- “refuse”, based on the past tense avante and established as if it had developed from avta.

For ʒ [ɣ]:

[g] proceeded to stage [ʒ]. This became [ı̯], before y; [u̯] before w; and before r, l was lost with lengthening of the preceding vowel. Thus magyā > mai(y)a; dagwe > lau(w)e, but magra > mára; stegrā > thḗra; soglā > sóla. It is said that AQ shows that in fact [ʒ] > ı̯ after ī̆, ē̆ but ei then > ē; similarly [ʒ] > u̯ after ū̆, ō̆, but ou then > ó (OP1: PE19/46).

g proceeded to stage ʒ and then became vocalized, the vocalic character assumed being dependent on the preceding vowel, except where it was followed by y or w.

• Before y: g > ʒ > ı̯ irrespective of the preceding vowel.
• Before w: g > ʒ > u̯ irrespective of the preceding vowel.

Thus magyā > maı̯ya, maia; ndagwē > ndau̯we > nauwe, naue (OP2: PE19/94).

Thus:

• before y: ʒ > i.
• before w: ʒ > u.
• before l, r: ʒ vanished with compensatory lengthening.

The last sound change is identical to the vocalizations of ʒ in Primitive Elvish, and had the added twist (as noted in OP1) that perhaps eʒl, eʒr > eil, eir > ēl, ēr and oʒl, oʒr > oul, our > ōl, ōr as part of the general rule whereby [ei], [ou] generally became [ī], [ū], but these more complex phonetic developments are indistinguishable from “vanished with compensatory lengthening” (though they might be detectable by analysis of ancient Sarati writings). In addition to the examples given above, we can see these phonetic developments from the 1930s and forward:

• ᴹ✶sagrā [> *saʒrā] > ᴹQ. sára “bitter” (Ety/SAG).
• ᴹ✶suglu [> *suʒlu] > ᴹQ. súlo “goblet” (Ety/SUK).
• √MAG [> *maʒya] > Q. †maie “well” (PE17/162).
• ✶tagra [> *taʒra] > Q. tára “lofty” (PE17/186).
• ✶ugrā [> *uʒrā] > Q. úra “torrid” (PE22/160).

There are a couple of examples from The Etymologies of the 1930s that indicate that the vanishing of ʒ from g with compensatory lengthening might have been a more widespread phenomenon in this conceptual period:

• ᴹ√REG > ᴹQ. réna “edge” (Ety/REG).
• ᴹ✶yagwē > ᴹQ. yáwe “ravine” (Ety/YAG), not the expected **yauwe.

These two examples seem to show vanishing with lengthening before w and n. For a discussion of the first example, see the entry on how [ŋ] vanished between [i], [u] and [n] with compensatory lengthening. For the second example, in the discussion in OP2 Tolkien wrote in ink: “nauwe, naue > náwe” and in the margin “The product of aiı̯, āı̯, and of auu̯, āu̯ fell together, aiı̯, āı̯”, but this was all deleted (OP2: PE19/94 note #126). So perhaps ᴹQ. yáwe represents an alternate development whereby yauwe [jauu̯e] > yāwe [jāu̯e]. Tolkien sometimes gave the proper form of Q. Maia as Máya (VT42/13), which might represent a similar sound change maiya [maiı̯a] > māya [māı̯a]. However, it seems the normal phonetic developments were aiya, auwa > aia, aua, not áya, áwa.

Conceptual Development: Analysis of these phonetic developments in the 1910s are complicated by the fact that voiced and voiceless spirants were part of the phonetic inventory of Primitive Elvish. However, in the Qenya Phonology from the 1910s Tolkien indicated that ƀ [β] generally vocalized to u before other consonants and finally:

ƀ + consonant [>] u̯; ƀ final [>] u (PE12/24).

It is nearly impossible to distinguish a ƀ that had development from b from an original primitive ƀ in Early Qenya, but there is at least one example of such a vocalization:

• ᴱ√FABA > ᴱQ. fauka “fat, large” (QL/37).

Tolkien likewise said:

Before consonant ... ʒ > vanished with compensatory lengthening (PE12/18).

Here Tolkien is specifically discussing ʒ that had developed from a voiceless velar spirant [x], but it seems the same sound changes were true for original ʒ and (probably) ʒ derived from g. For example:

• ᴱ√FOƷO > ᴱQ. fôle “stealth, a secret, secrecy” (QL/38).

Again, distinguishing ʒ that had development from g from an original primitive ʒ is extremely difficult in Early Qenya.

Determining whether there were any special phonetic developments of ʒ before y and w is complicated by the existence of the voiced palatal or palatalized spirant ʒ̑ (probably [ʝ]) and labialized velar spirant ɣ͡w (probably [ɣʷ]) in the phonetic inventory of Primitive Elvish in the 1910s. Distinguishing [ɣj] from [ʝ] and [ɣw] from [ɣʷ] in Early Primitive Elvish is nearly impossible. Generally [ʝ] > [j] and [ɣʷ] > [w] (PE12/16) but determining the surrounding vocalic effects (if any) is extremely difficult.