In Ancient Quenya, an s was voiced to z between vowels and before voiced consonants, a change described in both the Outline of Phonetic Development (OP1) from the 1930s and the Outline of Phonology (OP2) from the 1950s. The sound changes in the 1930s were:

s. Initially this remained unchanged. Medially in AQ and the earliest PQ it also remained unchanged, except in certain combinations discussed below. Later, but in the classical period it became voiced to [z] between vowels. This pronunciation as [z] remained the standard TQ pronunciation, serving as it did to distinguish early classical s, th after the development of th, þ > s (but never to z) (OP1: PE19/33).

[s]. Later during the classical period, but after the fixing of spelling so that sign for medial [s] remained same as that for (always voiceless) initial [s], Q s became voiced medially between vowels, or between a vowel and a voiced sound (nasal or continuant) ... The earliest of these voicings, possibly already in CE, or at least AQ, was in sm, sn > zm, zn ... While s was still voiceless sr > sř > ss (similar to developments after stops); but sl appears to have remained as such, and asyllabic. Hence it later became zl > ll. Later forms have zar, z-l ... sy, sw. Before voicing of medial [s], these combinations produced (long) voiceless y, and w: [ɧ] and [ƕ]: that coalesced with the products χy, χw (< khy, khw), as they did initially (OP1: PE19/49).

As indicated above, the sound change whereby [sr] became [ss] was an ancient one. If s came into contact with r later on it was voiced; the same was true in the 1950s (OP2: PE19/102). Although not listed above, s was also voiced before voiced stops, but these sound changes occurred in Primitive Elvish whereas other voicings occurred only in child languages (PE18/39), and likewise in the 1950s (PE18/102). In the quote above, s was not voiced before y, w; however, this is not reflected in contemporaneous examples from The Etymologies of the 1930s where we see sy > ry:

• ᴹ√PIS > ᴹQ. pirya “juice” (Ety/PIS).
• ᴹ√SUS > ᴹQ. surya “spirant consonant” (Ety/SUS).

Thus it seems that in the 1930s, s became z before voiced stops in Primitive Elvish, before nasals in Ancient Quenya, between vowels in Parmaquesta [PQ] and before liquids and semi-vowels (l, r, y, w) at some indeterminate point. The relative lateness of the intervocalic change is supported by the fact that [z] continued to be the normal pronunciation even into Tarquesta, and [z] > [r] only in Lindarin [precursor to Vanyarin] (OP1: PE19/33). This was not, however, reflected in the contemporaneous examples from The Etymologies of the 1930s, which consistently show r rather than z; see the discussion on how [z] became [r] for further details.

The sound changes of s > z between vowels and before nasals, liquids and semi-vowels were essentially the same in the 1950s, but the phonetic developments occurred in late AQ rather than early PQ:

s remained unchanged initially. In the earlier AQ it also remained unchanged medially between vowels. But already in both Vanyarin and Noldorin & Telerin somewhat before the classical PQ (written in Feanorian characters) it became voiced to z between vowels. A special sign for z was thus provided in the Feanorian alphabet; but (since cases in which s = z were clear) older s-spelling was often retained (OP2: PE19/73).

Later but already accomplished in AQ s became voiced medially (a) between vowels, (b) between a vowel and a voiced continuant (nasal, or r, l) ... The change of sm, sn > zm, zn was the earliest ... after the voicing of intervocalic s > z, sr > zr > rr ... sl remained as such until the voicing; then zl > ll ... Before y, w. Here before further change sy, sw > zy, zw (OP2: PE19/101-2).

As initially written in OP2, Tolkien attributed these sound changes to PQ, but this was revised to AQ (OP2: PE19/101 note #159). Tolkien may have pushed these changes backwards in time when he decided the Quenya rhotacism [z > r] was characteristic of the Ñoldorin Quenya dialect rather than Vanyarin, to allow for more time for distinct developments in the two dialects before the Exile.

There is a note from the 1960s indicating the change of intervocalic s > z > r was not universal, stating that this change did not occur when the s preceded a stressed vowel (VT44/20 note #4). Examples include ósanwë (VT41/5) and asambar(o) (VT48/20). As pointed out by Wynne, Smith, and Hostetter (VT44/20 note #4), this resembles the Germanic Verner’s Law where s > z > r was likewise prevented, though by a preceding stress rather than a following stress.

Conceptual Development: In the Early Qenya of the 1910s and 1920s, s was voiced before voiced continuants and nasals:

s gave z and then r before l, r, m, n, w̑, y̑, ƀ, ʒ, giving ll, ´r, rn, rm, rw, ry, rb [or] rw, rg [or] ´r (Qenya Phonology, 1910s, PE12/19).

However, intervocalic s survived:

• ᴱ√NASA > ᴱQ. náse “loathing” (QL/64).
• ᴱ√LESE > ᴱQ. lese- “gather” (QL/53).
• ᴱ√USU > ᴱQ. usu- “escape” (QL/98).

In fact, in Early Qenya a final [s] became [r] as opposed to later conceptual stages where it survived. Thus the Early Qenya developments (s surviving between vowels but > r finally) are nearly the opposite of the sounds changes in the 1930s and later (s surviving finally but > r between vowels). See the entry on the Quenya rhotacism for more details.

@@@ shift to MQ