In Ancient Quenya of the 1950s and 1960s, the combination dy became ly both initially and medially. Tolkien described this change in several places, included the Outline of Phonology (OP2) from the 1950s:

In Q *del- seems to have become *led, by dissimilation. The past form [lende] clearly shows *led, while lelya may also be derived from *ledja, since dj became ly medially in Quenya (Quendi and Eldar, circa 1960, WJ/363).

Phonetically medial dy > ly (not ry), though this phon[etic] development [is] often set aside (OP2: PE19/70).

Phonetically [medial] dy > ly (owing to greater tensity[?] of d + y), cf. initial ly < dy or < gy (OP2: PE19/94).

The dentalization of k-series before (initial) y was early: it was also shared by Telerin. Thus gy > dy or đy (if the development was after the spirantalization of the voiced stops). The result was in any case ly. Cf. ly (not ry) as the true phonetic development of medial dy. Medial gy became ı̯y, which coallesced with preceding vowels or formed diphthongs: thus CE magyā > maiya, maia. This was because the opening of the medial stops was much earlier than in case of the initials; so that gy already to ʒy > ı̯y before dentalization (OP2: PE19/75).

The last of these quotes provides quite a few clues to the ordering of this and other sound changes. Initially, this sound change could only result from gy- > dy- after the dentalization of palatalized velars; this is because dy- was not a valid initial combination in the Primitive Elvish of the 1950s and 60s, as noted in the second version of the Tengwesta Qenderinwa (TQ2: PE18/91). However, medially, the change gy > dy did not occur, because medially spirantalization of voiced stops occurred before dentalization. Thus it seems that the order of sound changes was:

1. Spirantalization of voiced stops (medially).
2. Dentalization of palatalized velars.
3. Spirantalization of voiced stops (initially).
4. The change of dy or đy to ly initially and medially.

The exact order of the last two sound changes was not specified by Tolkien. The initial development to ly was probably tied to the rule that initial [d] became [l]. Tolkien did list ly among the valid initial combinations in Quenya in the 1950s:

ly was derived from gy; and in case only of lye “you (sg.)” by analogy from le (after nye) (OP2: PE19/70).

However, ly as an initial consonant was a later insertion, in red ball point pen (PE19/80 note #54) and not part of the original composition; see the discussion of Conceptual Development below.

Conceptual Development: In Tolkien’s earliest conception of the languages, the Primitive Elvish of the 1910s and 1920s had only a single voiced palatal stop, which Tolkien represented as j or d͡y as shown in the Qenya Phonology of the 1910s (PE12/15); this was probably a pure voiced palatal stop [ɟ]. The form DY- was used to represent this sound initially in roots in the Qenya Lexicon (QL/105-6). Its initial development was generally to y-, but in a few cases to di-: ᴱQ. yenye(n) vs. ᴱQ. die “yesterday”, both from the root ᴱ√DYĒ² (QL/105). The Qenya Phonology indicates its medial development was to ẏ (PE12/16). Ẏ appears often medially in the Qenya Lexicon and consistently produces -y-, though this sound might also have been derived medially from a primitive voiced palatal spirant [ʝ], represented by ʒ̑ or zy (PE12/15-16).

Note that the corresponding Gnomish phonetic development was to g-; compare: ᴱQ. yenye(n)/ᴱQ. die vs. G. gîr “yesterday” (QL/105, GL/38) and ᴱ√DYELE > ᴱQ. yelin vs. G. Gilim “winter” (QL/106, GL/38). Since primitive initial y also became g in Gnomish, the exact correspondences are hard to work out. Regardless, the phonetic development of primitive voiced palatal stop [ɟ] from the 1910s seems to most closely resemble the sound changes for gy- in the 1930s.

In the 1930s, there is no evidence of dy > ly. Forms with initial gy and dy (which were both valid primitive initial combination in The Etymologies of the 1930s) all developed into initial y-:

• ᴹ√DYEL > ᴹQ. yelwa “loathsome” (Ety/DYEL).
• ᴹ√DYELEK [a deleted root] > ᴹQ. yelka “sword?” (EtyAC/DYELEK).
• ᴹ√GYEL > ᴹQ. yello “call, shout” (Ety/GYEL).
• ᴹ√GYER > ᴹQ. yerna “old, worn” (Ety/GYER).

In addition, Tolkien’s description of these analogous initial developments in the Outline of Phonetic Development (OP1) from the 1940s indicated that spirantalization of voiced stops occurred before dentalization initially as well as medially, so that gy- > ʒy- > y-:

Initial groups ... gy became a spirantal y, but this fell in PQ with original y (OP1: PE19/34).

The combination ly was not listed among initial consonant clusters in the 1940s (OP1: PE19/37-38). Likewise, Tolkien’s description of the medial development of dy in OP1 resulted in ry:

[d] became [r] as between vowels before y, w (OP1: PE19/46).

This medial development is consistent with examples in The Etymologies: ᴹ√MAD > ᴹQ. marya “pale” (Ety/MAD).

In the 1950s, the first layer of composition for OP2 shows the same developments as in OP1: gy > y initially (PE19/75 note #37) and dy > ry medially (PE19/94 note #123). Furthermore, ly was a later addition (in red ball point) to the list of valid initial consonants in OP2 (PE19/80 note #54). However, the development of dy > ly was well established by 1960 (at least medially) because it is mentioned in the Quendi and Eldar essay (WJ/363, see quote above).

Neo-Quenya: In prior versions of the lexicon, I advocated replacing earlier form derived from primitive initial gy- to show ly- in Neo-Quenya, but now I am less happy with these modifications. The initial combination ly- is incredibly rare in Quenya, appear only in the pronoun: Q. lye. As noted above, the pronoun itself did not develop from gy- > dy-; it was produced by analogy with -n(yë) (PE19/80). The texts that I’ve seen that make more extensive use of ly- in Neo-Quenya frankly that look strange and wrong.

I now recommend ignoring Tolkien’s late-50s innovation of initial gy- > dy- > ly-, and assume that spirantalization preceded dentalization both initially and medially so that gy- > y-; this allows the preservation of more of Tolkien’s forms without revision. However, I do advocate using medial -dy- > -ly- since (a) it is better attested and (b) affects far fewer words; the occasional ry-variants can be more easily explained as reformation by analogy with other words. This does mean that lye is a weird isolate in Quenya, but I think that’s tolerable.

@@@ as opposed to Late Period, also medial ðy > ry (marya)