In Sindarin and Noldorin voiced stops became spirants after the liquids [l] and [r]; voiceless stops underwent a similar development, though the mechanism was different. These combinations went through a nearly identical development in Welsh (WGHC/§105ii). In fact, Tolkien compared Sindarin to Welsh to describe this sound change:

The much changed Sindarin of Middle-earth turned the stops to spirants after l, r, as did Welsh: so *alkwa > alpa (Telerin) > S. alf (spelt alph in my transcription) (VT42/7). [Although Tolkien uses a voiceless stop in this example, it seems this rule applies to voiced stops as well.]

This sound change is most obvious in the case of lb, rb > lv, rv (written lf, rf finally) and rd > rdh:

• ᴹ✶golbā > N. golf “branch” (Ety/GÓLOB).
• ᴹ√KHER + ᴹ√BES [> *kherbessē] > N. herves “wife” (Ety/BES, KHER).
• ✶gardā > S. gardh “region” (WJ/402).

The development of ld > ldh [lð] is obscured by the fact that it seems the resulting [lð] became [ll], as suggested by David Salo in Gateway to Sindarin (GS/§4.86, GS/§4.157). There are unfortunately no published examples that show the full development, so the exact series of sound changes is unclear:

• ✶kuldā [> *kolða?] > S. coll “hollow” (WJ/414).
• ᴹ√MEL > meldon [> *melðon?] > N. mellon “friend” (EtyAC/MEL); this etymology is from deleted notes but probably remains valid.

In the case of lg, rg > lgh, rgh [lɣ, rɣ], the resulting [ɣ] typically developed into a vowel, such as becoming [i] medially. The clearest examples of this phonetic development are from Noldorin (with likely intermediate developments added for clarity):

• ᴹ√(N)DER + ᴹ√GYEL [> *ndērg(y)el > *(n)dīrgel] > ON. Dirghel > N. Diriel (Ety/DER, GYEL).
• ᴹ✶stalgondō [> *sthalghondo] > N. thalion “hero, dauntless man” (Ety/STÁLAG).

Where the resulting [ɣ] becomes final after short final vowels vanished, the final [ɣ] became [a] instead. Again, the clearest examples are from Noldorin:

• ᴹ✶stalga > ON. sthalga [> *thalgh(a)] > N. thala “stalwart, steady, firm” (Ety/STÁLAG).
• ᴹ✶targā > ON. targa > targh > N. tara “tough” (combining the etymologies from Ety/TÁRAG and PE22/34).

There seems to be at least one Sindarin example of [-lga] > [-lɣ(a)] > [-la] (with the likely intermediate development added):

• ✶phelgā [> *felgh(a)] > S. fela “mine” (PE17/118).

In both Sindarin and Noldorin, voiced stops can follow liquids in some compounds:

• S. guldur “sorcery” = gûl + dûr (PE17/125).
• S. bardor “home land” = bâr + dôr (PE17/164).
• N. meldir “friend [m.]” = mela- + dîr (Ety/MEL).
• S./N. Elbereth “Star-Queen” = êl + bereth (PE17/22, Ety/EL).
• S. Borgil “Red Star” = born¹ + gil (Let/426).

It is possible that the spirantalization of voiced stops after liquids was no longer an active sound change in Sindarin and did not apply to later compounds. This might actually be the case in Sindarin names from Gondor, but another likely explanation is that the second element of these compounds originally began with a nasalized stop, and the spirantalization was prevented by the intervening nasal which was later lost. This explanation can be applied to most if not all of the examples above, such S. bardor whose primitive form is given as bar-ndor (PE17/164). For further information, see the discussion of the liquid mutation.

Conceptual Development: The Gnomish and Early Noldorin developments are complicated by the fact that voiced spirants were part of the phonetic inventory for Primitive Elvish in the 1910s and and 1920s. This means Gnomish and Early Noldorin rdh could be a preservation of primitive [rð] rather than a development from [rd]. As suggested by Roman Rausch in his Historical Phonology of Goldogrin (HPG/§2.3), there are hints that the Gnomish developments might have been different from later Sindarin/Noldorin. In particular, there are examples of medial ld in Gnomish that cannot be explained away by compounds:

• ᴱ✶ŋolđō > G. Golda “Gnome” (GL/41).

Here it seems primitive [lð] > [ld]. However, the usual development of primitive [ld] in Gnomish was to [ll]:

• ᴱ✶nalda > nalla > G. nâla “yellow-lily” (GL/59).

This makes the Sindarin/Noldorin development of [ld] > [lð] > [ll] problematic, because it is unlikely we both have [ld] > [lð] and [lð] > [ld]; perhaps [ld] > [ll] directly, and then later [lð] > [ld]. Regardless of the details it is clear the phonetic developments during this early conceptual stage were not completely the same as later Sindarin/Noldorin. Lack of examples in the Early Noldorin of the 1920s also makes it difficult to determine the sound changes at this stage, though there are examples showing that at least [rd] > [rð]:

• ᴱ✶ʒirdǝ > ᴱN. girdh “entrails” (PE13/161).
• ᴱ✶ur’-irda > ᴱN. olerdh “pitiless” (PE13/144).

GS/§4.86 WGHC/§105ii @@@