In both Sindarin and Noldorin, combinations of homorganic nasals and voiceless spirants became long nasals: mph, nth, ŋch > mm, nn, ng (the long nasal ŋŋ became ŋg). It seems that in Sindarin (but not Noldorin), lth underwent a similar change to long ll. In all these cases, the long consonants were initially voiceless, but were later voiced (with the possible exception of ll, see below). Tolkien explicitly described this sound change in a couple places:

Medially lþ, mph, nþ, ŋχ (though for long without change of spelling) became long voiceless l, m, n, ŋ, English transcription lh, mh, nh, ngh. In late Beleriandic Sindarin these became voiced ll, mm (m), nn, ng (= ŋg) unless followed still by a stressed syllable (phonetic notes from 1962, PE17/131).

Medially however nth (nþ), nch (ñx), mf (mp with bilabial f) and lth (lþ) became long voiceless n, ñ, m, l though the old spelling was mostly retained (beside nh, ñh, mh, lh), and among those to whom Sindarin became a language of lore, as were the men of Gondor who were or claimed to be of the Númenórean race, the spirant was reintroduced from the spelling. In true Sindarin of the Elves or Elf-friends of the early ages the final form was often reintroduced medially. In the transcription of Elvish Sindarin in The Lord of the Rings, ll is used as in the manner of modern Welsh for medial voiceless l; as in mallorn < malhorn < malþorn < malt “gold” and orn “tree” (The Rivers and Beacon-hills of Gondor, late 1960s, VT42/27).

For nasal combinations there are plenty of examples of this sound change in both Noldorin and Sindarin. Some representative examples:

• ✶mantinā > manthen > S. mannen, passive participle of S. mad- “to eat” (PE17/131).
• ✶dankĭna > daŋχen > S. danghen > S. dangen “slain” (PE17/133).
• ON. gampasse > N. gammas “s-sign” (EtyAC/GAP).
• ON. thintha > N. thinna “fade” (Ety/THIN).

The developments (phonetic and conceptual) for lth > ll are more complex. There are some notes in Tolkien’s description of the Noldorin use of the Feanorian Alphabet that indicate Tolkien was already considering this sound change in the late 1930s:

Since in late Exilic and Toleressean medial lth j3 became voiceless ll; medial lh from l + h (s) is often written unhistorically j3 (Feanorian Alphabet, 1930s, PE22/32).

However, examples in The Etymologies from the mid-1930s (mostly written before the Feanorian Alphabet document) consistently show medial lth was preserved: N. toltho “fetch” (Ety/TUL); N. nelthil “triangle” (Ety/NEL, TIL); N. pelthaes “pivot” (Ety/PEL, TAK); N. malthen “of gold” (Ety/SMAL). The last of these may be contrasted with S. mallen “golden” in Cormallen “Golden Circle” (LotR/948; RC/625). The “canonical” example of lth > lh > ll is mallorn “golden tree”, and Tolkien demonstrated the sound change in the etymology of this word in a couple places:

• OS. malþorn > malhorn > S. mallorn (VT42/27), from the quote above.
• malt + orn > malh-orn > S. mallorn (PE17/50).

The word mallorn did not appear in The Etymologies, but it did appear in Lord of the Rings drafts from the 1940s (TI/226, 233), so perhaps Tolkien introduced this sound change between writing The Etymologies and starting work on The Lord of the Rings.

It is unclear whether or not ll should be voiceless or not in mallorn. The quote above from 1962 says that ll was later revoiced, but the quote from the late 1960s says that “In the transcription of Elvish Sindarin in The Lord of the Rings, ll is used as in the manner of modern Welsh for medial voiceless l” (VT42/27). This statement is quite sweeping, and implies that all double-l in Sindarin are voiceless. However, as Carl Hostetter pointed out (VT42/31, note #70) this is contradicted by Tolkien’s pronunciation notes in The Lord of the Rings appendices: “consonants written twice, as tt, ll, ss, nn, represent long, ‘double’ consonants” (LotR/1115), with no mention of voicelessness. At a bare minimum ll of other origins, such as in S. mellon “friend”, were almost certainly not voiceless, and it is possible that ll in mallorn was also regularized back to a voiced combination.

Furthermore, despite clearly describing this change in several places, Tolkien does not seem to have fully adopted the lth > ll phonetic development, even in the Sindarin period. There are quite a few examples of medial lth from names appearing in the 1950s and 1960s: S. Belthil, S. Belthronding, S. Brilthor, S. Gilthoniel, S. Melthinorn. These names can be explained as archaic forms (they all date to the First Age), but another interesting example, S. tolthui, appears as an alternate form of tollui “eighth” in The Rivers and Beacon-hills of Gondor notes quoted above:

The forms canthui, enchui, tolthui are those of the southern Sindarin dialect adopted by the Noldor (VT42/27).

The form tolthui is contrasted with its “normal” form tollui (VT42/25), but that normal form itself was first written as (deleted) tolthui (VT42/31, note #63). Tolkien’s vacillation is understandable in the context of two other statements from the same document: “the spirant was reintroduced from the spelling” and “the final form was often reintroduced medially” (VT42/27, from the quote above). It seems that in some circumstances the sound lth was restored, either drawing from tengwar spelling (j3), or by analogy from similar words ending in final -lt. Differing pronunciations may have persisted as dialectical variations, as with tollui vs. tolthui.

Conceptual Developments: The medial combinations mf, nth, nch were fairly common in the Gnomish of the 1910s, so it seems this sound change was not part of Tolkien’s earliest ideas for the language. As noted by Roman Rausch in his Historical Phonologies of Ilkorin, Telerin and Noldorin around 1923 (HPITN/§4.1.5), there are examples of medial combinations nc, nt, mp changing to (long) nasals (probably via mph, nth, ŋch) starting with the Early Noldorin of the 1920s:

• ᴱN. engin plural of ᴱN. anc “spear” (PE13/137, 159).
• ᴱN. ennir plural of ᴱN. ant “face” (PE13/137, 160).
• ✶rankind- > ranχin > reı̯ngi(n), a verb form of ᴱN. rhang- “to slay” (PE13/152).

But there are several examples of medial nth in this period as well: ᴱN. cranthir “perfect” (PE13/161), ᴱN. huntha- “to burnish, polish” (PE13/148), ᴱN. panthig “coin” (PE13/124). It seems that Tolkien had not yet firmly decided on this sound change, or perhaps there were some less-than-obvious special conditions that inhibited this phonetic development. There are examples of lth in both Gnomish and Early Noldorin as well, which is consistent with the theory that the sound change lth > ll was not introduced until the late 1930s or early 1940s.

Neo-Sindarin: A number of Neo-Sindarin writers propose that Noldorin forms showing lth should be updated to ll to reflect the Sindarin phonetic rules described here. I personally think the lth and ll pronunciations can coexist in Neo-Sindarin as dialectical variations. In the entries in this lexicon, I consistently use the lth unless recording late Sindarin forms as written by Tolkien; this likely reflects the Sindarin tengwar spelling (see above). It is relatively straightforward to change lth to ll for those who prefer that pronunciation. The reverse transformation would be much trickier, since you would need to know which ll should not be changed back, such as in S. mellon “friend”.

If you do use the ll pronunciation from lth, I recommend sticking to long voiceless [ɬɬ] (as in Welsh) to keep it distinct form voiced [ll] of other origins.

GS/§4.112 GS/§4.118 GS/§4.201 @@@ but not -lθ- < -lt-: heltha-