In Ancient Quenya, wherever a nasal n or m appeared after a voiceless stop or aspirate, that nasal was unvoiced with the ultimate result of t (from n) and w (from m). If the preceding sound was an aspirate, it simplified to an ordinary voiceless stop. One variation in this process was pm, phm, both of which became pp instead of pw. Tolkien described this change in the Outline of Phonology (OP2), first for voiceless stops in a set of red-ink revisions to this document he wrote probably in the late 1950s or early 1960s (PE19/85):

Voiceless before nasals. pn, tn, kn; and pm, tm, km.

In these combinations the Q. tendency to intensify (and slightly aspirate) the voiceless stops, which accordingly were not voiced in contact with voiced following consonants, but tended to unvoice them, is seen most clearly. It was perhaps reinforced by the “etymological” tendency to preserve as far as possible in suffixal derivatives the identity of basic medial-consonants, of which p, t, k were the most important.

We also see that as usual in Q. the stoppage was tensest in case of dentals and weakest in case of labials. Thus the development appears therefore to have preceded in this way: pn, tn, kn* > pʰň, tʰň, kʰň with strengthening of p, t, k and unvoicing of the nasal. Similarly: pm*, tm, km > pʰm̌, tʰm̌, kʰm̌. The tenser n then > t; the weaker m > ƕ; producing

pʰt, tt, kʰt > ꝑt, tt, ht
pƕ, tƕ, kƕ > pp, tw, kw.

* These [kn, pm] were very often transposed to ŋk, mp (especially latter)

He abbreviated this description for aspirates, but the developments were basically the same (PE19/87):

[Aspirates] before nasals. The results were same as for voiceless stops, see C (d) above. The nasals did not voice voiceless stops or aspirates, but were themselves unvoiced, in natural phonetic development unaffected by grammatical or etymological analogies. Most of the possible combinations were actually of infrequent occurrence, since ph was itself infrequent medially, while the sequences of aspirate + nasal were largely reduced in number prehistorically by metathesis (as thn > nth) or by the scarcely distinguishable process of substituting nasal infixion for nasal suffixion. No certain case of phm is known.

As indicated above, the frequency of these stops-nasal combinations were reduced by metathesis, particular of kn > ŋk and pm > mp. Where such metathesis did not occur, the end results were essentially:

• pn, tn, kn and pʰn, tʰn, kʰn > pt, tt, kt > (later) ꝑt, tt, ht
• pm, tm, km and pʰm, tʰm, kʰm > pp, tw, kw.

As noted above, the combination pʰm is purely theoretical, and did not occur in practice. In the combinations pt and kt, the p and k became (or remained) slightly aspirated ultimately resulting in a voiceless spirant; this was also the ultimate result of the primitive combinations pt, kt in Quenya. The aberrant change of pm > pƕ > pp is probably a mirror of more ancient phonetic developments whereby [pw], [pʰw] became [pp], [pʰp] (PE19/86, 88); in any case such combinations were extremely rare (perhaps non-existent).

Tolkien reaffirmed these developments in various notes from the late 1960s, where he said:

better: let changes before nasals be relative[ly] late and after other changes ... tm, tn > tw (nt) > nt, tt (VT49/54 note #10). [This was first written as (rejected): tm, thm, dm > dm, tm̌h, đm > nm, tƕ, zm > nw, tw, rm (VT49/46) and there was a marginal note saying “what a muddle”]

Q k, t, p were not voiced before medial nasals

... km > km̃ > kƕ > kw
... kn > kþ > ht
... tn > tt
... tm > tm̃ > tƕ > tw
... pm > pp
... pn > pþ > pt (ft) (PE22/150).

Since metathesis of stop + nasal was common, there are relatively few examples of stop + nasal phonetic developments in Tolkien’s later writings where such a metathesis did not occur, but a few can be found:

• ✶lukma > Q. luqa [lukwa] “heavy transport wain” (PE17/28) [beside lunka produced by metathesis].
• ✶matnā > Q. matta “food” (PE22/136).
• ✶pathnā > Q. †patta “smoothed” (PE19/88) [ultimately reformed by grammatically analogy to pasta].

Conceptual Development: In the Qenya Phonology of the 1910s, Tolkien described no special special phonetic developments for voiceless stop + nasal combinations, probably because at this earliest conceptual stage, the “transposition law” (metathesis) for voiceless stops + nasals was universal rather than sporadic (PE12/25).

In the Outline of Phonetic Development (OP1) from the 1940s, however, he described phonetic developments similar to those above, but only for aspirates + nasals (PE19/44):

[Aspirates] before nasals. Nasals did not voice aspirates, but were themselves unvoiced (in pure phonetic development, unaffected by analogy). Most of these combinations were rare, since they were largely got rid of prehistorically by metathesis (or by substitution of nasal infixion for suffixion); or by avoidance of a nasal suffix.

No cases of phm occur; and no clear case of phn. Medial ph was not frequent. Probably:

thn > thň (with voiceless n) > tth > tt or analogical st.
khn > khň (with voiceless n) > kht > ht.
but thm > þm̌ (with voiceless m) > thw, TQ sw (n.b. not hw).
khm > χm̌ (with voiceless m) > hw.

These aspirate developments in the 1930s were not 100% identical to those Tolkien describes after 1960, since thm, khm became sw (< þw), hw instead of tw, kw; these same developments were seen in the first drafts of OP2 from the early 1950s (PE19/87-88 note #88). Furthermore, in OP1 voiceless stops were instead voiced before nasals, and their phonetic development merged with those of voiced stops + nasals, ultimately becoming nasal pairs:

[Voiceless stops] before nasals, which in Q. exercised a strong voicing influence on preceding consonants, p, t, k were first (already earliest AQ) voiced to b, d, g. But in the earliest PQ b, d, g, whether original or thus derived from p, t, k were nasalized to m, n, ñ. Since ñ did not occur as a suffixal consonant, and m was seldom used after labials (p, ph, b) we are concerned only with pn > bn > mn; tn > dn > nn; tm > dm > nm; kn > gn > ñn; km > gm > ñm (PE19/43).

The most notable example of this phonetic development in the 1930s was:

• ᴹ✶tekmā > tegmā > teñma > ᴹQ. tengwa “letter; writing, grammar” (Ety/TEK; PE19/43).

This voicing of voiceless stops before nasals was mentioned in the second version of the Tengwesta Qenderinwa (TQ2) around 1950 (PE18/104) and can also be seen in the first (unrevised) layer of Outline of Phonology (OP2) from the 1950s (PE19/85-86 note #79 and 87-88 note #88). It seems, however, that when Tolkien revisited the etymology of tengwa in the late 1950s in his notes on Words, Phrases and Passages in the Lord of the Rings, he became dissatisfied with this phonetic development, saying:

The phonology of k, t, p + nasals in Quenya needs revising. In changes so far mapped k, t, p + m > g, d, b > nm, nm, mn > nw (> ngw), nw, mn, make derivation from nasal... (PE17/44).

In these same notes, he revised the root of tengwa from √TEK to √TEÑ, probably because he no longer considered the km > ñm > ñw > ñgw developments to be viable. The connection between this note in PE17/44 and the revisions of the phonetic developments of stops + nasals in OP2 was suggested by Christopher Gilson (PE19/86, second half of note #79). While the exact timing of these conceptual developments isn’t clear, it seems likely that the new paradigm for stop + nasal developments was created after the note on PE17/44, probably circa 1960 or shortly before then.

One more relevant passage from the late 1960s is this:

tengwa is not from tekmā, since -mā is not passive but instrumental, and prob[ably] by rev[ised] phonology -km- does not > ŋm- > ngw. but √TEŊ “indicate, mark, signify”. teŋmā > tengwa ? “sign, token” (PE22/149).

This late note seems to show that Tolkien may have had reservations about nasal unvoicing after voiceless stops into the late 1960s. At minimum he needed to remind himself of the new revised paradigm, and the “?” in this last quote may have meant he was still unsure of it, although in this specific set of notes he went on to confirm the OP2 paradigm, as shown in the quote from PE22/150 given above. See also the “what a muddle” note from VT49/54 given above, also from the late 1960s. Tolkien made further green-ink revisions to the relevant sections of OP2 around 1970, still working on refining these ideas.

Neo-Quenya: For purposes of Neo-Quenya, it may be best to simply ignore the unvoicing of nasals after voiceless stops and retain the OP1 paradigm whereby the stops were voiced and eventually nasalized. There are nearly a dozen words from The Etymologies of the 1930s that depends on this older paradigm, and a few derivations in Tolkien’s later writings as well, such as Q. ongwe < okma (PE17/170) and Q. Utumno < Utupnŭ (MR/69). There are also a good half-dozen words from first layer of OP2 that could be salvaged if the old paradigm was retained (PE19/86, second half of note #79). Conversely, for the two words given above that depend on the new paradigm, luqa “wain” has a more common variant lunca via metathesis and matta “food” might be derived from mattā instead.