All the Eldarin languages show some tendency towards metathesis (consonant reversal) of stops followed by sonants (nasals or liquids), as sonants + stops were preferred combinations. One of the longest descriptions of this sound change appears in the first version of Tengwesta Qenderinwa (TQ1) from the 1930s (PE18/54-55):

{Quendian >>} The Eldarin dialects of Quendian appear from the beginning to have shown a predilection for medial consonant groups consisting of a continuant + stop, and especially for the sequences l, r, or homorganic nasal + stop.

Probably already in CQ began the tendency (which persisted in the derived languages, especially Qenya) to metathesis, whereby suffixal or “suppression” combinations were rearranged in the preferred order. Thus for pl, pr, tl, tr, kl, kr and other similar groups, where l, r followed a stop (or aspirated stop) transposition might occur: e.g. pl > lp, kl > lk. Thus ²√TAK “make” — *taklō “craftsman, wright” > *talkō.

Similarly suffixal kn, pn (pm), tn were frequently transposed to ŋk, mp, mp, nt. But here we have also to reckon with the fact that nasal infixion was in CQ also employed as a morphological device parallel to suffixion of n. Cases of transposition of km, tm are rare and doubtful.

Contrary to the general rule it appears that, while st was much favoured in CQ, ps, ks were preferred to sp, sk. Variation between ps and sp (zb), and ks and sk (zg) are not infrequent.

In no Quendian language did these metatheses become a fixed rule.

This tendency towards metathesis is stronger in Quenya than it is in the other Eldarin languages, and there are Quenya examples where the sound change must have occurred after the Primitive Elvish period, since it happened after Quenya-only sound changes like the Quenya syncope. Examples include:

• ✶lepene > lepne > Q. lempë “five” vs. S. leben (VT42/24; VT47/10, 24).
• ✶neter(e) [> *netre] > Q. nertë “nine” vs. S. neder (PE17/95; VT42/26).

There are also examples of metathesis occurring only in Quenya and not in Sindarin/Noldorin, particularly for stops + liquids:

• ✶aklar(a) > Q. alcar “glory” vs. S./N. aglar (PE18/36, 87; Ety/AKLA-R).
• ✶oklā > Q. olca “bad, wicked” vs. S. ogol (PE17/170).
• ᴹ✶akrā > ᴹQ. arka “narrow” vs. N. agor (Ety/AK).
• ✶satrō > Q. sarto “companion” vs. S. sador (PE17/183).

Finding examples of metathesis in Sindarin/Noldorin is a bit difficult. As noted above, one challenge with analyzing these cases is that nasal-infixion was another formational pattern in Primitive Elvish, so it is hard to say whether a specific example is the result of stop-nasal metathesis or nasal infixion. The clearest Sindarin/Noldorin examples are all stop-nasal metathesis, and the actual sound changes occurred in Primitive Elvish:

• √GAP > gapna > ✶gampa > Q. ampa, S. gamp “hook” (VT47/20).
• ᴹ√STAK > staknā > ᴹ✶stankā > ᴹQ. sanka, N. thanc “cleft, split” (Ety/STAK).

In Sindarin there are no clear examples of stop + liquid metathesis, so it is possible this was a Quenya-only sound change. However, Tolkien did give at least one Primitive Elvish example of this metathesis: *taklō “craftsman, wright” > *talkō from the first quote given above (TQ1: PE18/55). If stop-liquid metathesis is a Quenya-only sound change, it was still probably quite ancient:

But the chief reason, no doubt, was the strong predilection which Quenya showed for the sequences of sonants: m, n, ñ; l, r before stops, as against those in which the sonants followed. Transposition also occurs in Quenya in ancient forms of tr, tl, etc. > rt, lt. Of this an example occurs in C.E. *netere which in Quenya appears as nerte (VT42/26).

Another challenge in analyzing Quenya metathesis is that it was not universal. For many examples of metathesis, there are similar counter-examples where metathesis does not occur:

• Yes: ᴹ√PAT > ᴹ✶pantā > ᴹQ. panta “open” (Ety/PAT).
• No: √MAT > ✶matnā > Q. matta “food” (PE22/136).
• Yes: √UB > ✶ubmē > Q. umbë “premonition” (PE22/168).
• No: √KHAB > ✶khabnā > Q. hamna “mound” (PE19/91).
• Yes: √SATAR > ✶satrō > Q. sarto “companion” (PE17/183).
• No: ᴹ√MIT > ᴹ✶mitra > ᴹQ. mitsa “small” (EtyAC/MIT).

Where metathesis does not occur, other sound changes may obscure that fact: tn > tt, bn > mn, tr > ts. The last example ᴹQ. mitsa was deleted along with its root, but the underlying sound change (tr > ts) was discussed by Tolkien in both Outline of Phonetic Development (OP1) from the 1940s and the Outline of Phonology (OP2) from the 1950s (PE19/43, PE19/86).

Examples of stop + sonant metathesis are more common for voiceless stops than they are for voiced stops. In the case of voiced stops + nasals, the frequency of such metatheses is obscured by nasal-infixion, as noted above. In the case of voiced stops + liquids, Tolkien himself said these changes were rare:

[Voiced stops] before r, l ... Metathesis occasionally occurred in these combinations, notably in dl > ld (not ll); and in bl > lb (OP1: PE19/46).

As with the voiceless stops metathesis occurred at an early (possibly even CE) period of voiced stops followed by r, l. But examples in Quenya are rarer in the case of the voiced stops, and chiefly limited to dl, bl > ld, lb. The metathesis of dl (? ðl) > ld seems also to have occurred again later (OP2: PE19/95).

In the 1970s, Tolkien deleted the bl > lb metathesis in green ball point pen (PE19/95 note #127), meaning the only remaining metathesis of this type was dl > ld. There are barely any attested words that demonstrate these voiced stops + liquid metatheses: the only potential example I have found is the etymology of Q. Elda. At various points Tolkien derived this word from primitive ✶edelā (PE17/139, PE17/152), and thus it might involve dl > ld metathesis after the Quenya syncope. But even in this case (a) there were many other etymologies of this word and (b) the switch might be explained by other means, such as variation between the roots √DEL¹ and √LED (WJ/360, 363).

Tolkien also mentioned metathesis of aspirates + sonants.

[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 (OP1, 1940s, PE19/44).

[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 (OP2, 1950s, PE19/87).

The groups phr, phl; thr, thl; khr, khl ... Only thr, thl were of frequent occurrence. In the case of thl metathesis occasionally reappears, so that þl > lþ > lt (OP2, 1950s, PE19/88).

It seems that, like voiced stops, the only metathesis of aspirates and liquids was the dental th + l: thl > lth > lt. I have found no words exemplifying these aspirate metatheses, aside from the usual Quenya past tenses which are more likely to represent nasal-infixion. Even the most common case of aspirates + nasals would be fairly obscure, since nasals became voiceless stops before aspirates in Ancient Quenya, so that: phm, thn, khñ > mph, nth, ñkh > pph, tth, kkh > pp, tt, kk (PE19/44, 89).

Finally, on a couple occasions, Tolkien also mentioned nasal + liquid metathesis, possibly also a Quenya-only change since both ml, mr appear in Sindarin:

[Nasals] before r, l: Here the developments were closely similar to those observed (above) where a stop preceded l, r. But after nasals certain metatheses were anciently even more frequent, and were more liable to reappear again later, namely ml > lm; mr > rm; nr > rn. This was less frequent in case of nl (OP1: PE19/47).

[Nasals] before r, l. In these combinations the developments were similar to those already observed when a stop preceded r, l. But after nasals certain metatheses were anciently more frequent than after stops, and were repeated again at later periods far more often. The “favoured” metatheses were: ml > lm; mr > rm; nr > rn; less frequent nl > ln > ld. Cf. holmo < khomlō “from the heart” us[ed] as adverb: sincerely, [?heartily] (OP2: PE19/97).

Aside from the one example of Q. holmo, there don’t seem to be any obvious examples of this nasal + liquid metathesis.

To summarize:

• Stop + Nasal metathesis likely occurred in Primitive Elvish and was shared with the Sindarin/Noldorin branch, but continued to be a factor after Quenya-only sound changes like the Quenya syncope. These metatheses with nasals seem to have occurred for voiceless stops, voiced stops and aspirates.
• Stop + Liquid metathesis may have been a Quenya-only phonetic development, though Tolkien sometimes described it as occurring in Primitive Elvish. It was most common for voiceless stops + liquids. For voiced stops and aspirates it may have been limited to dentals + l: dl, thl > ld, lth, though Tolkien mentioned bl > lb as a possibility in earlier writings.
• Tolkien also mentioned nasal + liquid metathesis, but it may have been Quenya-only and seems to be rare.

One final form of metathesis common to Quenya was the sound change whereby [sp], [sk] became [ps], [ks]. But this seems to be a later sound change, and is discussed in a separate entry.

Conceptual Development: Stop + sonant metathesis is mentioned in the very early Qenya Phonology from the 1910s as a “transposition law” (PE12/25), though at this early period the change seems to be limited to voiceless stops:

The transposition law

Stop (voiceless) + liquid (l, r) or nasal (ȵ, ŋ, ñ, n, m) (also certain combinations of liquids and sonants, see above) became, by a process of epenthesis, liquid or nasal + stop. This has no certain cases of occurrence with any spirants. The nasal became assimilated to the then following stop, i.e. pm, pn, pȵ, pŋ, etc. — all gave mp.

The exact causes for this regular and universal change are a little obscure. The following considerations probably assisted it.

The combinations mp, nt, nk, nq, lq, etc., were ones particularly favoured by Qenya and the existence of nasal (and probably liquid) infixion as a morphological method provided a great number of derivations made by the change. Forms made by infixion and suffixion would coexist.

The phonological basis is very like[ly] a change such as follows: Take t + n. tn > tň (where ň is voiceless n) > ňň > ňt. The change ňň > ňt being paralleled in Qenya dialects and Qenya itself by that of mm > mb, nn > nd.

This early stage is the only place where Tolkien described the actual mechanism of the metathesis. According to this early “transposition law”, metathesis was a universal change, as opposed to later conceptual stages where the sound change was more sporadic. Despite this “universality”, clear examples of metathesis are hard to find, but they do exist:

• ᴱ✶tet’l-asse > ᴱQ. teltasse “awning” vs. G. teglos (GL/70).
• ᴱ√ẎATA > ᴱQ. yarta “yoke” vs. G. gadr “joint, link of chain” (QL/105; GL/36).
• ᴱ√VETE > ᴱQ. velte “beard” (QL/101).

There are also numerous mp, nt, nk words derived from roots ending in p, t, k, but these might be the result of nasal-infixion, which was likely a formative pattern at the earliest conceptual stages given the frequency of nasal-infixed past tenses.

The presence of vocalic ṇ, ṛ, ḷ at these early conceptual stages also complicates the picture. There are examples where metathesis seems not to occur, but they all involve syllabic final ḷ, such as: ᴱQ. etl, litl, patl, rotl, sutl.