In Quenya the combinations sp, sk were frequently transposed to ps, ks: universally in the case of medial ps, sporadically in other cases. This sound change happened both initially and medially, and the initial transpositions introduced the new initial clusters ps-, ks- to Quenya. Although established in Early Qenya, the frequency and nature of these sound changes evolved over Tolkien’s life.

Conceptual Development: Tolkien mentioned these transpositions all the way back in the Qenya Phonology of the 1910s:

An “s” transposition period must be assumed that must be held to have been early discontinued and to have been uncertain in its area. sq > x is rare, sc > x is commoner than sc. st is never transposed; sty is never transposed; sp is almost unknown, ps being favoured (PE12/19).

The combination sp does not appear at all in Early Qenya (or in later iterations of the language for that matter), but sk is fairly common. There are, however, a few examples of transposition to ks in Early Qenya:

• ᴱ√M(B)ASA > ᴱQ. maksa- “cook” (QL/59).
• ᴱ√OSO¹ > ᴱQ. oksa “joint” (QL/71).

Tolkien did not introduce initial sk-, sp- into Primitive Elvish until the 1920s (PE14/64). At this conceptual stage, these initial clusters became spirants, and did not undergo metathesis:

• ᴱ✶skapya- > ᴱQ. hapta- “eject” (PE14/66).
• ᴱ✶skantá > ᴱQ. hanta “a blow with an axe” (PE13/147).

In the Outline of Phonetic Development (OP1) from the 1940s Tolkien mentioned the initial metathesis of sp, sk for the first time:

The same development was common in the case of sp, st, sk which thus became [ꝑ, þ, χ] > f, th (s), h initially. But as will be seen medially Q. favoured the sequences ps, ks rather than sp, sk (while favouring st). Thus at an early period initial sp, sk (not st) became ps, ks and these combinations endured in classical PQ and were the most usual representations of Eldarin sp-, sk-. In TQ though the spelling is retained ps, ks initial are commonly rendered by simple s or ps, ks > ups, iks (OP1: PE19/36).

Despite Tolkien’s statement that these “were the most usual representations of Eldarin sp-, sk-”, all of the initial sp-, sk- became f-, h- in The Etymologies from the 1930s:

• ᴹ✶spangā > ᴹQ. fanga “beard” (Ety/SPÁNAG).
• ᴹ✶spindē > ᴹQ. finde “tress, braid of hair” (Ety/SPIN).
• ᴹ✶skalnā > ᴹQ. halda “veiled, hidden” (Ety/SKAL¹).
• ᴹ✶skarwē > ᴹQ. harwe “wound” (Ety/SKAR).

A few examples of initial ps-, ks- do appear in the 1930s and 40s, however:

• ᴹQ. psára “peel” (PE22/51).
• ᴹQ. psar- “rub” (PE22/119).
• ᴹQ. ksenna “odour” (PE22/51).
• ᴹQ. ksara- “yearn, long for” (PE22/98).

None of these examples are clearly connected to roots beginning with sp-, sk-. In fact for one of these (psar- “rub”) Tolkien made a point that it was not related to the root √SPAR “hunt” (PE18/94). There is also one example of initial ts-: tsette “fly” (PE22/51), which was mentioned in at least one other place as a rare initial combination:

A notable feature of Quenya was the frequent reversal of sp, sk (sph/skh) initially and medially to ps, ks, the latter (with rarer ts) being tolerated initial combinations in Q. down to end of Classical period (PE18/93).

Medially sp is entirely absent in the 1930s and 40s, but sk did occasionally appear, as in ᴹQ. liske “reed” (PE19/51) and ᴹQ. askat- “break asunder” (SD/310). Medial metathesis to ks became common, however, most notably in The Etymologies:

• ᴹ✶mazgā [> maska] > ᴹQ. maksa “soft, pliant” (Ety/MASAG).
• ᴹ√MISK > ᴹQ. miksa “wet” (Ety/MISK).

The OP1 scenario remained essentially the same in the Outline of Phonology (OP2) from the 1950s:

sp, st, sk. For these initial groups a development to f, þ, h identical with that of ph, th, kh but probably coalescing with these only at the spirantal stage, is well evidenced, & occurs in many words and stems ... Only, however, in the case of st is the spirantal development to þ (later TQ s as in the case of original th) the normal one in Quenya. Of sp and sk the most frequent development in PQ was ps-, ks- (OP2: PE19/78).

In OP2, Tolkien explicitly mentioned that skw was immune to metathesis, as in Q. usquë:

sk was normally in older words transposed to ks, but (rare) skw was maintained (OP2: PE19/87).

Again, despite the statements above, the spirantalization of initial sp-, sk- was the norm in the 1950s and 60s. Where metathesis from primitive roots beginning with sp-, sk- occurs, it was after a vocalic augment, and thus may represent medial metathesis instead:

• √SPIN > Q. ipsin “fine thread” (PE17/17).
• ✶skalā > (archaic) Q. †ixal “cast shadow” beside Q. hala showing spirantalization (PE17/184).

Nevertheless, words with initial ps-, ks- (x-) also appear in the 1950s and 60s: Q. psar- “rub” (PE22/155), Q. xiétë “passing, impermanent” (PE18/94). Medial sp remains absent, but sk (or sc) also continues to appear in some words: Q. rusco “fox” (PM/353), Q. ascénima “visible, easily seen” (PE17/148). There are also some sk/x variants: uskare/uxare “doing wrong” (PE17/151), rusca/ruxa “wroth” (PE17/188).

The last example indicates that sk metathesis could occur even from s derived from þ [θ]. The word ruxa is derived from the root √(U)RUÞ, and the likeliest development is ruþka > ruska > ruxa (ruksa). If so, sp, sk metathesis likely remained an active sound change in Quenya for a long time. The notes above indicate the sound change began at least by Parmaquesta [PQ] but if þk > sk > ks holds in the above example then the sound change remained active into (Ñoldorin) Tarquesta [TQ].

Neo-Quenya: For purposes of Neo-Quenya, I think it is best to assume that medial metathesis of sp is universal but sk is sporadic. Despite Tolkien’s statement that initial sp and sk metathesis was common, there are no actual examples. I think it is better to assume that the initial clusters ps-, ks- were introduced sometime in late AQ or early PQ (with rarer ts-) for newer words, perhaps aided by metathesis of sp, sk after vocalic augments. In any case, these initial combinations seem to be unusual (especially ts-) and in Tarquesta pronunciation initial [ps], [ks] became [ups], [iks].