In Sindarin and Noldorin any palatalized velars (with an added y-sound) at the beginning of words lost this palatalization and became a simple velar. There is a similar blending of palatals with velars in Proto-Keltic (WGHC/§85, §88, §92i), though this real-world development is closer to the Gnomish phonetic history than Sindarin/Noldorin (see below). A clear description of this change appears in the second version of the Tengwesta Qenderinwa (TQ2) from around 1950:

Since T., N. both convert ky-series into simple series (T. > dental, N. to k-series), initially T. n also = CQ ñy (PE18/103).

These sound changes also appear for the phonetic development of initial palatalized velars for Noldorin in the Comparative Tables from the 1930s (PE19/18-19). These same tables make it clear the changes did not occur medially, except in the case of [ŋgj] > [ŋg] (PE19/23). A possible example of this medial development also appears in The Etymologies: ON. ango < ᴹ√ÑGYŌ (Ety/ÑGYŌ), though here the waters are muddied by the fact that the primitive form likely started with a syllabic ṇ̃.

Since the Telerin and Noldorin/Sindarin developments were distinct, this sound change could not have occurred in Ancient Telerin, although Tolkien did vacillate quite a bit on the exact Telerin developments. Since primitive initial [ŋj-] (ñy-) became [j-] (y-) in Sindarin, we also know that [ŋ] vanished before the loss of [j] (y):

[initial] ŋ̃ at first remained until ŋ̃w > m ... ñy > y not nil. This shows loss of y in ky-series was later than Noldorin, Telerin labialization of kw-series (PE18/104).

Probably the best-known example of this changes is S. celeb “silver” derived from primitive ✶kyelepē, and where [kj-] > [k-] (Let/426, UT/266). The Quenya and Telerin cognates also illustrate the typical developments in those languages: Q. tyelpë ([kj-] > [tj-]) and T. telpe ([kj-] > [t-]). Examples not involving initial ky- > c- are hard to find in Sindarin, but there are a variety of attested examples in Noldorin:

• [kj-] > [k-]: ᴹ√KYELEK > N. celeg “swift, agile” vs. ᴹQ. tyelka (Ety/KYELEK).
• [gj-] > [g-]: ᴹ✶gyernā > N. gern “worn, old (of things)” vs. ᴹQ. yerna (Ety/GYER).
• [skj-] > [sk-]: ᴹ✶skyapat- > ON. skhapa(t) > N. habad “shoe” vs. ᴹQ. hyapat (Ety/SKYAP).
• [kʰj-] > [kʰ-]: ᴹ√KHYAR > N. harad “south” vs. ᴹQ. hyarmen (Ety/KHYAR).

The last example appears in Sindarin as well (LotR/1115), though Tolkien did not explicitly describe its etymology.

There is one example where this change might have occurred medially: in the development of the past tense of the verb S. caw- “to taste” from the root √KYAW:

• akjāwē-n > S. agauwen > S. agowen (PE22/152).

Here we might have had [akjāwen] > [akāwen] and later the [k] voiced to [g] as it always did after vowels. However, it is equally likely that the consonant in the past tense was reformed by analogy with the present. Other examples, like S. cai “hedge” < ✶kegyā (UT/282) and N. breig “wild, fierce” < ᴹ✶mrekjā (Ety/MERÉK, EtyAC/MERÉK), instead imply that the sound change did not occur medially. Otherwise, the [i] in these forms is hard to explain, since it must have developed from the preserved [j].

Conceptual Development: There is a similar development in Gnomish where palatal consonants became velars. The earliest version of Primitive Elvish had a palatal c-series rather than the later ky-series. Tolkien mentioned a similar sound change explicitly in the Qenya Phonology from the 1910s, but at first he posited the palatal c-series merged with the dental t-series:

(3) [c-series] was related to (4) [t-series] exactly similarly, and have in many dialects fallen almost entirely in (4), especially Noldorin [Gnomish?] (PE12/16).

Examples from the Gnomish lexicon make it clear that Tolkien actually had palatals become velars:

• ᴱ√DYULU > G. gaul “burden” vs. ᴱQ. yulu- “to carry” (GL/38).
• ᴱ√TYAVA > G. caf- “to taste” vs. ᴱQ. tyava- (GL/24, QL/49).
• ᴱ√HYAÐA (or χ̑ṇđ) [> *χanna-] > G. †hanna- “to mow, cleave” vs. ᴱQ. hyanda- (GL/48, QL/41).

Note that dy, ty, hy are one of the representations Tolkien used for primitive palatal sounds in Early Period Primitive Elvish, beside c, j, χ̑ (IPA [c], [ɟ], [ç]). This Gnomish development is closer to the real-world development of Proto-Keltic, since it was based on the same pure palatal sounds present in Primitive Indo-European. Aside from the shift from primitive palatals to primitive palatalized velars, the Gnomish developments are essentially the same as in Noldorin/Sindarin, indicating this particular phonetic evolution was well-established in Tolkien’s mind.

See OS. [j] was lost after initial velars for discussion.

HPG/§2.1, HPG/§2.4