This root appeared in notes probably dating to the early 1960s as the basis for words for “speech”, probably replacing another root √KARAP of similar meaning, and with a deleted variant √PATAK (PE17/126). The root √PAKAT itself was probably a restoration of a (hypothetical) early root ✱ᴱ√PAKATA, the likely basis for “speech” words in the Gnomish Lexicon of the 1910s such as G. pactha- “utter, speak, talk” and G. paithron “orator” (GL/63).
Primitive elvish
pak
root. close, shut
pakat
root. *talk, speech
karap
root. *talk, speech
The root √KARAP appeared with various forms and meanings throughout Tolkien’s life. Perhaps the earliest iteration was ᴱ√KᴬRPᴬR [KṚPṚ] “pluck” from the Qenya Lexicon of the 1910s (QL/45), with derivatives in the contemporaneous Gnomish Lexicon as well such as G. carp “bundle, bunch” and G. crib- “gather, pluck” (GL/25, 27). The verb form ᴱ✶kṛp- > ᴱQ. karpe “pluck” reappeared in the Early Qenya grammar of the 1920s (PE14/58).
The root ᴹ√KARPA appeared unglossed in the first version of Tengwesta Qenderinwa (TQ1) from the 1930s as an example of a KALTA-stem root, but whether it was related to earlier √KṚPṚ is unclear. Possibly also related is N. carab “hat”, appearing in N. Amon Carab which was an earlier name of S. Amon Rûdh.
Finally, the root √KARAP appeared in notes probably dating to the early 1960s, serving as the basis for various “speech” words such as Q. carpa-/S. carfa- “talk, speak, use tongue”, but that root was probably replaced by √PAKAT (PE17/126). Given the muddled history of this root, I would mostly avoid its use for purposes of Neo-Eldarin, though it is needed to justify N. carab “hat”.
This root appears in a list of roots from 1959-60 as the opposite of √LAT “open, unenclosed, free to entry”, with a single derivative Q. pahta “closed, shut, private” (VT41/6). It seems to refer to the state of being closed, as opposed to √SKOL which refers to the act of closing. In the much earlier Gnomish Lexicon of the 1910s, primitive ᴱ√PAKA had derivatives like G. paga “court, paved floor” and G. pagra- “pave” (GL/63), but this seems to be unrelated semantically to later √PAK “close, shut”.