I'm moving this discussion to a new thread, since it may bore everyone but jheem and me!

jheem: The common Germanic cake is not related to to cook.
wordnerd: etymonline says, "cake: ... Not related to L. coquere "to cook," as formerly supposed." I'm curious: what changed the former supposition?
jheem: Well, if the cake words in Germanic are native and not loanwords, they could not be related to Latin coqueo 'to cook' as initial Latin c is cognate with initial Germanic h: cf. capio 'to take, seize' ~ to have, centum '100' ~ hundred, canis 'dog' ~ hound. The medial labio-velar -qu- being cognate with k is probably problematic, too.

But doesn't that prove too much? On that logic linguists couldn't have 'formerly supposed' a relationship, because (I presume) they have long been aware of the consonent shifts.

(As I understand your logic, it says at most that cookie cannot be a descendent from the same latin term that gave us cook; it does not preclude a 'cousin' relationship.)

This, from The Oxford Dictionary of Etymology, suggests to me a connection between cook and cookie, although that source does not explicitly claim a connection:

• cookie: ~ Du. koekje, dim. of koek cake
• cook: OE cōc - popL. cōcus,for L. coquus, which is directly repr., with short vowel, by OS kok (Du. kok)

But doesn't that prove too much? On that logic linguists couldn't have 'formerly supposed' that words were related, because (I presume) the linguists have long been aware of the consonent shifts.

Now I'm confused! I think the former view was that words like English to cook and German kuchen and cookie were indeed Latin loanwords, which would invalidate the consonantal shifts. But now it seems that cook was a loan but cookie was not. I haven't really read any of the pertient literature, just a couple of different etymological dictionaries review of that literature.