Everybody acquainted with colloquial English knows that Yes has alternations in pronunciation: it may lose its final [s] and add a centralizing offglide to become Yeah, and it may pick up an alternative final consonant, an unreleased [p] (simulating the sudden closure of the lips at the end of the utterance), to make Yep. No also gets a final unreleased [p] sometimes, hence the spelling Nope (notice that in each case there is a conventional spelling of the [p]-final pronunciation for use when direct reporting speech, e.g. in novels). But my colleague Heinz Giegerich just pointed out to me a surprising constraint on the final-[p] pronunciations: for a long time those pronunciations have been current only as single word utterances.
In particular, he noted (on receiving an email from a Chinese student who agreed to a meeting by writing "Oh, yep", and noticing that it seemed odd) that the [p]-final pronunciations don't seem to occur when preceded by the interjection oh.
Now, a hundred commenters are going to waste their time and yours by pointing out below (without even reading as far as this) that they have heard Oh, yep, so I'm wrong. I can't do anything about them; but for you, who have continued reading into the third paragraph, let me make it very clear that of course I agree the constraint is not absolute. Naturally in a case of casual speech like this you get plenty of individual variation. But even raw Google hit counts are pretty convincing evidence that Heinz is right on the level of unusualness. It is true that you can find about 160 thousand Google hits on the web for oh yep. But the point is that you can find about 24 million for oh yeah, and about 60 million for oh yes. You can also find about 60 thousand Google hits on the web for oh nope; but you can find about 24 million for oh no. These are really substantial differences — three orders of magnitude. I suspect that they are bigger than can be accounted for by the observation that yep and nope are found only in direct reported speech while yes and no occur in purely written English.
One small additional experiment confirms this. The Wall Street Journal corpus of 1987-89 newspaper prose, beloved of computational linguists because it offers 44 million words of easily searchable and freely available raw text for experimenting on, contains 4 cases of people quoted as saying "Yep" and 11 of "Nope" (note the capital letters there), but none at all of yep or nope. Which means that there are zero occurrences of oh yep and oh nope (though there are 24 of "Oh, yes" or "Oh, yeah", and 15 of "Oh, no").
Just to forestall one distraction, let me note that a conversational turn may consist of of the one-word utterance "Yep" followed by another independent clause, of course: we find "Yep, except for the fish" in the WSJ corpus. But that's really two utterances, phonologically and syntactically (the second is expanding on the first, so it's like I completely avoid alcohol. Except at weekends). Yep and nope are always the first word in their sentence, and I think they have to be regarded also as the only word in their sentence (certainly the only word in their phonological phrase).
There's more: in the Google results there are clear signs that oh yep and oh nope are special in that they have a sociolinguistic signalling status: the hits for them are all slang-peppered teen blogs, Facebook pages (there is a special one just for oh nope!), rap lyrics, and so forth. These are not just alternates for oh yes and oh no; they are slangy signals for membership in what we could call (in an allusion to my embarrassing overestimate of the recency of the latest semantic value of the word random) the rando generation.
Heinz Giegerich had noticed (without noticing that he had noticed), from his tacit grasp of the statistical structure of the set of all spoken utterances he had heard since he learned English a few decades ago, that he had been hearing Yep and Nope his whole English-speaking life, but he had not been hearing Oh, yep! or Oh, nope!. He was right.
It's not that the language couldn't develop in a way that would increase their frequency; it has now started to. But it was just that for a long time it hadn't. And I, for one, hadn't noticed that, but now that he points it out, I realize I agree with him. Yep and Nope appear to have evolved as one-word utterances, and originally (it seems) they hardly ever occurred in longer utterances featuring utterance-final occurrences of the words yes and no. And that seems quite surprising to me.