Discussions of "semi-modal" promise and threaten (see my previous post) generally discuss the two verbs together, under the assumption that they function similarly, the only major difference being the evaluation by the speaker (negative or positive/neutral) of the expected event. However, there are a couple of fairly striking differences in their occurence patterns which I still can't think of any explanation for.
A search of the Corpus of Contemporary American English produced the following results:
"threatens to": 1192
"promises to": 2273
"threatening to": 2335
"promising to": 847 (scarcely any examples of promise2)
"threatened to": 3738
"promised to": 3860 (few examples of promise2)
What's odd here is the difference in frequency in the progressive. Since the corpus doesn't distinguish between different usages of the verbs, I checked a sample of 100 results for each of the variations and counted the number of examples that are to be interpreted in the "semi-modal" sense. For both threatening to and promises to this was around 35-40%. For promising to (which was far less frequent to begin with) I found a total of 3 examples.
This suggests that there is an aspectual difference between the two verbs which makes it difficult to use promise2 in the progressive. Furthermore, promise2 -- in contrast to threaten2 -- appears fairly frequently with the verb complement to be.
This also seems to be the case with the regular meaning of promise and threaten: I can say "The child promises to be good", but "The child threatens to be bad" is odd. This may be related to the highly formalized performative nature of promise, which always seems to indicate a verbal act. Threatening, however, does not have to be verbal (one can threaten with a gesture). Furthermore, a promise generally seems to refer to an action which will take place in the future; while a threat may concern something that is already in its beginning phases.
What I don't see is how all of this fits together. Granted that the verbs have different situational semantics in their original sense, I'm not sure what effect this would have on the "semi-modal" usage. And, as noted in previously, this is not irrelevant to the question of their meaning, since there are correlations between modality and aspect. Even if we interpret them as having future meaning (analogous to "be going to") rather than phasal (analogous to "begin), it doesn't explain why the two verbs should behave so differently from each other.
Fortunately I'm not a linguist so I can leave these little puzzles for someone else to solve.