> As a sideline, has anybody else noticed that in the construction
> 'I have to', people often pronounce 'have' as if it were 'haff'?
> Perhaps this, too, is for emphasis.  I can't think of any other
> situation in which it happens.
>	' Clean your room!' Do I *haff* to?'
Actually, lots of people have noticed it; it's a standard topic in ESL classes, because English learners have to learn that the pronunciation is different from what one might expect. For example:
   I have to shovel the driveway.       /haeft@/
   I have two shovels in the driveway	/haevtu/
The reason is that this have to is a periphrastic modal construction (periphrastic means composed of a phrase -- several words -- instead of just a different inflected version of a single word, e.g Latin perfect passives are periphrastic because they use an auxiliary verb), and those virtually all have special pronunciations.


       want to   =   wanna   =   /wan@/ with flap /n/
 (be) going to   =   gonna   =   /g@n@/ with flap /n/
      ought to   =   oughta  =   /Ot@/ with flap /t/ = /d/
    used to (no standard eye dialect) = /yust@/ with /t/
     (compare is used to Verb (with) with used to Verb)
All of these (except used to, which is a periphrastic imperfect tense) refer to modals. Ought to is should, have to is must, going to is will, and so is want to, but in the other sense of will (= willing). Modals need some kind of paraphrases in English, since they're defective verbs and can't form past tenses, for instance.

That is, there isn't any way to use must in the past, so we use had to instead. Also, negatives compound differently with them: You don't have to do it is not the same as You must not do it, for instance.

All of these honorary modals have peculiarities of pronunciation that mark them as special, and the devoicing of the final /v/ in have to before the voiceless /t/ is an example of that.

-- followup --

>And "hafta" isn't a "fast speech" version of "have to"?
No, not really. The form represented in "eye dialect" by hafta is the only pronunciation of the modal paraphrase represented in spelling by have to. As has been pointed out,
These are the things I had to do.
is ambiguous. This shows that there are two senses, since where would the ambiguity come from otherwise? The two senses can be made unambiguous by separating the had from the to in one case only:
I had those things to do.
I had to do those things.
Both would relativize to
Those are the things I had to do.
But they're not the same in meaning.

As to "fast speech", that just means one of the ordinary phonological instantiations of "underlying" phrases, more or less like the written version. This can get pretty arcane (as any non-native speaker can testify), but it's different from canonical speech, and have to in its sense of must is canonically pronounced hafta /haeft@/, always with an /f/, never with a /v/, no matter how slowly or rapidly one is speaking. Whereas the putatively similar phrase have two in I have two shovels is normally pronounced with a voiced /v/, even before the voiceless /t/, unless the speech is very fast indeed.

  - John Lawler       Linguistics Department and Residential College     University of Michigan

    "Language is the most  massive  and  inclusive  art  we know,  a           - Edward Sapir
      mountainous and anonymous work of unconscious generations."       Language (1921)

More English Grammar   More About Language   The Eclectic Company   The Chomskybot