I have not studied the primary data on the so-called "double modals" of which Ben Zimmer speaks, though I have of course been hearing about them occasionally for decades (if you ever give a lecture about the syntax of the English auxiliaries, people who can't think of anything else to ask will ask you in the question period whether you have anything to say about double modals). As a member of a distinguished department of Linguistics and English Language, I might be assumed to have done some research on their origins in their emergence in the American South or their possible antecedents in the British Isles; but I have not. I am off duty at the moment. But I just want to air a speculation. I think there might be no double modals at all. I think it might be just a matter of the emergence of a small number of new adverbs with a rather strong preference for being used before certain modals. One new adverb was clearly spawned from a modal when may and be merged to form maybe. That is now standard, and can be placed wherever adverbs of probability go. I think it is possible that might (in origin, the preterite form of may) has also turned into an adverb, of rather limited distribution, in certain non-standard dialects in the South. That would account for all the sequences Ben mentioned: might could, might should, might oughta, might would, might would've; all of them.
The thing is, nobody seems to get random double modals. There are between eight and a dozen modal verbs: definitely can, dare, may, must, need, ought, shall, and will, and a few other marginal items: the had of the idiom had better, the would of the idiom would rather, the is of the is to sequence that is roughly synonymous with must, and possibly for some speakers also the used of the used to construction (though for me the latter does not have auxiliary behavior at all, so it can't be a modal auxiliary; it all depends on whether you can say Used you to live near there? and It usedn't to be allowed). Some of the modals have quite peculiar restrictions: (i) shall is almost extinct for many Americans; (ii) must has no preterite so it is limited to present tense; (iii) ought takes a to-infinitival complement for many but not all English speakers; and (iv) dare and need are limited to non-affirmative contexts (though they have non-modal regular verb twins that are not). But ignoring those peculiarities, there are 64 logically possible two-modal combinations of the eight most basic modals. Here they all are for your perusal:
| can dare | can may | can must | can need |
| can ought | can shall | can will | dare can |
| dare may | dare must | dare need | dare ought |
| dare shall | dare will | may can | may dare |
| may must | may need | may ought | may shall |
| may will | must can | must dare | must may |
| must need | must ought | must shall | must will |
| need can | need dare | need may | need must |
| need ought | need shall | need will | ought can |
| ought dare | ought may | ought must | ought need |
| ought shall | ought will | shall can | shall dare |
| shall may | shall must | shall need | shall ought |
| shall will | will can | will dare | will may |
| will must | will need | will ought | will shall |
(If you want to bring in had and would and is in addition, then there are 112 - 11 = 110 combinations.)
How many of these are actually attested? My (admittedly very shallow) acquaintance with the dialects exhibiting the so-called double modals suggests that approximately none of them are. People report might being used in a way that might suggest it is an adverb favoring the position between subject and predicate, and they report a sprinkling of other combinations; but nothing systematic.
Notice, in I might have one out back you can't tell whether might is a modal or an adverb. That would be one of the key facts that led to the emergence of the adverb: if you misanalyze it as having the same structure as I possibly have one out back rather than I could have one out back, you have assumed that might is an adverb.
So like I say, this isn't research I'm doing here, I'm just speculating; I'm not on duty qua linguist. But I'm thinking it is just possible that double modals don't exist. Research could be done that might would bear on the question.
Update: I've revised a little of the above text since first writing it. And Ben Zimmer has kindly provided me with a list of the modal-modal combinations in Marianna di Paolo's data (1989). In addition to the cases where might occurs before a verb phrase beginning with a modal (she has might could, might oughta, might can, might should, might would, might had better/might better, and might supposed to, also incorrectly listing might've used to as a double-modal use), there are cases where may is used in a very similar way, i.e., in a way that suggests it could just be an alternant for maybe (she gathered may could, may can, may will, may should, may supposed to, and may used to; she also incorrectly cites may need to, but that is just may with the regular verb need).
However, apart from the may/might examples there are just these cases:
should oughta
can might
used to could
musta coulda
would better
could might
oughta could
better can
Plenty of food for thought there; all I'm pointing out is that (as di Paolo herself noted) it's not a broad representative sample of all the logically possible combinations.
Oh, one other thing: she also gathered as case of might woulda had oughta. Now that's more like it! If cases like that spectacular beauty were commonplace (and they might be for all I know), then I wouldn't remain skeptical very long about double modals.
Posted by Geoffrey K. Pullum at November 20, 2007 05:28 AM