Complex Predicates in Urdu such as in (1) have a completive interpretation in the perfect, and a habitual one in the imperfect: a pattern that fits in well with most current theories of aspect and genericity.

(1)     Nadya-ne gaarii  calaa  l-ii
        N-Erg    car.Nom drive  take-Perf.F.Sg
        `Nadya has driven a/the car.'

However, complex predicates in the imperfect receive an entirely unexpected reading with exactly and only the light verbs "le" (take) and "jaa" (go).

(2)     Nadya gaarii  calaa  le-tii          hai 
        N.Nom car.Nom drive  take-Impf.F.Sg  be.Pres.Sg
        `Nadya does/will drive a car.'

This reading entails not only that Nadya is able to drive a car, but that she actually does so. It contrasts with the run-of-the-mill modal construction ("sak" `can') in the following way: while the modal could be used to describe a context in which Nadya in principle knows how to drive a car, but hasn't done so for about 20 years, this could not be the case for (2). The "le"-construction entails that Nadya does in fact exercise her ability.

A further twist to the story is provided by passives (most commonly used in the negative) which appear to provide a reading very close to the disposition/ability reading of (2). In the passives, furthermore, this reading is not limited to the imperfective, but also appears in perfects and futures.

In the talk, I will lay out the data in some detail, propose a syntactic account for the rather complex interaction of aspectual light verbs and passives with the tense/aspect system of Urdu, and review the various semantic options that have become available in the literature on generics/habituals/dispositions, within which the above data might fall (stage vs. individual level, a generic operator, or deontic vs. epistemic modality). I conclude that the bulk of the literature actually has very little to say about these kinds of readings and propose a direction in which a full-fledged semantic account might proceed (as based on the syntactic account I provide).