Questions in Hindi
Question formation in Hindi has drawn considerable attention in the literature, mainly because of work by Dayal (1994, 1996, 2000) and Mahajan (1990, 2000). Local wh-movement in Hindi seems to behave like local rel-movement in having the option of pronouncing either the lower copy or the higher copy; the phonological condition PF Relaxation is also operative.
(1)
Furthermore, if question formation involves crossing a finite CP, like long-distance rel-movement in Hindi, then the highest copy must be pronounced in order to derive a matrix question reading. In (2a), the only possible reading is the embedded question reading. In order to derive a matrix question reading, the wh-phrase in (2b) must be moved into the matrix clause. As demonstrated in (2c), there is an alternative way to achieve the matrix question reading—namely, to use the scope-marking strategy.
(2)
Following Mahajan’s (2000) proposal, I assume that (2b) and (2c) in volve two distinct constructions, which coexist in languages like Hindi and German. Mahajan suggests that (2b) involves long-distance scram bling, which is distinct from successive-cyclic wh-movement. As shown above, successive-cyclic A ¯-movement moves through Spec,CPs to reach its scope position; therefore, it leaves copies behind. However, long distance scrambling does not work the same way; it does not proceed through Spec,CPs and does not leave copies in those positions.
A question arises at this point: even if we assume that long-distance scrambling does not leave any copies along the way, why do we not have the choice of pronouncing the original copy as in (3)?
(3)
The answer to this question is that, like rel-movement in Hindi, long distance wh-movement in Hindi is constrained by Minimize Mismatch, according to which PF and LF prefer to pronounce the same copy. As a result of this economy condition, PF and LF will choose the higher copy inside the matrix clause, yielding (2b). The grammaticality of (2b) sug gests that wh-phrases in Hindi, unlike those in Punjabi, are not clause bound. A question arises at this point: why is long-distance rel movement in Punjabi not constrained by Minimize Mismatch? I suggest that, in Punjabi, PF Relaxation constrains both local and long-distance rel-movement; therefore, the economy condition Minimize Mismatch does not come into play. In sum, PF Relaxation and Minimize Mismatch operate in complementary distribution, as illustrated in table 1 This complementary distribution suggests that it may be possible to unify PF Relaxation and Minimize Mismatch or at least derive them from the same initial premises. I leave this unification for further work.
Table 1
