Constraint on Multiple Realization
So far, we have seen five options for realizing elements of a movement chain in Punjabi rel-movement. If the lowest copy is pronounced, (1a) results; if the intermediate one, (1b); and if the highest one, (1c). If the lowest copy and the highest copy are pronounced, (1d) results; if the highest copy and the intermediate copy are pronounced, (1e).
(1)
However, some options for multiple realization are unavailable.
(2)
Examples (2a–b) are not legitimate options. They show that multiple copy pronunciation within a given CP domain is not permitted. In addition, without any overtly realized rel-phrase in rel-movement, the sentence is ruled out as in (2c/c’). The ungrammaticality of (2c) suggests that a null operator is not available in Punjabi rel-movement; if it were, we would expect to find a structure like (3).
(3)
In addition, in the following section, I show that a null operator is also not available in Punjabi wh-movement and that recoverability of deletion requires at least one copy of a movement chain to be overtly realized. Note that the copy theory used here for deriving (1a–e) also generates the ungrammatical sentences in (2a–b). Now the question is how we can rule out these ungrammatical constructions. I propose that multiple-copy pronunciation in rel-movement in Punjabi is constrained by a PF economy condition, the CP-Phase Constraint.
(4)
In other words, two copies of the same element occurring in the same CP phase cannot be pronounced at the same time. According to Chomsky (2001), CP and vP are two phases in the derivation process. However, I have demonstrated above that rel-movement in Punjabi is an instance of successive-cyclic movement through Spec,CPs. Only CP phases seem to matter for the realization of chains.
(4) is an economy constraint from the PF point of view; pronouncing one copy of a movement chain in a given CP phase is more economical than pronouncing two. Given this constraint, it is also true that pronouncing zero copies is even more economical than pronouncing one. However, that possibility is constrained by recoverability of deletion. If a movement chain does not cross a finite clause, pronouncing zero copies will violate recoverability of deletion. If it does cross a finite clause, then pronouncing zero copies in a given CP domain is a possible PF output as long as a copy of the movement chain is pronounced in another CP phase as in (1a–c).
According to the CP-Phase Constraint, we can predict that in a movement chain schematized, repeated here as (5), pronouncing both Copy1 and Copy2 will violate the constraint since both copies ap pear in the same CP domain.
(5)
The prediction is borne out, as shown in (6). Examples (6a) and (6b) are (2b) and (2a), respectively, in which two copies of a movement chain are pronounced in the embedded CP domain, violating the PF economy constraint (4) (the CP-Phase Constraint).
(6)
The CP-Phase Constraint can further predict the ungrammaticality of (6c/c’) in Punjabi, as well as (7) in German and (8) in English.
(7)
(8)
In this section, I have shown that multiple realization of a rel movement chain is subject to both PF and locality constraints. In the following section, I focus on wh-movement in Punjabi.
