Theory of Wh-Movement
Insights into the pied-piping mechanism should come from the domain of wh-movement, where a variety of categories can be displaced. A movement is essentially limited to DP. Whatever principle of Universal Grammar (UG) lies behind this restriction on A-movement would make an additional feature superfluous, so that if some feature is involved in pied-piping in general, morphophonological realization of that feature is likely to be null in the domain of A-movement. In wh-movement, on the other hand, the working of the pied-piper feature will not be superfluous, since deciding which category to copy for Merge is not a trivial question. One major goal of this chapter is to present evidence that there are indeed wh-movement cases in which the pied-piper feature is morphophonologically realized.
As the background for discussion, let us consider the features that take part in wh-dependencies. On the basis of the analysis of f-feature agreement, Chomsky (2000, 2001, 2004) proposes that for the Agree operation to apply, the goal as well as the probe must be active in the sense of containing uninterpretable features. In the A-movement system, the probe is a set of uninterpretable f-features, and the goal a set of interpretable f features and structural Case. Chomsky (2000, 128) suggests that the wh movement system will mimic the A-movement system rather closely if the probe is an uninterpretable Q-feature of C and the wh-phrase contains an interpretable Q-feature and an uninterpretable wh-feature, as in (1).
(1)
It should be mentioned, however, that Chomsky (2001, 48n57) leaves completely open what kind of features are involved in wh-movement. In Watanabe 2002, I propose, given the analysis of loss of wh-movement in Old Japanese, that what renders the goal active is an uninterpretable focus feature, which exhibits focus morphology but has no inherent semantic content. (See Watanabe 2004a on the morphological realization of this feature in negative concord.) In Watanabe 2004c, I argue further that the Q-feature of C is interpretable. It follows that for Agree to apply, only the goal needs to be active.
(2)
This is a natural hypothesis in the bottom-up derivational system. The idea is that once an uninterpretable feature is thrown in, the computational system gets alerted. In this sense, the requirement that the probe be active is superfluous.
To review the part of the arguments in Watanabe 2004c about the Q feature of C that will become relevant later in this chapter, there are wh in-situ languages such as Chinese and Lakhota in which only wh-adjuncts are sensitive to wh-islands, as in (3).
(3)
Tsai (1994, 1999) proposes that Chinese should be analyzed as employing unselective binding for argument wh-in-situ and LF movement for adjunct wh-in-situ. Lakhota receives essentially the same analysis, as argued by Bonneau (1992). In the framework proposed by Chomsky (2000, 2001, 2004), this means that only a wh-adjunct undergoes Agree, driven by an uninterpretable feature of the wh-phrase. (Also see Cheng and Rooryck 2002.) At the same time, there is no evidence that the interpretability of the Q-feature of the interrogative C differs depending on whether it is associated with an argument wh-phrase or an adjunct wh-phrase. Furthermore, wh-in-situ in general must always undergo Agree in languages like Japanese where the wh-island effect is detected even for argument wh phrases, as in (4).
(4)
Again, there is no evidence that the status of the Q-feature varies cross linguistically. On the other hand, it is not implausible to assume the presence of an additional uninterpretable feature on wh-adjuncts in Chinese and Lakhota or on wh-phrases in general in Japanese. The morphosyntax of wh-phrases differs between Chinese and Lakhota on the one hand and Japanese on the other, as discussed by Tsai and reviewed later on. Thus, (5) should be chosen as the system of agreement for wh-movement.
