Abstract
This paper is an attempt to show that wh-words can be divided into two kinds. One is an NP wh-word, and the other is a non-NP wh-word. The former leaves a trace, and its distribution is constrained by the binding theory, while the latter does not leave a trace, and its distribution is constrained by a mechanism other than the binding theory. In this paper, I will show that the binding theory is superior to ECP in constraining the distribution of an NP wh-word and investigate the constraint on the distribution of a non-NP wh-word.
