Online long-distance dependency formation requires the following "component processes": identifying the dependent element (e.g., Wh-filler), identifying the controlling element (e.g., gap), holding the dependent element in working memory, linking the dependent element to the controlling element, and retrieving the filler from working memory. Many previous studies have shown that upon encountering the dependent element, the parser launches an active search for the controlling element, and the parser attempts to form and resolve the dependency as soon as possible.

Through various studies on the processing of wh-question in English and Japanese, this study attempts to reveal the nature of each component process and factor responsible for the active dependency formation. We show that the processing of long-distance dependencies is driven by the need to satisfy grammatical requirements encoded in the dependent elements (wh-elements).

Specifically, we will try to establish the following:

(1) The parser posits the controlling element for different wh-phrases in different positions, because different wh-phrase has different grammatical requirements (e.g., who/what is linked to the gap in argument position but why is linked to TP or VP).

(2) The wh-filler-gap dependency formation process is given a priority over well-known structural preference in syntactic ambiguity resolution.

(3) When the presence of the controlling element for the wh-phrase (e question particle in Japanese wh-question) is predicted, the parser tries to link the wh-phrase to the predicted Q-particle, rather than linearly local Q-particle, in order to resolve wh-Q dependency as soon as possible.