Wh-questions on subjects differ from other argument extractions in not having subject-auxiliary inversion. This means that in subject wh-questions the linear order of the constituents is the same as in declaratives so it is difficult to tell whether the subject has moved out of position or not (see [#!heycock/kroch93gagl!#] for arguments for and against moved subject). The English XTAG treatment of subject extractions assumes the following:
This type of `blocking' account is not applicable to subject wh-questions because there is no obvious candidate to do the blocking. Similarity between subject wh-questions and echo questions is also lacking. At least one account of echo questions ([#!hockey94!#]) argues that echo questions are not ordinary wh-questions at all, but rather focus constructions in which the wh-item is the focus. Clearly, this is not applicable to subject wh-questions. So it seems that treating subject wh-questions similarly to other wh-extractions is more justified than an in situ treatment. Given these assumptions, there must be separate trees in each tree family for subject extractions. The declarative tree cannot be used even though the linear order is the same because the structure is different. Since topicalizations are not allowed, the <wh> feature for the extracted NP node is set in these trees to +. The lack of subject-auxiliary inversion is handled by the absence of the <invlink> feature. Without the presence of this feature, the <wh> and <inv> are never linked, so inversion can not occur. Like other wh-extractions, the Sq node is marked <extracted>=+ to constrain the occurrence of these trees in embedded sentences. The tree in Figure 14.2 is an example of a subject wh-question tree.