Interaction with adjoined Comps

The XTAG analysis now has two different ways of introducing a complementizer like that or for, depending upon whether it occurs in a relative clause or in sentential complementation. Relative clause complementizers substitute in (using the tree $\alpha$Comp), while sentential complementizers adjoin in (using the tree $\beta$COMPs). Cases like ((208)) where both kinds of complementizers illicitly occur together are blocked.

(207)0(207

(208)

*the book [$\epsilon$_w<<14161>>i [that [that [Muriel wrote $\epsilon$_i]]]] 

This is accomplished by setting the S_r.t:<comp> feature in the relative clause tree to nil. The S_r.t:<comp> feature of the auxiliary tree that introduces (the sentential complementation) that is set to that. This leads to a feature clash ruling out ((208)). On the other hand, if a sentential complement taking verb is adjoined in at S_r, this feature clash goes away (cf. (209)).

(208)0(208

(209)

the book [$\epsilon$_w<<14172>>i [that Beth thinks [that [Muriel wrote $\epsilon$_i]]]]