How does parsing with unification-based grammars differ from parsing with pure context-free grammars? The difference of course lies in the categories which need more structure than simply being atoms.