Labeling via feature inheritance and sharing meets freezing effects

This paper deduces the freezing effects from Chomsky’s (2008, 2013, 2015) Labeling Algorithm. According to this Algorithm, when a minimal projection and a non-minimal projection merge, a minimal projection determines the label of the merger. When two non-minimal projections merge, there are two ways of implementing the labeling. One is via the trace convention; traces are ignored for the labeling algorithm. The other is via feature-sharing; the prominent features that are shared by two non-minimal projections also provide a label for the merger. We suggest in this paper that feature sharing is implemented by feature inheritance from a higher head to the head of its complement. Given the Labeling Algorithm, the freezing effects are to be accounted for in this paper. It is widely known that there is an asymmetry in sub-extraction from subject and object. The latter generally allows sub-extraction out of it, whereas the former does not. We argue following the long tradition of previous studies on this topic that when an element is base-merged as the complement of a head, it allows sub-extraction out of it. However, when the element which is labeling-wise unresolved is merged with a non-minimal projection, it disallows sub-extraction out of it. The freezing effects in various constructions follow from the system of labeling via feature inheritance and sharing.