Covert morphological structure and the processing of zero-derived words

Abstract

English makes use of a wide-spread pattern of word class alternation known as ‘zero-derivation.’ This involves pairs of homophonous forms which are semantically related, yet differ in part-of-speech (e.g. a knot vs. to knot). Many theories have been proposed to describe the relationship between these forms, with some proposing that to knot is covertly derived from a knot in the same way as government is from govern; an alternative view is that these pairs are instead two forms of a single lexeme with no inherent word class. We explore these claims in the context of morphological processing, using three delayed priming lexical decision tasks. The results suggest that some pairs are in a covert derivational relationship, in which the derived form is morphologically more complex than the base. However, not all such pairs are related this way, as some instead behave like inflectional relatives belonging to a single, underspecified lexical entry. Together, the experiments offer support for a mixed model, in which the grammar distinguishes between different kinds of zero-related pairs based on their underlying morphological relationships – a covert distinction to which morphological processing may be sensitive.