Lemmas and lexemes: the evidence from blends.

An analysis of 166 word blends provides support for the claim that word frequency effects are located at the phonological level of lexical access. The traditional structural approach to blends has been to view them as involving a sequence of two words where word(2) completes an incomplete word(1), as in yes/right-->yight. The proposal here is that blends are better viewed as homologous to sublexical substitutions. Among other advantages, this approach allows one to distinguish a target word from an intruder. Only those targets which are phonologically related to their intruders are subject to a word frequency effect; others are not. Copyright 1999 Academic Press.