Leading Up the Lexical Garden Path: Segmentation and Ambiguity in Spoken Word Recognition