Our commentary outlines a number of arguments questioning an autonomous model of word recognition without feedback. Arguments are presented against the need for a phonemic decision stage and in support of a featural level in a model including feedback.