Coproduction and direct perception of phonetic segments: a critique