Does visual word identification involve a sub-phonemic level?