Such errors, since they precede the first reinforcement of a correct response by "mmhm," do not depend on the reinforcement value of "mmhm"...