A criterion is developed for the choice of n experimental runs in multiresponse situations after N runs are already available. The criterion is applied to three examples involving two particular non-linear multiresponse models and the results are disussed.