From the assumption that there are more inacceptable than successful variants observable a difference in the reliablity of confirmations during the tolerance learning process is inferred:
Inferences from successful examples are more reliable than inferences from failures. Inacceptabe variants only give weak confirmations for tolerance hypothesis (and therefore relevance hypothesis). Consequently, they are not included in the evaluations cycles. Unsuccessful alternatives never are the basis for relevance hypothesis. They do not establish relevance hypothesis. They support and complement inferences from the acceptable examples.
Solely when it comes to verifying a relevance hypothesis these inacceptable example come into play. Inacceptable examples possibly contradict the current relevance hypothesis. In this case, they give valuable information as now the relevance hypothesis is under doubt by contradicting inacceptable variants.
Inacceptable example add weight to factors if a certain value is observed in successes and not in failures. This observation resists any attempt to quantification, so it is not processed in adaption cycles and tolerance hypothesis.