The specific properties of softwares for interpretation of physiological signals are discussed. Problems in the presentation of inaccurate knowledge and in the assessment of the quality of software realization of this knowledge are identified. Ways of describing inaccuracies by using the factors of certainty, inaccurate logic, linguistic approximations, and three-value logic are considered. Using the well-known approaches to testing the ECG and external respiration analyzing systems as examples, the authors show it necessary to develop special methods to evaluate the efficiency, selectivity, and stability of inferences for slight-magnitude conclusions.