The paper presents research focused on the decision-making process of multiple classifiers, conditioned by the characterisation of attributes provided by supervised discretisation. This transformation of the input domain imposes a specific distribution of data and features, exploited by the homogeneous ensembles of estimators based on the informativeness of attribute domains in a dataset. The committees of inducers aggregated decisions through several defined voting scenarios. The procedure was applied to two classifiers that worked on selected publicly available datasets with different properties. Performance was studied with particular attention given to characteristics and irregularities of input domains before and after discretisation, sensitivity of learners to various data forms, and consequences of the employed voting schema.