(eng): Stochastic conditional independence plays an important role in the application of probability theory into the field of artificial intelligence. From the comparison of complexity of models based on probability distributions and those based on belief functions it is obvious, that it is even more important in the latter framework. In this contribution we compare two conditional independence concepts (conditional non-interactivity and conditional independence) from various points of view. We will concentrate not only to their formal properties, but also to their unconditional versions, their relationship to stochastic conditional independence, number of focal elements of basic assignments satisfying the respective conditional independence constraints, the complexity of their checking, their consistency with marginalization and, naturally, also their mutual relationship.