In this paper we analyze the behavior of liquids in daily-life situations. This behavior may be described within the qualitative physics (QP) paradigm in a way that mimics people’s common sense. However, it may be questioned to what extend such a common sense like description can be used by a robot to perform liquid manipulation. We argue that the symbolic representations used in QP do not provide sufficient accuracy to solve such a control problem. The main reason for this is the critical dependence of the behavior of liquids on the spatial and temporal properties of the situation they are in. In order to deal with this dependence we propose to supplement a qualitative reasoner with an analogical simulation. In this way, the behavior of a liquid can be simulated instead of inferred. The result of a simulation can be interpreted by virtual sensors and passed to the reasoning module. We believe that this approach implements an important aspect of common sense namely the use of visual feedback to solve simple problems without reasoning.