Automation in experiments carried out on animals is getting more and more important in research. Computers take over laborious and time-consuming activities like recording and analysing images of the experiment scene. The first step in an image analysis is finding and distinguishing between the observed animals and then tracking all objects during the experiment. In this paper four tracking methods are presented. Quantitative and qualitative figures of merit are applied to confront those methods. The comparison takes into consideration the level of correct object recognition during different disturbances, the speed of computation, requirements as to the frame rate and image illumination, quality of recovering from occluded situations and others.