Spatial learning and memory in rodents is most often assessed in the Morris water maze. Neurobiologists have to distinguish behavioral patterns to unravel underlying neuronal systems. We analyzed swim patterns of mice videotaped before and after training with a multi-trial procedure in the water maze. In addition to traditional parameters, the animals' position in relation to trained and other possible platform locations was estimated five times per second by an image analysis system. This parameter, cumulative distance to platform, was correlated with time spent in the platform quadrant but not with latency to and crossings of the platform location. We detected a subgroup of animals with concentric patterns within the group of spatial/persistent patterns. Random patterns were classified as well. Swim patterns before training were not predictive for the one after training. In summary, image analysis systems have made it very convenient to quantify behavior. Using their capacity, we have further improved the analysis of swim patterns, revealing animals' different approaches to solve a problem.