The Morris water maze (MWM) is the most frequently used apparatus in behavioral neuroscience. In rodents, its application ranges from learning and memory studies to molecular, genetic and focal lesion studies, often combined with each other. Despite its widespread use, we still lack a methodology to analyse rodents' cognitive performance in the MWM. Until now, in fact, the great majority of behavioral analyses relied on escape latency as the main indicator of a complex spatial strategy. Although latency correlates with an improved performance in the MWM, it is not an accurate measure of cognitive performance because the same escape latency could result from very different spatial strategies.

In the last decade, technological advances have allowed us to track rodents' swim paths and record their spatial and temporal coordinates. Many quantitative variables and different parameters can now be extracted to describe the complex MWM behavior more accurately than latency does, but because they oversimplify the animal's strategy, they do not represent the complexity of the actual behavior. More recently, a new computing approach has shown that complex behaviors are better described by using many parameters at the same time. This method involves a factor analysis of different quantitative variables. It allows hypothetical variables (so-called 'factors') associated with characteristic swim paths to be identified, such as thigmotaxis, extended or restricted search. However, being mainly a descriptive tool, even this 'bottom-up' method has a limited discrimination power among the numerous behaviors observed in the MWM.

To overcome these limitations, we implemented a 'top-down' method to categorize rodents' swim paths by combining a tracking system (EthoVision) with the statistical method of Discriminant Analysis. We defined the following regions of interest (ROIs): a circular zone around the platform ('critical zone'), an outer annulus ('periphery'), and the total arena. All the measures included in the analysis were related to these ROIs. According to the above definitions, we first classified 1,061 swim paths, performed by 37 rats in the same MWM paradigm, into eight distinct categories: thigmotaxis, peripheral-circular search strategy, peripheral-jagged search strategy, extended or brief-extended search strategies, first-orientation-then-reach search strategy, approach targeting, and direct finding of the platform. We then extracted 28 dependent variables and performed a Discriminant Analysis to obtain a Classification Function (CF). By means of the CF, we obtained 97.7% correct automatic recognition. Furthermore, when the Discriminant Analysis was restricted to only two categories at a time, we reached 100% correct classification, no matter which categories were analysed. The CF also allowed us to classify newly recorded paths with the same statistical efficacy.

Naturally, the eight categories codified are not the only ones required to classify a complex MWM behavior - but this system is flexible enough to recognize any a priori defined swimming behavior, regardless of the starting point and platform position. In conclusion, this new methodology produces an automatic 'gestaltic' evaluation of different explorative searching strategies. A new genesis has begun in the study of cognitive strategies of rodents in the Morris water maze.

Paper presented at Measuring Behavior 2002 , 4^th International Conference on Methods and Techniques in Behavioral Research, 27-30 August 2002, Amsterdam, The Netherlands

© 2002 Noldus Information Technology bv