The development of animal models of human cognitive decline, represents a critical step to a better understanding of the neurobiological basis of memory function. Consequently, behavioural analysis of memory function in rodents cannot simply be restricted to the use of a rough recipe containing key notions like the water-maze, the radial-maze or contextual fear conditioning. In other words, the use of a spatial learning task is not sufficient to assess that this task is supported by hippocampal function. Additionally, successful performance of a memory task may result from various learning strategies belonging to different memory systems, and each could be differentially affected. As noted by Olton and co-workers, the best strategy for optimising the development of satisfying animal models is to design behavioural tasks that focus on specific cognitive processes. 
Accordingly, the first part of this presentation proposes a behavioural analysis based on the earliest paradigms devised by Eichenbaum's lab, and aimed at modelling in aged mice, the age-related cognitive decline in humans. Evidence is provided that two forms of memory can be assessed using the same task, and that one of these memory forms, supposedly representing the animal equivalent to human declarative memory, is affected in aged mice. Hippocampal lesions appears to parallel the general feature of the deleterious effect of ageing. Moreover, the specificity of the effects is strengthened by pharmacological evidence, showing that the administration of retinoic acid, a vitamin A derivative, strikingly reduces the memory deficit in aged mice without affecting any other behavioural parameters. 
The second part of the presentation opens the discussion on the notion of dependency between memory systems. On the basis of lesioning studies, it is generally assumed that memory function is not unitary, and that there are at least two memory systems, each with independent information-processing modules. We are focusing on the description of behavioural experiments documenting that hippocampal activity may be differentially affected with respect to the memory task involved. In other words, learning tasks engaging the procedural memory system were shown to modify neurobiological or neurophysiological markers of hippocampal activity in a way opposite to the one triggered by hippocampal-dependent memory task. Finally, we consider an example demonstrating a dissociation between hippocampal physiological activity and behavioural expression. This is based on evidence that the hippocampus can still support "a declarative knowledge" while the behavioural expression of the memory is impaired.
Although the theoretical framework representing the basis of these experiments is not discussed, we claim that behavioural analysis of memory processes in rodents needs to be carefully dissected in order to be of predictive value for human memory.