An instrumental step in assessing the validity of animal models of chronic cognitive disorders is to document disease-related deficits in learning/memory capacity. The water maze (WM) is a popular paradigm because of its low cost, relatively simple protocol and short procedure time. Despite being broadly accepted as a spatial learning task, inference of generalized, bona fide “cognitive” dysfunction can be challenging because task accomplishment is also reliant on non-cognitive processes. We review theoretical background, testing procedures, confounding factors, as well as approaches to data analysis and interpretation. We also describe an extended protocol that has proven useful in detecting early performance deficits in murine models of neuropsychiatric lupus and Alzheimer's disease. Lastly, we highlight the need for standardization of inferential criteria on “cognitive” dysfunction in experimental rodents and exclusion of preparations of a limited scientific merit. A deeper appreciation for the multifactorial nature of performance in WM may also help to reveal other deficits that herald the onset of neurodegenerative brain disorders.