Rapidly developing research in neurophysiology has challenged classical cognitive models based on behavioral evidence. Studies looking more closely at the relationship between cognitive function and the brain structure have shed new light on how the mental processes are physically implemented in the brain. Regardless of whether the neural correlate of cognition is dispersed (the activity of a particular neuron is not representative) or distributed (the level of individual neurons is selective of a concrete feature), it is essential to establish a cognitive ontology that instantiates the structure-function mapping of the brain. The core of the present work relies on the next systemic assumption: at some level, different parts of the normal, healthy brain subserve functions. Consequently, functions should predict the structure and the structure should predict the function. Direct inference or What are the neural correlates of a cognitive operation? and reverse inference or What is the function associated with a brain area activation? are dealt with this systemic and computational light. Needless to say, the task ahead is arduous. Anyhow, important steps are being taken towards true brain inspired architectures in cognitive systems. http://brainmap.org, a database for querying and retrieving data about brain structure and function over the internet, is available to be utilized for testing empirically architectural assumptions. We present a methodology, exemplified by an algorithm, to build cognitive ontologies that integrate cognitive and anatomical models of the brain.