Modern neuroscience aims at describing brain functions in terms of a neuropsychophysiological model. Despite its general success, this program is stuck at the higher cognitive level. Intuitively known phenomena of subjective experience need to find their neuroscientific description. The present approach to this challenge is implicitly grounded in a number of beliefs taken for granted, including the supervenience hypothesis (1-1 relation between mental states and physical states of the brain), the neural code hypothesis (that elements of brain states have definite semantics), and the attractor hypothesis (supporting the idea of robustness and self-perpetuation of neural code). Theoretical understanding and experimental verification of these hypotheses is a precondition for solving the challenge, that calls for a paradigm change in neuroscientific research. The talk will entertain a computationally grounded approach, presenting examples of facts, preliminary results and a roadmap to solving the challenge.