Studies of sensory processing in awake mammals are often limited to simple perceptual decisions, such as detecting the differences between two similar stimuli. However, such sensory tasks are usually limited by peripheral information, and do not tap into the special processing capabilities of the cortex. I hypothesize that neuronal responses in auditory cortex represent ethologically-relevant quantities that optimally summarize knowledge about the current scene and that allow behaviourally-relevant predictions of its future development. In order to study the role of such mechanisms in controlling behaviour, it is necessary to develop new technological and methodological tools.I will develop a semi-natural living environment for rats that will make it possible to train multiple animals to perform behavioural tasks while continuously recording the electrical activity of their brains. Neural activity will be recorded continuously using telemetry inside the environment, or periodically outside it with chronic imaging techniques. Brain activity will be manipulated with optogenetic techniques. This methodology will minimize human intervention, increasing the reproducibility of behavioral and electrophysiological data collection while reducing the number of animals used.Various amount of information about the states of the environment will be communicated to the rats with sounds. I will apply a new theory that rigorously quantifies the balance between information and reward. The theory will make it possible to deduce what the rats believe about the environment from their behavior, and to correlate these beliefs with neural activity.Hearing disorders are a major cause of reduction of quality of life, especially in the elderly population. Better understanding of auditory processing in real-world scenarios is a crucial step for the future development of better tools and therapies.