The understanding of the fundamental relation between brain resting activity and the response to stimuli is a long-standing fascinating question. Recent experiments have shown that the spontaneous brain activity is characterized by avalanches showing absence of characteristic size, result successfully interpreted in the context of criticality. However, in order to support the idea that the brain acts close to a critical point it is crucial to evidence the existence of long-range correlations. The temporal organization of neuronal avalanches in the rat cortex in vitro is characterized by the alternation between states of highly correlated activity and almost quiet periods, leading to a dynamic balance between excitation and inhibition. MEG data confirm that brain activity at large scale is temporally correlated evidencing a special role of α waves in the temporal organization of avalanches. The fundamental question concerning the relation between spontaneous and evoked activity is addressed by means of the coarse-grained Wilson Cowan model. An approach inspired in non-equilibrium statistical physics allows to derive a fluctuation-dissipation relation, suggesting that measurements of the correlations in spontaneous fluctuations in the brain activity alone could provide a prediction for the system response to a stimulus. Theoretical predictions are in good agreement with MEG data for healthy patients performing visual tasks.