Consciousness fades during deep nonrapid eye movement (NREM) sleep early in the night, yet cortical neurons remain active, keep receiving sensory inputs, and can display patterns of synchronous activity. Why then does consciousness fade? According to the integrated information theory of consciousness, what is critical for consciousness is not firing rates, sensory input, or synchronization per se, but rather the ability of a system to integrate information. If consciousness is the capacity to integrate information, then the brain should be able to generate consciousness to the extent that it has a large repertoire of available states (information), yet it cannot be decomposed into a collection of causally independent subsystems (integration). A key prediction stemming from this hypothesis is that such ability should be greatly reduced in deep NREM sleep; the dreamless brain either breaks down into causally independent modules, shrinks its repertoire of possible responses, or both. In this article, we report the results of a series of experiments in which we employed a combination of transcranial magnetic stimulation and high-density electroencephalography (TMS/hd-EEG) to directly test this prediction in humans. Altogether, TMS/hdEEG measurements suggest that the sleeping brain, despite being active and reactive, loses its ability of entering states that are both integrated and differentiated; it either breaks down in causally independent modules, responding to TMS with a short and local activation, or it bursts into an explosive and aspecific response, producing a full-fledged slow wave.