Abstract Respiratory patterns share bidirectional links with brain functions: they are modulated by sensory stimuli, attention and emotions, are affected in some cognitive disorders, but they also can influence perception, emotions and cognition. In particular, brain activity undergoes drastic changes when switching between vigilance states, such as transitions between wake and sleep - as do the overall respiratory rate. However, how the fine features of respiration, beyond its rate, accompany these transitions remains unclear. To address this question, we equipped freely-moving mice with both intra-nasal pressure sensors and hippocampus-targeted electrodes. The unprecedented accuracy of the respiratory signals in mice spontaneously alternating between wake, non-rapid-eye-movement (NREM) and REM sleep revealed that periods of respiratory pause with low or no airflow, are interspersed within phases of exhalation and inhalation. If pause durations impacted the respiratory rate in individual states, the effect of pauses differed across sleep and wake. This stemmed from strikingly different patterns across states: mainly pauses after inhalation during wake, mainly after exhalation in REM, and a mixture of both for NREM sleep. We verified that respiratory patterns are distinctive signatures for states by building an accurate machine-learning algorithm relying solely on respiration information for the prediction of vigilance states. Our experiments demonstrated that the information of missing pauses after inhalation and of breathing variability were instrumental to precise REM sleep prediction. Moreover, results highlighted that these vigilance state-respiration relationships can be generalized across animals. In agreement, kinetic indicators for exhalation and inhalation co-varied across states and were spared by animal-to-animal variations, while the duration of pauses after inhalation stood as an isolated, state-discriminant feature. Finally, dynamical analysis revealed that distinct breathing features adapt with different kinetics at the transition time points between different states, possibly accompanying distinct cortical changes. Our work therefore clarifies how different features of respiration, and in particular pauses in nasal airflow, are associated to the specific physiology of individual vigilance states and suggest new links with brain functions.