Synchronization of oscillations is a fundamental mechanism of rhythmicalrelations in nonlinear systems of different nature.In particular,the effectsof synchronization play a key role in information processing and transmission in Central Neural Systems.These processes are accompaniedwith formation and destruction of synchronous neuronal oscillating patternsarising from groups of neurons simultaneously generating action potentials(spikes).An important real world application is the neural drive of respiration in humans.Breathing is maintained and controlled by a network of automatic neurons in the brainstem that generate respiratoryrhythm and receive regulatory inputs.Breathing complexity therefore arises from respiratory central pattern generatorsmodulated by peripheral and supra-spinal inputs.Very little is known on the brainstem neural substrates underlyingbreathing complexity in humans.We used both experimental and theoretical approaches to decipher these mechanisms inhealthy humans and patients with chronic obstructive pulmonary disease(COPD).COPD is the most frequent chronic lungdisease in the general population mainly due to tobacco smoke.In patients,airflow obstruction associated withhyperinflation and respiratory muscles weakness are key factors contributing to load-capacity imbalance and henceincreased respiratory drive.Unexpectedly,we found that the patients breathed with a higher level of complexity duringinspiration and expiration than controls.Using functional magnetic resonance imaging(Fmri),we scanned the brain of theparticipants to analyze the activity of two small regions involved in respiratory rhythmogenesis,the rostral ventro-lateral(VL)medulla(pre-B(o)tzinger complex)and the caudal VL pons(parafacial group).Fmri revealed in healthy controls higher activity of the VL medulla suggesting active inspiration,while in patients higher activity of the VL pons suggesting active expiration.COPD patients reactivate the parafacial to sustain ventilation.These findings may be involved in the onset of respiratoryfailure when the neural network becomes overwhelmed by respiratory overload.We show that central neural activitycorrelates with airflow complexity in healthy subjects and COPD patients,at rest and during inspiratory loading.We finallyused a theoretical approach of respiratory rhythmogenesis that reproduces the kernel activity of neurons involved in theautomatic breathing.The model reveals how a chaotic activity in neurons can contribute to chaos in airflow and reproduceskey experimental Fmri findings.