The timing and magnitude of airway narrowing in central apneas is unknown. We have developed a method of apnea classification that relies on the transmission of cardiac airflow oscillation to indicate airway patency. Using a theoretical model, we showed that the amplitude of the cardiac airflow oscillation is proportional to airway diameter for small lumens. While in the majority of central apneas the amplitude of the cardiac airflow oscillation remains nearly constant, in a subset of events the waveform decreases with time, suggesting airway narrowing. We hypothesized that this is not a random occurrence but reflects a critical period of airway instability during central apnea. To test this hypothesis we studied 41 preterm infants. Of 4,456 central apneas, 585 had a decrease in the amplitude of the cardiac oscillation. The amplitude of the cardiac airflow oscillation during an apnea was recorded to provide a dynamic measure of changes in airway diameter with time. To allow for comparisons between patients the amplitude of each cardiac airflow oscillation was expressed as a proportion of the maximum amplitude observed in each infant. We then compared the amplitude at multiple successive 0.5 s intervals with the amplitude of the cardiac airflow oscillation observed at the apnea outset using ANOVA. We found a significant decrease in cardiac airflow oscillation after only 1 s irrespective of the apnea duration (3 to 16 s). We conclude that airway narrowing during central apnea is not a random occurrence but appears shortly after the onset of the apnea. We speculate that the phenomenon is secondary to passive airway relaxation.