Background: Accurate prediction of extubation outcome could result in a significant reduction of respiratory morbidity in premature neonates. Objectives: To assess whether the respiratory muscle time constant of relaxation (τ) predicted extubation outcome in mechanically ventilated, premature infants. Methods: Forty-six mechanically ventilated infants with a median gestational age of 26 (interquartile range [IQR] 25-29) weeks were prospectively studied. τ was calculated from the reciprocal of the slope of the decline in airway pressure as a function of time. Measurements of τ were done during 5-10 min of a spontaneous breathing test (SBT) prior to extubation. During the first and last minute of the SBT, τ1 and τ2, respectively, were assessed, and the difference between them was calculated (Δτ). Results: The median τ2 was significantly higher in infants whose extubation failed (20.7 [IQR 12.9-34.7] s/cm H2O) than in infants whose extubation succeeded (8.2 [IQR 6.2-17.8] s/cm H2O, p = 0.002). The median Δτ was significantly higher in infants whose extubation failed (10.3 [IQR 4.4-23.9] s/cm H2O) than in infants whose extubation succeeded (-1.63 [IQR -5.7 to 0.3] s/cm H2O, p = 0.001). Extubation failure was associated with τ2 (p = 0.011) and Δτ (p = 0.010) after correcting for postmenstrual age, patent ductus arteriosus, and intraventricular haemorrhage. Receiver operator characteristic curve analysis demonstrated that Δτ predicted extubation failure with an area under the curve of 0.937. A Δτ of +1.02 s/cm H2O predicted extubation failure with 94% sensitivity and 83% specificity. Conclusions: The respiratory muscle time constant of relaxation during an SBT was significantly greater in infants whose extubation failed and could be used to predict extubation outcome in prematurely born infants.