Depicting the salvageable tissue is increasingly used in the clinical setting following stroke. As absolute cerebral blood flow (CBF) is difficult to measure using perfusion magnetic resonance or computed tomography and has limitations as a penumbral marker, time-based variables, particularly the mean transit time (MTT), are routinely used as surrogates. However, a direct validation of MTT as a predictor of the penumbra threshold using gold-standard positron emission tomography (PET) is lacking. Using (15)O-PET data sets obtained from two independent acute stroke samples (N=7 and N=30, respectively), we derived areas under the curve (AUCs), optimal thresholds (OTs), and 90%-specificity thresholds (90%-Ts) from receiver operating characteristic curves for absolute MTT, MTT delay, and MTT ratio to predict three penumbra thresholds ('classic': CBF <20  mL/100  g per min; 'normalized': CBF ratio <0.5; and 'stringent': both CBF <20  mL/100  g per min and oxygen extraction fraction >0.55). In sample 1, AUCs ranged from 0.79 to 0.92, indicating good validity; OTs ranged from 7.8 to 8.3  seconds, 2.8 to 4.7  seconds, and 151% to 267% for absolute MTT, MTT delay, and MTT ratio, respectively, while as expected, 90%-Ts were longer. There was no significant difference between sample 1 and sample 2 for any of the above measurements, save for a single MTT parameter with a single penumbra threshold. These consistent findings from gold-standard PET obtained in two independent cohorts document that MTT is a very good surrogate to CBF for depicting the penumbra threshold.