The purpose of this study was three-fold: (1) to characterise both the on- and off-transient oxygen uptake (V̇O2) kinetics in endurance runners during moderate-intensity treadmill running; (2) to determine the degree of symmetry between on- and off-transients; and (3) to determine the reproducibility of V̇O2 kinetic parameters in endurance runners. Twelve endurance-trained runners [mean (SD) age 25.2 (4.7) years, body mass 70.1 (9.7) kg, height 179.5 (7.5) cm, ventilatory threshold (VT), 3,429 (389) ml.min−1, maximal V̇O2 (V̇O2max) 4,138 (625) ml.min−1] performed two multiple square-wave transition protocols on separate days. The protocol consisted of six (three transitions, 15 min rest, three transitions) square-wave transitions from walking at 4 km.h−1 to running at a speed equivalent to 80% of the V̇O2 at the VT (80%VT). To determine the reproducibility, the protocol was repeated on a separate day (i.e. a test-retest design). Pulmonary gas-exchange was measured breath-by-breath. The V̇O2 data were modelled [from 20 s post-onset (or offset) of exercise] using non-linear least squares regression by a mono-exponential model, incorporating a time delay. The on- and off-transient time constants (τon and τoff), mean response times (MRTon and MRToff) and amplitudes (Aon and Aoff) were obtained from the model fit. On- and off transient kinetics were compared using paired t-tests. The reproducibility of each kinetic parameter was explored using statistical (paired t-tests) and non-statistical techniques [95% limits of agreement (LOA, including measurement error and systematic bias) and coefficient of variation (CV)]. It was found that the τon [12.4 (1.9)] was significantly (P<0.001) shorter than τoff [24.5 (2.3) s]. Similarly, MRTon [27.1 (1.9) s] was shorter than MRToff [33.4 (2.2) s]. With respect to the reproducibility of the parameters, paired t-tests did not reveal significant differences between test 1 and test 2 for any on- or off-transient V̇O2 kinetic parameter (P>0.05). The LOA for τon (1.9 s), τoff (2.3 s), MRTon (1.2 s), MRToff (3.2 s), Aon (204 ml.min−1) and Aoff (198 ml.min−1) were narrow and acceptable. Furthermore, the measurement error (range, 4.3 to 15.1%) and CV (1.3 to 4.8%) all indicated good reproducibility. There was a tendency for τoff to be more reproducible than τon. However, MRTon was the most reproducible kinetic parameter. Overall, the results suggest that: (1) a multiple square-wave transition protocol can be used to characterise, reproducibly, both on- and off-transient V̇O2 kinetic parameters during treadmill running in runners; (2) the phase II time constant is independent of V̇O2max, and (3) asymmetry exists between on- and off transient V̇O2 kinetic parameters.