In the search for better predictors of cardiovascular events, pulse wave velocity (PWV) has gained considerable interest. We compared three single-location methods to locally estimate PWV based on simultaneous measurements of pressure (P), velocity (U) or arterial diameter (D): the PU, ln(D)U and QA-method. First, the performance of these methods was analyzed using 3D fluid-structure interaction simulations (FSI) in a tube and patient-specific carotid artery. We demonstrated that the outcome was dependent on whether the methods were tested in the ideal conditions of a 3D axisymmetrical and reflection-free tube or in the more realistic setup of a carotid artery. The three single-location PWV methods performed similarly in the tube (4.29 m/s for PU, 4.44 m/s for ln(D)U and 4.38 m/s for QA) while the carotid data showed that the PU-method dramatically overestimates PWV (9.16 m/s), and the ln(D)U and QA-method underestimate (3.86 and 3.84 m/s, respectively). The erroneously high wavespeeds from the PU-method were attributed to wave reflections, which was confirmed by measurements in 37 healthy adults. This in vivo study showed similar discrepancies between the 3 single-location techniques as present in the carotid simulations, with the difference between the PU- and ln(D)U-method related to the magnitude of wave reflection.