Three methods are in widespread use to build vulnerability curves (VCs) to cavitation. The bench drying (BD) method is considered as a reference because embolism and xylem pressure are measured on large branches dehydrating in the air, in conditions similar to what happens in nature. Two other methods of embolism induction have been increasingly used. While the Cavitron (CA) uses centrifugal force to induce embolism, in the air injection (AI) method embolism is induced by forcing pressurized air to enter a stem segment. Recent studies have suggested that the AI and CA methods are inappropriate in long-vesselled species because they produce a very high-threshold xylem pressure for embolism (e.g., P50) compared with what is expected from (i) their ecophysiology in the field (native embolism, water potential and stomatal response to xylem pressure) and (ii) the P50 obtained with the BD method. However, other authors have argued that the CA and AI methods may be valid because they produce VCs similar to the BD method. In order to clarify this issue, we assessed VCs with the three above-mentioned methods on the long-vesselled Quercus ilex L. We showed that the BD VC yielded threshold xylem pressure for embolism consistent with in situ measurements of native embolism, minimal water potential and stomatal conductance. We therefore concluded that the BD method provides a reliable estimate of the VC for this species. The CA method produced a very high P50 (i.e., less negative) compared with the BD method, which is consistent with an artifact related to the vessel length. The VCs obtained with the AI method were highly variable, producing P50 ranging from −2 to −8.2 MPa. This wide variability was more related to differences in base diameter among samples than to differences in the length of samples. We concluded that this method is probably subject to an artifact linked to the distribution of vessel lengths within the sample. Overall, our results indicate that the CA and the AI should be used with extreme caution on long-vesselled species. Our results also highlight that several criteria may be helpful to assess the validity of a VC.< Réduire