Discussion
The dehydration-rehydration-dehydration experiments conducted here indicate that embolism spreading is not random. While the water potentials of individual conduits remain impossible to determine, our experimental data suggests that pit membranes between adjacent conduits appear to work as fairly efficient safety valves when dehydration levels are rather low, ensuring that individual conduits have a high fidelity to a specific water potential threshold at which gas will invade the conduit. However, once sufficient embolism has accumulated in the xylem the spreading of embolism into water filled conduits does not appear to be strictly driven by a fixed pressure threshold between neighbouring conduits. When more than 60% of the xylem is embolized, gas spreads rapidly into water-filled conduits upon dehydration. This could be due to the amount of embolised conduits present, but also gas movement across embolised or sap-filled conduit lumina, pit membranes, and cell walls triggering embolism. These results have important implications for understanding the nature of embolism formation during dehydration as well as the impact of pre-existing embolism on the determination of relative embolism resistance.