Thermometric sap flow sensors are widely used to measure water flow in roots, stems and branches of plants. Comparison of the timing of flow in branches and stems has been used to estimate water capacitance of large trees. We review studies of sap flow in branches and present our own data to show that there is wide variation in the patterns and timing of sap flow of branches in different parts of the crown, owing to the course of daily solar illuminance. In contiguous forest, east-facing and upper branches are illuminated earlier than west-facing and lower branches and most capacitance studies do not include adequate information about branch sampling regimes relative to the overall pattern of crown illuminance, raising questions about the accuracy of capacitance estimates. Measuring only upper branches and normalising these results to represent the entire crown is dangerous because flows at the stem base likely peak in response to maximum crown illuminance (and transpiration) and this will differ compared to the timing of peak flows in upper branches. We suggest that the magnitude of flow lags between branches and stems needs further study, with careful attention to branch position and method application before a robust understanding of capacitance, particularly in woody tissues of large trees, can be formed. We did not detect flow lags in the world’s tallest and largest tree species Sequoia sempervirens and Sequoiadendron giganteum, despite measurement along large pathlengths (∼57 and 85 m), which raises questions as to why large flow lags are often recorded for much smaller species. One conspicuous possibility is the different methods used among studies. Constant-heating methods such as the thermal dissipation probe (and also heat balance methods) include heat capacitance behaviour due to warming of wood tissues, which delays the response of the sensors to changing sap flow conditions. We argue that methods with intrinsic heat-capacitance present dangers when trying to measure water-capacitance in trees. In this respect heat pulse methods hold an advantage.