Hydrograph separation is a widely applied technique that uses the stable isotopes of water (²H and 18O) or other tracers to quantify the contribution of different water sources to streamflow. For its successful application it is critical to adequately characterize these sources (end‐members). In most small catchment studies, water samples are collected from end‐members at one or a few locations that are assumed to be representative for the entire catchment. We tested this assumption by reviewing 148 papers that used the stable isotopes of water to investigate hydrological processes in catchments up to 10 km². We assessed the typical spatial variability in the isotopic composition of different hydrological compartments when they were sampled at five or more locations across a catchment. The median reported spatial variability was largest for snowmelt and soil water, followed by throughfall and shallow groundwater. To determine how this spatial variability might affect isotope‐based hydrograph separation results, we used three‐component hydrograph separation for two real rainfall‐runoff events and a synthetic rainfall‐runoff event and adjusted the isotopic composition of the end‐members (throughfall, soil water, and shallow groundwater) by the median observed spatial variability. The estimated maximum contributions of the three components differed by up to 26% from the reference scenario. This suggests that caution is needed when interpreting hydrograph separation results if they are based on samples taken at one or only a few locations. Above all, these results show that the assumption of negligible spatial variability may not be valid for small catchments.