Phenological data have become increasingly important as indicators of long-term climate change. Consequently, long-term homogeneity of the records is an important aspect. In this paper, we apply a breakpoint detection algorithm to the phenological series from the Swiss Phenology Network (SPN). A combination of three statistical tests is applied and different constraints are tested with respect to the choice of reference series. Breakpoint detection is only possible for a fraction of the series due to the shortness of some series and the lack of suitable reference series. Spring phases are more likely to be suitable than fall phases because of their higher spatial correlation. Out of nearly 3000 phenological series with at least 20 data points, only about 5% were found to be significantly inhomogeneous, although a visual validation indicates that many mid-sized breakpoints remained undetected. The detected breakpoints were compared with metadata and more than half of them could be attributed to a change of observer.