Abstract. Many climate records show the occurrence of large amplitude (10–15 K), millennial-scale warming events during glacial times, the Dansgaard-Oeschger (DO) events. So far these events have almost exclusively been investigated by means of linear time series analysis. The scope of this paper is to test if the assumption of linearity is fulfilled during DO events. By means of a surrogate-based Monte Carlo method, I here demonstrate that the 60 000-year long δ¹⁸O-record from the NGRIP ice core from Greenland allows to reject the null hypothesis of linearity beyond any reasonable level of doubt. Instead, the ice core record supports the interpretation that the events represent regime switches between different states of operation of glacial climate. As a conclusion, future studies on DO events should focus on the development and the application of more adequate (i.e., nonlinear) methods of time series analysis.