Fractal features of the electroencephalogram (EEG) signal are useful for characterizing the temporal complexity of non-stationary signals. The present study proposed an event-related complexity analysis (ERC) method for detecting the time-locked changes in the diversity of neural activity. This information is important for investigating higher cognitive functions, such as language. ERC method was applied in this study to explore the neural correlates of semantic processing in the context of written German sentence comprehension. Experiment results show that ERC method can provide more information compared with ERP, and thus is valuable in the investigation of cognitive processes.