Averaging is widely used in EEG research to improve the signal-to-noise ratio, particularly in structured tasks like the oddball paradigm. However, in complex cognitive tasks such as abstract reasoning, variability in response times makes time-locking difficult and undermines the effectiveness of averaging. This calls for alternative methods that can capture neural dynamics on a single-trial level. In this thesis, I apply the Hidden Multivariate Pattern (HMP) method—a technique that models EEG data as a sequence of cognitive events estimated directly from the signal. These events can then be used as anchors for further analysis. I test the method on an abstract reasoning task where participants infer the final item in a sequence of symbols based on underlying rules. This task is cognitively demanding and allows for long, self-paced responses, making it a suitable case for single-trial analysis. Previous applications of the HMP method have focused on simpler tasks; this is the first to apply it in a more complex setting. The aim is to evaluate whether the HMP method can extract consistent, interpretable time courses that reflect meaningful cognitive processes.