Abstract
Recent studies have suggested that humans and animals switch between different modes of processing during perceptual decision-making tasks. However, meaningful variations in internal processing remain difficult to discover and characterize. Here we utilize a data-driven clustering method based on modularity-maximation to identify consistent spatial-temporal EEG activity patterns across individual trials and relate this activity to behavioral performance. Subjects (N = 25) performed a motion direction discrimination task with six interleaved levels of motion coherence. We identified two subtypes of trials, Subtype 1 and Subtype 2. Subtype 1 contained event-related potentials with significant positive amplitude in occipital but negative amplitude in frontal areas over the 300 ms of stimulus presentation, whereas Subtype 2 exhibited the opposite pattern. Further, these patterns extended beyond the window of stimulus presentation. Surprisingly, even though the first subtype occurred more frequently with lower motion coherences (Z = -4.06; p = 4.72 x 10-5), it was nonetheless associated with faster response times (RT: Subtype 1 = 926 ± 3ms; Subtype 2 = 952 ± 4ms; t-value = -6.97; p = 3.29 x10-12). Modeling with the drift diffusion model suggested that the first subtype was characterized by higher decision boundary (DB Subtype 1: 1.5 ± 0.20; Subtype 2: 1.55 ± 0.20; t-value = -3.81; p = 0.001). These results demonstrate that brain activity measured with EEG can be used to distinguish subtypes of trials differing in their underlying internal processes, opening up a new way to identify brain states relevant to cognition and behavior.