Recent findings indicate that neural representations of behaviors are distributed throughout the brain. These distributed neural representations are likely to accompany the transmission of behavioral information across large-scale brain regions, often mediated by the propagation of brain oscillations. Yet, it remains unknown whether the brain-wide patterns of oscillatory amplitude can represent more naturalistic behaviors, and whether they are related to the brain-wide patterns of oscillatory propagation. Using an open human electroencephalogram (EEG) dataset recorded during video-game play (behaviors: shooting, collecting, crashing), we introduced activation states, momentary brain-wide patterns derived from oscillatory amplitude envelopes. The results showed that brain-wide patterns of activation states reliably predicted the likelihood of each behavior during gameplay, and cross-validated decoding recovered behavioral occurrence from single-trial brain-wide activation patterns. We then quantified large-scale oscillation propagation using the temporal consistency of propagation directionality. We found that the spatial patterns of propagation consistency were strongly correlated with concurrent brain-wide patterns of activation states, indicating that where oscillation amplitudes spatially organize, propagation organizes similarly across the whole brain. Together this study shows novel findings that (1) distributed EEG amplitude patterns are predictive enough to decode naturalistic behavior and (2) large-scale propagation provides a complementary signature that tracks the same brain-wide organization. From these results, we propose a Dual-State Oscillation Model (DSOM) engaging coupled brain-wide activation states (amplitude organization) and propagation states (directional transmission), which may provide a novel framework for linking distributed neural representations to large-scale communication dynamics to elucidate how brain networks coordinate naturalistic behaviors.