Fig. 1

Data acquisition and analysis pipeline. (A) Illustration of the task. Two participants played the music game as a team, in which they both needed to tap an iPad screen when visual cues (notes) reached the yellow judgment line. In the Team Flow condition, the music was typical, and participants sat adjacent to each other. In the Team Only condition, the music was scrambled. In the Flow Only condition, a cardboard separated the participants so they could not see each other’s feedback. (B) The trial sequence: each pair of participants played 18 trials – six songs (S1-S6) under the three conditions in the pseudorandomized order presented. (C) The two-stage dimensionality reduction pipeline. Preprocessed EEG time series of 128 channels went through power spectral analysis, resulting in the power spectral density (PSD) of each channel in the theta (4–7 Hz), alpha (8–12 Hz), beta (13–30 Hz), and low-gamma (31–50 Hz) frequency bands. Next, for dimensionality reduction, 40 latent components per frequency band (from 128 channels) were identified using non-negative matrix factorization (NMF). Finally, all these NMF components were transformed into a 7-dimensional latent space through linear discriminant analysis (LDA) to maximize the distances between the participation × condition classes.