We clustered subjects individual time points into group-wise task states. In each task, states appears to be combinations of, or single, resting-state networks (Figure 5). Moreover, across tasks, similar states appeared (Figure 6). For example, in the Motor and Working memory tasks, 7 of the 8 states were highly similar, with the Motor task having a state that was highly specific to the ventral sensory motor network, likely because, in the Motor task, subjects are presented with visual cues that ask them to move their tongue, while subject presumedly are not moving their tongue in the Working Memory task. Interestingly, our method is able to pick up the tongue movement state, and, while all the other states were similar to the working memory task state, the tongue state did not exist in the working memory task. The same phenomena was observed in the Language task; while states similar to the overlapping states in the Working Memory and Motor tasks were observed, two unique tasks were observed (Figure 7).