“…During "brain control," actuator movements are causally linked to an ensemble of neurons in M1, "direct neurons" (Taylor et al, 2002). Such direct neurons change their firing properties during neuroprosthetic learning (Ganguly et al, 2011;Arduin et al, 2013;Gulati et al, 2014Gulati et al, , 2017Athalye et al, 2017Athalye et al, , 2018; in contrast, the vast majority of neurons in the larger M1 network, "indirect neurons," generally become less task coupled (Fetz, 2007;Ganguly et al, 2011;Koralek et al, 2012Koralek et al, , 2013Arduin et al, 2013;Clancy et al, 2014;Gulati et al, 2017;Silversmith et al, 2021; but see Liu and Schieber, 2020). It remains unknown how neural activity in M2, i.e., indirect neurons which shape M1, plays a role in neuroprosthetic learning.…”