“…Previous studies have established that across‐trial variability in neural activity decreases during neural computations such as in response to the onset of a stimulus (Churchland et al, 2010). Decrease in across‐trial variability has also been correlated with broad spatial tuning after stimulus onset (Chang, Armstrong, Armstrong, & Moore, 2012), expected reward (Falkner, Goldberg, & Krishna, 2013), attention (Cohen & Maunsell, 2009; Herrero, Gieselmann, Sanayei, & Thiele, 2013; Hussar & Pasternak, 2010; Mitchell, Sundberg, & Reynolds, 2007; Sendhilnathan, Basu, Goldberg, Schall, & Murthy, 2019), task familiarity (Qi & One, 2012), sensorimovement learning (Mandelblat‐Cerf, Paz, & Vaadia, 2009), movement preparation (Churchland, Yu, Ryu, Santhanam, & Shenoy, 2006), task engagement (Hussar & Pasternak, 2010) and behaviour (Churchland et al, 2010; Falkner et al., 2013; Purcell, Heitz, Heitz, Cohen, & Schall, 2012; Steinmetz & Moore, 2010). Taken together, these studies raise the possibility that neural computations suppress the chaos in the system making it more ‘stable’ following an input (Abbott, Rajan, & Sompolinsky, 2011; Churchland et al, 2010).…”