Variance as a Signature of Neural Computations during Decision Making

“during Decision Making”

“during Decision Making”

“Lastly, we consider a recurrent neural network that is exemplary of a class of dynamical ‘‘attractor’’ models of decision making (e.g., Albantakis and Deco, 2009; Bogacz et al., 2006; Machens et al., 2005; Usher and McClelland, 2001; Wang, 2002).”

“during Decision Making”

“This is the same as drift diffusion with an upper and lower stopping bound (Bogacz et al., 2006; Ratcliff and Rouder, 1998), so long as the evidence for one alternative is evidence against the other (Gold and Shadlen, 2007).”

“For example, the irregular discharge of neurons bears on theories of coding, synaptic integration and circuit function (Shadlen and Newsome, 1998; Softky and Koch, 1993).”

“Together, the mean and variance of neural responses furnish rich insight into the limits of perception, motor control and decision making (Faisal et al., 2008; Glimcher, 2005; Parker and Newsome, 1998; Shadlen and Newsome, 1998).”

“Lastly, we consider a recurrent neural network that is exemplary of a class of dynamical ‘‘attractor’’ models of decision making (e.g., Albantakis and Deco, 2009; Bogacz et al., 2006; Machens et al., 2005; Usher and McClelland, 2001; Wang, 2002).”

“…call the residual covariance, The second term is the covariance of the conditional mean spike count, R ( t, e ) which we call the rate covariance, The total covariance therefore separates into a residual component and a component arising from structured changes in the underlying firing rates [45],”

“…the decision, providing a direct observation of the single-trial trajectory itself—a different order of evidence from the trial-averaged firing rates and the variance signatures of bounded diffusion previously derived from single-neuron spike counts [12], or from behavioral model-fitting alone.”

“The reduction of the global noise level during stimulus is likely to be related to a reduction in the neural variability, that is expected to quench due to stimulus presentation as observed at single-cell level 44,45 .”

“Because LIP activity in the random-dot motion task has been shown to reflect a process of evidence accumulation (Churchland et al., 2011; Huk & Shadlen, 2005; Steinemann et al., 2024; Vivar-Lazo & Fetsch, 2026), the neural DV can be used to infer certain dynamic features of this process.”

“The stochastic sampling closely matches the stochastic nature of neural dynamics, characterized by the large, structured neuronal response variability (Shadlen & Newsome, 1998; Churchland et al., 2011; Orbán et al., 2016; Echeveste et al., 2020; Zhang et al., 2023).”

“…area (LIP) gives a clear indication of a terminating threshold — a point of minimum variance just before the response (Roitman and Shadlen, 2002; Churchland et al., 2011; Kira et al., 2015) — implying that a downstream area terminates the decision when the fi ring rate reaches a critical level.”

“Previous studies have demonstrated the significant role this variability plays in neural computation [41, 42].”

“Because of this, models accounting for behaviour and/or neuronal activity often introduce a random process at some level in order to explain empirical variability (e.g. Churchland et al., 2011; Deco et al., 2009; Kanashiro et al., 2017).”

“Previous literature has reported evidence that the activity in the prefrontal cortex is contained in a lowdimensional manifold that can be extracted projecting the firing rates from the high-dimensional recording space onto few latent variables that represent the majority of the variance in the recordings (7,13,38).”

“…within a condition, which may further constrain the nature of the underlying neural processes (Bollimunta, Totten, & Ditterich, 2012; A. K. Churchland et al., 2011; M. M. Churchland et al., 2010; Latimer, Yates, Meister, Huk, & Pillow, 2015; Morcos & Harvey, 2016; Seidemann, Meilijson,…”

“…observed on multiple levels of neural activity, including cellular recordings (Arieli et al., 1996; Monier et al., 2003; Finn et al., 2007; Churchland et al., 2010, 2011; Hussar and Pasternak, 2010; Scaglione et al., 2011; Chang et al., 2012; White et al., 2012; Goris et al., 2014;…”

“However, differences between the two areas emerged in terms of the secondorder statistics of spike trains, which has been used previously to differentiate evidence-accumulation from presaccadic bursts in area LIP (Churchland et al., 2011).”

“Most previous studies on sensory decisionmaking examined the activity of neurons in the parietal and frontal cortex (Churchland et al., 2011; Ding and Gold, 2012; Gold and Shadlen, 2000; Kiani and Shadlen, 2009; Kiani et al., 2008; Kim and Shadlen, 1999; Roitman and Shadlen, 2002; Shadlen and…”

“…that decision termination is mediated by application of a threshold to the neural representation of accumulating evidence (Gold and Shadlen 2007; Churchland et al. 2011; P L Smith and Ratcliff 2004) and this also holds in experiments that do not allow for free responses, as in our…”

“Using the law of total variance which is derived by subtracting the amount of variability due to spiking from the total measured variance, an analysis of LIP data during a decision making task supports the evidence accumulation hypothesis (Churchland et al., 2011).”