We have investigated the hypothesis that perceptual cues and memory of trial history are integrated in the decision-making process underlying the countermanding task. Our analyses of the responses of neurons Y-27632 purchase in PMd of monkeys performing a countermanding arm task show the influence of recent trial history on both the performance of monkeys and on the variability of neuronal responses in PMd. We show that the behavior of the monkeys becomes increasingly more conservative

(longer RT) when a Go trial was recently preceded by one or more Stop trials and increasingly hastier (shorter RT) when it was recently preceded by one or more Go trials, as previously reported (Rieger and Gauggel, 1999; Emeric et al., 2007; Verbruggen and Logan, 2008; Nelson et al., 2010; Mirabella et al., 2006). We show that the behavioral performance is linearly correlated with changes in the variability of the neural response. To validate the possible signature of trial history in neural response variability, we performed an additional theoretical study using a mean-field approximation of a spiking neural model. We show that changes in the strength of a modulatory input that reflects trial history accounts for the observed changes in behavior and neural response variability, suggesting the existence of a trial history-monitoring system in the brain. Our study provides a neural correlate for task

history and its impact on the neuronal substrate of decision making and is a further example of how adaptive behavior is monitored and orchestrated in the brain (Walton et al., 2004; Ito et al., 2003). One of the weaknesses see more of using VarCE as a measurement of the across-trial variability lies in the estimation of the scaling factor ϕ. We computed it separately for each neuron (see Experimental Procedures), and the obtained distribution of the values Electron transport chain of ϕ was consistent with the ones previously reported for the neocortex

(Figure S2G) (Shadlen and Newsome, 1998; Nawrot et al., 2008). To check the robustness of our results to variations in the value of ϕ, we repeated our analyses (Figure 2B) but setting the same value of ϕ for each neuron. We observed that the difference in VarCE between history conditions is independent on the value of ϕ used (Figure S2H). Similar to VarCE, the Fano Factor (spike count variance divided by spike count mean) has been used to calculate the across-trial variability of neural responses. Although in most cases both measurements are considered to be equivalent, for significant changes in mean FR, the VarCE has shown to be more robust than the Fano Factor (Churchland et al., 2011). However, our conclusions hold for both the Fano Factor and the VarCE (see Figures S2I and S2J) and are further supported by the equivalent histogram obtained from the interspike interval observed in a Go trial preceded by different sequences of trials, i.e.