Figure 4D clearly demonstrates

the time specificity of population coding during the cue processing period, whereas population coding in the delay period is more time stable. Again, there is no evidence for cross-generalization of coding during the cue or associated delay period to the target-related response (Figure 4E). The results so far suggest that information concerning selleck screening library trial type is maintained through the delay period as a stable low-energy state. Although the population response differentiates between the three alternative contexts, the underlying code does not resemble patterns observed during cue processing or the expected target. This nonstationary coding scheme contrasts with classic models of WM that posit persistent maintenance of the initial input representations (Miller et al., 1996; Wang, 2001) or preactivation of the expected target/memory probe (Rainer et al.,

1999). We suggest that the postcue state could reflect a temporary reconfiguration of the tuning profile in prefrontal cortex for flexible behavior, i.e., to discriminate choice stimuli www.selleckchem.com/products/fg-4592.html according to context for “go”/“no-go” decision making. A systematic reconfiguration of the network state in prefrontal cortex would also be expected to alter the response characteristic of the network to fixed input (Mongillo et al., 2008; Sugase-Miyamoto et al., 2008). Indeed, we find that the population response to the neutral stimulus clearly differed as a function TCL of trial type (Figure 5A), even though the same neutral stimulus was used for all trial types (see Experimental Procedures). This suggests that the activation profile of the network is patterned according to trial type. To visualize the separation of activity states driven by the fixed neutral stimulus, we plot four independent estimates of the activity pattern associated with each trial type (color coded) onto the first two dimensions determined by MDS (Figure 5B; the full time course is captured in the Movie S3 available online). Data points clearly cluster as a function of trial type at 250 ms after stimulus

onset, reflecting systematic activity states that differentiate the response to the fixed neutral stimulus according to context. We propose that cue processing establishes a state in PFC that temporarily tunes prefrontal neurons to respond according to the current task context, i.e., to decide the appropriate behavioral response to choice stimuli. In a final set of analyses, we examined responses to the three choice stimuli that, according to the rule established by the current cue, could serve as either a “go” or “no-go” signal for the behavioral response. We defined stimulus 1 as the stimulus serving as a target with cue 1, but a distractor with cues 2 or 3, and similarly for stimulus 2 (target with cue 2) and stimulus 3 (target with cue 3).