Frequency-dependent representation of reinforcement-related information in the human medial and lateral prefrontal cortex

Abstract

The feedback-related negativity (FRN) is a commonly observed potential in scalp electroencephalography (EEG) studies related to the valence of feedback about a subject's performance. This potential classically manifests as a negative deflection in medial frontocentral EEG contacts following negative feedback. Recent work has shown prominence of theta power in the spectral composition of the FRN, placing it within the larger class of "frontal midline theta" cognitive control signals. Although the dorsal anterior cingulate cortex (dACC) is thought to be the cortical generator of the FRN, conclusive data regarding its origin and propagation are lacking. Here we examine intracranial electrophysiology from the human medial and lateral prefrontal cortex (PFC) to better understand the anatomical localization and communication patterns of the FRN. We show that the FRN is evident in both low-and high-frequency local field potentials (LFPs) recorded on electrocorticography. The FRN is larger in medial compared with lateral PFC, and coupling between theta band phase and high-frequency LFP power is also greater in medial PFC. Using Granger causality and conditional mutual information analyses, we provide evidence that feedback-related information propagates from medial to lateral PFC, and that this information transfer oscillates with theta-range periodicity. These results provide evidence for the dACC as the cortical source of the FRN, provide insight into the local computation of frontal midline theta, and have implications for reinforcement learning models of cognitive control.

Macaque anterior cingulate cortex deactivation impairs performance and alters lateral prefrontal oscillatory activities in a rule-switching task

“…Authors from these studies suggested that the dACC was required to integrate history of both reward and errors to guide effective choice-a hypothesis that is supported by our finding that reaction times increased in both trial types with dACC deactivation, even though the performance on prosaccade trials remained relatively intact. Our results also indicate that the cognitive control signal which guides effective responses with rule representations and/or the history of outcomes may be achieved through a theta/alpha input from the dACC to the dlPFC (Smith et al, 2015). Taken together, our findings suggest that feedback-based cognitive flexibility and other aspects of cognitive control reflect the concerted effort of an extended network, in which the dACC may broadcast different signals through different oscillatory rhythms.…”

“…Given that the animals must maintain the current rule throughout the session and to update it whenever necessary, we cannot rule out that the possibility that the effect found during the ITIs was still task-related, even though no decision regarding the saccadic response was made then. Since dACC theta can propagate feedback information to the LPFC (Smith et al, 2015) and modulates gamma activities in the LPFC during attentional processes (Voloh et al, 2015), it is not unexpected that dACC deactivation disrupted the theta-gamma interaction within the LPFC (Figure 7). Although the animals remained capable of updating the task rules based on feedback (Figure 2,3), they did commit more perseverative errors in the Stable post-switch stage during dACC deactivation ( Figure 4), which may reflect a subtle deficit in feedback-related cognitive flexibility.…”

“…Additionally, dACC theta was found to carry rule information prior to the onset of the stimulus that determines the direction of the saccade (Womelsdorf et al, 2010). Because task information can be transferred from the dACC to the lateral PFC through theta activities (Smith et al, 2015), the reduction in task-related theta/alpha activities in the dlPFC may reflect a loss of dACC input on either cognitive control or rule representation. Given that the decrease in theta/alpha activities coincided with longer reaction times but not necessarily errors, and that the animals improved significantly from the Early to the Stable stage even with dACC deactivation, the rule information must be preserved in other regions in the network such as the dlPFC, although its guidance over the action may be weakened.…”

DOWN state in the anterior cingulate and prelimbic areas in rats during immobility

“…Compared to PL, AC receives a significantly larger percentage of intra cortical afferents and input from anterior, as well as "relay" nuclei of thalamus 59 , while PL receives more dense projections from the middle nuclei of thalamus 58,59 . Functionally, AC mostly plays a critical role in memory consolidation processes 19,20,35,41,46,47 while PL is necessary for retrieval of fear 4,11 .…”

Estimation of narrowband amplitude and phase from electrophysiology signals for phase-amplitude coupling studies: a comparison of methods

“…Most studies dealing with PAC apply one of the following time-frequency approaches to extract narrowband representations of brain activity and to calculate their envelope amplitude and instantaneous phase: band-pass filtering of the activity signal followed by Hilbert transform (we will call this the Hilbert approach) (Cohen, 2008;Kramer et al, 2008;Voytek et al, 2010Voytek et al, , 2013McGinn and Valiante, 2014;Voytek et al, 2015a,b;Niknazar et al, 2015;Ninomiya et al, 2015;Smith et al, 2015;Xu et al, 2015;Blain-Moraes et al, 2015;van Wijk et al, 2015); or convolution with a continuous wavelet kernel (we will call this the wavelet approach), sometimes preceded by band-pass filtering (Maris et al, 2011;Canolty et al, 2012;Florez et al, 2015;McGinn and Valiante, 2014;Kajihara et al, 2015;O'Connell et al, 2015;Mizuhara et al, 2015;Sweeney-Reed et al, 2016;Vandenbroucke et al, 2015;van Driel et al, 2015). The two approaches are formally equivalent, since both of them can be formulated as linear convolutions with the activity time series, and one can make them to be identical with an adequate choice of parameters (particularly those of the band-pass filters) (Bruns, 2004;Kiebel et al, 2005).…”