I’m presenting a poster at the 2012 Cosyne meeting on February 25. My poster is titled:Decision making and working memory in a parietal-prefrontal loop model. My abstract is below:
Decision making and working memory in a parietal-prefrontal loop model
John Murray, Xiao-Jing Wang
Working memory and decision-making involve a distributed interacting network of brain areas, with the parietal and prefrontal cortices (PPC and PFC) at the core. However, the differential roles of these areas and the nature of their interactions are poorly understood. To examine these issues, we model both cognitive functions in a loop circuit model of interacting modules PPC and PFC. Within each module, excitatory populations selective
for choice options compete through mutual inhibition. Populations send long-range projections between modules onto excitatory and inhibitory cells. Stimulus input enters into PPC, reﬂecting the dorsal visual pathway. We identify conditions where both areas display persistent activity during working memory, and intervening distractors are represented in PPC but ﬁltered in PFC. These dynamics are observed experimentally and functionally desirable: PPC encodes saliency and PFC ensures robustness according to behavioral demands. Feedback from PFC to PPC can ﬂexibly gate whether each stimulus is ﬁltered or maintained in working memory. We propose the concept of pathway-speciﬁc excitation-inhibition balance for long-range projections. In this regime, only differences
in activity are propagated and integrated downstream, after the local computation is completed. Balance thereby enables gating and serial computation within the distributed network. The same circuit model is applicable to interactions between functionally distinct cell types, ‘target selection’ cells in PPC and ‘response’ cells in PFC. It provides a mechanistic explanation of the experimental observation that reaction time correlates with the separation time of target selection cells and the onset time of response cells. We explore when the two modules receive different, potentially conﬂicting, inputs, relevant to multisensory and reward-biased decision-making. Integration of these inputs depends on the relative strengths of local and long-range connections. With strong local and
weak long-range connections, the network can generate ‘conﬂict states’. We examine conﬂict dynamics and how conﬂict may be resolved.