Coherence from Context - Two-Point Neuron Models for Contextual Integration in Visual Information Processing

Abstract

The integration of contextual information is crucial for coherent perception and cognition. The morphology and conductance properties of cortical pyramidal cells suggest that they operate as ''two-point neurons'' (TPNs), asymmetrically combining driving sensory input at basal dendrites with modulating context at apical compartments. We present a mechanistic rate-based TPN model of layer V pyramidal cell populations and their local PV, SOM, and VIP interneuron circuits. The model rests on asymmetric feedforward-feedback integration and pooled inhibition to implement local competition and global cooperation supporting the selective amplification of coherent signals. Fitted to multi-compartment pyramidal cell simulations, it reproduces their basal-apical input-output map and, unlike static transfer functions, shows how apical amplification unfolds over time under interneuron control. Asymmetric apical-basal coincidence integrates context during processing and tags TPNs for reward-driven learning with a local four-factor rule. A pool of TPNs thereby learns to use flanker context to decide ambiguous orientations. Context-driven VIP cells gate apical integration and learning through dis-inhibition. We connect TPN populations with joint spatial and feature selectivity into networks. Contextual signals from lateral recurrence and top-down feedback yield coherence formation and evidence propagation. This propagation disambiguates visual motion and requires apical-basal coupling. For larger-scale simulations, we show how established simpler models follow from reducing the TPN's complexity while retaining its core principles. Overall, context in the TPN amplifies sensory evidence but cannot create it without feedforward input. Finally, we discuss when dynamics and interneurons matter, the model's testable predictions for VIP silencing, and implications for artificial neural networks.