Allen Institute Openscope - Global/Local Oddball project

This dataset was collected for the Global Local Oddball project, as part of the Allen Institute's OpenScope project.
The brain rapidly adapts to unchanging environments and is most excited by surprising events. A leading model to explain this phenomenon is predictive coding. Predictive coding proposes that the brain compares incoming sensory information against a prediction signal. This prediction is based on an internal model that is generated in higher-order cortex. This model reflects the brain’s assumptions about the statistics of the environment. If an incoming sensory signal matches the prediction, the two signals cancel. Expected sensory data thus are “explained away” and leave the brain unexcited. In other words, predictive coding is subtractive. Whenever the sensory signal does not match the prediction, subtraction results in a larger value, called the prediction error. This error signal initiates excitation in the lower-order cortex that propagates feedforward up the cortical hierarchy (i.e., V1, RL, LM, AL, PM, and AM). Prediction errors then instigate updates to the internal model to improve future predictions. Associated prediction update signals flow back down the hierarchy. Consistent with this model, optogenetic silencing of top-down inputs from frontal to visual cortex largely eliminates prediction error signals. However, the precise circuit mechanisms that generate these signals are largely unknown. Specifically, by recording from multiple neuropixels across the visual cortical hierarchy, we aimed to uncover what information is carried by layer 2 and 3 spikes that feed forward vs. by layer 5 and 6 spikes that feed back, using recently established analytic tools.