Complexity of Defensive Behaviors

 I am reading the review "Embracing Complexity of Defensive Behaviors" by Headley, Kanta, Kyriazi and Pare from Rutgers University from Neuron, 2019. There's an interesting task called the Barnes maze where a vigilant rat is placed in a box with an escape route. It engages the scanning of the environment and problem solving faculties. I am collecting different experimental set ups for fear conditioning and filing them away in my mind. I quote the paper under consideration: "In a standard conditioning chamber, aversive conditioned stimuli (CS) appear to reflexively elicit behavioral freezing. However, if the chamber also allows rats to avoid the anticipated shock, either by moving to a different part of the chamber or by stepping onto a platform, trained rats eventually show marked trial-to-trial variations in defensive strategy. They may first freeze and then escape, or escape without first freezing, and they do so at variable latencies from CS onset (Bravo-Rivera et al., 2014; Kyriazi et al., 2018)". These behaviors are complex and probabilistic. It would be interesting to quantify the neural dynamics behind these different neuronal manifestations where the stimulus-response associations are more loose. I think the Barnes maze would heavily engage the pre-frontal cortex and it would be interesting to record the neural activations longitudinally to quantify the non-stationary dynamics in the neural circuits. I worked on Neuropixel probe data during Neuromatch Academy 2020. These ephys probes can be placed all over the brain, permitting the temporal characterization of neural activity all over the brain. But what would you then do with the data? PCA would give you directions of greatest variance in the brain activity space and how intensely each direction is activated in time and which brain regions or neurons tend to co-vary, reducing them to a single component. You could do a network analysis of brain activity to make a time series of which brain regions co-activate over binned periods of time. You would have a network with either strenghtening and weakening or disappearing and appearing connections (by thresholding). You could quantify the behavior in some way (but what is the best way?) and do regression analyses to relate neural activity to behavior. You could search the literature to link different more controlled experiments that give causal roles to various brain processes and see what are the actual dynamics of these brain regions during the experiment, thus using literature and text search to map out what the animal is thinkind at each moment in time. For example, there was an fMRI study in humans where the task was to play a PacMan-like game with a predator. When the threat was far away, the ventromedial prefrontal cortex activated and when the threat was near, the periaquedactal gray activated. This reveals a role for each brain structure. But this is fMRI with its very rough activity blogs, the spiking dynamics of actual circuits are much more nuanced. Would timeseries analyses be appropriate? I am reading a book on neural timeseries analysis by Tohru Ozaki and this stuff is mighty complicated. In the Barnes maze, the sequential and temporal activation of neurons could be very complex spanning large bouts of time. This is really a cognitive task. And how can you temporally characterize cognition? These are really recurrent networks integration information temporally, where previous activity patterns continuously feed into the system and feed it over time. These are some complex dynamics! What will we gain if we quantify them?