During quiet waking periods a rat generates sequences of activation in the hippocampus that predict its future motion in two-dimensional space. Termed trajectories, the mechanism by which these sequences are generated is not yet known, despite very active research. The theory suggested here is that they are part of a process of the animal simulating its future motion. Simulations are hypothetical events or scenarios, and there is substantial evidence of simulation in humans. In language understanding, for example, evidence suggests we simulate the action implied by a statement, as part of understanding it. Recent evidence suggests that for motion in two-dimensional space, rats may also be simulating their future actions.
In rats, hippocampal trajectories may have different functions in different behavioral contexts, but the prevailing thinking has long been that all the trajectories originate in the hippocampus. I will suggest that a subset of trajectories may originate in areas outside the hippocampus, as part of a process of simulated motion. Simulated motion is the execution of a motor routine without activation of the motor neurons that would cause actual physical motion. I suggest that the simulated motion updates a path integration system, which has been hypothesized as existing mainly in the medial entorhinal cortex. In turn, this activates the place cells in the hippocampus whose place fields constitute the observed trajectory.
The principal contribution of this work is a computational model of simulated motion that generates trajectories similar to those observed experimentally. The model integrates prior computational models of the functioning of the hippocampus and the adjacent entorhinal cortex. To model the motor routines that drive both physical and simulated motion, I developed extensions to Petri net software. Petri nets are a formalism that is well-suited to abstracting asynchronous, distributed processes such as those occurring in the brain. Methodological contributions include the extensions to open source Petri net software, which help make Petri nets a better candidate for use in computational neuroscience. In addition, the computational model is well-supported with automated tests, which makes it a practical example of how complex computational neuroscience models may evolve over time to incorporate more and more findings, increasing their usefulness.