 This study investigates the temporal evolution of coupled systems by decomposing predictive information about a target system into amounts quantifying internal and transferred information using self-entropy, transfer entropy, cross entropy, and conditional self-entropy. The linear Gaussian approximation is used to compute these measures in benchmark systems with known dynamics, and their behavior is studied during head-up tilt and paste-breathing protocols to infer causal effects and interpret alterations with changing experimental conditions. This article was authored by Luca Faze, Alberto Porta, and Giannomenico Nallo.