Transfer entropy (TE) is a model-free measure of effective connectivity in neuroscience, based on information theory. It quantifies the direction and strength of information flow between brain regions, without assuming a specific model of interactions. Unlike traditional methods such as Granger causality, which rely on linear models, TE is inherently non-linear and can detect complex, non-linear interactions. This study demonstrates the applicability of TE in analyzing electrophysiological data, particularly in magnetoencephalography (MEG) recordings. TE was tested on simulated data and real MEG data from a motor task, showing its ability to detect effective connectivity in non-linear interactions and in the presence of linear cross-talk. TE is robust against linear cross-talk, which is common in MEG data due to volume conduction. The method was also tested on linearly mixed signals, which are typical in EEG and MEG recordings. TE was found to be effective in detecting effective connectivity even when interaction delays varied, and it was shown to be more sensitive to non-linear interactions than linear methods. The study highlights the potential of TE as a useful tool for analyzing effective connectivity in neuroscience, particularly in non-invasive electrophysiological measurements. The method was implemented using a data-efficient estimator and a statistical test to assess significance. The results demonstrate that TE can accurately detect effective connectivity in various scenarios, including those with linear and non-linear interactions, and with different interaction delays. The study provides a framework for applying TE to neuroscience data, emphasizing its advantages over traditional model-based approaches.Transfer entropy (TE) is a model-free measure of effective connectivity in neuroscience, based on information theory. It quantifies the direction and strength of information flow between brain regions, without assuming a specific model of interactions. Unlike traditional methods such as Granger causality, which rely on linear models, TE is inherently non-linear and can detect complex, non-linear interactions. This study demonstrates the applicability of TE in analyzing electrophysiological data, particularly in magnetoencephalography (MEG) recordings. TE was tested on simulated data and real MEG data from a motor task, showing its ability to detect effective connectivity in non-linear interactions and in the presence of linear cross-talk. TE is robust against linear cross-talk, which is common in MEG data due to volume conduction. The method was also tested on linearly mixed signals, which are typical in EEG and MEG recordings. TE was found to be effective in detecting effective connectivity even when interaction delays varied, and it was shown to be more sensitive to non-linear interactions than linear methods. The study highlights the potential of TE as a useful tool for analyzing effective connectivity in neuroscience, particularly in non-invasive electrophysiological measurements. The method was implemented using a data-efficient estimator and a statistical test to assess significance. The results demonstrate that TE can accurately detect effective connectivity in various scenarios, including those with linear and non-linear interactions, and with different interaction delays. The study provides a framework for applying TE to neuroscience data, emphasizing its advantages over traditional model-based approaches.