This paper introduces a new frequency-domain approach called partial directed coherence (PDC) to describe the relationships between multivariate time series, specifically focusing on the direction of information flow. The method is based on the decomposition of multivariate partial coherences derived from multivariate autoregressive models. The authors discuss the application of PDC and compare its performance with other methods for determining neural structure relations from simultaneous neural electrophysiological signals. PDC is shown to reflect a frequency-domain representation of Granger causality, providing a more detailed understanding of the functional connections between neural structures. The paper also reviews the concept of directed coherence (DC) and its generalization to multiple neural structures, highlighting the shift from mutual synchronicity to functional connectivity. Examples are provided to illustrate how PDC provides direct structural information for multivariate autoregressive models and to demonstrate its use in revealing changes in information flow during specific brain activities, such as spindle episodes during slow-wave sleep.This paper introduces a new frequency-domain approach called partial directed coherence (PDC) to describe the relationships between multivariate time series, specifically focusing on the direction of information flow. The method is based on the decomposition of multivariate partial coherences derived from multivariate autoregressive models. The authors discuss the application of PDC and compare its performance with other methods for determining neural structure relations from simultaneous neural electrophysiological signals. PDC is shown to reflect a frequency-domain representation of Granger causality, providing a more detailed understanding of the functional connections between neural structures. The paper also reviews the concept of directed coherence (DC) and its generalization to multiple neural structures, highlighting the shift from mutual synchronicity to functional connectivity. Examples are provided to illustrate how PDC provides direct structural information for multivariate autoregressive models and to demonstrate its use in revealing changes in information flow during specific brain activities, such as spindle episodes during slow-wave sleep.