The nature of information flow from one area of the cerebral cortex to anot
her is poorly understood. Frequency-dependent measures of information flow,
based on multivariate autoregressive modeling of field potential time seri
es, have shown promise for understanding information transactions between c
ortical areas (Liang et al., Neuro Report, 11 (2000) 2875-2880). In the pre
sent contribution, a time domain measure of information flow between two ar
eas, called the directed transinformation (DTI), is described and applied t
o investigate causal influences directly from the field potential time seri
es. We show that the DTI, as a generalization of mutual information, can be
measured in a rather natural way, such that the interdependence of two tim
e series is the sum of flow from X to Y, flow from Y to X, and instantaneou
s flow. We demonstrate the usefulness of this technique on both simulated d
ata and multichannel local field potentials from macaque monkeys. Compariso
n with the frequency-dependent measure is also made. (C) 2001 Elsevier Scie
nce B.V. All rights reserved.