The approach involves comparisons of streams of data known as time series, such as fluctuations in the stock market index and changes in employment levels. Because they consist of many pictures of the rise and fall of a value taken at regular time intervals, time series are comparable to movies. Given two movies, the comparison starts with frames from each of the movies taken at the same point in time. The second movie is then backed up one frame or more. Changes in those earlier frames in the second movie may predict changes that show up in a later frame of the first movie. Granger causality helps determine whether this link is coincidence or results from one process influencing another process.
Now this is what I think of as "interdisciplinary" - using ideas and methods from one domain of research to inform another.
I've never come across a paper that uses Granger causality to make causal claims about neural mechanisms, but I expect that this will prove to be a fruitful line of inquiry in neuroscience in the future.
Currently, nearly all neuroscience research that employs brain imaging (such as EEG or fMRI) is limited to reporting what parts of the brain "light up" during cognitive processes. While this information has given us interesting results, I'm not yet convinced that this is anything more than high-tech phrenology. The ability to reveal causal chains in brain imaging data is a big leap in cognitive neuroscience.