The study of the interaction between brain areas poses significant challenges both on the experimental and on the signal processing sides of the analysis. We propose to extend the
ideas behind dimensionality reduction to the multi-area problem, as a way to summarize the high-dimensional data recorded in the brain. We review the key concepts behind dimensionality
reduction and proceed to study the interaction between visual areas 1 and 2 using canonical correlation analysis. We find that these areas communicate in a way that preserves stimulus decoding.