Pittsburgh Brain Activity Interpretation Competition 2006 Methods Description Theory Based Analysis of Emotion Relevant Show Content and Brain Activity: A Canonical Correlation Approach

Based on entertainment/drama theory, the theory of humor, and the neuroanatomy of emotion this project developed a tentative coding scheme in order to recode the content of the competition’s movie stimulus. The additional content codes had to satisfy two criteria in order to be useful: (1) The coded content had to be more relevant for the subjects’ entertainment experience, and (2) the coded content had to correlate with the competition’s feature ratings in order to improve the forecasts. The rational of our approach is the belief that adding relevant and theoretical based content variables suppresses noise in the feature ratings data and thus leads to better forecasts. For data analysis and forecasting we applied a dynamic linear canonical correlation approach combined with a principle component analysis of the brain activity data for season one and two across all subjects. Although our results can certainly be improved through more intensive statistical optimizations, we could demonstrate that our content codings are systematically related to the provided brain activity data. Unfortunately we failed to show substantive correlations between our content codings and the competition’s feature rating data. Hence, the additional information did not significantly improve the forecasts. Nevertheless, we believe that a better understanding of the entertaining media stimulus and an optimized, more reliable and (entertainment) theory based measurement of content provides a valuable approach for a better analysis and interpretation of brain activity data. We will further develop our content analysis and demonstrate its strong relationship to the human brain’s emotional circuits.