We use computational linguistics techniques to study the use of emotion and reason in political discourse. Our new measure of emotionality in language combines lexicons for affective and cognitive processes, as well as word embeddings, to construct a dimension in language space between emotion and reason. Applying the method to the U.S. Congressional Record for the years 1858 through 2014, we show that emotionality in U.S. politics was stable until the early 1970s, after which time it increased significantly. More polarized politicians use more emotional language in their speeches. Emotionality increased after floor speeches became televised.