Nonverbal cues, specifically gestures, are a central element of political speech. The effects of political speeches can be altered dramatically through the use of gestures and vocal emphasis. However, studying politicians’ nonverbal cues during political speech is difficult due to its visual nature. As a result, we have little systematic understanding of how politicians communicate nonverbally, even though we know that this has great effects on voters’ perception of political speech. This project introduces a methodological framework for the comparative quantitative analysis of politicians’ body language based on the computational analysis of video recordings. The framework is based on pose estimation models, a class of methods developed in Computer Vision to trace body key
points such as wrists and elbows throughout videos. I show how to use pose estimation to study politicians’ body language during political speech. The proposed framework has four significant advantages: First, it can be adapted for the measurement of a variety of aspects of body language, such as arousal, power-posing, or the detection of other predefined gestures. Second, the method can be applied to video recordings from different parliaments. Compared to other solutions, it allows cross-country and cross-cultural studies of politicians’ body language. Third, the framework yields intuitive and easy-tounderstand measures that can be communicated to wide audiences. Fourth, it is based on open-source software and can be applied by researchers with limited computational resources. The presented work creates new opportunities for applied researchers to start exploring politicians’ body language across multiple contexts.
