This is a three-part project concerning methods for the study of political semantics -how the 'meaning' of political concepts is represented and organized in memory and implications for political attitudes and behavior. The first chapter proposes a framework for the estimation of group differences in memory representations of political concepts and applies it to evaluate partisan representational differences in the U.S. The second chapter proposes a memory-centered approach to the study of ideology along with the requisite methods for its implementation. The third chapter centers on word embeddings, a deep learning method to estimate word representations from large collections of text. Along with a conceptual overview, it provides practitioners with a series of tests to perform model comparison and validation, including a novel Turing-style test