Measuring Meta-Interpretation

American legal interpretation has taken an empirical turn. Courts and scholars use corpus linguistics, survey experiments, and machine learning to clarify meanings of legal texts. We introduce these developments in »issue-level interpretation,« concerning interpretive theories' application to legal language. Empirical methods also inform »meta-interpretive« debate: Which interpretive theory do interpreters use; which have they used; and which should they use? We demonstrate the relevance of machine learning to these meta-interpretive debates with insights provided by a word embedding that we trained on a corpus of over 1.3 million U.S. federal court decisions.