Researchers are increasingly using language to study emotion, yet we know little about how language measures relate to other measures often used to study emotion. Across 3 large, multimodal datasets, we tested whether different language dictionaries relate to self-report, observer report, facial expressions, and vocal cues of emotion. 

We find that language measures of valence (and to a lesser extent, language measures of discrete emotions and emotion frequency) are correlated with a range of other measures used to study emotion and may therefore be a useful option when self-report and behavioral coding measures are unavailable or impractical for a given research question.