Identifying and defining words has long been a thorny issue, both at the language specific level and especially cross-linguistically (Haspelmath 2011). Some issues with wordhood can be resolved by recognizing that morphosyntactic and prosodic evidence may point to non-isomorphic phonological and grammatical “word” constituents (Dixon & Aikhenvald 2002). Others have noted that for some languages, constituency tests point to no obvious—or no single—phonological word or other prosodic constituent (Schiering et al. 2010). In this presentation, I share ongoing work about constituency in Zenzontepec Chatino (Otomanguean), following the methodology of Tallman (2020). To approach the question, no a priori constituents are assumed, and sentences are flattened out into planar structures. Constituency tests are applied, their convergences are tallied, and the probability that the observed convergences could be due to chance is then calculated. In a pilot study of 17 languages of the Americas, Zenzontepec Chatino displays the strongest evidence for a (phonological, and general) word. Empirical studies like this are necessary for developing a more refined cross-linguistic notion of wordhood and constituency.