Years ago I saw Nic Danis present an interesting paper where he used relatively simple proofs to demonstrate that two different types of feature systems (one SPE-like , one with a special place node) defined the same inventory of segments, but different sets of natural classes. Later on, Scott Nelson showed me an earlier paper where he compares natural classes generated by privative and equipollent (i.e., binary) feature systems in a few contexts. Nelson’s paper, while model-theoretic, uses a toy inventory of three features and demonstrates (“proves” might be too strong) its observations by exhaustion: he just looks at what the toy system can or can’t do. I recently wrote a squib where I try to prove some of Nelson’s observations more generally. That is, I simply imagine there is a fixed inventory of feature predicates F such that f ∈ F the privative feature f and the equipollent feature +f have exactly the same interpretation. Then, it is relatively easy to show that privative and equipollent feature systems generate the same set of possible segments for any F we forbid underspecification and inconsistency (i.e., segments which are both +f and −f) in the equipollent system. Thus, any privative feature system defines the same set of segments as some complete, consistent equipollent feature system. At the same time, it is also relatively easy to show that any (non-empty) privative feature system defines a smaller set of natural classes than even a complete, consistent equipollent feature system. A reviewer said everyone knew this already, and I agree the results are sort of intuitive, but they weren’t actually results until now. This squib appears in Radical, with brief commentary from two of my colleagues.