Michael Kremer complains that my first-order presentation of Russell’s logicist reduction “misrepresents the technical achievement of Russell’s theory of classes” – which he identifies with the “ontological economy” of eliminating commitment to classes (sets). I disagree. Russell did attempt to eliminate classes, and he connected this attempt to other interesting projects. These projects aside, however, the purported elimination was not, in my opinion, a genuine achievement, and the relevant issues have little to do with first-order vs. second-order symbolizations. Kremer’s discussion focuses on what he calls Russell’s “contextual definition of classes,” the purpose of which is “to eliminate the notion of classes.” 1. F( {x: Gx} ) =df ∃H(∀y(Hy ↔ Gy) & F(H)) This characterization of the definition is tendentious. What (1) defines is a notation containing complex singular terms for classes. By identifying contents of sentences containing such terms with those of sentences that don’t, Russell eliminates the terms from the primitive vocabulary of his theory. Whether or not an ontological reduction is achieved depends on how predicate quantifiers are understood. On standard second-order interpretations, there is no reduction, since they range over sets. By contrast, Russell took the quantifiers to range over “propositional functions.” Kremer tells us little about what these are -- for good reason. Russell’s discussion vacillates between taking them to be formulas, properties, or functions from nonlinguistic arguments to nonlinguistic propositions. Mostly, he seems to have inclined to the former.1 So, if strict historical fidelity is the issue, then Russell should be taken to have seen himself as reducing classes to expressions. But this is problematic.