This work investigates the syntactic and semantic distribution of reflexive forms themself vs. themselves. We focus on whether the use of -self vs -selves is affected by specificity of the antecedent, distinguishing seven levels for a singular antecedent: quantified indefinites (some/any), quantified universals (each/every/no), generic kind definites, specific indefinites, specific definites, unknown specific person, named person. We administered an online acceptability judgment survey to 1,127 participants, recruited on Twitter and Prolific, who rated sentences on a 5-point Likert scale from “very natural” to “very unnatural,” in a randomized order, and then completed a short demographic survey. We find that antecedent type, -self/-selves form of the anaphor, and social variables including age, gender, and endorsement of prescriptive language ideology all affected how participants rated the sentences; themself was generally rated higher than themselves for most antecedents, contra previous findings. Our results seem to indicate that themself is preferred by many English speakers for singular antecedents, and also that the singular use of themself/themselves is a sociosyntactically active and salient variable across those speakers.