Sign languages are known to display the same general grammatical properties as spoken languages (‘Universal Grammar’), but also to make greater use of iconic mechanisms. In Schlenker et al.’s ‘Iconic Variables’ (Linguist Philos 36(2):91–149, 2013), it was argued that loci (= positions in signing space corresponding to discourse referents) can have an iconic semantics, in the sense that certain geometric relations among loci (subset and relative complementation, as well as high/low position relative to the signer) are preserved by the interpretation function. Here we ask whether plural and height specifications of loci display the formal behavior of phi-features in remaining uninterpreted in focus- and ellipsis-constructions (as in the bound readings of, e.g., Only Mary admires herself, or of Mary admires herself, and John does too). Data from ASL and LSF show that plural and height specifications may indeed remain uninterpreted in these constructions; furthermore, there are cases in which a single high locus is construed iconically and left uninterpreted in the course of ellipsis resolution. We argue that our data are compatible with two theories. According to the Strong View, plural and height specifications of loci display exactly the behavior of spoken language features. According to the Weak View, our data just show that plural and height specifications share the behavior of features and other non-assertive elements in being separable from the referential terms they specify. Our LSF data are compatible with the Weak View; our ASL data might provide support for the Strong View. While our aim is merely to open the debate about the featural status of iconic specifications, the question is of some importance: if features are innate and primitive elements of grammar, and if some of them have an intrinsically geometric semantics, the signed modality might play a greater role than is usually thought at the very core of Universal Grammar.