I argue against the anti-agreement hypothesis for argument ellipsis (e.g. Saito 2007, Sener & Takahashi 2010, Sato 2015), with new evidence from Kaqchikel.

(1) Anti-Agreement Hypothesis for Argument Ellipsis: (e.g. Saito 2007, Sener & Takahashi 2010) Argument ellipsis is possible iff the argument is not Agreed with.

Argument ellipsis (AE) refers to null nominal elements with non-pronominal, i.e. sloppy / quantificational, interpretations available in some languages. Otaki et al. (2013) show that Kaqchikel lacks AE for both transitive objects and subjects, and conclude that Kaqchikel’s lack of AE supports (1); Kaqchikel has agreement with both subjects and objects. However, (1) also predicts that those arguments that are exceptionally not Agreed with will exceptionally allow AE (see Sener & Takahashi on Turkish). I demonstrate that Kaqchikel Agent Focus is an environment in which arguments are not Agreed with. Nevertheless, AE is not exceptionally available, contrary to (1). Moreover, the causal relationship between AE and Agree is called into question given Preminger's (2014) analysis of the Agent Focus construction, and the general logical of feature-valuation, whereby failure to Agree does not induce ungrammaticality. I conclude with a discussion of the consequences of the lack of AE in Kaqchikel Agent Focus clauses for other theories of AE, and directions of future research.