Argument ellipsis refers to the phenomenon of unpronounced nominal arguments that are interpreted as full noun phrases. One particularly influential account claims that argument ellipsis is possible if and only if the argument is not φ-agreed with — the Anti-Agreement Hypothesis (Saito 2007, Şener and Takahashi 2010). Otaki, Sugisaki, Yusa, and Koizumi (2013) show that null subjects and objects in Kaqchikel are not interpreted as full noun phrases, and therefore not the result of argument ellipsis. They say that these facts lend support to the Anti-Agreement Hypothesis, given that Kaqchikel verbs agree with both subjects and objects.

In this paper, we consider additional syntactic environments not considered by Otaki et al. and show that the interpretation of Kaqchikel null arguments in fact cannot be held up as support for the Anti-Agreement Hypothesis. The empirical testing ground we introduce is the Agent Focus construction in Kaqchikel, where transitive verbs only exhibit φ-agreement with one argument (see e.g. Preminger, 2011, 2014, Erlewine, 2016). We show that, in Agent Focus constructions, even arguments that are not Agreed with disallow argument ellipsis. This fuller set of data from Kaqchikel thus provides empirical and conceptual evidence against the Anti-Agreement Hypothesis for argument ellipsis, contrary to the conclusions of Otaki et al.

Keywords: argument ellipsis Kaqchikel Mayan agreement Agent Focus