Tis paper examines the properties of inclusive generic constructions, focusing on languages where the inclusive generic pronoun is a null category. We investigate empirical data from a set of languages with and without agreement to test Phimsawat's (2011) hypothesis that the inclusive generic pronoun lacks all phi-features, and therefore has the least restricted reading, due to there being no restriction on the reference. We show that this hypothesis cannot hold true universally, as phi-features trigger agreement in inflecting languages. We show that there is a correlation between presence of agreement and restriction to human reference for null inclusive generic pronouns, based on comparison of a set of languages without agreement (Tai, Mandarin Chinese, Korean, Vietnamese, and Sinhala) with a set of languages with agreement (Finnish, Brazilian Portuguese, Hebrew, Basque, and Tamil). An explanation in terms of feature architecture is proposed for this correlation. A prediction for generic PRO is discussed and shown to be inconclusive or false.
