To account for the tense-wise and person-wise uneven distribution of null subjects in Hebrew, we argue first that Hebrew finite T can have either a full phi set – with person and number, a partial one – with number only or no phi set at all. Second, Hebrew pro is argued to lack a person feature. Feature matching with T consequently fails when T has [person], but succeeds when T lacks [person], albeit with the subject interpreted impersonally. Third, ‘control' of pro involves the assignment of a person feature to pro, rendering it capable of reference. Fourth, the fact that Hebrew pro can only be first or second person is an illusion: The speech act participants are associated with a functional head SAP0, to which the first and second person pronouns cliticize. Thus, there is no independently-referring pro at all in Hebrew.