Ariel (1985, 1988) has argued that discourse anaphora is determined by reference to the notion of accessibility in memory storage. Under the assumption that mental representations (specifically those of NPs) are accessible to addressees in varying degrees, the claim is that speakers choose between referring expressions so as to mark such accessibility differences for the addressee's convenience. Thus, different referring expressions (e.g., definite descriptions, demonstrative pronouns, pronouns) mark different degrees of accessibility. Definite descriptions mark relatively low accessibility, demonstrative pronouns and pronouns mark relatively higher degrees of accessibility. The linguistic coding of degrees of accessibility is claimed to derive from three universal principles: Informativity, Rigidity and Attenuation, such that more informative, less ambiguous and more highly pronounced, longer forms retrieve less accessible referents. This article argues that precisely the same mechanism is responsible for the distribution of sentential anaphoric expressions. I focus on Hebrew zero/pronoun choices, which are realised in a variety of forms and inflectional morphemes. The conclusion is that both intuitive grammaticality judgments and distributional patterns in texts corroborate the accessibility claim. Thus, the richly informative, rigid and fully articulated 1st and 2nd person pronouns mark relatively low accessibility. The present tense inflection, which is uninformative, ambiguous and attenuated, marks an extremely high degree of accessibility. Other markers are used for a variety of intermediate degrees of accessibility.