Anaphoric Islands

Abstract

Abstract Following Postal, alternative grammar‐based explanations of so‐called inbound and outbound anaphora were offered including more general notions of Lexical Integrity, all of which assumed that both inbound and outbound anaphora were categorically disallowed. However, it was soon observed that judgements of ‘outbound anaphora’ were gradient and context‐sensitive. Such counter‐examples were relegated to some (undefined) mechanism of ‘pragmatic amnestying’, inexplicably unavailable for ‘inbound anaphora’, which remains ill‐formed in any context. A general prohibition on pronouns occurring within words does seem to be operative in English (though, crucially, not in other languages) but outbound anaphora, as was first argued in Ward et al., is in fact grammatically unrestricted. As with anaphora in general, its felicity is shown to be a function of the accessibility of the referent, that is, the discourse entity to which an anaphor is used to refer. Subsequent work identified a number of formal and functional factors that increase the accessibility of discourse entities and therefore the felicity of anaphors being used to reference those entities. This analysis is supported by a corpus of naturally occurring tokens of outbound anaphora as well as a series of psycholinguistic studies that reveal that topicality and contrast facilitate comprehension of anaphors with word‐internal antecedents. One direct consequence of this work is that references to grammatical constraints on outbound anaphora have all but disappeared from the literature.