Prosodic prominence shifts are anaphoric

• Prominence shifts are anaphoric, and not reducible to the accessibility of the words or meanings involved in an utterance.
• A crucial test case differentiatinganaphoric and accessibility accounts involves coordinated antecedents.
• Coordinated antecedents led to variability in production of prominence.
• This variability is shown to be due to different anaphoric choices.

This paper presents evidence that shifts in prosodic prominence are anaphoric and require a contextually salient antecedent, similar to pronouns. The argument is based on a series of experiments looking at prosodic optionality in dialogues in which there are multiple potential antecedents embedded in a contextually salient coordinated structure. By looking at the interaction with adverbs that restrict the choice of antecedent, we show that the observed prosodic variability reveals different anaphoric choices, and hence different speaker intentions. The results are incompatible with the hypothesis that prominence shifts can be explained purely in reference to low-level facilitation due to repetition of the linguistic structure or accessibility of it referent, and are not reducible to existing accounts of prominence in terms of predictability.