In let alone ellipsis, the remnant typically stands in prosodic contrast with its correlate (e.g., John can’t run a MILE, let alone a MARATHON). To interpret the remnant (a marathon), the processor must locate the contrasting correlate phrase (a mile) in the prior clause from among other same-category competitors. Experimental and corpus research find that the most local possible correlate is vastly preferred (Locality Bias; Harris & Carlson, 2015), a pattern that has also been observed for other clausal ellipsis structures. This study uses pupillometry to examine the real-time effect of prosody in signaling non-local correlates, finding an immediate processing cost when Locality is violated, even when the non-local correlate is contrastively accented. The findings support the claim that default locations of focus disrupt focus-sensitive processes during online auditory comprehension, and further suggest a complex interplay between overt and default locations of focus, information structure, and ellipsis resolution.