Analyses of the that-trace effect have historically been divided between those that attribute the effect to constraints on the phrase structure representation and those that attribute it to constraints on linear adjacency, with some of the latter attributing it more specifically to prosodic constraints. If a prosodic approach is correct, we expect to find cases where the that-trace effect can be ameliorated by manipulating the prosody, while leaving the syntactic configuration essentially intact. We examine a prosodic account by means of carefully designed acceptability experiments in two different prosodic contexts with audio stimuli. Our results suggest the examined cases of prosodic amelioration are at best illusory. The first case, narrow contrastive focus, does increase acceptability of that-trace sentences, but this is part of a more general phenomenon in which contrastive focus improves all cases of embedded subject extraction, with or without that. The second case, contraction between that and a following auxiliary, does not appear to lead to any amelioration at all. These findings remove what would be very striking evidence in favor of a prosodic approach, but otherwise leave open the question of the proper analysis of the that-trace effect.