There are three competing analyses of Left-Branch Extraction in Slavic languages: sub-extraction, distributed deletion, and remnant movement. This paper argues that Left-Branch Extraction does not share a well-known property of remnant movement derivations, namely an anti-reconstruction effect known as Barss' Generalization. When applied to inverse linking constructions, it can be shown that the remnant movement analysis predicts the absence of such readings when a numeral undergoes Left-Branch Extraction. The fact that this prediction is not borne out casts doubt on the validity of the remnant movement analysis.