In accounting for the so-called copy raising (CR) construction, movement approaches have traditionally assumed that the embedded subject of the construction is moved to the matrix subject, leaving behind its pronominal copy. This simple analysis, appealing it may be, encounters issues with authentic data where there is no coindexation relation between the two subjects. This paper suggests that the variations of the CR have to do with the interactions among lexical properties of the CR verbs, argument realization, and interpretive constraints.