Of the three logically possible approaches to contrastive left-dislocation (CLD) (base-generation cum deep anaphora; movement cum surface anaphora; elliptical clausal juxtaposition cum resumption), two are represented prominently in the recent literature. Ott’s (2014) account treats CLD uniformly in terms of clausal juxtaposition, the first clause being stripped down to its contrastive topic via an ellipsis operation said to be akin to sluicing. He argues that this analysis is superior to Grohmann’s (2003) approach, featuring movement within a single prolific domain and late spell-out of a resumptive element. Using data mainly from Hungarian and Dutch, we reveal problems for Ott’s biclausal account that undermine its apparent conceptual appeal and compromise its descriptive accuracy. We show that the ellipsis operation required is sui generis, that the approach fails to assimilate the crosslinguistic variation attested in the availability of multiple CLD to known cases of parametric variation in the left periphery, and that it undergenerates in several empirical domains, including P-stranding and “floated” arguments. Grohmann’s movement-cum-surface-anaphora analysis as it stands also cannot handle all these data, but it can be fixed to fit the facts. For Ott’s analysis, no patches seem available. Some further empirical properties of CLD appear underivable from either of these approaches. For these, the base-generation-cum-deep-anaphora analysis can be considered.