The retroactive interference paradigm has been used in a variety of settings to investigate the nature of the representation of memory. Much of the research using this paradigm is methodologically flawed because it involves treatment comparisons which are inappropriate. It is argued that comparisons must be made between presentation conditions having the same interpolated activity and that, in addition, evidence is required that the differences are not confounded by acquisition level differences if one is to infer mode-specific interference and representation. The methodological issues are discussed in detail and the utility of the retroactive interference design is questioned. Studies employing the design are reviewed, and several conclusions are drawn: (1) there has been no unambiguous demonstration of visual mode-specific interference, (2) there has been no clear demonstration that imagery instructions produce memories that are more susceptible to visual than to auditory interpolation, and (3)no clear demonstrations are yet available that memory for spatial location is more susceptible to visual interference than memory for letters.