Retroactive interference in episodic memory has been shown to be time-dependent: The shorter the time interval between an original and an interpolated learning phase, the stronger the interference effect is predicted to be. Although this temporal gradient of retroactive interference (TGRI) has been replicated repeatedly, some null results cast doubt on its robustness. At the same time, the TGRI is viewed by many authors as a central piece of evidence for the claim that consolidation in episodic memory occurs whenever hippocampal resources are not required for the encoding of new information. In contrast, temporal distinctiveness theory attributes the TGRI to differences in the discriminability (i.e., retrievability) of items in memory. In two preregistered studies, we aimed to scrutinize both issues of replicability and theoretical explanations. In a first replication study, we closely adhered to the original procedure introduced by Ecker et al. (2015) but conducted the study in an online setting instead of a more controlled lab environment. A joint analysis of the original and our replication data indicates a successful replication of the original finding. In a second study, we applied a newly developed multinomial processing tree (MPT) model to disentangle storage and retrieval contributions to free recall performance within the same paradigm. The parameter estimates from the MPT analysis allow for an effective test of both the consolidation and the temporal distinctiveness hypothesis. Thereby, our results help clarifying the cognitive mechanisms underlying the TGRI.