Processing information depletes a limited cognitive resource. This resource can recover passively over time, but an open question remains: if resources are spent to process information that soon becomes irrelevant, are these resources immediately released? I present an analysis of the Source of Activation Confusion (SAC) model of human memory and its ability to explain the primacy effect in working memory – the tendency for items at the beginning of a list to be better remembered than those in the middle or end. The SAC model attributes the primacy effect to a limited resource pool that is depleted during the encoding of each item into memory, with resource recovery happening passively over time. The model's core mechanisms are challenged by two experiments that reveal the primacy effect can be reset with the application of a forget signal, suggesting resource recovery is not entirely passive. This means that either the passive resource recovery assumption needs to be reformulated, or that the primacy effect is not a result of resource depletion. Given the broad range of other findings the model can explain, abandoning it is premature. Instead, I propose that in addition to recovering passively over time, memory resources can be manually released when the information encoded with them is no longer relevant. This reformulation is important because it significantly limits the range of possible substrates for the proposed resource. I discuss implications for related theories and phenomena such as ego depletion.