Retrieval practice of previously studied material can impair subsequent memory for related unpracticed material. An emerging view holds that such retrieval-induced forgetting (RIF) may affect episodic recollection, but not the context-free familiarity of the affected items. Here, a survey of accruing recent findings of RIF in recognition tests shows that the impairment of unpracticed material depends vitally on baseline memory strength. Therein, the absence of RIF under specific conditions, previously taken as evidence for the immunity of familiarity, can be predicted on grounds of exceedingly low baseline levels. Similarly, differential RIF effects on the parameters of dual-process recognition models can be explained by baseline differences, suggesting that RIF might impair any sub-process that substantially contributes to overall recognition accuracy. By contrast, the strengthening of practiced material appears independent of baseline levels and does not predict the magnitude of RIF, in accordance with an inhibitory causation of the forgetting. In summary, the inventory presents RIF in recognition as a subtle proportional impairment, future illumination of which may demand increased attention to baseline memory levels.