We investigated whether healthy older adults are able to use an episodic encoding strategy known as unitization, which allows for subsequent associative retrieval based on familiarity, to overcome their associative memory deficit. Young and healthy older participants were presented with word pairs either together with a definition that allowed to combine the word pairs to a new concept (high unitization condition), or together with a sentence frame (low unitization condition). In Experiment 1, an age-related reduction in performance on a standard associative recognition test was observed in both conditions. This deficit was unexpectedly not reduced, but tended to be larger in the high than the low unitization condition. According to receiver-operating characteristics, this difference was due to a reduction of recollection, but not familiarity, in the high unitization condition. Instead of a standard recognition test, Experiment 2 used a 2 alternative forced choice (2AFC) test designed to maximize the contribution of familiarity to associative recognition. Although the disadvantage of older adults in the high versus the low unitization condition was abolished, there was still no performance advantage for the high unitization condition. Event-related potentials (ERPs) recorded during the encoding phase of Experiment 1 suggest that, while young adults engage in predictive processing during unitization, older adults do not engage in such predictive processing, which may prevent them from using unitization to their advantage in the subsequent associative memory test. We discuss the task characteristics that have an impact on the effect of unitization conditions on associative memory in older adults.