Memory for emotional items is often better than memory for neutral items. In three experiments, we examined whether this typical finding is due to the higher semantic relatedness inherent to emotional items, a confound in previous studies. We also controlled for other possible confounding variables, such as imagery. In Experiments 1 and 2, participants encoded lists of emotional and categorized neutral words equivalent in semantic relatedness, as well as lists of random neutral words with lower semantic relatedness. In Experiment 3, the lists were mixed, containing words from all the conditions. Surprise free recall was tested after a 40- to 55-min retention interval. Free recall of emotional words was better than that of random neutral words, replicating the classic effect. Importantly, categorized words were recalled better than random neutral words, and not worse than emotional words. These results emphasize the important role of semantic relatedness in the classic effect and suggest that organizational processes operate alongside arousal-related ones to enhance memory for emotional material.