Verbal overshadowing reflects the impairment in memory performance following verbalization of nonverbal stimuli. However, it is not clear whether the same mechanisms are responsible for verbal overshadowing effects observed with different stimuli and task demands. In the present article, we propose a multiprocess view that reconciles the main theoretical explanations of verbal overshadowing deriving from the use of different paradigms. Within a single paradigm, we manipulated both the nature of verbalization at encoding (nameability of the stimuli) and postencoding (verbal descriptions), as well as the nature (image transformation or recognition) and, by implication, the demands of the final memory task (global or featural). Results from 3 experiments replicated the negative effects of encoding and postencoding verbalization in imagery and recognition tasks, respectively. However, they also showed that the demands of the final memory task can modulate or even reverse verbal overshadowing effects due to both postencoding verbalization and naming during encoding.