Abstract <p>A direct comparison of the neural correlates of verbal and nonverbal working memory (WM) is complicated by the fact that neurocognitive studies addressing these two types of WM use tasks that differ in a number of parameters, including WM processes of interest (e.g., material retention or processing), stimuli, task instruction and/or structure, and difficulty, and the tasks are therefore incomparable. We developed verbal and nonverbal versions of the Sternberg and n-back tasks, equivalent in all aspects except the possibility to activate auditory-verbal and/or articulatory representations of material being remembered. To achieve this aim, we used characters of the artificial font BACS2sans as nonverbal stimuli to match the Cyrillic letters used as verbal stimuli. Testing the four tasks in fMRI and MEG studies on the same group of participants (<i>N</i> = 35) revealed no differences in response accuracy and time between the verbal and nonverbal versions of the 2-back task. For both versions and both performance measures, there was an effect of WM load (2-back vs. 0-back). In the Sternberg task, the effect of material (verbal vs. nonverbal) disappeared after the set size was balanced appropriately (2 symbols and 4 letters; 4 symbols and 8 letters). The WM load also affected response accuracy and time. Dependence of the performance on WM load indicates the validity of the developed tasks. The results have also shown that, besides matching design and instructions, the verbal and nonverbal versions of the tasks are generally comparable in difficulty, which opens up good prospects for the use of these tasks in future neurocognitive research on WM.</p>