Countless experiments have been devoted to understanding techniques through which memory might be improved. Many strategies uncovered in the literature are thought to act via the integration of contextual information from multiple distinct codes. However, the mnemonic benefits of these strategies often do not remain when there is no clear link between a word and its multisensory referent (e.g., in abstract words). To test the importance of this link, we asked participants to encode target words (ranging from concrete to abstract) either by drawing them, an encoding strategy recently proven to be reliable in improving memory, or writing them. Drawing provides a compelling test case because while other strategies (e.g., production, generation) shift focus to existing aspects of to-be-remembered information, drawing may forge a link with novel multisensory information, circumventing shortcomings of other memory techniques. Results indicated that while drawing’s benefit was slightly larger for concrete stimuli, the effect was present across the spectrum from abstract to concrete. These findings demonstrate that even for highly abstract concepts without a clear link to a visual referent, memory is reliably improved through drawing. An exploratory analysis using a deep convolutional neural network also provided preliminary evidence that in abstract words, drawings that were most distinctive were more likely to be remembered, whereas concrete items benefited from prototypicality. Together, these results indicate that while the advantageous effects of drawing exist across all levels of concreteness, the memory benefit is larger when words are concrete, suggesting a tight coupling between the drawing benefit and visual code.