It was a pleasure to see a letter to the editor in Anesthesiology that mentioned the learning advantage of pictures compared to words.  This advantage is known as the picture superiority effect. I was skeptical of the stated learning advantage of the picture superiority effect when the authors describe that the picture superiority effect results in a 650% increase in retained information after 3 days. To pursue the evidence behind this claim I reviewed the citation provided by the authors.  Unfortunately, the citation was a book that provided only an infographic. This infographic referenced another book, but that too did not provide a citation to the original work.   Subsequently, I sought peer-reviewed studies with measures of the size of the picture superiority effect. Randomized controlled studies support the picture superiority effect but with a substantially lower effect size. For example, the benefit of pictures over just words for recognition memory at 48 h was 25% and, at 1 week, it was 12% (see table 2 in Rajaram and Pereira-Pasarin, Hits, Noncollaborative condition).  Another relevant study found that the picture superiority effect was 35% when learners used shallow processing but 0% when learners engaged in deep processing (see table 1 in Rajaram and Pereira-Pasarin, Noncollaborative condition).  In a study that compared memory for words to memory for words plus pictures (i.e., not exactly a test of the picture superiority effect) using a 2-day delay, the authors found a 515% effect when the learners were instructed to use shallow processing (e.g., shape, curve) but only a 20% effect when the learners were instructed to use semantic processing (e.g., meaning). Last, the picture superiority effect can even be reversed under certain conditions. In summary, the picture superiority effect appears real and can be an important element in learning, but the magnitude of the effect seems to vary with the details of the experiments. One concerning finding is that larger picture superiority effect effects were found with shallow processing, but much smaller, or no effects were found with deep processing.

This is of particular concern because infographics are meant to convey meaning, and extracting meaning requires deep processing.