Butcher, K. R. (2006). Learning from text with diagrams: Promoting mental model development and inference generation. Journal of Educational Psychology, 98(1), 182-197.
Butcher (2006) starts with this primary observation: most educational materials rely on pictures, diagrams, and text to communicate information. Given this fact, Butcher identifies two gaps in the literature. First, while many studies have found a positive relationship between the use of pictures and recall or memorization, it is less clear if pictures have a similar impact on deeper levels of learning. Secondly, while principles like the coherence principle can help guide multimedia choices, the ideal format of individual learning materials has not been as consistently identified by the literature. Butcher chooses to focus her study on diagrams. (This includes schematic diagrams like flowcharts or Venn diagrams, and more simple iconic diagrams). Butcher makes a strong argument that this specific topic is under-addressed in the literature, despite the fact that visual elements are such a critical part of course instruction (even more so in online education, I would add).
According to Mayer (2001), the coherence principle states that “learning materials should include only relevant multimedia and should avoid irrelevant pictures, sounds, or words” (4). Applying the coherence principle to diagrams means that we would get rid of any parts of a diagram that are not necessary. Simplifying the diagram can reduce the demands placed on students’ working memories. However, simplifying can also mean an increase in abstraction, so which will be more useful to students? A realistic diagram that is more complex? Or a simple diagram that is more abstract? This is the question the author seeks to explore as she compares functional/simple diagrams and detailed/structural diagrams.
Butcher sets up two experiments. In Experiment 1, she tests the hypothesis that high knowledge learners benefit from detailed diagrams while low knowledge learners benefit from simpler diagrams. Students were provided with a pre-test, and then information in one of three formats (text only, text and simple diagram, and text and detailed diagram). A detailed post-test (including student drawings and explanations, memory questions, and inference questions) was then administered. The author found that both types of diagrams helped students develop a mental model of the concept, but simplified diagrams were best. Contrary to expectations, higher knowledge students didn’t benefit more from detailed diagrams.
Experiment 2 added an element of self-explanation in order to identify why simplified diagrams were effective. Butcher found that simplified diagrams created “more integration inferences” than complex diagrams or text-only instructions. In other words, diagrams helped students generate correct inferences, which is a key marker of deeper comprehension. In my opinion, experiment 2 was more abstract and the results less useful than experiment 1. However, as Butcher points out, many other studies stop once they identify a relationship between multimedia and recall, and so experiment 2 does provide evidence that diagrams/pictures can also influence deeper understanding by encouraging inference generation.
Butcher’s work has specific practical applications for instructors and designers, even though the article itself is quite technical and thus not as accessible as it could be. In particular, the recommendation to favor simple diagrams over “realistic” diagrams could be met with some resistance (I think we all have a tendency to assume is “more is better” from time to time). Being able to point to the research conducted by Butcher can help inform and justify best practices in those instances.
Lastly, I appreciate Butcher’s nuanced overall conclusion, which is that we cannot make a blanket recommendation to always choose simple diagrams or always choose complex diagrams. Instead, we should choose diagrams that make the key functional relationships within the diagram the most clear. In other words, consider the learning outcomes and ensure that the diagram only has what is necessary (no more, no less) to achieve those learning outcomes. Even though this is not an unexpected conclusion in the world of instructional design, it is nuanced nonetheless and strongly supported by the experiments conducted in Butcher’s study.