Graphical Literacy for STEM Students: Testing the Effects of Context and Scaffolding During Learning

Courtney Stevens
Professor of Psychology
Willamette University

Need: Graphical data displays are ubiquitous in introductory STEM courses and textbooks (e.g., Brown et al., 2015; Kump et al., 2010; Myers, 2013; Reece et al., 2014) as well as in popular media and news reports (Peden & Hausmann, 2000; Zacks et al., 2001). Unfortunately, prior studies indicate many college students struggle with graphical literacy, particularly when multiple variables are present in the data display (Bowen & Roth, 1998; Shah & Hoeffner, 2002; Shaughnessy, 2007; Stevens & Witkow, 2014; Stevens et al., 2016). Thus, there is a need to identify effective methods to support basic graphical literacy skills, which are important both for early STEM learning and for a scientifically literate populace.

Guiding Question: Can graphical literacy skills be improved in first- and second-year college students through an online training program and, if so, what program features will yield the greatest benefits?

Method: Here, we developed several versions of a short, online graphical literacy training program. 356 first-and second-year college students were randomly assigned to one of four different training programs or a control condition. All graphical literacy training programs included practice (with corrective feedback) answering 12 graphical literacy questions, similar to those used on a final graphical literacy assessment. The four training programs varied in the presence or absence of two design features, each of which was informed by different learning theories. In particular, we experimentally manipulated training conditions according to whether scaffolding (progressing from easier to more difficult graphs and questions) and/or context (providing background information about the variables displayed and their meaning in relation to a research question) was provided. The final assessment required students to interpret data and draw conclusions from multivariate bar graphs. This assessment was also completed two weeks later to assess longer term maintenance of any gains. As well, evaluation was conducted in two distinct settings, a 2-year community college and a 4-year college, in order to test for possible moderating effects of institutional setting in program impact.

Outcomes: Results indicated that context and scaffolding were each independently associated with better performance on the final graph-reading assessment. This was true even though all training conditions included an equivalent number of total questions and corrective feedback. Preliminary analyses indicated gains persisted for at least two weeks and were of similar magnitude in students attending both 2- and 4-year colleges, suggesting both durability of gains and generalizability to different college student populations.

Broader Impacts: These findings support the use of structured online graphical literacy training materials to support a key foundational skill important for STEM learning. Given the importance of graph reading skills for scientific literacy more broadly, effective methods for training graphical literacy will be valuable for all students. More broadly, these findings imply ways in which college courses can be modified to structure the learning environment to better support STEM skill learning. Project materials will be distributed free of charge for use by college instructors through posting to pedagogical web sites as well as for pre-college students through distribution to college-access programs.


Courtney Stevens, Willamette University, Oregon; Melissa Witkow, Willamette University, Oregon; Tova Hershman, Willamette University, Oregon; Allie Spiekerman, Willamette University, Oregon; & Chagall Ford-Roshon, Willamette University, Oregon