“May I suggest you consult the senior authors of your paper who should know how to label a scientific graph properly so that our readers can understand them even if they are non-specialists” read the e-mail from an unnamed scientific journal’s editor in chief (EIC). With my eyes and mouth wide open, I read it again: “…There is still not enough clarity about the labeling of the figures. …May I suggest …”
My manuscript had made it successfully past the peer review process based on the scientific contribution it made to the field of animal camouflage, but it turns out that was not enough for publication: the EIC was kindly informing me that it was going to take 56 more labels and 84 tick marks on Figure 3 of my article to get an official acceptance.
Take a blank sheet of US-letter size paper and write your first name 56 times. Whether it is Eli or Kananinoheaokuuhomeopuukaimanaalohilo, it will take a lot of ink! Now imagine how much more clutter there would be if that sheet of paper was no longer blank and had 28 plots on it. In the case of Figure 3, all plots had identical axes, so I did what any scientist primarily concerned with conveying her message to the reader would do: maximized the data-ink ratio by labeling the plots only once on top of the page.
The data-ink ratio concept was introduced by Edward Tufte, perhaps the most well known information design and data visualization expert. It denotes the “proportion of a graphic’s ink devoted to non-redundant display of data-information”. Jac-Luc Besson, another expert with a focus on effective communication of information, quantifies the same principle via his second and third laws of effective graphical displays: maximize signal-to-noise ratio, and use effective redundancy. Science photographer and MIT scientist Felice Frankel highlights that well thought-out graphics benefit authors and readers because “visually representing an idea in science not only requires an author to clarify her thinking and improve her ability to communicate with others but also leads to new insights into her work”.
Photographs, plots, charts and other visual representations of data arguably constitute the most important part of scientific communication. In fact, often times a reader’s exposure to a scientific paper is limited to a quick skimming of its abstract and figures. As a result, there is an implicit expectation that figures will convey the key concepts of the work, telling a reader exactly what she needs to know.
Unfortunately, this is rarely the case. An author spends anywhere between a few months to a year (or more) working on the write-up of research that may have taken much longer to develop. This prolonged and intense exposure sometimes misleads authors into thinking that their graphics are sufficiently self-explanatory; but to a first-time viewer they may appear cluttered, confusing or even incomplete. What an author decides to include in their data presentation, then, is a measure of how well they understand their data and their audience.
That was exactly my argument when I wrote to the editor that the 28 plots in Figure 3 had identical axes and therefore individual labels would be redundant. I only had the courage to write such a response because a few weeks prior to submitting my manuscript, I had attended Felice Frankel’s Master Class on Engineering and Science Visualization, sponsored by the Mechanical Engineering Department. Ms. Frankel not only presented mind-blowing before-and-after versions of scientific figures and photographs, but also talked about the resistance of most journals towards accepting less traditional visualizations. In fact, one mechanical engineering professor who had helped organize the workshop said that as the editor of a journal, he often asked authors to include top and right axes with tick marks in addition to the standard horizontal and vertical axes in a two-dimensional plot, because under-labeling of the figures was more likely to confuse readers than over-labeling. Ms. Frankel’s workshop convinced us all that this is not necessarily true and more often than not, less is more.
Following the workshop, I submitted my manuscript with figures that I thought had optimal layouts and data-ink ratio. The rest is history. The EIC asked for 56 more labels and many tick marks on Figure 3 so that “readers could understand them even if they were non-specialists,” and I obeyed. In case you are wondering whether this was an open-access publication that many non-specialist readers may come across; it was not. In fact, I had targeted this journal because of its highly specialized reader-base. In the end, I had to obey the editor and add the requested labels.
The axis labels and tick marks I had to add to Figure 3 are minor details when compared to other challenges in visualization and presentation of scientific data. Luckily, MIT understands and supports the need to improve scientific communication and consequently push journal standards. There are lectures and workshops in any given semester with emphasis on producing better scientific photographs, finding the best forms of graphic visualization and basic design principles. Our libraries have all publications of Tufte, Besson and Frankel, in addition to many other info-graphics resources like appointments with Felice Frankel. Hopefully, journals will move to more modern and concise graphics in the future.
Appointments with Felice Frankel can be made by emailing her at email@example.com. She preferentially schedules students in Mechanical Engineering, Chemical Engineering, and Materials Science and Engineering, but all are welcome.
This is a new column for students to share their experiences in research. Write to firstname.lastname@example.org and email@example.com if you are interested in contributing!