Introduction

2015 marks the 30^th anniversary of a comic strip created by Alison Bechdel, in Dykes to Watch Out For. In this strip, two women discuss the criteria for what would become known as the Bechdel Test – for one of the women, a film has to have 1) two female characters, 2) who talk to each other, 3) about something other than a man, in order for her to watch it. Their exchange indicates that this is a rare combination indeed.

Source: http://dykestowatchoutfor.com/the-rule

My initial goal for this visualization was simply to see how modern movies fare under this test. My hypothesis was that many films would still fail the test, but I was not sure at which criterion they would fail most significantly. I also wanted to see if there was any difference between commercially successful films and films that had received critical acclaim. In creating my dataset, however, I found an additional aspect to explore. The BechdelTest.com website allows users to rate films based on the established criteria, but also allows other users to agree, disagree, or otherwise comment on a rating. Even a film that was rated as passing all three criteria often had a long list of comments disputing the quality or content of the conversation.

Materials

I decided to limit my scope to five years of film. Since 2015 is not yet over, I used the time period 2010 to 2014. I obtained the list of Academy Award for Best Picture nominees and top ten films at the box office for each year, and used the ratings from BechdelTest.com to create a dataset with the following variables for each film:

• Year (year of the film’s release)
• Bechdel 1 (Y for a pass, N for a fail)
• Bechdel 2 (Y for a pass, N for a fail)
• Bechdel 3 (Y for a pass, N for a fail)
• Bechdel Overall (Y if passed all three criteria, N if failed)
• First Billed (M for a male credited first, F for a female credited first, obtained from IMDB.com)
• Contention (Y if there was serious argument over a film that passed the test, N if there was no serious argument, and ‘failed’ if a film did not pass and therefore the variable was not applicable)

I created separate files for award nominees and box office winners, as each is a separate observational unit. Note that I did not include the First Billed variable in my visualization. I originally intended to investigate whether the variable had any impact on a film passing the Bechdel Test, but so few films credited a female first that it became a moot point.

Methods and Discussion

My initial thought was to graph results by year, but the time span I chose is too short to make a meaningful time series visualization. There was no reason for movies to suddenly change their content between 2010 and 2011. What I really wanted to show was how many films passed the first criterion, then how many passed the second, and how many passed the test overall, with each stage showing a decreasing portion of passing films. I therefore decided on a stacked bar graph since Few advises the use of stacked bars “only when you must display measures of the whole as well as the parts” (2005, p. 14).

I used Tableau Public, version 9.1 to create my visualization. After uploading the Academy Award nominee dataset, I changed Bechdel 1, Bechdel 2, and Bechdel 3 to measures and used the Create Group function to group each by Y (pass) and N (fail). The three Bechdel variables were added to Columns, and Number of Records was added to Rows. I also added the Bechdel variables to the Color property of the Marks card so that the most effective pre-attentive element was used to indicate pass/fail for Bechdel criteria.

I followed the same steps to create a similar graph from the dataset of box office winners. Once I had the two graphs I added both to a common dashboard. I chose Desktop as the size of the dashboard, based on the audience I had in mind for usability testing. I planned to recruit colleagues, and at our place of employment many people only have desktops, and those that have laptops often leave them in docking stations and use a large monitor for most of the work day.

Since the visualization was rather bland at this point, I created an icon to represent each of the Bechdel criteria, using MS Word web images. I added these to the dashboard using the Image option.To minimize chart junk, I placed the icons between the two graphs, rather than repeating them below each graph.

To further reduce unnecessary ink, I removed ‘2010-2014’ from each graph and added it to the title of the overall dashboard. As each graph covers the same time period, there is no need to repeat this information. I also removed most of the borders and grid lines.

I decided to add a third color to represent films that passed with contention. I added the Contention variable to the Color property of the Marks card, for the final bar of each graph only.  Though the visualization is about women, in order for productive discussions to take place, it is essential that people of all genders be aware of these issues, thus I changed the colors to options from Tableau’s Color Blind 10 palette so as not to exclude males.

In addition to a dashboard title, I also added text boxes explaining the criteria of the Bechdel Test and the contention category. The Bechdel criteria were listed on the left, as English speakers read left to right, and thus would see this explanation before the graphs. I added the information on films that passed with contention to the right of the graphs, in a text color similar to the color used to identify films passing with contention.

The final product shows that despite the low bar set by the Bechdel Test, many films fail it outright, and very few pass without contention. There is also no significant difference between the test performance of Oscar nominees and that of top box office earners. Somewhat unexpectedly, Oscar nominees do slightly worse.

Post-Usability Testing Revisions

After accumulating usability feedback from a focus group and remote participants who used the visualization to complete a series of tasks, I made the following revisions:

• Added more text to describe the origin of the data.
• Staggered the graphs and criteria descriptions so that for each criterion, the Oscar nominees and box office hits are next to one another, for easier comparison.
• Changed the title of the color legend from ‘Bechdel Score’ to ‘Criterion Rating’ to indicate that each bar shows only how films perform under a particular criterion and not the overall test.
• Change the color scale so that blue stands for a pass, dark red is a fail, and orange is a pass with contention, in order to match colors more closely with user expectations.
• Deleted the icons representing each criteria, as they hindered interpretation of the visualization more than they helped.
• Thickened the grid lines and added percentages as labels to enable users to make exact comparisons.

Future Directions

There were several excellent recommendations from my usability testing participants that I was unable to incorporate at this time. These included comparing films from different countries, comparing films by genre, and including the results of a reverse Bechdel Test (2 named males who have a conversation that is not about women). This last is most interesting to me. The Bechdel Test alone is not sufficient to determine whether a film is feminist, anti-feminist or neutral, but a graphic that showed the same films and whether they pass Bechdel and/or the reverse Bechdel would provide a means of direct comparison between the representation of men and women in film.

References

Few, S. (2005). Effectively communicating numbers: Selecting the best means and manner of display. Boise, Idaho: ProClarity Corporation.