Effects of Splitting Text into Multiple Columns

Experiment

Introduction and Hypothesis

We conducted the experiment between three groups, manipulated two independent variables, and measured 3 dependent variables, including subjective satisfaction.

Independent variables:

Width of the entire text area
Treatments: 600, 800, 1000 pixels

Number of columns within the text area
Treatments: 1, 3

The three subject groups were split based on the value of the first independent variable. One group of subjects worked with a 600-pixel wide window, another - with an 800-pixel wide window, and the third - with a 1000-pixel wide window. Both of the second independent variable values were used with each group. Each group read text in both the single-column format and the three-column format. The order in which these two formats were presented to the subjects was random.

Dependent variables:

Time to Correct Completion
Measurement: time required to read the text and correctly answer the questions

Number of Incorrect Answers Given
Measurement: a total number of incorrect answers given after reading one article

Subjective Rating
Measurement: subjects' satisfaction with a page format on the scale from 1 to 9

Based on psychological theories and personal experience, our hypothesis was that for wider pages, the comprehension rate and subjective satisfaction would increase and the number of errors would decrease as text was split into multiple columns. We anticipated a minor improvement for narrow pages and a significant improvement for typical width pages. We predict a better than linear rate of improvement as the page width increases up to some limit, with the rate slowing down as the page width becomes unusually great.

Pilot Study Results

The pilot experiment was conducted with two subjects. In this phase of the study, the subjects were not to be split in groups, and each subject were to read a total of six articles: a single-column and a three-column format version in a 600, 800, and 1000-pixel wide window. After the pilot experiment, we found that subjects got tired of reading so much text and took longer time than we anticipated. Therefore, the decision was made to do a study between groups.

After evaluating the results, we found that, as expected, the completion times were slightly less when the text was split into three columns. The rate of errors was significantly lower when the text was split into three columns. However, one subject showed a strong preference for the single-column format and the other subject rated the three-column format much higher.

Subjects

All the subjects in the experiment were randomly chosen young college students or working professionals, who we believe to be typical users of word processors, web browsers, and other applications that display text. A total of 30 subjects participated in the experiment.

Materials

We conducted the experiment between three groups of subjects, for which three versions of a test application were developed in Microsoft Visual Basic. The difference between the three versions was the width of the application window: 600, 800, and 1000 pixels. Therefore, one group of subjects worked with a 600-pixel wide window, another - with an 800-pixel wide window, and the third - with a 1000-pixel wide window. Each of the three versions of the application can be downloaded from the Appendices page. Below we describe the functionality of our test application and offer a number of screenshots.

Procedures and Problems

During the experiment, the application first provided simple instructions to the subjects and then displayed in a single-column format an article randomly chosen by us from a news web site. When the subjects finished reading the article, they clicked on a button, and the application displayed three simple multiple-choice questions that were formulated to test the reading comprehension. A correct answer to a question was signaled by a checkmark symbol, and the subjects were allowed to change their answers and return to the article text until they answered all the three questions correctly. After that, the subjects clicked on a button, and the application displayed in a three-column format another article from the same source. As with the first article, when the subjects finished reading it, they clicked on a button, and the application displayed three simple multiple-choice questions. After the subjects answered them correctly, they clicked on a button and were asked to rate their satisfaction with the two different formats and provide general comments. This concluded the experiment.

The test application kept track of the time the subjects required to read the articles and correctly answer the questions, as well as the number of times they chose incorrect answers.

Below are the screen shots of the test application: text displayed in a single-column format, text displayed in a three-column format, and the multiple-choice questions. You can click on the images to see them in full size.