Reading Comprehension and Rate: One Column vs. Three Columns

Experiment

Introduction and Hypothesis

Most Web pages display text in one-column layout that spans across a page, a format that took root when computer monitors were capable of displaying only 80 characters across the entire screen. However, current computer monitors are far more capable. It is unclear whether or not it is beneficial to the readers, in terms of reading comprehension and rate, to break up a wide one-column text into several narrower columns.

Our experiment seeks to investigate the effects of multi-column text layouts on reading comprehension and rate. Subjects are asked to read two different passages, in one-column and three-column formats, of approximately the same length and difficulty. Each passage is fully contained within one page, that is, the subject can read each passage without scrolling.

Variables

The independent variable of this experiment is the number of columns in which the text is displayed. For this experiment, the independent variable has two treatments- one column and three columns.

The dependent variables are the time, in seconds, the subject takes to read each passage and answer questions about that passage, and the level of comprehension. The level of comprehension is measured by how accurately the subject answers the questions related to each passage. Scores are assigned using the following formula:

Score = 640 - Sum((Correct answer - Actual answer)²)
640

Correct answers have values of 1, which means No, the question could not be answered based on the information presented in the text passage, or 9, which means Yes, the question could be answered based on the information presented in the text passage. Actual answers, which ranges in value from 1 to 9, are the level of confidence in which the subject believes each question can be answered, based on the information presented. The highest possible sum of squares is 640, if the user answers all ten questions completely wrong. In order to get an easy to understand score, the sum of squares is subtracted from 640, and then divided by 640. Higher final score indicates higher level of comprehension.

In addition, at the conclusion of the experiment, the subject is asked whether the one-column or the three-column layout is preferred.

Hypothesis

We hypothesize that the reading time would be greater with the three-column layout, and that the level of comprehension would be greater, as well, with the three-column layout.

Pilot Study Results

We conducted pilot study of our experiment on five of our friends and family members. This small group of people represented our targeted test audience, which is college students or recent college graduates who are comfortable with using a computer, are familiar with the Windows 95/98/NT user interface, and have no difficulty reading text passages in English. Based on the results of the pilot study, we estimated the amount of time an experiment will take, broken down into stages:

Reading the instructions and filling out the background questionnaire - ~1 minute
Reading either the one-column or the three-column text passages and answering questions related to the text - ~5 minutes
Answering the subjective question and the statement of purpose and appreciation - <30 seconds

Based on feedback from our subjects and classmates from our class presentation, we made the following adjustments to our material, which consisted of our testing software running on computers using Windows 95/98/NT operating system.

Changed the "Yes/No scale" to a "1-9 scale" for answering questions, with 1 being No and 9 being Yes, and 5 being Maybe
Allowed subjects to toggle between each text passage and questions, i.e. the subject can read the one-column text, try to answer questions about the passage, and return to the text. This was done in order to minimize the impact of short-term memory on test results
Extended the choices on subject's date range, in response to feedback received during class presentation
Added margins around text passages to ease reading
Eliminated the initial statement about reading "one-column and three-column text passages," and instead stated, "you will be reading two passages..." This was done to minimize subjects' initial bias

Subjects

Twenty undergraduates and recent graduates from the University of Maryland participated in our study. Our subjects consisted of sixteen males and four females, sixteen of which are between the age of 19 to 24. Among our subjects, thirteen are computer science majors, and one of each of the following majors: physics, business management, chemical engineering, communications, geology, interior design, and decision information systems. Prior to the experiment, our subjects filled out a questionnaire regarding their proficiency in using Microsoft Windows operating system, and reading in English. Most of our subjects reported in the questionnaire high proficiency in using Microsoft Windows operating system, and fluency in English.

Materials

The following materials were used to conduct the experiment:

Hardware:

Intel Pentium-based PC
17" monitor with resolution of 1024x768

Software:

Microsoft Windows 98 operating system
Custom made testing software written using Microsoft Visual Basic, divided into the following sections:

Documentation:

Subject consent form

Procedures and Problems

The subjects required no special training to participate in our experiment.

The subjects were presented the subject consent form at the beginning of the experiment, as well as the instruction and questionnaire section of the testing software. Then the subjects were presented with either the one column text passage or the three column text passage, depending on which version of the testing software was administered. Our subjects read the passage, and attempted to answer the five questions related to the passage.

When all five questions were answered, the subjects were allowed to go onto the next text passage of the alternate layout and set of questions. The time the subjects take to read the passages and answer the questions were measured and recorded for each layout by the testing software. The subjects' responses to the questions relating to the passages were stored by the testing software as well, for comprehension level analysis. Upon the subjects' completion, they answered one question about which layout they preferred, one column or three columns.

Problems Encountered

We encountered a few problems while conducting our experiment:

Many of our subjects displayed frustration and boredom in reading the somewhat lengthy text passages, and as a result, may not have answered the questions to the best of their ability. We threw out results from three subjects who answered the questions randomly without reading the passage.
After reading one passage, some subjects learned to answer questions for the second passage by simply toggling between the text passage and the questions, without actually reading the text passage first. This may have caused some dramatically different reading times for the two passages, even though both contain almost equal length text of same difficulty. However, we feel that this effect is minimized by having the one-column text appear first half the time, and the three-column text first the other half.