Reading Comprehension and Rate: One Column vs. Three Columns

Conclusions

Impact for practitioners

The use of multi-column page layouts has been shown by this experiment to improve reading time by as much as 25%.  The impact of this finding can affect designers of web pages, web browser, and mobile units.  With the ability to read through a document faster, users can become more productive.  
The use of multi-column page layouts is shown by this experiment to have no significant difference in comprehension level.  These findings can affect designs for any text layout that requires users to comprehend.  Designers of learning software, books, or web browsers could use this information to design a screen layout without having to worry how the columnar screen layout of the text is going to affect the comprehension level of users.

Suggestions for future researchers

Some suggestions for future researchers are as follows:

• look for all possible extraneous variables

• use larger sample size

• diversify to better mimic the population

• choose suitable reading material

While we tried to control for many possible extraneous variables, controlling for all variables takes much planning and the ability to "look beyond the surface."  Small, seemingly insignificant variables or hidden variables can pop up at any time is not carefully controlled for.  
A larger sample size will increase the likelihood of achieving reliable results.  Our sample size of twenty subjects allows for a few unreliable data items to completely throw off the results, while a few erroneous data items will not have a great effect when using a large sample size.  Also, if using large sample sizes, subjects can be divided into two groups (one-column and three-column), with each reading the same passage (in one-column format and three-column format).  This way, any differences in the passages are controlled, since each group is reading the same passage. 
Diversifying will control any differences in the subject's character.  Since the population is well diversified, using a sample that mimics the population will help to achieve reliable results.  Our experiment was not very diversified as 80% of our subjects were of the age 19-24, over 75% were male, and over 50% were Computer Science students.  Using a sample size of diversified races, age, skill level, and gender will control any character biases that can affect results.
Choosing suitable reading material for this experiment is essential to achieving reliable data.  Typically, subjects will be weary to honestly participate in this type of experiment if the material is not of interest to them.  Of course, finding passages that are suitable and of interest for all subjects is probably impossible, but choosing a passage that can accommodate most subjects will help to provide more reliable results.

Refine the theory or develop a new one

Our results contradict past experiment results that state there are no significant differences in reading rate between one-column and three-column formats.  Our results show there is a statistical difference between the two formats.  Therefore, we offer a refinement to the theory that reading rate is affected by the number of columns text is broken into; more precisely, reading rates of three column formats are greater than reading rates of one column formats.
We also offer a refinement to the theory about test scores.  Previous studies have shown that reading comprehension is greater with a one-column format than a three column format.  Our results show that people do not comprehend any better, given either format; thus reading comprehension level is not affected to the layout of the text.  

Department of Computer Science: Direct questions and comments to the student editorial team