Reading Comprehension and Rate: One Column vs. Three Columns

Discussion

Reading Completion Times

The results reflect that the average reading time for a one-column formatted passage is greater than a three-column formatted passage;  this result disproves our hypothesis that the one-column format would yield a faster read time.  Users, on average, took 28% more time to read a the one-column format as opposed to the three-column format.  The order in which subjects received the test (one-column before three-column, three-column before one-column) did not matter, as subjects would generally do better on the three-column format, given any ordering.  The column preference of the subject does not seem to matter as subjects preferred the three-column format by a relatively even 11-9 margin.  The results are statistically significant as shown through the t-test results. In fact, many subjects noted that the three-column format seemed shorter in length, even though the length of the passage was the same as the one-column format.  There are several possible explanations for why our hypothesis was incorrect.  First, the shorter three-column format allows the users to more easily maintain their current position while reading a line of text.  The longer one-column format requires users to maintain their position on a single line for a longer period of time, forcing more concentration and allowing more room for time consuming errors.  Second, three-column formats may allow for easier scanning as there is little information in each line and people who scan can more easily pick up the main ideas of each line without losing their eye position.  Lastly, shorter lines allow for easier retention of data in short term memory, so the users may not have to go back and reread something they missed, causing a greater time to reading completion.

Test Scores

The results of the test score for both treatments were not as hypothesized.  We believed that test scores based on the three-column format passage would be significantly higher than those based on the one-column format.  The reason behind this thinking was that subjects could make better use of short term memory with a three-column format.  The differences in test scores between one-column and three-column formats turned out to not be statistically significant.  This evidence suggests that page formats do not affect the level of comprehension of users.  The average score of the test based on the one-column passage and the one based on the three-column passage are virtually identical.  The t-test confirms this observation of the two formats yielding identical results by showing a relatively large p-value.  There is no statistically significant difference between one-column test scores and three-column test scores.  The test score data may not have yielded statistically significant results because of the nature of the test itself.  The questions tested meta-cognition of the passages presented (do you know that you know the answer) rather than cognition of the passage (do you know the answer).  Both test examine the reading comprehension level of the subject, but the meta-cognition test may require some extra skill that may affect the results.  While this kind of testing was used to make the subjects' task easier, and it worked well in pilot studies, the actual use of it in the experiment may have caused an uncontrolled variable to have crept in.

Observations

Several observations were made while watching subject's participate in experiment. These observations include:

• Use of finger: Subjects would often use a finger to point to the screen while reading text in the one-column format.  This occurred rarely when reading the three-column layout, but occurred often when subjects were reading the one column layout.  This may suggest that subjects had a harder time keep their eyes focused on the line because it was longer than the three-column layout

• Change of Strategy: After completing the test, some subjects informed us that they had changed their strategy for the test.  After reading through the first passage, some would go more into a "scanning mode" for the second passage, rather than a "full read" mode.  This strategy did not appear to affect results as the order of passages was randomized.

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