Wednesday, April 25, 2012

The Influence of Reading Speed and Line Length on the Effectiveness of Reading from Screen

Because so much of our reading these days occurs on screen rather than in print, I felt that this article lent some extra insight into the ways our minds process reading in a continuous form, rather than in print. The need to scroll, rather than turn pages, gives us more control over our reading environment. Dyson and Haselgrove explore best practices in their article, The Influence of Reading Speed and Line Length on the Effectiveness of Reading from Screen.

The authors introduce the paper with an evaluation of existing literature on reading and the best practices regarding line length. The authors note that the use of skimming has increased since the advent of the Internet, due to the large amount of text now available to readers. As of 2001, few studies examined the efficacy of reading from a screen. The study compares reading from screen at normal and fast speeds and focuses on line length as a typographic variable to explore its effects on comprehension and reading rates.

This study established that the amount remembered, a criterion of comprehension, increased signification when reported speed of reading decreased from around 300 words/min to about 150 words/min. However, one of the conclusions of traditional studies on reading print conducted by Tinker in the 1930s and 1940s was that the fast reader tends to comprehend better. Both negative and positive correlations between reading rate and comprehension have been reported as these are two different types of correlations: between-individual and within-individual. When an individual increases his or her reading rate, his or her comprehension decreases, which is a within-individual negative correlation. However, people who naturally read fast also tend to demonstrate a high level of comprehension, which is a between-individual positive correlation.

Comprehension was worse and reading was slower in the screen condition. However, this result was attributed to participants' lack of familiarity with computers and reading from screen. When speed of reading has been increased, this has surprisingly resulted in better comprehension when reading from screen than print. However, this result could be accounted for by a speed-accuracy trade-off, as the rate of reading print was significantly faster than reading from screen. There is a trade-off between reading rate and comprehension when readers were trained to read from screen at a faster speed. More recent studies found that mild time pressure, encouraging people to read slightly faster than normal from screen, can improve comprehension. The improvement is explained in terms of increased "mindfulness'', i.e. mild time pressure increases motivation and effort. The relationship between reading rate and comprehension therefore appears to be dependent on the extent of reading acceleration. A small
increase in reading rate may not impair comprehension but at some point increased reading speed appears to be at the expense of comprehension.


Rayner and Pollatsek (1989) have interpreted the rather confusing results from studies by Tinker and co-workers (summarized in Tinker, 1963) and come to a more specific conclusion on line length. They deduce that Tinker's work identified an optimal line length of 52 characters per line. The explanation given for the legibility of this moderate line length is that it is the outcome of a trade-off between two opposing factors. If line lengths are too long, the return sweeps to the beginning of the next line are difficult. If the lines are too short, readers cannot make use of much information in each fixation. Also, eye movement studies have revealed that readers decrease their saccade length, make
more "xations, and increase fixation duration, when small windows are used. Therefore, reading short line lengths seems to be particularly inefficient.  The authors examined three categories: Comprehension, Reading Rate, and Scrolling Patterns.

The participants' comprehension was highest when reading at 55 characters per line on the screen. Comprehension was higher at normal reading speed. A series of correlations was carried out to assess whether there was any relationship between the time an individual spent in reading a document within each reading speed and their comprehension (between individual correlation). At the fast reading speed, with all line lengths combined, there was only one significant negative correlation between reading rate and incidental questions (Pearson correlation coefficient, R= -0.23, p<0.02). At the normal reading speed, there was a positive correlation between reading rate and overall comprehension (R = 0.21, p<0.03).

Among the participants, comprehension was highest for main factual questions, and lowest for recognition questions. Comprehension scores for each question type are shown in Figure 3. The standard error bars for title, main idea and structure questions are larger as the data points are based on fewer questions. The data are averaged across the two reading speeds as there is no interaction between speed and question type. Using Newman}Keuls, structure questions were found to be significantly more difficult than all other question types (p<0.05) and main factual questions were answered better than recognition questions (p<0.05).

The authors' analysis on reading rate including reading speed and line length as factors found the expected main effect of reading speed (F(1,34) = 56.06, p<0.0001), confirming that participants in the fast reading condition were reading faster, and there is a clear separation between the two speeds. There was also a main effect of line length (F(2,68)"19.63, p(0.0001) and an interaction between reading speed and line length (F(2,68)"7.18, p(0.002). The data are illustrated in Figure 4. At the fast reading speed, documents of 55 and 100 cpl are read significantly faster than documents of 25 cpl. At the normal reading speed, 55 cpl documents are read faster than 25 cpl documents. All differences are significant at p < 0.05 using Newman-Keuls.

Finally, the authors found the best scrolling methods for online reading by tracking the participants' movement through the documents. In general, scrolling up or back tracking was not very common. The authors found no correlation between scrolling patterns and line length. The authors measured scrolling by speed of scrolling, size of scrolling movements and frequency of scrolling.

There were no significant effects in relation to speed of scrolling. However, there was a main effect of speed on scrolling length (F(1,34) = 4.25, p<0.05), with participants using shorter scrolling movements when reading at the fast speed. There was also an interaction between speed and line length on scrolling length (F(2,68) = 9.35, p<0.001). Scrolling lengths are similar in all line lengths at the normal reading speed. However when reading fast, longer scrolling movements are used at 25 cpl. At this line length there is little difference in the size of the scrolling movements at the two speeds. Whereas at 55 and 100 cpl participants use longer scrolling movements when reading at a normal speed, compared with a fast speed. There is also an interaction between speed and line length (F(2,68) = 5.02, p<0.01) on scrolling frequency. As with scrolling length, frequency of scrolling is similar across line lengths at the normal speed, but when reading fast, scrolling is less frequent at the short line length.

There is a main effect of speed on the time spent in pauses (F(1,34) = 59.38, p<0.001). Participants spend less time pausing when required to read fast. There is also a main effect of line length F(2,68) = 5.15, p<0.01). A follow-up test with Newman Keuls identifies a significant difference between
100 cpl and the two shorter line lengths, 25 and 55 cpl, (p<0.05), but no difference between 25 and 55 cpl. Pauses are longer at the longer line length. When comparing thetime spent pausing in relation to scrolling, there is a main effect of speed (F(1,34) = 52.95, p<0.001). At the fast speed, relatively less time is spent pausing compared with scrolling. There are no significant di!erences between line lengths, having taken into account the amount of scrolling required (number of lines).

I would include graphics for scrolling rate, but I did not understand them. The authors found that at fast reading speeds participants are generally making smaller scrolling movements; these movements are correlated to faster reading times. The authors concluded that although 55 characters per line is the optimum line length, further research can usefully explore the effects of manipulating scrolling patterns to determine whether changing individuals' patterns of pausing and scrolling can affect comprehension. The current results suggest that such experiments should also include factors that may increase reading difficulty.


Dyson, Mary C. and Mark Haselgrove (1999) The Influence of Reading Speed and Line Length on the Effectiveness of Reading from Screen. University of Reading Department of Typography and Graphic Communication. Retrieved from:


  1. With regards to reading from a screen, I have found the service Instapaper extremely useful for increasing my productivity. ( Instapaper allows you to save any article or webpage to read later and strips all ads and distractions from the page - increasing the readers focus on the task at hand.

  2. Great post - interesting and useful. I'd like to read more about this.

  3. Was there anything about words per minute at different line lengths?