The Effect of Screen Size on Readability Using Three Different Portable Devices

Introduction

Background

Desktop computers have become a part of everyday life. Computers are used for word processing, e-mail, entertainment, and information gathering. There are as many different uses for computers as there are users. The offspring of desktop computers, portable devices, are quickly becoming as common as the desktop. Portable devices, like laptop computers, are prevalent in all aspects of society, from the business world to the classroom. The convenience of ‘taking your work with you when you go’ is rapidly changing the way people do their jobs and live their lives. Now, new devices are becoming even smaller and more portable. As manufacturers create smaller and smaller devices, with increasing power and memory, at lower costs, portable devices will become ubiquitous. Beyond the initial novelty factor, many of these devices may offer their users increased efficiency and convenience. However, if they are to be more than just toys for the technologically intrigued, portable devices must meet the needs of their customers in a comfortable and reasonably familiar fashion. A heavy device will be left behind. A complicated interface may cause users to stop trying. Poor screen display will probably cause frustration, poor readability, and decreased use.

Generally, readability is measured through comprehension tasks and reading speed tasks. An example would be timing a subject while he read and answered questions about a passage. Legibility is measured through identification tasks. For example, asking a user to click a button when a certain character or string of characters is flashed on a screen. Sometimes, however, it is possible to gather information about the readability of a display by using a legibility task, such as those performed in proof reading. (Mills and Weldon, 1987) Visual search tasks, which are a form of legibility tasks, have been shown to generate eye movements similar to those in normal reading, while avoiding the linguistic demands of proofreading. (Ziefle, 1998)

Previous Research

Many studies have been done comparing the readability of text on computer screens. Some of these made comparisons among computer display types. For example, Miller, et al., (1997), compared two different visual displays, one static, the other dynamic, and found no statistically significant difference in search times. Other studies compared the readability of text on a specific terminal with the readability of paper documents. The results of these studies were mixed (Mills and Weldon, 1987). Nevertheless, a large number of studies have shown that users read significantly slower from the computer screen than from paper, while comprehension was unaffected. Mills and Weldon suggest that this may indicate that subjects maintain a rate of speed that allows for a consistent level of comprehension. Kruk and Muter (1984) discuss a correlation between a reduced amount of information on a computer screen and slower reading rates. However, they are unable to identify one single cause of slower reading from a video screen. Instead, they identify two possible causes. The first is format. A smaller number of characters per line and fewer lines per page reduces reading rate. Readers are unable to process large chunks of information at one time because the chunks are not there. This slows the reading speed. The second cause is vertical spacing. They identify text with little interline spacing, for example single-spaced text, as another cause of slower reading rates. Crammed text causes readers to struggle to decipher the text and makes reading more difficult.

Other studies have investigated different aspects of readability and legibility of screen displays. Bednall (1992) studied the effect of screen format on searching a visual list. Working with a list of names and telephone numbers, she studied the most effective way to format that list for rapid scanning and retrieval. One of her main findings was that increasing white space between lines improved search times, thus reconfirming part of the experiment done by Kruk and Muter(1984) . Piolat, Roussey, and Thunin (1997) studied the difference between presenting text page-by-page vs. scrolling text in comprehension and location of information. One of their significant findings suggested that page-based systems were better than scroll-based systems for maintaining a “sense of text”. Dyson and Kipping (1998) also compared the differences between scrolling and paging as part of their study on line length. Using only arrow keys and not the mouse, they found that users spent more time and key clicks moving text line-by-line (even when holding the key shifted multiple lines at a time) than moving text page-by-page. They also found that users who scrolled went backwards in the text more frequently than subjects changing text page-by-page. A number of other studies investigated the efficiency of reading with either continuously scrolling text (the Times Square Marquee model), or text controlled by the user that appears sentence by sentence. (Rahman and Muter, (1999); Juola et al. (1995)) These two studies are interesting because they investigate how people read and process text displayed on a very small screen. They are of particular relevance to cellular telephone web displays and other very small display situations.

In the study mentioned previously, of both line length and text movement, Dyson and Kipping (1998) found that users preferred text at about 55 characters per line (cpl), although reading rate actually increased as cpl increased; 25 cpl produced the slowest reading rate and 100 cpl the fastest. Text was left justified and on tests where fewer than 100 cpl were used the unused portion of the screen was white. Dyson and Kipping identify two possible causes of the faster rate in their experimental method. The first is that glare may occur on pages with shorter lines, thus slowing down the reading rate. Second, is that more text on the screen reduces cognitive learning time because subjects are more easily able to put information into context. This is similar to Kruk and Muter's findings on processing large chunks of information at one time. The authors did a follow-up experiment where they used gray on the unused portion of the page instead of white. Again, there was a difference between subjects preferences and their reading rates. However, the follow-up experiment did not mirror the methods of the first experiment, and the effect of line length on reading rate was not as pronounced. In their conclusions, Dyson and Kipping suggest “that line length should be considered as a significant factor, in relation to performance (reading rate) and as a criterion for judging ease of reading.”

Portable Computing Devices

The influx of products with various screen sizes, and the effect screen size has on readability led us to look at three types of small portable devices. These were a palmtop computer, an electronic book, and a laptop computer. There are many different brands of palmtop computers. Manufacturers include Casio, NEC, and Compaq. However, the most commonly known manufacturer is 3com (manufacturer of the Palm Pilot^TM.) Electronic books are relatively new on the market. At this time there are three main models to choose from, NuvoMedia's Rocketbook, Librius' Millennium EBook, and SoftBook. Like the palmtop computer, laptops are manufactured by many different companies. Each has a different design and internal configuration, although most laptops are PC based running some version of the Microsoft Windows operating system. The choice of which brands to use for our experiment was based on availability. For that reason we choose a Toshiba Satellite 4015CDT^TM laptop computer, a Rocket e-Book^TM, and a Palm V^TM. The Toshiba weighs approximately 7 lbs. Its screen can vary in resolution from 640x480 to 800x600. A full screen of 12 point type has approximately 27 lines with approximately 116 cpl. The Rocket e-Book^TM weighs 22 ounces, just under 1 pound. Information, including HTML, documents and ASCII files can be downloaded directly to the device. The resolution is 477x318 and the normal font size is 12 points. In a vertical orientation the screen has approximately 27 lines with approximately 40 cpl. The Palm Pilot^TM is the smallest of the three devices. It weighs significantly less than a pound and will fit in a shirt pocket. It comes with built-in applications including to-do lists, a date book, an address book, and a memo pad. Information can be downloaded or uploaded onto a PC or a Mac. The standard resolution of the Palm V which we used is 160x160 . A full screen of standard 10 point type has 11 lines and approximately 35 cpl. The readability of text presented on small screens, like the screens of the three devices aforementioned, is of concern to manufacturers and users alike. Users are choosing to use devices with smaller screens, despite their potential readability drawbacks. Users select these smaller devices because they are extremely portable, lightweight, and convenient to use. (Schilit, et al. 1999)

Experimental Model

Our study compares the readability of text on the displays of these three portable devices using a visual search task that focuses on the legibility of the text while still giving information about readability. We have designed the experiment to allow for two sets of comparisons. In one test the length of the line remains the same as the screen size differs and no scrolling or paging is required on any device. In the other, the length of the line increases as the screen size increases and scrolling or paging is required on all three devices. We predict that in the first experiment, where text size is identical, reading rates will be similar for all three devices. In the second experiment, previous research suggests that as the length of the line increases, measured in characters per line, the reading rate will also increase, and the time to task completion, or performance time will decrease. Research also indicates that when paging or scrolling is not factored out of the equation, subjects reading on devices which minimize the need for scrolling or paging will have faster reading rates than those reading on devices where more scrolling or paging is needed. However, the studies also indicate that subjective satisfaction will favor the device that is most similar to text based reading.

For all three devices, we tried to normalize the appearance of the screen, including the font size and type, and the contrast. However, we were unable to change or normalize the number of pixels per character. For smaller devices this causes some graininess in the font appearance. We tried to normalize the contrast by using the backlighting on the devices where available. In addition we limited the movement of text to paging only. These choices will be discussed further in the description of the experiment. With our study, we hope to help users make a well-informed choice when considering readability and screen size. We also hope to help designers build small devices that provide clear text readability.

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